Fiber-Optic Sensor-Based Structural Health Monitoring with Machine Learning: A Task-Oriented and Cross-Domain Review.

Area of Science:

• Engineering
• Materials Science
• Computer Science

Background:

• Structural health monitoring (SHM) is crucial for aging infrastructure management.
• Fiber-optic sensors (FOSs) offer advantages like electromagnetic immunity and durability for SHM.
• Machine learning (ML) advances enable insights from complex sensor data.

Purpose of the Study:

• To systematically review FOS-based SHM systems integrated with ML across various infrastructure sectors.
• To examine sensing principles, deployment, data characteristics, and ML strategies.
• To identify challenges and emerging directions for intelligent SHM.

Main Methods:

• Systematic literature review following PRISMA 2020 guidelines.
• Categorization of FOS technologies (point-based, quasi-distributed, fully distributed).
• Evaluation of ML approaches for SHM tasks (detection, localization, prognosis).

Main Results:

• FOSs effectively capture strain, temperature, and structural responses.
• ML is applied to damage detection, localization, severity assessment, and environmental compensation.
• Key challenges include data volume, variability, limited labeled data, and generalization.

Conclusions:

• Integration of FOS and ML offers a powerful approach for intelligent SHM.
• Future research should focus on physics-informed learning, transfer learning, and digital twins.
• Addressing current challenges is vital for robust and scalable next-generation infrastructure monitoring.