Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 47, 3, pp. 515-535, Warsaw 2009 OPTIMAL DESIGN OF FIBER-REINFORCED COMPOSITE DISKS Krzysztof Dems Jacek Wiśniewski Technical University of Lodz, Department of Technical Mechanics and Informatics, Łódź, Poland e-mail: krzysztof.dems@p.lodz.pl The paper is devoted to the modeling, analysis and optimization en- countered in the design process of two-dimensional structural compo- nents made of fiber-reinforced composite materials subjected to service loading. The problem of optimal layout of reinforcing fibers in the com- posite in order to obtain the assumedmechanical properties of the disk is considered.The case of creation of linear and curvilinear fibers is discus- sed. The adequate model, relevant optimality conditions for this type of design problem are derived and the optimization procedure based on the evolutionary algorithm is proposed. The problem considered in the paper is illustrated by some numerical examples. Key words: composite material, fibers layout, optimal design, evolu- tionary algorithm 1. Introduction Fiber-reinforced composite materials are a group of modern construction ma- terials from which products used in many areas of technical applications are made. These materials are characterized by very good mechanical properties associated with their small weight. The optimal design of these structures is a very complex process. To fulfill the assumed properties of the composite structures, we canmodify some their structural parameters, such as mechani- cal properties of the matrix or reinforcing fibers, percentage participation of fibers in the structure, fiber shape and orientation, etc. Each of these parame- ters influences themechanical properties of the compositematerial and can be treated as the design variable during the optimal design. However, as shown in previous Author’s papers (Dems, 1986; Dems andWiśniewski, 2006, 2007) 516 K. Dems, J. Wiśniewski the full advantages of the composite materials can be also obtained when the reinforcing fibers are optimally oriented or shaped in thematrix with respect to the assumed objective behavioral measure of the structure under actual lo- ading conditions. The problemof optimal fibers arrangement in the composite material is discussed in the present paper. The paper constitutes the results of further investigation in the area of designing composite structures. The results can be treated as a starting point for optimal design of real composite structures subjected to service load.Thus, it allows for avoiding expensive experimental testing, which can be reduced to the final phase of structural design. 2. Object of analysis A thin, two-dimensional and linearly elastic disk shown in Fig.1 is conside- red. The disk has a uniform thickness and it is supported on the boundary portion SU with prescribed displacements u 0 and loaded by body forces f0 within a domain A and by an external traction T0 acting along the boundary portion ST . Fig. 1. Two-dimensional composite disk subjected to service loading The material of the disk is a composite made of a matrix reinforced with a ply of long and unidirectional fibers of an arbitrary shape and assumed cross-section (Fig.1). Let us assume that: • The matrix is homogeneous, isotropic and linearly elastic. Young’s mo- dulus and Poisson’s ratio of the matrix are denoted by Em and νm, respectively. Optimal design of fiber-reinforced composite disks 517 • The reinforcing fibers are homogeneous, orthotropic and linearly elastic. Themechanical properties of fibers are characterized byYoung’smoduli Ew1 and Ew2 with respect to thematerial axes 1-2, coinciding with the fiber direction and the direction perpendicular to the fiber, Poisson’s ratio νw12 as well as shear modulus Gw12 in the 1-2 plane. • The fibers are regularly spaced and perfectly aligned in thematrix with global density ρw, and their orientation at any point of the composite material is denoted by the angle θ with respect to the x-axis of global coordinate system xy. • The bonds between thematrix and the fibers are perfect. The angle of fiber orientation θ defines the layout of reinforcing fibers at any point of the composite material (Fig.1). This angle can be constant in the composite, and