You are on page 1of 64

ANALYSIS OF C A B L E STRUCTURES

by

NEWTON'S METHOD

by

RONALD I A N SPENCER MILLER

B.A. (1965)
B.A.Sc. (1967)
The University of British Columbia

A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF

THE REQUIREMENTS FOR T H E D E G R E E OF

MASTER OF A P P L I E D SCIENCE

In t h e Department

of

CIVIL ENGINEERING

We accept this thesis as conforming

to the required standard

The University of British Columbia

April 1971
In presenting this thesis i n partial fulfilment o f the

requirements f o r an advanced degree a t the University of British

Columbia, I agree that the Library shall make i t freely available

for reference and s t u d y . I further agree that permission f o r

extensive copying of this thesis f o r scholarly purposes may be

granted by t h e Head o f my Department or by h i s representatives.

It i s understood that copying or p u b l i c a t i o n of this thesis for

financial gain shall n o t be allowed without my written permission.

Department of CIVIL ENGINEERING

The University of British Columbia,

Vancouver 8, Canada.
Abstract

The analysis of structures which contain catenary cables

is made difficult by the non-linear force-deformation relationships

of the cables. For a l l but the smallest d e f l e c t i o n s i t i s not

possible to linearize these r e l a t i o n s h i p s without causing signific-

ant inaccuracies.

Newton's Method solves non-linear equations by solving a

succession of linearized problems, the'answer converging to the

solution of the non-linear problem. Newton's Method so used to

analyze cable-containing structures results i n a succession of linear

stiffness analysis problems. As a result, conventional stiffness

analysis computer programs may be modified without great difficulty

to solve cable s t r u c t u r e s by Newton's Method.

The use o f Newton's Method to solve cable structures forms

the body of this thesis. The two basic innovations necessary, which

are t h e p r o v i s i o n o f methods for calculating the end-forces of a

cable i n an a r b i t a r y p o s i t i o n , and for evaluating the stiffness

matrix of a cable, are presented. Also discussed are the co-ordinate

transformations necessary to describe the cable stiffness matrix and

cable end forces i n a Global Co-ordinate System.

The v i r t u e s o f t h e method a r e d e m o n s t r a t e d i n two example

problems, and the t h e o r e t i c a l basis f o r Newton's Method i s examined.

Finally, the value o f t h e method presented is briefly discussed.


TABLE OF CONTENTS

Abstract

List of Figures

Acknowledgements

Page

1. The P r o b l e m 1

2. The M e t h o d 5

3. C a b l e End F o r c e s 12

4. The C a b l e S t i f f n e s s Matrix 21

5. The C a b l e C o - o r d i n a t e S y s t e m 26

6. Advanced Topics

1. Non-linear Behaviour of Non-Cable Structural

Components 31

2. S p e c i f i e d Cable Tensions 33

3. Miscellaneous Problems 3^

4. C a b l e Loads 3^

7. Examples

Example 1 36

Example 2 38

8. Discussion 44

Bibliography 46

Appendix 1 i

Appendix 2 vii

Appendix 3 x
LIST OF FIGURES.

Page

Fig. 2.1 Example Problem 7

Fig. 2.2 Path o f S o l u t i o n to Example 10

Fig. 3.1a The Cable Co-ordinate System 12

Fig. 3.1b Forces i n the Cable Plane 13

Fig. 3«lc Dimensions i n the Cable Plane 13

Fig. 3.2 Element o f a Catenary Cable 14

Fig. 4.1 Degrees o f Freedom i n the Cable Plane 22

Fig. 4.2 Degrees o f Freedom f o r a General Cable 25

Fig. 5*1 Cable and G l o b a l Co-ordinate Systems 30

Fig. 7.1 Guyed Tower 39

Fig. 7.2 A x i a l Force a t 750' L e v e l Versus Initial

Cable S t r e s s 40

Fig. 7.3 Bending Moment a t 750' Level Versus

I n i t i a l Cable S t r e s s 40

Fig. 7.4 L a t e r a l D e f l e c t i o n a t 1,000* L e v e l Versus

I n i t i a l Cable S t r e s s 41

Fig. 7.5 L a t e r a l D e f l e c t i o n a t 750' L e v e l Versus

I n i t i a l Cable S t r e s s 41

Fig. 7.6 S t r e s s i n Higher Windward Cable 42

Fig. 7.7 S t r e s s i n Lower Windward Cable 42


AC KNOWLEDGEMENTS

I should l i k e t o e x p r e s s my a p p r e c i a t i o n f o r t h e g u i d a n c e

and help given b y my s u p e r v i s o r , D r . R.F. H o o l e y t h r o u g h o u t t h e

writing of this thesis. I should also l i k e t o thank t h e N a t i o n a l

Research Council o f Canada f o r t h e i r financial support, and t h e

University o f B r i t i s h Columbia Computing Centre f o r t h e use o f

their facilities.

Finally, I should l i k e t o thank Miss Sarah Fenning f o r her

painstaking efforts i n typing this thesis.

A p r i l , 1971
Vancouver, B r i t i s h Columbia.
Chapter 1. The Problem.

There a r e many structures which involve cables - suspension

bridges, guyed towers, transmission lines, aerial tramways,

cable-supported roofs and numerous others. F o r some o f these

problems the cables a r e so taut that they may be t r e a t e d as bars:

for others they a r e so t h i c k they must be t r e a t e d a s beams: for

s t i l l others, they a r e so c l o s e l y spaced they may be t r e a t e d as

membranes. There remain, however, a large number of structures

wherein the cables may be a n a l y z e d under the assumptions of

catenary behaviour: that cables are subjected to a loading per

unit length which i s constant i n i n t e n s i t y and d i r e c t i o n , and

that they are completely flexible i n bending.

These problems are d i f f i c u l t to solve, f o r unlike many o f

the problems encountered i n structural analysis, their load-

deformation r e l a t i o n s h i p s are markedly non-linear. As a cable

is stretched, i t becomes stiffer, and as i t i s r e l a x e d i t becomes

more flexible. Moreover, this non-linearity i s significant for

all but the smallest deflections. The f r i e n d l y assumptions r e -

quired for stiffness analysis cannot be made, f o r even i f we det-

ermined the stiffness matrix f o r the structure i n its initial

configuration, that stiffness would change so markedly as the

deformations increased that t h e answer we found would be quite

unreliable.

We may be a b l e to simplify the formation o f the stiffness

matrix by making further assumptions: f o r instance, we may


- 2 -

assume that the cables are inextensible a x i a l l y , or that they

are reasonably taut ( i n which case the cables follow homely

parabolas, instead of esoteric catenaries). Nevertheless,

the central problem will remain - the cables are non-linear.

This thesis presents a method f o r the a n a l y s i s of s t r u c t u r e s

containing catenary cables. The method may be simplified

(and restricted) by the assumption of inextensible behaviour

and/or parabolic cables. As presented, i t solves those

problems where cables may be treated as subjected to constant

loading per unit length, completely flexible i n bending, and

linear - elastic axially. Non-cable components o f the

structure are analyzed by conventional small strain-small

rotation theory stiffness analysis.

In the development of this thesis, three methods f o r

the a n a l y s i s of the problems described i n the previous para-

graph were investigated. The three methods are discussed

briefly below.

The first method was simply to treat each cable as a

series of pin-ended bars. Live and dead loads were applied

at the joints, and t h e s t r u c t u r e was analyzed using large

deflection theory f o r the bars. Results were often satis-

factory, b u t two disadvantages were apparent: f o r cables with

little sag the s t i f f n e s s matrix was poorly conditioneds and

the amount of computation involved i n t h e method was quite

high.

The second method was to treat each cable as one member

and to apply the load i n increments, treating the cable

stiffness as linear f o r each increment. This method was also


- 3 -

often s a t i s f a c t o r y , b u t i t , t o o , had a drawback: the

accuracy o f the s o l u t i o n depended on t h e s i z e of the load incre-

ment chosen. The o n l y way t o ensure an accurate a n s w e r was t o

perform analyses with successively smaller load increment sizes,

until the solutions converged. I f only one s o l u t i o n was made,

the accuracy was indeterminate. F o r some s t r u c t u r e s , which

behaved almost linearly, only a few i n c r e m e n t s were required,

whereas f o rhighly non-linear structures the load had t o be

built up i n many small increments.

