Application of minimum cross entropy to model-based monitoring in diamond turning

This paper discusses model-based monitoring as one of key steps toward building an in-process diagnostic tool for the supervisory control level in diamond turning. In particular, the concept of information-theoretic measure called cross entropy, and Kullback's principle and some of the fundamental properties of minimum cross entropy are reviewed first. Estimation of surface profile from a non-linear force model is shown to be consistent with the minimisation of cross entropy. Model-based monitoring of tool chipping is then demonstrated. The experimental results indicate that monitoring strategy based on the minimum cross entropy guarantees consistent and optimum classification results in the presence of varying cutting conditions. In addition, the cross entropy as a criterion for indicating the dynamic quality of machine-tool is explained.