Structural damage recognition and residual capacity prediction using computer vision for reinforced concrete members

• Main Author: Tang, Jun Yuen
• Other Authors: Li Bing
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Civil engineering::Construction technology
• Online Access: https://hdl.handle.net/10356/158441
• Institution: Nanyang Technological University

This project presents the results of using computer vision-based technology to relate the damage levels and damage states of columns to quantitative load estimates in structural components for automated façade inspections. Image-processing and machine learning regression techniques were utilized to create estimation models able to predict the damage states and levels of the columns based on superficial crack patterns in the images. The model was trained using a database of 190 individual specimens which produced 639 images at different loading stages. Various textural and geometric attributes of surface crack patterns were defined and evaluated for their usefulness in building the estimation models. Statistical error measures and cross-validation techniques are used to quantify the prediction accuracy and training/ test methods were considered relative to the actual field application scenarios. The results show that the estimated models work well across ranges of geometries, loading patterns, loading types, concrete strengths, and reinforcement details. The physical size of the practical columns can be approximated by keying in the scaled dimensions of the specimens into the estimation model.