NASA TECHNICAL NOTE

N A S A TN_ D-5782 -

e. 1

A FINITE ELEMENT APPROACH TO THE STRUCTURAL INSTABILITY OF BEAM COLUMNS, FRAMES, A N D ARCHES

by JeweZZ M . Thomm
8

George C. Marsha ZZ Space FZight Center Marsh&, A h . 35812

i
1

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION

WASHINGTON,

D. C.

MAY 1970

TECH LIBRARY KAFB, NM

-~
~~

1. 4.

REPORT NO.

GOVERNMENT ACCESSION NO.

lllllll0132513 llllrllllllllll~llllll
I

- NASA TN D-5782
T I T L E AND S U B T I T L E

15. REPORT D A T E

A Finite Element Approach to the Structural Instability of Beam Columns, Frames, and Arches
7.
AUTHOR(S)

16: PzO:tz:

ORGANIZATION CODE

J e r r e l l M. Thomas
9.
PERFORMING O R G A N I Z A T I O N N A M E A N 0 ADDRESS

i
I I
I

8 . P E R F O R M I N G O R G A N I Z A T I O N REPOR

M-351
WORK U N I T NO.

10.

George C. Marshall Space Flight Center Marshall Space Flight Center, Alabama 35812
12. SPONSORING AGENCY N A M E A N 0 ADDRESS

981-10-10-0000-50-00-08
11 1.
CONTRACT OR GRANT NO.

13. T Y P E OF REPOR-; & P E R I O D COVERE

National Aeronautics and Space Administration Washington, D.C. 20546

~~

Technical Note
SPONSORING AGENCY CODE

' . I .

15. S U P P L E M E N T A R Y NOTES

Prepared by Astronautics Laboratory Science and Engineering Directorate

16. ABSTRACT

A nonlinear s t i f h e s s matrix for a beam column element subjected to nodal forces and to a uniformly distributed load is developed from the principle of virtual displacements and the bifurcation theory of elastic stability. Three cases of applied load behavior a r e considere The buckling load of a uniform circular arch is calculated as an example.

1 7 . ' K E Y WORDS

18. O l S T R l B U T l O N S T A T E M E N T

Unclassified

- Unlimited

19. S E C U R I T Y C L A S S I F . (of this report)

Unclassified

iI

20.

SECL

IF. (of this pege)

21.

NO. O F P A G E S

22.

PRICE*

Unclassified

58

$3.00

TABLE OF CONTENTS

INTRODUCTION....

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

1
3 3

GENERALTHEORY.................................... Basic Assumptions and Limitations Method of Analysis Development of Element Stiffness Matrix Description of Element Displacement Functions Development of Equilibrium Equations Bifurcation Theory of Instability Conventional Stiffness M a t r i x , K Geometric Stiffness Matrix, ,G Load Behavior Matrix, L. Coordinate Transformation Master Stiffness M a t r i x . Stability Criterion Primary Equilibrium State.
N N

. . . . . . . ... . . ........ . ......... .. ..... . .. .. . . . . .. . .... . . . . . . . . .. ....... . .. . ... .. . . .. .... .. . . . . . ..... . .. . . . ... .... . . ... .. . . . .. . . .. .. . ....... . . . . .. . . . .. . ... .. . . . . . . . . . .. .. .. . . . . . . .. .. ..... . .. .. . . . ... .. .. . . . . .. . ... .. .. . . . . .. . . .. .. . . .... . ... .. . . .. . . .. .. . . .. . .. . .. . . . . .. . . . .. . .. . ... . . . . . . . . . . .. . .... . . . .... . . . . . . . . . . .. . . . . ... APPLICATIONOF THE THEORY ...........................
EXAMPLE P R O B L E M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CONCLUSIONS........................................

4
5 5 5 7 16 20 21 28 32 35 37 38

39

41 44 46

REFERENCES

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

iii

111ll1llIll I I1 I I

L I S T OF ILLUSTRATIONS
Figke Title Page
5

i .
2.

3.

............................. Loads Remain Normal to Element. ..................... Loads Remain Parallel to Original Direction ..............
Beam Column Element

10

11
12
32 35

4.
5.

6.

7.

............. Coordinate Transformation .......................... General Structure ................................ Uniform Circular A r c h . ............................
Loads Remain Directed Toward a Fixed Point

42

L I S T OF TABLES
Table I. Title Comparison of Arch Buckling Loads Page

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

44

ACKNOWLEDGMENT
This analysis was done at the suggestion of D r . C. V. Smith of Georgia Institute of Technology. M r . John Key assisted in checking certain portions of the analysis. M r . Dan Roberts of Computer Sciences Corporation did all of the programming f o r the example problem. The contributions of these people a r e gratefully acknowledged.

iv

L I S T OF SYMBOLS
Symbol A
E

Meaning Cross sectional area of element Modulus of elasticity of material Force applied to element node Force applied to element node referenced to undeformed element coordinate system Proportionality constants between forces, F , and lateral load, p Geometric stiffness matrix defined by equation ( 4 6 ) Moment of inertia of element c r o s s section Conventional stiffness matrix defined by equation (45) Total element stiffness matrix (see equation (44)) Element stiffness matrix in global coordinate system Nonlinear stiffness matrix defined by equation ( 9 1 ) Length of element Load behavior stiffness matrix Nodal force behavior stiffness matrix Lateral load behavior stiffness matrix Distance from element node to point through which the F X nodal forces are directed

