Gear defect detection based on EEMD method and BP neural network

GUO Mian， CHEN Hong-fang
（Mechanical Engineering and Applied Electioncs Thchnology Institute， Beijing University of Tehcndogy，
Beijing 100124，China）

Abstract： Aiming at overcoming the difficulties in noise analysis，such as the complexity of the gear noise signals，and the interference by outside noises，a new noise analysis method was proposed based on ensemble empirical mode decomposition（EEMD）algorithm，time synchronous averaging（TSA）and back propagation（BP）neural network. EEMD was used to extract useful signals from the original signal based on the gear mesh frequency and multiplication of the mesh frequency. TSA was used for further de-noising. Then the feature values of the tested signals after de-noising were calculated. Discriminative features among different gear defect type were selected and taken as the input of BP neural network，the type of gear defect was effectively identified. The results of our experiments indicate that the proposed method based on EEMD and TSA enhances the denoise effect，and effctive features are obtained after the de-noising. Moreover，the inference of the gear defect type based on the BP neural network could avoid the disadvantages of humans' subjective judgments，and achieves accurate identification results.
Key words： gear defect detection； ensemble empirical mode decomposition（EEMD）； denoising； back propagation（BP） neural network； time synchronous averaging（TSA）