Estimation of cBN grinding wheel condition using image sensor

Abstract

The tool condition used in machining process is generally determined by the wear status, which is an important factor of resulting in the machining efficiency of the manufacturing processes. This paper presents a method to estimate the tool condition of grinding wheel using an image sensor with machine learning techniques. Statistical features, mean, standard deviation, and entropy were selected from the simplified wear model. The selected features were extracted from the wheel surface images of an electroplated cubic Boron Nitride (cBN) wheel on a commercial CNC machine, and compared to the tool condition represented by the number of grinding passes. Finally, the statistical features were fused by machine learning techniques to estimate the wear condition of the grinding wheel. Results show a sufficient correspondence between the estimated and true tool condition with 0.9590 of the coefficient of determination (R2) based on support vector regression model.