Professional Documents
Culture Documents
DOI 10.1007/s00158-007-0207-2
RESEARCH PAPER
Received: 30 July 2007 / Revised: 24 September 2007 / Accepted: 17 October 2007 / Published online: 12 January 2008
© Springer-Verlag 2007
Abstract In this study, two optimality criteria are pre- single element. Plane stress finite elements have been
sented for optimum design of composite laminates used by Kasap and Oral (1997) for weight optimiza-
using finite element method. Thickness of the layers tion of composite laminates with curved fibers. Parnas
and fiber orientation angles in each finite element are et al. (2003) presented a methodology for the opti-
considered as the design variables. It will be shown mum design of laminated composites with curved fiber
that the optimum design of composite laminates with courses using Bezier curves for fiber orientations and
varying fiber orientations and layers thicknesses may be Bezier surfaces for layer thicknesses. Nonlinear pro-
found by using these optimality criteria in an efficient gramming methods (Lin and Yu 1991; Fukunaga and
way, without performing the sensitivity analysis. Vanderplaats 1991), multilevel optimization algorithms
(Watkins 1987; Weiji and Boohua 1986; Watkins and
Keywords Design optimization · Optimality criteria · Morris 1987; Schmit and Mehrinfar 1982; Kam and
Composite structures · Composite laminates · Lai 1989; Sadr et al. 1989; António et al. 1995; Mota
Thickness optimization Soares et al. 1995), and genetic algorithms (Le Riche
and Haftka 1996; Kodiyaiam et al. 1996) have also
been used for optimization of composite structures. For
more studies on optimization of laminated compos-
1 Introduction ite structures, the reader is referred to the references
at the end of this article (Hyer and Lee 1991; Fang
Nowadays, composite laminates are increasingly used and Springer 1993; Morton and Webber 1997; Burns
in areas where weight saving is crucial. Laminates are and Cherkaev 1997; Kalamkarov 1997; Verijenko and
usually optimized by varying the layer thicknesses and Johnson 1996; Landriani and Rovati 1991; Muc 1988;
fiber angles to improve the performance of the struc- Rao et al. 1991; Fukunaga and Vanderplaats 1990;
tures at the lowest possible weight for the cases where Kam and Snyman 1991; Miravete 1990; Gürdal and
manufacturing cost is not a primary consideration. Sev- Olmedo 1993; Olmedo and Gürdal 1993; Hyer and
eral studies related to the optimization of laminated Charette 1991).
composite structures are available in the literature. The efficiency of the nonlinear mathematical pro-
Fukunaga and Sekine (1993) derived an optimal re- gramming methods of optimization depends very
lation for fiber angles and layer thickness ratios in a much on the performance of the sensitivity analysis
(Haftka and Prasad 1981; Adelman and Haftka 1986;
Haftka and Adelman 1989; Zynczowski 1992; Grandhi
P. Khosravi (B) · R. Sedaghati 1993). Design sensitivity analysis for rectangular or-
Department of Mechanical and Industrial Engineering,
Concordia University, Montreal, Quebec, Canada thotropic laminated plates have been presented by
e-mail: peyma_kh@encs.concordia.ca Pedersen (1987) and Reiss and Ramachandran (1987).
R. Sedaghati Adali (1983) and Kengtung (1986) presented the de-
e-mail: sedagha@encs.concordia.ca sign sensitivities with respect to the fibers angles.
