Nuclear Engineering and Design 57 (1980) 333-368

© North-Holland Publishing Company

ON A SIMPLIFIED INELASTIC ANALYSIS OF STRUCTURES *

J. Z A R K A , J.J. ENGEL and G. INGLEBERT

Laboratoire de M~canique des Solides, Ecole Polytechnique, F-91128 Palaiseau, France

Received 9 November 1979

In this paper two main problems are considered: the derivation of cyclic constitutive relations during inelastic regime
where hardening, softening and creep can occur, and the development of the eventual periodical state in the structure dur-
ing cyclic thermodynamical loadings.
We give a very simple and practical framework to solve these problems in one unique manner.
Its essential feature consists in the introduction of a family of internal parameters which characterize local inelastic
mechanisms and the family of transformed internal parameters which are linearly linked to the previous ones through a
symmetrical non-negative matrix and are indeed the opposite of the associated residual stresses. Thanks to that, the treat-
ment of the local plastic or viscoplastic yield conditions can be easily made from only the classical simple purely elastic
(or viscoelastic) analysis.
This property allows important results during cyclic Ioadings: conditions for elastic shakedown, plastic shakedown,
ratcheting and bounds for the limiting state.
Several examples are given in the text.

1. Introduction

The effects o f cyclic loads and temperatures upon the performance and time-life o f engineering components
are o f fundamental interest to nuclear designers. Two main problems have to be considered:
(i) The derivation o f cyclic constitutive relations during inelastic regime (local constitutive equations).
(ii) The development o f the eventual periodic state in the structure (asymptotical global evolution boundary
values problem).
This paper concerns these particular problems.
To describe inelastic constitutive relations for materials (here in fact we shall limit ourself only to metals),
usually two different approaches are possible **"
(i) phenomenological relations, valid for only a few experiments, can be guessed and developed in the macro-
scopical approach,
or
(ii) the different local processes leading to inelastic strains can be analyzed and the behaviours can be deduced
after some computations in the microscopical approach.
An intermediate approach can be proposed *** in the regime where hardening, softening and creep have to be
considered. It is also based on the fact that the fundamental volume element we have to represent is a polycrystal
i.e., an aggregate o f crystals, b u t a simplification o f local behaviours is assumed.

* Expanded version of Invited Paper L3/2", presented at the 5th International Conference on Structural Mechanics in Reactor
Technology, Berlin (West), 13-17 August, 1979.
** A complete review of these approaches was recently given in [18].
*** Some authors, [1], [15,16] and [22], have introduced it, hut indeed it does not differ very much from the microscopical
approach.

333
334 J. Zarka et al. / On a simplified inelastic analysis o f structures

With some special constitutive laws, one has then to look at what will happen within the structure during a
cyclic thermomechanical loading. Different classical approaches are possible:
(i) One may take the real structure and the real particular cyclic loading or one may look at a scale model
and the corresponding scaled loading; this experimental approach is, in general, very expensive and will be limited
to that particular test.
(ii) One may use the big computer codes (based on finite element method for example) which are now effi-
cient; the numerical approach is also very expensive and cannot give a positive answer except for a relatively small
number of cycles.
(iii) One may utilize the available elastic shakedown theorems or the displacement bounding principles; this
theoretical approach is very complicated to use and says nothing about the number of cycles, plastic shakedown or
influence of the initial state.
(iv) One may extrapolate, like it is already recommended in some engineering codes, directly from the beha-
viour of certain particular elementary structures to much more complex structures; this engineering approach is
very simple but also dangerous.
Here, we shall describe our own framework to solve these two problems in one unique manner. We shall, in
fact, show how we may reach simple and practical answers. For example, we shall be able to obtain bounds of
the (or the real) limiting state with only elastic analysis.

2. Global constitutive relations

2.1. Experimental facts and principle o f the description

Under applied cyclic stress in the homogeneous uniaxial condition, some important facts are classically under-
lined (fig. 1):
(i) The eventual existence of a steady cyclic state for which the stress and strain change periodically; there is
elastic shakedown (where the sample will sustain a large number of cycles before failure) or plastic shakedown
(where low cycle fatigue will generally occur).
(ii) Some asymmetrical stress cycles may induce ratcheting, (for which progressive plastic strain increments
will drive to incremental collapse), even if during some lower mean stresses there may be shakedown.
(iii) Some cycles of loading are necessary to reach the eventual limiting state.
(iv) The eventual ratcheting or the cyclic inelastic strain may be the result of both plastic and viscous pheno-
mena during cyclic tests where varying frequencies or rest-periods are imposed.
On the other hand, the material is at our macroscopical level a polycrystal i.e., the fundamental volume ele-
ment is an assembly of crystals (or grains) which contain themselves some various components and some structural
defects. Its global behaviour will be associated with the local behaviours of its different constituents but will pre-
sent particular properties too [18].
Following that picture, we shall just say that the volume element we consider is symbolized by a discrete
assembly of local subelements or mechanisms with simple mechanical properties.
We shall distinguish four types of such elements; the three first ones, similar to the classical rheological models:
(a) sliders or friction blocks (for instantaneous strains with a threshold),
(b) dashpots (for viscous strains), and
(c) in parallel slider-dashpots (for delayed viscous strains with a threshold), are linked between them by
(d) linear springs (for linear elastic strains).
We shall take all these subelements as perfect.
As we shall see it later, hardening and softening will only be the result of creation and redistribution of residual
stresses within the volume element [ 11 ] and however a reasonable description of the cyclic behaviour will be
obtained.
 J. Zarka et al. / On a simplified inelastic analysis o f structures 335

O" ~ax
arna~ 1) 4~
II//111I
s

DE

%,.
,,_z if(2}
b) 2) _~_
11
a~

O'm,x

D/2/S ........ i i

cl
$
Fig. 1. Typical behaviour under applied cyclic stress (a) elastic
shakedown, (b) plastic shakedown, (c) ratcheting. Fig. 2. Elementary rheological models.

In order to help to the understanding of the framework at first, we shall give some examples of assembly of
uniaxial theological models, but we do not recommend to try to find if such an assembly can be associated to any
set of general constitutive relations that, later, we could easily write down a priori.

2.2. Some examples

2.2.1. Basic equations for classical uniaxial non linear rheological models (fig. 2)

(a) Slider
(i) The applied local stress o (l) must be plastically admissible at any moment i.e., it has to satisfy
- S o <~ u (1) < S o (1)
where So is the threshold for gliding.
This implies that o (1) belongs to the immobile convex set Co -= [-So, So].
(ii) When a (l) is strictly inside Co, there is no glide & = O; but when 0 (1) = +So, we can have any & >~ 0 and when
o O) = - S o , any & ~< 0.
These 3 cases may be symbolically written in the more simple form:
d~E a~Co (0 (1)) (2)
i.e., & is an external normal to the convex set Co in o (1) [13].

(b ) Linear dashpot
o (2), being the applied local stress and ?7 the damping factor, we have
=- o(2)/~, (3)
which implies that/~ = O, only when o (2) = 0.
336 J. Zarka et al. / On a simplified inelastic analysis of structures

(c) In parallel slider-dashpot

Due to the internal kinematic constraint in the group, the displacements of the slider and the dashpot are
always identical. If 0 (3) is the applied stress at the level of the group, we have:
1
~" = - [(o (3) - So)+ )a (3) +So(] , (4)
~7
i f x ~<0,
where (x)is the positive part of x - = { x0
ifx~>0 ;
[
i f x <~0,
and )x( the negative part of x - ~x
ifx/>0.
tu
We notice that -? = 0, when - S o ~< 0 (3) ~<S0 i.e., when 0 (3) E D o (5)
the immobile convex set [-So, So].

2.2.2. First example

Let us consider the elementary assembly which is given in fig. 3, it is composed of two sliders and three springs.
(i) The global stress, o, is applied to the assembly; the global strain, e, is the sum of the global inelastic strain,
e p, and the global elastic strain, ee; here we have:

6e=o/E, eP=Otl, with* ' a = ( al)a: ' (6)

where 01, as are the local inelastic strains.

The knowledge of lt~ is sufficient to describe the internal state of the assembly; they are thus its "internal
parameters."
(ii) At the level of each slider, the local stresses are given by

ot 1) = o - k1(~1 - ~ : ) / (7)
o~ l) = k l ( a t - a2) - k2a2 = k l a l - (kl + k2) as J
(iii) By definition we shall set:
Oi l) - h i 1 ) O - & l , o~ I)-~A~ l ) O - & 2 ' (8)

with here for "the localization factors"

IA(I) = ( A t l ) ~
~,A~)]
respectively
At D = 1, A~.l) = 0 , (9)

and

= = B(l)Ja. (10)
&2 -kl kl +k2 a2
We can easily see that the B (1) matrix is a symmetrical strictly positive matrix. ~&, which are here linked to ~a by a

* A back superior 'T' to a quantity means the field in the volume element, i.e., local quantity for the structure. On the contrary
a back superior "w" to a quantity means that we consider this quantity field in the whole structure (cf. section 3.1).
 J. Zarka et aL/ On a simplified inelastic analysis of structures 337

~2

-l-
i
I
I
~:~._,
I
i
..................... .. ..... [
II

if"
-[ t : , a
2S~

Fig. 3. First example. Fig. 4. Mobile convex setd(a) associated to the first example
and velocity field.

one-to-one linear relation, will be called "the transformed internal parameters."

