Extended fragility surfaces for unreinforced masonry walls using vision-derived damage parameters

Abstract

This paper introduces a novel methodology for identifying the post-earthquake damage states of unreinforced masonry walls using visual damage features. Current guidelines provide a qualitative description of surface cracking and crushing to determine the damage states. However, evaluating the damage states according to the qualitative damage description is highly affected by the experience and judgment of the inspector. Therefore, the present study quantifies the qualitative description of guidelines to measure the probability of reaching different damage states using 3D fragility surfaces. To this end, 333 images of 168 damaged unreinforced masonry walls at different drift ratios between 0.0 and 2.5 percent are collected, and the magnitude of cracking and crushing are measured. Three image processing filters are employed to measure various types of cracking and crushing areas, which are the main characteristics described by the guideline for damage state identification. Then, a new methodology for generating the 3D fragility surface, entitled the Box-counting method, is introduced to study the influence of cracking and crushing on the damage states. The fragility surfaces are developed using the information on crack pattern length and crushing areas that enable determining the damage states of the masonry walls just by looking at the surface damages. The Tree-based machine learning regression learners are also employed to formulate the 3D fragility surface. The proposed framework of this paper can be used for risk assessment of unreinforced masonry walls and is a reliable basis for generating multi-variable 3D fragility surfaces.