Directed-Graph-Learning-Based Diagnosis of Multiple Faults for High Speed Train With Switched Dynamics

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Railway Engineering

Background:

• High-speed trains (HST) exhibit complex switched dynamics (traction, coasting, braking) susceptible to multiple faults.
• Existing fault diagnosis methods struggle with distributed systems, asynchronous switching, and complex fault interactions.

Purpose of the Study:

• To develop a robust methodology for distributed multiple fault isolation and closed-loop fault estimation in HST.
• To address challenges posed by asynchronous switching and electromagnetic interferences in switched dynamic systems.

Main Methods:

• Directed-graph-quantum-learning-based multiple-agent systems (MAS) for characterizing joint fault effects.
• Single-integrator agents with edge agreement and persistence conditions for time-varying fault modeling and asymptotic consensus.
• Robust fault estimation design using switched Lyapunov functions and average dwell time for asynchronous switching.

Main Results:

• Sufficient conditions derived for solving multiple fault isolation using randomized learning techniques, ensuring a directed spanning tree and cycle edge in the fault topology.
• Asymptotic consensus achieved in fault modeling through agent-based approaches.
• Switched estimators designed to guarantee closed-loop system asymptotic stability under fault conditions and interferences.

Conclusions:

• The proposed methodology effectively addresses distributed multiple fault isolation, modeling, and estimation in HST with switched dynamics.
• The approach demonstrates robustness against asynchronous switching and electromagnetic interferences, ensuring system stability and reliable operation.