Abstract
A planetary gearbox with large transmission ratio is a critical component of a wind turbine (WT). When the planetary gearbox suffers from failure, the regular variation of circular pitch vibration energy contains fault pattern information. Although deep learning-based fault diagnosis models have the capability for automatic feature extraction, they rarely consider emphasizing the energy variation characteristics related to the planetary transmission structure in the data representation stage, which degrades the fault diagnosis performance. To solve this problem, this paper describes a novel data representation method, circular pitch cyclic vector (CPCV) and, on this basis, proposes a WT planetary gearbox fault diagnosis method using CPCV and a bidirectional gated recurrent unit (BGRU). Firstly, this paper utilizes root mean square value as the energy evaluation metric for calculating the CPCV. Then, the mapping relationship between circular pitch energy and the ring gear mesh phase is captured using the BGRU through its bidirectional learning capability. Finally, a fully connected layer with softmax is used for fault classification. The proposed method is evaluated by analyzing the in-service WT vibration data collected from a wind farm in eastern China. The experimental results and comparison analysis illustrate the superiority of the proposed method.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have