Numerical simulation provides critical support for geo-disaster reduction. However, the calibration of parameters for modelling rocks remains a major challenge that undermines the trustworthiness of simulation results, particularly within the discrete element method (DEM). To this end, built on the DEM-based universal distinct element code (UDEC), we established the mapping from modelling parameters to modelled properties using artificial intelligence (AI) and implemented the inversion of modelling parameters from modelled properties by integrating the grid search method. Accordingly, we developed the open-source software, UPCal, that enables us to output modelling parameters by inputting experimental data (target modelled properties). For validation, we collected experimental data from 99 rock types and obtained their corresponding UPCal-calibrated modelling parameters. UDEC-simulated results are highly consistent with experimental data, confirming the robustness of UPCal. Furthermore, we applied the UPCal to simulate rockslides triggered by water-weakening processes, demonstrating its significant practical potential. The methodology and UPCal may be transferable to other numerical methods, such as the finite-discrete element method (FDEM), discontinuous deformation analysis (DDA), and peridynamics (PD).

Artificial Intelligence,Parameter Calibration,Numerical Simulation,Rock Mechanics,Discrete Element Method