%e Office of Naval Research

SContract
N00014-67-A-0109-0003

Task NR 064-496 '-I

Technical Report No. Zi 1.,

".4i

S2~ $
,9
b

THE LATERAL BUCKLING o0? A STRAIGHT OR

CURVED BEAM SUBJECTED TO PURE.BENDINGI'

by 7
F. E. Vnnslager

4•

SEP 6 i9 I-

Department of the Aerospace and Mechanical Engineering Sciences

UNIVERSITY OF CALIFORNIA, SAN DIEGO , .4
La Jolla, California
July 1968 .... .-
Reproduced by the •- . •',"' A101 iW~l•
, .-. Mlh
C ARI H0LlSE for pub",cr.....,'
for Federal Scientific & Technical I ,,L,-, .I-- -

Information Springfield Va. 22151 . . -"

 Table of Contents

Page
Abstract .................................... 1

Nomenclature ............................... iii

1. INTRODUCTION .............................. 1

2. BASIC -T,'UATIONS ........................... 2

3. METHOD OF SOLUTION ..................... 3

a. Curved Beam ......................... 3

b. Straight Beam ........................ 18

4. EXPERIMENTAL RESUJLTS .................. 20

5. SUMMARY ....................... ............ 31

APPENDIX ................................... 33

*REFERENCES .............................. 35

ii
 THE LATERAL BUCKLING OF A STRAIGHT OR
CURVED BEAM SUBJECTED TO PURE BENDING-

by

F. E. Vanslagert
Department of the Aerospace and Mechanical Engineering Sciences
University of California, San Diego
La Jolla, California

ABSTRACT

This paper presents a new analysis of the classical problem of

the lateral buckling under pure bending of a curved- beam with circular

axis. Although a solution to this problem, and to the co-iresponding

straight beam problem, is well-accepted in the literature, the present

analysis gives buckling~mornents which are significantly different than

the-usually accepted results. The difference arises from an application

of new boundary conditions tiat are shown to be the only boundary con-

•ditions consistent with the equilibrium of the buckled beam under a

pure bending moment. Thi'sinew analysis provides generally higher

bending moments. A short exp.erimental program was performed, and

the results tend to confirm the new analysis.

The results presented in this paper were obtained in the course of

research sponsored under Contract No. N00014-67-A-0109-0003
Task NR 064-496 by the Office of Naval Research, Washingtm, D.C.
t Research Assistant

I 1l
 Nomenclature

a kT A k arbitrayy constantsi
bk, Bk arbitrary constants

,C torsional rigidity aI. nut the centroidal axis

E Young's modulus

I,1 second moment of the rrea about the x and y axes

x y
k an eigenvalue of eq. (3) or'(23)

an eigenvalue of eq. (3)

L length o' a straight beam

M, M bending moment vector

xM My M s components of the bending moment vector

n an integer

Ps Pk-', P axial force

R mean radius

s coordinate along the centroidal axis

u displacement in the x direction

U work of deformation

v displacement in the y direction

V potential energy

W 1 , W2 work of external forces

x, X coordinate in the radial direction

y, y coordinate transverse to the beam

beam angle

iii
 unit twist about the centroidal axis

v an angle change

6 a length change
A change in radius of curvature
I

c an arbitrarily small constant

Xx, x curvature changes in the x and y directions

V Poisson's ratio

p deformed radius of curvature

angular displacement of the cross *section

*i

iv- --,
A
 1. INTRODUCTION

This paper is concerned with the well-known problem of the

lateral buckling under pure bending of a straight beam or of a' curved

* beam with a circular axis. A sketch of a curved beam of rectangular

cross section with mean radius R and opening angle a is shown in

Figure 1. It is assumed that the beam is loaded by moments M acting

a
II
R

A-A
A
Figure 1.

at s 0 and s ; aR and is otherwise unloaded. In order to have a con-

servative system, it is also assumed that the direction of the applied

moment vector does not change during the buckling process. When the

applied moment M reaches a critical value, the planar configuration

becomes unstable, and the beam begins to deform laterally. Section

A-A shows the displacement of a typical cross section of the buckled

 beam; the initial position is indicated by the dashed rectangle.

