The gearbox is a key component in wind turbines, and the fault diagnosis of gearboxes in wind turbines is a significant process of reliability management. Therefore, a SAE-BP transfer neural network is proposed in this paper for fault diagnosis of gearboxes in wind turbines. The proposed method is conducted by two processes. Firstly, a source task data is served as the training process to pretrain the SAE-BP neural network. The final learned network structure is the transferable weights or parameters that contain the feature information. Then, the learned weights are transferred into a target task with different working and fault conditions as the initial weight of a neural network model. To extract more fault-sensitive features, fast Fourier transform (FFT) is introduced to transform the raw data into a frequency domain. Several comparison experiments are conducted to validate the proposed method, and the results show that the proposed method achieves higher classification accuracy.
|Number of pages||11|
|Journal||International Journal of Performability Engineering|
|State||Published - 1 Jan 2019|
Bibliographical noteFunding Information:
This work was funded by the Program of Chongqing Municipal Education Commission (No. KJZH17123), Research Start-Up Funds of Chongqing Technology and Business University (No. 1856018), Program of Chongqing Municipal Education Commission (No. KJQN201800830), MOST Science and Technology Partnership Program (No. KY201802006), and National Key Research & Development Program of China (No. 2016YFE0132200).
© 2019 Totem Publisher, Inc. All rights reserved.
- BP algorithm
- Intelligent fault diagnosis
- Sparse autoencoder
- Transfer learning