Abstract:Aiming at the problem that the gear needed to be stopped for many times to take pictures in gear fatigue test, an improved wavelet domain threshold denoising method was proposed, the acoustic emission signal of the meshing gear was analyzed and processed, and the feature vector was extracted as the input feature vector of the support vector machine, and the identified fault signal was used to determine the defective gear teeth according to the gear rotation period. Compared with the records of photos taken after unpacking, defective gear teeth which were calculated finally coincided with the photo records. The method reduces the influence of inaccurate data collection of gears caused by shutdown and taking pictures, and also reduces the workload of workers, and the method has been effectively applied in subsequent experiments.