Tool strain–based wear estimation in micro turning using Bayesian networks

Abstract

Estimation of tool wear in micro turning is important as it enhances the process fidelity and the surface quality of the job. In this work, a simple process is demonstrated that estimates the tool wear from strain data near the cutting edge of the tool tip for micro turning operations. The tool strain for tool with six different wear lengths, collected using fiber Bragg grating sensor, was preprocessed to generate a probability distribution. The strain and tool wear data were used as the training dataset. This training dataset was subjected to maximum likelihood estimation algorithm to obtain the conditional probability distribution table required for the functioning of a suitable Bayesian network. The Bayesian network was tested for estimation of tool wear using strain data as priors for three different experiments. The maximum error in tool wear estimation using this procedure was ∼6 µm.