JOURNAL OF THEORETICAL AND APPLIED MECHANICS 42, 2, pp. 381-397, Warsaw 2004 STABILITY OF COMPOSITE PLATES WITH NON-UNIFORM DISTRIBUTION OF CONSTITUENTS1 Bohdan Michalak Department of Structural Mechanics, Łódź University of Technology e-mail: bmichala@p.lodz.pl This contribution deals with stability of certain composite plates with a deterministic material structure which is not periodic but can be ap- proximately regarded as periodic in small regions of a plate. The for- mulation of an approximate mathematical model of these plates, based on a tolerance averagingmethod, was discussed inWoźniak andWierz- bicki (2000), where the plates under consideration were referred to as heteroperiodic. Key words: plate, modelling, non-periodic structure, stability 1. Introduction Themain objects of considerations in thepaper are thin composite annular plates made of two families of elastic beams with axes intersecting under the right angle. A homogeneous elastic matrix fulfils regions situated between the beams (Fig.1). Buckling of annular homogeneous plates was investigated, for example, by Waszczyszyn (1976). Eigenvalues of circular plates resting on elastic founda- tions were determined by Gomuliński (1967). Woźniak and Zieliński (1967) investigated some stability problems of circular perforated plates. The aimof this contribution is to propose andapply amathematicalmodel of heteroperiodic plates. In order to apply the general modelling procedure given in Woźniak, Wierzbicki (2000) we have to solve a whole family of the periodic variational cell problems, where every such problem is related to a 1 The researchwaspresentedontheXthSymposium”Stability ofStructures” inZakopane, September 8-12, 2003. 382 B.Michalak Fig. 1. A scheme of the analysed plate small region in which the plate, with a sufficient tolerance, can be treated as periodic. In this contribution, a certain approximate solution to the periodic cell problems for the composite plates under consideration are proposed. The- se solutions are based on some heuristic assumptions and lead to a system of equations with functional but slowly-varying coefficients for the averaged displacement vector field. The derived equations are dependent on the mi- crostructure size in contrast to the equations obtained by the method of no- nuniform homogenization, Bensoussan et al. (1978). Following Woźniak and Wierzbicki (2000) we can observe that the mathematical modelling of media which are periodic and related to a certain curvilinear coordinate system, see Lewiński and Telega (2000), is not able to describe composite plates under consideration with a constant cross section of the beams. 