The classical solution to this problem, obtained by S. Timoshenko

and published in Bull. Polytech. Inst. Kiev in 191.0,. is widely accepted,

viz [1], [Z]*, and, to the author's knowledge, it has never beendisputed.

In-the first part of the present paper, a different solution is obtained for

both the circularly curved beam and the straight beam. The last part of

the paper describes an experimental verification of the new solutions as

opposed to the classical solutions. The discrepancy between the two

solutions is shown to-be due -to the boundary conditions prescribed onthe

loaded ends s = 0 and s = ceR. The only conditions that are consistent

vith,the basic equations [1] are shown to be those of zero geometric

constre-int- i. e., "free edges", and the solution for these conditions is

de.vied in section 3 below.

Z4 BASIC EQUATIONS

The linearized strain-displacement relations for thetransverse

curvature X the in-plane curvature change x and the unit twist ~

are [l], [3],

x= Sd2" ds v
R 2 X y= T'
u d2
d•2
u ,
' •=d +
1 ±_•d (v
(la)
~x ds +ds2
+~ ds R ds

where u and v are linear displacement components and rp is a

rotation as iihown in Figure 1. The stress-strain relations are given by

M M M
x El ' 1 (I b)
x y
* * __ _ _ _ Ely C
Numbers in brackets refer to references in the List of References. .

2 9
wvhere M M , and M are the components of the m6oinnt vector
x y S

M in the x, y, and s directions, I and I are the secon&dmoments,

x y
of the cro-,s section about the x and y axes, and, i' the torsional

rigidity about the centroidal axis. The equations of eqi!ibriui [ 1[] for

the buckled beam loa•d onl_ by end :moments M areý

x (s) M (P M
y (s)sM sS = Md -dS (2)

Substitution of X and
x
P from eqations'(lb) and (2) into, equations

(la), and elimination of v betmrý-e .A• •efirst and third ofiequations

(la) gives the classical results

ds2

where

k2 1 / El (
0 (M -- M (4)
x

The general solutions of this equation can be written as

= a~ s + b
oa
s o (5)

(Pk aka sin ks + b kcos ks

3. METHOD OF SOLUTION

a. Curved beam

The classical approach is to apply the "simply supported"

boundary conditions that (= 0 at s = 0 and s = aR, which leads to

3
 if
gk= ak sinks where k- and n"", 1, iZ,

or, alternatively, to apply the "built-in" boundary condiftions, that

d = 0 at s = 0 and s = aiX, which leads to

ds_
-= 0

ok = bk cos ks where k 7j and n 0, 1l, 2,... (7)

However, an inspection of equation (4) shows that both of these

solitions lead to a inite positive buckling moment for an infinitely I

long beam, since equations (4), (6), and. (7) show that k -. 0

(M - C/R or EI /R) as a'- -. Another objection is that, as shown

by equations (3) and (4), the minimum buckling moment M = 0 occurs

when k = ,and corresponds
11 to a rigid body rotation about a dliariieter*,

Therefore, since equations (6) and (7) would give k -for o = iT and
1, the buckli' gmoment would change from a small negative value

to a small positive value as the beamt, angle a passes through 7r.A

plot of the nege.tive buckling moment vs. beam angle ce would, there-

fore, show a discontinuous jump to the next higher mode at a = 7r. And,

finally, since the rigid body rotation solution with k = - should be in-

dependent of the angle a, and since equations (6) and (7) show that
1

RI
k = 1 occurs only for discrete vaudes'of a, we conclude that equaiions

(6) and (7) are probably not theý correct solutions to the rFoblem.

What, then, is wrong with this classical approach? For an

One could equally well take k = -- as the fundamental rigid body

R
rotation but this has somne notational disadvantages.

4
. *1
answer we must take a closer look at the statement of the problem. It

is simply the question of determining the magnitude of the equal and

oppositely directed end moments M, at which the beam begins to later-

ally deform in some unknowm:fashion. Equation (3) is, therefore,

derived by assuming that the deformed beam is in equilibrium with an

end moment M which moves with the beam, but whose direction

remains fixed in space. Since no geometric boundary conditions are

implied in the derivation of equation (3), the ends are essentially "free",

and any solution of equation (3) is potentially a solution of the problem.

