Accelerating uncertainty quantifiations in geohazard analysis

Physics-aware AI simulators to accelerate probablistic analysis in geohazards

Geotechnical systems are inherently uncertain due to the spatial variability of geomaterials, requiring probabilistic simulations for reliable analysis and design. We develop Random Graph Network Simulators (RGNS) to accelerate stochastic simulations by learning the underlying physics of geomechanical processes from high-fidelity numerical models. This framework enables efficient uncertainty quantification, reliability analysis, and probabilistic hazard assessment that would otherwise be computationally prohibitive.

(Choi & Lee, 2026)

Probabilistic landslide hazard zoning based on exceedance probability of post-failure distance.

References

2026

  1. Accelerating stochastic simulation of post-failure landslide runout using a random graph neural network-based simulator
    Yongjin Choi and Seungjun Lee
    Computers and Geotechnics, 2026