Surface roughness and forces prediction of milling Inconel 718 with neural network

The objective of the article is to indicate a tool that allows finding the relationship between cutting forces and surface roughness and variable cutting parameters and cutting tool wear. The tests were carried out on a workpiece of Inconel 718, machined with cemented carbide milling cutters. The article compares artificial neural networks and multiple regression. Multi-layer Perceptron networks with backward error propagation were used. Based on the conducted research, it was found that in the case of Inconel 718 machining, multiple regression is not a suitable tool for testing the relation analyzed in this article. The correlation (R²) for multiple regression is about 0.6 for forces and 0.25 for roughness. Neural networks have a correlation coefficient (R²) higher than 0,9.