Study on coupling fault diagnosis of the planetary transmission gearbox by using wavelet transform and 1DCNN

Considering the susceptibility of planetary transmission to strong noise interference, a fault diagnosis method is proposed based on Morlet wavelet transform and 1D Convolutional Neural Network (1DCNN). Firstly, the acceleration captured from the transmission is dealt with angle resampling to reduce the dependence of speed on signal characteristics, and band-pass filtering is performed, which parameter is determined by using the fast spectral kurtosis algorithm. Subsequently, the filtered data is decomposed into several multiple intrinsic mode function (IMF) components with the EEMD algorithm and reconstructed according to the correlation coefficient parameter. Then reconstructed data is subjected to further noise reduction processing using wavelet transform and the amplitude in the scale parameter with the highest kurtosis as the input to 1DCNN for training and fault diagnosis classification. The experimental results show that the proposed method can effectively reduce noise interference, the fault diagnosis accuracy of the processed data is relatively high. The fault diagnosis accuracy of the processed data is significantly higher than that of the raw data.