F
+

G = [GI
N

I
N

K = [K] K f = [ Kf]*
N

K = [ K]
N

K O = E KO1
N

L
N

L = [L]

kF =

LF1

L ILJ P

L I S T OF SYMBOLS (Continued)
Symbol M M=M Meaning Bending moment applied to element node Bending moment applied to element node referenced to undeformed element coordinate system Unit vectors in element coordinate system Unit vectors in global coordinate system Lateral load applied to element Nodal force o r moment Column vector of nodal forces and moments Column vector of nodal forces and moments referenced to global coordinate system Nodal displacement o r rotation Column vector of nodal displacements Column vector of nodal displacements in global coordinate system

m, n

iii,
P

Q
N

Q = {Q}

g = {a}

R
T = [TI

Distance from element centerline to point through which lateral forces are directed; also radius of circular arch Coordinate transformation matrix defined by equation (81) Strain energy of the element Displacement in x direction Displacement in Z direction

U
U U

vi

LIST OF SYMBOLS (Continued)

Symbol
V W W

Meaning Volume of the element Displacement in z direction Displacement in E direction Element coordinates Global coordinates Constant coefficient Angle between element and global coordinate systems Displacement functions defined by equation ( 6 ) Change in strain energy, virtual work of external forces Strain of element in x direction at any point Strain of element midsurface in x direction Rotation of element node Curvature of beam element centerline

x, z

z, z
CY

Y
6 U , 6V

k
E

xx

Stress in x direction at any point in the element Displacement functions defined by equation ( 5 ) Unit rotation vector

dl

vii

L I S T OF SYMBOLS (Concluded)
Symbol Meaning

Matrices defined by equations (55) through (61)

Subscript
1, 2

Meaning Denotes left and right nodes respectively of the elements; also denotes a particular displacement function, y Denotes a particular @ displacement function Denotes a particular element of a general structure Denotes a particular node of the element or one of the displacement functions, y or @ Indicates no summation on double subscript, i Direction of the coordinates x and z

9-x;

rmc
1

F i r s t and second derivatives with respect to x Meaning Denotes displacements of the structure just prior to buckling Denotes displacements of the structure just after buckling Inverse of a matrix Transpose of a matrix

Superscript
0 I
-1

viii

A FINITE'ELEMENT APPROACH TO THE STRUCTURAL INSTAB l L l T Y OF BEAM COLUMNS, FRAMES, AND ARCHES SUMMARY
From the principle of virtual displacements and the bifurcation theory of elastic stability a stiffness matrix is developed for a beam column element with s h e a r , moment, and axial load applied to the ends ( nodes) of the element and a uniformly distributed load applied along the span of the element. The stiffness matrix developed relates the forces applied at the nodes to the displacements of the nodes and is a function of the magnitude of the applied load. Three types of behavior of the applied loads are considered as separate cases. These are: the loads remain normal to the deformed element; the loads remain parallel to their original direction; and, the loads remain directed toward a fixed point.
A method is given for using the element stiffness matrix to predict the buckling load for a structure which may be represented by beam column elements. A s an example, the buckling load of an arch for each of the three load-behavior cases is calculated and compared to known solutions.

INTRODUCTION
In recent years interest in the solution of structural problems by the finite element method has greatly increased. This interest can be attributed to the fact that the digital computer has made possible the solution of a great number of complex, yet practical, problems by this method. Those active in publishing results in the field include a group at the University of Washington and The Boeing Company [ 1-91, J. H. Argyris at the University of London [ I O , 111, and R. H. Gallagher and his associates at Bell Aerosystems Company [ 12- 141. Basically the method consists of representing an actual structure by an idealized structure made up of elements for which relationships between the forces applied to points (nodes) on the boundary of the element and the displacements of the nodes is known. This relationship is conveniently represented by

a n element stiffness matrix. If all of the element stiffness matrices are transformed to a common (global) coordinate system, the sum of.forces applied to the elements meeting at each node is equated to the externally applied force at that node, and the displacements of elements which meet at each node a r e equated, a s e t of equations relating the externally applied forces to the nodal displacements results. When this s e t is expressed in matrix form, the matrix relating the externally applied forces to the nodal displacements is called the master stiffness matrix. Then for any given combination of known applied external forces and nodal displacements the above relationships may be algebraically manipulated to determine internal loads and displacements throughout the structure. Once a sufficient variety and quality of element stiffness matrices a r e available to idealize a particular structure, the remainder of the above procedure to solve a particular problem becomes tedious but quite mechanical. Hence the advantage of the computer in solving such problems is obvious. Although the use of such techniques in solving complex structural problems is now an everyday occurrence, a great deal of developmental activity i n the field still persists for several reasons. First, there is no such thing as a correct stiffness matrix for a particular element which excludes other stiffness matrices for that same element as incorrect. A number of correct stiffness matrices may be developed for the same element if different assumptions a r e made for the derivation. Each of these matrices may have its own advantages for some particular application and all of them may converge to identical answers if the structure to be analyzed is broken into small enough elements. It is the search for the most advantageous element for particular applications o r for allinclusive elements which gives r i s e to much of the current research. Second, many derivations have been based on geometrical considerations only and have sometimes led to incorrect stiTfness matrices o r matrices with poor convergence characteristics. The recent trend has been to base derivations on basic elasticity principles and to use more c a r e in assuring that adjacent element deformations conform to each other not only a t the nodes but along the complete boundary 15, 15-19]. Third, extension of the finite element techniques to include a larger class of problems has received a great deal of recent attention. In particular, the solution of nonlinear problems including large deformations and structural instability has been of interest [ 1-3, 5, 8, 9, 13-15, 17, 201.

This report was motivated to some extent by all three of the above considerations. The objective was to develop a beam column element, beginning with fundamental principles, which would account for the applied load behavior and could be used for structural stability analysis.

