Evaluation of neural models applied to the estimation of tool wear in the grinding of advanced ceramics

Abstract

Grinding wheel wear, which is a very complex phenomenon, causes changes in most of the shapes and properties of the tool during machining, reducing the efficiency of the grinding operation and impairing workpiece quality. Therefore, monitoring the condition of the tool during the grinding process plays a key role in the quality of workpieces being manufactured. In this study, diamond tool wear was estimated during the grinding of advanced ceramics using intelligent systems composed of four types of neural networks. Experimental tests were performed on a surface grinding machine and tool wear was measured by the imprint method throughout the tests. Acoustic emission and cutting power signals were acquired during the tests and statistics were obtained from these signals. Training and validating algorithms were developed for the intelligent systems in order to automatically obtain the best estimation models. The combination of signals and statistics along with the intelligent systems brings an innovative aspect to the grinding process. The results indicate that the models are highly successful in estimating tool wear.