Application of wavelet-neural networks to diagnosis of incipient gear fault

Firstly in the paper,the vibration from surface of gearbox is processed by the synchronous average sampling technique based on test of geared system running.Then the features of gear fault information are extracted by wavelet compact and entered into the neural networks as the input characteristic vectors.Three kinds of different states of gear running,i.e.the normal condition,the condition of tooth crack and the condition of tooth surface spalling,can be identified effectively by use of function of pattern identification of neural networks.