Sunteți pe pagina 1din 2

PAMM · Proc. Appl. Math. Mech. 2, 214–215 (2003) / DOI 10.1002/pamm.

200310092

Schieck, B.

A Shortcoming in the Geometrically Non-Linear Shakedown Theorem

The original static shakedown theorem of Melan [1], valid for geometrically linear theory, was extended for geometri-
cally non-linear theory e.g. by Polizzotto and Borino [2], who presented a proof for large rotations with small strains.
However, a counterexample to this extended Melan’s theorem has been found. The reason of the failure is investigated
and is corrected by an additional condition in the theorem. The outline of the proof is given.

1. Geometrically non-linear extension of Melan’s theorem and counterexample

Originally, the static shakedown theorem by Melan [1] was established for geometrically linear theory combined
with ideally elastic - ideally plastic materials. According to this, shakedown occurs if there exists a fictitious
time-independent self-stress state such, that the stresses due to variable loads superposed on the fictitious time-
independent self-stress field do not violate the yield condition. However, for geometrically non-linear theory the
stresses can’t be superposed additively any longer, but elastic and plastic deformations can be superposed. Thus,
for large deformations Melan’s shakedown theorem had to be modified by replacing the fictitious time-independent
self-stress field by a fictitious time-independent plastic strain field. E.g. Polizzotto and Borino [2] proposed this,
and they presented a proof for moderately large rotations with small strains.
But somewhere in their proof there must be something doubtful, because one
can construct a simple counterexample. Let us consider a bar with rectangular
cross section and simple supports at its ends. Its length l shall be very large
compared with the height h and the width b of the cross section. Let us assume
b = 2 h and ideally rigid-plastic material with the yield strength Re . This bar
is loaded by a vertical line load p that varies between zero and its maximum
value. Then, the limit load p max is determined by the bending momentum of
M = p l2 /8 and the full plastic resistance momentum of W = b h2 /4 , what
results in p max = 2 Re b h2 /l2 . In the present case with b = 2h the limit load
becomes p max = 4 Re h3 /l2 . If the load reaches or exceeds this limit, the bar
collapses. However, there exists a plastic deformation by a suitable torque load,
which causes that in the middle part of the bar the width of the cross section
becomes the height and vice versa. Then the limit load is p max = 8 Re h3 /l2 .
But usually, there is no chance for the bar to gain this limit, because the torque
that puts the cross section into the upright position is absent.

2. Corrected theorem and proof

The experience with the counterexample rises the idea, to correct the static geometrically non-linear shakedown
theorem by the additional condition, that the actual plastic deformations shall approximate the fictitious time-
independent plastic strain field. Then, the theorem reads: ”Shakedown occurs, if there exists a time-independent
plastic strain field Êp , such that the fictitious elastic stress response Ŝ due to the loads and Êp does not violate the
yield limit. In the case of not infinitesimally small deformations the time-independent plastic strain field Êp must
be chosen such that the plastic strain rates Ėp in the actual configuration reduce the deformation gradient difference
∆F = F − F̂ between the actual and the fictitious configuration.”
In the following the concept of the proof will be outlined. The commonly used presentation of the theory of
finite elastoplasticity is applied for convenience of the reader. In the proof two different configurations are considered:
the actual configuration with deformation gradient F , Green strain tensor E , 2nd P.-K. stress tensor S and plastic
deformation rate Ėp ; and the fictitious configuration with the time independent plastic strain field Êp , deformation
gradient F̂ , Green strain tensor Ê and 2nd P.-K. stress tensor Ŝ. ∆F = F − F̂ , ∆E = E − Ê and ∆S = S − Ŝ
denote the differences between the corresponding properties in both configurations.
The proof is based on the virtual work principle, where in both configurations equal volume and surface loads
pV and pS are applied, but the deformations u and û are different. However, the virtual deformations may be
Section 6: Damage and Fracture 215