In general, however, arbitrary boundary conditions will give rise to

shearing forces on the boundaries which were not included in the deriv-

ation of equation (3). Therefore, additional principles must be invoked

in order to determine what are the correct boundary conditions for the
problem. It turns out that, since k = is a solution to the pobler,

it is sufficient to require that the principle of stationary potential

energy be satisfied; i. e., in the vicinity of the true solution, t,,e Woa',

of deformation- must be equal to the work of the applied moment M.

In evaluating the potential energy for the beam, we will assume

that I is very large in comparison to I and C. The work of defor-

y x
-nation U will then be given by

U j EIZ
=-
aR
X0
2 ds
K× + C
d5+-
o'R
P2
pdsds• ,
o0 uan0

or, by using equaticon,, '1) and (2),

5
1,

U 2 2 R C 2 ds (8)
2E,1J sds + z C~~~) JO .dvs
x 00

The wc.rk of the applied moment can be split into two parts. The first

part is the work due to the "shortening" (or "lengthening") 6, which

causes an angle change y approximately given by

aR 2
z: • ds
R ZR ,'' ds

Since the applied moment stays constant during the buckling process,

'this part of the work of the applied moment will be simply given by My.

By using the relationship between v and p obtained from equations (1)

and (2), the Work due to the "shortening" can be written as

W1 MY= C~~f ds .s (9)

2MR( - TR

The second part of the work of the applied moment comes from.the

"rotation" of the ends due to the apparent opening or closing of the

ring. This rotation is equal to the integral along the length of the

beam of the curvature changes ( in the original plane) caused by the

buckling. The projection of the bending curvature on the original

plane contributes to this curvature change an amount

sin•p (p
x x El
x

The last expression has been obtained from equations (1) and (2). In

"addition, due to the swinging of the cross section through an arc

* 6 4
 °, "[

there is a small change in curvature of amount

1 1 _ 1 1 R(cos - )
R - p R R -A 2R2
ZR

Integrating these c'-rvature changes along the length of the beam to get

the total angle change and multiplying by the applied moment M, we

get that the work W% due to the "rotation" of the ends is given by

c•-R
w2 =(~~-~~)f
2pdi (10)
x

Subtracting W and W from the work of deformation. U, we find,

2

after some rearrangement, that the potential energy V can be written

as

M El a M a d
TEX )f P-M + s (11)

Using the definition of k given by equation (4), this can also be

written as

SMC [kaR 2
-aR -

2(M -) o 0

In this form it can be readily seen that the Euler-Lagrange equation

corresponding to the minimization of the potential energy V is the

equilibrium equation (3). This is as it should be, indicating that the

potential energy has been correctly derived.

We can now apply the principle of stationary potential energy.

We take a varied state in the form

7
 Si

*= Ok. + CO -1 (13)

where Pk is the assumed "true" solutionto equation (3) under the applied

moment determined by k from equation (4), p is the "true" solution

oi equation (3) with k replaced by t,, and E is an arbitrarily smia'll

constant. Substituting (13) into (12)-and requiring that the pQtential

energy be stationary at the true solution Ok, we obtain

_2 (M -
R
C) •w(•)
, V (& = k0 'IR ds
p4O• J-
&-R I'
d9 dps / ds = 0.
MC C -- 0 O Ok ds d sd
(14)
, We now take another varied state in the form(4

"= , k (15)

where now the applied moment is determined by t,, and the roles of

cp1 and have been reversed. Then, requiring the potential energy

to be stationary at the true solution pt, we obtain

R) aC2f __ kds k d s =0.

MG a 0 j 0oP~,
(16)

Thus, if k 2 , we
i obtain from equations (14) and (16)
oaR
a0 Pk(t,
ds = 0

(17)
oaR d•pk dqo4
0s

These orthogonality relations between the eigenvectors are analogous.

