Anda di halaman 1dari 402

H. Eschenauer, N. Olhoff, W.

Schnell

Applied Structural
Mechanics
Fundam entals o f Elasticity, Load-Bearing Structures,
Structural O ptim ization
Including Exercises

With 179 Figures

Prof. Dr.-Ing. H. Eschenauer


University o f Siegen
Research Center for Multidisciplinary Analyses
and Applied Structural Optimization FOMAAS
Institute o f Mechanics and Control Engineering
D - 57068 Siegen / Germany

TA
t

S ? 1 3

/ ?<77

Prof. Dr. techn. N. O lhoff


Aalborg University
Institute o f Mechanical Engineering
DK - 9220 Aalborg East / Denmark
Prof. Dr. Dr.-Ing. E. h. W. Schnell
Technical University o f Darmstadt
Institute o f Mechanics
D - 64289 Darmstadt / Germany

ISBN 3-540-61232-7 Springer-Verlag Berlin Heidelberg New York


Die D eutsche B ibliothek - C IP-E inheitsaufnahm e
E schenauer, H ans A.: A pplied structural m echanics: fundam entals o f elasticity. load bearing
structures, structural o ptim ization; including exercises / H. E schenauer; N. O lhoff; W. Schnell. B erlin; H eidelberg; N ew Y ork; B arcelona; B udapest; H ong Kong; L ondon; M ilan; Paris;
S anta C lara; Singapur; T okyo: Springer, 1997
ISBN 3-540-61232-7
NE: O lhoff, N iels; Schnell, W alter
T his w ork is subject to copyright. All rights are reserved, w hether the w hole or part o f the m aterial is
concerned, specifically the rights o f translation, reprinting, reuse o f illustrations, recitation, broadcasting,
reproduction on m icrofilm or in o ther w ays, and storage in data banks D uplication o f this publication
or parts th e re o f is perm itted only under the provisions o f the G erm an C opyright Law o f Septem ber
9, 1965, in its current version, and perm ission for use m ust alw ays be obtained from Springer-V erlag.
V iolations are liable for prosecution act under G erm an C opyright Law.
S pringer-V erlag B erlin H eidelberg 1997
P rinted in G erm any
T he use o f general descriptive nam es, registered nam es, tradem arks, etc. in this publication does not
im ply, even in the absence o f a specific statem ent, that such nam es are exem pt from the relevant
p rotective law s and regulations and therefore free for general use.
Product liability: T he p ublisher cannot guarantee the accuracy o f any inform ation about dosage and
application contained in this book. In every individual case the user m ust check such inform ation by
consulting th e relev an t literature.
T ypesetting: C am era-ready by editors
SPIN: 10508157
61/3020-5 4 3 2 I 0 - P rinted on acid -free paper

P r e fa c e

T he present E n g lish -lan g u ag e work is a com pilation of th e tw o -v o lu m e 3rd


edition ( in G erm an ) of E lastizitatsth eo rie (1993, 1994) published by B I W issenschaftsverlag M annheim , Leipzig, W ien, Zurich. Since th e first edi
tion of th is book h ad appeared in 1983, th e fundam ental concept of this
book has rem ain ed u n altered, in sp ite of an increasing am ount of stru ctu
ra l-a n a ly tic a l com putation softw are (e.g. F in ite E lem ent M ethods). T he
im portan ce of co m p u ter-to o ls, m ay th is b e supercom puters, parallel com pu
ters, or w orkstations, is beyond discussion, however, th e responsible engineer
in research, developm ent, com putation, design, and planning should always
be aw are of th e fact th a t a sensible use of co m p u ter-sy stem s requires a re
alistic m odeling and sim ulation and hence respective knowledge in solid
m echanics, th e rm o - an d fluiddynam ics, m aterials science, an d in fu rth er
disciplines of engineering an d n a tu ra l sciences. Thus, th is book provides th e
basic tools from th e field of th e th eo ry o f e la s tic ity for students of n atu ral
sciences an d engineering; besides th at, it aim s at assisting th e engineer in
an in d u strial environm ent in solving cu rren t problem s and thus avoid a
m ere black-box thinking. In view of th e growing im portance of product lia
bility as well as th e fulfilm ent of extrem e specification requirem ents for
new products, th is p ra c tic e -re le v a n t approach plays a decisive role. A part
from a firm handling of softw are system s, th e engineer m ust b e capable of
b o th th e g eneration of realistic com putational m odels an d of evaluating th e
com puted results.
Following an outline of th e fundam entals of th e theory of elasticity an d th e
m ost im p o rta n t lo a d -b e a rin g structures, th e present work illu strates th e
tran sitio n a n d in terrelation betw een S tr u c tu r a l M echanics an d S tr u c tu r a l
O p tim iza tio n . As m entioned before, a realistic m odeling is th e basis of
every stru c tu ra l analysis and optim ization com putation, an d therefore nu
m erous exercises a re attach ed to each m ain chapter.
By using tensor notation, it is attem p ted to offer a m ore general insight into
th e theory of elasticity in order to m ove away from a m ere C artesian view.
An arb itrarily shaped solid described by generally valid equations shall
be m ad e th e object of our investigations ( M ain C hapter A ). B oth th e condi
tions of equilibrium an d th e stra in -d isp la c e m e n t relations a re presented for
large deform ations ( nonlinear th e o r y ); th is knowledge is of vital im p o rt
ance for th e treatm en t of stability problem s of th in -w a lle d lo a d -b e a rin g

structures. W h en deriving th e augm ented equations as well as th e corre


sponding solution procedures, we lim it our considerations to th e m ost essen
tial aspects. All solution m ethods a re based on th e HOOKEAN concept of
th e lin e a r-e las tic solid. As exam ples of lo a d -b e a rin g structures, disks,
plates an d shells w ill be treated in m ore detail ( M ain C hapters B ,C ). Fi
nally, an introduction into S tructural O ptim ization is given in order to illus
tra te ways of determ ining op tim al layouts gf load-bearing structures ( M ain
C hapter D ).
In th e scope of th is book, th e m ost im p o rtan t types of exercises arising
from each M ain C hapter are introduced, an d th e ir solutions a re presented
as com prehensively as necessary. However, it is highly recom m ended for the
reader to te st his own knowledge by solving th e tasks independently. W hen
treatin g stru ctu ral optim ization problem s a large num erical effort generally
occurs th a t cannot be hand led w ithout im proved program m ing skills. Thus,
at corresponding tasks, we restrict ourselves to giving h in ts an d we have
consciously avoided presenting details of th e program m ing.
T h e authors would like to express th e ir g ratitu d e to all those who have as
sisted in preparing th e c a m e ra -re a d y pages, in tra n slatin g and proofreading
as well as in draw ing th e figures. At th is point, we would like to th a n k Mrs
A. W ach ter-F reu d en b erg , M r K. G esenhues, and M r M. W engenroth who
fulfilled th ese tasks w ith perseverance an d great patience. W e fu rth e r ac
knowledge th e help of M rs Dipl.-Ing. P. N euser an d M r Dipl.-Ing. M. Seibel
in proofreading.
Finally, we would also like to express our th an k s to th e publisher, an d in
p articu lar to Mrs E. Raufelder, for excellent cooperation.

H ans E schenauer
Siegen/ GERM ANY

A pril 1996

Niels Olhoff
A alborg/D E N M A R K

W alter Schnell
D arm stadt / G ERM ANY

C o n te n ts

L ist o f s y m b o ls

xra

I n t r o d u c t io n

F u n d a m e n ta ls o f e la s t ic it y

C h a p te r 2 to 7
A .l D e f in it io n s F o r m u la s C o n c e p ts

2.1
2.2
2.3
2.4
2.5

3
3.1
3.2
3.3
3.4
3.5

4
4.1
4.2
4.3
4.4
4.5

5
5.1
5.2
5.3
5.4

T e n s o r a lg e b r a a n d a n a l y s is
Term inology - definitions
Index rules an d sum m ation convention
T ensor of first order ( v e c to r)
Tensors of second a n d higher order
C urvilinear coordinates

S t a te o f s t r e s s
Stress vector
Stress tensor
C oordinate transfo rm atio n - principal axes
Stress deviator
E quilibrium conditions

S t a te o f s t r a in
K inem atics of a deform able body
S train ten so r
S tra in -d isp la c e m e n t relations
T ransform ation of principal axes
C om patibility conditions

C o n s t it u t iv e la w s o f l in e a r ly e la s t ic b o d ie s
B asic concepts
G eneralized H O O K E -D U H A M E L 's law
M aterial law for p lane states
M aterial law for a unidirectional layer ( U D -la y e r) of a
fibre reinforced com posite

5
6

7
10
13

18
18
20
21
24
25

26
26
29
30
31
31

31
31
32
35
37

VIII

6
6.1
6.2

6.3
6.4
6.5
6.6

6.7

7
7.1
7.2
7.3
7.4
7.5

Contents

E n e r g y p r in c ip le s
Basic term inology an d assum ptions
Energy expressions
P rinciple of v irtu al displacem ents ( P v d )
P rinciple of v irtu al forces ( P v f )
R eciprocity theorem s and U n it-L o a d -M ethod
T reatm en t of a variatio n al problem
A pproxim ation m ethods for continua

39
39
40
44
44
46
46
47

P r o b le m f o r m u la t io n s in t h e t h e o r y o f lin e a r
e la s tic ity

48

Basic equations an d b o u n d a ry -v a lu e problem s


Solution of basic equations
Special equations for th re e -d im e n sio n a l problems
Special equations for p lane problem
C om parison of sta te o f p la n e stress a n d sta te o f plane
s tr a in

48
49
49
50
51

A .2 E x e r c is e s

53

A-2-1
A-2-2
A-2-3

53
60
63

A-3-1:
A-3-2:
A-3-3:
A-4-1
A-4-2
A-4-3
A-4-4:
A-5-1:
A-6 -1 :
A-6 - 2 :
A-6-3:
A-7-1:
A-7-2:

Tensor rules in oblique base


A nalytical vector expressions for a parallelogram disk
A nalytical vector expressions for an elliptical hole in
ellip tical-h y p erb o lical coordinates
M O H R 's circle for a state of plan e stress
P rin cip al stresses an d axes of a th re e -d im e n sio n a l state
of stress
E quilibrium conditions in ellip tical-h y p erb o lical coordi
n ates ( continued from A -2-3)
D isplacem ents and com patibility of a rectan g ular disk
P rin cip al strain s from stra in gauge m easurem ents
S train tensor, principal strain s a n d volum e d ilatation of
a th re e -d im e n sio n a l sta te of displacem ents
Strain-displacem ent relation an d m aterial law in ellipti
cal-h y p e rb o lica l coordinates ( continued from A-2-3)
Steel ingot in a rigid concrete base
D ifferential equation and boundary conditions for a
BERNOULLI beam from a variatio n al principle
B asic equations of linear therm oelasticity by HELLING E R /R E IS S N E R 's variatio n al functional
A pplication of th e principle of v irtu a l displacem ents for
establishing th e relations of a triangular, fin ite elem ent
Hollow sphere under constant in n e r pressure
Single load acting on an elastic h a lf-s p a c e - A pplica
tio n of L O V E 's displacem ent function

66

67
70
71
73
74
76
78
80
82
83
86

89

Contents

P la n e lo a d - b e a r in g str u c tu r e s

IX

93

C h a p te r 8 to 10
B .l D e f in it io n s F o r m u la s C o n c e p ts

93

93

D is k s
D efinitions - A ssum ptions - Basic Equations
A naly tical solutions to th e homogeneous bipotential
equation

8.1
8.2

93
95

9
9.1
9.2

P la te s
D efinitions - A ssum ptions - Basic E quations
A nalytical solutions for s h e a r-rig id plates

99
99
107

10

C o u p le d d is k - p la te p ro b le m s
Isotropic plane structures w ith large displacem ents
L o ad -b earin g structures m ad e of com posite m aterials

113
113
118

10.1
1 0.2

B .2 E x e r c is e s

123

B-8-1:
B- 8 -2 :

123
128

Simply supported rectan g u lar disk under constant load


C ircular an n u lar disk subjected to a statio n ary tem p era
tu re field
B-8-3
R otating solid an d an n u lar disk
B-8-4
C lam ped q u a rte r-c irc le disk under a single load
B-8-5: S e m i-in fin ite disk subjected to a concentrated m om ent
B- 8 - 6
C ircular an n u lar C F R P -d isk u nder several loads
B-8-7: In fin ite disk w ith an elliptical hole under tension
B-8 -8 : In fin ite disk w ith a crack u n d er tension
B-9-1: S h ear-rig id , rectan g u lar plate subjected to a trian g u lar
load
B-9-2: S h e a r-stiff, se m i-in fin ite plate strip u n d er a boundary
m om ent
B-9-3: R ectan g u lar plate w ith two elastically supp orted bound
aries subjected to a tem p eratu re g rad ien t field
B-9-4: O verall clam ped rectan g u lar plate u n d er a constantly
distrib u ted load
B-9-5: R ectan g u lar plate w ith m ixed boundary conditions un
der distrib u ted load
*
B-9-6: C lam ped circu lar plate w ith a constant circular line load
B-9-7: C lam ped circu lar ring plate w ith a line load at th e outer
boundary
B-9-8: C ircular plate und er a distributed load rested on a n ela
stic foundation

131
133
137
139
145
151
153
155
157
167
170
172
177
179

Contents

C e n tre -su p p o rte d circular plate w ith variable thickness


un d er constant pressure load.
B-10-1: Buckling of a rectan g u lar plate w ith one stiffener
B-10-2: C lam ped circular plate under constant pressure consi
dered as a coupled d is k -p la te problem

183

199

B-9-9:

C u r v e d lo a d - b e a r in g s tr u c tu r e s

188
195

C h a p te r 11 to 14 C .l D e f in it io n s F o r m u la s C o n c e p ts

199

11

199
199
209
213

G e n e r a l f u n d a m e n t a ls o f s h e ll s

11.1
11.2
11.3

12

Surface theory - description of shells


Basic theory of shells
S h e a r-rig id shells w ith sm all curvature

M e m b r a n e t h e o r y o f s h e ll s

12.1
12.2
12.3
12.4
12.5
12.6

G eneral basic equations


E quilibrium conditions of shells of revolution
Equilibrium conditions of tran slatio n shells
D eform ations of shells of revolution
C onstitutive equations - m aterial law
Specific deform ation energy

13

B e n d in g t h e o r y o f s h e ll s o f r e v o lu t io n

13.1
13.2

Basic equations for a rb itra ry loads


Shells of revolution w ith a rb itra ry m eridional shape T ra n sfe r M atrix M ethod
B ending theory of a circular cylindrical shell

13.3

14

T h e o r y o f s h a llo w s h e ll s

14.1
14.2
14.3

C h aracteristics of shallow shells


Basic equations an d boundary conditions
Shallow shell over a rectan g u lar base w ith constant
principal curvatures

214
214
215
218
220
221
221
222
222
228
233
241
241
242
245

C.2 E x e r c is e s

247

C -ll-l:

247

F u n d am en tal quantities an d equilibrium conditions of


th e m em brane theory of a circular conical shell
C-12-1: Shell of revolution w ith elliptical m eridional shape sub
jected to constant in tern al pressure
C-12-2: Spherical boiler under in tern al pressure an d centrifugal
force

251
253

Contents

XI

C-12-3: Spherical shell und er w ind pressure


C-12-4: H anging circular conical shell filled, w ith liquid
C-12-5: C ircular toroidal ring shell subjected to a uniform ly distrib u ted b o undary load
C-12-6: C ircular cylindrical cantilever shell subjected to a tran sverse load a t th e end
C-12-7: Skew hyperbolical paraboloid ( h y p a r s h e l l ) subjected to
deadw eight
C-13-1: W ater ta n k w ith variable wall thickness under liquid
pressure
C-13-2: C ylindrical pressure tube w ith a shrinked ring
C-13-3: P ressure boiler
C-13-4: C ircular cylindrical shell horizontally clam ped at both
ends subjected to deadweight
C-13-5: Buckling of a cylindrical shell und er ex tern al pressure
C-13-6: F ree vibrations of a circular cylindrical shell
C-14-1: Spherical cap under a concentrated force a t th e vertex
C-14-2: Eigenfrequencies of a hypar shell

255
258
260

288
290
293
296

D S tr u c tu r a l o p tim iz a tio n

301

264
267
272
276
281
283

C hapter 15 to 18 D .l D e f in it io n s
15
15.1
15.2
15.3
15.4
15.5
15.6

16
16.1
16.2

17
17.1
17.2
17.3

F o r m u la s C o n c e p ts

F u n d a m e n t a ls o f s t r u c t u r a l o p t im iz a t io n
M otivation - aim - developm ent

Single problem s in a design procedure


Design variables - constraints - objective function
Problem form ulation - task of stru ctu ral optim ization
D efinitions in m ath em atical optim ization
T reatm en t of a S tructural O ptim ization Problem ( S O P )

A lg o rith m s o f M ath em atical P rogram m in g ( M P )


Problem s w ithout co nstraints
Problem s w ith constraints

S e n s i t i v i t y a n a l y s is o f ^ t r u c t u r e s
P urpose of sensitivity analysis
O verall F in ite D ifference ( O F D ) sensitivity analysis
A n alytical an d s e m i-a n a ly tic a l sensitivity analyses

301
301
301
302
303
306
307
309
310
310
314
321
321
322
322

XII

18
18.1
18.2
18.3

Contents

O p t im iz a t io n s t r a t e g ie s
Vector, m ultiobjective or m u lticriteria optim ization P A R E T O -o p tim ality
Shape optim ization
A ugm ented optim ization loop by additional strategies

325
325
329
334

D .2 E x e r c is e s

337

D-15/16-1:

337

E xact an d approxim ate solution of an unconstrained optim ization problem


D-15/16-2: O ptim um design of a plan e tru ss stru c tu re - sizing
problem
D-15/16-3: O ptim um design of a p a rt of a long circular cylindrical boiler w ith a ring stiffener - sizing problem
D-18-1:
M athem atical treatm en t of a Vector O ptim ization
Problem
D-18-2:
Simply supported column - shap e optim ization problem by m eans of calculus of variations
D-18-3:
O ptim al design of a conveyor b elt d ru m - use of
sh ape functions
D-18-4:
O ptim al shape design of a satellite ta n k - treatm ent as a m u lticriteria optim ization problem
D-18-5:
O ptim al layout of a point-supported sandw ich panel m ade of C F R P -m a te ria l - geom etry optim iza
tion

342
347
352
355
360
364
370

R efe r e n c e s

375

A
B
C
D

375
376
377
378

F und am en tals of elasticity


P la n e lo ad -b earin g structures
C urved lo ad -b earin g structures
S tru ctu ral optim ization

S u b jec t in d e x

383

L ist o f sy m b o ls

Note: The following list is restricted to the m ost im portant subscripts, notations
and letters in the book.
Scalar quantities are printed in rom an letters, vectors in boldface, tensors
or m atrices in capital letters and in boldface.
1. I n d i c e s a n d n o t a t i o n s
The classification is limited to the m ost im portant indices and notations. F u rth er
term s are given in the tex t and in corresponding literature, respectively.
i ,j , k , .. .

latin indices valid for 1,2,3

a , 3, p ,...

greek indices valid for 1,2

Index for a layer of a lam inate

Xj

subscripts for covariant components

x1

superscripts for contravariant components

( ii )

indices in brackets denote no sum m ation

prim e after index denotes rotated coordinate system e.g. ox .x #


com ma denotes partial differentiation w ith respect to the
quantity appearing after the comma, e.g. u#x

'

superscript prime before symbol denotes deviator, e.g. 't*

vertical line after a symbol denotes covariant derivative rela


ting to curvilinear coordinates !;*, e.g. v jj
b a r over a symbol denotes virtual value, e.g. Fj
roof over a symbol denotes the reference to a deformed body

tilde denotes approxim ation

asterisk right hand of a sm all letter denotes physical compo


nent of a tensor, e.g. a*

asterisk right hand of a letter denotes extrem al

asterisk right hajid of a capital letter denotes the c o m p le


m e n t a r y of work or energy, e.g. U*

n a b la-o p erato r

a4
V

differential operator
fl B

intersection of A and B

A c: B

A is a subset of B

for all

point

XIV
2.

List of symbols
L a tin le t t e r s

determ inant of a surface tensor

radius of a spherical or a circular cylindrical shell

a 0 ,a a

c o - and contravariant base vector of a surface in arb itrary


coordinates

a3

norm al unit vector to a surface


,a

c o - and contravariant components of a surface tensor

a ,b

semiaxis of an ellipse

determ inant of the covariant curvature tensor

b ap i

, b^

co -, contravariant and mixed curvature tensor


volume dilatation

e;

orthonorm alized base ( Cartesian coordinates )

ejjk , e'jk

perm utation symbol

volume force vector

f ,f

objective function, - vector

weight per area unit

determ inant of the m etric tensor

gj,g

inequality constraint function, - vector

gi.g*

c o - and contravariant base ( arbitrary coordinates )

g- , g 1J

c o - and contravariant m etric components, metric tensor

height of a boiler

h j, h

equality constraint function, - vector

hc , h|cu , hkl

core height, distance of the k. layer

buckling value

k =

t^

from the m id -p la n e

shell param eter

norm al unit vector

parabola param eter

vector of external loads ; vector of control polygon

Pj

p seudo-load m atrix

p[f ( x )j

preference function

p ,p

circumferential and norm al loads

Px >P,j >P

points

external loads of a cylindrical shell


distance perpendicular to axis of rotation
radius of curvature

distance to axis of rotation along the curvature radius


load vector
position vector to an arbitrary point of the m id -su rface or a
body

List of symbols
i j , rk

XV

orthogonal vectors

coordinate in meridional direction

Sj

vector of search direction

t , tk

w all thickness, layer thickness ( k = 1 , . . . n )

stress vector

t 1; t x , t y , tz

com ponents of a stress vector

11

state vector of a cylindrical shell, state variable vector

u ,v

displacements in m eridional and in circumferential direction

displacement vector

vq

displacements tangential to the m id -su rface

displacement perpendicular to the m id -su rface

w;

weighting factors, penalty term s

w*

approxim ation for deflection

design variable vector

x 1; x , y , z

Cartesian coordinate system, EUCLIDIAN space

x;

shape param eter

y,yj

transform ed variables

complex variable

Zj

state vector at point i of a shell of revolution

area, surface ; concentrated force at a corner

strain -stiffn e ss m a trix ; m atrix of A -conjugate directions

Bik

B -sp lin e base functions, B E R N ST E IN -polynom ials

rotation m a trix ; coupled stiffness m atrix

C j, C

transfer m atrix of a shell element, to tal transfer m atrix

C ^1

elasticity tensor of fourth order

elasticity m atrix

flexibility m atrix

tension stiffness of an isotropic shell

Dx ,

, DXJy

s tr a in - or shear stiffnesses of an orthotropic shell

Djjkl

flexibility tensor

E ,E

Y O UN G ' s m odulus, elasticity m atrix

^ap-yl

plane elasticity tensor

F ,F

objective functionals

F j, F

concentrated forces ; load vector

F(x )

im plicit representation of a surface

sym m etrical flexibility m atrix - mixed transform ation tensor


- system m atrix

XVI

List of symbols

shear modulus

Gj

penalty function

Gj

operator of inequality constraints

m ean curvature

Hk

operator of equality constraints

elasticity tensor of a shell

H , Hk'

HESSIAN matrices

integral function

I j , I 2 , 13

invariants

unity m atrix, - tensor

JACOBIAN m atrix

compression modulus

bending stiffness of an isotropic shell

GAUSSIAN curvature

Kx ,

, Kxj?

K * , K y , K xy , H

bending and torsional stiffness of an orthotropic cylindrical


shell
stiffnesses of an orthotropic plate

bending stiffness m atrix

differential operators

LAGRANGE -function

boundary moment
m om ent tensor

Mxx ,

, Mx^

bending and torsional moments

M aP

p se u d o -re s u lta n t m om ent tensor

N ^
N

,N
N
Nxx W xt?
N
N
_xx yy xy

m em brane force tensor


1
1
1I

norm al and shear components of m em brane forces

Nx^

effective inplane shear force

Pj

polynomials

transverse shear forces

Qx ,

effective transverse shear force

boundary force

Rj

penalty param eter

Rj

polynomial approxim ations

Rj , R2

radii of principal curvatures

R.

shape function of a shell surface

IR

n dimensional set of real numbers

stress tensor

List of symbols
S

XVII

shear stiffness m atrix

tensor of n - t h order ( n = 0 , 1 , 2 , 3 , 4 . . . )

Tj

CHEBYCHEV polynomials or -functions

transform ation m atrix

specific deformation energy

U*

specific complem entary energy

potential for field of conservative forces

volume

tensor of deformation derivatives, deformation gradient

V.

strain tensor (sym m etrical p a rt of V )

Va

tensor of infinitesim al rotations ( antisym m etrical p a rt of V )

weight

W , W*

external work, com plem entary work

feasible design space, subset

3.

Greek letters

sem i-angle of a cone

or

optim al step length

otrp

coefficient of therm al expansion

otap

strains of the m id-surface of a shell

oo , Pj

LAGRANGE multipliers

$ap

distortions of the m id-surface of a shell

Pj

components of a rotation m atrix

P'J

th e rm al-elastic tensor

Yjj

strain tensor

Yxy , Yxz , Yyz

shear strains in Cartesian coordinates

^ofi >1

strains, distortions

shear deformation
shear strain

variational symbol

Sj , Sy

K RO N ECK ER's tensor in curvilinear and Cartesian coordi


nates

S-

M A X W ELL's inijuence coeffients

, 8

factor of the step length , strain vector

ijk sijk
Eaf:

perm utation tensors

yy zz
i > <pip e <9<9

strains in Cartesian and spherical coordinates

vector of free thermal strains


coordinate perpendicular to m id-surface

XVIII

List of symbols

T);

slack variable

coordinate in latitude direction, latitude angle

/Cj, k 2

decay factors

K j, k 2

principal curvatures, variable exponents of an ellipse function

Kctp

tensor of curvatures

Xj

eigenvalue

Xj

vector of auxiliary variables

X , (i

LAME constants

(i

decay factor for a conical shell

v , v , \jy
, I;

POISSON's ratio
curvilinear coordinate system, GAUSSIAN surface param eters

m ass density

Pap

tensor of curvatures ( shallow sh e ll)

xx yy >aZz

norm al stresses in Cartesian coordinates

I I I 1 ii i

principal stresses

stress vector

oM

m ean value of norm al stresses

tim e

r 1J

stress tensor

Tx y

Tx z Tyz

shear stresses in Cartesian coordinates

<p

coordinate in m eridional d irection, meridional angle

X, ^

physical com ponents of the bending angles of a shell of


revolution

LOVE function
' ^x 'Jy

o ,X
, <J2

slope of cross-sections, bending angle


eigenfrequency, eigenfrequency param eter
coordinates of a spherical shell ( starting from the bound
aries )

G R E E N -L A G R A N G E 's strain tensor

Ejjk ,

CHRISTOFFEL symbols of first and second kind

A,

LA PLA CE-operator, modified LAPLACE operator

0 ( ,5)

tem perature distribution

0^1

th e rm al-elastic tensor

I l e , FL

external, internal potential

n * , 11*

external, internal com plem entary potential

n , II*

total potential, to tal com plem entary potential

A IR Y 's stress function

modified function, penalty function

In tr o d u c tio n

T h e classical fundam entals of m odern S tructural M echanics have been


founded by two scientists. In his work D iscorsi, G alileo GALILEI (1564 1642) carried out th e first system atic investigations into th e fractu re pro
cess of b rittle solids. Besides th a t, he also described th e influence of th e
shape of a solid (hollow solids, bones, blades of g rass) on its stiffness, and
th u s successfully treated th e problem of th e T h e o r y o f Solids w ith U ni
f o r m S tren gth. O ne century later, R obert HOOKE (1635 - 1703) stated the
fundam en tal law of th e lin ear th eo ry of elasticity by claim ing th a t str a in
(a lte r a tio n of len gth) and stress (lo a d ) are p ro p o r tio n a l ( ut tensio sic
v is ) . On th e basis of th is m aterial law for th e T h e o r y o f E l a s t i c i t y ,
E dm e M A R IO T T E (1620 - 1684), G ottfried W ilhelm von LEIBNIZ (1646 1716), Jakob BERNOULLI (1654 - 1705), L eonard EU LER (1707 - 1783),
Charles A ugustin COULOMB (1736 - 1806) an d others trea te d special pro
blem s of bending of beam s.
U ntil th e beginning of th e 19th century, th e T h e o ry of B e a m s had alm ost
exclusively been th e focus of th e T h e o ry of E la s tic ity an d S tre n g th .
Claude - Louis - M arie - H enri NAVIER (1785 - 1836) developed th e general
equations of elasticity from th e equilibrium of a solid elem ent, a n d thus
augm ented th e beam theory. Finally, he also set u p a torsion theory of th e
beam . Hence, he m ay quite ju stly be seen as th e actu al founder of th e
T heory of E lasticity. N A V IER's disciple B arre de DE SA IN T -V E N A N T
(1797 - 1886) augm ented th e work of his teach er by contributing new th e
ories on th e im p act of elastic solids. His contem porary, th e outstanding
scientist an d engineer G ustav R obert KIRCHHOFF (1824 - 1887), derived
w ith scientific strictn ess th e plate theory nam ed a fter him . T h e first m a th
em atical treatm en ts of shell stru ctu res w ere contributed by m ath em atician s
an d experts in th e theory of e la stic ity as Carl F riedrich GAUSS (1777 1855), CASTIGLIANO (1847 - 1884), M OHR (1835 - 1918), A ugustin Louis
B aron CAUCHY (1789 - 1857), LAM E (1795 - 1870), BOUSSINESQ (1842 1929), and, as m entioned above, NAVIER, DE S A IN T -V E N A N T and
KIRCHHOFF. A com plete b e n d in g theory of shells was derived system ati
cally by A ugustus E dw ard Hough LOVE (1863 - 1940) on th e basis of a
publication by ARON in 1847.
D uring th e 19th century, num erous works have been published in th e field
of S t r u c t u r a l M echanics w hich cannot be described in detail here. However,
based on th e a b o v e -s a id one m ight assum e th a t this discipline is an old
one, the problem s of which have largely been solved. As a m a tte r of fact,

1 Introduction

th is surm ise m ay have been tru e u n til recently. However, th e continuous de


velopm ent of th e sciences and th e technology, especially during recent
years, calls for an increased exactness of com putations, in p a rticu la r in th e
construction of com plex system s an d plants an d in lightweight construc
tions, respectively. Owing to th e introduction of duralum inium and other
advanced m aterials like com posites, ceram ics, etc. into th e lightw eight con
structions, th e num ber of publications in th e field of shell an d lightweight
stru ctu res has w itnessed a su b stan tial increase. In [C.6 ] it is shown th a t th e
am ount of publications has doubled per each decade since 1900. Proceeding
from about 100 papers in th e year 1950, one counted about 1000 publications
in 1982, i.e. th ree p er day. Thus, th e references to this book can only com
prise a very lim ited selection of textbooks and publications.
T he still continuing im portance of S tructural M echanics also stem s from
th e fact th a t th e relevance of structures th a t are o p tim a l w ith respect to
bearing capacity, reliability, accuracy, costs, etc., is becom ing m uch m ore
ap p aren t th a n in form er tim es. E specially in th e field of str u c tu r a l o p tim i
za tio n , considerable progress has been achieved during recent years and
this has prom pted increased research efforts in underlying branches of solid
m echanics like fractu re an d dam age m echanics, viscoelasticity theory, plastom echanics, m echanics of advanced m aterials, contact m echanics, an d sta
bility theory. Here, th e application of com puters an d of increasingly refined
algorithm s allows treatm en t of m ore an d m ore complex system s. In this

Fig. 1.1: Integration of m a th e m a tic a l s tr u c tu r a l o p tim iza tio n procedures


(MSOP ) into th e design process

1 Introduction

context, one should m ention th e large am ount of novel finite com putation
procedures (e.g. F in ite E lem ent M ethods [A.1, C.25]) as well as th e Algo
rith m s of M athem atical Program m ing applied in stru ctu ral m echanics. One
can thus ju stly claim th a t all of th e a b o v e -n a m e d m ore novel fields and
th e ir solution approaches are all based on th e fu ndam entals of elasticity
w ithout w hich th e cu rrently occurring problem s cannot be solved an d evalu
ated. T h e field of S tr u c tu r a l O p tim iz a tio n increasingly moves away from
th e stage of a m ere t r i a l - a n d - e r r o r procedure to enter into th e very des
ign process using m ath em atical algorithm s (Fig. 1.1). T his developm ent
roots back to th e 17th century, a n d is closely connected w ith th e nam e
G ottfried W ilhelm LEIBNIZ (1646 - 1716) as one of th e last universal scho
lars of m odern tim es. His works in th e fields of m athem atics an d n atu ral
sciences m ay be seen as th e foundation of an aly tical working, i.e. of a cohe
ren t th in k in g th a t is a decisive assum ption of stru c tu ra l optim ization.
LEIBNIZ provided th e basis of th e differential calculus, and he also inven
ted th e first m echanical com puter. W ithout his achievem ents, m odern opti
m ization calculations would yet not have been possible on a large scale.
Here, one m ust also nam e one of th e g reatest scientists L eonard EU LER
(1707 - 1783) who extended th e determ ination of extrem al values of given
functions to practical examples. T he search for th e extrem al value of a
function soon led to th e developm ent of th e v ariational calculus w here
entire functions can becom e extrem al. Hence, Jakob BERNOULLI (1655 1705) determ ined th e curve of the sh o rte st fa llin g tim e (B ra c h isto c h ro n e ),
and Issac N EW TO N (1643 - 1727) found the solid body o f revo lu tion w ith
the sm a lle s t resistance. Jean Louis LAGRANGE (1736 - 1813) and Sir W il
liam Row an HAM ILTON (1805 - 1865) set up th e principle of th e sm allest
action an d form ulated an integral principle, a n d thus contributed to the
perfection of th e variatio n al calculus th a t still is th e basis of several types
of optim ization problems. M any publications on engineering applications
over th e previous decades utilize th e variational principle. LAGRANGE,
CLAUSEN an d DE SA IN T -V E N A N T h ad alread y trea ted th e optim al shape
of o n e -d im e n sio n a l beam structures subjected to different load conditions.
T ypical exam ples here a re th e buckling of a column, as well as th e canti
lever beam for which optim al cross-sections could be found using th e vari
ational principle. T his requires th e derivation of o p tim ality criteria as ne
cessary conditions; these are E U L E R 's equations in th e case of unconstrai
ned problems. If constraints are considered, as, e.g., in solution of an isoperim etrical problem, LAGRANGE's m ultiplier m ethod is used.

A Fundamentals of elasticity
A.1 D e fin itio n s - F o rm u la s C o n c e p ts
2

T e n s o r a lg e b r a a n d a n a ly s is

2 .1

T e rm in o lo g y d e fin itio n s

T he use of th e index n otation is advantageous because it norm ally m akes it


possible to w rite in a very com pact form m ath em atical form ulas or system s
of equations for physical or geom etric quantities, w hich would otherw ise
contain a large num ber of term s.
C oordinate transform ations constitute th e basis for the general concept of
tensors w hich applies to a rb itra ry coordinate system s. T h e reason for th e
use of tensors lies in th e rem arkable fact th a t th e validity of a tensor equa
tion is independent of th e p articu lar choice of coordinate system . In th e fol
lowing we confine our considerations to quantities of th e th r e e - d im e n
sional EUCLIDEAN space. We introduce th e following d e fin itio n s:
A sc a la r characterized by one com ponent (e.g. tem perature, volum e) is
called a te n so r o f zeroth order.
A vecto r characterized by three com ponents ( e.g. force, velo city ) is called a
te n so r o f f i r s t order.
T he dyadic product of two vectors, called a dyad ( e.g. strain , s tre s s ), is a
te n so r o f second order characterized by n in e com ponents.
T ensors o f h ig h er o rder appear as well.
N otation of ten so rs o f f i r s t order:
a)

Symbolic in m atrix n o tatio n : a =

b ) A n a ly tic a l:

a = ax e x + ay e y + az e z
3

or

a = a e1 , +, a 2e , +, a 3e , V
y a 1e ;
i= l

w ith ex ,ey , ez as base vectors in a C artesian coordinate system . T he sub


scripts a re indices, an d not exponents. In index notation th e expression
a1
( or aj ) ( i = 1 ,2 ,3 ) denotes th e to tal vector ( see Fig. 2.1).
N otation of ten so rs o f second order:
a ) Sym bolic in m a trix n o ta tio n :

T =

*21

*22

*23

*31

*32

*33

2 Tensor algebra and analysis

Fig. 2.1: D efinition of th e base vectors

T = t n e1e1 + t J2 e1e2 + ... + t 33 e3 e3

b ) A n a ly tic a l:
or

T =

t ij e l e J >

w here e 1 e J is th e dyadic product of th e base vectors. In index notation th e


expression tjj denotes th e total tensor.
2.2 I n d e x r u l e s a n d s u m m a t i o n c o n v e n t i o n
( i ) Index ru le
If a le tte r index appears one an d only one tim e in each term of
an expression, th e expression is valid for each of th e actual va
lues, th e le tte r index can take. Such an index is called a free
in d ex.
( ii) E IN S T E I N 's s u m m a tio n co n ven tio n
W henever a le tte r index appears tw ice w ithin th e sam e term , as
subscript a n d /o r as superscript, a sum m ation is im plied over
th e range of th is index, i.e., from 1 to 3 in th e th r e e - d im e n
sional space ( L atin indices u s e d ), and from 1 to 2 in th e two dim ensional space ( G reek indices u s e d ). Such an index is called
dum m y.
(Hi) M a x im u m rule
Any le tte r index m ay never be applied m ore th a n twice in each
term .
E x a m p les o f ( i ) :
a* +

b* =

<=>

a1 +

b1 =

a2 +

b2 =

a3 +

b3 =

2.2 Index rules and convention

t
T .

=T

<=>
t

'1

= dT
dXj
T

i '2

=d ^
dx2

Note: Com m a im plies p a rtia l differentiation w ith respect to th e coordi


n a te ^ ) of succeeding indices. T h e rules ( i ) - ( Hi) apply for these in
dices as well.
E xa m p le s o f ( ii) :

a = a* e 4 = a 1 ej + a 2 e 2 + a 3 e 3 ,

th ree - dim ensional space ,

a = aQ e = a 3 e1 + a 2 e2 ,

two - dim ensional space ( surface ) ,

T = ty e 1 es* = tjj e 1 e 1 + t 12 e 1 e 2 + . .. + t 33 e 3 e 3 ,
t; - tj - 1; +

12

13

df = f j d x 1 = ^ j - d x
'

dX

,
+ -^ -d x
dx2

+ -M r-d x 3 .
dx3

A tten tio n : As it is of no im p o rtan ce w hich notation a doubly appearing


index possesses, th is s o - c a lle d d u m m y index can be arb itra rily renam ed:
i

21 E 6 . Su G- E
1
J

6 i . . .
K

E xception: No sum m ation over p aranthesized indices, i.e.


*

/ (ii)

a i = a i Vg

In

> a f = a i Vg

etc

E xa m p le s o f ( Hi) :
Following expressions are m ean in g less:
ci- 1i ! =

baa cos $ a

= 1

T he following expressions are also m eaningless, as th e free indices have to


be th e sam e in each term:
.1 , i I

O:
^ + bk = 0 , k p ia = Bp
2.3

T e n so r o f firs t o rd e r (v e c to r)

Base vectors (F ig .

2 .2 )

, e2 , e3 ,

e; =

o rth o g o n al b a se w ith th e u n i t v ecto rs

gj =

b a se in a r b itra ry co o rd in a tes w ith th e b a se v ecto rs g j , g 2 , g 3 .

M e asu re o r m e tric co m p o n en ts

Si

Sj Si j gji gj gi .

(2.1a)

2 Tensor algebra and analysis

Fig. 2.2: O rthonorm alized and


a rb itra ry base
M etric tenso r
S ll

(g n ) =

S l2

S l3

622

S23

S3 3
D eterm inan t of th e m etric tensor

Due to ( 2.1a) th e m etric


tensor is sy m m e tric a l.

det ( g ) = |g | = | g - | .

(2.1b)

(2.1c)

Scalar product x - y of th e vectors x = x* g ; an d y = y J g. ( Fig. 2.2 )


(2.2a)

x y = gij x1y J
Length of a vector x

(2.2b)

d = | x ] = / g i j xi xJ
Angle ip betw een vectors x an d y
SijX 1 y

cos ifi

(2.2c)

m n /
^gmn y y VgpqXP
Xq
Vg

C o v a ria n t and c o n tr a v a r ia n t base


An arb itra ry base gj ( i = 1 , 2 , 3 ) is given in th e th ree - dim ensional
EUCLIDEAN space. We are searching a second base gJ so th a t th e following
relation exists betw een th e base v e c to rs:
(2.3 a)
w here K R O N E C K E R 's delta is defined by
4 = 1
I

1
0

for i = i
for i ^ j .

(2.3b)

2.3 Tensor of first order

If th e base gj is known, th e base gJ can be determ ined by m eans of th e nine


equations ( 2.3a). T h e base g; iscalled th e covariant base and gj th e contra
v a rian t base.
C ovariant m etric com ponents
gij =

Si gj = gji -

(2.4a)

C ontravariant m etric com ponents


gij =

g 1 g j = g ji

(M b )

R ule of exchanging indices


g 1 = g J gj .

( z -5a)

gi = gjj g J .

( 2-5b)

4 = gi j gj k -

(^c )

O th er determ ination of th e co ntravariant base vectors

l
S

g2 x g 3
[ g i- g 2 . g 3 ]

g 3 x gi
3
[gi.g2 .g3 ] S

gi x g 2
[g1. g2.g3]

,9R)

w here [ g j , g 2 , g 3 ] is th e scalar trip le product of th e th re e covariant base


vectors g , , g 2 , g 3
T r a n s fo r m a tio n behaviour
A fundam en tal ( d e fin in g ) property of a tensor
is its behaviour in connec
tion w ith a coordinate transform ation. In order to investigate this transfor
m ation behaviour, th e following task shall be considered:
An in itia l base g ; or g 1 ( i = 1 , 2 , 3 ) is given together w ith a new base gj,
or g 1' ( i ' = 1 , 2 , 3 ) generated by an arb itrary lin ear transform ation w ith
prescribed transform ation coefficients (3\, . A dditionally, a vector be given in
the initial base by its com ponents v 1 or v; . Its com ponents v;- or v 1 shall
now be d eterm ined in th e new base.
R u le s o f tr a n sfo r m a tio n
T ransform ations of bases
g,- = p I ^

s' = Pj gJ ,

S k = / Jk g i'

gk = ^ g

(%-7b)

T he following relations are valid


P l-P f = %

Pl'P-.=t.

(2.8)

10

2 Tensor algebra and analysis

T ransform ations of tensors of first order


Vj, = P y v i

v 1 = (31 v 1 ,

(2.9a)

v.

v j = p[. v 1' .

(2.9b)

= /3V ,

P h ysica l co m p o n en ts of tensors of first order (vecto r)


a* = a
2.4

or

a* = a;

(2.10)

T e n s o rs o f se c o n d a n d h ig h e r o rd e r

D e fin itio n s :
Two vectors x an d y are given in th e EUCLIDEAN space. W ith th at we are
form ing th e new product
T = xy .

(2.11)

T he notation w ithout dot or cross sh all indicate th a t it is n eith er a scalar


product nor a vector product.
D epending on w hether th e covariant base vectors g ; or th e contravariant
base vectors g 1 are applied here, one obtains four kinds of descriptions for a
tensor of order tw o :

(2.12)

T = tlJ g ; gj = tj gi g J = t J gl gj = U g gJ
According to th e position of th e indices one denotes
tjj

as covariant com ponents ,

t'J

as contravariant com ponents ,

tlj

as m ixed contravariant-covariant com ponents ,

t;J

as m ixed covariant-contravariant com ponents of th e tensor T .

Form al generalization of tensors


T (0) = t

tensor of zeroth order ( scalar ) 3 =

1 base e le m e n t,

T (1) = tsgi

tensor of first order ( v e c to r)

31 =

3 base elem ents,

T (2) = ti j g ig j

tensor of second order ( dyad ) 3

T (3) = t ijkg ig jg k

tensor of th ird order

T (4) = t,jklg i g j g k gi tensor of fourth order

2
3

9 base elem ents,

= 2 7 base elem ents,

3 4 = 81 base elem ents.

2.4 Tensors of second and higher order

11

T ra n s fo r m a tio n ru les
For a tran sfo rm atio n of a vector base g ; into a new vector base g ;, equa
tions (2.7a) an d (2.7b) are used:
Si = P\ gj-

an d

g i. = p\. gj .

T he tensor T can be given eith er in th e old base gj or in th e new base gj,


T = tk

1 g k,

g l, = t ,J g ; gj

(2.13)

T h e tran sfo rm atio n form ulas read as follows


t ij = p[, p\. tk' V
From

follows

or

t' J' = & p { t kl .

t~ g 1 g J = tk, v g k

= / /?J t^ j.

or

In a sim ilar way one obtains


com ponents of th e tensor T.

t , r = ft) fi\, t kl

(2.14)
(2.15)
(2.16)

th e transfo rm atio n form ulas of th e m ixed

Note: It is w orth m entioning th a t tensors are actually defined by th e rules


by which th e ir com ponents transform due to coordinate tran sfo rm a
tions. Thus, any quan tity T w ith 3 2 = 9 com ponents is th en a n d only
th en a second order ten so r if its com ponents transform according to
(2.14) or (2.16) in connection w ith an a r b itr a r y coordinate transfor
m ation.
P h ysica l co m p o n en ts o f a ten so r o f second order
T he physical com ponents for orthogonal coordinate system s can be deter
m ined as follows ( for non - orthogonal coordinate system s see [ A .8 ] ) :
t*ij -

tJ

Vg( i i) / g ( j j ) >

t *1j =

'
(2.17)

t* J =

t;i

t+ij = tq

/^TiTT

7 (ii) / g (jj)

S y m m e tr ic a l a nd a n ti s y m m e tr ic a l ten so rs o f second order


Any tensor of second order can always be presented as a sum of a sym m e
tric a l an d an an tisy m m etrical ( or skew - sy m m e tric a l) te n s o r :
tij =
w ith

+ (at >ij

(2.18a)
(2.18b)
(218c)

12

2 Tensor algebra and analysis

P e r m u ta tio n ten so r or e ~ tensor


As perm utation tensor a tensor of th ird order is defined
E.

ijk
~= V &

u i = 4 r ' l
a

(* *

w ith th e perm utation symbol


ik
(
e;.k = e'J = <

+ 1
- 1

for

i k } cyclic
{ i , j , k } anticyclic ,
{ i >j >k } acyclic

(2.20a)

Perm utation symbol in two dim ensions


ejj =

e 12 = +

,
(2.20b)

Vector product as application of th e e ~ ten so r

(*1)

x x y = ekimxV g mE igenvalu es and eigenvectors o f a sy m m e tr ic a l ten so r


- P rin c ip a l a x is tra n s fo rm a tio n
Lem m a:

For any sym m etrical, real valued, th re e -c o lu m n m atrix T th ere


always exist th ree m utually orthogonal principal directions
( eigenvectors) a and th re e corresponding re al eigenvalues A
( w hich not necessarily have to be different from each o th e r ).
T hese eigenvectors an d eigenvalues a re governed by th e follow
ing algebraic eigenvalue problem, w here I is th e u n it tensor:
(T -A I)a =

or

(t* - X 6*) a- = 0 .

(2.22a)

D eterm ination of th e eigenvalues:


tj - A
det ( t-j - A j ) =

t2 - A

(2.22b)

C haracteristic equation of ( 2.22b) :


A3 - Ij A2 + I 2 A - I 3 =

(2.22c)

T he roots A = Aj, An an d Am of th is cubic equation are in v arian t w ith


respect to transform ations of coordinates. Substituting sequentially these
eigenvalues into (2.22a) an d solving for a , we obtain a j, a j j and a m -

2.5 Curvilinear coordinates

13

T h e quantities I j , I2 , I 3 in (2.22c) a re invariants defined by [ A.8 ]:

2.5

h = t| ,

(2.23a)

i2 = i ( t W - t j t o ,

(2.23b)

I 3 = det ( t*) .

(2.23c)

C u r v ilin e a r c o o rd in a te s

Base vectors - m e tr ic ten so r


In th e th re e - dim ensional space a vector r can be presented in C artesian
coordinates x 1 an d in curvilinear coordinates indicated by ( l ( i = 1 ,2 ,3 )
( Figs. 2.3 an d 2.4 ).

Fig. 2.3: Position vector in


orthonorm alized base

Fig. 2.4 : C urvilinear coordinates


and base vectors

Position vector r of a point P


r = r ( x i)

* = *(')

(2.24)

B ase vectors
dT
e i = Vd x~' i = r 'i

Si

(2.25)

R elation betw een base gi ( ( l ) and orthonorm alized base &


(2.26)
L ength of a line elem ent
ds 2 = d r d r

y F irst fundam ental form of a surface.

(2.27a)

Indicating th e derivative w ith respect to th e curve p ara m eter t by a dot, the


length of th e curve betw een t 0 and tj is given by :

dt .

(2.27b)

14

2 Tensor algebra and analysis

Volume elem ent


dV = f g d* d 2 d 3 .

(2.28)

P a rtia l base derivatives - CH RISTO FFEL symbols


8k

elk

, i = r u 8j .

( 219a)

= ~r L&-

( S9h)

CHRISTO FFEL sym bols of th e first kind


r ijk = Y ( gjk,i + gki,j ~ gij,k ) '

f 2'30)

CHRISTO FFEL sym bols of th e second kind

r m = g kmr ,jk

(Z31)

R u le :

T h e CHRISTOFFEL symbols can be expressed alone by the


m etric tensor an d its derivatives.

Note:

T h e CH RISTO FFEL symbols do not have tensor character.

F or th e CH RISTO FFEL symbols of th e first kind (2.30) th e following re la


tions hold:
1)

Tyk = Fyik

2 )

r m n s + r m sn = T r r f '

interchangeability
of th e f i r s t two indices ,

(2.32a)

interchangeability
of th e la st two indices .

(2.32b)

For th e CH RISTO FFEL symbols of th e second kind, th e following relations


are derived from (2.31) using (2.30):
1)
2

r k = r k

interchangeability of subscripts (s y m m e try ) ,

r ' = gik dgik = a O n / f O


,j

2 s

'

(2.32c)
(2.32d)

C o v a ria n t d eriv a tiv e s


T ensor of first order
a ,j = a 1|j g i
w ith

a |j = a 'j +

By anology

a^

(2.33)
ak .

(2.34a)

= ai J - . T y a k .

(2.34b)

2.5 Curvilinear coordinates

15

Tensor of second order


I
a*.l
ljlk

a.*
i ni m
.1l k a m
ij ,k

= A

+ r L

ln jm
..k 3 ,.1 m

*"1

(2.35a)

(2.35b )

+ ' L * " -

G radient of a scalar funktion $


( 2.36a )

v = g rad $ = V $ # |j g J .
G radient of a vector v

(2.36b)

G ra d v = V v = v ^ g * g j .
Divergence of a vector v
div v = V - v = v j |j - --L yg

(2.37a)

( /g vj ) .

Divergence of a ten so r T of second order


(2.37b)

Div T = V - T = tkl|k gj
R otation of a vector v

(2.38)

r o t v = V x v = v J|i ( g 1 x g j )
LAPLACE operator
A

= V 2 $ = div grad $ = * |! = - ) L ( / i gjk


yg

k) ,

(2.39)

'

B ipotential operator
AA $ = V 4 $ = V 2 ( V 2 $ ) = $|!j =
(2.40)

GAUSSIAN theorem
J J J H /g
V

d ( 3 = J J v j n jd A .
A

(2.41)

2 Tensor algebra and analysis

16

E xa m p le: A p p lic a tio n o f the p re v io u s fo r m u la s to c y lin d ric a l coordinates


Single-valued relations betw een C artesian coordinates x 1 and cylindrical
coordinates
read as follows ( see Fig. 2.5 ):
1

x = cos (

-l

-2

x = ( sm (

-3

x f

(2.42a)

Position vector
(2.42b)

= f c o s (2 e l + sin e 2 + e 3 .

r ( 7

C ovariant base vectors according to (2.25)


g j = cos
dr
Bi = r ,i

62

+ s in e 2 ,

= -^sin ^ ej + ^cos^e2 ,

(2.43)

"*7

e3
C ovariant m etric com ponents according to ( 2.4a)
Sij 6 i
For exam ple:

g 22 = B2 '

62

= 7

) 2 sin 2

+ ( f

cos2

= ( f

C ovariant m etric tensor


1

(gy) =

(? ?

According to ( 2.5c), because of


>

(2.44)

g^ = gj

O rthogonal base

C ontravariant com ponents from (2.5c)


g

( )

i
g(ii) = 1

S(ii)

for i f j

2.5 Curvilinear coordinates

17

C ontravariant m etric tensor


1

( g ij)

o U

1)- 2

(2.45)

D eterm in an t g of th e covariant m etric tensor


1

I gij I

1 )2

(2.46)

CHRISTO FFEL sym bols of th e first kind according to (2.30)


For exam ple:

T 221 = y { g

2 1 ;2

+ g 12

>2

- g22;1) = - ? r (

A ) = f*

CHRISTO FFEL sym bols in m a trix notation


'

-e

e1

(A p) =

'
1

(2.4 T)
'
( A j2) =

(A ja)-

>

CHRISTO FFEL symbols of th e second kind according to (2.31)


For exam ple:

T 22 = glk f 22k =

( - ^ 1) +

-|-0

- 0

= -^1.

CHRISTO FFEL sym bols in m atrix notation


0

(*i> =

0
0

(2.48)
0

(Aj)

1 )- 1

1)"1
0
0

18

3 State of stress

LAPLACE operator according to (2.39)


a # = ^ ( / g g J k ^ , k ) lj = y T

^ ^ gJl^ d +

[ ( / g gU ^ , i ) , i + ( / g g 2 2 ^

= | i [ ( e ^ . 1),i + ( ^ ,

e1

e1^ , n +

^ ,1

^ ,2 2

A $ $ n + ^1 ^ , 1 +

S ta te o f s tr e s s

3.1

S tre s s v e c t o r

2 ),2

, 2 ),2

g J2 ^ ,2

+ gJ 3 # ,3)

+ ( / g g33^ . 3 ),3

+ (e^,3),3] =

+ e1^ 33

^ !^ 2

^ 1 22

^,33

(jg.4s;

T h e essential objective of stru ctu ral analysis is th e calculation of stresses


and deform ations of bodies. As shown in Fig. 3.1 we m ake a cut through th e
body, w hich is in equilibrium under ex tern al loads in th e form of volume
forces f ; , surface tractions p ; and concentrated forces F k . A resulting force
A F is tra n sm itte d at every elem ent A A of th e cut.

F ig . 3.1: C ut through a body

F ig . 3.2: Resolution of th e stress vector

3.1 Stress vector

19

According to N E W T O N 's principle of actio = r e a c tio , reaction of the


resulting force A F is found on th e sam e plane of th e opposite p a rt of th e
body, in th e form of an opposite directed force of th e sam e m agnitude. We
assum e th a t th e relation A F / A A in th e lim it of A A > 0 tends to a fin ite
value, and we call th is lim iting value
stress vector

t=

lim 4M r = 4 r -
aa-o^A
dA

(8.1)
1
'

Here it is assum ed th a t only fo rces ( and no m o m e n ts ) are tra n sm itte d at


any point of th e cut.
T he stress vector t can be resolved into a p a rt p erpendicular to th e surface
of th e cut, th e so - called n o rm a l stre ss of th e value cr, and into a p a rt ta n
gential to th e surface, th e sh ea r stress of th e value r ( Fig. 3.2).
S ig n convention'.

Stresses on cut planes w ith outw ard norm als pointing


in th e positive ( respective n eg ativ e) coordinate direc
tions, are tak en positive in th e positive ( respective
negative ) coordinate directions ( Fig. 3.3 ) .

Stress vectors on th e positive cut planes of th e cubic elem ent in C artesian


co o rd in a te s:
XX

II
>>

xy
xz J

T
yx
cr
yy
r
yz J

Tzx

(3.2)

ii

<7

zy
L

(7z_z

In th is context, <7 .. ( i = x , y , z ) are norm al stresses and t - ( i , j = x , y , z )


are sh ear stresses.

"positive"
cut planes
"negative"
cut planes

1
cut planes

k c yy
T^vx

positive
stresses

X
M

negative
stresses

F ig. 3.3: Sign convention for th e stresses

Ih

20

3 State of stress

3.2

S tre s s t e n s o r

T he stress vectors can be assem bled in m a trix n otation as th e so -c a lle d


stress tensor S . In C artesian coordinates it reads

"
ft x j t y ! t z1

^xx

Txy
L

xz

Tyx
yy

Tzy

Tyz

^zz

Tzx

(3.3)
J

T he superscript T indicates tran sp o se of a m a trix


T he im p o rta n t CAUCHY's form ula in arb itra ry coordinates is w ritten

or

t = Sn

(3.4a)

t1 = 7 * 1 1 ,

(3.4b)

I n words-. T his form ula gives th e stress vector t at a given surface or cut
plane in term s of th e stress tensor S an d th e u n it outw ard nor
m al vector n for th e surface or cut.
T he stress vector t acts on th e infin itesim al area dA of th e inclined cut
plane characterized by th e u n it outw ard norm al vector n ( Fig. 3.4 ):
" n

y
, nz.

" cos a '


cos /?
cos

r t X i

t =

ty
A .

( 3.5a,b)

T he rem ain in g infinitesim al surfaces of th e tetrah ed ro n result from th e pro


jection of dA w hich can be w ritten as follows in index notation w ith x = x 1,
y = x 2, z = x 3 :
dA

dA cos a dA n

dA cos B = dA n
H
y
dA,Z = dAcos 71 = dA n ;
dA y

dA- = dA n- .

(3.6)

F ig. 3.4: Stress vector t at a tetrah ed ro n in


C artesian coordinates

3.2 Coordinate transformation - principal axes

21

R elationships betw een C artesian an d o th er coordinates will be given later.


In C artesian coordinates th e re is no difference betw een a covariant and a
contravarian t base, an d for th is reason th e indices can always b e lowered.
C om ponent notation in C artesian coordinates
t.

ffx x n x

ty =

r xy n x + < T y y n y +

Tzy * z >

4.

Tx z n x +

^zz n z

+ Tyxny +

Ty z n y

Tz x n z

(3.7)

Note: Shear stresses are pairw ise equal to one another, i.e., the stress ten so r
S is s y m m e t r i c

r xy = r yx

(3.8)

r yz = r zy

T he sym m etry reflects satisfaction of m om ent equilibrium conditions.


3 .3

C oordinate tran sfo rm atio n principal axes

We consider a C artesian coordinate system x 1 and a rotated system x* ( see


Fig. 3.5).
Stresses in a ro tated system according to (2.14)
T,J

Symbolic n otation
S' = B

(3.9)

BT .

rotated coordinate system

22

3 State of stress

A rrangem ent of transform ation coefficients in a ro tation m atrix B


X

3'
3 J

2'

B =

ft
X

(S.10)

P r in c ip a l stresses, p r in c ip a l axes
P rincipal stresses ( see (2.22) and (2.23) )
(3.11)

( Tj a ^ ) n * = 0

C haracteristic equation
cr3 - Ij cr2 + I 2 a - I3 =

(3.12)

w ith th e invariants for anv direction and for th e principal stresses cr.
( i = I , I I , I I I ):
G-T,

yy
Txy
Txy

(Jr -f- (Trr H- (J-hi


yy

ayy

yz

sum of norm al stresses ,

C j C jj +

<7n CTjjj +

( 3.13a)

c rm

<7j ,

Ti

(3.13b)
o

h =

XX

T*y

T xz

r xy

yy

Tyz

a .ZZ

xz

(3.13c)

yz

Note: It can be shown th a t th e th ree roots of (3.12) com prise th e m axim um


and th e m inim um norm al stress appearing on all possible cut planes
through a given point. T h a t is w here th e nam e p rin c ip a l stresses is
coming from. For th e sym m etrical stress tensor th e principal stresses
are always real. T he directions of principal stresses of different m ag
nitudes are always unique an d m utually orthogonal.
State o f plane stress in C a rtesia n coordinates
D e fin itio n :

o zz = r xz = r yz =

(3.14)

Stress tensor
a

S =

r xy

xy
a yy

(3.15)

3.3 Coordinate transformation - principal axes

23

Fig. 3.6: C oordinate transform ation


T ransform atio n coefficients according to Fig. 3.6
/3j = cos a

(3\ = cos ( y - a ) = sin a ,


(3.16)

/3j = cos ( +

a ) = - sin q

/?2

= cos a .

Form ulas of transfo rm atio n for any ro tatio n a of th e coordinate system


* 'x ' = \

( CTxx + yy ) + \

( ffxx " ^yy ) COS 2 a + Txy sin 2 a ,

y'y' = i ( <7xx + <7 y y ) - i ( Tx x - Ty y ) COS2Q Txy sin


Tx 'y '

( <Ty y - f f x x ) s i n 2 Q

(3.17)

T zxy
v COs 2 q '

P rincipal stresses

ai
1
ii J

+ T.y

= - ( crx x + a y y
2 l

(3.18)

T he directions of th e principal stresses follow from th e extrem al condition


to be
ta n 2 a* =

2 t . ...

xy
a x x - a yy

(3.19)

an d from th is th e principal directions

a* and

a* + it or a* and a * +

7T

T he principal directions are orthogonal to each other.


M axim um sh ear stress
xx

= i/(s
D irection

yy

a** = a*

+ r xl y.

= ^

~ U

(3.20)

24

3 State of stress

Fig. 3.7: M O H R 's stress circle


M O H R 's circle
T h e form ulas for transform ation of th e plane state of stress lead to the
M O H R 's circle ( c x,x ,
, r x ,y, = r )
a xx 4- a yy

(a

+ T = ( 7 - 70 f + T2 =

(3.21)

w ith th e distance of th e centre M on th e er-axis


1

2 '

xx

(3.22a)

+1 Oyy

and th e radius of th e circle

4- txy
3 .4

= r max

(3.22b )

S tre s s d e v i a t o r

D efin itio n :

'r* = t* - a M <5

(3.23)

w ith the m ean norm al stress o M


M

( x x

y y + a zz ) ~

( a l + Cjj + <7n I ) 3

( 3 .2 4 )

P h ysica l in te r p re ta tio n :
T he stress deviator 'r j expresses th e deviation of th e state of stress from the
m ean norm al stress.
Since 'Ij =
from

0,

th e principal values of th e stress deviator follow in analogy to

( 3 .1 2 )

v 3 + 'I 2 V - ' I
w ith

I 2 I2

3 crM ,

Ij

I 2 <7m +

(3.25)

|
I3

<tm .

(3.26)

3.5 Equilibrium conditions


3 .5

25

E q u ilibrium conditions

T he conditions of force e q u ilib riu m are stated w ith regard to th e u n d e


fo rm e d configuration of th e body in th is section [ A .ll, A.15, A.16, A.17 ].

xx+

3x

F ig. 3.8: E quilibrium for an infinitesim al volume elem ent in C artesian


coordinates
1

) C artesian coordinates ( Fig. 3.8 )


d <7.

dX
dT

d ryx
d

a..

dT

dX

d r zx
dZ
d T.

dX

d
d

(3.27a)

d z ^ + fy = 0

dz

+ f =

o .

fj ( i = x , y , z ) a re th e com ponents of th e vector of volume forces.


A bbreviated notation
+ fi =
2

(3.21b )

) C urvilinear coordinates
- j li +

or

Div S + f =

(3.28a)

0
0

(3.28b)

3 ) C ylindrical coordinates ( * t ,

d Tyr
d

1 d r ,
dr
+ J - J ! +
r d ip
dZ

f d cr
r

dip

dTn + J-1 d rj etcz


r d y ?
d

= ip , ( 3 z ) ( Fig. 2.5 )
1

+
r (Va r r - a

dr
9
+ -V 542- + r
r
__zz_
i z

+ fv =
f

rz

) + fr =

tptp'

=
z

(3.29a)

26

4 State of strain

F ig. 3.9: Spherical coordinates


( ? = * , ? = <>, t 3 = <P)

T w o -d im en sio n a l case : Polar coordinates r, ip


d<TTT , 1

4. 1 (

+ ~r Tdip
7n + T (

ffW

) +

fr =

0 >

(3.29b)
d T

JL +

c)r

1 d<r

dip

' r ^r

4 ) Spherical coordinates r, 1?, ip ( Fig. 3.9 )


dT
v<) dtp
\ ^
r sin

dI

3-----r Tr^COt ^ +

+
( 2 a r r - <r,,
- cripip ') + fr
j- v
i/I?
dr

dT

dr

d i?

r sin i?

ip

+ - T

(3.30)

, +

+ y c o t 0 (<rw - crw ) + {* =
dT

jr<_

dI

dT Q

+ -r - Vdl>^ +

iff. .

-----
r s in v

d ip

_|--------- _
r

_|------- T
rip

S ta te o f s tr a in

4.1

K in e m a tic s of a d efo rm ab le b o d y

QOt
ip v

_(_ f

q .

ip

D escription of th e deform ation of a body w ith LA G RA N G E's n o tatio n :


T h e displacem ent of a m aterial point of a body B is observed as a
function of th e in itial state.
We distinguish betw een th e in itial state t = t 0 (w ith o u t ~ ) and th e defor
m ed sta te of th e body t = t ( w ith ~ ).

4.1 Kinematics of a deformable body

27

Position vector f of th e m aterial point P of th e deform ed body B ( Fig. 4.1)


? ( * ) = r U 1) + v U 1)

(4.1)

w ith th e position vector r ( * ) and th e displacem ent vector v ( * ) of the


sam e m aterial point P of th e undeform ed body B.
D ifferential increase d v of th e displacem ent vector v
dv = d r - d r = V d r ,

(4.2)

w here V is th e ten so r o f the d isp la cem en t d eriva tives.


A ccording to (2.25) th e base vectors g ; an d g ; result from th e total differen
tia l of th e respective position vectors
d r = ~ 7 T df 1 = r (i df 1 = g i d ^ 1 ,

(4.3a)

dr =

(4.3b)

of

d^1 = r (i d* = g; d f 1 .

T hese in fin itesim al changes of th e position vectors lead to th e points Q and


Q adjacent to P an d P ( see Fig. 4.1).
In accordance w ith th e rules for th e transfo rm ation behaviour of tensors
( Sections 2.3, 2.4 ) th e base vectors of th e deform ed body can be expressed
by those of th e undeform ed body and vice v e r s a :
Si=f3\Sj ,

(4.4a)

Si = P \ gj -

(4.4b )

28

4 State of strain

T he displacem ent vector can be w ritten as follows


v = v g ; = v gj = r - r .

(4.5a)

D ifferentiation of (4.5a) w ith (2.SS) leads to


(4.5b)
Since th e base of th e undeform ed body or th e base of the deform ed body
can be used altern ately in order to illu strate th e displacem ent vector v , we
have two different covariant derivatives of th e displacem ent com ponents as
a resu lt ( according to (4.5b) ) . Here, one line stan d s for the covariant deri
vative applied to th e base of th e undeform ed body, a double line stands for
th e covariant derivative related to th e base of th e deform ed body. W ith the
KRO NECK ER sym bol we obtain th e relation betw een th e base vectors of
th e deform ed body and th e undeform ed body [ A.7 ]:
(4.6a)
(4.6b)
T h e elem ents of th e transfo rm atio n tensor th e n read
(4.7a)
(4.7b)
Corresponding tran sfo rm atio n relations are valid for th e line elem ents d r
and d f in analogy w ith th e base vectors. If we define a m ixed tensor of or
der two according to (2.12)
(4.8a)
(4.8b)
for th e line elem ents th is results in
dr = F dr ,

( 4.9a )

dr = F d f .

(4.9b)

By m eans of (4.2) an d due to (4.5a) we obtain th e following to tal differen


tia l of th e displacem ent vector d v :
dv = d f - dr = ( F - I ) dr = V dr .

(4.10a)

According to (4.2), V is called th e ten so r o f the d isp la cem en t d e riva tive s


or th e d efo rm a tio n g ra d ie n t.

4.2

Strain tensor

29

Due to (4.10a), th e to tal differential d v can be w ritten in o ther notation by


m eans of (2.S6b)
(4.10b)

d v = Grad v dr .

In C artesian coordinates th e relatio n (4.10b) for a tim e independent dis


placem ent vector
u(x,y,z)
v (x ,y ,z)
v(x,y,z)=
. w(x,y,z) .
reads as follows in m atrix notation

dv =

du
dv
dw

d U

dU

du

d X
d V

d V

dZ
dV

dW

dz
dW

^y

dZ

d X
d W
dX

4. 2

dx
(4.10c)

dy

dz

S tra in te n so r

T he sta te of stra in of an elastic body is obtained by subtraction of th e


squared line elem ents of a deform ed an d an undeform ed body. T hus, we ob
ta in a m easure of how th e distances of single points have changed due to a
load [ A .7, A .8 ].
We w rite
dr dr - dr dr = d s 2 - ds2 = ( g ; gj - gj g j ) d* dj =
= ( gy - gy ) d f 1 d f
i = 2 7y

(* U a )

w here 7 y a re th e com ponents of th e s tr a in ten so r.


Accordingly, they can be determ ined as follows

7ij = T ( Sij Sij) '

(AUb)

Expressing th e m etric com ponents of th e deform ed body by those of the


undeform ed body, we obtain th e G REEN - LA G RA N G E's com ponents of
stra in f A.6 1:
+ vjli + v l i vk D-

(4.12a)

L in earized com ponents o f s tra in by neglecting th e quadratic term s in


(4.12a):
T < vi

i)

(4.12b)

30
4 .3

4 State of strain
S t r a i n d i s p l a c e m e n t r e l a t i o n s

- C artesian coordinates
From the tensor of th e displacem ent derivatives V follows as the sym m etric
p a rt the lin e a r stra in tensor. According to th e rules (2.18), it becomes in
C artesian co o rd in ates:
J_ 7
-y
2 ' xz
2 ' xy
J_

VS= | ( V + v T)

'yz

(4.1S)

w here due to ( 4.12b)

II

_ dU
xx ~ d x
dV

dU

yy ~

7xz

<3U

zz ~~ ~d~Z

d Z

+
+

dV

<3X
dw

(4.14)

dX

<3w
<3V
7yz = <3z +
>y

7 ij ( i J = x , y , z ) are th e so -c a lle d technical shears, and


th e norm al strains.

Special c a s e s :
- C ylindrical coordinates ( Fig. 2.5 ) ( r , u ; (p, v; z , w )
<3u
<3r

_ 1 <3U
r d(p
iu
7rz _ ~d~z

<3W
<3Z

v>z

<3V
dr

3w
~dV

_ <3V
<3z

- Spherical coordinates ( Fig. 3.9 ) ( r , u ; i?,v;

1 <3W
r <3y
p ,w )

du
dr
1

d-W

= ---- :---- n I -------- 1r s i n 17 ^ d t p

a ,

U S in 17 + V COS 17 I ,
>

1 d V

~d
1

dU

r sini?

dip

+
'

dr

w
r

1 iu
, dV
7ri> r <3i?
dr

1 ( dv
<3w .
0\
T qI
h ~ FTsin V ~ W COS V 1 .
in v V d tp
dv
/
V

( i = x , y , z ) are

4.5 Compatibility conditions


4 .4

31

T ra n s fo rm a tio n of prin cip al axis

T he principal strain s a re determ ined in analogy to th e principal stresses.


C haracteristic equation according to ( 2.22c) :
A3 -

A2 + I 2 A - I 3 = 0 .

( 4.17a)

T he first invariant corresponds to th e so - called vo lum e d ila ta tio n e :


Ip = e = d i v v = v 3|j =
4 .5

j .

(4.17b)

C o m p atib ility conditions

T he lin ear stra in - displacem ent relations ( 4.12b) form a system of six
coupled, p a rtia l differential equations for th e th ree com ponents Vj of th e
displacem ents for given values of th e stra in tensor. Thus, th e system is
kinem atically redundant. In order th a t there will exist a displacem ent
vector Vj subject to given values of th e six m utually independent com po
nents of th e s tra in tensor, it is necessary th at th e th re e com ponents of the
displacem ent
vector satisfy th e following com patibility conditions ( DE
SAINT V E N A N T ):
Tijlkl + Tki|ij - 7ii|kj - 7kjLi = Tijlki ^ n ,km = 0 .

(4.18)

M echanical in te rp re ta tio n :
T h e interior coherence of th e body has to be preserved after th e defor
m ation, i.e. m aterial gaps or overlaps m ust not occur.
For a two - dim ensional state of stress or stra in th e com patibility condition
in C artesian coordinates reads as follow s:
d2e

d2e
d2 7
f - + ----------------------- = o

(4.19)

ay2

5
5.1

C o n s tit u tiv e la w s o f lin e a r ly e la s t ic b o d ie s


B asic concepts

In th e following we a re going to deal w ith bodies for w hich th ere exist reversibly unique relations betw een th e com ponents of th e stra in tensor and
th e stress tensor, an d we furtherm ore assum e th a t these relations are tim e
in d epend en t. T h e behaviour of th e bodies is denoted as elastic, i.e. th ere are
no perm an en t strain s ep] after rem oving th e load of th e body ( Fig. 5 .1 ). T he
bodies considered shall furtherm ore, as it is usual in th e classical elasticity
theory, be m ad e of a lin e a r ly elastic m aterial such th a t th e ir constitutive
law expresses lin ear relationship betw een th e com ponents of th e stress ten
sor a n d th e s tra in tensor ( range 0 - A - in Fig. 5.1). Such bodies are usually
called H O O K E A N bodies.

5 Constitutive laws of linearly elastic bodies

32

F ig . 5.1: <7 ,e - diagram of a real


m aterial w ith a lin ear elastic range
= lim it of proportionality
= elastic lim it
upper yield point
= lower yield point
elastic - plastic state
= ultim ate stress lim it
= plastic strain

A
B
C
D
D- E
F
epi

For a g re a t num ber of problems in practice th is assum ption is feasible, even


if we have to consider non - lin ear stra in - displacem ent relations ( e.g. geo
m etrical non - linearities for th e post - buckling of plates and shells ).
5.2

G eneralized
H O O K E -D U H A M E L 's
elastic, isotropic m aterials

law

for

7 xy

Tx y

'x z

therm o

C a r t e s i a n co o r di na t es

i r IK XxX - ^ K v + O l + a T e

XX

= ^ r [cr - v (cr + a )] + a n &


yy
e 1 yy
zz
xx

zz H=c If o zz - v i Va

xx

+ a y y ')I1 + a
T (9

xz

(5.1)

G Ty%

with
YOUNG's modulus ,

POISSON's ratio ,
E
2(1 + v)

sh ear m odulus ,
one dim ensional th e rm a l expansion coefficient ,
difference betw een fin al an d in itial te m p e ra tu re .

= T, - T0
Symbolic notation
.

V .with

~l~ 1/
1/
g
S -^ g -sI+ a TI

s = sum of norm al stresses

(5.2)

5.2 Generalized HOOKE - DUHAMEL's law

33

S olving (5.1) w ith re sp e c t to stresses yields


_

i/

i + v ( exx + i _ 2 v e /

a xx -

_ r.
Txy ~~

l + v Veyy + l - 2 v e /

a yy ~

l - 2v Qt

1 - 2v T

_
T ~

l + i/ ( zz + 1 - 2 v e ) _ 1 - 2 v Q t &

7x y

7x*

(5.S)

Tyz = G 7 yz

Sym bolic n o ta tio n


V
,
Vc +

S = 2G
w ith
-

V
T
------ e l 1 - 2v

1 +
1

v
2

Clrr 0 1

(5.4)

e = volum e d ilatatio n .

C u rv ilin e a r co ordinates in in d e x n o ta tio n

According to ( 5.1) it follows th a t


"lid =

Tkl - " E kl T i +

a T Ski = ^ i j k l T

+ a T Ski

( 5.5a)

w ith th e flexibility tensor of fourth order


^ ijk l ~

2E

^ k Sjl 7

S;i Sjk )

Sij Ski

(5.5b)

Solving (5.5a) w ith regard to stresses leads to


( 5.6a)

Ta = Cykl7 k l - l8

w ith th e elasticity tensor of fourth order


^ ijk l

/-< /

G (g

ik _jl

il

jk

g + g gJ + x

2V
ij k l .
-2 - g J g )

(5.6b)

and th e therm o - elastic tensor of second order


J3'J = P g V

E a,T
1 - 2 //

ij

( 5.6c)

O ther notatio n of ( 5.6a) :

Tv = c*iJkl(7kl - aTgkl)
w ith

ik J 1 ,

(5.7a)
il J k '

C*Jk' = A6 gk l + M ( g ' V + e V , )

(5.7b)

34

5 Constitutive laws of linearly elastic bodies

and th e LAME constants


2v
1 - 2v

T he relations betw een th e different specific elasticity constants can be


draw n from th e following ta b le :
A
A

tfj,

/i = G =
(3

,v

A( 1 - 2 v )
2v

(l +

f j, (E - 2f i )
Ev
(1 +

v) { \ - 2v)

E
(l +

+ 2 /i)
+ /i

i/)

A
(A+

i/)(l-2 i/)A

fi)

V
E

- E

,v

A
A

v =

fi, E
E

E - 2 /i
2 /i

T a b le 5.1: E lasticity properties


T he linearly elastic constitutive equations shall be augm ented by another
system of equations which allows a physical interpretation, and which is
applied in elastoplastic structures. Therefore, we split both th e stra in and
th e stress tensor in a sp h e ric a l-sy m m e trica l and a deviatoric p a rt accord
ing to th e following relations :
m

T'k
m

Tk

rm

, m

= y e \ + 7k ,
i 1 rm , m
= y s \ + Tk

^
\I
)

(5.8)

In (5.8) th e know n expressions for th e sum of strain s e = 7 ^ or th e sum of


stresses s = r k occur which are th e first invariants of the stra in tensor or
th e stress tensor according to ( 4.17b). Substituting (5.8) in th e generalized
H O O K E 's law leads to th e following two equations
m
m = k

s-

e=-^(JM T 3 aT

w ith
K

=
= s /3

- -

com pression modulus ,


m ean value of th e norm al stresses
volum etric th e rm a l expansion coefficient .

(5.9b)

5.3 Plane states

35

W ith (5.9a) a change of th e shape w ithout a change of volum e ' 7 ^ = 0 is


described physically according to (5.8), w hereas (5.9b) describes a change of
volume w ithout a change in shape ( ' 7 = 0 > 7 = 1/3 e 8 ). B oth re
lations (5.9) give th e proportionality betw een strains and stresses for lin
early elastic m aterials. It has to be em phasized once again th a t an isotropic,
linearly elastic body only possesses two m utually independent m aterial pro
perties, an d m ost often E and v are chosen.
POISSON's ratio v can be m ore closely lim ited from (5.9b) neglecting all
effects of th erm al stresses
(5.10)

'M

Since e an d crM always have th e sam e sign, v m ust be sm aller th a n 1/2. Ac


cording to (5.10), e = 0 for v = 1 / 2 , w hich corresponds to an incom press
ible m edium ( constant v o lu m e). v 0,3
0,33 is valid for steel and light
m etals.
5 .3

M a te r ia l law for plan e s ta te s

a ) S tate o f p la n e stress
- C artesian coordinates
D e fin itio n :

= T

(5.11)

0 .

yz

S train - stress relations


T? ( ax
= -f^ (c r

V ^vv ) +

yy

- v o

a T

) + a (9

xx'

( 5.12)

yy

yz

Stress - stra in relations

a
yy
o-- =
w ith

E
- v

E
l-i.
0

K x

2~ Lyy +

V yy - ( !

0 ]

t xy

(5.13)
T

= Tj

x ,y

= G 71'xy

( 1 + v )aT &1

^ xx

(x ,y )

V )aT

yz

= 0

- T0 .

Symbolic notatio n of (5.13)


a = E [ e - ee )

(5.14)

5 Constitutive laws of linearly elastic bodies

36
w ith

yy
rxy J

t/

0
0
1

- v

Cq -- CXrp &

yy
L 7'xy
X J

C urvilinear coordinates
T he equations (5.12) and (5.13) read in index notation
I-hvfi-v/T
a p y 6 T
1

7 exp

a p

^ a p -y S

(5.15a)

+ Ctrj, g a p

>.

(5.15b)

( 7 7 - Q T g7 & )

= E

w ith the plane elasticity and th e plane flexibility tensors of fourth order
otfi'yS

1 + vi/ (
2E

E
f
+7 v] )( s

ap 'y6

b)

-l-

^*7 ^/38

oc'y 08

2 ( 1h

&a8 ^/3y '

if

oc6 P'y

+g

if

,
2 is
+ t ^ ; s

ap

8 s.