then the fibers are rectilinearly spaced in the matrix or can vary through the composite domain. In this last case, the fibers are placed curvilinearly in the matrix. In general, the layout of reinforcing fibers can be described by the shape of the so-called directional fiber using, for instance, the polynomial, spline orBezier representation.As the result, the fiber orientation at any point of the composite depends on a set of fiber shape parameters b defining that particular representation, i.e. θ= θ(x,b) and it is treated as an angle between the tangent line to the fiber and x-axis of the global coordinate system, as it is shown in Fig.1. Fig. 2. Family of straight fibers obtained by translation of directional fiber in arbitrary direction In the first case, the reinforcing fibers constitute a family of straight fibers (Fig.2), and the shape of directional fiber is described by y(b)= bx+ c (2.1) 518 K. Dems, J. Wiśniewski where b and c denote the parameters of a straight line. The fiber orientation angle θ for a straight fiber at any point of the composite is related to the parameter b as follows tanθ= b (2.2) or it is directly definedby the angle between the fiber line and the x-axis of the global coordinate system (Fig.2). All other fibers in the family are obtained by translation of the directional fiber in the d-direction, according to the rule yi = y(b)+ iq (2.3) where i is the number of current fiber in the family and q is the distance between two adjacent fibersmeasured in the y direction. As a result, the one- -parameter family of reinforcing fibers of the same shape and constant fiber density is created. Besides the family of straight fibers, also a family of curvilinear fibers can be created. In this case, the fibers layout is described by the shape of the middle line of directional fiber (Fig.3 and Fig.4). Now, the shape of this line can be defined using an arbitrary parametric description, namely x=x(b, t) y= y(b, t) (2.4) where b is a set of fiber shape parameters defining the shape of the curvilinear directional fiber in the composite and t is a real parameter varying in the range 〈α,β〉. The fiber orientation angle θ at any point of the composite is defined now by tanθ= y(b, t),t x(b, t),t = y,t x,t and x,t 6=0 (2.5) where y,t and x,t denote derivatives of the function y(b, t) and x(b, t) with respect to the parameter t, respectively. For the case of curvilinear fibers, the layout of all other fibers in the family can be obtained by translation of the directional fiber in an arbitrary direction (Fig.3), or by its shifting in the direction normal to its middle line (Fig.4). In the case of creation of the fiber family by translation of directional fiber (Fig.3), the directional fiber is translated in the d-direction according to the rule xi =x(bk, t)+ idx yi = y(bk, t)+ idy (2.6) where i is the number of current fiber in the family, while dx and dy denote components of the distance vector between two adjacent fibers in the direction of translation. Thus, all fibers constitute a one-parameter family of fibers of Optimal design of fiber-reinforced composite disks 519 Fig. 3. Family of curvilinear fibers obtained by translation of directional fiber in arbitrary direction Fig. 4. Family of curvilinear fibers obtained by shifting of directional fiber in normal direction the same shape, but the local fiberdensity is varying in the composite domain. The fiber density at any point Mi(xi(t),yi(t)) of the composite is specified as follows ρw(i) = ρw(0) µx0,t−y0,t µx,t−y,t √ (x,t)2+(y,t)2 (x0,t)2+(y0,t)2 µ= dx dy (2.7) where ρw(0) denotes the fiber density in the assumed point M0(x0(t),y0(t)). The family of curvilinear reinforcing fibers can also be obtained by shifting the directional fiber in the direction normal to itsmiddle line, as it is shown in Fig.4. In this case, the local fiber density is constant in the composite domain, 520 K. Dems, J. Wiśniewski but each particular fiber in this family has a slightly different shape described by the following expressions xi =x(bk, t)+ iq ( − y,t√ (x,t)2+(y,t)2 ) (2.8) yi = y(bk, t)+ iq ( x,t √ (x,t) 2+(y,t) 2 ) where i is the number of current fiber in the family and q is the distance between two adjacent fibers. Moreover, the creation procedure of the family of curvilinear fibers by shifting the directional fiber has to be subjected to the following condition √ [ζ−x(t)min]2+[η−y(t)min]2 ­ √ [x−x(t)min]2+[y−y(t)min]2 (2.9) which prevents the fibers from intersection of their middle lines. 