Although the influence of duration of ground motion on the seismic response of aboveground structures is clearly recognized, its influence on underground structures remains unclear. To this end, this study performs incremental dynamic analyses under both short and long duration ground motions, to quantify the significance of the duration effect on the seismic fragility of subway stations. A two-dimensional soil-structure system is established on the basis of the Daikai subway station, consisting of an elastoplastic soil model and a concrete damage plasticity model. A set of thirty spectrally matched ground motions with varying significant durations (D5-95) are employed. In particular, using the center column total compressive damage index (DTCD) and peak inter-story drift ratio (IDR) as structural demand measures (DM), the percentage difference in fragility curves between long and short duration is evaluated by accounting for six suits of damage state thresholds. Correlations between the two DMs and D5-95 show that ground motion duration affects significantly the seismic fragility of subway stations. Overall, the duration effect is not detected in the minor damage state and becomes more pronounced in the collapse state, suggesting that the duration effect increases as the damage state threshold increases. The median collapse capacity for long duration ground motions is up to approximately 60% or 37% lower than that for short duration ground motions, when a peak IDR or DTCD are adopted, respectively. The results of this study highlight the great importance of properly considering duration when selecting earthquake records for seismic fragility assessment of subway stations.

Owing to the repeatedly observed long-duration ground motions (GMs), the duration effect is becoming increasingly critical for the seismic design and assessment of important infrastructures. In this study, dynamic analyses are performed to evaluate the influence of GM duration on the inelastic seismic response of subway stations. Based on the Daikai subway station in Kobe, Japan, a two-dimensional numerical model incorporating a concrete damage plasticity model and a soil nonlinear model is developed. Twenty-five spectrally matched records with different significant durations (D 5-95 ) are selected for dynamic analyses. The responses are evaluated based on various engineering demand parameters (EDPs) including the internal force, drift ratio, element damage index, and the total damage of the structural member. The results show that for weak earthquakes (PGA = 0.15 g) the duration has a negligible influence, while for strong earthquakes (PGA = 0.45 g) the duration effect is particularly pronounced due to the significant cyclic degradation of stiffness and strength of the material. Compared to the location-dependent maximum internal forces and the element damage index, the maximum drift ratio and the total compressive damage index are suitable EDPs, since they correlate well with D 5-95 . On average, for every 10-s increase in D 5-95 of strong GMs used in this study (PGA = 0.45 g), the maximum drift ratio of the side wall and center column increases by 0.12 % and 0.2 % while the total compressive damage index of the side wall and center column increases by 0.01 and 0.04, respectively. The duration effect is more pronounced for the center column implying that damage to the weak structural member is exacerbated by long-duration GMs. The increased column stiffness or elastic soil behavior may reduce the duration effect. This study highlights that strong GMs having similar amplitude and response spectrum with different significant durations can significantly affect the inelastic seismic response of subway stations.