Multi-objective Optimization of EDM Process with Performance Appraisal of GA based Algorithms in Neural Network Environment

Abstract

Abstract This work focuses on the material removal rate and electrode wear rate of A2 steel in the EDM process, which are considered as to be the important governing parameters for higher productivity and accuracy. The purpose of this investigation is to minimize the electrode wear rate and maximize the material removal rate by controlling the machining parameters. In this work, an experimental investigation has been carried out by considering four machining control parameters such as discharge current (Ip), pulse duration (Ton), duty cycle (τ or Tau) and voltage (V) by using the full factorial design methodology. Artificial Neural Network (ANN) model has been developed to correlate the data generated from experimental results. To obtain an efficient ANN model and to achieve minimum prediction error, ANN architectures, learning/training algorithms and numbers of hidden neurons are generally varied. However, so far the variations have been made in a random manner. For this reason a full factorial design integrated Analysis of Variance (ANOVA) has been utilized to investigate the influence of control factors on response. Developed ANN model equation of material removal rate and electrode wear rate were subsequently used as the fitness functions in Genetic Algorithm (GA) based multi-objective algorithms. Three advanced GA based optimization techniques, i.e., Non-dominating Sorting Genetic Algorithm-II (NSGA-II), controlled NSGA-II and Strength Pareto Evolutionary Algorithm 2 (SPEA2) were attempted for the considered EDM process. Furthermore, a non-dominated set of solution was obtained to have diversity in the solutions for the EDM process. The result obtained using the ANN method and optimization techniques were confirmed using confirmation experiments.