3. Model of fiber-reinforced composite material The microscopically non-homogeneous composite material is treated on a macro-scale level as a plane, homogeneous, orthotropic and linearly elastic material. In conformity with this assumption, the relevant model of the com- posite is built (Fig.5). The purpose of the modeling process is to determine the extensional stiffnessmatrix D for thatmodel in the global coordinate sys- tem and to express its components in terms of mechanical and geometrical properties of the matrix and the reinforcing fibers. Fig. 5. Fiber-reinforced composite material: (a) real composite, (b) model of composite Optimal design of fiber-reinforced composite disks 521 The extesional stiffnessmatrix D for the assumedmodel of the composite material in the global coordinate system xy can be expressed by (German, 1996) D=T−1CT−⊤ (3.1) Thematrix C denotes here the stiffnessmatrix for the composite with respect to the material axes 1-2, coinciding with the fiber direction and the direction perpendicular to the fiber, and has the form C=        E1 1−ν12ν21 E1ν21 1−ν12ν21 0 E2ν12 1−ν12ν21 E2 1−ν12ν21 0 0 0 G12        (3.2) whereas the matrix T denotes the transformation matrix from the global co- ordinate system xy to the material axes 1−2, and it is expressed as follows T=    cos2θ sin2θ 2sinθcosθ sin2θ cos2θ −2sinθcosθ −sinθcosθ sinθcosθ cos2θ− sin2θ    (3.3) Thecomponents ofmatrix Cdependon the so-called engineering constants for orthotropic lamina, where E1 and E2 are apparent Young’s moduli in the fiber direction and in the direction perpendicular to the fibers, respectively, while ν12 and ν21 aremajor andminor Poisson’s ratios, and G12 denotes the in-plane shear modulus. Using the model of lamina presented by Halpin and Tsai (1969) these engineering constants have the following form E1 =Ew1ρw+Em(1−ρw) E2 =Em (1+ ξχρw 1−χρw ) where χ= Ew2 Em −1 Ew2 Em + ξ ν12 = νw12ρw+νm(1−ρw) (3.4) ν21 = E2 E1 ν12 G12 =Gm (1+ ξχρw 1−χρw ) where χ= Gw12 Gm −1 Gw12 Gm +ξ Gm denotes here the shear modulus for the isotropic matrix, while the para- meter ξ is a measure of fiber reinforcement of the composite that depends on the cross-section of fibers and packing geometry. The values of ξ for typical 522 K. Dems, J. Wiśniewski Table 1.Values of ξ for typical fibers value of ξ forE2 value of ξ forG12 circular fibers ξ=2 ξ=    1 for ρw ¬ 0.5 1+40ρ10w for ρw > 0.5in square array quadratic cross-section ξ=2 ξ=1fibers in diamond array rectangular cross-section ξ=2 a b ξ= √ 3ln a bfibers in diamond array fibers obtained by Adams and Doner (1967a,b) and Foye (1966) are given in Table 1. The matrix T is considered as the matrix function of fiber orientation angle θ. When the fiber line is described by (2.1), its components are simply defined by sinθ= tanθ√ 1+tan2θ = b√ 1+ b2 (3.5) cosθ= 1√ 1+tan2θ = 1√ 1+ b2 or they can be explicitly defined by the angle between the fiber line and the x-axis of the global coordinate system. For the case of curvilinear fibers, sinθ and cosθ follow from (2.5), and are expressed in the form sinθ= tanθ√ 1+tan2θ = y,t √ (x,t)2+(y,t)2 (3.6) cosθ= 1√ 1+tan2θ = x,t √ (x,t) 2+(y,t) 2 4. Analysis of structural behavior Under the applied load, the structure shown in Fig.1 undergoes some defor- mations described by the displacement field u, the strain field e and the Optimal design of fiber-reinforced composite disks 523 stress field σ. Thus, the behavior of the structure can be described by the equilibrium equation given in the form (Zienkiewicz, 1971) divσ+f0 =0 (4.1) as well as the kinematical relation between strain and displacement fields e=Bu (4.2) where B is a lineardifferential operator relating thedisplacementfieldwith the strain field.A linear stress-strain relation is assumed in the formof generalized Hooke’s law σ=De (4.3) where D denotes the extensional stiffness matrix for the orthotropic model of the compositematerial, and it is given by (3.1)-(3.6). Besides, the structure is subjected to the boundary conditions expressed as follows { σn=T0 on ST u=u0 on SU (4.4) where n denotes the normal unit vector on the external boundary S of the disk. To solve the set of equations (4.1)-(4.4), the finite elementmethod (FEM) is proposed. The structure domain is divided into a set of finite elements connected atnodes each to another. In thepresentpaper, the two-dimensional four-nodequadrilateral elements areused.Thedetailed descriptionof thefinite element methods is presented, for instance, in Zienkiewicz (1971). 