Machining and Chatter Avoidance: Predictive Analytics and Uncertainty Analysis

The focus of this dissertation is on the topic of chatter avoidance during machining and the prediction of specific cutting forces and maximum tool temperatures during machining using machine learning. Self-excited regenerative vibration or “chatter” is a significant obstacle in machining which results in poor surface quality. To avoid chatter, a 2D diagram of the depth cut limit vs. the spindle speed, called Stability Lobe Diagram (SLD), is used. The SLD depends on the cutting parameters and structural dynamics parameters. Theoretically, chatter can be avoided using the physic-based SLD. But in practice, there is a gap between the empirical results and what the theory supports due to the uncertainties associated with the in-process structural dynamics parameters. using a multivariate Newton method, given the empirical data sets. The first part of the dissertation focuses on the inverse problem in chatter avoidance where the in-process structural dynamics parameters are extracted using a multivariate Newton method, given the empirical data sets. The SLD and the cutting parameters are assumed to be known and given. Using this knowledge, the structural dynamics parameters are obtained using the inverse approach. In addition, the uncertainty in the value of each structural dynamics parameter derived through the inverse approach is also presented. The results derived from the algorithm are used to discover the sensitivity of the boundary with respect to each parameter. The last part of the dissertation covers the prediction of the specific cutting force and the maximum tool temperature during machining using machine learning models.