POLYMER COMPOSITES, cilt.32, sa.12, ss.1988-2000, 2011 (SCI İndekslerine Giren Dergi)
In this study, the layer optimization was carried out for maximizing the lowest (first) fundamental frequency of symmetrically laminated square and rectangular laminated composite plates resting on rigid point supports distributed in eight different arrangements. Genetic Algorithm (GA) was used for searching the optimal stacking sequences of laminated composite plates, which is maximizing the first natural frequency defined as an objective function (fitness function). The first natural frequencies of the composite plates with various stacking sequences was calculated using the finite element method (FEM). However, the calculation of the first natural frequency for each new lay-up sequence and point-support arrangement needs a longer time than the evaluation period of GA. In order to reduce the searching time of the optimal lay-up sequence, the artificial neural network models were proposed and trained with small training and testing data composed of the natural frequencies of the composite plates calculated for random lay-up sequences, layer number, eight-point support arrangements, and four-plate length/width ratios using the FEM. The GA combined with these trained neural networks predicted successfully the optimal layer sequences without yielding a local optimum on the contrary the Ritz-based layerwise optimization method (Narita and Hodgkinson, Compos. Struct., 69, 127 (2005)). The maximized fundamental frequencies were always better than those obtained for typical cross-ply, angle-ply, and quasi-isotropic lay-ups. The effects of the rigid-point support locations, the optimum fiber orientation angles and the layer number on the maximum lowest natural frequency are also discussed. POLYM. COMPOS., 32:1988-2000, 2011. (C) 2011 Society of Plastics Engineers