(iv) With eq. (9), the global plastic strain e p is equal to

eP = ~A~I)~, - I/l(1)Tl~. (I I)
f

(v) Since ol 0 and cr~I) have to be plasticallyadmissible: ol I) E C~ = [-St, S,] and o~ I) E C2 = [-$2, $2],
&t and &2 have respectivelyto belong to the mobile convex sets~'I(o),d'2(o)which are subjected to simple trans-
lations:
e,(o) - I-s, + At')o, s, + At')o] -- c, +At')o,
~2(o) = [-s~ + A~)o,S~ +A~')o] = G +AP)o. (12)
In the (fib ~'2) plane, the transformed internal parameters l& belong then to the mobile convex set e (o), the rect-
angle whose center is in A(1)a and whose side lengths are (2S1, 2S2), (fig. 4).
(vi) For each slider r, the flow rules (eq. (2)) will be written &r E ~cr(orC1)), i.e.
- S t < o~ l) < Sr =" &r = 0 ,
o~ l) = St =" &r >1 0 ,
a~ 1) = -Sr ~ &r < 0 ,

or when introducing eq. (8) and assuming known the global stress,
- S r +3(1)o < &r < S r +A~ l)q =~ &r ~ 0

&r = - S t +A~l)° =~&r > 0 &r E --OxIt~r(O)(&r). (13)

&r = Sr + A~l)o =~&r <<-0
&r is thus an internal normal to the mobile convex set Cr(o) in &t-
(vii) Simultaneously for the two sliders we can write down the flow rules
lt~ E - a~I,6<o)(la), (14)
and for the velocities of the transformed internal parameters
l~t E - - B ( 1 ) a x I t o . ( a ) ( & ) , (15)

We have shown in fig. 4 the particular meanings of eqs. (14) and (15).
338 J. Zarka et al. / On a simplified inelastic analysis of structures

kl
Ct!

SZ d

Fig. 5. Second example.

2.2.3. Second example

Let us now consider the assembly which is shown in fig. 5. We have still two sliders but they are now connected
differently. It is very simple to obtain the following properties [23]:
(i) la - ( a 2a)l are the two internal parameters of the system.
(ii) The global stress o being assumed known, tile local t'~resses at the l~vel of each sl

oil) kl klk2 k2 klk2

= - - O -- - - (a I -- 0/2) , O~ 1) -- O+ (0/1 -- 0/2), (16)
kl + k2 kl + k2 kl +k2 kl +k2
and can be written as in eq. (8) with:
At l ) = k t l ( k , + k : ) (>0); A~ l)=k21(k, +k2) (>0), (17)

&2 ka +k~ -1 1 /\0/2:

but this time the symmetrical matrix B (l) is singular and we always have the compatibility condition:
~: &l + &2 ; 0 . (19)
The transformed internal parameters l& are linearly linked to 10/but the relation is not one-to-one. For a "com-
patible" t& (satisfying eq. (19)), there are an infinite number of solutions 10/such that eq. (18) is verified.
(iii) Eq. (1 I) which expresses the global plastic strain is still valid. The global elastic strain is, here, given by
e e = o/(kl + k2). (20)
(iv) The evolution laws for 10/and l& keep the same shapes as in eqs. (14), (15).
The one differency with the previous example lies only in the singularity of the B (1) matrix which implies eq.
(19) (fig. 6).

2.2.4. Third example

In fig. 7, we have now a more complex assembly of two sliders and three springs. We easily obtain when anal-
yzing the local stresses at the level of each slider:
AI l) = 1 (2>t3); A~ l) =-ka/(k2 +k3) (<0), (21)

,,,, =(~1 -~, ). (22)

-kl (klk2 + k2k3 + k3ki)/(k2 + k3)
The Ar(l) have opposite signs; the matrix B (1) is strictly positive. All the other evolution equations are still valid.

2.2.5. Fourth example

We shall consider the assembly of one dashpot and one in parallel dashpot-slider and one spring as shown in
 J. Zarka et al. / On a simplified inelastic analysis o f structures 339

~2 Ik ('l1Tl~/, SI

/
s/

~m-.~-~--- . s"/

' k3 ,
~T
• / $/ _I~ A
I
~S
Fig. 6. Mobile convex set @(a) associated to the second Fig. 7. Third example.
example, velocity field and compatibility condition~.

fig. 8. The spring symbolizes hardening while the dashpot characterizes the softening. The internal parameters of
this assembly are (~), the displacements of the two inelastic models. Always with the same procedure * as in the
previous examples, we obtain:
(i) The local stresses:
a(2) = A(2)o - fl , 0(3) =A(3)o - ~" , (23)

with

A (2)=0; A (3)=1, (24)

and

The matrix q~ is singular and we have the compatibility condition

~: ~ + ~, = o . (26)
(ii) The evolution laws

= o(2)/n~ = -t~/n~ = S i n s , (27)

1 1
? = - - [<o(3) - $1) + ) o (3) + $ 1 ( ] = - - [ ( o - ~ - S1) + ) o - ~ +$1(] (28)
rh rh
and
(iii) The velocities shown in fig. 9.

2.3. General f o r m u l a e (for a more precise presentation see [23])

Let us now forget the previous uniaxial inelastic models but let us keep in mind only the formalism.

• We compute the local stresses at the level of each inelastic model; we have two terms, one is linked to the applied stress o and the
other one characterizes the residual stresses due to the inelastic strains.
340 J. Zarka et aL / On a simplified inelastic analysis of structures

I I

i
.,cO I ° i

lJ2 kl
\P'" 1 ^
10[ - $| I I~+ $I •

///////2////,
Sl
Fig. 8. Fourth example. Fig. 9. Velocity field associated to the fourth example.

2.3.1. Internal parameters

(i) We shall assume that the internal state of our volume element can be characterized by a family I x of internal
parameters (scalars or tensors)

IX= 1~ , (29)
13,
(ii) and that these parameters are directly linked to the local inelastic strains of the three different types:
10t is the vector made of the instantaneous plastic strains with a threshold,
1/3 is the vector made of the delayed viscous strains without a threshold, and
17 is the vector made of the delayed viscoplastic strains with a threshold.
The total number of parameters will be assumed finite and designed by n.

2.3.2. L o c a l stresses on the inelastic subelements

At the level of these inelastic subelements or mechanisms, we shall assume that the local stresses are respectively
given by

lo( ) / o- =IAo- X, (30)

~o(3)J L'A(3)J 'i,
where
o is the global stress tensor * applied on the volume element,
lo(]) is the vector made o f all the local stresses (scalars or tensors) for subelements of type 1 **, so for la(2)
and 1o(3) and

* Or any generalized forces as for rods, shells, plates ....

** I.E., mechanism with instantaneous plastic strain and a threshold.
 J. Zarka et al. / On a simplified inelastic analysis of structures 341

lA (1) is the vector made of all the localisation factors (scalars or tensors) for the subelements of type 1, so for
IA(2) and IA(3).

12__ I~ ,
t?

with I&, l~ and 15,, the opposites of residual stresses (scalars or tensors) at the level of inelastic elements, will still
be called the transformed internal parameters.

2.3.3. Transformed internal parameters

We shall assume that the transformed internal parameters l,(, are linked to the internal parameters IX through
a linear application:

E B (3 17

where ~ is a non-negative symmetrical matrix.

Two cases will have to be considered following to:
(i) q~ is strictly positive and thus regular:
Ij~r = ~ 1 x :=~ |X = q~-I 1)~ (one-to-one relation),

or
(ii) °B is singular of rank (n - p ) :
.~" must belong to the linear subspace of compatibility ~ .
To such a compatible l.~, it may be associated an infinite number of solutions IX such that
p
iX = 61.~" + ~ / a t I X i , (32)
1

where I~. is one particular solution ofq~6l.~ " = t~,; IX are p linearly independent solutions of the homogeneous
system c~|X l = (0), and t~l are undetermined factors.

2.3.4. Evolution laws for the subelements

(a) Instantaneous plasticity

We shall assume, for each such a subelement or mechanism, the perfect-plasticity behaviour:
(i) Co, an immobile convex set, defines the threshold * (Io(l) must belong to it);
(ii) the associated normality flow rules are satisfied:

l~ ~ ~q, Co(iO(1)), (33)

i.e., l& is one external normal to the convex set Co in IO(1).

