Fault diagnosis of piston compressor based on Wavelet Neural Network and Genetic Algorithm

An improved wavelet neural network (WNN) for diagnosis of machine fault is proposed combining WNN and genetic algorithm (GA). With the property of global optimal search of GA and the temporal-frequency localization of WNN, the networks could avoid falling into local infinitesimal values. Firstly, the original parameters of WNN is obtained by making use of the GA, and then the gradient descent algorithm is employed to train the WNN to speed up the training process, so that the drawback of lower speed for only using GA to train the WNN could be overcome. Finally, the improved WNN is applied to the fault diagnosis of piston compressor, in which the results show it is superior to the common WNN in the aspects of precision and convergence.