TURBO-GENERATOR VIBRATION FAULT DIAGNOSIS BASED ON AFSA-RBF NEURAL NETWORKS

摘要

The learning process of radial basis function(RBF) neural networks with artificial fish-swarm algorithm(AFSA) based possesses the characteristics, say, it is insensitive on the initial weights and parameters, and exhibits an excellent capabilities to avoid the local extremum and obtain the global extremum. In this paper, a new turbogenerator vibration fault diagnosis method is proposed based on RBF neural networks with AFSA optimized defined as AFSA-RBF neural networks. In the method, the decision table of turbo-generator vibration faults diagnosis serves as learning sample to train AFSA-RBF neural networks. The well-trained neural network is applied to diagnose turbo-generator vibration fault, the results shows that the proposed method possesses better convergence speed and diagnosis precision, and is an ideal pattern classifier.