2. Preliminaries Introduce a polar coordinate system in a physical space denoted by Oξ1ξ2ξ3. Throughout the paper the indices α,β,... run over 1, 2 and a verti- cal line before the subscripts stands for the covariant derivative in the polar Stability of composite plates with... 383 coordinate system. The summation convention holds for all aforementioned indices. Setting x ≡ (ξ1,ξ2) and z ≡ ξ3 it is assumed that the undeformed plate occupies the region Ω ≡ {(x,z) : −h/2 ¬ z ¬ h/2, x ∈ Π}, where Π is the plate midplane and h is the plate thickness. The orthogonal Carte- sian coordinate system Oy1y2, with the vector basis eα (α=1,2), is a local coordinate system in an arbitrary cell ∆(x) (Fig.2). Fig. 2. An arbitrary cell ∆(x) of the plate The considerations are based on the well-known second order non-linear theory for thin plates (Woźniak et al., 2001): — strain-displacement relations εαβ =u(α|β) καβ =−w|αβ (2.1) — constitutive equations nαβ =DHαβγδεγδ m αβ =BHαβγδκγδ (2.2) where Hαβγδ = 1 2 [gαδgβγ +gαγgβδ +ν(�αγ�βδ + �αδ�βγ)] D≡ Eh 1−ν2 B≡ Eh3 12(1−ν2) — equilibrium equations n αβ |α +pβ =0 m αβ |αβ +(nαβw|β)|α+p=0 (2.3) 384 B.Michalak The displacement vector field of the plate midplane is denoted by u(ξα, t)=uβ(ξα, t)gβ +w(ξ α, t)g3 ξ α ∈Π (2.4) and the external surface loading by p(ξα, t)= pβ(ξα, t)gβ +p(ξ α, t)g3 ξ α ∈Π (2.5) Setting the external surface loading pβ = p = 0, we obtain equilibrium equ- ations (2.3) in the form m αβ |αβ +nαβw|βα =0 (2.6) This direct description leads to plate equations with highly-oscillating coeffi- cients, which are too complicated to be used in the analysis of stability pro- blems and numerical calculations. 3. Modelling procedure By a heteroperiodic plate we shallmean amicroheterogeneous plate which in subregions of Π, much smaller than Π, can be approximately regarded as periodic. The characteristic feature of every periodic plate is that there exists a representative cell ∆. The edge length dimensions of the cell ∆ are equal to the periods of the heterogeneous material structure of this plate. Now we define ∆(x) := { y=x+ηαlα(x), η∈ ( −1 2 , 1 2 )} x∈Π∆ (3.1) where lα = |lα| are the cell length dimensions, Π∆ := {x∈Π : ∆(x)⊂Π}. Denoting by l(x) the diameter of ∆(x) and define l = sup l(x) as a meso- structure parameter, we assume that l is sufficiently small compared to the smallest characteristic length dimension LΠ of Π (l �LΠ) and sufficiently large compared to the plate thickness h (h � l) (Fig.2). In this case, every ∆(x) defined by Eq. (3.1) will be called a cell with the center at x. Now we assume that a certain cell distribution ∆(·) has been assigned to Π. The averaging formula can be now generalized to the form 〈ϕ〉(x)= 1 |∆(x)| ∫ ∆(x) ϕ(y) dy x∈Π∆ (3.2) Stability of composite plates with... 385 In order to derive an averaged mathematical model for the plate under consi- deration we will adapt the tolerance averaging method developed by Woźniak and Wierzbicki (2000). In the framework of the method for periodic plates, we introduce the concept of a slowly varying and periodic-like function for the tolerance system T =(F,ε(·)). The continuous