Comparative Analysis of Multi-objective Algorithms for Machining Parameters of Optimization of EDM Process

Recently, several evolutionary algorithms have been formulated with multi objective optimization capabilities. Evolutionary algorithms (EAs) are gaining popularity with the increasing computational resources. Moreover, in the field of non-conventional manufacturing processes, evolutionary algorithms are emerging as a powerful tools for their highly efficient population based optimal searches. However, in most cases selection of algorithms is based on empirical understanding and no standard resources exist for comparing the performance of such algorithms relevant to the manufacturing domain. This paper compares results of five advanced evolutionary algorithms- Non-dominated Sorting Genetic Algorithm-III (NSGA-III), Strength Pareto Evolutionary Algorithm-II (SPEA-II), Multi-objective Evolutionary Algorithm Based on Decomposition (MOEA/D), Pareto Envelope-based Selection Algorithm-II (PESA-II), and Passing Vehicle Search (PVS) algorithm. The performance of EAs are compared using three cases of EDM process. In each case, solution sets for all five optimization methods are recorded. These solution sets are used to plot Pareto optimal plots for visual comparison of performances. To quantitatively ascertain the performance of an algorithm based on the generated solution sets, seven performance metrics are considered—Generational Distance, Inverted Generational Distance, Spacing, Spreading, Hypervolume, and Pure Diversity which are coded using MATLAB. The combination of these performance metrics determines the cardinality, accuracy and diversity of solution sets in each case. Preliminary studies have shown that NSGA-III has better performances measure in overall terms among the five algorithms. Thus, the results of this study will help researchers in selecting appropriate optimization technique based on the established performance measures of that algorithm.