S ta te o f plane s tr a in in C artesian coordinates

D efin itio n s:

ezz = -y'xz = -y'yz =

(5.16)

Strain - stress relations


1 + t/
E

e 1 + v
yy
E

U 1 " K

*/ 0 y y ] +

[(1 - * 0 % , - ^

(!

xx] + (1

J ' W

+ v ) a T

> 7 x y = - G ^ xy

, 7 XZ = 0 ,

(5.17)

-yl y z 0 .
Stress - stra in relations
a xx = - ( T + T X r ^ ^ l [ ( 1 " " ) xX + " yy - ( 1 + " )T 0<>]
^yy = ~(I + y f ( l -

2 t/ ) [ ( 1

~ ^ )yy + v e xx - ( ! + " K

0] -

Ev
(1 + v ) ( l - 2 v Y
Txy

^ 7 xy

Txz

Tyz

**
0

(5.18)

5.4 Material law for a UD-layer


5.4

37

M a t e r i a l l a w fo r a u n i d i r e c t i o n a l l a y e r ( U D l a y e r ) o f a
fib re rein fo rced c o m p o site

T h e m aterial law for a UD - layer reads as follows according to (5.15) w ith


out tem p eratu re term s
IS'
T

17a 'p 'Y S '

=E

(5.19)

ly s

Here, indices equipped w ith a dash refer to th e m aterial coordinate system


in th e UD - layer.
P lan e elasticity tensor of fourth order for a U D -la y er in th e
o rdinate system ( see Fig. 5.2):

a ."(3* 7 ' 6

(E 1

) =

E 1' 1' 1' 1'


e 2' 2 ' 1' 1 '

2' 2 '
E 1#1<
e 2'2 2 ' 2'

e 1 ' 2' 1 1 '

e 1' 2 ' 2 ' 2 '

[ 1 ~ VEy yl' l/y y


=
1

V\'2' E 2'
- vy y vy i.
0

_ 1' 1 ' 1 ' 2 '


E
_ 2' 2 ' 1' 2'
j
g l ' 2 ' 1' 2 '

vy i . Er
- vy y vy y

e 2.
1

- V y 2, V p y
0

, 2' - co

(5.20)

<*vr .

w here th e m aterial properties have th e following m eaning [ B.10 ]:


E ,,
e

2-

YOUNG's m odulus in - direction parallel to th e fibres ,


Y OUNG's m odulus in ( -d ire c tio n p erpendicular to th e fibres ,
POISSON's ratio p erpendicular to th e fibres in case of a loading pa
rallel to th e fibres ,

F ig. 5.2: M aterial coordinate system for a UD -lay e r

5 Constitutive laws of linearly elastic bodies

38
v 2.y

POISSON's ratio parallel to th e fibres in case of a loading perpendi


cular to th e fibres; it holds th a t
VW

Gr 2 ,

= V1'2' ^ 2 '/ V

Shear m odulus parallel and p erpendicular to th e fibres .

R otation of th e UD - layer by an angle a ( see Fig. 5.3 ) is obtained by appli


cation of th e transfo rm atio n form ula for a tensor of fourth order (g en era li
zation of (2.14) ):
E '"

= / $ / / $ / E - r *v

(S S I)

T ransform ation coefficients according to (3.16) in m a trix notation:

( )

A -

cos a

sin a

- sin a

cos a

(5.22)

Substitution of th e com ponents of th e elasticity tensor by sim plifying the


notation y ie ld s :
g l 'l ' 2 ' 2 '
e 2 ' 2' 1 ' 1 '
E 1' 2' 1' 1'

E1' 1' 1' 2'


e 2 ' 2 ' 1' 2'

2 ' 2 2 ' 2 '

E
g l ' 2 ' 22'

El ' 2 ' l ' 2'

i'i'
E11

E1' 2'

E1'3'

E 2'1'

E2' 2'

E 2'3'

E 3'1'

E 3' 2'

E 3'3'

(5 1 3 )

Com ponents of th e elasticity tensor for th e ro tated vector base re a d th e n as


follow s:
- ,1 1

,2 2

-3 3

- ,1 ' 1 '

=E

,2 '2 '

=E

_ 3 '3 '

= E

- ,2 '2 '

cos a + E
4

_ 1 '1 '

cos a + E
,

1 . 2 '

+ -rA
4

. 4

sin a
. 4

sin a

+
,

.1 '

2_

.1 '

2 .

A sm

+ A sin

2a ,
2a ,

(5.24a)

. 2_

sm 2 a ,

F ig . 5.3: R otation of a UD - layer


by an angle a

39

6.1 Basic term inology and assum ptions

F ig . 5.4: C om ponents of th e elasticity tensors vs. th e fibre - orientation a


of a High T ensile fibre ( H T - fib r e ) [ B.10 ]
M aterial values: E 1# = 143 0 0 0 MPa , E2, = 5140MPa ,

E 12 == E 21
E 13 == E 31

II
BM
(O

G r 2 , = 5 280MPa , v V2, = 0.28 , v V l, =

= 4 l-

A 1' -= E2'1' +

A2' sm
* 22 q ,
A

E1'1' + A1' + A

2'

sin a ] sin

(5.24b)

. 2' . 2 I
E 2'2' - A 1' - A sin a j sin 2 a

E 23 == E 32

w ith

0.01

E3

'3'

> a 2' = E 1' 1' + E 2'2' - 2 A 1'

E n e r g y p r in c ip le s

6.1

B a sic te r m in o lo g y an d a ssu m p tio n s

O ur consideration of solid bodies in th is section is based on th e following


assum ptions [ A.9, A.15, A.16, A.18]:
a ) T he processes produced in a stressed body are reversible, i.e. no dissipa
tive effects ( e.g. plastic d efo rm atio n s) occur. We lim it ourselves to the
scope of th e classica l elasticity theory.
b ) T he deform ation process takes an iso th e rm a l course, i.e. th ere is no in
teractio n betw een deform ation and tem perature.

40

6 Energy principles

c ) T he load process is quasi - static, i.e. th e kinetic energy or the forces of


in e rtia can be neglected.
d ) T he state of displacem ent of a solid body is described according to a
LAGRANGEAN approach.
e ) T he theorem of m ass conservation ( dV = dV ) and th e volum e forces in
the deform ed and undeform ed bodies ( f = f ) a re equal.
6 .2

E nergy expressions

F irst, we consider th e uniaxial state of stress of a rod subjected to a single


force F. T h e relation betw een force and displacem ent can be assum ed to be
nonlinear as well as lin ear ( Fig. 6 .1 ). T h e external work done by th e nor
m al force F against th e displacem ent <5u is given by

( 6.1)

6 W = F(5u .

Here, we use the differential 6 for the changes of state, e.g. deform ation dif
ferentials, strain differentials. For these quantities it is assum ed th a t they
are virtual ( not existing in reality ), infinitesim ally sm all and geom etrically
com patible. Eq. (6.1) illustrates the area of a th in strip w ith the w idth 6 u
and th e height F in a force-deform ation diagram (Fig. 6.1), where term s of
higher order have been neglected. T he total work of th e single force results
from an integration over th e deform ation differentials
U

W =

( 6.2)

F<5u

iup
,r

~ i ->
dV

dx

a)
Fig.

6.1 :

N onlinear and linear force - deform ation curve of a rod subject to


a single load

6.2 Energy expressions

41

In Fig. 6 .1 , th e a rea W* represents th e com plem entary work, because W and


W* co m p lem en t one an o th er and th e ir sum is represented by th e rectangle
F u = W + W*.
By analogy to (6.1) an d (6.2) th e following holds for th e com plem entary
work
6 W* = u<5 F

(6.3)

or
F

(6.4)
F= 0

In th e case of a lin ear force - deform ation curve F = c - u (Fig. 6 .1b ) an in


tegration over th e deform ation differentials can be carried out. Thus, we ob
ta in

= W * = | f - u .

(6.5)

T he ex tern al work is stored as so - called in te r n a l energy or deform ation


energy in th e rod. Substituting th e increase of deform ation 6 u by <5edx in
( 6.1), we can w rite ( u is denoted by u now, because it cannot be changed
w ith th e final value u ):
6 W = F(5u = -^-A(5edx = (7 (5 e A d x = (7 (5 e d V = 6 U
A

( 6.6)

If we divide by th e volum e elem ent dV = Adx, we obtain the expression for


the specific d e fo rm a tio n energy
6U = a6e

(6.7)

and by analogy, for th e specific co m p le m e n ta ry energy we obtain


<SU* = eSc t

( 6.8)

T he relatio n betw een th e stress a and the stra in e is given by a non - linear
curve sim ilar to th e one shown in Fig. 6.1a. If a lin ear a, e - curve exists, by
analogy to (6.5) we obtain th e following for the specific deform ation energy
and the specific com plem entary energy
U = U* = - i- a e

(6.9)

T he expression is now extended to a three - dim ensional elastic body sub


jected to ex tern al forces (volum e forces f, distributed surface tractions p,
and concentrated forces F k ) ( Fig. 6.2 ).

42

6 Energy principles

F ig . 6.2 : E lastic body subjected


to external forces
In vector notation, the external work can be w ritten as follows
SW =

( 6.10)

|YT <5vdV + |p T <5vdS + F T <5v

w ith
-T

vector of volume forces ,

P T = ( P x . Py , P z )

vector of surface tractions ,

displacem ent vector of an elastic body ,

FT = ( F ^ ,F j ,...,F f )

vector of concentrated forces


= (F x.Fy ,F ,)i ,
vector of displacem ent vectors for points
of action of concentrated forces

o
V =

,o

w
T r a n s itio n to isotropic, lin e a rly elastic body
Specific deform ation energy and com plem entary energy
TT

TT>k
I
t
U = U*
= yI f f t e = -^e
a.

( 6.11)

Introduction of H O O K Es law
U =

l eT C e ,

(6.12a)

U* = y trT D t r

(6.12b)

6.2

Energy expressions

43

w ith vectors for th e strain s and stresses in C artesian coordinates defined by

( x x y y zz > ^ x y T y z ^ z x )

( x x y y z z *^ x y ^ y z T z x )

leads to th e e la stic ity m a tr ix


1-1/

1-1/

l-i/

E
( l + (/ ) ( 1 - 2 . )
0

I
|
I

II 1 " ^2 "

" i "

= C

(6.13a)

- 2v
2

and th e fle x ib ility m a tr ix


1

- V
1

1
1
1

0
*0
II

D = i

1,
1

11
1*

(6.13b)

2 ( 1 + 1/ )
0
0
0
2(1+1/)
0
0
0
2(l+i/)

T he expressions (6.12) are b ilinear forms which are positive definite be


cause U > 0 and U* > 0 .
In usual in d e x n o ta tio n
U

= U * = - ^ T ij 7ij

(6.14)

(6.15a)

U + = l D i j kl r 1Jr k I .

(6.15b)

Consideration of thermal influences

u = y Tj ( "Tij - QTSij ) = } c 1]kl 7ij 7kl - p Ti\e ,

(6.16a)

U* =

(6.16b)

j T1J( 7 ij + a T g y 6 ) = i - D ij k lr u r kl + a T r \ .

In (6.16), the quadratic tem perature term s are neglected [ B.ll.

44
6 .3

6 Energy principles
P rin c ip le o f v ir tu a l d is p la c e m e n ts ( P v d )

T he v irtu a l work <5W of ex tern al forces is equal to th e increase of v irtu a l


stra in energy <5U according to (6.6):

Jt *6 7^ dV

S W = <SU =

(6.17)

v
W ith th e stra in energy U =
U

in te r n a l poten tial IIj

J l l d V = IT

(6.18a)

and th e work of the e x te rn a l forces ( here w ithout concentrated forces ) of a


conservative system equals th e negative of th e p o tential IIe of th e ex tern al
forces
W = J f Vj dV + J p Vj dS = - IIe ,
v
s

(6.18b)

th e to tal p otential of th e elastic body reads as follows

n = n;+ ne.

( 6.19)

P rinciple of statio n arity of th e v irtu a l to tal potential


<511 = <5(11; +

n e) = 0 .

(6.20)

T his im plies an extrem um value of th e to tal potential


II = IT + n e = extrem um

( 6.21a)

G REEN - D IR IC H L E T 's principle of a m inim um (v a lid for linearly elastic


behaviour of m a te r ia l) [ A.8 ]
II = IT -l- I le = m inim um .
6 .4

( 6.21b)

P rin c ip le o f v ir tu a l fo rc e s (P v f)

T he com plem entary v irtu al work 6 W * of the external forces is equal to the
increase of th e com plem entary v irtu al energy <5U*:
6 W * = 6U* = J ^ ij <5rij dV .

(6.22)

T otal com plem entary p o tential follows by analogy to (6.19)

n* = nf + n*

( 6.23)

6.4 Principle of virtual forces

w ith in te r n a l com plem entary potential

45

U* = JlJ*d V = II1*
v

(6.24a)

and e x te r n a l com plem entary potential II* defined as th e negative of the


ex tern al com plem entary work W*
W * = J v jf* dV + J v j p dS =
v
sv

-n*

(6.24b)

P rinciple of sta tio n a rity of th e v irtu a l to tal com plem entary potential

<sn* = <5(n i* + n*) = o .

(6.25)

T his im plies an extrem um value of th e to ta l com plem entary potential


II* = II* + II* = extrem um .

( 6.26a)

T he CASTIGLIANO and M ENABREA principle ( valid for linear - elastic


behaviour of m a te r ia l) [ A.9 ]
II* = II* + II* = m inim um .

F irst theorem by CASTIGLIANO

( 6.26b)
v- = ^ H ( F )

/ s.27a)

SF1

T heorem by M ENABREA

= 0 ,

(6.28)

c) ( F )
w here th e index R refers to th e reaction forces.
Second theorem by CASTIGLIANO

i
f iU f v .)
F =

(6.27b)

G eneralized variatio n al functional by HELLINGER and REISSNER [ A.5,


A.17, A.18, A.19 ]:
n R = J { H (7 ij) - f 'v j +

+ v j . ) - 7 ij]} d V -

(6.29)
J P q Vjd S + J ( v i0 - Vj) p1dS

where p0 , v ;o are prescribed loads or displacements at the boundary.


HELLINGER REISSNER stationary principle:
tf n R = 0 .

(6.30)

46
6 .5

6 Energy principles
R e c i p r o c i t y t h e o r e m s a n d U n i t L o a d m e t h o d

T heo rem by B E T T I

In words: If two sets of loads are acting on an elastic body, th e work of


loadset 1 against th e displacem ents due to loadset 2 is equal to
th e work of loadset 2 against th e displacem ents due to loadset 1.
If displacem ents a t points 1 an d 2 are expressed by M AXW ELL'S influence
coefficients <L, th e T h eorem by M A X W E L L follows proving the sym m etry
of th e coefficien ts:
(6.31b)
In words: T he displacem ents at a point i due to a u n it load a t another
point j is equal to th e displacem ents a t j due to a u n it load at i.
( I t is assum ed th a t th e displacem ents of the points are m easu
red in th e directions of th e applied fo rces.)
U nit-Load m ethod
T he U n it-L o a d m ethod plays an im p o rtan t role in elasto-m echanics. By
m eans of th is m ethod, th e deform ations of an elastic body at a certain
point can be calculated [ A.18, A.19 ]:
v irtu al static group
(6.32)

real kinem atic group

6 .6 T r e a t m e n t o f a v a r i a t i o n a l p r o b l e m
*

a tta in s an extrem um value. T his im plies statio n arity of I

<51 = 6 J F ( x , y , y ' , . . . , y ( n ) ) d x = 0

(6.34)

For an integrand function F containing only derivatives up to th e second


order of th e unknow n function y = y ( x ), follows from p a rtia l integration:

6.7 Continuous approximation methods

<51 = J ( 1 F _ d d F
J Vdy
dx 5 y '
xi >

47

j d ^ d _ \ g dx
dx2 ?>y" >
-

==> E U L E R s eq u atio n

(6.S5)
+

==>

( dF

iF

I 2

y w ~ t e W T) 6y\ +
'Y J
resid u al ( physical )
b o u n d a r y c o n d itio n s
J

dF
^ y

j-

6y

= 0.
1

1
,. , / '
. \
e sse n tia l ( g eo m etric )

boundary conditions
6.7

C o n tin u o u s a p p ro x im a tio n m e th o d s

T he following approaches belong to the group of continuous m ethods [A .3]


( as opposed to discrete m ethods like th e F in ite E lem ent M ethod, or the
Boundary E lem ent M ethod).
1) M ethod by R A Y L E IG H - R I T Z [K M ]
Point of reference
* variatio n al expression (6.33)
x 2

1 = J F ( x ,y , y ' ,.. . , y ^ )d x

E x tre m u m .

xi

Choice of a set of linearly independent approxim ation functions


N

y * ( x ) = 2 * n y n* ( x ) >
n= 0

( 6-S6)

w here th e y* m u st a t least satisfy th e geom etric boundary conditions.


D em and of a m inim um :
61
d a_
A ssum ing a q u adratic form of the functional, th is leads to a linear system
of equations for determ ination of th e coefficients a n .
2 ) M ethod by G A L E R K IN [ A.6, A.16 ]
Point of re fe re n c e
> variational functional in th e varied form according
to (6.35).
Choice of a function y * ( x ) in analogy to the RITZ approach (6.36). Func
tions y *( x ) m ust fulfill a ll boundary conditions.
D em and of a m inim um :
Fulfilling of th e GALERKIN equations
x2

J L (y * )y ^ d x = 0 ,

(n = 1 ,2 ,3 ,.,.,N )

xi
w ith L ( y* ) as th e differential equation for th e problem ( see (6.35) ).

(6.38)

48

7 Problem formulations

This leads to a system of linear equations w ith respect to the coefficients


an if a quadratic form of th e functional ( lin ear differential e q u a tio n ) is
assum ed. In th e case of functionals of higher order th a n quadratic, we get a
non - linear system of equations.

7 P r o b le m f o r m u la t io n s in t h e t h e o r y o f lin e a r
e la stic ity
7.1

B a s i c e q u a t i o n s a n d b o u n d a r y v a l u e p r o b l e m s

- three equilibrium conditions (3.28a) ,


- s ix stra in - displacem ent relations (4.12) ,
- s ix equations of th e m aterial law (5.5) or (5.6) ,
i.e. altogether 15 equations for 15 unknow n field quantities (6 stresses t j, 6
strains
, 3 displacem ents v; ).
Problem of elasticity th e o ry : solution of basic equations w ith given bound
ary conditions
> b ou n d a ry - value p roblem s.
W e distinguish betw een three kinds of b o u n d a ry -v a lu e problems'.
- F irst b o u n d a ry -v a lu e problem
On the to ta l surface S of a body B, th e tra c tio n s t J are given (F ig. 7.1a).
T he following is valid for th e com ponents of th e stress vector
t(S) = ( TJ n! )s

- Second bou n d a ry -v a lu e problem


T he displacem ents of th e to tal surface S of th e body B are given (Fig.
7.1b ). On th e surface th e following displacem ents are given:
vi(S) = ( vi)s

(7.2a)

- M ixed b o u n d a ry -v a lu e problem
On one p a rt St of th e surface S of th e body B, th e tractions are given, and
on th e rem ain in g p a rt Sd of the surface th e displacem ents are given ( Fig.
7.1c ). T he boundary conditions th en read

F ig . 7.1: Illustration of b o u n d a ry -v a lu e problems

7.3 Special equations for three-dimensional problems

49

< k ) = <TS n i ) s , .
vi<Sd> = ( vi>sd 7.2

( >

S o lu tio n s o f b a s ic e q u a tio n s

L A M E - N A V IE R 's e q u a tio n s - solving w ith regard to th e displacem ents


u,

i .

A v + e
A + fi

|i

3A +

2n

--------------
A + fj.

A + fi

f = 0

(7.3)

w ith th e LAM E constants p , A according to ( 5.7b) a n d th e volum e d ila ta


tion e according to (4.17b).
T hese are th ree coupled, p a rtia l differential equations of second order for
th e three unknow n displacem ent com ponents v1.
B E L T R A M I - M IC H E L L 's equations - solving w ith regard to the stresses
A t* +

^ - s | ij

(7.4)

*
g* A s = 0
1+ v

w ith s as th e sum of th e norm al stresses according to ( 3.13a).


These are six coupled, p a rtia l differential equations of second order for th e
six unkow n stress com ponents t 1j.
7.3

S p e c ia l e q u a tio n s fo r th r e e - d im e n s io n a l p ro b le m s

Solved w ith respect to d isp la cem en ts [ A.5, A.10, A .ll, A.12]


Use of th e LOVE function x leads to
AA x = 0 .

(7.5)

T his bip o ten tial equation has an in fin ite num ber of solutions, e.g. feasible
solutions in cylindrical coordinates are for th e axisym m etric case [ A.9 ]
X = r 2 , In r , r2 In r ; z , z , z3 ; z In r , R , i , In ^
K
H
w ith

, z In ( z + R )

(7.6)

R = / r2 + z2 .

All linear com binations of (7.6) w ith a rb itrary constants a re solutions of


(7.5) as well.
D isplacem ents in th e axisym m etric case

50

7 Solution procedures

Stresses in the axisym m etrical case

o <p<p =

1 - 2i/
2 -1 /
1 1
1 -

7.4

( 7.8a)

2 v dZ

1 -

2i/

i/
2

d
dZ

(7.8b)

(7.8c)

dZ
d

* - V dz*
y -

(7.8d)

j j L

1 - " a z2

d r

S p e c i a l e q u a t i o n s fo r p l a n e p r o b l e m s

a ) S tate o f p la n e stress
- Solved w ith regard to displacem ents
N A V IER's equation
a \/3
1 + 1/ J 3 \ a
1 + 1/
^ila , 1 (O
M ^ + T ^ r ^ l / 3 2T ^ r a T 0 I
g~
=

r.

(7.9)

Coupled system of two p a rtia l differential equations for th e two unknow n


displacem ent com ponents v.
Introduction of a displacem ent function S'
= S f .

(7.10)

For vanishing volume forces f th e POISSON's equation is obtained from


(7.9)
y f p = AS? = V V = (1 + u ) a r e

(7.11)

w here S' is called th e th erm o - ela stic d isp la cem en t p o te n tia l [ A.13].
- Solving w ith regard to stresses
a il

B v _

761

ap-y6T Imi/ + a T

0 _ |6

(7.12)

Introduction of A IR Y 's stress function in (7.12) assum ing conservative vol


um e forces ( f = - v | ) [ A.4, A.8 ]
t

y6

= e

fir x|
, ir
e $ \OT + V g7

provides th e bipotential equation


S>

(7.13)

7.5 Comparison of plane states

or
b)

51

(7.11)

AA $ = - E a T A - (1 - u ) A Y
S ta te o f p la n e s tr a in

- Solved w ith regard to displacem ents


Analogous to (7.11)
A

- -j +- ~ a T e
1 -

(7.15)

- Solved w ith regard to th e stresses


In analogy to (7.14)

7.5

1- 2u
1- v

E
1 - i/

AA$ =

C o m p a ris o n
s tra in

of s ta te

o f p la n e

(7.16)

s tre s s a n d

s ta te

o f p la n e

Since m any problem s can be described in C artesian coordinates, we would


like to list th e notations for both two - dim ensional states in these coordi
n ates again.
State of plan e stress

State of plan e strain

T he stresses m ust fulfill th e equilibrium conditions by analogy to (3.27a)


do

dr

dx

dr
d X

dy

+ fx = 0 ,
x

do
+ V- + f = 0 .
d y
y

(7.17)

T he boundary conditions (7.1b, 7.2b) can be given in a m ixed form


(y o x x cos a +

xy

sin a yIS
)L = t xO ,1
(7.18a)

(' o y y sin a -I t x y cos a -MS


)L = t yO
or

Ur- = u

Vr. -= v .

Here, equations (7.18a) are valid at


acting w ith th e com ponents tx0 and
which boundary displacem ents a re
(7.18b) is valid. T hese quantities, but

(7.18b)

points on S, where external loads are


t 0 p er surface unit. For points on S at
given by th e com ponents u 0 and v 0,
also th e tractions may equal zero.

52

7 Problem formulations

For 0 = 0, th e m aterial laws (5.12) and (5.17) read


= ^ - [ ( 1 - " 2 K x ' ( ! + V ) a yy} ,

ey y = i ( <7y y ~ l/<7x x ')'


T'xy

yy

g T*y

i/2 )<7

- i / ( 1 + v ) 0 xx\ ,

(7.19)

Q Tx y

7xy

an d th e corresponding com patibility condition according to (4-19) follows to


be
.2
b 7
d2 e
d 2 e.
JC L
(7.20)
+
a x dy
One should pay atten tio n to th e fact th a t 7 xy is th e tech n ica l sh e a r s tr a in
w hich differs from th e tensorial quantity 7 qj6 by a factor of 2!
Substituting th e m aterial law (7.19) into (7.20) leads to
A (V a x x + a yy )' =

A (' a X X + a V V ') =

(7.21)
df

-(1 +/)(

dX

= +

_i
(
1 - u )) 1V a x

_____X

^>y )

According to this, th e sum of stresses as th e first in v arian t of th e stress


state is a harm onic function. If th e ex tern al load is know n at th e whole
boundary, th e stresses can be determ ined from (7.17), (7..18), an d (7.21)
w ithout considering th e displacem ent field.
By introducing A lR Y 's stress function $ according to (7.13)
d2 $

d y '1

+ V

J,J'

ax'

+v

r xy
^

a2 $
axay

(7.22)

th e equilibrium conditions (7.17) are identically fulfilled w ith

^a x

= -f

^V
b^ y = - f y

From (7.21) result


A A $ + (1 - v ) A V = 0

AA<2> + 4 A V = 0

(7.23)

If no volume forces are present, th e biharm onic equations (7.23) take the
sam e form for both states
AA< =

bA $ + 2 _ 4 *
2
2
^x4
a x <ay

0 .
^y4

(7.24)

A .2 E x e r c ise s
E x e r c i s e A -2 -1 :
An oblique base gj ( 1 = 1 , 2 , 3 ) expressed by th e orthonorm alized base
vectors e i ( i = l , 2 , 3 ; | e i | = 1) is given:
51 = e i -

5 2 = el + e2 i
63 =

e i +

e 2

+ e 3

a ) D eterm ine for th e vectors


x = 2et + 2 e2+ 2 e3
th e scalar product x
A-l )

,y = - e j

- e2

y , | x | , and th e angle

+ 4e3
ip betw een th e vectors (Fig.

F ig . A - l : Oblique base w ith two


vectors x an d y

b ) State th e contravariant m etric tensor an d th e co ntravariant base vectors,


c)

Apply th e vector a = g , + 2 g 2 + g 3 to th e contravariant base g .

d ) D eterm ine th e physical com ponents of th e vector a .


e ) For th e given transfo rm atio n m a trix betw een two bases
1
1
1

-1 -1
0 -1
0
1

( i ' = 1,2,3),

express th e o ld base g^ in term s of th e new base g i ; , and vice versa.

54

2 Tensor algebra and analysis

S o lu tio n :
a ) We first determine the required quantities in the orthonormalized b a se :
x = 2 e j + 2 e 2 + e 3 = x1 ej

x1 = 2

y 1 = - 1! ,

y = - e, - e 2 + 4 e 3 = y ej

x2 = 2

x3 = 1 ,

y 2 = - 11 ,

y3 = 4A .

Scalar product in orthonormalized base (2.2a) ( gjj >


.
x - y = x i y..J ei
ej = x y

+ x2y2 + x3 y3 = 2 ( - l ) + 2 ( - l ) + l - 4 = 0 .

8ij

Length of vector x (2.2b)


| x | = -/x - x = -Jx x3 e ; ej = -Jx x 3!+ =
= / ( x 1)2 + (x 2)2 + (x 3)2 = / 4 + 4 + 1 = / 9 = 3 .
Angle b e tw e e n v e c t o r s x a n d y (2.2c)
X1y J?Sy

1 1

2 2

x y + x y

cos <p =

/ x 1xJ Sy / y k y 's kl

3 3

+ x y

/ ( x 1)2 + ( x 2)2 + (x3 )2 / (y1)2 + (y2 )2 + (y3f

= 0

<P= 2

3-/l8

i.e.

y 1 x .

Transformation into the oblique base then yields :


el= S i
e 2 82 _ 8 l

e 3 = 83 S 2

Vector in covariant base


x

- g 2 + g 3 = x 8i
-J

y = - 5 g 2 + 4 g 3 = y gj

= 0

y1= 0

x2 = 1

y3=

= -5

Covariant metric components according to ( 2.1a,b)

8i ' gj gy

(g ij) =

Sn

Si 2

g 13

*12

S22

S 23

g 13

g 23

g 33

= el el = 1 - 1 = 1

gj2 = g i g 2 = e i ( e i + e 2 ) = 1 >

813

= g l " g 3 = e l ( e l "b e 2 "b e 3 ) = ^ >

= 1,
4

Exercise A-2-1

622 = g 2

82

= ( e i

+ e 2) =

+ e 2 ) ( e i

623 = 82 83

=( e l e 2 ) ( e l + e 2 + e 3 ) = ^

g 33 = g 3

83

=( el

The metric tensor then

reads

(g ij) =

1 1
1 2

1
2

+e 2 + e3

1 +

55

1 =

2 >

e3)

= 1+ 1 + 1 = 3

^= ^ >

) (e i+

e2 +

(1 )

Formulation of the s c a l a r p r o d u c t for the covariant base yields due to ( 2.2a):


x

y =

-1 -j
g;jx y =

1 j
_2_j
_3_j
gjj* y+g2jx y +g3j x y =
_ 1_ 1
_1_2
_1_3
gnx y+g12x y +gi3x y +
21
_2_2
_2_3
+ g2i x y+ g22x y + g23x y +
+ gji^3y1+ g32x 3y2 + g33x 3y3 =

0 - 0 + 1- o - ( -

+ 1

0 - 4 +

+ 1 1 - 0 + 2 1 - ( - 5 ) + 2 -1 4 +
+ 1 1 0 + 2 1 (-5) + 3 1 - 4 =
- 1 0 + 8 - 1 0 + 12 = 0 .
L e n g th of vector x in the covariant base

I2

XI =

j _________ _1

8 l l X X + gl2 X X +
_3_1
_3_2
+ g 3J X X + g 32 X X

0 + 0+ 0 + 0 + 2 / 9

l-

g l 3 X X + g21 X
-3 -3
+ g 33 X X =

+ g22X X

l + 2- l - l + 0 + 2- l -

-2

+ g23X

l +3- l -

l = 9;

herewith, the length is proved to be an invariant .

In oblique base the a n g le <p follows from ( 2.2c)


- 1

-J

gijx y

cos <p =

m n

/ g ki x k x l

,y

3 /1 8

b ) The contravariant base can be calculated by two methods :


F irst way :

According to (2 .5a )

g 1 = g 1Jgj

and to ( 2.5b)

g. =

g.. g J .

56

Tensor algebra and analysis

The scalar product of the base vectors is determined by ( 2.5c)


.i
ij
il
k 8 6kj 8 Ski

i2
8

i3
Sk2

Sk3

and further
.1
11
k 8 Ski
2
k

12

13

Sk2

Sk3 >

21
22
,2 3
8 Ski ^ 8
Sk2 ^ 8 Sk3

_3
31
,3 2
, 33
k 8 Ski + 8
Sk2 + 8 Sk3

g
g
g

11

12

12
22

13

13

23

23
33

8 11

Si 2

813

812

822

823

813

S23

833

0
1
0

Forming the inverse of ( g~ ) yields the contravariant matrix ( glJ ):


1 1
1 2
1 2

( s ij) = ( S i j T 1

1
2
3

-1

From the rules of matrix inversion follows

( I-1
1
(8 ii)
=

2 2
2 3

1 1
2 3

1 1
2 2

1 2
1 3

1 1
1 3

1 1
1 2

1 2
1 2

1 1
1 2

1 1
1 2

The determinant is

8 = 8 , J

4 -3

= 6 + 2 + 2

The contravariant metric tensor now becomes


2
-1
0

(g ij) = (g ijr 1 = |

0
-1
1

-1
2
-1

From ( 2.5a) follows herewith


g
g

g gj

= g

11

12

13

g i + g g 2 + 6 g3 = 2 g i
12
, 22
, 23

S2

S g t + g g 2 + g g 3 = "gi + 2 g 2 - g 3 ,
3

g =g

13

23
g l + g

33

g2 + g

g3= - g 2 + g 3 -

0
0
1

Exercise A-2-1

57

The contravariant base can be expressed by the orthonormalized base as


l

C heck:
According to ( 2.3a)

gj gJ =

with ( 4 ) yields
g1 = ei ( ei - e2) = 1

El

or with ( 3 ) and ( 2.5b) with ( 1)


El

E 1 = ( g 1 + g 2 + g 3 ) ( 2 E i - E 2 ) = 2 S i S2 = 2 ~ 1 = 1

El

g 2 = e j ( e 2 - e 3 ) = ( g 1 + g 2 + g >) ( - g 1 + 2 g 2 - g 3 ) =

- S j + 2g| - S3 = - 1 + 2 - 1 = 0 .
Second w a y :

The contravariant base vectors can be calculated by forming the vector products
according to (2.6):
E

82

X8 3

2
7

S i >S 2 E 3 ]

[ E 1. E 2183J

1
1

= S2 x S3 =

83 x gj

8 ] x E2

[ 811 E 21E 3 ]

[ S i j E 21E 3

= 1

ei
1

e2
1

e3
0

= gj - ( g 2 - g j ) = 2g j - g 2 ,

58

Tensor algebra and analysis

= S3 x Si =

e l e2 e3
1 1 1 = e2 1 0 0

= g 2 - Si - ( g 3 - s 2) =

= - g t + 2g2 - g 3 ,

g = *1 x S 2 =

el e 2 e3
1 0 0 =
1 1 0

e3

= g3 _ g2

According to ( 2.4b), the contravariant metric components become


gJ = g g*

g = K K = ( 2 g j - g 2 ) ( 2 g l - g 2 ) = 4 g u - 4 g 12 + g22 = 4 - 4 + 2 = 2 ,
12
1
2 /.
w

X
g = g ' g = ( 2 g t - g 2) ( - g j + 2 g 2 - g 3) =
= ~ 2 gu + 4 g12 - 2 g13 + g12 - 2 g22 + g23 =
= - 2 - 1 + 4- 1 - 2 - 1 + 1 - 2- 2 + 2 = -1 ,
etc.
c ) The vector a in the covariant base
a = g x + 2g2 + g 3 = a'gj

( &1 1 ,

2 , a =1)

shall be applied to the contravariant base.


Analogously to ( 2.5b) it holds that
a.1 = &ij
e -a

ai = 8 n a + gi2 a + gi3 a
1

aj = l - l + l - 2 + l - l = 4 ,

a2 = 8 i 2 a + g22 a + g23 a

a = 1 1 + 2 - 2 + 2 - 1 = 7 ,

a3 = Ei 3 a + g23 a + g33 a

a3 = l - l + 2 - 2 + 3 - l = 8 .

Therefore, the vector in the contravariant base is


i
a = aj g = 4 g

_
+

3
7g+ 8g .

d ) The physical components of the vector according to (2.10) become


a

1/
= a Vg(ii)

or

*
/ (i i)
a* = a iVg

With the vector in co- and/or contravariant base


a = g x + 2g2 + g 3 = 4 g + 7g + 8g ,

Exercise A-2-1

59

it then follows that


a* = 1 T in = 1A " = 1

a* = 4 / g i i = 4 / 2 ,

a*2 = 2 A 7 2 = 2 A

a* = r y g 22 = 7 A

**3 = l l / l h 3 = A

a* = 8 A

= 8

= 8A

e ) The new base g ; , is expressed by the old base g j by means of the


transformation matrix

p f .

1
1
1

-1
0
0

-1
-1
1

i' = 1 , 2 , 3

new base

For this, we use ( 2.7a)


S i - = P-, gj

gj > P jr g i + Pi / g z + Pi* g 3
1

gi

g2

g3

el

e 2

e 3

g 2' = P2 - g l + P2 - g 2 + P2 - g 3 = g l - g 3 = e 2 ~ e 3 >
3' = ^3 ' 1

^3' g

2 b

^3 ' 3 = 1 b 3 = ^ e l b

e2 + e 3

The old base can be expressed by the new base as


gl

1
1
"+ g2' + "o~g 3

1' b g 2 ' >


1

g3 "2"2' b"2"g3'
The transformation matrix then follows to be
k'

V2

V2

-V 2

V2

In order to check this matrix, we form according to (2.8)


-1
0
0

-l
-1
1

0
1

V2

V2

0 -V 2

V2

1
0
0

0
1
0

0'
0 = S:.
1

>

60

Tensor algebra and analysis

E x e r c i s e A -2 -2 :
In order to d eterm ine th e state of stress in a th in p arallelogram disk ( Fig.
A -3), a tran sfo rm atio n in an oblique coordinate system
( a = 1 , 2 ) is ad
vantageously carried out for establishing th e required equations.
a ) D eterm ine th e base vectors, th e co- an d th e contravariant m etric tensor,
and th e CH RISTO FFEL symbols.
b ) How does th e bipotential equation
= 0 (<? = A IR Y 's stress func
tion ) read in th e oblique coordinate system , and how do we determ ine
th e associated stresses from it ?

F ig. A -3 : P arallelogram disk

S o lu tio n :
a ) According to Fig. A-4 , we express the relation between C artesian and oblique
coordinates as follow s:
x 1 = Ic 1 +, Ic 2 cos a ,
2

x = !; sin a

, ( a = const ) .

Base vectors
According to ( 2 .26 ) , for the coordinates in the plane the relation between a base
g Q ( 5 ) and the orthonorm alized base e Q is

_ dx

ea '

F ig. A- 4 : C artesian and oblique coordinate system

Exercise A-2-2

61

Herewith, the base vectors are calculated as follows:


g1

dx

dx

df;1

1 ^2

^1

df;1

dx1

dx2

2 = d ? " * 1 + d ? " * 2 = COSOel + s m a e 2


C o varia n t m e tric ten so r
According to / 2.1a), the covariant metric components become
&ct/3

ga

gp

812

^21

= gl

g 2 = el

( c o s a e j + s i n a e 2 ) = cos a

g 22 = g 2 ' g 2 = cos a ( e ! e ! ) + sin a ( e 2 e 2 ) =


regarding

e2 =

Thus, the covariant metric tensor is


1

cos a

cos a

(2 )

As ( 2 ) shows, the metric tensor of a non-orthogonal base is completely occupied.


Determinant of the metric tensor according to ( 2.1c)
8 ~ | 8a/j I = 1 " cos2 = sin2 ~ const '

(3)

C o n tra v a ria n t m e tric tensor


The contravariant metric component g^ can be determined according to ( 2.5c) by
inversion of the covariant metric tensor ( gap ) .
This yields
U P ) = (g a/g )"1 = - t V

- cos a

-cos a

(4 )

CHRISTOFFEL symbols of the first and second kind from (2.30) and (2.31)
0)07

8 q^3 = const .

r c*/3 - 0

(5 )

This is always valid for rectilinear coordinates.


b)

According to ( 2.4O), the following expression is valid for the bipotential


operator A A C>:

/r

.p

62

2 Tensor algebra and analysis

W ith ( 3 ) the bipotential operator is simplified to

AA=gV
AA = g

11

12

+ g

11

+ 8

11

lm + g g
. 1211 + 8

22

11 _

im + g g

2111 + 8

,1212 + g

f2 U2 + g

12

22

12

1221+g

g
22

g 2121 + 6 6
21 ,

22

21

21

22

,2212 + g

1121 + g

11

21

21

21

12

12

22

g ,2211 + g

11

12

21

11

+ g

12

12

11

21

(6 )

22

,2221 + 8

1122

. 12 22 +
,2122 +

22 ^

,2222

By using the sym m etry g ^ = g ^ and SCHW ARZ's perm utation equation we
can further sim plify the fully expanded expression in ( 6 ) :
*

AA = g

11 11 .
, . n 12
, . 12 12
g l l u + 4g g >1112 + 4g g >1212 +

11 22 .
g

+ 2g

12 22 ^

,1122 + 4 8

22 22
,1222 + g

,2222

Application of ( 4 ) yields
A A ffi =

[ >lm + 4 ( - cos a ) m 2 + 4 cos2 a m 2 +

+ 2 ,1122+ 4 ( ~COSa) ,i222 + ,222 2 ] - 4


^

= [ ,1111 - 4 cos a f1U2 + ( 2 + 4 cos a ) l m

4 c s a |1222 + ,2222 J . 4

The bipotential equation then reads


,llll - 4 cos a 1U2 + 2 ( 1 + 2 cos2a ) U22 - 4 cos a 1222 + 2222 = 0 . ( 7 )
W ith as A IR Y 's stress function, the stresses follow from (l,.13) by m eans of
the two - dimensional perm utation te n so r:
y6

yo

6r ^ I

S IIo r

Analogous to ( 2 .19) and ( 2 .20 ) the tw o -d im e n sio n al perm utation tensor


responds to
7a

= e

yo

/g

+1

/g

-1

cor

1
and with ( 3 )

sT<T =

sm a

(8 )

0
W ith ( 5 ), the covariant derivative ( 2 .34b) changes into a p artial derivative

I c jt

= 1 , o ) ! = , o r - ,p r o/ rJ = , o r
v

J \t

Exercise A-2-3

63

Each component of the stress tensor can then be obtained from (7.13)
11
T

It _ I
= E

x
12
T

22

>

12 12

12

fiT

12

21

12 11 I
E

fU

| 21 +

11 12,1
E

|l2 +

12 1 2 , 1
E

l 22

,22
12 21 _
+

21

,22 . 2
sm a
11 21

E E

|u +

E E

. 2
sin a

11 22 _

ffi^ j

12 22 _
^

22

,21

26 2t ^
21 21 _
22 21
21 22 ^
22 22
= E E ,t = E E ,11 + E E
,21 + E E
,12 + E E
,22

^
=

1 2t
=

11

11 11,1

| iT =

22
X

21 21
= E

1
(1 1

sm a

,U

E x e r c i s e A -2 -3 :
An infin ite strip under constant tension crx has a crack which can b e pre
sented as an elliptical hole ( see Fig. A -5). T he geom etry is described by an
elliptical - hyperbolical coordinate system *, 2. T he relationship betw een
C artesian and elliptical - hyperbolical coordinates is given b y :
x = c cosh cos

x 2 = c sinh

sin 2

w ith c = focal d istance to th e origin


ant^ 0 <

< 0 0

; 0 < < 2 rr .

F ig. A -5 : E lliptical hole in


an in fin ite strip

D eterm ine
a ) the covariant base vectors, th e covariant m etric tensor, its determ inant,
and th e co ntravariant m etric ten so r ;
b ) the CH RISTO FFEL sym bols of th e first an d second kind ;
c ) the physical com ponents of a vector v .

64

Tensor algebra and analysis

S o lu tio n :
a ) For determining the covariant metric components we esta b lish the base vec
tors according to ( 2 .26)
dxh

a?

e/3

dx

dx

df;

df;

g j = j- e j -|
dx

S2

j -I

df;*

1 2

1 2

j- e 2 = c sinh \ cos \

+ c cosh \ sin !; e 2 ,

( 1)

dx

j e 2 = - c cosh 5 sin ^ e j (- c sinh \ cos !; e 2

df;

T he metric com ponents then follow from ( 2 .1a)


_

_ ^

d x v dx

z ap - e a

~^p

~ *

. 1 . 1 . 2 . 2
O
X dx , OX dx
2 , . .2 ^.1 2-2
,2 -1 . 2 -2.
g n = j- j + T r = c (sinh cos g + cosh g sin Z ) =

dl

dZ

dZ

dl;

2, . . 2 J , ,
. 2 2.
. ,2 J , . 2 2.
2 , . .2 1
. 2J,
= c (sin h Z (1 sin Z ) + (1 + sinh Z ) sin Z ) = c (sin h Z + sin Z ) >
. 1 . 1
. 2 . 2
dx dx
dx dx
2.
-2 1 . 2 _2
. . 2 1 2 2.
822 = T^2 T 3 + TI2" T72 = c ( cosh 5 sm r + sinh I cos g ) =

df;

di;

df;

df;

2 . 2 1
. 22
= c (sin h I + sin \ ) = gu ,
% 2 d>x 2
dx

dx1 dx1

gl2

d^1 d?2 + d
2

df

= c ( sinh ^ cos \ ( - c o s h !; sin ? ) + cosh i; sin 5 sinh \ cos 5 ) = 0 .


Since g12 = g21 = 0, the elliptical - hyperbolical coordinate system is orthogonal,
i.e. the two families of curve param eter lines meet each other at right angles.
T he determ inant of the metric tensor follows from ( 2 .1c)
g = d e t(g Q/3) = g u gu = c4 ( sinh2 f;1 + sin2 f;2)2

(2 )

T he covariant metric tensor then is

(gop) =

w ith

g ll

811

0
(3 )

-fg = c ( sinh2 i;1 + sin2 f;2) .

Since gap is an orthogonal basis ( diagonal m atrix ) according to

a/3
g g/?T

a
7*

Exercise A-2-3

65

the non - vanishing contravariant metric components are inverses of the elements
of the covariant metric components
g

(aa)

=
. T
g(aa)

for

a = 1,2 ,

i.e. the contravariant metric tensor is


l//g

(4)

(s* ) =
0
b)

i//g

The CHRISTOFFEL symbols of the first kind can be calculated from (2.30)
S a / ^ 7 ))

6 7 a , (3

rOt

==>

1 d8 u
2 d?1
1 d8 u
2 d?2

A . - K1

r1 211r121 >
^221

2 ^g 21,2 +

8 l 2,2

822 ,1 )

1,
T = 2:

112 = o '( g 12.1 + 821,1 -

r1 212 r1 122

821,2

812,2)

= C S ln? COS? .
2

99

= c cosh 5 sinh I

~ ^ T

222

2 .

-S
1

^ ( g 22,i +

~~ COSllf> sinh ?

4 ^g u

8 u >2)

r !22 =

2 ^g 22,2 + 822,2

822, 2)

2 . 2
C Sm
C S

^822

'

The CHRISTOFFEL symbols of the second kind follow from (2.31) with ( 4 )
r 6 - </7 r
8

a/3

8 = 1:

rn= 8
^2

17

a /3 7

T-

__ ^

r ii7 = e

11

= e 17 r12T = g

rm + g

r121 +

12

t. t

r U2 = - ^ c sm hc cosh^ ,

r122 =

c* sin f cos r ,

y8

r121
1 r2
1 12
T-1
22

1T

^227

11 ~

221

12

^222

1 2 . , 1
, 1
y~~ c sin^ 5 cosh ^

66

8=

3 State of stress

9
27
rn = g r
ri2- g

27 _

=g

21 _
22
1 2 . ..2
_2
r u] + g r n2 = - ~ = c * i cos? ,

127 8

yg

21 _

r i21 + g

1 2 . , _1
,_ 1
r i22 ~ r= c sinhl; cosh^ ,

22 _

yg

121 112 }
^2

22

27

rL227 g

21

^221

+ 8

22

r 222

c2

c sin 5 cos \

W ith the abbreviations


1 2 . , 1
, ,.1
1 sinh 2 ^
- = c sinh \ cosh I = -----= = -------fg
2sin h l l + sin2 f
1 2 . 2
Jl
j= c sin I cos \

ft

1
sin 2 ^
=
------r -25- = B
2 sinh2 I 1 + sin2 i f

the CH RISTOFFEL symbols of the second kind read in m atrix form


A
B
c)

r2
a@ -

-A

-B

(5)

The physical components u , v of a vector v are calculated from ( 2 .10) :

v* = vQ/ g ^
u = v! V g

11

= vi g

v* = v = v2 y g 22 = v2 g

-V 4

'v 4

V1 = g
v2 = g

V4
V4

11

(6)
v

E x e r c i s e A -3 -1 :
For th e sta te of plane stress
crx x = 60 M Pa , aj j = - 2 0 M Pa , r

= 30 M Pa ,

determ ine by m eans of M O H R 's circle the m agnitude an d orientation of th e


principal stresses, and the m agnitude of th e principal shear stress.
S o lu tio n :
For the given c x x , c yy und t xy we can determine the stresses on arb itrary sections
from M O H R's circle.
For drawing the circle shown in Fig. A-6 we first plot the given stresses on the
c - a x is due to the signs. At these points we then plot the shear stresses xxy an d /
or ryx as ordinates. Here, as opposed to the usual rule of signs, a shear stress is

Exercise A-3-2

67

considered positive, if it rotates clockwise. This special sign convention, which is


only valid for MOHR's circle, is necessary, if we want to display the rotation
direction of an arbitrary angle a between the coordinate systems in the circle.
Thus, i xy is to apply down, and i yx to apply up in the sketch. With A and A ,
two points of the circle are fixed. Their connection line is the diameter of length
2 r of MOHR's circle, and provides the centre point M. Now, the circle can be
drawn.
The principal stresses dj and/or Ojj following from the intersection points of the
circle with the e -a x is , lie under the angles a* and/or a* + rt/2 . The highest
shear stress xmax can be read as the highest ordinate of the circle, i.e., i max = r
( cf. / 3.22b) ). The corresponding section has the angle
a** = cc* +

with respect to the x -a x is . From the drawn circle the results can be read for the
given task :
T

T,'m ax

Scale: |------ 1 = 20 MPa .


dj = 70 MPa ,
djj = - 30 MPa ,
2a* 37 ,

2a** 127 .

Fig. A-6: MOHR's circle

E x e rc is e A -3 -2 :
T h e stress com ponents of a th re e -d im e n s io n a l sta te of stress

are given.
D e te rm in e :

a ) P rin cip al stresses ,


b ) P rin cip al axes ,
c ) Stress deviator ,
d ) Invariants of th e stress deviator .

68

State o f stress

S o lu tio n :
With the invariants (3.13)
Ij = 100 + 60 + 10 = 170 [MPa
L

O
CM

100
l

O
CM

^2

60

60 0

0 10

10 0
+

0 100

= 4800 + 600 + 1000 =


2
6400 [ MPa ] ,

100
l

O
CM

20/3^

60

10

= 48000 [MPa]

According to (3.12), the principal stresses follow from


a3 - 170 a + 6400 a - 48000 = 0 .
As no shear stresses are acting on the section z = const in the example, the com
ponent Ojjj = czz = 10MPa is a first root of this cubic equation. Dividing by (o 1I1 )i we come to the quadratic equation
c2 - 160 a + 4800 = 0
and with that to the following eigenvalues
120 MPa ,

Or

= 80 y 6400 - 4800 =
40 MPa .
b ) By substituting the eigenvalues into the homogeneous equation system (3.11)
the orientations of the principal stresses follow. With Cj = 120 MPa this yields:
2 0 -fz n*

= 0,

2 0 -fz n* + (60 - 120) n*

= 0,

(100 - 120) n* +

(10 - 120) n* = 0 .
From this, it follows that

and herewith the eigenvector

n* = 0

n* = C

n* = / F i

1
0

Normalizing the eigenvector to the length 1


|n * | = 1 = C2 [ ( / 3 f + l 2 + 02]

Exercise A-3-2

69

Fig. A -7 : The in itial system and the system of


principal axes

yields
II

-1 /2
0
V3/2'
, and in analogy one obtains n*j = / 3 / 2
1 /2
nn*lII _ 0
1
0
0

As the components of the normal vectors correspond to the cosine directions rela
tive to the initial system, we can draw the principal axes in the x , y , z - system
(Fig. A-7).
In order to check, we can carry out the transformation from the initial system to
the system of principal axes. With the rotation m atrix according to (3.10)
^
B =

-T

follows from the matrix product (3.9)


T 5 '

~i i

S' = B S B

100

0
0

2 0 / 3^

120

40

10

2 0 / 3^

60

0
3,

0
0

0
0

JII

the diagonal matrix of the principal stresses,


c)

With (3.24)

M = - j ( 100 + 60 + 1) = " X

MPa .

we calculate according to (3.23) the stress deviator


1 0 0 -* !
s =

2 0 /iT
0

2 0 / 3

130
3

6 0 -* p

1 0 -f>

20/3^

2 0 /3

10
3

140
3 .

70

d)

State o f stress

According to ( 3.26) , we obtain the invariants of the deviator as


,

_ 130
1 3

10
3

140
3

0,
9700 r

\2

'I 2 = 6400

[ M Pu | ,

170

'I , = 48000 - 6400 ^


6

+ 2 .

1330000
[ MP a f
27

In order to check, we directly calculate the invariants of the deviator from 'S.
Thus, we get for example

'i 3 = r s

140
3

^
20

2 0 /3 -

y^T

1330000
[ MP a f
~ 27

E x e r c i s e A -3 -3 :
For th e in fin ite strip w ith an elliptical hole ( see A -2 -3 ), establish th e equi
librium conditions in elliptical - hyperbolical coordinates.
S o lu tio n :
The equilibrium equations follow from ( 3.28a) :
<%

+ ( =

The two equilibrium equations then re a d :


= 1:

x ^

+ f1 = 0 ,

a. - 2:

x%

+ f2 = 0 .

The covariant derivatives may be determined from ( 2.35b) :

Poc\
Poe a -yP U3 ary
\P~ T ,p+ TPyT + V l T

Fully written, this yields for

- = l:

+ x21|2 + f1 = 0

r 1+ r l 1r 11+ r i12 r21 + r ; i r + 4 x 12 +


_1 12 , _1 22
2 11 _2 12
+, t21,2 +, r2
1r + r22t + r21
1 + r221 +

X11,! +

x 21, 2

.1
f

+ 3 A xn + 4 B X12 - A X22 + f1 = 0 .

Exercise A-4-1

71

Using the abbreviations A and B for the components of the m atrix of the
CHRISTOFFEL symbols from A - 2 - 3 , one obtains

( 3 sinh 2 t 1' +

2 (sinh2 i;1 + sin2 2)

+ 4 s i n 2 V 2 - sin h 2 5 1 t 22) + f* = 0 .
In analogy, one obtains for a. = 2:

+ 3 sin 2 2 t 22 - sin 2 1;2 t ) + f2 = 0 .


Finally, the physical com ponents according to ( 2 .17) should be introduced into
these equations.

E x e r c i s e A -4 -1 :
For a disk, which is clam ped along th e boundaries x = 0 , y = 0 as shown
in Fig. A-8 , and has th e opposite edges free, th e strain s have been deter
m ined from stra in gauge m easurem ents during th e action of an external
load. T h e results can be approxim ately described by

( k = constant factor)

x
a

Fig. A - 8 : Disk w ith two clam ped


an d two free edges

a ) C alculate the displacem ents u and v and th e sh ear


of th e disk .
b ) Is th is a com patible state of strain ?

over th e dom ain

Xy

72

4 State of strain

S o lu tio n :
a ) For the given strains the following strain-displacem ent relations (4-14) are
valid:
_ d \l

= k ( ^ x 2y + a y 2) ,
Exx ~ "dx
dv
, a 2
y y = d ^ = k b y x '

(i)

Integration of ( 1 ) yields
x y + ay x
a 3
/
\
, ai _1_ 3
, .
v ( x , y ) = k-r- y x + g ( x ) .
b 3

(2)

The unknown functions f ( y) and g ( x ) can be calculated from the following


boundary co n d itio n s:
1)

u(0,y) = 0

u ( x , 0) = 0

f(y) = o ,

2)

v (0,y) = 0

v(x,0)=0

g(x) = 0 .

(3)

Thus, we obtain for the displacements


u ( x , y) = k ( ^ | x 3y + a y 2x ) ,
(4a)
v ( x , y ) = k j l y 3x .
In dimensionless form ( 4a ) may be written as
^

" (M i

-i)

' a3 b + b2 a )

v(* L ) = v 1 y X
Va b /
33 b3
w a
with

u0 = v0 = k a b

(4b)

The displacements u and v at the free boundaries are presented in Fig. A-9.

u
u
v
1
v

Fig. A-9 : Displacements of the


free edges of the disk

E xercise A-4-2

73

Finally, with ( 4a ) and according to (4-14) the shear strain becomes


du , d v
Yx y =

d^ +

1 , 7

d3T = y k U

3
X

a 3 ,

+ 6 a x y + y )

(5 )
3 b ' a3

ab

ajjS/

b ) According to (4-19), the compatibility condition is


\2
a

N2
Ex x

dy2

Ey y

dx2

Txy

^xdy

'

Inserting of ( 1 ) and ( 5 ) yields


^ - [ k ( ^ x 2y + a y 2 )] + ^ ( k ^ y 2 x ) 2

I 1

dx dy

/ h a

a a\ I

[ i k(^x3+ 6axy +y3)l = 2 a k

- 2ak = 0 .

It is hereby proved that the state of strain is compatible.

E x e r c i s e A -4 -2 :
A stra in gauge rosette as shown in Fig. A-10 has m easured th e following
strain s in th e directions 1 to 3

e2 ~ ~
e3 ~

eo

~ o

D eterm ine th e principal strain s in term s of direction a n d m agnitude.

F ig . A -1 0 : S train gauge rosette

74

4 State of strain

S o lu tio n :
Suppose th a t the direction 1 lies under the yet unknown angle a against the prin
cipal axis I. Applying the transform ation form ula for stresses ( 3 .17), which is
analogous for strains, and taking into consideration th a t no shear strains occur
between the principal directions, the following is valid for the directions 1 and 3 :

El = y ( EI + EI l) + ^ ( EI eu ) c o s 2 -

( la)

e3

(lb)

= - i - ( I + n ) - i ( EI - EI l ) c o s 2 (x .

From this follows


EI + EII = E1 + e3

>

( EI EH ) cos

= h -H

(2)

Inserting ( 2 ) into the relation for the direction 2 , we obtain


t j- tt

S2 =

t ~ tt

E, -j- Eo

E1

+ ------ cos[ 2 ( a + 45 " )] = ----

En

tan 2 <x

2 2 - 1 - e3

and from this

t a n 2 a = ------------------- .

From the given values it becomes


ta n 2 a = + - fz

------> a = 30 .

From

EI + EII = 0

( EI ~ EII) cos 60 = 2 Eo

we then obtain the principal strains to be


ei

= 2 eo

eii

= - 2 eo -

E x e r c is e A -4 -3 :
In C a rte s ia n co o rd in a tes a d isp la c e m e n t vecto r is given as
X

u
V

y + 4z
4 / 2 x + 3z

D e te rm in e
a ) th e s tr a in te n so r X >
b ) th e p rin c ip a l s tra in s w ith re g a rd to m a g n itu d e a n d d ire c tio n ,
c ) th e v o lu m e d ila ta tio n .

Exercise A-4-3

75

S o lu tio n :
a ) The strain tensor \ follows from (4-13)
1_/ dji , dv \
2 dy
dx '
dv
dy

du
dx
V = k
sym.

dw\
dx
dv/\
dy

2 /2
2
3

V = k

2 /2

1 / du .
2 dz
1 / dv .
2 dz
dw
dz

From this we obtain the normal and the shear strains


Ex x =

v1x y = 0

yy = k
ezz = 3 k

Txz = 4 / 2 k ,
Y

1y z

= 4k

b ) The principal strains are determined from the corresponding eigenvalue


problem (2.22) :

det(V s - X I) =

Ex x ~ X

2 Y'xy

Y
2 'x

Y
2 ' xy

Syy-X

Y
2 'y*

i-Y

Y
2 'y*

E -X
zz

2 /2 k

k-X

k-X

2k

2k

3k-X

2 /2

= 0.

Calculation of the determinant yields the characteristic equation


(k - X) (X2 - 4 kX - 9 k 2) = 0
with the roots
Xj = k

X2 3 = (2 / l 3 " ) k ,

i.e., the principal strains


Ej = k

Ejj (2 -|- \ 13 ) k

Ejjj = (2 /13 ) k .

The d irectio n s of the principal strains are calculated by the corresponding


eigenvectors. This will be shown for the case of Ej = Xj = k, where we insert this
root into the homogeneous system of linear equations ( 2.22a) :

76

4 State of strain
0 a lx + 0 a ly + 2 -^2 k a Jz _ 0
a,lz = 0
0 ' a lx +

0 a ly +

2 k a lz = 0

2 / 2 k a lx + 2 k a ly +

2 k a jz = 0

a lx = C

arb itrary ,

-^2 a^x /2 C .

ly
Thus, it yields the eigenvector

1
aj = C

w ith arbitrary C 6 R

0
By analogy, the reader may calculate the two other eigenvectors and obtain
1

a2 - C

,
L

72

( l

a3 - C
1

'

w ith arb itrary C 6 IR.


c)

According to ( 4 .17b), the volume dilatation is equal to the first invariant


e = I,1 = exx + e yy + e zz k +1 k 1 + 3 k = 5 k ,

which may be alternatively calculated as


e = EI + EH + Em = k + ( 2 +

k + (2 /is " ) k = 5 k .

E x e r c i s e A -4 -4 :
T he stra in - displacem ent relation for th e norm al stra in e22 and th e m ateri
al law of an isotropic, in fin ite disk w ith an elliptical hole are to be deter
m ined by using th e elliptical - hyperbolical coordinates from E x e rc ise
A -2-3.
S o lu tio n :
T he physical com ponents of the strain tensor ( tensor of second o rd e r) are first
calculated according to ( 2 .17) :

la p = y ap l/g(Q)g(/J/J) .
Inserting ( 5 ) and ( 6 ) from A - 2-3 yields the following for e22 :

Exercise A-4-4

122 - e22 = Y zz/g22 g22 = Y22 ^

= v2|2 ^

(1 )

+ A vi - B v2) - r - = [T 72-(g1/4y) + A g 1/4u - Bg1/4v

*?

77

/?

, . ,2,1
. 2,2.1/2 1
c ( sinh c + sin t ) 'vv/)

1 . _ ,2

, . ,2 1

1 .
"2 "S

. 2 2.-1/2

- sin 2 \ c ( sinh c, + sm t, )

,1 , . , 2 , 1 , . 2,2 .-1/2
c(sm h !; + sin t, )
u

v
c2 ( sinh2 (} + sin2 f;2 )

. .2 ,1

. 2 ,2 ,-1 /2 1

= (sinh !; + sin !; )
1

, . ,2 ,1

. 2 2,1/2 d V

sin 2 1; v + (sm h !; + sin 1; )


2

j- +

,5

1 . , _ , 1, . ,2,1
. 2,2,-1/2
+ sinh 2 5 ( sinh i; + sin \ )
u 1
,2 / . ,2 ,1
. 2,2 ,-1/2
sin 2 !; ( sinh i; + sin !; )
v

c ( sinh2 51 + sin2 2)
1

[/ sinh2 51 +

c (sinh2

+ sin2 !;2)

sin2 1;2

sinh 2

^ / sinh2 1;1 -f sin2 I;2


1
22

dv

sinh 2 I;1

/sinh2 I;1 + sin2 i;2 d I;2

^(sinh2!;1 + sin2 2)

(2 )

The material law for the state of plane stress without consideration of tempera
ture terms reads according to (5.5)
76
a /3 ~

with

^aP'yS

( 3a )

1 + V,
2 E ^go"f gp6 ^

^p

E goP

Insertion yields
76

1ap =

2 E ^gfv'yg/?^

gafig/3^)

E 8a^g7 JT

We now obtain the material law for y22 :


7

^ 22

^27625

&26 g 27 ) _ r&22&-yS

With the covariant metric components

78

5 Constitutive laws

finally follows
f

22 =

822611

[~

1 + V,

I 2E ^622 822 + 822822) _ 622 822 j x

2,1

I 22

2E (8 + g ) - 8 j x
1

4,

-ggt

Here, the physical components of y22 , t


I
Y22

V6

e22

22

22

and t

2 , 2.2 22

= p c ( slnh * + sin ? )

^22

11

22

- VI

11

11

(4 )

are still to be introduced:

= i ( o 22 - vou ) .

22 -

E x e r c i s e A -5 -1 :
A square - shaped steel ingot is em bedded in a rigid concrete base and both
are s tre s s - f r e e at room tem p eratu re (Fig. A -ll). T he upper end of the in
got is free. T he ingot only is now subjected to a constant tem p eratu re in
crease 0 (coefficient of th erm al expansion a T ) .
a)

W hich stra in occurs in th e z -d ire c tio n of th e ingot, provided th a t the


vertical boundaries are frictionless ?

b ) How does th e stra in change if, additionally, th e boundaries at x = const


are m oved so th a t no contact occurs ?
S o lu tio n :
a ) We solve the problem by means of
HOOKE-DUHAM EL's law (5.2)

v = 1- w s - i s I + a T 0 1 From the problem formulation the condi


tions follow
- stress-free surface
aZZ = 0 ,

( l a )

- impeded strains
xx = yy = 0 ,

(lb)
Fig. A - l l : Steel ingot

Exercise A-5-1

79

shear free state of strain


Txy = Txz = Tyz = 0

( lc )

According to (5.1), the relations for the strains in component notation read
Exx = i ( x x

- v (yy + )1 + a T 0

( 2a )

Eyy = k [oyy

~ v(<5 + 0* * )]+ T

( 2b)

Ezz

V(xx

( 2C)

+ 0yy)l+ T

Inserting the conditions ( la,b ) into ( 2a,b) then yields the two equations
x x VOy y
" V 0 XX +

E a T 0

yy =

- E aT 0

( 3a )

( 3b )

From ( 3a,b ), axx and cyy are calculated as


E a T 0

ox x = ay y

--------
.
l v

( 4 )'

The strain in the z - direction follows from ( 2c ) and ( 4 ) :

Ezz = i K x + yy) + T 0

=>

Ezz =

aT 0

b ) In this case, for free boundaries at x = const


xx =

(5 )

is valid instead of sXX = 0 .


Then follows from the equations ( 2a,b,c )

Exx

-^ yy

Ey y = %
Ezz

+ T 0

+ T 0 .

= -^ y y + T 0

Furthermore, no strain s
3yy

>

( 6a )

( 6b)

( 6c)

occurs in the y-direction; thus, from ( 6b ) follows

E Of.y 0

and from ( 6a ) or ( 6c )
Exx = (1 + v)otT 0 = szz .
The strain in the z - direction is now smaller than in a ).

80

Energy principles

E x e r c i s e A -6 -1 :
A B E R N O U L L I-E U L E R b eam w ith one end clam ped and th e o ther sim ply
supported, is subjected to a tran sv erse load q ( x ) p er u n it length of th e axis
of th e beam ( plan e beam b en d in g ) ( Fig. A-12).
T he differential equation and th e b o undary conditions for th e plane beam
are to be determ ined
a ) by m eans of th e principle of v irtu al displacem ents,
b ) by solution of th e corresponding v ariatio n al p rin cip le.

w(x)

8w(x)
F ig . A -1 2 : Clam ped, sim ply supported b eam und er uniform ly distributed
load
S o lu tio n :
a ) At first, we have to state the strain energy of the beam with d/dx ( ) :
U = - | j E I y w xx2dx = n i .

(la)

For the work of the external forces we have


1
W = - I le = J ^ q ( x ) w d x .
o

(lb)

From ( la ) and ( lb ) we obtain the total p o ten tia l


1
n = II. + n e = J [ -i- E Iy w xx2 - q ( x ) w dx .
o

(2)

The minimum principle (6.21b) is now applied:


II = ITj + IIe = minimum .
The stationarity, see (6.20), implies that the variation of the total potential must
vanish for arbitrary kinematically admissible variations 8 w

Exercise A-6-1

81

1
811 = 8 { J [ 4 E I y w.xx2 - q ( x ) w ] dx J = o
0
1

(3)

8 1 1 = JI

(4)

E I y w , x x 8 w >xx - q^( vx ) )8 w

dx = 0 .

After double partial integration, ( 4 ) can be written a s :


1
8!! = J [ ( E I yWiXX)(Xx - q ( x ) 8 w dx -

( E l y w ixx) x 8w

x= 0

11
+

E I y W, x x 8 w ,x x=0

(5)

= 0.

Vanishing of the integral for arbitrary variations 8 w leads to the differential


equation for the elastic line of the beam
( E I y w ,xx),xx = < l ( x )

( 6 )

The boundary co n d itio n s follow from the boundary terms. In order that these
vanish, we must have at both x = 0 and x = I,
either
and
either

8w = 0

or

(' E I y w , x x )A x = 0^ z = 0 , 1

(7a)

8w

or

E Iyw ixx = - M y = 0 .

( 7b )

=0

The boundary conditions for the clamped end x = 0 are:


for

x = 0

>

w = 0

= 0 .

Hence, at x = 0 we have 8 w = 0 and 8 w


and My ^ 0 in general.

(8a)

= 0, and according to ( 7a,b ), Qz =/= 0

The boundary conditions for the simply supported end are:


for

x = I

w = 0

(8b)

My ~ w .XX = 0 .

Here, the condition w = 0 implies that 8 w = 0 and thus, by ( 7a ) that Qz ^ 0 in


general. The second condition My = 0 is obtained from ( 7b ) as, in general, w_x =/=
0 and thus 8 w x =/= 0 at a simply supported end.
b ) We naturally obtain the same result for the example if we apply the general
solution (6.35) to the problem of variation (6.34)
According to ( 3 ), the basic function reads
F ( X - W ' W,xx) =

2i

(9)

E I yy W , x x 2 - q ( X ) W

From (6.35) now follows E U L E R s d ifferen tia l equation:


v sw

A.

aw

( E I y w .xx).xx

This differential equation is identical with ( 6 ).

- q(x ) = 0

( 10)

82

Energy principles

As b o u n d a r y c o n d i tio n s we obtain according to ( 6.35):

\ x2
/N L _/_dP_N
V dw

,xj

)x1

[ ( E I y W, x x ) , x S w ] x=0 = 0

[ E I y W, x x S w ,

x=0

= 0

( 11 )

= 0

[ ( E I y W, x x ) , x S w

= 0

E I y W, x x S w , x j x =; = 0 '

In ( 11 ) the boundary conditions ( 7a,b ) as well as ( 8a,b ) are included ( see dis
cussion in Section 6.6 ).

E x e rc is e A -6 -2 :
A linearly elastic body is subjected to a statio n ary tem p eratu re field 0 ( ) .
Set up th e variatio n al functional by HELLINGER and REISSNER for sm all
strains an d show th a t one obtains th e basic equations of linear th e rm o elasticity as necessary conditions from th e corresponding v ariational pro
blem. Volume, surface an d in e rtia forces are to be neglected.
S o lu tio n :
Substitution of the expression for strain energy U from ( 6.16a) into the variatio
nal functional ( 6.29 ) yields
nR =

J { ~2 C'jkl ^ij ^kl -

PT l ! 0 + ^

( 4

+ Vjli) - 1 . j ] } dV

This functional contains altogether 15 unknown functions Yy . r'J , Vj. From the
problem of variation

8 n R = 8 J fiR dV = 0
V

follow in analogy w ith ( 6.35) the corresponding E U L E R s equations as necessary


co nditions:
1.

A ilp = o

>

- = ( v | + v.l. )

S train-displacem ent equations according to ( 4 -12b) .

^nR

2.

-z z - =

ij

ijkl

x = C

ij

Tk l - 13g

'ij

3-

M aterial law according to ( 5.6a) .

| ^ f l | s= 0

1^ = 0

Equilibrium conditions according to ( 3.28 a) .

E xercise A-6-3

83

E x e r c i s e A -6 -3 :
S tructural com ponents in m echanical and civil engineering often have a
com plex shape. F u rtherm ore, they are subjected to m ultiple loads and th e ir
boundary conditions can be quite com plicated. S tarting from the energy
principles of th e theory of elasticity th e F in ite E le m e n t M ethod (FEM) was
developed for determ ining strain s and stresses of such com ponents, and
taking th e principle of v irtu al displacem ents as a basis, displacem ents are
introduced as unknow ns and can be derived from th e equilibrium condi
tions. We proceed in such a way th a t th e supporting fram ew ork is divided
into so -c a lle d fin ite elem ents ( bar, beam , disk, plate elem ents e tc .) w ith
approxim ate expressions chosen for th e displacem ent fields in each elem ent
( for details, see e.g. [ A .l, A.2, A.21 ]).
W ith the exam ple of a disk u n d er tem perature loads th e basic relations for
the m ethod should be determ ined. Therefore, as shown in Fig. A-13, a tria n
g u lar elem ent is taken out of th e dom ain of th e disk a n d is described in a
C artesian coordinate system .
At each of th e th ree nodes 1, 2, 3 of th e tria n g u la r disk elem ent two nodal
displacem ents are adm issible an d can be assem bled in a row vector as fol
lows :

dT = ( u 1, u 2 , u 3 , v 1, v 2 , v 3)T .

F ig. A -1 3 : T ria n g u la r elem ent w ith 3 x 2 nodal displacem ents of the


corners
a)

T he stra in -d is p la c e m e n t equations are to be stated in m a trix notation


in th e elem ent - inherent system . Therefore, a linear displacem ent ap
proxim ation ( constant strain s ) of the following form will be a ssu m e d :
u ( x . y ) = a1 + a 2 x + a3 y ,
v ( x , y ) = a4 + a 5x + a gy .

84

Energy principles

T h e relatio n betw een strain s and nodal displacem ents is to be given


w ith a p . . . , a 6 as free values.
b)

T he equilibrium of th e elem ent is to be form ulated w ith th e help of th e


principle of v irtu a l displacem ents.

Solution:

a) The strains in the element - inherent system (upper index e ) are given for the
state of plane stress according to ( 4-12b):
(T ^ )e = y ( v Q|^ + v^|a )e .
In matrix form this relation reads
C = ( D v ) e ,

(la)

where

d
dx
0

yy

d
<^y

xy J

o
(lb)
<^y
_d_
dx

If we introduce the assumed linear displacement field with the six free values
a1, . . . , afi, we can write the following vector in column notation for the displace
ments :
II

with

If

(2a)

1 x y 0 0 0
0 0 0 1 x y
"5

b
Further, we obtain the vector of the nodal displacements as
(3a)

de = ( O a f

with

Cf

1
1
1

X1
X2

*1

X3

y3

y2

(3b)
i

Vi

X1
x2

X3

y3

y2

Exercise A-6-3

85

We now insert the relations ( 2a ) and ( 3 a ) into ( l a ) and obtain the following re
lation between strains and nodal displacem ents:
(4)

b ) We sta rt from the D U H A M E L - N E U M A N N f o r m of H O O K E 's law for disks


in ( 5.15b) :
aP

-po/M f

= E

[T7 5 - T a7 0

In m atrix notation this law reads, see ( 5.14)


a = E [ e - ee ]

(5)

with

1- v

Bg o(fj> 0

The equilibrium conditions can now be derived from the principle of virtual dis
placements. Thus, according to ( 6.20 ), a body is in static equilibrium, if the virtu
al total potential is equal to z e ro :
s

n .

Here, ITj and II are the internal and external potential, respectively.
W ith the m aterial law ( 5 ) it follows from ( 6.12a) that the internal potential for a
therm ally loaded disk is given by
T

8 Ee-28

E e fl d A .

(6)

We now insert ( 4 ) into ( 6 ) and obtain


e

n> =
with

. T . e

T . e . e

y ( d ) K d + ( d ) k0
"2

= .(

EVdA

as element stiffness m atrix and


V

E e0 d A

as a vector of the te m p e ra tu re forces.

(7)

(8)

86

7 Problem formulations

W ith th e principle of virtual displacements we obtain after introducing th e virtual


work of nodal loads ( vector p ) from ( 7 ):
s n * + s n * = s ( d T )e ( K d +
p

_ ,e -e

= K d

kg

- P )e = o

_e

+ ke .

After determ ination of th e nodal forces, we can now establish the equilibrium for
the entire domain. Therefore, the variation of th e virtual energy for the entire
structure has to vanish. This energy is obtained by sum m ation over the elements
( see [ A .l, A.2 , A.21 , C.2 5 ] ).
Rem ark-. I t should be emphasized th a t this example only serves to illustrate th e
way of developing a triangular finite element. Such an elem ent with
constant strain is only a very sim ple one. M ost softw are system s con
tain more accurate and improved elements.

E x e r c i s e A -7-1:
D eterm ine th e stresses in a hollow sphere ( outer radius a, inner rad iu s b )
un der constant internal pressure p from th e basic equations of th e theory of
elasticity.
S o lu tio n :
A spherical - sym m etrical state of stress is given, in which
dcp

= V= w = 0 ,
0 iptp

( 1)

0t/v
QQ ,1 iptp

In view of ( 1 ), the stra in - disp lacem en t equations ( 4.I6) can be reduced to

^<pip

(2)

Since y^ = 0 the current strains are principal strains.


Combination of the two equations ( 2 ) yields the com patibility condition
r r = ( T S v v ) . r

( 3 )

Equation ( 3 ) contains the two unknown strains srr und


. We now replace by
m eans of ( 1 ) the principal strains by the m aterial law according to ( 5.1) ( spheri
cal coordinates are orthogonal, therefore the same structure as in C artesian coor
dinates ) :

Exercise A-7-1

r r

E I rr

^iptp

1
E I ^iptp

87

2 V v>v> 1

(4 '
^ ( ^rr ^tp<p ^

^i9t? '

Inserting ( 4 ) into ( 3 ) yields


a

rr

2 \i a

tptp

= 1| r f ( l - v ) oV>> - v ror Ji J ,rr .

( 5 )/

Furthermore, we need the equilibrium conditions (3.30), which by means of ( 1 )


can be reduced to
rr,r

+ f ( rr "

) =0

( r2 r r ) , r = 2 l 0 v v '

or

( 6 )

With (5 ) and ( 6 ) we now have two differential equations for err and ov v . If we
replace ov v in ( 5 ) by means of ( 6 ), we obtain one differential equation for er r :
( 2 r r ) , r r - 2Orr = 0 -

( 7a )

This differential equation corresponds to the BELTRAMI differential equation


(7-4) in spherical coordinates for an axisymmetrical state of stress. This ordinary,
second order differential equation leads to the following EULER differential equa
tion
r 2 rr,rr + 4 r 0 r r, r = 0 -

( 70 )

the general solution of which can be determined with the aid of a power approach
orr = A rn

(8 )

The characteristic equation is


n(n-l) + 4n=0
with the roots
nx = 0

n2 = - 3 .

The solution of ( 7b ) then becomes

. .

Ao

rr = Al +

(9)

with the two constants Aj and A2 to be determined by the boundary c o n d i


tions. These read as follows:
rr ( r = b ) = P

> rr ( r = a ) = 0 '

From (10 ) we obtain with ( 9 )

( 10 )

88

7 Problem formulations

Fig. A-14: Stresses in a hollow sphere


with a /b = 2

The stresses are thus found to be


\3
rr =

- P

Cv v

c&0

2 /a\3

+2

(H)

They are presented in Fig. A-14 for the case of a / b = 2. The maximum tensile
stress occurs at the inner surface and is given by

+2

( 12)

W max

If the stresses in a t h i n - walled sphere are to be determined, one starts from a


medium radius
r m -i2 (a + b) .
With the wall thickness t = a - b we obtain
a = V + ^
and thus for t / rm

rm

1 the approximations

(r)3

It now follows from (11) that for a thin - walled sphere the tangential stresses
are constant over the thickness, and from ( 12) we obtain
< p v 0 i9i9

Prm
21

Because of its importance in boiler design this relation is called the boiler f o r
mula.

Exercise A-7-2

89

E x e r c i s e A -7 -2 :
A concentrated force F acts upon a n elastic h a lf - s p a c e as shown in Fig.
A-15.
D eterm ine th e strain s and stresses w ith th e help of L O V E 's displacem ent
function x ( r ,z ).

F ig . A-15: E lastic h a lf -s p a c e subjected to


a concentrated force

S o lu tio n :
We first seek a LOVE's displacement function x ( r ,z ), which
a ) satisfies the bipotential equation (7.5)
AAx = 0 ,

(l)

b ) fulfills the boundary conditions


zz( r >z = 0) = 0 ,

( 2a )

i rz ( r , z = 0 ) = 0 .

( 2b )

We choose a linear combination of two solutions according to (7.6)


X

= C j-J t 2 + z2 + C2 z In ( z + / r 2 +

) .

(3)

With this approach (1 ) is satisfied. Cj and C2 are arbitrary constants to be de


termined from proper boundary conditions. We first determine the displace
ments from ( 7.7a,b) with R = - / r2 +

90

Problem formulations

1
1 - 2v

(C 1 + C2 ) i J - C I 5 ( _
(4)

1
1 - 2v

( 3 - 4 v ) C j + 2( 1 -

v)c2

(C 1 + C 2 ) z2

1 - 2v

R3

and the stresses from (7.,8a-d)


2G[ l-2v
~vv

R(z+R) +

I - 2J ) r 3

1 - 32v ^ C1 + C2) R

2G L
[v( C1
^ - T1 -^ 2v R( z2+ R) J
1 +' C2
2 ' )R3

= - 2 G[( c. -

t^

- )

+ t4

Tr,

= - 2 G [ ( C1 - J Z 2 ^ )|T3 +

(5 )

z
t ( c . + c ! ) R5

r z2
. * . ( C1 + C2) R5

Substitution of the shear stress xrz into ( 2b ) yields


Xr z ( r - Z = 0 ) =

- l ~ 2v c
2v Cl -

Thus, it follows that

- -r

zz

3G

z3

(6 )

1 v ( l - 2 v ) Rs '

With ( 6 ) the boundary condition ( 2a ) is fulfilled. In addition, the integral of the


normal stresses czz acting on arbitrary sections z = const ( z > 0 ) must be equal
to the magnitude of the concentrated force F , i.e.
F = - Jf e ZZ dA = - Jf czz 2 re r dr .

r= 0

( 7)

Inserting ( 6 ) into ( 7 ) yields the integral for the concentrated force


eCjGz3 ^
v(l-2v )

r dr
( r 2 + z2 f t 2
r= 0
2

By using the substitution r + z = v the integral can be solved. The constant is


then found to be

_ v( 1 - 2 v)
1 ~~

2ttG

which with the use of ( 4 ) and ( 5 ) leads to the final expressions for the displace
ments and stresses:

Exercise A-7-2

(,
2 tc ^
=

o ^

1
R(z+R)

_ 2v)

R3

3 z r2
Rs

91

(9a)
(9b)

R ( z + R)

_3F
2 n RS

( 9c )

_ J r rz
rz ~
2 tt R 5 '

( 9d )

The relations ( 9a-d ) are called BOUSSINESQ's formulas.


Finally, let us discuss the resu lts:
- Resultant stress
From ( 9c,d ) we find the simple relationship
El

( 10)

Thus, according to Fig. A-16, the resu lta n t stress at points of cut planes
const has the magnitude
/2
=

, 2

V z z + Trz

3F

3F

271 ( r 2 + z 2f

2 tt r 2 + z 2

cos p

Z =

(ii)

and points towards the origin 0.


If we lay a spherical surface of diameter d tangentially to the boundary plane z = 0
at the origin 0 , the following is valid for every point on this surface
.,2
2
2
,2
2
R = r + z = d cos p .

Fig. A-16: Resultant stress from a concentrated force acting on a halfspace

92

Problem formulations

Thus, according to (11 ), in the intersection points between any plane z = const
and the spherical surface, the magnitude of the resultant stress is the same and
given by [ A.16]
a

3F
2nd

- Displacements of the boundary plane


For the displacements of the boundary plane z = 0, we obtain from ( 8a,b )
u ( r , z = 0 ) = - ( l ~ 22V^ + v) F ,

(1 2 a)

w ( r , z = 0) =

(12b)

F .

Thus, the theory yields singular values for the displacements u and w at the point
of force application, and according to ( 9 ) similarly for the stresses. These singu
larities vanish if the force is distributed over a small area of the surface.

