Ghostscript wrapper for D:\Digitalizacja\MTS88_t26_z1_4_PDF_artykuly\02mts88_t26_zeszyt_2.pdf M E C H AN I K A TE OR E TYC Z N A I ST O SO WAN A 2, 26 (1988) CON STRAIN TS IN SOBLD M ECH AN ICS. AN APPLICATION OF N ON STAN DARD AN ALYSIS EWARYST WlE R Z BI C KI Instytut Mechaniki UW 1. Introduction The concept of constraints in solid mechanics is usually utilized to formulate special cases of constitutive relations identifying with certain restriction imposed on pairs {%, T ) of a motion % and a stress T . To be a constitutive relation such restriction must have a spe- cial form, i.e. it must fulfil certain neccesary conditions stated in the general theory of constitutive relations. The following N oll axiom is exactly one of these conditions, [9, p. 160]. The principle of determinism for simple materials. The stress at the place occupied by the body- point % at the time / is determined by the history %' of the motion of the body up to the time t, i.e. T ( x (x,t) !t ) = &(v x Kx, • );'*)• Here ś F( • ) denotes a sufficient regular mapping of histories V%' of a gradient V% of a mo- tion %, body - points X onto symmetric Cauchy stress tensors. The above principle of determinism will be called here a classical principle of determi- nism. However, there exist real materials for which forementioned defined principle leads to the theory not consistent with experiment. In such situations more general or alternative formulations of the principle of determinism should be applied. F or example, if admissible motions of a body are subjected to constraints of the form: w(Z JV^ rV%(Z , / )) = 0, (1.1) where u>( • ) is sufficient regular function with values in R" then the following statement holds, [9 p. 176]. Principle of determinism for simple materials subject t o constraints. The stress at the place occupied by the body- point X at the time t is determined by the history %* of the motion % up to the time t only to within an arbitrary tensor that does no work in any motion compatible with the constraints. That is: T( X(X, 0 , 0 = 11 M ech . T eorct. i Stos. 2/ 88 378 E. WIERZBICKI where the mapping G() need be sufficiently regular and defined only for arguments V / such as to satisfy the constraints, N being a stress for which the stresspower vanishes in any motion satisfying the constraints, i.e. tr(ND) = 0 for each symmetric tensor D such that: The principle of determinism for simple materials with constraints is a generalisations of the classical principle. If there are no restrictions of the form (1.1), i.e. w() = const., then N = 0 and both principles coincide. The principle of determinism can be formulated in mechanics also in more general form, describing more general classes of physical situations. For example in [14] it can be found the following formulation of constraints: wOt.Vz, ...,v'a) = o and in [1] we deal with constraints: where p, q are natural numbers and numbers in scopes under the letters denote suitable timederivative. In the paper we apply nonstandard analysis as a mathematical tool derive new consti tutive relations of mechanics from the known constitutive relations. Fundamental concepts of this approach are based on [7, 11, 12]. The aim of the paper is to prove that by applying concepts and methods of nonstandard analysis the principle of determinism for consti tutiverelations with constraints can be obtained from the classical principle of determi nism. This proof will be realized by a certain specification of nonstandard constitutive relations which are consistent with the classical principle of determinism. We are to show that this approach has the following attributes: (i) it eliminates from the axioms of mechanics the principle of determinism with constraints, (ii) it has a clear physical interpretation being based only on the classical principle of determinism, (iii) it leads to a description of physical situations which cannot be described neither by the classical principle of determinism nor by the principle of determinism for constitu tive relations with constraints. In the paper the concept of a constitutive relation is understood in more general sense than that in most of the papers on this subject. Namely after a certain specification the constitutive relations will be treated as constitutive relations for the internal forces descri bing material properties of bodies — or as constitutive relations for the external forces describing interactions between a body and its exterior, cf. [11, 15]. CONSTRAINTS IN MECHANICS 379 2. Physical foundations Let be given the following objects: (i) the set F of