GENERAL
THEORY
STABILITY
IN
of
boulldnr~-\XlW
the mrrcnt
partly
proble1n.
shape
pIastit
At
UNIQUENESS
ELASTIC-PLASTIC
1.
JJ111:theory
OF
SCOl%
CJlr
_iNl)
SOLIDS
I.\r%l:
solitls
a
of strms
arc* suIqxMx1
detcrminecl
already,
the positional
changes and rotatiotw of material clcments are neglected,
the
bol~~~t~a.r~-~.al~~e
probfcm aiways has a. unique solution when the ~vorkl~~~r~leIlingis
~r~oIlot(~~~icand the yield function and plastic potenti~~l are idcnt,icat (XELAS
I!>38 ; a somewhat more rigorous and general proof can be found iu HILL I!%o,
pp. 53-G).
However,
the
really
typical
plastic* problems
involve
changes
in geometry
that cannot be disregarded, and here the question of uniqueness has remained
solids
open until very recently, when :ni answer was given for rigid-plastic
(HILL 195Sa). Broadly speaking, it is found that a unique solution is certain
only when the rate of hardening exceeds a determinable critical \;Llue, depending
on the particular situation.
Tn the present paper a similar result is proved for
an extensive class of elastic-plastic
solids, includin, a metals, with mechanical
properties
over,
the
336
A general theory
237
the question
of stability
standpoint,
namely via the energy criterion.
Previous work is
in that either the merhanical properties, the field equations. or the
boundary
conditions,
are often not, properly formulated
for a situation whcrc
rotations of material elements are large compared with the strains. Furthermore,
the occurrence of instability
existence of infinitesimally
unjustified
has invariably
near positions
with the
be quite
the present
intention
to review a confused
literature
nor to
attempt any correlation with experiment but to make a fresh start and establish
a broad basic theory free at least from the objections mcntioncd.
2.
The typical
traction
OF STRESS
considered
surface element
here is that
is assigned,
the change
irrespective
of
but the rate of change of the nominal traction based on the configuration
at
that instant. To formulate this boundary condition we introduce the unsymmetrical
nominal
stress tensor
with
certain
fixed
rectangular
axes.
Fj = 1i sij,
Fi is the nominal traction
condition is then that
SFi := li Ss,.
at t.
pj = 1; i,
is prescribed.
In equilibrium
are satisfied
of elements
through
The boundary
(1)
forces.
at stage
the association
If buO denotes
288
Now
R. Hm.
the material
in terms
prop&es
oii ~sw&t~d
v.
in
tht
with rectangular
I+om
first instanw.
axes
the transformation
rotating
ruk
for
where
In an clement
pllfrly
lo~&ti
whose response
il~finitrsimal
c~hange of stress
is
clastir (i.e. an clement. that has cithcr ncv?r yirldcd or is now fX1.H Iy 1111after being plastir). thr isothermal relatioiks l&wren stress-rate
and stzailt-
to be l~c~ln~~ell~(~usand linear,
as
A general
theory
recoverable
strain is finite,
subsequent
analysis*.
and indeed
the simplest
=K(E
K in thr
is
-
EP)
(7)
and K is a symmetric
when
2.79
solids
hypothesis
ir +&J
in elastic-plastic
matrix depending
n (i + 80) > 0
. . . . . . . . . . . . . . . . GO
n is stipulated
change.
moduli,
the equation
n (i + 6~) = 0
cffertively defines those stress-rate vectors that are tangential
a.t the current stress point, and thereby t.he surface clement.
the relations
of ir, equation
identically
(8) conforms
by a rigid-body
to the minimum
that
when
n(ir-+Off)> 0
. . . . . . . . . . . . . . . <o
This inequality
we satisfy
by requiring
nKn
> 0 in a plastic
element.
(9)
That is, the K quadratic form is positive for devintoric vect,ors ; this has the interpretation that a strain-rate in direction n constitutes 1oad;ng. Then, in a plastic
element,
,-+~u=~E-
nKc
____~_
I
i
Y > 0,
by
when
n (15 +
Bu)
0
(8)
. . . . . . . . . . . . . . . . <0
unreal materials.
