Design Exploration of Non-crimp Fabric for Resin Transfer Molding by the GP using AIC and TOL

This paper reports on a study in which the architecture of non-crimp fabric (NCF) was optimized to improve its stiffness and permeability. The selection of design variables in shape optimization is extremely important, but in many cases the appropriate criteria to use in choosing those design variables are unclear. This study investigated a proposed method for extraction of effective design variables. In the proposed method, Pareto solutions are first obtained by preliminarily optimizing a simple NCF model with a few design variables. The effective design variables, including dependent variables, are then extracted from the obtained Pareto solutions. In-plane and out-of-plane stiffness and permeability values are also calculated in the optimization process using the homogenization method in liquid and solid phases as objective functions. Further, the relationship between the design variables and the objective functions are determined by applying genetic programming (GP) with the akaike information criterion (AIC) and tolerance (TOL) to the Pareto solutions. The results of analysis indicated that constriction and waviness of the fiber bundles improved the performance of the NCF. Consequently, the model was again optimized, with special focus on constriction and waviness, to ensure its utility as a design guideline. The optimized model simultaneously improved the elastic modulus and permeability coefficient of the trade-off relationship—thereby verifying the efficacy of the proposed method.