chosen equally, δu = δû, what results in


     
 T 
pV · δu dV + pS · δu dS = S · F ∇δu dV = Ŝ · F̂T ∇δu dV . (1)
V S V V

By the choice of ∆u̇ = u̇ − û˙ = 0 as virtual displacement δu = δû (∆u̇ = 0 since in the actual configuration there
is plastic yield but in the fictitious configurations there is no plastic deformation rate) we obtain with ∆(2)E =
1 T
2 ∆F ∆F
         ˙  
˙
0 = S · FT ∆Ḟ dV − Ŝ · F̂T ∆Ḟ dV = Ŝ · ∆(2)E + ∆S · ∆Ė − ∆S · F̂T ∆F dV . (2)
V V V
 ˙
By time integration, where the term Ŝ · ∆(2) E is integrated by parts, and by introducing the estimation parameter
θ = O (||∆E||)  1 , e.g. θ < 0.3 , we get with C
| as tangent elasticity tensor

       
1 −1 (2) ˙T ˙ (2)
∆S · C · ∆S (1 + O (θ)) dV + Ŝ · ∆ E dV −
| ∆S · F̂ ∆F + Ŝ · ∆ E dV dt
2
V V t V




∆W1 ∆W2 −∆W3
 
= const. − ∆S · Ėp dV dt . (3)
t V


∆Wp

∆W1 is a quadratic form and therefore it is always nonnegative. ∆W2 may become negative, however for structures
with unstable deformation paths that can be stabilised by a suitable supporting construction (e.g. by stiff springs)
∆W1 + ∆W2 has a lower bound. ∆W3 behaves like an alternating series with absolutely decreasing terms, if we
split the time integration over the load history into loading and unloading intervals. Let us say, loading means that
˙ ˙
Ŝ · ∆(2) E dV > 0 and unloading < 0 , the sign is determined by the direction of Ŝ that is prescribed by the load
V
history in the fictitious configuration, and ∆(2)E = 12 ∆FT ∆F is decreasing due to the theorem. Similar it is with
  ˙
˙ ˙
∆S F̂T ∆F dV = F̂ (∆F ∆S) dV , where the sign is determined by the direction of F̂ that is prescribed by the
V V
load history in the fictitious configuration, and ∆F ∆S is decreasing due to the theorem (with ∆S limited by the
yield surface). Thus, ∆W3 is convergent according to the Leibniz criterion of alternating series, and therefore, ∆W3
is bounded.
Now we have seen that the left hand side of eqn (3) is lower bounded. But one can show that under the
conditions of the theorem ∆S · Ėp > 0 holds. In the case of small strains this is known as the material stability
condition; but for large strains this must be proved explicitly. The author did this, but due to lack of spare here
the proof must be postponed to a comprising paper. From ∆S · Ėp > 0 it follows that the left hand side of eqn (3)
is monotonously decreasing, although it is lower bounded as it was shown before. Thus at infinite time the plastic
strain rate Ėp must tend to zero, what means that shakedown occurs.

Acknowledgements

The author gratefully acknowledges financial support by the Fachhochschule Lübeck and scientific support by Prof. em. Dr.-Ing.
H. Stumpf at the Ruhr-Universität Bochum, Germany.

3. References

1 Melan, E.: Theorie statisch unbestimmter Systeme aus ideal plastischem Baustoff. Sitzb. Akad. Wiss., Wien, IIa, 145
(1936), 195 pp.
2 Polizzotto, C. and Borino, G.: Shakedown and steady-state responses of elastic-plastic solids in large displacements.
Int. J. Solids Structures 33.23 (1996), 3415-3437.
3 Schieck, B.: The transition from spatial covariant notation to co-rotational formulation in finite elastoplasticity. Lecture
notes of the GAMM conference in Regensburg, April 1997. ZAMM 78 (1998), S703-S704.

Prof. Dr.-Ing. B. Schieck, Fachhochschule (University of Applied Sciences) Lübeck, FB M+W, D-23562 Lübeck,
Germany; e-mail: schieck@fh-luebeck.de

S-ar putea să vă placă și