. . . . . . . . . . . < 0.
I 0
nKE>O
(10)
where
(yn.
when
(nh'nd h)lin
interest
uniqueness
theorem
that a boundary-value
; this
solution
and extremum
problrm
is solvable
principle
then apply
R. HILL
240
with h = 0.
Now in fact the only elastic-plastic
term
nor in postulating
any particular
stress-dependence
of K itself
in the
elastic range, though we may still envisage measurable changes in K the to plastic
straining,
espcrially through changes in thr st.ate of anisotropy.
Equations
(7).
to the generally
accept&
relations.
of liylrostatk
(11)
which is a function
of velocity
gradient.
equal
to 3 iii ki/
,;..r]
I1
)=
9
t 2Xi,
(jE.
,iii
--
(J7),ihSi)
I.
In an elastic
clcnlent
A (b
In a plastics element,
This lemma
for h 2
is used in deriving
~ Bu) A E
AC K AC.
(1")
0,
thr uniqucncss
crit,crion
(Sec+on
5).
IJroqf.
The equality in (18) holtls when both strain-ratrs
call for additional
holds when n K E* -I o
loading, from (11) and the symmetry of K. Th(~ inqualit,y
and nKE < 0. lor then the left sitk of (13) is
produce
equality
unloading.
only when
In an elastic
element
element,
241
identity
l].
(14)
for h > 0,
L(B+Bu)AE+
A<KAe-----
(15)
When
which
EKE _
!KEr
nKn
+ hI
by (10) and the symmetry of K. If n K E* > 0 this is just the right side of (15)
If, however, n K E* < o the right side exceeds
which is therefore an equality.
the left by (nKE*):!,(nKn
+ h).
When nKE < 0 the left side of (15) is equal to
A (EKE) If nKE*
2:;);.
side of (15) is
A (EKE) -- ;wh
which
exceeds
side of (15)
the
left
is just
since
A (c K E) and
is not
less than
the
If
n K E < 0 the
left
(being
equal
right
when
nKAE=O).
5.
UNIQUENESS CRITERION
V as the present
O=
=
where
2(v)
surface
and volume,
A$AvdS==lls,A(~)ric
A(ti
+Bo)AedL-Z(Av)
v.
by a
from
I(. 111I.l.
242
:I suflicient
condition
for
uriiqucncss
A (i
is tlirrdorr
+ 6~) AC /I I
S(Av)
that
the uniqueness
condition
maximum
being
When
the criterion
bkTHElllUh1
(16) is satisfied.
extremum
the given
values
property.
on
So
the
in class
w1
(18)
implied.
6.
foilowing
243
solids
to give explicitly
(g+j)
/3 = Max
where
- >f3
5%
the u&bra&
can be rearranged
in elastic-plastic
l-)HINC!IPLE
the unique
solution
is characterized
velocity
by the
fields taking
functional
1
E(v)dL--
has an absolute
minimurn
when
AV :==v* -
ASA from
in the left-hand
and
(3)
I? being given
(19)
Jh
With
Jhis,.
the
usual
transformation.
= 1PAVtS.
integral.
From
(3) this
becomes
(&+&)AE~V+H(A~)<A
Combining
the
last
three
equations
and using
(16),
which
holds
for w -= A v
by hypothesis,
This
proves
A
the
IJ
(&$&)EdV-~(V)-2
extremum
principle.
true
(NILL
I
When
fivC&
geometry
nominal stress-rates
1956, pp. 65-B).
and
>O.
changes
not
are
negiigi blr
distinguished)
it
even
is not unique
and (16)
is not
K. IIILL
244
satisfied,
the
extremum
It
variation
principle
is essential
and (19).
(6 +
n K
to keep
0 u) E is equal
By contrast,
Further
insight
ness criterion
of solids.
