Application of Wavelet Neural Network on Turbo-Generator Set Fault Diagnosis System

To improve the limitation of applying traditional fault diagnosis method to the diagnosis of multi-concurrent vibrant faults of turbo-generator set, a novel diagnosis approach integrating the wavelet transform with neural network is proposed. The effective eigenvectors are acquired by binary discrete wavelet transform and the fault modes are classified by neural network. The back-propagation (BP) algorithm is used to fulfil the neural network structure and parameter initialization. By means of choosing enough practical samples to verify the wavelet neural network (WNN) and the information representing the faults is inputted into the trained WNN, and according to the output result the type of fault can be determined. Actual applications show that the proposed method can effectively diagnose the multi-concurrent vibrant faults of turbo-generator sets and the diagnosis result is correct. The method can be generalized to other devices´ fault diagnosis.