Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 41, 2, pp. 365-381, Warsaw 2003 OPTIMUM DESIGN OF COMPOSITE CHANNEL CROSS-SECTION COLUMNS UNDER COMPRESSION Arkadiy Manevich Department of Computational Mechanics and Strength of Structures Dniepropetrovsk National University, Ukraine Zbigniew Kołakowski Department of Strength of Materials and Structures, Technical University of Łódź, Poland e-mail: kola@orion.p.lodz.pl Thin-walled open cross-section bars are effective structural members (with respect to overall stability). But the weight efficiency of bars with flat walls is constrained by their low stiffness in local buckling. The use of composites and multi-layer materials is one of perspective ways to overcome this shor- tage. The minimum-weight problem for such laminated bars has not been studied until now. In the paper a solution to the structural optimization pro- blem for three-layer channel cross-section columns under compression is pre- sented. The optimization problem is formulated as a nonlinear programming problem. Basic constraints are conditions of overall and local stability. In ad- dition, strength, wrinkling and geometrical constraints (onminimal thickness of skin layers andmaximal total thickness) have been accounted for. For the givenmaterial properties the optimal dimensions of the cross-section and the optimal thickness ratio (for the steel and composite layers) are determined. The three-layermembers are shown to be muchmore efficient in comparison with columnsmade of homogeneousmaterials. The sensitivity of optimal pro- jects to changes in the material properties (in particular, to changes of the modulus of elasticity in tension and in shear) is also studied. Key words: optimization, thin-walled structures, channel, composite 1. Introduction The minimum weight design problems for thin-walled bars-beams under compression or bending with respect to their stability were investigated in a 366 A.Manevich, Z.Kołakowski few papers already in the 60-70-s on the basis of various theoretical models. A review of the literature concerning the optimal design and some newworks in this field can be found, in Adali (1995), Birger andPanovko (1968),Walker et al. (1996), Życzkowski and Gajewski (1983), and in proceedings of recent international conferences (Zaraś et al., 2001; Gutkowski andMróz, 1997). Thin-walled bars of open cross-section are effective structural members (with respect to overall stability), however the weight efficiency of bars with flat walls is constrained by their low stiffness in local buckling. The use of composites and multi-layer members is one of perspective ways to overcome this shortage. Over the past years an interest was exhibited to new types of structural elements, in particular, to metal-composite-metal members which were developed for applications in manufacturing of parts for cars, train and other machines (Ashafi et al., 2000). But the minimum-weight problem for such laminated bars with respect to their stability and strength has not been studied systematically until now. In the paper a solution of the structural optimization problem for com- posite and three-layer channel cross-section columns (metal-composite-metal members) under compression is presented. The solution is based on the clas- sical plate theory and the linear stability theory and employs two program packages: • WARST package for calculation of critical stresses of compressed and bent plate assemblages composed of an arbitrary number of orthotropic layers (see Kołakowski and Kowal-Michalska, 1999) • LMRG package for solution of the general nonlinear programming pro- blem by the linearized method of reduced gradient (Manevich, 1979). Results of the solution are verified by analyzing the obtained optimal channels using ANSYS 5.7 with account of shear deformation. 2. Formulation of the problem and method of solution A simply supported thin-walled member of channel cross-section loaded centrally by the compressive force P is considered (Fig.1). Two cases of ma- nufactured channels are studied: case I – the member is made of a composite material (Fig.1a) Optimum design of composite channel... 