Hyper-reduced POD formulation for the hydro-mechanical assessment of tailings dams

The reliable assessment of tailings dam response under transient hydro-mechanical loading is a key challenge for mining infrastructure safety and risk management. High-fidelity numerical models capable of representing coupled groundwater flow and deformation in partially saturated soils provide valuable insight into internal states of the dam, but their computational demands often limit their use in operational settings, such as scenario analysis or near–real-time monitoring.

We consider a transient, nonlinear hydro-mechanical finite element model describing groundwater flow in unsaturated soils and apply a proper orthogonal decomposition (POD)–based reduced-basis formulation to accelerate simulations. While POD effectively reduces the number of unknowns, the computational cost of assembling nonlinear operators remains tied to the full-order mesh dimension, limiting efficiency gains. To address this bottleneck, hyper-reduction techniques are investigated that construct reduced approximation spaces for the nonlinear terms themselves, with the goal of alleviating computational cost relative to standard full-order finite element simulations.