Development of parallel explicit finite element sheet forming simulation system based on GPU architecture

Sheet forming simulation is very important for vehicle body design. Due to the increase of complexity and scale of the CAE model, a tradeoff between the accuracy and efficiency become the bottleneck for application. Therefore, a parallel explicit finite element (FE) based on graphics processing unit (GPU) architecture for sheet forming is developed. Implementation details with computer unified device architecture (CUDA) are considered in this work. A pre-index strategy is suggested for parallelization of nodal force assembling. Parallel reduction method is introduced to calculation of the global time step. To ensure the reliability and accuracy of the GPU-based program, double precision floating and intrinsic functions are implemented for the explicit FE computing. The simulation results based on a commercial NVIDIA GTX285 device can obtain about 27X speedup than on a Intel Q8200 CPU, which demonstrates the efficiency of the parallel sheet forming simulation system.

► A parallel explicit FE based on CUDA architecture for sheet forming is developed. ► A pre-index strategy is developed for parallelization of nodal force assembling. ► Parallel reduction method can be used to calculation of the global time step. ► Techniques to ensure computation accuracy on GPU are developed.