4.i
to the orthogonality ielations for the buckling of a straight beam under

an axial force (see the Appendix). By integrating equations (17) by

parts and using equation (3), one can obtain the natural homogeneous

boundary conditions on the Pk These are

d 0o aR

Pk ] 0 =
and (18)

[d'pk aR
The orthogonality conditions are sufficient to determine the

eigenvalues k uniquely when it is remembered that the rigid body

rotation solution must be an allowed solution to the problem. In order

1.
to emphasize that k = is a solution of this buckling problem, we

will rewrite equations (5) in terms of odd and even parts, with this

rigid body rotation solution explicitly included as follows:

•00 = A ( -• + B

1i R -
= AS_( 2 + B 1 cos R - • (19)

Asin k s- 2R)+ Bcos k s -

Pk k - Bk

When the orthogonality conditions (17) are applied to equations (19),

one obtains, after some manipulation,

9'
 IR
1 +
(A Ak+ B B k)
si kR
IR - - k sin k +LR
(A A k -)BB Bk)i =
0, •i

k- -2- (A01B~k k +0, R)

dins'i k L &R sin

(A Ak + B Bk) 1 + (AA k +1
1k i ~ R1 k kR
cvR- (20)
as• sin (k +t) - -
sinA Bk-) 0,
(A4Ak + BtBk) k-) - (AAk- B6Bk) k+t -0

aR sin (k+ +')t.

sin (k -t) + 2Ak=
k- A
(A Ak + B6Bk)

The solutions (19) that satisfy (20) and have real coefficients are of

the form

= Ak sink (s _• ) cos k(s_ R)] (21)

where the eigenvalues are given by (IZ

2niT+ 1
Rn 0,
o , d2, •3, ...

for all of the

Whichever sign is used in equation (21), it must be used
At the
eigenfunctions in order for them to be mutually orthogonal.
Ak and
isolated points where a is a multiple of ir, the coefficients

Bk can be assigned in an arbitrary manner. However, since we are

isolated points
concerned here with the behavior of a real beam, these

are of no particular importance.

Figure 2 shows a plot of the three lowest eigenvalues from

10
 -9-

-8 M

kR

-4- W40

-2R
0-
(M=orL0) 7 _ E2,-

2 .,,4-CLASSICAt, n=I El] 2Zn=

/ Eq22, n= I

4 /
5 /
k21
7 ! 8 ''(.

Figure 2. The three lowest eigenvalues of equation (3) vs.

* beam angle a.
equation (22) vs. beam angle c. For comparison purposes one of the

classical curves from equation (6) or (7) is also included. The ordinate

of this graph is shown positive d o wnwanrds in order to emphasize that

values of kR less than 1 correspond to a positive (closing) moment

and values of kR greater than I to a negative (opening) moment. * It

can be seen that the classical curve causes the moment to change sign

and is asymptotic to a positive value, while the new curves are asymp-

totic to kR = 1 which ce:responds to zero buckling moment. The small

rectangles on this figure refer to experimental results which will be

discussed below.

These uniquely determined values of k are next used to deter-

mine the value of the buckling moment M from equation (4). However,

since equation (4) is quadratic in M, extranO,,s solutions may be en-

countered. To guard against these extraneous solutions, we note that the

upper curve in Figure 2 should only give rise to positive moments and

the lower curve to negative moments. In addition, the magnitude of the

buckling moment should monotonically decrease with increasing a. This

arises from the fact that a curved beam can buckle into the true buckled

shape of a shorter curved beam, with the remaining portion having only

a rigid body displacement. This "unbent" portion will not change either 4-

the potential energy or the boundary conditions; thus the longer beam

1
This behavior would have been reversed had k beezi chosen
as the fundamental rigid body rotation. R

12
 could buckle at the same buckling moment as the shorter beam.

Applying these considerations to the buckling moments obtained

from Figure Z and equation (4), one obtains a typical buckling moment

curve as shown in Figure 3. The discontinuity in the positive buckling

moment curve arises from the fact that it is composed af portions of

two different curves. The lower curve passes thru zero at = ir and

increases until it reaches a maximum at a Z=r. The upper curve would

begin increasing for a greater than 2Zr; thus the decreasing moment

requirement results in a discontinuity at a = Zu, where the buckling

C El
moment decreases from M i to M - . However, this discon-

tinuity is only of theoretical interest, since any real curved beam must

necessarily have a less than Zrr by some small amount. In passing,

it should be noted that these buckling moment curves are the minimum

possible for lateral buckling of a curved beam. The addition of any

lateral restraints on the ends of the beam can only serve to increase the

buckling moments.

