MULTI-SENSOR DATA FUSION FOR SPECIFIC ENERGY ESTIMATION IN A SURFACE GRINDING PROCESS

Abstract

Specific energy is one of the key parameters for evaluating the performance of grinding processes. This paper presents an approach to estimate specific energy in grinding processes by fusing data from multiple low-cost sensors. Six sensors were adopted to measure total machine current and voltage, grinding motor current and voltage, vibration and acoustic emission. Time and frequency domain analysis are conducted to extract features from raw sensorial data which partially represent the grinding specific energy variation. An artificial neural network (ANN) model was established to fuse the extracted features and determine the mathematical correlation to the specific energy. The approach was experimentally tested on a commercial surface grinding machine. The results of experimental validation indicate that the specific energy for the grinding process under study can be accurately estimated with the developed approach with an average accuracy of 72.94%.