Deep learning enabled seismic fragility evaluation of structures subjected to mainshock-aftershock earthquakes

Abstract

Mainshock-aftershock earthquakes have gained significant attention since accumulated damages induced by multiple shocks are likely to cause failure of structures. This paper presents a deep learning approach based on a Gated Recurrent Unit (GRU) network for assessing the seismic fragility of structures under mainshock-aftershock scenarios. The GRU network is utilized to create a surrogate model that captures the nonlinear relationship between seismic responses and mainshock-aftershock earthquakes. Subsequently, seismic fragility analysis is conducted based on double incremental dynamic analysis, employing the trained GRU network. A single-degree-of-freedom system with Bouc-Wen hysteretic behavior was investigated to demonstrate the proposed approach. The results indicate that the approach shows a substantial reduction in computational costs and holds promising potential for evaluating the seismic fragility of structures exposed to mainshock-aftershock earthquakes.