The third method investigated, which i s t h e method

described i n the remainder of this thesis, had none of the

drawbacks of the f i r s t two m e t h o d s . Any d e s i r e d degree o f

accuracy could be o b t a i n e d , a n d t h e amount of computation

required was relatively modest.

In the i n i t i a l configuration, the 'unbalanced forces'

acting on the s t r u c t u r e were evaluated (the unbalanced forces

are simply the external loads minus the i n t e r n a l forces resist-

ing them). The l i n e a r behaviour of the structure was then

represented by a s t i f f n e s s matrix (each cable, like each beam,

being represented by a s i n g l e member matrix) and t h e l i n e a r

deflections due t o t h e u n b a l a n c e d forces were calculated. In

this new deformed p o s i t i o n a new s e t o f unbalanced forces was

determined, a n d a new stiffness matrix was f o u n d (since the

stiffness of the structure w a s n o t t h e same i n t h e d e f o r m e d

position a s i t was i n the i n i t i a l position). The d e f l e c t i o n s

due t o t h e new s e t o f unbalanced forces were c a l c u l a t e d and

added to the previous deflections. This procedure was carried

out until the unbalanced forces were effectively zero.


_ 4 _

Mathematically, this i s Newton's method. A good descrip-

tion may be f o u n d i n Livesley ( C h . 10.3. p . 241) ( 1 ) . The

procedure will be d i s c u s s e d i n more d e t a i l i n the next chapter,

and the mathematical implications o f Newton's method a r e r e -

viewed i n Appendix 3«
- 5 -

Chapter 2. The Method.

The method proposed h a s much i n common w i t h stiffness

analysis as i t c o n s i s t s of a series o f ever finer approxim-

ations to the solution, t h e change from one a p p r o x i m a t i o n t o

the next being found by s o l v i n g a linear s t i f f n e s s problem.

As i s usual i n stiffness analysis, the p o s i t i o n of the

structure i s defined by a s e t o f generalized co-ordinates,

one degree o f freedom being assigned t o each possible deform-

ation d i r e c t i o n o f each joint. Wherever there i s a point

load on a cable, o r wherever there i s a change i n the d i r e c t i o n

or i n t e n s i t y of the uniform load on a cable, a joint must be

defined.

As described i n Chapter 1, t h e m e t h o d proposed i s an

iterative procedure consisting of the following basic steps:

(0) Choose a deflected shape which will serve as the

starting point o f the i t e r a t i o n . The most con-

venient initial position will be t h a t a t which a l l the

non-cable components of the structure are unstressed,

when such a position exists.

(1) In the deflected shape, calculate the unbalanced

loads (UBL), which are just the external loads minus

(a) the cable end f o r c e s , and (b) t h e end forces

developed by t h e non-cable members, including the

effects o f member loading i fpresent. Also calculate

the s t i f f n e s s matrix f o r the structure i n this deform-

ed position.
- 6 -

(2) Solve f o rthe incremental deflections due t o t h e

unbalanced loads, using the stiffness matrix just

found. Add these incremental deflections to the

previous deflections.

(3) Repeat steps (1) a n d (2) using t h e new d e f l e c t e d

shape, until the unbalanced loads are negligible.

The unbalanced loads represent t h e amount by which the

structure i s out o f equilibrium. When we evaluate the un-

balanced loads i n step ( 1 ) , we have i n fact performed an exact

solution o f the s t r u c t u r e , b u t under a different loading from

that i n which we are interested. We have an exact solution f o r

the l o a d i n g which consists of the applied loads minus the un-

balanced loads. When we apply the unbalanced loads to the

structure, and add t o t h e p r e s e n t deflections the incremental

deflections they cause, t h e new u n b a l a n c e d loads a r e much

smaller, representing only the error i n the solution. As t h e

unbalanced loads approach zero, the l o a d i n g f o rwhich our

deformed position i s an exact solution approaches the loading

whose effects we w i s h to study. Furthermore, the closer the

unbalanced loads are t o zero t h e more rapidly do t h e two load-

ings approach coincidence. We can thus find a solution to

within any* arbitrary small tolerance f o rerror, limited only

by our computational techniques.

* This i s perhaps more valorous than discreet. See C h a p t e r 8.


- 7 -

Let us consider a simple single degree of freedom

problem which illustrates the major features of the method.

In figure 2.1 a 1000 foot long inextensible cable which weighs

one pound per foot i s shown a t t a c h e d at i t s left end to a support

and at i t s right end to a spring. The ends are constrained to

remain at the same height. The spring has a s t i f f n e s s of 1 k i p /

foot and i s unstrained when t h e cable span, L, i s 1000 feet.

One kip i s applied to the r i g h t hand end of the cable, and i t is

desired to find the equilibrium position.

IOOO' long inextensible cable


weight* 1 ibyft.

Example Problem

Fig. 2.1.

For this cable, the h o r i z o n t a l component of tension, H,

is related to the span, L, by equation (3«5)i which may be

simplified to:

2.1
Hsinhy = . 5
Where 2.2
• Q005L
H
- 8 -

Differentiating e q u a t i o n 2.1, we find:

dH _ HcoshY 2 3
dL Lcoshy-^inhy
The generalized degree o f freedom a c t s to the right on

the r i g h t hand end o f the cable. T h e n a t any s p a n L, the un-

balanced load e q u a l s the external l o a d minus t h e resisting

forces due t o the cable and the spring:

UBL = 1 - [ H + ( 1000- L )] (kips) 2.4

Where H i s f o u n d by s o l v i n g e q u a t i o n 2.1. The stiffness,

k, e q u a l s the cable stiffness plus the spring stiffness:

k - -^r-*
dL
1 ( k/ft.) 2 , 5

The s o l u t i o n now p r o c e e d s as follows:

(0) We cannot s t r e t c h the cable t o the position L = 1000

feet without causing an infinite force, so we choose

as the initial p o s i t i o n L = 999 feet, corresponding

to a d e f l e c t i o n § of -1 foot.

Cycle 1.

(1) At L = 999 feet ( S - l ' ) . UBL = - 4.4463 kips


k= 4.2318 k/ft.
(2) The incremental deflection i s : " \• 6
3 = -1.0507'
4.2 318
The new deflection i s : -|.0-1.0507 = - 2.0507'

Cycle 2.

(1) A t § =-2.0507', UBL*-1.4445 kips .


k = 2.10 21 k/ft.
I 444«S
(2) The new deflection i s : -2.0507- g |Q2| =
~ 2 7 3 7 8
- 9 -

C y c l e 3.

(1) At § =- 2.7378', UBL = - . 1489 kips


k= 1. 7151 k/ft.
(2) The new d e f l e c t i o n i s : - 2.73 78- = -2.8246

Cycle 4.

(1) At 8 =-2.8246*. UBL =-.001435 kips


k = 1.6824 k/ft.

(2) The new d e f l e c t i o n i s : - 2.8246- .001435 = -2.8255


1.6824

Cycle 5.

(1) At § =-2.8255', UBL= .00000013 kips


Which i s s m a l l enough f o r most p r a c t i c a l purposes.

So our s o l u t i o n i s L = 1000'-§ = 997.1745'

Note that t h i s procedure e x h i b i t s q u a d r a t i c convergence -

as the d e f l e c t i o n approaches the s o l u t i o n , the incremental

d e f l e c t i o n approaches the true error i n deflection. In other

words, over the s m a l l d e f l e c t i o n s calculated as the s o l u t i o n i s

neared, the s t r u c t u r e remains almost l i n e a r . In the f i r s t solu-

t i o n , the e r r o r i n the d e f l e c t i o n was c u t by a f a c t o r o f 2.36, from

1.8255 f e e t to 0.7748 f e e t . In the second s o l u t i o n i t was cut

by a f a c t o r o f 8.8 (from .7748 t o .0877) i n the t h i r d by 103,

and i n the f o u r t h by 10,900.

The path f o l l o w e d i n the s o l u t i o n i s shown graphically

i n F i g u r e 2.2.
- 10 -

2
L , feet
I
9 99 998
0 L -L
5 ?3,4 997

9- -i
l i n e a r stiffness
© start 1

-3

-4
o START
-5

Path of S o l u t i o n to Example
Fig. 2.2.

The example c h o s e n was very simple, but i t exhibits

all the important aspects o f t h e method. F o r more general

problems there may be many cables a n d many n o n - c a b l e elements,

the cables will n o t be inextensible, and the l o a d i n g will be quite

complicated; b u t t h e same m e t h o d will converge readily to a solu-

tion f o r almost any s t a r t i n g point.

