Milling stability map identification and machining parameter optimization using Bayesian inference

This dissertation describes a physics-guided Bayesian learning approach for statistically modelling and optimizing machining processes under a state of uncertainty. This approach uses a series of automatically-selected cutting tests to refine uncertainties about the machining system's dynamics and cutting force and identify higher productivity cutting parameters. The algorithm is evaluated experimentally and compared to the cutting tool manufacturer’s recommendations, both in laboratory conditions and in an industrial setting to optimize the machining process for a large aluminum component. These results show that the proposed Bayesian model can quickly identify both highly-productive machining parameters and accurate information about the underlying system in a small number of cutting tests, providing a robust solution for optimizing machining parameters without requiring any specialized equipment or measurement tools.