5. Optimization problem Themechanical properties of composite disksdependonmechanical properties of the matrix and reinforcing fibers, percentage participation of fibers in this material, fiber shape and orientation as well as on the cross-section of fibers. Thus, each of these parameters can be treated as a design variable during the optimization procedure. However, full advantages of the composite disk can mainly be obtained when the reinforcing fibers are optimally arranged in the matrixwith respect to the assumed objective behavioralmeasure under actual loading conditions of the structure. 524 K. Dems, J. Wiśniewski The problem of optimal design of fibers arrangement in the composite, so that the structure should satisfy assumed requirements in the range of mechanical properties can be formulated in general as the minimization or maximization of the proper quality index of structural behavior, expressed in the global form as follows min (or max) Fc = ∫ A Γ(σ,e,u,b) dA+ ∫ ST Ψ(T0,u) dST (5.1) subjected to the global or local behavioral constraints ( ∫ A Γ1(σ,e,u,b) dA+ ∫ ST Ψ1(T 0,u) dST ) −G0 ¬ 0 (5.2) and/or the constraint imposed on the total cost of the composite structure [cwρw+ cm(1−ρw)]V −C0 ¬ 0 (5.3) where Γ ,Γ1 and Ψ,Ψ1 are continuous functions depending on the state fields induced in the deformed structure, while cw and cm are the costs per unit vo- lume of the fiber andmatrixmaterials, respectively. The fiber orientation of a straight fiber or parameters defining the shape of a curvilinear fiber are selec- ted as components of the vector of design variables b during this optimization process. The optimization task, defined by (5.1)-(5.3), will be performed with the aid of the evolutionary algorithm shown in Fig.6. This method based on the imitation of the evolution processes occurring in the nature (Michalewicz, 1996) still finds the growing interest in engineering design problems. The evo- lutionary algorithm is a simple, powerful and effective tool used for finding the best solution in a complicated space of design parameters and it is not limited by a restrictive assumption about the search space. This method only needs the information based on the objective functional, which is itsmain advantage in comparison to the methods based on gradient information of the objecti- ve functional and constraints. Besides, in contrast to deterministic methods, which often fall into a local optimum, the evolutionary algorithm always finds the global optimum or the solution close to this optimum. It should be added that in the case of the optimization problem with constraints, the penalty function approach (Findeisen et al., 1980) is applied in the proposed algorithm. Using this approach, the constrained problem is transformed to an unconstrained one as follows minZ(b,α)=min [ Fc(b)+ 1 2 ng ∑ i=1 αi[max(0;Gi(b))] 2 ] (5.4) Optimal design of fiber-reinforced composite disks 525 data: LP – number of generations N – number of chromosomes in population pc – crossover probability pm –mutation probability step 1: GENERATIONOF INITIAL POPULATION (floating point representation) ∧ j=1,...,N chj → bj = [b1,b2, . . . ,bp] where bi = bi(max)+r(bi(max)− bi(min)) step 2: EVALUATIONOFCURRENTPOPULATION: ∧ j=1,...,N v(chj)≡Fc(bj) step 3: if k>LP then STOP else go to step 4 step 4: OPERATIONOFCURRENTPOPULATION      deterministic selection: choice chromosomes the best v(chj) to modification heuristic crossover: ch′ = r(ch2− ch1)+ ch2 where: Fc(ch2)­Fc(ch1) non-uniform mutation: chj = [b1, . . . ,bi, . . . ,bp]→ ch′j = [b1, . . . ,b′i, . . . ,bp] step 5: CURRENTPOPULATION= NEWPOPULATION and go to step 2 Fig. 6. Flow chart of evolutionary algorithm where α is a vector of positive coefficients of penalty functions, and ng is a number of inequality constraints