Abstract: By analyzing the characteristics of CNC fault diagnosis and in the view of FANUC 7 CNC system, this paper establishes and compares two kinds of fault diagnosis ANN——the bidirectional associative memory (BAM) and back propagation (BP), which are suitable for CNC system fault diagnosis. The results show that both methods have special virtue. Through a complement of each other to form an integrated model, the fault diagnosis system will be more efficient. Furthermore, the application of Fuzzy Neural Network in CNC fault diagnosis is discussed. Applying Fuzzy Neural Network to recognize MACS 500 CNC system fault as an example, the validity and feasibility of this method are proved. At the same time, the framework of CNC diagnosis expert system (CNC DES) developed here and several relatively key techniques are introduced, such as the object oriented knowledge representation, the diagnosis reasoning mechanism and the case based reasoning. Additionally, this paper describes the situation of applying this system to CNC system fault diagnosis.

Key words: CNC system, fault diagno sis, art ificial neural netwo rk s, expert system