Identification of grinding wheel wear states using AE monitoring and HHT-RF method

Abstract

The grinding wheel wear is inevitably accelerated for difficult-to-machine materials due to their high hardness and wear resistance properties. However, it is difficult to precisely estimate the wheel states and implement appropriate dressing strategies during the actual grinding process. Previous methods to identify the grinding wheel wear status online suffer from inadequate accuracy and weak real-time performance. In this work, a novel identification approach of grinding wheel wear states was proposed using the Hilbert-Huang transformation (HHT) of acoustic emission (AE) signal and the random forest (RF) algorithm. The AE signal was decomposed into the sum of multiple inherent mode functions (IMFs) via the empirical mode decomposition (EMD) method. The IMFs with a stronger correlation with the original AE signal were selected by the correlation coefficient. The Hilbert transformation was performed to obtain its marginal spectrum. The maximum, root mean square, and spectral centroid of each IMF were calculated as feature values. The mapping model between grinding wheel status and feature values was built using the RF method. The optimum model parameters were studied to obtain the accurate identification results. The grinding experiment results of the bearing steel showed that the comprehensive accuracy reached 93.3 %. The effective monitoring scheme could be provided for more precision grinding practices.