in the constrained problem. The evolutionary algorithm starts from random selection of the initial po- pulation of N chromosomes. Each chromosome is a coded vector of design parameters b defining the shape or orientation of reinforcing fibers and de- scribes onepossible solution to thegivenproblem.Thispopulation isprocessed by threemain operators of the evolutionary algorithm. They are deterministic selection, heuristic crossover and non-uniformmutation. Applying these three operators, a new population of solutions is created and the single cycle of the evolutionary algorithm, which is known as a generation, comes to the end. Each successive generation contains better „partial solutions than in the pre- vious generations, and converges towards the global optimum.This procedure is continued until the best solution is found according to the assumed stop criterion or the specified number of generations is attained. As it is shown inFig.6, all chromosomes in the current population are eva- luated using the objective functional Fc in the second step of the evolutionary algorithm. The values of state fields appearing in the objective functional are calculated using the finite element method in the analysis of the structural behavior (Section 4). 526 K. Dems, J. Wiśniewski 6. Optimal design of composite disks To illustrate the applicability of the proposed approach to the optimal design of fiber-reinforced composite disks, some simple examples are presented in this Section. The particular forms of optimization problem (5.1)-(5.3) will be introduced in the successive examples, in order to obtain the optimal layout of reinforcing fibers in the composite structure. As the first example, let us consider the problem of optimal strength design of a composite disk for which the optimal fiber shape and orienta- tion should be derived. Let us assume that the allowable stress level for the composite structure after homogenization is defined by the longitudinal ten- sile strength Rr1, longitudinal compressive strength Rc1, transverse tensile strength Rr2, transverse compressive strength Rc2 and shear strength Rs. For an orthotropic material, the strength function Φ corresponding to the stress field σ induced in the deformed structure is assigned using the bi-axial theory (Tsai and Wu, 1971). In conformity with this theory, the value of the strength function Φ at any point of the composite structure should satisfy the following relationship Φ(σ)=F1σ1+F2σ2+F11σ 2 1 +2F12σ1σ2+F22σ 2 2 +F33τ 2 12 ¬ 1 (6.1) where the strength coefficients Fi and Fij appearing in (6.1) have the form F1 = 1 Rr1 − 1 Rc1 F2 = 1 Rr2 − 1 Rc2 F11 = 1 Rr1Rc1 F22 = 1 Rr2Rc2 F33 = 1 R2s F12 =− √ F11F22 2 (6.2) whereas σ1, σ2 and τ12 are components of the stress field with respect to material axes of the composite, coinciding with the fiber direction and the di- rection perpendicular to the fiber.They are obtained using the transformation rule in the form    σ1 σ2 τ12    =    cos2θ sin2θ 2sinθcosθ sin2θ cos2θ −2sinθcosθ −sinθcosθ sinθcosθ cos2θ− sin2θ       σx σy τxy    (6.3) The main idea of the strength design results in reduction of values of the strength function, which violate criterion (6.1), and then in redistribution of the stress field in the domain of the structure. Thus, the global measure of Optimal design of fiber-reinforced composite disks 527 local strength criterion can be introduced as a behavioral functional and the optimization problem is formulated in the following form minFc(b)= n √ √ √ √ ∫ A [Φ(b)]n dA (6.4) where A is the area of structure domain of unitary thickness and b denotes a vector of design parameters defining the layout of fibers in the composite, and n denotes an even integer number. When n → ∞, the functional is a strict measure of the maximum local values of strength criterion. In practice, the value of n does not exceed the upper bound limit following from numerical constraints. In all presented examples, the value of nwas set to 20. The above formulated problem can be illustrated by a simple example. Let us consider a disk supported along its lower boundary and loaded by linear- ly varying traction on the upper boundary (Fig.7). The disk has a unitary thickness and it ismade of the compositematerial withmechanical