The grinding wheel wear condition monitoring method based on multi-sensor information hierarchical fusion for high-speed cylindrical grinding

Abstract

In high-speed cylindrical grinding, the grinding wheel is prone to wear, resulting in increased grinding force and vibration, which causes grinding burn, chatter, and a decline in workpiece surface quality. Thus, determining the grinding wheel status to carry out dressing is vital. Therefore, a multi-sensor information hierarchical fusion method is proposed for grinding wheel wear state monitoring. Firstly, the information curve grinding wheel wear is constructed by surface roughness, power signal, wear plane area rate, and material removal amount to reveal the grinding wheel wear law. Then, the wear states are quantified by the wear plane area rate. Then, multi-domain feature extraction, optimization, and dimensionality reduction are performed on the sensing signals (Acoustic Emission and Vibration) to realize multi-sensor information data-level fusion. Finally, a BPNN-D/S model is established to realize multi-sensor information features and decision-level fusion based on Back Propagation Neural Network (BPNN) and Dempster-Shafer Evidence Theory (D/S). In the bearing outer ring cylindrical grinding experiment, the credibility of identification is more than 93 %, according to this proposed method for high-speed grinding wheel wear.