Comparative study of statistical and soft computing-based predictive models for material removal rate and surface roughness during helium-assisted EDM of D3 die steel

Abstract

This research work proposes mathematical models, based on artificial neural network (ANN) with back-propagation algorithm, adaptive neuro-fuzzy inference system (ANFIS) and response surface methodology (RSM), for prediction of material removal rate (MRR) and surface roughness (SR) of helium-assisted electrical discharge machining of D3 die steel. The helium gas-assisted die-sinking EDM with perforated electrode was carried out by an EDM machine. For the present experimental work, discharge current, pulse on time, duty cycle, electrode rotation and discharge gas pressure were selected as process factors, while MRR and SR were chosen as process responses. Analysis of variance (ANOVA) was done to examine the adequacy of the developed model. The fit summary confirmed that the quadratic model is statistically appropriate and the lack of fit is insignificant. Root mean square error and absolute standard deviation, obtained through RSM, were also used for developing the model and for its predicting abilities through ANN and ANFIS. The experimental and predicted values of MRR and SR during the process, obtained by RSM, ANN and ANFIS, were found to be in accord with each other. However, the ANFIS technique proved to be more fitting to the responses as compared to the ANN and the RSM. The optimum value of the MRR at 28.54 mg/min and the SR at 4.21 µm was obtained with optimal process parameters by optimization of developed statistical models using genetic algorithm.