A Neural Network-Based Multi-Sensor On-Line Process Monitoring in Stone Machining

Abstract

The purpose of this research was to analyze the capacity of different types of machining process signals in tool wear classification of highly non-homogeneous and anisotropic materials, such as stone. The variable physical and mechanical properties of the workpiece material have a significant influence on the selection of cutting parameters. Improper values of cutting parameters can negatively impact tool wear dynamics and potentially result in tool or workpiece breakage due to higher cutting forces. Therefore, servomotor currents, cutting forces, and acoustic emission signals were measured during drilling of three types of stone samples using nine combinations of machining parameters and drill bits with four different wear levels. The capacity of features extracted from those signals to classify tool wear level correctly was analyzed using the artificial neural network algorithm. The features extracted from acoustic emission signals achieved the highest classification accuracies and were insensitive to the type of stone sample.