function Φ(·)∈F , defined on the periodic plate region Π, will be called slowly varying if ∀x,y∈Π ‖x−y‖¬ l ⇒ |Φ(x)−Φ(y)| ¬ εΦ (3.3) The continuous function f(·) defined on Π will be called a periodic-like func- tion if for every x ∈ Π∆ there exists a ∆-periodic function fx(·) such that for every y∈Π∆ ‖x−y‖¬ l ⇒ |f(x)−fx(y)| ¬ εf (3.4) We shall write Φ(·)∈SV∆(T) if Φ(·) and all its derivatives are slowly-varying functions, and f(·)∈ PL∆(T) if f(·) and all its derivatives are periodic-like functions. The periodic-like function f(·)will be called an oscillating periodic- like function if the condition 〈cf〉(x)∼=0 holds for every x∈Π∆, where c(·) is a positive value ∆-periodic function. Now definitions (3.3), (3.4) can be generalized, and after interpreting the symbol ∆ as a cell distribution ∆(·), the definition of slowly varying and periodic-like functionswill be given by Φ(·)∈SV∆(T) ⇔ {∀x∈Π∆ : Φ|P(x)(·)∈SV∆(x)(T)} (3.5) f(·)∈PL∆(T) ⇔ {∀x∈Π∆ : f|P(x)(·)∈PL∆(x)(T)} for a certain region P(x) such that ∆(x)⊂P(x)⊂Π; the symbol f|P(x)(·) denotes here the restriction on the function f(·) to P(x). Let f(·) be an integrable functiondefinedon Π such that 〈f〉(·) is a slowly varying function, 〈f〉 ∈ SV∆(x)(T). We assume that averaged values 〈f〉(x), x ∈ Π∆ have to be calculated with some tolerance determined by a certain tolerance parameter ε〈f〉. The function f(·) will be called a ∆-heteroperiodic function if for every x∈Π∆ there exists a ∆(x)-periodic function fx(·) such that ∀x∈Π∆ 〈|f−fx|〉(x)¬ ε〈f〉 (3.6) Aheterogeneous platewill be called heteroperiodic if all material properties of this plate can be described by heteroperiodic functions. Otherwise, by a hete- roperiodic platewemean aplatewhich in small regions (small neighbourhoods of ∆(x)) can be approximately regarded as a periodic one. 386 B.Michalak 4. Averaged description The tolerance averaging applied to the plate under consideration is based on two additional modelling assumptions. The assumption ofmacromodelling states that every cell ∆(x) (Fig.2), within a certain tolerance, can be treated as nondiscernible with a rectangular cell shown in Fig.3. Fig. 3. The rectangular cell The conformability assumption states that the deflection w(·) of the plate midplane is in a small region P(x) (Eq. (3.5)) a periodic-like function, w(·)∈ PL∆(x)(T), that means the deflection is conformable to the plate structure. This condition may be violated only near the boundary of the plate. Bearing in mind the lemmas of the tolerance averaging method (see Woźniak and Wierzbicki, 2000), the conformability assumption can be represented by the decomposition w(ξβ, t)=w0(ξβ, t)+ w̃(ξβ, t) (4.1) where w0 = 〈w〉, w0(·)∈SV∆(x)(T), w̃(·)∈PL1∆(x)(T) is called thedeflection disturbance and satisfy the condition 〈w̃〉∼=0. Substituting the right-handside ofEq. (4.1) into equilibriumequation (2.6) and applying the tolerance averaging, we arrive