A further point may be observed from this example -

because of the quadratic convergence o f t h e method, we can save

time by u s i n g a starting point as c l o s e as p o s s i b l e t o the solu-

tion. I f we have solved a structure f o r some loading, and now

wish to solve f o r a d i f f e r e n t but s i m i l a r loading, we will be

well advised to start the s o l u t i o n procedure at the deflected


- 11 -

shape resulting from the earlier solution. The unbalanced loads

in this position will be just the d i f f e r e n c e between the two load-

ings, and will be quite small i f the two loadings are similar.

For instance, i f we wish to solve the example just comp-

leted without the one kip load applied at the right end, the soluti

(L=996.54 feet) i s considerably c l o s e r to the previous solution

(L=997.17 feet) than i t i s to the initial point (L=999 feet).

Thus, as might be visualized from Fig. 2.2. starting at the pre-

vious solution point will result i n considerably faster convergence

than starting at the initial point.

By s o l v i n g the non-linear problem as a series of linear

problems, we are able to utilize familiar methods to solve each

linear problem - in particular, we use the methods of stiffness

analysis. In order to modify a stiffness analysis computer

program to solve cable problems we need make only two major mod-

ifications! at each stage of the iteration we need (1) to eval-

uate the vector of unbalanced loads, and (2) to find the stiff-

ness matrix.

Moreover, the contributions of the non-cable elements

to the unbalanced load vector and the stiffness matrix are

already known. Assuming that these elements behave linearly,

their stiffness matrices are constant and the forces they develop

are simply the products of their stiffness matrices with their

deflections.

For the cables, however, things are not so simple. It is

necessary to find the end-forces developed by each cable; and to

find their d e r i v a t i v e s , which comprise the cable stiffness matrix.

These problems are discussed i n the next two chapters.


- 12 -

Chapter 3» Cable End Forces.

Since the loading on each cable i s constant i n direction,

the cable lies i n a plane. This plane, hereinafter called the

"cable plane" i s readily found since (1) i t contains both ends

of the cable, and (2) i t contains the vector representing the

load on t h e c a b l e .

It i s convenient t o use a c o - o r d i n a t e system i n the cable

plane when calculating the cable end f o r c e s and s t i f f n e s s matrix:

this co-ordinate system will be called the "cable co-ordinate

system". As shown i n F i g . 3.1a, the y-axis i s opposite to the

direction of loading, and t h e x - a x i s i s perpendicular to i t .

Forces and dimensions a r e shown i n F i g . 3«lt>, a n d i n F i g . 3»lc.

The actual directions o f t h e x and y axes are discussed i n

Chapter 5 - f o r now, suffice i t to say that they are readily

found.

yj
direction of loading on cable

The Cable Coordinate System


Fig.' 3.1a.
- 13 -

F o r c e s in the C a b l e Plane

Fig. 3.1b. a = Cable area


E= effective modulus
of elasticity
U S L = unstressed length
C - curve ( s t r e s s e d )
length

D i m e n s i o n s in the C a b l e Plane

Fig. 3.1c
- 14 -

H(y'+dy')

Element of a Catenary Cable

Fig. 3-2.

The r e l a t i o n s h i p s between the f o r c e s o f F i g . 3«lb and

the dimensions and p r o p e r t i e s o f the cable as shown i n F i g . 3.1c

may be r e a d i l y d e r i v e d .

Summation o f the y f o r c e s o f F i g . 3 . 2 y i e l d s :
Hy' + wdC = H ( y + dy' )
1

so dy' = -77-dC
n

y
H dx

solving , y' = j h(^ -+


S n
x
A) 3.1

y = "S-cosh(^- + A ) + B 3.2
- 15 -

We have two b o u n d a r y conditions

@ x = 0 y=0

@ x = L y=h

Which g i v e : =sinh"T w h wL
A w L 3.3
2Hsinh 2H
L 2H -I

B = "-^-coshA 3.^

Knowing t h e shape o f the c a b l e , and k n o w i n g t h a t t h e

component o f c a b l e tension i n the x - d i r e c t i o n i s the constant

value Hf i t i s r e l a t i v e l y s t r a i g h t f o r w a r d to c a l c u l a t e the

forces a t any p o i n t a l o n g the cable and t h e l e n g t h o f t h e c a b l e

itself.

The stressed length of the cable i s :

C = |V+^sinh ] 2
3.5

The y-components o f t h e end t e n s i o n s a r e :

w _ wh , . t u wL , Cw _ ,
Vo --~2~ c o t n
2TT ~2~ 3*
A n d
u I n
\i - wh .. wL . Cw __
V, - - y c o t h -grr + 3.7

The end t e n s i o n s themselves are simply:

To S* [|HH"++ Vv . ' ]|' '


'. 3.8
2 2
0

And
3-9

* In f a c t , x was c h o s e n n o r m a l t o the load i n order to y i e l d

this simplification.
- 16 -

And interestingly:

T, - T = wh
0

The elastic elongation of the cable, A» I s


found from:

0 o

A = H L r ^ c o t h v ^ + ± + J ^ s i n h ^ l 3.10
" a E[_2HL 2H 2 2wL H j

The unstressed length of the cable i s o f course the

actual length minus the e l a s t i c elongation:

USL = C - A 3.H

In general, we will know the cable properties w . a ,

and E , a n d we will know the positions o f t h e ends of the cable

(and so h and L ). I f we also know the value of H i we could

directly solve equations 3.6 a n d 3.7 f o r t h e y-components of the

end tensions.

Unfortuanately we seldom know the value of H » hut instead

know some other quantity: the unstressed length of the cable,

USL: o r t h e s a g a t some value o f x i o r one o f t h e e n d t e n s i o n s To

or T, . In this case we u s e t h e known q u a n t i t y to find H , and

then use H to find V 0 a n d V\ .


- 17 -

Just a s we are going t o u s e Newton's method to solve the

set of simultaneous non-linear equations which define the equil-

ibrium position o f the entire s t r u c t u r e , s o we will now use i tt o

solve f o r H '> H i s related t o o u r known q u a n t i t y (USL, s a g , T 0

or Ti ) by a n o n - l i n e a r equation, and Newton's method solves non-

linear equations.

If we l e tK stand f o r t h e known q u a n t i t y , the following

procedure (similar mathematically t o t h e example discussed i n

Chapter 2) w i l l find H .

(1) Guess a value of H ,

(2) C a l c u l a t e K*, the value that the quantity K would

have i f i t were based on the guessed value of H .

(3) Define f(H) = K-K*, the error i n K .

(4) Carry out the Newtonian sequence:

i+ 1 f(H')
H H' +

V dH J

The guessed value of H will converge quadratically to the

correct value, a t which timeflH) = 0 . We can say that conver-

gence has occurred when the absolute value o f f(H) i s l e s s than

some arbitrarily small function of K , f o r instance:

|f(H)|<. 000001K
- 18 -

rl K
It i s only necessary, then, to determine-^—. This task
On
is simplified by using the following functions:

oE o -JL
a-f 3-t

A-WJL+A -n - sinh2T sinhY 2

A
' H M
'/ ' 2 €y €^ 2

=j3y? c
scn2

Using these functions, we can re-write the•governing

equations:

sinhX
1
3.1
y =

y = -tL ( coshX - cosh A ) 3.2

A = 3.3

c = 3.5

Vo = aySL(€-^gcothy ) 3.6

v, = aySL(€ + /Scothy ) 3.7

A• §(ev^r
L i
h +
T ^y )
+ r 3.10

And sag sag =£ * - y 3.12


- 19 -

We c a n now d i f f e r e n t i a t e to find 4 5 * T h e
results are:
an

Case 1.

If t h e known v a r i a b l e K i s t h e USL:

3.13

Case 2 .

If t h e known v a r i a b l e K i s t h e s a g a t x

dsoq -i - \ L I"*. ^J-2Y* dA \ coshA HA


+ ——+sinhA- 3.14
dH H dH
Where:

dA xr, - ff( l - y c o t h X ) 3.15

Case 3^ .