367 case II – the member is constructed from two steel skins and a composite layer between them (Fig.1b). Fig. 1. Two cases of manufactured channels: composite material (a); steel+composite+steel (b) The length L and the material properties of all layers are considered to be given, and the optimal dimensions of the cross-section and the optimal thickness ratios (for the steel and composite layers) are to be determined. The composite is considered as a structurally orthotropic material, and the classical plate theory is employed in the buckling analysis. The optimization problem is formulated as a nonlinear programming pro- blem. For a given value of the weight per unit length Q the cross-section geometrical parameters for which the critical load is maximal are sought. The bendingmode in the symmetry plane, flexural-torsional modes and local mo- des (Fig.2) are taken into account, and the minimal value of critical stresses for these three types of buckling modes determine the critical load. In addi- tion, the strength,wrinkling (local buckling of external layers) andgeometrical constraints were accounted for. Fig. 2. Considered buckling mode: (a) overall flexural mode; (b) flexural-torsional mode; (c) local mode Critical stresses for all bucklingmodeswere calculatedbyprogramWARST whichwasworked out for thin-walled girders built of plateswith closed or open 368 A.Manevich, Z.Kołakowski cross-sections. Theplate elements under consideration can bemulti-layer walls made of orthotropic materials. The classical laminated plate theory (Jones, 1975) isused in the theoretical analysis, the effect of sheardeformation through the thickness of the laminate is neglected. Thematerials they aremade of are subjected to Hooke’s law. In order to account for allmodes of global and local buckling, aplatemodel of thin-walled structures is employed. Sinusoidal deflection along the length is assumed and the amplitude changing in the transverse direction is determined by numerical integration. During the integration of the equations, Godunov’s orthogonalization method is employed. Themost important advantage of this method is that it enables us to describe the complete range of behaviour of thin-walled structures from all global (flexural, flexural-torsional, lateral, di- stortional buckling and their combinations) to the local stability (Kołakowski and Kowal-Michalska, 1999). The column global buckling occurs at one sinu- soid half-wave along the column length, whereas the local buckling takes place at the number of half-waves m> 1 with bk ≪ ℓ. Due to the symmetry or antisymmetry conditions only a half of the pro- file was considered. For the overall bending mode (in the symmetry plane, see Fig.2a) at the point x = 0 (center of the wall) the symmetry condition was imposed, and the number of halfwaves m was assumed equal to 1. For flexural-torsional modes (Fig.2b) the antisymmetry condition and m=1we- re assumed. For local modes (Fig.2c) the symmetry condition was imposed but there was assumed that m > 1 (for more details see Kołakowski and Kowal-Michalska, 1999; Kołakowski et al., 1999). Critical stresses for wrinkling of the external layer (skin) were calculated according to (Birger and Panovko, 1968) σw =              0.91 3 √ E1E2G2 1−ν21 when t2 2t1 > 0.4 3 √ E1E2 (1−ν21)G 2 2 0.58 √ 2E1E2t1 (1−ν21)t2 when t2 2t1 < 0.4 3 √ E1E2 (1−ν21)G 2 2 (2.1) where E1 is the elasticity modulus of the skin, ν1 Poisson’s ratio of the skin, E2, G2 are the effective elasticity modulus and shear modulus of the core, respectively. The strength constraint was taken in the form σ ¬ σy, where σy is the yield stress. Geometrical constraints were imposed on the minimal thickness of skin layers and the maximal total thickness. The thickness constraints had the form: t1 ­ t1,min and t=2t1+ t2 ¬ tmax. Optimum design of composite channel... 369 The nonlinear programming problemwas solved by the linearized method of reduced gradient (Manevich, 1979). This method, which uses the idea of change of the independent variables set in the vicinity of the allowable region boundary by means of linear operations with sensitivity matrices, effectively overcomes familiar difficulties arising at solving nonlinear programming pro- blems (in particular, connected with zigzag-type motion in the boundary vi- cinity). As a rule, 20-30 iterations were required to achieve the optimumwith a relative error of order 0.001. For generality of the analysis, nondimensional parameters of the load P∗, weight Q∗ and stress σ∗ were used P∗ = P ·106 EL2 Q∗ = Q ·103 ρgL2 σ∗ = σ ·103 E (2.2) where E stands for