Wang, Siqi (Xihua University), Wu, Yunpu (Xihua University), Zhou, Zongmin (Xihua University), Allen, Paul (University of Huddersfield), Tucker, Gareth (University of Huddersfield), Lei, Xia (Xihua University)

Prototype-Driven Moment Exchange for Data Augmentation in High-Speed Train Fault Diagnosis

Scheduled for presentation during the Regular Session "S32b-AI-Driven Traffic Monitoring, Safety, and Anomaly Detection" (FR-EA-T32), Friday, November 21, 2025, 14:10−14:30, Southport 2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

Reliable fault diagnosis in high-speed train systems remains challenging due to complex operating conditions and limited labeled fault data. To address this issue, we propose Prototype-Mixed Moment Exchange (PMoEx), a feature-space data augmentation method that enhances sample diversity by exchanging statistical moments—mean and variance—between samples, modulated by class-level prototypes. Unlike existing approaches such as mixup and MoEx, PMoEx explicitly integrates class-conditioned statistics, enabling the preservation of discriminative features while mitigating inter-class interference. Experimental results on a simulated high-speed train fault dataset demonstrate that PMoEx consistently improves diagnostic performance under data-scarce conditions, outperforming conventional augmentation methods.