It is assumed that the reader is familiar with the conventional matrix and index notation used throughout the report. Since there is no stringent space limitation, the work is presented in an unusual amount of detail with the presumption that the development could s e r v e as a guide f o r the development of curved or three-dimensional elements.

GENERAL THEORY Basic Assumptions and Limitations

The usual Euler-Bernoulli beam assumptions a r e made for developing the beam column element stiffness matrix. These are basic engineering assumptions and may be listed as follows:
I . The material of the element is homogeneous and isotropic.

2.
3.

Plane sections remain plane after bending. The stress-strain curve is identical in tension and compression. Hooke's law holds. The effect of transverse shear is negligible. The deflections are s m a l l compared to the c r o s s sectionai dimensions.

4 .
5.
6.

7 . The loads act in a single plane passing through a principal axis of inertia of the c r o s s section.
8.

No initial curvature of the element exists.

No local type of instability will occur within the element.

9 .
IO.

Loads are applied quasi-statically.

The above assumptions limit the application of results to structures that lie in a plane and are loaded in the s a m e plane. Out-of-plane and torsional buckling are not treated.

Method of Analysis
The work done during a virtual displacement of the element is equated to zero, as presented by Hoff in Reference 2 1 and derived m o r e rigorously by Langhaar in Reference 22, to obtain the equilibrium equations for the element. The bifurcation concept of elastic stability as presented by Novozhilov in Reference 23, Chapter V , is used to postulate that two possible sets of,displacements which satisfy the equilibrium equations may exist under the same magnitude of external load if the magnitude is such that a structure is unstable. Each of these sets of displacements is substituted in turn into the equilibrium equations. The resulting sets of equations are combined to obtain a relationship between the nodal forces and the nodal displacements during buckling. When placed in matrix form this relationship becomes

where Qi is the column matrix of nodal forces , is the column matrix of nodal displacgments, K is the usual beam element stiffness matrix, G is the geometric stiffness m a t r i x r a n d L is the load behavior stiffness matrix. G and L both contain the magnitudezf external load as a factor. Thus an e l e k e n t syiffness matrix can be derived which is a function of the applied load including the applied load behavior.
N

A number of such elements may be combined to represent a particular structure, s o that equation ( I) applies to the entire structure and K + G + L is the master stiffness matrix. Boundary conditions may be applied To r&ucg the s i z e of the m a s t e r stiffness matrix. For instability to exist the determinant of the master stiffness matrix must vanish. Hence, an eigenvalue problem is formulated where the eigenvalues are the magnitudes of the applied load at which the structure is unstable.

Development of Element' Stiffness Matrix

Description of Element. Consider a beam column element as shown in Figure 1 which is subjected to nodal forces and moments and a uniformly distributed load, p, applied along its length. It is desired to determine a stiffness matrix for the element which may be used to calculate the stability of structures made up of such elements. The stiffness matrix is to account for the fact that the components of the nodal forces o r distributed load, o r both, may be a function of the element displacements. Three cases a r e to be considered: the loads remain normal to the deformed elemeht; the loads remain parallel to their original direction; and, the loads remain directed toward a fixed point.

FIGURE I.

BEAM COLUMN ELEMENT

Displacement Functions. If it is assumed that the lateral displacement of the beam element of Figure i may be represented by

and the longitudinal displacement is given by

may be determined in terms of the nodal i displacements from the element boundary conditions, then the six unknown constants,
a !

w w, w
X

=wi

at x = O
at x = O
a t x = L
=L

=-el
=w2

w,
u

-e2 at x

=ui

atx=O a t x = L

=u2

The resulting displacement functions a r e

w=wi 1-3-+2

9 L3 +w2(3$2$

+ e , ( - x + 2 L

-2) L2 + e2($

5)

= w . @.(XI
1 1

These displacement functions have been used by others for the beam column element (see Reference 5, for example). However, it should be mentioned that a n inconsistency a r i s e s when the cubic function is used for the lateral displacement of the present element which has a distributed load. This is illustrated by taking the third derivative of w, which should be the equation for the shear load on the element, and observing that a constant results. But the element has a distributed load, and the s h e a r should obviously be a linear function, not a constant. Since the change in shear over the length of an element becomes negligible in the limit as the element becomes smaller and smaller, this inconsistency is not unacceptable and it will be seen that adequate results are obtained. The correct fourth order function could not be used f o r w because there is no other boundary condition available for evaluating another constant in equation ( 2 ) . This will be discussed further in the conclusions. Development of Equilibrium Equations. The element equilibrium equations will now be developed from the principle of virtual displacements,
6U-6V=O
(7)

where 6 U is the change in strain energy and 6V is the work done by the external forces during a virtual displacement.
For the uniaxial state of stress assumed to exist in a beam the change in strain energy is given by

6U

J ax6 exdv
V

Here, by

is the strain of the beam at any point of the c r o s s section and is given

where

xx

is the s t r a i n of the beam mid-surface, z is the distance of the point in

the beam from the mid-surface, and

K

xx

=W,

xx

is the beam curvature.

E =

The nonlinear strain-displacement relationship

2 + -1 u,
2

xx

u,x+2wx

is used for the mid-surface strain. The u , ~ term in this expression is not
X

known to have been used in previous derivations of nonlinear beam element stiffness matrices. Substituting equations ( 9 ) , ( I O ) , and (11) into equation
(8),

6 U = s [,,cxx6t L

xx

+ E I K 6K dx xx xx

where

d =6u, +d(iw,;) xx X

+d(&;)

= 6 u , x + w , x 6w, x + u , x 6% x

u , +~I w,: L

+I

zu,

2 ,
X

tu,
X

+ w , dw,
x x

+ u, ~ u , ~ ) + - E I w , ~ dx ~ ~ w ,
X

+ higher

order terms.

