Slope reliability analysis of tailings dam under coincidence of rainfall and pseudo-static motions

Tailings dam failures pose catastrophic risks to downstream communities and ecosystems, yet conventional stability assessments rarely account for the coincident effects of rainfall and pseudo-static motions or the spatial variability of soil properties. This study proposes a reliability analysis framework integrating unsaturated seepage simulation, random field modeling, and an XGBoost surrogate model to evaluate transient factor of safety (FoS) and failure probability under coupled hazards. A tailings dam case study shows that the timing of minimum FoS under advanced rainfall shifts from rainfall cessation to during-rainfall when cumulative rainfall exceeds 100 mm. Neither moderate rainfall (40 mm) nor pseudo-static action ( k h =0.05) alone induces instability, but their concurrency drives the dam to a critical state. Incorporating horizontal pseudo-static force uncertainty increases failure probability by up to 2 orders of magnitude under 40 mm rainfall. Beyond conventional wisdom, our results identify a 100-mm rainfall threshold governing peak risk timing under advanced rainfall patterns, and quantify the significant failure probability amplification driven by pseudo-static uncertainty. The XGBoost surrogate model achieves a validation R 2 of 0.78 for FoS prediction and ~ 95% computational savings versus traditional Monte Carlo simulation, providing a robust, generalizable tool for reliability assessment of tailings dams and other geotechnical systems under compound hazards.