at the equation [ 〈BHαβγδ〉(ξτ)w0|γδ(ξ τ , t) ] |αβ + [ 〈BHαβγδw̃|γδ〉(ξτ, t) ] |αβ −Nαβw0|βα(ξ τ , t)= 0 (4.2) where Nαβ = 〈nαβ〉. According to the conformability assumption, we have to Stability of composite plates with... 387 assume that the forces in the plate midplane are determined by the periodic- like function nαβ(·)∈PL∆(x)(T). Hence, these forces can be represented by the decomposition nαβ(ξβ, t)=Nαβ(ξβ, t)+ ñαβ(ξβ, t) (4.3) where Nαβ(·) ∈ SV∆(x)(T), and ñαβ(·)∈ PL1∆(x)(T) is a fluctuating part of forces nαβ(·), such that 〈ñαβ〉∼=0. MultiplyingEq. (2.6) by anarbitrary ∆(x)-periodic test function δw, such that 〈δw〉 = 0, averaging this equation over ∆(x), x ∈ Π∆, and using the tolerance averaging formulae (see Woźniak and Wierzbicki, 2000), we obtain a periodic problem on the cell ∆(x) for the ∆(x)-periodic function w̃x(·), given by the following variational condition 〈δw|αβBHαβγδw̃x|γδ〉(ξτ , t)+ 〈δw|βnαβw̃x|α〉(ξ τ , t)= (4.4) =−〈δw|αβBHαβγδ〉(ξτ)w0|γδ(ξ τ, t) which has to hold for every test function δw. The approximate solution to the above variational cell problem will be assumed in the form w̃x(y, t)∼=hα(y)Vα(x, t) (4.5) where y ∈ ∆(x), x ∈ Π∆; hα(·) are postulated ∆(x)-periodic functions such that 〈hα〉=0, and Vα(·, t) are new unknowns which are assumed to be slowly varying functions, Vα(·)∈SV∆(x)(T).The functions hα(·), called shape functions,dependon themesostructureparameter l such that l−1hα(·)∈O(l), lhα |γβ (x)∈O(l), max |hα(y)| ¬ l2, y∈∆(x). Substituting the right-hand sides of Eq. (4.5) into (4.2) and (4.4) and setting δw= hα(y) in (4.4) on the basis of the tolerance averaging relations, we finally arrive at the governing equations for the considered plates [ 〈BHαβγδ〉(ξτ)w0|γδ(·, t) ] |αβ + [ 〈BHαβγδhµ |γδ 〉(ξτ)Vµ(·, t) ] |αβ −Nαβw0|αβ =0 (4.6) 〈BHαβγδhµ |αβ 〉(ξλ)w0|γδ(·, t)+ 〈BH αβγδh µ |αβ hτ|γδ〉(ξ λ)Vτ(·, t)+ +Nαβ〈hµ |β hτ|α〉Vτ =0 where the underlined term depends on the mesostructure parameter l. In Eq. (4.6)2 we have assumed that the fluctuating part ñ αβ(·) of the forces nαβ(·) is very small compared to their averaging part Nαβ(·), and hence 〈hµ |β nαβhτ |α 〉∼=Nαβ〈hµ|βh τ |α 〉. 388 B.Michalak Taking into account Eq. (4.5), the plate deflection can be approximated bymeans of the formula w(ξβ, t)∼=w0(ξβ, t)+hα(y)Vα(ξβ, t) (4.7) The presented model has a physical sense when the basic unknowns w0(ξβ, t), Vα(ξ β, t) are ∆(x)-slowly varying functions, w0(·) ∈ SV∆(x)(T), Vα(·)∈SV∆(x)(T). The characteristic features of the derived length-scale model are: • The model takes into account the effect of the cell size on the stability of the considered plate. • The governing equations have averaged coefficients that are slowly vary- ing functions. The simplified model of the stability of plates with non-uniform distribu- tion of constituents can be derived from the length-scale model, Eq. (4.6), by passing to the limit l→ 0, i.e. by neglecting