Multi-objective Optimization of Process Parameters in Electric Discharge Machining of Inconel 718 Using Machine Learning Techniques.

Abstract

Nickel-based alloys such as Inconel 718 are increasingly used in the automotive and aerospace industries due to their high strength, corrosion resistance, light weight and low wear rate. The process parameters were pulse on, pulse off, voltage and current. The response variables consisted of the material removal rate (MRR) and the surface roughness (SR) of the workpiece can be significantly improved by choosing the optimal machining parameters. The main objective of this research was to develop a model using machine learning techniques to predict the process parameter value for the optimum value of material removal rate and surface roughness, and also to develop a mathematical model based on the regression derived for the response. Regression models were created to predict how the in- process parameters affects output process parameters variables such as surface roughness, material removal rate. Regression models were performed using a full factorial design of experiments. The validity of the developed model was assessed using analysis of variance. Tests were performed on an EDMN450CNC compression EDM machine with Inconel 718 workpiece with a copper electrode used in the molding industry. The results showed that the pulse on time had a 47.76% as the most important parameter affecting the surface roughness, followed by the percentage of peak current at 26.7% and the interaction of the input parameters peak current against pulse on time and pulse on time vs pulse on time significantly affects surface roughness to roughness in ratios of 3.18% and 6.19%. Peak flow had a significant effect on material removal rate (78.13%), followed by pulses in time with 21.344%. The multi-objective optimization problem was solved using a supervised machine learning technique (Multiple Regression Predictive models) and a full factorial design of experiments with the help of a design expert software version 13. From the full factorial design of experiments, the optimal value of multiple responses of the process optimizer was Ra 3.80. µm and MRR were 52.0 mm 3 /min for I, T-on and T-off 15A, 100 µs and T-off 15A, I 100 µs and T-off 20 µs, respectively. 80 µs and Ra was 3.01 µm I, T-in and T-out at 6A, 100 µs and 20 µs, respectively. Result obtained from multi-objective optimization by using desirability function Ra = 4.339µm and MRR= 55.440 mm3 /min at I, T-on and T-off was 15A, 100µs and 20µs where obtained.