* There might be many inequalities to characterize Co which may have corners. The origin must be a center of symmetry for the
convex Co.
342 J. Zarka etal. / On a simplified inelastic analysis o f structures

As we showed it in the examples when introducing l& for all these subelements, we can write at any time t:
(i)
lot E Q(o) , (34)

with
Q(o) = e ( 0 ) + 1A(Oo. (35)

e ( o ) is a mobile convex set which is subjected to a simple known translation and which is in fact built locally
for each such a subelement, and
(ii)
lot E - 8,I~6(o)(&), (36)
i.e., 10t is one internal normal to the convex set e ( o ) in Jot.

(b ) Viscosity
We shall assume that there is for each such a subelement a potential 60 such that:
(i) 60 is only a convex function of o (2),
(ii)
= 860(0(2))/80 (2) , (37)
and
(iii)
= 0, only if o (2) -=-0 . (38)
This implies we have locally a perfect viscous behaviour without a threshold. In fact we shall assume later the
linear viscosity as for example in the form
1~ = 10(2)/|~ , (39)

When introducing Ill, for all these subelements, we can write

l/~ = _alw(IA(2)o _ 1~)/81/~ ' (40)

or
1/~= (IA(2) o _ I/])/ir/" (41)

(c) Viscoplasticity
For each such a subelement, we shall assume that there exists a convex viscoplastic potential I2 such that:
(i) ~2 is only a convex function of o (3),
(ii)
~' = ~~(o(3))/8o (3) , (42)
and
(iii)
"~ - 0, only if o (3) ~ D ~ , (43)
an immobile convex set which defines the threshold.
We thus have locally a perfect viscoplastic behaviour.
Here also, when introducing l~, we shall obtain for all these subelements
I,~ = --SlI2(IA (3)o -- 5')/8t~, (44)
* Do must also have the origin as a center of symmetry.
 J. Zarka et al. / On a simplified inelastic analysis of structures 343

and
l~ = 0, only if 1~ ~ cb(o) ' (45)
with the mobile convex set given by
+ IA(3)o . (46)
The properties and the building of C~(o) are identical to those of f~(o).

2.3.5. Global evolution o f the volume element

(a) Global inelastic strain rates tensor

In order to keep the idea that all the subelements are linked through a linear elastic matrix, as it was proved in
[10,6,2,21] and as we got it in eqs. (11), (28), we shall take:
e~ =--IA(I)T It~ + IA(2)T 1/~ + IA(3)T 1,~,= IAT 1•. (47)

Moreover, so, the local normality and convexity rules induce those for the global behaviour.

(b) Global elastic strain rates tensor

We shall assume at last that
~e = c?~0 , (48)
where cl?~is the symmetrical positive matrix of elastic coefficients.

(c) Global strain rates tensor *

= ~e + ~p. (49)

Once given the initial values of Ix and o, (IXo, Oo), for any loading path in stresses o(r) or-in strains e(r) (or
mixed components in o, e) for r E [0, T], the global behaviour of the volume element can be analyzed.
It is worthwhile noting that hardening and softening at this global scale are only the result of stresses redistri-
bution within the volume element since all the plastic and viscoplastic subelements are "perfect" [11 ].
Two fundamental properties of this framework for the inelastic constitutive laws have to be underlined:
(i) A simple analysis during cyclic loadings will be possible, even if we may think we need to follow step by
step the volume element and its parameters;
(ii) A similar simple analysis, for any structure which is made of such materials, will be also available.
They are due to the introduction of the transformed internal parameters and their particular linear relation
with the internal parameters.

2.4. Global cyclic behaviour

2.4.1. Position
(i) Many important theoretical papers were published about shakedown [4,5,7-9,12,14,15,19,20]. Here, we
shall thus not prove again the classical statical fundamental Melan's theorem for a perfectly plastic or viscoplastic
structure.
(ii) Due to the viscous and viscoplastic subelements we need to fully know the cyclic loading path during one
period and not only its bounds. Here we shall take a loading path in stresses.
(iii) Subelements of types 2 and 3 are formally similar, the elastic domain for the first ones being only 0. That

* A thermal component could be added too.

344 J. Zarka et al. / On a simplified inelastic analysis of structures

is the reason why from now, we shall just keep linearly viscous subelements in 2, the other ones will be put with
3.
2.4.2. Purely linearly viscous global behaviour
Let us assume, at first, that la and 17 are all kept at the 0 value, the global behaviour will only be characterized
by the linear viscous subelements. We must solve the system of linear differential equations:
I~(t) = (l A(2)o(t) - B (2) 113(t))/rl . (49)
The submatrix B (2), which was introduced in the matrix q~ (32), is also symmetrical and non negative. It has
non negative eigen values ~k. The general solution of eq. (49) is thus:

l~(t) = I~v(t) + ~ Pk(t) exp(--Xkt/rlk) , (50)

k

where l/3v(t) is a periodical particular bounded solution of eq. (49a) and Pk(t) is a polynomial function of t whose
degree is equal to the order of multiplicity of ~.k minus 1.
When B (2) is strictly positive all the ~,k are positive; for t going towards infinite, )/3(t) will go towards l/3v(t).
When B (2) is singular, some ~.k are nul; eventually/3(t) will be unbounded for t going towards infinite.
We shall assume, from now, that B (2) is always strictly positive, or otherwise there is no viscous subelement.
2.4.3. General response
(i) For the volume element, let us set
tj3(t) ---l/3~(t) + l~31(t), (51)
where l/3v(t) is the periodical bounded solution of eq. (50), and let us define the "new transformed parameters
12~ such that

r,1 I'l
121 =t1~1/ =q~ 1~1 and
r<,vl
1 2 = 1 " , +]B(2)l/3v] . (52)
L'*,J L' J LE' v J
(ii) The evolution laws for the subelements will be changed into:
1-Subelements
~&~ -a,I, c~(o, lad (~al), (53)
with
1~, E e"l(o, I/3v), (54)
and
@1(o,lt3v)--=@(0) + IX(1)o - C I" ~v. (55)
2-Subelements
l/~1 = --151/17/ and 1/~= 1/~v + l/~l . (56)
3-Subelements
l~, = _aIi2(A(3)o _ El/3v_ l~l)/Ol~t I , (58)
and
l~, = 0 , if 15'1 E ~ l ( o , l~v), (59)
 J. Zarka et al. / On a simplified inelastic analysis of structures 345

with
c~,(o, I fly)-c/)(O) + 1-4(3)o - EI/3v •
"~ (60)

2.4.4. Eventual "elastic" shakedown

Our definition for "elastic" shakedown is here that all (la, 1/31and 17) will tend towards constant values when
t is going to infinite.
Two cases must be considered [23] :
(i) cB is regular,
The condition for "elastic" * shakedown is very simple to express
There is a non void intersection S t o f convex sets

~l(o(r), l/3v(r)) (61)

in the &l space and a nonvoid intersection c~ I o f convex sets C~l(o(r), i~v(r)) in the I~[ space when r E [0, T].
This condition implies indeed that by taking any I&~ E ~1, t ~ = lO and any l ~ E ~ l , we have found a con-
stant residual stress field

l';J
which is plastically admissible i.e. the condition of the generalized Melan's theorem.
But, since the convex sets St and ~ t are locally built, here we have a local condition.
(it) cB is singular.
The generalized Melan's theorem is always true; but this time l ~ must also belong to the global subspace of
compatibility Z? in the l~, space. The condition for "elastic" shakedown will have now a global character:
There is a constant new transformed parameter If(*1 such that

IX~= II~[ E , ~ , (62)

L'*;J
with l&~ strictly included in St # ~, 1~ ~ 0, l~l strictly included in ~ l # ~b.
For both cases, generally, it is not yet possible to give the number of cycles necessary to reach the limiting
state; this number may be infinite.
2.4.5. Eventual "plastic" shakedown
The two cases have still to be considered:
(i) c~ is regular.
When condition (61) is not satisfied, there will always be plastic shakedown.
(it) cl~ is singular.
When condition (62) is no more satisfied, it is not yet possible to givean a priori knowledge of the limiting
state. However when there is no collapse (i.e. the intersection between ./~ and Ql(o(z), I/3v(r)), for T E [0, T] is
non-void what is the necessary condition for the existence of I&l(t)) there will always be a periodical solution
IXl (t) while IX 1(t) may be periodical (plastic shakedown) or not periodical (ratcheting).
Here also, in general, the number of cycles to reach the limiting state cannot yet be given.

