Fast GPU simulation of reinforced concrete at the scale of reinforcement ribs by the discrete element method

The Performance Analysis of the Thermal Discrete Element Method Computations on the GPU

“…Kernel 2 performs the contact detection on the thread per particle basis. The implementation [9] based on the hashed grid is employed in this research. To avoid storing the grid by using a dense array and to save the GPU memory, each cell is mapped into a fixed-size hash table.…”

“…The curve, representing the reinforcing bar with ribs, is rotated about the y axis to define the interface surface between the concrete and the reinforcing steel bar. Sufficiently fine discretization is required to accurately describe the interface surface, which leads to large numbers of particles [9].…”

“…The implementations of the considered DEM models have been validated comparing the obtained solutions with the experimental measurements. The implemented BPM has been validated solving the uniaxial tension of reinforced concrete problem in [9]. The validation of the TBPM against the experimental measurements of temperature has been presented in Figure 2.…”

“…Thus, DEM can simulate the transition process from the continuum to discontinuum by changing the bond types between the particles without the need for specialized elements or remeshing lattices. Now, various DEM models are not limited to the analysis of granular materials [4,5], but can be effectively applied to studying rock cutting [6], coupled multi-physical problems [7], sea ice failure [8] and the fracture of reinforced concrete [9].…”

“…In 3D simulations, containing 20 000 particles, an average speedup of 19.66 has been achieved on NVIDIA Tesla K40c card. Pacevič et al [9] have implemented the BPM based on the contact bond in the GPU code for simulating the damage and fracture of cohesive solids. Zheng et al [27] have presented a GPU-based DEM-FEM computational framework implemented by CUDA FORTRAN for simulating the tire-sand interaction.…”