- Research Article
- 10.1049/icp.2022.3082
Fault diagnosis of rolling bearing under multiple working conditions based on multi-scale kernel convolution neural network
- Jan 13, 2023
- IET Conference Proceedings
- X.-L Li + 2 more +2
As the key component of wind turbine, the rolling bearing of gearbox plays an important role in the running process of wind turbine. However, in real-world situations, the working environment and conditions of wind turbine are very bad, resulting in a high failure incidence probability of gearbox rolling bearings. As mechanical vibration signals are typically non-stable, non-linear, and non-periodic, failure mode and its mapping relation to the raw signal signs become very complex. This paper aims to design a convolutional neural network based on multi-scale kernel depth and angular domain re-sampling intelligent fault diagnosis model to accurately diagnose the rolling bearing fault of gearbox. The model can effectively reduce the difference of vibration signals caused by rotational speed changes, and the inception model is used to divide the deep convolutional neural network into multiple branches to extract the features of rolling bearings in uniform or variable rotational speed states. After extracting the features, information fusion and fault classification are carried out. This combination model has excellent feature learning ability and classification performance. Moreover, the proposed method can adapt to fault diagnosis under variable speeds and multiple working conditions, and the robustness and accuracy of fault diagnosis are greatly improved.
Read more