(14)

The virtual work of the external forces acting on the element is given by:

where the force components depend upon the load behavior. cases will be considered. Case I.
Pz = P
pX

Three load behavior

Loads Remain Normal to the Deformed Element (Fig. 2)

-PKx

xi

= F . + F e xi zi i

F zi = Fzi - F xi e i
Z. = M.
1
1

Then

pz

= p cos 9 H p

4
I
I

FIGURE 2.

LOADS REMAIN NORMAL TO ELEMENT

1 . Loads Remain Parallel to Their Original Direction (Fig. 3) Case 1

P, = P
p
X

=o

Fxi - Fxi
FZi = FZi

M.1 = M.1

10

FIGURE 3.

LOADS RENIAIN PARALLEL TO ORIGINAL DIRECTION

6V = F . 6 u . + F . dwi + M. 6 8 . i - s p6wdx x1 1 z1 l L
Case 1 1 1 . Loads Remain Directed Toward a Fixed Point (Fig. 4) Pz = P
P, =
PU

(19)

Ui Fxi = F . + F Z i ~ x1
FZi = FZi - Fxi

Wi 1

k. = M. 1 1

Fxi

F Z iu ) % 6 ~ +. FZi

Fxi?)6w

FIGURE 4.

LOADS REMAIN DIRECTED TOWARD A FIXED POINT

12

Substituting equations ( 14), (17) , ( 19), and ( 21) into equation (7), we obtain the following results: Case I.

{ [EA(USx

+z I w , 2 + -u 3

2)6u,x + EA(u,xw,x)6W,x

x
+ EIw ,xx 6w,

2 'x

-1

dx = Fxi ;FZi Bi

) (

6ui + FZi

- Fxi ,)6wi

Case 11.

xx

dx = F . 6 u .

+ F . 6 ~+. Misei+
L

p6wdx

Case 111.

F r o m the displacement functions, equations ( 5) and ( 6 ) u,

X

= u. y.
1

l,x
(25)

w, X = w i w, xx = wi

cPi,x
cPi,-

.
13

And the above variations can be written, 6u 6u, X 6w

= y . 6Ui
1

Yi,x 6ui

= Gi6Wi

6WYx = @ i , x d W i
6w 6WYn - @i,xx i

Upon substituting equations (25) and (26) into equations (22) through (24) and recalling that Oi = wi 2, the following equations are obtained:
+

Case I.

) 6 u i - ( F zi - F xi wi + 2) 6 w i

Case 11.

- F x1 .6u . - F z1 . 6 w1 .-M.6 w i+2 1 1

p@.6w.dx=O 1 1

14

Case 1 1 1 .

1
L
+( u.y. J J , X wk qJ k,x)'i,x

6wl
(

+ EIw. J @. 1,xx CPj , x x

+ FZi %)hi Ui - FZi - Fxi

P Swi + - u. y. y . 6ui R J J 1

"')

6wi - Mi 6wi+

(29)

For independent virtual displacements , 6u. and 6w

1

i'

the equilibrium

equations are obtained from equations (27) through ( 2 9 ) . Case I.

j , x i,x

I 3 + -2w w $ @. y + -2u u k j k,x J,X i,x ky j k , xy j , xy i,x

EAuj yj , x wk CP k , x CP i , x + E I wJ. @ I., =

L

CPj y =

-P(@i)]h

Case 1 1 .

[..

('jyj,xYi,x

+-w 2 k wj @ k,x

3 @. J , X yi , ~ + 2 % ~ j ~ k , x ~ j , x ' i , x

15

I I II 111111 I

111

._.. _.. . - . ..

.._._ . -

bA
F

Case 1 1 1 .

I
j'

( j '

,x 'i,

ZWk"j 'k, x 'j

,x 'i,

% ;

"j ' k ,

xj '

,x 'i,

+ F -= - Qzi xi 1

Wi

- o

(35)

Equations ( 3 0 ) through ( 3 5 ) a r e the final equilibrium equations for the beam column element. Bifurcation Theory of Instability. Let a solution of the equilibrium equations be uo w o According to the bifurcation concept of instability, at the
i' i'

instability load magnitude there is another set of displacements, arbitrarily close to the first set, which. also satisfies the equations of equilibrium. Denote this second set of nodal displacements by uo t. u1 w ? + w1 .
1

i ' i

Upon substituting this new solution into the equilibrium equations, there results for Case I

16

----

-. ........_...

... ..- -.

.,,

y @ @ j , x k,x i,x

(37)

Similar results are obtained for Cases I1 and 111, If these equations are expanded, if the q? state t e r m s (which themselves
1

satisfy the equations) are canceled, and if only linear t e r m s in the arbitrarily small q! state are retained, the following sets of equations result.
1

Case I.

FA
+

($'j,xyi,x

wj W k g k , x

+.

J,X

y. + 3Y(:uly 1,x J k, x j '

,x 'i,

+pw!@. y. d x - F 0' = O zi i J JYX 1

@ y +wi j\ i,x k,x j,x
F .e! = 0 x1 1

(38)

'+.i , x @.

J,X

k,x

+ EIW!+ + J i , x x j , x x dx
(39)

Case 11.