B Plane load-bearing structures


B .l D e fin itio n s - F o rm u la s -C o n c e p ts
8

D is k s

8.1

D e fin itio n s A s s u m p tio n s B a sic E q u a tio n s

D is k s are pla ne loa d -b e a rin g s tr u c tu r e s th e thicknesses t of w hich are


sm all in com parison w ith th e other dim ensions ( Fig. 8.1) an d w hich are
subjected to loads acting in th e m id -p la n e . All stresses are assum ed uni
form ly distrib u ted over th e thickness, i.e. they do not depend on z . We
therefore have a State of P lan e Stress for which the m ost im p o rtan t basic
equations in C artesian an d in polar coordinates a re sum m arized in th e fol
lowing, w here th e thickness of th e disk is assum ed to be constant.
a)

Isotropic disk i n C a rte sia n coordinates

B ipotential equation ( d is k e q u a tio n ) according to (7.14):


A A $ = - E a T A 0 - (1 - i / ) A V
w ith

d/dx = (),x

,
yy

(8.1)

d/dy = () , y .
LAPLACE operator due to ( 2.S9) ,

= $ (x ,y )

A IR Y 's stress function due to (7.22) ,

0 = 0 ( x , y )

T em p eratu re difference relative to the initial


s tre s s -fre e state ,

V = V(x,y)

P otential of volum e forces .

94

D isks

Stresses from derivatives of A IR Y 's stress function due to (7.22)

CTxx

* yy +

CTyy

xx

Txy

( 8.2)

$ yxy

S train - displacem ent equations


exx = U,x

eyy = V.y

(8.S)

7xy = U,y + V,

M aterial law due to (5.12)


e xx = i r K

V(Jy y )

a T

(8.4)

ey y = i r K y ~ I/7*x) + T O 0 >
"Y

'xy

----- T

*y

or due to ( 5.1S)
E

1-

V [ eyy +

(8.5)

v e xx ~ t 1 + ^ ) T 0 ]

^ 7 xy

Txy

w ith th e YOUNG's m odulus E, th e sh ear m odulus G, and the PO ISSO N 's


ratio v.
b)

Isotropic disk in polar coordinates

B ipotential equation using (2.49) w ith d / d r

* * *

* rrrr + J

) . t

( $ ,T ~ 2 ^/Ttptp ) +

j .

Stresses by (7.13)
a u = r3- ^ '<pv
W + 7r $ ,r + V ( r . V ) ,

r <p

,T T

J_<
2

,ip

( r lP )

_ J_<f
r

,rv>

+ $ ,VVVV) =

- ( i - ) ( y + 7 v , r + ) v

).<p

~ j g ( * ,rr ~ 2 * , r V V ) +

- - E T ( 0 , r + T O0.r + ^ O 0 W ) -

% V >

, d/ dt p

= - ( JL #
\ r

,V ) , t

( 8.6)

8.2

A nalytical solutions

95

S train - displacem ent equations due to (4-15)


err = U,r
1

e<pV = Y v,v> + 7

( 8.8)

= T U'<p +

M ateria] law due to (5.12) and (5.1S)


Crr

e<p<p -

1 ,
E ^

o.

(8.9)

1
G TrV

or
E
[
1-u2 [

avv -

E
[
1 I/2

Trip
8.2

( 8.10)

II
O
r*
s

a rr =

A n a ly tica l
e q u a tio n

so lu tio n s

to

the

hom ogeneous

b ip o te n tia l

a ) C a r te s i a n c o o r d i n a t e s
-

Approach w ith power series expansion

i k
# = ?i ?k aikX y

( 8.11)

w ith the free coefficients a ik an d a rb itra ry integer exponents i and k


( including z e ro ).
Since according to (8.2) th e stresses resu lt from second derivatives of the
stress function, term s w ith
i + k < 2
do not contribute. Power series w ith
i + k < 4
fulfill th e basic equation for a rb itra ry constants aik , because only deriva
tives of fo u rth order occur in th e bipo ten tial equation. T he m ost im p o rtan t
special cases a re listed in T able 8.1.

96

8 Disks

$(x,y)
2

yy

Tx y

com m ent

ao2 y

2 a 02

an xy
2
a20X

-a u

2 a20

constant tension in y - direction

ao3 y3

6 a 03y

pure bending m om ent Mx

pure bending m om ent My

a3 0 X

a3 0 X

constant tension in x - direction


constant shear

T a b le 8.1: States of stress in power approaches


For

i + k > 4

AA<P = 0 is only fulfilled if single constants a ik satisfy th e necessary coup


ling conditions.
-

A pproach w ith F O U R IE R series e x p a n sio n

P eriodic fu n c tio n s
A load is given as a periodic function along a boundary, or it varies peri
odically.
FO U R IE R expansion of a boundary load q ( x ) , 0 < x < /
q ( x ) = a 0 + 2 an cosQ Hx + 2 bn sinQ n x
n
n
w ith

= ~ ~ f~
1 r
a0 = T J q ( x ) d x
o

(812a)

(n = 1 , 2 , 3 , . . . ) ,
,

2 r
an = j - J q ( x ) c o s n x d x ,
* o

(8.12b)

2 rl

b = T J q ( x ) s i n a n x d x .
o
E xpansion of a stress function in FO U R IE R series in case of an odd func
tion
* ( x , y ) = 2 ^ n ( y ) sinanx
n

( 81S)

T ransform atio n of th e homogeneous bip o ten tial equation leads to an ordi


n ary differential equation w ith constant coefficients for every $ ( y ):
* n , y y y y - 2 n * n , yy + n

- ( d/dy = ,y ) .

(8.14)

8.2 Analytical solutions

97

Solutions to (8.14) :
*n(y)=
or

m=l

* n ( y ) = - ^ - ( A n c o s h a n y + Bn a n y c o s h a n y +
+ Cn sin h n y + Dn a n y sin h a n y ) .

(8.15)

N o n -p e r io d ic fu n c tio n s
Load described by th e FO U R IE R integral form ula
^ oo

q(x ) =

+ oo

J [ cos a x J q ( ( )

^ 0

-o o

-f-oo

oo

(8.16)

+ v / [ sinax
w 0

cos a ( d ] d a +

sin a d ] d a .

-o o

A pproach w ith com plex stress fu n c tio n s

In stead of th e re a l variables x an d y, th e complex variable z = x + i y


its com plex conjugate ~z = x - i y a re introduced. Because of
x = ~ (z + z )

y = y i ( z - z ) ,

and

(8.17)

follow th e derivatives
4.

dx

dz

a
ay

.( d
HVdazz

-V

dz

_ _ a!_

ax2

a2

O ' " A_ A~

Sz2

dzdz

'

a2

d"z2
(8.18)

d \
a z ^/

d2 _
ay2

d2 . a2
az2 T \

s2
az2

and the LAPLACE operator


2

A = - ^ + - ^
ax2
ay2

4
azaz

'

A pproach for a stress function in complex n otation ( cf. [ B.2, B.5 ] ):


$ ( z , z ) = -i-["zy>(z) + z t p ( z ) + j ip ( z ) d z + J ip ( ~z ) dz ] .
Equations for determ ining states of plane stress an d th e d isp lacem en ts:
<7** + ^yy = 2 (<P' + <P') = 4 R e y / ( z ) ,
^yy - *** + 2 i 7 xy = 2 ( z <p" + V ) ,
2 G ( u + i v ) = - z i p ' - ip + Ktp

} (8.19)

98

D isks

w ith

= f , v
l + i/

for sta te of plane stress ,

= 3 - 4v

for state of plane stra in .

T he superscrip t prim e ' denotes derivatives w ith respect to z or z . i p an d ip


are two a rb itra ry an aly tical functions by m eans of w hich all stresses and
displacem ents can be calculated.
GOURSAT already presented th is solution at th e tu rn of th e century, and
KOLOSOV im proved th is procedure w hich was augm ented an d presented in
detail by MUSKHELISHVILI [ B .6 ].

b)

P olar coordin ates

- A x i s y m m e t r i c a l states of stress $ = ^ ( r )
D ifferential equation from (8.6) w ith

,rr rr

S olution:

+
#
f

,r r r

<? = C0 +

d/dr = (

+ 4 r 3^

j.2

rr

),r

, V = 0 , <9 = 0

=0.

(8.20)7

'

r 2 + C2 In ^ + C3 r 2 In

(8.21)

w here a denotes th e reference length.


Stresses
aIT

,T

a(fitfi

= ^

f TT

- R a d i u s - i n d e p e n d e n t states of stress
D ifferential equation from (8.6) w ith
<
Solution :

+ 4

=Ttfi0 .

(8 .22)'

'

$ = $( <p)

d / d ip = ( ),v

= 0 .

(8.23)

$ Cj ip + C2 + C3 cos 2 <p + C4 sin 2 tp .

(8.24)

Stresses
=
-

* vv =

R a d ia tin g states o f stress

Stress function

TtV = - ( t

* ' V ).t -

(8

r iifi = 0

^ = r g ( y >) + h ( r ) ,

( 8.26a)

w here h ( r ) denotes a n axisym m etric state of stress according to / 8.20) .


D ifferential equation for g ( ip )
S .v v v v + 2 &,w + 6 = 0
Solution :

( 8-26b)

r cos <p + C2 r sin tp + C3 r tp cos tp + C4 r tp sin ip ( 8.27)

9.1

D efinitions - A ssum ptions - Basic equations

99

Stresses due to (8.7)


a

vv

=0

(8.28)

T lip = 0

- N o n - a x is y m m e tr ic sta tes o f stress


F u rth e r solutions are obtained from (8.6) by m eans of separation approach
es of th e form r 11cos n ip ( n > 2 ) [ see Exercise B-8-4 ].
-

C om plex stress fu n c tio n

T ransform atio n of (8.19) into polar coordinates w ith z = x +


yields
aTT +
= 4Rec//( z) ,
a TT + 2 i T n? = 2 ( Z tP " + V ) e

iy = re '

2it?

2 G ( u + i v ) = ( - z tp' - ip + ni p) e

(8.29)
-it?

w here u, v are th e com ponents of th e displacem ents in th e r - an d


tion, respectively (se e [ B. 5] ).

P la tes

9.1

D e fin itio n s A ssu m p tio n s B a sic e q u a tio n s

direc

A plate is a stru ctu re like a disk w ith sm all thickness t in com parison w ith
o ther dim ensions. T h e plane w hich halves th e plate thickness is called the
m id -p la n e . As shown in Fig. 9.1 a ), th e plate is subjected to surface loads p
perpendicular to th e m id -p la n e . A n a rb itra ry load is resolved vertically
an d parallel to th e surface. T he in - p la n e forces can th e n be dealt w ith by
m eans of th e disk theory ( Ch. 8 ). T he in terest in this chapter is restricted
to th e influence of th e tran sv erse loading on th e plate. T h e thickness of the
plate is assum ed constant in th e following.

F ig . 9.1 : a ) D im ensions an d loads of a plate


b ) Sign convention for stress resu ltan ts of a plate elem ent

100

Plates

a ) P l a t e s i n C a r t e s i a n c o o r d in a t e s
- S h e a r -e la s tic , isotropic p la te
D isplacem ents of an arb itra ry point P at a d istance z from th e m id -p la n e
( cross - sections rem ain plane, see Fig. 9 .2):
u ( x , y , z ) = zi px ( x , y ) ,
v ( x ,y ,z) = z^y(x,y) ,

(9.1)

w ( x , y , z ) w ( x, y , z = 0)
w ith th e bending angles ipx an d ipy
S train - displacem ent
a / d y - ( )-y)

relations

from

(4.14)

w ith

XX= Z ^X,X>

Txy = ( ^ X,y + ^ y,X)Z ,

eVyy
= zil>
,
Xy,y
V
r v . v l,

-7tlyz
vz

ez z = 0 ,

n/'zx vr #x + *x .

(9 .1 )

(d/dx=

( ) , x,

= il>
+ w,y
,
Yr y
v
.V

Stress resu ltan ts ( defined per u n it length of a line y = const or x = const


in th e plate m id - p la n e ) :
+t/2
+t/2

=J

M xx =

yy =

xxzdz

J* cryy z dz

Mxy = Myx

J r *y z d z

torsional m om ents

(9.3 a)

-t/2

+t/2
Qx

bending m om ents

-t/2

-t/2

+t/2
T XZ

dZ .

Q y =

-t/2

tran sv erse shear


forces

dZ

-t/2

T he sign convention consistent w ith ( 9.3a) for th e stress resu ltan ts is show n
in Fig. 9.1 b ).
D efinitions ( 9.3a) w ith th e m aterial law (5.12) an d (9.2) lead to th e stress
resu ltan t - deform ation relations
K (^x.x +

M yy = K v(ip
Yy <y + v ib
x ,x )/ >
l-i/
Mxy = My x = ^ - K ( ^ x + ^ r y) ,

(9.3b)

Qx = Gts(^x + W,x) Qy

= G t 8 (V<y +

w ith th e plate stiffness


sh ear thickness t < t .

Wi y )

K =

E t3
12 ( 1 - v l )

, th e sh ea r m o d u lu s G , a n d th e

9.1 Definitions - Assumptions - Basic equations

101

E quilibrium conditions
Qx ,x + Qy,y + P
M xx,x +

= 0

M xy,y Q x =

My*,* + M yy^y

Qy

(9.4)

0 >

0 .

T he relations (9.3a) an d (9.4) result in eight equations for th e eight un


knowns (fiv e stress resultants, th ree deform ation quantities).
R e d u ctio n o f the equations
T he com binations
$ = ih
TV

.X

+1 ih
TV.V

= lb
^ytx ib ,y

(9.5)

are th e basis for th e derivatives ( see [ B. 5] ):


W.x = - ^ x + ^ r ( ^ x - 1 i L ! ? . y ) .
(9.6)
+ 1 ZJLy

w.y = ^ y + C T r K -y T

T his yields th ree equations for th e th ree unknow n functions w, $ an d S ':

KA< = P
- Aw =

A! ? - KS V =

w ith the sh ear influence factor

(9.7)

l-i/ K
2
Gt

102

9 Plates

By m eans of a n additional auxiliary function


K
w_ = w - 7V=nr
^
j t

(9.8)

two uncoupled equations a re derived


K A A ws = p
(9.9)

- A !? = 0

Hereby, a p a rtia l differential problem of th e six th order is generated. T hree


quantities can be prescribed a t each boundary [B .5].
- Sheairig id , iso tro p ic p la tes w ith te m p e ra tu re g ra d ien t
For such plates th e sh ear stiffness G ts > o o , i.e. th e term s m ultiplied
K / G t s can be neglected. From (9.6) follows:
Vx = ~ W fX ,

Vy = - W,y .

by

(9.10)

T his m eans th a t after deform ation a norm al to th e m id - p la n e rem ain s a


norm al. T hus, no sh ear deform ation occurs in cross direction ( 7 yz = 7 XZ =
0 ) = > K IR C H H O F F 's P la te T h eo ry.
M aterial law - stress resu lta n t - displacem ent equations due to (9.3)
M XX =

K K

X +

" W y y

( !

^ ) T l 0 ] .

Myy = - K[ w yy + v w xx + (1 + i / J o j , 1] ,
M xy =

K ( 1 -

(9.11)

" H x y

w ith th e co n stan t tem p eratu re grad ien t 10 ( x , y ) through th e thickness of


th e plate.
T ransverse sh ear forces from (9.4)
Qx = - K ( A w ) _ x - ( l + , ) a T K 1e x ,
(912)
Qy = - K ( A w )

- (1 +

)aTK

Note-. As th e shear deform ation vanishes, no law of elasticity for Q x and


Q y as in (9.3b) exists.
Basic equation of K IR C H H O F F 's p la te th eo ry
K A A w = p - a T ( l + i / ) K A 10

(9.13)

T he above equation is a p a rtia l differential equation of fourth order. At


each boundary only two boundary conditions can be fulfilled.

9.1 Definitions - Assumptions - Basic equations

103

B o u n d a ry co n d itio n s at a b o undary x = c o n s t:
-

Free boundary
Mxx = 0

or

Qx = Qx +

w (xx + v w yy + ( l + i / ) a T 10 = O ,

M xy,y =

( 9.14a)
1

1
r

W, x x x +

w here

( 2 -

" ) W,y y x +

( 1 +

" ) T

Q x = - K [ w xxx + ( 2 - v ) w yyx

>
x =

0 -

(9M b)

+ (1 + v ) a T l0 x ]

(9.14c)

is one of th e KIRCHHOFF s effective tra n sverse sh ear forces.


-

Simply supported boundary

or

w = 0

Mxx = 0

( 9.15a)

w =0

Aw = 0.

( 9.15b)

Eqs. ( 9.15b) are called NAVIER s boundary conditions.


-

C lam ped boundary


w = 0

w x = 0 .

( 9.16)

Analogous boundary conditions can be form ulated for a boundary y = const.


-

C orner force
A = 2 Mxy = - 2 K ( l - i / ) w Ky

(9.17)

D eterm ination of m axim um stresses


a
-

xxm ax

^2

o-

yymax

= 6 -^ ^
j2

xymax

= 6 ^f^.
|2

(9.18)

T ra n sv e rse ly v ib ra tin g iso tro p ic plate

D ifferential equation for free transverse vibrations


(9.19)

w here r denotes tim e and p = g t denotes th e m ass per u n it plate a rea (g


m ass density of th e plate m aterial).
Product approach due to D. BERNOULLI for th e calculation of n atu ral vi
brations :
w = w ( x , y ) T ( t ) .

(9.20)

104

9 Plates

D ifferential equation for the tim e -in d e p e n d e n t vibration m ode w ( x , y ) :


A A w = A4 w .

(9.21)

For an overall sim ply supported plate as exam ple, the n atu ral angular fre
quencies wmn are calculated from
,4

2
MWmn

K
R efer to th e s h e a r-rig id plate (9.14) to (9.16) for th e b o u n d a ry co n d itio n s
to (9.21).
-

S h e a r - r ig id , o rth o tro p ic p la te

M aterial law - stress resu ltan t - deform ation equations according to (9.11)
M XX = - K X( Vw , x x + I /yW , y y )/

M yy = - K y( vw yy - f i /x w , x x )/ ,

(9.22a)

Mxy = - 2 Kx y w #xy

w ith th e stiffnesses

K*

I 2 ( i - i / x i/y )

Ky

12(l - vx vY )

K*y

12

(g m )

T he equilibrium conditions are th e sam e as in th e case of th e isotropic


plate ( see ( 9.4b) ).
H U B E R s differential equation
K xw

rxxxx

+ 2 II w ,xxyy + K yw

,y y y y

= pF '

(9.23)

w ith th e effective to rsio n a l stiffn e ss


2 H = 4 Kxy + v x K y + v y K x .

(9.24)/

b ) P la te s in p o la r c o o r d in a te s
- S h e a r-r ig id , iso tro p ic c irc u la r plates

(hr

D ifferential equation due to (2.40) an d (2.49) w ith d / dr = ,r , d / dtp =


A A w = '( w ,rr H1
-2K - aa T

1
r

w,x

i
,
Hf 2- w ,ipip )>

^ ^ HP
'

9.1 Definitions - Assumptions - Basic equations


M aterial law M = - k [ iv , r r
M

Wfi

= -K

stress re su lta n t-d isp la c em en t relations


+

Y w ,r +

( 1 -

Vr

>

^ 2 W,V V )

i/) K ( w

105

' Tv

f'W.rr +

"7
j.2

,<p 1

( 1 +

v ) T 1 0

(9.26)

v ) T ^

Effective transverse sh ear forces


Qr = - K

( A w ),r +

( 7 W,rv - ^2 W, v \ v +
(9.27a)

+ (1 + v ) a T ( 10 r cos ip - 10 v - ^

- k

(9.27b)
+ (1 + i / ) a T ( 10 r s i n v> - lG 'V E ^ V )] .
-

T ra n sv e rse ly v ib ra tin g c irc u la r p la tes

D ifferential equation for th e tim e - independent vibration m ode w ( r , tp) according to (9.21) [ B.8 , B.9 ]
( W,rr +

W, r + ^ W | W )

A4 W

A4 =

W ith

pLW
K

(9.28)

S eparation of th e vibration m ode


OO
w ( r , y > ) = R ( r ) - $ ( i p ) = 2 R n ( r ) c o s n ip
n= 0
>

(9.29)

BESSEL s differential eq u atio n s:


d2 R
! dR
/
211
+
^
+
d r2
r
dr
V
d2 R
d r2

i
r

dR
dr

/ 2
VU

n^
r2

)/ Rn = 0

n
^7 ) ) Rn = 0
n2

(9.30a)

(9.30b)

T his type of differential equation is dealt w ith in [ B.3 ].


c)

P l a t e s in c u r v i l i n e a r c o o r d i n a t e s

E quilibrium conditions

= 0,

M ^ L - Q = 0 .

QL

+p

(9.31)

106

9 Plates

M aterial law

stress re su lta n t-d isp la c em en t relations

M0* = - K E*aPyS ( w L + a T a 4 !0 )

(9.32)

with
K
a.ap , a

plate stiffn e ss,


qc(3

com ponents of th e m etric ten so r in th e m id - plane of th e p la te ,

1
CX
& (( )

tem p eratu re g ra d ie n t,

E *aPl/6 =

v . ( a &7 a.PS + 3.a6

) + v a^ a?6

plane elasticity ten so r.

D ifferential equation
(9.13)

K A A w = p - a T (1 4- v ) K A *0

w ith th e LAPLACE operator A given by (2.39) in term s of th e applied cur


v ilinear coordinates.
E n e r g y e x p re ssio n s
-

C a rte sia n coord in a tes ( d / d x = ( ),x , d / d y ( ),y ) from (6.16a) [ A.9]

ni =

JJ {\ K[WL

+ w 2yy + 2 ( 1 - t / ) w ;2xy + 2 v w xx w<yy ] +

+ K a T ( l + i / ) ( w xx + Wy y ) 1 j d x d y
(9.33)
ni

= J J { i K[ ( W,xx +

W,yy f -

2(1 - 0(WfXXWfjy " Wxy)] +

+ K q t (1 + ^ ) ( w jXX + Wy y) 1 j d x d y .

Polar coordinates
K

W rr +

= ( ),r ,

+ ( 1 - v ) ( W,r*p - f W,rV W, v + ^2 W,2J

2 t/ ( w W f W

d/ d<p

- w i I T W fr )

+ l1w 2
-

(d/dr

(9.34)

-f
w .Ifiifi w ,r -f 2irw r
2 /')|J J| rd co
dr .
r
^

C u rv ilin e a r co ordinates ( w ith a = | aQ^ | )

E i = J J ( y KE*^

wL wU + O y K a ^ E ^ w ^ e l ^ d ^

107

9.2 Analytical solutions for shear-rigid plate:


9 .2

A n a ly tic a l s o lu tio n s fo r s h e a r - r i g i d p la te s

a ) C a rte sia n co o rd in a tes


-

S im p ly su p p o rted p la te s tr ip (

d/dy = 0

D ifferential equation from (9.13)

, d/dx

( )

AAw = w xxxx = ^

)
.

Pn
. n 7Tx
s in ------= y
^(nTr)4

Solution:

(9.36)

n 7TX

sin ( x ) = 2 P n si

for

w ith

pn = j -

J p ( x ) sin

dx

( n = 1,3,5,...) .

R ectangular plate w ith sim p ly supported b o undaries ( dim ensions a, b ;


Fig. 9.3 )

D ifferential equation from (9.13) ( d/d x = ( ) x , d/d y = w y )

p (x,y)

A A w = w , x x x x +' 2 w #xxyy + w ,y y y y

S olution: Double series expansion according to NAVIER


/

V V

mwx

nw y

w( x .y) = Zm Zn wmnsma ~ sln_b


t
j
Load

( m-n =

. Q

r
\
W
m7rx . n w y
p f x . y j ^ ^ Z j PmnSln
a sin b
'
m

E xpansion coefficients

(9.37)

Pmn

(9.38a)

21

2c
2d

fcr

p0(x>y)

2d

JC

F ig . 9.3: P late under a uniform ly distributed load over a rectan g u lar subdom ain

108

9 Plates

A plate subjected to a uniform ly distrib u ted load p0 ovei


dom ain as shown in Fig. 9.3 will be considered as an
tion.
W e first expand th e constant load p0 in th e y -d ire c tio n
/

p(y) =

V
niry
Z P n s m b~
n

w ith
b

v+d

2 f
niry ,
2 f
niry ,
Pn = Y J Pos l n - b
dy = b"
Pos l n - b ^ dy =
n
vv-d
-d

2 b
(
n7r y \
= - b 7 T po l - cos b i

iV~fd
iV+d
,

Iv - d

P0 0 . n 7r v . n 7r d
= 2 -----2 sin r s i n r
nw
b
b

(
to o
\
(n = 1 , 2 , 3 , . . . ) .
v

Ensuing, this p ( y ) is expanded in th e x -d ire c tio n


/
x v V
m 7 r x . n7ry
P ( * , y ) = Z Z P m n S ln- - T ~ S m b m n
a
2 f
mirx ,
p = p s i n
dx .
rmn
a J *
a

...
w ith

T h e calculation yields
. m w u . mwc
j s m - sm
m ni
a
a
Po

Pmn =

16

n7rv . n 7rd
- r D
D

a n - j - s m

(m ,n = 1,2,3,...
H erew ith, we obtain w ith ( 9.38b)
\ \ '
Pmn
. m7rx . m r y
w = > > -------------------------------- =-s i n --------- s i n =-=*-

K K

)]

Tw o special cases:
F ull load

----- >

It follows t h a t :
^ Po W

w = t;

c = u = a /2 ,

pmn = 16
_
m
m ni
^

d = v

= b /2

an d
.

m 7r x .

n 7r y

---------------- ;;---------------- 5- s i n ---------s i n 5

(m ,n = 1

9.2 Analytical solutions for shear-rigid plates

109

H erew ith, for a quad ratric plate ( b = a ) th e m axim um deflection ( found at


th e centre ) becom es
( a

a \

I6 P0 a ,
K tt6 I 4

1
300

1 +
1
300
9 324

)
' /

O ne can discern a fast convergence, p articu larly as th e higher term s have


alte rn a tin g signs.
Single load F at th e point u , v
W e ex ten d th e expansion coefficients
m 7r c
nw d
a

s in -------- sin j
4 . ,
. m7ru . n w v
a
b
p
= t- 4 c d p . s in -------- sin r ------------------------ t------mn
ab
0
a
b
m ire n r d
a
b

in such a way th a t th e rectangle can b e reduced to a point. W ith th e lim it


ing value
lim

si n/ cc

c - o

an d

= 1

sin/id

l i m ----- = 1

d -.o

KC

lim 4 c d p0 = F ,
c > 0
d 0

...

we obtain
-

4 F . m 7 r u . n 7 r v .

mn

= ts in -------- sin r
ab
a
b

i o q

(rn,n = 1 ,2,3,...)
v

Plates w ith two parallel, sim p ly supported boundaries and other bound
aries a rb itra ry

D ifferential equation ------ > (9.13) .


Solution approach according to L E V Y :
w ( x , y ) = 2 wn ( y ) s i n
n

( 9M )

T ransform atio n of (9.13) into an o rd in ary differential equation w ith con


sta n t coefficients ( d /d y = ( ),y ):

Wn ( y ),yyyy 2 ( ^
Homogeneous so lu tio n :

^ * ( Y ) , + ( ) ' wn ( y ) = ^

g ( Y) -

(Ml)

110

Plates

P la tes w ith a r b itr a r ily su p p o rted boundaries

If th e surface load can b e considered as a product sim ilar to (9.37), closed


form solutions for plates w ith m ixed supports can still be found. W e are
going to explain th e solution approach by th e exam ple of an overall clam p
ed plate according to Fig. 9.4.
Since no solution is know n for th e overall clam ped plate w hich fulfills both
th e differential equation an d th e b o undary conditions, we sep arate th e pro
blem into th e following th ree subproblem s according to usual m ethods of
stru ctu ral engineering, and obtain th e solution using th e superposition p rin
ciple :
0 T he overall sim ply supported plate under uniform ly distributed load
w ith th e solution w0 following from (9.39).
1

An overall sim ply supported p late w ith a yet unknow n m om ent d istri
bution MXXj
> solution W j.

2 An overall sim ply supported plate w ith a yet unknow n m om ent d istri
----- *solution w2 .
bution M
F rom th e geom etric boundary conditions - th e bending angles have to van
ish at th e supported boundaries in th e superposition, i.e.,
,

b
X. Y :

W 0 ,x +

W l,x +

W 2 ,x =

0 *

(MS)
vb,y + w i,y + w 2,y = 0 -

From (9.43) follow th e previously unknow n m om ent distributions, a n d from


w = w0 + Wj + w2
we obtain th e general solution.
In case of non - sym m etrical support th e p a rtia l solutions becom e m ore
com plicated an d th e num ber of geom etric boundary conditions increases.
H erew ith, a solution in closed form can be only theoretically established at
th e expense of m ore work. Thus, th e use of an energy m ethod would be
m ore effective [ B.8].

F ig . 9.4: P la te w ith a ll ed g es cla m p ed

9.2 Analytical solutions for shear-rigid plates


-

111

O rth o tro p ic p la tes [ B.5, B.9 ]

F rom th e series expansion wh = ^


acteristic equation
m

eAmY sin m ^ x follows th e char

W ith th e four roots

th e solution procedure depends on th e radicand. W e distinguish betw een


th ree types of p la te s :
2
1. Type: H > Kx Ky =

plate of high stiffness against torsion.

Since th e bending stiffnesses are always positive, th e in ner root is less th an


H. Hence, all four roots are real. T h e solution is valid for Kx ( or Ky ) x 0.
T his corresponds to a plate w ith a negligible bending stiffness in th e x - ( o r
y - ) direction. T h is assum ption is valid if th e plate is of very high stiffness,
w hich m ay be achieved by m eans of b o x - ty p e ribs in th e y - ( o r x - ) direc
tion ( Fig. 9 .5a) .
2. Type: H Kx Ky

approxim ation according to (9.35).

W e find th e double roots


\

_ \
mi

m2

_ m 7T / H
a ] / K,

_ m 7r 4 / Kx

a ]/ Ky

m3

m 7 r 4 / Kx
a ] / Ky

'

T his type occurs w ith a crosswise reinforced concrete plate as shown in Fig.
9.5b.

F ig. 9.5: O rthotropic plate profiles

112

9 Plates
2

3. T y p e : H < Kx Ky = plate of low stiffness against torsion.


In this case, th e roots are com plex conjugate

For negligibly low stiffness against torsion H O w e obtain


A

i , 2 ,3,4

=
a

J i [ K
V
y Ky

= i 1
a 2 r

y Ky

T his solution occurs, e.g. in cases of plates stiffened by bending profiles th a t


have very low torsional stiffness ( Fig. 9.5c ).
b)

P o la r co o rd in a tes

- A x is y m m e tr ic a l load case
T h e loads an d boundary conditions a re independent of <p * p = p ( r ),
w = w (r).
From (9.25) we obtain E U L E R 's differential equation
w ,rrrr + \

rrr ~ \

W , + ~

W ,r =

IY

(9.46)

Homogeneous solution:
wh = C0 + q r 2 + C2 ln { - + C3 r 2 l n ^

(9.47)

w ith a suitable reference length a .


- N o n -s y m m e tr ic a l load case
Load

p ( r ,v3) g ( r ) ^ P n c o s n ip

( n in te g e r) .

(9.48)

E xpansion approach for deflection

w ( r ,</?) = ^ w n ( r ) c o s nip .

(9.49)

T ransform ation of (9.25) into an o rd in ary differential eq u atio n :


Wn = ^ P n S ( r ) -

( 95 )

n = l:

solution ( 9 .4 7 ) for the axisymmetrical load case ,


Q
whl = Cj r + - ^ + C3 r3 + C4 r In
,

(9.51a)

n>2:

whn = Cln rn + C2nr n + C3nr 2+n + C4nr 2- n .

(9.51b)

n =

10.1 Isotropic, plane structures with Large displacements

10

C o u p l e d d i s k p l a te p r o b l e m s

10.1

I s o tr o p ic , p l a n e s t r u c t u r e s w ith la rg e d is p la c e m e n ts

113

- B a sic eq u a tio n s
In th e previous chapters we considered elastic structures w ith sm all dis
placem ents. T his sim plifying assum ption is not always fulfilled; especially
in cases of th in - w a lle d structures subjected to larger com pressive loads, th e
deform ations m ay becom e large com pared w ith th e thickness. T h e equilibri
um conditions m ust th e n be form ulated for th e deform ed state of th e stru c
tu re and term s of higher order m ust be taken into account in th e s tra in deform ation relations. T his corresponds to th e g eo m etrica l n o n -lin e a r ity .
Here, th e m aterial law is considered to be linear. F urtherm ore, th e lem m a
of m ass conservation ( g dV = g d V ) is assum ed to rem ain valid as w eh as
equality of th e volum e forces in th e deform ed an d undeform ed sta te ( f =
f ) . T h e s tr e s s - f r e e initial state (LA G R A N G E fo rm ulation) is tak en as a
basis. W ith these assum ptions, th e equilibrium conditions read as follows
[ B.l, B.2, B.4 ]:
[K

v i |k ) T j k ] | j

+ fi = 0 .

(10.1)

T h e stra in - displacem ent relations have been introduced in C hapter 4, and


we obtain according to (4.12a)
^

= 1 ( 4

+ 4

+ v k l i v k lj)

( l0 )

As th e stra in s are assum ed to be very sm all in com parison w ith th e defor


m ations, n o n -lin e a r term s can be neglected in th e com patibility conditions.
T h e six equations of th e m aterial law are adopted in th e ir usual form ( 5.5a)
or ( 5.6a) .

F ig. 10.1 : P la n e lo a d -b e a rin g stru ctu re under tem p eratu re and surface loads

114

10 Coupled D isk-Plate Problems

In C hapters 8 an d 9, disks an d plates w ere considered separately because


th e ir loads w ere assum ed to act eith er in th e m id - p la n e or p erpendicular
to th e m id - p la n e of th e structure. Now, we extend our consideration to th e
coupled d is k - p la te problem. Besides, it will be assum ed th a t th e plane
stru ctu re is subjected to an arb itra ry tem p eratu re field ( Fig. 10.1)

(io.s)

0 U a ,C) = 0 ( O 4
with

e ( C ) = \ [ 0xu a ) + 2 U a ) ] ,

1&{ C ) = Y [ e i ( 0 - e a ( 0 ]
and an ex tern al surface load p ( ( a ). At its boundaries, it m ust satisfy pre
scribed bo undary conditions. In th e following, we w ill re strict ourselves to
th e s h e a r - r ig id p lane structure, but assum e large deform ations.
T reatm en t of th e problem by m eans of th e H E L L IN G E R -R E ISSN E R energy
functional:
-

Stress resu ltan ts - tensors

In additio n to th e m om ent tensor Ma^, we introduce th e tensor of in - p la n e


forces Na/3 (m e m b ra n e ten so r). According to Fig. 10.1 w ith ( 3 (, they
read:
+
2

+ -2

N /3 = J TPd (

Ma/3 = J r a p ( d ( .

_t

_t

( 10A )

S train -d isp lacem en t equations

A plane lo a d -b e a rin g stru ctu re is subjected to strains ' f a p an d distortions


l la p
f th e m id -p la n e . T h e to tal strain s at an a rb itra ry point can be super
posed from these two parts. According to (10.2), th e following s tr a in - d is
placem ents relations a re valid
\ / 3

V J / 3

Vp L

W ,a

W.p

'

( 10-5a)

A ssum ing va : w, only th e n o n -lin e a r term w a wi(5 is ta k en into considera


tion in ( 10.5a).
T he disto rtio n of th e m id -p la n e l ^ a p is obtained from th e relations of th e
s h e a r-r ig id plate, w here th e cross sectional rotations are expressed by th e
angles of th e bending surface w,a (K IR C H H O F F 's norm al hypothesis). T he
following relation is valid for th e d istortion of th e s h e a r-rig id plate
\ p

= - wL

(10.5b)

10.1 Isotropic, plane structures with large deformations


W ith t ap an d l'fa/3 we form th e to tal stra in
1*13 = \ p
-

+ ( \p

115

aj3

( 10-5c)

M aterial law

T h e m ate ria l law is used in th e form for a sta te o f p la n e stre ss of the


body ( r 33 k 0 ). W ith th e tem p eratu re field (10.S) and th e stra in tensor
com ponents (10.5) we obtain th e following relations for th e stress resul
tants
r f * = t E ^ 4 ( \

s - a T a7, <9) ,

M a/? = g - EaPyS ( \

(10.6a)

6 - a T a 7, *0 ) ,

(10.6b)

w here, after substitution of ( 10.5b), th e equation ( 10.6b) becomes identical


w ith (9.32) an d w ith th e elasticity ten so r
defined in connection w ith
(9.32).
V a ria tio n a l fu n c tio n a l
W e shall now derive th e differential equations an d boundary conditions for
th e coupled d is k - p la te problem by m eans of an energy functional w ithout
work contributions from volum e forces, boundary loads, an d boundary dis
placem ents [ E T 2 ] . If we substitu te th e deform ation energy U of (6.16a) into
(6.29) for th e th re e -d im e n s io n a l body, th en first follows
nK =

+ T gije ) ] d v

(*-7a)

and w hen regarding th e plane stru ctu re as a tw o -d im e n sio n a l body

= J {^

IccP ~ \

(Icfi + a T g ap & )}dV

( 10'n )

In ( 10.7b) we have o m itted all term s w ith r 33 because of th e th in -w a lle d


stru ctu re ( r 33 0 ) an d w ith 7 a3 because of th e s h e a r -rig id behaviour ( rya3
~ 0 ) of th e p lane structure. By introducing into ( 10.7b) th e stress resu ltan ts
according to (10.4) an d th e tem p eratu re field according to (10.3), th e func
tional becom es
D

N0^ N74 +

( 10.8 )
+ ^

Da/374M ^ M 74 + 2 a T N 0 + 2 a T M 1 ] ] d A .

116

10 Coupled D isk-Plate Problems

W ith
^ap-yS = E ^ap-yt =

^^

&P6 + &ct6 &Py ) V &ap &yS

and th e plate stiffness K defined by (9.3), th e functional takes the form

n K = J J { y eCT/?T* L ( vJ p + vp L + w ,a w ,p ) +

+ f

C*^
B * * * w | ^ w |7, + K a T &ap E

|7/ 00
w |7,

(10.9a)

V*aPy6$\ap$|7, + 21aT< e]}fi d^1d2


Abbreviated th e functional (10.9a) h as th e form

n n = J F ( ^ ; va . vJ ^ ; w . w , a . w U ; ^ U ) dA -

(10M)

where th e th ree functions va , w , # are unknow n.


T he ex tern al p o ten tial /7e in ( 6.18b) for th e work of th e surface loads p ( ( a )
is given by

( 10.10)

77e = - W = - J p w d A .

T h e to tal p o ten tial is now superposed from ( 10.9a) and (10.10)

n = nK+ ne = nK- w .

( 10.11)

In C artesian coordinates th e to ta l p o ten tial of th e coupled d is k - p la te pro


blem is expressed as

n = JJ { 1[

( v -y + T wi } + # -yy( u - + Y w- } "

* x y ( U,y +

V, x

W, x W y ) ]

,xx + w,yy+ 2 ( l - i / ) w xy + 2 i / w f w >yy +


(10.12)

+ K a TTv( 1 + i / /) (Vw , x x + w, yy )/ *0
4 ' $ ,2x x + S#yy
'2 - 2 v $ , x x $ , y y - 2** V
( 1* +1 1"/ )) $ 2
2E
4 q t

( ^ ,x x

^ ,y y ) 0@ } d x d y

JJ

p w d x dy

10.1

117

Isotropic, plane structures w ith large deformations

From th e sta tio n a rity condition 6 II = 0 ( see (6.20) ) in anology w ith ( 6.S5)
now, th e equilibrium condition (10.1Sa), th e com patibility condition (10.13b)
an d th e bo u n d ary conditions follow as :
VO N K A R M A N ' s d ifferen tial equations
K A A w = p - K a T ( l + i / ) A 10 + t 0 4 ( w , # )
AA $

(10.13a)
( 10.13b)

= - E a TA 0 - ^ - O 4(w ,w )

or in index notatio n ( see [ Bl, B.2, B.8 ] )

K wG = p - KM 1 + " )
*15

= - E a T e \l

le\l + 1^ wL * U -

- y a>1w ^ w |ap .

(10Ma)
( 10.14b)

(10.13a) = equilibrium condition of th e forces in th e z -d ire c tio n in cases of


large deform ations,
(10.13b) = com patibility condition of th e coupled d is k -p la te problem .
T he operator 0 in (10.13) is defined in C artesian coordinates by
0v4 (' f , g6/) = f ,xx,yy
g - 2

'

f,xy,xy
g + 1 f ,yy,xx
g .

(10.15)
1

B o u n d a ry conditions for boundaries y = const or x = c o n s t:


-

Simply su p p o rted boundary


w = 0,

w = 0 ,

yy

0.

( 10.16a)

w x = 0 or

= 0

=0 .

(10.16b)

F ree boundary
M xx = 0

or

or

C lam ped boundary


w = 0

M vv x=x 0

M yy = 0

,* N x x w #x + N x y w /V +1 Q
=0
~x
_
, N xyw ,x + N
x'yy w,y + Q
v*y = 0 .

(10.16c)

Note: Due to th e equilibrium considerations for th e deform ed elem ent, the


tran sv erse force conditions contain additional contributions from in
plane com pressive and sh ear forces in ( 10.16c) .

118

10 Coupled D isk-Plate Problems

F ig. 10.2: P late under in - p la n e compressive an d shear forces

F ig. 10.3: C haracteristics of a


bifurcation problem

Special case: B a sic equation fo r p la te buckling


For p = 0 , l6 = 0 an d th e operator fully w ritten, th e differential equation
( 10.13a) reads
K M w = t (v$ #yy wixx + ^ /xx
w

/yy

- 2 # #xy w#xy ')

(10.17)'

If we introduce, by m eans of (8.2), th e in - plane forces Nx = t <rxx a n d Ny =


t c yy as well as th e shearing force Nxy = t r xy for the derivatives of the
stress function, an d if we tak e th e com pressive forces to be positive, we ob
ta in th e following differential equation for plate b u ck lin g :
K A A w + Nx ( x , y ) w xx + Ny ( x , y ) w yy + 2 N xy( x , y ) w xy = 0

(10.18)

T he solution of th is equation leads to a bifurcation at a critical load ( Fig.


10.3)

10.2

L o a d -b e a rin g structures m ade of com posite m aterials

T he use of structures m ade of com posite m aterials will be steadily increas


ing because of th e possibility of ta ilo rin g th e ir characteristics. Thus, very
dem anding technical requirem ents can be specified for com posites, which
cannot be achieved w ith conventional single - com ponent m aterials.
O ur m ain interest h ere will be directed tow ards com posite m aterials w ith
glass an d carbon fibres. It is characteristic for a com posite m aterial ( Fig.
10.4 ) th a t th e fibre com ponents of a single layer ( l a m i n a ) are all oriented
in th e sam e direction an d em bedded in a m a trix m aterial. We call such a
layer a unidirectional layer or, in short, a U D -lay er. C haracteristic p aram e
ters of a U D -lay er (Fig. 10.4) are

10.2

Structures made o f com posite materials

119

fibre orientation

F ig. 10.4: M ultilayer com posite consisting of stacked single layers


- layer thicknesses t k ,
- fibre angles a k ,
-

volum e percentage of th e fibres (pF

and all m a te ria l d a ta for m a trix an d fibre m aterials.


F u rth e r to th e orthotropic plate ( Ch. 9.1) we sh all now consider an aniso
tropic, plan e stru ctu re m ad e u p as a lam in ate consisting of several layers
( Fig. 10.5). Here, stra in s yap and distortions lryap are tre ate d together as
discussed in 10.1.
M aterial law

stress r e s u lta n t-s tra in relations

For a plane stru c tu re m ade u p of several layers, we assum e a lin ear stre ssstra in behaviour as was done for th e previous isotropic plane structures. In
case of a com posite structure, however, th e stress curve exhibits c e rtain dis
continuities at th e boundaries betw een th e single la y e rs; here, th e stress rem id-plane

F ig. 10.5: P late m ade of several


layers

120

10 Coupled D isk-Plate Problems

su ltan ts in th e single layers rem ain constant. For th e lam in ate itself they
depend on th e thickness coordinate
because th e com ponents of th e tensor
of elasticity differ from one layer to another. T h e stress resu ltan ts of the
lam in ate follow from th e equilibrium conditions by m eans of sum m ation of
th e stress resu ltan ts of all single layers [ B .ll ]:
N0"3 = 2 kNa/3 - A^ v

+ B a^

M'-" = > > < " =

+ K'

\ v - N gf

- M g ,

(10.19)

,a3/33

w ith

> a.@\xv

k
yCcPfiv _ ^

aPiii/
= 2kB '

ga3/33 _ ^

ga3/33

( 10.20)

k
a /3

M'

For a physical in terp retatio n of th e relations betw een stress resultants and
strains we w rite (10.19) w ith (10.20) in an appropriate symbolic n o ta tio n :
A
B
0

' N '
M
Q

B
K
0

0
0
S

N,e
7

( 10.21)

M,

w ith
A 13

A 12

13

^22

-^23

m a trix of m em brane stiffnesses


sym.
K-1
1

33

K,
12

K.13

^22

^23
K.33

12

B 13

22

23
B 33

<A = Ai i) .

m atrix of bending stiffnesses

K
sym.
11
sym.

m a trix of couple stiffnesses

10.2

Structures made o f com posite m aterials

121

m a trix of shear stiffnesses

^11 ^12
sym. S.22

(S.. = S..),
v u
ji '

[ ^1] >^22 ^12 ]

vector of in - p la n e disk (m e m b ra n e ) forces,

m t

= [ m u , m 22, m 12]

vector of plate m om ents,


vector of sh ear forces,

Q t = [ Q i .Q2]

0T

1 T

vector of strain s,

I ell >e22>e12J
r

vector of distortions,

7 ~~[ KU>K22K12J

vector of sh ear deform ations.

7 T = [ 7 1, 7 2]

For calculation of th e m atrix com ponents we need th e com ponents of th e


elasticity m a trix E for a single layer, which we obtain from a tran sfo rm a
tion according to (5.21) presented in m atrix notation
(10.22a)

JE
w ith
COS CL,

sin a.

sin 2 a.

. 2
sin a,.

cos a.

- sin 2 c l

sin2ak

- sin 2 a,.

Er
1

U1 '2 'U2'1'
V Y 2' ^2*

1 - Vy 2 , V2, v

cos 2 cl

2 ' 1

(10.22b)

Ex,

ul ' 2 ' V2 'l'


E,

1 vi'2'vr v

E lasticity m a trix of
a UD - layer (5.20) .

0
T he m aterial param eters can be determ ined by m eans of th e relations by
TSAI and HAHN [ B.10, B .ll ].
As (10.21) shows, disk and plate actions occur coupled in a plane stru ctu re
m ade of com posite m aterial. In addition, as a resu lt of th e transform ation,
th e single com ponents of th e stiffness m atrix depend on th e fibre angle c*k.
T he com ponents of th e subm atrices in (10.21) th erefore are based on th e
lam in ate design an d th e fibre orientation.

122

10 Coupled D isk-Plate Problems

S tra in -d isp la c e m e n t equations according to (10.5c)


1a[3

'l'a/3 =

(10.2S)

^ ^a P

w ith th e strain s

\ p

+ ^L )

an d th e disto rtio n s

\ af3 = \ ( i p a \p

+ ipp \a )

(10.24 )

Shear deform ation


o

1
(10.25)

w ith

an d

7 a = ipa + w | q

va

in -p la n e displacem ents of th e m id -s u rfa c e ,

ip

bending angles ,

displacem ent p erpendicular to th e m id -su rfa c e .

E quilibrium conditions for th e u n d efo rm ed state


From (S.28a)

N0^

From (9.31)

Q| +

= 0

(10.26)

=o
(10.27)

In (10.21), (10.23) an d (10.25) th e strain s are expressed by m eans of defor


m ations. Substituting these relations into (10.26) and (10.27) th e n leads to a
system of five coupled differential equations for th e unknow n deform ations
v a , w an d for th e angles ipa of ro tations of th e cross section f B.7, B.10,
B.11 ].

B.2 E x er cises
E x e r c i s e B -8 -1 :
A sim p ly supported disk (le n g th I, height h, thickn ess t ) is subjected to a
constant load q per unit length as show n Fig. B -l.
a)

Set up an approxim ate A IR Y s stress function in form o f a power series,

b ) D eterm ine th e stresses in th e disk from the stress function.


c ) Check that th e E q u i l i b r i u m a t Large is fulfilled .
d ) C alculate and com pare th e displacem ents o f th e d isk to th o se o f a
BERNOULLI -b e a m ( h I).

F ig. B -l: Sim ply supported rectan


gular disk under constant
load q

S o lu tio n :
a ) Based on ( 8.11) we consider the following power series assum ption for the
AIRY stress function:
fl> = a. x y

- constant shear

+ a20X

- constant com pression in y -d ire c tio n

+ W

- pure bending m om ent Mz


2

+ a2i x y

- a yy linear in y and xxy linear in x

+ a23x2 y3

~ X X ( x 2 > y ) - y y ( y 3 ) ' Xx y ( X , y 2 )

+ a05y

x x ( y 3

The single aik are free coefficients.

(1 )

124

b)

Disks

The stresses are calculated by means of (8.2) and ( 1 ):

0 XX <D*yy

a-osy + 6 a 23x 2y + 2 0 aosy3

( 2a)

*20 2a21y + 2a23y

( 2b )

- a n - 2 a21x - 6 a 23x y 2

( 2c )

yy

, XX

xxy = -
xy

The b o u n d a ry c o n d itio n s for the disk are formulated as follow s:

yy

^ 3,20

= ~T :

yy( x - ^ ) = 0 :

,1

r
xy

( 3a )

= 0

( 3b)

6a03y+ 4 a23*2y
= ,
23
J +
T Z2ao5y3
U d 05

(3c)

( x , - -jj-) = 0 :
2
h

= :

= ~ jf

oa0 3 y ^ Y

= 0:

V ( X 2 )

"4 a23 F

2 a 20 a2 i h - i a23h3

xx( 2 y )

a2i b

I
3
2
> - a n - 2 a 21x - - a 23h x

= 0 .

(3d)

Substitution of ( 1 ) into the bipotential equation ( 8.1) yields


A A 0 = 0 /xxxx + 2 0

/xxyy

>

Boundary condition

a 23 =

au

( 4 )/

_ 5 a 05

a 20 =

From ( 3 a ) + ( 3 b )
From (3 d )

+1 0 #yyyy = 2 4 a2,,3 Jy +1 120a 05


yJ = 0

= 0

.
3
2
and a 21 a 23 h .

( 3d )m ust be fulfilled for all x.

From ( 3a ) - ( 3b ) + ( 3d ) a 23 = - 9 - .
h t
From ( 3 a )

a 2i = _ | j L .

From( 4 )

Since ( 3 c ) cannot be fulfilled for all y, it is dem anded approxim ately th a t the re
su ltan t force F vanishes:

x h/2
Fx =

J 0xx( 4 , y ) t d y = .

-h/2
This condition is fulfilled for all arbitrary a 03, a 23 and a05.

(5)

Exercise B-8-1

125

In addition it is therefore demanded that the moment Mz ( ox x ) of the stresses at


the boundaries vanishes:
h/2
M.

(6)

= JD( 4 ,y)tydy=
- h/2

Y2
2
3
2 2
( 6 a 03y2 + - a23I2 y2 + 20a ogy4 ) t d y =

h/2
= t ( 2 a 03y3 + Y a 23/ 2 y3 + 4 a osy5 )
-h /2

*1 1
V3 , 1
72 V3 I
1 1 2 a 03
" * " 8 a 23 * "
+

03

h3t V 10

V5

4 a os a

= 0

4 J '

Hence, in dimensionless notation the stresses become


S- xx
ti / t

h2 1012

_ i , 2 L U - 4f y f
4
I2 ' h
' h '

f y ) a- l L - l .
Vh )

L*y

2 h

2 h

(7b)

\2

q/ t

f _
h V h / 2 h Z

(7a)

_ f i l i a l
hZVh/'

(7c)

Fig. B-2 shows the approximate stress distribution for some cuts of a disk with
1/ h = 1 and \ x/Z, tj = y / h .

Fig. B-2: Stresses in the disk


with Z/ h = 1

126
c)

8 Disks
Checking the E q u ilib r iu m at Large

T he resultants of the shear stresses at the supports m ust be in equilibrium with


the load q .
h/2

h/2

f Xx y ( | y ) dy - 1

- h /2

h/2

f xx y ( _ 4 y ) dy = 2 t
- h /2

f Xx y ( 4 y ) dy = qZ
- h /2

h/2

h/2

2t Ji^[lh2z3Zy2]dy=iH !h2zy_2Zy3)
- h /2

-h /2

d ) The strains can be calculated from the strain - displacement relations (8.3) and
from the m aterial law ( 8.4) for the state of plane stre s s:
xx = u ,x = 4 - v(o xx - v oyy') ,*

(v8a )J

( 8b )

yy

=v

,y

=4-(o

E V yy

f Xy = U, y+ VfX =

- v o ),
xx 7

( 8C )

^ t xy .

Substitution of csx x , csyy , t xy from ( 7a,b,c ) into ( 8a,b,c ) leads t o :


q

I2

/ h2

E^h [5 H i o

u,x

u = i k { ! 6( T j 4

2)

h3 ) ]

+ x J y " 4y v ( 2y - 2 h y - y J l

'

+ T ) ) , - 4 >3

(9a)

- v ( 2 y 3 - - | h 2y - y ) ] x + f ( y ) } ,

Eh3 1

,y

{ 2y3 f h2y - y - v [ 6 ( ^ - - ^ + x2 ) y - 4 y 3]}


f 1 4

3,2 2

h
(9b)

_ v [ 3( i k ~

+ *2) y2 - y4] + e ( x )}

We now form the derivative of u with respect to y, and of v with respect to x :

ik lK w 4

+) -

< *

-14]*+ ^ 1

Exercise B-8-1

127

From ( 8c ) follows :

~ T + t ) - 1 2 y 2x - v ( 6 y 2 - | h 2 ) x + ^ )
.

_ 6 v x y

dg(x)l

2(l+v)

q [ 3,2

6 x y

( 10 )

Collection of term s reduces the equation to the following:


3 12 , 2
2x - h x
5

32
3 ,2
dg(x) df(y)
- - v h x - I x + v + ^
2
2
d x d y

= 0 .

, x
( 11 )
v /

Relation ( 11) can only be fulfilled if the term depending on y equals a co n stan t:
^

= Cl

f ( y ) = C l y + C2 .

T he sam e is valid for


dg(x)
3 , 12,2
, 3 .2
3,2
_
dx ~ 2x + - h x + - v h x + - Z x - C ,
=>

g(x)

x4 + - | h 2 x 2 + - | v h 2 x 2 + - | / 2 x 2 - Cj X + C3 .

If we substitute functions f ( y ) and g ( x ) into ( 9a,b ), we obtain the following


displacem ents:
,2

Eh3t U

h
10

I
x
-4r T
+ 3 ) y - 4 y' 3 -

(l2a)

/
3
3 .2
- v ( 2y - j l i y - i J l x
q J 1 4
3
V = Eh3 t l 2 y _ 4

2 2

- y x4 +

h3
[ / h2
- 2 y - Vl 3 U h 2 x2 +

+ Cjy+C,),

I2
2\ 2
41
T +X Jy - y j ( 12b )

v h2 x 2 + | l 2 x 2 - C 1x + C3 } .

T he following boundary conditions and conditions of sym m etry are employed for
determ ining the constants C ( i = 1 , 2 , 3 ) :
u ( 0 ,y )

= 0 ,

/ I
h ,
v( 2 - - 2)

= 0

t 1

v ( - 2 - 2 ^ = 0

From ( 13b,c )

>

Cj = 0 .

(1 3 a )
(1 3 b )
(1 3 c)

128

8 Disks

Thus, the displacements become


\ ( h2
10

r
1

12

z 2

-4

L
3 . 2 2
h3
- 4h y T y " v

4 , / 6 , 3

~ 2 X + U

/ 3
3 2
h3 'I 1
2 \
3
+ X J y - 4y - v ( 2 y - y h y - T J xJ

\,2

2 ,3

+ 4V h X +

2/
y

h2

l l o

,2 2

z 2
-

,
+

, 4/ 3

2\
41
x ) - y 1
1

X _ h ^ 32 + 80 v ) ~
5 ,4

~32l

2 .2 / 3

" h 1 U

{l 4b )
\\

+ I6 V i '

For com parison w ith standard beam theory we transform ( 7a ) as follows :


0xx = ^ [ 6 ( * 2 - V ) y + j p ( 3 h 2 - 20 y2 ) y ] >
where the underlined term corresponds to the bending stress oxx in the corre
sponding B E R N O U LL I-beam . The second term is sm all for h /Z <K 1 .
T he displacement function v ( x , y ) is structured accordingly :
(
\
q / 1 4 , 3 2 .2
5 ,4
\
V ( X y ) = i l ? T ^ 2 X + 4 X 1 " 3 2 * + ......... J '
The above term s are identical w ith the deflection function of the corresponding
B E R N O U L L I-beam ; for h / I 1 , all further term s can be neglected.

E x e r c i s e B -8 -2 :
A circular, an n u lar disk as shown in Fig. B-3 is assum ed to be supported
frictionless at th e outer boundary an d is subjected to a constant, axisym m etric tem p eratu re distribution over th e thickness t

D eterm ine th e stresses and th e ra d ia l displacem ent function of th e disk by


m eans of th e displacem ent potential 'I'.

F ig . B - 3 : Circular, an n u lar disk


subjected to a statio n ary
tem p eratu re field

Exercise B-8-2

129

S o lu tio n :
We proceed from POISSON's differential equation (7.11):

A ' F = ( l + v ) a T0(r).
Due to the axisym m etric shape, boundary conditions and tem perature distribution
of the disk, the derivatives w ith respect to <p vanish. After m ultiplication by r 2 the
p artial differential equation is transform ed into an ordinary differential equation
of second order of EU LER type w ith variable coefficients,

+ v K ^ .

(1)

T he homogeneous solution follows with the assum ption 'Fh = r* as


( 2a )

Yh = Ai lni + V

After choosing an assum ption for the right - hand - side type, the p articular solusolu
tion reads
ur _ 1 + v o l!.
P36
T a a4

(2b)

We then have the general solution

f = i , l n f + A! + i ^ T 0 , ^ .

(3)

By means of the m aterial law and the strain - displacement relations for the plane
case, the stresses are calculated as
E
1 d
1 + v r dr
E d2T
1 + v dr2

(4a)

1 + V

7 + | < 1 + ) T e . ( T ) T

(4b)

T he shear stresses Tr^ vanish identically because of the axisym m etry.


T he boundary conditions read as follows:
r r ( a ) = 0

rr(b ) = 0 -

F ig . B - 4 : Superposition of tw o load cases

( 5a,b )

130

8 Disks

As ( 4a,b ) shows, only one constant value of Aj will occur in (5 ). Thus, one
boundary condition can be fulfilled a t th e outer or a t the inner boundary only. We
obtain an additional condition if we - by analogy to statics - superpose a 1
system onto the 0 - system (F ig . B-4). If we apply boundary condition ( 5 a ) , an
equal tensile stress p occurs a t the outer boundary ( 0 -sy ste m ). We now subject
the outer boundary of the disk to a com pression - p ( 1 - sy stem ) which has to
act opposite to + p in the 0 - system so th a t this boundary is not subjected to a
stress any longer.
According to (3 a )

According to A-7-1

J 1) _ o r
rr

l+ v

i C2

2 C 1 +

( 6a )

~T

Note: The third constant C3 equals zero


for <> since the logarithm ic p a rt viol
ates the required periodicity of v in the
circumferential direction.

+ (l+ v )

At the boundary r = a

At the boundary r = a

arr ^ ( a ) = const = p .

( 6b )

1) ( a ) = - P

m ust be valid.
In order to fulfill the boundary conditions ( 5 ) we superpose the radial stresses
and obtain

J ) +, 0J1)
_ -_
with

l+ v

A* = Aj - 1 ^ V C2

0r* +
+ (1+v)

C* = 2 C

(7)

l+ v
E

Inserting into the boundary conditions ( 5 ) we obtain

rr(b ) =

r r 0 ) ( b ) +

(8a)

(a) = p - p = 0 ,

rr ( a ) = !r } ( a ) +

^ ' ( h )

( 8b )

= 0 '

Eqs. ( 8a,b ) provide the equations for determining the constants A* und C*
i t - C* =

A* =

_ c* = - (1 + v )

with

T0a( 4 )

4 4

c* = (l+v)

t M

1 + P2 ) b2

+ p 2 + p4 )

p =: b / a .

The following expressions are obtained for the radial and the tangential stresses:

l E T e a {i + (p2 + p 4 ) [ i - ( 4 ) - ( 4 } ,

t 0 ^

1 + ( p 2 + P4 ) [ i + ( 7 )

] - 5 ( 4

(9a)

(9b)

Exercise B-8-3

131

Fig. B-5 presents the non-dim ensional stresses in dependence on the radius ratio
p . Without the need for further integration, the expansion u ( r ) can be determi
ned from (9a,b ), the second strain - displacement equation (8.8) for e , and the
material law (8.9):
u ( r )

v o rr) + r T 0

The reader is asked to check the results ( 9a,b ) by means of the solutions to the
bipotential equation (8.6) . For further examples of thermal stress problems refer
to [ A.13 , B.8J.

E x e r c i s e B -8 -3 :
A circular disk w ith constant thickness t rotates
w ith a constant an g u lar velocity w ( Fig. B - 6 ).
D eterm ine th e location and th e m agnitude of th e
m axim u m stresses for a full disk ( 0 4 r = a )
and for an a n n u la r disk ( b ^ r ^ a ) .

F ig. B - 6 : R otating circular disk

132

8 Disks

S o lu tio n :
During rotation a centrifugal force at a distance r
fr = p r o 2

( la )

occurs as a D 'A LEM BERT inertia force a t the element.


dV

Since

t, = ----
r

dr

this force can be derived from a potential, i.e.


V = - y p u 2r2 .

( lb )

Thus, from the bipotential equation ( 8.6)


A A <D = - ( 1 - v ) A V
A V = V rr + y V r = - 2 p ij2

with

( 2a )

the following is obtained from ( 8.20) in view of the axisym m etry


r 4 <D(m r + 2 r3 <D>rrr - r 2 <Djrr + r r =

2 ( 1 - v ) p u 2 r4

( 2b )

The homogeneous solution is given by ( 8.21) . The particular integral is obtained


by an approxim ation of the right - hand - side type
<Dp = C r 4

------>

<Dp = ^ 32^ P u2r 4

(3)

If we - due to the uniqueness of the displacement v - assum e the constant C3 to


be equal to zero (according to [ E T l ,2 ]), and if we consider (8.7), the to ta l solu
tion for the stresses reads :

1*
rr

= T ,r +

^ , , C2 , 1 - V

V =

2 C 1 +

2C + C,

= 2 Cl + p

P U

+ v
g

= .ix+

8 P<J

V = 2Cl ~ ~ ? +

1 + 3v

C2

= 2 Ci

7 2

-------^

22
P

<4,)

r ,
22

3(1 v)
w

22

2 + 5

1
~

2 2

2P

22
r

(4b)

pu r

a ) In case of a so lid d isk the constant C2 m ust equal zero because of th e n o n existing singularity a t the origin.
From

orr ( a ) = 2 Cj - 3 ^ Vp <j2 a2 = 0

follows

2 C. = - ^ V p ij2 a 2

and thus

crr = b -V p <j2 ( a2 - r 2 ) ,

<PH> =

3 + v
2 21 + 3 v
g P a
g

( 5a )
22
P <J r .

/ _, ,
( 5b )

Exercise B-8-4

133

The maximum values are attained at the axis r = 0 and become


3 T V 2 2
rrm ax

W m ax

/ r- \

p u a .

( 5c )

b ) For the case of an a n n u la r disk, the constants Cj und C2 follow from the
boundary conditions
rr(b ) = 0

>

rr(a ) =

0 -

( 6 )

These conditions yield a system of two linear equations for Cj and C2 Having
determined these, we obtain the stresses
3 + v
p w 2( a 2 + b2 8

- r2 ) ,

3 + v
2( 2 , ^2 , a2 b2
- p U (a + b
8

vv

1 + 3v 2^
r j.

( 7a )
{>
7u \
(7b)

The maximum circumferential stress is found at the inner boundary r = b


2/ 2 , 1 V 2\
3 + V
4 _ p U lv a + 3 ++ V^ b )

a.

Wm*x

, S

( 7c )

The maximum radial stress results from


2 ,2

r r ,i -

as

rrm ax

- p u2( 2
8
-

---- 5 "

rr

r-

- 2r ) = 0

> r* = / T b

/* \
3 + v
2/
,
r* ) = p u a - b ) .
>

For b = 0 the solution is equal to that obtained in a ).

E x e r c i s e B -8 -4 :
A q u a rte r - circle an n u la r disk (o u te r radius
a, in n e r radius b, thickness t ) according to
Fig. B-7 is clam ped a t A-A. A force F acts
in th e ra d ia l direction a t th e free end B-B
of th e disk.
D eterm ine th e stresses w ith in th e disk by
assum ing th a t th e norm al stresses in the
tangential direction vary proportionally
w ith s i n (p (a s is th e case in th e theory of a
curved b eam ). Discuss th e results.
F ig. B -7 : C lam ped q u a rte r-c irc le
a n n u la r disk subjected
to a single load

134

8 Disks

S o lu tio n :
T h e given q u a rte r-c irc le a n n u la r disk is subjected to a n o n - ax isy rn m etric sta te of
stress .
T h e b ip o te n tia l equation ( 8 . 6 ) in p o la r coordinates r , 9 ( 0 = 0 , V = 0 ) rea d s :

A A <P = , rrrr + f ,rrr - 4 ( . - 2 9

) +
(1 )

4 i_3 (' . r - 2

,tt p tp

) 4 44 (tv 4 itp
4 Itfilfiip
) = 0
> '1 ^4
Iip ip
w ''
iipipifiip
ip >

Since the norm al stresses are varying proportionally with


sum ption is chosen :

sin

= f ( r ) sin 9 .

9 , the following a s
(2)

Substitution into (1 ) yields

A A = ( f 4 f f rrr -

+ ^ f r - ^ f ) sin

= 0 ,

i.e. the differential expression in the paranthesis has to v a n ish :


r 4 f ,rrrr

+ 2 r 3 f ,r r r - 3 r" f + 3 r f r - 3 f = 0 .

(3 )

Eq. ( 3 ) is a differential equation of the EULER type.


The solution assum ption f ( r ) = r" leads to the characteristic eq u atio n :
n4 - 4 n 3 + 2 n 2 + 4 n - 3 = 0 .
The solutions are:

nj

^3

* ^4 I *

Thus, we obtain the general solution of the differential equ atio n :


f ( r ) = Cj r 4- C2 4 Cj r -f- C4 r In r

( 4a )

or, according to ( 2 ) , the AIRY stress function


= ( Cj r3 4- C2 y + C3 r 4 C4 r In r ) sin 9 .

( 4b )

From ( 4 b ) one obtains the stresses in polar coordinates (8.7) as :

2C2
(V
+ r 1) sin 9
r3
o

,rr

c4\ .

2C2

( 6 C j r + i ++ -T.
) s 1isin
n 99 ,
r3

2C2
XTV>

r3

C4 )
Icos cp
t J

(5a)

( 5b)
(5c)

T h e b o u n d a r y c o n d i t i o n s a re :
rr ( a *V )

>

r r ( b >? ) =

( a > ) = 0

xrV> ( b >f ) = 0 -

( 6a )

( 6b)

Exercise B-8-4

, (

r > 0

t d r

135

b
We can now determine the constants C j , C2 and C4 by m eans of ( 6a,b ) and ( 7).
Substitution of ( 6a ) into ( 5a ) yields
C2
C4
2 C,l a - 2 -4- +
a = 0 ,
c

(8a)

(8b)

2 G, b - 2 i + -7- = 0 .
1
b3
b
From ( 8 a ,b ) we determine
l _

C,
2T2
a b

(9)

Condition ( 7 ) w ith ( 5c ) becom es:

Jk v ( r ) tdr = vf _ ( 2Ci r ~

^ r
r

^ i )cos_otdr =

= - t ^ Cj r + 2"
, 2
Cj ( a

2
a2 - b2
b ) -j- C2tl~
2 a 2 ib 2

= F

r')

, a _ F
C* In.
4 b

( 10)

Substitution of ( 9 ) into (10 ) leads t o :


2 ,2

2 ,2
^C
9
"2

a o o 4T- C
^9
2 ,2
"2 a 92 _,29
a b
a b

C4 =

- ^C4 In
t)
_
t

a2 - b2

2C

2 a2 b2

( 11 )

We then proceed w ith ( 9 ) and (11) substituted into ( 8 ) :


2 C,2 2L ,1 22a b

2 C 2 (a2 - b2)

2,2

+
1

a b

ln b

2 , 2 r,

a b

C2 =

and, further.

C4 =

(12a)

2 - b2 - ( a2 + b2 ) l n ^
(a 2 + b 2 ) F
|a 2 - b 2 - (a 2 + b2 ) l n ^ j t

(12b)

136

8 Disks

By substituting ( 12,a,b,c) into ( 5a,b,c) one obtains the stresses :

ta

(13a)

sm cp ,

^2

a2

F
ta

3 - - ^
a
a2
a2

sm cp ,

(1 3 b )

cos cp

(13c)

( 1 + a 2) b l b

ta
i - 1 - ( 1 + T 2 ) lnF

For presenting the results, the stresses are normalized as F / ( b - a ) t . This deli
vers the following stress distributions where the abbreviation p = b / a is used :
o (a-b)t
=

a +P2( f )
=

P= - 1 + ( 1 + P = , , P

10

0.25

0.5

0.7 5

Fig. B-8 : Circumferential and radial stresses in the quarter - circle disk

Exercise B-8-5

137

V ( - b)'

>j T - i > ( f ) - ( , + '>, ) 7 .


vv=
F
= (1 - P )
p ! _ 1 + ( l + p2)ln(>

V ( a_b)1
= ------- p------=
T he maximum stresses a
rrm
m axim um a t <p = 0 .

ax

,,

or c,

W m ttx

+p2( ? ) - ( 1 + p ) f

(? . 1+(1+(?)ln(>

are obtained at <p = , and


2

TV

has its

Fig. B-8 presents the distribution of the norm al stresses over the cross section for
different ratios of the dimensionless radii p . If p increases,
approaches a
linear distribution corresponding to th a t of a straight beam.

E x e r c i s e B -8 -5 :
A sem i - in fin ite disk y > 0 , - o o < x < + o o ( thickness t ) is subjected to a
concentrated m om ent M at th e origin 0 as shown in Fig. B -9 a .
a ) D eterm ine th e stress function for th is load case by using a force couple
as shown in Fig. B-9b an d by applying th e stress function (ra d iatin g
stress sta te ) $ = ( F /x t ) r ip cos (p for a concentrated force F as
form ulated in [E T 2,A .16].
b ) W hich stresses occur in th e sem i - in fin ite disk?

__

y
F ig . B - 9 : Semi - in fin ite disk u n d er a concentrated m om ent M
Solution:
a ) In order to determine the stress function , we substitute the prescribed mo
m ent by a force couple as shown in Fig. B-9b. Then the stress function for F in
Oj can be w ritten as :
-<D(x - s , y ) .

138

8 Disks

If we superpose the above stress function by the corresponding stress function for
F in 0, i.e. ( x , y ) , we obtain
, = - ( x - s , y ) + ( x , y ) .

(l)

According to the rules of the differential calculus the partial derivative is defined
lim

-* 0

- ( x - E , y ) + ( x , y ) _ d ^
E
d' x

(2)

The difference in ( 1 ) can, by applying ( 2 ) , be w ritten in the form

(3a)

1 = S , X
After transform ation into polar coordinates (c)/c)r ( ) r , d/ d
difference reads :
= e ( r cosq> - 9 V ^ - )

( ) ^ ) the

(3b)

In [ E T 2 ] , the stress function of a single load on a s e m i-p la n e is derived as:


= ^

r <p cos tp .

(4)

The p artial derivatives of ( 4 )

= J f * cos * F
\
v = ^ r ( cos v ~ v sm 9 '
yield by substitution into ( 3b ) the following stress function :
_

j =

With F e

Fe r

. 2

[ <p cos cp - sm <p cos (p + <p sm cp J =

sm

cp

Fe r

[ <P sin <P cos 9 J

cos

cp

(5 )

F ig. B -1 0 : R adial stresses in a semi - infinite disk under a concentrated m om ent

Exercise B-8-6
b)

139

The stresses are determined according to (8.7)

rr = T l , r + ^ l , w

(6)

Since fl>,1 ,r = 0,1 it follows from (v 6 /) t h a t :

arr = i 2

1 ,<p<p

By
( cos2<p sin2 (P ) ] = ^ [ 1 - cos 2 ^ ] - i,w = ^ 2 s in 2 <p ,

1 ,p

2M

an = ^ s i n 2 cp

we obtain

0 < <p < y

i.e.,

or r > 0
tension

,
,

^
j

= zrifi = 0 ,
< <p <

t3r r < 0
compression

Fig. B-10 shows the distribution of the radial stresses in this semi - infinite disk.

E x e r c is e B -8 -6 :
A circular, a n n u la r disk m ade of carbon fibre reinforced plastic ( C FR P ) is
subjected to an outer tractio n po, an inner pressure p j , th e centrifugal forces
fr = pru>2, an d a tem p eratu re field 0 ( r ) . T he lam in ate is an a n g le -p ly
com posite consisting of two layers w ith a fibre angle a and th e to tal
thickness t ( t 1 = t 2 = t / 2 ) according to Fig. B -ll.
T he following is to be d e te rm in e d :
a ) the EU LER differential equation for th e corresponding disk problem ;
b ) th e distribution of th e ra d ia l a n d tan g en tial stresses in dependence of
th e fibre angle a a n d th e rad ial expansion of th e disk ;
c ) T he resu lts obtained shall be discussed on th e basis of th e following nu
m erical v a lu e s :
G eom etry:
O uter radiu s a = 0.25 m , in n er radius b = 0.0625 m , thickness t = 2 10 m
M a teria l c h a ra c te ristic s o f the la yer ( fibre T300, m a trix 914C)
( S tandards see Section 5.4 ) :
E j, = 132 680 M Pa ,
vw

>

E2, = 9059 M Pa ,
=

6 kg / mm,3 ,

G r 2 , = 4268 M Pa ,
<
p = 0.60 ,

a T l, = 0.23 10~ 6 / C , a T2, = 53.25 10~ 6 / C .


Loads :
p. = 0 MPa , po = 1 M Pa , revolutions n = 1000 m in 1 .

140

8 Disks

F ig. B - l l : C ircular an n u lar C FR P - disk u n d er several loads


T em p era tu re fie ld :
<9(r) = 0 o(-g -)

w ith

= 1 C .

Solution :
a ) According to (3.29b), the equilibrium conditions for a circular disk can be ex
pressed as follows in polar coordinates:
r r + r t 5n - , r +

r V , r + tW

+ 2V

(la)

+ F fr =

+ rf*

= 0.

(lb)

All loads are axisymmetric so that both the stresses and the displacements are
independent of the circumferential coordinate <p . Thus, all derivatives with respect
to tp vanish in ( la,b ). Furthermore,
y IV>, and v are equal zero. The simpli
fied equilibrium condition ( l a ) then reads :
(2 )

rr+ r0n , r ev v + r fr = 0

In a similar manner, the strain - displacement equations (8.8) for the axisymmetrical load case can be written as :
zwp

( 3a,b )

r '

Finally, we have to determine the material law for an a n g le-p ly lam inate. This
is achieved by rotating a unidirectional layer ( UD - layer ) ( see Section 5.4 ) . For
pure disk action, we have according to ( 10.21) :
Nr r '

^11
=

-^12
A22

sym.

^13
A 23
^3 3

a Tr

' Er r '
E ifiip

- f r(C> -

Tv-

TrV -

( 4a )

Exercise B-8-6

141

The elasticity matrix in ( 4a ) is fully occupied and its components depend on the
angle of rotation. By special choice of the fibre orientations in a m ulti-layered
composite ( see Fig. B -ll) and constant uniform layer thicknesses symmetrical
with respect to the m id-plane, one obtains a simplified material law for the state
of plane stress. We can prove for a k = a that A13 = A23 = 0 , and thus
( 4 a ) becomes with the assumptions made,
0

rr

__

E 11 ( a )

Ei2 ( a )

E 2l ( a )

rr

( O

Tr

(4b)

*TV

E v>v>

In ( 4b ) we have introduced the stresses corresponding to the stress resultants by


means of Nrr = er r t etc. This way, the elements of the elasticity matrix corre
spond to those in ( 5.24a,b) . Furthermore, aTr and a T(f> denote the coefficients of
thermal expansion in r - and <p- direction.
At this point it should be mentioned that with the chosen fibre position, the ele
ments of the elasticity matrix are independent of the angular coordinate <p ; the
assumptions concerning strains and displacements are thus confirmed.
Equations ( 2 ), ( 3a,b ) and ( 4b ) provide five equations for the determination of
the five unknowns er r , ov v , sr r , sv v and u . In the following, this system shall
be solved with respect to the stresses ( see Section 7.4). In a first step we there
fore transform ( 4 b ) with respect to the strains srr und
and obtain :
Er r

D11 (

rr

Tr
+

D 2l ( a )

(p<p

D12 (

D 22( a )

ipffi

with components of the compliance matrix denoted by


are given by:
D ( a ) = - 22^
111 ;
A ( )
with

E u (a)

D22 ( )

A()

( 4c )

tv

( y , 8 = 1 , 2 ) . These

Ei2 ()
E 12 ( a ) ~ E 21 A ( )

A ( a ) = Ei:1 ( a ) E22 ( a ) - E12 ( a ) .

From ( 3a,b ) with ( 4c ) we obtain


=

>

4- r

U,r = d ? ( r ^ ) = 'tptp

r %>>*

( 0 21

0 12 ) 0 ^ + ( Tv. - Tr ) 0 +

D n ) rr + ( 0 22

( 5a)
+ r ( Di2 rr,r
We substitute from ( 2 ) into ( 5a )

D22W ,f + ttT<p 0 , r ) = 0 '

142

8 Disks

and obtain the following differential equation :


^22 r

3 ^22 r rr,r +

rr ,rr +

( ^ 2 2 ~ 11 ) rr =

(5b)
2

= ( a Tr - a T(c, ) e ( r ) - a X(c, r 0 r + ( D12 - 3 D 22) p r Id


After substitution of the temperature field and its derivative
0 ( r ) = 0 o( ^ ) 4

(r),r = ~ j ~ (

)3

Eq. ( 5b ) takes the form


^22 r

r r , r r + 3 b )2 2 r 0 r r ,r + ( ^ 2 2 ~ ^11 ) rr =

(5c)
=

( a T r _ 5 a T ifi ) 0 ( "b" )

"*" ( 12 ~ 3 ^ 2 2 ) P

10

b ) Equation ( 5c) is an inhomogeneous EULER differential equation. We derive


the homogeneous solution by means of the assumption

(6)

rrh = C r " .
which yields the characteristic equation
X + 2X + (1 with the roots

Dr

Xj 1 + -|/ ^

X2 1

22

pD11
is the degree of anisotropy.
Thus, the homogeneous solution is
rrh = V * 1 + C2rA2

(7 )

with the free constants Cx and C2 .


We now determine the particular solution by means of an assumption for the
rig h t-h an d side. Thus, we obtain
( a Tr

b a Tv>) 0

r \4

(*>

(^12 3 ^22 ) p r2 o
( 9 D22 D jj )

The total solution consists of parts ( 7 ) and ( 8 ) :

(8 )

Exercise B-8-6

143

The constants are obtained from the boundary cond itions


rr ( b ) = Pi -

( 10a )

rr ( a ) = Po '

( 10b )

After substitution of ( 9 ) with ( 7 ) and ( 8 ) into ( 10a,b ) the constants can be de


termined from the linear system of two equations
_

2 2 2J )PP<W2 ,.2 ________________


(l 0
D1
122 - 3^ D
( a Tr 5 a T <p ) 0 ) >.2

IY

( 9 D22D22 ~ Dll
D1:1))
(

b - - ( 2( 255 1D22^ -D
d uj )

( 0 12 3 0 22 ) P w

( 9 D ,, - D )

VP ~

C2 = A { - [ P i - 4 ^

( a Tr - 3 i y ) 0 ^ )

(25 D - D )

3 D ^ )pt02 b2 - ( ^

( 9 0 22 Du )
, [

( 0 12 ~ 3 0 22 ) P 10

+ l P
w here

( 9 D22- D u )

>

5 gT ^ ) e o | aA1 +

(25D22
2

T? '

Du )

( a Tr ~ 3 a TV ) 0 a4 ]

a " ( 2 5 n B - D 11)

? J a 1

A =

bll>>A2[ ( F ) Al- ( t ) 1
Introducing the abbreviation p0 = a / b, the complete solution for the radial
stresses reads :

2
( 0 12 ~ 3 0 22 ) P 10 / r \ 2
( 9D2 2 - Di i )

(H )
( a Tr _ 3 a T ^ ) 0 ( 1
(25D22- D u )
V^

where Bj and B2 are given by

>2

Bi = ^ r 7 r l p i p0
Po Po

(,

A2 )f

-P o + U -P o

)[

+
D _

*1

, (,

2
(D 1 2 - 3 D 2 2 )p io

( 99 D
D 2_
,, - D ,, )
V 22
11

( a Tr - 5 Tb ) i
(25 D22 fDne )r *

U ) l]'

A i ^ f ( D1 2 - 3 D 2 2 ) P

2 _ PoAl- P o 2 ^ Pi _ P o + ^ 1 _ P ^

( 9 D 22- D u )( T r ~ 5 g Ty> ) 0

+ (25D3J-DU)

2 ,

" +
(

a, \ 4 1 1

U ^ J ] -

A fter d eterm in atio n of orr , th e rad ial stresses can b e calculated from ( 2 ) :

o tptfi =

r r + r 0 rr ,r + P r 10 '

The radial expansion can then be calculated from ( 3b ) and ( 4c ) :

( 12 )

144

8 Disks

u(r= a)
[mm]
0.030
0.025

Radial deflection under


external traction Utr
centrifugal force Ucf
temperature

0.020

F ig. B -12: Radial displacement of a circular hollow C F R P -d is k

u
a JV d2
2

30 0.470

II

45 1.000

IV

F ig. B -1 3 : Circumferential stresses in a CFRP - disk

V
oOo

III 60 2.129
3.829

Exercise B-8-7

145

Fig. B-12 shows that the radial deflection depends on the angle a ; the smallest
value occurs at a* = 60". The circumferential stresses in dependence of the de
gree of anisotropy -J Dn / D22 are presented in Fig. B-13. Here, it is obvious that
the circumferential stresses are approximately constant for -J Dn / D22 ~ 2,0 so
that the high stresses at the inner and the outer boundary respectively can be re
duced together.

