Tracking changes in grinding wheel condition using reduced features of acoustic emission signals

Abstract

In this paper, a universal method is proposed to accurately identify wheel state during grinding brittle materials. In contrast with conventional methods relying on adequate repetitive experimental data under the same process conditions, only history acoustic emission (AE) data in the same life cycle of a grinding wheel are needed for current wheel state identification. During grinding process, AE spectra samples are acquired sequentially on a series of nodes with equal interval. Samples on each node are categorized into the same class of wheel state. In the theoretical part, AE spectrum is proved to be suitable for feature representation of the degradation of wheel condition. Linear discriminant analysis is used to project AE spectra samples into a two-dimensional feature space, and the changes of the projection gives a clue to the evolution of wheel state. Two types of commonly used diamond wheels for optical surface grinding, which are straight wheel and cup wheel, were used to verify the general applicability of the method. The experiments were carried out on different machine tools and the two wheels also possessed different degradation process, yet for all of them, the evolution of wheel state can be accurately identified. It was proved that the proposed method can adaptively trace wear stage change of diamond grinding wheel.