Structural Health Monitoring Using Deep Learning

This research addresses the critical need for more accurate, efficient, and objective methods for detecting and measuring cracks in concrete structures, a task traditionally prone to human error and limited by outdated image-processing techniques.

By developing an advanced deep learning-based approach, the project eliminates manual parameter selection in deep learning modelling and achieves precise crack-width measurement in millimetres using an innovative laser calibration method.

Additionally, the creation of a comprehensive dataset for deep learning model training overcomes a significant gap in crack detection resources. The highly accurate deep learning models developed here surpass traditional methods, offering a scalable framework for integration into structural health monitoring standards. This work not only enhances reliability in damage assessment but also paves the way for expanding these methods to detect a broader range of structural damage, with promising implications for safety and maintenance in infrastructure management.