Fault diagnosis algorithm of electric vehicle gearbox based on SDEA-Bi GRU

This paper suggests a hybrid method that combines the strengths of a bidirectional gated recurrent unit with a stacked denoising autoencoder to enhance the precision and effectiveness of diagnosing transmission faults in electric vehicles. The bidirectional gated recurrent unit makes advantage of these deep features for efficient fault pattern identification and classification. The results revealed that the hybrid algorithm had the best feature extraction ability for gear fault signals, and the signal features extracted by the algorithm were more concentrated and crossed each other less. The neurons in the hidden layer of the stacked denoising autoencoder was 180, and the number of neurons in the bidirectional gated recurrent unit was 160, and the hybrid algorithm performed best when the neurons in the hidden layer was 180 and the neurons in the bidirectional gated recurrent unit was 160. The hybrid algorithm performed best when the number of neurons was 160. The hybrid algorithm had the highest diagnostic accuracy for the faults, with the highest diagnostic accuracy of 97.98% in the balanced samples and 94.86% in the unbalanced samples. The hybrid algorithm constructed in the study effectively improves the diagnostic accuracy of transmission gear faults in electric vehicles.