A novel fault diagnosis technology and its application based on neural network multi-sensor information fusion

To solve the traditional fault diagnosis can not be adapted to the complicated system, a kind of new multi-sensor fusion fault diagnosis method is presented. The method applies the theory of genetic algorithms and fuzzy logic to the BP (back propagation) neural network. Combined with BP and GA, it walks in several steps. Firstly, the best individual is chosen in current population and trained in order to make object error quickly fall and determine the search direction. Secondly, the best individual crosses with the other individual after BP training. Thirdly, the current best individual that is chosen in crossover reproduction and the original best individual are trained in next cycle. Experiment results show that the fault diagnosis accuracy is improved effectively by this method.