Multi-objective Optimization in Drilling Kevlar Fiber Reinforced Polymer Using Grey Fuzzy Analysis and Backpropagation Neural Network–Genetic Algorithm (BPNN–GA) Approaches

An integrated approach has been applied to predict and optimize multi-performance characteristics, such as optimum thrust force (Fz), torque (Mz), hole surface roughness (Ra), delamination (D) and hole roundness (R), in drilling process of Kevlar fiber reinforced polymer. The experiments were performed by varying drill point geometry and drilling process parameters, i.e., drill point angle, feed rate, and spindle speed. The quality characteristics Fz, Mz, Ra, D, and R were the smaller the better. Taguchi orthogonal array (OA) L18 was used as the design of experiments. Grey fuzzy analysis was first applied to obtain a rough estimation of the optimum drill point geometry and drilling process parameters. Backpropagation neural network (BPNN) model was developed and utilized to predict the optimum Fz, Mz, Ra, D, and R. Genetic algorithm (GA) was performed to search for global optimum of drilling process parameters combinations. The analysis of the effect of drill point angle, as well as drilling process parameters, on the individual performance characteristics was conducted by examining both the percentage contribution of drill point geometry and drilling process parameters on the total variance of three responses individually, and the response graphs. The results of the confirmation experiment showed that the BPNN based GA optimization method could accurately predict and also significantly improve the multiple performance characteristics.