properties given in Tables 2 and 3. Fig. 7. Composite disk subjected to load and boundary conditions Optimization problem (6.4) was solved for two classes of the fiber shape. In the first case, the composite material was reinforced with one family of the straight fibers, and the angle of fiber orientation θ was selected as the design parameter, i.e. b= {θ}. Next, the design problemwas considered for the case of one family of curvilinear fibers for which the shape of the middle line of directional fiber was described by Bezier’s curve (Kiciak, 2000). In this case, the coordinates of four nodes of the Bezier polygon were chosen as the design 528 K. Dems, J. Wiśniewski Table 2.Material data of components of composite E [GPa] ν G [GPa] ρ [%] fibers (graphite HS) 220.0 0.22 8.90 45 13.8 matrix (epoxy) 3.5 0.38 – 55 fibers array quadratic cross-section fibers in diamond array source: Adopted fromGerman (1996) Table 3.Allowable stress level for composite (graphite HS/epoxy) Rr1 [MPa] Rc1 [MPa] Rr2 [MPa] Rc2 [MPa] Rs [MPa] 1531 1390 41 145 98 source: Adopted fromGerman (1996) parameters, i.e. b= {xi,yi}, i=0,1,2,3. The family of curvilinear reinforcing fibers was obtained by shifting the directional fiber in the direction normal to its middle line. The optimization problem discussed here was solved using the proposed evolutionary algorithm,discussedalready inSection5.Toanalyze thebehavior of the disk, its domain was discretized into 196 two-dimensional four-node quadrilateral elements. The results of the optimization process are given in Table 4, and the optimal layouts of reinforcing fibers in the disk domain are shown in Fig.8a and 8b, respectively. Table 4.Results of optimization process design parameters value of local strength function optimal disk with one family θ=85.17◦ Φmax =0.92 of straight fibers Φmin =−0.13 optimal disk with one family of curvilinear fibers described by Bezier function P0 (0.317,0.000) P1 (0.131,0.075) Φmax =0.84 P2 (0.142,0.205) Φmin =−0.12 P3 (0.148,0.280) reference disk with one family θ=90.00◦ Φmax =1.15 of straight fibers Φmin =−0.11 These optimal solutions were also compared with the reference solution for the disk reinforced with one family of straight fibers parallel to the y-axis (Fig.8c) in order to qualify the results of optimization. One can observe that Optimal design of fiber-reinforced composite disks 529 Fig. 8. Plot of strength function in: (a) optimal disk with one family of straight fibers, (b) optimal disk with one family of curvilinear fibers described by Bezier representation, (c) reference disk with one family of straight fibers the distribution of strength function Φ in the reference disk is considerably re- duced after the optimization process, and the values of local strength function violating criterion (6.1) are eliminated for both optimal solutions (cf. the plots shown in Fig.8 and maximal values of Φ presented in Table 4). It must be also noted that the distribution of the stress field in the optimal structures is practically uniformwhich allows for better usage of the constructionmaterial or allows increasing the allowable upper level of service load. As the second example, the problem of optimal layout of reinforcing fibers in the composite for the case of mean stiffness design of the disk is di- scussed.Thus, theminimization of thework done by external forces is selected as the objective functional, and the optimization problem can be written in the following form minFc(b)= ∫ ST u ⊤ T 0 dST (6.5) where b is a vector of design parameters defining the shape of themiddle line of the directional fiber. To illustrate this kind of design problem, the composite disk shown in Fig.9 was considered. Similarly as in the first example, the disk has a unitary thickness and it is composed from polyester matrix reinforced with graphite 530 K. Dems, J. Wiśniewski fibers symmetrically placed with respect to the x axis. The material data of the components of the composite material are given in Table 5. Fig. 9. Composite disk subjected to load and boundary conditions Table 5.Material data of composite components E [GPa] ν G [GPa] ρ [%] fibers (graphite HS) 220.0 0.22 8.90 50 13.8 matrix (epoxy) 3.2 0.41 – 50 fibers array quadratic cross-section fibers in diamond