It is perhaps helpful to plot the eigenfunctions given by equation

(Zl) in order to show the types of conditions that result at the ends of

the beam.. Figure 4 is such a plot for the case of buckling under a

negative moment. Various values of a are represented between the

limiting cases of a -' 0 and a - 2Tr. Figures 5 and 6 present the

same information for buckling under a positive moment and for a rigid

body rotation about a diameter. In these plots it is helpful to remember

13

.4
 It

'160 I
140 a

i R
100

~60-

""0 20.

*
0

'-.40]
D J;1'00.w

"-60 / E

-120 .

-140"

-160 I

-180

-14014

AI Figure 3. The theoretical buckling moment vs. beanm angle

ce for a curved beam of 5"1 0. D., 2P"I. D., and
*.081 " thick, with E =5.5 x 106S psi and V =0.3.

14
 k-2
0 a -2v?

/a •k = 73

k= 3
0 a =

//•l?.a k= 5

•_. 2aO k C•O0

0- a --,0

Figure 4. Lowest eigenfunction. ok of equation (3) for buckling

under a negative moment (i. e. , n = +1), for various
beam angles at.

15
 I--

kk

k I

k -3

4'4

a=-
o0_ _'

Fiur 5ienunti
Lwet p f qaton(3 fr uk ling
undr apoitie
omei (~e, n -1, fr ar17u2
bemage
I1
 a k= I
a 27r

k-!1

51a 3W/

a.l. kk= II=•

0 al k I--

k= I

171
0 1 a --o o.

Figure 6. Eigenfunction p, of equation (3) for a rigid body

rotation about a diameter (i. e., n= 0), for various
beam angles a~.

17
that k represents the number of full sine waves that would occur if

the beam angle were 27r. These curves were obtained by using the plus

sign in equation (21); use of the minus sign would essentially interchange

the ends. Since (p can be related to v by equations (1) and (2), these

curves also represent the deflection of the beam centerline to within an

arbitrary rigid body motion. Note that, as would be expected, the eigen-

functions approach the same shape for positive or negative buckling as

the beam angle approaches zero. Also, the lowest classical shape, for

either positive or negative bending moments, would simply be half of a

sine wave regardless of the value of (.

b. Straight beam

One can obtain the buckling moment for a straight beam of length

L by taking the limit of the curved beam equation (22) as R -. c and

oaR - L. This gives k = L which is twice the accepted classical

value '4). Because of this disagreement, it is necessary to present

the analysis although it is similar to that for the curved beam.

The equations for the straight ,beam can be readily obtained

1
from equations (1) and (2) by omitting those terms which contain •

In this manner one obtains

ds+ k2 = 0 (23)

where

k
= -- (24)
x

18
The solutions to (23) can be written in the form

00 A (s 0- ~2 + B0

= A sink(s L) + BL cos k(s L

9k k k( 2jB ~k (S2

where L is the length of the beam. The potential energy V is then

derived in the same manner as before and results in

V = ~
m2
xO
L 2
pds + f
j-
L
LD) ds .(26)

It can be readily seen that the Euler-Lagrange equation of this functional

is the equilibriu-n equation (23). Application of the principle of station-

ary potential energy again leads to the orthogonality conditions

L
0
ds
o0,'k o
(27)
L (dIlk d9'-) ds = 0

We note that, for the straight beam, the trivial solution o is the

solution which must be an allowed solution to the problem. Therefore,

an application of the orthogonality conditions to the solutions (25)

results, after some manipulation, in

kL.
AA sin - 0 (28a)
o k 2

1 kL L kL kL
2AoA
1 sin - cos -- + 2BB sin 0 (28b)

19
 sin (k - sin (k + t,)
+ k k At A k-
k k+t
0 , (28c)
(,k+B k) k- tk

(sin (k - t) sin (k + t)
~A~A +BBk 2-i +-A~ B.Bk-j+, 2 0.(28d)

The only solution to the system (28) which -has zero potential energy

for the trivial solution pop is

Po =BB
-0 0
(29)
= A sin k L(--.+ B cos k(s
pk k k2(c- k 2)

where-
Zn T
k L n = 0, J1, 1=2, (30)

Note that since the A and B are arbitrary, equations (29) and (30)
k k
contain the limit of equations (21) and (22) (as R -c and cR - L)
as a special case.