Ex in case I (modulus of the orthotropic composite in the longitudinal direction) and for E1 – in case II (modulus of the isotropic external layer), ρ is the density (in the case of a three-layered channel – the density of the external layer). The following effective parameters of load and weight were used for comparison of the optimal members made of various materials P∗eff =P ∗ E E0 Q∗eff =Q ∗ ρav ρ0 (2.3) where E0, ρ0 are the elasticity modulus and density of an isotropic mate- rial which the given composite material is compared with, ρav is the average density for the three-layered channel. Two equivalent formulations of the optimization problem are possible (in the given nondimensional parameters): • for a given Q∗ two nondimensional cross-section parameters (b1/L,b2/b1) are determined yielding themaximum of P∗ (minimal va- lue of critical loads for all possible bucklingmodes) • for a given P∗ the optimal geometrical parameters are determined yiel- ding the minimum of Q∗. In thiswork, the first approachwas chosen, as a rule,which ismore convenient for the sensitivity analysis. All optimal nondimensional parameters of the compressed member are defined by specifying the single nondimensional parameter Q∗ (for given pro- perties of thematerial); when the length Landmodulus Ex (or E1) are given, and the dimensions of optimal cross-sections are to be determined. 370 A.Manevich, Z.Kołakowski For testing the algorithm, calculations of optimalmembers of the isotropic material were carried out. Solution to this problemwas obtained inManevich and Raksha (2000) with using the Vlasov theory for overall critical stresses (flexural and torsional-flexural modes) and the analytical solution of the buc- kling problem for local modes (by conjugation of analytical solutions for all constituent plates). Results of both solutions – Manevich and Raksha (2000) and obtained with using theWARST package – coincide for the isotropic ma- terial. 3. Results of the solution 3.1. Composite channel Two composites with the following characteristics are considered: • compositeA: Ey Ex =0.0796 G Ex =0.04316 νxy =0.3 Ex =1.393 ·1011 Pa ρ=1.56 g cm3 • compositeB (6-layer graphite/epoxy composite): Ey Ex =0.1417 G Ex =0.05157 νxy =0.3 Ex =0.37385 ·1011 Pa ρ=2.2 g cm3 Ex, Ey, G are, correspondingly, the modulus of elasticity in the longitudi- nal and transverse directions and the modulus of shear, ρ is the density (Exνyx =Eyνxy). Comparison of the optimal cross-sections for these materials and for the isotropicmaterial enables us to estimate the influence of ratios Ey/Ex,G/Ex, and also Ex/E0 on optimal parameters and (with account of ρ/ρ0) their com- parative efficiency. Values of Q∗ were chosen in the range (0, ...,1.2) with the step 0.1 (in this range the critical stresses for optimal members from usual materials do not exceed the yield stress). Optimum design of composite channel... 371 The active constraints are found to be the buckling constraints on the torsional-flexural mode and one (or two) local mode(s); note that the critical stresses for the overall flexural buckling were higher by 15-25% (similarly to the optimal channels of the isotropic material (Manevich andRaksha, 2000)). In Table 1 there are compared cross-section dimensions, critical loads and critical stresses for thin-walled members with the length L = 1m, obtained for three cases: a) isotropic material – steel with Young’s modulus E0 = 2 · 1011Pa and density ρ0 =7.8g/cm3, for Q∗ =0.2 b) composites A and B at the same Q∗ value (i.e. with the same cross- section area) c) composite A at Q∗ = 0.865 and composite B at Q∗ = 0.71, i.e. at the sameweight (for the effective weight parameter Q∗eff according to (2.2), equal to 0.2). Table 1.Comparison of the optimal cross-section of thin-walled channels made of steel and composites (length L=1m) h b1 b2 P ∗ P σ Material Q∗ [mm] [mm] [mm] [kN] [MPa] Steel 0.2 1.4 78.2 34.1 0.190 38 189.8 CompositeA 0.2 1.82 58.35 25.76 0.099 13.8 68.7 CompositeA 0.865 4.8 96.3 42.1 1.099 153 177 CompositeB 0.2 1.75 61.1 27.0 0.107 4.004 20.02 CompositeB 0.71 4.03 94.0 41.1 0.864 32.3 45.51 The optimal profiles of the isotropic member and channels of composites with the same area (Q∗ = 0.2) for L = 1m are shown in Fig.3, and the optimal profiles of the same weight (Q∗eff =0.2) are compared in Fig.4. As in this analysis we used the classical plate model based on Kirchhoff’s hypothesis, it is interesting to test the results obtained by comparison with more precise models accounting for the shear deformation (which can have si- gnificant effect on buckling loads (Whitney, 1987)). Calculations of the critical stresses of the optimal channels by the FEMwith using ANSYS 5.7 were car- ried out. The shell finite elements (SHELL 91) were employed. On the loaded edges forces were applied at the gravity center, and boundary conditions of the simple support were satisfied. 