A dual response surface-desirability approach to process modeling and optimization of Al2O3 powder-mixed electrical discharge machining (PMEDM) parameters

Abstract

This paper presents an effort to model and optimize the process parameters involved in powder-mixed electrical discharge machining (PMEDM). Aluminum oxide (Al2O3) fine abrasive powders with particle concentration and size of 2.5–2.8 g/L and 45–50 μm, respectively, were added into the kerosene dielectric liquid of a die-sinking electrical discharge machine. The experiments were carried out in planing mode on a specially designed experimental set up developed in laboratory. The CK45 heat-treated die steel and commercial copper was used as work piece and tool electrode materials, respectively. Response surface methodology, employing a face-centered central composite design scheme, has been used to plan and analyze the experiments. Based on the preliminary and screening tests as well as the working characteristics of selected EDM machine, discharge current (I), pulse-on time (Ton), and source voltage (V) were designated as the independent input variables to assess the process performance in terms of material removal rate (MRR) and surface roughness (Ra). Suitable mathematical models for the response outputs were obtained using the analysis of variance technique, in which significant terms (main effects, two factor interactions, and pure quadratic terms) were chosen according to their p values less than 0.05 (95 % of confidence interval). Having established the suitable regression equations, a search optimization procedure, based on the use of desirability functions, optimizes the process performance in each machining regime of finishing (Ra ≤ 3 μm), semifinishing (3 μm ≤ Ra ≤ 4.5 μm), and roughing (Ra ≥ 4.5 μm). The results are sets of optimum points which make the MRR as high as possible and keep the Ra and all machining parameters in their specified ranges simultaneously. Finally, the modeling and obtained optimization results were also discussed and verified experimentally. It was shown that the error between experimental and anticipated values at the optimal combination settings of input variables are all less than 11 %, confirming the feasibility and effectiveness of the adopted approach.