Ground Motion-Dependent Rapid Damage Assessment of Structures Based on Wavelet Transform and Image Analysis Techniques

Abstract

Rapid and accurate evaluation of the damage state of structures after a seismic event is critical for postevent emergency response and recovery. The existing rapid damage evaluation methodology is typically based on fragility curves incorporated into earthquake alerting platforms. However, the extent of damage predicted solely based on the fragility curves can vary significantly depending on ground motion characteristics. This paper presents a methodology for damage assessment of structures while accounting for temporal and spectral nonstationarity of ground motions using continuous wavelet transform and image-analysis techniques. The methodology involves the establishment of a prediction model for wavelet transform of ground motions and damage state of a structure using convolutional neural networks. The methodology is demonstrated in this paper through two case studies: (1) a low-rise nonductile concrete building frame in California and (2) a four-span concrete box-girder bridge in California. The proposed methodology identified damage states with an accuracy greater than 75% in both cases.