|
Non-contact detection of concrete crack propagation using image processing technique |
|---|---|
| รหัสดีโอไอ | |
| Title | Non-contact detection of concrete crack propagation using image processing technique |
| Creator | Than Zaw Toe |
| Contributor | Itthisek Nilkhamhang, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2568 |
| Keyword | Crack segmentation, Crack propagation, Flexural bending test, Crack length measurement, Crack width measurement, Crack monitoring |
| Abstract | Cracks in concrete structures are a critical indicator of poor structural health, posing a significant threat of catastrophic failure, especially in seismically active regions. Understanding how flexural cracks initiate and propagate in beams under load is essential forimproving structural safety. However, traditional monitoring methods rely on manual inspection, which is labor-intensive, prone to human error, and poses significant safety risks topersonnel conducting measurements during hydraulic load press experiments. To overcomethese limitations, this research develops and validates an automated, non-contact, visionbased system designed to accurately detect, monitor, and analyze crack behavior in concretebeams under hydraulic press loading. The objective is to provide precise, continuous observation of crack evolution, enhancing both measurement accuracy and operator safety. Thesystem integrates the advanced AI and computer vision techniques. It employs the SegmentAnything Model (SAM) for segmentation of the concrete specimens from background images. A crack detection pipeline is proposed based on OpenCV that is used in conjunctionScikit-learn for further analysis and measurement. The flexural bending test was conductedon multiple concrete samples in a controlled laboratory environment at CONTEC, wherereinforced beams were subjected to loading from a hydraulic press. The goal of this testwas to determine whether each beam can safely support applied loads and comply with design codes. The proposed non-contact, vision-based system monitored the concrete beams to detect vertical flexural cracks as the load increased. Key performance metrics include thesystem’s ability to provide accurate, continuous measurements of crack formation and propagation, as well as width and length analysis. Results indicated that the proposed visionbased system successfully detected crack initiation and continuously tracked propagationof cracks. In conclusion, this automated, non-contact approach provides a robust, safe, andhighly accurate alternative to manual inspection. By minimizing human error and enhancingoperator safety, the system significantly improves the reliability of structural health monitoring during load testing, offering a highly scalable solution for future structural evaluationsand civil engineering applications. |
