ABSTRACT
In cohesive soil, the tunnel structures are susceptible to unexpected deformation and damage, especially under sequential natural hazards such as earthquakes followed by flooding. These Multihazard scenarios are vital during lifecycle and pose challenges for maintaining structural integrity, functionality, and long-term serviceability. This study outlines tunnel resilience modeling and prediction approach for post-earthquake (PEQ) flood events in cohesive soil. It utilizes a macro-modeling approach by adapting PEQ geometric nonlinearities of the Tunnel structures and assessing them for flood-induced hydrostatic loads. Insights from 108 simulations are analyzed to evaluate the Tunnel’s resilience and develop a curated dataset for predicting deformation. The study recommends two predictive models: CatBoost and Model Agnostic Meta-learning (MAML) based on M-CatBoost models. The CatBoost model is found to be the best-performing model. The precise modeling and prediction approaches are aimed to add value to the predictive maintenance and disaster preparedness of tunnel infrastructures exposed to complex Multihazard conditions.