A sufkient
for
all pairs
tinuous
where
some
by
be two
for
to a,zy velocity
tij
the
where
in
(16)
n h E -., 0,
of the plastic
zone.
where
A (kj;
and
in outline.
velocity
depending
is ttierefure
that
taking
only
the
general
the
class
(non-linear)
be
also
continuous
on strain
second
we
values
see
that
history,
derivatives.
l,outld:trv-valIIe
prcscrihed
inlinit~esitnally,
and uniquc-
more
IA
gradient
to the
fields
principle
prohlc~m.
on So.
Us
,it, inlplies
gmdicnt.
on the
3Xi),
hercb only
h tias to be included
nKtj.
thCYJWlll
t,he differeuce
gradient
of
licttl.
value
of velocity
:-
though
functionals
zone
f0I a still
solutions
differing
IllCall
the matter
distinc~t
uniqueness
pairs
sign
sectiondly
velocity
A v is now
the
containing
;uitl
of continuous
ve1ocit.y
part
of the
veloc.it\ _ gradient.
of the
could
function
(23-l), with
stress-rate
derivatives
the
solution,
idea
(20)
corresponding
Now,
basic
first
for
between
the terIn
irrespective
by studying
the
nominal
condition
specializing
each
of the plastic,
in the loading
E is a function
there
distinction
functiond
zone,
(+a11 be gained
continuous
of
to
plastic
between
Supposing
the
to E K E. Only
WC tles~ribe
relations
in respect
in the 1Iniqueness
in the
where
in mind
zone
0, is it equal
~z~yz&ere
exists
not.
In tlie elastic
E>
with
principle
may
of cr/ry two
fields,
ii vdue
Front
(21)
101
atld
245
unique
solution
a transf(jrmation
theorem
If, now,
gives the
principle
(23)
gradients
such that n K E = 0.
7.
STABILITY CRITERION
Suppose that a part of the surface of the body is rigidly constrained and that
constant
nominal tractions
(dead loading)
are maintained
on the remainder
t~ou~~lout, any movement from the position of equilil~riunl at stage t. Let u be
an arbitrary virtual displacement,, during the course of which the nominal stress
changes from sij to sij + 6S, (still referred to stage t). In a further infinitesimal
displacement du the increase in internal energy exceeds the work of the external
forces
by the amount
where S,
volume
and position
at stage t.
The total
path.'
excess
(24)
where the inner integral is taken along any path leading to each geometrically
possible configuration
infinitesimally
near the initial one, These configurations
need not of course be positions of equilibrium since they
attained (if at all) by free motion folIowing a transitory
However,
quasi-
statically under the dead loads ; in this event the integral vanishes, as may be
shown by direct transformation
or equivalently from the virtual work principle.
Of course the integral can vanish in other circumstances also.
In a plastic solid the change in stress is not a single-valued
function of the
total ~lisplacement gradient. and SO (a.& if only for this reason) the integral in
(24) is strongly path-dependent.
=2mong the paths leading to any one final position
the most critical for stability is that which calls for the smallest increase in internal
energy in the body as a whole. In attempting to determine this it seems desirable,
at the outset, to approach the problem by calculating the least increase in energy
w.
246
in bringing each element
sincbc tllis would generally
IfILl.
t,he straining.
These
stage)
in these
~Ol~~I~onents proportiotial
to the ,I,%
t0 second
order,
Iti terms
Of f*ompOncnts
a rigorOus
proof
referred
to tftc
to second
order,
Xij
lij
eik
ejk
and
fOr the work Of flistortion per unit initid volurtte on the Optitnum path.
In this
it is t0 be understood that Ssij and afcj/U, arf related in the same way as sij
namely thr0ugh (3) atld (10) if the elemf:nt is plast,ic,
through
atlfl
This expression fur the work is f-x-idcntly obtaincfl dso from
(6) if it is elastic.
:i path iti wlklt the tfisplaccmettt
field tluW.lglt0ut tltt~ tWfty is illcrtWcLl prctpftiOtt;~tcl>* atid ~Ot~tit~uffusI~ (possible if the tnrtteriat is (!Otttl~r~ssibl~~). It follows
that \vf. have f~ert:tittl_v not ~~~.~~~r~stittt~~te~
tltp lc:tst work.
