Sensors Information Fusion System with Fault Detection Based on Multi-Manifold Regularization Neighborhood Preserving...

Area of Science:

• Electrical Engineering
• Control Systems
• Signal Processing

Background:

• Electrical drive systems are critical in high-speed trains, necessitating robust performance monitoring.
• Sensor information fusion systems require accurate fault detection to prevent severe consequences of closed-loop faults.

Purpose of the Study:

• To propose an optimal neighborhood preserving embedding (NPE) method, termed multi-manifold regularization NPE (MMRNPE), for detecting faults in electrical drive sensor information fusion systems.
• To enhance fault detection by integrating local and global sensor information and optimizing manifold combinations.

Main Methods:

• Developed MMRNPE by extending locality preserving embedding to utilize both designated and paired points' Euclidean distances.
• Fused multiple manifolds to extract distinct features and allocated parameters across these manifolds.
• Incorporated information entropy to balance manifold contributions and prevent overweighting of single manifolds.

Main Results:

• Experimental validation demonstrated the effectiveness of the MMRNPE approach in fault detection.
• MMRNPE showed superior fault detection representation compared to the standard NPE method.
• The method successfully distinguished faulty states from normal operations in sensor data.

Conclusions:

• MMRNPE is a promising technique for reliable fault detection in high-speed train electrical drive sensor systems.
• The proposed method offers improved performance by leveraging multi-manifold regularization and information entropy.
• Accurate fault detection is crucial for the safety and reliability of high-speed train operations.