If t h e known v a r i a b l e K i s tension T 0

3.16

Where:

3-17

Case 3i-»

If t h e known v a r i a b l e K i s t e n s i o n T,

dT, dV, 3.18


H dH
dH
Where:

3.19
dV
d -4>-rv
T h i s method i s not i n f a l l i b l e : there a r e two ways i n

which i t can f a i l . Firstly, the cable equations are not solvable

for any g i v e n cable. We have d e f i n e d w as t h e l o a d per u n i t


- 20 -

length of cable: thus i f we stretch a cable to double its

original length we are constrained to double the total load on i t

(which is quite reasonable for, say, ice loading). This extra

load in turn produces further s t r e t c h i n g , which produces further

loading etc. This effect i s completely negligible except for

very heavily loaded very flexible cables, i n which i t can get out

of hand. Fortunately such cables exist only i n the imagination.

The other danger i s more real: i t is possible that at

some stage i n the s o l u t i o n the value of H will become path-

ologically small, or even negative, i n which case the procedure

will succumb to numerical ailments. The antidote is simply, at

each iteration, to limit the new value of H to be no less than

half the previous value.

In p r a c t i c e , the solution for the cable end forces is

easier done than said. In Appendix 1 a subroutine which finds

cable end forces (written in G-level /360 Fortran) is reproduced,

along with three minor subroutines which calculate —j-.


dH

For cables which are defined by known sag or tension

values i n the initial p o s i t i o n of the structure, the solution

for the initial cable end forces also gives the unstressed

lengths, upon which the c a l c u l a t i o n s of cable end forces are

based at subsequent stages i n the solution.

0
- 21 -

C h a p t e r 4. The C a b l e S t i f f n e s s Matrix.

As was d i s c u s s e d i n C h a p t e r 1, t h e s t i f f n e s s o f a c a b l e

changes as i t i s deformed. When we r e f e r to the s t i f f n e s s matrix

o f a c a b l e , we mean t h e s e t o f d e r i v a t i v e s o f c a b l e end f o r c e s

with respect to cable end movements e v a l u a t e d i n the present

position o f t h e cable-. This i s called a "tangent" stiffness

matrix, and i s a n a l a g o u s t o a t a n g e n t modulus.

Within i t s own c a b l e plane, each cable has f o u r degrees

of freedom: two a t e a c h e n d . I f we a s s i g n t h e s e degrees o f f r e e -

dom a s shown b y t h e numbered a r r o w s i n F i g . 4.1, t h e n , ignoring

for t h e moment t h e p o s s i b i l i t y o f d i s p l a c e m e n t s out of the plane,

the cable s t i f f n e s s matrix is:

3F,, 3F, 3F, 9F,


35, 3S 2 35 3 354
3F 2 dFz 8F 2 3F 2

35, 35 35
Hr
2
3 S 3 4

3F 3
3F3 8F 3 a F 3

35, 35a 35 3 3S4


3F 4 3F 4 3F 4 3F 4

35, 35 2 . 35 3 35 4 4.1

Where F, i s t h e f o r c e i n direction 1 , 5 , »• I s
deflect-

ion i n direction 1 , etc.


- 22 -

If we now make the substitution:

•a§, aS 3

ah =• s§ a8< 2
3F,=- aH
aF = 2 avo

aF 3 = aH
8F 4 = av,
Our matrix will become:

aH aH - a H - a H
aL ah aL ah
-avo -a v.. avo avo
aL ah 3L ah
-aH -aH an aH
aL ah aL ah
-av, -av, av. av,
aL ah aL ah 4.2

Degrees of Freedom in the Cable Plane


Fig. 4.1.
- 23 -

Upon evaluating the d e r i v a t i v e s , i t i s found that i n

general this matrix i s n o t symmetric, and t h a t the set of forces

represented by each column i s not self-equilibrating. This un-

usual behaviour i s due t o t h e f a c t that we are applying a non-

conservative load to the cable. By d e f i n i n g t h e l o a d per unit

length of cable as the constant value w , we have assumed that

if the cable i s stretched the t o t a l load increases.

This does n o t mean t h a t i n practice our cable equations

are inadequate: the e l a s t i c elongation i s small compared to the

length of the cable, and t h e i n c r e a s e i n loading on t h e cable i s

in t h e same small ratio to the t o t a l load. We do f a c e a problem,

however: conventional stiffness a n a l y s i s programs use symmetric

matrices, a n d i f we wish t o modify such a conventional program

to handle cables, we will save a l o tof trouble by u s i n g symmetric

matrices f o r our cables.

Now t h e asymmetry i n the cable matrix i s small, and i n f a c t

is virtually negligible i n most cases. Accordingly, we will make

minor modifications to the matrix which will render i t , though no

longer strictly exact, symmetric.

This i s readily achieved by:

Replacing:
avo by
-8H
3L 8h

avi
by
aH
aL ah
- 24 -

And:
av. av*
by
8h an

Where:
av* av. a Vc
ah ah ah

This g i v e s us a new approximate matrix:

aH an -aH -aH
aL ah ah
aH av* - aH - av*
ah ah ah ah
[ "] =
K
-aH -aH aH aH
aL ah aL ah
-aH -av* aH av*
ah ah ah ah
h.3

Using the approximate matrix does n o t hinder the solution

procedure. Firstly, i t t u r n s out that f o r cables within (and

somewhat b e y o n d ) t h e r a n g e o f e n g i n e e r i n g u s a g e , t h e approxim-

ations are small. Secondly, Newton's method does n o t require

the correct matrix: a c l o s e one will do. (1), (2), (3)« ( I n the

example i n Chapter 2, f o r i n s t a n c e , any positive finite value could

have b e e n u s e d f o r the cable s t i f f n e s s , ^ ^ , and Newton's method


d L
w o u l d have i n e v i t a b l y l e d to the correct solution, though conver-

gence m i g h t have b e e n s l o w ) . I t i s of course necessary to eval-

uate the unbalanced l o a d e x a c t l y , but this i s independant of cable

stiffness.

L e t us now t u r n b a c k t o what was ignored at the start of

this chapter: the p o s s i b i l i t y of cable displacements out of the


- 25 -

cable p l a n e . Two more degrees of freedom are r e q u i r e d to

d e s c r i b e these displacements, and are numbered 3 and 6 i n

F i g . 4.2. As f o r a pin-ended bar i n t e n s i o n , the s t i f f n e s s i n


u
these d i r e c t i o n s i s simply y- , so the approximate matrix becomes:

3H 3H 0 -3-H -3H
0
3L 3h 8L ah
3H 3 V* -av* 0
8h 3h
0
ah ah
-
0
3H
0
-3H
X 0
an
O
8H
" \
0 0
3L 8h 8L ah
-8H dH 3V*
0 0
ah ah ah
0 0 -\ 0 0 \
4.4

The terms i n the m a t r i c e s of equations 4.2., 4.3.., and

4.4., are d e r i v e d i n Appendix 2. In the remainder of t h i s thesis

i t w i l l be assumed that the approximate matrix Kca i - used.


s

Degrees of Freedom for a General Cable


Fig. 4.2.
- 26 -

Chapter 5. The C a b l e C o - o r d i n a t e System.

A t t h e end o f C h a p t e r 2 we s e t o u t t o f i n d t h e two new

f e a t u r e s w h i c h w o u l d e n a b l e us t o c o n v e r t an o r d i n a r y stiff-

ness a n a l y s i s program i n t o an improved v e r s i o n capable of hand-

l i n g cable structures. T h e s e two f e a t u r e s w e r e t h e capacities

t o f i n d t h e c a b l e end f o r c e s and t h e c a b l e s t i f f n e s s matrix, and

t h e y have been p r o v i d e d i n Chapters 3 and 4 .

L i k e many m o d i f i c a t i o n s , t h e y do n o t f i t d i r e c t l y into

the e x i s t i n g framework, f o r t h e y work i n terms of cable co-

ordinates, and s t i f f n e s s analysis p r o g r a m s work in global co-

ordinates. To a d a p t them we n e e d a t r a n s f o r m a t i o n m a t r i x , and

to get t h e t r a n s f o r m a t i o n m a t r i x we n e e d t o know t h e directions

of the cable c o - o r d i n a t e axes.

The l o a d p e r u n i t l e n g t h o f c a b l e , W , may be s p l i t into

components p a r a l l e l to the g l o b a l X , Y , and Z a x e s : w x iW y i

and respectively. Thus t h e c a b l e l o a d i n g may be represented

vectorially as:

W = Wy

5.1

Now y , t h e d i r e c t i o n o f t h e y - a x i s i n the cable co-

o r d i n a t e s y s t e m must be o p p o s i t e to the d i r e c t i o n of loading.