4. EXPERIMENTAL RESULTS

In order to verify the trend of the results given by equations

(22) and (30), a short experimental program was undertaken. Since

there is a relatively wide disagreement between these results and the

classical values, the intent of the experiment was not to provide a

rigorous proof of the new results but -rather to be of sufficient accuracy

to show the disagreement with the classical results.

20
 .1

Before the buckling moment of a curved beam has been reached,

the stress distribution is simply that of the well-known plane stress

problem of the pure bending of a curved beam [5]. Therefore it was

decided to make the test specimens of a photoelastic material in order

to have visual proof that the loading was giving the correct theoretical

stress distribution before buckling has occurred. The specimens were

therefore cut from photoelastic material PS-ZB, and were circular

ring sectors of 5" O.D., 21" I.D., and .081" thick. The specimens

were then ep6xy bonded to metal strips which were hinged at the center

of the circular ring sector. Figure 7 shows an unstressed full ring test

specimen in the transmission polariscope with dark field illumination.

In this case the metal strip was actually a strap hinge; in other tests

it was found necessary to use metal strips bolted to a hinge, so that the

center of rotation could. be adjusted until the axisymmetric plane stress

solution was obtained upon opening or closing the hinge. Figure 8 shows

the same specimen after pulling upward on the top of the strap hinge to

give a measured opening moment of -16.0 in-lb. It can be seen that the

correct axisymmetric stress distribution has been obtained except, of

course, in the immediate vicinity of the bonded ends. The first dark

fringe is just at the inside radius of the specimen. For this geometry

and material, each dark fringe corresponds to a principle stress differ-

ence of approximately 970 psi.

Figure 9 shows the same specimen at incipient buckling, and

21
 kk

Figure 7. Stress free full ring test specimen in the transmission

polariscope with dark field illumination.

.3l

Figure 8. Full ring specimen with an opening moment of -16.0

in-lb. The stress optical coefficient is 970 psi/fringe.

22
 f-

Figure 9. Specimen at incipient buckling with an opening moment

of approximately -20.6 in-lb.

Figure 10. Post-buckled specimen with an opening moment of -20.8

in-lb. The dark radial bands are inflection points.

23
!r

Figure 10 shows the post-buckled specimen with an opening moment of

-20.8 in-lb. The four radial dark bands in Figure 10 were experiment-

ally seen to correspond to the inflection points where 9 = 0. This post-

buckled shape is, therefore, in fairly good agreement with the topmost

theoretical curve shown in Figure 4. Note thatthe classical post-buckled

shape of equation (6) would have no inflection points and the classical ¶

shape of equation (7) would have either one or two inflection points, de-

pending on the boundary conditions. The strap hinge arrangement was

intended to be loose enough laterally, so that the ends would behave like

the "free" edges required by equation (3). However, it is possible that 4

St. Venant's principle applies to this problem, since it was observed

that a small deflection of the ends did not seem tucontribute to the lateral

buckling, in contrast to the problem of buckling under an axial force.

Figures 11 and 12 show the pre- and post-buckled stress dis-

tributions for a three-quarter ring specimen which buckled under an

opening moment of - Z7.7 in-lb. The three inflection points in Figure 12

can be seen to be in reasonable agreement with the second theoretical

curve of Figure 4. Figures 13 and 14 show the same information for a

half ring specimen which buckled at -41.8 in-lb. Some difficulty was

encountered in maintaining the plane stress solution for this specimen,

and any slight departure from the plane stress solution caused a con-

siderable reduction in the buckling moment. In addition, the lateral

restraints had to be maintained so loosely that the buckling process

24

F:÷
 Figure 11. Three-quarter ring specimen with an opening
moment of -24.6 in-lb.

Figure 12. Post-buckled specimen with an opening moment of -28.8

in-lb. The specimen buckled at -27.7 in-lb.

Z5
 kL

Figure 13. Half ring specimen with an opening moment

of -21.9 in-lb..

Figure 14. Post-buckled specimen which buckled at an

opening moment of -41.8 in-lb. '
2 t

26
itself caused a considerable change in the strcss distribution. Con-

sequently, the buckled shape in Figure 14 is only in rough agreement

with the third theoretical curve of Figure 4.

