Development of neuro-fuzzy models for hole drilling on Ti-6Al-4V and Inconel 718 using electrical discharge machining

The aim of this study is to develop ANFIS models for prediction of input-output relationships in hole drilling EDM process. It is a nontraditional machining process preferred to produce holes on difficult-to-cut materials, particularly aerospace alloys, in a fast and accurate way with a good surface finish. There are many parameters in this process, and their effects on the process outputs are very complicated. It is usually not possible to define such complex relationships by means of conventional modeling techniques. Fuzzy logic and neural networks are intelligent modeling techniques to predict the response of a process in accordance with the given inputs.For this purpose, an Adaptive Neuro-Fuzzy Inference System (ANFIS) has been implemented to develop neuro-fuzzy models. The experimental data were obtained by making several holes on specimens of Ti-6Al-4V and Inconel 718 using copper and brass electrodes (Ø2 mm) with input parameters of current, pulse-on and pulse-off times, and capacitance. The output parameters were material removal rate, electrode wear rate, and surface roughness. The comparison between experimental and ANFIS results reveal that developed models can predict the values of process outputs for given input parameters within the lowest error range.