A data-driven predictive model of the grinding wheel wear using the neural network approach

Abstract

Advanced manufacturing depends on the timely acquisition, distribution, and utilization of information from machines and processes. These activities can improve accuracy and reliability in predicting resource needs and allocation, maintenance scheduling, and remaining service life of equipment. Thus, to model the state of tool wear and next to predict its remaining useful life (RUL) significantly increases the sustainability of manufacturing processes. there are many approaches, methods and theories applied to predictive model building. the proposed paper investigates an artificial neural network (ANN) model to predict the wear propagation process of grinding wheel and to estimate the RUL of the wheel when the extrapolated data reaches a predefined final failure value. The model building framework is based on data collected during external cylindrical plunge grinding. Firstly, usefulness of selected features of the measured process variables to be symptoms of grinding wheel state is experimentally verified. Next, issues related to development of an effective MLP model and its use in prediction of the grinding wheel RUL is discussed.