A Broad Learning Aided Data-Driven Framework of Fast Fault Diagnosis for High-Speed Trains

This paper proposes a new fault detection and diagnosis (FDD) architecture for high-speed trains, whose core is a modified broad learning system (BLS). This architecture is a data-driven realization, which enables fast and accurate FDD by effective feature extraction for online implementation. Under the proposed architecture, multiple FDD methods can be developed because of its inherent scalability.