372 A.Manevich, Z.Kołakowski Fig. 3. Optimal profiles of channels with the same area (Q∗ =0.2) for L=1m: (a) steel; (b) compositeA; (c) compositeB Fig. 4. Optimal profiles of channels with the same weight (Q∗ =0.2) for L=1m: (a) steel; (b) compositeA; (c) compositeB Table 2. Comparison of the critical loads for two optimal channels obta- ined in the present analysis and by ANSYS 5.7 Present analysis ANSYS 5.7 Optimal Overall Overall variants Q∗ Local (Euler) Local (Euler) of channel buckling buckling buckling buckling [kN] [kN] [kN] [kN] CompositeA 0.865 153 (m=4) 180.3 145.9 (m=4) 181.7 CompositeB 0.71 32.3 39.2 30.5 39.5 Optimum design of composite channel... 373 Typical results of such comparison are presented in Table 2, where the values of critical loads for the local mode and the overall flexural (Euler) mode, obtained in the present analysis and by using ANSYS 5.7 are given for two optimal channels fromTable 1. We see that the critical loads in bothmodels are close to those in local and overall modes (for the local modes ANSYS 5.7 gives a little lesser value, for the Euler overall modes – a slightly greater value). So, we can conclude that, at least for the adopted material properties, our analysis is valid. The optimal configurations of channels of the isotropic material and the compositewith the samearea are essentially different. The composite channels have a larger thickness, but lesser width of the web and flanges. If to compare channels of the same weight then all dimensions of the composite channel become larger than those of the isotropic channel – the thickness gets larger by several times. The comparative efficiency of the composite channels is determined, first of all, by Ex/E0 and ρ/ρ0 ratios.The optimalmember of compositeA, forwhich the modulus Ex lower than that of steel only by 30%, and density – almost by 4 times, has the critical force 4 times larger than that of the optimal steel channel of the sameweight. At the same time the channel of compositeB, for which themodulus Ex equals about 0.19E0, turns out to be less efficient than the isotropic steel channel. It has been revealed inManevich andRaksha (2000) that optimal channels of an isotropicmaterial have apractically constant flangewidth to awebwidth the ratio b2/b1 (in the whole range of Q∗ considered) equal to b2/b1 =0.42- 0.43. For the optimal composite channels this ratio it also almost constant, and practically the same (b2/b1 =0.43-0.44). Other nondimensional parameters depend upon the parameter Q∗. In Fig.5 cross-sectional nondimensional parameters are given versus Q∗ for com- positeA and for the isotropic material. Note that these curves do not depend on the value Ex and are determined only by the ratios Ey/Ex, G/Ex and therefore are close to composites A andB (but differ from the corresponding dependencies for the isotropic material). InFig.6 curves for the parameter t/Lversus Q∗ for the isotropic and com- posite materials are presented. Values of the nondimensional load parameter P∗ and critical stresses σ∗ versus Q∗ are given in Fig.7 and Fig.8. The presented optimal nondimensional parameters enable us to determine all cross-sectional dimensions when L and Ex are given. Let us estimate the influence of the elasticity modulus in two directions and shearmodulus on the critical load of optimal composite channels. Nondi- 374 A.Manevich, Z.Kołakowski Fig. 5. Cross-sectional nondimensional parameters b1/L, b2/L, t/b1, t/b2 versus Q∗ for compositeA (broken line) and for the isotropic material (full line) Fig. 6. Parameter t/L versus Q∗ for the isotropic and composite materials mensional load parameter (2.2) depends upon two parameters of thematerial: η1 = Ey/Ex, η2 = G/Ex. Calculations of the sensitivity of the critical load to changes in these parameters for the optimal channel of composite A at Q∗ = 0.865 (i.e., Q∗eff = 0.2) were carried out. Calculations were performed in two variants: • cross-section remained invariable; • cross-section dimensions were changed so that the cross-section became the optimal for the composite with changed properties (i.e. at simulta- neous optimization of the channel with the same area). Optimum design of composite channel... 