Data was also obtained for buckling under a closing moment.

However, because of a shortage of photoelastic material, the larger

specimens were simply cut down to make the shorter specimens; and,

due to an unfortunate oversight, no data was obtained on the approxi-

mately full ring specimen. The three-quarter ring specimen, however,

is shown in Figures 15 and 16 for pre- and post-buckling under a closing

moment. The absence of an inflection point is in agreement with the

second theoretical curve of Figure 5. Figure 17 shows the half ring

specimen with a closing moment of approximately 49 in-lb. A slight

increase in this moment appeared to cause the specimen to simultan-

eously buckle and fracture. Because of this tensile strength limitation,

no data on shorter specimens was obtained.

The experimentally obtained buckling moments were used to

calculate values of k for these five specimens. These values are

indicated by the small squares on Figure 2. In addition, the buckling

moments themselves are shown on Figure 3. The amount of agreement

between the data points and the theoretical curves should be considered

fortuituous, since there were uncertainties in both the experimental

procedure and the material properties of the specimens.

Figures 18 and 19 show the pre- and post-buckled stress

27
N

1.

F igure 15. Three-qujarter ring specimen with a closing

moment of +Z6.0 in-lb.

Figure 16. Post-buckled specimen with a closing

mnoment of +Z8.0 in-lb.

2.8
Figure 17. Half ring specimen immediately before failure with
a closing moment of +49.0 in-lb.

29
 ii

Figure 18. Straight beam specimen

with a calculated
applied moment of 20.0
in-lb.

I
J

Figure 19. Post-.buckled specimen.

30
distributions for an initially straight beam of the same material. The

beam dimensions were 7½" long by 0.8" high and .081" thick. The

specimen was epoxy bonded to long transverse metal strips; the right

hand strip was clamped to the testing apparatus and the left hand strip

remained free. The bending moment was then applied by hand with care

being taken to obtain the plane stress solution before buckling had occured.

The apelied bending moment was, therefore, calculated from the observed

stress pattern, using the dimensions of the beam and the stress optical

coefficient of the material.

The specimen remained straight above the classical buckling

moment and had a:- tendency to return to the straight form when a very

small buckle was manually introduced. The specimen finally buckled

when the calculated applied moment reached a value of 20 in-lb, which

is almost exactly twice the lowest classical value for these dimensions;

this is in agreement with equation (30). Note the similarity of the

buckled shape of Figure 19 to the bottommost theoretical curve of either

Figure 4 or Figure •. This is possibly due to the fact that +he straight

beam is actually significantly curved before buckling finally occurs.

5.. SUMMARY

The basiz differential equation for the lateral buckling of a beam

subjected to pure bending is derived under the assumption that there are

no lateral forces acting on the ends of the beam. Therefore, the

classical approach of assigning relatively arbitrary boundary conditions

31
 will, in general, violate this assumption, and there will be shearing

forces on the boundaries which were not included in the differential

equation. In contrast, the method presented in this paper insures that

the principle of stationary potential energy (i. e., virtual work) is satis-

fied and the resulting orthogonality conditions lead to a unique set of

boundary conditions. The experimental data agrees with these new

results rather than the classical results - in magnitude of the buckling

moment, as well as in the post-buckled shape of the beam.

I'

ri
I"

32
 Io

APPEND]X

1. Orthogonality Relationships for a Straight Beam Subjected to an

Axial Force.

For an initially straight beamý subjected to an axial force P

whose direction of action remains fixed in space, the equilibrium

equation for the deflection v of the beam is

d4 v
EI--•-+ dsv - 0
Pds2 .(1)

Since the force P is the only agent doing work on the beami, the

potential energy expression for the beam, in the case of an inextensible

centerline, is given by

V= 0 ' )\ds
ds - j-I (cds) ds (2)