Y . ' 1

+W!W0 ~ J

y. k @k , x @. J,X 1

+~3 q u Ji yk , x y j , x 'i, x

I)

dx = 0

17

~[EA(u~w~'iy kx , x Yj , x + w ji u O k 9. 1 , x +. J , X Yk , x

+ E I WJ ~ +i,=

+j , =

dx

Case 111.

i,x k,x j , x

+wi

j\

'@ . @. Y 1,x J , X k , x

a These equations may be expressed in matrix notation saas follows:

three cases

or
[Kil (q'} = ( 0 )
where [K1] = [ K 1 + [GI + [Lp] + [LFl

18

s
[GI (q'} =

3EA

$ yk,xyj,x'i,x
$
y

1
dx

1s
1
I

EAwL'k,x'j,xYi,x

---------EAw0$

--I
dx

k i , x k,x j , x

' L

Is

- - - ----EA$$

i,x

@.

J,X

k,x

for all cases. The L matrices are different for each case.
N

Case I.

[Lpl

{si} =

(47 1

[L,l{ql}

(48)

Case 11.
[ L 1 = [L,I=[Ol P

(49)

Case 111.

Conventional Stiffhess Matrix

5,

Equation (45) represents the conven-

tional beam column e l e m e n t a t r i x , which is suitable for use when there is no interest in nonlinear effects o r instability. The elements of this matrix may be derived from the general expression, equation ( 4 5 ) . F o r example,

from equations ( 5) and ( 6 ) . Then,

Therefore,

dx=
0
N

EI(-j!$+%)
0

I2EI d x = - L3

Other elements of K a r e obtained similarly to yield:

20

UI

uz
LE L AE L
0 0

AEL

Fxi
0 0
0

AE -L
0

Fx2
FZi

12EI L3

--12EI
L3

6EI -6EI -L 2
L 2
L 2

[Kl =
0

0
0

12EI --L3

12EI 6EI L3

6EI L 2
FZ2

(52)

0
0

MI M2

Geometric Stiffness Matrix G.

-~
N

Equation (46) is the general form of the

geometric stiffness matrix which accounts for the effect of loads existing i n the element on the stiffness of the element. F o r example, it is well known that the axial load in a beam column has an appreciable effect on the lateral stiffness. The geometric stiffness matrix is an adjustment to the conventional stiffness matrix to account for such effects. Such matrices have also been referred to in the literature as stability coefficient matrices and incremental stiffness matrices. Elemehts of the G matrix will now be derived.
N

Evaluation of
0

@ k , x @j , x Y i , x d x .

For

i =1

21

when

k= I , j = I,

For

- 9

i=2 d

Other values of j and k are evaluated similarly to yield

22

~---..--

1 1 ~ . , ~ . , 1 1 . 1 . 1 1 1 1 1 1 1 1 1 1 1 . 1

111.111111111111.11.11111.1111111111111111111

1111111 II

1111 I I I I I 1111 1 1 1 1111111 11111 I I

II 1 1 1

L 'k.x
0

'

IOL

j , x Y2 , x d x =

-- I

iOL

I -

iOL

4 - -I 30
30
4 -

(54)

30

L Evaluation of
0

'i,x'j,x

'k,x

dx

For

i=I

j = I , k= 1

L
0

when

j =

I, k = 2

6 -- 6 5L2 5L2

J 0

dx = 'l,xCpk,xYj,x

i -

i0L

-- i

(55)

IOL

1 __

IOL

-- i i0L
23

i
IOL

-L-IOL

.f 0

'3,x'k,x'j,x

dx =

L
0 '4,x'k,xyjyx

I
IOL

-L-IOL

dx =

Evaluation of
0

yk , 'Yj

Yi , dx

Then,

24

. ..

.. .. ,.

Evalllation of

1
0

@.
1YX

@.
JYX

k,x

dx

The above matrices a r e now multiplied by the cyo state displacements as k indicated in equation ( 4 6 ) ,

-6
-5L2
6 5L2

6 5L2
6 ----

I IOL

I IOL

5L2

1 -- I IOL IOL

I
iOL

---I
IOL

4 30

I 30

25

-IOL

I --

I -- I IOL 30

4 30
L

Similarly,

wo
k

L 'I,x'k,x'j,x

dx = [wl 0 w2 0 el0

-e2J 0 5L2
7

-6

6 5 u

6 5L2

-- 6

5L2 i0L (64)

I -- I IOL

I IOL

-- I IOL

26

The above matrices a r e combined to form the G matrix (Equation ( 46) ) :

N

27

[GI =

Load Behavior Matrix. The effect of applied load behavior on the element stiffness is obtained by adjusting the 5 matrix with the k m a t r i x , where [ L l = [ L 1 + [LFl P'

(75)

The L matrix will now be derived from equations ( 4 7 ) throug,, (51) for each of the t&ee load behavior cases. Case I.

28

Then,

-0
0 0 0 0.

I 2 I -2 0
0 0 0

L
12 12

L
12
12

L -A .o
0 0

0 0

(761
0
0

0
0

-0

L may be written directly from equation (48) in view of equations ( 3 8 ) -F and ( 3 9 ) ;

0 0

- FZ2
0

[LFl

=P

FXl

(77)
" 2
0
0

Q
0

0
0 0

0
0

where F's are proportionality constants between the indicated forces and p. The magnitudes of the applied forces' are assumed to remain in a constant ratio to each other and to the lateral load, p, during loading of the element. This

29

assumption is made here and in the Stability Criterion section for illustrative purposes. The assumption is easily altered to study buckling under p r e s s u r e with specified nodal forces o r buckling under nodal forces with specified pres sure. Case 1 1 .
[ L 1 = [LFl = P

io1
Case 111.

dx

="[(x-r+3 X2 R

x3

) uj +($ -$) "11

L
0

--L

6R

0 0

0 0

0
0

--L 3R
0

[L I'P P

0 0 30

L may be written directly from equation (51) ; -F

[LFl = p

IO
4

-- F Z 2
R
0

0
4

FXl

1
0

0
4

Fx2

1
0

31

Coo r di nat e Transformation

The developments thus far have been in a coordinate system which was oriented s o that the x-axis coincides with the element centerline. To combine several elements f o r solution of a particular problem, it is necessary to obtain the stiffness matrices of the elements in a common or global coordinate system. This system w i l l be denoted by z as shown in Figure 5.

x,

FIGURE 5.

