Cloud and incremental dynamic analysis (IDA) are the two most commonly used methods for seismic fragility analysis. The two methods differ significantly in the number of ground motions and whether these motions are scaled. This paper designed a random selection procedure to thoroughly discuss the influence of ground motion combinations encompassing different numbers of motions on the Cloud-based and IDA-based seismic fragility analysis for underground subway station structures. Focusing on a shallow-buried single-story station structure, a nonlinear dynamic time-history finite element analysis model of soil-structure interaction was developed. 400 ground motions were selected for random combination to perform Cloud-based seismic fragility analysis, and 20 ground motions were selected for random combination to perform IDA-based analysis. The results showed that the number of ground motions has a significant influence on the seismic fragility analysis in both Cloud and IDA, especially on the prediction of damage probability for higher seismic performance levels and when the PGA exceeded 0.3 g. In regions with a low probability of strong earthquakes, this paper recommended using no fewer than 10 and 220 ground motions in the IDA-based and Cloud-based seismic fragility analyses, respectively. In regions with a high probability of strong earthquakes, the optimal number of ground motions should be increased to 300 for Cloud-based analysis and 15 for IDA-based analysis.