A vector in this direction is thus:

y = w,

- w.
- 2 ? -

L e t C be the l i n e f r o m t h e end of the c a b l e at the origin

of the cable co-ordinate system to the o t h e r end of the cable.

(Fig. 5.1). C i s represented by i t s three components:

c
c=
c 2

5 - 3

Now the c r o s s product o f two vectors i s perpendicular

to both o f them. The c a b l e z -axis i s , of course, perpendicular

to the cable plane, and s i n c e b o t h y" and C l i e i n the cable

p l a n e , Z must be p e r p e n d i c u l a r t o e a c h o f them. Thus we write:

z = CXy
5 . 4

The cable X -axis i s , of course, perpemdicular to the y

and Z -axes, and so i s f o u n d by:

x = y X z
5 . 5

Equations 5.2., 5.4., and 5.5., define vectors parallel

t o the cable co-ordinate axes. I t i s convenient to normalize

these v e c t o r s b y d i v i d i n g e a c h t e r m by the l e n g t h of the vector,

A subscript 1 w i l l denote a n o r m a l i z e d vector. (The components

of a normalized vector are, of course, the d i r e c t i o n c o s i n e s of

the v e c t o r ) . For example, the l e n g t h of the W vector (equation

5.1.) is w , where:

V/,
2 ^ 2 2
w = Wx + W y
+
w
z

5 . 6
- 28 -

So the normalized cable loading vector would be

'w /w ^
x

Vl /W y

w /w z

5.7

The quantities L and h are simply the x and y components

of C , and a r e r e a d i l y found by:

L = Xi- C 5.8

h =y,-c 5.9

Where • represents dot product.

In F i g . 4 . 2 a r e shown the s i x degrees o f freedom of a

cable. I n terms o f them, the cable end f o r c e s ( F i g . 3.16) are

represented by the vector:

-H
Vo
0
H
V,
0 5.10

and the s t i f f n e s s matrix i s as i n equation (4.4). To transform

them to global co-ordinates we will use a 6x6 transformation


- 29 -

m a t r i x , J_TJ . This matrix i s composed o f two identical 3x3 sub-

m a t r i c e s ft] a r r a n g e d on t h e diagonal:

0 0 0
0 0 0
0 0 0
o o p
0 0 0
0 0 0!

5.11

Where the three columns of the sub-matrix are the vectors

x\, y t , and Z :
t

5.12
[T.] = [ x x y x z t ]
In global co-ordinates, the cable end f o r c e s a r e

5.13
q I ' I • c

And the s t i f f n e s s matrix i s :

K CO 9
T c
T 5.14

We now know a l l we need i n order to convert a stiffness

analysis program to solve cable structures. I n the next chapter

we will consider some ways t o extend the v e r s a t i l i t y of the

method.
- 30 -

Cable and Global Coordinate Systems

Fig. 5.1.
- 31 -

Chapter 6. Advanced Topics.

With what has been discussed i n the previous five chapters

we a r e now able to analyze many cable structures. In this chapter

we will consider some r e f i n e m e n t s which will make t h e method more

general.

1. Non-linear Behaviour o f Non-Cable Structural Components.

In Chapter 2 i t was assumed that the non-cable components

of t h e s t r u c t u r e were linear - their stiffness matrices would

not change as they deflected. There a r e many cases where this

assumption i s not justified, f o r instance: i f the material

stress-strain relationship i s non-linear, o r i f t h e member under-

goes large rotations, or i f there i s an i n t e r a c t i o n between axial

force and bending stiffness.

In principle, these non-linearities pose no p r o b l e m - we

can simply handle these elements just a s we handle the c a b l e s , by

using a tangent stiffness matrix, a n d e v a l u a t i n g t h e member e n d

forces a t each successive deformed position of the structure. In

practice, this approach c a n be q u i t e difficult, but f o r at least

one of the non-linearities mentioned above there i s a simpler

procedure.

In frame analysis, the effect of axial f o r c e on the bending

stiffness o f a beam i s g e n e r a l l y handled by a d i f f e r e n t kind of

matrix: a "secant" matrix. "Stability f u n c t i o n s " (4) define the

secant matrix i n terms of the anticipated axial force. Using this

scheme, a solution i s performed using the linear matrix. This

solution gives an estimate of the axial forces, and a secant

matrix i s built based on t h i s estimate. A solution i s now p e r -


- 32 -

formed based on this new matrix, and a better estimate of the

axial forces results. The procedure i s continued until success-

ive solutions converge, which i s usually achieved after just two

or three solutions.

Now this secant matrix i s not the tangent matrix we want,

but i t is a lot closer to i t than the linear matrix i s , and as

was mentioned i n Chapter 4, i t i s not necessary to have the

correct matrix: a close one will do. Moreover, the secant matrix

lets us find the member end forces: i n i t s deformed position

we calculate the axial force i n the member (which i s no problem

since the axial stiffness i s constant) and then the secant matrix

based on the axial force. Multiplying the secant matrix by the

member deflections now gives the true (exact) non-linear bending

moments and shears.

The point to note about the secant matrix is this: i f the

guessed value of axial force i s correct, then the secant matrix

is linear, with respect to bending and shear deformations. The

axial f o r c e s i n the members tend to converge rapidly on their

final values, so the beams b e h a v e quite linearly by the time

the final solution is approached.

There i s one danger in this method: i f the axial loads

are too great, the stiffness matrix will become singular. It is

possible that i f a structure were loaded almost to its critical

load the Newtonian sequence would wander beyond the critical load

and suffer a premature demise. This problem occurs rarely, and

can be o b v i a t e d by a p p l y i n g the load i n steps: for small steps

the error i n each stage of the iteration i s reduced, and the


- 33 -

deflections never deviate f a r from the solution. In other words

we will first solve the problem f o r , s a y , 50 percent of the r e -

quired load, then f o r 75 percent using the previous solution as

the starting point, then f o r 85 percent starting at the solution

for 75 percent, and so on.

Beam-column effects are often significant i n the behaviour

of guyed towers. Stability functions were used, as described

above, i n the second example of Chapter 7.

2. S p e c i f i e d Cable Tensions.

If we know the unstressed length of a cable, we can find

its end f o r c e s f o r any deformed position. I f , i n the initial

position o f t h e s t r u c t u r e , we know the sag of the cable or the

tension at either e n d , we c a n use t h e methods of Chapter 3 "to

find the unstressed length (by f i r s t f i n d i n g H). Frequently,

however, the cables i n a structure are tensioned to predetermined

values after the structure i s erected. In this case we do n o t

know the position o f the s t r u c t u r e , f o r i t deforms from i t s un-

stressed position as the cables are tensioned.

The solution to this problem i s surprisingly simple: at

each successive deformed position found during the solution pro-

cess, we re-tension the cables to their specified tensions (by

changing their unstressed lengths). When c o n v e r g e n c e i s achieved,

we calculate the true unstressed lengths of the cable.

This a n a l y s i s , o f course, considers only the loads applied

at the time of the cable tensioning. Once the unstressed lengths

of the cables a r e known, o t h e r loading cases (wind, snow, e t c . )


- 34 -

are readily handled i n the usual fashion. The second example

considered i n Chapter 7 employed this method of specifying cable

tensions.

3. Miscellaneous Problems.

The effects on cables of temperature changes, end slippage,

and turnbuckle adjustment are readily handled by changing their

unstressed lengths.

4. Cable Loads.

The loads on a cable due to i t s own weight and the weight

of accumulated ice are readily evaluated. For wind loading the

evaluation i s more difficult.

Firstly, the wind loading acts perpendicular to the cable.

Since the cable i s curved, the direction of the wind loading

varies along the length of the cable, which is contradictary to

the theory of catenary cables. We avoid this embarrassement in

a rather direct manner: i f a cable i s reasonably taut, we treat

the direction of each element of the cable as being the same as

that of the line between the ends of the cable: C i and apply the

wind load normal to C , ( F i g . 5»1«) I f the cable has a large sag,

we simply treat i t as a s e r i e s of shorter cables, each of which

has a low sag.