* In fact, the asymptoticalglobal behaviour will be viscoelasticsince we must add the particular periodical viscous solution as in
eq. (52).
346 J. Zarka et al. / On a simplified inelastic analysis o f structures

2.4.6. Practical rules during radial cyclic loadings for a regular Q3 matrix
The cyclic loading will be said radial if we can write:
O(T) = )t(T) Omi n + ( l -- ~.('/')) O m a x , (63)
where ~,(z) is a periodical function of time varying monotonically between 0 and 1 and then between 1 and 0, and
Omtn and Oma x are two constant global stress states.
(t) It is evident that two particular loading rates, which will give direct bounds for the real one, may be con-
sidered:
A "static" loading such that I~Plwill always be very small at any time (T is very "large"); this will imply that:
1~ will always be near from IA (2)0 and 15, will always approximately belong toC~ (o). (64)
In other words in the volume element, we shall have to consider only the plastic subelements 1 and 3 with
their internal parameters (Its, 17); the internal parameters 1/3of subelements 2 will be known afterwards by the
condition I/~ ~ IA(2)o:
113= (B(2))-I(1A(2)o - Cla - ETIT) (65)

and by eq. (47) we shall deduce e p.

A "dynamic" loading such that IdPl will be almost infinite at any time (T is very "small"); this "instantaneous"
loading will produce 13= 0 and 7 -~ 0 i.e., only the subelements 1 will take part in the deformation of the volume
element.
So, for these two loading rates we need only to analyse a volume element with subelements of type 1 and for
which only the monotonical variation between Ornin, Omax and then Omax, Omin is important.
(ii) In [24], we gave some practical rules to obtain bounds for an elastic plastic structure with kinematical
hardening. The procedure will be still similar when the matrix qfl is strictly positive (its submatrices

B(1) and (B(I) DT )

\D B (3)

are also strictly positive) and one short abstract will be given in the annexes of this paper. It is evident that for a
regular matrix qfl only the stress amplitude Omax -- Omin will be important (the mean stress (Oma x + O'min) will not
take part).

2.5. Examples o f global inelastic behaviours within this framework

In all the examples we shall write down "a priori" general equations, we shall just keep in mind the mechanisms
at the local scale.
From the previous analysis, we already know that there may be ratcheting or collapse only when qfl is singular.
If then, we need to represent experimental results, according to the eventual observation of the ratcheting or
collapse we shall or shall not take a regular matrix.
Now, in the following elementary examples, we shall see some properties which will allow us to interpret more
and more complex behaviours. A combination in serial of such simple behaviours may allow to interpret any of
the known cyclic phenomena. We shall consider only examples of rate independent inelastic behaviours since in
the next section we shall treat a viscoplastic structure.
In order to simplify the presentation, here we shall assume that the loading is always biaxial and if the first
component will vary periodically with time, the second one will be kept constant at its initial value:

o= 02 (66)
0
 J. Zarka et aL / On a simplified inelastic analysis of structures 347

This special loading belongs to the radial cyclic loadings described by eq. (63).

2.5.1. One one-dimensional subelement 1

We just say that the unique local subelement 1 has a onedimensional characteristic; it is the case when, for
example, we consider the slip of one particular slip mechanism * in a single crystal.
In that case we shall write that the most general behaviour within our framework is given by
(i) local stress:
O(1) = A t l ) o l + A i l ) o 2 - &, (67)
(ii) parameters:
& =Bi])(~, with Btl) >~ 0 , (68)
(iii) flow rules:
-S<~o(')<~S or Atl)ol-S+Ai')o:<~&<~AI')oI +S+Ai')o2,
or
&Ee(al,o2), dtE-a~(ol,o2)(l&), (69)
(iv) component of the global plastic strain tensor along the direction of Ol
eV (70)
(there may be other components not only along 02 but also in the other directions: e~ = A!])c~.)
From eq. (68) we can see that only the perfect plastic behaviour (Bt~) -- 0) or the kinematical hardening
behaviour (Bill) > 0) can be obtained.
The number of cycles of loading to reach the limiting state (shakedown or ratcheting of collapse) is always
equal to 1.

2.5.2. Two one-dimensional subelements 1

In the sections 2.2.2., 2.2.3. and 2.2.4., we have already shown particular cases of volume elements where two
such subelements were put. The most general behaviour will be now given by
(i) Local stresses:
first element ol l) =At~)ol +At~)o2 - &l, second subelement o~ 1) =A~1)Ol +A~)o2 - &2. (71)
(ii) Parameters:

with B (1) a non negative matrix.

(iii) Flow rules:
l& E d ( o l , o2), (73)
the rectangle whose center is
(Alll)O'l +'All2)02 ,A~I1)0-1-{-All2)02)

* The slip mechanism is, as it is well known, defined by a slip plane direction and a slip direction the corresponding ~ is in this
case the relative glide (scalar) along the slip direction between two planes which are distant from unity of length.
348 J. Zarka et al. / On a simplified inelastic analysis of structures

and whose sides are 231, 2S 2 and

l& E --axP~(ol,o2)(l&).
(iv) Component of the global plastic strain tensor along the direction Ol:
el (74)

The matrix B (1) is regular:

The limiting state will depend only on the stress amplitude m o I = O'lmax -- Olmin.
For increasing values of this stress amplitude we shall obtain elastic shakedown, partial plastic shakedown and
total plastic shakedown (fig. 10).
When A t~)/> 0, A ~ ) >i 0 and Bt l) 1> 0, the limiting state is always reached after the first cycle (fig. 10). On the
contrary, when A t~)A ~) < 0 and Bt ~) < 0 (or A t~)A ~) > 0 and Bt12) > 0), the elastic shakedown state (when

22

~(O'max.)
\

F a'1

t C(°'min')l

~2 a

~rmi..

Fig. 10. Regularmatrix B(1): elastic shakedown,partial plastic shakedown,total plastic shakedownafter one cycle.
 J. Zarka et al. / On a simplified inelastic analysis of structures 349

,, ,/Z'._ .,¢[--'f ......

Fig. 11. Regular matrix B(1): elastic shakedown after an Fig. 12. Regular matrix B ( 1): plastic shakedown after a finite
infinite number of cycles. number of cycles.

~l = ~(Olmi n, O2) (3 ~(Olmax , 02) # ~b) is reached whether after one cycle of loading or only after an infinite number
of cycles (fig. 11) while the plastic shakedown state is reached after any finite number of cycles (fig. 12).

The matrix B (O is singular:

This implies that BI~ ) = 5 ~ with 8 -- +1 or - 1 , and that there is a subspace of compatibility Z?:

(8~/(8,~)/8t~,)) ,~, - ,~ = o)
There is a solution (no collapse) only when ~ N O(ol(r), o2) # q~ at any time r.
The limiting state (elastic of plastic shakedown or ratcheting) is always reached after the first cycle.
The mean stress Ol,,, = (o l max + oitn!n)/2 takes also part; so according to o l m for a given stress amplitude Aol, it
may be elastic shakedown (when Z? N e (Olma x, a2) n Q(o] rain' O2) # ~b), plastic shakedown, ratcheting or collapse
(fig. 13).

2.5.3. One two-dimensional subelement 1

In order to have a picture of this subelement we may think of a slider which is moving on a plane. The most
general behaviour will correspond to:
(i) Local stresses:
ai ')=Ai~)O, + A I ~ ) a 2 - & , , ot 1)=A[~)O, + A [ [ ) 0 2 - 5 2 . (75)

~,(o-,,,,,,.)

Fig. 13. Singular matrix B(1): elastic shakedown, plastic shakedown, ratcheting and collapse are possible.
350 J. Zarka et al. I On a simplified inelastic analysis of structures

F'A~IAr'IE'~RCS sr4A×, ,3,5

A,I --? SM~N, -- 3 0

Sl ~6 ~2

Fig. 14. Two-dimensional subelemeni 1. Singular matrixB

(l) : Plastic shakedown.

(it) Parameters:

= // /=~ ) a (76)

Oil) F l o w rules:
We shall, for example, assume that the two-dimensional threshold Co is the circle:
(otO) 2 + (o~O) 2 ~< S 2 • (77)

In the l& plane, we shall thus get that ~ ( o i , 02)is the circle centered in (All)oi +Atl2)o2, A ~ ) o l +A~12)a2) and
whose radius is S and that
la E -~',Is~(oi,o2)(l~). (78)
(iv) Component of the global plastic strain tensor along the direction al
elp = A t il)al + A ~ ) a 2 . (79)
The eqs. ( 7 5 ) - ( 7 9 ) are similar to eqs. ( 7 1 ) - ( 7 4 ) (only the shapes of the two respective convex sets e are dif-
ferent).
The global behaviours will thus be similar (fig. 14 and 15); the most important difference is in the non-constant
hardening modulus during monotonic loading which is induced by the rotation of the internal normal on e ( o i , o~).

2.5.4. Two two-dimensional subelements 1

Now let us assume we want to find complex behaviours based on two surfaces as they were introduced for
example in [17,3] but however let us assume we want still to stay in our framework. We may, for example, take
two two-dimensional subelements 1 but with a special B (1) matrix *"

• With this matrix, it will be possible to give a plane representation of the different cases of loading.
 J. Zarka et aL / On a simplified inelastic analysis of structures 351

. . . . . . . . . . . I E~I I. I , ~I/I//III, III(///',,': ,, ~.

