Machine learning-based instantaneous cutting force model for end milling operation

Abstract

Cutting force is the fundamental parameter determining the productivity and quality of the milling operation. The development of a generic cutting force model for end milling operation necessitates a large number of experiments. The experimental data contains multiple outliers due to noise and process disturbances lowering prediction accuracy of the model. This paper presents a novel approach combining the mechanistic model and the supervised neural network (NN) model to predict instantaneous cutting force variation during the end milling operation. The approach proposes training of an NN model using datasets generated from the mechanistic force model instead of using experimental data. The methodology generates a large number of datasets for the training of an NN model without conducting rigorous experimentation. A set of NN architectures were developed, and an appropriate network was derived by comparing performance parameters. A series of end milling experiments were conducted to examine the efficacy of the proposed approach in predicting cutting forces over a wide range of cutting conditions.