A fault diagnosis online method of railway track based on knowledge transfer learning
Zhenwei Li, F. Lv, J. Yang, Xiaoming Li · IET conference proceedings. · 2023
In the complex environment of actual train operation, due to the lack of railway track fault sample data, the fault diagnosis online model cannot be directly established. However, under the environment of track inspection vehicle and actual train operation, the railway track has the same health status space and most fault types. Therefore, this paper proposes a railway track fault diagnosis online method based on knowledge transfer learning. First, the railway track fault data under the track inspection vehicle environment is taken as the source domain, and the fault data is processed as knowledge and rules, and the expert knowledge rule base based on Belief Rule Base reasoning (BRB) is established. Then, the noise distribution of actual train operation is simulated by triangular distribution, and the fault space of railway track in complex environment is obtained as the target domain. Finally, the knowledge rules in the source domain are transferred to the target domain, and a three-level iterative optimization transfer weight learning method is designed to realize the accurate diagnosis of railway track online fault. The effectiveness of the proposed method is verified by experiments.