Fig. 15. Two-dimensionalsubelement 1. Singularmatrix B(1): Ratcheting.

(i) Local stresses:

first subelement
ot ') = Al;)o, + Al'?o~ - ~,, (80)
second subelement
4 ') -- A t ; ) o , +A~'?o, - ~3, 41) = A~)o, + Al~)o2 - ~,. (81)
(ii) Parameters:

]hBtl) hBt~) hZBl 1) hkBll)l = B(I) 1°~ '

(82)

LkBt~) ksi~) hkSi~) k~Bi~)J

with Bi~)si~) - (8i~)) 2 > o and hk > O.
The particular B (1) matrix is singular of rank 2.
The subspace of compatibility .~ is defined by the two linear relations
&a =h&l, &4 =k&2 (83)
the limited space

:(°, t
52

will thus be sufficient.

(iii) Flow rules:
352 J. Zarka et al. / On a simplified inelastic analysis of structures

We shall take for the two-dimensional threshold Ca and C2 respectively:

(oil)) 2 + (o~l)) 2 ~<S~ (circle),
(at l)/h) 2 + (o~O/k) 2 <~S~ (ellipse), (84)
These two conditions, expressed for

lot, = ,
&2
show that lot, must belong to the intersection of the two circles:
~I(OI, O2) centered in (At])ol +At~)o2,A~)oI +A~)02), radiusS, (85)
and
e2(01,02) centered in ((At])oa + A~)a2)/h, (A~)oa + A~)o2)/k) , radiusS2 (86)
The normality flow rules, when using t&,,
ietl)~
(87)

will be written here:

--~X/~l(O'l, O2)( O/ ) ; \kot4 t ~ --~I/~2(O1' O2)( O/ ) •

(88)
&2!
From that we shall deduce:
.
lot,=
=(ei ) 8t ) al ) o )dot') (89)

(see in fig. 16 the velocities fields).

(iv) Components of the global plastic strain tensor in oa, 02 directions:

[Ai;)., +A i% +All).. i (90)

Monotonic loading
(i) In order to have a solution, the intersection of 81(Ol, 02) and 52(oi, 02) must be non void. The condition
for no collapse (or the "ultimate" yield surface in the (ol, 02) plane) is thus:
[(A t~) - A ~ ) / h ) 01 + (AI~) - A ~ ) / h ) 02] 2 + [(A~1) - A~l)/k) at + (A~ 1) - A ~ ) / k ) a2 ]2 < ($1 + $2) 2 . (92)
(ii) We can follow the evolution of the elastic domain in the (el, o2) plane during a monotonic loading (see
fig. 17). This elastic domain changes in shape and size (it is the intersection of two ellipses); it "seems" to be
moreover subjected to a rotation. This evolution is however quite different from that assumed in [17].

Cyclic loadings
(i) As long as 16t' stays at any time inside one of the circles ( e l , ~2) and on the boundary of the other one, we
find again exactly the behaviour 2.5.3. with only elastic or plastic shakedown (since Bt~)B~[ ) - (B~[)) 2 > 0).
 J. Zarka et al. / On a simplified inelastic analysis o f structures 353

_: ,~2

/ I
/ %
/
/
/
/ \
\
// \

\
J (2) (31 {4)
1 6

/ I

/ T'I : °,
'°'x--L/ i
I
I
I I I I I I I I
!
I
!
/
/
/
0 ~ 1 ~2 --3--4--S-6//
\ /
\ /
~' /
\,, /
/
t /

Fig. 16. Two-dimensional subelements 1. Velocity field. Fig. 17. Two-dimensional subelements 1. Evolution of the
elastic domain during a monotonic loading.

PARAHETRES SHAXs 3 5
At -I SH~N= - 2 5
A2 t T , 0
A3/H -I A

B2zl 2 ~ e e . \ T

l: . . . . . ill~
al

Fig. 18. Two-dimensional subelements 1. Ratcheting

354 J. Zarka et al. / On a simplified inelastic analysis of stnwtures

(ii) As soon as l&' belongs at some particular times to the intersections of both two boundaries of the circles
~1, ~2, the behaviour will be different and ratcheting may be obtained (fig. 18).

2.6. Effects o f temperature

Without changing anything to our framework, we may take very easily some influence of temperature if we
assume that only the matrices IA, q0 and c ~ will stay constant.
This hypothesis, in fact, implies that the local elastic properties * of the matrix which links the inelastic sub-
elements are independant from temperature.
When the temperature range is not too large (a few hundred degrees), this hypothesis is reasonable.
The local inelastic subelements can have general coupling with temperature:
(i) For subelements 1., we can only have that the size and shape of the thresholds Co(O) is linked to the tem-
perature.
(ii) For subelements 2, we can say that the viscous damping factors r/(0) are functions of temperature **
(iii) For subelements 3, both the thresholds Do(O) and the viscoplastic potentials ~ ( o (3), 0) can be tempera-
ture dependant.
The convex sets ~ a n d c~ will be then not only subjected to a translation but also to a change of size; so, for
example, insteadof eq. (35) we shall have
~(O, 0) = ~(0) + IA( l)o, (92)

however their buildings will still be local.

3. Global evolution of structures

3.1. Problems

A structure is symbolized by a volume V with boundary ~ V. It is subjected to some thermomechanical load-

ing, for ~"E [0, T] :
wxd(r) body forces per unit mass in V, w0d(r) temperature change in V, WF~/(r) surface forces on a part ***
~Fi V o f 3V and Wu](r) displacements on the complementary part ~ui V.
Here our first problem will be to obtain its global evolution when it is assumed that the constitutive laws are
as given in section 2.3.
We shall show, when introducing "new transformed parameters" w.(,, and a new matrix q0', that the global
evolution of the structure will be strictly given within our same framework as in section 2.3.
Then, we shall prove that a special procedure of analysis is very efficient when the matrix q~, at the local state
of the structure, is regular. With this procedure, it will just be necessary to make a purely elastic analysis of the
structure in order to obtain bounds of solutions during cycling loadings.
At last, to illustrate this framework we shall give a special example o f application: the viscoplastic three-bar
system.

* As in any thermoelastic problem, when there is a coupling effect between temperature and elastic constants, the problem is non-
linear and must be solved step by step with a new rigidity matrix.
** The purely viscous analysis in 2.4.2. can no more be realized so easily, we reach a nonlinear problem.
*** We shall assume that the geometrical parts OFiV and OuiV are always constant (this implies a constant rigidity matrix for the
elastic analysis).
 J. Zarka et al. / On a simplified inelastic analysis of structures 355

3.2. Global evolution of the structure

3.2.1. Purely elastic response of the structure

Had the structure been made of a purely elastic material with the same elastic constants matrix wc~ as the
real one, we should have obtained the fields:
(i) woel(r) which is Statically Admissible (S.A.) with wXd(r) in V and F~i(r ) on OFiV, i.e.,
ao~(r)laxj +ox~i(r) = o , in V,
o~](r) n/=F~/(r), on aFiV. (93)
(ii) weel(r) which is associated to woeS(r) by the elastic constitutive laws and which is Kinematically Admissible
(K.A.) with Wu~(r) on OujV, i.e.,
weel(r) = wc'/~woel(r) + we°(r) , (94)

where e°(r) is the dilatation strains field associated to 0(r), and there exists one displacement field Wue~(r) such
that
u~l(r)= Wu)l(r) , on auiV,
and
2e~ = au~l/Ox] + au~l/axi . (95)

3.2.2. Global evolution of the structure

The actual response is, in fact, given as in section 2.3.; we shall then have for the global total strain t~eld within
the structure:
We(r) = wee(,/-)+ e°(r) + eP(r), with wee(,/-)= wc~wo(r), (96)

where Wo(r) is the actual global stress field.

By putting
We(r) = woel(r) + Wp(r) ' (97)

is is well known that the residual stress field Wp(r) for the structure is such that

Wp(r) is S.A. with O in V and 0 on aFiV, I (98)

wQ/tWp(r) + weP(r) is K.A. with 0 on aujV. )
This means that Wp(r) is associated to the global plastic strain field WeP(r) by
Wp(r) = BoWep(r), (99)
where the linear operator Bo is indeed a matrix when the structure is discrete or discretized (as we shall assume
it here) and corresponds to the solution of an elastic problem with wc~ffas the elastic constants matrices field,
wep(r) as initial strains field and the homogeneous boundary conditions, O in V and 0 on ~FiV and 0 on auiV.
Since eP(x, r) is given by eq. (47), we deduce
Wo(r) = BoWA T W X ( r ) " (100)
When introducing eq. (97) in eq. (30) we f'md locaUy, in every point in the structure
Ilo(l)
Io(2)] = IAoel(r) _ (1~-_ lAp(,/,)) , (101)
/
Io(3) j
356 J. Zarka et al. / On a simplified inelastic analysis o f structures

which implies, always with the same principle of our framework, that for the structure a New Natural transformed
parameters family has to be introduced *:
wf(,= wf( _ WAWp. (102)

From eqs. (31) and (100), we can write down

w x ' = qO'wx, (103)

where qS' is a new nonnegative symmetrical matrix, but with a much bigger dimension than qS, which is linked to
the structure.
The global evolution of the structure will then be given by equations similar to those introduced in the con-
stitutive relations. The threshold conditions for subelements 1 and 3, will still be treated easily since Woet(r) is an
a priori known solution of the elastic problem and the new convex sets ~_'(woel), ~'(Woel) images of eqs. (35),
(46) are locally built from:
~'(woel) = ~'(0) + wA(l)wo°l,
c~'(woel ) = c~'(O) + WA(Owoel . (104)

During cyclic loadings, the results given in section 2.4. are identically valid. The most important difficulty will
only be in the regularity or singularity of cB'.
As in [24], we shall now prove that if the small matrix qfl is regular, the big matrix qfl' will also be regular and
we shall only need an elastic analysis to obtain qfl,-l.

