An online self-constructing wavelet fuzzy neural network for machine condition monitoring

Abstract

The subject of machine condition monitoring is charged with developing new technologies to diagnose the machinery problems. A problem with diagnostic techniques is that they require constant human interpretation of the results. Fuzzy neural networks show good ability of self-adaption and self-learning, wavelet transformation or analysis shows the time frequency location characteristic and multi-scale ability. Inspired by these advantages, a wavelet fuzzy neural network (WFNN) is proposed for fault diagnosis in this paper. This fuzzy neural network uses wavelet basis function as membership function whose shape can be adjusted on line so that the networks have better learning and adaptive ability. An on-line learning algorithm is applied to automatically construct the wavelet fuzzy neural network. There are no rules initially in the wavelet fuzzy neural network. They are created and adapted as on-line learning proceeds via simultaneous structure and parameter learning. The advantages of this learning algorithm are that it converges quickly and the obtained fuzzy rules are more precise. The results of simulation show that this SWFNN network method has the advantage of faster learning rate and higher diagnosing precision.