3Zji
&pi,
ill>ll
A general theory
Equation
of uniqueness
ant1 stability
in elastic-plastic
by
E (v) rl LY=
(ti + Bo) E d V -
247
solids
as either displacement
.Z (v) > o
(26)
Now it was shown that the boundaryfor all fields v vanishing at t.he constraints.
value problem set by rigid constraints
on So and given nominal traction-rates
on
SF has a unique solution when
A(C,+fkr)A~tlV-
where
A v is the
differencc
of any
differences
of the
Z(Av)>o
pair
of fields
corresponding
(27)
vanishing
stress-rates
at the constraints.
and strain-rates,
but are not necessarily related by (10) in the plastic zone, on account of the nonlinearity
inherent
in the different
behaviour
in loading and unloading.
Consequently
(26) and (27) arc distinct criteria.
Nevertheless.
there is a definite
conncxion
in that (26) is always satisfied when (27) is (but not \-ice versa) ; for
(27) reduces to (26) when one field of a pair vanishes identically.
A partly plastic
state in which the boundary-value
problem has a unique solution is therefore
certuiuly stable ; and even when the solution is not unique the state lnuy still be
stable.
This is the general statement
and proof of a corljccturc
by SIIANLE~
(1946,
1947).
The situation
clearly
and h/p has the same value everywhere in the plastic zone.
Section 5, and supposing that. ,?I < IS,, (26) becomes
h
where
->a
ap
in class
H, (w) = II,
(w) --
isotropic
w,
of
(28)
from (17)
must change
with further
l&dS
the velocity
= s
dcformatior~
E(v)dV>
actual
modes.
x.
COllPARISON
\ilTII
~~~c;IU-~I..AS~~~~~
~IIFX)H)
COttsider t.he littiititig form of the prcviutts theorems whett t,tie clast ica ttlduli
are iti~cfit~itel~ ittcreaseci, the currtttt strtsstx itttcf r3t.t. of ltar(l(~ttirt~ being lteld
fixed. K is itow taken t.O he positive-deiittitc.
In the uniqucncss
criterion (16) the functiotd
2 is ittclepcttdcttt of the ttduli
whil(h the asymptotic
form of If is
n K n never vanishes.
H,, is ttever ttegatiw.
and \:tnishcs ottly for it~torlll)rcssible W fields such
itnrf q 3! n in the plastic zone (the l)ro~~or~i[)t~~li~~
For thcsc special fields
and tnagnitude).
since
by the
Schwarz
while fOr all others If,, and N c;~n bc ttde arbkraril?; large anri positive
ing the moduli.
Consequently
the sutIiciettt
condition
for uniqueness
the requirement
that
h$ <!I -- z(w) > 0
.I-
inequality,
by ittcrcasapproaches
.4 general theory
the
outset
in the plastic
in elastic-plastic
249
solids
by h E = (n &) n
for which reason
to becomc
consistent
at least when
were different
principle.
velocities
As the elastic
moduli arc increased, the value of the functional is unbounded except for incomprcssible fields v such that E - 0 in the elastic zone and EK n in the plastic zone,
wit,h a non-ne,@Yve
factor
of proportionality,
hE2dV -
Only
such
special
v fields
E(v)
satisfying
the
when
2
its limiting
value
is
Pvns,.
.!
velocity
boundary-conditions
need
be
consider
loads
and
the
set
constraints.
of
virtual
the
modes
functional
for
a rigid-plastic
solid
under
the
given
becoming
obtained
for a rigid-plastir
solid treated
on its merits
I 957d).
functionals
rigid-plastic
classes
differ.
REFERENCES
HILL,
R.
1950
1951
19.x
1937
RkLAN,
E.
.~jlIANLEY, 1.
R.
Mrrthenrnticd
Them1 of Plmticitq
(Clarendon Press, Oxford).
Phil. Mq. 42, 868.
.I. ML&. Plqs. Solids (a) 5, BO ; (b) 4, 247.
J. ;Ilcch. Phys. Solidr (a) 5, 1% ; (b) 5, 30%; (c) 5, 229 ; (d) 6, 1.
9,
1938
I?!.-Archill
1946
1947
116.