Thre research on multi-parallel mixed artificial neural network in fault diagnosis system

摘要

The paper put forward a banausic ANN system considering the character of complex mechanical sister. The network system is named as multi-parallel mixed artificial neural network and is applied for the full diagnosis of mechanical system. Making use of the research based ago, the same mechanical signal at different character domains is deal with and is parallel separates and is parallel identified by the network. At last, the fault is colligate and identify network. The character domains include time domain, frequency domain, axes track and wavelet analysis results, and so on. This network pick-up the signal characters by distributing fuzzy subjection function. The BP network and self-adaptive resonance theory network (ART2) is applied for faults separation and faults identification in this network. After training, the simulating diagnosis is finished by this network. In the diagnosis of four typical faults, the dianostic degree of accuracy is 100percent.