Compression–separation history bonded-contact model for improved UCS–BTS matching in DEM of brittle rock

Traditional bonded-particle models in the discrete element method often struggle to reproduce both unconfined compressive strength and Brazilian tensile splitting strength of brittle rock using a single micro-parameter set, because the same bond formulation is commonly applied to closure-dominated and opening-dominated loading paths. This study proposes a compression–separation history bonded-contact model to address this limitation. The model introduces closure-driven stiffness hardening to represent progressive microcrack closure under compression, together with a history-dependent damage rule that degrades normal and tangential bond strengths more strongly under opening-dominated kinematics. A two-stage surrogate-assisted calibration procedure is then used to identify the model parameters from laboratory UCS and BTS data. The framework is validated against three matched sandstone specimen groups spanning a wide experimental UCS-to-BTS ratio range. Quasi-static acceleration is assessed using an inertial number criterion and a rate-insensitivity check, enabling loading rate increases of up to 10 5 for UCS and 10 4 for BTS. Across all groups, the surrogate predicts DEM peak responses with errors below 5%, and the final calibrated simulations reproduce the experimental UCS-to-BTS ratios with absolute relative errors below 10%. The results show that the proposed bonded-contact model improves simultaneous DEM representation of compressive and tensile rock strength while providing an efficient and reproducible calibration framework.