Abstract: A synthetic multi signal features is presented based method for multi-abnorma1ity recognition using neural networks. The principle of pattern classifier with multi-input parallel processing and its learning algorithm are given, and the feature extraction for intensive tool wear and chatter is described. The examination experiment is also performed. It is shown that the new method is superior to previous ones, capable of parallel diagnosis of multi-complex abnormalities with easy set threshold.

Key words: Fault diagnosis, Machining, Neural network