A novel heterogeneous CPU/GPGPU-accelerated 3D CDEM and its application to modeling deep roadway excavation

Abstract

To improve the accuracy and computational efficiency of the CDEM for deep coal mine roadway excavation modeling, this study proposes a heterogeneous CPU/GPGPU-accelerated solver that integrates a mixed continuous–discontinuous media algorithm. The solver employs an explicit time integration method combined with a modular approach for 3D tetrahedral solid finite elements and fracturable penalty springs, which model rock fracture behavior and the transition from continuum to discontinuum in rock masses. To maximize computational efficiency, the solver uses a hybrid CPU/GPGPU framework with SIMD parallel techniques, achieving up to 600-fold speedup on a single GPGPU. The solver’s accuracy is validated for both quasi-static and dynamic problems, and its scalability across different hardware accelerators is demonstrated. Modeling results from the 22nd mining area of the Quandian coal mine show significant shear deformation and crack evolution in the soft rock, particularly at the intersection of the roof slab and sidewall, where stress concentration and large deformation are most pronounced. These findings validate the efficiency and reliability of the proposed method for simulating and analyzing underground excavation processes.