Knowledge Updating for Automated Tool Condition Monitoring in Machining

Abstract In this paper, a knowledge updating algorithm for automated tool condition monitoring is developed. The monitoring system uses fuzzy classification in multiple principal component directions of the learning data to provide fast, efficient and reliable means of developing the automated monitoring system for machining. The principles and details of implementation of the Multiple Principal Component (MPC) Fuzzy Neural Network were discussed in a previous paper (Li and Elbestawi, 1994). This paper focuses on the knowledge updating (continuous learning, retraining, or adaptation) of the system from new experimental samples. New learning data are used to tune the classification parameters at the existing neurons and to add new neurons into the network as necessary. Such a system can improve the classification performance by updating new class information. Experiments in turning showed that the success rate of tool condition classification was improved by the retrained system.