states y of the mechanical system under consideration, i.e. assume that F is an open set in a certain topological space, (ii) the set 0* of admissible evolutions R B t »• y(t) E F of states of the mechanical system under consideration, i.e. the set of righthand side differentiable functions of real variable; assume that this differentiation is well defined, (iii) the dual pairing (W,{, •>, W) of linear topological spaces in which W is the space of time rates y, W is the space of reactions Q of the system and (y, Q) is the power of the reaction Q for the rate y, y e W, n e W, (iv) the set Ht of histories y (() : Rt > F of the evolution['y( •) s 0 1 of the system states up to the time t, defined for each t e R, i.e. for t sR,seR,s ^ 0. As a basis for our considerations the following requirement will be postulated. Principle of determinism. For each timeinstant t e R a reaction q{t) of the system is uniquely determined by the history y(t) e H, of the evolution y( •) e 0* up to the time t and by the rate y(t) of change of a system state in the time t, i.e, e(0 = 0>y(0,y('))- (2.1) Introducing above and applying below concepts such as the state of the system, a reac tion of the system, the rate of changing of a state of the system, etc., can have a different physical interpretation, which can be found in [11]. In a description of a mechanical system the concept of constraints is used in situations where it is impossible to receive so many informations to be sufficient to describe it by a constitutive relation satisfying the classical principle of determinism. Accepting here as a fundamental requirement the classical principle of determinism has then a superior authority with respect to other ones. The approach using in the paper is in agree with above premises because the concept of constraints is here a natural consequence of the classical principle of determinism. 3. Tools from nonstandard analysis Let 9Ł be a nonempty set. From all sequences of points of SC we shall distinguish the set C, C cz !%N, elements of which will be called the converging sequences. For each con verging sequence (x,,)neN we assing exactly one point limx„ e $Ł which will be called a limit °f (,x„)„sN. We will also say that each sequence (x„)„lN e C converges to the limit limxn. We assume that the operation lim: C > SC fulfils the following conditions: (i) each subsequence of a sequence converging to x, x e f . i s a sequence converging to x, (ii) the constant sequence with values equal to x, x e SE, converges to x, (iii) each sequence not converging to x, x e SŁ, contains a subsequence whicn in turn does not contain any subsequence converging to x. u* 380 E . WlERZBICKI cThen the pair (#", lim) will be referred to as 1/space, cf. [6 p. 339]. Let CA, A t stands for the set of all converging sequences with values in A and let P(3T) be the power set of X. Define two sequences d„: P{%)> P{SE), int„: P(SQ »• P ( f ) , of operations, setting: (xXeN e CA}, cln+l(A) = cUcl,,^), nei\T, and: int,1+1(^() = inttint,,^), neN, for every A eP(2F). It is easy to verify that each pair ($Ł, cl„) n eN, is a stepspace, cf. [3], i.e. for each « e JV the operation cln fulfils all conditions defining a closure operation in a topological space (possible except the requirement that cl;; must be equal to cl,,). It is easy to introduce a topological structure in each i'space by defining the closed sets as the sets D containing limits converging sequences of points belonging to the set D, cf. [2 p . 90]. This topology will be denoted by r. If the operation lim fulfils the additional condition (iv) if limx„ = x and limxj! = xn, neN, then there '[exist sequences (rii)ieN, (ki)ieII of natural numbers for which limxj!{ = x, then cl„ = cl,„ for each pair (n, m) eN2 and cl = cl„, neN, is then a closure operation in topological space (5Ł, r, cf. [2, p. 90]. Similarly to such topological concepts as: the monad, the standard part operation, the Flimit operation, we are going to define, for any neN and for any Z/space, new con cepts of Mmonad, nstandard part operation and F limit operation. To this aid let the pair (SŁ, lim) be a JJspace and let 9C, lim be objects in a certain full structure 501. Let *9Jl be an enlargement of 9JI. We have *3f 6 *2ft and lim e *9K (here and below we write lim instead of *lim). The pair (*ŁŁ, lim) is considered here as a QL'space. For x eSŁ and neN define : ^ ) s n {*A: A e P(3Ł), xeA= intnA}. (3.1) Denoting by f.iT(x) the monad of x in the topological space (SC, x) it is easy to verify that the following inclusions: Monx(x) 3 Mon2(x) a Mon3(*) =3 ..., as well