Artificial intelligence (AI) and virtual/augmented reality (VR/AR) are facilitating objective and fast assessment of infrastructures. Computer vision advancements are also transforming traditional methods into automated information modeling and decision support systems. These advancements offer new opportunities to combat growing challenges that threaten infrastructure systems. In particular, climate change, aging structures, and population growth have intensified threats to infrastructure, requiring methods for evaluating infrastructure quickly after a disaster. VR uses cameras and sensors to provide images of the current state of a structural system, including the pattern of concrete cracks. A novel framework for analyzing the degree of damage in cracked reinforced concrete shear walls (RCSWs) is presented in this paper by leveraging virtual reality (VR) technology. An automated and unbiased approach is enabled by converting images of crack patterns on the surface of concrete structures into graphs. It is possible to extract relevant information from graphs, including graph features, to quantify the extent of damage using graph theory. A machine learning algorithm is then trained using these features to predict the extent of the damage. To validate the approach, the framework was applied to data collected from three RCSWs that were subjected to quasi-static cyclic loading. ML was used to predict the secant stiffness and its descending trend during each load cycle in the experimental test. The VR-based approach achieves high R2 scores of 0.90, 0.91, and 0.99 in a machine learning regression, indicating the framework's success.
|