Sie sind auf Seite 1von 10

Journal of Applied Mathematicsand Physics(ZAMP)

Vol. 23, 1972

Birkh~iuserVerlag Basel

On One-Dimensional Finite-Strain Beam Theory"


the Plane Problem 1)
By Eric Reissner, Dept. of Applied Mechanics and Engineering Sciences, University
of California, San Diego, La Jolla, California, USA

Introduction

The following is concerned with a consistent one-dimensional treatment of the


class of beam problems dealing with the plane deformation of originally plane beams.
Our principal result is a system of non-linear strain displacement relations which is
consistent with exact one-dimensional equilibrium equations for forces and moments
via what is considered to be an appropriate version of the principle of virtual work.
Having a consistent system of equilibrium and strain displacement equations
it is further necessary to stipulate, or rather to establish by means of an appropriate
set of physical experiments, an associated system of constitutive equations. We
discuss the nature of this aspect of the problem, including a solution of its linearized
version, but without arriving at the solution of the general problem.
The principal novelty of the present results is thought to be a rational incorporation of transverse shear deformation into one-dimensional finite-strain beam
theory. A case may he made that the theory, with this effect incorporated, is of a
more harmonious form than the corresponding classical theory, where account is
taken of finite bending and stretching, while at the same time it is postulatedfollowing Euler and Bernoulli-that the transverse shearing strain is absent, with the
corresponding force being a reactive force.
As an application of the general work a solution is given of the problem of
circular ring buckling, including consideration of the effects of axial normal strain
and of transverse shearing strain on the value of the classical Bresse-Maurice L6vy
buckling load.

Kinematics of Beam Element

We consider an element ds of a one-dimensional beam with equations x = x (s)


and y=y(s) before deformation. We designate the tangent angle to the beam curve
by ~P0 and write cos ~oo = x' (s) and sin Cpo= y' (s), where primes indicate differentiation
1) A report supported by the Officeof Naval Research and the National ScienceFoundation,Washington, D. C.

796

Eric Reissner

ZAMP

with respect to s. We note that Cpo is also the angle between the normal to the beam
curve and the y-axis.
Due to deformation the points x=x(s) and y = y ( s ) of the undeformed beam
curve are changed to x(s)+u(s) and y(s)+v(s). We now assume that transverse
elements which were originally normal to the beam curve do not necessarily remain
so but end up enclosing an angle 89~ - Z with this curve. At the same time we designate
the angle enclosed by such an element and the y-axis by q~. We then have a geometrical situation as shown in Figure 1. We note in particular, in addition to the

.,,-

x+u+(x+u)'ds

-.-

x~l+u(s)

x(s) - -

y(s)+v(s)

h "~-%

y(s)

I
f

Figure 1

angle X, the relative change of length e of the beam curve element ds, and the change
o f the angle ~00 into an angle q0, and we read from the deformed beam element, as
relations between Z, cO, e, u and v,

_x'+u'
_
= cos(~0 + Z),
l+e

~y'+v'
=
l+e

sin(q~ + Z).

(la, b)

Dynamics of Beam E|ement


We now consider the deformed beam element, with normal and shear forces N
and Q and with a bending moment M, in accordance with Figure 2. Together with
this we assume force load intensities Px and py and a moment load intensity m, per
unit of undeformed beam curve length, also in accordance with Figure 2.
We then read from Figure 2 as component equations of force equilibrium in the
directions of x and y,
(N cos q~-Q sin q0)'+ p~= 0,

(2a)

(N sin ~o+Q cos ~0)'+ p y = 0.

(2b)

Vol. 23, 1972

On One-Dimensional Finite-Strain Beam Theory: the Plane Problem

797

O'~Q'dst~k M+M'ds

,~2__*%~

mdsl

/'%

__

Figure 2

At the same time we obtain as equation of moment equilibrium


M'+(1

+e)(Q cos z - N sin x ) + m = 0 .

