Optimal design of filament winding using neural network experimental design scheme

The goal of this paper is to design the filament winding in composites mate rials based on our previous developed technique. The most significant param eters in design and construction of composites prepared by filament winding are resin temperature, fiber tension and winding angle. The three variable s depict nonlinear relationship; thus a nonlinear modeling technique is req uired. The proposed methodology enjoys many of the advantages claimed for t he artificial neural network (ANN), random search optimization, fuzzy class ification and information theory for the sequential design filament winding . The neural network is used to construct a model based on the currently av ailable experimental data. Random search generates a number of candidates o f the next batch of experiments. Fuzzy classification and information analy sis are defined to balance the need of better classification and the releva nce of each class in optimization. The test results of the proposed method show that the abilities of the proposed methodology handle multivariable ex perimental design and also help experimenters discuss complex trade-offs be tween practical limitation and statistical preferences in the experiment.