Loading...

Post-Earthquake Damage Assessment of Structural Walls Using Image Processing and Machine Learning Techniques

Asjodi, Amir Hossein | 2023

61 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 56180 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Mohtasham Dolatshahi, Kiarash
  7. Abstract:
  8. The main objective of this research is to use the visual features of damage in evaluating the post-earthquake status of the damaged structures. Structural health monitoring, as a comprehensive approach, includes methods for identifying the damaged structures, estimating damage location, measuring damage extent, and methods for reconstructing and repairing structures. The present study is specifically focused on the second and third stages of structural health monitoring, including “estimating the location of damage” and “measuring the extent of damage”. Various methods for evaluating the condition of earthquake-damaged structures are available in technical literature; one of the most common is visual inspection. Visual inspection incorporates a wide range of assessments, including checking the appearance of the damaged components, examining the surface damage, such as cracks and crushing, and conducting in-situ tests. Among these methods, exploring the visual features of surface damage, such as the extent of cracking and crushing in structural components, is of great importance. Although most codes and standards over the past decades focused on limiting the crack pattern characteristics as a potent damage indicator, determining the state of damaged structures based on crack patterns heavily depends on the experience and judgment of the inspector. Therefore, a new generation of studies has focused on introducing computer-aided tools as an alternative to engineering judgment. In addition to compensating for the shortcomings of visual inspection, advances in computer science have also significantly improved processing speed and attracted much attention from engineering communities. For example, machine learning and image processing techniques are state-of-the-art computer-aided methods used to quantify crack patterns rapidly and extract the main features for the structural health monitoring stream. Image processing algorithms and machine learning models can promptly assess a wide range of damaged areas and detect and measure various types of damage. Consequently, the extracted damage features are used as inputs to statistical models to determine the remaining strength of a damaged member. This study investigates the feasibility of using visual features of damage patterns in evaluating the post-earthquake status of damaged structural components. At a glance, damage patterns generally include information on crack patterns and crushed areas. Therefore, by studying the characteristics of these patterns and employing computer-aided tools such as image processing techniques and machine learning algorithms, the extent of damage in a structural member can be estimated, and the remaining strength can be determined. The proposed framework of this study will be extended to various types of structural components, such as reinforced concrete shear walls and unreinforced masonry walls
  9. Keywords:
  10. Damage Assessment ; Image Processing ; Machine Learning ; Cracking ; Crush ; Post-disaster Inspection

 Digital Object List

 Bookmark

No TOC