as the equality: H Mon„(jc) = fix(x) iitsN hold. The i'space {SC, lim) will be called «Hausdorff, neN, if JC = y is implied by Mon„(x) = Mon„(y). It is easy to see that if L'space (SC, lim) is HHausdorff, for a certain neN, then the topological space (#\ T) is a Hausdorff space. Now let i'space be nHausdorff for a certain neN. Then in every nmonad Monn(x), x e a:, there is exactly one standard point. For each pair (x, y) e 2Ł x *2Ł we shall write s t n j = x if y sMon„(x). The aforemention operation st„: *% >• % will be considered as the nstandard part operation. The domain of st„ is equal to (J{Mon„(x): x e SŁ). A sequence (x„) „e*N of points of *SC will be called ^converging if there exist a point C O N ST R AI N T S I N M E C H AN I C S 381 I , X E I , and a hypernatural number X o e *N \ N , such that the relation x, e M on„ (*) holds for every v e *N \ N , v < X o . Points from M on n(x) will be considered as Ą - Iimits o f (x„)„e*N - The concepts of n- H ausdorff L'- space, «- monad. «- standard part operation, Fn- limii, operation will be used below only in the case of n = 1. In the sequel instead of a 1- monad a 1- Hausdorff space, etc., we shall use the terms: a monad, a H ausdorff X'- space, etc., respectively. N ow let T stands for a fixed topological regular space and 2 r be the set of all closed subsets of T . Let define a convergence in 2T setting (A„)„ eN e C iff for some A e 2T the following statements holds: (i)lim sup^ n = A, i.e. each neibourhood of any point from A has a nonempty inter- sections with almost every set A,„ n eN , (ii) liminL4„ = A, i.e. each neibourhood of any point from A has a non- empty inter- section with infinite number of sets A n , n eN . The set 2r with the convergence of sequences of sets defining above, determines a certain L'- space, [6 p. 188], which will be denoted here by (2T, lim). An important result, [10], is that this Z/ - space is H ausdorff (i.e. 1- Hausdorff) an d: MonOO = {Be *(2T) :°B = A}, Ae 2T, (3.2) where °B stands for the standard part of the set B. It means that the standard part opera- tion in / / - space ( 2r , lim) is equal to the standard part operation of (closed) subsets of T . Moreover, .F- lim^,, = Mon(^4) provided that: (3A0 6 *N \ N )(Vn e *N \ N )[[n < Ao] => [A = °An]], for each f- converging sequence (A n ) nB * N of closed subsets A,,e*(2T). 4. From microconstitutive relations to macroconstitutive relations. N ow we are going to formulate the method which enable us to obtain new constitutive relations from the known constitutive relations. The known constitutive relations are here relations satisfying the following form of the classical principle of determinism (2.1): where function ^ , :fx W x H t -> W , for every t e R, is defined by
) is a starting point of our considerations. I n the sequel arguments t and y(y, yU) e H t , t e R, will be treated as parameters; for the sake of simpli- city they will be omitted. So (D) has a form: Q = q>(y,y)\ f:Fx W - * W . (4- 1) Let us assume that the set U(y) s dom ^(y, • }, for every y eT , is open in W . I n a parti- cular case Eq. (4.1) reduces to g = q>(y). Let 0 be a set of functions ć piF- tW - »• W which are assumed to describe physical situations defined by (D). H ence we conclude that the set 0 depends on parameters t and / °(• )• In agreement with physical premises, 0 is an infinite set. Every function 382 E , WlERZBICKI
(y,y); (p:*Fx*W - **W ',
• ) = *U(y). Every function cp,qs *&, is an internal relation but not necessary stan-
dard.
Let us assume th at the set F is a topological H ausdorff space satisfying the first axiom
of countability. F or every state y, y e T , we denote by {o
n
(y))
nsN
the neibourhood- basis
of y in F. I n the space 2W I of all closed subsets of W we shall introduce a L'- space struc-
ture setting T : = W in i'- space (2T, lim). I t is possible to introduce such structure by
means of considerations of Sec. 3, provided that W is regular. Let us define sequences
{0t\ {y, w))
neN
Getting
# ? ( y, «0 • {g = KY> W ) - (y> w) e ). It is a macroidealisation of physical situation described
by microconstitutive relation given by (4.2). It is important that Mn{y, w) is a closed set
in W but not necessary bounded. Microconstitutive relations #j and ę
2
will be considered
as nondiscernible if they generate the same macroconstitutive relation. Equality of
C O N ST R AI N T S I N M E C H AN I C S 383
classes 3i((pi) and 3r(c?2) is equivalent to nondiscernibless of microconstitutive relations
c>! and cp
2
. Introducing parameters t an d ym it can be form ulate the following proposition .
