Study on planetary gear fault diagnosis based on variational mode decomposition and deep neural networks

Planetary gear failures occur frequently in working conditions at low speeds, large loads, and closed operating environments, which makes the identification of faults a difficult task. A fault diagnosis method for planetary gear based on power spectral entropy of variational mode decomposition (VMD) and deep neural networks (DNN) is proposed herein. The three-axial vibration signals of a planetary gear are collected and decomposed into narrowband components with different frequency centres and bandwidths based on VMD. Power spectral entropy (PSE) is used as the original feature to represent the magnitude and distribution of the spectral amplitude of each component. A DNN based on an automatic encoder (AE) and back propagation neural network is used to realise the reduction of original signal features and the classification of gear states. The achieved overall recognition rate is 100% after the training of neural networks with training samples. The experimental results indicate that the proposed method is capable of extracting the sensitive features and recognising the fault states.