Abstract
The Electro-discharge Machining (EDM) is a thermoelectric method in which material is removed from electrically conductive material by a series of rapid sparks between the workpiece and tool. Due to the non-contact process, it is suitable for brittle, hard, or complex-shaped materials. Modelling the EDM process numerically to obtain a single crater becomes complex due to many factors involved during machining, such as capacitance, voltage, flushing pressure, tool speed, and environment. Previously, a research effort has been made to develop the Plasma Flushing Efficiency (PFE%) to model the crater size. The inclusion of PFE% in the model has improved the accuracy of the prediction. Even though the model gave valuable insight, still the predictive accuracy remained limited. The current study presents the extension of previous research on the numerical modeling and simulation of micro-EDM for the prediction of crater size. The primary aim of the present work is to integrate an ANN for crater size prediction based on key process parameters, capacitance, and voltage. The model is developed by the grid search method to identify the optimal combination of activation function and neurons. The model with optimal hyperparameters showed higher predictive accuracy with R 2 > 99% for test and train data. Moreover, the models demonstrated significantly improved accuracy, with prediction errors falling below 1% for most data points, and a maximum error of ~5% occurred in a single instance, even with the limited data set.