Now f o r wind acting perpendicular to a cable, the drag force

per unit length i s :

w
Lf T P<lv C 2
a

6.1
- 35 -

Where P i s the density of the a i r ,d i s the cable diam-

eter, v i s t h e wind velocity, and C d i s the c o e f f i c i e n t of drag

for the cable. A reasonable value for C d i s 1.2. A i rat s.t.p.

weighs .08071 lbs/ft. 3

It has been shown (5) that i f the wind direction i s a t an

a n g l e 7) to the plane perpendicular to the cable (that i s ,the

plane perpendicular to C ) the drag i s s t i l l perpendicular to

the cable and has magnitude:

6.2

The wind velocity may b e r e p r e s e n t e d by i t s three global

components V x , v y , and V 2 , so that vectorially i t i s :

V
6.3

The direction of the drag, w rf » i s found by:

w d = C X V I C 6.4

And when n o r m a l i z e d i s written W, Cos Tj i s simply:


'dl

6.5

So the cable loading vector due t o wind i s :

6.6

wind = W
d rog
P C O s 2 7
? W
<"
- 36 -

Chapter 7. Examples.

Example 1.

T h i s simple example has been solved by others (6), (7). A

cable spans 1,000 f e e t h o r i z o n t a l l y between f i x e d supports, the

midspan sag b e i n g 100 feet. The cable weighs 3.16 lbs/ft, is


2 6
0.85 in area, and has an e f f e c t i v e modulus of 19x10 psi. A

v e r t i c a l load of o i s then placed 400 f e e t from the l e f t support.

As s o l v e d by Frances and O'Brien (7) the loaded p o i n t moves

from x = 400', y = 96.0495' to x = 397.180', y = 114.509'. The

problem was s o l v e d by the methods presented h e r e i n , u s i n g a number

of d i f f e r e n t i n i t i a l p o s i t i o n s . The convergence c r i t e r i o n used

was t h a t the unbalanced f o r c e s should a l l be l e s s than one pound.

The procedure converged to the same f i n a l p o s i t i o n as t h a t found

by Frances and O'Brien, r e g a r d l e s s of the i n i t i a l p o s i t i o n chosen.

The number of i t e r a t i o n s r e q u i r e d to achieve convergence f o r each

i n i t i a l p o s i t i o n i s shown i n Table 7.1.

It i s apparent upon examining the range of s t a r t i n g p o i n t s

used that the method i s not p a r t i c u l a r l y v u l n e r a b l e . For the

case where the s t a r t i n g p o i n t was x = 400', y = -50' the vertical

s t i f f n e s s i n the i n i t i a l p o s i t i o n was so s m a l l that the first

s o l u t i o n l e d to a value of y which was about 700' too low! Never-

t h e l e s s , the c o r r e c t s o l u t i o n was e v e n t u a l l y found, though twenty

i t e r a t i o n s were r e q u i r e d . For an i n i t i a l p o s i t i o n which was in

any sense reasonable, only f i v e or s i x i t e r a t i o n s were r e q u i r e d .


- 37 -

CONVERGENCE OF EXAMPLE 1.

CASE INITIAL POSITION NUMBER OF ITERATIONS REQUIRED

TO ACHIEVE CONVERGENCE.

X y

1 400 100 10

2 400 0 15

3 400 -50 20

4 400 -96.0495 7

5 400 -110 5

6 400 -120 5

7 400 -200 6

8 400 -300 6

9 350 -110 6

10 390 -110 6

11 410 -110 5

12 450 -110 8

Table. 7.1.
- 38 -

Example 2.

In t h i s example the e f f e c t s o f v a r y i n g the i n i t i a l cable

tensions and t h e b e n d i n g s t i f f n e s s i n t h e mast o f a g u y e d t o w e r

are i n v e s t i g a t e d . The t o w e r i s 1,000' h i g h (unstressed) and is

anchored at i t s t o p and m i d - p o i n t s b y f o u r c a b l e s a t e a c h level.

The c a b l e s a r e a n c h o r e d 700' from the tower. Areas, weights,

and l o a d i n g a r e as shown i n F i g . 7.1.

Four d i f f e r e n t slenderness ratios f o r t h e mast were consid-

ered: l / r = 310, l / r = 269, l / r = 240 and l / r = 219, where

1 = total tower h e i g h t . The c a b l e s w e r e s e t t o t h e p r e d e t e r m i n e d

tensions u n d e r t h e i n f l u e n c e o f c a b l e and mast d e a d l o a d s only.

The a n t e n n a l o a d s a t t h e t o p o f t h e m a s t and t h e w i n d l o a d s were

t h e n a p p l i e d , and t h e d e f o r m e d shape and member end f o r c e s found.

Beam - c o l u m n e f f e c t s on t h e mast w e r e c o n s i d e r e d as d e s c r i b e d in

Chapter 6.

The c a b l e s w e r e i n i t i a l l y t e n s i o n e d , f o r each slenderness

ratio of the mast, t o 10, 20, 30 and 40 k s i a t t h e i r bottoms,

giving a total o f 16 cases c o n s i d e r e d . Some t y p i c a l results

under the t o t a l l o a d a r e shown i n F i g s . 7.2 to 7.7.

The a x i a l f o r c e i n t h e mast was u n a f f e c t e d b y c h a n g i n g the

slenderness ratio, but increased markedly w i t h i n c r e a s i n g cable

pretension, as shown i n F i g . 7.2.


- 39 -

Note: the values of l/r = 310,269,


240,219 correspond to
r ( inches ) = 1500,2000,2500,3000
2 2

Note: similar cables out-of-plane

Guyed Tower

Fig. 7.1.
- 4 0 -

< 2 5 0 I 1 1 1 1 1_
10 20 30 40 50
Initial c o b l e s t r e s s (ksi)

Fig. 7.2.

The b e n d i n g moments i n t h e mast increased as t h e b u c k l i n g

l o a d was a p p r o a c h e d . Thus i n F i g . 7 . 3 we s e e t h e moment inc-

reasing as t h e p r e t e n s i o n or the slenderness ratio increased.

Fig. 7.3.
- 41 -

The deflections followed a predictable pattern: higher

initial guy t e n s i o n s decreased the deflections a t t h e guy attach-

ment points, which were r e l a t i v e l y unaffected by the tower slender-

ness ratio, but increased the deflections a t t h e 750' level as the

buckling load was approached. The d e f l e c t i o n s a t t h e 250' level

were quite small due t o t h e h i g h e r stiffness o f the lower half of

the mast. The d e f l e c t i o n s at the 1,000' and 750' points are plot-

ted i n Figs. 7.4 and 7.5.

to
>
1 6
O
O
2 5
_ +-
- a »
o o »

o
l/r = 3 1 0 , 2 6 9 , 2 4 0 , 2 1 9
v
O

10 20 30 40

Initial cable stress (ksi)

Fig. 7.4.

o
m
N 20
l/r = 310

5 *" 1 5

o a>
v
l/r = 2 6 9
o •*-

- 10 l/r = 2 4 0
•4-
a l/r :
219
a
JL
10 20 30 40

Initial cable stress (Ksi)

Fig. 7.5.
- 42 -

Finally, t h e guy tensions on the windward side increased

with initial tension, as might well have been expected. The

tension i n the higher cable was almost unaffected by the varying

stiffnesses o f t h e mast, while that i n the lower cable increased

with increasing slenderness o f the mast, as indicated i n Figs. 7.

and 7.7. " r • 219


l/r = 3 1 0

60

a. o
Q. O ^
3 —
•o in 50
e *•
*
tt T3
tt c

<D —
£ X
40

10 20 30 40

Initial cable tension stress (ksi)

Fig.. 7.6..

Fig. 7.7.
- 43 -

This example i s n o t i n t e n d e d t o be a case study, but

rather to point out the f a c i l i t y with which t h e method can handle

otherwise i n t r a c t a b l e problems. Over t h e 16 solutions performed,

on average o f 2.5 iterations were required to set the pretensions

in the cables, and on average of 3*25 iterations were required

to solve the l i v e - l o a d condition. The full s e t o f 16 solutions

were performed i n about 70 seconds on an IBM 360-67.


_ 44 -

Chapter 8. Discussion.

In general, t h e Method proposed herein will converge from

the initial position to the nearest stable equilibrium position.

The closer the i n i t i a l and f i n a l positions, the faster t h e Method

will converge.

If the structure starts a t , or erroneously wanders into,

an unstable configuration, t h e s t i f f n e s s method will break down.

It i s rather hard t o make a reliable prediction as t o whether the

Method will converge or not f o r a given structure. We may* s a y ,

however, that i fa structure has a well-defined position of stable

equilibrium, and i s n o t u n s t a b l e close to this position, then f o r

an initial position reasonably close to this point t h e Method will

converge to i t . Given the present state of mathematical knowledge

regarding Newton's Method, we c a n s a y no more. More o p t i m i s t i c a l l y ,

we can point out that, i n t h e problems so f a r p r e s e n t e d to i t , the

Method has n o t f a i l e d t o converge on a s t a b l e solution i f one

existed.