3.3. Particular procedure for a regular local matrix°B

Since the local q0 is assumed regular and since its dimension is rather small, we can directly invert it and have
qfl-~. This operation has to be eventually made in any volume element of a non homogeneous structure.
We can then write
IEP = I a X l x = I A T ~ - I I ~ " , (105)

and, when using the definition eq. (102), we obtain:

IeP = IATCJfl-11.~" + I A T q f l - I l A I p . (106)

As we have recalled it, in eq. (98), the strain field

wqffWp + Wep = w(c//~ + ATqfl-IAj w0 + W(A TQ3-1 ~, ) (107)

must be Kinematically Admissible with 0 on OuiV, and the residual stress field wp must be Statically Admissible
with 0 in V and 0 on OFiV. From eq. (107) we can see, for a known field w~, and then for a known initial strain
field W(A Tq8-*/?') we have to solve an homogeneous elastic problem with now a new elastic constants matrices
field c//u = cl?~ + W(ATqf-IA)" Symbolically as in eq. (99), we shall write
Wp = B,oW(ATo,B-, X' ) . (108)
As I x = q8 -l 1~ and w.~- = w)?, + WAWp, we can then deduce with eq. (108),
wX = qfl,-i w.~., . (109)
This proves in fact since we build for any w~,, wx, that a sufficient condition to have a regular matrix q8' is
that the local matrix qfl is regular.

*"In [24] we introduced for the first time these new transformed parameters.
 J. Zarka et al. I On a simplified inelastic analysis o f structures 357

iF
rr7 rTi
kt

IzZ /J I1 II1

Sr
Fig. 19. Three-bar system. Fig. 20. Elementary behaviour of each bar.

Moreover, we have shown how, with a classical elastic analysis, it is possible to come back from w~,, to wx.
Since the matrixCB' is now regular, during cyclic loadings, only elastic or plastic shakedown of the structure
can be obtained and direct bounds, as given in section 2.4.6, can be proposed.

3.4. An example of structure: The three-bar system

This classical example is given, once again, in order to illustrate our framework. The simple close form analysis,
which will be made, can be easily extended to the numerical analysis of any complex structure; only an elastic
computer code will be needed.

3.4.1. Description of the structure and of the thermomechanical loading

(i) Its consists of three bars fixed to two rigid supports. Two identical bars (index 1) are symmetrically placed
with respect to the third one (index 2) (fig. 19).
(ii) We apply a constant non negative axial force to the supports and moreover the two bars 1 are subjected to
the change of temperature 01 (r) assigned to have the form
01(I") = AT sin ~ot, (1 I0)
while the bar 2 is kept at a constant temperature.

3.4.2. Constitutive hypothesis

For each bar, we shall assume that its global behaviour is the result of an in serial assembly of one slider-
spring, one dashpot-spring and one spring (fig. 20).
So, we obtain for bar r (r equal to 1 or 2)
(i) Internal parameters

IXr =
~,
(ii) The local stresses and transformed parameters
a(1) --- o r - h , a r = or - ar, /
(111)
0(2) = Or -- k r ~ r = Or - ~r ,

(112)
'X,= ~r = 0 kJ\~,I
where Or is the global applied stress to the bar.
By hypothesis, the qSr are regular, i.e., hrk r > O.
358 J. Zarka et al. / On a simplified inelastic analysis of structures

(iii) The global plastic strain

cP = IA~lSr - ar + ~r. (113)
(iv) The evolution laws
ar ~ -0qJoAor)(&r), (114)
(115)
where ~(Or) is the segment [Or - St, or + St].
(v) The global thermal and elastic strains
e 0 = arO r , (116)
e e = Or/Er, (117)
where ar is the thermal expansion coefficient.

3. 4. 3. Purely elastic analysis o f the structure

(i) The global equilibrium equation requires that the stress field

woo,=(;i:)
will be Statically Admissible with the given load F, if
2S~o~ I + S 2 o 2el = F = S ' l o ~ I + S 2 o 2el. (118)
Here, we shall assume that S'I = $2, and we shall write eq. (118) in the form
O~! + 031 = FIS2 - 2OA. (119)
(ii) The rigid supports impose that the strain field will be KinematicaUy Admissible when:
o~llE1 + a,O, = o~llE2 . (12o)
With eq. (110)we thus find the elastic stress field.
= (O~i(r) ~ = ( 20 AE1/(E' + Ez) - OT sin O°r ]
WOe|(T ) (121)
\o~l(r)/ \2OAE2/(E1 +E2) + oT sin o ~ r / '

where
E1E2
aT - ar A T =Ea r A T , (122)
E l + E2
with
E = E1E=/(EI + E2) • (123)

3.4.4. General response o f the structure

(i) The total strain field in in fact:

(124)
e21 \oJ& + e~
 J. Zarka et al. / On a simplified inelastic analysis of structures 359

This field will be K.A. when

o i / E I + alO1 + ePl = 02/E2 + e~ . (125)
(ii) The global stress field

O2

will be S.A. with F when

o, + 02 = 2OA • (126)
(iii) When introducing the residual stress field Wp = ( ~ ) such that wa = woeJ + wp the conditions (98) become:
wp S.A. with 0: Px +02 = 0 , (127)
and
Ic/?~lp + lep K.A. with 0:PilE1 + e~ = p2/E2 + e~, (128)
Whose evident solution is:
~E E
1( e~ ~ = Bo we p . (129)
wp = ( E -El\e~/

(iv) The new transformed parameters for the structure (102) will be

1)?, = IR1 _ 01
=,. w~-, = - qa'wx = qY
F'
Otl

, (130)
I.¢; = iX2 _ 02

with

E + h, E -E -E 1

IE
-E
-E
E + k,
-E
-E
-E
E + h2
E
E
-E

E + k2 1 (131)

Of course, since the matrixqff is only of dimension 4, its direct inversion can be made; we shall however follow
the special procedure given in section 3.3.
Let us first introduce the linked constants

l / k = 1/kl + 1/k2 , 1/h = l / h , + 1 / h 2 , (132)

a = E/El , b = k/k, , c = h/h,

and

1 / v = 1 / e + 1/k + I/h .
360 J. Zarka et al. / On a simplified inelastic analysis of structures

The eqs. (106) will be written as:

ePl~i [cot ,lh + b~l/k + (e/h + b/k) p, =[&',/h, + ~'llk, + ( l / h i + 1/kl)
RcP / /
~e~/ \(1 - c) u2/h
~' + (1 - b) {32/k+
"' ((1 - c)/h + (1 - b)/k) P2 \&'2/h2 + ~2/k2 + (l/h2 + I/k2) 02
(133)

and the Kinematically Admissible condition eq. (107) will drive to

(l/El + l/hi +1 / k l ) P l + 6t'l/hl +~'l/kl = (l/E2 + 1/k2 + l / h 2 ) p z + 6t'2/h2 + t~'2/k2. (134)
From which we deduce
t

W(A Tc~-lk') = e~' (135)

a'2/h, + Y d k , / '

wo=(pl)=(-G G ](e~ 't] (136)

P2 G -Gl\e(/ '
and at last

wx=tlX2j-- =q,_lw2, '

with
'(1 - G / h i ) / h , -G/hik, G/h,h2 G/h,k2
-Glh dq (1 - GIk,)lk, GIh2k, GIk,k2
(137)
qo'-' = Glhlh2 GIh,k, (1 - GIh )lh, -GIh,k
.GIh,k Gl#,,k, -GIhJ¢ (1 - CI1<,)lk..
It is easy to see that our hypothesis on the regularity o f ~ l and q~ 2 implies the regularity of the q3' matrix.
But it is also easy to verify that even if one of these matrices is singular, for example ~ 1 but hlk2h 2 > 0, the
matrix q3' will still be regular. This shows that the sufficient condition, q~l and q~2 are regular, is not necessary
to ensure a regular qs' matrix.