160 P. Khosravi, R. Sedaghati
where λ is the Lagrange multiplier. Minimization of L recurrence relation is found by multiplying both sides
with respect to the design variables tip and θip (KKT of (10) by tip and taking the rth root:
conditions, (Arora 1989)) leads to
(tip )η+1 = (tip )η . (λētip )1/r
η (15)
∂L
=0 (6) where η + 1 and η are the iteration numbers and r is
∂tip
the step size parameter. The value of λ at the optimal
design is determined by minimizing the sum of the
∂L squares of the residuals at iteration η:
=0 (7)
∂θip
Resη = (1 − λētip )2 (16)
From (3), (5), and (6) we have i, p
⎛ ⎞
where i, p means that the sum is on all the thickness
∂e m
∂ ∂u
Ai ρ p − λ ⎝ ⎠=0
i j
+ (8) design variables in all elements. Thus,
∂tip j=1
∂u j ∂tip
i, p ētip
λ= 2
(17)
where m is the number of the nodal displacement de- i, p ētip
∂
grees of freedom. Considering (4), the term ∂u vanishes
j Substituting (17) into (15), we finally obtain
and (8) is simplified to
1/r
∂ei i, p ētip
Ai ρ p − λ =0 (9) (tip )η+1 = (tip )η . 2
. ētip (18)
∂tip i, p ētip η
which is actually the “minimum strain energy” criterion Equations (11) and (14) contain the derivatives of the
in the element level. Equation (12) can be rewritten strain energy of the ith element with respect to the
in the following form to make it easier for recurrence thickness and fiber angle of layer p. To find these deriv-
relation: atives, we start by writing the force–strain relationship
in the xyz local coordinate system of the ith element
1 = (1 + ēθip ) (13) (Berthelot 1999):
⎧ ⎫ ⎧ ⎫
◦
where ⎪ Nx ⎪
⎪ ⎪ ⎪ εx ⎪
⎪
⎪ ⎪
⎪
⎪ ⎪ ⎪ ◦ ⎪⎪
∂ei ⎪ Ny ⎪
⎪ ⎪
⎪
⎪
⎪ εy ⎪
⎪ ⎪
⎪
ēθip = ⎪
⎪ ⎪
Nxy ⎪ ⎡ ⎤ ⎪ ⎪ γxy ⎪
◦ ⎪
(14) ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
∂θip ⎨ ⎬ An Bn [0]3×2 ⎨ ◦⎪ ⎬
Mx ⎣ ⎦ κ
= Bn Dn [0]3×2 . x
(21)
Equations (10) and (13) are the optimality criteria, ⎪
⎪ My ⎪⎪ ⎪
⎪ κ y◦ ⎪
⎪
⎪
⎪ ⎪ [0] [0] F ⎪ ⎪
which should be satisfied to get the optimum solution. ⎪
⎪ Mxy ⎪
⎪
⎪
2×3 2×3 n ⎪
⎪
⎪ κxy◦ ⎪⎪
⎪
⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪
An iterative relation (called “recurrence relation”) is ⎪
⎪ Qy ⎪⎪ ⎪
⎪ γ̄ ⎪
⎪
⎩ ⎭ ⎪
⎩
yz ⎪
⎭
usually used to satisfy the optimality criteria. The first Qx γ̄xz
162 P. Khosravi, R. Sedaghati
n
∂ Dn 1
= z − zi(k−1) Eik
2 2 4 Design optimization methodology
∂tip 2 ik
k=1
p In this study, the objective of the optimal design is to
− zik − zi(k−1) Eik + zip
2 2 2
Eip (36) find the fiber angle and the thickness of each layer
k=1 to minimize the total mass of the structure subject
to the stress constraint. The optimization problem is
Equations (35) and (36) can be written in the following solved using a two-level optimization procedure. At the
simple form: first level, the fiber direction of each ply in each finite
element is optimized using (19). At the second level,
∂ Bn 1 the thickness of each ply is optimized, using (18) to
= An − A p + zip Eip (37)
∂tip 2 minimize the total mass of the material. Both of these
∂ Dn equations are applied using the strain energies com-
= Bn − 2B p + zip
2
Eip (38) puted at the final load level. The new thicknesses are
∂tip
then scaled to satisfy the stress constraint in the form of
in which A p and B p are found using an equation similar failure criterion. In the case where the failure occurs at
p a lower load than the applied load, the thicknesses are
to Eq. (23), substituting nk=1 by k=1 .