(3)

We note, for future use, the possibility of deducing from (2a, b) the relations

N'-qCQ+n=O,

Q'+q/N+q=O,

(2*a, b)

where n = Px cos cp+ py sin (p and q = py cos (p - Px sin cp are components of load
intensity in the directions of N and Q, respectively.

Constitutive Equations
We postulate that the material of the beam is elastic and that we have the
existence of axial and transverse force strains s and 7 and of a bending strain ~, in
such a way that constitutive equations for beam elements may be written in the form

N=f~(~, 7, K),

Q= fQ(~, ~, ~c),

M = f~(~, ~,, ~).

(4)

We are ignorant, at this point, not only in regard to the form of the functions f in (4),
but also in regard to definitions for the components of strain e, 7 and tc which enter
into the constitutive equations (4)2).
2) However,we expect that e~ e, 7~ Z and ~c~ q/-~o~, for sufficientlysmall strain.

798

Eric Reissner

ZAMP

Defining Equations for Strain


In order to obtain equations for strain we consider a virtual work equation
of the form
S2

(N 6e+Q 67+M 6~c)ds


Xl
82

= J (Px 6u+p, 6v+m&o) ds

(5)

Sl

+ [(N cos (p - Q sin ~o) 6u + (N sin p + Q cos ~o) 3v + M 6q~]~,


and we stipulate, as Principle of Virtual Work, that equation (5) be equivalent
to the dynamic equations (2) and (3) in the interior of the interval (s1, s2), given
that be, 37 and 6K are appropriate expressions for virtual strains.
Since we know the form of the dynamic equations but do not at this point
know expressions for virtual strains we use equation (5), in conjunction with (2)
and (3), to deduce expressions for virtual strains.
Introduction of (2) and (3) into equation (5) gives a relation of the form
$2

(N 6~+O 67+M 6~c)ds


Sl
82

= - j {(N cos ( p - Q sin go)' 6u+(N sin rp+Q cos rp)' gv


~,
+ [ M ' + (1 + e)(Q cos x - N sin )~)] 6q~} ds

(6)

+ [(N cos q0-Q sin q0) 6 u + . . - + M bq)]~,


and in this we may now consider N, Q and M as arbitrary differentiable functions ofs.
In order to utilize (6) we integrate by parts, thereby eliminating all derivatives
of N, (2 and M as well as the boundary terms on the right. In this way we obtain
S2

j (N 6~+Q b T + M 6~c)ds
SI
$2

= ~ [(N cos q0- Q sin ~0) bu'+ (N sin ~0+ Q cos q~) 6v'

(7)

S1

+ M 6q0' - ( 1 + e) (Q cos z - N sin )~) 6qo-1ds.


The arbitrariness of N, Q and M means that (7) implies the virtual strain displacement relations

6~ = (6u') cos q~+ (6v') sin qo+ (~5qo) (1 + e) sin X,

(8a)

@ = (6v') cos qo- (6u') sin ~o- (6qo) (1 +e) cos X,

(8b)
(9)

It remains to take the step from virtual strain displacement relations to actual
strain disNacement relations.

Vol. 23, 1972

On One-DimensionalFinite-StrainBeam Theory: the Plane Problem

799

One of these actual strain displacement relations follows directly from equation (9) in the form

~=~0'-~0o.

(10)

A correspondingly simple derivation of expressions for e and y is clearly not


possible through direct use of (8 a, b). Remarkably, we may obtain e and y by using
(8 a, b) in conjunction with the geometrical relations (1). To do this we observe that
equations (1) imply the following relations between virtual quantities
6 u' = (6 e) cos (q) + Z) - 6 (q9+ Z) (1 + e) sin (q~+ Z),

(11 a)

6v' =(6e) sin (q~+Z)+6(qo +Z) (1 +e) cos (q0 +Z).