3.4.5. Direct bounds

(i) "Static" loading: (OA is first applied "very slowly" and then the temperature of bars 1 varies "very slowly").
Since at any time,/7] and/7~ will both stay null in each bar r, we shall just find a material with hr as the kine-
matical hardening modulus and with a new elastic Young's modulus
E'r = Erkr/(Er + kr) . (138)
The "static" elastic stresses induced by the thermomechanical loading will thus be eqs. (121)-(122):

w ,el (°1,el~/
s ( 2trAE1/(EI
' ' +E2)
' - [Elff,/(Ei
' ' + E,)](OT/E)
' sm
"
Os = ,el = , . . . . . . . ~or]/ " (139)
\o2 s ] \ 2trAE2/(//I + E2) + [EIE2/(E1 + E2)](OT(E) sm ~ r
As in [24], we shall find easily the limiting states (elastic or plastic shakedown regimes) (fig. 21).
 J. Zarka et al. I On a simplified inelastic analysis of structures 361

_ . . . . . . .

; IX.X,, ,
..... ' r - - ~~,0~) % ~ . ,. '~'
m,nm.
I •

t ",, !,,,3, z

Fig. 21. Limiting states for increasing stress amplitude partial (1), (2), and total (3) plastic shakedown.

(ii) "Dynamic" loading: (OA is first applied "instantaneously" and then the temperature of bars 1 varies
"very quickly"). This time,/31 and/32 will stay null; we find again a material with h as the kinematical hardening
and E as the Young's modulus.
The "dynamic" elastic stresses induced by the loading will thus be the same as in eq. (121).
wo• = Woel (140)

Always with the same procedure, we shall find the limiting states.

3.4.6. General loadings (OA i$ instantaneously applied and then the temperature o f bars 1 varies according to eq.
(119)).
(i) Purely linear viscoelastic response:
As in section 2.4.2., we first determine the purely viscoelastic response and look at the limiting state w~v(t) for
t going towards ~". We find that

(141 )
wily \/3v2/ \(2o~/k)(1 - b ) ( a +bE/k)~(1 + E l k ) - (OT/k)(1 - b ) f ( t )
with
f(t) = [(wr//k) cos cot - (1 +Elk) sin ¢ot]l[co2r~21k2 + (1 + E l k ) 2 ] , (142)
and
1/~ = 1/~1 + 1/~2 (143)
In order to simplify the analysis, from now we take b = k/k1 = ~?fih.
(ii) General response
As in section 2.4.3., by setting

r°?
wo = w/3v + w/31 and
L/
W ex_~a= w ~ '¢ 1 + ¢'~' ~v 1
0
362 J. Zarka et al. / On a simplified inelastic analysis o f structures

we obtain

(144)
\&2 /

where

w&,l \&;l!

and
'1(o, ~v) is the rectangle centered in (Ov a, av 2) and of sides (2St, 2S2) with (145)
Ovl = 2OA[1 -- a + (1 -- b) E/k]/(1 + Elk) - oTg(t) ,
Ov2 = 2OA(a + bE/k)/(1 + Elk) + omg(t ) , (146)

and
g(t) = [[coirl21k2 + 1 +Elk] sin cot + (EIk)(wrffk) cos cot]l[co2rl21k 2 + (1 + Elk)2], (147)

w~l = ( l ~ l t = ( - ~ ' l l b / ~ ). (148)

~: ~-Y~(1 - b)/~
If we introduce as generalized loading parameters 2oA and OT sin cot, the associated global generalized strains
will be

= B' . (149)
P -1 1
~2
The integration of eqs. (144) and (148) can be numerically made. The condition for "elastic" shakedown is
expressed by:
Do T = [(1 + co2r12/k2)l(co2172/k2 + (1 + E/k) 2 ] v 2 0 T ~< Min(S1, $2) - S , (150)
i~e., the convex sets ~'1(o, 13v)have a non-void intersection. D is a monotonic function of co, increasing from
k/(E + k) to one as co goes from 0 to oo. Whatever co might be, elastic shakedown will be impossible for
kOT/(E + k) > S or will always occur for OT < S.
In figs. 22, 23, 24 we give some examples of cyclic toadings on the structure.
We can observe that the general response of the structure is (of course) a function of the frequency of the
loading; for given (OA, OT) it could vary from "elastic" shakedown to "plastic" shakedown following co.
The case of a perfectly plastic material can be practically obtained for very "small" values of h ~ and h2;
ratcheting may then be get, as in fig. 25.

4. Conclusion

We have described our general framework to solve (in the same time) the problem of global constitutive rela-
tions of materials and the problem of global evolution of structures where hardening, softening and creep can
occur.
 J. Zarka et al. / On a simplified inelastic analysis o f structures 363

PARAHETERS PARAPtI[TERS A
ol(t) t ~h (t)
BAR I BAR 1
El 6~0 El 8800
E'TAt 200 I~"TAI 200
kl 1200 k~ 1209
Sl t2 01 12
~2 (t) hi 090
OAR 2 BAR 2
E2 t ~dlOO E2 I L:~e
9; ~ ,... . . ~
S2 10
h2 1000

E.~EA 4 0 .
73.8S
i~i( t )
h2

E.~E
A
1"e

~ 00
73.80
i ++
k 442. t k +42.1
LOAC,'rNS LOAt)ZNG
SZ~ A S SZG A e
8Z9 T SZ~ T 1~
OHEOA OMEGA 028
GCAL~S SCAL.£$
T.T,HE, TZH(,
ONE PERZOD ONE PERZOD
S'tR(.E8, STRESSES,
(;E l * E 2 ) / 2 ~
$TRAZNS~ STRAINS,
s/+mm 5/1000

oT s i n a i GTsin~*t

2Y,
- Et. ~"

z'/.
"( y . . . . . . 1"],. . . . . . . ,,P ,.P

Fig. 22. Viscoplastic behaviour of the structure. Elastic Fig. 23. Viscoplastic behaviour of the structure. Plastic
shakedown. shakedown.

pXd~IE~KRS

OAR I
El 0 ~ 0
I[TAI 21~
kl 1200
$1 12
hi I
BAR 2
[2 t2008
E'rA2 t 10.00
kZ 700
02 10
hz t ~I(')
ETEA 460B
73.66
k 442. I
LOADINS
sis A 0
SZG T t s
OHEGA 39
S~,LES
TIHF.,:
ONE P[R.TOD
S'tRESSF-S'~
CE t +E2~/2000
$TR&ZNS'
S/t.e

oTlin~t

Fig. 24. Viscoplastic behaviour of the structure. Ratcheting.

364 J. Zarka et al. / On a s#nplified inelastic attal),sis of structures

Its essential feature consisted in the introduction of a family of internal parameters IX (local inelastic strains)
which characterizes the local inelastic mechanisms and a family of transformed internal parameters iX (- asso-
ciated local residual stresses) which are linearly linked to the previous ones through a symmetrical non negative
matrixq0.
Thanks to these l~,, the treatment of the plastic or viscoplastic thresholds (local yield conditions) can be
easily made; the position of the field JX can in fact be known from only the simple purely elastic or linear viscous
analysis of the system(i.e., fundamental volume element or structure). The uses of this important property were
shown mainly during cyclic loadings: condition for shakedown, bounds for the limiting s t a t e . . . We have also
proved that, if the local cB matrix in the constitutive equations is regular, we find again a regular global q~' matrix
for the structure and, we need only a classical elastic analysis to build the inverse ~ , - 1 , this allows us to go easily
from one field to the other one and vice versa.
In order to illustrate this framework, we have also given several examples of global behaviours and we have
described the application of the proposed systematic treatment of structures on the particular three bar system.
However, we must say that there are other experimental facts which cannot be represented easily within this
framework:
(a) an initial deformation of some materials may influence their cyclic curves; this could be the result of the
modification of the texture but also of some changes of phases or twinnings.
(b) Damage is occuring within the material during cyclic loadings.
For that, it would have been necessary to assume non constant IA, q0, c ~ matrices which would thus imply
again a classical step by step analysis.
We must also say that, until now, we are only in position to qualitatively describe the behaviours of materials;
in order to achieve a quantitative representation of one special material it would be useful to dispose of a syste-
matic analysis of constants identification based on experimental data.
Nevertheless, if we could find such an agreement, the analysis of any structure is afterwards elementary. That
is why we are absolutely convinced that it is always better, even if we must neglect some aspects of the real beha-
viour, to enter into our proposed framework.
At last, we must say that in general, we are not yet able to explicit the number of cycles of loading which are
necessary to reach the limiting state. This serious limits the applications in designing of the framework, since we
may get fatigue rupture before reaching the stationary state.
All these problems will be the objects of following papers.