the parameter l, which is placed in the underlined term. Hence, we arrive at the local model governed by [ 〈BHαβγδ〉(ξτ)w0|γδ(·, t) ] |αβ + [ 〈BHαβγδhµ |γδ 〉(ξτ)Vµ(·, t) ] |αβ −Nαβw0|αβ =0 (4.8) 〈BHαβγδhµ |αβ 〉(ξλ)w0|γδ(·, t)+ 〈BH αβγδh µ |αβ hτ|γδ〉(ξ λ)Vτ(·, t)= 0 Thismodel can be treated as a certain homogenizedmodel, in which through the tolerance averagingmethod one can calculate an approximate value of the averaged stiffnesses modulus. 5. Applications We shall investigate the linear stability of plates for polar-symmetric buc- kling. Assume that the matrix and walls of a plate are made of two different isotropic homogeneousmaterials. The bending stiffness of the walls is denoted by B1 and that of the matrix by B2 = α1B1, Poisson’s ratio respectively by ν1 and ν2 =α2ν1. Moreover, the loadings p are neglected. On the leading assumption, the physical components of shape functions, for the cell shown in Fig.3, will be taken as h〈1〉(y)=h1(y)= s1(y1) [ 1− (2y2 b2 )2] (5.1) h〈2〉(y)= ρh2(y)= s2(y2) [ 1− (2y1 b1 )2] Stability of composite plates with... 389 where s1(y1)=    a2 [ 4 a2 ( y1− 1 2 l1) 2−1− 2 3 l1−2a l1 ] y1 ∈ 〈1 2 b1, 1 2 l1 〉 a2 [ − 4 (l1−a)2 (y1) 2+1− 2 3 l1−2a l1 ] y1 ∈ 〈 − 1 2 b1, 1 2 b1 〉 a2 [ 4 a2 ( y1+ 1 2 l1 )2 −1− 2 3 l1−2a l1 ] y1 ∈ 〈 − 1 2 l1,− 1 2 b1 〉 (5.2) s2(y2)=    a2 [ 4 a2 ( y2− 1 2 ∆ϕρ )2 −1− 2 3 ∆ϕρ−2a ∆ϕρ ] y2 ∈ 〈1 2 b2, 1 2 l2 〉 a2 [ − 4 (∆ϕρ−a)2 (y2) 2+1− 2 3 ∆ϕρ−2a ∆ϕρ ] y2 ∈ 〈 − 1 2 b2, 1 2 b2 〉 a2 [ 4 a2 ( y2+ 1 2 ∆ϕρ )2 −1− 2 3 ∆ϕρ−2a ∆ϕρ ] y2 ∈ 〈 − 1 2 l2,− 1 2 b2 〉 5.1. Governing equations for the length-scale model Using Eq. (4.6) with shape functions given byEq. (5.1), (5.2), we obtain a system of governing equations for polar-symmetric buckling. These equations, describing thebuckling of theplate in the framework of the length-scalemodel, take the form (〈BH11γδ〉(ρ)w0|γδ),11+ (2 ρ 〈BH11γδ〉(ρ)w0|γδ ) ,1−2〈BH22γδ〉(ρ)w0|γδ + −ρ(〈BH22γδ〉(ρ)w0|γδ),1+(〈BH 11γδh1|γδ〉(ρ)V1),11+(〈BH 11γδh2|γδ〉(ρ)V2),11+ + 2 ρ (〈BH11γδh1|γδ〉(ρ)V1),1+ 2 ρ (〈BH11γδh2|γδ〉(ρ)V2),1−2〈BH 22γδh1|γδ〉(ρ)V1+ −2〈BH22γδh2|γδ〉(ρ)V2−ρ(〈BH 22γδh1|γδ〉(ρ)V1),1−ρ(〈BH 22γδh2|γδ〉(ρ)V2),1+ −N11w0,11−N22w0,1 =0 (5.3) [C11(ρ)+N11〈(h1|1) 2〉+N22〈(h1|2) 2〉]V1+ +[C12(ρ)+N11〈h1|1h 2 |1〉+N 22〈h1|2h 2 |2〉]V2+B 111(ρ)w0,11+B 221(ρ)w0,1 =0 [C21(ρ)+N11〈h1|1h 2 |1〉+N 22〈h1|2h 2 |2〉]V1+ +[C22(ρ)+N11〈(h2|1) 2〉+N22〈(h2|2) 2〉]V2+B211(ρ)w0,11+B222(ρ)w0,1 =0 390 B.Michalak where the following denotations have been introduced B111(ρ)= 〈BH1111h1|11〉+ 〈BH 1122h1|22〉 B211(ρ)= 〈BH1111h2|11〉+ 〈BH 2211h2|22〉 B221(ρ)= 〈BH1122h1|11〉+ 〈BH 2222h1|22〉 B222(ρ)= 〈BH1122h2|11〉+ 〈BH 2222h2|22〉 (5.4) C11(ρ)= 〈BH1111h1|11h 1 |11〉+2〈BH 1122h1|11h 1 |22〉+ +4〈BH1212h1|12h 1 |12〉+ 〈BH 2222h1|22h 1 |22〉 C12(ρ)=C21(ρ)= 〈BH1111h1|11h 2 |11〉+ 〈BH 1122h1|11h 2 |22〉+ +4〈BH1212h1|12h 2 |12〉+ 〈BH 2211h1|22h 2 |11〉+ 〈BH 2222h1|22h 2 |22〉 C22(ρ)= 〈BH1111h2|11h 2 |11〉+2〈BH 1122h2|11h 2 |22〉+ +4〈BH1212h2|12h 2 |12〉+ 