(llb)

We now use (11a, b) in order to eliminate 6u' and 6v' in (8a, b). In this way
we obtain
6 e = (6 e) cos Z - (6)0(1 + e) sin Z,

(12 a)

67 =(6e ) sin X+ (6)0 (1 +e) cos Z-

(12b)

The form of (12a, b) is such that we can now go from virtual strains to actual strains.
The results are
e=(1 +e)cos Z - 1,

7=(1 +e) sin Z.

(13a, b)

Having (13a, b) we can further express e and y in terms of u, v and cp. Introduction of (13 a, b) into (1 a, b) gives first
x' +u' =(1 +a) cos (P-7 sin q),

(14a)

y ' + v ' = ( 1 + ~) sin (p+7 cos ~o,

(14b)

and then, by inversion


= (x' + u') cos q) + (y' + v') sin ~o- 1,
7 = (Y' + v') cos q) - (x' + u') sin ~0.

(15a)
(lgb)

We finally note the possibility of rewriting the moment equilibrium equation (3)
somewhat more simply with the help of the strain components e and y as in (13),
in the form
M'+(I+e) Q-vN+m=0.

(3*)

Observations on the Problem of Experimentally Derived Constitutive Equations


In order to see the nature of the problem of experimentally establishing the
nature of the functions f in equations (4) we consider the problem of an originally
straight beam, with x = s , y = 0 and ~00=0 , fixed at the end x = 0 and subject to
given displacements u(a)= v a, v(a)= va and cp(a)= ~0a at the other end. We assume

800

Eric Reissner

ZAMP

absent distributed loads and have then from equations (2a, b)


Ncos~o-Qsinq~=X~,

Nsin~o+Qcosq)=Ya,

(16a, b)

where X. and Y. are two constants of integration the mechanical significance of


which is evident.
To proceed further we consider the moment equation (3*) as a differential
equation for (p, by writing
Q= Yacosqo-X~sin(o,

N = Yasinqo+XaCOS(O,

(17a, b)

and by considering the constitutive equations involving N and (2 partially inverted


in the form
e = f~(N, Q, ~o'),

7 = f~(N, Q, (p')

(18a, b)

so that M = f M ( e , 7, to)= f~(n, Q, q/)= g(~o, qo').


The resultant second-order equation for q~must be solved subject to the boundary
conditions (p(0) = 0 and cp(a) = qo,, with which (p = qo(x; X~, Ya, q0o).
Having (p we find u and v from (14a, b). The boundary conditions for u and v
are satisfied upon setting
a

[(1 + e) cos q~- ? sin ~o] d x = u a Aft a,

(19 a)

[(1 + e) sin (0 + 7 cos qo3 dx = v a .

(19 b)

0
a

We now measure Xo, Y, and M~ as functions of uo, v a, % , and of a, giving a


set of three relations X ~ = f x o ( u a, v,, (pa; a), etc. The remaining task then is to deduce
from the form of these three experimentally determined functions fxo, fYo and f u ~
the form of the desired three functions fN, fe, fM in equations (4).
The linear case. We consider a range of stresses and strains within which
{N Q M} = [C] {E7 ~c}

(20)

with a view towards determining the elements CN~,..., CM~ of the three by three
matrix [C].
From equations (17) follow the linearized relations
Q=Q_.,,=Y~,

N=N.=X.,

(21a, b)

and the moment equation (3*), again with boundary conditions q)(0)=0 and
qo(a)= q)a, is reduced to
M' + Qa 0.

(22)

-~"

Equations (19) for the translational edge displacements become


a

j' e d x = u,,

~ (q) + 7) d x = va.

(23 a, b)

Vol. 23, 1972

On One-DimensionalFinite-StrainBeamTheory:the Plane Problem

801

In order to solve the problem as stated in (20) to (23) we partially invert (20) in
the form
{g 7 M} = I-C*] {Na Qa (P'},

(24)

and write (22) in the form C}~ q/'+ Qa=0, with solution
X

O,X--X 2

~P= Pa ~ -+Qs

2C~t~

(25)

We then have further, from (24),


Ms = Gnu
, Ns+(CuQ-ga)
,
1 Qs+ C*~ a -1 (Pa

(26)

and, upon making use of (23 a, b),


9 a N s + C~q
* a Qa+ C ~* qG,
G -_ C~N

(27 a)

a3

a2

(27b)

We now stipulate knowledge of a matrix [B], as a result of experiment, such that


{Ua ?)a (tOa}:" [B] { N s Qa Ms}.