P roposition. F or every microconstitutive relation f, cp e IF Q, there exists in $Jl a m acro-
constitutive relation M^ {ip){yw, • ) generated by q>, i.e. there exists in SIR a set of reactions,
closed in W ", uniquely determined by y(t), y(t) and y{t\ A relation @?&(yV\ • ) n ot
depend on a choice of a microconstitutive relation from the class n((p), i.e. it is the same
for each pair of microconstitutive relations. So (4.2) implies:
e
(t) e m® (V
v
\ y(t), HO)- (G D )
Above proposition will be considered as the general principle of determinism an d the
family of multifunctions:
Fsy - + Af'Xy) m {w e W : (y, w) e dom@f"(y">, • )}, (4.7)
where:
dom # ?&> (yt o, • ) • {(y, w)eFx W : ®f<<\ y«\ y, w)^ 0},
will be formed constraints. We will describe below physical situations for which constraints
(4.7) do not depend on the history / °, y' 0 e H
t
. So, we shall also define A
t
(y) m Af
l
\ y).
F rom now on and from Eq. (4.4) we conclude that evolutions y(- ) e&>, satisfying for
every / 6 R the condition y{t) e A,(y(t), exist. So, for each t e R and y e F the set A
t
(y)
is the set of all rates y of state y at the time t. H owever, macroconstitutive relations as well
as constitutive relations from the set 0 not necessary have physical sense.
5. From the general principle of determinism to the principle of determinism for
constitutive relations with constraints
The formalism presented in Sec. 4 leads from microconstitutive relations satisfying
the classical principle of determinism (D) to the macroconstitutive relations satisfying
the general principle of determinism (G D ). The idea of such passage is in splitting the
set *? 0 of microconstitutive relations into disjointed classes. To every class is assigned th e
value of the operation ę - » fl^fi)(• ) on an arbitrary element tp of this class. This 'mapping
is one to one and t h e operation mentioned above is additive if at least one from the com-
ponents is standard, i.e.:
^*&+ *v)(.) _ gt"^ \ • )+ ^3I(ł *!)( O-
The operation 0 - * *& together with the choice of the set 0 of constitutive relations leads
to the set *0. The choice of the operation 0 • * *0 seems to be natural, because 0 an d *0
represent the same physical object in different structures 501 and *9Jl respectively. The
restriction of the considerations to the set W
Q
, which is the domain of th e operation
0t"l^ { • ), has a character of a regularisation assumtion and is made only for securing
mathematical correctness of th e proposed approach. N ow the question arise: Wh at con-
stitutive relations already known in mechanics can be obtained on that way from a certain
microconstitutive relation q>, W is a function for
which every element of the family {dom ip(y, • ): y e F} is open. I t is n ot so easy to obtain.
a result related t o the question for more wide class of constitutive relations. I n Sec. 6 we
shall obtain results for certain special cases of constitutive relations, namely we shall
found solutions to the following problem :
P roblem . Let t be a fixed time instant, / e R, and let be given:
(i) constraints Fay - > A
s
(y) <= W ,seR,
(ii) the family of functions ip
s
:Fy.W x.H
s
- * W ,seR, sufficiently regular and that
for every s e R an d for every pair (y, y( s> ) eFxH
s
inclusion:
A
s
(y) c doxny>
s
(y, • , / s ) )
holds. We are t o find a microconstitutive relation which generate th e macroconstitutive
relation :
fl(0 Bft(y(t), y(t), y
w
)+N *
Mm
(y(t)). (5.1)
I n Eq. (5.1) N
AtW ))
(y{t)) is a cone, normal to the set A
t
(y(t)) in a poin t y(t) e A
t
(y(t))
}
defined as follows. Let A c W an d w e W . F irst we define a cone tangent to A at a point
H>, setting, [8]:
A3W ±W
no
where lim inf is taken in the H ausdorff sense [4, p. 147].
A cone n orm al to A is the set defined by:
N
A
(w) m {
Q
eW : • R denote the known Gateaux differentiable function, [At] is assumed to
be a nonempty convex closed set. This result is equivalent to the following principle
of determinism:
The principle of determinism for potential constitutive relations with holonomic con
straints in Hilbert spaces. The reaction Q{t) of the system at the time t is determined by
a state y(t) of the system with an accuracy to an additive term Q having nonnegative power
(y> 6) 5= 0 on every rate y,y sW, admissible by constraints, i.e. on every rate belonging
to the set Tyt](y(t)).
As before in the forementioned principle of determinism inequality