Abstract:In view of the unique sensitivity of acoustic emission signals to earlystage cracks in gears, it is of great significance to identify the characteristics of early gear acoustic emission signals. The theory and principle of wavelet transform were introduced, the gear fatigue test platform was established. By using the wavelet threshold denoising, the gear acoustic emission signals were preprocessed under different operating conditions, the signal of high energy frequency band was extracted and the characteristic parameters in time domain and frequency domain were acquired, which was used as input of BP neural network to identify the acoustic emission signals under different operating conditions.The experimental results show that the characteristic parameters extracted from the high energy band signal have higher recognition rate than the full band signal after denoising, which provides a reference for the analysis and detection of early gear fault signals.