COORDINATE TRANSFORMATION

The figure shows that vectors in the two coordinate systems transform according to the relationships m = fi cos p

+ ii sin p

n = ii cos ,8 - GI sin p
o = w

- .

32

Writing in matrix notation, we see that

Thus the displacements and forces of the element transform according to

FZ

I
I

FZ2

G 2

33

or

If the nodal forces and displacements are related by the stiffness Ki (equation (44) ) , then
N

where Ki has been rearranged to conform to the Q and q matrices. stitutini equations ( 82) and ( 8 3 ) into equation ( 8 z ) ,
N

Upon sub-

But in this case,

34

where

if is the
N

stiffness matrix of the element in the global coordinate system.

Master Stiffness Matrix

The stiffness matrix f o r complete s t r u c t u r e is obtained by combining the element stiffness matrices in the global coordinate system. This section describes the method of accomplishing this combination. Consider p a r t of an overall structure schematically represented by . I s in Figure 6 are intended to represent the total load vector Figure 6. The &
1

FIGURE 6.

GENERAL STRUCTURE
35

at the point of application,

Load-displacement relationships f o r each of the three elements in the global coordinate system are expressed from equation ( 87) as follows:

By combining the above relationships with the fact that

gi is the same for

both elements b and c, and noting that

and that similar relationships hold for the other nodes, the master stiffness matrix is obtained:

36

_-0 0 0

Kai-i.i-l

+ Eb i - 1 , i - 1

Kbi-i,
j

%i, i - 1

Kbii

Kci, i + 1
Kci+l,i+l

+Ec i i

0
0

K . ci+l,i

txd i + l , i t

-_-

Note that this is a relationship between the externally applied loads and the nodal displacements of theassembled structure in the global coordinate system.

Stabi Iity C r i t e r i o n
Elements with stiffness matrices of the type given by equation (44) may be assembled into a m a s t e r stiffness matrix to represent a structure subjected to the critical (buckling) magnitude of applied loads. Thus,

[[Kl
where,

[Eol]{G1) = (0)

p[liol =

[GI + [Ep] + [ E , ]

(91)

and the null matrix of applied external loads indicates that the structural stiffness has vanished under the critical load magnitude ( a physical interpretation of instability). Boundary conditions may be applied to reduce the size of the s e t of equations ( 9 0 ) and the new reduced s e t again denoted by the equations (90 )

matrices above are obtained by assembly of element "P matrices as described in the preceding sections, The E matrix is m o r e "F conveniently obtained by direct application of equations (48) and (51) to each node of the assembled structure in the global coordinate system.
A nontrivial solution of equations ( 90) will exist only when the determinant of the matrix + p E, vanishes;
N

G, and The E, "

This is an eigenvalue problem where the magnitudes of applied load, p, at which instability will occur are the eigenvalues.
37

111.1111111

111111111

11111III

II

Primary Equilibrium State

It should be noted at this point that the foregoing solution is contingent upon a knowledge of the primary equilibrium state. That is, the G matrix cannot be formulated until the q odisplacements are known in t e r m s the applied loads. Given a s e t of applied loads, the nonlinear equilibrium equations (30) through ( 3 5 ) may be solved f o r the q odisplacements. An iterative procedure f o r incorporating this solution into the stability problem is given in Reference 1 4 .
Another approach which is consistent with many classical stability problem solutions is to linearize equations ( 3 0 ) through ( 3 5 ) for the purpose of obtaining the primary state solution. If this is done these equations become for all three cases

(93)

A l l t e r m s in the above equations have previously been evaluated and expressed in matrix form except the p@.t e r m which is as follows:
1

-0

38

or
0
0

L 2

{P@> = P

L 2
12 12

L2 -

L2 -

Equations ( 9 3 , 94) may then be expressed in the form

From this expression for each element, the m a s t e r stiffness matrix for the entire structure is assembled as before. Then for the complete structure,

A further simplification f o r obtaining 9' which has also been used extensively in classical problems is the use of a membrane solution for The structure is assumed to take no loads in bending before buckling. F o r many practical problems such a solution can readily be obtained by inspection, elementary equilibrium considerations, o r from the literature. For simplicity this approach is used in the example problem of this report.

zo.