Appendix

Practical bounds f o r an elastic plastic structure during radial cyclic loadings *

A. 1. H y p o t h e s i s

(i) We shall assume here that there are only subelements 1 within the fundamental volume element.
(ii) Its B (1) matrix is regular, and that
(iii) The cyclic loading is radial (see section 2.4.6).
We know, there will always be elastic or plastic shakedown response.
It is obvious that the treatment of the structure with the much bigger matrix q~' will be strictly similar.

* This is a short transcription of J. Zarka and J. Casier's text with the definitions of that present paper.
 J. Zarka et al. / On a simplified inelastic analysis o f structures 365

A.2. Elastic shakedown behaviour

A.2.1. Condition
In the tS space, the convex set Q(t~) is subjected to a simple radial translation between the two limiting positions
~(Omirt) and ~(Omax).
The condition for elastic shakedown is that the convex set Qp = e(Ornin) ~ C(Omax) is a non void set. As it
was previously shown, since ~(o) is locally built, the elastic shakedown condition is a local one; we need to take
each subelement 1, say I, and we need to see if all the local convex sets 6"(Omin,/) and C'(Omax,/) have a non void
intersection (fig. A.1 .).
All the limiting states, which correspond to different initial states of the volume element, belong to el.
(i) If we do not know this initial state any such ~ field tS E ~l is possible. Since e] is locally built, we can gene-
rate it very easily and since B (1) is regular, we can come back to the fields let = B(1)-]tS and thus we can find the
bounds of limiting plastic strains.
(ii) If the initial state ]So is known, better bounds of the associated limiting state can be produced.

A.2.2. Practical procedure for a known initial state

Very often, elastic shakedown behaviour can be reached after the first cycle of loading; then, one should make
the inelastic analysis of the volume element during the first cycle and look if this is the case.
If this is not the case, an infinite number of cycles would be generally necessary.
(i) Let us call Is(l) the state which is reached after the cycle: (o: Omin )~ O'max ~ Omin) ; Is(l) belongs thus to
~ (Omin).
(ii) Take the normal projection tS l, of iS(l) on t~l; this will also be possible by a local building (fig. A.2.). 16q
will be a bound for the limiting state (indeed this is only a lower bound in the IS space with the classical norm or
a lower bound in the ic~ space with the new norm [B ( 1)- 1] T B (1)- 1; very often it will however be a good approxi-
marion of the limiting state).

:
~'a.i.
^ ;: | .......
~,10
~
.... I
~(~.i..) ~0.x) ~(°...,)
C (Ornin.} C (°max.) t I ~. :

^ ")
ol(1.~ a,.';

~'lt J")

i (1) L-
A crma~..~ ~

/ l
Fig. A.1. Condition for elastic shakedown. Fig. A.2. Projection of l&(1) on ~1.
366 J. Zarka et al. / On a simplified inelastic analysis o f stnwtures

C(Crm,,)
X^ !
C fo-,~.~,)

. I C max

eCo-.,o.)
:
:~"c(o-,,,,~.)

~/ / / ¢

Fig, A.3. "Lower" bounds during plastic shakedown.

A.3. Plastic shakedown behaviour *

A.3.1. L o w e r b o u n d
If we take a cycle whose extremities correspond to states of minimum distance between the disjoint convex
sets, ~(Omin) and ~(Omax), we shall have a "lower" bound in the J& space.
We consider in turn subelements I, for whom C(omin,/) O ~'(Omax, D = • we define l&max(/) and l&min(/) by
the local building (fig. A.3.) and we thus deduce l(A&(L)(/)).
For all other points we take I(At~(L) -------0.
We come back to the la space, to find I(Ao~(L)).

A.3.2. "Upper'" b o u n d
If the amplitude of the radial loading is "very" large during the first half of the cycle, the monotonic solution
will correspond to IA (1)0 = ldt for all subelements.
This will imply I& = B(1)-llA(1)O"
We take then in turn all the subelements (even those for which ~'~(Omin) ('1 C~Omax) # •) and locally define

* It is possible to treat analytical other thresholds and to have other important properties [24].
 J. Zarka et al. / On a simplified inelastic analysis of structures 367

l&max(/) and I&min(/) such that

-i~'qJ~(amax)(l&max(/)) = n ( l ) - l ( / , J) Ia(1)(J) O,

and
_ ~ t ~ , ( a m i n , l)(l~min(/)) = _ B ( 1 ) - I ( / , j ) IA(1)(j) O,

(l&max and I~mi n will be undetermined for subelements for which l~(/) = B ( l ) - i ( i , j ) IA(O(j ) 0 = 0).
So we obtain I(A&(U) and thus I(Aa(u)).

A.3.3. Examples of one-dimensional thresholds

(i) "Lower" bound:
For subelements r for which IA(rl)(omax - Omin)l > 2St (only possible when A~ l) # 0) we take:
A&~L) = A~l)(omax - Omin) - 2 S ~ A r / I A r l .

For all other subelements we take A6~ L) = 0.

(ii) "Upper" bound:
l& = +B( l ) - n A ( 1 ) 0 .

For subelements r for which It~rl # 0,

A&(o) = A(l)(omax - Omin) - 2Srbtr/[d r [ , we take

(it may be possible even for some Ar(1) = 0).

For all other subelements for which I&rl = O, we take A&(U) = 0

Acknowledgement

We are thankfull to E.D.F.-SEPTEN for supporting this work.

References

[1] J.F. Besseling, Nat. Aera Res. Inst. Report $410 (1953).
[2] H.D. Bui, Evolution de la fronti~re du domaine ~lastique des m~taux avec ~crouissage plastique et comportement ~lastoplas-
tique d'un agr6gat de cristaux cubiques, Thesis, Paris, (1969).
[3] Y.F. Dafalias and E.P. Popov, J. Appl. Mech. 43 (1976) 645.
[4] O. Desbordes, Adaption des structures 61astiques parfaitement plastiques, S6minaire Mat6riaux et Structures sous Charge-
ments Cycliques, Palaiseau, Ecole Polytechnique, 1978.
[5] B. Halphen, Probl~mes quasistatiques en viscoplasticit6, Thesis, Paris (1978).
[6] R. Hill, J. Mech. Phys. Solids, 15 (1967) 79.
[7] W.T. Koiter, General theorems for elastic-plastic solids, Progress in Solid Mechanics, vol. 1 (North-Holland, Amsterdam,
1960).
[8] J.A. KiSnig, Bul. Acad. Sci. Theor. Appl. Mech. III (2) (1972).
[9] F.A. Leckie, WRC, Bulletin 195 (1974).
[10] J. Mandel, Contribution th~orique ~ l'~tude de l'~crouissage et des lois d'~coulement plastigue, Proc. 11 th International
Congress on Applied Mechanics, Munich, 1966.
[ 11 ] J. Mandel, Equations de comportement d'un syst~me 61astoviscoplastique dont l'6crouissage est dQ h des contraintes r~si-
duelles, C.R. Ac. Sci., t. 284, Paris, 1977.
[12] J. Mandel, J. Zarka and B. Halphen, Mech. Res. Comm. 4 (1977) 309.
[13] J.J. Moreau, Rafle par un convexe variable, S~minaire d'Analogie Convexe, Montpellier, 1972).
[14] G. Maier, Mec. 4 (3) (1969) 250-260.
368 J. Zarka et al. / On a simplified inelastic analysis o f structures

[15 ] Z. Mroz, A description of workhardening of metals with application to variable loading, in: Foundations of PLasticity, Ed.
A. Sawizuk (1972).
[16] Z. Mroz, On the theory of steady plastic cycles in structures, Proc. SMiRT-1, 6, part L, (1972) 489-501.
[ 17 ] Z. Mroz, On the description of plastic hardening of materials with application to variable and cyclic loading, Proc. Syrup.
Franco-Polonais, PWN 1976.
[18] Z. Mroz and J. Zarka, Relations de comportement des m~taux so us chargements cycliques, S6minaire Mat6riaux et Struc-
tures sous Chargements Cycliques, Palaiseau, Ecole Polytechnique, 1978.
[ 19 ] B. Nayrolles, Structures alge3briques des theories classiques de plasticitY, dans: S~minaire Plasticit~ et Viscoplasticit~, Eds.:
D. Radenkovic et J. Salenqon (McGraw-Hill Ediscience, 1972).
[20] Q.S. Nguyen, Contribution ~ la th~orie macroscopique de l'~lastoplasticit~ avec ~crouissage, Thesis, Paris 1973.
[21] J.R. Rice, J. Appl. Mech. 37 (1970) 728.
[22] O.C. Zienkiewicz, G.C. Nayac and D.R.I. Owen, Composite and overlay models in numerical analysis of elastoplastic conti-
nua, in: Foundations of Plasticity, A. Sawczuk, Ed. (Noordhoff, Groningcn, 1973).
[23] J. Zarka, J. Appl. Mech. 3 (3) (1979) 291-326.
[24] J. Zarka and J. Casier, Cyclic loading on an elastoplastic structure, Practical rules, Mechanics today, vol. 6, Ed.: J. Nemat-
Nasser (1979).