We now consider that we have a complete set of solutions vk to

equation (1) which satisfy the same natural boundary conditions

giving rise to (1) and (2); i. e., either the moment or the slope

is zero on the boundaries, and, either the shear or the displacement

is zero on the boundaries. We now apply the principle of stationary

potential energy at the eigenstates vk and v . Taking a varied state

about the true state vk in the form v'"= vk + CvV, we obtain from

equation (2),

"" Ld?¢ d2 dvd dvs

(v*) 0 ElIJ( v"~~) ds f~('. ( 3)
E0 T 0 d o-s
0s d

33

r
 Taking a varied state about the true state v, in the form v-ý = v v,
41 k
we obtain

' •' IEI •(ek dt, P•

pt dk dv,

____
r,= 0 10 (-jS-2K ) (-d ;F -ds JLdO
(--as- (j - ) ds. (4)

Therefore, if P. - , we obtain from equations (3) and (4) the

following orthogonality relations for the eigenfunctions:

IL d 2 vk d vt

Sdvk \dv
jy- d
)y~ -ds= 0

l'o

0, "344

I'
I"

r,1

i3
 REFERENCES

1. Timoshenko, S., and J. M. Gere, Theory of Elastic Stability,

McGraw-Hill, Znd ed., 1961, p. 315.

2. Bleich, H.H., Buckling Strength of Metal Structures, McGraw-

Hill, 1952, Ch. IV.

3. Love, A. E. H., A Treatise on the Mathematical Theory of

Elasticity, Dover Publications, 4th ed., 1944, Ch. XIX.

4. Timoshenko, S., and J. M. Gere, Theory of Elastic Staility,

McGraw-Hill, Znd ed., 1961, p. Z57.

5. Timoshenko, S., and J. N. Goodier: Theory of Elasticity,

McGraw-Hill, Znd ed., 1951, p. 61.

35
 Unclassified
Secunt y Cls!sitiratinn

IDOCUMENT CONTROL DATA - R & D

"Security clasilfication of title. bod, of bsftr Ii, andealn2
• nn.,flsI n t • i• ntered •lien
h tL. .,t-Ial reputt 8. cl.lasliled,

ORIGIJA• TONG AC TI'SITY lCorpoatie Author)

SUnc lass ified
CLASSIFICA T$ON
-,a. R'rORT SECUR.TV
•
UNIVERSITY OF CALIFORN;_A, SAN DIE6O Unclassifie
Zb. GROUiP
La Jolla, California 9Z037
3 REPORT TITLE

THE LATERAL BUCKLING OF A STRAIGHT OR CURVED BEAM SUBJECTED

TO PURE BENDING
4 DESCRIPTIVE NOTES (7T'pe -t eport and inclusivr dates)

Research Report
5 AU TNOR451 (Frst name. middle initial, last name)

F. E. Vanslager

£ REPORT DATE 17a. TOTAL NO. OF PAGES 7b. NO. OF REFS

July 1968 40_ 5

Sa. CONTRACT OR GRANT NO. 90. OPIGINATOR'S REPORT NUMBERIS)

N00014-67-A-01 09-0003
b. PROJECT NO. No. 2
NR 064-496
C'. 9b. OTHER REPORT NO(S) (Any other numbeCs that may be assigned
this report)

d.
10 DISTRIBUTION STATEMENT

Reproduction in whole or in part is permitted for any purpose of the

United States Government. Distribition of this document is unlimited.
II SUPPLEMEN.TARY NOTES I2. SPONSORING MILITARY ACTIVITY

Office of Naval Research

Washington, D. C.
13 ABSTRACT

This paper presents, a new analysis of the classical problem of the

lateral buckling under pure bending of a curved beam with circular axis.

Although a solution to this problem, and to the corresponding straight beam

problem is well-accepted in the literature, the present analysis gives buck-

ling moments which are significantly different than the usually accepted results.
The difference arises from an application of new boundary conditions that are

shown to be the only boundary conditions consistent with the equilibrium of the
buckled beam under a pure bending moment. This new analysis provides

generally higher bending moments. A short experimental program was per-

formed, and the results tend to confirm the new analysis.

DD FOcRa.14 3 Unclassified
 Unclassified
Security Classification
14. LINK A LINK 8 LINK C
.EY WORDS
ROLE *.T ROL= WT ROLE WT

V Buckling,

Lateral buckling of beams,

Buckling, experimental,

Buckling under pure bending,

Curved beam,
Elastic stability

UI .

I ((tfl,{|l, f{.ll
Unclass ified