Quantifying seismic damage in RC walls with image analysis

Abstract

This study presents an image-assisted method for seismic damage evaluation of RC walls, integrating image processing, feature ranking, and machine learning. The method utilizes features from surface crack images, such as crack patterns and ratios, combined with design parameters, to predict damage levels and states. Seismic damage evaluation tools based on damage state, strength degradation, and drift ratio are introduced as indicators for quantifying structural damage. The approach is tested using 450 crack images, and feature selection was applied to identify the most important predictors. The results demonstrate high accuracy with an R-squared of 0.87 and an RMSE of 0.28.