Expert system-supported fuzzy diagnosis of finish-turning process states

Abstract Developing an effective method for on-line machining condition monitoring has been of great interest with the advent of automated machining systems. By effective it is meant that reliable and timely diagnosis of machining process states, such as tool breakage, severe wear and chip hazard, should be provided under various work conditions in a practical workshop environment. This is difficult as the machining process, especially finish-turning process, is complex, random and uncertain in nature, and influenced by numerous process parameters. In an attempt to tackle the problem, a new approach based on fuzzy state diagnosis is presented in this paper by introducing a series of fuzzy feature-state relationship matrices to quantify the strength between each key signal feature identified from cutting force-tool vibration data and various actual machining process states. The knowledge-intensive fuzzy feature-state relationship matrices are off-line developed with the support of a knowledge-based expert system that is constructed by a well-established machining reference database, expert intelligence on logic reasoning and decision-making, and experimental results of signal characteristics under various work conditions. These matrices, once established, can be on-line implemented to generate an integrated fuzzy feature-state matrix (ten features and nine states in this work) which is the essence for a fast and reliable diagnosis of machining process states. Finally, a detailed case study is worked out to demonstrate the work principle of the methodology presented in this paper.