Mechanistic force model for machining process—theory and application of Bayesian inference

Abstract

This work discusses the Bayesian parameter inference method for a mechanistic force model for machining. Bayesian inference methods have gained popularity recently owing to their intuitiveness and ease with which empirical knowledge may be combined with experimental data considering the uncertainty. The first part of the paper discusses Bayesian parameter inference and Markov Chain Monte Carlo (MCMC) methods. MCMC method effectiveness has been further analyzed by (1) changing the number of particles in MCMC estimation and (2) changing the MCMC move step size. The second part of the paper discusses two example applications as nonlinear mechanistic force model coefficient identification. The Bayesian inference scheme performs prediction of the cutting force coefficients from the training data. Using these coefficients and input parameters to the model, the cutting force is predicted. This prediction is validated using experimental data, and it is demonstrated that with very few parameter updates the predicted force converges with the measured cutting force. The paper is concluded with the discussion of future work.