〈BH 2222h2|22h 2 |22〉 Eliminating the internal variables V1 =A 11w0,11+A 1ρw0,1= B211K1−B111K2 K3K2−K21 w0,11+ B222K1−B221K2 K3K2−K21 ρw0,1 (5.5) V2 =A 22w0,11+A 2ρw0,1= B111K1−B211K3 K3K2−K21 w0,11+ B221K1−B222K3 K3K2−K21 ρw0,1 where K1 =C 12(ρ)+N11〈h1|1h 2 |1〉+N 22〈h1|2h 2 |2〉 K2 =C 22(ρ)+N11〈(h2|1) 2〉+N22〈(h2|2) 2〉 K3 =C 11(ρ)+N11〈(h1|1) 2〉+N22〈(h1|2) 2〉 we obtain the equilibrium equation in the form (C1(ρ,N αβ)w0,11 ),11+C2(ρ,N αβ)w0,11+(ρC2(ρ,N αβ)w0,11 ),1+ +(ρC3(ρ,N αβ)w0,1 ),11+ρC4(ρ,N αβ)w0,1+(ρ 2C4(ρ,N αβ)w0,1 ),1+ (5.6) −N11w0,11−N22ρw0,1=0 Stability of composite plates with... 391 where C1(ρ,N αβ) = 〈BH1111〉+B111A11+B211A22 C2(ρ,N αβ) = 2 ρ2 〈BH1111〉−〈BH2211〉+ ( 2 ρ2 B111−B221 ) A11+ + ( 2 ρ2 B211−B222 ) A22 (5.7) C3(ρ,N αβ) = 〈BH1122〉+B111A1+B211A2 C4(ρ,N αβ) = 2 ρ2 〈BH1122〉−〈BH2222〉+ ( 2 ρ2 B111−B221 ) A1+ + ( 2 ρ2 B211−B222 ) A2 5.2. Governing equations for the local model Now we consider buckling of a plate in the framework of the local model. Thismodel can be derived directly from the length-scalemodel Eqs (5.3)-(5.7) bypassing l→ 0, i.e. by neglecting termswith themesostructure parameter l. Hence, we arrive at equilibrium equations (〈BH11γδ〉(ρ)w0|γδ),11+ (2 ρ 〈BH11γδ〉(ρ)w0|γδ ) ,1−2〈BH22γδ〉(ρ)w0|γδ + −ρ(〈BH22γδ〉(ρ)w0|γδ),1+(〈BH 11γδh1|γδ〉(ρ)V1),11+(〈BH 11γδh2|γδ〉(ρ)V2),11+ + 2 ρ (〈BH11γδh1|γδ〉(ρ)V1),1+ 2 ρ (〈BH11γδh2|γδ〉(ρ)V2),1−2〈BH 22γδh1|γδ〉(ρ)V1+ −2〈BH22γδh2|γδ〉(ρ)V2−ρ(〈BH 22γδh1|γδ〉(ρ)V1),1−ρ(〈BH 22γδh2|γδ〉(ρ)V2),1+ −N11w0,11−N22w0,1 =0 (5.8) C11(ρ)V1+C 12(ρ)V2+B 111(ρ)w0,11+B 221(ρ)w0,1 =0 C21(ρ)V1+C 22(ρ)V2+B 211(ρ)w0,11+B 222(ρ)w0,1 =0 with the denotations given by Eq. (5.4). 392 B.Michalak Eliminating the internal variables V1 = A 11w0,11+A 1ρw0,1= = B211C12−B111C22 C11C22− (C12)2 w0,11+ B222C12−B221C22 C11C22− (C12)2 ρw0,1 (5.9) V2 = A 22w0,11+A 2ρw0,1= = B111C12−B211C11 C11C22− (C12)2 w0,11+ B221C12−B222C11 C11C22− (C12)2 ρw0,1 we obtain the equilibrium equation in the form similar to Eq. (5.6) (C1(ρ)w 0,11 ),11+C2(ρ)w 0,11+(ρC2(ρ)w 0,11 ),1+(ρC3(ρ)w 0,1 ),11+ (5.10) +ρC4(ρ)w 0,1+(ρ 2C4(ρ)w 0,1 ),1−N11w0,11−N22ρw0,1=0 where C1(ρ) = 〈BH1111〉+B111A11+B211A22 C2(ρ) = 2 ρ2 〈BH1111〉−〈BH2211〉+ ( 2 ρ2 B111−B221 ) A11+ + ( 2 ρ2 B211−B222 ) A22 (5.11) C3(ρ) = 〈BH1122〉+B111A1+B211A2 C4(ρ) = 2 ρ2 〈BH1122〉−〈BH2222〉+ ( 2 ρ2 B111−B221 ) A1+ + ( 2 ρ2 B211−B222 ) A2 5.3. Illustrative example Now we will investigate a special case of polar- symmetrical buckling of an annular plate. Assume that the cell length l1 = ∆ϕρ, Poisson’s ratio ν1 = ν2 = 0 and the beam thickness a = ml1. Hence, all averaged plate stiffenesses are constant, and for the local model equilibrium equation (5.10) have the form C̃1w 0 ,1111+ 2 ρ C̃1w 0 ,111− 1 ρ2 C̃1w 0 ,11+ 1 ρ3 C̃1w 0 ,1−Nρw0,11− 1 ρ Nϕw 0 ,1 =0 (5.12) Stability of composite plates with... 393 where C̃1(ρ)= 〈BH̃1111〉+ B̃111Ã11+ B̃211Ã22 Ã11 =A11 Ã22 = 1 ρ A22 B̃111 =B111 B̃211 = ρB211 (5.13) 