Sustainable EDM production of micro-textured die-surfaces: Modeling and optimizing the process using machine learning techniques

The present research explores the role of die-sinking EDM parameters, such as peak current, pulse duration, and gap voltage, and their proper selection for the cost-effective fabrication of negative micro-pillar-textured surfaces using a die-material of H13 steel alloy. The developed artificial neural networks (ANN) models of surface roughness and overcut outperform the response surface methodology models in predicting responses as higher 'R-2 values' and lower 'mean squared error' with ANN models. A lower peak current of 2 A, lower pulse duration of 55.46 mu s, and intermediate gap voltage of 37.83 V is selected from metaheuristic optimization approaches such as teaching-learning-based optimization and particle swarm optimization that are efficient and comparable with the desirability function-based approach while finding optimum parameter values. Further, sustainable fabrication of micro-textured H13 die surfaces is carried out on large areas using selected optimized parameters with a form tool electrode. The negative micro-pillar pattern surface demonstrates an average overcut of similar to 40 +/- 15 mu m in comparison to the dimensions of the form tool. This research emphasized the sustainability of the EDM process by utilizing reusable dielectric fluid, prioritizing optimal parameters for high-quality fabrication using low electrical-energy utilization, and enabling large-scale production using a form tool.