A Systematic Survey of Unsupervised Deep Learning Techniques for Structural Health Monitoring and Predictive Maintenance

Key Points

• The aim is to systematically review unsupervised deep learning techniques in structural health monitoring and predictive maintenance.
• Reviewed various unsupervised deep learning techniques including autoencoders and clustering methods.
• Evaluated anomaly detection models and common datasets used in the field.
• Identified evaluation metrics and challenges in implementing these techniques.
• Found significant diversity in techniques for SHM and PdM, including autoencoders and clustering.
• Highlighted challenges such as data quality and dataset availability.
• Suggested future research directions for improving unsupervised learning applications.