(28)

Having (26) to (28) we may then successively determine the elements of the matrix
[C*] in terms of the elements of [B]. To see this we write
a

~Ps-

C~vIN
a-- 2 C*O_
a
C* K Ns q 2 C~K Qa + ~

Ms,

(26')

and have then from the relation ~0a = B~u Ns + B ~ Qs + B~M Ms that
a -B~,
C~

a-C* e
2 C*~ -BYQ'

a C* u _
C*

B~M,

(29)

,
from which CM~, C~Q and C}N follow in succession in terms of elements of [B].
We next introduce (26') into (27a, b) and compare the resultant relations with
corresponding relations in (28). In this way we obtain the remaining six elements
C~N, etc. of the matrix [C*] in terms of the elements of [B].
Finally, having [C*] we find the elements of [C] by returning from (24) to (20).
Buckling of Circular Rings

As an application of the foregoing we consider the classical problem of in-plane


buckling of a circular ring of radius R, subject to a uniform normal pressure p.
We wish to obtain a buckling-load formula which incorporates the effects of (1)
the symmetrical deformation of the ring prior to the onset of buckling, (2) axial
strain associated with the buckling mode, (3) transverse shearing strain associated

802

Eric Reissner

ZAMP

with the buckling mode. We will be concerned, in particular, with the question of
appropriate constitutive equations.
Inspection of Figure 2 indicates that for uniform normal pressure p, per unit
of deformed beam curve, we have as expressions for the load intensity components q
and n in the force equilibrium equations (2* a, b)
q=p(l+e)cosz=p(l+e),

n=-p(l+e)sinz=-p?,

(30a, b)

together with an absent moment load intensity m in equation (3*).


We further have, with K as in equation (10) and with R d~0o = ds, that q0' = R - 1+ ~c.
Therewith the equilibrium equations (2*a, b) and (3*) may be written in the form
N ' - ( R -1 +~c) Q - 7 p = 0 ,

(31 a)

Q'+(R -1 + t o ) N + ( 1 + ~) p---0,

(31b)

M'+(I+~) Q-yN=0.

(31c)

In complementing (31) by constitutive equations we have no difficulty in deciding


that suitable relations involving tc and e are of the form
M=Dtc,

e=CN.

(32a, b)

In stipulating a relation involving ~ we find it necessary to concern ourselves


with the question whether ? would be determined by the force Q tangential to the
deformed cross section or by a force Q, normal to the deformed centerline. Evidently,
we have Q, given in terms of Q and N by the relation Q, = Q cos z - N sin ;( or,
approximately, by Q, = Q - N ?. If we stipulate that y = B Q , we arrive at a relation
for ,/in terms of Q and N, of the form ?=BQ/(1 +BN)3). If we use Q instead of Q,
at the outset we have instead that 7 = BQ. We may subsume both relations to one
of the form

(32c)

Be

Y= 1 + 2 B N '
and consider in the end the two limiting cases 2 = 0 and 2 = 1.
Having equations (31) and (32) we now consider the stability of the state
N=-P,

~=-~p,

Q=M=?=~c=0

(33)

for which, evidently, in view of (31 b) and (32b)


P = (1 - ~e) R p,

ep = CP.

(34)

e = - ~p + q ,

(35)

We now write
N = - P + N1,

and linearize (31) and (32) in terms of Q, M, 7, ~:, N1 and el.


3) This, t~gether with (3 2 b ), is e~ective~y equiva~ent t~ c~nstitutive equati~ns ~f the f~rm Q=(~'/B) + (~?~/C)
and N=(8/C)+(y2/2 C)-

Vol. 23, 1972

On One-Dimensional Finite-Strain Beam Theory: the Plane Problem

803

Equations (31) become

N;-R-1Q-p?=O,

(36a)

Q' + R -1N 1 - P tc+ p ~1= 0 ,

(36b)

M'+(1-ee) Q+P~=0.

