Detection of Small-Amplitude Hunting Motion in High-Speed Trains Under Extreme Data Imbalance

The hunting motion can seriously affect the safety of high-speed trains. Small-amplitude hunting is an intermediate state between the normal state and large-amplitude hunting, so its detection offers early warning of subsequent hunting instabilities. Because the normal data is much more than the failure data, the hunting data of high-speed trains are extremely im-balanced. In this paper, a classification method CTransVAE-GAN is proposed for data enhancement. Experimental results show that the sample generation capability of CTransVAE-GAN was significantly higher than that of other methods. CTransVAE-GAN fits the data distribution of the hunting data and enhances the dataset for hunting detection for high-speed trains.