APPLICATION OF THE THEORY

This section is intended to give a step by step approach for applying the theory to the solution of practical problems. For problems with several elements

39

either a general computer program would need to be developed o r a relatively simple computer program for each application could be developed as w a s done for the example to follow later. One begins by dividing the structure to be analyzed into discrete elements, the number depending upon the desired accuracy. From the known properties and loading on each element the K matrix can be calculated from equation (52) and the L and L matrices determined from the appropriate equations (76), (77), "P -F ( 7 8 ) , (79) , o r ( 8 0 ) . Note that the critical magnitude, p, of the applied loading remains as the unknown to be determined.
N

The qo primary state is now determined from a known membrane solution o r from equztion (97). The K matrix in equation (97) is assembled by using equations (81) and ( 8 6 ) to tr&sform elements to the global coordinate system and equation (89) for combining the elements. Boundary conditions a r e applied and equation (97) is solved for qo. Once the qo state is determined the G matrix can be found from equatl'uon ( 7 4 ) , using the definitions (53) throcgh ( 6 1 ) . The K' matrix defined by equation ( 4 4 ) is now known for each element. These a r e agsembled into a master stiffness matrix by equations ( 8 6 ) and ( 8 9 ) , and boundary conditions are applied to the resulting set of equations to obtain equations ( 9 0 ) . Eigenvalues of the characteristic determinant, equations (92) , which is obtained from equation ( 9 0 ) , a r e the magnitudes of the buckling load. The mode shape for each buckling load c.an be determined by substituting each eigenvalue in turn into equations ( 9 0 ) and solving f o r the relative amplitudes of displacements, 6 i'

EXAMPLE PROBLEM

Circular Arch With Uniform Pressure

The above application procedure w a s applied to the uniform circular arch shown in Figure 7. The known membrane solution for a circular a r c h under uniform pressure is

40

for all elements.

The m G . matrix is formulated as indicated by equation (74) :

EA

4 1

I Il I

I Il l l l1l 11l1l1l l Ill I I

-I zw +

MOMENT OF INERTIA = I = 0.314159 i n 4 (13.1 .m44 YOUNGS MODULUS = E = lo7 p s i (6.9 x 1010 N/m )

FIGURE 7. UNIFORM CIRCULAR ARCH Other elements of the elements

,G matrix a r e calculated similarly to yield for all

3
L L
0

-3 L

0
0

Fxl Fx2
FZl

3 - -3
L
0 -

0
6 -

6 -

5L
6 -

5L

I 10

I 10
10

[GI = P R
0 O

5L

-- 6 _ 5L -- i
10

-I _
io
4L 30

I L 30

(99)

Fz2
MI

I io

M 2 30 30 10 O 10 where the four t e r m s in the upper left corner can be traced directly to the term in equation ( ii ) u

I -- i

4L. -L - -

,x

42

Now for all elements the K matrix is given by equation (52)

N

matrices for the three load cases a r e given by equations respectively, and all L matrices a r e null. -F
N N

, the L "P (76) , (78) , and

(79) ,

The K , G, and L matrices are now all rearranged to the order, -F

I I I

4 2

I---I I
I I
Ki2

x, y by the operation,

- -

A l l element matrices a r e now rotated to the global coordinate system,

1
N

FXl

FZ I

Fx2

FZ2

M2

as indicated by equation ( 8 6 ) . The T matrix is of course calculated separately

for each element a s indicated by equation (81) where p is the angle measured clockwise from the element x axis to the iT axis. The element matrices a r e now combined to form the master stiffness matrix as indicated by equation ( 8 9 ) . The size of this matrix is reduced by applying the boundary conditions,

uI =
where

wi=~i=y=wl= 1= ~o

(102)

Y,wl' 81 are displacements of the right end of the last element.

From

43

this reduced matrix the characteristic determinant ( 92) is formed and eigenvalues are calculated. The above operations were performed on a computer for idealizations of 2, 3, 6, 9, and 12 elements. A hand calculation was also made for the two-element arch. Results of the analysis are shown in Table I. A s can be seen from the table, comparison with known results is excellent and there is rapid convergence to the exact solutions. The exact solutions w e r e obtained from the analysis by Wempner and Kesti; [ 241. The 12-element solution for Case I1 has also been previously obtained by the finite element method in Reference 14.

TABLE I.

COMPARISON OF ARCH BUCKLING LOADS Buckling P r e s s u r e (lb/in)

Number of Elements 2
3

~-

~-

Case I 94. I 57.1 57.4 56.6 56.4 56.87

1 Case 1

Case I 1 1 94. I 67.4 64.4 63.6 63.3 63.46

94. I 64.7 62. 6 61. 9 61. 6 60. 95

6 9 12 Exact solution
~~____

CONCLUS IONS
The results of the example problem above show that the theory developed h e r e may be quite useful in solving practical engineering stability problems .

44

The implicit assumption of uniform shear on the element mentioned in Development of Element Stiffness Matrix section does not appear to adversely affect the results. However, if desired, this assumption could possibly be term of eliminated by the technique introduced by Pian [ 191. Also, the u ,x equation (11) apparently has a negligible effect on the results for the particular example solved, since the exact solutions used for comparison do not contain effects of this term.
A very important item to note is that the present theory applies to unconservative systems (repreqented by Case I ) as well as conservative systems since the principle of virtual work was used. It is known, however, that certain unconservative systems a r e stable in the static sense (considered h e r e ) , but unstable in the dynamic sense. (See Reference 25, page 152 through 156. ) Thus the present analysis could not be expected to adequately treat this type of problem, and great c a r e should be used in applications to unconservative systems.

Fruitful extensions of the present theory would likely be found in the area of curved and three-dimensional elements and in a general automation of the theory application.

