Research On Non-stationary Early Fault Diagnosis Of Planetary Gears

As one of the important parts of rotating machinery,planetary gears are widely used in high-speed railway trains and various large-scale construction machinery and equipment.The safety and reliability of its operation are extremely important.Therefore,the fine diagnosis of early faults of planetary gears has theoretical significance and practical application value.However,the early fault features of planetary gears under strong background noise are relatively weak,and it is difficult to extract and classify fault features.If traditional methods are used,it is easy to cause misdiagnosis or missed diagnosis.In response to the above-mentioned key issues,this article has conducted the following research:(1)Aiming at the problem that the fault features of planetary gears are difficult to extract under strong noise interference.This thesis first adopts the signal preprocessing method of artificial fish school to optimize the variational modal decomposition,determines the number of modal decompositions and the penalty factor,and decomposes the planetary gear signal into eigenmode components that do not alias each other.Secondly,the decomposed components are reconstructed according to the correlation coefficient law.Finally,taking advantage of the weak energy impact of noise in the fractional domain,two filtering methods,fractional Fourier transform and fractional wavelet transform,are used to filter the reconstructed signal.The simulation results show that both the fractional wavelet transform and the fractional Fourier transform can achieve the denoising effect on the signal;the denoising effect of the fractional wavelet transform is better than that of the fractional Fourier transform;the energy-based fractional Fourier transform The transform algorithm is better than the fractional Fourier transform algorithm based on peak search.(2)Aiming at the problem that planetary gear failures are difficult to accurately identify under strong background noise.Moreover,the traditional one-dimensional convolutional neural network is prone to feature loss during fault classification,and it is difficult to accurately classify and recognize.This thesis uses a two-dimensional convolutional neural network model for fault classification and recognition.First use the wavelet packet to solve the normalized energy value of the signal filtered by the fractional Fourier transform and the fractional wavelet transform respectively;then convert the one-dimensional wavelet energy value into a two-dimensional feature matrix for the training of the diagnostic model;finally use The test set is tested in the diagnostic model.The simulation results show that the two-dimensional convolutional neural network can effectively realize fault classification and recognition.In addition,the accuracy of planetary gear fault classification based on FRWT and 2D-CNN is better than the accuracy of planetary gear fault classification based on FRFT and 2D-CNN.