An Improved GPU-Parallelized 2D/3D Elastoplastic-Damage-Fracture Joint Framework for Combined Finite–Discrete-Element Program

The combined Finite–Discrete-Element Method (FDEM) has demonstrated significant advantages in simulating the fracture and fragmentation process of brittle materials. However, the use of simple elastic-damage model alone is insufficient to meet the requirements for simulating the deformation of materials with ductility, and the computational cost of the discrete-element makes it a bottleneck in further development of FDEM. In this work, the 3D elastoplastic-damage-fracture joint mechanical model is proposed, and a combined algorithm has been implemented in the new FDEM package, INterdisciplinary Finite–Discrete-Element Program (INFIDEP). To improve computational efficiency and scale, a universal General-Purpose Graphics Processing Unit (GPGPU) parallelization INFIDEP code has been developed using the Compute Unified Device Architecture (CUDA) in C/C +  + . An improved parallel framework has been proposed, and a simplified contact detection algorithm that is more robust for non-uniform elements has been adopted. Using GPGPU-INFIDEP, we simulated Brazil’s tensile strength (BTS) tests under quasi-static loading conditions and Taylor bar impact tests. The BTS tests demonstrated consistent stress–strain curves and accurate deformation trends. The speedup ratio stabilized at around 400–500 times on the NVIDIA A6000 GPU platform. The Taylor bar impact tests simulated based on the joint mechanical model well reflect the deformation trend of the specimen head and the impact fracture mode, which cannot be simulated by traditional FDEM and 2D plane-strain models. These results show that GPGPU-INFIDEP offers a valuable and powerful numerical tool for studying continuum–discontinuum deformation problems of brittle and elastoplastic materials in rock engineering.