The advantages o f t h e Method may be summarized:

(1) The r e s u l t i s an exact solution of the equations

chosen to describe the behaviour of the structure.

(2) The Method c a n be a p p l i e d to structures which contain

many cables (slack or taut) a n d many non-cable elements.

* By t r a n s l a t i n g t h e m a t h e m a t i c a l proof o f convergence given by

Goldstein (3) (Appendix 3) into structural terminology.


- 45 -

(3) Already existing stiffness analysis programs can be

adapted to the Method without great difficulty.

And i t s disadvantages:

(1) Because of the generality of the Method, i t does not

solve certain restricted types of problems as effic-

iently as more specific methods.

(2) The Method may break down (though this appears to

be a very rare occurrence).


- 46 -

Bibliography.

Livesley R.K., "Matrix Methods of Structural Analysis",


Pergamon Press, 1964.

I967.
John F., "Lectures on Numerical Analysis", Gorden and
Breuch,

K a n t o r o v i t c h L . V . a n d A k i l o v G.P., "Functional Analysis


i n N o r m e d S p a c e s " , M a c M i l l a n , New Y o r k , 1964.

Press, I965.
Home and Merchant. "The S t a b i l i t y o f Frames", Pergamon

R e l f E . F . a n d P o w e l l C.H., " T e s t s on Smooth and S t r a n d e d


W i r e s I n c l i n e d t o t h e Wind D i r e c t i o n , a n d a C o m p a r i s o n o f
R e s u l t s on Stranded Wires i n A i r and Water". Associated
R e s e a r c h C o m m i t t e e , R & M 30?, L o n d o n , J a n . 1917. Results
reproduced i n :
N a t i o n a l R e s e a r c h C o u n c i l o f Canada, Report MER-1
"The A n a l y s i s o f t h e S t r u c t u r a l B e h a v i o u r o f G u y e d

1956.
A n t e n n a M a s t s Under Wind a n d I c e L o a d i n g " , Ottawa,

M i c h a l o s J . and B i r n s t i e C , "Movements o f a C a b l e Due t o


Changes i n Loading", Trans.. A S C E , V 1 2 7 , 1962. Part 11.

O ' B r i e n W.T. a n d F r a n c i s A . J . , " C a b l e M o v e m e n t s U n d e r Two


Dimensional Loads", J . S t r . D i v . A S C E , V 9 0 , No. S T 3 • J u n e
1964. Part 1.

Row, I967.
Goldstein A.A., "Constructive Real Analysis", Harper and
- i -

Appendix 1. Listing o f FORTRAN S u b r o u t i n e to Calculate

Cable-End Forces.

The following is a listing of a Fortran Subroutine which

u s e s Newton's Method to c a l c u l a t e the end-forces of a cable

as a function o f known USL, s a g a t some p o i n t x, o r end tension

T o r T. .
o 1

Also included are three function sub-programs to calculate

the d e r i v a t i v e o f t h e known f u n c t i o n w i t h r e s p e c t to the h o r i -

zontal components of cable tension (H).


- i i -

APPENDIX 1

C-**** SUBROUTINE TO DETERMINE CABLE END FORCES

SUBROUTINE CABPOS<W,EL,V,USL,AREA,E,H,A,B,T0,Tl,SAG,X)
I M P L I C I T REAL*8tA-H,0-Z)

C***» FOR GIVEN : w = LOAD IN POUNDS PER LINEAR FOOT


c***« EL HORIZONTAL LENGTH
C **** V = VERTICAL LENGTH
c«*** USL = UNSTRETCHED LENGTH OF CABLE
c**** AREA AREA OF CABLE
SAG SAG OF CABLE AT X
c**** E = MODULUS OF CABLE
TO = TENSION AT BEGINNING
c**«* Tl TENSION AT END
c***«
C**«* NOTE: I F SAG.NE.O CALCS WILL BE BASED ON SAG
C***« I F SAG.EQ.O AND TO OR T l NE.O CALCS WILL BE BASED ON
C***« TENSION, OTHERWISE CALCS WILL BE BASED ON USL
C****
C***« THIS PROGRAM USES A NEWTON-RAPHSON METHOD TO CALCULATE THE
C***« POSITION OF THE C A B L E , AND RETURNS VALUES OF:
C**** USL
C*«** H = HORIZONTAL TENSION
C**«* T 0 , T 1 {NOTE: IF VERT COMP < 0, T I S SET < 0)
C**«* SAG AT X
C**«* A,B = CONSTANTS IN THE CABLE EQUATIONS:
C***«
C«*** Y = H/W*COSH(W*X/H+A)+B
C**** Y* = SINHtWX/H+A)
C**#*
NITER = 0
C* * « * GUESS A VALUE OF H, I F NECESSARY
IFtH.EQ.O.>H=W*EL/2.
IFC H.LT.(W*EL/20.))H=W*EL/20.
AE=AREA*E
1 CONTINUE
IFtSAG.NE.O.)G0 TO 7
IFCTO.NE.O.O.OR.Tl.NE.0.0)G0T011
CALL DUSLDHtW,EL,V,AE,H,USL$,DERI V)
F=USL-USL$
C**«* F = ERROR IN CALCULATED USL
IF{DABS(F/USL).LT.1.D-8)G0T02
G0T08
7 CALL DSAGDH(W,EL,V,AE,H,X,A,B,SAG$tDERIV>
F=SAG-SAG$
C***# F = ERROR IN CALCULATED SAG
IF(DABS(F/SAG).LT.1.D-6)G0T02
GOT 08
- Ill -

11 I F ( T O . H E . 0 . ) C A L L DTODH (W, E L , V, AE,H, T$ , D E R I V , & 1 2, & 1 3)


C A L L DT1DH ( W , E L , V , A E , H , T $ , D E R I V,& 12,&13)
12 F=ro-rf
C **** F = ERROR I H C A L C U L A T E D T E N S I O N AT B E G I N N I N G
I F (DABS ( F / T O ) . L T . 1.D-6) G 0 T 0 2
G0T08
13 F=T1~T$
c* * * * F = ERROR I N C A L C U L A T E D T E N S I O N RT ESD
I F (DABS ( F / T 1 ) . L T . 1.D-6) GDT02
8 DELTAH-F/DERIV
IF{(DELTAH+H/2.).LT.O.)DELTAH=-H/2.
H=H+DELTAH
KITEB=BITER+1
IF(NITER.LT.13)G0T01
W R I T E ( 6 , 1 0 0 ) H,F
100 F O R M A T ( ' CABPOS: NO SOLS AFTER 12 I T E R A T I O N S . H=•,F9
* • F=',F9.6))
C * * * * WRAP UP - G E T UNKNOWNS
2 I F ( S A G . N E . 0 . ) G O TO 9
IF(TO.HE.0.0.OR.T1.NE.0.0)GOT015
C A L L DSAGDH ( W , E L , V , f c E , H , X , A , B , S A G $ , D E R I V )
SAG=SAG$
GO TO 10
9 CALL DUSLDH(W EL,V,AE,H,USL$,DERIV)
f

USL=USL$
GOTO10
15 C A L L DSAGDH (W,EL,V,AE,H,X,A, B, SAG, D E R I V )
CALL DUSLDH(W,EL,V,AE,H,USL,DERIV)
10 CONTINUE
S0=DSIHH(A)
S 1 = D S I » H (W*EL/H+A)
T0=H*DSQRT ( 1 . +S0*S0) * D S I G N (1. DO,SO)
T 1=H*DSQRT (1. +S 1 * S 1 ) * D S I G N ( 1 . D0,S1)
WRITE(7,101)NITER,H
101. FORMAT(13,' I T E R A T I O N S . H=',F13.5)
RETURN
END
- iv -

SUBROUTINE DUSLDH (W , EL, V, AE, H , US L$, DERIV)


