Comparison of artificial intelligence techniques for cutting tool condition monitoring

Abstract. In manufacturing, tool wear monitoring is crucial as it directly influences production quality and operating costs. Inaccurate replacement strategies can result in increased costs and substandard parts. Various Artificial Intelligence (AI) methods have been proposed to monitor tool wear using cutting signals, but a comprehensive performance comparison is lacking. This paper evaluates three distinct AI approaches: Artificial Neural Networks (ANN), Support Vector Machines (SVM), and K-Nearest Neighbours (K-NN). The selection of these approach is based on their learning mechanisms. Each method is optimized using a GridSearch algorithm and their real-time wear monitoring capabilities are compared. The results shows that all AI techniques monitored tool wear with similar precision, making it challenging to draw a definitive conclusion in this regard. The choice of the most appropriate AI method is heavily dependent on the manufacturing environment. For large-scale manufacturing under similar cutting conditions, K-NN and SVM are a good choice. The ANN is better suited to all scenarios, but particularly where there are substantial fluctuations in cutting conditions or, in general, larger databases.