BKiHtaiHu SCC'ETitS LIBM/
JUN 241986
P. G.Hodge, Jr.
Department of Aerospace Engineering
and Mechanics,
University of Minnesota,
Minneapolis, MN 55455
K.-J. Bathe
Department of Mechanical Engineering,
Massachusetts Institute of Technology,
Cambridge, MA 02139
E. N. Dvorkin
Universidad de Buenos Aires,
Facultad de Ingeniera,
Buenos Aires, Argentina
Causes and Consequences of
Nonuniqueness in an
Elastic/Perfectly-Plastic Truss
A complete solution to collapse is given for a three-bar symmetric truss made of an
elastic/perfectly-plastic material, using linear statics and kinematics, and the solu-
tion is found to be partially nonunique in the range of contained plastic deforma-
tion. The introduction of a first-order deviation from symmetry and/or the inclu-
sion of first-order nonlinear terms in the equilibrium equations is found to restore
uniqueness. The significance of these effects is analyzed and discussed from
mathematical, physical, modelling, computational, and engineering points of view.
1 Introduction
In the linear theory of elasticity the solution to a "well-
defined boundary-value problem" is known to exist and be
unique. However, for an elastic/perfectly-plastic material
neither of these facts is obvious. Indeed, if the load is equal to
the so-called "yield-point load" the solution ceases to be
unique; beyond this load no solution exists.
The mathematical description of a boundary-value problem
for an elastic/perfectly-plastic material involves partial dif-
ferential equations and nonlinear constitutive relations.
Further, if actual answers are required, it is necessary to ap-
proximate the structure with a numerical model. Thus the dif-
ficulty in combining mathematical rigor and physical intuition
in the discussion of such questions as uniqueness is com-
pounded by the necessity of distinguishing between the true
nature of the continuum model and aspects introduced by the
numerical model.
However, many features of the general continuum are pre-
sent in much simpler structures where they can be more clearly
discussed. As an illustration of this approach, the present
paper is concerned with a particular simple three-bar plane
truss.
We begin by defining a "well-defined equilibrium problem"
for trusses as one in which the applied forces and
displacements at the joints are such that:
(a) overall equilibrium is not violated;
(b) overall displacement constraints prescribe a unique
allowable rigid-body motion (usually zero);
(c) at each joint in each of two independent directions
either the displacement or the applied load is prescribed.
It is not difficult to prove that the solution to any well-defined
Contributed by the Applied Mechanics Division for presentation at the
Winter Annual Meeting, Anaheim, CA, December 7-12, 1986, of the American
Society of Mechanical Engineers.
Discussion on this paper should be addressed to the Editorial Department,
ASME, United Engineering Center, 345 East 47th Street, New York, N.Y.
10017, and will be accepted until two months after final publication of the paper
itself in the JOURNAL OF APPLIED MECHANICS. Manuscript received by ASME
Applied Mechanics Division, November 27, 1984; final revision August 1, 1985.
Paper No. 86-WA/APM-8.
equilibrium elasticity problem is unique. Indeed, if two solu-
tions exist and are denoted by primes and double primes, then
it follows from the principle of virtual work that
W"/' -F, ")(e,-' - e, ") = E
Joint
,(P;' -P,")-(u,-' -u,") (1)
where F, and e, are the bar force and elongation, respectively,
and Vj and u
y
are the respective force and displacement vec-
tors at joint./'. If each solution satisfies condition (c) above and
we write the scalar product in terms of components in the in-
dependent prescription directions, then either P'?=P" or
u'=u" in each term on the right. Further, if each bar is
elastic, then
F, = k,e, (2)
where the stiffness k, is positive. Thus equation (1) becomes
Ebars^(e/'-e,")
2
= 0 (3)
which clearly requires that each e,' = e," and hence each
Fj'=Fj". Equilibrium equations then show each P
y
'=Pj",
and truss kinematics plus condition (b) above lead to each
u / -Uj•" =0. Therefore, the solution is unique.
Let us generalize the material behavior of equations (2) to
an "elastic/perfectly-plastic" truss where each bar behaves
elastically under small bar forces but can elongate indefinitely
when Fj reaches a certain limiting value Y
f
with similar
behavior in compression. Thus during any sufficiently small
time interval each bar is
EITHER
Elastic
OR Plastic
(F,
2
< F,
2
AND AF, = £,Ae,) (4a)
(F,2=r,
2
F,Ae,>0) (4b)
Consider first any well-posed equilibrium boundary-value
problem where any prescribed displacements are zero and the
loads Py are such that every bar is elastic. Clearly the same
unique solution will hold as in the ideally elastic truss.
Now consider this same truss under the set of loads XPy
where X is slowly increased from X = 1. So long as F,-
2
< K,-
2
for
each bar, the truss will remain fully elastic and the unique
solution will increase in proportion to X. However, for some
critical value X„ one or more bars will have their force
Journal of Applied Mechanics JUNE 1986, Vol. 53/235
Copyright © 1986 by ASME
Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 07/07/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use