Machine learning for multi-dimensional performance optimization and predictive modelling of nanopowder-mixed electric discharge machining (EDM)

Abstract Aluminium 6061 (Al6061) is a widely used material for various industrial applications due to low density and high strength. Nevertheless, the conventional machining operations are not the best choice for the machining purposes. Therefore, amongst all the non-conventional machining operations, electric discharge machining (EDM) is opted to carry out the research due to its wide ability to cut the materials. But the high electrode wear rate (EWR) and high dimensional inaccuracy or overcut (OC) of EDM limit its usage. Consequently, nanopowder is added to the dielectric medium to address the abovementioned issues. Nanopowder mixed EDM (NPMEDM) process is a complex process in terms of performance predictability for different materials. Similarly, the interactions between the process parameters such as peak current ( I p ), spark voltage ( S v ), pulse on time ( P on ) and powder concentration ( C p ) in dielectric enhance the parametric sensitivity. In addition, the cryogenic treatment (CT) of electrodes makes the process complex limiting conventional simulation approaches for modelling inter-relationships. An alternative approach requires experimental exploration and systematic investigation to model EWR and overcutting problems of EDM. Thus, artificial neural networks (ANNs) are used for predictive modelling of the process which are integrated with multi-objective genetic algorithm (MOGA) for parametric optimization. The approach uses experimental data based on response surface methodology (RSM) design of experiments. Moreover, the process physics is thoroughly discussed with parametric effect analysis supported with evidence of microscopic images, scanning electron microscopy (SEM) and 3D surface topographic images. Based on multi-dimensional optimization results, the NT brass electrode showed an improvement of 65.02% in EWR and 59.73% in OC using deionized water. However, CT brass electrode showed 78.41% reduction in EWR and 67.79% improved dimensional accuracy in deionized water. In addition to that, CT brass electrode gave 27.69% less EWR and 81.40% improved OC in deionized water compared to kerosene oil.