Mechano-electric system fault diagnosis based on wavelet analysis and neural networks

According to the trend of intelligence fault diagnosis and the variety of information provided by the equipment, the idea of integrated neural networks used in fault diagnosis is put forward in this paper for the first time, and the correlated questions about the new idea is also researched, such as how to model, how to realize, etc. It provides a new method for fault diagnosis. Integrated neural networks can be used to diagnose faults from different levels and different aspects, so it is conform to the actual situation well. In the paper wavelets´ applications in fault diagnosis is given. Here wavelet is used as a tool of signal processing. By applying wavelets analyzing, one kind of new feature vector reflecting the faults is gotten, which can be used by neural networks as its training mode. This paper presents an intelligent methodology for diagnostics of incipient faults in rotating machinery. A fault diagnosis system is developed for rotating machinery. In this system, the wavelet transform techniques are used in combination with function approximation model to extract fault features used in the fault diagnosis of rotating machinery