In case of symmetric laminates (Bn = [0]) where increased by a scale factor greater than one, and vise
symmetry is preserved during the thickness optimiza- versa. The two-level procedure is then repeated until
tion, p is selected among the layers in half thickness of convergence based on (20 is achieved.
the laminate ( n2 + 1 ≤ p ≤ n). In this case, one can eas- In this study, failure is assumed to occur in accor-
ily find the following relations using a similar method as dance with the Tsai–Hill ply failure criterion (Tsai and
above: Hahn 1980):
∂ An σ1 2 σ2 2 τ12 2 σ1 σ2
= 2Eip (39) + + − − 1 ≤ 0 (44)
∂tip σ1 f σ2 f τ12 f σ12f
Parnas et al. (2003) and shown in Fig. 5) is presented. based on the pure deformations, which are usually a
The plate is subject to a distributed load of 36 N/mm at function of the global nodal displacement vector.
the end and is made of two layers with [5/ − 30] lay-up It should be mentioned that variable-thickness com-
and the following material properties: posite structures are fabricated by ply drops and splic-
ing, which are not necessarily cost-efficient due to their
E1 = 157 GPa , E2 = 8.5 GPa , high manufacturing cost. As a result, such optimization
G12 = 2.6 GPa , υ12 = 0.3 , methods are justified only for those critical aerospace
σ1 f (Tensile) = 1349 MPa , σ1 f (Comp.) = 1349 MPa , structures in which tradeoff in cost, durability, stiffness,
σ2 f (Tensile) = 36 MPa , σ2 f (Comp.) = 36 MPa , and weight is not a primary concern. Ply drops and
τ12 f = 36 MPa splicing can also affect the strength of the structure
significantly; however, this is not included in the stress
Initially, each layer is assumed to have a thickness of constraint. More detailed studies are required to im-
0.85 mm, and the initial plate volume is V0 = 850 cm3 . prove the proposed method with more realistic stress
In this example, both fiber directions and layer thick- constraints.
nesses are considered as the design variables. However,
in each layer, the fiber direction is considered to be the
same for all elements with the same x coordinate. Us-
References
ing the proposed optimality criteria and the two-level
optimization procedure explained before, the thickness
Adali S (1983) Design sensitivity analysis of an antisymmetric
and fiber directions of each layer are optimized, and angle ply laminate. Eng Optim 7:69–83
results are shown in Figs. 6 and 7. Results are also com- Adelman HM, Haftka RT (1986) Sensitivity analysis for dicrete
pared with those found using the sequential quadratic structural systems. AIAA J 24:823–832.
António CAC, Marques AT, Soeiro AV (1995) Optimization
programming method and Bezier surfaces and curves of laminated composite structures using a bilevel strategy.
(Parnas et al. 2003). The proposed method converges in Compos Struct 33:193–200
nine iterations and needs 28 function evaluations. How- Arora JS (1989) Introduction to optimum design. McGraw-Hill,
ever, according to Parnas et al., the sequential quadratic New York
Berthelot JM (1999) Composite materials: mechanical behav-
programming method converges in 16 iterations, and
iour and structural analysis. Springer, Berlin Heidelberg
obviously, more function evaluations are required to New York
perform the sensitivity analysis. Again, it may be seen Burns T, Cherkaev A (1997) Optimal distribution of multimater-
that the proposed method leads to the optimum result ial composites for torsional beams. Struct Optim 13(1):4–11
Díaz AR, Sigmund O (1995) Checkerboard patterns in layout
in a more efficient way, without performing the sensi- optimization. Struct Optim 10:40–45
tivity analysis. Fang C, Springer GS (1993) Design of composite laminates by a
Monte Carlo method. J Compos Mater 27(7):721–753
Felippa CA (2003) A study of optimal membrane triangles
with drilling freedoms. Comput Methods Appl Mech Eng
192(16):2125–2168
6 Conclusions Fukunaga H, Vanderplaats GN (1990) Optimum design of lam-
inated composite structures. In: de Wilde WP, Blain WR
The present study provided a design methodology for (eds) Composite materials design and analysis. Springer,
Berlin Heidelberg New York, pp 493–497
minimum weight design of composite laminates using
Fukunaga H, Vanderplaats GN (1991) Strength optimization of
optimality criteria. Fiber angles and layer thicknesses laminated composites with respect to layer thickness and/or
in each finite element were considered as the design layer orientation angle. Comput Struct 40(6):1429–1439
variables, and two optimality criteria were presented Fukunaga H, Sekine H (1993) Optimum design of composite
structures for shape, layer angle and layer thickness distri-
and applied in a two-level optimization procedure. It
butions. J Compos Mater 27(15):1479–1492
was shown that, by using these optimality criteria, opti- Gáspár ZS, Lógó J, Rozvany GIN (2002) On design-dependent
mum design of the composite laminated structures with constraints and singular topologies. Struct Multidisc
minimum mass is found efficiently without performing Optim 24(4):338–342
Grandhi RV (1993) Structural optimization with frequency
sensitivity analysis.