E x e r c is e B -8 -7 :
A n in fin ite disk possesses an elliptical hole. T h e disk is subjected to a uni
ax ia l tensile stress cr0 , see Fig. B-14. No ex tern al loads ac t a t th e boundary
of th e ellipse.
C alculate th e stresses in th e disk by th e com plex solution m ethod. For this
purpose, use th e conform al m apping

z = f(() =
...
w ith

a +b
c = -
2

C(c +
,

a -b
m = p
a+ b

0 <m < 1

to tran sfo rm th e area outside th e elliptical hole in th e z - p la n e into an area


outside th e u n it circle ( | { | = 1) in th e com plex { - p la n e .

F ig. B -1 4 : In fin ite disk w ith an elliptical hole u n d er u n iax ial tension a 0
S o lu tio n :
C om m ents on conform al mapping
The equation z = f ( ) assigns any point z = x + iy = r e 1?9of the complex z domain to a point ^ = { + ii) = R e 1 0 of the complex
domain (see Fig. B-15).
For further details on conformal mapping refer to, e.g., [A.8, B.6 ].

146

8 Disks

Fig. B-15: Mapping from the complex


vice versa

- domain to the complex z-dom ain and

First, we will show that the given equation


* = f ( G) = c ( c + f )

( 1)

maps the area outside the u n it- 1 -circle in the - domain into the area outside
an ellipse in the z-dom ain. Thus, the circle with radius 1 is transformed into an
ellipse.
The equation for the circle with radius 1 in the
C = e, e

domain reads :

( R = 1) .

(2 )

Substitution of ( 2 ) into (1 ) leads to


, .
ct i \
/ i
-i .
z = x + iy = f ( e ) = c ( e
+ me
)
and, after some transformations by means of the MOIVRE formulas, we obtain
i
\e +e
(l + m )

i
i -i
/.
\ e e
+ i ( l - m ) -------------

= c[ ( 1 + m ) cos 0 + i ( l m ) sin 0 ]
or

(3 )

x = c ( l + m) c os 0 ,

(4a)

y = c(l - m )sin0.

(4b)
2

Eq.

( 3 )and (

4 ) are

the

For

0 = 0 we

obtain

the

parametric relations of the ellipse~ = 1.


a
b
large sem i-axis

x(0) = a = c(l + m).


The small sem i-axis results from 0 =
y (- x- ) = b = c ( l - m )

Exercise B-8-7

147

R ela tio n s fo r the sta te o f plane stress


The following approach holds for the stress function in complex notation and
fulfills the bipotential equation derived in [A.5, B.6 ]:
( z . z ) = 4 - [ z<P ( z ) + z < p ( z ) + J l ' ( z ) d z + J i I ( z ) d z ] .

(5)

Here, <p and ip are arbitrary_analytical functions to be determined by means of


boundary conditions. 9 (zf), <J) ( "z) are the corresponding complex conjugate func
tions with "z = x - i y .
Eq. (5 ) provides the following formulas for the stresses and deformations (8.19) :
~ 2(

yy = >yy
0

yy - a

xx

+1 2 i x x y = , x x - , yy

2 G( u + iv) = - z
where x =

(6a)

+ <P ) .
2

(<p + + ' )

' - i ] > + xip ,

(6b)
( 6c )

is valid for the state of plane stress.

1+v

Boundary conditions
a ) Boundary of the hole
The boundary conditions at the boundary of the hole result from the fact that the
load vector must equal the stress vector t at the boundary.
In accordance with Fig. B-16 we formulate the equilibrium conditions considering
the + /- sign of dx :
x-direction :
*

^x

l ^ d s t = x x ^ 1 +
dy
-rds -

dx
dx
dy
yx
ds = ....
'yy -rds + *y

y-direction :

tv d s t = Oyy( - d x ) t + Txyd
yt
xy

v =

dx
dy
dx
_
dy
.
a
i - t = - ------
yy ds
*y ds
-xx ds
-xy ds

Fig. B-16: Equilibrium at the boundary of the hole

148

8 Disks

Since the boundary load is zero it holds that :

t = tX + i t y = f ( C,. y - i X) =
=>

C> - i C> = const .


>y
<x

Since the choice of constants is not yet limited, we can set them zero :

<y

,x

.7 d
\ dz

d
oz

i( -

d
dz

d \
r = )
dz /

or, with ( 5 ) at the boundary of the hole

. d
dz

- 2 l ^ =

<p(z) + z < p ' - f 4 , = 0 -

(7)

Using the chain rule we transform the formulas for the stresses and displacements
into functions of :
z = f(C)

<p ( z ) = <p
<p(z)
p[ [ff(( C)]
c)]

=>

=<P1 ( C ) = <Pi.

4j ( z ) = * [ f ( 0 ]
dip
dz
d2 tp
dz2

d<Pj dC

<

t ffl
- d (

~ dO

dip

dz

<Pl ( 0

dz

' Vi'}

. f' J dz

f l3

~ {'

From ( 6a ) through ( 6c ) we then obtain the transform ed relation :


Cxx 41- 0 yy

= 2 '

( ! )

(te)

y y - x x + 2 i x xy =

2G (u + iv)

= -

- (Jjj + x

( 8c )

At the boundary, we have, according to ( 7 ):

Pi + J TVi +

+1

= 0

( 9a)

According to [B.6 ], ( 9 a ) can be replaced by its conjugate form. We now achieve a


simplified notation in the form
Pi + +1 = 0 >
where

f=c(c + ^ )

f' = c ( l -

( 9b )
,

f" = 2 c m ^ .

( 10)

D e t e r m i n a t i o n o f th e s t r e s s f u n c t i o n b y m e a n s o f L A U R E N T - s e r i e s

In the following, power series shall be established for the functions <Pj und i]q
[B.6 ]. Since the stresses in equations ( 8a ) and ( 8b ) m ust converge at infinity, no
term with positive exponents are allowed to occur in the series. F irst, the follow
ing L A U R E N T -series are set up for the derivatives :

Exercise B-8-7
oo

oo

n= 0

n= 0

149

( 11)

with An and Bn as complex constants.


Integration yields:
( 12a )
n= 2

o B
<K = ? 0 C + B1l n C + ^ - Z ^ T

(12b)

+ B-

n= 2

By differentiation we obtain from (1 1 ) :


oo

(12c)
n 0

D eterm ination o f the constants


a ) C onditions at in fin ity
By means of ( 8a ) and (11), we state for the stresses at infinity :

The imaginary part of A Q only represents a rigid body displacement of the disk
and is therefore set zero, i.e.

Correspondingly, it follows from ( 8b ) that

Hence, we have obtained the results for A0 and B0 .


b ) B oundary o f the hole R = 1
For boundary condition ( 9b ) we derive the remaining relations for the determina
tion of the constants :

- 2i V 1

2 i

n= 1

2 -4i

+ m e

3 6i

+ m e

150

8 Disks
.

cpj

A0

Aj

-i

-2i

A2 e

2 -,

+ A3 e

-3i

+ . -. ,
2 \

-i

-3i

-^r <P, = A0 m e l S + AQ( l + m ' ) e l u + AQ( l + n / J e 3 It' + . . . +


+ Aj i n + A , ( l + m 2 ) e 2l& + A , ( l + m 2 ) e il@ + . . . +
+ A2 m e l& + A2 ( l + m 2 ) e 3i@ + A2 ( l + m 2 ) e 5i S + . . . + . . .

and with ( 12a,b )

Pi =

-i

A0 e

- 1 A , 0 - A2 e

^ B o e '+ iB jO -B .e

A3

--x -e

2i

A4

-2i

3i

~ 5 e - - >
4

- e

- 3 i

Now, all of the above equations are substituted into ( 9b ) and rearranged with re
spect to the e - functions. After that we get
0

B, = 0

B2 = - ^ ( 1 - m + m2 )>

= 0

B;2 n

cop
= 2 ^ ( l + m2 ) ( m - l ) ( 2 n - l ) m 2
2

Bo2 n

+ l

for

for

We thus obtain

n > 3 ,
for

n>2

n > 0 .
(1 3 a)

l + (m -2)-U ,

P i

K =

+ ( l - m + m2 ) - ^ +
n 2

+ (1 + 4

n=2

D e te r m in a tio n o f the stresses

The circumferential stress can be calculated by means of equation ( 13a ) :

. I - I B .
rr + ^ Re ( ff> )
^ Re

xx + yy

_
-

0Re

C2 + m - 2
_

r2
/ Ty2

2i

- P r (R e
0
I 'D 2
(Re

0R e

E ----

K1- ^ )

R2 e2i + m - 2
t >2 2i
Ri
m

n W yy2

2i

+ m - 2)(R e
- m) _
- j 2 2 i
~
- m j \ (z R
e
- m)\

2i

(13b)

Exercise B-8-8
4

2 -2i

__i T > R + m R e

2 2i

- mR e

- m + 2m

t>2 2i .
R - m Rr >2e - 2i - m R
e
+ m2

= n

2 -2i

- 2R e

151

^4

R4 + 2 m - m 2 - 2 R2 cos 2 0

i ------------------------- 5-----

R - 2 m cos 2 0 + m

At the boundary of the hole R = 1, we have atT 0 and therefore the stress
becomes
1 + 2 m - m2 - 2 cos 2 0
1 - 2 m cos 2 0 + m 2
In the special case of a circular hole we have a = b and, correspondingly, m = 0 .
For 0 = n / 2 we obtain the maximum stress
m ax = 3 0 0 -

The stresses at arbitrary points r, 0 are calculated from ( 8b ) . These calculations


sh all not be perform ed here. F urther examples pertaining to th e increase of
stresses at holes and notches can be found, e.g., in [ A.15, B.6 ] .

E x e r c i s e B -8 -8 :
An in fin ite disk subjected to a u n iax ial tensile stress <x0 possesses a crack
w ith th e len g th 2a (Fig. B-17). T h e state of stress w ith in th e disk is de
scribed by th e com plex stress functions
a o z + <7o a

f )

- 1

fP =

y,
w ith z = x + i y .
Check w hether th e corresponding stresses ful
fill all bou n d ary conditions

k -2 a !

F ig. B -1 7 : In finite disk w ith a crack

152

8 Disks

S o lu tio n :
In order to apply equations (8.19) , we calculate the derivatives
,

0 , 0

=T

<P
*

= ----2a

(la)

- 1

(lb)

T/2

KtT
(1c)

3/2 '

m
With (la ,b ,c ), the stresses follow from (8.19)
ax x +1 ayy

= - o + 0

(2a)

M
_Z_

_ Z _

a
yy -

x x + 2 i T xy =

0 +

3/2

(2b)

i m - '
In order to calculate the stresses themselves, the rig h t-h an d sides must be separ
ated into their real and imaginary parts. For checking the boundary conditions the
following considerations of the limit values are sufficient.
1) For large z und , i.e. the outer boundaries, holds
o x x +1 ayy

(3a)

yy xx + 2 i t xy = 0 >

(3b)

and it follows that

Xx =

0 > yy = o

> Txy = 0

in accordance with the given load.


2)

Along the x -a x is , i.e. z = = x , the relations (8.19) become


xx +

yy -

- 0 +

(4a)

2 0

Ml ) - 1
yy -

xx +

2 i T xy =

(4b)

For reasons of symmetry, we have along the x -a x is


a)

= 0 -

The above is valid both for the non - loaded crack surfaces I -5-1 < 1 and along the
axis I IT I > L

Exercise B-9-1

153

Solution of ( 4a,b ) now yields


for

b)

< 1 ,

_x
a

M - i - i
The stress cxx along the crack remains constant. At the crack tips x = a, a sin
gularity occurs, and for large x 0XX approaches zero (lo a d -fre e outer boundary).
)
yy =

_x
a

for

i-

<i

>1.

The stress 0yy is equal to zero at the surface of the crack |


crack tips x = a , 0yy (a s well as 0 XX) exhibits a singularity.

| < 1, but at the

Thus, the stresses derivable from the functions given in the problem formulation
fulfill all boundary conditions.

E x e r c i s e B -9 -1 :
An isotropic, s h e a r - r ig id rectan g u lar plate (dim ensions a , b , thickness t ) is
sim ply supported a t x = 0 an d x = a , clam ped at y = 0 , and free a t y = b
(Fig. B-18). T h e plate is subjected to a transverse, tria n g u lar load
p ( y ) = P0 ( 1 _ f - )
D eterm ine th e com plete solution to th e plate equation, and form ulate the
boundary conditions.
Po

154

9 Plates

S olu tion :
The differential equation (9.13) ( *0 = 0 )
K M w = Po( l - [ )

(1)

has the homogeneous solution with the double root Xn =


w.h = Z-J'
Y ( A*n cosh a
+ C*
n sinh

n+ B*
a i ^ c o s h ^a
+ D*n ^ a

( see(9.42) )

, .
(2 )

sinh ^ a) s' i n a^ .

This solution fulfills both the homogeneous differential equation and the boundary
conditions at the boundaries x = const:
w ( 0 , y ) = Mxx ( 0 , y ) = w ( a , y ) = Mxx ( a , y ) = 0 .
In order to determine the particular solution, the load in z-direction is expressed
as a FOURIER series as discussed in the beginning of Section
9.2,
P ( x , y ) = p0 ( l - ^ ) = ( l - | - ) ^ p n s i n ^

(3a)

with
2 f
. nitx ,
4P 0
Pn = a J po s l n ^ - dx = i n r
0

1 c
* n = l,3,5...

(3b)

Using the following series expansion for the particular solution


w

<*>

it follows from the differential equation that

( ) * . ( - f ) - & ( - f )

' . - < -

<*>

The complete solution is then obtained from ( 2 ) and ( 4 ) :


4 Po

Wh + Wp = - ^ 2 [ ( l - f ) ( ^ ) 5 + An COSh^ + Bn ^

cosh

+ Cn sinh I ?

+ Dn ^

sinh ^

. n 7TX
sin -

(6 )

Owing to the sine-expansion, the complete solution fulfills the boundary condi
tions at x = const, too.
For the determination of the 4 x n unknowns An, . . . , D , we have two boundary
conditions for each of the boundaries y = const., and these apply for each value
of n = >
for

y = 0 acc. to (9.16)

for

y = b acc. to (9.14a,b) ----->

>
M

w( x, 0) = 0,

w (x, 0) = 0,

(7a)

( x, b) = 0,

Q ( x, b) = 0.

(7b)

Exercise B-9-2

155

The stress resultants are obtained from (9.11) according to ( 7b ) or, alternatively,
in analogy with (9.14c) with d/dx = ( ) x , d/dy ( ) y:
M y y = ~ K ( w, y y +

v w ,xx) >

Qy = - K ( w yyy + ( 2 - v ) w xxy) .

Exercise B -9 -2 :
A sim ply supported sem i - in fin ite
plate strip ( width a, thickness t )
as show n in Fig. B-19 is subjected
to a load in form of a uniform ly
distributed m om ent along th e
boundary y = 0. T h e m om ent per
u n it len g th of th e b o undary is
denoted by M 0.
D eterm ine th e influence of th e
plate sh ear stiffness on th e deflec
tion.
F ig. B -1 9 : Semi - inf in ite plate strip subjected
to boundary m om ent M0
S o lu tio n :
We now deal with a shear -e la s tic plate, and refer to the basic equations (9.7).
Since no surface load acts, the basic equations reduce to
K A <D = 0

A w = - <D ,

Y - AT = 0 ,

(1 )

where the shear influence factor is abbreviated as


_L _ 1 - v K
xs
2 G ts
In order to solve the coupled partial differential equations, we introduce into (1 )
the following approximation series of products with a seperation of variables,
= S fn ( y ) sin otn x ,

( 2a )

V -

( 2b )

2 g n ( y ) cos otn x ,

w = Jh
n

(y)sina x ,

( 2c )

11

where a n = n it/a . These approximations take into account the simple support at
the boundaries x = const. The circular functions are now separated from ( 1 ) , and
for each n one obtains a coupled system of ordinary differential equations with
constant coefficients, and with the abbreviations d / d x = ( ) x , d/d y = ( ) it
follows in this case

156

9 Plates

fn , y y - n fn

= <

b n ,yy n^n

(3a)
(3b)

>

(3c)

g n - ^ ( g n , y y - !gn) = As solutions to these differential equations we have

(4a)

fn = Cn e

(4b)
h = ( A + ^ L y ) e- ay
n
' n
2a

with

(4c)

= XB + n >

where only the decaying parts of the semi - infinite plate strip are of interest.
We now expand the boundary moment M0 as a FOURIER series
Mn
v =

sin
n a x .

(v5 ')

In order to state the boundary conditions, we have to express the moments in


terms of the solutions( 4 ) . From (9.3) and (9.6) follows that
M yy = K ( * y . y + v * * .* ) = K

" w . y - v w ,xx + ^ ( . y y

v .x x ) +

+ ^ ( ^ x y - v T xy) =
= K { [ - n An + V n An +

+ VV ) ] e' nX +
(6a)

V n ) Cn e_any + ^ Xn n ( 1 V) Dn e ^

Mxy = ^ K ( *y>x + *x<y) . l ^

} sin a

K { - 2 w (Xy + K. [ 2 <D(Xy + I z v

(v T xx - Y - y y )' J J =

(6b)

The boundary conditions then yield :


w ( x ,0 ) = 0
M yy ( x i 0 ) = M 0

An = 0 ,
Mn
K

(7a)

Cn [ 1 + c l > ( l - v ) ] +
+ ^ Xn ( l - ' ' ) D n ,

(7b)

E xercise B-9-3

Cn

_K
-2^C n+ ^ ( - a ^ - X ^ ) D n =0.
2 + Gt,

Mx y ( x , 0 ) = 0

157

( 7c)

T he solution of ( 7b ) and ( 7c ) w ith the sh e a r area of the rectangle


= t

and

_1_

1-v

E t3

12 (1 - v )

2 (l + v) 6 _ t 2

5t

10

yields

C =

M n A

= TT n

K
1+ 2:

(8)

i 7 7 2 - [ ( 1 - v K X + f t2 X
n + ^n 1

For the deflection in ( 2 c) follows


Hows w
\ ith the solution for h n in ( 4 c )
w = 2_ sin a i x a n Jy e 0,1 y

(9)

2K n

In the sh e a r - r ig id case ( G t g > oo , l / x s 0 , X oo ) , only the first


sum m and rem ains in the denom inator of An in ( 8 ) . In com parison, we now focus
on the first term n = 1. Considering the exam ple of a = 10 n t , we obtain
0(1

_ n _

lo t

10

10

t* + V
V lot )

*2

and the factor A 1 is calculated as 0.98 .


Then,

woh

1
0.98

1.02 .

Thus, the effect of taking into account the finite shear stiffness in our plate ex
am ple w ith a = 10 7r t , is th a t the deflection increases by only 2% in com parison
w ith th a t of the sh e a r-rig id plate.
Hence, from an engineering point of view, K IR CH H O FF's plate theory generally
provides sufficiently precise solutions for thin plates ( in the present exam ple t / a
= 1 /lO n 1 / 3 0 ) .
T he influence of the fast decaying p a rt of the solution ( Xn a n ) on the resultant
forces is discussed comprehensively in [B.5].

E xercise B -9 -3 :
A re c ta n g u la r p la te o f c o n s ta n t th ic k n e ss t ( le n g th a, w id th 2b, v = 0,3) is
sim p ly s u p p o rte d a t th e b o u n d a rie s x = 0 , a . A s sh o w n in Fig. B-20, tw o
tu b e s w ith th e s a m e b en d in g stiffn ess E Iy a re w elded to th e p la te alo n g th e
b o u n d a rie s y = b . T h e p la te is su b je cted to a te m p e r a tu re g ra d ie n t
th ro u g h th e p la te th ic k n ess. T h e te m p e r a tu re g ra d ie n t d ep en d s on y, a n d is
g iven b y:
1 ( y ) = 0 1 ( l - m | - ) ,
w h ere

T - T
j = h__ 1. = const

and

0 < m < 1 .

158

9 Plates

T h e tem peratu res of both tubes are equal to th e tem p eratu re of th e plate
m id -p la n e which is everyw here equal to ( T j + T 2 )/2. ( T] and T 2 denote
th e tem peratures of th e plate surfaces along th e x - a x i s ).
a ) Derive th e differential equation for th e deflection of th e plate under the
given tem p eratu re field and boundary conditions by m eans of an energy
principle. T h e tubes can be assum ed to be rigid w ith respect to torsion.
b ) Set up th e general solution of th e differential equations and th e system
of equations for determ ining th e integration constants.
c ) Discuss th e influence of th e elastic su p p o rt from th e two tubes ( p a ra
m eter X = ( E I y ) / ( a K ) , 0 < A < o o ) on th e deflection and th e stresses
of a plate w ith th e ratio a / b = 1 and l6 = 6 X = const ( m = 0 ) .

Vt
b

Section at y = 0 :

F ig. B -2 0 : R ectangular plate w ith two sim ply supported and two elastically
supported boundaries subjected to a tem p eratu re grad ien t 10 ( y )
through its thickness
S o lu tio n :
a)

D ifferential equation and boundary conditions

The derivation is carried out by means of the principle of stationary ( minimum )


total potential energy (6.20). Due to the absence of external static forces in the
current example, the potential of such forces is zero. We now establish the n o n vanishing energy expressions for our problem:
- D eform ation energy of the plate
The deformation energy of the plate consists of elastic energy due to pure bending
and thermal deformation energy due to the temperature gradient through the plate
thickness. The total deformation energy for the plate is given by (9.33)

Exercise B-9-3

159

+ b

U pl

f f f w /XX +1 w /y y /) - 2 '( l - v )' fV w

J
J [v
y = -b x = 0

/XX

w, y y - w

,xy

2 J) +1
(i)

+ 2 ( 1 + v ) a T 10 ( w (XX + w>yy) ] d x d y
with the notations d/ dx = ( ) x , d/dy = ( )

- E la stic energy o f the tubes


Based on [A.8, A.18, A.19] we have for either of the tubes
n

2E

JL(,]
x= o

Considering My ( x ) = - E I , w ( x , b ) xx yields:

E IV
T1 = __ ?- J w ( x , b ) x dx

(2)

The total energy of the system then reads:

n = npl + rit ==>


n = !

J
r =

j
- b

X =

[ ( W.xx + W,yy) - 2 ( 1 - ' ' ) ( W,xxW, y y - W,xy2 ) +


0

+ 2 ( 1 + V) 0T 30 ( w xx + w>yy )] dx dy +
a

Ely
H
T

J w(x-b)l dx

J w(x,-b)*

dx

x = o

and in abbreviated form


n = J J F ( x , y ) dx d y + J f ( x , b ) d x + J f ( x , - b ) d x

(3)

with
F(x,y) =
f f w 'XX +1 w ryy /^ - 2 (v l - v ) 'V
fw
2 1 ''

.xx

w , y y - w , xvy 2 /) +

+ 2 ( 1 + v ) cCp *0 ( W(XX + w yy )] ,
f ( x , b ) =

w ( x , b ) xx

These functions are independent except at the boundaries.


According to (6.20), the variation of the total energy of the system has to vanish,

160

9 Plates

We formulate the EULER equation in analogy with (6.35) as the necessary condi
tion :
dF ( J L . )
+
V dvr y A,
dw Vdw
dw.,
.y
(4 )
(

dF . .)
+ ( . dF ) + ( dF )
=0 .
dw.xx \ vx
\ dw,Xy 'tXy
' dw.vV
vv
yy \yy

Note that 8 I lt only enters in the boundary integral.


With

dw

dF
= 0
dw.x

= 0

dF
= 0
dw v

we obtain
dF
dw. XX

dF
dW,Xy

K
2 4 (1 - ^ ) WfXy *

dF
dw.

**

2 ( W ,xx +

W,yy)

2 (1

V ) W ,yy +

2 ( 1 +

V) a T ^

K
- 2

2 ( W ,xx +

W,yy) -

2 ( 1 -

'O 'A x x +

2 0

V) a T ^

Substitution into ( 4 ) yields the differential equation that describes the problem
w ,xxxx +

2 w , x x y y + W ,y y y y +

or, abbreviated,

C1 +

v ) T ( ^ .x x

^ .y y ) = 0

A A w + ( 1 + v ) a T A *0 = 0 .

(5)

- B o undary co n d itio n s
The variation additionally yields the boundary conditions in analogy with (6.35).
Boundary x = 0 , a :
dF
1

8w

d F
)
K dw,xx
-|x

Vdw,xy J,y J

= 0

(6)

Swx ^ - = 0 .
,X dw.xx
Owing to 8w = 0 ; Sw x ^ 0 (sim ply supported at the boundary x = 0 , a ) ,
boundary conditions at x = 0, a a re :

the

w = 0 ,

( 7a )

w,xx + v w,yy + ( l + v ) a T 10 = O.

(7b)

Boundary y = b
8w - (
)
Vdw.xx A

dF
+ dw..

xy ,x

(8)
8 W y

y dw.vu

= 0 .

Exercise B-9-3

161

Owing to Sw ^ 0; 8 w = 0 ( tubes torsionally rigid, boundary y = b ) , the


boundary conditions at y = b are:
w ,y = 0

( 9a )

EI

w ,xxxx -

W ,y yy ~ ( 2

~ V ) W,xxy

According to ( 9b ), the transverse force


to the beam.
b)

'

(* +

V) T

0 ,y =

( 9b )

at the boundary istransferred as a load

Solution o f the differential equation

Since A *0 = 0 in the present problem, ( 5 ) can be reduced to the bipotential


equation :
AA w = 0 .

( 10 )

The solution to (10) consists of two partial solutions :


w = wx + w2 .

(11)

P artial solution Wj: Solution of the plate strip


2

w i,xxxx =

* wi = A + B x + C x + D x

with the constants A , B , C , D.


Boundary conditions for the plate strip according to ( 7a,b ) for x = 0 , a :
w1 ( 0 ) = 0 and
wl,xx()

wt ( a ) = 0 , ( 12a )

+ ( ! + v ) otj. Xe = 0 and wt

( a ) + ( 1 + v ) otT *0 = 0 .

( 12b )

From ( 12a,b) we determine the constants after which we obtain Wj as


Wj = j ( l + v ) a T 0 j ( a x - x 2 ) ( l - m

) .

(13)

P artial solution w2 :
LEVY's approximation (9.40) :

w2 = ^ Yn ( y ) sin n ^ X .

(14 )

This approximation (14 ) identically fulfills the following boundary conditions:

w2
(

> y

a n d

w2

( a

> y

>

W2 , x x ( . y ) + ' ' w 2 , y y ( - y ) = 0
and

15a )

^ 15b J

W2 , xx( a - y ) + VW2 , yy( a - y ) = -

Thereby, the total solution w = w, + w also fulfills the boundary conditions


(7a, b) .
1
2
Substitution of equation (14) into (10) yields for the n - th te rm :

162

9 Plates

The solution of ( 16) then reads :


Y (y )

(1

v ) n r 8 1( A

, m

h ! S

B S S I . i n i S S L

+ C ..M. S I i + D . i 5 I c c h ^ )
w ith the yet unknown constants An , . . , Dn .
Now we can form the total solution w = Wj + w2 as :
w = ( l + v)ccT e i | y x ( a - x ) ( l - m ^ ) +

( A OTh i S

+ B l S , i n lli a

( 1? )

n 1, 2
, _ . , niry , _ n n y
. n ^ y ^ . nTi xl
+ n sln^
n a"^ cosh
J sin | .
We wish to express all term s in (1 7 ) by means of one sum m ation sign, i.e. we
w rite a FO U R IER series for the first term ( see [ E T 2 | 8.2.4] ) :

f ( x ) = i2- x '( a - x )' = 2 an


n s i n i a

n
The FO U R IE R coefficients are then calculated as follows :
a

2 l r / \ nr r x ,
an J ( x )sln a

1 1 /
an =
x ( a -

n7TX

x ) sin - d x =

n TCX J

x sin - d x

n7TX

sin - dx .

After some calculations we obtain:

2a
3

and the to tal solution thus reads:


w

(1

v ) a T 0 1

00

Ian ( 1

\
-

) +

An cosh~

n - 1,2

n7iy

n 7 ty

n rc y

, _

n rc y

_ nrcy

+ Bn - j - sinh - j p - + Cn sinh - g + Dn - cosh

. n7TX
sin -

(19)

In order to determine the constants, we have to form the derivatives of ( 19) up


to the fourth order ( w ith a n = ( n i r ) / a ) , and then to substitute into th e bound
ary conditions ( 9a,b ) .
First, let us consider the boundary conditions ( 9a ):
,
an m
y = b:
k + An sinh a n b + Bn ( sinh a n b + a n b cosh a n b ) +
"
+ Cn cosh a n b + Dn ( cosh a n b + a n b sinh a n b ) = 0 ;
y = - b:

( 20a )

- An sinh a n b - Bn ( sinh a n b + a n b cosh a b ) +


n

+ Cn cosh a n b + Dn ( cosh a n b + a n b sinh a n b ) = 0 .

( 20b )

Exercise B-9-3

163

For establishing the transverse shear force boundary condition ( 9b ) it is addi


tionally required to expand
(1 +

= (1 + v ) a T 0 1( - ^ )

in a FOURIER series :
= 2 y
sin a. x .

a a n (v - 1 ')n - 1
b
b
n1=
=l1,2
,2
Inserting into boundary condition ( 9b ) now yields:
y = b:

a
f a_ ( 1 - m ) + A_ cosh a b + B_ a b sinh a b +
a nL nV
/

n ' n n
n
'

( 20c)

+ Cn sinh an b + D_Ti a n b cosh a Ti b


11 An sinh a n b + Bn ( 3 sinh a n b + a n b cosh a n b ) +
+ Cn cosh a n b + Dn ( 3 cosh a n b + a n b sinh a n b ) j (2 - v ) - a**^ + An sinh a n b + Bn ( sinh a n b + a n b cosh a n b ) +
L otnb
+ Cn cosh a n b + Dn ( cosh a n b + a n b sinh a n b ) j +
2m

7b a a n4
y = -b :

(_l)n _1]} = ;

El
^ a a n | an ( 1 + m ) + An cosh a n b + Bn a n b sinh a n b Ka
Cn sinh a n b - D_n a n b cosh

{[

( 20d )

An sinh a n b Bn v( 3 sinh a n b ' + n a b cosh a n b ') +'

+ Cn cosh a n b + Dn ( 3 cosh a n b + a n b sinh a n b ) j - (v 2 - v /y


) - 311
- An sinh a n b - B ( sinh a b + a b cosh a n b /) +
<xnb
+ Cn cosh a n b + Dn ( cosh a n b + a n b sinh a n b ) j +
2m
baa*

( - 1 )" - i ] } = o .

Eq. ( 20a 4- d ) presents a linear system


four free constants can be determined.
c)

of four equations by means of which the

Special case o f m 0

Because of symmetry we must have Cn = Dn = 0 . Eqs. ( 20a,c ) then yield the


following reduced system of equationsfor the integration constants with X =
Ely / K a . ( ( 20b,d ) lead to the same equations ) :
An sinh a n b + Bn ( sinh a n b + a n b cosh a n b ) = 0 ,

( 21a)

164

9 Plates

X a a n a n + An cosh a.n b + B_n otn b sinh a n b


( 21b)

- 11 An sinh a n b + Bn ( 3 sinh a n b + a n b cosh a n b ) j - ( 2 - v ) | An sinh a n b + Bn ( sinh a n b + a n b cosh a n b ) j J = I


Transformed :
A_n sinh an b + Bn (' sinh a n b + a n b cosh a n b ') = 0 7,

( 22a )

X a a n Lf a n + An cosh a n b ' + n B_n a b sinh a n b J1 - 2 B_n sinh a n b = 0 .

( 22b )

Eq. ( 22a ) leads to :


sinhanb + onbcoshanb
-B _
sinhanb
From ( 22b ) follows :
coshanb , .
.
/j
V\ (' sinh a n b + a n b cosh a n b ') +
sinhanb

X a a n a n + X a a n Bn

+1 a n b sinh a n b

2 B_ sinh a b

The integration constants can now be stated as:


^ii
B_ =
coshan b + a n b +
An

Bn \( 1

o l

sinhan b

X aan

b coth a n b /) .

N um erical evaluation:
- D eflection fu n c tio n at y = 0:
OO
w = (1 + v ) o T 0 1 ^

( a n + An ) sin

The given numerical values lead t o :


F - 1
4

ub =
n7r ub = n 7T
cl

m = 0

a3 = 0.0048 a

a2 =

l ) X = 0 ( E I y = 0):

Bn = 0

An = 0

w ( 1 + v ) a T 0 j a2 ( 0.12 91 sin
0.1291 a

2 ) X = 1: Bj =

+ 0.0048 sin - )

= 0.0058 a ,

cosh k + jr + sinhn
A3 = - Bj ( 1 +

71 coth 7 1 )

b3

a3

a4 = 0

= - 0.0241 a2 ,

Exercise B-9-3
3 7TX

( 1 + v ) a T Gj ( ay + Aj ) s i n ^ - + a3

= ( l + v ) otj, 0 3 a^ 0.1050 s i n ^a- +* 0.0048 sin


3 ) X = oo : B. =
1

* 291 a = 0.00878 a2 ,

COSh 7T+ 7T

Aj = - Bj (1 + 7t coth n ) = - 0.0364 a2 ,
B3 0
>

A3 ~ o

w ( 1 + v ) a T j a' 0.0927 sin ^ a + 0.0048 sin ~


a

Fig. B-21 presents the curves for the dimensionless deflections


( X! 0 )

w( x , 0 )
( l + v ) a T 0 3a2

at y = 0

M om ents Afxx at y 0:
M xx = K [ W, xx +

V W , yy +

(!

+ V ) T 0 l] .

PQ
w ,xx(x >) = - ( !

+ v ) t 0 1 2

( an +

2
siniLi

The FOURIER series


00
x2
n= 1

n=l n n
5 k x

3 ttx

7t:x
i
4 /I SI

s .m
1------
7t v
a
3

presents the constant one in the interval 0 < x < a .


This leads to
00
W, x x ( x < ) = - ( !

+ ' ( ) T fl1 1 +

\2

An ( ^ r )

ii - 1

sin^

0.05
0.10
0.15
w(x.0)

Fig. B-21: Dimensionless deflection w ( x , 0 ) of the plate

165

166

9 Plates

and with
00

^,yy(x,) = - ( 1 + v ) a T 0 1 J

(An + B j (

\2

) sin^

the m om ents at y = 0 yield the expression:


00 r

Mxx = K ( ! + ' ) ) a T e i ^

1/

(1 - v ) A n - 2 v B n

\2

) s in ^ ^ .

n= 1

Here, we apply the sam e assum ptions as in the case of th e deflection functions,
i.e.,
1 ) X = 0:

2= 0

(l-v )A 1-2vB1 \
a

MXX = 0 .
2 ) X = 1:

( l - v ) A

M
3 ) X = 00 :

2 v B j

0.202 sin

0.202

K ( 1 + v ) 0^ Gj .

( l - v ) A j - 2 v B j -A- = -0 .3 1 4

- 0.314 sin

K ( 1 + v ) a T 0j .

The curves for the dimensionless bending moments

K (l + v )a T 0 ,
at y = 0 are presented in Fig. B-22.
W ith vanishing stiffness of the tubes ( X = 0 ) at the boundaries the plate de
forms w ithout stresses.
a /4

3 a /4

a /2

1=0

0.1
0 .2 0 2 ^ -^

0.2

0 . 3 1 4 ^ < <' 1 = \

0 .3

?u=

0 .4
'

Fig. B-22 : Dimensionless bending moments Mxx ( x , 0 )

Exercise B-9-4

167

E x e r c is e B -9 -4 :
A thin, rectangular plate with all boundaries clamped (dim ensions a, b,
thickness t ) is subjected to a uniformly distributed load with the intensity
P = P0 ( Fi& B' 23 )Determine the deflection function by means of the RITZ method, using the
trigonometrical double series as an approximation function
*

V1

w = Z

m= l n=l

1 ' 108

2m7rx \ ( .

2n7ry \

j ^ l - c o s ----

with amn as free coefficients.


Po

S o lu tio n :
For this task, we first write the internal potential ( strain en erg y ) of a rectangu
la r plate according to (9.33), expressed in term s of the approxim ation function
w*:

ba
n i = i Kj J ] ( W*XX + W*yy)2
00

" V) ( W*xxw,yy - wf xy) ] d x d y

(0

t3
with the plate stiffness K = ----------- and the notation d/d x ( ) . d/d y
( S.
1 2 (1 - v )
v ''x

v Ay'

F or the external potential ( potential of the external forces ) we have, in term s of


w*,

ba

IIe = - J J p0 w* dx dy .
0 0

(2)

W hen applying the R IT Z method, the approxim ation function has to fulfill at
least the geometrical boundary conditions. These a r e :
w* ( 0 , y ) = w* ( 0 , y ) = 0

w* ( a , y ) = w* ( a , y ) = 0 ;

( x , 0 ) = w*y ( x , 0 ) = 0

w * ( x , b ) = w*y ( x , b ) = 0 .

(3 )
W'

168

9 Plates

The principle of virtual displacements (see (6.20) ) is now used as a necessary


condition for determining the unknown coefficients amn

an

_ d(ni + ne) ^ o

^amn

^amn

For the given approximation function

(x.y) =

m= 1 n=l

we now calculate the derivatives


oo

oo

00

00

. 2iti7t:x
22ni t y \
*
X X a ______
2mn; c in
2m7t x/I/ I. r n c _____
W?x = Z Z amn
sln- S I 1 - COS~ b ~ j >
m = 1 n1
X

*
V1 V1
/
W*y = Z
amn v
m = 1n = 1
oo

oo

2irmx \ 2n7t . 2nn:y


a / b S^n b
*

2 2

V NT
4m ti
2m7tx /
w*r X X _______ ____
Z ____
Z amn
2 COS a V
m
n=ln=l
a
oo

oo

oo

o o

V V1
f
Z
A j amn V
m= 1n= 1

2n7iy \
K / 1

2iti7t:x \ 4n k
2nn:y
COS a / 72 COS K *

COS

*
V NT4mnTi
. 2m7tx . 2niry
w ,xy /z_i Z/i a-------mn
a bi sm ;a Sln b
m In r 1

/r \
( 5a)
/
\
( 5b )
. x
( 5c )

For simplicity, the solution based on a single term approximation ( m , n = 1)


only, will be presented in the following.
Substituting ( 5 ) into the expression II = 11; + I I e for the total potential, gives in
view of ( 1) and ( 2 ):

0 0
2 t t x \ 4 7r 2

+ an ( 1 - cos 1 . v cos
b2

2 i t y l2

J 2 167I4
2i t x /
2nx \
27ty/,
27ty \
h 1 - v ) I an ^ 2 - ^ 2 cos
A1 - cos- i r - y cosb- V1 COS_b_ J
2

16 7t4 .

_ a u ~ 2T 2 sm

2/

2 7t X \ .

2/

2 7t y \ 1 1 ,

b a
J J po 3- ! ^ 1 - cos^ ^ ) ( 1 - c o s - ^ - ) d x d y .
0 0

^ r ) s m { b- AJ j dx dy

Exercise B-9-4

169

Integration leads to the following expression :


II = 27i4 K a b a j 1( - ^ +
a

^ ) - p0 a n a b
b

a b

(6)

The application of equation ( 4 ) as a necessary condition for determining the un


known coefficient a jj now leads to

an
4 7r4 K a b a n ( - T + ~TT2 + A ) Poab = 0
11 ^ a4 ' a2b'

^a ll
11

, 4V( 3 , 2 , 3 V
KVa4 + a2,2
b + b,4

4n

T he single term approxim ation w* to the deflection function is thus found to be


*/

Po

w ( x y ) = 7 7 ^ 7 3

2 ir x \ ( ,

2 rr y \

- cos- b ~ ) -

Va4 -I" a2,b 2 + ,b 4 j

Special case o f a quadratic plate (a = b):


w* ( x , y ) = _ ? 4 - ( 1 _ cos^
32ir K V

) ( 1 _ cos^
a

)
a

'

with a maximum deflection of


fc

*/ a

W ma x

a \

Po ^

( ~2 ~ 2 ^ ~

8 k 4K

1
_

^ W '

More precise values can be achieved by employing a f o u r-te rm approach. Com


parison of the maximum deflection obtained in the two cases with

w0

Poa4
it K

yields the following values :


approach

one-term

four-term

1 ,2

1 ,2

WL x / W0

0.125

0.12205

This exam ple clearly shows a relatively fast convergence of the R IT Z method.

170

9 Plates

E x e r c i s e B -9 -5 :
D eterm ine th e m axim um deflection due to a uniform ly distributed load p0
by m eans of th e GALERKIN m ethod for a rectan g u lar plate w ith two clam
ped and two sim ply supported edges (Fig. B-24).
sim ply supported

F ig . B -2 4 : R ectangular plate w ith m ixed


boundary conditions
clam ped

S o lu tio n :
The GALERKIN method requires fulfillm ent of all boundary conditions. In order
to comply with this demand, we form ulate beam solutions for b oth coordinate di
rections. These solutions enable us to fulfill the corresponding boundary condi
tions of the plate. For this purpose, the eigenfunctions of the vibrating beam or of
the beam subjected to buckling prove to be useful, since they also comply with
desirable properties of orthogonality.
In the present task, we will proceed from the basic equations for a vibrating uni
form beam (bending stiffness E I, mass density p, length a ) [ A.20, B.4]:

d w

,2

p o w

To xT + j l To xT ^ = 0 -

(0

By means of the separation approach


w ( x ,x) = 2 X

( x ) sin u

m 111

111

(2)

Eq. ( 1 ) is transform ed with ( 2 ) into an ordinary differentia]equation


stan t coefficients for Xm ( x ) with d /d x ( ) :

Xm ,xxxx(x ) - ^ T Xm ( X ) = 0 -

w ith con

( 3 )

/
2
4 / p Wm

where Xm = a y

, ( m = 1, 2 , . . . ) denote eigenvalues.

T he solution of ( 3 ) reads :
X m( x ) =

C lm sin Xm | -

C 2m c o s X m | -

C 3m s i n h X m T

C 4m c o s h X m | -

(4)

Exercise B-9-5

171

The eigenvalues are calculated from the boundary conditions for a simply supported-clam ped beam :
w(0) = 0

, w>xx( 0 ) = 0

w(a) = 0

y Xm( 0 ) = 0

w x ( a ) = 0 > Xm( a ) = 0

, XB1#xx( 0 ) = 0 >

(5a)

, Xm x ( a ) = 0 .

(5b)

We substitute ( 4 ) into ( 5a ) and ( 5b ), and hereby obtain a homogeneous, linear


system of equations for the constants. The necessary
condition (for no n -triv ial
solutions ) that the determinant of the coefficients must vanish, leads to the char
acteristic equation for the eigenvalues,
tan X
m = tanh
m X

(v 6 ')

Eq. ( 6 ) possesses an infinite number of real-valued solutions Xm . Furthermore,


by means of ( 5a,b) all constants Cjm in ( 4 ) can be reduced by an arbitrary
factor, so the eigenfunctions can be written
v = sin Xm--------,
x
s*nr
r sinh X_x .
Xm
m
ma
smh Xm
m a

f(71
17 \
v

As mentioned above, the eigenfunctions are orthogonal, i.e.


( 0

for

m =/= n ,

for

m ^ n

X mX n d x

x 0

(8 )
a
, xx

,xx

2 , xx dx
Jf x m

x= 0

In th e present example, the same boundary conditions apply for the y -d ire c tio n ,
and we can therefore use th e analogous eigenfunctions Yn ( y ) for this direction.
Thus, the product series approxim ation for the plate deflection function reads as
follows :
w * ( x , y ) = 2 2 WmnXm ( x ) Yn ( y )

(9)

m n

If the distributed load is expanded w ith respect to the eigenfunctions

...
with

= 2 2 P m n Xm ( X) Yn ( y )

( 10a)

J J p X m Yn dxdy
= -id, ------------------

m n

m"

,
( 10b )

JJx^Xfdxdy

and if ( 9 ) and ( 10a ) are substituted into (6.38), we obtain

K2

J J Xi Yk ( A A Xm Yn ) dx dy

- ^ ^ pmJ J X m Yn X; Yk dx dy = 0

By considering the orthogonality this yields :


w*

mn

P mn 1.3 ^6__________

K ( I j I 6 + 2 I 2 I5 + I 3 I 4 )

(11)

172

9 Plates

where
Il = J X m ,x x x x Xm d x , I2 = J X m ,x x Xm d x , I3 = J x ^ d x
0

K .yyyyX A -

^ =

(d /d x = ( ) , x ),

J\

,yy Y dy ,

J Y* dy

( d/dy - ( )#y ) .

Limitation to a sin g le -te rm approxim ation of the form


w* =

w^

X j Yj = w* (sin X j f

yields w ith

sinh X j |- ) ( s m X ^ -

sinh X , | - )

Xj == 3.9266

and after substitution:

Pn a

w* = 0.00198 -Y__ .
Proceeding from w*/X = w*ry = 0 , the maximum deflection is determined as

Pna
w*max = 0.00223

at x = Jy = 0.383 a .

This value differs from the exact solution by ss 3% only.

E x e r c is e B -9 -6 :
A circular plate clam ped at th e outer boundary r = a is subjected to a
constant circular line load q a t a radius r = b ( b < a ) from th e centre
point ( Fig. B-25 ).
a ) D eterm ine th e general solutions for th e inn er ( r < b ) and th e outer p a rt
of the plate. How m any free constants are obtained for th e solution of
this problem?
b ) How do th e boundary conditions an d th e tra n sitio n conditions re ad ?
c ) C alculate th e deflection functions for both th e in n e r and th e outer p a rt
of th e plate. How large is th e m axim um deflection of th e plate ?
d ) Prove th a t th e rad ial and th e tan g en tial m om ents in th e inner p a rt a re
equal.

Fig. B -2 5 : C lam ped circular plate w ith a constant circular line load

Exercise B-9-6

173

S o lu tio n :
a)

The circular plate is divided into an inner and an outer part with the fol
lowing axisymmetrical loads corresponding Fig. B-26.

E>

( r

~ i)

M
inner section

Fig. B-26: Free-body diagram for the outer and inner part of the circular plate
According to (9.^7) , the solution for the outer part ( r > b, index o ) is
wo ( r ) =

C0 +

C i r2 +

C 2 l n i : + C 3 f 2 l n -I: *

( ! )

and for the inner part ( r < b, index i ) we obtain


vi ( r ) =

C4 +

(2)

C5 r

Solution ( 2 ) takes into consideration that for r * 0 the deflection must be


finite.
When solving this problem, we altogether obtain six free constants C0 , . . . , C5 .
b)

B oundary conditions and tra n sitio n conditions

Boundary conditions for the outer part ( r > b ):


w0 ( a ) = 0
M rrc ( b ) =

wo>r( a ) = 0
.

Q ro( b )

( d/dr ( ),r ) ,

- q -

(3a)
(3b )

Boundary conditions for the in n er part ( r < b ):


Mrr. ( b ) = M ,
where M denotes the yet unknown radial moment at r = b.
In addition, the following transition conditions have to be fulfilled for r = b :

(4)

174
c)
-

9 Plates
D eflection f u n c t i o n s

Outer p art

F irst, we determine the derivatives of w0 ( r ) in ( 1 ) :


wo,r

= 2Cl r +
C2

- ZCj - ^

A w o -

w o , r r + T Wo , r = 4 C 1 + 4

-y + C3 ( 2 r l n - A + r )

(6a.)

+ C3 (2ln-- + 3 ) ,

(6b)

C3 ( ln T

(6c)

+ 4)

Substitution of ( 6 ) and (1 ) into the boundary conditions ( 3a,b ) yields :


wo ( a )

= 0 = C 0 + Cj

w o ,r ( a ) = 0 = 2Cl a + ^
Qr0 ( M

= - K

= - 1

^ w

(7a)

a2 ,

(7b)

+ C3 a >

(8a)

o ,- 4 K C 3 i ,

M rrc ( b ) = M = - K [ w o rr + w o r ] =

2 C 1( l + v ) - % l - v ) + C 3( 2 ( l + v)ln-jl- + 3 + v )

= -K

2Cl ( !

v ) - ( 1

Eq. ( 8 a ) yields :

- v ) + C a ( 2 ( 1 + ' ) 1 T

C3 =

+ 3 + v ) =

(8b)

( 9a)

We divide ( 7b ) by a , and ( 8 b ) b y ( l + v ) and s u b tra c t:

C2( ^

+ ^ tS ) + ^ l1

+ v ) l n ^ + 3 + v )1 = k ( i + ^ ) -

By substituting ( 9a ) we obtain :

C = _____ 1_____ {

X.

a2

lK ( l + v)

b2 1

bq
4K

(9b)

' - T T v C 2^ + ' ) l n + 3 + - ) ] }

Solution of ( 7b ) with respect to Cj yields :

c, = - | ( J + , )
For C q one obtains from ( 7a )

C2

bq

2 ? 8K '

(9c)

Exercise B-9-6

175

By that, the deflection function of the outer plate can be determined, however, still
in dependence on the yet unknown radial moment M at r = b.
-

Inner p a r t:

Derivatives from ( 2 ) :

w; r = 2 C s r ,

( 1 0 a)

wi>rr = 2 C s .

(1 0 b )

Substitution of (10 a , b ) into the boundary condition ( 4 ) yields :


Mrr. ( b ) = M
r

M = - K ( w . rr + ^ w . r ) = - K ( 2 C 5 + 2 v C 5 )
M
2 K ( l + v)

(lla)

The second constant C* can be determined by means of the transition condition


(5a):
w a ( b ) =

w i ( b ) =

C4 +

C 5 b2

Mb

(11b)

^4 - W" aa V~
t- 2 K ( 1 + V ) '
( b ;) +

C 4 =

Thus, the deflection of the inner plate according to ( 2 ) is

( 12)

Wi ( r ) = Wa ( b ) - ^ K ^ j ( r 2 - b2) '

In order to write the final expressions for the deflection functions, we determ ine
the radial moment M by means of the transition condition ( 5b ) .
According to ( 6a ) and ( 9c ) :
3,r ( b ) = 2 C 1b + ^

+ C3 ( 2 b l n | + b ) =
(13)

= C,

2( r - ^ ) + 2C3bl n^

Herein, we substitute C2 and C3 from ( 9a,b ) :


M

..<*>) = ( F - y - r T T T ^ t K ( l + v)
a2

b2 l + v

bjq
1 1 ^ 7 ( 2 ( l + v ) l n | + 3 + v)
4K

+ 2 ^4 K

and after reform ulation we obtain


V

r ( b ) =

v+

-b
AI M
b2+ a 2 K
k{
i 2

b -a

bq
4

(1 + v)

2(1 + v ) l n | - 3 - v]} + | ^ l n |

(14)

176

Plates

Equations ( 10a ) and ( 11a ) provide :


vi , r ( b ) =

Mb
K (l + v ) '

( 15)

Substitution into the transition condition and some algebra finally lead to :
( 16)

Eqs. ( 9a ~ d ) substituted into ( 1 ) yields the following expressions for the outer
deflection function:
,( 0

a2 b2 ( l + v )
f
M
b2( 1l + v ) 4+ a2
a2((1l - vv)) I K ( l + v )
Ti _ ( 2 ( l + v ) l n ^ + 3 +

bq
4K

v ) ] j ( |- - ^
+

+ lni) +

bq
4KV 2

With M given by (16 ) follows :

The deflection function for the inner plate then becomes :


u2

The maximum deformation wmax for the circular plate is obtained from (18 ):
Vi( 0 )
d)

qa b
4K~

( 19)

R adial and tangential m om ents

The moments in the inner part are calculated by means of (9.26) :


M rr

= - K [ Wi,rr + T Wi, r] -

MW = - K [ T Wi.r + ' ' Wi.xx]According to ( 10a,b ) it holds that


wi . r ( r ) = 2C5 r

wi i r r ( r ) = 2 Cs .

Substitution of the above derivatives yields :


Mrr

= - K ( 2 C 5 + 2 ^ C 5r )

= - 2 C BK ( l + v ) ,

MW = - K ( 7 2 C 5 r + 2 v C 5) = - 2 C 5 K ( l + v )

Mrr = *LVV = M

The two moments are constant for 0 < r < b and equal to the moment M ( see
(1 1 a )).

Exercise B-9-7

177

E x e r c i s e B -9 -7 :
A circular ring plate m ade of steel ( E , v ) is clam ped at th e in n er boundary
r = b ; th e plate is subjected to th e following harm onically varying line load
at th e outer b o undary ( Fig. B-27 )
t
\
Fi
q ( a > ^ ) = ~ ^ costfi = q i c os v>a ) W rite th e differential equation an d
the b o undary conditions for th e
circular plate.
b ) D eterm ine th e deflection function
for th e circular plate.

F ig. B -2 7 : C lam ped circu lar ring plate


w ith a line load at th e outer
boundary

S o lu tio n :
a) Based upon the differential equation (9.25) for the shear-rigid, isotropic circu
lar plate
AA = ( 4
v dr

+ i A + - L 4 ) 2. = - E < i i )
T dT
r2 dtp J
K

and as we have p ( r , tp ) = 0 (n o surface load) in the present case, the problem


is governed by the homogeneous differential equation
AA w = 0 .

( 1)

The boundary conditions for the problem read :


w(b,ep) = 0 , w
M i i ( a >'P)

= 0

r ( b, cp) = 0 ;
Q r U . ' P )

= q ( a ,<p) = qjcostp .

( 2a,b )
(2c,d)

b ) Deflection fu n c tio n w(r,<p)


Due to the harmonic form of the line load, we assume that the
deflectionfunction
w ( r , <p ) can be written in the following harmonic form with a separation of
variables,
w ( r >P) = wi ( r ) cos <p .

(3)

178

9 Plates

This form corresponds to the first term ( n = 1 ) in (9.49), which upon substitu
tion into the plate equation (9.25) transforms this partial differential equation in
to the ordinary differential equation (9.50) with n = 1 for the function Wj, and
the solution Wj for the homogeneous equation is given in ( 9.51a). As ( 1) is ho
mogeneous, its solution w thus takes the form:
Co
3
1
w ( r , c p ) = C j r + + C 3 r + C4 r In -M cos cp .

(4)

In order to determine the four free constants, we calculate the derivatives


C! - ^ + 3C 3 r2 + C4 l n + C4
C2
C4
2 p r + 6 C 3 r + 1?
w,lplp = Cl r
Aw

(5a)

C O S cP

(5b)

C O S (p

+ ~F+C3r + C4rlnTT

= w jrr + i w r + i w w

(5c)

COS <p ,

8 C
*3, r

+1

2C a

cos <p

( 5d )

The radial moment is obtained from (9.26) with ( 5 a - r c )


M

= -K

+ ^ T w, r + 7 w, ^ ) ]

I c2

c4

KI 2 - j -f* 6 C3 r -f- h
+ A
cos cP
2 ^ ( l - v ) + 2C 3 r ( 3 + v ) + C4 - p ( l + v) j cos<p.

(6)

By means of ( 9.27a) the effective transverse shear force is determined a s :


Q = - K ( A w )>,t
Qr = - K
C- ' 44 ,,

C rj

F + ~y

Qr = - K

4--

rr V( ri w,v> t \ 2 w

2C4
+ 8 C3 ~

C 2

+ -j

l-v (
r v
C4

C1

F"
X1

+ C3 r + in -g- ) jc o s

^ i 4^ C

\Jiifi 1J ,

c 2

, p

^3 ~

C4
r

r
b

+ 2(3 + v)C3 - ( V i c

( 7)

Substitution into the boundary conditions yields :


( 4 ) in ( 2a ) >

b Cj 4 -

( 5a ) in ( 2b )

Cl ~ ^4 +

(6)

in ( 2 c )

1* b C3 0 ,
3 b

C3 +

C4 =

( 8a )
( 8b )

0 >

^ - 5- ^ C 2 + 2 a ( 3 + v ) C 3 + (1 + v ) ^ = 0 ,
fl

3.

( 8 c)

Exercise B-9-8
(7)

179

in ( 2 d ) >

- K |

~ *

C 2 +

i ^

2 ( 3

v ) C 3 -

2 ~V

^ ^ 4

c o s tP =

(l ( a > (p )

<l i c 0 8 1

(8d)

c 2 + 2 ( 3 + v ) C3 _ ( i _ ^ ) C4 = _ | .

Cj can be eliminated from ( 8a ) and ( 8b ) :


C, = - 5 ? - b2C,
C, =

2C2

The relations ( 9 ), ( 8c ) and ( 8d ) form a system


nation of the constants C2 , C3 and C4 :
T T ^2

_ C4

C2

2(3 + v ) a C 3

(1 + v ) 4 - C 4

( 1 - v ) TT C 2

2(3 + v )

( - 3 + v ) - i - C4

(9)

of linear equationsfor determi

_ ^ b ^3

( 1 ~ V) ^ T

- 2 b C3 - C 4 = 0 .

- 3 b2 C, - C.

C3

0 ,

0,

( 10)

= - ^ .

a,

Solution of (10 ) yields the following constants :


=

q i ab2 ( 3 + u ) a 2 - ( l + u ) b 2

( 3 -f-v)a4 + ( l - v)b4
_

qi a ( l - u ) b 2 + ( l + u ) a 2

qi a

8K ( 3 + v )a4 + ( 1 - v ) b 4

4K '

Finally, ( 8a ) becomes :
r, _
1_

c2
b2

1.2 ^
3

_ q i a (3 + v) a4 - 2 ( 1 + v ) a 2b2 - ( l - v)b4
_ 8K
(3 + v )a4 + ( l - v )b4

We can now calculate the deflection function w ( r , <p ) from

4).

Assuming that v = 1 / 3 , with a = 4 b we obtain a maximum deflection at the


boundary given by
3

r( a . 0 ) = ( l 6 1 n 4 - - ^ ) ^

b3

1178 \

E x e r c is e B-9-8:
A s h e a r- r ig id circu lar plate (ra d iu s a , thickness t ) as shown in Fig. B-28 is
resting on a linearly elastic foundation (fo u n d atio n stiffness k [ N / m 3 ]), and
is subjected to a constant surface load p0 .
a ) Express th e to tal p o ten tial of th e given sy ste m .

180

9 Plates

b ) Derive th e differential equation an d th e boundary conditions for the


plate by a calculus of variations approach.
c ) C alculate th e m axim um deflection at th e centre of th e plate by m eans
of th e R IT Z m ethod applying w* = a 0 + a2 r2 ( where a Q and a 2 are
free coefficients ) as an approxim ation to th e deflection fu n c tio n .
N u m eric a l v a lu e s :
E = 2.1 10s M Pa, t = 0.02 m , v = 0.3, a = 0.5 m , k = 0.487 10 N / m 3.
Po

F ig. B -28: C ircular plate subjected to a uniform ly distributed load p0 and


restin g on an elastic foundation
S o lu tio n :
a)

Total potential

The total potential energy is the sum of the elastic energy of the plate l i p , the
elastic energy of the foundation l i b , and the potential of the external load IIe :

n = n p + nb + n .

(i)

The elastic energy o f the circu la r plate can be obtained from (9.35). Since we
are dealing with an axisymmetric problem all derivatives ( )iV 0. Thus, we
obtain the simplified form with d /dr ( )>r

n_ = i - Et3
p

2 1 2 ( l - v2) J [ ( W,rr + T W, r ) "

W,r W,rr dA

(2)

K
The elastic energy o f the fo u n d a tio n is
2 tt a

2 7T

J w 2 rdr d<p.
n b = -^ J J ( k w )wdrrd<p = i k J
00
y) = 0 r= 0

(3)

Exercise B-9-8

181

Finally, the p o te n tia l o f the ex te rn a l forces is

2-7Ta
n e = - J J p 0 wrdrdcp .

(4)

After integration over <p, (1 ) with ( 2 ) and ( 4 ) yields the total potential energy
n =

b)

a
71J{K[(w.rr+ T w.r) -

w,r w fr +

kW2 -

2 pQw ] r dr .

(5 )

D ifferential equation and boundary conditions

In the following, we will determine the differential equation and the boundary
conditions of the problem. The calculus of variations is employed to achieve this
goal.
The basic requirement is that the variation of the total potential energy must
vanish, and in view of ( 5 ) and (6.34) this may be expressed a s :

S II = 7 r s J ' F ( r , w , w r , w r r ) d r = 0

(6a)

with the basic function


F ( r , w , w r , w r r ) = K [ r w n + 2 w r w#rr + T w r - 2 ( 1 - v ) WfI. w r r ] +
+ krw

- 2p0r w .

(6b)

Based on (6.35) , we obtain as a necessary condition from ( 6 a ) the EULER


differential equation
dw

V dW/T Jz

( 7)

Vdw h r

with the boundary conditions


dF
dw,r

( dF 'I 1Sw
^ dw,rr JT i
dF
Sw
dw.rT

If we substitute the derivatives

(8a)

= 0 .

(8b)

= 0

182

9 Plates

into ( 7 ) , we obtain the differential equation for the present problem as


2 k r w - 2 p0 r - 2 k ( v w<rrr + i w>rr - w>r ) +
+ 2 K [ ( 2 + v ) w (rrr + r w rrrr] = 0
,2
1
. 1
. k
Po
W, r r r r + F W, r r r - ^ W,rr + 7 W, r + K W = ^ '

r ct\
(9'

Eq. ( 9 ) is a differential equation of the BESSEL type [ B.3 ].


By substituting the derivatives into ( 8a,b ) , the boundary conditions result as
K ( 7
N.

W,r - T

. . ..

0,

( 10a )

= 0.

( 10b )

W , r r - W,rrr ) Sw

...

...

= Qr

K ( W, r r +

T W , r ) S w ,r

= M

Eq. ( 10a,b ) allows us to establish different combinations of boundary conditions,


c)

Solution by m eans o f the R IT Z method

The following approximation for the deflection function is given :


w*(r) = ^ + ajr2 .

(lla)

The approximation must comply with the essential ( geometrical ) boundary con
ditions, i.e.,
w* ( a ) = 0 ,

( 12a )

w* ( 0 ) =

( 12 b )

Substituting (lla ) into (12a) ((12b) is automatically fulfilled) gives:


a0 -f a2 a = 0 ------Thus, the approximation function only depends on one free coefficient, say, a 2 :
w* ( r ) = a2 ( r2 - a2 ) .

( lib )

The derivatives
w * = 2a2 r

w *r = 2 a 2

substituted into ( 5 ) yield :


a

n = 71:J { K [ ( 2a2 + 2 a 2) 2 ~ 2 ^ r V^ 2 a 2 2 a 2 r ] +
0
+ k a2 ( r4 - 2 r2 a 2 + a4 ) - 2 p0 a2 ( r2 - a2 ) } r dr =
a

= 7 r : J { K a ^ 8 ( l -| -v)r-(- k a 2 ( r 5 - 2 r3 a2 + a 4 r ) o

- 2 p 0a2 ( r - r a ) } d r .

Exercise B-9-9

183

After integration we obtain

n = 7i;|a2[4Ka2( l + v ) + ^ k a 6] + a2 p0-|-a4 J .
From the extrem um condition (6.37) , i.e.
a2 [ 8 K a ( l -f- v ) -(- k a ] + - ^ p 0 a

= 0

(13)

= 0 follows th at

Po*

a^

16K(1 + v ) + - |k a 4
and thus the approxim ative deflection function is determined as
Poa

' CO

16K (l + v ) + |- k a 4
For the given numerical values we obtain
w j = w* ( 0 ) = 1.296

k '

T he exact solution can be found in [ MARKUS, G.: Theory and Calculation of axisym m etrical Structures ( in G erm an). Diisseldorf: W erner-V erlag, 4th edition
1986] as :
ber o t - i ^ b e i ot

ft
wo = k

with

a =

In (14 ) ,
guments,
and Y j (
1950). A
KELVIN

( b e r a - -L ^bei'ot)ber ot+(bei g -f
d

1 -v ,

( 14)

b er' a )b e i a

1 =

ber and bei denote modified cylindrical functions w ith complex ar


named after LORD KELVIN ( see Tables of BESSEL Functions Y0 ( z )
z ) for Complex A rgum ents, New York, Columbia University Press
BESSEL function with im aginary argum ents consists, e.g., of the two
functions :

I q ( x 7 T ) = ber x bei x .
T he m axim um deflection calculated by means of (14 ) leads t o :
w0 = 1 .2 3 9 ^ .
Hence, the ratio between the approxim ate and the exact value is

10. _
= 1.046 , i.e. the agreem ent is rather good.

E x e rc is e B -9 -9 :

Consider a thin, circular, rotatio n ally sym m etric s h e a r-r ig id plate w ith ra
dius a ( Fig. B -29). T h e plate has nonuniform thickness described by the
hyperbolic function

lo-atr"*.

184

9 Plates

for th e distan ce of th e lower plate surface from th e plane upper surface of


th e plate. Here, tQ denotes th e plate thickness at th e outer boundary, an d x
is a free shape param eter. At th e centre of th e plate ( r = 0 ) th e thickness
function tends to infinity, an d a point su p p o rt can be assum ed h ere. T his
local behaviour contradicts th e assum ption for a th in plate, an d therefore
th e results in close proxim ity to th e centre point of th e plate present a
rough approxim ation only.
,P o

t(r)

F ig. B -2 9 : C entre - supported circular plate w ith nonuniform thickness


a ) Derive th e differential equation for a plate w ith variable th ic k n ess.
b)

D eterm ine th e deflection functions for th is type of circular plate ( a =


0.15 m , tQ = 0.005 m ) corresponding to th e th ree values of th e shape pa
ra m e te r x = 0.5; 2.0 an d 3.0. C om pare th e results to those for a plate
w ith constant thickness t Q = 0.005 m .

S o lu tio n :
a)

Form ulation o f the differential equation

We disregard the fact that the m id-surface of the plate is not plane, and proceed
from the equilibrium conditions for a circular plate subjected to an axisymmetric
load ( see [ E T 2 ] ) :
( r Qr ),r +

P ( r ) r

0 *

( r M r r ) , r - MW - Q r r = By denoting the slope of the deflection function by


and circumferential moments read :

( la )

( lb )
( d / d r = ( ) ), the radial

Mrr = K ( r ) ( * fr + v | ) ,

(2a)

MW = K(0 (? +

(2b)

Exercise B-9-9

185

with the radius - dependent plate stiffness


K( r) =

E t> >

K(>- )
Substitution of ( 2a,b ) into ( lb ) leads to the differential equation
4i

+ l ( l +
r v

K (r)-f r U
K( r )

- i f l - K Cr_ k v r ) ^ = M l l
r
K( r )
' V K( r )

If we substitute the prescribed function for the plate thickness


/ \ 3
* ( 0 = .( ! )

(3 )
( '

( >

into the plate stiffness K ( r ) , we obtain


K = K0( i ) ' X
with

(5a)

E to ~ = const .
K = ------0 12 (1 - v )

The derivative of K with respect to r is


K ( r ),r = - K o f ( i r -

( 5b)

The relations ( 5a,b ) are now substituted into ( 3 ), and we herewith obtain a dif
ferential equation of the form

or, alternately,

rx+2

r < L - | - ( l - x ) r i L - ( l - t - v x ) i p = -^.
( 6b)
t^o a
The above equation is an EULER differential equation for the slope
of the de
flection. On the rig h t-h an d side, the yet unknown transverse force Q r occurs
which consists of the constant surface load p and of a concentrated load F at the
centre of the plate (reaction force from the support).
To determine Q r ( r ) in ( 6b ), we integrate once ( la ) with p - const and obtain
r Qr ( r ) = c i -

Here, we determine the integration constant Cj from the boundary condition


Q r ( a ) = 0, and find C x = ( p a 2 ) / 2 . Substituting this and solving the above
equation for Q r ( r ) yields
Qr ( 0 = | ( E^

- pr)

(7)

Eq. ( 7 ) is now substituted into ( 6b ), and we obtain


rx + 2

r2 * ,rr + ( ! - x M , r - (1 + v x ) * =

-pr)-^
.x+1

_2

1 .1/ iU
p r x+1 a 2
2K,A
a
as a differential equation which describes the given problem.

a
p r x-f-3
x+3\

186
b)

9 Plates
Deflection

The solution to ( 8 ) consists of a homogeneous and a particular p a r t.


Homogeneous solution:
By assuming that
ip = C rA

(9 )

the homogeneous equation yields :


X2 - x X - ( l + v x )

= 0.

( l O)

|- 1+ v x .

(ll)

The roots in (10 ) yield


Xj 2 i: J
Particular solution:
In order to determine the particular solution, we assume it to be of the rig h thand-side type. For the first particular solution we write
iL

= A rX+1

=*

= A (x + l ) r X ,
W'r
W i = A ( x + l ) x r X 1 .

( 12)

Substitution of this function and its derivatives into ( 8 ) leads t o :


r 2 A ( x + l ) x r x ' + ( l - x ) r A ( x + l ) r X - ( l + v x ) A r x+1 = ^r- p r
>

A [ ( x + l ) x + (1 - x ) ( x
A x ( l - v) =

+ 1) - (1 + v x ) ] =
^ 5

2 K0 a,x~

2 x ( l - v ) K 0 ax 2
Thus, the first particular solution is :
nrx+1
^P,
=
~oZT\
rl
2 x ( l - v )1is
K 0ax -2 '

^13 )

For the second particular solution we assume that


,

*M = B r

=*

= B(* + 3 ) ' ' 4 .


V r r = B (* + 3 ) ( X + 2 ) r

. ,
( )

...
'

Substitution of ( 14 ) into ( 8 ) then yields :


r2 B ( x + 3 ) ( x + 2 ) r X+1+ ( l - x ) r B ( x + 3 ) r X+2- ( 1 + v x ) B r I + 3 = - ^ _ r x+3
2K0ax

B = - 2 [ 8 + ( 3 - v ) x ] K 0ax

Exercise B-9-9

187

The second particular integral is then written as


p rx+3
2 [ 8 + ( 3 - v ) x ] K 0 ax

p2

(15)

and the total solution then reads:


4> = C j / 1 + C2 / 2 +

pr x+l
2 ( l - v ) x K 0 a11 2

pr,x + 3
2 [ 8 + ( 3 - v ) x ] K 0 ax

( 16)

The free constants Cj and C2 are determined from the following boundary con
ditions:
1)

*(0) = 0.
According to (11 ), X2 must always be negative,i.e. the solution will always
take the form C2/ + 2 I , and therefore we must have C2 = 0 .

2)

Mrr ( a ) = 0 ,

i.e. the radial moment at the outerboundary has to

vanish.

Considering ( 2a ) and ( 5a ) , Mrr can be determined as follows :


(17)

= K M If we form the derivative of ( 16 ) and substitute it into ( 17), we obtain


Mr r ( r = a ) = K0

X C aAl-1 +

( x + 1)p a2
2 (1 -

+1
ia ( c r

(x + 3 )p a 2

Kq

2[8+ (3-

pa
2(l-v)xK0

v )x

]K0

pa
2[8+(3-v)x]K0

)j =

x2 + 2 ( 2 + v ) x + 4 ( l + v )

Cl =

a>i - 3 ( X 1 + v ) K 0 I

( 18)

(8 + (3 -v )x )(l-v )x

The slope of the deflection function can now be calculated by means of (16 ), and
the deflection itself is obtained by integration,
J* = - w,r

w = - J i } j ( r ) d r + C3 .

( 19)

The boundary condition for determining the remaining free constant C3 reads
w(0) = 0

C3 = 0 .

After integration of (19 ), we obtain the following function for the deflection:
x+2
Ai -j-1
Pr
+
(0
^ Xj + 1
2 ( l - v ) x (x +2) K0 a3
( 20a )

x-2

x+4

__________ P______________

2 [8 + ( 3 - v ) x ] ( x + 4 ) K 0 ax
and, in shorthand notation:
Al + i

x+ 4

( r ) = ( K j r Al + 1 - K , r x + 2 + K, r'

)p .

( 20b)

188

10 Coupled D isk-Plate Problems


150 r[m m ]

x=0.5

tn=const

Fig. B-30: Deflection of a point-supported plate with variable thickness


where
K, =

x 2+ 2 ( 2 + v ) x + 4 ( l + v)
a*1 3 (Xj + v ) Kq [ 8 + ( 3 - v ) x ] ( l - v ) ( ^ + l )

1
2 ( l - v ) x a x 2 ( x + 2) Kfl

x_____________
3 ~ 2[ 8 + ( 3 - v ) ] ( x + 4 ) K0 ax
Fig. B-30 shows curves for the deflection for the shape parameters x = 0.5 ; 2.0
and 3.0. These curves may be compared with the solution for a plate with central
support and constant wall thickness
Po
64 K

r4 + 2 a2 r2 (
+ 4 In
Vl +v

which is also shown in the figure. This expression cannot be determined as a spe
cial case ( x = 0 ) of ( 20b ), since the characteristic equation (10 ) possesses a
double root X = 1 for x = 0.

E x e r c i s e B -1 0 -1 :
A s h e a r -rig id rectan g u lar plate (d im en sio n s a, b ; thickness t ) w ith sim ply
supported edges and w ith a stiffener along its centre lin e y = 0 (cross sec
tional area As , bending stiffness E I ys) as shown in Fig. B-31, is subjected to
an ax ial load Nx = c o n s t.
a)

E stab lish the differential equations for buckling o f th e tw o plate parts


in dicated by (T) and (2) in the figure.

Exercise B-10-1

Ix =

const

IWrfl

i
i

CD

J____________

189

(h
A s > EIys

F ig. B -3 1 : R ectan g u lar p late w ith stiffener


b ) Buckling m odes th a t are sym m etrical an d an tisym m etrical w ith respect
to th e longitudinal stiffener m ay occur which can b e described by
m eans of th e tra n sitio n conditions. Discuss th e characteristic buckling
equation of th e stiffened plate.
S o lu tio n :
a) Each of the two plate parts (T) and (2) can be viewed as a rectangular plate
with simply supported longitudinal boundaries at x = a / 2. The corresponding
differential equation reads according to (10.18)
K A A Wj + Nx W; xx = 0

( i = 1 ,2 number of the plate parts ) ,

( 1)

where Nx is considered positive if compression and d/d x ( ) x .


For the solution we use LEVY's series approximation ( 9.1fi) :

with

w ( x f ) = I wm ( j ') silli!T L
m
m = 2, 4 ,6 . . . .

(2)

Substitution into (1 ) yields an ordinary differential equation of fourth order for


wm ( y ) ( d/dy = ( ),y )
- 2 w ( y ) , ( = ) + . m ( , ) [ ( ^ ) * - ^ ( J ? ) 2] = 0
(3a)
With the shorthand notations llj =

t nx =

Wm ( y \ y y y y ~ 2 & l W m ( y l y y +

( 1 ^

,( 3a ) reads :

) Wm ( Y ) = -

( 3b )

Herewith, we have transformed (1 ) into an ordinary differential equation with


constant coefficients. The assumption wm ( y ) = e*y leads to the characteristic
equation
X4 -2 1 1 12 X2 + 0 14 ( l - ^ - ) = 0

(4a)

190

10 Coupled D isk-Plate Problems

with the roots


1 ,2

(4b)
x 3,4 =

nx ) = i / &i ( nx - i) = i x :

The solution to ( 3b ) then reads


wm ( y ) = Cl eKiy + C2 e' Kiy + C3 eiK2y + C4 e ' iK2y .

(5a)

If the exponential functions are substituted by trigonometrical and hyperbolical


functions, ( 5a ) is transformed into :
Wm ( y ) = Acosh Xj y + B sinh Xj y + C cos x2 y + D sin x2 y ,

( 5b )

and thus the general solution of ( 1) including ( 5b ) reads :


w ( x , y ) = [A cosh Xj y + B sinh Xj y + C cos x2 y + D sin x 2 y ] sin m^ X ,

(6)

where the constants A, B, C and D are to be determined by means of the bound


ary conditions.
Because of symmetry, solutions for the plate parts (T) and (2) have the same
form ( 6 ).
b ) The plate as a whole may exhibit buckling modes that are symmetric and an
tisymmetric with respect to the stiffener (see Fig. B-32).
- S ym m etrica l buckling
The corresponding boundary conditions read as follows :
0.

w(x, - ) = 0

(7a)

The bending moment Myy has to vanish at the boundaries y = b / 2 which im


plies, according to (9.11) :
My y ( x ^ ) = - K M

x lf),yy + v w ( x

) . 1 = 0

Assuming this boundary is to remain straight, we therefore demand that


W( X| ) , x X = 0

Ij

antisym m etrical
b uckling

[
|

and thus

W( x !"),yy = 0

sym m etrical
buckling

Fig. B-32: Symmetrical and antisymmetrical


buckling of a plate

Exercise B-10-1

191

Considering ( 2 ) this means :


wm ( ^

~2 \ y y

( 7b )

0 '

With symmetrical buckling, the buckling shape at the stiffener has to have hori
zontal tangents in the y-direction and we therefore have
w ( x , ) fy = ,
i.e. according to ( 2 ) :
wm ( 0 ),y = 0 .

( 7c)

Finally, we have to formulate the transition condition between the single plate
parts and the_stiffener ( see Fig. B-33). Assuming that q = q ( x ) isthe stiffener
loading and Qy. ( i = 1 , 2 ) are KIRCHHOFF's effective transverse forces corre
sponding to (9.14) , we demand th a t:
q = % - Q y 2-

(8)

It follows from ( 9.14c) that


Q y = - K 1[ w, y y y + v( 2 - v >) w > x x y J1,
and from ( 8 ) follows :
cyi

Q
^y2 = -K

i.yyy

2 ,yyy + ( 2 - v) (<

1 ,xxy

I ,xxy ) ] y _ 0

= q

For reasons of symmetry we have for y = 0 th a t:


w l,y

and

W2,y

w i.y y y

w 2 ,yyy '

Thus, the following is valid for the stiffener loading at y = 0:


q ( x ) = - 2 K w, yyy .

(9)

The stiffener loading q resulting from the difference of the effective transverse
shear forces relates to the displacement w via the differential equation of the
bending of a BERNOULLI-beam with an additional axial load [A.19]
F x = Nx
h

d l|=
2

Fig. B -33: Free -body diagram of the transition between plate and stiffener

192

10 Coupled D isk-Plate Problems

i.e.,

q ( x ) = E I ygWfXXXX + N w
h

with the bending stiffness E Iyg , the cross sectional area As , and the height h of
the stiffener. Comparison with (9 ) yields :
E l ys w.l / x x x x +1 Nx -r^
h w,1 /XX -f1 2 K w,1 /

yyy y =

( 10 )

0 ,
o

and by means of ( 2 ) follows


i ^ ) 2 \ E I y S ^ ) 2 ~ NXx ] W- ( ) + 2 K w m(), yyy = 0
By including the TIMOSHENKO parameters [ B.8 ]
EI,ys
Kb

8=

bh

we obtain

(* ? ) - n | =

- < s)

The tra n sitio n condition (10 ) divided by K becomes


b fl> wm V
( o )/ +1 2mV
w ( o/ , y)y y

(W)

With ( 7a -f- d ) we now have four equations for determining the four constants of
(5b).

+ b<D C

- 2 x2

b
co sx 2^

b
sin x 2^

ii

o
II
a

+ 2 xj B

+ *2

ii

' ( Ua )

xl B
b<D A

2 .
b
x2 sin x2 2

2 cos x2 b2 C
x2

v Xj
b A +. Xj2 sinh
i, Xj
b B
Xj2 cosh

Substitution yields the four homogeneous equations :


b
x2
b D = 0
cosh Xj ^ A
+ sinh Xj ^ B + cos X2 C +, sin

The system determinant is:


cosh Xj ^
2 ,
b
Xj cosh Xj

sinh Xj
2 . ,
b
Xj sinh Xj

0
b<D

2x.

* 2 cos>2 2
0
b<D

2 .
b
- x 2sin x 2 ^

= 0 .

( lib )

2 x

The vanishing of the buckling determ inant delivers the so-called buckling condi
tion :
Xj X2 ( Xj - X2 ) +
( Xj tan x2^ - x2 tanh Xj^- ) = 0 .
(12 )
According to ( 4b ) and with the above introduced abbreviations we can formulate:

Exercise B-10-1

193

7r2 K
Since we devote our main interest to the critical load Nx = NXcrjt = k jwrite
^

^ _ rrn _
K ~ * b
with the buckling value k. Therefore, with a = a / b we have

*v)-

<13>

In ( 13) the buckling value k is still contained in the values Xj and x2 . This
equation is therefore solved by tr ia l- a n d - e r r o r .
Fig. B-34 presents the k -v alu es for different stiffness ratios y in dependence on
the dimension ratio a .
For

T = TOT =

8 = bK =

one obtains the garland - shaped curves of the unstiffened plate as already describ
ed in [ET2] with a minimum at k = 4 . For
T= 0

= 0.2

the k -v a lu e drops below 4 which implies that the stiffener under compression
possessing no bending stiffness ( E Iys K 0 ) but a certain cross - section ( As > 0 )
has to be supported by the plate. An arrangement of this type would of course be
unsuitable.