375 Fig. 7. Nondimensional load parameter P∗ versus Q∗ for the isotropic and composite materials Fig. 8. Nondimensional critical stresses σ∗ versus Q∗ for the isotropic and composite materials In the first case the sensitivity coefficients were calculated separately for those buckling modes that govern the carrying capacity – the overall mode (torsional-flexural) and the local mode with minimal critical stresses. In the second case the sensitivity coefficients referred simultaneously to all critical modes. Formore generality, the relative sensitivity coefficients, i.e. coefficients k1 and k2 in the relations δP P = k1 δη1 η1 δP P = k2 δη2 η2 were calculated. These coefficients are presented in Table 3. In the case of an invariable cross-section the influence of the parameters η1 and η2 on the overall stability is found to be lesser than on the local buckling (especially for the parame- ter η1). So, when η1 is increased by 10% the local critical load increases by 4.4%, but the overall critical load – only by 0.58%; for the parameter η2 376 A.Manevich, Z.Kołakowski these values equal 4.4% and 1.7%, respectively. This is clearly since the overall buckling is governed mainly by the modulus Ex. Table 3. Sensitivity coefficients for the critical load with respect to para- meters η1 and η2 Coefficient Invariable cross-section Optimized overall buckling local buckling cross-section k1 0.058 0.44 0.086 k2 0.17 0.44 0.214 If the cross-section is optimized simultaneously with the varying parame- ters η1 and η2 the sensitivity coefficients lie between their values for theoverall and local modes obtained for the fixed cross-section. The parameter η2 has more influence on the critical load than η1: when η1 is increased by 10% the critical load increases by 0.86%,when η2 is increased by 5% the load increases by 2.14%. 3.2. Three-layered channel Calculations were carried out for the following core material: 6- layer graphite-epoxy material with the parameters Ey/Ex = 0.0796, G/Ex = 0.04316, Ex = 1.393 · 1011Pa, ρ = 1.8g/cm3 (material A, Sec- tion 3.1). Values of the weight parameter Q∗ were considered in the interval (0,...,0.35) for which the use of the linear stability theory for elastic materials can be justified. The computations showed that the strength- and/or geometrical constra- ints are of principal importance for the optimization problem. In distinction from the isotropic thin-walled bar, the solution to the weight optimization problem with respect only to buckling constraints for usual elastic materials has no sense (it results in unrealistic thin external layers and very high criti- cal stresses considerably exceeding the yield stress for usual materials). Only when the geometrical and/or strength constraints are imposed we come to realistic optimal projects. We considered the following values of the relative minimal skin thickness: t∗min = t1,min · 10 3/L = 0.1; 0.15; 0.2; 0.25; 0.3 and dimensionless yield stress values (for the skin): σ∗y =σy ·10 3/E1 in the range (1.0; 2.2). Optimal bars with constraint on minimal thickness of external layers Consider the results of the solution obtainedwith the only geometrical con- straint on theminimal thickness of skin layers t∗min =0.2mm.This constraint Optimum design of composite channel... 377 is always active, so for all these optimal projects t1/L = tmin/L = 0.0002. In Table 4 the optimal nondimensional cross-section parameters and critical values of the load and stresses for several values of the weight parameter Q∗ are presented. Table 4. Optimal nondimensional parameters of the 3-layered channel (t∗min = t1,min ·10 3/L=0.2) Geometrical parameters Critical load and stress values Q∗ b1/L b2/L h/L P ∗ σ∗av σ ∗ 1 0.10 0.0375 0.0296 0.00357 0.237 0.687 0.940 0.15 0.0468 0.0341 0.00492 0.5 0.884 1.226 0.20 0.0540 0.0376 0.00614 0.832 1.049 1.464 0.0602 0.0404 0.00686 1.222 1.193 1.673 0.25 0.1170 0.0512 0.00409 1.289 1.434 1.975 0.30 0.1246 0.0544 0.00483 1.792 1.591 2.204 For Q∗ = 0.25 there exist two local optima, both presented in