A Multiscale Numerical Approach to Long‐Term Stability Analysis in Rock Engineering

Long‐term stability of rock engineering structures is controlled by progressive strength degradation under sustained loading, yet direct experimental determination of long‐term strength (LTS) is severely constrained by impractically long testing durations and strong material variability. To address this limitation, this study proposes a feasible numerical framework for long‐term stability analysis by linking time‐dependent microscale damage evolution to engineering‐scale strength parameters. The four‐dimensional lattice spring model (4D‐LSM), incorporating creep, plasticity, and progressive bond fracture, is calibrated using conventional creep test data, in which the maximum bond‐fracture threshold governs long‐term failure time. Time effects are introduced through temporal evolution of model input parameters, enabling numerical creep tests to reproduce stress‐time‐to‐failure behavior and predict LTS. The numerically obtained LTS data are fitted with an empirical stress‐time relationship and further converted into time‐dependent equivalent Mohr‐Coulomb cohesion and friction angle. These time‐varying parameters are then implemented in a classical strength reduction finite element framework, and a machine learning model is adopted to establish the correlation between the factor of safety and the degrading strength parameters. The proposed approach facilitates the practical prediction of long‐term stability of rock engineering structures using temporally evolving strength parameters derived from 4D‐LSM simulations, thereby providing a mechanistically sound and computationally efficient tool for the assessment of delayed failure in rock engineering.