20

16

15

o
J3
"cS
>
Ml

c
12
o

10

3
03

2.85

2
3
a = a /b

Fig. B-34: Buckling value of a plate with one stiffener dependent on the ratio
a = a /b

194

10 Coupled D isk-Plate Problems

- A n tisy m m e tric a l buckling


Eq. ( 9 ) is valid only if the buckling mode is symmetric with respect to the stiffener (Fig. B-35a). However, an antisymmetrical buckling mode as shown in Fig.
B-35b can also occur. If this is the case, the boundary conditions ( 7a,b ) remain
unchanged, i.e.
wm ( 2 )

Wm ( ! ) , y y

( 14a)
(14b)

Furthermore, we have to consider the transition conditions at y = 0:


w m ( 0 )

(14c)

>

wi n ' -( 0 )/ , y y = 0 .

( 14d )

The above conditions correspond to those of a plate simply supported at the


boundaries y = 0 and y = b / 2 . The present buckling case can therefore be redu
ced to that of a rectangular plate with the width b / 2 and all edges simply sup
ported. In analogy with [ ET2 ] we obtain the buckling value:
k = (2a
mi Y
Vnij + 2a )

( 15)

with mj as the number of longitudinal waves in the case of antisymmetrical buck


ling. mj here generally differs from the value m of the symmetrical buckling.
a)

Nv

Fig. B-35: Buckling modes of the


stiffened plate
a ) Symmetrical buckling mode
b ) Antisymmetrical buckling mode

Exercise B-10-2

195

Fig. B -3 6 : Buckling values in dependence of bending stiffness


The buckling value belonging to the antisymmetrical buckling is independent of
the dimensions and of the stiffness (if we neglect the torsional stiffness of the
central stiffener), since during buckling this stiffener coincides with a nodal line
of the buckling mode and is therefore not subject to deflection.
If we present the k - values of both buckling modes in dependence of the bending
stiffn e ss y of the stiffener for a specific width ratio (e.g. a = a / b = 2 ) and a
specific cross-sectional area As of the stiffener (e.g. = 0.2), we obtain the dia
gram in Fig. B-36.
By increasing the stiffness of the stiffener, both the supporting force and the k values increase; when the stiffness of the supporting stiffener reaches a specific
value -(*, either a symmetrical buckling mode with deflected stiffener, or an anti
symmetrical buckling mode with an undeflected stiffener can occur, k* = 16 is
valid both for the former and latter buckling mode ( see Fig. B-36). An increase
of the stiffness of the stiffener beyond the above value y* is obviously useless,
since the plate would nevertheless show antimetrical buckling, and the k -v alu es
would remain unchanged, y* is called least stiffn e ss.

E x e r c is e B -1 0 -2 :
A th in circu lar p late (ra d iu s a, thickness t, Y O U N G s m odulus E ) is
clam ped along its outer boundary as show n in Fig. B-37 an d is subjected to
a constant surface load p th a t leads to large deflections.
T his problem is to be treated as a coupled d is k - p la te problem, and th e de
flection is to be d eterm ined by m eans of th e GALERKIN m ethod w ith a
sin g le -te rm approxim ation. T he results are to be com pared w ith those ob
tain ed by lin ear theory.

196

10 Coupled D isk-Plate Problems

rrrrrij

i'

=*t

Fig. B -37: C lam ped circular


plate under uniform
pressure

S o lu tio n :
We proceed from von KARM AN's equations in operator notation (10.13)

AA w = ^ + ^ 0 4 ( w, ) ,

(ia)

AA = - ^- 04 ( w , w ) .

(lb)

This system of nonlinear, partial differential equations, which is coupled in w and


, is rew ritten in tensor notation according to ( 10.14) :
17^

wU

|7
I70

- k + k
_E

wU

I
2

-J

Uj*
I

'tJ' v

(2a)
(2b)

The problem is m ost conveniently solved in polar coordinates for which reason we
transform the covariant derivatives in the equation by means of ( 2.4O)

12 21 I _i
= + J L 12 12 I _ 1
K
K Wl22l l l + Wl21ll2 +
21 12 I . I , 21 21 I _|
+ Wll2 I21
Wl11 122

E 12 12 1
1
12 21 1
1
"2"[ Wl22Wlll + W|21 WI12

(3a)

(3b)

21 12
21 21
+ Wll2Wl21 + w l11 w | 22
T he m etric tensors, their determ inant, and the CHRISTOFFEL symbols required
for the further treatm ent of the problem, have already been determ ined for
cylindrical coordinates ( see 2.5). If we now substitute the perm utation symbols
(2.20b) and the covariant derivatives ( 2.34b) (e.g. w| jj = w ,i|i e^c- ) bito the
r ig h t- h a n d sides, and if we take into consideration th a t the dependence on the
angle coordinate i;2 <p vanishes in the derivatives owing to the axisym m etrical
shape and loads, the equations reduce with i;1 = r and d /d r = ( ) to

{ r [ l ( r W-r), r L [ r = ^ + ^ ( w . r ,r ),r ,

(4a)

{ r [ T ( r , r), r ],r [ r = - f [ ( w , r )2l r .

(4b)

Exercise B-10-2

197

In order to solve this nonlinear system of differential equations, we first integrate


( 4b ). This yields
-r

= ~ : J r ( J ^ - dr) dr + c i r ( 1 + 21ni ) + 2C2 r +

(5)

Eq. ( 5 ) contains terms with three constants which can be determined from the
following conditions. First, we use the disk stresses according to (8.7)
" T 1 => '

I 6* )

Sv = V
= *-'
Substitution of equation ( 5 ) into ( 6a,b ) yields :
a
o

<p<p

<>>

^ J r ( J ^ L dr ) d r + C1( l + 2 l n i ) + 2 C2 + 5 * f

= -E

7 l r ( J ^ - dr) drl'r + Cl (

3 + 2l n i ; )

2C2 - 7 --

( 7a)

Owing to the condition that the stresses in the middle of the disk ( r = 0 ) cannot
become infinite (fin itn e s s condition), the constants Cj and C3 must vanish.
The remaining constant can be calculated from the boundary condition
u (a ) = 0 .

(8 )

For this purpose we proceed from the material law (8.9) and then replace the ex
pansion <p<p by the radial displacement u, using the corresponding linearized
strain-displacem ent equation of the axisymmetrical stress state in (8.8). (Higher
order terms with respect to the deflection w in the strain - displacement equation
vanish in this case.) It then follows that
u = re w

= i ( w

v 0 rr)

By introducing ( 7a,b ) w e o b tain :

U = r { i l Y j r ( j ^ T ^ d r ) d r ]'r +

(9)

In order to evaluate ( 9 ) , we require an approximation for the deflection w. For


this purpose, we choose the solution of a clamped plate with small deflection
/
_2 \2
w* = c ( l - - j ) .
( 10)
Here, c is a free coefficient describing the maximum deflection at the centre. The
above approximation function fulfills all boundary conditions. After calculation we
obtain

198

10 Coupled D isk-P late Problems

From ( 8 ) then follows


u (r = a) = a

c2

5 - 3

, 2 C 2 (1

v
+

T7

v ) ] _
I

.2

5 -3 v
a2 12(1 - v ) '

Using the approximation w* according to (10 ) , we now state with ( 5 ) an approximation for :
4 6
, 8\
2 5 - 3v
0i** = E 2( r _
( 12)
2
r
+
^
4
)
+
E
r
C
' 6(1 - v) '
a2
a
Da
r
V a4
Further calculation is then carried out by means of GALERKIN's equations,
stated according to (6.38) . By means of ( 4a ) we write the operator
L ( w *,

) = K { r [ i ( r w f r ),r ],r },r - p r - t ( w * * ) , r ,

( 13)

which we substitute into

2x a
J" L ( w *,
0r= 0

J"
<p=

) w* r dr d<p = 0 .

After integration we finally obtain a cubic equation in terms of


2 23 - 9
21

1 - v

-(f)3

16
3 ( 1 - v 2)

Fig. B-38 presents the numerical evaluation of the calculation with v = 1/3. In the
case of higher loads ( p > 5 ), the nonlinear theory leads to smaller deflections
than the linear theory, owing to the membrane forces in the plate which increase
the stiffness against transverse deflection and thus facilitate a higher load bearing
capacity [B.8].

(!)
Fig. B-38: Comparison of maximum deflections obtained by linear and by non
linear theory (dotted lines denote deviation of the pressure for special
deflection values in the scope of the linear and nonlinear theory)

C Curved load-bearing structures


C .l D e f in it io n s - F o r m u la s C o n c e p ts
11

G en era l fu n d a m e n ta ls o f sh e lls

11.1

S u rfa c e t h e o r y d e s c r ip tio n o f s h e lls

11.1.1 R e p r e s e n t a t i o n o f s u r f a c e s
W e assum e th a t th e re exist one-to-one relationships betw een th e curvilinear
coordinates ( GAUSSIAN surface p a ra m e te rs ) ( a = 1 , 2 ) and th e C artesi
an coordinates x1( i = 1 , 2 , 3 ) of th e points P of a surface ( Fig. 11.1). T his
correlation is expressed by
(11.1a)

x
or, in vector n otation w ith th e position vector r
[C .5 ,C .ll] as
r = r ( f ) = x 1( f ) e i

of a point on th e surface
(11.1b)

D ifferentiability w ith continuous first derivatives is assum ed along w ith


non-singularity of th e JACOBIAN m a trix ( functional m a tr ix ):
1

dX

dX

n 1

d X

,x

n 2

Fig. 11.1 : D efinition of th e p aram etric represen tation of a surface

( 11.2 )

200

11 General fundam entals of shells

If ( 2 = const for variable


(11.1b) describes space curves em bedded on
th e surface, an d these curves are called 1-lin e s. In analogy, w ith
=
c o n st, one obtains 2-lin e s. T hese
an d 2-lin e s constitute a cu rv ilinear
coordinate m esh on th e surface ( Fig. 11.1).
a)

Surfaces of revolution

Definition 1:

One obtains a surface o f revolution ( Fig. 11.2), if a two-dim ensional curve m positioned in th e x 1, x 3 - plane ( x 2 = 0 )
1

x = r

3/

, x = 0 , x x ( t)

is ro tated around th e x3- axis as axis of revolution.


1
2
Using th e polar coordinates ( = r , f = d as GAUSSIAN param eters, th e
vector r reads:
r =r

( r , $ ) = r cos $ e j + r sin $ e 2 + x 3 ( r ) e 3

(11.3a)

or according to ( 11.1b)

(11.3b)

Note:

In ( 11.3 a,b ), r describes th e projection of r an d not its value. T he


r-lin e s ( $ = c o n st) are called m eridional curves; for i) = 0 we ob
ta in th e zero-m eridian m . T he $ - lines ( r = c o n s t) are called pa
rallels of latitude. B oth types of lines cover th e surface w ith an
orthogonal p aram etrical m esh (Fig.11.2).

Special case :
- Spherical surface
If th e m eridional curve t) = 0 of a surface of revolution is chosen as a
circle w ith th e rad iu s a an d centre point in th e origin (Fig. 11.3), a sphere
is described. A fter introducing an angle tp, one obtains
1

x = asiny>

x = 0

x = a cosy) .

(11.4 a)

T he position vector r of th e spherical surface th e n reads:


r = r ( y) , $ ) = ( a si ny) ) c o s tf e j + ( a s i ny) ) sin i? e 2 + a cos i p e 3 ( 11.4b)
w ith th e com ponents
x 1 = ( a sin y>) cos

, x 2 = ( a sin y>) sin

, x3 = a cosy) .

(11.4 c)

11.1 Surface theory - description of shells

Fig. 1L2: G eneration of a surface


of revolution

Fig. 11.3:

201

P aram eters of a
spherical surface

On th e surface of th e earth , th e geographical <p, ^-coordinate system is ge-

7T

nerally applied, w here Tp ip denotes th e latitu d e and i? th e longitude


of a respective point. T h e circle of latitu d e <p = 0 describes th e equator.
b) Ruled surfaces
Definition 2\ T h e term ruled surface ( derived from surface reglee =
lin ear su rface) or rad ial surface denotes each surface generat
ed by m oving a straig h t line g along a guide-line d ( direc
trix ) ( Fig. 11.4). T h e single positions of the straight line are
called th e generatrices g of th e ru led surface.

Fig. 11.4 : G eneration of a ruled


surface

Fig. 11.5: Skew hyperbolical


paraboloid surface

202

11 General fundamentals of shells

Such a m ovem ent can be described by defining th e p a th y = y ( ) of a


point P 0 of th e straig h t line, an d a u n it vector z ( ^ 1) w hich points in th e
direction of th e straight line. T he ruled surface is th en form ulated as

r = r U) = yU1) + 2* (f1) .

(U-5)

w here 2 is th e distance of a point P on th e surface (Fig. 11.4) to th e point


P 0 of intersection of th e inh eren t g en eratrix g an d th e directrix d .
F or th is purpose, it is assum ed th a t
^ 0 ,
di x z = -yM, x z 7i.e. th e g en eratrix m ust not point into th e direction of the directrix. T he
( 2 - lines (* = c o n st) are th e lin ear g eneratrices of th e surface. A ccording to
(11.5) th e d ire c trix r ( ^ 1, 0 ) = y ( ^ ) occurs for ( 2 = 0 in th e fam ily of th e
- lines ( ( 2 = const ).
Special cases:
- Cylindrical surfaces
We obtain a special type of a ruled surface if all generatrices are parallel to
each other, a n d if thus th e vector z is constant. In th is case equation (11.5)
reads:
r = r ( ^ ) = y ( ^ ) + ^2z
(11.6)
Surfaces of th is type are called cylindrical surfaces. D epending on th e form of
th e directrices we obtain different types of cylinder surfaces, e.g. elliptical cy
linder su rfa c e :
y ( i?) = a cos i? e j + b sin $ e 2

an d

z = e3 .

(11.7)

- Skew hyperbolical paraboloid surface


A so-called skew hyperbolic paraboloid is generated according to Fig. 11.5 by
m oving th e straig h t line g along th e x1- or x 2 -a x is , respectively, as genera
trices. Here, one also uses th e term conoid surface, th e explicit representation
of which r e a d s :
3
f
l 2
x = ix x .
ab

,..D >
/ 11.8 a)

T he param etrical represen tatio n of th is surface is


r = r ( t ) = ( 1e 1 + 2 e 2 + - ^ Y

e 3

w ith th e dim ensions a , b an d th e height f ( Fig. 11.5 ).

(11.8b)

11.1 Surface theory - description of shells

203

c) T ra n sla tio n or slid in g su rfa ces


Definition 3:

A tran slatio n or sliding surface is generated by a p arallel dis


placem ent of one curve g along a second, a so-called guide
curve or d irectrix d. T his surface can be described as follows:
(11.9a)
w here th e curves shifted in parallel are called th e generatric
es g or sliding curves of th e tran slatio n surface ( Fig. 11.6).
In a m ore extended definition ( ( > ), tran slatio n surfaces
are rep resen ted by
r = r ( f ) = y(^) + z ( f )

11.1.2

(11.9b)

F u n d a m e n t a l quantities of first a n d second o rd e r

Base vectors on the surface


a

( 11.10)

T he base vectors a Q are lying in th e tang en tial plane w hich touches the
surface a t a point w ith th e position vector r ( ) (F ig. 11.7a). F rom the
base vectors ( covariant surface tensors of order o n e ) one obtains, in analo
gy to (2.1 a), th e com ponents of th e covariant m etric tensor or surface te n
sor.
C o m p o n en ts o f the co va ria n t m e tric or surfa ce tensor
F undam en tal quantities of first order - f i r s t fu n d a m e n ta l fo r m of surface
theory &aP = r

T .P

a a

' a p

( 11.11)

204

11 General fundam entals o f shells

Fig. 1L7: C ovariant a n d contravariant base vectors


D eterm inan t of th e surface tensor
a = I &a0

1=

a ll

a i2

a 21

a 22

( 11.12)

Length of an arc elem ent


ds =

(11.13)

dt

( t = curve param eter, ( ) = d / d t ).


A rea of a surface elem ent
dA = / a d* d f

(11.14)

C ontravariant com ponents of th e surface tensor from (2.5c)


a c*7 a 1 P = 6 o:P

(11.15a)

C ontravariant base vectors (Fig. 11.7b))


B _ 01
3. 1 3

(11.15b)

U nit vector p erpendicular to th e surface ( Fig. 11.8)


(11.16)
F u n d am en tal quantities of second order ( Fig. 11.8 ) - tensor of curvature

dr-da3

OiJfP

= ~bapded(>

w ith th e c u rv a tu re co m p o n en ts

(11.17)
I

ft i 3^ , 3.^ 1

ya

(11.18)

11.1 Surface theory - description of shells

205

Fig. 1L8: C urvature of a surface

bll

b12

b 21

b 22

b = lb<J =

(11.19)

C urvature in a point of a curve of th e surface


1 _
_
r
K
' a Q, d r d ^

( 11.20)

C haracteristic equation for principal curvatures Ki ( i = 1 , 2 )


- b^a

rk

aP

+ = k - 2Hk + K = 0
a

( 11.21)

Invariants
TT

J0 ! ,

"2

b/ ?= T b

&f3 = K

m ean curvature ,
GAUSSIAN curvature

K = l b l = T

(11.22a)
. (11.22b)

CHRISTO FFEL symbols in surface theory


p a _ _1_ ciq ,
_
.
1 &eP n + a7e>P V r -e >

1 Pi ~ 2 a

Special c a s e s :

(11.23a)

^3 =
r3 -u

1 a/3 ap

(11.23b)

ri 33a = ir 33
a ir 33
3 uo

GAUSS-WEINGARTEN derivative equations

aa,p f ap a7 + bQ/Sa3
a J/3

3# or

b a/3a 3

b a a /3

(11.24)

206

11 General fundamentals of shells

Exam ple: A p p lic a tio n to su rfaces o f rev o lu tio n


R eferring to Fig. 11.9, th e arc length s and th e angle i? are chosen as
GAUSSIAN surface p aram eters According to (11.3a) th e p aram etric represen tatio n of th e surfaces of revo
lution th en r e a d s :
r ( s , t ? ) = r ( s ) cos $ e j + r ( s ) sin i? e 2 + z ( s ) e 3

(11.25)

W ith ( 11.10) th e covariant base vectors resu lt as


a i = r ,s = r ,s cos1?e i + r

sin ? ? e2 + z s e 3 ,

(11.26)
- r sin 0

a2 =

+ r cos 0 e 2 ,

w here th e derivatives w ith respect to s are denoted by d/ ds = ( )


d/dO = (
. Accordingly, a 2 is parallel to th e x 1, x 2 -plane.

and

W ith
( dr ) + ( dz ) = ( ds )

or

r 2g + z 2 = 1

th e derivative of z is
1 - r

(11.27)

W ith th e chosen m easuring direction of s , z decreases w ith increasing s ,


i.e., z < 0 . Thus, th e negative sign is valid for z.

Fig. 11.9: C oordinates a t a surface of


revolution

11.1 Surface theory - description of shells

207

T he com ponents of th e covariant m etric tensor th e n become


2

2 Q .

- 2Q.

2 ,

ail = S 1 ' ai = r,s C0S * + r,s Sln * + Z,s = r,s + Z,s = 1 a J2 =

a2 =

r g cos i? ( - r sin i?) + r

sin i? ( r cos i?) = a 21 = 0 ,

2 . 2
2
2.
2
a 22 = a 2 a 2 = r sm v + r cos v = r
Thus, one obtains
1

r2

(11.28)

Due to a 12 = a 21 = 0, th e coordinate lines in tersect each o ther perpendicu


larly; i.e. s an d d form an orthogonal system .
By m eans of th e d eterm in an t
a = l a< J =

r2

= r

(11.29)

th e co n trav arian t m etric tensor can be calculated by inversion


1
( * ) =
0

0
, o
J/ r .

(11.30)

In order to d eterm ine th e curvature tensor, th e derivatives of th e base vec


tors are required:
a i,i = r ,.

= r,ss c o s tf e , + r ;8S s i n t f e 2 + z e 3 ,

ai,2 = r,tfs

= r,Bsin,?ei + r ,Bc o si? e 2 = a 21 ,

a2 2 = r ^

= - r cos &

- r sin i?e2

We obtain from th e scalar trip le product (11.18)

bn

r #S B cos i?r

sini?
, SB

zi S I

r 8 cos i?

r sini?

- r sin i?

r cos i?

208

11 General fundamentals of shells

or after differentiation of (11.27) with


r

r
1 - r:

Ai

= r ,

r #s

'

r ,sa

2
1 - r .s

1 - r ,s

The other components are analogously determined a s :

o
(11.31a)
0

~r / 1 - K

or the mixed tensor of curvature

( b p ) = a7 b ^

(lim b )

/ 1 r .:
0

and
. V l - ' r- r

Since a12 = a21 as well as bJ2 = b21 vanish, the coordinate lines are also
curvature lines, i.e. lines with extremal curvature.
The mean curvature results from ( 11.22a ) as
rr

A Z eL 1
/

1- A

+ r

2rJ

's

- 1

(11.33a)

1 rl

and the GAUSSIAN curvature follows from ( 11.22b)


^

IS

^ , 86

(11.33b)

K = b = ~

The CHRISTOFFEL symbols are determined by means of (11.23a). We ob


tain for instance:
r,i

22

,
a

' **2 A, 2

1
=

d
^

\
a A2,2

, 2.
(r } =

a 22,A)

1
2

11
&

22,1

11.2 Basic theory of shells

209

T he fu rth e r values are w ritten w ithout d etailed calculation using ( 11.23a) :

( 0

0
II

r
_S
r

-rr.

r'
r

(11.34a)

A A
/

- r

1 - r:
(UMb)

AA
A A

11.2

B a s i c t h e o r y o f s h e lls

[0 .9 ,0 .1 0 ,0 .1 1 ,0 .1 2 ,0 .1 4 ,0 .1 7 ,0 .1 8 ,0 .1 9 ]

G eom etry o f sh ells


T he shell continuum according to Fig. 11.10 is described by m eans of the
m id-surface of th e shell which halves th e shell thickness t a t each point.
T he shell space is presented by th e GAUSSIAN surface p aram eters ( of
th e m id-surface, and by a coordinate ( p erpendicular to th e m id-surface.
Position vector rp of an a rb itra ry point P of th e shell space:
i p U ^ C ) = r ( ) + Ca 3 U )

(11.35)

C ovariant base in th e three-dim ensional space:


gj = r P j

(11.36)

Eq. (11.36) w ith (11.35) and (11.24) y ie ld s :


Eoc = ( 6 ~ C b f ) a p = n P
a

(11JS7)

e 3 = a3
T he three-dim ensional base e is transform ed into the base
of th e midsurface by m eans of th e shell tensor or shell sh ifter fj, introduced by
NAGHDI [ C.17 ].

210

11 General fundamentals of shells

Fig. 11.10: Coordinates an d base vectors


C ontravariant base in th e three-dim ensional space:
a
g

, -l.o
= (^

) p

P
*

= a

(11.38)

D eterm inan t p of th e shell ten so r:


1-CbJ

~ (b \

-Cb I

l - ( b ]

i - c ( b ; + b^) + c2( b ^ - b > b ? ) =


1 - 2 H C + KC

(11.39a)

w here H is th e m ean curvature, an d K is th e GAUSSIAN m easure of cur


v ature ( ll.ZZa,b) . T he la tte r expression also denotes th e ratio betw een th e
space m etric g an d th e surface m etric a [ E T 1 , 2 ] , i.e.,
(11.39b)

11.2 Basic theory o f sh ells

211

K in e m a tic s o f a sh ell c o n tin u u m


According to Fig. 11.11, the base point PQ on the mid-surface of the shell
allocated to P is transformed into PQ (th e state of deformation is denoted
by ~)The position vector rp consists of the following p a rts:
( 11.40a)

F p U ^ C ) = rp( f C) + v p U ,C)
with the vector of displacement v p
v p U ,C) = v ( ) + C w ( )

The total distortion (norm al rotation and shear deformation)


by the vector w ( Q).

(11.40b)
is described

The following assumptions are made:


a)

Plane cross sections remain plane after loading (see (11.40b)).

b)

Normal stresses t 33 in the -direction are neglected (thin-walled


shell), i.e. 733 = 0 .

Displacements of a shear-elastic shell are described by five independent


components vQ, w , and wQ:
v = ( va + ( w j a + w a 3 .

( 11-41)

This relation denotes a space tensor. Its components vi arereferred to the


spatial base g 1 at the point P of the undeformed shell.

F ig . 11.11: K in e m a tic s o f th e sh e ll c o n tin u u m

212

11 General fundamentals of shells

S tr a in ten so r
CAUCHY-GREEN s stra in tensor according to (4-12b) lim ited to sm all
strains ( lin ear th e o ry ):
( 11.42)
A fter transform ation to th e m id-surface of th e shell we obtain
a ap = T ( v i e

( HAS )

V/? L - 2 K p w )

Norm al an d shear strain s of th e m id-surface of th e s h e ll;


Pop = y ( w J P + W/?L - K v e\p

b6 v p\ot
I + 2 b a.f b op w ')

(11.44 )

A lterations of curvature an d to rs io n ;
'Tsa =
=

(11.45)

+ W,c + ba Ve )

Shear strain s .

S tre ss R e s u lta n ts
In analogy to th e plate problem, th e three-dim ensional shell problem is re
duced to a tw o-dim ensional problem of th e m id-surface of th e shell. R esul
ta n t forces and m om ents are introduced instead of th e stresses w hich are
obtained by integrating th e stresses over th e shell thickness.
+t/2
M em brane forces

is
ppP
g T ae dC

N^ =

(11.46a)

- t/2
+t/2
T ransverse sh ear forces

p. T

oc3 j >
dC ,

(11.46b)

- t/2
+t/2
P ae
M o T CdC

M om ents

(11.46c)

- t/2
N^ and M ^ a re nonsym m etrical because of p P .
In the theo ry of shallow shells one can approxim ate p *
+ t/2
M aP =

TaP( d (

(11.47a)

- t/2
+ V2

a p

p
-

t/2

CtP 1/
dC

(11.47b)

11.3 Shear-rigid shells with weak curvature

213

~ aB
N
is called th e sym m etrical pseudo tensor of resu ltan t m em brane
forces, an d it is valid th a t

( 11.48 )

= Nq/3 + hP M ae
Q
E q u ilib riu m c o n d itio n s [ E T 2 ]
N\
=

- Q bf + PP = 0

(U.48a)

two equilibrium conditions of th e resu ltan t forces in th e m id-surface;


Q l +

(11.48b)

+ P = 0

equilibrium conditions of forces p erpendicular to th e m id-surface;


(11.48c)

- O' = 0
=

two equilibrium conditions of th e resu ltan t m om ents.

C o n stitu tiv e eq u a tio n s fo r iso tro p ic sh ells [C.5]

(11.49a)

= G t a %

(11.49b)

(11.49c)

M aP
stra in stiffness

bending stiffness

sh ear stiffness

Gt =

Et
1 - v2

Et
12(l-i/)

(11.49d)

Et
2 ( l + i/)

and th e elasticity tensor


xjOr/376 _ 1 - 1 / , a-y _/36 , _aS _/3-y ,
2 V _<*/? _-y6 \
n
^ la, a
i a a
1
d, d, i
2 '
1 - 1/
11.3

(11.49e)

S h e a r-rig id s hel ls w i t h w e a k c u r v a t u r e

Here, th e following additional assum ption is m ad e:


-

T he sh ear deform ation due to tran sv erse forces is neglected. T his


m eans th a t points lying on a norm al to th e undeform ed m id-surface
after deform ation lie on a norm al to th e deform ed m id-surface ( n o r
m a l hy p o t h e s i s ) .

214

12 Membrane theory of shells

T his is one of th e principle assum ptions behind BERNOULLI s beam th e


ory an d KIRCHHOFF s plate th eo ry ( see C hapters 9, 10) . It follows th a t
7

= - i2 - ( w a + w rOC + beoc v g ') = 0 ----- >

F urtherm ore, for th in shells N

ot

~ - w , a - heot v g

is su bstituted by N

( 11.50a)'

'

, th a t m eans

= N ^ .

(11.50b)

A sim ilar sim plification is valid for th e curvatures according to (11.44)

+ w /jL) = - [ WU + | ( bve X + t K O U

Kp
(11.51)

Neglecting th e term s w ith b in (11.48a) an d (11.51) leads to th e basic


e q u a tio n s :
- Equilibrium conditions ( 11.48a,b,c)
N ^L + P ^ O

Qia +
M *|

- Q ^ O

(11.52)

Strain-displacem ent relations (11.43,11-44)

= 1 - ( VJ p +

- 2hap w )
(11.53)

Kap = 0 * 0 - - wL

' }

C onstitutive equations (11-49 a,c)

(11.54)
M ^ = K

Equations (11.52), (11.53) an d (11.54) for th e shear-rigid shell provide 3 +


3 + 3 + 6 = 15 relations for th e determ ination of th e 3 N^ + 3 M ^ 43 a ap + 3 ga/3 + 2 vQ + w = 15 unknowns.

12 M e m b r a n e th e o r y o f s h e lls
12.1 G e n e r a l b a s i c e q u a t io n s
A ssum ptio n '.

T he stresses t ^ are uniform ly distributed over th e thick


ness, i.e. only so-called m em brane forces occur, but no bend
ing m om ents an d no sh ear forces are found.

12.2 Equilibrium conditions of shells of revolution

215

Equilibrium conditions
+P^ = 0
N ^ /j + P

=0

( 12.1)

These are three equations for three unknown resultant forces N ^, i.e. the
membrane theory is sta tic a lly determ inate. The resultant forces
can
be calculated from the equilibrium conditions ( 12.1) alone and the defor
mations from (12.2) and (12.3).
- Strain-displacement relations
a *p

= i ( vi e + V/?L - 2 ba^ w )

(***)

- Constitutive equations
N* = D H W %
with

(l j a)

defined by ( 11.49e) or, alternatively,


%e

1
= eT

xtY^

-i-v

( 12.8b)
with

^>ap l8

( &ocS

+ ac7 &p s ) ~ v &ap a7S

- Specific deformation energy according to ( 6.14) or ( 6.15b)


U

1 2. 2

f t D

(12.4)

E quilibrium con d ition s of shells o f revolution

Coordinates:

s or
i?

ip

in meridional direction,
in circumferential ( latitudinal) direction.

Derivatives:

d / d s = ( ),s ,

d/di? = ( ),tf .

- Equilibrium conditions
( r N ss),s
( rNs ,) ,B

+
+ N^ , ^ + C0SV,N ^

+ rpB = 0

(12.5a)

+ r Po

( 12-5b)

216

12 Membrane theory of shells

of revolution
Kj = centre point of curvature
Case 1: A xisym m etrical loading
P = 0

IT

'* =

Introducing ds = r t dip , one obtains


( r N w ) , v " r i cos<^ Nw + r r iPv> = 0 -

( 12.6)

N
N
ie . +
= p
rl
r2
A fter elim ination an d integration, (12.6) leads to
<p

N (pip

i
r2 sin2 (p

J Fl r2 ( p cos tp -

p^ sin <p) sin <p &<p

(12.7a)

tp 0

N vv
,, =

r2
, *
p

N tfiip

(12.7b)

Case 2: N o n -sy m m etrical loading


E xpansion of loads in circum ferential direction by m eans of FO U R IER
s e rie s :
OO
P8( s , t f ) = 2 .Psm
( s ) cos m &
m= 1 m
oo
=

2 p ^ m ( s ) s i nm1,
m= 1

2 P m ( s ) cosmi y
m= 1

oo

p(s, tf)

(12.8)

217

12.2 Equilibrium conditions for shells of revolution


A sim ilar product expansion is chosen for th e re su lta n t fo rc e s:
OO

N_

^^ N ss (V s )> cos m $ ,>


m= 1

N.d-d

2 NM m ( s ) cosmi:>
m= 1

(12.9)

oo
N.

2 Ns^m ( S ) s i nm1 ,
m = 1

Substitution into (12.5) th e n yield th e following set of ordinary differential


equations:
( rN ss ).

C O S ^ N _m +

r P sm
_

( r N s^m ) - m N Wm + COS^ Ns ^m + r Pdm


'm
N

>

( 12.10)

= 0 .

N.
=

Introduction of angle ip an d elim ination of


n ary differential e q u a tio n s:

Pn

in (12.10) yield two ordi

(N
) + (1 H -)cot<z>N
+ m
r. pm c o t y - r. p
v Wm ><v
\
r2 /
w m
r 2 sin ip
1 m
1
(
)
+ 2 cot w N ,
v v v m >.<p
r

( 12.11)

N
H T'Pm = m r .
- r, p ,
sin <p
1 sin p
1

Special shells:
1) Spherical shell ( r j = r2 = a , r = asin<>)
( s i n y ) N v v )iV -I-

- c o s ^ N ^ + a sin <p p^, = 0 ,

(12.12a)

( s i n ^ N ^ )tv

+ cos ip

(12.12b)

+ asin^p^

= 0 ,

Nw + N

= Pa

(12.12c)

B o ile r F o r m u la : Spherical shell subjected to in tern al pressure p0 = const


Stresses

Ng
a._,_ =
'w
t

= aJdd

21

(12.13)

218

12 Membrane theory of shells

2 ) C ircular cylindrical shell ( r = r 2 = a , rt Atp = d x )


Nxx ,x + Y Nxtf ,0 + px

^xtf ,x + Y

= o ,

f0 + Ptf
N

(12.14b)
(12.14c)

C ylindrical shell w ith closed ends subjected to in tern al


pressure p 0 = const
=

Nxx
Poa
t = 2t

N
t

>

II

e,
Stresses

= pa

(12.14a)

>

e*

B o ile r fo r m u la :

(12.15)

3 ) C ircular conical shell w ith sem i-angle a ( r = s sin a , cos <p = sin a ,
ds = T jdip] see Fig. 13.3)
( s Nss ),s + gjn Q

(s

(tf+ N S0 + s p^

),s

sjn Q

+ s pg

= o

(12.16a)

(12.16b)

= s p ta n a

(12.16c)

1 2 .3 E q u i l i b r i u m c o n d i t i o n s o f t r a n s l a t i o n s h e l l s

Hyperbolical shell
C onsidering th e class of tran slatio n shells we restrict our tre a tm e n t to th e
special case of a hyperbolic shell. T his type of shell has a w ide-spread appli
cation especially in th e design of cooling towers. Due to th e ir negative
(hyperbolic) curvature, th ese shells display a bearing behaviour th a t differs
decisively from th a t of shells w ith a positive (elliptical) cu rv atu re (e.g. spheri
cal shells, elliptical shells of revolution).

Fig. 12.2: C oordinates of an axisym m etrical hyperbolical shell

12.3 Equilibrium conditions of translation shells

219

In order to illu strate th is difference, we w ill proceed from th e equilibrium


conditions of th e shell of revolution (12.5). If we solve ( 12.5c) in term s of
th e resu lta n t force
an d sub stitu te th e solution into (12.5a,b), we obtain
a system of two first order p a rtia l differential equations in term s of th e
force resu ltan ts N
a n d N ^ . W ith ds = r t dp th is system re a d s :
( r 2 s i n v N w )#v + ^ c o s ^ N ^ + r , ! ^

^ = r i r2 ( p c o s y - p ^ s i n ^ )

,
(12.17)

( r 2 s i n v N v # )<v + r j c o s y - r a NW f#

- - r i r 2( ptf s i n p + p >#) .

By introducing th e following substitutions


2

U = r 2, sin p~ N tpip
V = r2sin 2p

I
j

(12.18)

the equations are transformed into the sim ple form


2

r
v , + f 2 s in y U v

= r2 sin v3(pcosv? - p^sinv?) ,

(12.19a)

V"f+

= - r i r 2s i n V ( p # s in y + p ^ )

(12.19b)

We now consider only the homogeneous part of the two differential equa
tions ( 12.19). V can be eliminated by differentiating ( 12.19a) w ith respect
to p and (12.19b) with respect to i), and subtracting the equations. We
then have
2
r2 sin 2 p U w + r2 r 2 U w +
sin p ( ^ ~ sin p \ v U v = 0 .
( 12.20a)
W ithin the classification of linear partial differential equations of second
order w e write
A U ,w + C l) ,w + a U v = 0 ,
w here A, C, a
nan t

(12.20b)

a re functions of p. D epending on th e sign of th e discrim i

D = A C = rjT jS in 2^

eq. ( 12.20b) exhibits a different solution behaviour, w here the decisive fac
tor is w h eth er th e product of th e radii of curvature rt r 2 is positive or ne
gative. T h e following cases sh all be co n sid ered :
a) Tj r 2 > 0

D ifferential equation of an elliptical ty p e ( spherical


shell, elliptical shell of revolution, e tc .)

b) Tj r 2 < 0

D ifferential equation of th e hyperbolical type ( hyper


bolical s h e ll)

220

12 Membrane theory of shells

M echanical in te r p r e ta tio n :
A principle way of solving th is problem is to re-transform th e p a rtia l diffe
ren tial equation into two ordin ary differential equations by using a sep ara
tion approach ( see [ C.18, C.19 ] ) :
- S o lu tio n s o f d iffe re n tia l eq u a tio n s o f the ellip tica l type w ith r t r 2 > 0
(sp h eric a l shell, elliptical shells of revolution) a re such th a t discontinu
ities of th e boundary values occuring in th e case of point supports do not
propagate into th e in n er regions but a re confined to a narrow boundary
zone.
- S o lu tio n s o f d iffe re n tia l eq u a tio n s o f th e hyperb o lical type w ith r^ r 2 < 0
display a com pletely different behaviour. T hese solutions a re associated
w ith curves on th e shell surface, so-called ch aracteristics, along w hich dis
continuities of boundary conditions propagate over th e en tire shell [ C.2 ].
T his problem occurs in p articu lar w ith hyperbolical shells w ith single sup
ports. In th is case, th e m em brane theory is not sufficient for determ ining
th e state of stress; bending deform ations an d resu ltan t m om ents m ust th e n
be considered by an extended theory.
12.4

D e f o r m a t io n s o f s h e lls o f r e v o lu t io n

Strain-displacem ent relations due to (12.2)


ipip

=
( u ' V + w )' ,
^
v

(12.21a)

Hs

= 7 (v

,tf

(12.21b)

TW

= T u .0 +

+ U cos <P + w sin tp ) ,


V

7 7

- 7 c os ^

(12.21c)

Special sh ells :
1)

Spherical shell
(1212a)

= t ( u -* + w )
=
'V

( ^

v^ + u c o t^ + w )

= T ( - i I ^ u. * + %

- vcoM

(12.22b)

'

(1212c)

2 ) C ircular cylindrical shell


(12.23a)
= t ( v .* + w )

(12.23b)
(1213 c)

12.5 Specific deformation energy

221

3 ) C ircular conical shell


(12.24a)

s >
= f

s ' sm a

+ U + W cot Q )

_1_ y

s sin a

(12.24b)

>8

(12.24c)

12.5 C o n stitu tiv e eq u ation s m aterial law


R egarding tem p eratu re fields <9 ( ip, i? ) constant over th e thickness
= D [ ew +
N

D [ e*> +

" ( x + t ' ) a T &] ,


v e v v

C1 +

^ ) a T <S*]

w ith

= D -v

(12.25)

Et

D =

1 - i/

or

= rrr(N

Et

+ a

(12.26)

1^,0

_ 2(1 + v )
Et

Substitution of f 12.21) into (12.26) generally leads to o rd inary inhom ogeneous


d ifferential equations of first order w ith variable coefficients. T hese equations
have th e general fo rm :
g

+ P(x)y + Q(x) = 0

w ith th e general so lu tio n :


y = e- J'P(x)dx[ - J q ( x ) eJ'P(x)dx dx +

c]

(12.27)

12.6 Specific deform ation energy


G eneral exp ressio n in m em b ra n e th e o ry according to (12.4)
U = ~n
m& +
2 (' Ntf#
&&&

U = 2 W t\K o +

tptp ww
tfitfi +

<p&urff
ted )

" 2 " NW N*> + 2 ( 1 + v ) K # \

(12.28 a)

( 12-28b)

222

13 Bending theory of shells of revolution

13

B e n d in g th e o r y o f s h e lls o f r e v o lu tio n

13.1

B a s ic e q u a t io n s fo r a r b it r a r y lo a d s

D erivatives: d/ di p = ,v or d / ds = ,s , d / d f l =

- E quilibrium conditions according to (11.52) an d Fig. 13.1:


- rriP v

(13.1a)

,0 + r j e o s ^ N ^ + T j S i n ^ Q ^ = - r r l P 0

(13.1b)

= - r r lP

(13.1c)

(v r N v v
+ ri ^ip-S.S - Tj C O S ^ N ^ + r Q v
>.*
( r V

),lp + ri

( rtO

>H> + ri

( rM

>.*

Fig. 13.1:

'' r

ri sin V N W

- Tj COS if M m

= TTi %

(13.Id )

+ ri

+ ri cos V

= r r iQ^

(13.1e)

Surface param eters an d sign convention for load com ponents


an d stress resu ltan ts of a shell elem ent

13.1 Basic equations for arbitrary loads

223

S tra in -d isp la c e m e n t relations due to (11.51)


S8

= X,. ,
= - r ^

( ls -2a)
+ ^

X,# + ( r

(1M >)

)>B - 2 ip cos v?

w ith th e two an g u lar d istortions ( x ws

( 13.2cJ

>$

w$) according to f11.50a)

u - w
X = ---------

in m eridional direction ,

( 13.3a)

v sin tp - w ^
ip = -----------------'

in circum ferential direction .

( 13.3b)

Special relations for w eakly curved shells of revolution w ith (11.53) [ C.4 ]:
6c <p<p = V
(
) P ,
fj w 'v Ji<

(13Aa)

= " t ( t w . w + - j r v ,.'Pc o sv ) >

( 1SAb)

= - 2 ( w,

(13Ac)

w f# )

- M aterial law
Nss

= D [ ess + v e M - ( 1 + v ) a T 0 ] <

(!3.5a)

NW

= D [ M + " ss - ( 1 + " ) a T 0 ] -

( 1S5b)

Ns*
M ss

K K s

(13.5c)

- ( ! + l' ) a T1e>]

Mw = K [ k m + v k m - ( 1 + v ) ^ 1} ,
^

= K ^ - k

Et
w ith D = -------------- ,
1 - i/2

rr
Et3
K = -------- 5
12 (1 - v 2 )

according to (11.50a)
6
1

(ISM)
(13.5e)

224

13 Bending theory of shells of revolution

E quations (12.21), (13.1), (13.2), (1S.S) an d (13.5) altogether define a system


of 19 equations for 19 unknow ns ( 8 resu ltan t forces, 6 strain s an d curvatu
res, 2 ang u lar distortions, 3 displacem ents ). T hey allow to calculate th e
stress an d deform ation states of shells of revolution.

Case 1: A xisym m etric loads


A ssum ptions:

= Q, = 0

v =

^ =

= 0

(13.6)

- E quilibrium conditions from (13.1):


( t N v v ). v

~ ti cos V n #o + T%

= -thPv>

( i s -7a)

( r Q v ),v

r N w - ri s i n ^ Nw

= rriP .

( 137b)

- Tj COS ip

( 1S-7c)

),V

( rM W

r r i Q v,

- S train -d isp la c e m e n t relations from (12.21), (13.2) an d (13.3)

e<
p<
p = 4 ^ %

w )

(13.8a)

(13.8b)

" ( u cos V + w sin tp ) ,

Kss = X,.

or

(13.9a)

Kvtfi = X, ,

cos w
"w =

(13.9b)

- M aterial law from (13.5 a,b,d,e)


=

D [ ew

=
M

K [L ac<p<p

<>

II

v e M

(1

i / ) a T ]

(13.11a)

(1

v ) a T ]

(13.11b)

(1

i / ) a T 1 ]

(13.11c)

V K

( 1 +

v ) a T * ]

k m

(13.11d)

13.1 Basic equations for arbitrary loads

225

Special sh ells:
1) C ircular cylindrical shell subjected to axisym m etrical loads
D erivative: d / d =

w ith =
3,

- E quilibrium conditions

= 0 -

(13.12a)

Q x .f - Nw + a p = 0 ,

(13.12b)

( 13.12c)

Nxx,f + a Px

= 0

aQ x

E lim inatio n of th e tran sv erse force from ( 13.12c) an d substitution into


(13.12b):
- M M , + a N w = a 2p

(13.13)

w
a

(13.Ua)

- Strain-displacem ent relations


exx
a u *

(13.14b)
B ending angle

* = T w-f

(13.14c)

- M aterial law
+ Um

- (1 + v ) a T ]

+ " exx

- (1 + v ) a T ]

Mxx = K[xx

+ vkm

- ( 1 + v ) a T 1]

+ ^xx

- (1 + v ) a T * ]

II

^>

= D fe
L xx

Nxx

(13.15)

Eqs. (13.14) together w ith (13.15) in (13.13) yield th e boiler equation:

< 4

+ 4k w =

w ith

a P
/ii
\ 2 l/-\
. i 4
0/^
j / - (1 + i / ) a a T & + 4 k a a T &

k4 = 3 ( l - i / 2) ( - y )

( k = decay factor )

(IS.16a)

( 13.16b)

226

13 Bending theory of shells of revolution

Solution for a cylindrical shell w ith sem i-infinite length subjected to


boundary loads M , R ( see Fig. 13.2):
a D
K cos k + M ( c o s k - s i n K ^ ) J e

2 k2 K

a D
R cos
K

2/c

N.

2)

M ( cos k

- sin

R sin k/ + M ( cos k + sin

Qx

R ( cos

( sin

() -

( 13.17a)

je

K^ ,

( lS .l7 b )

) j e

*^,

( 13.17c)

k )

M sin

j e *^

ff5J7d,1

Spherical shell - M ethod by GECKELER

T his m ethod utilizes th e fast decay of boundary disturbances in a circular


cylindrical shell ( x ~
T he essential MEISSNER equations are the
sta rtin g relations for th e approxim ation m ethod [ C.7 ].
T he following two uncoupled differential equations for the bending angle x
an d the tran sv erse force

are obtained ( d / d ip ~ ( ) V )'


(13.18a)

Q ,tptptptp

+ 4 a Q

E ta
K

w ith

- v

v can be neglected in th e case of thin-w alled shells


decay factor given by (13.16b) can be used.
3)

( 13.18b)

1. T hen, the

C ircular conical shell - M ethod by GECKELER

In this case, it is also assum ed th a t a fast decay of th e disturbances from the


boundaries occurs. T he corresponding differential equation reads ( d / d s = ,s) :
Qs

With

+ 4 k , QS = 0

-4
/,
2 \ cot a
K, = 3 (1 - V ) -

(13.19)

13.1 Basic equations for arbitrary loads


C y lin d rica l sh ell

227

S p h erica l sh e ll

(lS.20a,b)

2K k

Xc

2K k

R + M

.
A rn =
s

Xoc

R a3 . 2
M a2 .
------- 5- sin w. H---------- 5- sin ip,
2K k
2K k
Yl

Ma
Ra
2
s
i
n
^
i
T
OT
2 K/c

(lS.21a,b)

Fig. 13.2: D isplacem ents of th e boundaries for cylindrical and spherical


shells in dependence on th e b o undary loads R an d M
Here, th e decay factor depends on th e variable s . Owing to the lim itation to
narrow b o u n d ary zones ( r 2 c o n s t), we can assum e th e decay factor to be
approxim ately constant in these areas. Thus, we obtain th e sam e solution as
w hen considering spherical shells.

Fig. 13.3: Substitute of a conical shell by spherical shells at th e boundaries

228
4)

13 Bending theory of shells of revolution


Com bined shell structures - Solution by M ethod o f T h eo ry o f S tr u c
tu re s [ B.4 , C .4, C.24]

T h e force-quantity procedure, th e so-called M ethod o f T h e o ry o f S tru c tu re s


for th e an aly tical layout of com bined structures can also be used successfully
in shell statics. A fter a p artitio n in g into single substructures ( 0
1 -,
2 -System e tc ..), com patibility conditions have to be form ulated a t th e loca
tions of tran sitio n betw een th e subsystem s.
A p p ro x im a te d e te r m in a tio n o f boundary d istu rb a n c e s fo r conical shells:
Cone-shaped jo in t u n its can be replaced by spherical shells w ith tangential
joining. T h e w all thickness t of th e substituting spherical shell is equal to
th a t of th e conical shell, an d th e rad iu s a of th e su bstituting shell is equal
to the distance r 2i = s; ta n a a t th e boundary of th e conical shell (fru stu m )
( i = 1 or 2) (Fig. 13.3).

1 3 .2

S h e lls o f r e v o lu t io n w i t h a r b i t r a r y m e r id io n a l s h a p e T r a n s fe r M a trix M e th o d

Shells of revolution w ith arb itrarily variable m eridional contours constitute


an im p o rta n t group of com ponents ( casings, com pensators, tu rb in e disks,
car wheels, e tc .).
T h e stru ctu ra l behaviour of th is type of shells can only be calculated ite
ratively. One way of solving th e problem is to assum e the shell to be as
sem bled of single elem ents of shells of revolution, in the following abbrevi
ated as sh ell elem en ts. Here, a tra n sfe r procedure shall be introduced pro
ceeding from th e basic equations ( IS.7J to ( 1S.11) of a shell subjected to
axisym m etrical loading. T hese equations can be w ritten as a system of dif
ferential equations of first order, i.e. th e state equations.
If, on th e righ t-h an d side, we substitute into
equations
an d k .. v*a th e law
th e strain-displacem ent relations ( 1S.8), th e n
equations of first order w ith six unknow n state

( r N ^ Xv

th is system of differential
of elasticity (lS .llb ,d ) and
we obtain six differential
quantities:

i ( u cos ip + w sin ip ) cos tp +


+ y(u

+ w ) cos y

rr,

(13.22)

= D ~ ( u cos <p + w sin ip ) sin <p +


+

+ w )s i n + r Nw - r ^ p

13.2 Shells of revolution with arbitrary meridional shape

(' r M w )>,<p

x c o s ip + i/ cob tp x iV

= K

+ rriQ v .

j)*N w-------r- v ( u cos tp + w sin ip) - w

= u - r,x
ri

229

,
(13.22)

,
ri

M _

v x cos ip

If we assem ble th e sta te q uantities in a sta te vector


y T = (r N w - r(V

rM W u - w - *)

(13.23)

we obtain th e following symbolic notation for (13.22):


y . v = A y + B y ,* , + p

(13.24)

w here the m atrices A , B an d p are given b elo w :

-1

rl
2
D cos ip

t ri
\
siny1+ vjcosip

ri
D cosi^sinv?

(h
\
siny> +v)sm<p

rl

1 rl
D r

0 -L.1

r,
2
K cos ip

A =
VCOS,ip

- 1 - v sin ip
r

1
h
- VCOS ip
T

In the 6 x 6-m atrix B only th e following term s do not v a n is h :


b 14 = D t/cos tp

b 24 = D^sin</?

b 36 = K u c o s t p

230

13 Bending theory of shells of revolution

T he load vector p read s:


P

r r i P v . - r r j P , 0, 0, 0, 0

T he shell is now subdivided into th e above-m entioned sh ell ele m en ts w ith


sm all angles A</^ in such a way th a t th e elem ents of the m atrices w ithin
each single elem ent are assum ed to be constant ( Fig. 13.4).
T his ta sk can be solved by substituting th e first derivative for th e i-th
tw ill by the difference quotient:

All quantities at point i are expressed by values a t point ( i - 1 ) . E quation


( 1S .24 ) th e n reads
yi = y i-i + ^ ( 1

- B i) _1( A j y ^ i + P i )

(13.25a)

Owing to th e suitable stru ctu re of m atrix B ; ( b 14, b 24, b 36 ^ 0, all re m ain


ing by = 0), a p o tential series expansion of th e inverse of ( I - B j) yields
th e following id en tity :
(1 - B . J 1 = 1 +

Fig. 13.4: S h ell elem en t for th e tran sfer procedure

13.2 Shells of revolution with arbitrary meridional shape

231

E quation (13.25a) th e n takes the form


Yi = yi-1 + A V i(I - B j ) ( A jy . J + p . )

(13.25b)

T he sta te
vector y ; still contains th e rad iu s r of th e
i-th subelem ent. As
th e rad iu s
m ay differ from one subelem ent to another,
it m ust be elim ina
te d from th e vector w hen using th e tra n sfe r m atrix procedure. In addition,
the load quantities m ust be included into th e vector, and we therefore re
place y ; by a new state vector z; defined as
ZT = ( Nw > Q*,. Mw . u > w * X , i ) ;

(1S.26)

F rom (13.251) we th e n obtain th e transfo rm atio n for th e i-th su b ele m en t:


z i = C i z i-i

(13.27 a )

T he tra n sfe r m a trix C j shall not be w ritten explicitly here as it can be deriv
ed from ( IS.25 b) .
T he conditions of continuity an d com patibility expressing th a t a t the point of
tran sitio n betw een two elem ents equal forces an d m om ents are tran sferred
and th a t equal deform ations m ust occur, finally yields the tra n sfer procedure
betw een th e boundaries i = 0 and i = n :
zn =

II1 C j Z j ^

= Cz0

(13.28)

T he above m a trix equation represents a set of lin ear equations containing six
equations w ith 2 x 6 unknow n state quantities a t both boundaries. By giving
2 x 3 = 6 boundary conditions at the beginning and a t th e end of the shell,
one obtains a solvable set of equations for th e boundary quantities.
E x te n s io n to sh ells w ith large deflections
If large deform ations are to be tre a te d by a purely linear m ethod, th e sin
gle step procedure proves to be very suitable. Here, the load is applied in
crem entally, an d th e to tal tran sfer m atrix is recalculated a fte r each incre
m ent. W hen using the tra n sfe r m a tr ix procedure, one proceeds from the
equilibrium conditions of th e undeform ed structure, where the position vec
tor r ; for th e i-th shell elem ent is assum ed to be constant, b u t shall be tre
ated as a function of th e displacem ents u an d w . Since th e position vec
tors Tj of th e deform ed stru ctu re cannot be determ ined analytically, m atrix
C can only be calculated for an u n d efo rm ed stru c tu re . Thus, equation
( 13.27a) becom es :
~ 0
,0 , 0 . 0 . ~ 0
z i =
, A p ) z i_j
.
Thus, we obtain as a tra n sfe r rule (Fig. 13.5):
^ k

^k /

k- 1

z i = c i(rj

k- 1

. k x^ k
. A p ) * i_.

(13.27b)

13 Bending theory of shells of revolution

232
w ith

Arj

k/ k

k 'i

= A r ; ( uj , Wi

) .

T he increm ental procedure com prises th e following s te p s :


Step 1:

T h e stru c tu re is considered unloaded and is subjected to th e load


increm ent A p (in d e x 0 ) .

Step 2:

T he re su lta n t forces an d m om ents as well as th e deform ations


a re calculated according to th e lin ear theory.

Step 3:

Forces and m om ents are sum m ed up, and th e contour subjected


to th e load is determ ined on th e basis of th e deform ations. T he
deform ed contour is th en taken as th e sta rtin g point for th e next
load step
> Step 1 ( index 0 > l , 2 , . . . , k

in ( 2 ) ) .

k
T his procedure is repeated u n til th e sum of th e load steps A p equals the
to ta l load to be applied. T hus, th e n o n lin ear load-deform ation-curve is ap
proxim ated by piecewise lin e a r sections as shown in Fig. 13.6.
Since th e equilibrium is established by th e deform ed structure, a correction
is not carried out, an d thus th is procedure has th e disadvantage th a t the
approxim ated solution deviates from th e exact solution w ith increasing
loading. O n th e oth er hand, th is procedure is characterized by num erical
stability an d by a sim ple realization since th e stru c tu ral analysis program
does not require any m anipulation.

Fig. 13.6: Shell elem ent of two


successive load steps

Fig. 13.6: L inear and nonlinear


increm ental procedure

13.3 Bending theory of a circular cylindrical shell

233

13.3 Bending th eory o f a circular cylindrical shell


D erivatives : 6 /
a)

w ith =

d / dtf S

General shear-rigid theory for an isotropic shell by FLUEGGE [C.4]

- Equilibrium conditions from (11.52) after elim ination of


+ Nx ,

N x x,*

= -ap x

: l
= a p*

+
X ( Mxx

+ Mx^

+ M^x j # +

'QQ ) -

= - ap

(1S19)

- R esu lta n t forces - displacem ent relations from (11.53) and (11.54)

Nxx = T [ U' +
XT

+ W)1

D 1- v ,

K( \ - v ),

>

= Ta ~ 22 ( u ^ + V.) ^ 3 ---- Z( w, ^ - V,f) D 1- v


2
( u,tf + v , f ) >
= a
= t *K *

+ w + v n -t> - a

Mxx = - - a2 - [ W, +

(13.30)

" 2 v -* w + " w . )

- 2v,tf - W)]

Mxtf = Mtfx = - T - ( 1 - / ) ( W. - V.f) .


K
= - ~ T ( W.w " 2 v ,tf w + w , )

On th e basis of (13.29) and (13.30) we obtain w ith in the general bending


theory of shear-rigid circular cylindrical shells the following coupled sy
stem of th ree p a rtia l differential equations for th e displacem ents u, v, w :
2

, 1 v
u + j

, 1+ v
+ j

,
+ i/Wff

a Px
= -------g -

^ ,,,,.1

(13.31a)

1+ v
2

u . ftf

v ,-d-d +

V <&

2 ^ [ ( ^

+ w , r 2 k ( wi W - wrtf + wifftf)

2 v . w l

= -

a2 p
D

( 1SSlb)

234

13 Bending theory of shells of revolution

UU,( + V,t? - 2 k ( v , * w -

+ w +
2

k (

AA

- 2

w -

a p
D

( lS.Slc )

,2

w ith

2 v w

( 1S.SU )

a.2 D

12 a 2

B o u n d a ry c o n d itio n s u sin g p r in c ip le o f to ta l p o te n tia l [ ET2 ]


- Clam ped boundary a t x = c o n s t:
( I S M a)

u = v = w= w, = 0 .
- F ree boundary at x = c o n s t:
n

= n x # = m xx =

- C lam ped boundary a t

q x

= o

( 1S.S2 b)

= c o n s t:
( I S M a)

u = v = w = w^ = 0 .
- Free boundary a t d c o n s t:
N

(lSJSSb)

In cases of shear-rigid shells, only four boundary conditions ( the differen


tia l equation is of th e eighth o rd e r) can be fulfilled. T h ree of th e existing
five boundary stress resu ltan ts are re-defined as effective ones ( sim ilar to
KIRCHHOFF s plate th e o ry ), nam ely th e effective tra n sve rse shear forces
M 1? x , f

M xl? ,0

(lS.Si )

and the effective in - p la n e shear force


N
= iN
+
1 XI?
N xl? +

b)

(1S.S5)

S im p lifie d D O N N E L L 's th eo ry [C.3,C.15]

- Equilibrium conditions w ithout ex tern al lo a d s :


Nxx.f + N ,x ,, = 0 ,
Nx**

(1S.S6)

=
)

13.3 Bending theory of a circular cylindrical shell

235

R esu ltan t force - displacem ent relations


N xx

W )]

= ( V,tf + w + i / u f )

N*>

= N. x = T J T i i ( u^ + \ t )
(1337)

M xx

Mw

K
- r
a

( W,

>

T ( W,w + " w ,ff ) -

From (13.36) we obtain, by su b stituting (13.37) , a sim plified, coupled set of


three differential equations for th e d isp lacem en ts:

1
V

1- V
2

+ V

v ,w

u .e*

u ,f

+ v,* +

, 1

+ 1/

v .e* + " w .f

1 - V

v ,ff + w ,*

w + k AA w

(13.38)

= 0

Solution w ith respect to w yields one differential equation of eighth order :


k A A AA w + (1 - i/ )w g g 0

(13.39)

or a coupled system of two differential equations of fourth order for th e dis


placem ent w an d AIRY s stress function $ ( sim ilarly to th e coupled diskplate p ro b le m ) :

3.

AA w +
t

(13.40 a )
(13.40 b)

T he corresponding boundary conditions are analogous to (13.32) - (13.35).

236

13 Bending theory of shells of revolution

c)

S o lu tio n o f closed sh ells u n d e r b o u n d a ry loads


- C om plete th eo ry
W ith px =
= p = 0 , (lS.Sla,b,c) are transform ed into a system of ho
mogeneous differential equations. In th e case of a closed shell, all displace
m ents m ust be functions of th e circum ferential angle i?, since after one
rotation a t & = 2w th e sam e values as in th e in itial point & m ust occur.
T he separation approach using FO U R IER expansion series
oo

u =

oo

^ u m c o sm ti > v =
m= l

oo

2 vm sin r n '^ , w =
m =l

wm c o s m d

(13.41)

m= l

yields from (13.31) a coupled system of o rd in ary differential equations w ith


constan t coefficients for th e unknow n functions um( ), v ( ) , wm( ) ( ( =
x / a ) . T h e given problem is th e n treated fu rth e r by applying exponential
ap p ro x im atio n s:
Um = UeAC

vm = V e ^

Wm = W eAf

(1*4*)

T his leads to a homogeneous system of equations w hich only possesses


non-trivial solutions provided th a t th e d eterm in an t of the coefficients va
nishes. If higher order term s ( k 1 ) a re neglected, one obtains th e charac
teristic equation for th e unknow n eigenvalues A :
A8 - 2 ( 2 m 2 - i / ) A6 +

k4

- y L + 6 m 2(m 2 - l ) ] A 4

2 m 2 [2 m 4 - ( 4 + i / ) m 2 + ( 2 +

(13.43)

) ] A2 + m 4 ( m 2 - 1 ) 2 = 0 .

T his fo u rth order equation in A has four com plex roots. We thus obtain
solutions for
m > 2 ,
m = 0 and

(1S.440-)
m = 1 .

(13.44b)

T he to tal solution consists of th e single solutions of ( IS.440-) an d (13-44b)


(se e [ E T 2 ] for m ore d e ta ils ).
- S im p lifie d th eo ry
Here, we use th e eighth order differential equation (13.39).
W ith
w = We^cosmi?

(13-45)

we obtain th e characteristic equation


A8 - 4 m 2 A6 +

( ^ ^ -1- 6 m 4 ) A4 - 4m 6 A2 +

m8 = 0

(IS. 46)

Com parison of th e eigenvalue equation (IS. 46) w ith (IS. 4S)


shows th a t in
the sim plified theory only th e highest term s in th e coefficients are re tain
ed. T he two theories yield th e sam e resu lts for m = 0 and m > 2 . For the
case of m = 1, however, th ere is no agreem ent. M O RLEY has solved this
problem according to (13.30) by introducing higher order constitutive laws
[ C.16].

13.3 Bending theory of a circular cylindrical shell


d)

237

T h e o ry fo r fa s t decaying b o u n d a ry d istu rb a n c e s

If an a rb itra ry loading is given a t th e boundary, one has to calculate all


p a rtial am plitudes by m eans of th e total solution, sim ultaneously consider
ing th e b o u n d ary conditions. T h e fast decaying p a rtial solution (larg e A) is
predom inantly rem oved v ia th e circum ferential force
an d th e bending
m om ent Mxx, an d one thus obtains an approxim ative theory w ith respect
to th e large roots. Here, th e following sim plifying assum ptions are v alid :
Mxtf =
duces to

= 0 is set in th e equilibrium conditions. T hus, (IS. 29) re


+ Ntfx,tf

= 0 >
= 0 ,

i
-

Mx x , - N^

(* /* s

().( .

().*)

T he stra in exx (12.2Sa) an d th e sh ear stra in 7 XJ> (12.23c) are set


zero. It th e n holds for th e derivatives of th e displacem ents th a t
u

= 0 ,

(IS.48)

u ,tf + v ,( = 0
-

= o

(1S.47)

If th e influence of PO ISSO N s ratio is neglected { v = 0 ) for th e m em


brane force, the sim plified m aterial law (1S.S7) reads
EEl tl ,

XT

+ W) (13-49)
M

T he seven equations ( IS.47) to ( 13.49) allow calculation of th e seven un


knowns
u

Nxx

NXJ>

Mxx

Owing to ( IS.48) , no m aterial law can be given for Nxx an d Nxtf. T hese two
m em brane forces a re obtained from ( IS.47) .
Substitution of (13.49) into (13.47) yields after some re-calculation the
six th order differential equation
w . + 1 i r ^ w - = 0

Introducing (13.45) leads to th e eigenvalue equation


2
A4 + 1 ~ v = 0

( l s -5 )

(13.51)

238

13 Bending theory of shells of revolution

w ith th e roots
^ 1 , 2 ,3,4 = Ki i ^ i

w ith

Ki ==
Ki

jr n

A*i == v/ ~2
/
2 V

~ kk '

(1S.52)

In the scope of this approxim ation th e b o undary disturbances thus decay


independently of th e num ber of circum ferential waves, i.e. in th e sam e
m a n n e r as in th e axisym m etric case (see Case 1 in Section 13.1).
e)

T h e o ry fo r slowly decaying b oundary disturban ces

T his theory plays an im p o rtan t role in th e case of sm all roots in th e ei


genvalue equation, since these roots extend over a large a re a of th e shell.
In this context, a special approxim ation theory has been developed which
is called T h e o ry o f Flexible Shells or Sem i-M em brane T h e o ry [C.1]. As
th e theory om its th e bending forces, it should m ore suitably be term ed
Se m i-B ending T h eory. In th e to tal solution we have shown th a t, in th e ca
se of sm all roots, the m om ent
gains a decisive influence. A
corresponding approxim ation theory can thus be determ ined on th e basis
of the following assum ptions:
-

For the conditions of equilibrium , Mxx = Mxtf = 0 is set. Hence,


(13.29) becomes
^ x x .f +

= 0
_1
a

1
a

N*? -

= 0 ,

(13.53)

T he stra in
an d th e shear stra in 7 xj? vanish. T his requires state
m e n t of th e following couplings betw een th e d isp lacem en ts:
v + w = 0 ,

I
u .o +

v ,e

(13.54)

Considering ( 1S1>4) an d neglecting v in th e m em brane force, th e m ate


ria l law (13.30) reduces to
1VT

N XX =

(1S..55)
+ W)

T h e seven equations (13.53) to (13.55) allow calculation of th e seven un


knowns
u

N xx

>

N xtf

Nw

Solution th en leads to

1"k

v2
+

w .dddddddd +

2 w ,dddddd +

w ,dddd

(13.56)

13.3 Bending theory of a circular cylindrical shell

239

W ith (1S.45), th e eigenvalue equation follows from (1S.56)


1 ------v a\ 4 r, rn
4 / 2
1 \2

( m 11

(1S.57)

If, in accordance w ith DONNELL s approxim ation, w is neglected against


w w , we obtain from
2

1 , v A4 + m 8 = 0

(lS.58a)

th e sm all roots a s :
k 2 if i 2 w ith

' 5 , 6 , 7 ,8

k2

1 v

( lS.58b)

T he sem i-bending th eo ry can be fu rth e r sim plified if th e bending-stiff shell


(b en d in g m om ents are tran sferred in circum ferential direction only) is re
placed by a m em brane shell stiffened by discretely positioned ring stiffe
n e d (Fig. 13.7).
Eqs. (13.53) to (18S5) th e n yield for each shell field ( t he bending stiffness
K of th e m em brane shell is assum ed to be zero):

v.e

Ux

= -^ -N

Nxx,e

.
a

xx

( 1SJ59)

r= M

Nx*,f = 0

According to th a t, th e in-plane sh ear Nxj? has to be constant in each field


( sh ea r fie ld t h eor y ) , w hile Nxx is lin ear w ith respect to x . T he sh ear in
th e longitudinal direction is th e n changed a t th e stiffener ring. If th e ring
is also considered as a shell w ith th e length / r , we obtain from th e second
equation of (13.53) w ith (13.54) an d (13.55)
N xtf,x

((N ^ W r-fN ^ a

K
3

V,

)
(IS.60)

a
i
ft

i+
Js

Fig. 13.7: Stiffened Shell

240

13 Bending theory of shells of revolution

or, w ith th e bending stiffness of a ring E ^ , = l r K r , th e step condition at


th e tran sitio n from th e i-th to th e ( i + 1 )-th field

E Ir

+ 2v

+ V'M )

(1S.61)

a
E quations ( 1S.59) a n d (13.61) suggest assem blage of th e essential field
quantities u , v , Nxx, Nxtf in a state vector, an d to solve th e problem by
m eans of th e tra n sfe r m a trix procedure as described in Section 13.2 [ E T 2 ].
f)

O rth o tro p ic c y lin d ric a l sh e lls

In analogy w ith th e orthotropic plates considered in Section 9.1, th e corre


sponding m aterial laws can also be stated for orthotropic shells. Here, th e
principal stiffness directions are p erpendicular to each o th er (e.g. sandw ich
shell, shells m ade of fibre com posite m aterials, stiffened s h e lls ). A ssum ing
DONNELL s sim plifications, th e m aterial law reads as follows:
D,
n xx = ~ - u ,e +
N t? t?

- T

- ( v .#

N.XI?

( u ? + v . ( )
=

K.
r W.

(13.62)

a
K

M.t? t?

M XT?

( v itf + w )

2 W,T?T?
_

xt?

D epending on th e given m aterial or on th e considered construction, th e


stra in stiffnesses Dx , Dtf, D^ , th e sh ear stiffness Dxj9 , the bending stiffnesses K , K T?, K , as well as th e torsional stiffness K XT? have to be calcu
lated or to be determ ined by experim ents.
Substitution of (13.62) into (13.29) yields a system of equations th a t is a n a
logous w ith (13.31) an d which contains eight independent characteristic v a
lues for th e stiffness as p a ra m e te rs : Dx , D ^ , D ^, Dxtf, K x ,
, K xj9 .
Depending on th e problem form ulation, th e system can be solved by m eans
of approxim ations of th e type (13.45). F u rth e r details concerning stiffened
shells can be found in [B.7, B .9, C .6, E T 2 ] .

14.1 Characteristics of shallow shells

241

14 T h e o r y o f s h a llo w s h e lls
14.1 Characteristics of shallow shells
Shallow shells possess a very large c h a ra c te ristic shell radius or, in other
words, a very sm all, non-vanishing shell curvature. Therefore, a typical beha
viour of such shells also occurs, nam ely th e su p p o rt of transverse loads on
th e m id-surface by m eans of m em brane forces. T his effect has already been
described w ithin th e scope of m em brane th eo ry in C hapter 12.
In addition, th e th eo ry of shallow shells does not neglect com pletely the
tran sv erse forces a n d bending m om ents, but considers them in th e equa
tions of equilibrium of forces p erpendicular to th e m id-surface, as well as
in th e equilibrium of m om ents. Thus, we a re no longer dealing w ith a sta
tically d eterm in ate system , as was th e case in th e m em brane theory, and
th e com putational effort for solving th e shell problem therefore increases.
In th e following, however, it will be shown th a t th e effort does not exceed
an acceptable lim it in com parison w ith a tre a tm e n t by th e com plete shell
theory [ C.7, C .8, C.20 ].

Fig. 14.1:

T ypical form s of shallow shells


a) E lliptical paraboloid over a rectan g u lar base
b) Hyperbolical paraboloid
c) Shells w ith horizontal boundaries over a rectan g u lar base
( so a p -film shel l s)

242

14 Theory of shallow shells

Typical examples of im portant shapes o f shallow shells ( Fig. 14.1)


Let an e llip tic a l paraboloid su rfa ce be extended over a rectan g u lar base
( Fig. 14.1a). T h e explicit form can be derived from [E T 2] an d by a coordi
nate transform ation as
b* )

Fig. 14.1b presents th e form of a h yp erb o lica l paraboloid. In Section 11.1.


th is ty p e of surface has alread y been tre a te d und er th e heading of surfa
ces. In explicit notation these surfaces can be described as follows:
a b
Finally, Fig. 14.1c illustrates a so-called so a p -film sh e ll, i.e. a shell w ith
horizontal boundaries extended over a rectan g u lar base.

14.2 B asic equations and boundary conditions


The following notations of approximation are valid (projections onto the
plane are denoted by
) [ C .ll]:
N* * W 3

OtJ3
M'

(14.1)
P

~ P

V +

P
W zl

(14.2)

Equilibrium conditions according to ( 11.52)


N "t + /
Q lo + N ^ U

=
+ P

M l. - <f

0
(UJ)

= 0

- Strain-displacement relations due to (11.5S) and (14-1), (H-%)

af3

- Constitutive equations due to (11.54)


o(5
N'
D H W q 7 S
M aP = K H W 8-yS

+ z U wL ) >

(14.4)

(14.5a)
(14.5b)

243

14.2 Basic equations and boundary conditions

1 2
Fig. 14.2: Projection of a m id-surface onto th e x , x -plane E of th e threedim ensional space
R eduction of th e num ber of basic equations follows in analogy w ith (7.1S)

Nap
and

= ea l / S$\i6 - Pap

(14.6)
(14.7)

PaP\a = p'3 .

We obtain two coupled differential equations w ith w an d $ for a curvi


lin ea r system of c o o rd in a te s:

,aP ,
K a ^ a 75w |Q/375 - e 7 / 5 z L
# \
Lys -- pP + P'
ap9

lap

= 0
(Hi)

+ E

= o

Eqs. (14.8) expressed in C artesian coordinates ( d / d x = ( ),x , d/ dy = ( ),y)


read as follow s:
K A A w - Q4 ( z , ) = F j ( x , y )
(14.9)

A A + E t Q 4 ( z , w ) = F2( x , y )
w ith

AA = (

)>xxxx + 2 (

) xxyy + (

) yyyy

bipotential operator,

244

14 Theory of shallow shells

0 (f)g )
a nd

(xx&,yy+ ^.yyS,xx

^^.xy^xy '

F 1( x , y ) = p - z xxJ p x dx - z yyJ p y dy

F2(X>y) =

( j P x d x ),yy +

(J P y dy ),x x

"j
,

(14.10)

~ " ( Px ,x + Py.y) -

Special c a se s:
1) C urvature an d disto rtio n vanish in one direction
-

z,xx = z,xy = 0

>

z,yy = y = fin ite .

T he differential equation (1 4 3 ) can be sim plified w ith


K A A w - Ky $ xx

= p ,

A A $ + E t Ky w xx = 0

px = py = 0:
( 14.11a)
( 14.11b)

T he above equations correspond to th e differential equations of a cylindri


cal shell ( see ( lS.^OaJb) ).

2) No cu rv atu re or distortion occur in b o th directions


>

z ,xx = z

,yy =

z ,xy = 0

T he system of differential equations th e n splits into th e two uncoupled


differential equations
KAAw = p ,
AA#

= 0

(14.12a)
.

(14.12b)

T he first relatio n ( 14.12a) is th e differential equation of KIRCHHOFF s pla


te theory ( 9.1S) w hile ( 14.12b) is a special case of th e differential equations
of th e theo ry of disks (8.1) following from th e com patibility condition.
B o u n d a ry c o n d itio n s
At each of th e four boundaries of th e reference plane, boundary stress re
su ltan ts ( Nxx, Nxy, M xx, Mxy, Qx , or Nyy, Nxy, Myy, Mxy, Qy ) or boun
dary displacem ents or -slopes ( u, v, w, w x , w ) can be described. Howe
ver, since th e order of th e system of differential equations only possesses
four boundary conditions, so-called effective tran sv erse sh ear forces (1S.S4)
an d one effective in-plane shear force have to be introduced in analogy
w ith KIRCHHOFF s plate theory ( see 9.1). T he effective forces re ad as fol
lows :

14.3 Shallow shell with constant principal curvatures

245

Q
xe = Q
x +1 M x y . y
^Qy e = Q + M xy,x

} '

<U,S>

In case of a shallow shell, th e effective in-plane sh e ar force Nxye can be


replaced by th e sh ear force Nxy. In order to avoid confusion w ith projected
forces according to (14.1) , th e effective forces a re h ere indicated by ( ) .
T he following boundary conditions m ay be form ulated for a boundary x =
c o n s t:
- C lam ped edge
u = v = w = w x = 0 .

(14.14a)

- Simply supported
Nxx = Mxx = v = w - 0

(14.14b)

- Free edge
Nxx = Nxy = Mxx = Q Xe = 0

14.3

(14.14c)

Shallow shell over a rectangular base with constant principal


curvatures

T his tye of shell can often be found in civil engineering applications, e.g.
as a ty p ical roof construction extended over a rectangular base (le n g th 2a,
w idth 2b). T h e m id-surface of th e shell is defined by z = z ( x , y ) , w here
th e following ch aracteristical values are a ssu m e d :
z

z,xx = Kx = const

/yy

= k = const
y

z /xy
v =0

(14.15)

T he shell is sim ply supported a t all boundaries, an d is subjected to a ver


tical surface load p ( x , y ) ( px = py = 0 ) .
F rom th e system (14-9) we obtain w ith (14-15)
KAAw - k

>yy

y $ xx

(14.16a)

= P

A A # + E t ( / c x w yy + Ky w xx ) = 0

(14.16b )

Using an auxiliary function i p ( x , y ) and th e approaches


w = A A ip ,

(14.17a)

$ = - E t 0 4 ( z , V ) = - E t ( / c x 7p yy + Ky V>(XX) = 0

(14.17b)

246

14 Theory of shallow shells

( 14-1^ ) is now transformed into a partial differential equation of eighth


order. The differential operator 0 transforms into a modified LAPLACEoperator with constant coefficients A *ip = k ip
The approaches for w and < (14-17) identically fulfill ( 14-16b) . Substitution into
( 14-16a) then yields

( 14-18)

A A A A ip + ~ A * A * ip = ^ K
K

The approaches for the auxiliary functions ip (14-18) are also substituted
into the relations for the stress resultants (14-6) and (l4-5b), and we thus
obtain in Cartesian coordinates
= -E tA * ,S ,
N =

<f.xx = - E t ( - = , l f ' V+ S ^ )

= - E t i * ^ xx ,

Nxy = - ^,xy = E t ( y ^,yxxx + Kx ^,yyyx )

= E t A * ^ >xy ,

Mxx = - K A A (^,xx + tf.yy) Myy = - K A A ( ^ yy + ^ J

Mxy = - K ( l - ^ A A ^ x y

(14-20)

In the follwoing, a solution shall be given for a shell that is simply


supported at all edges. For this purpose we draw on the treatment of the
simply supported, shear-rigid plate (see Section 9.2). This problem was sol
ved using a FOURIER double series expansion that strictly fulfilled both
the KIRCHHOFF plate equation and the boundary conditions. The shallow
shell is treated analogously by choosing a FOURIER double series expan
sion for the auxiliary function ip:
OO OO
iH * ,y ) =

m=l n=l

sin v

s i n - n -x

(U-M)

where ipmn are free FOURIER-coefficients ( m , n = 1 , 2 , 3 , . . . ) .


It can be shown that the above approach fulfills the boundary conditions
of the simply supported shell according to ( 14.14b)-

C.2 E x e r c ise s
E x ercise C - ll- 1 :
A circu lar conical surface constitutes a
special case of an elliptic conical sur
face, an d belongs to those conical sur
faces th a t can be described by moving
a gen eratrix (p a ra m e te r) along a di
rectrix y ( j?) (c irc le w ith rad iu s a ) p a
rallel to th e x ^ x 2-plane (see Fig. C -l).
T he position vector r of a point P on
the surface reads in param etric pre
sentation;
r= r(s,$) =

s sin a cos

-I- s sin a sin &e2 +

Fig. C-l: C ircular conical surface

-(- s c o s a e 3
w ith

s ,$
a = const

GAUSSIAN param eters ,


semi-angle of a cone .

D eterm ine
a ) th e fundam en tal quantities of first an d second order ,
b)

th e equilibrium conditions for th e m em brane theory of a circular coni


cal shell.

S o lu tio n :
a)

Fundam ental quantity o f fir s t order - surface tensors

By means of the given parametric representation of a circular conical surface

(,#)=

s sin a cos $
s sin a sin &
s cos a

(1)

we determine from (11.10) the covariant base vectors:


dr
de

where E,1 * s , I;2

It then follows
a ,1 = r fS =

sin a cos 8
sin a sin &
cos a

a 2 ~ r ,tf

- s sin a sin &


s sin a cos &
0

( 2a,b )

248

11 General fundamentals of shells

By means of ( 2a,b ) and according to (11J1), the covariant components of the sur
face tensor ( f i r s t fu n d a m e n ta l fo r m fo r the surface ) are calculated as :
a a /3

all = al ' al = 1

s* sin* a
0

The covariant surface tensor thus reads :

(3a)

s2 sin2 a

and the determinant due to (11.12)


(3b)

a = \ &ap\ = s2 sin 2

The diagonal form of ( 3a ) ( a 12 = 0 ) implies that the parametric lines are mutu
ally perpendicular (orthogonal mesh). The contravariant surface tensor can be
obtained by forming the reciprocal values of the elements of the principal
diagonal, i.e.

(4)

( * * ) = ( a p ) _1 =

6) F undam ental quantity o f second order - curvature tensor


The curvature tensor constitutes the second fu n d a m en ta l fo rm for the surface.
The single components are calculated by means of ( 11J8) :
[ * a . 0 . " V * 2]

/ f ---------

b =
with the derivatives

1.1

3a 1
3s

f s sin a cos 3 '


- s sin a sin-3

(0 \
0

2,2

33

\0 J

1.2 _ 2,1

3s

f - sin a sin 3 ^
sin a cos 3
V

One obtains the components of the curvature tensor by formulating the scalar
triple products :
0
sin a cos 3
- s sin a sin <

0
sin a sin 3
s sin a cos 3

0
cos a
0

0 ,

Exercise C-ll-1
sin a sin &

sin a cos &

sin a sin &

cos a

- s sin a sin d

s sin a cos

- s sin a cos
b 22

- s sin a sin
b, =

sin a cos A
- s sin a sin &

sin a cos &

sin a sin $

cos a

s sin a cos &

249

s sin a cos a ,

The curvature tensor thus reads

( ha p )

s sin a cos a

(5a)

(5b)

with the determinant b = | bQ^ | = 0 .

The form of the fundamental quantities allows us to draw the following con
clusions :
-

a 12 = 0 and b 12 = 0 mean that the parametric lines are simultaneously lines


of principal curvature.

b jj = 0 implies that the curvature is zero along the parametric line s .