George C. Marshall Space Flight Center National Aeronautics and Space Administration Marshall Space Flight Center, Alabama 35812, May I , 1968
981- I O - I O - 0000-5 0- 00- 008

45

Il 1 Ill I1 I

REFERENCE S
1.
G r e e n , B. C. : Stiffness M a t r i x f o r Bending of A Rectangular E l e m e n t with Initial M e m b r a n e S t r e s s e s . S t r u c t u r a l Analysis R e s e a r c h M e m o r a n d u m No. 45, T h e Boeing Company, Seattle, W a s h . , Aug. 1962. K a p u r , K. K. and H a r t z , B. J. : Stability of Plates Using the F i n i t e E l e m e n t Method. J. Eng. Mech. Div. ASCE, vol. 92, 1966, pp. 177195. L u n d e r , C. A. : Derivation of a Stiffness M a t r i x f o r a Right T r i a n g u l a r Plate i n Bending and Subjected t o Initial S t r e s s e s . M. S. T h e s i s , Dept. of A e r o n a u t i c s and A s t r o n a u t i c s , Univ. of Washington, Seattle, Wash. , Jan. 1962. 4. M a r t i n , H. C. : Introduction t o M a t r i x Methods of S t r u c t u r a l Analysis. McGraw-Hill Book Co., Inc. , 1966. M a r t i n , H. C. : On the Derivation of Stiffness M a t r i c e s f o r t h e Analysis of L a r g e Deflection and Stability P r o b l e m s . Conference on M a t r i x Methods i n S t r u c t u r a l Mechanics , Wright-Patterson AFB, Dayton, 1965. Melosh, R. J. and M e r r i t t , R. G. : Evaluation of S p a r M a t r i c e s f o r Stiffness Analyses. J. A e r o s p a c e Sci. , vol. 25, 1958, pp. 537-543. T u r n e r , M. J . ; Clough, R. W.; Martin, H. C . ; and Topp, L. J.: Stiffness and Deflection Analysis of Complex S t r u c t u r e s . J. Aeron. Sci. vol. 23, no. 9 , Sept. 1956, pp. 805-824. T u r n e r , M. J. ; D i l l , E. H. ; Martin, H. C. ; and Melosh, R. J. : L a r g e Deflections of S t r u c t u r e s Subjected t o Heating and E x t e r n a l Loads. J. A e r o s p a c e S c i . , vol. 27, 1959, pp. 97-106. T u r n e r , M. J. ; Martin, H. C. ; and Weikel, R. C. : F u r t h e r Development and Applications of the Stiffness Method. P r e s e n t e d at AGARD S t r u c t u r e s and M a t e r i a l s P a n e l ( Paris, F r a n c e ) , J u l y 1962. Published i n AGARDo g r a p h 72, P e r g a m o n Press, 1964, pp. 203-266. A r g y r i s , J. H. : Matrix Analysis of Three-dimensional E l a s t i c Media, S m a l l and Large Displacements. AIAA J. , vol. 1, 1965, pp. 45-51.

2.

3.

5.

6.

7.

8.

9.

IO.

46

REFERENCES (Co nt inued)

11.
A r g y r i s , J. H. : Recent Advances i n Matrix Methods of S t r u c t u r a l Analysis. P r o g r e s s i n Aeronautical S c i e n c e s , P e r g a m o n Press , 1964. Gallagher, R. H. : A C o r r e l a t i o n Study of Methods of Matrix S t r u c t u r a l Analysis. AGARDograph 69, Jul. 1962. Gallagher, R. H. and Padlog, J. : D i s c r e t e E l e m e n t Approach t o S t r u c t u r a l Instability Analysis. AIAA J. , vol. I, no. 6, Jun. 1963, pp. 1437-1439. Gallagher, R. H. ; Padlog, J. ; Galletly, R. ; Mallett, R. : A D i s c r e t e E l e m e n t P r o c e d u r e f o r Thin-Shell Instability Analysis. AIAA J. , vol. 5, no. I , Jan. 1967, pp. 138-145. A r c h e r , J. S. : Consistent Matrix F o r m u l a t i o n s f o r S t r u c t u r a l Analysis Using Influence-coefficient Techniques. AIAA J. , vol. 3 , 1965, pp. 1910- 1918. Melosh, R. J. : B a s i s f o r Derivation of M a t r i c e s f o r the D i r e c t Stiffness Method. AIAA J. , vol. 1, no. 7 , Jul. 1963, pp. 1631-1637. Oden, J. T. : Calculation of G e o m e t r i c Stiffness M a t r i c e s f o r Complex S t r u c t u r e s . AIAA J. , vol. 4, no. 8, Aug. 1966, pp. 1480-1482. P i a n , T. H. H. : Derivation of Element Stiffness M a t r i c e s . v o l . 2 , no. 3 , Mar. 1964, pp. 576-577. AIAA J.

12.

13.

14.

15.

16.

17.

18.

19.

P i a n , T. H. H. : Derivation of Element Stiffness M a t r i c e s by Assumed S t r e s s Distributions. AIAA J. , vol. 2, no. 7 , Jul. 1964, pp. 1333-1336. Navaratna, D. R. : E l a s t i c Stability of Shells of Revolution by the Variational Approach Using D i s c r e t e Elements. ASRL TR 139-1 , BSD TR 66-261, Jun. 1966. Hoff, N. J. : T h e Analysis of S t r u c t u r e s . J o h n Wiley and Sons, 1956. J o h n Wiley

20.

21. 22.

L a n g h a a r , H. L. : Energy Methods in Applied Mechanics. and Sons, 1962.

47

REFERENCES (Concluded)
23. Novozhilov, V. V. : Foundations of the Nonlinear Theory of Elasticity. Graylock Press (Rochester, N. Y. 1 , 1953. Wempner, G. A. and Kesti, N. E. : On the Buckling of Circular Arches and Rings. Proceedings of the Fourth U. S. National Congress of Applied Mechanics (Am. SOC. of Mech. Eng. , N. Y. , 1962) , vol. 2. Timoshenko, S. P. and Gere, J. M. McGraw-Hill Book Co. , 1961.
:

24.

25.

Theory of Elastic Stability.

48

NASAyLangley, 1910

- 32

M351