Machine learning-based optimization of geometrical accuracy in wire cut drilling

Wire cut electrical discharge machining (EDM) equipment is run by computer numerically controlled (CNC) instruments and it is widely used in various industries such as aerospace, medical, and electronics. Thus, producing tight corners or very intricate patterns, wire EDM's increased precision allows for intricate patterns and cuts. Not only dimensional but also geometrical precision of products does play a very important role in today's industry. To the best of our knowledge, despite the dimensional precision, the geometrical precision has been studied by few researchers. Employing machine learning techniques, such as artificial neural networks (ANN) and non-dominated sorting genetic algorithm (NSGA), this research tries to minimize the geometrical deviation of parts produced by wire cut machining. To do so, firstly, samples have been produced based on the design matrix which contained input parameters, namely wire velocity, pulse time, and feed rate. The desired deviation from cylindricity, circularity, and symmetricity are investigated using NSGA and ANN. Then, the best and optimal combination of parameters are offered, which shows that the combination of ANN and NSGA has a significant effect on finding the optimum machining parameters. This study could supply a new viewpoint for studying geometric accuracy in wire cut drilling.