Application of genetic neural network in fault diagnosis of rotating shaft systems

ZHU Wen-cai， HU Hai-gang， ZHU Ming-he， PANG Hong-lei
（Marine College， Ningbo University， Ningbo 315211， China）

Abstract： In order to solve the problems of slow convergence rate and falling into local minimum easily of BP neural network，a diagnosis method to combine the neural network with the genetic algorithm was investigated. Firstly，a simulation platform for marine diesel engine shafting was designed. Then，by analyzing the torsional vibration signal in the decomposition of wavelet packet when marine diesel engine's rotating shaft system failed，the energy spectrum entropy of wavelet packet was extracted as the feature vector of failure patterns. Finally，four kinds of operation condition were identified by genetic neural network. The experimental results show that GA-BP can get higher forecast accuracy than the conventional BP in the task of simulation，which is suitable to the condition monitoring and fault diagnosis of rotating shaft system.
Key words： fault diagnosis； BP neural network； genetic algorithm