Regression Models with Multiobjective GA for EDM Parameters Optimization

Abstract

Over recent years, regression model is a well known modeling technique used to model the real world application. This paper conducted computational experimental study using two types of regression models; second order polynomial regression (SOP) and multiple linear regression in optimizing machining process parameters of cobalt-bonded tungsten carbide (WC/Co) electrical discharge machining (EDM). Multiobjective genetic algorithm (MOGA) is widely known in optimization researches. Therefore, combination of conventional modeling (regression) and modern optimization (MOGA) techniques, MLR-MOGA and SOP-MOGA are examined to observe the capability of these two techniques in maximizing removal rate (MRR) and minimizing surface roughness (Ra). Four parameters are considered to create correlation with the machining performances. The best removal rate and surface roughness values are obtained from MLR-MOGA; 168.212 mg/min and 0.693 µm respectively. Nevertheless, SOP-MOGA produced viable results. The results of MLR-MOGA and SOP-MOGA benefits the machine operators or engineers when various combination of machining parameters can be selected based on the desired requirements.