Implementation and performance analysis of the Transient Solver in STREAM3D-GPU

Abstract

Whole-core transient simulations with high-fidelity, pin-resolved details are vital for multi-physics reactor applications, driven by the increasing need for accuracy in nuclear reactor design. However, these simulations are computationally intensive. To overcome this challenge, a GPU-based Method of Characteristics (MOC) Transient Fixed Source Problem (TFSP) solver has been developed and incorporated into the STREAM Transient code, significantly enhancing the speed of neutron transient simulations. By leveraging GPUs' parallel processing capabilities, this approach substantially reduces the computational time for direct whole-core transport calculations. The GPU-optimized solver was validated using the C5G7-TD transport transient benchmark, its effectiveness in simulating heterogeneous problems, including both step and linear control rod movements. The demonstrating results show that the 2D transient transport solver performs well, with further optimization of time steps.