Diagnosis of manufacturing defects in a gear pair using wavelet analysis of vibration and acoustic signals and an ANN-based inference technique

This paper presents a method for detecting manufacturing defects in a spur gear pair based on the wavelet transform. A tool mark on the gear tooth and unshaved gears are considered for the diagnosis. Wavelet transform provides a variable resolution time-frequency distribution from which periodic impulses in vibration and acoustic signals due to the meshing of defective teeth can be detected. The study reveals periodic impulses corresponding to the rotational frequency of the gear with a dent on its tooth, which is measured in the discrete wavelet transform (DWT) signals. The results are compared with feature extraction data and results from spectrum analysis, which show that the DWT is an effective tool for gear fault diagnosis. This paper also presents artificial neural network (ANN) diagnostics. Three algorithms: a feed forward with back propagation network (FFBPN), a radial basis function network (RBFN) and a probabilistic neural network (PNN), are used for the purpose and compared. Experimental results show that the FFBPN trained with features extracted from the DWT-processed signals gives good results over the other two networks.