Artificial neural network and genetic algorithm for multi-objective optimization in drilling of glass fiber reinforce polymer-stainless steel stacks

Glass fiber reinforced polymer (GFRP)-stainless steel stacks used in the aircraft structural components. The assembly process of this components requires mechanical joining using bolt and nut. The conventional drilling process is commonly used for producing hole to position the bolt correctly. Drilling is the complex machining process due to the variation in geometrical chance along the cutting edge. Thrust force and hole surface roughness are responses that used to evaluate the performance of drilling process. The quality characteristic of these responses are "smaller-is-better." The aim of this experiment is to identify the combination of process parameters for achieving required multiple performance characteristics in drilling process of GFRP-stainless steel stacks materials. The three important process parameters,i.e., point angle, spindle speed, and feed rate were used as input parameters. Point angle was set at two different levels, while the other two were set at three different levels. Hence, a 2 × 3 × 3 full factorial was used as design experiments. The experiments were replicated two times. The optimization was conducted by using the combination of backpropagation neural network method and genetic algorithm method. The architecture of developed BPNN network had 3 input layers, 1 hidden layers with 8 neurons and 2 output layers. The activation functions of the hidden layer, the output layer and the network training were tansig, purelin and trainlm respectively. The minimum thrust force and hole surface roughness could be obtained by using point angle, spindle speed and feed rate of 118°, 2380 rpm, 61 mm/min respectively.