University of Twente Student Theses
Investigating limitations of computer vision structural health monitoring : a case study on the UT campus footbridge
Turków, M. (2022) Investigating limitations of computer vision structural health monitoring : a case study on the UT campus footbridge.
PDF
1MB |
Abstract: | Computer vision-based solutions are a promising tool for structural health monitoring (SHM) and asset management of civil engineering infrastructure. Computer vision (CV) systems offer many advantages over existing sensor technologies: non-contact, long-distance and precise measurements without complex installations, with lower cost and labor intensity. However, the robustness of CV-SHM remains significantly unexplored. Insights into limitations and challenges of various CV-SHM methods must be investigated, before automated input-output systems can be developed for quick analysis and asset management. The paper presents the results of literature review about applications of computer vision in short and long term bridge monitoring and damage detection. Moreover, paper presents findings from two sets of measurements collected with a use of consumer grade cameras (GoPro Hero 5/8) to extract structural displacement of 1) aluminum beam in a laboratory set-up and 2) steel girder footbridge subjected to a variety of different loading scenarios. The cameras are modified with a mounted varifocal zoom lenses recording at high resolution (720 x 1280 px) and high frame rate (240 FPS). Based on the data collection process and analysis of the results a set of conclusions is drawn, especially with regard to limitations of proposed computer vision-based structural health monitoring methodology and its potential in dynamic analysis, namely identification of modal shapes. The proposed methodology has proven successful in identification of three natural frequencies in laboratory setting within 0 - 120 Hz frequency spectrum. In the field conditions, the researcher was able to identify the natural frequencies at lower bandwidth (0 – 5 Hz). Both in laboratory and field environment, it has been proven successful to extract the dynamic signature of the structure in the form of 1st and 2nd mode shapes. The results obtained via image processing, namely template matching, are compared against accelerometry findings. No significant differences have been found between accelerometers and image processing. Furthermore, limitations in terms of hardware specification, environmental factors, operational conditions, and others are discussed. |
Item Type: | Essay (Bachelor) |
Faculty: | ET: Engineering Technology |
Programme: | Civil Engineering BSc (56952) |
Link to this item: | https://purl.utwente.nl/essays/93547 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page