The curvature at a point P of the surface can be calculated according to (11.20)


b

&

i f

Jll

1
1

a ^

OB

R ,

1
R

1
2

hi

= o

(6)

_^L = _ J _ c o t a
3>oo
22

The two invariants describe the curvature properties of a surface ( see ( 11.22a,b) ):
H = 4 a a/3'

a/0

K =

mean curvature ,
GAUSSIAN curvature .

This yields
H = K = 0 .

s cot a ,

( 7a )
(7b)

Surfaces with an equal measure of GAUSSIAN curvature K = const can be


mapped isometrically onto each other, i.e. they are developable on each other.
Owing to the fact that K= 0 due to ( 7b ), the circular conical surface can be
developed on the plane, just as is the case with any cylindrical surface.

250

11 General fundamentals of shells

b ) Equilibrium conditions fo r the membrane theory o f a circular conical shell


We proceed from the equations (12J)
N1c, + PP = 0
N<* V

+ P - 0

As an example, the first equilibrium condition ( /9 = 1 ) , i.e.,


n 11| i

+ N21|2 + p1 = 0 ,

(8a)

shall be written in expanded form. The resultant normal forces N 11, N21 are ten
sors of the second order, and their covariant derivatives are to be formed accord
ing to ( 2.35b) :

N i + r l N'1 + r !e N1' + n21,2 + r 2e N'1 + ^

+ P* = 0

( 8b )

In a first step, the CHRISTOFFEL symbols of the surface have to be determined,


using ( 11.23a) :
r /?7 2

( a e/0'7 +1 7 e>P

a fl7,p
)
/
0 7 -e

One thus obtains the following CHRISTOFFEL symbols:

(rie) =

0
0
0 s sin a

( r^ ) =

0
1/s

1/s
0

( 9a,b )

By substituting ( 9a,b ) into ( 8b ) one obtains :


WU . m21 , 1 11
. 2 -kT22 ,1
n
N j + N f2 + -g- N - s sin a N + p = 0 .

( 10)

Finally, the physical components are introduced into (10 ) by ( 2.17) :


N*U s Ngs = Nu

(11a)

(lib)

N*12 ^ NSJ? = N12ssinoc ,


N*22 s Nw = N22 s2 sin2 a

(11c)

From (10 ) and (11 ) now follows


d N
d ( N_, \
i
i
+ 7 i - ( - n i f ^ ) + T N. - T B + p.
d s
a N Si
ds
or

88

sin a

d N 8$
d "d

(\ s N88 /)f8 Hsin


r~0CN8$#$
, , - N 1,71,7+ s p*1

-
= 0

( 12)
= 0

The above equation is identical with equilibrium condition ( 12.16a) where ( ) s s


d / ds and ( ) ^ ^ d/ dO .

Exercise C-12-1

251

Finally, the equilibrium condition ( 12.16c) is checked, i.e.,


N 11 b jj + N 22 b 22 + p = 0

* N 22 s sin a cos a + p = 0 .

Using ( 11c ) it follows that


1
s sin a cos a + p =
N $$ 2 * 2
s^si n^a

N&&

- p s t an a .

E x e r c i s e C-12-1:
A shell of revolution w ith an elliptic m eridional shape (Fig. C-2) is subject
ed to a constant in tern al overpressure p0 .
D eterm ine th e m em brane forces in th e shell.

Fig. C - 2 :

Shell of revolution w ith


elliptical m eridional shape

S o lu tio n :
We take from analytical geometry the radius of curvature Tj for a point P of the
ellipse
1

a 2 b2
( a 2 sin 2 9 + b 2 cos 2 <p) 3^2

and the distance r 2 PN to the axis of revolution


2

Assuming that
N.
vv

( a 2 sin 2 9 + b 2 cos 2 9

) 1//2

= 0 , we obtain according to ( 12.7a)

( a 2 sin2 <p + b 2 cos 2 <p) X^2


a 2 sin 2 9

a 4 ,2
b
2 P0cos
( a 2 sin 2 9 + b 2 cos 2 9 )'

v?= 0
By means of the substitution
2
b2
sin
9 = z , 2 sin 9 cos 9 d 9 = d z ,
- b2
the integral can be transformed into a basic integral.

d9

252

12 Membrane theory of shells

Fig. C-3: Equilibrium at large


However, the above results can be obtained more easily if we consider the equili
brium at large for a thin top section, cut symmetrically from the shell of revolu
tion at arbitrary angles 9 (see Fig. C-3). The vertical load F results from the
pressure acting on the horizontal projection of the shell (circular surface of radius
since the horizontal components of p 0 counterbalance each o th e r):
F = 7t r 2 ( <p) p 0 .
From the eq uilibrium at large follows that
2 n r ( <P)

sin <p = F = 7t r 2 ( <p) p 0 ,

and by assuming that r ( 9 ) = r 2 sin 9 , one obtains the membrane force in the
meridional direction

N
= P2 12
99
and the membrane force in the latitudinal direction by ( 12.7b)
w

_
r2Po

r2 Pp r2 _
/,
Tj 2 ~~P r2'

At the top (/ 9 = 0 \) holds

99

and a t the equator (


Pna
N9 9 = - ^ -

"

with rt = r 2 = a-g2 - th a t

P0__
a2
N

r2 \

2Tj

2b

n ) follows w ith Tj = b 2 , r 2 = a that


=
/
a2 ^
> NW = P 0a ( 1 - f j ^ )

For a > 72 b , i.e. in cases of more shallow shells, a compressive stress occurs in
the circumferential direction at the equator. An elliptic shell bottom reduces its
diameter when subjected to overpressure. In the special case of a spherical shell
with Tj = r 2 = a = b , the boiler formula (12.13) is verified in the form:
n

tfitp

Pna
n
y_
tW
2

A spherical shell subjected to internal overpressure only exhibits tensile stresses.


The same applies for a cylinder.

Exercise C-12-2

253

E x ercise C-12-2:
A spherical boiler ( radius a , w all
thickness t) subjected to in tern al
overpressure p0 is supp o rted in b ear
ings a t its top a n d bottom points (Fig.
C-4). T h e boiler ro tates around th e
v ertical axis A-A w ith a constant an
g u la r velocity w.
D eterm ine th e ro tatio n al speed for
unset of yielding, assum ing th a t only
a m em brane sta te of stress exists an d
th a t th e deadw eight can be neglected.
Fig. C -4: Spherical boiler

N um erical values:
-3

a = 1m

t = 2 10 m

<ry = 360 M P a ( yield s tre ss) , p0 = 0.8 M P a , g = 7.86 kg/ m .

S o lu tio n
Besides the internal overpressure, a centrifugal load occurs in this problem. With
r = a sin 9 , the resulting load components in the meridional and the normal di
rection become:
2 . o
(la)
p = p 0 + ptco a s m 9 ,
p^ = p t o a sin

cos

( l b)

Substitution of ( la,b ) into ( 12.7a) yields

N 99 = - -2
I ( p cos
sin 9 __J

- p sin
*

) sin

d 9 ~ %
sin 9

9=0

I p 0 cos

sin

dip .

9=0

Fig. C - 5 :

Components of the centri


fugal load

254

12 Membrane theory of shells

All terms with to vanish so that the meridional resultant force N


on the internal pressure p0 . After integration we obtain

( _ p c^ 2 9 + c )

N.
vv

(2)

sin2 <p

Since the meridional resultant force N


N

only depends

sin2 <p '

has to be finite for <p = 0 , we get


finite

>

(p = 0

C =

Substitution of C into ( 2 ) yields :


N

vv

(3a)

The resultant forces in the latitudinal direction are calculated by means of equa
tion ( 12.12c) and by superposing the two load cases:
N m > = ^ f - + p t u 2a 2 sin2 <p .

(3b)

The stresses in the latitudinal and meridional direction then become:


* 22 <p
- ^ Poa
r + P 22 a 22 s.n

o v v = Poa
- ^

The maximum stress occurs at iz / 2 . Following the von MISES hypothesis, the
maximum stress can be expressed as follows :
3r = / o 2 +

02

10 2

<

0 j.

(4 )

With ui = 7rn , relation ( 4 ) allows us to calculate the rotational speed n for un30
se t of yield in g :

26.4 rev/sec

Exercise C-12-3

255

E x ercise C-12-3:
C alculate th e m em brane forces in a spherical shell (radius a) subjected to a
w ind pressure described by th e approxim ate distribution
p = - p 0 sin ip cos $ .
T an g en tial frictional forces occur in p ractice but will be neglected here.
side view

top view

Fig. C - 6 : Spherical shell subjected to w ind pressure load


S o lu tio n :
Assuming that p^ = p^ = 0, the equilibrium conditions (12.12) read :
sin <P( Nw ),v + cos <PNw + (
sil><P( Nv* ),9 +

2cos <p

)f<, - cos

9 Nw

+ ( Nw )'#

0,

= 0 ,

(1 )

= - p 0a s i n 9 cosf> .

Nw + N ^
By a product approach according to ( 12.9)
Nw

= ( 9 )cos

N ^

= Y ( 9 )Bin , Nw = 0 ( 9 ) c o s ,

(2)

we transform the system of partial differential equations ( 1) into a system of or


dinary differential equations (
( ) ) :

9 0 ' + cos 9 0 +
sin 9 Y 1+2 cos 9 V - 0

sin

V cos

90

= 0,
=

0 + 0

( 3a )

0,
- p 0a sin 9

(3b)
.

( 3c )

By eliminating from ( 3c )

= - O - p0a sin

9,

we obtain
sin

9 O'

s in 9 + '

2cos 9O +

+ + p 0a sin

9 cos 9 = 0

+ 2 c o s 9 V + O + p 0 a sin 9

,
(4)

256

12 Membrane theory of shells

The form of ( 4 ) suggests introduction of the sum and the difference of the un
known functions as new functions :
F1 = + ' r , F

= - T .

(5 )

If we now divide (4 ) by sintp , (5 ) yields by addition and subtraction, respective


ly, of the two equations (4 )
F l,2 + ^1,2F 1,2 + F l,2 = 0
with

Xj 2 = 2 cot <p ^

where the index

implies

Pj

= p 0 a ( cos tp 1 ) ,

and the index

(7)

implies - .

The ordinary inhomogeneous differential equations of the first order with variable
coefficients ( 6 ) have the following solutions according to ( 1Z.Z7) :
F ..:. =

(8 )

The integrals are evaluated bymeans of ( 7 ) :


J'Xj d(p = ^ ( 2 cot tp -f s-^ ^ ) d <p= 2 In sin tp -f In tan

J1

Ai dw

= e

2 In sin id + In tan w / 2

. 2

- sin

(p

tan

In a similar way we determine


<p

-fAjdv

e J

COt "2

= ---- 5
sin tp

<p

f>2dV

, eJ

-2

. <P

-f>2dv

= sin tp cot , e J
2

tan ~2

= .9
sin 2 (p

For Fj we then obtain


F, =

C, 1

cot
p . a ( cos tp + 1) sin 2 <p tan
d<p .
.
J u
I
sin <p

By means of
1 -f cos tp = 2 cos 2 ^

sin 2 tp = 4 sin 2 ^

co s2

>

the integral can be determined as follows :


J^ co s tp -f l ) sin 2 tp tan

dip = J li cos 3

sin 3 ^ d t p =

<Pd<p =

1
- COS tp
COS3 tp

If we substitute
o

2 V.

(9)

Exercise C-12-3

257

we o b tain fro m ( 9 )

^ .
/
1
i \ 1 1 + cos m
Cl1 + 0Pna ( cos V ~ To cos <P) J----sin
J3-----.
9
an d analogously

F2 = [C2 -

,
1 i
P0 a ( cos<P ~ T COS

1-

cos 9
sin39

'

S ubstitution in to ( 5 ) a n d solving lead s, a fte r intro d u ctio n o f tw o new in teg ra tio n


c o n stan ts D j = C j + C 2 a n d D 2 = C j - C2 , to

= i ( F 1 + F2) =
= y

D j -(- D 2 cos 9 + 2 p Qa c o s 9 ( c o s 9 - y c o s 3 9

Y = - y ( Fj - F 2 ) = y

>

( 10)

D2 + Dj cos 9 + 2 p 0 a ( c o s 9 - y c o s ^ ) ]

In o rd er to ensure finiten ess o f th e re s u lta n t forces a t th e to p ( 9 = 0 ) , we de


m a n d th a t
D j + D2 + 2 p0 a y

( 11)

Since s in 3 9 occurs in th e d enom inator, n o t on ly th e n u m e ra to r b u t a lso its first


a n d second derivative have to v a n ish a t th e p o in t 9 = 0 . W e o b tain , fro m th e
second equation (
) , fo r th e firs t deriv ativ e o f th e te rm in sq u are b rack ets

10

| - D j sin 9 + 2 p Qa ( - sin 9 + c o s 2 9 sin 9 )

9= 0
a n d for th e second derivativ e
/
o
3 \
D j cos 9 -|- 2 p 0 a ( - cos 9 - 2 cos 9 sin 9 + c o s 9 J

= 0

J <p = 0

Whereas the first condition is fulfilled directly for


for 9 = 0 yields :
- D j + 2 p 0a ( - l + 1 ) = 0

9 = 0,

the second derivative

Dj = 0

a n d thus, according to ( 11 ) ,

D2 = - y P o a
W ith ( 1 0 ) a n d ( 2 ) th e follow ing ex p ressio n s for thie
e m e m b ran e forces a re ob
tain ed :
\ COS <P

cos3 v )

258

12 Membrane theory of shells

Fig. C-7: Support of the spherical shell at the ground


The wind load p ( tp , $ ) possesses a resultant F in the x-direction which can be
equilibrated by the resultant of the shear forces
at the cut (p = n / 2 . At
other cuts defined by <p, components of Nvv contribute to the eq u ilib riu m at
large. However, since the shear forces at the two semi-spheres act in the same di
rection and therefore add up, their resulting force has to be provided by the
ground through a stiffening ring (Fig. C-7). Without this or a similar type of
support, the spherical shell would be blown away. Thus, the support disturbes
the membrane state of the shell which can therefore only be considered as an ap
proximation.

E x e r c i s e C -1 2 -4 :
A hanging conical shell ( height h, conical sem i-angle a ) supported as de
picted in Fig. C-8 is filled w ith liquid of m ass density g.
D eterm ine expressions for th e m em brane forces in th e ranges I and II
show n in Fig. C-8. T h e deadw eight of th e shell can be disregarded.

Fig. C-8: H anging conical shell


filled w ith liquid
S o lu tio n :
The loads are axisymmetrical, and can be written as follows for the two ranges :
Range I :

p = 0 ,

pg = 0

Range II:

p = p g z = p g ( h 1 - s cos a ) ,

ps = 0

,
.

(la)
(lb)

Exercise C-12-4

259

The expressions for the membrane forces can be determined by means of the
equilibrium conditions (12.16) for the axisymmetrical load case
^ - ( s N88) = Nw

(2a)

(2b)

= p s tan a

Range I:

N88 =

N* * =

(2c)

In order to determine the constant C , we proceed from the equilibrium at


large at the transition between range I and II. We demand according to Fig. C-9
that
( Nsgcos a ) 2 7rhj t a n a = -|-p g it ( h 2 ta n 2 a ) h j

N.

, 2 sin a

Tb - P g h 1l cosz5--a

(3)
We determine the constant C from the boundary conditions for s = Sj = hj / cos a
with ( 3 ) as follows :
2 sin a
cos2 a

C = l - Psh ;

Substitution into ( 2c ) then yields the following expression for the membrane force
Nss in range I :
N 88 = 46- Pw g6 -

h 31

s3a

sin a

(4)

Range I I : By including ( lb ), we obtain from ( 2b )


N

s p g ( h j - s cos a ) ta n a

and from ( 2a ) after integration


N8 8 = ^ ( V

i a + C ) tan a .

(5a)

Fig. C -9: E q u ilib riu m at large for


range II of the conical shell

260

12 Membrane theory of shells

At the boundary s = Sj we have


2 sin a
-I- C tan a = - j r pgh ^i
1 cos 2 a

C = 0

Thus, we determine the following expression for the membrane force Ngs in range
II:
(5b)
Ngg = ^
( 3 hj tan a - 2 s sin a ) .

E xercise C-12-5:
A section of a casing has th e shape of a circu lar toroidal shell as show n in
Fig. C-10 ( rad iu s of th e circu lar section a, rad iu s from centre point r0 , w all
thickness t) .
At th e boun d ary tp tp0 th e shell is subjected to a uniform ly distributed
boundary load N0 acting in th e tangential direction.

Fig. C-10: Section of a casing w ith toroidal shell shape


a ) D eterm ine th e m em brane forces and th e stresses in th e shell.
b)

S tate th e basic equations for determ ining th e displacem ents u and w for
th e section of th e casing.

Solution :
a)

We proceed from the equilibrium conditions for shells of revolution with arbi

trary contours (12.6) subject to an axisymmetrical loading ( p . = 0 ;


( r Nvv ).v ~ ri cos <P

+ r rj p v = 0 ,

U-V7

= 0):
(la)

Exercise C-12-5

261

With the angle tp relative to the axis of rotational symmetry, the radius of curva
ture rj = a , the distance r = a sin tp + rQ from the centre line, and the auxiliary
radius r2 = a + rQ/ sin tp resulting from the projection onto the centre line, the
following system of equations is obtained:
+

[ (a s in < p

N
a

r j N

^ J ^

rQ + a sin

tp

- a cos

tp N * ,

0,

( 2a )
( 2b )

Differentiation of ( 2a ) and transformation of ( 2b ) yield


Nvv, a cos V + ^ VVlV ( a sin tp + r ) - N ^ a cos tp = 0
N.
(p(p,<p

a cos <p
rQ -f- a sin <p ( N w - N w ) = 0 ,

(3a)

tv ( a sin
a sin tp

(3b)

tp

N..

v>v>

If we substitute ( 3b ) into ( 3a ) , we get


N

a cos

Nt p t p ,t p +

* tp (\ i +
iin

a sin tp

v>tp =

a cos tp

------- (- cot tp N
= 0 .
j
r + a sin tp
T] w
P(<P)

The general solution of the differential equation of type N


reads

i(fi + P ( <p )

N tptp = Ce"^P( v ) dv
Evaluation of the integral leads t o :
JP(<p)d<p =

rp + T s i n y

d<p + J Cot v d<p =

dip + J cot ip dip = In (


-(- sin tp
N

+ sin tp ) + In ( sin <p )

(4)

_ C e -[ln (^ - + sinvj) + ln(8inv )]


<p<fi

or

N tptp = C

Boundary condition :

4-

sin tp
N

sin tp

C * _____
1_________
sin tp (r Q + a sin tp )

( tp = tp0 ) = - N0

(5)

(6)

262

12 Membrane theory of shells


- Nn = C* ;--------i----- 7 ;----- r
0
sm <P0(r 0 + a sin (p0)

From ( 5 ) follows that


'

C* = - s i n >0(r 0 + a sin <^>0) N 0

Thus we obtain
s i n <P0 ( r p +
v v
and

by

s i n <p ( r 0 +

in c lu d in g (

3b )

. V

a s i n ' tp

T h e s tr e s s e s a r e g iv e n

^
0

a v>v>

a s in y 0)
a s i n (p )

( )

by
N* ,
ow = - p -

= _p>

and

(9 )

b ) With axisymmetrical loading and support conditions, we apply the following


strain-displacement relations (12.21) with i / c )0 ( ) ^ = 0 and v = 0:
=

u cos <p + w sin <p

--------------s.
M. = --------------------------r
C , ,

Y^

=0

( 10a )

u cos ip + w sin
r
a sin
<p -f, rQ

v( 10b )
( 10c )

According to (12.26) the constitutive equations read :

ew = t < Nw - v N >

( lla )

ew = i t ( Nw - v N w )

( llb )

Solution of

(10 )

with respect to w yields :

( 10a ) >

w s

a - u

( 10b ) >

w=

( a sin *P + ro ) ~ u cos *P
sin <p

By comparing we obtain
a sin <p - n iV>sin <p =

( a sin <p + rQ) - u cos <p

u v - u c o t < p = sw a - s ( a + -i ^

( 12)

Exercise C-12-5

263

W e now su b s titu te ( 11) in to ( 12 ) a n d get

u v -ucotq. = ^ T[(Nw - v N w ) - ( N w - v N w ) ( +

E t

1X

( a + v a + v ) - N ( a + v a + r-2 ) ]
sin tp y
sin tp / J

vv\

-)] =

(13)
'
'

F in ally , su b stitu tio n o f th e m em b ran e forces ( 7 ) a n d ( 8 ) in to ( 13 ) yields :

Nn [

si n<P0 ( ro + asin(P0) U ( ! + v ) + v - ^

U c o t (p = - = * --------------------------- ;------ 7------------- ;------t---------------------E 11


sin tp ( r Q + a sin 9 )
(14)
sin<P0 ( r o + a s i n y 0 ) [ a ( l + v ) +
a s in 2 tp

The linear, first order differential equation ( 14 ) reads in abbreviated form


u ,*> + P (P)u = Q(<p)
with P ( tp ) = - cot tp and Q( tp ) = right-hand side of ( 14 ) .
With ( 1S.S7) , the general solution is
u(p) = e - / P W d v [ - j Q ( < p ) e / P ^ ) d v d<p + C] .

( 15 )

Calculation of the integrals :

e/p(p)dv> _ e-/cot(pdp _ e~ Insintp _


e -fp(<p)d<p _ e/cotv dvJ

_ ginsintp

sin tp

_ sjn(p

j Q ( < p ) e / P M dv,d<p =
_

N /

" o lo fI a (vl + v ) + v
sin
tp ,
E t J ( rQ
a sin tp) sin 2tp ^

/ fa(l-t-v) +

^ 0*
0I v

sin0 tp
Et J
a sin 2tp
jj

with lQ sin <P0( rQ + a sin (pQ) .


In order to determine the constants of integration , we write the boundary condi
tions at point A
u(<P = f ) =

( 16)

Thus we obtain the meridional displacement u ( tp ) by means of which we can de


terminethe normal displacement w ( tp ) from ( 10a ) . For reasons ofbrevity, the
integrals will not be determined here.

12 Membrane theory of shells

264

E x e r c i s e C -1 2 -6 :
A thin-w alled circu lar cylindrical shell w ith one end clam ped as show n in
Fig. C -ll is subjected to a sinusoidal distrib u tio n of tan gential m em brane
forces at its free end w ith th e shown vertical force F r as resultant.

Fig. C -ll: C ircular cylindrical shell


subjected to an end load
a ) How large a re th e m em brane forces ?
b)

D eterm ine th e vertical displacem ent w of th e bottom point A of the


free end of th e shell.

c)

Check th is displacem ent by m eans of th e first theorem of CASTIGLIA N O .

S o lu tio n :
a ) We assume that the vertical force FR at the free end of the shell stems from
the following sinusoidal distribution (see Fig. C-12):
- Nxj? = k sin R
Then
F*

7t /2

J -

tt /2

Nx# sin S adS = 4 k a | sin'RdR = 4 k a ^ -

(1)

lI s

must hold. From (1 ) follows that k = f ^ / 7ra, and according to (12.14) with
= px = p = 0 we obtain the resultant forces as follows
Nv, v, = 0

F
sin R
N x t ,f = -----7 t
a

Owing to the boundary condition Nxx (x =


nishes. Thus, the final result reads
F
N xtf, = - tt:a sin &

F
Nx x = Ja - ^7 t xa c o s f t + Cl . v( )'

0) =

0 , the constant Cj ( & ) va


=

(2)

Fig. C-12: Relationship between vertical load FR


and tangential membrane forces NXJ?

Exercise C-12-6

265

This corresponds to the solution that would be obtained by the elementary beam
theory. By defining the moment of inertia for a thin-walled circular section as Iy
= 7ra t and the bending moment of the cantilever beam My = - FR x , the nor
mal stress at sections x = const is
Mv

FR*
7 ra3

a cos

x 4" cos $ =
t

ca^

N XX
t

b ) The deformations are calculated by means of the equations of the constitutive


equations (12.26) after substituting the strain-displacement relations of the circu
lar cylindrical shell (12.23) :
(3a)

= E F ( Nxx - v N *>)
Et ( n w - v n

+ W

4- V

+ v ,x

(3b)

(3c)

^ ( 1 + v )
E t

After substituting the resultant forces ( 2 ) into equations ( 3 ), we calculate the


axial displacement u by integrating ( 3a ) , the tangential displacement from ( 3c )
and, finally, by a simple transformation the radial displacement w from ( 3b ). We
then obtain
1
Et

_E
j 2 ^ - c o s S + C2 ( ) ]
7ra

(4 a )

1
Et

_ 2 ( 1 + v ) 's t ; x s in ^ +

^ 3

1
Et

^ J ( 2 + v)xcos - - ^

! T s in ^ - "ir C2,.?+ C3 ( ^ )]

(4b)

(4c)

^ - c o s + f C2 M - C3(1?]

The two arbitrary functions C2 ( & ) and C3 ( & ) only allow the fulfillment of two
boundary conditions, e.g. u ( / ) = v ( i ) = 0 , instead of the fo u r boundary condi
tions for the clamped boundary u ( / ) = v ( / ) = w ( I ) = w x ( /) = 0 . Thus,
( 4a ) yields

u (I) = 0

( 4b ) yields

v (I) = 0

C3( ^ ) = 2 (

1+

(5a)

C2 ( ) = - ^ 7 2 ^ COS&

V) fn ra * sin^ -

c 3(&) = ^ s i n [ ^

Tta*5

2 (1

sin + -j*

7ta

+ v)-J

We then obtain the radial displacement w from ( 4c ) with ( 5a,b )


EtSKr c os (2 + v)--

jl

6a

J L
2 a 2

4*r-sin$
(5b)

12 Membrane theory of shells

266

Since no function is available to fulfill w ( I ) = 0 , this condition cannot be com


plied with. It holds that
w (I) = -

L it

cos =(= 0
a,

Sim ilarly,

( / ) ={= 0
1

The membrane theory cannot meet these essential boundary conditions and there
fore only yields an approximate solution as we have already seen in various exam
ples. In order to fulfill the essential boundary conditions w ( / ) = w x ( / )
= 0, a
bending solution has to be superposed onto the approximate solution.
From ( 5c ) we obtain for the displacement of point A :
w ( x = 0 , = 7t ) = wma x=

2(1 + v ) 4 - ]

(6)

When compared to TIMOSHENKO beam theory, the first term represents the
contribution from bending, and the second term the contribution from shear de
formation.
c)

C om parison by m eans o f the Theorem o f CASTIG LIANO

The displacement of the point of load application can be calculated by meansof


the first theorem of CASTIGLIANO ( 6.27a) as follows:

= d U* ( F-*) = a u ( F j )
dFl
d Fl

(7)
V

Equation ( 7 ) applies to a linearly elastic structure. In the present case, the defor
mation energy according to (12.28b) can be employed. For the circular cylindrical
shell x s <p , s o :
U=

t J [ NXX+ N ^ - 2 I / N XXN ^ + 2 ( 1 + v ) N x% ] d A

(8a)

According to ( 2 ), NM = 0 holds in the present case, i.e. ( 8a ) reduces to :


U = i k t J [ Nxx + 2 ( l + v ) N x2* ]d A

(8b)

A
We then obtain the displacement w by ( 7 ) with ( 8b ) as

x = 0

= 0

N*

2 ( 1 + v ) N -

i5 w

] d 8 'i '

'

<9)

We now substitute into ( 9 ) the resultant forces Nxx and Nxi? from ( 2 ) and their
derivatives:
W= f t

J J
x = 0

[ J t

x2cos2 + 2 ( 1 + v ) J ^ s i n 2 dx d

= 0

After integration we obtain the same result as given in ( 6 )


max

E17T L A
3 a 3- +

2 (1 + v ) 4 - l

Exercise C-12-7

267

E x e r c is e C -1 2 -7 :
A type of shell often found in civil and m echanical engineering is a ruled
shell as shown in Fig. C-13. Its mid-surface has the form of a special hy
perbolical paraboloid which is generated by moving a straight line g along
a rectangle ABCD. The rectangle lies in the x 1, x 2-plane and has the side
lengths lv l2 . The straight line moves along the line AD and along the hy
potenuse of the triangle BEC. This so-called skew hyperbolical paraboloid
shell is also termed a h y p a r shell and its parametric description is given
by
r ( x , y ) = x e j + y e 2+ ^

(x =

, y s f

, c = ^ )

Fig. C -13: Coordinates of a hyper


bolical paraboloid shell
a)

Set up the equilibrium conditions of this shell according to membrane


theory.

b)

Determine the resultant forces and moments for a shell subjected to


the deadweight g per unit surface area, i.e. its physical load compo
nents in the global Cartesian coordinate system x l ( i = 1 , 2 , 3 ) are
given a s :
Pi = P2 =

P3 = S

S olu tion
a) Equilibrium conditions
First, the fundamental quantities of first and second order as well as the CHRISTOFFEL-symbols have to be determined. Proceeding from the given parameter
description
r ( x , y ) = x e 1 + y e 2+ - ^ e

the base vectors are determined as :

al = r,x = ei +
2 r , y ~

e2 +

Te3 c

e 3

( la )
(lb)

268

12 Membrane theory of shells

Similarly, the metric tensors are calculated according to (11.11), the determinant
according to (11.12), and the covariant tensor of curvature according to (11.18):
xy
c2
xy
c2

= M

/( a

1+

(2a)

(f)2

= 1+ (i)! + ( i f

(2b)

xy
c2

1 + (t )2

\ _( aQ
( ) \1 1
)

xy

( 2c )

1 + ( i f

C2

(3)

aP
c Va

y
c2a

(C >

0
WV-

From ( 11.23a), the CHRISTOFFEL-symbols of the second kind result as

( 0

c 2a

c2a

(4)

In order to formulate the equilibrium conditions (12.1)

Nli + n21|2+ p1 - 0
A

+ n212 + P2 = 0
11

19

99

N bn + 2N b 12 + N

(5)

b22 + p = 0 ,

the covariant derivatives of the stress resultants are required. With the relations
( 2.35b) and ( 4 ) they become
11

2y

n 12|1 = n 1 2 1 + - ^ n 22 + - ^ - n 12
'
era
c a

( )

Exercise C-12-7

269

Substitution of the derivatives ( 6 ) into ( 5 ) yields :


,,ll

, , , 2l

12 ,

3y

N ,l + N , 2 + 2 _ N
, T12

, t22

, 3x

N ,1 + N , 2 +

, Tn

+~T~N

,,12 ,

T - N

+p

,,2 2

+ - f N
, , 12

-N

=0

+P

(7)

= .

+ p

The solution of this system of equations requires a transformation into physical


components. Owing to the occuring non-orthogonal surface coordinate system (m e
tric ( 2a ) is fully occupied ), the relations (2.17) cannot be used for determining the
physical components. On the basis of [ C.6 , C.ll ] we therefore define, as physical
components of a stress vector xIJ, the components of the stress vector in the di
rection of the unit vectors that are parallel to the base vectors and that are thus
not perpendicular to a tetrahedron cut plane. We obtain from the equilibrium of
the tetrahedron
g(ii)
(ii)

(8a)

In transition to the shell, ( 8a ) yields the physical components of the membrane


forces
N*^ =

W )

(8b)

Substitution of ( 8b ) into ( 7 ) requires formation of the following derivatives :


d
dx

Z 5 l"
<)y

a 22

,2
JLiL
c4 a / ^ Van a22

d
dx

L l.
c4 a / ^ Van a22

6
dy

c2 a / a
22

"11

I ___
c2 a / i :

By introducing the physical components of the surface loads


P* = V a(aa) p , p* = p ,
and by denoting the physical components of the membrane forces by subscripts
N*11 Nxx

N*22 =

Nyy

N*12 = Nxy

Eqs. ( 7 ) finally yield the equilibrium conditions of the skew hyperbolic parabo
loid shell (d/dx = ,x , d/dy ,y):

270

12 Membrane theory of shells

77 =

2=

r r N + f / r = o , ( * , )
- | - N >y + P* = 0

. (sc)

Eqs. ( 9a,b ) are a system of first order partial differential equations with variable
coefficients. Eq. ( 9c ) yields the membrane shear force
Nxy = - f a p *

By formulating the derivatives ( note ( 2b ) )


Nxy,x = - f p * - f a P * ,

Nxy(y = - | p * - f a p *

and by substituting them together with the metric ( 2a ) into ( 9a,b ) , we obtain
after re-formulation two uncoupled differential equations for the two unknown
membrane forces:

( Nyyi r a n~),y = ~1T


b)

p* + f

p*,y - ^ p * 1 .

V* + f a

P*,x ^

V*2

(ia)

( 10b )

R esu lta n t m em brane forces

First, the physical load components in the global Cartesian coordinate system
x1 ( i = 1 , 2 , 3 ) have to be decomposed into components both in the direction of
the local surface parameters and perpendicular to them.
a 3 is calculated from a j and a 2 (la ,b ) by forming the vector product according
to (11.16):
e l

+ e3 >

(1 1 a)

*2 ^ 2

cfe 3

(11b)

a l

y
a3

X
^

+ e3

( H e )

The above vector equations constitute the transformation between the local base
vectors and the base vectors in the Cartesian coordinate system. The latter vec
tors can be written in abbreviated form as
a,,

= 3 1 .^

Correspondingly, the vector can be written in different bases. The covariant com
ponents of the load vector with (2 3 a ) read, for instance,

With (2 )
t i t i = 81

the transformation coefficients 3^ are determined by inverting ( 3 {. ) ;

Exercise C-12-7

X2

i + TCT

(< ) = !

_2
C

2L
c

c_ 2
V2
2
C

271

( 12)

By substituting the load components we obtain with (2.10) the physical compo
nents of the load vectors:
P* = ~ h

cE

/ a 22
a

(1 3 )

P* = - g -

Substitution of these transformed loads into ( 10a ) yields after re-formulation


\2
( 14)
By integration we obtain

2y

In ( x + -/c2 + x 2 + y 2 ) -f C ( y )

( 15 )

A ^ ( |y

From the boundary condition Nxx ( x , 0 ) = 0 , the integration function C ( y ) fol


lows as
C ( y ) = - In / c 2 + y 2

and thus

(1 6 a)

In-

N xx = 2 6 y

/ 1+ ( t ) =

l + VT -

From ( 10b ) with the boundary condition Nyy ( x , 0 ) = 0 , one analogously ob


tains the membrane force in the y-direction:

Nyy = 2 g x

In ------------------

/A

(16b)

r + / r

Eq. ( 9c ) finally yields the membrane shear force


N

(16c)

TIMOSHENKO [C.24] and other authors have treated the same problem by pro
jecting the forces onto the x,y-plane, and then formulating the equilibrium. Their
results can be transformed, by respective measures ( e.g. Nxx = -/ a n / a 22 N n TIM)
into eq. (16 ). Given the prescribed boundary conditions, the load at the bounda
ries x = 0 and y = 0 only acts via shear. Thus, boundary stiffeners are required, a
fact that leads to incompatibilities between the deformations of the stiffeners and
of the shell boundaries. For this reason, the membrane solution has to be augmen
ted by a solution from bending theory. Further examples are treated in [ C.2 , C.8 ].

272

13 Bending theory of shells of revolution

E x e r c is e C -13-1:
A circular w ater ta n k (ra d iu s a, height h ) h as a linearly varying wall thick
ness ( t 0 = m axim um w all thickness)
t ( x >= y

1" f )

as show n in Fig. C-14.

Fig. C-14: W ater ta n k clam ped at


th e bottom
G iven values :

a = 4.0 m , h = 5.0 m , t Q = 0.35 m , v = 0.3 ,


E = 2.1 105 M P a , e g = 1 104 N / m 3 .

a ) Derive th e differential equation an d th e boundary conditions for th e


circular w ater ta n k by m eans of a variatio n al principle.
b ) D eterm ine th e ra d ia l displacem ent w by a R IT Z approach. For th is pur
pose,

W -M lJ V tF

(*->.

sh all be chosen as coordinate functions for th e approxim ation of w , and


th e calculation sh all be perform ed using a tw o-term approach .
Note: T h e deadw eight of th e ta n k can be disregarded. T h e assum ptions of
th e technical shell theory are valid.
S o lu tio n :
a ) The total potential energy is composed of the deformation energy of the shell
and the potential energy of the external loads (see [C .ll]). With the approxima
tion N P N^ , we obtain the total potential energy expression
11 = |

J ( N^ > +

M^ v ) dA j ( p " va + P w ) dA

( X)

Exercise C-13-1

273

For a cylindrical shell we write in physical components :


II

J"(E xx exx +

e^,j

+ 2 N xij

uxx+

+ 2 M x^ o x#) d A - J ( Px u + p ^ v + p w ) d A

(2)

An axisymmetrical load case is given in the present problem, and the longitudinal
force Nxx vanishes. Thus, ( 2 ) reduces to
11 = 2" J ( N<WeW + Mxxu x x ) dA - J p w d A
A

(3)

With (13.14)

'

(*)

arid
Nw

= D ( l - v2 ) ^ - = E t ( E ) f

( 4a)

( 4b)
From ( 3 ) follows that
n = n(E,w,wfa)=

2t

h/a

J { y [E t(f) + K(

) ] _ Pw J dA . ( 5 )

By ( 5 ) we have determined a variational functional for which we now have to


find an extremum according to (6.34) Therefore, we formulate

We then obtain an EULER differential equation in accordance with (6.35) as a


necessary condition:
( J L )

+ ^ = o

Et5 - " p = 0

"

( 6 )

For a constant wall thickness t follows


,
w . +

Et 2
_ pa4
! T a W~ K
4k 4

as the differential equation of a circular cylindrical boiler (13.16 a ) .


We obtain as boundary conditions
^
8w t
dw,
= const

=0


w = 0
a 2

or

8w c = 0 ,
(
>

( 7a )
y J

274

13 Bending theory of shells of revolution

= const

= 0

= 0

or

Sw

= 0 .

( 7b)

b ) In order to calculate the radial displacement w by means of the RITZ me


thod, we employ the energy expression ( 5 ) . For this purpose, we introduce the
linearly increasing pressure
p = gp h ( 1 and the varying bending stiffness

With dA =

2 ? ia d x =

2 7 t a 2d i

we

o b ta in

h /a

n = 2 7t I -7T E t n ( 1 - T1f= o
- g p h ( l - ^ - 5 ) a 2 w d ^

( 8)
.

The application of the RITZ method (cf. Section 6.7) requires that we choose an
approximation to w with linearly independent coordinate functions in such a way
that the essential, i.e. geometrical, boundary conditions are fulfilled. According to
(6.36) we choose an approximation
N

w* = 2 cn fn ( 0
n= 1

( n = 1, 2 , . . . , N ) ,

(9a)

where the coordinate functions in the problem formulation are given as


(9b)
The coefficients cn are the free, yet unknown coefficients.
The approximation ( 9 ) obviously fulfills the geometrical boundary conditions
( w ( 0 ) = w ^ ( 0 ) = 0 ). In addition, the dynamic boundary conditions are also
satisfied since K ( x = h ) = 0.
Based upon (6.37)
u Cn

= 0

, n = 1 , 2 , . . ,,N ,

we derive a linear system of equations for determination of the coefficients c^


with
u
h/a

h /a

{= o

( 10)

Exercise C-13-1

275

For a two-termed approximation n = 1 , 2 , this system of equations reads:

h/a
n = 1:

ci j K t 1 - i r ^ fi +

r 5 >3f ? . ] d^ +

=0

h/a
+

c2 J

[ Eto (1

h I ) f2fj +

a2 ( 1

h ^ ) f 2, *1,

dt, =

=0

h/a

= g p h a 2 J (1 -

h/a
n = 2:

: 1 J" [ E *0 ( 1 " } f ^ ) fi f2 + -^2'( 1 ~ F ^ ) fl . f2 , ]


f=0

= 0

h/a
= gpha"

J (!-F^f2dt.

= 0

0.5

2 w*
[10' m]

Fig. C-15:

[10 N/m]

[10 N ]

Approximate displacement w *, membrane force


, and bending
moment M* of a cylindrical tank with variable thickness

276

13

Bending theory o f sh ells o f revolution

After integration and solution of the linear system of equations we obtain the fol
lowing coefficients:
_ g p (1 - v 2 ) h5
1008
Cl ~~
E t2
468 + 87 X + X2
_

g p ( l - v 2) h 5

E t2

w ith

2 1 (3 6 - X)
468 + 87 X + X2

X = t 1 - V2 ) ( ^ V )

By ( 9a ) we thus approxim ate the radial displacem ent w* as

a t 36 - x ) ( i - F * >
E t2

Vh

(1 1 )

468 + 87 X + X2

Finally, the curves for


and M *x are calculated by m eans of ( 4a,b ) . Fig. C-15
presents the w*-curve and the approxim ations for the resu ltan t forces of the
numerical example.

E xercise C-13-2:
A reinforcing ring 2 (cross-section b 3 1 , b <K 1) is to b e positioned in the
m iddle of a thin-w alled, long pressure tube 1 m ad e of sheet steel ( radius a,
w all thickness t) . F or th is purpose, th e ring is w arm ed up in such a way
th a t it can be slided into its position on th e unloaded tube ( see Fig. C-16).
At a te m p e ra tu re T 2 = 50 C th e ring ju s t fits th e tube in stress-free contact.
Cooling of th e ring to th e tu b e te m p eratu re of
= 20C leads to shrinking
of th e ring.

F ig . C -16: P ressu rized tu b e w ith sh rin k ed re in fo r cin g rin g

Exercise C-13-2

277

D eterm ine th e re su lta n t q uantities


a n d Mxx in th e tube as well as th e
stresses in th e tu b e a n d th e ring, w hen th e tube is subjected to a constant
in tern al overpressure p .
N um erical values:

a = 1 ra , b = 1 m , t = 1.5 10

-2

p = 1.5 M P a , a T2 = 1.1 - 10~S/ C , v

0.3

E , = E = E = 2.1 10s M P a .

S o lu tio n :
The problem w ill be solved by means of the well-known, so-called M eth o d o f
T h e o r y o f S t r u c tu r e s (Section 13.1.4). For this purpose, we p artitio n the pressu
re tube and the reinforcing ring into three subsystem s ( 0 , 1
and 2 - s y
stem ) according to Fig. C-17. We can now formulate the com patibility conditions:

w() + W(1) + w W

w () +

w ^

(ia)

(2 2 )

(2)

(lb)

xS0) + x ^ + x<2) = x(20) + x(21} + X2

Here, th e subscript denotes tube 1 or ring 2, respectively, (including tube ele


m ent ). The parenthesized superscript refers to the 0
, 1
and 2 -system .
We can now compile the values of deformation for the tube and the ring, where
the m em brane solution follows from ( 1 2 . 1 4 ) , ( 1 Z . 2 3 ) and ( 1 2 . 2 6 ) . The values for
the partitioned tube subjected to the boundary force and boundary moment M are
derived from Fig. 13.2:
50 C

AF=

"0" - s y s te m

Z22

Z33

'1" - s y s te m

3 V Z22
M
M
2" - s y s te m

Fig. C-17: Partitioning o f the pressure tube in single subsystem s

278

13

Bending theory o f shells o f revolution

Tube 1:
( 0)

R a3

(*)

Et

x() = 0

(2) _
1

2K/c3

(l)

R a2

2 K x2

M a2

2K x2

( 2) _ M a
KX

(2a)
( 2b )

Ring 2:
v() = w ( 01) + w(2) = A + P b a " = A +
2
2
+ EA
4E t

v2

(3a)
(l) _ 2 R a 2 _ R a 2
W2 = E A
2 E tb

w<2> = 0
(3b)

xx (0)
2

<(1)
^2

.(2 )

^2

A in ( 3a ) denotes the shrinking measure.


We now substitute ( 2 ) and ( 3 ) into ( 1 ) , and obtain a system of linear equa
tions by m eans of which we can determine the unknown boundary lo a d s :

pa2 _
Et

0 -

Ra3

Ma2

2K x3

2 K x2

Ra2
2K x2

Ma
Kx

A +

Ra2

pa

+ 0

(4a)

= 0

(4b)

R = M
a

(5a)

Eq. ( 4b ) leads to

and ( 4a ) correspondingly to
A -

M =

3 p a

4 Et
ax
2K x2
E tb

(5b)

The shrinking measure A has to be determined by an additional calculation. For


this purpose, we separate ring 2 from tube element 1 according to Fig. C-18 and
in sert the forces acting on the single parts. Then, the following circumferential
strains are determ ined:

JVL
Et

a T2 ( ^ 2 -

Ps ( a + 2 t )
3Et

( 6a )

^1 ) =

a T2

(6b)

Fig. C-18: Free-body-diagram of ring and tube


elem ent

Exercise C-13-2

279

where pB denotes the shrinking pressure and 0 = T 2 - T i the temperature


difference. After the ring has been mounted and cooled to T j , the circumferential
extensions and hence strains in ( 6a,b ) must be equal, i.e.,
(7 )

Substitution of ( 6a,b ) into ( 7 ) yields with


p sa
. a _ n p)
Pna _ -E---P
c
3E t
T2 u
Et

1:
_

3 Et
4

(8 )

a T26

We then calculate the circumferential strain from ( 6a ) with ( 8 ) as


i

and with s
A

4 a T2

, the shrinking measure A is determined as


4 a a T2 0

(9 )

By substituting ( 9 ) into ( 5b ) we obtain the boundary m om ent:


M _ 3

P i
<->
E T + a a T2
aK
E tb
2 k2 K

We are now able to calculate the circumferential membrane force


and the
bending moment Mxx from ( 13.17c) . For this purpose, the membrane solution wp
of a circular cylindrical shell subjected to internal pressure has to be superposed.
The total deformation then reads as follows:
pa
Et

2 /t K

R cos K %+ M ( cos K 1; - sin ^ ) ] e

( 10)

Mn

[N ]

[ 1 0 3N /m ]

Fig. C -19:

K^

Bending moment M xx and circumferential force


pressure tube

in the

13 Bending theory of shells of revolution

280

Now we replace R by ( 5 a ) and substitute it into the relation for the circumferen
tial force Nm :
w = pa - ^

M ( cos K i; +sin K 1; ) e- K^

( 11a )

From (13.17c) we obtain for the bending moment Mxx with R =


opposite to the assumed direction of R )
Mxx = - M ( cos /cl; sin /c i; ) e~ K^

M (here

( lib )

Fig. C-19 depicts the curves of the resultant moments and forces for the given
numerical values.
Finally, we calculate the stresses in the tube and the ring :
- Tube 1
Longitudinal stress :

oxx = ^

xx + P ^

The second term in ( 12a ) only applies for a tubeclosed


this case an additional longitudinal load occurs.

( 12a )
at both ends, since in

Equation ( lib ) substituted into ( 12a ) yields the maximum stress


max ~

- 6 M xx( f ; = 0 )
t2

pa
+ 2t

6M , pa
t2 + 2t

Circumferential stress :
6 MXX1
= j v ^2

pa
=

2/c2

I T

M ( cos K

+ sm

, - Kt
K 5 )e

v ^^2 ( cosK _ si n/ c l ; ) e _K^


From
d<3*>
d?

1 = 0

----- I - 0.35

we obtain
= 0 35 ) !, mlx
Numerical values:

o vv
m 156 M P a ,
max
o,,
v% ax rs 102 M Pa .

- Ring 2
Circumferential stress :
~ Pa , P a . 2 R a _ E g T2 e
^ 2 ~ 3 t + 4 t + 4 b t
4
Numerical value :

Ow2 82 M Pa

pa
41

kM
bt

Exercise C-13-3

281

E xercise C-13-3:
A pressure boiler m ad e of steel
consists of a circu lar cylindrical
shell (ra d iu s a, w all thickness
t ) closed a t each end by two
sem i-spherical shells (Fig. C-20).
T he boiler is subjected to a
constant in te rn a l overpressure p
(th e deadw eight of th e boiler
can be neglected).
D eterm ine th e curves for th e
stress resu ltan ts both in th e cy
lindrical shell a n d th e semispherical shells.

N um erical v a lu e s :

p = 10 M P a

Fig. C -20: P ressure boiler


a = 2 m

, t = 0.1 m

1
E = 2.1 10 M P a , v = y

S o lu tio n :
Owing to the symmetry we only consider one half of the pressure boiler. As in
the previous exercise C-13-2, we partition the spherical shell from the cylindrical
shell and mark the single loads according to the 0 , 1
and 2 -system s in
Fig. C-21.
(0)

(0 )

r m

"2" - s y s t e m

Fig. C-21: Partitioning of the pressure boiler in subsystems

282

13 Bending theory of shells of revolution

Here, the compatibility conditions for displacements and rotations at the interface
become:
<> + > 4. W!> = <?> +
4 " + 4

(la)

4 + X<> + *?>

(lb)

- 4 > + 4 + 4

We substitute the single deformation values for the boundary loads (see Fig.
C-21); the deformations of the 0 -system are membrane solutions of the cylin
drical and the spherical shell:
pa2 ,
2Et ^

R a 3 , M a2
pa2 ,
2Kx3
2K/c2 _ 2 E t

,
R a3 , M a2
/
) + 2 K k3 + 2 K k2 *

. R a2
Ma
, R a2 , M a
0 + j K ^ - K ^ = 0 + ^ s
+ in r

x
*

/
( 2b)

Eq. ( 2b ) immediately yields


M = 0

( 3a )

Owing to the fact that the semi-sphere and the cylindrical shell exhibit the same
deformation behaviour at their boundaries when subjected to boundary forces,
and because no twisting angle x of the boundaries occurs subject to internal
compression, the compatibility of the deformations can be introduced by the
transverse boundary forces alone. Eq. ( 2a ) then leads t o :
R = - W

'

( 3b)

The curves for the resultant forces as a function of \ x / a can be determined by


means of the relations (13.17) :
NW = p a ( l - j e 't f c o s K^ )

Mxx = f t t * e~ K ( s i n K t

Qx

Nxx =

e_K f ( cos K \ - sin

I) ,

We then calculate the resultant forces in the semi-spheres by means of (13.18) :


= ^2 " ( 1 + 2"e KU1 c o s KOj ) ,
- P
l
Qv, =
| ^a e. Kltwl(
c o s K u 1 - sinKtij) ,

= 8k2 6

1 Sm K Wl

~ ^2

Fig. C-22 shows the behaviour of the stress resultants around the transition
between the cylindrical and the semi-spherical shell.

Exercise C-13-4

283

m em brane solution *N00 [1 O N /m ]

x[m]
N<p<p lN H [ l O N / mm]
10000

0.1

0.2

Fig. C-22: Resultant forces and moments in cylindrical and semi-spherical shell

E xercise C-13-4:
A thin-w alled circular cylindrical tu b e m ad e of steel (radius a, w all thick
ness t) as shown in Fig. C-23 is horizontally su p p orted betw een two rigid
walls in such a way th a t th e cross-sections a t b o th ends of th e tu b e are
com pletely clam ped.
D eterm ine th e stresses in th e tube due to its specific deadw eight g g , after
rem oval of th e m ounting equipm ent w hich ensures an in itial stress-free
state of th e tube. Use th e following num erical values :
I = 10 m

E = 2.1 10* M P a ,

a = l m
v = 0.3 ,

= 1 1 0 2m

g g = 8 104 N /r n

284

13 Bending theory of shells of revolution

Fig. C-23: C ircular cylindrical tube clam ped horizontally a t both ends
S o lu tio n :
The complete solution is determined by superposition of a membrane solution
(Ch. 12) and the solution of the boundary disturbance problem (Ch. 13).
M embrane solution (denoted by superscript 0)
Using the abbreviated notation y for the
deadweight p g t per unit area of the mid
surface, the following surface loads are act
ing on the shell (see Fig. C-24).
Px = 0

ptf = y sin & ,


p

Y cos 8 .

Fig. C-24: Components of the dead


weight within the shell

We obtain the following resultant forces by substituting the loads into the equili
brium conditions ( 12.14) and by defining 1; =
(la)

N - . = - y a cos 9 ,
N x i?

- ( 2 Y a ( + Dj ) sin $ ,

NXX = ( f a ^2 + Di + D2 ) cosd

(lb)
( 1c)

Based on (12.23) and (12.26), we write


V

= | f ( nxx- v N )

V* + W

= t ( NW

u I ,.
U^ + v-f

... 2 ( l + v ) a
m

v N xx)
XJ?

(2a)
(2b)
(2 c )

Exercise C-13-4

285

Substituting (1 ) into ( 2 ) and integrating, we obtain the membrane displacements


r

u = t [ Y a ( T

2
+

D i T

D2?

+ Ds]0" *

( 3a)

= t [ ^ a (&4 - T ( 4 + 3 v ) ) + Di ( T - 2 ( 1 + v ) 0 + D2 T +
+ D3 I + D4 j sin 9

o
w

a
Et

[ Y

1 )

(3b)

(3c)
+ D2 (

] cos

The integration constants D; ( i 1 , . . . , 4 ) can only be determined from the


complete solution of the problem.
B ending solution ( denoted by superscript 1 )
Since the membrane solution depends on the circumferential coordinate 8 via cos 8
or sin 8 , respectively, the bending solution of a shell clamped at its boundaries
possesses terms with m = 1 only. The eigenvalue equation thus reduces to the fol
lowing characteristic equation dealt with in detail in [ ET2 | 11.3.2 ]:
X8 - 2 ( 2 - v ) X 6 +

=0

(4 )

with the shell parameter k defined by ( 13.31d) .


The characteristic equation has the roots
^1,2 = 2 - v ^ / ( 2 - V )2 - L ^ ~ = 2 - V i ^ ^ Since 1

- (2 -v f

( 2 - \j )2, these roots may be approximated by

xi,2 ,3, 4 = 14 >14 with ^i = y i | A k ~

(5)

The characteristic equation ( 4 ) has four additional eigenvalues X5 e 7,8 =


The corresponding solutions are already included in the membrane solution ( 3 ) ,
and therefore they do not need to be considered in the homogeneous solution.
The shell shall have a sufficient length so that no mutualinfluence of the boun
dary disturbances occurs. We therefore exclusively considerthe boundary i; = 0,
and by including ( 5 ) we obtain the following homogeneous solution:
u1 = ( A 1el '*if + Aa e " 1,,lf) e " f c0s

v1 = ( B 1>'i + B2e i |*C) e " '*<s in ,


w1 = ( C j e ^ i f + C2 e i '1i f ) e " 'ii f c0s

(6 )
.

286

13 Bending theory of shells of revolution

The complex constants A j, B j, Cj ( j = 1 , 2 ) are coupled to each other via a ho


mogeneous system of equations. The first equation ( for m = 1, k 1) yields

(x - V

) AJ +

+ XJCJ = 0

+ ( - 1 r Lx ) Bi +

cj

= 0

and for j = 1 with Xj = - Pj + i Pj, we obtain the following dependencies of the


constants:
Aj =

1
4p*

1
Bi = 4p2
Since X2 = -^ 1

- 1 + 2 v pj + i ( 1 -(- 2 v pj )

Cj = (

+ i *2 ) Cj ,
(7 )

^ + i(2 + v ) j C j

= ( p i + i p 2 ) Cj

P j, the conjugate complex relations for j = 2 foil

A2 = ( 0Cj - i a 2 ) C2

B2 = ( P j - i P2 ) C2

(8)

If we substitute ( 7 ) and ( 8 ) into ( 6 ), all displacements depend on Cj and C2


only.
Boundary conditions
If we consider the boundary 5 0 only in the case of the membrane solution,
the two boundary conditions for 5 = 0 and 5 = 11 a have to be replaced by two
symmetry conditions for 5 = I / 2 a . We thus obtain from

N * ( i ) = 0

und

u ( 2 i ) = 0

with ( lb ) and ( 3a )

The remaining four constants C j, C2 , D2 and D4 result from the four boundary
conditions
u(0)

= u ( 0 ) + u1 ( 0 )

= 0

v (0)

= v ( 0 ) + v1( 0 )

=0

w(0)

= w ( 0 ) + w1( 0 )

=0

w, e ( )

= w ,e ( 0 ) + w ^ ( )

After carrying out the numerical calculation with the given values we obtain the
circumferential force as
= - 8 + ( 47.3 cos 12.91 + 5.02 sin 12.91 ) e 12 9f cos A .

Exercise C-13-4

287

Fig. C-25 shows the membrane forces according to ( 1 ) and the bending moments
Mxx und
acting along the top longitudinal line 9 - 0 of the shell. One can
see how fast the bending disturbance has decayed already at a distance of ~ 0.4
m from the boundary. The stresses are calculated from
Nxx , 6 Mxx
x*=

~ J 2 ~

~T~

( 9)

T 5-

'T '

-4N
[N /m ]

Fig. C-25: Membrane forces and bending moments along the top longitudinal line
of the cylindrical tube under deadweight

288

13 Bending theory of shells of revolution

The maximum stresses at the boundary are due to ( 9 )


o

xxmax= 1.31 ,x
v 0.92 = 2.23 MPa.

llmax= 0.39 *>0.28 = 0.67 M P a

o ,,

,
.

The numerical values show that both the longitudinal and the circumferential
stresses due to the boundary disturbances are of similar magnitude as the mem
brane stresses.

E x e r c i s e C -1 3 -5 :
w :

A circu lar cylindrical shell ( a , I = 4 a ,


t a / 400) is subjected to a constant
ex tern al pressure p (Fig. C-26).

'

F orm ulate th e basic equation for shell


buckling in analogy w ith th e basic
equation
of
plate
buckling
( see
(10.17)).

_ _x_
a

D eterm ine th e n th e critical load for


th e special case of a shell which is
sim ply supported at b o th ends.
Fig. C-26: C ircular cylindrical shell
under external pressure
S o lu tio n :
We proceed from the simplified basic equations for a shear-rigid shell ( DON
NELL's theory). Before buckling the initial stress state prevails within the shell
^xx = 0

^yy

P a

^xy ~

f )

At buckling the component Nyy w;M must be included in the equilibrium condi
tion in the radial direction of the deformed shell . Then, u and v can be eliminat
ed, and we obtain in analogy with (13.39)
k A A A A w + ( 1 - v2 ) w

+ - ^ A A w |W = 0

(2)

as the basic equation o f shell buckling u n d er ex te rn a l pressure.


In order to determine the critical load, we put the coordinate
the cylinder. The approximation
w = W cos mTl:.a ^ cos -52!
I
3<

(m,n = 1,2,3...)

x in the centre
(3)

of

Exercise C-13-5

289

fulfills the boundary conditions of the simply supported shell in the longitudinal
direction
(

27

) =

' (

5 7

) =

and the condition of periodicity in the circumferential direction


w(2ita) = w(o)

(4)

By substituting ( 3 ) into ( 2 ) and by defining X = .m y a , we obtain the relation


for the critical load as
. . 2)X
^4
^ ( X 2 + n2 ) n 2 = k ( X 2 + n2 )4 + ( l - v2
or with

P = TT

P =

(5a)

+2n ^ + ( 1 v2) J 2 ( \ 2 . 2 \2
n
n (X + n )

(5b)

We now have to determine that combination of m and n for which p has the
smallest value. We can immediately see from ( 5b ) that X will attain its smallest
value for m = 1. The shell therefore buckles with one wave in the longitudinal
direction.
Assuming that

a =

>
2

we obtain

p = k n 2( 1 + a ) 2 + (1 - v2)n (l + a)
1-

kX4

(1+oQ2

( 6a)

(l+)2

Assuming that many waves occur in the circumferential direction ( n 1 ) , then it


is valid for long shells / a that
a 1 ,
and ( 6a ) then reduces to
_

l - v 2f

kX4

1 ,

3")

+ )

(6b)

The minimum value follows by differentiating p


dp
da

1X2

kX4
= 0
1 - v: - ( - A ) + 3 2

a* = X 4

3 (1 v2)

( 7)

Substituting ( 7 ) into ( 6b ) yields after elementary re-transformations:

p = !(*) = 7 T ? f f /d - ) (1

(8a)

or with v = 0.3
pcri t 0 . 9 2 E f ( ) S/2

(8 b )

290

13 Bending theory of shells of revolution

Eq. ( 7 ) delivers the corresponding number of waves as

* = = x v ^ i 5 y

" ^ t i / F

Using the given numerical values I 4 a and t = a / 400, we obtain


n 2 7 .3 ^ /4 0 0 = 36.5

n = 6.04 and p crit 7.19 E 10 8 M P a .

Owing to the necessary periodicity in the circumferential direction, the following


adjacent integer number at buckling is n = 6 ; in the numerical example follows
ot* ~ 0.017 and therefore : 1.

E xercise C-13-6:
D eterm ine th e eigenfrequencies of th e free vibrations of a circular cylindri
cal shell w ith sim ply supported ends as shown in Fig. C-26 (w ith o u t exter
n a l pressure p ) . A ssum e sm all vibration am plitudes (lin e a r th eo ry ) and
solve th e exercise using DONNELL s theory.
Note: T h e coordinate system has in th is case been m oved to th e lower
boundary.
S o lu tio n :
In order to treat the circular cylindrical shell, the basic equations (13.31) are sim
plified in accordance with DONNELL's theory (see (13.38)):
u. + J ^
^

u.*> + i T

+ v, M + ^ v ^

+ v w .f =

(la)

+ w

(ib)
2

Vu f + V + w + k A A w

( lc )

As loadings we write D'ALEMBERT's inertia forces:


. d2u
Px = - P t ^

t d2 v
Py = - e ^ r

A d2 w

f
{2)

where p denotes the mass density and r the time ( r is introduced in order to
avoid confusion with the wall-thickness t). This approximate theory neglects the
rotational inertia of the shell; its consideration would result in additional, very
high frequencies, whereas its influence on the lower frequencies considered here is
negligible ( see [ C.21, C.22, C.23 ]).
The eigenfrequencies are determined via separation of variables:
u=u(x,<p)sin(Jt

v = v ( x , <p ) sin u r ,
w = w ( x , <p ) sin g>r .

(3)

Exercise C-13-6

291

Substitution of ( 3 ) into (1 a , b , c ) and ( 2 ) yields elimination of time, and we


obtain the equations
Xu +

u +

~~2 ~

v i( + v w { = 0 ,
v.a + v +

vu +

v e- Xw 4-k A A w

= 0 ,

(4)

= 0

with the frequency parameter


.

(B)

In ( 4 ) , n 7 V, W are
position functionsu ( x , 9 ) , etc. The following assumption
regarding the form of these functions
U cos n

tt

cos m

x
V V sin n cp sin m a
r~
mx
w = tit
W cos n 9 sm

(6 )
,i

m
tu
a
with m
I
fulfills all_ boundary conditions for a shell with simply supported ends
( w = v = Mx = Nx = 0 for x = 0 and x = l ) . Substitution into ( 4 ) yields a ho
mogeneous system of equations for the unknown amplitudes U, V, W. By setting
the determinant of the coefficients equal to zero, one obtains the characteristic
equation of an eigenvalue problem
X3 - AX2 + BX - C = 0 ,

( 7)

where the ^efficients A, B, and C depend on the dimensions of the shell as well
as on m ^ffd n .
Numerical evaluation of ( 7 ) shows that each pair of values of m and n defines
one lower and two higher eigenvalues which exceed the lower ones by powers of
ten. The technically relevant lower eigenvalue can thus be approximated from ( 7 ):
xr = -

<)

The numerical values additionally show that Xj is associated with a pronounced


transverse vibration ( W U , V ), while longitudinal vibrations are associated
with X2 and X3 ( U,V W ). Based upon this observation, the lowest frequency
can be governed by a single formula.
If the amplitudes U and V are very small, the terms of
inertia forcesinthe lon
gitudinal and circumferential direction can be neglected inequation (4). Conse
quently, the displacements u and v can be eliminated from the first two equations,
using the same procedure as in (13.38). Thereby, (13.39) is augmented by the winertia term, and we obtain considering the vibration approach
k A A A Aw + (1 - v2 ) w IV - X A A w = 0

(9 )

With ( 6 ) this yields an approximation for the lowest eigenfrequency:


Xl = ( l ~ v 2 ) ( m 2 + 1 2) 2 + k ( a 2 + n2) 2
The results according to ( 8 ) and ( 10 ) are numerically almost identical.

( 10)

292

13 Bending theory of shells of revolution

The eigenfrequency equation (1 0 ) consists of two termes, where the first stems
from extensional vibrations and the second from bending vibrations. Fig. C-27
presents a numerical example, where both terms are drawn separately in depen
dence on n . The curves clearly illustrate that for a small number n of circumfe
rential waves extensional vibrations predominantly occur, and for large n ben
ding vibrations, respectively. Close to the minimum, the two terms are approxi
mately equal. Therefore, no further simplifications must be performed for the
simply supported shell. Elimination of the first term in ( 9 ) , for instance, cannot
be admitted since this would correspond to an inextensional vibration. If defor
mations which are incompatible with the assumption of inextensionalbending are
prescribed at the boundary, bending and extension will act jointly, and thus have
to be considered by a complete shell theory.
The minimum of Xj can be calculated by a formula. Eq. (10 ) implies that Xj in
creases with m . It attains its smallest value for m = 1, i.e. the shell vibrates
with one wave in the longitudinal direction. Xj then only depends on n . If the
actually discrete number of waves n is assumed to be continuous, (1 0 ) can be
differentiated with respect to n :
- 7 :; 2dXl 2, 2 = - ( 1 - v 2 ) 7 - # - - 2T4 + k = 0 .
d(m + n )
v
( m 2 + n 2) 4
It follows that
( m 2 + n 2) 2 = m 2 - | / ( l

V )

10

(11)

11

12

13

Fig. C-27: Lowest eigenfrequency of a simply supported cylindrical shell

Exercise C-14-1

293

Substitution yields (the two terms in (10 ) result in the same value)
X,

J M in

= 2 /( 1 - v 2 ) k m 2
' V

'

or, with the frequency parameter

( t )s

< >

This proves that the lowest frequency depends on all dimensions and on the ma
terial data. From ( 11) we determine the number of waves n assigned to the mini
mum. In practice, the adjacent integer value of n would occur. In our example,
the shell vibrates with nine circumferential waves; using the given numerical
values, ( 11) would give the value n = 9.22.
For a shell with free boundaries, inextensional bending may be assumed as a pos
sible vibration mode. For this case, Lord RAYLEIGH determined, by equalling ki
netic and elastic energy, that

nz + 1
We obtain the same result by defining m > 0 in (10 ) . Here, the differences in
dependence on n are a result of the DONNELL simplifications. For larger n ,
these differences are unimportant (X = k n 4) .

E xercise C-14-1:
A spherical cap ( Fig. C-28) is extended over a circular base (radius r0 ,
r 0 a), and is assum ed to be subjected to a co n stan t surface pressure
load p. T h e height of th e cap is given as f.

Fig. C-28: Spherical cap over a circular base

294

14 Theory of shallow shells

As r 0

a , th e m id-surface of th e spherical cap can b e approxim ated by

im plying constant curvatures everyw here, i.e.

= Ky ~

a)

State th e differential equation and obtain th e solution of th e hom o


geneous equation for an axisym m etrical problem.

b)

C alculate th e deflection of th e shallow spherical cap w hen subjected to


a concentrated force F a t th e top poin t and assum ing th a t r > 0 .

S o lu tio n :
a)

With the given assumptions, the system of differential equations (14-9) reads
= K

<2>)

AA -
a Aw =0

(v2 b )7

We now multiply ( 2b ) by a factor X and add it to ( 2 a ). We thus obtain


A A ( w + X) - X a ( w - - ) = ^
v

X E tK

'

.
K

(3)
v

'

If the underscored terms in ( 3 ) are equal, one can formulate a differential equa
tion for
F = w + X

(4 )

Introducing i as the imaginary unit, we write


X = ~ i/l2(l - v2 )
Et2

( 5a )

With the abbreviation k we obtain


Et
1T X = lV~ M

1 - v2 ) = ik'

( 5b)

By ( 5b ) and ( 4 ) , ( 3 ) transforms into

(6)

A A F - ik 2 A F - p
K

Here, our considerations will be restricted to the homogeneous solution of ( 6 ).


Then, the differential equation can be split as follows:
(A - i k 2 ) A F = A ( A - i k 2 ) F = 0

(7 )

In the present case it is sensible to use polar coordinates owing to the axisymmetry of shell and load. The LAPLACE-operator is then independent of the
angular coordinate S and hence

Exercise C-14-1

& = 4dr2+ F ' 7 dr-

295

We can thus determine partial solutions from the two differential equations:
AF

= 0

( 8a)

A F - i k2 F = 0

( 8b )

The solution to ( 8a ) can be stated immediately as


F1 = C1 + C2 l n r

(9a)

while ( 8b ) is a BESSEL differential equation [ B.3 ] of the form


l + l

_ ik2 F = 0

( 10)

Solutions to ( 10) are modified cylinder functions


of first and second
type,
I0 ( k r -/T ) and K0 ( k r -/T ) , respectively, that are linearly independent [ B.3 ]:
F 2 = C3 I0 ( k r / i ) + C4 K0 ( k r / T )

(9b)

where C3 and C4 are complex constants.


According to KELVIN, two new functions b e r ( k r ) and b e i ( k r ) can be
introduced that correspond to the real and the imaginary part of I0( k r -/T ),
respectively, as well as the functionsker ( k r )and kei ( k r ) which are equal to
the real and imaginary part of K0 ( k r -/T ) [ B .3]:
I0 ( k r - / T ) = b e r ( k r ) + i b e i ( k r )

,
(11)

K0 ( k r -/T ) = ker ( k r ) + i kei ( k


The reader is referred to standard tables, e.g. [B.3], for graphs of the KELVIN
functions.
The general solution to ( 6 ) then consists of a linear combination of F} according
to ( 9a ) and of F 2 according to ( 9b ). If we substitute the solution into ( 4 ) and
compare the coefficients, considering the complex constants, we obtain from (11)
the following terms
forthe bending w and for AIRY's stress function :
w = Bj ber

(kr)

+ B2 bei ( k r )

B3 ker ( k r ) +

+ B4 kei ( k r ) + Bs + Bg In r ,
2

<D = -------M

f- Bj bei ( k r )

( 12a )

+ B2 ber ( k r ) - B3 kei ( k r ) +

- *2 )
+ B4 ker( k r ) + B? + Bg l n r j

where B j. .. Bg are real constants.

( 12b )

296

14

Theory o f shallow shells

b) In the following, the deformation in the middle of th e shallow spherical cap


shall be considered. For this purpose it is assum ed th a t th e boundary of the
shell is very rem ote from the top point ( r > oo) , and th a t the displacem ent w
and its higher derivatives vanish a t the boundary.
Under the given assum ptions we write for r * oo ( d / d r ( ) ):
w = w ,1 = w ,TT 0

(' 1 3 a)>

In addition, for r = 0 , i.e. a t the top


w , w r , Nn , N

have to

point
be fin ite .

( 13b )

T he concentrated force F a t the top point ( r = 0 ) is equilibrated by a to tal verti


cal shear force V2 along any circle of radius r . Thus,
F
V = ---z
2 7t r

where

(1 4 )

V = Q + N
z

rr

A fter evaluation of all conditions, we obtain the constants as follow s:


Bj = B2 = B3 B4 = B5 = Bg = B7 = 0 , B4 = B g ,

Fa

/l2 (l-v 2)

2 7t

E t2

T he deflection function then reads


_ Fa

/ l 2 ( l - v2 )

2 it

kei ( k r )

( 15a )

E t2

The m axim um deflection occurs at the top point where the load acts. For r = 0
we have k e i ( o ) = - 7t / 4. T his y ie ld s:
wmax = - l / 3 ( l - v 2 )

(1 5 b )

In his fundam ental papers, REISSNER has treated problem s of shallow spherical
shells with a num ber of load cases. F or further details refer to [ C.20 ].

E xercise C-14-2:
T h e eigenfrequencies of a hyp ar shell projected against a rectan g u lar base
(Fig. C -29) sh all be determ ined. T h e distance f betw een base and shell is
assum ed to be sm all.
a)

Set up th e fu ndam ental equations for th e eigenfrequencies.

b)

W hich eigenfrequencies appear for th e special case of a sim ply sup


p orted sh ell? Derive an approxim ate form ula for th e lowest frequency.

Exercise C-14-2

297

Fig. C-29: H ypar shell against a rectan g u lar base


S o lu tio n :
a) For f a , b we can apply the fundamental equations from the theory of
shallow shells. In this example, we replace the external loads in the equilibrium
conditions (14-3) by D'ALEMBERT'S forces of inertia (density p , time t ) .
Neglecting the rotational inertia, we obtain
N ^ ll a

= V
p tf i t^2

=0

Here, the only difference relative to the equilibrium conditions for the shallow cy
lindrical shell (13.36) is that instead of the circumferential force a component of
the shear force occurs perpendicular to the shell.
The kinematic relations are obtained from (14-4)- By transforming the first equa
tion of ( 14-4) by means of ( 14-2) we write
a /J = i ( Val/J +

vp\-

2 z la/Jw )

P ap

= - WLp

( 2 )

The relations between the resultant forces, moments, and the strains are described
by the material law (14-5)
N^ D H ^ a 7,

m^

= k h W p 74

(3 )

We assume that the boundaries of the hypar shell belong to the linear generatri
ces. The mid-surface can then be described in Cartesian coordinates as

Z=S ( X' f ) ( y - | )

(4)

When expressed in Cartesian coordinates, all covariant derivatives simply become


partial derivatives. By introducing the dimensionless coordinates 1; = x / a and
T) = y / b , and denoting the derivatives by

298

14 Theory of shallow shells

we obtain with

z|xx = z|yy = 0

2f
z|xy - a b

the fundamental equations ( 1 ) through ( 3 ) as

T <3 .t + r cU

+ i b N*j = t , dt2
f

>

. -

Q >

(5 )

= -r- v
yy
b 'V

Ex x = T U , f

xy

(6 )
P

pxx = - i w.
^xy

^yy

b 2 ,fw

a b Wf{r}

^xx

^ ( Ex x

v Ey y )

N xy =

D ( ! -

M xx =

K (P x x +

M yx =

K ( 1 - v ) Pxy

v ) x y

* ^yy

^ ^ Ey y

v Ex x )

v Pyy)

> M yy =

K ( P yy +

v Pxx )

Here, the physical components of the originally tensorial quantities are denoted by
the usual indices x , y. The displacements of the mid-surface of the shell are de
noted by u and v .
The kinematic relations ( 6 ) differ from the kinematic relations of the shallow cy
lindrical shell (see (13.37)) by two terms only:
1)

In the case of the cylindrical shell the radial displacement is a part of the
circumferential strain syy , and

2 ) in the case of the hypar shell the deflection w contributes to the shear strain

Exercise C-14-2

299

By eliminating the transverse forces and by substituting ( 6 ) and ( 7 ) into ( 5 ),


we obtain the following three partial differential equations for the three displace
ments from the 17 equations (8 resultant forces and moments, 6 strain quantities,
3 displacements ) :
, 1 4 - v

1 v

U- + ~^ 2 ~ V.(r, +
a v

l 1 + Vn

tab d u
a (
.f
'I
p ta
b \ ur;77 - 4 a W, J = D dx2

I i-vbf,.

..

^ _ ptab

(8)
h2 ( b 2

,)
+ 2 ( 1 - v )-4r- ( a u + b v c - 4 f w ) = p t * b ^
ab
")
D dt
b) The following boundary conditions have to be fulfilled for a shell with all sides
simply supported:
u = w = 0

Nx y

Mx = 0

for

5= 0

and

5= 1

v = w = 0

Nxy = My = 0

for

T) = 0

and

r) = 1

(9 )

These conditions are satisfied if we assume displacement functions u , v , w of the


form
u = U sin m7t5 cos nTtr] sin o t ,

( 10)

v = V cos m7t5 sin m i] sin o t ,


w = W sin m7t5 sin nTtr] sin u t

( m , n integer )

Substitution of (10 ) into ( 8 ) then yields a homogeneous, linear algebraic system


of equations for the unknown amplitudes U, V, and W . Vanishing of the determi
nant of the coefficients leads to a cubic equation for the eigenvalue
2

ptab

where the solutions depend on the dimensions and the integers m and n . Numeri
cal evaluation shows that there exist one lower and two substantially higher ei
genvalues for each pair of values of m and n . The numerical values clearly show
that the lowest frequency corresponds to pronounced transverse vibration ( W
U , V ) , and we therefore obtain a good approximation to the smallest eigenvalue
provided that the terms of inertia forces tangential to the mid-surface of the shell
are neglected in ( 8 ) :
p t d ju

dx2

pt

d V
dx2

After substitution of (10 ) into ( 8 ), we can determine the amplitude ratios U / W


and V / W by means of the first two equations ( 8 ). Substitution of the ratios in
to the third equation of (8 ) finally allows us to approximate analytically the
smallest eigenvalue as

300

14 Theory of shallow shells

where a = a / b denotes the side aspect ratio. Multiplying (11) by a , and intro>tar
during the scaled frequency o 2 =
<j2 then yields:
D
=

+ v

>2

- 2 ) ( ^ ) 8(

<1 2 >

We now need to determine that combination of m and n for the given dimen
sions which provides the smallest values of tJmn .
If we, at this point, limit our considerations to values a < 1 ( all solutions for
a > 1 can be obtained by suitably exchanging the sides) , we can deduce from
( 12) that n always has to take the value 1. Calculations show that m > 2 is
always valid for the lowest frequencies. We can therefore approximately assume
m 2 oc2n 2 , and we thus obtain from (12 )
+

(1 3 )

In order to determine the dependence on m of the smallest value of m l , we as


sume the number of waves to vary continuously and differentiate:
d mi

mi n

Substitution into (13 ) and re-formulation yield:


Simi = [ 12^ 2 ( l - v 2 ) i - ( i ) 2 a 4] 1/

(1 4 )

Thus we have obtained the desired approximate formula for the lowest frequency
in dependence on the dimensions.

D Structural optimization
- C h a p te r 15 to 18 -

D .l D e fin itio n s F o rm u la s - C o n c e p ts
15

F u n d a m e n ta ls o f s tr u c tu r a l o p tim iz a tio n

15.1

M o tiv a tio n aim d ev elo p m en t

T h e previous chapters have introduced fundam entals for determ ining th e


stru c tu ra l behaviour required for th e dim ensioning a n d design of a stru ctu
re, i.e. calculation of deform ations, stresses, n a tu ra l vibration frequencies,
buckling loads, etc. In view of th e developm ent an d construction of m achin
es a n d system com ponents th e question arises w hich m easures m ust be ta
ken in order to reduce costs a n d to im prove quality an d reliability; in o ther
w ords th is m eans th a t an o p t i m i z a t i o n of th e p roperties is being aim ed at.
In term s of th is dem and, th e topic S t r u c t u r a l O p t i m i z a t i o n has emerged,
over th e p a st years, an extensive field of research th a t can be described by
th e following form ulation [D .29]:
m ay b e defined as th e ratio n al estab
lishm ent of a stru c tu ra l design th a t is th e best of all possible
designs w ith in a prescribed objective an d a given set of geom e
tric a l an d /o r behavioral lim itations.

S tr u c tu r a l o p tim iz a tio n

C urrent research in optim al stru ctu ral design m ay very roughly be said to
follow two m ain paths. Along th e first, th e research is p rim arily devoted to
studies of f u n d a m e n t a l a s p e c t s o f s t r u c t u r a l o p t i m i z a t i o n . B road conclu
sions m ay be draw n on th e basis of m ath em atical properties of governing
equations for op tim al design. T hese properties a re not only studied analy
tically in order to derive q u alitative results of general validity, but are also
often investigated num erically via exam ple problems. Along th e o th e r m ain
p ath of research, th e em phasis is laid on th e d e v e l o p m e n t o f e f f e c t i v e n u
m e r i c a l s o l u t i o n p r o c e d u r e s for optim ization of com plex practical stru ctu
res [ D .3, D.12, D.21, D .22, D.30 ].
T he constan t flow of general reviews, surveys of subfields, conference pro
ceedings, and new textbooks on o ptim al stru c tu ra l design testify th e strong
activity, recen t advances, an d increasing im p o rtan ce of th e field. A selec
tion of such recent publications is listed as references at th e end of this
book.

302

15 Fundamentals of structural optimization

15.2 Single prob lem s in a d esign proced u re


In this section, th e d istinctions betw een usual structural an alysis, redesign
or sen sitiv ity analysis, and optim ization o f structures w ill b e m ad e clear,
and th e basic steps pertaining to optim al design w ill b e outlined.
In a usual s tr u c tu r a l a n a ly sis problem , th e structural design is given, toge
ther w ith relevant properties of the m a teria l(s) to b e used and th e support
conditions for th e structure. Also, on e set or m ore o f loading is specified,
that is, com pletely sp ecified in d eterm in istic problems, or given in term s of
probabilities in probabilistic problems. For each set o f loading, th e relevant
set o f equilibrium (or state) equations, con stitutive equations, com patibility
conditions, and boundary conditions, are th en used for d eterm ining th e
structural response, e.g., th e state o f stress, strain and deflection, natural vi
bration frequencies, and load factors for elastic in stab ility or p lastic collapse
( see M ain Chapters A, B, C ) .

R edesign or s e n s itiv ity a n a ly sis (Ch. 18) refers to th e ty p e o f problem


w here som e o f th e design, m aterial, or support param eters are changed (or
varied), and w here th e corresponding changes (or variations) o f th e structu
ral resp onse are determ ined via a repeated (or special) analysis. It is w orth
noting that a conventional design procedure norm ally con sists of a series of
repeated changes o f th e structural param eters follow ed by an alysis, i.e., a se
ries o f redesign analyses, w hich is carried out u ntil a structure is found that
fu lfills th e behavioral requirem ents and is reasonable in costs. If th e chan
ges o f th e structural param eters prior to a given redesign an alysis are deter
m ined rationally from th e earlier an alyses as th e b est possib le ones, th e
procedure would identify on e o f optim al design. Such procedures are of
m uch higher sign ifican ce than th e traditional design procedure w here usual
ly th e design changes are only decided by guessw ork based on exp erience or
inform ation obtained from previous analysis, and a structure obtained by
th e traditional design procedure w ill therefore not n ecessarily b e b etter than
other possib le alternatives.
T h e label s tr u c tu r a l o p tim iza tio n id en tifies th e ty p e o f design problem
w here th e set o f structural param eters is subdivided into so-called p re a ssig
ned p a ra m e te rs and design va riables, and th e problem con sists in determ i
ning th e optim al values o f th e design variables such that th ey m a x im ize or
m in im ize a sp ecific function term ed th e objective fu n c tio n (c rite r io n or
cost fu n c t i o n ) w h ile satisfiyin g a set o f g eo m etrica l and/or b ehavioural re
q u irem en ts, w hich are sp ecified prior to design, and are called c o n stra in ts.
According to th e m anner in w hich th e design va ria bles are assu m ed to de
pend on the sp a tia l variables, optim al design problem s m ay b e roughly ca
tegorized as

(1 )

C o n tin u o u s (or d istrib u te d p a ra m eter) optim ization problems, and

( 2 ) d iscrete (or p a ra m eter) optim ization problems.