Table 4. These local optimahave essentially different cross-section dimensions –narrow andwideweb and flanges, and correspondingly large and smallwall thickness, but the values of the critical loads are rather close (two local optima also exist for other Q∗ values, but for lower Q∗ the first optimum is global, for the larger ones – the second, and transition from the first local optima to the second occurs approximately at Q∗ = 0.25). Note that at the first local optimum three-layered channels have equal critical loads for the flexuralmode (in the symmetry plane) and for the flexural-torsional mode. At the second local optimumthebars have equal critical loads for theflexural-torsionalmode and one of local modes (critical stresses for the flexural mode are found to be larger by up to 20%). In the last three columns of Table 4 the values of the load parameter P∗, nondimensional average (over the cross-section) stress σ∗av and the critical stress in the external (steel) layer σ∗1 are given.We see thatwith an increasing Q∗ the critical load P∗ quickly raises aswell as thecritical stress in theexternal layer, and this stress becomes larger than the yield stress for usual steels. In Table 5 cross-section dimensions, critical loads for the optimal three- layered channel of the length L = 1m for Q∗ = 0.2 and the optimal single- layered (steel, compositeA) channel of the same weight are compared. 378 A.Manevich, Z.Kołakowski Table 5.Cross-sectiondimensions and critical loads for the optimal single- and three-layered channel; Q∗ =0.2, length L=1m h b1 b2 P ∗ P Type of bar [mm] [mm] [mm] [kN] 1-layered, steel 1.4 78.2 34.1 0.19 38 1-layered, composite 4.8 96.3 42.1 1.099 153 3-layered, steel+composite 6.14 54.0 37.6 0.832 166.4 The optimal configurations of laminated channel columns essentially differ from those made of a homogeneous material. Big advantage of the optimal three-layered channel is due to high characteristics of the composite layer, i.e. low density and high elasticity modulus Ex. In order to test the results obtained on the basis of the Kirchhoff hy- pothesis, calculations were carried out with using ANSYS 5.7. Spatial (3D) elements were employed for the composite (core) and shell elements – for the steel (external) layers. For the optimal 3-layered channel, presented in the last line of Table 5, the critical force inANSYS 5.7was 167.3kN, i.e. the difference was less than 0.4%. Influence of constraint on minimal thickness upon optimal bars Let us consider now results of the solution for various values of t∗min. The characteristics of the optimal channels for the load parameter Q∗ = 0.2 at various values of t∗min are presented in Table 6. Table 6. Optimal channels for the load parameter Q∗ = 0.2 at various values of t∗min = t1,min ·10 3/L Geometrical parameters Critical load and stress values t∗min b1/L b2/L h/L P ∗ σ∗av σ ∗ 1 0.10 0.0606 0.0370 0.00663 0.867 0.971 1.376 0.15 0.0572 0.0374 0.00637 0.856 1.017 1.431 0.20 0.0540 0.0376 0.00614 0.832 1.049 1.464 0.25 0.0511 0.0376 0.00592 0.799 1.069 1.481 0.30 0.1052 0.0467 0.00263 0.667 1.275 1.665 An increase in theminimal thickness of the external (skin) layer results in perceptible changes in the geometrical parameters (b1/L,h/L) and reduction of the critical load. So,we can draw the conclusion that theweight efficiency of the three-layered channel largely depends upon the constraint on theminimal skin thickness. Optimum design of composite channel... 379 Influence of strength constraint upon optimal parameters Calculations of optimal channels with constraints on the ultimate stress were carried out. We assumed that Q∗ = 0.3, t∗min = 0.2 and varied the values σ∗y =σy ·10 3/E1 in a certain range. The results of the calculations are presented in Table 7. Table 7. Optimal channels for Q∗ = 0.3, t∗min = 0.2 at various values of σ∗y =σy ·10 3/E1 Geometrical parameters Critical load and stress values σ∗y b1/L b2/L h/L P ∗ σ∗av σ ∗ 1 2.2 0.1246 0.0544 0.00485 1.792 1.591 2.2 2.0 0.1116 0.0487 0.00558 1.675 1.436 2.0 1.8 0.0520 0.0416 0.00949 1.639 1.277 1.8 1.6 0.0381 0.0387 0.01140 1.488 1.131 1.6 1.4 0.0313 0.0360 0.01290 1.319 0.988 1.4 1.2 0.0264 0.0330 0.01460 1.143 0.846 1.2 1.0 0.0220 0.0298 0.01680 0.964 0.704 1.0 The strength constraint becomes active when σ∗y < 2.20 (for given Q ∗ and t∗min). At σ ∗ y ≈ 1.9 transition fromone local optimumto the other occurs, with a snap in the optimal parameters. The strength constraint strongly affects the optimal projects, and it should be taken into account. The same regards the geometrical constraints. Summarizing, we can conclude that three-layered members under com- pression may be much more efficient in comparison with columns made of homogeneous materials. References 1. Adali S., 1995, Lay-up optimization of laminated plates under buckling lo- ads, In:Buckling and Postbuckling of Composite plates, edit. G.J. Turvey, I.H. Marshall, Chapman andHall, 329-365 2. Ashafi N., Langstedt, Andersson C.