Vision-based probabilistic post-earthquake loss estimation for reinforced concrete shear walls

In this paper, a probabilistic methodology based on image analysis is proposed for earthquake-induced loss estimation in rectangular reinforced concrete shear walls (RCSWs) in compliance with FEMA P-58. The methodology is developed using a databank of 285 surface crack patterns gathered from the result of experimental research on 99 rectangular RCSWs with wide-ranging structural and geometric characteristics. For the complexity extraction, three fractal geometry indices of the damaged specimens are taken into consideration. Fragility functions are generated to determine the probability of exceeding a FEMA P-58 compliant damage state based on the various methods of repair for specific complexity indices. The fragility functions are obtained using four common probabilistic distributions including lognormal, Weibull, gamma, and beta. The fitness of the fragility curves is examined by two goodness-of-fit tests, namely the Kolmogorov-Smirnov test and Lilliefors test. Results show that the Weibull is the best-fitting distribution for the majority of damage states and succolarity possesses the best results in goodness-of-fit tests among fractal geometry indices. Empirical loss indices are also derived in this paper aimed at optimizing the goodness-of-fits. Finally, the proposed procedure is utilized for a sample specimen from the databank at various damage levels.