Related Experiment Video
Updated: May 14, 2026
06:17
Quantifying the Relative Thickness of Conductive Ferromagnetic Materials Using Detector Coil-Based Pulsed Eddy Current Sensors
Published on: January 16, 2020
Physics-Enhanced Orthogonal Sensing for Self-Supervised Anomaly Detection in Rolling Mills
Yifan Wang1, Bin Zheng2, Yehan Feng1
1School of Information Science and Technology, Fudan University, Shanghai 200438, China.
Sensors (Basel, Switzerland)
|May 13, 2026
Summary
This study introduces a physics-enhanced system for real-time monitoring of rolling mill guiding systems. It enables early fault detection in industrial settings, even with limited data, improving steel product quality.
Area of Science:
- Mechanical Engineering
- Artificial Intelligence
- Cyber-Physical Systems
Background:
- Rolling mill guiding systems are critical for steel product quality.
- Harsh industrial environments limit effective online monitoring and early warning systems.
- Scarcity of fault data hinders traditional supervised AI methods.
Purpose of the Study:
- To develop a novel "physics-enhanced" orthogonal-sensing cyber-physical architecture for rolling mill guiding systems.
- To enable effective online monitoring and early fault detection despite limited industrial fault samples.
- To integrate hardware and software for robust state monitoring.
Main Methods:
- Designed an embedded orthogonal sensing layout (P⟂V) to decouple vibration and force fluctuations.
- Formulated state monitoring as a self-supervised anomaly detection problem.
- Developed a two-branch network using CSD transformer and VQ-VAE for physical coupling and context extraction.
Main Results:
- Achieved an AUC-ROC of 0.952 with a 0.048 false alarm rate at 95% TPR.
- Demonstrated end-to-end processing latency of approximately 8 ms per window.
- Achieved a system-level fault response time of approximately 108 ms.
Conclusions:
- The proposed physics-enhanced architecture meets real-time industrial monitoring requirements.
- The self-supervised anomaly detection approach effectively addresses data scarcity issues.
- The system enhances the reliability and quality control of steel production.
