Electrical Discharge Machining (EDM) is one of the most efficient processes to produce high-ratio micro holes in difficult-to-cut materials in the Inconel 718 superalloy. It is important to apply a statistical technique that guarantees a high fit between the predicted values and those measured during analysis of test results. It was especially important to check which method gives a better fit of the calculated result values in case they were relatively small and/or close to each other. This study developed models with the use of the response surface methodology (RSM) and artificial neural networks (ANNs). The aim of the study was comparison between two models (RSM and ANNs) and to check which model gives a better data fit for relatively similar values in individual tests. In all cases, the neural network models provided a better value fit. This is due to the fact that neural networks use better fitted functions than in the case of the RSM method using quadratic fitting. This comparison included the aspect ratio hole and the thickness side gap data, the values of which for individual tests were very similar. The paper reports an analysis of the impact of parameter variables on the analyzed factors. Higher values of current amplitude, pulse time length, and rotational speed of the working electrode resulted in higher drilling speed (above 15 µm/s, lower linear tool wear (below 15%), higher aspect ratio hole (above 26), lower hole conicity (below 0.005), and lower side gap thickness at the hole inlet (below 100 µm).
Keywords: Inconel 718; artificial neural networks; difficult-to-cut material; electrical discharge machining; response surface methodology.