(36c)

Equation (32 a) remains as is and equations (32 b) and (32 c) become 4)

51= CN1,

7 = ( 1 - 2 BP) -1 BQ.

(37a, b)

We now use (32a), (34) and (37a, b) to write (36a, b, c) as a system of equations
for N1, Q and ~c, as follows
P
B
NI'- (1-~ 1 - CP 1--2BP) Q~-=O'
CP
1-cP)

Q'+(lq

Dtr

(38a)

~N1
--Pfc=0,

(38b)

( 1-CP-~ 1 - BP
2BP )Q=0.

It is evident that
and (38c) as one
Appropriate
where n =2, 3. . . . .

(38c)

(38 b), differentiated once, may be written with the help of (38 a)
second-order differential equation for Q.
solutions, for a complete ring, will be of the form Q = cos n s/R
From this follows as the equation for possible values of P,

rt 2
CP
R2 ~ 14 1 - C P

)( 1 -+ 1

,( I-CP-~ l _..)
+~
2Bp

1-CP

..),

1-2BP

R 2

(39)

=0.

Equation (39) may be written as a cubic equation for pR2/D, involving axial-strain
and transverse shear-strain parameters k~= CD/R 2 and k~,=BD/R 2. We will here
limit ourselves to a discussion of the case k~=0, with ky=-k, for which the cubic
equation reduces to a quadratic of the form

k(pR2/D)
1--2
k(PR2/D) ~ -PR2
D - - ( 1 4 l_2k(pR2/D)
k(pR2/D) ) = n 2 - 1 .

(40)

The smallest positive value of P follows from this for n = 2. We consider in particular
the cases 2 = 1 and 2 = 0.
4) We note the possibility that C and B, as well as D in equation (32a), may be considered to depend
on ee.

804

Eric Reissner

ZAMP

When 2= 1 we have from (40), in agreement with a recent result by Smith


and Simetses s).

PR 2
3D

l + 4 k -1-4k+O(k2)"

(40a)

When 2 = 0 the solution is


3D -

6k

14 ( k + l ) z

1 =~+1

= 1 - 4 k + O(k2).

(k+l)2

~-..-

(40 b)

For small k, say for k up to 2~, the values of PR2/3D given by the two different
formulas differ by very little. For larger k equation (40a) is the conservative one,
giving a larger shear correction effect than equation (40b). To illustrate, when
k = 88then (40 a) predicts that transverse deformability reduces the classical buckling
load by 50 %, while (40b) predicts a reduction by 42 ~.
s) j. Eng. Mech. Div., ASCE 95, EM3, 559-569 (1969).

Summary
The paper formulates a one-dimensional large-strain beam theory for plane deformations of plane
beams, with rigorous consistency of dynamics and kinematics via application of the principle of virtuai
work. This formulation is complemented by considerations on how to obtain constitutive equations, and
applied to the problem of buckling of circular rings, including the effects of axial normal strain and transverse shearing strain.

Zusammenfassung
Das Ziel dieser Arbeit ist eine eindimensionale Theorie mit endlichen Dehnungen und Schubform~inderungen, fiir ebene Verformungen yon urspriinglich ebenen Balken. Das wesentliche der Theorie
ist die genaue Vertr~iglichkeit der dynamischen und kinematischen Gleichungen, insoweit das Prinzip der
virtuellen Arbeiten in Frage kommt. Die vorstehenden Entwicklungen sind vervollstiindigt durch Betrachtungen iiber das Problem der Aufstellung yon Spannungs-Form~inderungsbeziehungen und durch
eine Anwendung auf das Knickproblem des Kreisringes einschliesslich des Einflusses yon Axialdehnung
und Schubverformung,
(Received: May 5, 1972)