Usually th e design variables of stru c tu ra l elem ents like rods, beam s, arches,
disks, plates, and shells are considered to vary continuously over th e length

15.3 Design variables - constraints - objective function

303

or dom ain of th e elem ent, an d such problem s th en fall into category (1),
w hile problem s of optim izing inh eren tly discrete structures like trusses,
grillages, fram es, or com plex practical structures belong to category ( 2 ).

15.3

Design variables constraints objective function

v a r i a b l e s m ay describe th e configuration of a structure, elem ent


quantities like cross-sections, w all-thicknesses, shapes, e tc ., an d physical
p roperties of th e m aterial.

D e s ig n

O ptim ization problem s can best be classified in term s of th eir design vari
ables. Based upon th e exam ple of a truss-like stru c tu re according to L.A.
SCHM IT/ R.H. M ALLET [D.41], an d N. O L H O F F /J.E . TAYLOR [D.29]
possible design variables can be divided into five (o r six, respectively) dif
ferent classes (Fig. 15.1). In th e following, th ese groups are briefly descri
bed in term s of th e degree of com plexity w ith w hich they en ter into th e
design process:
a) C o n s tr u c tiv e la y o u t

T h e determ ination of th e m ost suitable layout is on principle only possible


by investigation into all existing types a n d by com paring th e calculated op
tim a.
b) T o p o l o g y

T h e topology or arran gem en t of th e elem ents in a stru ctu re is often descri


bed by p aram eters th a t can be m odified in discrete steps only (e.g. num ber
of trusses at th e supporting stru ctu re of a reflector, num ber of sections of a
continuous girder). D ifferent topologies can also be obtained by elim inating
nodes an d linking elem ents.
c) M a t e r i a l p r o p e r t i e s

In term s of th e ir properties, conventional m aterials possess variables like


specific weight, YOUNG's m oduli, m echanical stren g th properties, etc.
w hich can usually only a tta in certain discrete values. For b rittle m aterials,
often th e stochastic ch aracter of th e properties has to be regarded.
d) G e o m e tr y - sh a p e

T h e geom etry of trusses or fram es is described eith er by th e nodal coordi


nates or by th e b a r lengths, w hile in th e case of load carrying structures
w ith plan e or curved surfaces (p lates, shells) th e geom etry is given by
spans, curvatures, an d th e thickness distributions. Usually, th e se variables
a re continuously variable quantities.
e) S u p p o r t s - l o a d i n g s

Design variables of th is type describe th e su p p o rt (or boundary) conditions


or th e distribution of loading on a structure. Thus, eith er th e location,
num ber, an d ty p e of support or th e ex tern al forces m ay b e varied in order
to yield a m ore effective design. T h e design variables of this category may
be continuous or discrete.

304

15 Fundamentals of structural optimization

f) C ross-sectio n s
T his class of design variables has been used m ost frequently in optim iza
tion tasks (cross-sectional areas, m om ents of in e rtia ).
W hen considering stru ctu ral param eters, we distinguish betw een indepen
dent an d dependent variables, as well as pre-assigned (c o n sta n t) p aram e
ters, w here a stru c tu re is uniquely characterized by statin g th e values of
its independent variables, th e so-called design variables. M ost often, cer
ta in cross-sectional ch aracteristics (thicknesses, d iam eters) a re em ployed
as independent variables, from which all o th er values can be calculated.
As previously m entioned, consideration of th e constant param eters and th e
determ in atio n of th e dependent variables are m ade in algorithm s for a n a
lysis of th e stru c tu ra l design.
T h e i-th design variable w ill be denoted by Xj, an d all n design variables
are com posed in a vector x which lies in th e design space, an n-dim ensional EUCLIDEAN space:
(15.1)

[ x i , x 2 , . . . , Xj, . . . , x j .

Any set of design variables defines a design of th e stru c tu re a n d m ay be


represented as a point in th e design space.

a)

Construction

b)

Topology
M aterial
properties

d)

G eom etry
shape

e)

Supports
loadings

i)

Steel

A lum inium

Com posite

A"

Cross-section

Fig. 15.1: C lassification of design optim ization problem s for truss-like


structures in term s of different types of design variables.

15.3 Design variables - constraints - objective function

305

M any designs from th e to ta lity of possible designs will generally not be ac


ceptable in term s of various design an d perform ance requirem ents. To ex
clude such designs as candidates for an o ptim al solution, th e design and
perform ance requirem ents are expressed m ath em atically as constraints
prior to optim ization. T h e constraints m ay b e of two following types:
- G eo m etrica l (or side) c o n s tra in ts a re restrictions im posed explicitly on
th e design variables due to considerations such as m anufacturing lim it
ations, physical practicability, aesthetics, etc. C onstraints of th is k in d a re
typically in e q u a lity c o n s tr a in ts th a t specify lower or upper bounds on th e
design variables, but they m ay also be e q u a lity c o n s tr a in ts like, e.g., lin k a
ge c o n s tr a in ts th a t prescribe given proportions betw een a group of design
variables.
- B eh a vio ra l c o n s tr a in ts a re generally n o n lin ear an d im plicit in term s of
th e design variables, an d th ey m ay be of two types. T h e first ty p e consists
of eq u a lity c o n s tr a in ts such as sta te an d com patibility equations governing
th e stru c tu ra l response associated w ith th e loading co ndition(s) u n d e r con
sideration. T h e second ty p e of behavioral co nstraints com prises in e q u a lity
c o n s tr a in ts th a t specify restrictions on those quantities th a t characterize
th e response of th e structure. T hese constraints m ay im pose bounds on
local quantities like stresses a n d deflections, or on global quantities such as
com pliance, n a tu ra l vibration frequencies, etc.
T h e constraints are form ulated in th e form of e q u a lity a n d /o r in e q u a lity
c o n s tr a in ts :
hi ( x )

gj(x)=<0

(i

l.---,q).

(15.2a)

(j = 1,..., p ) .

( 15.2b)

Each ineq u ality constrain t ( 15.2b) is represented by a surface in th e design


space w hich com prises all points x for w hich th e condition is satisfied as
an equality con strain t gj ( x ) = 0 . O ne distinguishes betw een feasible, or
adm issible, an d infeasible, or inadm issible designs. All th e feasible (o r ad
m issible) designs lie w ithin a subdom ain of th e design space defined by th e
co nstrain t surfaces as indicated in Fig. 15.2 for th e special case of a plane
design space (o n ly two design v ariab les). W e generally assum e th a t th e
constraints a re such th a t th e design space is open, i.e. th a t a set of feasible
designs exists.
T h e objective fu n c tio n , w hich is also term ed th e cost or th e c rite rio n
fu n c tio n , m u st b e expressed in term s of th e design variables in such a way
th a t its value can be determ in ed for any point in th e design space. It is this
function w hose value is to be m inim ized or m axim ized by th e optim al set of
values of th e design variables w ithin th e feasible design space, an d it m ay
represent th e stru c tu ra l weight or cost, or it m ay be tak en to represent som e
local or global m easure of th e stru ctu ral perform ance like stress, displace
m ent, stress in ten sity factor, stiffness, plastic collapse load, fatigue life,
buckling load, n a tu ra l vibration frequency, aeroelastic divergence, or flu tte r
speed, etc.

306

15 Fundamentals of structural optimization

a c t iv e c o n s tr a in ts

g(x) = 0
f e a s i b l e in itia l d e s ig n
f(*o )

is o - c u r v e f ( x ) = c o n s t
fo r t h e o b j e c t iv e f u n c t io n
in f e a s ib le s e t

g(x) >
Fig. 15.2: Concepts of stru ctu ral optim ization
T he objective function is in m ost cases a scalar function f of th e design
variables x defined as follow s:
f:=f(x)
15.4

(15.3)

P r o b le m f o rm u la tio n T a s k o f s tr u c t u r a l o p tim iz a tio n

T h e usual problem of o ptim al stru ctu ral design consists in determ ining the
values of th e design variables xi ( i = 1 , . . . , n ) such th a t th e objective func
tion a tta in s an extrem e value w hile sim ultaneously a ll constraints are sa
tisfied. M inim ization of th e objective function, i.e. M in f , is considered in
th e conventional m athem atical form ulation. If an objective function f is to
be m a x im iz e d , one sim ply substitutes f by - f in th e form ulation, since
M ax f < > M in ( - f ).
M athem atical fo r m u la tio n :
(15.4)

w ith

Rn
x
f(x )

g(x)
h(x)

n-dim ensional set of real num bers,


vector of th e n design variables,
objective function,
vector of th e p inequality constraints,
vector of th e q equality constraints.

F easible d om ain:

X : = j x e R n| h ( x ) = 0

, g (x ) < 0

(15.5)

S tructural optim ization generally deals w ith th e solution of N on-Linear O p


tim iza tio n Problem s (N L O P ).

307

15.5 Definitions in mathematical optimization


1 5 .5

D e f i n i t io n s in M a t h e m a t ic a l O p t im iz a t io n

D efinition 1: Global and local m in im a


1)

A global m in im al point x* e X is characterized by


f ( x* ) < f (x )

2)

x e X .

(15.6a)

A local m in im al point x* e X is characterized by


f(x*)<f(x)

x e X n U (x*),

(15.6b)

w here U ( x * ) denotes th e e - neighbourhood of point x*.

D efinition 2: C o n d itio n s fo r a m in im u m o f an u n c o n stra in e d problem


1)

N ecessary condition
V f(x*)= (^-,

v '

2)

v d Xj

-i!

d x2

*1)

d xn v

=0

'(15.7)

Sufficient condition

If (15.7) is fulfilled, a n d if th e HESSIAN m atrix


H * := H ( x * )

d2 f

(15.8)

.d X; d Xj

is strictly positive definite, th e n x* is a local m inim um of f ( x ).

D efinition 3: C o n d itio n s fo r a m in im u m o f a c o n stra in ed problem


D eterm ination of optim ality conditions by m eans of th e LAGRANGE function
^
L ( x , a , / 3 ) = f ( x ) + 2 a i hi ( x ) +
i=i

w ith th e LAGRANGE AN m ultipliers


1)

jSj ( x )

( 159)

j=i

N ecessary conditions for a local m inim um x* to the problem (15.4)

KUHN-TUCKER conditions [ D.26 ]:


q

V L ( x * ) = V f ( x * ) + 2 Q * v h i ( x *) + 2 / ?f V S j ( x *) = 0
i=i

j=i

( 1510a)

308

15 Fundamentals of structural optimization

Fig. 15.3: G eom etrical in terp retatio n of th e K U H N-TUCK ER-conditions


for a problem w ith th ree inequality constraints
and
hj(x*) = 0

(i = 1

q)

(15.10b)

gj(x*)<0

(j = l , . . . , p )

(15.10c)

pf>

(j =

i,...,p)

(m od)

For each value of j ( j = 1 , . . . , p ) it holds th a t /?* = 0 if gj ( x * ) < 0 and


th a t (3* > 0 if g j ( x * ) = 0. T h e values of th e LAGRANGIAN m ultipliers
a * associated w ith
th e equality constraints (15.10b) m ay both be positive
an d negative (or zero).
2)

Sufficient conditions

For a convex problem , th e KUHN-TUCKER conditions are also sufficient.


A geom etrical in terp reta tio n for a problem w ith th ree inequality con
strain ts (a n d no equality co n strain ts) is illu strated in Fig. 15.3. For this
problem, th e following equations m ust be valid for points A an d B accor
ding to (15.10) in order to be m inim um p o in ts :
a) Point A :

- V f ( x * ) = 0 * V g l ( x * ) + (3* V g 3 ( x * )

(15.11a)

T he negative g rad ien t of th e objective function does not lie w ithin th e sub
set defined by th e g rad ien ts of th e constraint functions, i.e. (15.10d) is vio
lated by eith er /3* or /5* . x ^ is not a m inim um point as th e function v a
lue f ( x * ) can be reduced in th e feasible set.
b) Point B :

- V f ( x * ) = (3* V g 2 ( x * ) + (3% V g 3 ( x * )

(15.11b)

T he considered point xjjj is a local m inim um point w ith a ll KUHNTU C K ER conditions satisfied. Note th a t for poin t B th ere is no direction
in th e feasible set in w hich th e function value f ( x * ) can be reduced.

15.6 Treatment of a Structural Optimization Problem (SOP)

309

15.6 Treatm ent of a Structural Optim ization Problem (S O P )


An optim ization problem can generally be treated by proceeding in accor
dance w ith th e T h r e e - C o l u m n s - C o n c e p t [D.12] com bined in a so-called op
tim ization loop (Fig. 15.4). T his concept is based on th e fu n dam ental con
cepts presented in 15.1 an d 15.2, and w ill be described briefly in th e follow
ing.
Column 1: S t r u c t u r a l M o d e l - S t r u c t u r a l A n a l y s i s
One of th e m ost im p o rtan t assum ptions w ith in an optim ization process
consists in tran sferrin g a real-life m odel into a stru ctu ral model. Thus, any
stru ctu ra l optim ization problem is based on a m ath em atical description of
the physical behaviour of a structure. In th e case of m echanical system s,
these are typ ical stru c tu ra l responses to static and dynam ic loads like de
form ations, stresses, eigenfrequencies, buckling loads, e tc. T h e sta te vari
ables required for th e form ulation of objective functions and constraints are
com puted in th e stru c tu ra l m odel by m eans of efficient analysis procedures
like Finite-E lem ent-M ethods, T ran sfer M atrices M ethods, etc. In order to
achieve a w ide range of application for an optim ization procedure, various
stru c tu ral analysis m ethods, e.g. hybrid m ethods, should be available.
Column 2 : O p t i m i z a t i o n A l g o r i t h m
M athem atical program m ing procedures a re predom inantly applied for the
solution of nonlinear, constrained optim ization problem s (N L O P ). T hese al
gorithm s a re based on iteratio n procedures which, proceeding from an in iti
al design x 0, generally yield an im proved design variable vector x k after
each iteratio n cycle k. T h e optim ization calculation is term in ated w hen a
predefined convergence criterion becom es satisfied during an iteratio n (for
m ore details see C hapter 16). Num erous studies have show n th a t th e choice
of th e m ost appropriate optim ization algorithm is problem -dependent, a fact
th a t is of p a rtic u la r im portance in term s of a reliable optim ization flow
and high efficiency (com putational tim e, ra te of convergence).
Colum n 3: O p t i m i z a t i o n M o d e l
F rom an engineering point of view, th e optim ization model constitutes a
bridge betw een th e stru ctu ral model an d th e optim ization algorithm s, and
is a very considerable colum n w ith in th e optim ization process. F irst, the
analysis variables are chosen from th e stru ctu ral p aram eters as those
quantities th a t a re v aried during th e optim ization process. T he d e s i g n m o
d e l to be determ in ed describes th e m ath em atical relation betw een th e an a
lysis variables a n d th e design variables. By additionally transform ing th e
design variables into transform ation variables, th e optim ization problem is
adapted to th e special requirem ents of th e optim ization algorithm in order
to increase efficiency an d convergence of th e o ptim ization calculation. The
e v a l u a t i o n m o d e l determ ines th e values of th e objective functions an d con
stra in ts from th e values of th e state variables. T he s e n s i t i v i t y a n a l y s i s
m odules com pute th e sensitivities of th e objective functions an d th e con
stra in ts w ith respect to sm all changes of th e design variables; all th is in
form ation is tran sferred to th e optim ization algorithm or to th e decision
m aker, respectively, for judging th e design.

310

16 Algorithms of mathematical programming ( MP)

j
I

d e c isio n
m aker

I
1

d ata

~~1

o p tim al
d esig n

transform ed
variable

o p tim izatio n
alg o rith m

optimization model

p.g.h

r)p

r)h

Sxdxdx

op tim izatio n
strateg ics

stru c tu ra l m o d e l
stru ctu ral a n a ly sis
u = u (y )

ev alu atio n
m o d el

m m rnz
se n sitiv ity
ainalvsis

Fig. 15.4: S tructure of an optim ization loop


T he accom m odation of th e m odules for th e se n sitivity analysis (se e C hap
te r 17) an d fu rth er im p o rtan t m odules for optim ization strategies (see C hap
ter 18) like M u lticriteria O ptim ization, Shape an d Topology O ptim ization,
M ultilevel O ptim ization, etc. w ith in th e optim ization model, w ill be m en
tioned in Section 18.3.

16

A lg o r ith m s o f M a th e m a tic a l P r o g r a m m in g (M P )

In the following, solution algorithm s for optim ization problem s cast in the
sta n d ard M P form (15.4) will be considered. O ne distinguishes betw een
o p tim iza tio n a lg o rith m s o f zeroth, f ir s t and second order depending w he
th e r th e solution algorithm only requires th e function values, or also th e
first an d second derivatives of th e functions. It w ill be assum ed in this
C hapter th a t in general th e functions f, h j , gj in (15.4) are continuous and
a t least tw ice continuously differentiable.
T he m ajority of solution algorithm s is of an iterative character, i.e. sta rtin g
from an in itia l vector x 0 one obtains im proved vectors
, x 2 , . . . , by suc
cessive application of th e algorithm . Iterative solution procedures are neces
sary since p ractical problem s of stru ctu ral optim ization are generally highly
nonlinear.
16.1 P r o b le m s w i t h o u t c o n s tr a in ts
(a )

M ethods o f one-dim ensional m in im iza tio n steps

Itera tio n rule:


Xj+1 = Xj + Qj Sj

( i = 1 ,2 , 3 , . . . )

(16.1a)

w ith arb itra ry search directions s; in th e design space a n d an optim al step


length Qj for th e i-th iteration step ( Fig. 16.1).

16.1 Problems without constraints

Fig. 16.1: O ne-dim ensional m inim ization step along a


straig h t line

311

Fig. 16.2: A pproxim ation of f ( a ) by


m eans of a quadratic poly
nom ium P 2 ( a )

In case of a variable step len g th a


x ( a ) = Xj + a S j ,

(16.1b)

a straig h t line occurs in th e design space, and we have the following form
of th e objective fu n c tio n :
f [ x ( a ) ] = f ( Xi +

QSj ) = f ( a )

(16.2)

w here th a t value cq of a is to be determ ined which m inim izes f in a given


direction Sj. T h is im plies a reduction of th e problem to a series of one-di
m ensional m inim ization problems w here th e resu lt depends on th e choice
of th e search direction Sj ( L in e-S ea rch -M eth o d ).
T h e following procedures can be applied for determ ining m inim al points
[ D.10, D.21, D .22, D .24, D.40 ].
-

Q uadratic polynom ia (LAGRANGEAN in te rp o la tio n ) P 2 (o: ) (Fig. 16.2),

Cubic polynom ia ( H ERM ITE in te rp o la tio n ) P 3 ( a ) ,


R egula falsi ,
Interval reduction by m eans of FIBONACCI-search [D.25].

By applying m ath em atical optim ization m ethods, it is intended th a t a


yields a good approxim ation of f in th e neighbourhood of th e point of m i
nim um Qj (Fig. 16.2). D ifferent convergence criteria can be used, for in
stance :

312

16 Algorithms of mathematical programming (MP)

or
|P2( S ) - f ( a ) |

< e

w ith a lim it of accuracy e ss 0.01 .

(16.3b)

P2( )
( b ) F irst PO W E LL m ethod o f conjugate directions - m ethod of 0th order
In order to calculate an iteration point x i+1, inform ation on th e previous
points X j ,..., Xj is used. F or th is purpose, we need th e notion of conjugate
directions. T h e vectors Sj and ^ are called conjugate if they satisfy the
condition [D.37]

SJ H s k = 0

(j ^ k ) *

( 16A)

w here H is th e HESSIAN m atrix (15.8). In PO W ELL s m ethod, H in ( 16.4)


is replaced by an approxim ation m atrix A , an d vectors th a t fulfill this
condition are called A-conjugate.
In case of a q u adratic function, th e conjugate direction
belonging to an
a rb itra ry tan g en t Sj always leads to th e centre of a fam ily of ellipses.
S o lu tio n strategy:
Proceeding from a n in itia l point x 0 , n one-dim ensional m inim ization ite ra
tions ( s te p s ) are carried out along n linearly independent search directions
s0 , . . . , sn , e.g. along th e n u n it vectors e 0 , . . . , e n of th e coordinate axes in
th e design space, i.e.
x i+i = x i + a i si

(i = 0 ,1.2,...,n) .

(16.1c)

Fig. 16.3: A pplication of th e F irst PO W ELL m ethod of conjugate directions

16.1 Problems w ithout constraints

313

T he search cycle is considered by th e ( n + 1 )-th one-dim ensional m inim i


zation step in th e direction of
sn+1 : = x n + 1 - x n_x ,

(16.1d)

w hich leads to point x n+2. For th e following search cycle, s t is elim inated,
th e index of th e rem ain in g directions 8 2 , . . . , 8 + 1 is decreased by one, and
th e described procedure is th e n applied to all subsequent search directions
(Fig. 16.3).
In m any cases th e convergence behaviour of th is procedure is insufficient
due to generation of alm ost linearly dependent search directions. C ertain
m odifications, however, lead to im provem ents [ D.18, D.38 ].
(c)

G ra d ien t m ethod o f steepest descent - m eth o d o f 1st order

By this m ethod we choose th e search vector s ; in th e direction of th e


steepest descent of f ( x ) at th e point x ; , i.e. in th e negative g ra d ien t direc
tion ( Fig. 16.4):
a^-V ffxj).

(16.5)

By m eans of (16.5) we determ ine th e optim um point along this direction,


using the ite r a tio n ru le
* 1+1 = * i + Qi si = *i Qi V f ( x i )

(i 1,2,3...)

(16.6)

w here th e optim al step length a ; can be calculated by one of th e m ethods


described und er ( a ).
(d)

F LETC H ER-REEVES-M ethod o f conjugate gradients [D .19,D .23] m eth o d o f 1st order

T h e first one-dim ensional m inim ization step is carried out in the direction
of th e steepest descent according to (16.5)
si = ~ v f ( x i)
an d we thus reach point x 2 (Fig. 16.5).
ta n g e n tia l
h y p e r p la n e

V f ( x 1+1)

f = co n st
- V f ( x i+1)

Fig. 16.4: M inim ization step in th e


direction of th e steepest
descent

314

16 Algorithms of mathematical programming (MP)

x,

Fig.16.5: FLETC H ER -R EEV ESM ethod of conjugate


gradients

Proceeding from this point, m odified search directions are g e n e ra te d :

(e)

Special Q u a si-N E W T O N procedure S Q N P - m ethod o f 2nd order

In this m ethod, th e search direction sk of th e k -th iteration is defined as


Sk = - H k V f ( x k )

(16.8)

w here th e m atrix H k (w hich is not th e HESSIAN) is calculated by m eans


of variable m etrics according to DAVIDON, FLETC H ER, PO W ELL ( D F P )
[ D-ll ]:

w ith

1^ = I ,

yk-i = v f (xk) - V f( xk-i) A* i

16.2
16.2.1

(16.9b)

step length along th e search direction sk l .

P r o b l e m s w it h c o n s t r a i n t s
R eduction to u nconstrained problem s

( a) G eneral rem arks on penalty functions


T hese m ethods have originally been developed by FIACCO a n d McCORMICK who chose th e nam e SUMT ( Sequential U nconstrained M inim ization
Techniques ) [ D.17 ].

16.2 Problems with constraints

315

T he optim ization problem (15.4) contains p inequality constraints gj ( x ) < 0


F orm ulation of a m odified objective function for each itera tio n :
p

^ ( x . R J ^ f U j + R i^G fg^x)]
j= i

(i = 1,2 ,3,...)

(16.10)

w ith a p en alty function G [ g j ( x ) ] an d th e penalty p aram eters R i .


Irrespective of explicit constraints, m inim ization is th e n carried out by
m eans of th e previously described m ethods for u nconstrained problems.
F unction G is chosen in such a way th a t during m inim ization for a series
of values for Rj th e solution converges tow ards th a t of th e original problem
w ith constraints.
( b ) F requently applied m ethods
- M ethod o f e x te r io r p e n a lty fu n c tio n s
P enalty objective function in (16.10):
G [ g j ( x )] := ( m a x [ 0 , g j ( x ) ] )

(16.11a)

M odified objective function


^ i ( x - Ri ) = f ( x ) + R i 2

( max[ . g j ( x )])2

( i = 1 >2 3 - -)

(16.11b)

j= i

is to be m inim ized for a set of increasing values of

-3

R; ( e.g. R t = 10 ,

R 2 = 102, R 3 = K f 1, . . . ) .
Owing to th e inequality constraints gj ( x ) < 0 ( j = 1 , 2 , 3 , . . . ) in ( 16.11a)
we o b ta in :
G [ gj ( x ) ] = 0

in th e feasible dom ain ,

G [ g j ( x )] = gj ( x )2 > 0

in th e infeasible dom ain .

F unctions f a n d $ a re identical in th e feasible dom ain, w hereas f is p e n a


lized in th e infeasible dom ain by a sum m ation of th e non-negative term s
(16.11b).
- M ethod o f in te r io r p e n a lty fu n c tio n s or b a rrie r fu n c tio n s
P enalty function in (16.10):
G [ g j ( x )] : =

(16.1Sa)

316

16 Algorithms of mathematical programming ( M P )

0 2( x , R 2 ) = M in

0 2(x ,R 2 ) = M in

g ito

x2

Fig. 16.6: M ethod of

a ) exterior penalty functions


b ) interior p en alty functions

g2(x ) = 0

16.2 Problems with constraints

317

M odified objective function


<P i ( x , R i ) = f ( x ) - R i | ; - 4 - T

(i = 1 , 2 , 3 , . . . ) .

(16.12b)

is to b e m inim ized for a set of decreasing values of Rj ( e.g. R x = 105,


R 2 = 104, R 3 = 103 e tc .).
In th is case, th e objective function f is penalized in th e feasible dom ain bysum m ation of th e positive term s.
For details on fu rth e r procedures refer to [D .24].

16.2.2 G en eral n o n lin ear problem s - d irect m ethods


In th e p ast, a su b stan tial num ber of direct algorithm s of M athem atical
Program m ing ( M P - algorithm s ) have been developed for th e solution of ge
neral, n o n lin ear optim ization problem s. W hen applied to problem s of com
ponent optim ization, these algorithm s m ust b e able to reduce th e num ber
of optim ization steps during optim ization since often extensive stru c tu ra l
analyses have to be carried out at each iteration. In addition, a sufficiently
good convergence behaviour as well as reliability an d robustness m u st be
dem anded of th ese procedures. T hese characteristics largely depend on the
degree of non lin earity of th e posed problem.
In th e following, two frequently used procedures sh all be briefly described
as ty p ical algorithm s:
( a ) S eq u e n tia l L in e a r iz a tio n P rocedure SLP
T he efficiency of lin ear m ethods can also be utilized for nonlinear design
problem s by successively solving lin ear sub stitu te problem s in th e form of a
so-called seq u e n tia l lin e a r iz a tio n [ D .20, D.35, D.38 ].
By introducing u pper an d lower bounds (hypercube, move lim its) for all
design variables, G R IFFIT H an d STEW ARD have augm ented th e range of
application to problem s w here th e solutions are not at th e intersection of
constrain ts but, m ore general, on a curved hypersurface [D.20]. T h e objec
tive function a n d constraints of th e nonlinear, scalar in itia l problem (15.4)
are expanded in a TAYLOR-series in th e vicin ity of a point x k . By m ain
tain in g th e lin e a r term s only we obtain
f(xk + Ax)

f(xk) + V f ( x k)A x,

(16.1Sa)

hj( xk + A x )

hj(xk) + V h j(x k )Ax (i = l,2 ,...,q )

(16.1Sb)

6j( x k + A x )

gj(xk ) + Vgj ( x k )Ax (j = 1 ,2 , . . ., p ) .

(16.1Sc)

318

16 Algorithms of mathematical programming ( M P )

In addition, th e design space of th e linearized problem is bounded by a hy


percube according to Fig. 16.7
k

*1 ~< X;1 -< X; u

X;

(16.14)

(i = 1,2,..,,n) ,

since th e TAYLOR-expansion is only valid for sm all A x . Here, it is conve


nient to use superscripts k for th e approxim ation steps, w hereas th e sub
scripts i denote th e num ber of design variables, an d u or I th e upper an d
lower bounds, respectively.

n o n lin e a r
in itia l p r o b le m

1.

a p p r o x im a tio n

2.

a p p r o x im a tio n

3.

a p p r o x im a tio n

Fig. 16.7: O ptim ization flow of SLP in th e tw o-dim ensional case

16.2 Problems with constraints

319

T he linearized problem according to (16.13) is solved by m eans of the


SIM PLEX -procedure by DANTZIG [D.20]. Since for th a t purpose a ll vari
ables have to be larger th a n zero, a lin ear tran sfo rm ation of variables has
to be carried o u t :
yi = A xi + ( xf - x u )

i = l,2,...,n.

(16.15)

T h e linearized problem th e n r e a d s :
M in

| cT y J =

cT y

w ith

c = V f xk ) ,

(16.16)

together w ith th e linearized co nstraints h ; ( y ) and g j ( y ) , an d th e hyper


cube according to (16.14) .
T he solution y * of th e linearized problem yields an im proved x k+1 for the
nonlinear problem. T h e hypercube is reduced by m eans of correction rules,
th u s th e side length of th e hypercube decreases during th e optim ization.
Fig. 16.7 illu strates how SLP works in th e tw o-dim ensional case. It becom es
obvious th a t th e optim ization flow strongly depends on th e choice of the
in itial hypercube.
( b ) A u g m e n te d L A G R A N G E -F u n c tio n Procedure LP N LP
E quality an d inequality constraints a re included in a n augm ented
LA G R A N G E -function. T h e u n constrained problem is th en solved by m eans
of search techniques [ D.36 ].
P IE R R E an d LOW E have developed th e optim ization procedure LPNLP
(LA G RA N G E Penalty M ethod for N on-Linear Problem s). Using the
L A G R A N G E -function (see (15.9)) directly as an objective function has
c ertain im m an en t disadvantages as th e KU H N -TU CKER-conditions are not
necessarily sufficient. Even if VL ( x * , a * ,/? * ) = 0 is valid, V 2 L* m ay not
be positive definite. Therefore, th e LAGRANGE-function defined in (15.9)
is augm ented by p e n a lty -te r m s w ith th e special w eighting factors w; ( i =
1,2,3)
La ( x , a , / ? , w ) = L ( x , a , / J ) + Wj Pj + w 2 P 2 + w3 P 3

(16.17)

w ith
pi = i [ M * ) ] 2 .
i=l

p2=
P3

Z[gj(*)]2
i eC

- ca = {j|/?.>o},

Z [ g j ( x ) ] 2 . Cb = { j | ^ j > 0

jecb

and gj > 0 } -

T he optim ization problem is th e n solved sequentially by u n constrained op


tim izatio n an d correction steps. For fu rth e r details refer to [D.36], Fig. 16.8.
illustrates th e flow -chart of th e LPNLP-procedure.

320

16 Algorithms of mathematical programming ( M P )

Fig. 16.8: Flow -chart of th e LPNLPprocedure


( c ) F u r th e r a lg o r ith m s
In th e following, some of th e algorithm s curren tly applied in stru c tu ra l op
tim ization sh all be briefly introduced.
S e q u e n t i a l Q u a d r a t i c P r o g r a m m i n g (S Q P )

[D.38]

B ased upon th e L A G R A N G E -function, th is m ethod utilizes th e sequential


linearizatio n an d q uadratic approxim ation of a nonlinear problem by
m eans of th e BFGS (BROYDEN, FLETC H ER , GOLDFARB, SHANNO)-form ula [D.21] of th e HESSIAN-matrix. T h e q uadratic subproblem is th e n sol
ved in ord er to g enerate a search direction; for th e one-dim ensional search,
th e optim al step len g th is d eterm ined by a p en alty function and a q u adra
tic interpolation.
M e th o d o f G e n e r a l iz e d R e d u c e d G r a d i e n t s (G R G )

A
by
in
by

[D .l]

subset of th e design variables is elim inated from th e objective function


m eans of active constraints. T h e reduced g rad ien t is th e n calculated
order to g enerate a search direction. T he op tim al step length is found
em ploying th e Q uasi-N EW TO N - Algorithm .
H y b rid

procedure

c o n s is tin g

of

SQ P-

and

G R G -m e th o d s

(Q P R L T )

[D.1,D.35]
In a first step, a search direction is d eterm ined by m eans of th e SQP-algorithm ; then, th e optim al step length is calculated by th e G RG -algorithm .
T hus, th e advantages of SQP a n d GRG are com bined in one single algo
rithm .

17.1 Purpose of sensitivity analysis

321

M ethod o f M oving A sym ptotes (MMA) [D.45]


Here, a sequential, convex aproxim ation of th e n onlinear problem is carried
out. First, th e problem is transform ed into a dual problem w ith specific
characteristics (separable, convex problem ) an d is th e n solved by m eans of
a conventional g rad ien t algorithm or by a N E W TO N -algorithm . T his solu
tio n point serves as th e sta rtin g poin t for th e subsequent approxim ation.
A com pilation of direct M P-algorithm s for th e solution of general, nonline
a r optim ization problem s can be found in num erous books, am ong others
[ D .24, D .33, D .39, D.40, D.46 ].

17

S e n s it iv it y a n a ly s is o f s tr u c tu r e s

17.1

P u r p o s e o f s e n s itiv ity a n a ly s is

T he objective of design se n sitiv ity a nalysis is to calculate gradients of the


stru ctu ra l responses an d cost functions w ith respect to sm all changes of th e
design variables. T h e determ in atio n of th e g rad ien ts of th e objective func
tion an d th e constraints is a highly im p o rta n t step in th e optim ization pro
cess (see Fig. 15.4), since these g rad ien ts a re not only a prerequisite for the
m ajority of optim ization algorithm s (see C hap ter 16), b u t th e y also provide
im p o rta n t inform ation on th e stru ctu ral sensitivity w hen changing arb itrary
stru ctu ra l param eters. T he choice of an appropriate m ethod of sensitivity
analysis strongly influences th e num erical efficiency an d thus has im pact
on th e en tire course of th e optim ization. For th is reason, th e trea tm e n t of
th e fund am en tals of stru c tu ra l optim ization sh all be d ealt w ith separately
w ith some rem arks on frequently applied techniques for d eterm ination of
gradients.
A com plete overview of sensitivity m ethods in stru ctu ral optim ization is
given in [D .2,D .21,D .22], In add itio n to sim ple num erical fin ite difference
procedures, an aly tical or sem i-analytical m ethods and form ulations derived
from th e variatio n al principle, are increasingly applied.
Proceeding from an m -dim ensional vector function tp (e.g. vector of objec
tive function an d co n strain ts)
V = v[x,u (x)]
w ith

x e Rn

design variable vector ,

u R u vector of sta te quantities ,


we obtain th e to ta l differential of (17.1) as:

(17.1)

322

17 Sensitivity analysis of structures


d ^ [ x , u ( x ) ] = A d x

[aij]mxll

w here A =

(YlJSa)

is called th e se n sitiv ity m a trix . T he following as

sum ptions are m a d e :


d

ip

d x

ip{

dxj

m x n

d ip

d Vi

dll

dll

auk

d X

m x nu

dUk
a Xj

(17S b )

In th e sequel, different ways of determ ining A w ill be introduced.

17.2 O verall F in ite D ifferen ce (OFD) s e n s itiv ity a n a ly sis


T he O verall F in ite Difference ( O FD ) - approach im plies th a t th e entries in
th e sensitivity m atrices are approxim ated by sim ple finite difference quo
tien ts :
V i [ x j > u ( x j ) ] - Vt1 [ x l u ( x ) ]

( i

* ------------------------- a j ; -------------------

l , . . . , m )

o = i

>

<m >

w ith
Xj =
Axj

perturbation of th e j-th design variable .

T he O F D -m eth o d is very easily im plem ented, an d is com pletely indepen


dent of th e stru c tu ra l model. T h e m ethod is also applicable for tasks
beyond th e field of optim ization. However, th e O F D -analysis has th e im
m an en t disadvantage th a t for n design variables, n + 1 com plete stru ctu ral
analyses of th e to tal stru ctu ral response are required, a fact th a t leads to
very extensive com putation tim es in case of larger optim ization problems.
In addition, th e occurrence of round-off errors does not allow th e relative
p erturbatio n e Axj / xj to be chosen arb itrarily sm all. Based on expe
rience, values of e ss 10s to 10"3 a re suitable.

17.3 A n alytical and sem i-an alytical se n sitiv ity an alyses


In order to reduce th e extensive com putational effort, an aly tical or semian aly tical m ethods a re increasingly em ployed for g rad ien t calculation
[D .2,D .15,D .21,D .22,D .30,D .31]. T hese procedures a re closely linked w ith
th e applied stru c tu ra l analysis procedure, an d th e ir realization thus renders
m anipulatio n s in th e source code necessary. T hese m ethods have originally
been developed for th e FE-m ethods, but they can be generalized to all
o th e r stru c tu ra l analysis procedures w hich transform th e differential equa
tions of th e considered m echanical system into a set of algebraic equations
(tra n sfe r m a trix m ethods, difference procedures, an aly tical solution m e
thods according to RAYLEIGH / RITZ, e tc .). Here, our considerations w ill
be lim ited to those lin ear system s th a t have a system m a trix equation of
th e form :

17.3 Analytical and semi-anlytical sensitivity analyses


Fu = r
w ith

323
(17.5)

F = F (x )

global system m a trix ,

u = u (x )

vector of state quantities ,

r = r (x )

load vector .

T he solution of (17.5) for th e sta te vector u subjected to a given load vector


r, is norm ally carried out by GAUSSIAN elim ination perform ed as a twophase process of factorization of th e system m a trix F w hich does not re
quire sim ultaneous m odification of r , an d thus m akes it possible to solve
(17) for ad d itio n al load cases, i.e. several rig h t-h and sides, w ithout m uch
ad d itio n al com putational effort.
By im plicit differentiation of (17.5) w ith respect to any of th e design vari
ables Xj ( j = l , . . . , n ) , rearran g em en t of term s, and m ultiplying th e equa
tion by d t p / d u , one can replace th e derivatives of th e sta te quantities in
(17.2) by th e following ex p ressio n :

d ll

d Xj

du

' d Xj

dXj

'

(m )

Pj
w here Pj is th e so-called pseudo load vector associated w ith th e design va
riable Xj.
W ith r know n an d u obtained by solution of (17.5) , com putation of the
pseudo load vectors Pj ( j = l , . . . , n ) in (17.6) only requires th a t th e de
sign sensitivities dr/ dXj an d dF /d X j of th e load and the system m atrix
are known. Note here th a t th e form er sensitivities vanish if th e load is de
sign independent.
If in (17.6) th e design sensitivities of th e global system m atrix d F / dxj are
d eterm in ed an aly tically before th e ir num erical evaluation, th e approach is
called th e m ethod of a n a ly tic a l sensitivity analysis, an d if th ey are deter
m ined by num erical differentiation, cf. (17-4), th e label se m i-a n a ly tic a l
sensitivity analysis is used. W hile th e an aly tical m ethod is expedient for
problem s w ith cross-sectional design variables ( see Section 15.3), it is usu
ally a form idable task to im plem ent th e m ethod w hen shape design vari
ables (se e Sections 15.3 a n d 18.2) are encountered. Thus, a large am ount of
an a ly tica l work an d program m ing m ay be required in order to develop
an aly tical expressions for derivatives of, for instance, various finite elem ent
stiffness m atrices w ith respect to a large num ber of possible shape p aram e
ters. For problem s involving shape design variables, it is m uch m ore a ttra c
tive to apply th e sem i-analytical m ethod because it is easier to im plem ent
as it trea ts different types of fin ite elem ents an d design variables in a uni
fied way.

324

17 Sensitivity analysis of structures

Eq. (17.6) can be tre a te d in two different w ays:


( a ) D irect m ethod ( Design Space M ethod)
F irst, we com putate th e pseudo load vector pj associated w ith each Xj as
described above, an d th e n d eterm ine from
F

<5u
dx.

( hr

<5F

= ( ^^ xj
r - ^ r^ xj
" ) = P i

<m >

th e corresponding g rad ien ts of th e state quantities >u/ <5Xj. Since th e form


of (17.7) is analogous
to th a t of (17.5), each of these gradients ( j
1 , . . . , n ) can be solved from (17.7) using th e sam e factorization
of th e glo
bal system m a trix F as was used w hen solving (17.5) for th e vector u .
T hus, (17.7) has to be solved for n right-hand sides p. for each load case r
in (17.5).
Finally, th e g radients d tp /<5u a n d dtp/ d x are calculated in order to esta
blish th e to ta l differential according to (17.2).
(b )

A u x ilia r y variable m ethod (S ta te Space M ethod)

Here, we introduce a n auxiliary variable vector A; , th e transpose of w hich


we define as th e following product according to (17.6) :
A

l ^ F

- 1( i = l , . . . , m ) .

(17.8a)

T his yields th e equation system

fTji = ( ^ ) T
for A: th a t is to be solved for th e m
).
\
..., m

<m h )
right-hand sides

<5u

( i = 1,

G enerally, an a ly tic a l an d sem i-analytical m ethods of sensitivity analysis


a re able to reduce th e com putational tim e to a fraction of th a t necessary
for th e O F D -ap p ro ach (S ection 17.2). T his is m ainly due to th e fact th a t
th e system m a trix of th e form er m ethods has to be factorized or inverted
only once for a sensitivity analysis.
D epending on th e num ber of equation system s to be solved in (17H) and
( 17.8b), th e d irect m ethod is p referred in those cases w here few er design
variables th a n constraints are defined in th e optim ization model, w hereas
th e auxiliary variable m ethod should be applied to problem s w ith a prevai
ling num ber of design variables.
A lthough th e sensitivity m ethods described above prim arily focus on g ra
d ient calculation for structures subjected to static loading, th e m ethodology
can be extended to o th er problem s in a straight-forw ard m anner.

18.1 Vector, multiobjective or m ulticriteria optim ization

18

325

O p tim iz a tio n s tr a te g ie s

In order to tre a t different types of optim ization tasks like shape an d topo
logy optim ization problem s as well as m u lticriteria or m ultilevel optim iza
tion tasks, specific optim ization strategies have to be integrated into th e
optim ization loop according to Fig. 15.4. T hese strategies are sub-parts of
optim ization m odeling, an d they tra n sfe r a rb itra ry optim ization problem s
into so-called su b s titu te pro b lem s by way of tran sform ation or decom posi
tion so th a t th e given task s can be solved by u sual scalarized p aram e ter
optim ization procedures. In th e following, two of these strategies w ill be
briefly treated , nam ely
- vector, m u lticriteria or m ultiobjective optim ization, an d
shape optim ization,
w here a tran sfo rm atio n into p ara m e te r optim ization problem s for both
strategies is carried out.
18.1

V e c to r, m u ltio b je c tiv e o r m u l t i c r i t e r i a
P A R E T O - o p tim a lity [ D.14, D .34, D .43, D.44 ]

o p tim iz a tio n

In contrast to problem s w here a single criterion governs, for m u ltic r ite r ia


o p tim iza tio n th e o ptim al design reflects sim ultaneous m inim ization on two
or m ore criteria. T h e labels vecto r o p tim iza tio n or m u ltio b jec tive o p tim i
za tio n are also used for such problems. Problem s of this kind a re of p a rti
cu lar relevance to practice where, in general, several stru c tu ra l response
modes an d failure criteria m u st be tak en into account in th e design process.
O rdinarily in vector optim ization th e re exists a tra d e -o ff am ong c r ite r ia ,
i.e. a change in design m ay resu lt in im p ro v e m e n t according to one or m o
re criteria, b u t only a t th e expense of a w orsening as m easured by others.
One altern ativ e is to apply th e concept of P A R E T O -o p tim a lity (se e Def. 1
below) according to which a given m u lticriteria o ptim ization problem m ay
have an y th in g from one to an in fin ity of PA R ETO - solutions. It is th en up
to th e designer to identify th e o ptim al design w ith in th is set. T his step re
quires th e application by th e designer of judgm ent or som e o ther basis of
choice.
T his state of affairs reflects th e fact th a t it is only possible to obtain a
unique o ptim al design if a single, scalar objective function f ( x ) is encoun
tere d in th e optim ization problem , cf. (15.j ) . T h is fact, however, suggests
an o th er option th a t is available for th e tre a tm e n t of a m u lticriteria optim i
zation problem, nam ely to in te rp re t it into a form w ith a single, sc a la r ob
jective function. As it is shown in Section 18.2, several options exist for
scalarization of m u lticriteria optim ization problems.
T he form of a Vector O p tim iza tio n Problem ( V O P ) is in analogy w ith
(15.4)
M in I f ( x ) | h ( x ) = 0
X e R

g(x)<o]
1

(18.1a)

326

18 Optimization strategies

w here f ( x ) is a so-called v e c t o r o b j e c t i v e f u n c t i o n of th e design variables


/fr(x)\
(1 8 .1 b )

f (x )
\ U X)

Problem s w ith m ultiple objective functions are characterized by th e occu


rence of an o b j e c t i v e c o n f l i c t , i.e. n o n e of th e possible solutions allows for
sim ultaneous optim al fulfillm ent of a l l objectives ( denoted by M in in
(1 8 .1 a ) ).

D efinition 1: F u n c t i o n a l - e f f i c i e n c y o r P A R E T O - o p t i m a l i t y [ D .34, D.43 ]


A vector x* X is th e n - an d only th e n - term ed PA R E T O -optim al or
functional-efficient or p-efficient for th e VOP (1 8 .1 ) if no vector x X
exists for which
fj(x)<fj(x*)

for all

je{l,...,m }

fj ( x ) < fj ( x* )

for at least one

an d

j {1,..., m } .

(m

Fig. 18.1 depicts, as an exam ple, a projection from th e tw o-dim ensional de


sign space X into th e objective function or criteria space Y. T he PA R E
T O -optim al solutions th e n lie on th e sections of th e a rc AB ( d X*
d Y * ) . T h e designer m ay now choose one of these solutions depending on
how he or she, from practical considerations, assesses th e relative m erits of
th e two objective functions.

Design space

Fig. 18.1: M apping of a feasible design space into th e criteria space

18.1 Vector, multiobjective or multicriteria optimization

327

D efinition 2: S u b s titu te problem a nd preference fu n c tio n - S c a la riza tio n


o f m u ltic r ite r ia o p tim iza tio n pro b lem s
N onlinear m u lticriteria, vector, or m ultiobjective optim ization problem s can
be scalarized, i.e. reduced to usual optim ization problem s w ith a single,
scalar objective function by form ulating a substitute problem [ D .6, D.14,
D .32, D .44].
T he problem
M in

p[f(x)l

(18.S)

i e R n

is called a scalarized substitute problem for a m u lticriteria optim ization


problem if th e re exists a x X such th a t
p [ f ( x ) ] = Mi n
ieR

p[f(x)]

Here, p is called th e preference fu n c tio n or scalarized objective function.


Preference functions:
- S u m o f w eighted objectives
m
,

n ii
x e R

(18.5a)

w ith weighting factors chosen by th e designer


m
(18.5b)
j= l

P Zj(x)

a)

I
lmin

Fig. 18.2: P reference function

f,
b)
a)
b)

Trade-off-form ulation
M in-m ax-form ulation

Z,

x2

328
- Sum

18 Optimization strategies
o f d is ta n c e fu n c tio n s

m
p [ f ( x ) ] : = ( 2 | f j ( x ) - yj r ) Vr

xeR"

(18)

j=i
w ith th e vector y designating given goal values or dem and levels for crite
ria fj ( j = l , . . . , m ) . Here, th e values of th e com ponents of y an d th e ex
ponent r ( l < r < o o ) are a t th e choice of th e designer.
- C o n s tr a in t-o r ie n te d tr a n s fo r m a tio n (T r a d e -o ff m e th o d )

P [ f ( x )] = fi ( x )

j =

x e Rn

(18.7)

w ith f j ( x ) as th e principal or m ain objective, an d f2 , - - -, fm as secondary


or sides objectives ( constraints ). y^ denotes th e corresponding respondence
levels w hich are chosen by th e designer.
- M in -M a x -fo r m u la tio n

p [ f ( x ) ] := Max[z.(x)]

w ith

z. ( x )

f. ( x ) - f.
J- ,
= -----=
fj

x Rn

_
f. > 0

j =l,...,m

(18.8)

w here lj denote values specified separately by th e designer for each objec


tive function. For fu rth e r details refer to [D.14,D.32].
E x t e n d e d M i n - M a x b y w e ig h te d o b je c tiv e s :

p [f(x )] := Max[wj-f.(x)]
j
3

x e Rn

(18.9)

w ith weighting factors analogously to ( 18.5b) .


Note th a t th e full form ulation of th e M in-M ax problem w ith w eighted ob
jectives is ( substitute (18.9) into ( 18.S) ):
M in (M a x [ w - f ( x ) l )
xeR "

(18.10)

'

T his M in-M ax problem can be given as th e equivalent B o u n d F o r m u l a t i o n


[D .6,D .30,D .31]
M in
P. JLG R n

(3

subject to

w. f ( x ) < /3 .

(18.11)

33

Here, (3 is an additional, scalar p aram eter term ed the b o u n d v a r i a b l e


w hich executes th e ta sk (18.10). Thus, (3 constitutes a variable upper bound
on each of th e weighted objectives (n o w transform ed
into constrain ts)
while a t th e sam e tim e subject to m inim ization since adopted as
th e objec
tive function of th e scalarized optim ization problem (18.11) .

18.2 Shape optimization

329

In fact, th e full set of P A R E T O -solutions, cf. Section 18.1, can be generated


by application of th e preference functions covering an appropriate range of
values for th e weighting factors Wj. Thus, th e designer's choice of values
for Wj is related to th e application of judgm ent in a PA R ETO - approach.

18.2 Shap e o p tim iza tio n


T he te rm shape o p tim iza tio n denotes th e o ptim al shaping of com ponents
by sim ultaneously considering given requirem ents. In order to achieve this
goal, functions have to be determ ined w hich describe th e shap e to b e opti
m ized ( shape fu n c tio n s ) . Hence, in general, shape optim ization problem s
lead to th e form ulation of objective fu n c tio n a ls F , an d sim ilarly, general
co n stra in t operators G have to be considered. As th e shape is continuously
varied during th e optim ization, th e respective m odel (e.g. p artitio n in g into
stru c tu re or shell elem en ts) m ust be ad ap ted accordingly, an d th is often
requires re-discretization.
(a )

In d ir e c t m eth o d s [ D .4, D .5, D .7, D .9, D.21, D .22, D .27, D .28, A .3]

T he above m ethods incorporate tw o s te p s :


1)

D erivation of o p tim a lity c o n d itio n s as necessary conditions for the


optim al design,

2)

F ulfilm en t of th e o p tim a lity c o n d itio n s by m eans of suitable solution


procedures.

T he curv ilin ear coordinates


define th e a rea A of a load-bearing stru ctu
re w ith th e b o undary r . T h e o ptim al shape function R w ith th e com po
n ents RJ ( j = 1 , 2 , 3 ) sh all be determ ined. T h e derivatives dR* / d ( a are
abbreviated as R-ja . In addition, EIN STEIN 's su m m ation convention (see
Section 2.2) will be used.
T he following considerations will be lim ited to such shape optim ization
problem s for which b o th th e optim ization objective and th e constraints can
be expressed in th e form of integrals [ A .3, A .6 , D.5 ]:
M i n F = Mi n

J f ( ( a , Rj , Rj Q ) dA

(j

1,2,3) ,

(18.12a)

w here

J f k( f

,Rj ,Rj, J d A = 0

(k = l , . . . , m ) ,

(18.12b)

J f z ( f , R j , Rj rQ) d A < 0

(Z = 1

r)

(18.12c)

are assum ed to be given.


In order to derive th e necessary conditions for th e present problem , th e in
equality operators ( 18.12c) are first transform ed into equality operators,
using slack variables ril :

330

18 Optimization strategies
J f z ( , Rj , RjfJ d A + ? = 0

(1 =

(18.13)

U sing th e LAGRANGEAN m ultipliers Ak , A; and th e abbreviation

m
* = f ( f ,Rj ,Rj, J + 2 \ f k ( f

k=l

(18.14a)

+ 2 V z U .Rj .Rj, J .
Z= 1

we obtain th e L A G R A N G E -functional as
I = f ^ ( f , R j,R\a , \ , ^ ) d A + 2
A

>7?

(UMb)

Z=1

T he EULER equations for th e variatio n al problem


M i n l U , R j , Rj (Q)Ak , V i 7 z )

(1815)

will now be very briefly set up a n d solved.


Owing to th e dem and th a t th e first v ariatio n <51 has to vanish (<51 = 0 ) for
a rb itrary variations of RJ , we obtain th e following p artia l differential equa
tion including boundary conditions w hen considering th e GAUSSIAN rule of
integratio n an d com ponent-w ise application of th e fu n d a m e n ta l le m m a of
v a ria tio n a l calculus:

r- dR1

fn

1 = 0 ,

V d R \ J ia

<5Rj d r = 0 ,

(18.16)

dR \a

w here na are th e com ponents of th e norm al u n it vector on th e boundary F.


<51 = 0 recovers th e constraints (18.12b) an d (18.13) for arb itrary adm issible
variation of Ak an d A; , an d v ariatio n of rjl yields
2 ^ = 0

(I = 1

r)

(18.17)

By separating (18.13) by m eans of (18.17) into th e cases of active ( 7?; = 0 ,


A; ^ 0 ) an d non-active (?;; =/= 0 , A; = 0 ) inequality operators, th e rjl can be
elim inated as follow s:
Ai f j f i ( f , R j , R j( Q) d A l < 0

(I = 1 , . . . , r ) .

(18.18)

W ith EU LER equations according to (18.16), th e relations (18.18), an d the


equality co nstraints ( 18.12b), all required equations are available for deter
m ining th e unknow n quantities R>, Ak, an d A; for th e present problem.
(b )

D ire c t m eth o d s

[ D.16 , D.47, D.48]

In th e direct m ethods th e shape optim ization problems are transform ed in


to p aram eter optim ization problem s w hich are th e n treated by m eans of
M P-algorithm s according to Fig. 8.6.

18.2 Shape optimization

331

O ne determ ines an o ptim al shape function R* for which the objective func
tional F a tta in s a m inim um
MinF(R)

Rer2

----- F ( R * )

(18.19)

w ith r 2 denoting th e set of all shape functions.


T he feasible v ariatio n al dom ain is defined by the co nstraint operators
H ., G .:
* j

Hi R = ^ i

(i = i , - - , q) >

G j R < Xj

(j = 1 , , ? )

T he unknow n functions R are approxim ated by suitable functions R , socalled shape approxim ation functions.
In recent years, th e progress of CA D -techniques in th e design an d con
struction d ep artm en ts has substantially increased th e im portance of ge
o m e tric a l m od elin g also in application to stru c tu ra l optim ization. Basically,geom etrical m odeling deals w ith com puter-based design an d m an ip u la
tion of geom etrical shapes [ D .8, D.30 ].
T he choice of suitable approxim ation functions for optim al geom etries is
problem -dependent. T he chosen function is to approxim ate th e course to be
followed as precisely as possible, a dem and th a t leads to a large am ount of
shape p aram eters and thus to increased com putational effort. A reduction
of this effort can be achieved by decreasing th e num ber of param eters,
which, however, requires some a-priori knowledge and experience concer
ning th e choice of a given type of approxim ation. If this inform ation does
not exist, optim ization should proceed w ith sim ple approxim ations to be re
fined w ith increasing level of knowledge.
In th e following, we w ill introduce som e of th e m ost im p o rta n t approxim a
tion functions for geom etric m odeling of shapes of co m p o n en ts:
l)

Shape fu n c tio n s depending on a single variable

T his type of approxim ation function is chosen if th e shape optim ization


can be reduced to optim ization of curves R th a t only depend on one coor
din ate ( , i.e. curves th a t can be described by eith er a continuous function
R or by th e sum of single continuous functions w ithin th e defined dom ain.
A general p o ly n o m ia l fu n c tio n describes th e dependence of a shape func
tion R J on th e local coordinate ( in th e following form :
R J ( f , x ) = Xj + x ^ + x 32 + . . . + xn n+1

(18.21)

M onotonically increasing or decreasing functions can be suitably approxi


m ated by polynom ia. However, for n > 3 very undesirable, strong oscillatory
behaviour generally occurs. In addition, a precise representation of particu
larly interesting boundaries, edges, or tran sitio n s is often not possible.

332

18 Optimization strategies

C H E B Y C H E V -f u n c tio n s are polynom ia w ith special properties [A .3]:

R j (e,x) = x1T0(O + x2T1(O + x3T2(O--- ,

( 18-22)

where Tj ( i = 0, ..., n - 1) describe th e CH EBY CH EV -polynom ia from th e


0-th to th e ( n - 1 )-th d eg ree T h e T j-polynom ia depending on ( are calcu
lated as follows for th e range of
< < u :
T0 = l ,

Tj = 2

1 > Ta = 2T1 - T 1 - T 0 ,
M

= ^Tj Tk l - Tk 2

Besides being orthogonal, th e CHEBYCHEV system of polynom ia also pos


sesses th e favourable properties u n if o r m con vergence an d o p t i m a l i t y .
However, even th e CH EBY CH EV -polynom ia only allow for a lim ited pre
cise rep resen tatio n of single dom ains.
A nonlinear, p aram etric B - s p lin e -f u n c tio n is defined by n + 1 control points
w hich define a so-called control polygon (Fig. 18.3). W ith th e exception of
th e startin g point and th e end point of th e control polygon, th e control
points do not lie on th e B-spline curve. T he curve r ( ( ) is defined by
n
r ( 0 =

Z P i Bik ( 0

( 18-ZSa)

i= 0

w ith

Pj

.
1
2
.
vector of th e i-th control point in th e given x ,x coordinate
system ,

B jk( ( )

m ixed function .

T he m ixed functions B; k( ( ) , or base functions of B-splines, are polyno


m ial - p ara m e te r functions of degree k - 1 w hich can be calculated by m eans
of th e following recursive fo rm u la :
Pi

Fig. 18.3: B -spline-curve of degree k = 3 w ith 9 control points

18.2 Shape optimization

333

Pi

n=8

Fig. 18.4: BEZIER-curve of degree n = 8 (9 control poin ts)

( 1 for t . < ( < t. , .


6^(0 = {
1 0 for all oth er (

BikU )

li+k-l

li

( 18-zsh)

Bik_1( 0 + t ti+k; /
Bi+lk_1( 0
i+k li +1

(18.23c)

T he quantities tj a re called course node quantities, and th ey assign th e va


lue of th e p aram eter ( to th e control points an d thus influence th e shape of
th e curve. For m ore details see [D.8].
BEZIER-curves are defined in analogy
B-splines, using n + 1 control p o in ts :
r ( 0 =
w ith

w ith

the

description

of the

2 P i B i k( 0

i=0

Pj

vector of th e i-th control point ,

Bik ( )

scalar m ixed function of degree k .

In contrast to th e non-periodical B-splines, th e p aram eters here range b e t


w een 0 < ( < 1. As is th e case w ith th e B-spline-curves, th e control points
of BEZIER-curves generally do not lie on th e curve, w ith th e exception of
th e first an d la st control point (Fig. 18.4). T he m ixed function Bik( ( ) is a
scalar polynom ial-param eter function, a so-called B ERN STEIN -polynom ium
of k-th degree, w eighted by binom ial coefficients [ D.8 ]:
Bik ( 0 = T ! ( k - i ) ( i ( 1 ^

(im h )

10.0

B: Kj = 6,0

k2 =

2.0

C: K, = 2.0

k2 =

5.0

D: Kj = 2.0

k2 =

2.0

E: Kj = 1.5

k2 =

1.5

k2 =

2.0

G: Kj = 1.0

II
<N
*

ii

k2 =

A: K, = 10.0

o
'ui

18 Optimization strategies

n
r,

334

Fig. 18.5: Influence of th e shape param eters on th e m odified ellipse [ D.13 ]


E llip se fu n c tio n s w ith va ria b le exp o n en ts can successfully be em ployed in
order to determ ine th e shape of boilers or to find optim al notch configura
tions [ D.13 ]. In m ath em atical term s these functions r e a d :

(fP + (fPa /

(18.25a)

vb

w ith th e shape param eters k v k 2 an d th e sem i-axes a , b .


T he p aram etrical representation of th e ellipse equation r e a d s :
x 1 = a ( sin ip) ^ 1 ,

x 2 = b ( cos ip) ^ 2

w ith param eter

ip.

( 18.25b)

Fig. 18.5 illustrates th e influence of th e shape p aram eters k v k 2 on the


shape.

18.3

A u g m e n te d o p tim iz a tio n lo o p b y a d d i t i o n a l s t r a te g ie s
[ D.3 , D.12, D.40]

Fig. 15.4 presents th e basic m odules of an optim ization model. T he sensiti


vity analysis treated in C hapter 17 as well as th e two optim ization stra te
gies m ulticriteria or an d optim ization (C h a p te r 18) contribute m odules th a t
are im plem ented into th e optim ization model, and they thus present im
p o rta n t elem ents of an effective tre a tm e n t of stru ctu ral optim ization pro-

18.3 Augmentation of the optimization loop by additional strategies

335

blem s. D ire c t m eth o d s are especially ap p ro p riate for solving m ulticriteria


an d shape optim ization problem s th e basic procedure of w hich is shown in
Fig. 18.6, w hile shape optim ization problem s can be processed by a m ere
augm entation of th e design model, m u lticriteria optim ization requires a
special evaluation model. Fig. 18.7 illustrates how th e optim ization loop is
augm ented by these additional m odules w ithin th e T h ree-C o lu m n s-C o n cept [D.12] discussed in Section 15.6.
In a sim ilar m anner, o th er m odules can be im plem ented into th e loop, e.g.
for optim ization w ith tim e-dependent p aram etric quantities, or for stocha
stic optim ization problem s. A n im p o rtan t future dem and on th e optim iza
tion process w ill be th e consideration of m ultidisciplinary aspects from the
fields of fluid- or aerodynam ics, therm odynam ics, h eat transfer, m anufactu
ring, etc. In th is context, th e te rm m u ltid is c ip lin a r y o p tim iza tio n has be
come general use [ D.42 ].

Param eter Optiin izatio n Problem


Min p[f(x)]
h;(x) = 0 , i = l,...,m
gj(x) < 0 , j = l,...,n
x, < x < xu

M in f [R(S ,x)]
hj(x) = 0 , i = l,...,q
gj(x) < 0 , j = l,...,p
Xj < X < xu

Fig. 18.6: D irect optim ization strategies

336

18 Optimization strategies

definition evaluation model

data input
determination
OS

decision maker

selection optimization algorithm

y0
initial design

optimal design

stmctural
parameters

yn -> *o

y = const

OS/SOP
Idesign model |
discretization
rule \

X
optimization
algorithm

P,f,g
dp df dg
dx dx dx

optimization model

'shape algorithm"
shape function

Eft)
OS/VOP
preference
model

1H'
structural model
stmctural
analysis

f x,u(x)f ,g[x,u(x)j
evaluatio n m odel

state variable
u(x)

sensitivity
analysis
8u
dx

OS : optimization strategy
SOP: shape optimization
VOP: vector optimization

Fig. 18.7: O ptim ization loop augm ented by m u lticriteria and


sh ape optim ization

D. 2 E x e r c ise s
E x e r c is e D -1 5 /1 6 -1 :
An uncon strain ed optim ization problem is given by th e objective function
f ( Xj , X j ) = 12

+ 4

- 12 Xj x 2 + 2 Xj

----- M in

Xj , x 2 e R.".

a ) D eterm ine th e m inim um of th is function using th e necessary a n d th e


sufficient conditions.
b ) Check th e exact resu lt by m eans of th e PO W ELL-m ethod, startin g w ith
x 0 = ( - 1 , - 2 )T as in itia l point.
c ) Apply also th e algorithm of conjugate g radients according to FLETC H E R -R E E V E S to obtain th e result. Let again x 0 = ( - 1 , - 2 )T b e th e
sta rtin g point.

S o lu tio n :
a ) We are confronted with an unconstrained optimization problem with a conti
nuously differentiable objective function possessing an exact solution.
According to (15.7) the candidate minimum point is obtained from the necessary
conditions
df

dXj

24 Xj - 12 x2 + 2 = 0 ,

^ - = 8 x2 - 12 x1

2 X1

By substituting x2 into the first equation one obtains:


24 Xj - 18 Xj + 2 = 0

>

*
X1 =

The corresponding function value becomes


f* = -JL
3
The HESSIAN matrix is calculated from (15.8)

H (x ) =

24

-1 2

-1 2

= 48

This proves positive definiteness, i.e. a minimum solution has been found.
b)

T h e sta rtin g v ector for th e P O W E L L -m ethod is given as :


\T

= ( - 1 >- 2 )

f0 = 2

338

15 Fundamentals of structural optimization / 16 MP-algorithms

F irst cycle
X

For the first search direction we choose s0 = ( 1 , 0 ) . Thus, we obtain according


to ( 16.1a) :

( -1 + a
+ a

( 1)
-2

By substituting ( 1 ) into the given function


f ( a ) = 12 ( - 1 -f a ) 2 -f 4 ( - 2 ) 2 - 12 ( - 1 + a ) ( - 2 ) + 2 ( - l + a )

da

= 2 4 ( - l + a ) + 24 + 2 = 0

12

which yields
_

il
12

and

f ( x j ) = 1.9167

-2

As second search direction we choose Sj = ( 0 , 1 )


( 1 ), leads to

, which, in accordance with


_ 13

ic2 = Xj + a si = I

12

12 I + aj

(2)

-2 + a
Substitution of ( 2 ) into the given function:
f(a) = 1 2 ( - 1 |) 2 + 4 ( - 2 + a)2 - 1 2 ( - g - ) ( - 2 + a) + 2 ( - g )

da

= 8 ( - 2 + a ) + 13=0
'
'

One thus obtains


( - 13 ^
12

and

_ 13
8

f ( x 2 ) = 1.3542

An additional search direction is determined by means of (16.1d)


( _ 13 \
12

.1 3
' 8 J

j_ \
12

3^
V 8 /

Then it follows
(

1 \

(3 )

Exercise D-15 /16-1

339

Substitution into f ( Xj, x2 ) and re-formulation yields


f() = - ( l + ) ( l 4 - f ) + 4 (-2 + } )1

1 4

) - ( 1

- *

'

From ( 4 ) one determines


( __157 \
147
_ 83
V 49 /

f ( x 3 ) = 1.319728

and

This concludes the first cycle.


Second cycle
X

The second cycle also proceeds from the search direction sQ = ( 1 , 0 ) . We get
( - 147
I +
^
_ 83
49

x4 = x 3 + a 80

2 (-

157
147

(5)

bf
- = 2 4 f - 4 ^ + ) + 12
+ 2 = 0
V 147
/
49

ba

a = 0.1377552

and finally from ( 5 )


( -0.9302720'
and

f ( x 4 ) = 1.092008

-1.6938775
We then formulate
f -0.0833333 \

( -0.9302720 'I

x 5 = x 4 + ocs2

(6)

+
V -1.6938775

I,

0.375

f ( 0) = 12 ( - 0.9302720 - 0.083333 a ) + 4 ( -1.6938775 + 0.375 a ) - 12 ( - 0.9302720 - 0.083333 a ) ( - 1.6938775 + 0.375 a ) +


+ 2 ( - 0.9302720 - 0.083333 a )
J*

dec

a = 0.438567

340

15 Fundamentals of structural optimization / 16 MP-algorithms

From ( 6 ) follows

( - 0.9668191
and

f ( x ) = 0.89566164

-1.5294147
The course of the optimization process clearly shows a very slow convergence to
wards the solution point. We therefore stop the treatment at this point and pro
ceed to c) .
c)

Algorithm of conjugate gradients according to FLETCHER-REEVES

starting point xfl = ( - 1 , - 2 )


We again proceed from the starting
tion is calculated from the gradient

. The starting direc-

/ 2 4Xj - 12XJ +
V f ( x 0) = V f ( x )
l0

8x2 ~ 12xi

The steepest descent direction is


s0

= - V f ( x o) = (

I )

Eq. (16.6) yields for the end of the first step

and thus
f ( 0 ) =

12( - 1 - 2 0 )2 + 4 ( - 2 + 4 o )2 - 12(-1 - 2 a 0) ( - 2 + 4 a 0) + 2 ( - l - 2 a 0) ,

jr
3H
nr0 = 0

a nU= 0.048077

The new point Xj reads


Xj =

( -1.0961 \

and

V f ( Xj ) =

V - 1-8077 )

V - 1-3084 /

The next search direction is calculated with (16.7)


2

|v fM

_ / 2.6140 \
Bl

V 1.3084 J +
( 2.6140 \

( -2.6140 \

(-2.6140 f+ (-1.3084f ( ~ 2
( - 2 ) 2 + 42

/ - 2 \
+ 0.4272
V 1.3084)
V 4 )

^ 4
( 1.7596 \
V 3-0172 J

Exercise D-15 /16-1

341

The next point is determined from


( -1.0961 \
x 2 = x i + a i si =

V -1-8077 )

( 1.7596 ^
+ ai

V 3.0172 )

and correspondingly f ( otj ). The minimum condition


= 0

yields

otj = 0.4334

One obtains
( - 0.3334 \

= ( -05 j

( 0 \

"d

V'(Xl) = ( o J '

Thus, the condition


T
s0 H s j =

/ 24 -1 2 \ ( 1.7596 \
- 2 , 4)1
0
V - 12
8 / \ 3.0172 )

is fulfilled.
Fig. D-l illustrates the single search steps for the FLETCHER-REEVES-method.
It is obvious that this method converges much faster than the POWELL-method.
By suitable modifications, however, convergence of the latter method can be im
proved.

Fig. D -l: Search steps for the FLETCHER-REEVES-method

342

15 Fundamentals of structural optimization / 16 MP-algorithms

E x e rc is e D - 1 5 /1 6 - 2 :
T h e tru ss stru ctu re shown in Fig. D-2 consists of 13 steel bars w ith th e
cross-sectional areas A; ( i = 1 , . . . , 13 ) an d 10 nodal points. A v ertical force
F = 100 kN acts a t node 3.
D eterm ine th e cross-sectional areas in such a way th a t th e w eight of th e
stru c tu re is m inim ized. T h e stresses in th e single bars m ust n o t exceed an
adm issible tensile stress of cr.
= + 150 M P a , and an adm issible com*adm

pressive stress of a

c adm

= -1 0 0 M P a .

As fu rth e r values are given:


l = 1 .0 m , E = 2.1- 10s M P a .
F6 = F
k

x,u

Fig. D -2: P lan e tru ss stru c tu re

a ) F orm ulate th e stru c tu ra l model, an d determ ine th e solutions for dis


placem ents an d stresses.
b ) In order to form ulate th e optim ization problem, define th e objective
function an d th e constraints w hen th e cross-sectional areas are used as
design variables x : = A = ( Aj )T ( i = 1 , . . . , 13).
c ) D eterm ine th e o ptim al solution of th e constrained optim ization problem
by m eans of an ex tern al penalty function approach.

Exercise D-15 /16-2

343

S olution:
a ) The relation between the nodal forces and nodal displacements is established
by means of the displacement method. This will be demonstrated for the forces F3
and F4 acting at node 2 and the corresponding displacements u2 and v2 E q u ilib riu m co n d itio n s :
Equilibrium conditions give the external forces
in terms of the b ar forces at node 2:
F3 = S6 + i / 2 ' S7 - i / 2

S5 ,

(1)
F4 = 1 / 2 S7 + 1 / 2 S5

E lasticity la w :
AI Z S
6 EAg 6

AI
- EA7

J7

J 21
> A L = EAg

(2)

Kinem atics:
AZ6 = 0 ,

AZ7 = T

-J* V2

( 3 )

Substitution of ( 2 ) and ( 3 ) into ( 1 ) yields


V*

F4 J

22

2J

with the element stiffness matrix


As -|- 2 -{2 Ag + A2
K* - F
22 2 T T l

-Ag + Ay

Ag + Ay
Ag + Ay

E
2 / 2 7 K 22

( 4 )

Analogous relations can be derived for the other nodes.


The total stiffness matrix K consists of the single stiffness matrices of the bars;
it relates the external forces to the displacements in the following linear equation:
f = Kv
(5a)
with the displacement vector

V = (u1,V1,U2,V2,U3,V3 ,U4,V4)U5,Vg)T

344

15 Fundamentals of structural optimization / 16 MP-algorithms

the force vector consisting of the 10 nodal forces


f = ( F i , F 2,. . . ,Fio)T ,
and the symmetric total stiffness matrix
k
k 2i

2V21

*11 =

CO

with

12

13

*14 ;

K 23

^ 4 i K 25

K 32

K 33

k 34 j

K 42

K 43

K 44 : K 45

5 1;

K52

K 53

S 4

A2 1 A4 1 A 8

/ 2 A 3 + A4 + As

A 4 - A5

A4 - AS

A4 + A5

A jj + 2 /2 ^ A 12 + A 13

A11

A13

A jj

A13

A11 + A13

-f2 Ag + Aj0 + Ajj

A10 + A11

- Aj0 + Ajj

A10 + A11 J

- A.
K 13 = K 31

4J
'10

K,51

"1 0

~A5
*^ 3 2

As

K 54 = K 45

10

A 5

- A 11

A ll rem aining K y v anish .

5J

Ml

K 55

A io

+ A4 + Ag -f A10

K 33 =

35

4 i

K 22 = see ( 4 ) ,

*^ 23

' A2 + A4 + Ag - Ajo

15

K 22

2V2 Aj + A^ + A4 +
+ Ag + A10

K,55

(5b)

Exercise D-15 /16-2

345

Assuming non-singularity of the stiffness matrix, ( 5a ) allows us to calculate the


displacements of the nodal points Uj, Vj ( i = 1 , . . . , 5 ):
K-1 f

(6)

Thus, the displacements of the end-point of each single bar is established, and we
can now, on the basis of the element stiffness matrices, determine the stresses
within the bars by means of the matrix relation between stresses and displace
ments :
a = Rv
( 7 )
Here, R is a (13 x 10)matrix of the form

_i _

" 2
0

" 2
0

0
J_
2
0
J_
2
0

J_

_1_

2
1

' 2
0

2
0

2
0

_1 _

_1 _

_1 _

~2

~2

2
0

13 x 10

(S )

Substitution of (6 ) into ( 7 ) then yields the relation required for calculating the
stresses:
a = R K 1f

(9 )

The equations for the structural model that is required for the optimization have
now beeen established.
b ) In the following, the equations of the optimization problem shall be set up. In
accordance with the problem formulation, the cross-sectional areas of the bars
shall serve as design variables, i.e. we define
x := A
According to ( 5b ), K depends on the design variables, i.e. K = K ( x ) . Given the
same bar material, weight minimization is equal to volume minimization; the ob
jective fu n ctio n of the structural volume is thus a linear function with respect to
the design variables:
f ( x ) : = V ( x ) = I T . x = f i iXi
i 1

with l-x denoting the bar lengths.

( 10)

346

15 Fundamentals of structural optimization / 16 MP-algorithms

For the bar stresses we formulate the following constraints


Si ( x )
S2 ( x )

i ( x ) tadm

( i = 1 , . . . , 13 ) ,

- 8 c j ( x ) : = cadm ~ j ( x ) ^ 0

(j = 1 , , 1 3 )

(11a)
.

(lib)

Finally, we demand non-negativity for the cross-sectional areas of the b a rs:


Xj > 0

for all

i = 1 , . . . , 13

( 12 )

c ) The constrained optimization problem is solved using an external penalty


function by means of which the task is transformed into an unconstrained pro
blem. With ( 16.11b) we state
'1
j=l
where

(x)])2

(i = 1 , 2 , 3 . . . )

( 13)

in the infeasible domain ,


f gj2( x )
max [ 0 , gj ( x ) ] = <
I
0
in the feasible domain

Here, the choice of a suitable initial value for the penalty parameter Rj is crucial;
for the present task we choose:
Rx = l ( f 5

The unconstrained problem (13 ) can


be solved by means of suitable algo
rithms; in the present case, the
PO W EL L -m ethod of conjugate gra
dients has been used, where a qua
dratic polynomium ( L AGRANGEinterpolation) has proved sufficient
for a one-dimensional minimization.
In addition it could be shown that
different initial designs ( Aj = 100, . . . ,
1000mm 2 ) virtually lead to the same
optimal solution.
The calculation, the scale of which re
quires the use of a computer, yields
the result that the force F is mostly
carried via bars 4 to 8 into the sup
ports 7, 8, 9 (denoted by bold lines in
Fig. D-3). Consequently, the remai
ning bars need very small cross-sec
tional areas only.
Fig. D -3: Optimized truss structure by
changing the cross-sections of
the bars

Exercise D-15 /16-3

347

E x ercise D -1 5 /1 6 -3 :
Fig. D-4 shows a section of a circu lar cylindrical shell C w ith a ring stiffe
n er S. T h e considered p a rt of a boiler is subjected to a constant in tern al
pressure p an d h as a constant in n er tem p eratu re 0 ic.
T he tem p eratu re distribution w ith in boiler an d stiffener has been determ i
ned by m easurem ents; for th e cylindrical section C we assum e a lin ear tem
p erature distrib u tio n over th e thickness w ith th e grad ien t 10 c = const in
the longitudinal direction

c ( z ) e c + z r
iC +

w ith

0C

c =

oc

iC

T he tem p eratu re distribution in th e ring stiffener is assum ed to be constant


over th e thickness, an d is approxim ated in th e m id-plane by a second-order
polynom ium in r w ith th e following fo rm :
s ( r ) = 0 iiC

w ith

( r - 1) ( r + 1 - 2 a ; ) /

(i - w)2

0 oS

e.c )\

= a/ b

Choosing as two design variables th e h a lf th ic k n e ss h o f the rin g stiffe n e r


an d the boiler th ic k n e ss t , th e section is to be dim ensioned w ith respect to
m inim um w eight, subject to the condition th a t th e m axim um reference
stresses in th e ring an d th e stiffener m ust not exceed a prescribed value.

.X .

-2h

U ---------

Fig. D -4 : Section of a ring stiffened circular cylindrical boiler under


pressure u n d th erm al load
a ) D eterm ine th e stress curves by m eans of th e T h e o ry o f S tru c tu re s.
b ) Form ulate th e expressions required for th e optim ization (objective func
tions and constraints), and determ ine th e design dom ain. T h e design va
riables Xj = h and x 2 = t are restricted by upper bound values of 20 and
40 m m , respectively. State th e w all-thicknesses of th e optim al design.

348

15 Fundamentals of structural optimization / 16 MP-algorithms

N u m erica l v a lu e s:
G e o m e try :

b = 0.5 m

a = 0.65 m

L o a d s:
&\c = 170 c

- 0 os = 50 c

l 0 c = 1 c / m m

. P = 2 M Pa

M a te ria l:
a TC = a TS = cxT = 1.11 10~ S/C , Ec = Es = E = 2.1 10s M P a
&c = Ps ~ 6 0.785 104 k g / rr? , v 0.3 , cQ s

= 200 M P a .

S olu tion :
a ) - Structural model and structural analysis

The stress state of the given stiffened boiler structure can be most conveniently
calculated by applying the compatibility between the single parts. Since the re
spective steps for establishing the structural equations have already been described
in detail (see C.13.1/2), only the most im portant aspects will be treated here.
In a first step, we separate the two semi-infinite cylindrical shells from the ring
stiffener. Owing to the different deformations of boiler and stiffener at the interfa
ce point, the required compatibility is induced by yet unknown boundary forces R
and boundary moments M. Each of the substructures shows deformations caused
by temperature and pressure loads (state 0 ), and by the forces R acting at the
boundaries (state 1 ), and the moments M (state 2 ). In the present case, the
parts of the boiler can be idealized as circular cylindrical shells subject to axisymmetric loads (pressure, temperature, boundary force R, and boundary moment
M), and the ring stiffener can be treated as a circular disk subject to internal
pressure, temperature and the radial force R. The boundary moment M of the cir
cular cylindrical shell does not effect the stiffener.
The deformations are calculated from the basic equations for the circular cylindri
cal vessel and for the circular disk ( see C-13-2). After determination of the defor
mations at the points of the substructures, we formulate the compatibility condi
tions

wc = wc 1 + ' o 11 + " c =

Xc =

'

0> +

io ,

11 = s -

( la)

(lb)

where wc and Ug denote the expansions of the vessel and the radial displace
ments of the stiffener, respectively; xc are the corresponding angles of rotation.
Conditions ( la,b ) constitute a linear system of equations for determining the un
known boundary quantities R and M. After some calculation one obtains:

Exercise D-15 /16-3

M = " i ^ R - Kc ( 1 + ' O T C l 0 C

349

( 2b)

with
iC J

12 L

(1 - w)

fe)l

%
e l<
Kc =
E c t'
C 1 2 ( 1 - v 2)

k4

= 3 ( 1 - v2 ) ( | - ) 2
'

Refer to C-13-2 for further details of determining the curves of stress resultants
and deformations.
- S tresses w ith in the p a rts o f the boiler
Cylindrical shell C
- Longitudinal stresses

x x (x ) = ^ { [

Rsin* x + ( m + (1 +

(3a)

( sin

Kx +

cos k x )] e KX - (1 +

- Circumferential
al stresses
V > ( x ) = 2 2 K c { R C S * X + [ M + C1 + v ) T C K C l 0 C

(3b)
( cos k

- sin x ) J e -KX + -^-p

- Reference stress according to VON MISES' hypothesis

orCn = Jy o2
2ipxx +1 a ip
Ring stiffener S ( disk )
- Radial stresses

axx a

tptp

>

(3c)
V

350

15 Fundamentals of structural optimization / 16 MP-algorithms

- Circumferential stresses
2
r
+ ^ ) + ES oTS0 i c [ ^ ( 1 ) +

a<p<p(r ) =

(4 b )

- Reference stresses
2

rr 1

<p<p

/0 + 0

(4c)

- 0 rr 0<p<p

The reference stresses provide the basis for defining the constraints for the
optimization.
b ) Definition of the O ptim ization model
In order to illustrate the design domain, only two design variables are considered
in the following: the half thickness of the stiffener ring Xj = : h and the shell
thickness x2 : t, both of which are combined in the design variable vector-.
T

x = (Xi,X2 )

( 5 )

As stated in the problem formulation, a pure weight minimization problem is to


be solved. We thus require the objective fu n c tio n to be the sum of the weights
of the two parts of the boiler:
(6 )
with the volumes of cylinder and ring stiffener given by
Vc ( x ) = 4 n b ( b - xi ) x 2
Vs ( x ) K 2 n Xj [ a2 - b2 ]

,
for x2 b

We now consider the c o n stra in ts that at each point x or r of the two boiler
parts, the reference stresses Cr have to be smaller than the maximum admissible
stress values:
Cylinder C
Cadm

- 1< 0

(7a)

Stiffener S

(7b)

Exercise D-15 /16-3

351

The two design variables are restricted to the intervals


0 < Xj < 20 m m

0 < x2 < 40 m m

(8 )

Now, the following structural optimization problem (15.4) with the scalar objective
function ( 6 ) and the inequality constraints ( 7a,b ) shall be solved:
Min { f ( x ) | g ( x ) < o }
xeR "

In order to solve this constrained problem, an algorithm can be chosen from MPalgorithms of zeroth, first, and second order. In the case of the actual non-convex
problem (Fig. D-5), the algorithm should perform as simply and robust as possi
ble; here, one of the penalty function methods (e.g. internal penalty function) or
the COMPLEX algorithm by BOX are very suitable zero-order methods (see
[D.24]).
Since only two design variables are considered, the determination of the optimal
design of the current problem can be carried out analytically. As shown in Fig.
D-5, the feasible domain X of the design domain is determined by the active con
straints of the reference stresses in the vessel and the stiffener ring ( 7a,b ), and
by bounding the wall-thicknesses of vessel and stiffener ( 8 ). In addition, the iso
lines of the objective function f ( x ) ( total weight W of the considered parts of
the boiler) are depicted in the diagram.
Fig. D-5 displays the optimal values for the design variables as
Xcpt = ( 4-2 - 11-8 )T .