〈BH̃1111〉=B1[m+m(1−m)+α1(1−m)2] In Eq. (5.13), B1 denotes the bending stiffeness of the beams, and α1 =Ematrix/Ebeams. Fig. 4. An annular plate subjected to constant compressive forces Wewill investigate the stability of the annular plate subjected to constant compressive forces distributed along the edges of the plate (Fig.4). Bearing in mind that the tensile forces Nρ and Nϕ are averaged parts of the middle surface forces nαβ, from the equilibrium equations formembrane forces in the midplane one gets, for pa = pb, the following condition Nρ =Nϕ =N. In this case, equilibrium equation (5.12) can be assumed in the form L[Lw0(x)]−γLw0(x)= 0 (5.14) 394 B.Michalak where, adopting a newdimensionless independent variable x= ρ/rz (rz is the external radius of the annular plate) L= d2 dx2 + 1 x d dx γ= N(rz) 2 C̃1 (5.15) Fourth order differential equations (5.14) can be replaced by two independent second order Bessel’s differential equations. The solution to these equations will be obtained as w0(x)=D1+D2 lnx+D3J0(λx)+D4Y0(λx) (5.16) where λ= √ −γ and J0(λx), Y0(λx) are Bessel’s functions. In the case of an annular plate clamped along the circumference, the bo- undary conditions have the form w0(x= η)= 0 dw0(x= η) dρ =0 w0(x=1)=0 dw0(x=1) dρ =0 (5.17) where η= rw/rz (rz – external and rw – internal radius of the annular plate). Substituting Eq. (5.16) into (5.17), we obtain the condition ∣∣∣∣∣∣∣∣∣ J0(ηλ) Y0(ηλ) lnη 1 −ηλJ1(ηλ) −ηλY1(ηλ) 1 0 J0(λ) Y0(λ) 0 1 −λJ1(λ) −λY1(λ) 1 0 ∣∣∣∣∣∣∣∣∣ =0 (5.18) fromwhichwe calculate the critical value of the coefficient λcr and the critical compressive force Ncr = (λcr) 2C̃1 (rz)2 (5.19) Introducing notations Ncr = scrB1/(rz) 2, where B1 = Ebeamsh 3/12(1−ν2), we derive diagrams of the parameter scr versus the ratio n= rw/rz. On the diagram in Fig.5 one can observe the smallest value of the critical parameter scr versus the ratio n for the ratio of the matrix and beams Young moduli α1 =Ematrix/Ebeams =0.5,where the ratio a/l1 wasusedas aparameter.The diagrampresenting the parameter scr for n= a/l1 =1.0 shows the parameter corresponding to the critical force for a homogeneous plate made of the same material as that of the beams, while the diagram for n = a/l1 = 0 shows the critical parameter for a homogeneous plate made of the matrix material. Figure 6 shows the critical parameter scr for n= a/l1 =0.1, where the ratio α1 =Ematrix/Ebeams is used as a parameter. Stability of composite plates with... 395 Fig. 5. The smallest value of the parameter scr of critical forces N versus the ratio n= rw/rz. The ratio a/l1 is used as a parameter. It is assumed that α1 =Ematrix/Ewalls =0.5 Fig. 6. The smallest value of the parameter scr of the critical forces N versus the ratio n= rw/rz. The ratio α1 =Ematrix/Ewalls is used as a parameter. It is assumed that a/l1 =0.1 6. Conclusions In this paper, the tolerance averaging method, developed byWoźniak and