I M P L I C I T REAL*8 (A-H,0-Z)
S/R TO FIND USL AMD D(USL)/DL FOR k CABLE
BE=V/EL
GA=W*EL/2./H
SHGA=DSINH (GA)
CHGA=DCOSH (GA)
DE=H/AE
EP=DSQST(BE*BE+SHGA*SHGA/GA/3A)
rH=DE*(BE*BE*GA*CHGA/SHGA+0.5+SHGft*CHGA/2./GA)
OSL$= (EP-TH)*EL
EIA=SHGA*CHGA/EP/G&-SHGA*SHGA/GA/GA/EP
PSI=BE*BE*GA* (CHGA/SHGA-S ft/SHGA/SHGA)-SHGA*CHGA/2./GA
1+(1.+2.*SHGA*SBGA)/2.
DERIV = EL/R* (- ET A - T E* D E* PSI)
RETURN
END
- V -

SUBROUTINE DSAGDH(W,EL,V,AE,H,X,A,B,SAG$, DE81V)


I H P L I C I T REAL*8 (A-H,0-Z)
c* * « * s/R TO FIND SAG, D (SAG) /DH, A AND B FOR A CABLE
C**** NOTE THAT THE EQUATIONS ARE SLIGHTLY REWORKED
C**** FROH THOSE IN THE THESIS IN ORDER TO INCREASE
C**** COMPUTATIONAL EFFICIENCY
AL=AE/EL
BE=V/EL
GA=W*EL/2./H
SHGA=DSINH (GA)
CHGA=DCOSH {GA)
A=DRSINH(GA*BE/SHGA) -GA
B=-EL/2./GA*DC03H (A)
SAG$=BE*X-EL/2./GA*DC0SH (2. *GA*X/EL*A) -B
DE=H/& E
DGADH=-GA/H
DADH=1./DSQRT (1.+GA*GA*BE* BE/SHGA/SHGA) *(BE/SHGA-GA*BE
1/SHGA/SHGA*CHGA)*DGADH-DGADH
DBDH=EL/2./GA/GA*DCOSH(A)*DGADH-EL/2./GA*DSINH(A)*DADH
DERIV=-1./W*DC0SH(2.*GA*X/EL+A)-EL/2./GA*DSINH(
12.*GA*X/EL+A)* (2,*X/EL*DGADH+DADH)-DBDH
RETURN
END
- vi -

S U B R O U T I N E DTODH (W , E L , V, AE, H, T $, DER IV , *, * )


I M P L I C I T R E A L * 8 (A~H,0-Z)
c **«* S / B T 0 F T E N S I O N AND D ( T E N S I O N ) / D H FOR A C A B L E
I H D

TMOLT=-1.
GO TO 1
ENTRY DT1DH(W,EL,V,AE,H,T$,DERIV,*,*)
TMULT=1.
1 CONTINUE
AL=AE/EL
BE=V/EL
GA=W*EL/2./H
SHGA = DSINH{GA)
CHGA = DCOSH (GA)
DE=H/AE
E P = D S Q R T ( B E * B E + S HGft*SHGA/G A / 3 A )
ETA=SHGA*CHGA/EP/GA-SHGA*SHGA/GA/G&/EP
PHI=BE*GA*GA/SHGA/SHGA
VV=AL*GA*DE*EL*(BE*THULT*CHGA/SHGA+EP)
T$=DSQRT (H*rI + VV*VV)
DVDH=TMULT*PHI-GA*ETA
D E R I V = 1 . / T $ * (H + VV*DVDH)
I F ( T M U L T . L T . 0 . 0 ) RETURN 1
RETURN 2
END
- v i i -

Appendix 2. Derivation of Terms i n the Cable Matrix.

In equations (4.2), (4.3) and (4.4) the stiffness matrix

for a cable was defined i n terms of several derivatives. In this

Appendix these d e r i v a t i v e s are evaluated.

In a d d i t i o n to the symbols defined i n Chapter 3t w e


will

require an additional function:

The unstressed length of a cable i s constant, so:

a^SL . a i £ i A ) , 0 A 2 a

a h dh

So :

ac . aA
ah " ah A2.2

From equation (3.5)

-3C - O t V BH_
ah ' p/
€ og -
ah A
2.3

And from equation (3.10)

A2.4
- v i i i -

Solving equations ( A 2 . 3 ) and (A2.4) f o r ^tL


3h

| y - ^ [ 2 y S c o , h y - ± ]
A2.5

Differentiating e q u a t i o n s (3.7) and (3.6) we find:

|^=a[ 8(coth /9/ )] + «t>-yr)) | i i


r r+ €

A2.6

3H 3V
2 y S c o t h * / j - 2<j> t

3h 3h +
3h
A2.7

As defined i n Chapter 4 , the approximate term:

3h " 2 3h 3h
A2.8

Or = a/Scothy + ^> A2.9


As i n equation (A2.1)

3USL 8(C-A)
3L 3L ' U A 2
- i o

So

3C 3A
3L " 3 L
A2.ll
- ix -

From e q u a t i o n (3.5)

8L

€ +
>
7
o£ 8l_ A2.12

And f r o m e q u a t i o n (3.10)

3A
a A2.13

Solving equation (A2.12) and (A2.13) f o r


dH
dL

dH
-a
dL A2.14

Or

dL ""^aF A2.15

And d i f f e r e n t i a t i n g equations (3.7) and (3.6) we find:

U-a[ 8( -%+ i , - * v ) ] ( « - ^ ) | H
y € + A2.16

And
9H , 3V,
A2.17

Knowing H . e q u a t i o n (A2.5), (A2.9) and (A2.15) are readily

evaluated to provide the cable s t i f f n e s s matrix o f equations (4.3)

or (4.4)
- X -

Appendix 3. Newton's Method.

Newton's Method (also called the Newton-Raphson method) is

one of the oldest, simplest, and best procedures for the solution

of non-linear equations. It i s somewhat surprising therefore, that

it was not until comparatively recently that mathematicians were

able to come to grips with i t .

Theorems d e s c r i b i n g the convergence of Newton's Method may

be found i n Kantorovitch (3), John (2) and Goldstein (3)» Quantit-

atively these theorems are of little use to us, but qualitatively

they are most valuable i n that they show that Newton's Method will

converge to a stable solution (assuming one exists) for a structure

provided only that the initial point is close enough to the solution.

Consider the f o l l o w i n g theorems* (based on Goldstein's

Theorem 1, Chapter C-4, page 143, which in i t s turn was based on

Kantorovitch's work**). In i t let:

be d e f l e c t e d shapes of the structure.

be the stiffness matrix at X


K
(x)
UBL, . be the unbalanced loads at X
(x)

Which we will merely state, and not prove. The insistent'reader

may verify for himself i t ' s isomorphism to Goldstein's theorem.

Some r e a d e r s might find Kantorovitch somewhat too h i r s u t e to be

readily digested, hence the references to Goldstein and John,


- x i -

Let a point X 0 (the i n i t i a l p o i n t ) Le given f o r which

K (x )
_1
0 exists.

Set 7)o = || K'Hxo) UBL (Xo) ll= the


length o f the incre-

mental deflection vector calculated at X 0 .

Define the sphere S such that i tcontains a l l X where:

II x- X o 11$ 27? 0

( S i s thus a sphere i n deflection space centered at Xo

and having a radius equal t o 27] ). 0

Define^ = 0 || K ^ ( X o ) ! ! = the inverse o f the smallest

eigenvalue o f K (x) 0 . (In a highly stable position of

the structurej3ois small and p o s i t i v e , i t becomes

larger as an u n s t a b l e position i s approached, and i s

infinite at a position of neutral equilibrium).

If a numberk exists such that:

II K(x)- K(y)ll> k l l x
^y'! fora l l x and y i n S

A n d !
/§o7? k$l/2 0

Then the structure has a p o s i t i o n of stable equilib-

rium i n S and the Newtonian sequence defined by:

x, =
+1 Xj - K'^Xj) UBL (Xj)

converges quadratically to i t .
- x i i -

Note that the closer the i n i t i a l position X 0 i s to the solu-

tion, the smaller w i l l be7) and


0 hence the sphere S . The s m a l l e r

the sphere S i s , t h e more n e a r l y will the s t i f f n e s s matrix vary

linearly with deflections across S and t h e s m a l l e r w i l l k be. Thus

the closer the i n i t i a l point i s to the solution, the better the

chances (and the f a s t e r the rate) of convergence.

What i s the e f f e c t o f u s i n g an approximate matrix? John

shows (Chapter 2.12) that i f the error i n the approximate matrix i s

bounded then the convergence criteria become harsher, b u t t h e same

general statement c a n be made: i f the i n i t i a l point i s close

enough to the (stable) solution then convergence i s assured.

You might also like