constraints—a review. AIAA J 32:2296–2303
One can easily see that the optimality criteria pre- Gürdal Z, Olmedo R (1993) In-plane response of laminates with
sented in this study are valid for geometrically non- spatially varying fiber orientations: variable stiffness con-
linear structures as well. The only difference is that cept. AIAA J 31(4):751–758
Haftka RT, Adelman HT (1989) Recent developments in struc-
the analysis of geometrically nonlinear structures is
tural sensitivity analysis. Struct Optim 1:137–151
performed through an iterative process (e.g., Newton– Haftka R, Prasad B (1981) Optimum structural design with plate
Raphson algorithm) and strain energies are computed bending elements—a survey. AIAA J 19:517–522
Design of laminated composite structures for optimum fiber direction and layer thickness, using optimality criteria 167
Hyer MW, Charette RF (1991) Use of curvilinear fiber format in Muc A (1988) Optimal fiber orientation for simply-supported
composite structure design. AIAA J 29:1011–1015 angle-ply plates under biaxial compression. Compos Struct
Hyer MW, Lee HH (1991) The use of curvilinear fibre format to 9:161–172
improve buckling resistance of composite plates with central Olmedo R, Gürdal Z (1993) Buckling response of laminates with
circular holes. J Compos Struct 18:236–261 spatially varying fiber orientations. In: Proceedings of the
Jog CS, Haber RB (1996) Stability of finite elements models AIAA/ ASME/ASCE/AHS/ASC 34th SDM Conference,
for distributed-parameter optimization and topology design. La Jolla, CA, pp 2261–2269, 19–21 April 1993
Comput Methods Appl Mech Eng 130:203–226 Parnas L, Oral S, Ceyhan Ü (2003) Optimum design of composite
Kalamkarov AL (1997) Optimal design of fiber-reinforced lami- structures with curved fiber courses. Compos Sci Technol
nates. ASTM Spec Tech Publ 1309:237–249 63:1071–1082
Kam TY, Lai MD (1989) Multilevel optimal design of laminated Pedersen P (1987) On sensitivity analysis and optima1 de-
composite plate structures. Comput Struct 31:197–202 sign of specially orthotropic laminates. Eng Optim 11:305–
Kam TY, Snyman JA (1991) Optimal design of laminated com- 316
posite plates using a global optimization technique. Compos Poulsen TA (2002) A simple scheme to prevent checkerboard
Struct 19:351–370 patterns and one node connected hinges in topology opti-
Kasap I, Oral S (1997) The effect of curved fiber courses in the de- mization. Struct Multidisc Optim 24:396–399
sign of composite laminates. In: Gutkowski W, Mroz Z (eds) Rao KP, Issac JC, Viswanath S, Murthy SS (1991) Optimum de-
Proceedings of the second world congress of structural and sign of composite laminates for strength by ranking. J Reinf
multidisciplinary optimization, Zakopane, Poland pp 675– Plast Compos 10:477–494
679, 26–30 May 1997 Reiss R, Ramachandran S (1987) Maximum frequency design of
Katili I (1993) A new discrete Kirchhoff–Mindlin element based symmetric angle-ply laminates. Composite structures 4(1).
