Neural Network Modelling and Multi-Objective
Optimization of EDM Process

Abstract

Modelling and optimization of EDM process is a highly demanding research area in the present scenario. Many attempts have been made to model performance parameters of EDM process using ANN. For modelling generally ANN architectures, learning/training algorithms and nos. of
hidden neurons are varied to achieve minimum error, but the variation is made in random manner. So here a full factorial design has been implemented to achieve the optimal of above. From the main effect plots and with the help of ANOVA results the optimal process parameters
for modeling were selected. After that optimal process modeling of MRR and TWR in EDM with the best of above parameters have been performed. In the 2nd phase of work three GA based multi-objective algorithms have been implemented to find out the best trade-ups between
these two conflicting response parameters. ANN model equations of MRR and TWR were used in the fitness functions of GA based multi-objective algorithms. A comparison between the Pareto-optimal solutions obtained from these three algorithms has been made on the basis of diversity along the front and domination of solutions of one algorithm over others. At the end a post-optimality analysis was performed to find out the relationship between optimal process parameters and optimal responses of SPEA2.