-H., Ostergren N., Hakansson T., 2000, A new lightweight metal-composite-metal panel for applications in the automotive and other industries,Thin-Walled Structures, 36, 289-310 380 A.Manevich, Z.Kołakowski 3. Birger I.A.,PanovkoYa.G. (Edit.), 1968,Strength, Stability, Oscillations, Vol. 2, Moscow,Mashinostroenie, 211-242 (in Russian) 4. Gutkowski W., Mróz Z. (Edit.), 1997,WCSMO-2. Structural and multi- disciplinary optimization, Proc. of the Second World Congress, 2, Zakopane, Poland, IFTR,Warsaw, Poland 5. Jones R.M., 1975, Mechanics of Composite Materials, International Student Edition, McGraw-Hill Kogakusha, Ltd., Tokyo, p.365 6. Kołakowski Z., Kowal-MichalskaK. (Edit.), 1999,Selected Problems of Instabilities in Composite Structures, Technical University of Łódź, A series of monographs, p.222 7. Kołakowski Z., Królak M., Kowal-Michalska K., 1999, Modal inte- ractive buckling of thin-walled composite beam-columns regarding distortional deformations, Int. J. Eng. Science, 37, 1577-1596 8. ManevichA.I., 1979,Stability andOptimumDesign of Stiffened Shells, Kiev- Donetsk, p.152 (in Russian) 9. Manevich A.I., Kołakowski Z., 2001, Weight optimization of compressed thin-walled bars of composite materials, In: Theoretical Foundations of Civil Engineering, Vol. IX, Edit.W. Szcześniak, OWPW,Warsaw, 109-114 10. Manevich A., Raksha S., 2000, Optimal centrally compressed bars of open cross-section, In: Theoretical Foundations of Civil Engineering, Vol. VII, Edit W. Szcześniak, OWPW,Warsaw, 484-489 11. WalkerM., Adali S., VerijenkoV.E., 1996,Optimal design of symmetric angle-ply laminates subject to nonuniform buckling loads and in-plane restra- ints,Thin-Walled Structures, 26, 1, 45-60 12. Whitney J.M., 1987, The effect of shear deformation on the bending and buckling of anisotropic laminated plates, In: Composite Structures, 4, Edit. I.H. Marshall, Elsevier Applied Science, London, 1.109-1.121 13. Zaraś J., Kowal-Michalska K., Rhodes J., (Edit.), 2001, Thin-walled structures. Advances and developments, Proc. of the Third Internat. Confer. on Thin-Walled Structures, Elsevier 14. ŻyczkowskiM.,GajewskiA., 1983,Optimal structuraldesignwith stability constraints, In: Collapse. The Buckling of Structures in Theory and Practice, Edit. J.M.T. Thompson, G.W. Hunt, Cambridge Univ. Press Optimum design of composite channel... 381 Optymalne projektowanie kompozytowych słupów o przekroju otwartym poddanych ściskaniu Streszczenie Cienkościenne słupy o otwartymprzekroju poprzecznym są efektywnymi elemen- tami konstrukcjami (z uwzględnieniem stateczności globalnej). Bardziej wydajne wa- gowo słupy o ścianach płaskichmają niższą sztywnością na wyboczenie lokalne. Uży- cie kompozytów imateriałówwielowarstwowych jest jedną z perspektywicznych dróg podniesienia ich obciążalności.Problemminimalnejwagi takich laminatowych słupów nie był do chwili obecnej jeszcze rozpatrywany. W prezentowanej pracy przedsta- wiono rozwiązanie zagadnienia optymalizacji konstrukcji trzywarstwowego ceownika poddanego ściskaniu. Zagadnienie optymalizacji sformułowano jako problem progra- mowania nieliniowego. Podstawowymi ograniczeniami są warunki globalnej i lokal- nej stateczności. Ponadto uwzględnionowytrzymałość, pomarszczenie i geometryczne ograniczenia (na minimalną grubość okładzin oraz maksymalną całkowitą grubość). Dla zadanych stałych materiałowych zostały określone optymalne stosunki wymia- rów przekroju i grubości (dla stalowej i kompozytowych warstw). Trzywarstwowa konstrukcja jest bardziej wydajna na ściskanie niż słup zmateriału homogenicznego. Analizowano także czułość optymalnej zmiennej projektowej na własności materiało- we (w szczególności zmianamodułów sprężystości na rozciąganie i ścinanie). Manuscript received November 26, 2002; accepted for print February 4, 2003