School of Engineering Department of Civil and Environmental Engineering 77 Exploring the Potential and Limitations of Artificial Intelligence Based Structural Health Inspections Supervisor: ZHANG Jize / CIVL Student: KWAN Sum Yi / CIVL Course: UROP 1100, Spring Artificial Intelligence is applied in structural health monitoring system to autonomously detect any potential structural damages, like cracking and corrosion, to prevent severe structural damage and maintain long service life of buildings and infrastructure. Deep learning semantic segmentation model is applied for crack recognition. Through trial-and-error, our model can satisfactorily detect cracks in unseen images with consistent background that resemble to the training dataset. We are investigating the possibility of recognizing crack under different complex backgrounds, and the use of Foundational Model to aid the detection process. Distributed Fiber Optic Strain Sensing for Civil Infrastructure Monitoring Supervisor: ZHANG Shenghan / CIVL Student: CHEUNG Hoi Yan / CIVL Course: UROP 1100, Summer Structural health monitoring is crucial to the structural life assessment, among which the identification and location of cracks is the most effective link in structural health monitoring. In traditional discrete sensing methods, strain gauges are distributed at each point at several distances. However, the difference is large, and the details of the cracks cannot be shown. The traditional strain sensor (point sensor) is discrete with the gauge length constant, such as short-gauge and long-gauge, including the technology are micro-electromechanical (MEMS), electrical and fiber optical sensors. The main difference between point sensor and distribute sensor is the spatial resolution which means the distance between independent measurements. Distributed Fiber Optic Sensing (DFOS) technology for testing reinforced concrete infrastructure enables fiber optics to be embedded or installed to study strain or other critical references for continuous measurement. This new approach provides a proven and efficient way to determine infrastructure capabilities and overcome the limitations of relying on traditional discrete sensing. We use decentralised fiber optic sensing (DFOS) technology instead of traditional methods. The fibres break easily when using the most sensitive optical fibers because they are so brittle. Use the stiffest fiber optics available, which may not be sensitive enough to test for cracks in steel and concrete. In this report, we focus on the performance of fiber optics to investigate the fiber optic sensor strain transfer mechanism under cracking scenarios.
RkJQdWJsaXNoZXIy NDk5Njg=