on Mindlin–Reinssner plate theory and assumed shear strain In: Marshall IH (ed) Analysis and design studies. Elsevier,
fields—part I: an extended DKT element for thick-plate London, pp 1476–1487
bending analysis. Int J Numer Methods Eng 36:1859–1883 Sadr H, Afzali M, Sabzevari H (1989) Calcul et optimisa-
Kengtung C (1986) Sensitivity analysis and a mixed approach to tion des structures composites stratifiees. Rapport CETIM
the optimization of symmetric layered composite plates. Eng Senlis
Optim 9:233–248 Schmit LA, Mehrinfar (1982) Multilevel optimum design of
Khosravi P, Sedaghati R, Ganesan R (2007a) Optimization of structures with fiber-composite stiffened panel components.
geometrically nonlinear thin shells subject to displacement AIAA J 20:138–147
and stability constraints. AIAA J 45(3):684–692 Sigmund O, Petersson J (1998) Numerical instabilities in topol-
Khosravi P, Sedaghati R, Ganesan R (2007b) Optimization of ogy optimization—a survey on procedures dealing with
stiffened panels considering geometric nonlinearity. J Mech checkerboards, mesh—dependencies and local minima.
Mater Struct 2(7):1247–1263 Struct Optim 16:68–75
Khosravi P, Ganesan R, Sedaghati R (2007c) An efficient facet Tsai SW, Hahn HT (1980) Introduction to composite materials.
shell element for corotational nonlinear analysis of thin and Technomic, Lancaster, PA
moderately thick laminated composite structures. Comput Valido AJ, Cardoso JB (2003a) Geometrically nonlinear com-
Struct (in press) posite beam structures: optimal design. Eng Optim 35(5):
Kodiyaiam S, Nagendra S, DeStefano J (1996) Composite sand- 553–560
wich structure optimization with application to satellite com- Valido AJ, Cardoso JB (2003b) Geometrically nonlinear com-
ponents. AIAA J 34(3):614–621 posite beam structures: design sensitivity analysis. Eng
Landriani GS, Rovati M (1991) An optimal design for two- Optim 35(5):531–551
dimensional structures made of composite materials. J Eng Verijenko VE, Johnson JD (1996) Design synthesis of compos-
Mater Technol 113:341–354 ite structures through the integration of optimization tech-
Le Riche R, Haftka RT (1995) Improved genetic algorithm for niques and finite element analysis. Computer aided design in
minimum thickness composite laminate design. Compos Eng composite material technology—international conference.
5(2):143–161 Computational Mechanics, Southampton, pp 197–206
Lin CC, Yu AJ (1991) Optimum weight design of composite Watkins RI (1987) Multilevel optimization of composite struc-
laminated plates. Comput Struct 38(5/6):581–587 tures. In: Proceedings of the fourth international conference
Lógó J (2005) New types of optimal topologies by iterative on composite structures. Elsevier Applied Science, Paisley,
method. Mech Struct Mach 33:149–171 Scotland pp 1393–1403
Miravete A (1990) Analysis and optimization of simple compos- Watkins RI, Morris A (1987) A multicriteria objective func-
ite structures. In: de Wilde WP, Blain WR (eds) Composite tion optimization scheme for laminated composites for
materials design and analysis. Springer, Berlin Heidelberg use in multilevel structural optimization schemes. Comput
New York, pp 529–547 Methods Appl Mech Eng 60:233–251
Morton SK, Webber JPH (1997) Heuristic design of compos- Weiji L, Boohua S (1986) Multilevel optimization procedure of
ite laminates for strength stiffness and multiple load cases. composite structures. Comput Struct 26:1357–1367
Aeronaut J 101(1001):35–45 Zynczowski M (1992) Recent advances in optimal structural de-
Mota Soares CM, Correia VF, Mateus H, Herskovits J (1995) sign of shells. Eur J Mech A Solids 11:5–24
A discrete model for the optimal design of thin compos- Zhou M, Shyy YK, Thomas HL (2001) Checkerboard and min-
ite plate-shell type structures using a two-level approach. imum member size control in topology optimization. Struct
Compos Struct 30:147–157 Multidisc Optim 21:152–158