array source: Adopted fromGerman (1996) The problem of mean stiffness design was discussed, as previously, for two classes of fibers shape, and due to symmetry of the problem, only one half of the disk was considered. To solve the design problem, the evolutionary algori- thm presented in Section 5 was used. The values of state fields and objective functional Fc were calculated using the finite element method with 100 two- dimensional four-node quadrilateral elements applied for discretization of the disk domain. In the first case, the composite material was reinforced with one family of straight fibers for which the angle of orientation with respect to the symmetry axis of the diskwas selected as the design parameter, i.e. b= {θ}. The results of optimization process for this case are given in Table 6 and the optimal layout of reinforcing fibers in disk is shown in Fig.10a. Optimal design of fiber-reinforced composite disks 531 In the second case, thematerial of the disk was reinforced with one family of parabolic fibers. Thus, the shape of directional fibers was assumed in the form x = ay2, where the coefficient a was treated as design parameters, i.e. b= {a}. All other fibers in the family were obtained by translation of the directional fiber in the x-direction.The results obtained after the optimization process for this case are also given in Table 6, and the optimal layout of fibers is depicted in Fig.10b. Table 6.Result of optimization process design objective parameters functional optimal disk with straight fibers θ=76.98◦ Fc =15.40J optimal disk with curvilinear fibers a=0.72 Fc =14.15J reference disk with straight fibers θ=158.0◦ Fc =38.76J Fig. 10. Optimal disk with family of straight fibers (a), optimal disk with one family of parabolic fibers (b) To evaluate the results obtained for optimal solutions, the mean stiffness of the reference disk reinforcedwith the family of straight fibers of orientation θ = 158◦ with respect to the symmetry axis of the disk (the most flexible disk) was also calculated. The optimal layout in the case of family of straight fibers increases the mean stiffness of the disk by 60%, while in the case of family of parabolic fibers, this stiffness increases by 68%, when compared to the reference disk. As the last example, let us consider the problem of optimal design of the composite disk with respect to the cost of its components. The disk is 532 K. Dems, J. Wiśniewski additionally subjected to the global and/or local behavioral constraints. Thus, the optimization problem can be formulated as follows minC(b)= [cwρw+ cm(1−ρw)]V (6.6) subjected to the global behavioral constraint ( ∫ A Γ(σ,e,u,b) dA+ ∫ ST Ψ(T0,u) dST ) −G0 ¬ 0 (6.7) where b denotes a vector of design parameters, cw and cm are the costs per unit volume of the fiber and matrix materials, respectively, while Γ and Ψ are continuous functions depending on the state fields induced in the deformed structure. To illustrate this kind of optimization problem, let us consider the compo- site disk shown inFig.11. Thedisk has a unitary thickness and themechanical properties as well as unit costs of the components are given in Table 7. Fig. 11. Composite disk subjected to load and boundary conditions Table 7.Material data of composite components E [GPa] ν cost [PLN] fibers (glass E) 75 0.25 30 matrix (epoxy) 3.5 0.38 12 fibers array circular fibers in square array source: Adopted fromGerman (1996) The minimization of the total cost of the disk reinforced with one family of straight fibers was considered. It was also assumed that the stiffness of Optimal design of fiber-reinforced composite disks 533 reinforced disk should be equal to 170J. The angle of fiber orientation θ and fiber density ρw were selected as the design parameters, i.e. b= {θ,ρw}, and they were determined during the optimization process. Table 8.Results of optimization process design parameters stiffness [J] cost [PLN] optimal disk θ=134◦, ρw =0.52 170.0 2.48 reference disk θ=90◦, ρw =0.70 170.0 2.96 Fig. 12. Optimal disk (a) and reference disk (b) reinforced with one family of straight fibers The design problem was solved using the proposed evolutionary algori- thm and the results, obtained after of the optimization process, are given in Table 8. In