Wierzbicki (2000) for heteroperiodic solids, is adopted to the analysis of sta- bility of composite plates with non-uniformdistribution of constituents. From the above considerations it follows that the tolerance averaging method can 396 B.Michalak be successfully applied to the formulation of the averaging model of the linear stability of such plates. The modelling approach is different from the known homogenization me- thods and leads to a model in which governing equations depend on the mi- crostructure size. It has to be mentioned that the results obtained in this contribution cannot be derived by using the homogenization method related to solids which are periodic with respect to a certain curvilinear parametriza- tion, see Lewiński and Telega (2000). It can be seen that the above modelling approach leads from equations, which have highly oscillating coefficients, to a system of equations with non- constant but slowly varying coefficients. A solution to these equations can be obtained by applying known typical numerical procedures. References 1. Bensoussan A., Lions J.L., Papanicolau G., 1978, Asymptotic Analysis for Periodic Structures, North-Holland, Amsterdam 2. Gomuliński A., 1967, Determination of eigenvalues for circular plates resting on elastic foundation with twomoduli,Arch. Inż. Ląd., 13, 183-203 3. Lewiński T., Telega J.J., 2000, Plates, Laminates and Shells. Asymptotic Analysis and Homogenization, World Sci. Publ. Co., Singapore-HongKong 4. Waszczyszyn Z., 1976, The critical load an annular elastic plate for assym- metric buckling [in Polish],Arch. Bud. Masz., 23, 79-93 5. WoźniakC. (edit.), 2001,Mechanikasprężystychpłyt ipowłok, inMechanika Techniczna, VIII, PWN,Warszawa 6. Woźniak C., Wierzbicki E., 2000,Averaging Techniques in Thermomecha- nics of Composite Solids, Wydaw. Pol. Częstochowskiej, Częstochowa 7. Woźniak C., Zieliński S., 1967, On some stability problems of circular per- forated plates,Arch. Inż. Ląd., 13, 155-161 Stateczność płyt kompozytowych z niejednorodnym rozkładem składników Streszczenie Celem pracy jest sformułowanie i zbadanie uśrednionegomodelu opisującego sta- teczność płyty kompozytowej z niejednorodnym rozkładem składników. Rozpatry- wana płyta ma określoną budowę, która nie jest periodyczna, ale która w małym Stability of composite plates with... 397 obszarze rozpatrywanej płytymoże być w przybliżeniu traktowana jako periodyczna. Przedmiotem analizy jest kolista płyta kompozytowa zbudowana z dwóch rodzajów sprężystych prętów, których osie są prostopadłe. Obszar pomiędzy prętami wypełnia jednorodny sprężysty materiał matrycy. Sformułowanie przybliżonego modelu mate- matycznego bazuje na koncepcji uśredniania tolerancyjnego przedstawionej w pracy Woźniaka iWierzbickiego (2000), gdzie ciało tego rodzaju nazwane jest ciałem hete- roperiodycznym. Manuscript received December 30 2003; accepted for print February 4, 2004