W
11000 [N]
10000

9000
8000
7000
6000
5000
4000
3000
2000
20

h [mm]

Fig. D-5: Design domain of the


ring-stiffened boiler

352

18 Optimization strategies

which according to equation ( 6 ) allows us to determine the optimal weight a s :


Wopt = 3185.5 N .
The inequality constraint functions gj ( x ) and g2 ( x ) , i.e the reference stresses
crr c and crr s , respectively, become equal to the admissible value (cradm =
200M P a ) at the optimal point, and thus the material is utilized optimally. If we
start the optimization calculation with an initial design
x 0 ( h t )T = (

.20 )T [ mm ] ,

we obtain a weight reduction of approximately 43% at the optimum point.

E x e r c i s e D -1 8 -1 :
Perform a mapping into the criteria space for a vector of the two objective
functions (criteria)
' w
.

a)

b)

'

f2 ( x ) .

x2 - 4 x + 5
1 2 _
, 29
_T x - 5 x +

Show the graphs of the individual objective functions in the design


space, and determine the domain of the functional-efficient set of solu
tions in the design space.
Determine the curves of the functional-efficient solutions in the crite
ria space, using a constraint-oriented transformation (trade-off method).

S o lu tio n :
a ) Presentation of the objective functions in the design space:

Fig. D-6 : Objective functions and domain of the functional-efficient set of solutions

Exercise D-18-1

353

Fig. D-6 shows that the curves have slopes of opposite signs in the dotted area;
according to Def. 1 in Section 18.1 there exist functional-efficient (or PARETOoptim al) solutions of the two functions.
b ) Functional-efficient set of solutions in the criteria space
The Vector Optimization Problem (18J ) can be transformed into a scalar, con
strained optimization problem by minimizing only one of the objective functions,
for instance f j ( x ) , and by imposing upper bounds on the remaining ones (18.7),
e.g.
fj ( x ) Min
V x 6 X ,
subject to

( )
fj ( x )

yj

where fj is denoted the main objective, and f2 , . . . , f secondary objectives. The


present task can be interpreted in such a way that, when minimizing f j , all re
maining components of the objective function are to attain prescribed values
y 2 ,. , ym . These constraint levels illustrate the preference behaviour.
If one is to precisely achieve the constraint levels in (1 ) , the given task corre
sponds to a minimization of the respective LAGRANGE-function (153):
Min

L ( x , p ) : = fx( x ) + 2 P j [ fj ( x ) ~ yj]

(2)

j= 2

with the necessary optimality conditions (15.10)


dL' = J ! + y p . L l o

(i = l , . . . , n )

dx

dL

dpj

fj ( * )

(3a)

(3b)

yj

The optimal values for Xj , . . . , xn and the corresponding LAGRANGEAN multi


pliers P2 , . . . , 3m are then calculated from ( 3a,b ).
For the present problem holds that

f(x) =

x - 4x + 5

fl ( x )
. f2 ( x )

1-2
c i 29
~2 X - 5 x + T -

and we thus choose according to ( 1 )


fj ( x )
subject to

----- > Min ,

f2 ( x ) = Vj

(j = 2 ,---,6)

Using the LAGRANGE-function ( 2 )


L ( x , P ) = fj ( x ) + p j f2 ( x ) - y.

Min

354

18 Optimization strategies

by means of ( 3a,b ) , the optimal values are determined as


R*

4^- = 2 x - 4 + / ? ; ( x - 5 ) = 0
dx
J

d/?j

= 4 -x 2 - 5 x +

- y = 0

yl

J 1-2

* x ?

Ji,2

2xj 1,0 - 4
UJl____

( 4)

5 - x *1(2

= 5

( 5)

Finally, results are listed for different values of yj :

y2

'1,2

y3 = 6.5

x*

3 1.2

r5 2 73
1

/ 42.78/1.22 \

f - 3.773
, 3*
={
3 1*3 \- 0 .2.27

( 37.0/1.0 \

{ -4 .0

= 5 3

8.0

6,

-5.0

y4 = 4.o

11,2

y5

2,0

y6

1.0

3* = oo
k?

6 1,2

= 5 i 72

, no real solution

'

This proves that only the constraint level of 2 < y. < 6.5 leads to unique functio
nal-efficient solutions. Fig. D-7 presents the 3*-values belonging to the different
constraint levels yj in the criteria space. The efficient boundary dY* (solid line)
of Y is valid for non-negative values of 3*-

c r ite r ia s p a c e
y 2 = 8 .0

RZ =-3.73

y3 =6.5

(33i =-4.00
y4 = 4-0
y5=2.0

y6 = 1 0

12 16 20 24 28 32 36 40 44 48

fj

Fig. D-7 : Functional-efficient boundary in the criteria space


The reader should check whether use of f2 as the main objective and ft ( x ) = yj
as a constraint leads to similar results.

Exercise D-18-2

355

E x ercise D -1 8 -2 :
A simply supported column as shown in Fig.
D-8 has variable, circular cross-sections ( radius
r ( x ) ) and is subjected to buckling. The length I
and the total volume V 0 are given.

Ill

a ) Set up a functional which governs the pro


blem of m axim izing the buckling load Fcrft.
b ) Derive the optim ality criterion for the pro
blem.

V0= const
A (x)

c ) Maximize the buckling load for the given


volume V 0 . Derive an equation for the opti
m al cross-section law r = r ( x ).
d ) Compare the optimal buckling load for a
column with variable cross-section with the
buckling load of a column with the same
volume and constant cross-section.

Fig. D -8 : Simply support


ed column
S o lu tio n :
a ) In order to establish a functional, we start with the following expressions des
cribing the problem :
Volume

V = j u [ r ( x ) ]2 dx = V0 = const
o
F_
w rX X +

Differential equation
for column buckling
With

I ( x ) K[ r (x )]
4

follows

Geometrical boundary conditions:

4 F C

-|------ ^

'xx 1 h E

w(0) = 0

(1)
(2 )

( 3)

w ( I) = 0

(4)

r ( x )4

The relations (1 ) to ( 3 ) are transformed into an integral expression of the form


I = J F ( x ) dx

----- > Extremum

( 5)

with F ( x ) as the basic function ( see (6.33) ).


Eq. ( 3 ) yields
2 w

w.xx + V-
r ( x)
with

2
4F
P = tcE

= o

[ r ( x ) ]4 = - p

2 W

(6 )

356

18 Optimization strategies

Substituting ( 6 ) into ( 1 ) and considering that p is independent of x, we can write


I

Tip
Vo

- 1

< >

x 0

Due to V0 = co n st, minimization of the left-hand-side term yields the maximum


value for the force F :
I
[ / - ^ - d x
I
f
x= 0

----- >

Min.

(8a)

,xx

With ( 8a ) we have established an u n c o n stra in ed m a them atical fo r m of our


originally constrained optimization problem.

b)

The basic function according to ( 8b ) read s:


F ( X -W -W, x x ) =

,xx

( 8 b )

In accordance with the rules of the calculus of variation one obtains EULER's
differential equation as the necessary condition:
I

d v i

With

dF \
=0
w /X X / # x x

JJ

dF
w

and

2 (V
1 (
2
I

dF
wf X X

(9 )

1/2

/X X

w
W

1/2

/XX

w :

Multiplication by w leads to

- F * z

Augmentation of the first term of ( 10a ) by ( w xx ) yields

,xx

>xx

Eq. ( 10b ) constitutes the optimality criterion for the present problem.

<10* )

Exercise D-18-2

357

c ) Based on ( 10b ), the optimal cross-sectional radius function r = r ( x ) shall


now be determined.
Using the abbreviation v = / -------
V
w ,xxv +

v ,xxw

we obtain from ( 10b ):

w , xx
0

--------*

( v ,xw -

v w ,x),x =

( U )

Taking into consideration that w ( 0 ) = v ( 0 ) = 0 , ( l l ) can be written


V, x W - V W , x =

(w ).x

( 12)

After integration of (12 ) we have


C2 W2

(13a )

W , x3 x

Since w ( x ) can only be determined up to a multiplying factor ( w ( x ) eigenmode ), one can choose c = 1, and thus
w = -

w , X3X

( 13b)

For the subsequent calculations ( 13b ) is reformulated in the following w ay:


-I/3

w , xx = ~ w

or

( w ,x

),x =

3 ( w2 /3 ),x

2 w , x w ,xx =

>

-I/3

" 2 w ,xw

( 14)

Integration of (14 ) with the integration constant a2 yields:


w(2x = 3 ( - w 2' 3 + a2 )

w, x = J J J a2 - w2/3

( 15)

After transformation of ( 15 ) we obtain:


Jd* =

j
-fz /

dw
a^T T T

(16a )

Now, introducing
w = u

dw = 3 u du

and integrating ( 16a ) , we get

,----------- + C
7/ aa 22 - u- 2

Solution of the right-hand-side integral yields


x = /F |

arcsin

^ 7 a 2 - u^ | + C

( 16b )

358

18 Optimization strategies

Re-substitution and factoring out leads to

The boundary conditions ( 4 ) provide the constants a2 , C :


x = 0:

w = 0

>

x = 1:

w= 0

>

C = 0
2

a2 =
n

w,xx

According to ( 6 ) we have

r 4 = - p2 W

Eq. (1 3b) leads to

w =

W
,xx

==>

=[L

or

= - w 1^ 3

w3173 = (12 W4/ 3 .

(i ,18, !)

w '

Decreasing the power of r in (18 ) to 3 , one obtains


r3 = f y * w

( 19)

Substitution of ( 19 ) into (17 ) then yields the implicit form of the equation for the
optimal cross-sectional radius function
x = / arcs in ------ / 1 ~
7C |
rQ r0 *
\r 0J

1
j

with

r4 =
0
3U3

^
E

. (20)
v

t0 is the largest radius at x = 1 /2 . If we solve the transcendental equation ( 20 )


with respect to r, we obtain

r = r0 f ( x )

( 21)

For the given volume V0 , one obtains r0 from ( 21) and ( 1 )


vo
lr ,
7 i f ( x) dx
0

( 22)

Eq. ( 20 ) finally gives the buckling load


37i3Er!j
F-

'

S i* 5

(23)

d ) Comparison of the optimalbuckling load according to (23 ) with the buckl


ing load of a column with the same volume but uniform circular cross-section.
Proceeding from ( 22 ) we obtain
r o = V o ( 7rJ f2( x ) d x )

( 24)

Exercise D-18-2

359

With ( 24 ) we obtain from ( 23 )


3 7T3 E
Fcnt
. =

16 r

(25)
( * / f 2( x ) d x )

It is demanded that the volume V0 be identical for both the column with constant
and with variable cross-section. Thus, VQ = * r^ I be valid for the column with
constant cross-section.
T
n
4
The area moment of inertia for rk = const
1y = T rk
Thus, we can write
Vo = 4 * 1 /

( 2 6)

By substituting ( 26 ) into ( 25 ), we determine the critical load as follows


F c n .t =
4

.2
Ely It'

= w F_c r i t ^

( t Jf2(x)dx)
crit c o n s t

If the cross-sectional radius function is chosen according to ( 20 ) or ( 21 ), respec


tively, the buckling load increases by 36 % in comparison to the critical load with
constant cross-section. Fig. D-9 shows the column designs.

c n tc o n s t

1.0 Fc n tc o n s t

L 3 6 F c ritc o n St

M a x im u m
b u c k lin g lo a d

>i<

' ~r ~'

A
Fig. D-9: Comparison of buckling loads for simply supported columns with the
same volume and circular cross-section

360

18 Optimization strategies

E x e r c is e D -1 8 -3 :
T h e essential com ponents of a conveyor b e lt d ru m a re th e belt, th e suppor
ting rollers, as well as th e drive an d guide d ru m (Fig. D -10a,b). T h e single
drum s consist of a d ru m casing ( 1 ) and a d ru m bottom ( 2 ). For th e pre
sent ty p e of construction, th e bottom is connected w ith th e sh a ft ( 4 ) v ia a
tension pulley ( 3 ).

Fig. D-10: Belt conveyor a) in teg rate system


b) conveyor b elt drum
c) surface load
T h e d ru m forces F d of th e conveyor b elt induce a surface load p ( i?) (see
Fig. D-lOc), w here th e pressure in th e direction of th e d ru m axis is assum
ed to be constant as a first approxim ation. T h e pressure distribution in th e
circum ferential direction is defined as a load depending on th e circum fe
ren tia l angle i?. T h e m axim um pressure occurs a t d = 0 , and sm aller
values occur a t th e points &= 7T/ 2 .
T h e coefficients of th e chosen pressure distribution
p(

) = p0 + px cos -d + p2 cos 2

resu lt from th e conditions th a t th e resulting pressure forces in th e guide


a re a correspond to th e d ru m forces, i.e.,

7T/2
F d = ra I

j
o

p ( ) cos

di?

(la)

Exercise D-18-3

361

and from th e condition th a t th e load for th e rem aining a rea a tta in s a m ini
m um via a root m ean square form ulation.
Thus, th e d ru m force F d = 650 kN leads to th e load:
p ( t f ) = (0.2117 + 0.3 3 2 6 co stf + 0.1411 c o s 2 tf) [ M P a ]

(lb)

T h e shap e of th e m id-surface of th e d ru m consists of portions of th e drum


bottom w ith constant thickness (id ealized as a circular rin g p la te) an d of
th e drum casing (c irc u la r cylindrical shell). T h e unknow n w all thickness
distributio n t ( ) a t th e tran sitio n s is defined by section-w ise lin ear and
constant approxim ation functions, using th e design va ria b les t j j t j . t j , ^
( see Fig. D - ll):
(V.ti.tj) = -

* ( > . * 3 ** 4 ) =

t ( ^ , t 4)

( t x - t2) ^

(*3

<P3

<PX <<P<<P2 ,

~ *4 )

P2

5 - uf'<P2 <<P<<P 3 ,

= t4

<
p 3 <<p < y

(2)

W e also have to consider u pper an d lower bounds for th e w all-thicknesses:


< t ( <p) < t u ( i p)

10 m m < t. < 200 m m

10 m m < t < 150 m m ,

10 m m < t 3 < 150 m m

10 m m < t 4 < 100 m m

(3a)
(3b)

Fig. D -ll: Shape function for a conveyor b elt drum

362

18 Optimization strategies

a)

E stablish th e optim ization m odeling relations for the task of designing


a conveyor belt drum , w hen th is task is treated as a m u ltic rite ria op
tim izatio n problem w ith th e objectives of m inim izing th e w eight W
and th e m a x im u m reference stre ss crrmax ( c radm = 30 M P a ) .

b)

D eterm ine th e optim al w all-thickness distribution of th e conveyor belt


drum according to Fig. D - l l .

S o lu tio n :
a) The objective functionals read as follows:
F i ( f , ) = / Pgd V W ,

(4a)

F 2 (<f>) = ma x [ o r (<f>,)]

( 4b )

where the reference stress is calculated by means of the VON MISES hypothesis :

(5 )

i? i? - <p v 0 *, , 4 x.tp d

The scalarized objectives dead-weight and m a xim u m reference stress result


from ( 4a, b) as vector functions of the variable shape parameters x =
( t j , t2, t 3 , t 4 ):
fj(*) = W (x) = pgV (x)
f2 ( x ) =

max
[r ( x , 9
<
p^ <. tp < 7t: / 2
0 < i? < *

(6a)

,
,0

(6b)

)]

The present multicriteria optimization problem is treated by means of the con


straint-oriented transformation according to (18.7). For this purpose, the seconda
ry optimization objective (minimization of the maximum reference stresses) is
substituted by the following constraints :
g x( x ) =

g2 ( x ) =

g3 ( x ) =

max

[or ( x ) , 9 , ]

Cradm < 0

[or( x ) , 9 ,] -

oradm < 0

9, < V < 92

max

< V < V-i


0<$ <n
max
0 < iJ < n

[0r( x ) , 9 =<P3,a] -

(7)

^radm 0

with or adm = 30 M Pa
In the structural analysis, the drum bottom is treated as an uncoupled disk-plate
problem, and the drum casing is considered as a circular cylindrical shell. For this
purpose, a special transfer matrix procedure has been used according to Section
13.2. The results were additionally verified by control computations by means of
an FE-software system [ A.21 ].

Exercise D-18-3

363

b) The present shape optimization problem has been solved by means of the opti
mization algorithms SLP and LPNLP (see 16.2.2 a,b ). Fig. D-12a,b illustrates the
efficiency of the above algorithms when using the constraint-oriented transforma
tion as optimization strategy for the treatment of multicriteria optimization pro
blems. Sequential linearization shows fast convergence; if active constraint limits
are imposed, about six to ten linearization steps are necessary, where the gradient
evaluations require the highest computational effort. Fig. D-12b shows that the
rate of convergence of the LPNLP-algorithm is lower than that of the SLP-algorithm.

Fig. D-12:

Optimization graphs in dependence on the number of function evalu


ations
a) Sequential Linearization SLP
b) LAGRANGE-multiplier-method LPNLP

Fig. D-13: Functional-efficient solutions for a conveyor belt drum


Fig. D-13 illustrates the functional-efficient solutions which clearly show the influ
ence of different admissible stresses on the shape of the optimized conveyor belt
drums. Proceeding from the weight-optimal design characterized by active stress
constraints, the increase of the variables tj and t2 leads to a substantial decrease
of the maximum reference stress with only a small increase of weight. Only in
those cases where the range of tj has been used to its full potential, the remain
ing variables gain influence on further stress reductions. Variable t4 in particular
causes a strong increase of weight without reducing the stresses in a decisive
manner.

E x e r c is e D -1 8 -4 :
C om ponent optim ization plays an im p o rta n t role especially in space tech
nology. As a typical exam ple, a satellite th a t is to be brought into its orbi
ta l position should have an extrem e lightw eight design for saving tran sp o r
tatio n costs; even sm all weight savings for single com ponents resu lt in a
su b stantial cost reduction. O ne of th ese com ponents is th e fu e l ta n k of th e
satellite w hich stores th e fuel for th e position control rockets over th e en t
ire life-tim e of th e satellite.
In th e present exam ple, th e calculation an d optim ization of a thin-w alled,
satellite ta n k subjected to constant in tern al pressure shall be dealt w ith ( a
q u a rte r section of th e com ponents can b e considered for reasons of sym m e
try (Fig. D-14) ).

Exercise D-18-4

x2'

'

365

500

400

300

200

100

100

200

300

400

500
x 'f m m ]

Fig. D-14: P rinciple sketch of th e contour of a satellite ta n k w ith boundary


conditions
T h e following design specifications and stre n g th verifications are given:
-

T h e construction space allows for a m axim um outer radius of r max =


436.9 m m . T h e ta n k height
h m a x m ust not exceed a value of 433 m m .

T h e ta n k is subjected to an in tern al pressure p = 34.4 bar. T h e dead


weight is to be disregarded.

T h e half-tan k m ust b e able to store a volum e V w hich is larger th a n a


m inim u m value Vmin = V = 215.2 liter. T h e following volum e con
stra in t is specified:

T he ta n k is m ad e of tita n iu m alloy w ith specified m ate rial ch aracteri


stics :
D ensity

g = 4.5 103 k g / m

YOUNG's m odulus

E = 1.1 10 s M P a ,

POISSON s ratio n

v = 0.3 ,

B reaking stre n g th

c B = 1080 M P a

366
-

18 Optimization strategies
T he stren g th verification is perform ed depending on th e sign of th e
p rincip al stresses in m eridional an d circum ferential direction an d in ac
cordance w ith th e following stress hypotheses of th e state of plane
s tre s s :
1)

If th e principal stresses cr1, <r2 have th e sam e sign, th e reference


stress is calculated from th e m axim um stress:
M ax ( <Tj, <r2 ) < a,r a d m

2)

If th e principal stresses have different signs, th e VON MISES equal


stress hypothesis is to be used:

T h e required thickness t r can be calculated from th e resu ltan ts an d from


th e adm issible reference stress ffradm- For th is purpose we define th e follo
w ing thickness c o n s tra in t:
e> = - T

r s o

T h e task is to m inim ize th e w eight of th e satellite ta n k subject to th e given


constraints by sim ultaneously determ ining a suitable m eridional contour
and a thickness distribution of th e tank.

S o lu tio n :
S tru ctu ra l A n a lysis
In the following, some general remarks shall be made concerning the stru c tu ra l
analysis. The minimum volume of the tank already occupies more than 80% of
the given construction space. This fact demands tank contours that smoothly fol
low their boundaries both at the poles and at the equator. At the equator, the
shape of the tank approaches a cylindrical shell curved in the circumferential di
rection; at the pole, the radii of curvature increase to such a degree during optimi
zation that a very shallow shell emerges. It appears that linear calculations pro
duce large displacements in proximity of the pole, exceeding the wall-thickness by
fax. As the displacements do not occur constantly over the arc length, the radii of
curvature of the deformed structure change substantially. According to the above
theory, the pole area shows a decisively larger curvature in the state of deforma
tion, which results in a violation of the conditions of equilibrium that were origi
nally formulated for the undeformed element. Thus, we used an augmented ap
proach for the structural analysis according to Section 13.2.
Shape O ptim iza tio n
The following shape optimization requires a mathematical description of the tank
shape as a function of free parameters by means of shape functions. The descrip
tion should be characterized in such a way that a large number of admissible
shapes can be achieved with a relatively small amount of parameters. The shape
functions have to comply with the following requirements:

Exercise D-18-4

367

The tank shape should not exceed the specified fitting space.

The meridional contour of the curve must be determined in such a way that
the given minimum volume is attained.

The tangent at the pole must be perpendicular to the axis of rotational


symmetry; the tangent at the equator must be parallel to the axis of rota
tional symmetry.

A curvature undercut (change of sign) is not admissible.

Simple shape functions can be achieved by using a circle or an ellipse as meridio


nal contour. These functions are, however, not suitable for the present problem
because the required tank capacity cannot be fulfilled, and because the equator
curve remains arched. A further disadvantage is the invariability of the curve
shape. The same applies if a so-called CASSINI-curve is used since it does not
possess any free parameters either. The shape is uniquely determined by the vo
lume, and thus optimization calculations for finding a more suitable contour can
not be carried out. Further investigations were carried out using cubic spline-functions as shape functions. These third order polynomials define a continuous curve
up to the second derivative, i.e. the derivative conditions at the pole and the
equator are fulfilled. The splines, however, are disadvantageous in as far as chan
ges of curvature can easily occur, and because the prescribed dimensions of the
construction space cannot be complied with. In addition, they often cause pro
blems in the structural analysis.
In order to avoid the above difficulties, we here choose a modified ellipse function
according to (18.22a) as shape function (see Fig. 18.5). The use of a modified
ellipse has the advantage that meridional shapes always exist for Kj > 1 and
K2 > 2 which satisfy the demands made with respect to the curvature shape and to
the tangent position. In the pole area, the ellipse function is replaced by a polynomium of fourth order.
T rea tm en t as a M u ltic rite ria O ptim iza tio n Problem
In the following, some results shall be presented for a pure shape optimization
and for a simultaneous shape and thickness optimization. When optimizing the
tank, one has to address two conflicting objectives: The weight W of the tank
shall be minimized, and the internal storage volume V shall be maximized. This
multicriteria optimization problem, too, can efficiently be solved by using the con
straint-oriented transformation according to (18.7). For the present problem, the
volume is introduced into the optimization model ( gt ) as an additional constraint
( secondary objective). For various desired volumes V0 , a scalar weight optimiza
tion is carried out, and thus the functional-efficient boundary is determined. The
following design variables will be considered:

or

Xj = Xj

1st ellipse parameter ,

x2 = x2

2nd ellipse parameter ,

x3 = t

thickness of the shell ,

x,. = t-

thickness of the i-th shell element .

The thicknesses t of the shell elements are used as additional design variables in
the transfer matrix procedure. Since the computational effort increases substanti
ally with the number of design variables (> 200), only the geometry variables are

368

18 Optimization strategies

transferred to the optimization algorithm. The optimal thickness distribution is


determined within each functional call, i.e. for each shape design, employing a
F ully-Stressed-D esign (FSD )-strategy, the use of which fulfills a p r io r i the
stress constraints and thus the thickness constraints. By this, the problem can be
reduced to the following design variables and constraints:
Xj = Xj

1st ellipse parameter

x2 = x2

2nd ellipse parameter

V
gj = 1 < 0
vo

volume constraint

,
,

The procedure of the Generalized Reduced Gradients was used as optimization al


gorithm according to [D.1],
Fig. D-15 compares the functional-efficient boundaries of the simultaneous optimi
zation to those of a pure shape optimization (without variation of the thickness).
It can be shown that the integration of an FSD-strategy into the structural analy
sis leads to a substantial improvement of the designs. Three functional-efficient
solutions ( I: V0 = 2001; I I : Vfl = 215.21; I I I : V0 = 2301) are depicted separate
ly, and Fig. D-16 shows the corresponding designs of the shape and the cross-sec
tions, of the radial displacements, of the membrane forces, and of the meridional
bending moments. It is in the responsibility of the decision-maker to choose the
most appropriate design.

f , = m a s s m [k g ]

Fig. D-15: Functional-efficient boundaries for a pure shape optimization


and for a simultaneous shape-thickness-optimization

Exercise D-18-4

Fig. D-16:

Optimal meridional contours and their respective displacements,


membrane forces, and meridional bending moments (s0 = entire
meridional arc length)
(For the shapes I and III only the maximum membrane forces are
shown)

369

370

18 Optimization strategies

E x e r c i s e D -1 8 -5 :
Fig. D-17 shows a spatial sketch of a parabolic radiotelescope reflector w ith
circular aperture. T h e reflector is assem bled from single panels w ith sand
wich stru ctu re, each of which consists of an alum inium honeycom b core and
top layers m ade of Carbon F ibre R einforced P lastics C F R P (see Fig. D -18).

Fig. D-17: Sketch of a parabolic reflector w ith circular ap ertu re an d panel


stru ctu re
T h e panel tre a te d in th e following is assum ed to be plane a n d rectangular,
an d a t several points it is supported a t th e rear tru ss stru c tu re by m eans of
adjusting devices (se e Fig. D-18). T h e num ber n of point-supports predo
m inantly depends on th e desired panel accuracy, i.e., on th e m axim um
tran sv erse displacem ent.

Fig. D-18: Point-supported, rectan g u lar sandw ich panel m ade of C FR P

Exercise D-18-5

371

In addition to the load cases deadweight, wind pressure, and concentrated


forces, the layout must also take temperature effects into account. The opti
mization objctives consist in finding a construction which, for reasons of dy
namics, is as light and stiff as possible, in order to increase the lowest ei
genfrequency of transverse vibrations.
A special rectangular CFRP-sandwich panel design with four, five, and six
point supports which is subjected to a constantly distributed wind load of
p = 1.384 10 4 M P a according to Fig. D-18, shall be investigated in the
form of a multicriteria optimization problem.
S o lu tio n :
O p tim iza tio n m odeling
In order to find optim a l compromise solutions, we choose, according to (18.1) ,
two objective functions, namely weight fj ( x ) : = W and maximum displacement
f2 ( x ) : = wmax. The present multicriteria optimization task will treated using
preference functions like (18.5) to (18.8). As a main strategy, however, a con
straint-oriented transformation (Trade-off method) will be applied (18.7) that de
fines the weight as the primary objective, and the maximum displacement as the
secondary objective, i.e. as constraint:
p [f (x)] = fj(x) = W

with

f2 ( x ) - gx( x ) < y2 = wmax

'

( X)

The design variable vector x here defines the fibre angle oc , the ply thicknesses
t j , the core height hc , and the sides ratio b / a :
x

= ( a i > - >a n i

i - > ! hc ; )

(2)

The panel weight is the sum of the layers and of the core:
f j ( x ) : = W = g a b { h c pc + i I iti [pF.<pF, + pM. ( l - <PF i)]}

(3 )

with pp M c defining the density of fibres, matrix, and core, respectively, and <pF.
denoting the fibre volume fraction.
The optimization modeling is also augmented by a number of inequality con
straints like a fibre breakage criterion, a fibre bonding criterion for the single
layers as well as a shear failure criterion for the core [B.9,B.10]. In addition, the
design variables are bounded by the following upper and lower constraint values:
0" < ock < 90'
^ min

^ max

(4 )

h,,
^ ll,, ^ h r
^ mill
c.
^,
S tru ctu ra l ana lysis
The maximum displacement wmax as the secondary objective fu n c tio n is here
determined from the following system of equations, using the FE-program system
ANSYS [A.21]:

372

18 Optimization strategies

( 5)

K(x)u = r

with

" max

K(x)

symmetrical total stiffness matrix as a function of the design


variable vector x ,

vector of displacements ,

load vector .

[mm]

( * ) =- 0.39

- ----- I( - at=0
i-1-MM-uuumii-i*=22.6mm
0.035 t=0.2 i.m
mm
(b\* - 0.450.030
\a )
0.040

0.025

/Vk= 5o- Va) =0.45


> 42.
x jx x
hj=72.6mm X*x jx x tj= 1.87mm^

t?=o.7:
mn

0.020

0.015
0.010

0.005
0.000

t----- 1------P ^ 'hk


0

20

40

60

80

100

120

140

160

W p^

Fig. D-19: Functional-efficient solutions of a panel supported at six points


a j [d e g r e e ]

W [N]
Fig. D-20: Optimal fibre angle

as a function of weight W

Exercise D-18-5

373

b /a

[-]

4 - p o in t su p p o r t
- 5 - p o in t su p p o r t
------6 - p o in t su p jo r t

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

'

-------------- -------------- i -----------1i - -o

40

80

120

160

W [N]
Fig. D-21: Optimal ratio b / a as a function of weight W
R e su lts
Some optimization results shall be presented and interpreted in the following. Fig.
D-19 shows the functional-efficient solutions of a panel supported at six points,
including different fibre orientations as well as layer - and core thicknesses. Fig.
D-20 illustrates a panel with one core thickness only, where the optimal fibre an
gles depend on the weight. It can be shown that, for a weight > 80 N, the fibre
angle is nearly constant for all types of support. The fibre angle is equal to about
45 in the case of four or six supporting points, whereas it is 30 for five sup
ports. According to Fig. D-20, a substantial change of the fibre angle occurs at the
panels with five and six point-supports and with a weight less than 60 N and
80 N, respectively. The dependence of the optimal ratio b/a on the weight is shown
in Fig. D-21. For each optimal weight, the panel with four supporting points dis
plays an almost constant ratio of w 1, while b /a ranges between 0.4 and 0.6 for
the other cases of support of the panel. The above results illustrate the importan
ce of the optimization investigations as they present an important decision tool to
the engineer for choosing a best possible design.

R e fe r e n c e s

The present work is based on the contents of the following two volumes ( in
German ):
[ETl] ESCHENAUER, H.; SCHNELL, W.: Elastizitatstheorie I - Grundlagen,
Scheiben und Platten. 2. Auflage, Mannheim, Wien, Zurich: Bl-W issenschaftsverlag 1986, 277 pages.
ESCHENAUER, H.; SCHNELL, W.: Elastizitatstheorie II - Schalen. Mann
heim, Wien, Zurich: BI-W issenschaftsverlag 1983, 269 pages.
[ET2] ESCHENAUER, H.; SCHNELL, W.: Elastizitatstheorie - Grundlagen, Flachentragwerke, Strukturoptimierung. 3. vollstandig iiberarbeitete und erweiterte Auflage, Mannheim, Leipzig. Wien, Zurich: Bl-W issenschaftsverlag
1993, 491 pages.
ESCHENAUER, H.; SCHNELL, W.: Elastizitatstheorie - Form el- und Aufgabensammlung, Mannheim, Leipzig. Wien, Zurich: BI-Wissenschaftsverlag
1994, 279 pages.

F u n d a m e n ta ls o f e la s t ic it y
C h a p te r 2 t o 7

[A.l]

BATHE, K.-J.: Finite Element Procedures in Engineering Analysis. Engle


wood Cliffs: Prentice Hall 1982

[A.2]

COOK, R.D.; MALKUS, D.S.; PLESHA, M.E.: Concepts and Applications of


Finite Element Analysis. 3rd ed. New York: Wiley & Sons 1989

[A.3]

COURANT, R.; HILBERT, D.: Methoden der Mathematischen Physik (in


German ). Berlin, Heidelberg, New York: Springer 1968

[A.4]

FLUEGGE, W.: Tensor Analysis and Continuum Mechanics. Berlin, Heidel


berg, New York: Springer 1972

[A.5]

FUNG, Y.C.: Foundations of Solid Mechanics. Englewood Cliffs: Prentice


Hall 1965

[A.6]

GALERKIN, B.G.: Reihenentwicklungen fur einige Falle des Gleichgewichts


von Platten und Balken ( in Russian ). Petrograd: Wjestnik Ingenerow 1915

[A.7]

GREEN, A.E.; ADKINS, J.E.: Large Elastic Deformations. Oxford: Claren


don Press 1970

[A.8]

GREEN, A.E.; ZERNA, W.: Theoretical Elasticity. 2nd ed. Oxford: Claren
don Press 1975

[A.9]

LANGHAAR, H.L.: Energy Methods in Applied Mechanics. New York, Lon


don: John Wiley 1962

376

References

[A.10] LOVE, A.E.H.: A Treatise on the Mathematical Theory of Elasticity. 4th


ed. New York: Dover Publ. 1944
[A.ll] LURIE, A.I.: Raumliche Probleme der Elastizitatstheorie. Berlin: Akademie-Verlag 1963
[A.12] MARGUERRE, K.: Ansatze zur Losung der Grundgleichungen der Elastizi
tatstheorie ( in German ). ZAMM 35 ( 1955 ) 242 - 262
[A.13] NOWACKI, W.: Thermoelasticity. London: Peigamon Press 1986
[A.14] RITZ, W.: Uber eine neue Methode zur Losung gewisser Variationsprobleme der mathematischen Physik (in German). J. Reine Angewandte
Mathematik 135 (1908 ) 1
[A.15] SOKOLNIKOFF, I.S.: Mathematical Theory of Elasticity. New York, To
ronto, London: McGraw-Hill 1956
[A.16 ] TIMOSHENKO, S.; GOODIER, J.N.: Theory of Elasticity. 3. ed. New York,
St. Louis, Toronto, London: McGraw-Hill 1970
[A.171 WANG, Ch.-T.: Applied Elasticity. New York, Toronto, London: Mc
G raw -H ill 1953
[A.18] WASHIZU, K.: Variational Methods in Elasticity and Plasticity. 3rd ed.
Oxford, New York, Toronto, Sydney, Braunschweig: Pergamon Press 1982
[A.19] WEMPNER, G.: Mechanics of Solids. New York, Toronto, London:
McGraw-Hill 1973
[A.20] ZIEGLER, F.: Mechanics of Solids and Fluids. New York, Heidelberg, Ber
lin: Springer 1991
[A.21] N.N.: ANSYS User's Manual for Revision 5.0. Swanson Analysis Systems,
Inc. Johnson Road, Houston, TX / USA, . . .

P la n e lo a d - b e a r in g str u c tu r e s
C h apter 8 to 10

[B.l]

ESCHENAUER, H.: Thermo-elastische Plattengleichungen; Beulen einer


Kragplatte. TH Darmstadt: Dr.-Thesis 1968 (in G erm an)

[B.2]

GREEN, A.E.; ZERNA, W.: Theoretical Elasticity. 2nd ed. Oxford: Claren
don Press 1975

[B.3]

JAHNKE, E.; EMDE, F.; LOSCH, F.: Tafeln hoherer Funktionen. 7. Aufl.
Stuttgart: Teubner 1966

[B.4]

MALVERN, L.E.: Introduction to the Mechanics of a Continuous Medium.


Englewood Cliffs: Prentice Hall 1969

[B.5]

MARGUERRE, K ; WOERNLE, H .-T .: Elastic Plates. Waltham, Toronto,


London: Blaidsdell 1969

[B.6]

MUSKHELISHVILI, N.I.: Einige Grundaufgaben zur mathematischen Ela


stizitatstheorie. Miinchen: Hase 1971

[B.7]

PLANTEMA, F.J.: Sandwich Constructions. New York, London: Wiley &


Sons 1966

[B.8]

TIMOSHENKO, S.; WOINOWSKY- KRIEGER, S.: Theory of Plates and


Shells. 2nd ed. New York, Toronto, London: McGraw-Hill 1959

References
[B.9]

377

TROITSKY, M.S.: Stiffened Plates - Bending, Stability and Vibrations.


Amsterdam, Oxford, New York: Elsevier Scientific 1976

[B.10] TSAI, S.W.; HAHN, H.T.: Introduction to Composite Materials. Westport,


Conn.: Technomic Publishing 1980
[B.ll]

VINSON, J.R.; SIERAKOWSKI, R.L.: The Behavior of Structures Compo


sed of Composite Materials. Dordrecht, Boston, Lancaster: Martinus Nijhoff
Publ. 1986

C u r v e d lo a d - b e a r in g str u c tu r e s
C h a p te r 11 to 14

[C.l]

AXELRAD, E.L.; EMMERLING, F.A.: Flexible Shells. Heidelbeig, New


York, Tokio: Springer 1984

[C.2]

DIKMEN, M.: Theorie of Thin Elastic Shells. Boston, London, Melbourne:


Pitman 1982

[C.3]

DONNELL, L.H.: Beams, Plates, and Shells. New York: M cGraw-Hill 1976

[C.4]

FLUEGGE, W.: Stresses in Shells. New York, Heidelberg, Berlin: Springer


1973

[C.5]

FLUEGGE, W.: Tensor Analysis and Continuum Mechanics. Berlin, Heidel


berg, New York: Springer 1972

[C.6]

FUNG, Y.C.; SECHLER, E.E.: Thin-Shell Structures. New York: Prentice


Hall 1974

[C.7]

GECKELER, J.W.: Zur Theorie der Elastizitat flacher rotationssymmetrischer Schalen ( in German ). Ing.-Arch. 1 (1930 ) 255 - 270

[C.8]

GIBSON, J.E.: Thin Shells, Computing and Theory. Oxford, London, New
York, Paris: Pergamon Press 1980

[C.9]

GOLDENVEIZER, A.L.: Theory of Elastic Thin Shells. Oxford, London,


New York, Paris: Pergamon Press 1961

[C.10] GOULD, P.L.: Analysis of Shells and Plates. Berlin, Heidelbeig, New York:
Springer 1988
[C.ll] GREEN, E.; ZERNA, W.: Theoretical Elasticity. 2nd ed. Oxford: Clarendon
Press 1975
[C.12] KOITER, W.; MIKHAILOV, K.G.: Theory of Shells. Amsterdam, New
York, Oxford: Proc. IUTAM Symp., North-Holi. Publ. Comp. 1980
[C.13] KRATZIG, W.B.: Thermodynamics of Deformations and Shell Theory.
R uhr-U ni Bochum, Inst. f. Konstr. Ing. Bau (1971) Techn. Wiss. Mitt.
7 1 -3
[C.14] KRATZIG, W.B.: Introduction to General Shell Theory. In: Thin Shell
Theory - New Trends and Applications, ed. by W. OLSZAK. Wien, New
York: Springer 1980, 3 -6 1
[C.15] MARGUERRE, K.: Zur Theorie der gekriimmten Platte grofier Formanderung ( in German ). Proc. 5th Int. Congr. Appl. Mech. ( 1939 ) 93 - 101
[C.16] MORLEY, L.S.: An Improvement on DONNELL's Approximation for
Thin-W alled Circular Cylinders. Quart. Journ. Mech. and Appl. Math. XII
(1959 ) 89 - 99

378
[C.17]

References
NAGHDI, P.M.: Foundations of Elastic Shell Theory. In: Progress in Solid
Mechanics, Vol. IV ed. by SNEDDON/HILL. Amsterdam: North-Holi.
Publ. Comp. 1963

[C.18] NIORDSON, F.I.: Shell Theory. Amsterdam, New York, Oxford: North Holland Series in Appl. Math, and Mech. 1985
[C.19] NOVOZHILOV, V.V.: The Theory of Thin Shells. Groningen: Noordhoff
1970
[C.20] REISSNER, E.: Stresses and Small Displacements of Shallow
Shells. J. Math. Phys. 25 (1946 ) 80 - 85, 279 - 300; 27 (1948),
(1959 ), 16 - 35

Sherical
240; 38

[C.21] SHIRAKAWA, K.; SCHNELL, W.: On Some Treatment of the Equations of


Motion for Cylindrical Shells Based on Improved Theory. Ing.-Arch. 53
(1983 ) 275 - 63
[C.22] SINARAY, G.C.; BANERJEE, B.: Large Amplitude Free Vibrations of Shal
low Spherical Shell and Cylindrical Shells. Int. J. N on-linear Mech. 20
(1985) 69-78
[C.23] SOEDEL, W.: Vibrations of Shells and Plates. New York, Basel: Marcel
Dekker Inc. 1981
[C.24] TIMOSHENKO, S.; WOINOWSKY-KRIEGER, S.: Theory of Plates and
Shells. 2nd ed. . New York, Toronto, London: McGraw-Hill 1959
[C.25] ZIENKIEWICZ, O.C.: The Finite Element Method in Engineering Science.
New York, London, McGraw-Hill Vol. 1: Basic formulation and linear Pro
blems 4th ed. 1988 Vol. 2: Solid and fluid mechanics, dynamics and nonlinearity 4th ed. 1991

S tr u c tu ra l o p tim iz a tio n
C h a p te r 15 t o 18

[D.l]

ABADIE, J.; CARPENTIER, J.: Generalization of the WOLFE Reduced


Gradient Method to the Case of Nonlinear Constraints. In: FLETCHER, R.
( Ed. ) : Optimization. New York: Academic Press 1969, 3 7 - 4 8

[D.2]

ARORA, J.S.: Introduction to Optimum Design. New York, Toronto, Lon


don: M cGraw-Hill 1989

[D.3] ATREK, E.; GALLAGHER, R.H.; RAGSDELL, K.M.; ZIENKIEWICZ, O.C.


( E d s.): New Directions in Optimum Structural Design. New York, Toronto:
John Wiley & Sons 1984
[D.4]

BANICHUK, N.V.: Optimality Conditions and Analytical Methods of Shape


Opimization. In: HAUG, E.J./CEA, J. ( eds.): Optimization of Distributed
Parameter Structures. Alphen aan den Rijn: Sijthoff und Noordhoff, 1987

[D.5]

BANICHUK, N.V.: Introduction to Optimization of Structures. New York,


Berlin, Heidelberg: Springer 1990

[D.6]

BENDS0E, M.P.; OLHOFF, N.; TAYLOR, J.E.: A Variational Formulation


for Multicriteria Structural Optimization. J. Struct. Mech. 11 (1983)
523 - 544

References

379

[D.7]

BERNADON, M.; PALMA, F.J.; ROUSSELET B.: Shape Optimization of an


Elastic Thin Shell under Various Criteria. J. Structural Optimization 3
(1991) 7 - 21

[D.8]

BRAIBANT, V.; FLEURY, C.: Shape Optimal Design using B-splines.


Comput. Methods Appl. Mech. Eng. 44 (1984 ) 247 - 267

[D.9]

CARMICHAEL, D.G.: Structural Modelling and Optimization. New York,


Toronto: John Wiley & Sons 1987

[D.10] CHENEY, E.W.; GOLDSTEIN, A.A.: NEW TON-Method and Convex Pro
gramming and CHEBYSCHEV Approximation. J. Num. Math. 1 (1959)
253 - 268
[D.ll] DAVIDON, W.C.: Variable Metric Method for Minimization.
Research and Development Report ANL-5990, 1959,

A.E.C.

[D.12] ESCHENAUER, H.: The Three C olum ns for Treating Problems in Opti
mum Structural Design. In: BERGMANN, H.W. ( e d .): Optimization: Me
thods and Applications, Possibilities and Limitations. Berlin, Heidelberg,
New York: Springer 1989, 1 - 2 1
[D.13] ESCHENAUER, H.: Shape Optimization of Satellite Tanks for Minimum
Weight and Maximum Storage Capacity. J. Structural Optimization 1
(1989 ) 171 -180
[D.14] ESCHENAUER, H.A.; KOSKI, J., OSYCZKA, A.: Multicriteria Design Opti
mization. Berlin, Heidelberg, New York: Springer 1990
[D.15] ESCHENAUER, H.A.; SCHUHMACHER, G.; HARTZHEIM, W.: Multidisci
plinary Design of Composite Aircraft Structures by LAGRANGE. In: Com
puters & Structures, 44, 4 (1992 ), 877 - 893
[D.16] ESCHENAUER, H.; WEINERT, M.: Structural Optimization Techniques as
a Mathematical Tool For Finding Optimal Shapes of Complex Shell Struc
tures. In: GIANESSI, F. (Ed.): Nonsmooth Optimization Methods and Ap
plications. Switzerland, Australia, Belgium: Gordon and Breach 1992,
135 -153
[D.17] FIACCO, A.V.; McCORMICK, G.P.: Computational Algorithm for the Se
quential Unconstrained Minimization Technique for Nonlinear Program
ming. J. Management Sci. 10 (1964 ) 601 - 617
[D.18] FLETCHER, R.; POWELL, M.J.D.: A Rapidly Convergent DescentMethod
for Minimization. Computer J. 6 (1963 ) 163 -168
[D.19] FLETCHER, R.; REEVES, C.M.: Function Minimization by Conjugate Gra
dients. Computer J. 7 ( 1964 ) 149 -154
[D.20] GRIFFITH, R.E.; STEWART, R.A.: A Nonlinear Programing Technique for
the Optimization of Continuous Processing Systems. J. Management Sci. 7
(1961) 379 - 392
[D.21] HAFTKA, R.T.; GURDAL, Z.; KAMAT, M.P.: Elements of Structural Opti
mization. Dordrecht, Boston, London: Kluwer Academic Publishers, 2nd ed.
1990
[D.22] HAUG, E.J.; ARORA, J.S.: Applied Optimal Design. New York, Toronto:
John Wiley & Sons 1979
[D.23] HESTENES, M.R.; STIEFEL, E.: Methods of Conjugate Gradients for Sol
ving Linear Systems. J. Res. Nat. Bur. Std. B49 (1952 ) 409

380

References

[D.24] HIMMELBLAU, D.M.:


McGraw-Hill 1972

Applied

Nonlinear

Programming.

New

York:

[D.25] KIEFER, J.: Sequential Minimax Search for a Maximum. Proc. Am. Math.
Soc. 4 (1953 ) 105 -108
[D.26] KUHN, H.W.; TUCKER, A.W.: Nonlinear Programming. In: NEYMAN, J.
( Ed . ): Proceedings of the 2nd Berkeley Symposium on Mathematical Stati
stics and Probability. University of California, Berkeley 1951, 481 - 492
[D.27] LUKASIEWICZ, S.: Local Loads in Plates and Shells. Alphen aan den Rijn:
Sijthoff and Noordhoff 1979
[D.28] MROZ, Z.; DEMS, K.: On Optimal Shape Design of Elastic Structures. In:
ESCHENAUER, H.; OLHOFF, N. ( ed s.): Optimization Methods in Structu
ral Design. Mannheim, BI-W issenschaftsverlag 1983, 224 - 232
[D.29] OLHOFF, N.; TAYLOR, J.E.: On Structural Optimization. J. Appl. Mecha
nics, 50 (1983 ) 1139 -1151
[D.30] OLHOFF, N.; LUND, E.; RASMUSSEN, J.: Concurrent Engineering Design
Optimization in a CAD Environment. In: HAUG, E.J. ( e d .): Concurrent
Engineering Tools and Technologies for Mechanical System Design. New
York: Springer-Verlag 1993, 523 - 586
[D.31] OLHOFF, N.; LUND, E.: Finite Element Based Engineering Design Sensiti
vity Analysis and Optimization. In: HERSKOVITS, J. ( ed. ): Advanced in
Structural Optimization. Dordrecht, The Netherlands: Kluwer Academic Pu
blishers 1995, 1 - 4 5
[D.32] OSYCZKA, A.: Multicriterion Optimization in Engineering. New York, To
ronto: John Wiley & Sons 1984
[D.33] PAPALAMBROS, P.Y.; WILDE, D.J.: Principles of Optimal Design. Cam
bridge, New York, Melbourne, Sydney: Cambridge University Press 1988
[D.34] PARETO, V.: Manual of Political Economy. Translation of the French edi
tion by A.S. SCHWIER (1927). London-Basingslohe: The McMillan Press
1972
[D.35] PARKINSON, A.; WILSON, M.: Development of a Hybrid SQP-GRG Al
gorithm for Constrained Nonlinear Programming. Proc. ASME Design En
gineering Technical Conferences, Columbus/Ohio 1986
[D.36] PIERRE, D.A.; LOWE, M.J.: Mathematical Programming via Augmented
Lagrangians. London: Addison Wesley 1975
[D.37] POWELL, M.J.D.: An Efficient Method for Finding the Minimum of a
Function of Several Variables without Calculating Derivations. Computer J.
7 ( 1964 ) 155 -162
[D.38] POWELL, M.J.D.: VMCWD: A FORTRAN Subroutine for Constrained Op
timization. University of Cambridge, Report DANTP 1982/NA4
[D.39] ROZVANY, G.I.N.: Structural Design via Optimality Criteria. Dordrecht,
Boston, London: Kluwer Academic Publishers 1989
[D.40] SCHITTKOWSKI, K.L.; HORNLEIN, H.: Numerical Methods in FE -B ased
Structural Optimization Systems. ISNM-Series, Zurich: Birkhauser 1993
[D.41] SCHMIT, L.A.; MALLET, R.H.: Structural Synthesis and Design Parame
ters. Hierarchy Journal of the Struct. Division, Proceedings of the ASCE,
Vol. 89, No. 4, 1963, 269 - 299

References

381

[D.42] SOBIESZCZANSKI-SOBIESKI, J.: Multidisciplinary Optimization for Engi


neering Systems: Achievements and Potentials. In: BERGMANN, H.W.
( E d . ): Optimization: Methods and Applications, Possibilities and Limi
tation. Berlin, Heidelberg, New York: Springer 1989, 42 - 62
[D.43] STADLER, W.: Preference Optimality and Application of PARETO-Optimality. In: MARZOLLO/LEITMANN ( E d s.): Multicriterion Decision Ma
king. CISM Courses and Lectures. New York: Springer 1975, 125 - 225
[D.44] STADLER, W.: Multicriteria Optimization in Engineering and in the Scien
ces. New York, London: Plenum Press 1988
[D.45] SVANBERG, K.: The Method of Moving Asymptotes - A New Method for
Structural Optimization. Int. J. Num. Methods in Eng. 24 (1987 ) 359 - 373
[D.46] VANDERPLAATS, G.N.: Methods of Mathematical Optimization. In:
BERGMANN, H.W. ( E d .): Optimization: Methods and Applications, Possi
bilities and Limitations. Berlin, Heidelberg, New York: Springer 1989, 22 - 41
[D.47] ZIENKIEWICZ, O.C.; CAMPBELL, J.S.: Shape Optimization and Sequential
Linear Programming. In: GALLAGHER/ZIENKIEWICZ ( E ds.): Optimum
Structural Design. Chichester, New York, Brisbane, Toronto: John Wiley &
Sons 1973, 109 -126
[D.48] ZYCZKOWSKI, M.: Recent Advances in Optimal Structural Design of
Shells. European J. of Mechanics, A/Solids 11, Special Issue (1992 ) 5 - 24

Subject index

A-conjugate directions
312
AIRY Is stress function
50, 93, 123, 235
Algorithm of conjugate gradients
313, 340
Aluminiumhoneycomb core
370
Analogy disk-plate
51
Antisymmetrical tensors of second order
11
Arc element, length of
204
Area of a surface element
204
Auxiliary variable method
324
Axisymmetrical state of stress
98
Axisymmetrical loads
216, 224
-, circular cylindrical shell
225
spherical
226
-, conical shell
226
B

B-spline-functions
332
Barrier function
315
Base, covariant
8
, contravariant
8
oblique
53
Base vectors, covariant
7, 13, 203
contravariant
9, 204
Basic theory of shells
209
Behavioral constraints
305
BELTRAMI differential equation
87
BELTRAMI-MICHELL s equations
49
Bending angle
214, 223, 225
Bending theory of circular cylindrical shell 233
-, of shells of revolution
222
BERNSTEIN-polynomium
333
BESSEL function
183
BESSEL's differential equations
105
BETTI, theorem
46
B^ZIER-curves
333
BFGS-formula
320
Bipotential equation
50ff, 93ff
Bipotential operator
15
Boiler equation
225
Boiler formula
217, 218
Boiler structure, stiffened
348
Bound method
328
Bound variable
328
Boundary conditions 103, 117, 147, 158, 234,
242, 244
-, elastically supported
158
-, NAVIER s
103
plate with mixed
170
Boundary disturbances of circular cylindrical
shells
228
fast decaying
237
slowly decaying
238

Boundary of a hole, equilibrium


Boundary-value problem, first

, mixed
-, second
BOUSSINESQ's formulas
BROYDEN
Buckling load, maximizing the
optimal
Buckling modes

147
48
48
48
91
320
355
355
189, 194

c
Carbon Fibre Reinforced Plastics CFRP
37
-, plate made of CFRP
118 ff
-, circular disk made of CFRP
139
Cartesian coordinates
25, 32, 35, 106, 107
isotropic disk
93
-, plates
100
Casing with toroidal shell shape
260
CASSINI-curve
367
CASTIGLIANO, theorems
45, 264
CASTIGLIANO and MENABREA principle 45
CAUCHY s formula
20
CHEBYCHEV-functions
332
CHRISTOFFEL symbols
14, 17
- of the first kind
14
- of the second kind
14
- in surface theory
205
Circular plate on elastic foundation
180
-, centre-supported
184
-, thin
195
Circular toroidal shell
260
Compatibility conditions
31, 73
Complementary energy, specific
41
Complementary work
41
COMPLEX algorithmby BOX
351
Complex solution method
145
Complex stress function
97ff
Composite materials
118
multilayer
119
Compression modulus
34
Conditions for a minimum
307
Conformal mapping
145
Conical shell
218, 221, 226
-, boundary disturbances of
228
Conical surface
247
Constitutive equations
214, 221
-, isotropic shells
213
Constitutive laws of linear elastic bodies
31
Constrained optimization problem
356
Constraint operator
329
Constraint-oriented transformation 328, 352,
363
Constraints
303
-, active
306

384

Subject Index

- , b e h a v io ra l
305
- , g e o m e tric al
314
- , p ro b le m s w ith
310
- , p ro b le m s w ith o u t
8 210
C o n tra v a ria n t b ase
204
C o n tra v a ria n t b a se vectors
9, 204
C o n tra v a ria n t c o m p o n e n ts
332
C o n tro l polygon
C on v ey o r b e lt d ru m
360
C o o rd in a te tra n s fo rm a tio n
21
25, 32, 35, 106, 107
C o o rd in a te s, C a rte s ia n
c u rv ilin e a r
13, 25, 36, 105, 106
c y lin d ric a l
16, 25, 30
e llip tic a l-h y p e rb o lic a l
63, 70, 76
obliqu e
53, 60
p o la r
98, 106, 112
sp h erica l
26, 30
113
C ou p led d isk -p la te p ro b lem
8
C o v a ria n t b ase
14
C o v a ria n t d e riv a tiv e s
9
C o v a ria n t m e tric c o m p o n e n ts
203
C o v a ria n t m e tric ten so r, co m p o n e n ts o f
C rite ria space
326
305
C rite rio n fu n c tio n
C u rv a tu re c o m p o n e n ts
204, 205
C u rv a tu re , G A U SSIA N
205, 208, 210, 249,
- , m ean
205, 208, 249
213
- , s h ea r-rig id sh ell w ith w eak
- , te n so r of
204, 249
C y lin d ric a l s h ell
218, 220, 225, 226, 264
233
- , b e n d in g th e o ry
202
C y lin d ric a l surfaces
C y lin d ric a l tu b e
283

D
D A N T Z IG
319
D E S A IN T V E N A N T
31
D eadw eight,
283
D ecay fa c to r
225, 227
D eflections, p la n e s tru c tu re s w ith large
113,
195
231
, shells w ith la rg e
D e fo rm a tio n energy, sp ecific
41, 215, 221
D e fo rm a tio n g ra d ie n t
28
D e rivatives, c o v a ria n t
14
D esign o p tim iz a tio n p ro b lem s
304
D esign space
304, 326
D esign Space M eth o d
324
D esign v a ria b le s
302 ff
D e te rm in a n t, te n s o r o f c u rv a tu re
205
, m e tric te n so r
8
- , s h ell te n so r
210
- , su rface te n so r
204
D ifferen tial e q u atio n , b o ile r
273
- , e llip tic a l ty p e
219
EULER
183, 210
- , h y p e rb o lic a l ty p e
219
- , V O N K A R M A N 's
117
- , BESSEL a
105
, coupled
243

317, 324, 330


D ire c t m e th o d
335
D irect o p tim iz a tio n stra te g ies
201
D ire c trix
128, 139
D isk, a n n u la r c irc u la r
93
C a r te s ia n c o o rd in a te s
131
- , c irc u la r ro ta tin g
151
- , in fin ite w ith a c rack
145
in fin ite w ith e llip tic a l hole
94
- , p o la r c o o rd in a te s
133
q u a rte r-c irc le a n n u la r
137
, sem i-in fin ite
123
- , sim p ly s u p p o rte d re c ta n g u la r
93
D isk e q u atio n
113 ff, 195
D isk -p la te p ro b le m , coupled
27, 28
D isp lac e m e n t d e riv a tiv e s, te n so r
89
D isp lac e m e n t fu n c tio n , L O V E S 's
50
D isp lac e m e n t p o te n tia l, th e rm o -e la s tic
27
D isp lac e m e n t v ecto r
328
D istan c e fu n ctio n s
223
D is to rtio n s
15
D ivergence, te n so r o f seco n d o rd e r
15
- , v e cto r
239
D O N N ELL s a p p ro x im a tio n
234
- th eo ry ,
5
D yad
6
D y ad ic p ro d u c t

E ffective in -p la n e s h e a r force
E ffective tra n s v e rse s h e a r force
103, 105,
E ig en freq u en cies
E ig en v alu es o f a s y m m e tric a l te n so r
E IN S T E IN s s u m m a tio n co n v en tio n
E la s tic e n erg y o f fo u n d a tio n
E la s tic e n erg y o f p la te
E la s tic -p la stic sta te
36, 43,
E la s tic ity m a trix
- p ro p e rtie s
33,
- te n so r
E llip se fu n c tio n s w ith v a ria b le e x p o n e n t

234
234
296

F easib le d o m a in
F IA C C O

307
314

12

329
180
180
32
121
34
213
333,
367
E llip tic a l-h y p e rb o lic a l c o o rd in a te s
63, 70, 76
E llip tic a l p a ra b o lo id su rface
241 ff
E n e rg y ex p re ssio n s
40, 106
E n e rg y fu n c tio n a l, H E L L IN G E R -R E IS S N E R
114
E n e rg y p rin c ip le s
39 ff, 80
E q u ilib riu m a t la rg e
123, 252
E q u ilib riu m c o n d itio n s 25, 213, 215, 218, 222,
225, 233, 234, 242, 267
E U C L ID E A N s p ac e
5, 8, 10,
304
E U L E R e q u atio n s
330
E U L E R 's d iffe re n tia l e q u atio n
81, 112, 185
E x c h a n g in g in d ices, ru le o f
9
E x te r n a l p e n a lty fu n c tio n
346

Subject Index
F IB O N A C C I-search
311
F in ite E le m e n t M eth o d (FEM )
83
F ir s t fu n d a m e n ta l fo rm o f surface
248
FLETCH ER
320
F L E T C H E R -R E E V E S
337, 340
F L E T C H E R -R E E V E S -m e th o d
313
F le x ib ility te n so r
33
- m a trix
43
F le x ib le sh ells, th e o ry of
238
F L U E G G E , sh e a r-rig id th e o ry
233
F o rc e -q u a n tity p ro c e d u re
228
F o u n d a tio n , e la stic e n erg y o f th e
180
96, 216
F O U R IE R series e x p a n sio n
F u lly -stre ssed design
368
F uel ta n k o f a sa te llite
364
F u n c tio n a l efficiency
326
F u n c tio n a l m a trix
199
F u n c tio n a l-e ffic ie n t b o u n d a rie s
368
F u n c tio n a l-e ffic ie n t set
352, 353
F u n d a m e n ta l form , first
203
F u n d a m e n ta l q u a n titie s, fir s t o rd e r
203, 247
, second o rd e r
203, 204, 248

385

H O O K E 's law
34
H O O K E 's law, D U H A M E L-N E U M A N N fo rm of
85
H O O K E -D U H A M E L s law
32
H O O K E A N b o d ies
31
H y b rid p ro c e d u re Q P R L T
320
267, 296
H y p a r shell
H y p e rb o lic a l p a ra b o lio d s h ell
241, 242, 267
218
H y p e rb o lic a l sh ell
I
6
In d e x ru le
329
In d ire c t m e th o d s
In flu en c e coefficients
46
101
In flu en c e fa c to r
78
In g o t
In v a ria n ts
13!, 22, 24
93
Is o tro p ic d isc in C a rte s ia n c o o rd in a te s
93, 94
Is o tro p ic d isk
103
Is o tro p ic p late, tra n s v e rse ly v ib ra tin g
233
Is o tro p ic shell, g e n e ra l sh e a r-rig id
213
c o n situ tiv e e q u atio n s

J
G A L E R K IN e q u a tio n s
47, 198
m e th o d
47, 170, 195
G A U S S -W E IN G A R T E N d e riv a tiv e eq u atio n s
205
205, 208
G A U SSIA N c u rv a tu re
p a ra m e te rs
200
249
- c u rv a tu re
323
- e lim in a tio n
210
- m e a su re o f c u rv a tu re
209
- su rface p a ra m e te rs
15
- th e o re m
G E C K E L E R , m e th o d by
226
G e n e ra l b e n d in g th e o ry
233
331
G e n e ra l p o ly n o m ia l fu n c tio n
320, 368
G e n eralize d R ed u ced G ra d ie n ts
G e n e ra trix
201
G e o m e tric a l c o n stra in ts
305
- , m o d elin g
331
G e o m e try o f shells
209
G O L D FA R B
320
GOURSAT
98
G ra d ie n t m e th o d , stee p e st descen t
313
G ra d ie n t, s c a la r fu n c tio n
15
, v ecto r
15
G R E E N -D IR IC H L E T s p rin c ip le
44
G R E E N -L A G R A N G E *s c o m p o n e n ts o f s tra in
29
G R IF F IT H
317

H
H alf-space
H E L L IN G E R - R E IS S N E R fu n c tio n a l
H E R M IT E in te rp o la tio n
H E S SIA N m a trix

89
45. 82,
114
311
307, 312, 320, 337

JA C O B I A N m a trix

199

K
183
K E L V IN fu n c tio n
26
K in e m a tics o f a d e fo rm a b le b o d y
K IR C H H O F F s effective tra n s v e rse s h e a r force
103
114
n o rm a l h y p o th e sis
102, 214, 244
- , p la te th e o ry
98
K O LO SO V
8
K R O N E C K E R s d e lta
307 ff, 319
K U H N -T U C K E R c o n d itio n s

L
L A G R A N G E fo rm u la tio n
- , n o ta tio n
307, 320,
- , -fu n ction
-a u g m e n te d
- , -fu n c tio n a l
311,
- , -in te rp o la tio n
- , -m u ltip lie r-m e th o d (L P N L P )
L A G R A N G E AN a p p ro a c h
307, 330,
m u ltip lie rs
L A M c o n sta n ts
L A M -N A V IE R S e q u atio n s
L A P L A C E o p e ra to r
15, 18, 97,
L A U R E N T -series
L ayout, c o n stru ctiv e
L e a st stiffness
L en g th o f a n a rc e le m e n t
L E V Y s a p p ro a c h
Line elem en t, len g th
Line lo ad , c o n s ta n t c irc u la r

113
26
353
319
330
346
363
40
353
34
49
106
148
303
195
204
109
13
172

386

Subject Index

L in e-S earch -M eth o d


L in e a r s tra in te n so r
L oad v e cto r
L o a d -b e a rin g stru c tu re s
L oading, a x isy m m e tric a l
- , n o n -sy m m e tric a l
L O V E fu n c tio n

311
30
230
93
216
216
49, 89

M
M apping, c o n fo rm a l
145 ff
M a te ria l law
115, 2 2 1 ff
- , p la n e states
35
- , U D -la m in a te
118 ff
37
U D -lay er
M a te ria l p ro p e rtie s
303
M a th e m a tic a l P ro g ra m m in g , a lg o rith m s
310
M atrix , fu n c tio n a l
199
JA C O B I AN
199
M ax im u m ru le
6
46
M A X W E L L , th e o re m
314
M cC O R M IC K
M ean c u rv a tu re
205, 208, 249
7
M easure co m p o n e n ts
M E IS S N E R e q u atio n s
226
M e m b ra n e th e o ry o f shells
214
M em b ran e th e o ry , g e n eral e x p ressio n s
221
M E N A B R E A , th e o re m
45
M e rid io n a l curves
200
7
M etric co m p o n e n ts
, c o n tra v a ria n t
9
, c o v a ria n t
9
M etric te n s o r
13
, d e te rm in a n t
8
M in-M ax, e x te n d ed
328
M in -M ax -fo rm u la tio n s
328
M in im a , g lo b al
307
- , lo cal
307
M odeling, g e o m e tric al
331
M o d ifie d ellip se
334, 367
M odulus, sh ea r
94
YOUNGs
94
M O H R 's circle
24, 66
M O IV R E fo rm u la s
146
M oving A sy m p to te s MMA, m e th o d
321
M u ltic rite ria o p tim iz a tio n
325, 367, 371
M u ltila y er c o m p o site
119
M u ltio b je ctiv e o p tim iz a tio n
327

N
N A G H D I-sh ifte r
N A V IE R 's a p p ro a c h
b o u n d a ry c o n d itio n s
- , e q u atio n
N o n -a x isy m m e tric a l s ta te of stress
N o n -sy m m e tric a l lo a d in g
N o rm a l forces, te n s o r of
N o rm a l h y p o th e sis
K IR C H H O F F s

209
107
103
50
99
216
212
213
114

O
O b jectiv e c o n flic t
O b jectiv e fu n c tio n
303, 305,
- fu n ction, v ector
O b jectiv e fu n c tio n a ls
O b liq u e b ase
O n e -d im e n sio n a l m in im iz a tio n ste p s
O p tim a l design, sim p ly s u p p o rte d colu m n s
O p tim a lity c o n d itio n s
329,
O p tim a lity c rite rio n
O p tim iz a tio n a lg o rith m
O p tim iz a tio n lo o p
a u g m e n te d
O p tim iz a tio n m o d el
O p tim iz a tio n p ro b le m , c o n stra in e d
- , c o n tin u o u s
d iscrete
, M u ltic rite ria
325,
M u ltio b jectiv e
n o n -lin e a r (N LOP)
O p tim iz a tio n stra te g ie s
310,
, d ire c t
O p tim iz a tio n , m u ltic rite ria
, m u ltio b jec tiv e
, sh ap e
325,
O rth o tro p ic c y lin d ric a l shells
O rth o tro p ic p la te s

326
317
326
329
53
310
359
353
355
309
310
335
309
356
302
302
367
325
307
325
335
325
325
329
240
104

p
P a n e ls
370
P a ra b o lic ra d io te le sc o p e re fle c to r
370
P a ra b o lo id , e llip tic a l
241
P a ra b o lo id , h y p e rb o lic a l
241 ff
P a ra b o lo id , skew h y p e rb o lic a l
267
P a ra m e te rs , G A U SSIAN
200
P A R E T O -a p p ro a c h
329
P A R E T O -o p tim a l so lu tio n s
326
P A R E T O -o p tim a lity
325
P A R E T O -so lu tio n s
325, 329
P e n a lty fu n c tio n
314
e x te rn a l
346
- m e th o d of e x te rio r
315
- m e th o d of in te rio r
315
P e n a lty -te rm s
319
P e rm u ta tio n sy m b o l
12
P e rm u ta tio n te n so r
12
P h y sic al c o m p o n e n ts
10, 11
P la n e s tra in , s ta te of
51
P la n e stress, s ta te o f
51, 93, 147
P la n e stru c tu re s w ith la rg e d e fle c tio n s
113
P la te b u ckling, b a sic e q u atio n
118
P la te s in C a rte s ia n c o o rd in a te s
99 ff, 110
- in c u rv ilin e a r c o o rd in a te s
105
, in p o la r c o o rd in a te s
104
- s h e a r stiffness
155
- s tn p , s e m i-in fin ite
155
- K IR C H H O F F s th e o ry
102, 244
- w ith m ix e d b o u n d a ry c o n d itio n s
170

Subject Index
c irc u la r o n e la stic fo u n d a tio n
180
circu lar, c e n tre -su p p o rte d
184
c la m p ed c irc u la r
172
c la m p e d re c ta n g u la r
167
e la stic en erg y o f th e
180
e n erg y ex p ressio n
106
re c ta n g u la r stiffened
189
, re c ta n g u la r
158
- , sh e a r-e la stic
155
, s h ea r-ela stic , iso tro p ic
100
- , s h ea r-rig id , iso tro p ic c irc u la r
104
, s h ea r-rig id , o rth o tro p ic
104
th in c irc u la r
195
, tra n sv e rse ly v ib ra tin g c irc u la r
105
P O IS S O N 's e q u atio n
50
P O IS S O N s ra tio
32 ff, 37, 38, 94
P o la r c o o rd in a te s
94, 98, 104, 106, 112
P o te n tia l energy, to ta l
272
- , v o lu m e forces
93
P O W E L L m e th o d
337, 346
P O W E L L m e th o d o f co n ju g ate d ire c tio n s
312
Pow er series e x p a n sio n
95
327
P referen c e fu n c tio n
277
P re ssu re tu b e
P r in c ip a l ax es
21, 22., 31
P r in c ip a l ax es tra n s fo rm a tio n
12
P r in c ip a l s tra in s
74
22
P r in c ip a l stresses
P rin c ip le o f s ta tio n a rity
44, 45
P rin c ip le o f v irtu a l d is p la ce m e n ts
44, 80
P rin c ip le o f v irtu a l forces
44
,
-,
-f
-,
-,

Q
Q u a si-N E W T O N p ro c e d u re SQ N P

314

R
98
98
322
47
46
189
167
311
233 ff,
235
206
272
167, 180, 182, 322
131
15
38

R a d ia tin g s ta te o f s tre ss
R a d iu s-in d e p e n d e n t s ta te o f stress
R A Y L E IG H
R A Y L E IG H -R IT Z s m e th o d
R e c ip ro c ity th e o re m s
R e c ta n g u la r p la te w ith stiffe n e r
R e c ta n g u la r p late, c la m p ed
R eg u la fa lsi
R e s u lta n t fo rce -d isp la c em e n t re la tio n s
R ev o lu tio n , su rfaces of
R I T Z a p p ro a c h
- m e th o d
R o ta tin g c irc u la r d isk
R o ta tio n o f a v ecto r
- o f a U D -layer
R u le d su rface

201

s
S atellite, fuel ta n k
S c a la r fu n ctio n , g ra d ie n t
S c a la r p ro d u c t

364
15
8

387

S em i-B en d in g th e o ry
238 ff
S em i-M em b ran e th e o ry
238
S e n s itiv ity a n a ly sis
302, 310, 321
a n a ly tic a l
322
- , O v e ra ll F in ite D ifference (O FD )
322
, s e m i-a n a ly tic a l
322
S e n s itiv ity m a trix
322
S e p a ra tio n a p p ro a c h
170
S e q u e n tia l L in e a riz a tio n P ro c e d u re SL P 317, 363
S e q u e n tia l Q u a d ra tic P ro g ra m m in g S Q P
320
Series ex p an sio n , F O U R IE R
95 ff
S h a llo w shells, th e o ry o f
242
SHA N N O
320
Shape
303
- fu n c tio n s
329, 331
- o p tim iz a tio n
325, 329 ff, 366, 367
- o f sh allo w shells
242
S h e a r field th e o ry
239
S h e a r force, effective in -p la n e
234
effective tra n s v e rse
105, 234
S h e a r m o d u lu s
32, 38, 94
S h e a r stiffn ess, p la te
155
S h e a r s tra in , te c h n ica l
30, 52
S h e a r-e la stic p la te
100, 155
S h e a r-rig id o rth o tro p ic p la te
104
- p late, a n a ly tic a l so lu tio n s
107
- sh ells w ith w eak c u rv a tu re
213
- F L U E G G E 's th e o ry
233
- is o tro p ic c irc u la r p la te s
104
Shell
- e le m en t
228
- o f re v o lu tio n , e llip tic a l m e rid io n a l 251
- s h ifte r
209
- s tru c tu re s, c o m bined
228
- tensor, d e te rm in a n t of
210
- , c irc u la r co n ic al
220 ff, 226
- , c irc u la r c y lin d ric a l
220 ff, 225,
264
, c irc u la r to ro id a l
260
- , c y lin d ric a l
226
, h y p a r
267, 296
, h y p e rb o lic a l
218
, h y p e rb o lic a l p a ra b o lo id
267
- , ru le d
267
, sh e a r-rig id w ith w eak c u rv a tu re
213
- , so a p -film
241, 242
s p h e ric a l
217, 220, 226, 255
S h ells o f re v o lu tio n w ith a rb itr a r y m e rio d io n a l sh ap e
228
- , b e n d in g th e o ry
222 ff
, d e fo rm a tio n s
220
d e fo rm a tio n e nergy
221
, e q u ilib riu m c o n d itio n s
215
- , w eakly c u rv e d
223
Shells
- , la rg e d e fle c tio n s
231
, b a sic th e o ry
209
, b o u n d a ry d istu rb a n c e s
228
, c h a ra c te ris tic s of shallow
241
- , c o n stitu tiv e e q u atio n
213
- , d e sc rip tio n of
199

388

Subject Index

g e o m e try o f
209
m e m b ra n e th e o ry
214
- , o rth o tro p ic c y lin d ric a l
240
sh allo w
241
217
sp ecial
stiffen ed
239
Shell stru c tu re s, co m b in ed
228
S ide c o n stra in ts
305
Sign c o n v en tio n
19
S IM P L E X -p ro ced u re
319
S im u lta n e o u s s h ap e -th ick n e ss o p tim iz a tio n 367
Single force, to ta l w ork
40
Skew h y p e rb o lic a l p a ra b o lo id
267
Skew h y p e rb o lic a l p a ra b o lo id su rface
201 ff
S lack v a ria b le
329
S lid in g surface
203
Slow ly d ecay in g b o u n d a ry d istu rb a n c es
238
S o a p -film sh ells
241 ff
S o lu tio n m e th o d , com plex
145
Sphere, hollow
86
S p h e ric a l b o ile r
253
S p h e ric a l cap
293
S p h e ric a l co o rd in a te s
26, 30
S p h e ric a l sh ell
217 ff, 226
S p h e ric a l shell, w in d p re ssu re
255
S p h e ric a l su rface
200
S ta te o f p la n e s tra in
36, 51
S ta te o f p la n e stre ss
22, 35, 51 93, 147
S ta te o f s tra in
26
S ta te o f stress
18
a x isy m m e tric a l
98
n o n -a x is y m m e tric a l
99
- , ra d ia tin g
98
- , ra d iu s -in d e p e n d e n t
98
S ta te S pace M eth o d
324
S ta te v ecto r
229, 231
STEW ARD
317
S tiffened b o ile r stru c tu re
348
Stiffness, le a st
195
S tra in gauge ro se tte
73
S tra in te n so r
29, 74
- , lin e a r
30
S tr a in , G R E E N -L A G R A N G E s c o m p o n e n ts 29
S tra in -d is p la c e m e n t re la tio n s
30, 214, 215,
223 ff, 242
S tra in -stre ss re la tio n s
35, 36
S tra in s, p rin c ip a l
74
S tress d e v ia to r
24
Stress fu n ctio n , A IR Y s
50, 93, 123, 235
- , com plex
97, 99
S tress, u ltim a te lim it
32
re s u lta n ts
114
- , te n so r
20
18 ff
v ecto r
S tre s s -stra in re la tio n s
35, 36
S tresses, p rin c ip a l
22
S tr u c tu ra l a n aly sis
302, 309
S tr u c tu ra l m o d el
309
S tr u c tu ra l o p tim iz a tio n
301, 306
S u b s titu te p ro b le m s
325 ff
S u m m a tio n c o n v en tio n , E IN S T E IN 'S
6

SU M T
Surface

314

204
- elem ent, a re a of
199, 209
- p a ra m e te r, G A U SSIA N
247
- te n so r
203
- tensor, c o m p o n e n ts
204
- tensor, d e te rm in a n t
199
- th e o ry
203 ff
- , b ase vectors
247
- , c irc u la r c onical
- , c u rv a tu re in a p o in t o f a curve
205
242
- , e llip tic p a ra b o lo id a l
248
- , first fu n d a m e n ta l fo rm fo r
201
- , ru le d
- , skew h y p e rb o lic a l p a ra b o lo id
201
203
- , slid in g
200
- , sp h erica l
203
- , tra n s la tio n
200, 206
- , rev o lu tio n
202
- , c y lin d ric a l
11
S y m m e tric a l te n so rs o f second o rd e r
T
102
T e m p e ra tu re g ra d ie n t, pla te
128
T e m p e ra tu re field, s ta tio n a ry
114
T e n so r
- , c o v a ria n t m e tric
203, 248
204
- , c u rv a tu re
- , c u rv a tu re , d e te rm in a n t of
205
27, 28
- d is p la ce m e n t d e riv a tiv e s
- , eigenvectors o f a s y m m e tric a l
12
213
- , e la stic ity
5, 7, 10
first o rd e r
10
h ig h er o rd e r
13
m e tric
212
n o rm a l forces
12
- , p e rm u ta tio n
5, 10, 15
- , second o rd e r
- , second order, a n tis y m m e tric a l
11
11
- second order, s y m m e tric a l
15
- second order, div erg en ce of
- second order, p h y sica l c o m p o n e n ts 11
29, 74
s tra in
20
stress
203, 247
- , surface
33
- , th e rm o -e la s tic
5
z ero th o rd e r
T h e o ry o f s tru c tu re s, m e th o d
228, 277, 347
32, 34
T h e rm a l e x p a n sio n coefficient
309
T h re e -C o lu m n s-C o n c e p t
303
T o p o lo g y
272
T o ta l p o te n tia l energy
44
T o ta l p o te n tia l, v irtu a l
40
T o ta l w ork o f single force
328, 352
T ra d e -o ff m e th o d
228 ff
T r a n s fe r m a trix m e th o d
9
T ra n s fo rm a tio n b e h av io u r
22 ff
- , coefficients
- , m a trix
22, 53, 69
12
- , p rin c ip a l axes

Subject Index

rules
9, 11
- of bases
9
- of tensor of first order
10
Translation shells, equilibriumconditions 218
Translation surface
203
Truss structure
342
283
Tube, circular cylindrical
u

UD-layer, material law


Unconstrained mathematical form
Unconstrained problem
U n i t - L o a d - method

37
356
307
46

Variational calculus, fundamental lemma 330


115
Variational functional
80
- principle
46, 333
- problem
5, 7
Vector
326
- objective function
325
- optimization
325, 353
- Optimization Problem
12
- product
27
displacement
15
-, divergence
15
-, gradient
8
length
230
-, load
15
-, rotation
229, 231
state
18, 19
-, stress
8
-, angle between
7, 13
-, base
170
Vibrating uniformbeam
rectangular plate
103, 105
290
-, circular cylindrical shell
296
shallow shell
44, 80
Virtual displacements, principle
Virtual forces, principle
44
44
Virtual total potential
Volume dilatation
31, 74
Volume element
14
93
Volume forces, potential
VON KARMAN's differential equation 117, 196
VON MISES' hypothesis
351
w

Water tank
Work, external

272
42

Yield point
YOUNG's modulus

32
32, 37, 94

389