this example, the disk was divided into 120 two-dimensional four- nodequadrilateral elements. Theoptimal solution, shown inFig.12a,was next compared with the solution for the same disk reinforced with one family of straight fibers parallel to the y-axis (Fig.12b). The optimal layout of fibers decreases the total cost of disk by 17%when compared to the reference design shown in Fig.12b. 7. Conclusion The full advantages of a composite disk subjected to a particular load can be obtained when fibers are optimally distributed and oriented in the structure with respect to the assumedmeasure of structural behavior. This optimal lay- out can be obtained using the proposed evolutionary optimization algorithm, which can be a very effective tool for finding a reasonable solution to a gi- ven problem. Thus, this algorithm can constitute an alternative technique for 534 K. Dems, J. Wiśniewski the classical gradient-oriented methods applied in optimization of structural elements, or can supplement them. The presented analysis can be treated as a starting point for computer- oriented optimal design procedures of real structures made of composite ma- terials subjected to actual loading conditions. Such a procedure can allow for avoiding expensive experimental testing, which can be reduced to the final phase of structural design. Acknowledgements This research work has been supported by Grant No. 50106032/3955 of the Mi- nistry of Science and Higher Education in Poland. References 1. Adams D.F., Doner D.R., 1967a, Longitudinal shear loading of a unidirec- tional composite, Journal of Composite Materials, January, 4-17 2. Adams D.F., Doner D.R., 1967b, Transverse normal loading of a unidirec- tional composite, Journal of Composite Materials, April, 152-164 3. Dems K., 1986, Sensitivity analysis and optimal design for fiber reinforced composite disks, Structural Optimization, 11, 178-186 4. Dems K., Wiśniewski J., 2006, Design of fiber layout in multi-layer compo- sites using hybrid algorithms,Foundations of Civil and Environmental Engine- ering, 6, 53-77 5. DemsK.,Wiśniewski J., 2007, Stiffness optimization of fiber-reinforced com- posites,Proc. 17-th International Conference on Computer Methods in Mecha- nics, CD, 1-8 6. Findeisen W., Szymanowski J., Wierzbicki A., 1980, Teoria i metody obliczeniowe optymalizacji, PaństwoweWydawnictwo Naukowe, Warszawa [in Polish] 7. Foye R.L., 1966, An evaluation of various engineering estimates of transverse properties of unidirectional composites, Proc. 10-th Nat. Symp. Soc. Aerosp. Mater. Process Eng., G31-G42. 8. GermanJ., 1996,Podstawymechaniki kompozytówwłóknistych,Wydawnictwa Politechniki Krakowskiej, Kraków [in Polish] 9. Halpin J.C., Tsai S.W., 1969, Effects of environmental factors on composite materials,AFML-TR, June, 67-423 Optimal design of fiber-reinforced composite disks 535 10. Kiciak P., 2000,Podstawy modelowania krzywych i powierzchni – zastosowa- nie w grafice komputerowej, WydawnictwaNaukowo-Techniczne,Warszawa [in Polish] 11. Michalewicz Z., 1996, Genetic Algorithms + Data Structures = Evolution Programs, Springer-Verlag, Berlin 12. Tsai S.W., Wu E.M., 1971, A general theory of strength for anisotropicma- terials, Journal of Composite Materials, January, 58-80 13. Zienkiewicz O.C., 1971,The Finite ElementMethod in Engineering Science, McGraw-Hill, London Optymalne projektowanie tarcz kompozytowych wzmocnionych włóknami Streszczenie Praca dotyczy zagadnień modelowania, analizy i optymalizacji zachodzących wprocesie projektowaniapłaskich, dwuwymiarowychelementówkonstrukcyjnychwy- konanych zmateriałów kompozytowych i obciążonych statycznie siłami działającymi w ich płaszczyźnie. Rozpatrzono problem projektowania optymalnego ułożenia włó- kienwzmacniającychwkompozyciepodkątemuzyskaniaprzezelementkonstrukcyjny wykonany z tegomateriałuwymaganychwłasnościmechanicznych.Rozważania doty- czyły sposobu projektowania oraz generowania rodziny prostoliniowych, jak również krzywoliniowychwłókien wzmacniających. Do poszukiwania optymalnych rozwiązań zastosowano opracowaną do tego celu metodę optymalizacyjną opartą na algorytmie ewolucyjnym. Rozpatrywany w pracy problem zilustrowano przykładami numerycz- nymi. Manuscript received May 11, 2009; accepted for print June 8, 2009