Building a multiscale framework: An overview of the NEAMS thermal-hydraulics integrated research project

GPU-based supercomputing enables a significant advancement in Computational Fluid Dynamics (CFD) capabilities for nuclear reactors. Pre-exascale GPU-based supercomputers such as ORNL's Summit allow for the first time to perform full core CFD calculations with URANS and LES approaches. Key to this has been the development of NekRS, a novel GPU-oriented variant of Nek5000, an open-source spectral element code in development at Argonne National Laboratory. NekRS delivers peak performance for key kernels on the GPUs and good scaling performance even on CPU architectures. Recent performance measurements showed that NekRS, when running on GPUs, outperforms the CPUs by 40x. This manuscript focuses on how these calculations and novel high-resolution experimental data improve the fidelity of traditional approaches such as systems analysis codes, subchannel codes, and Reynolds-Averaged Navier-Stokes. Supercomputing simulation alone cannot impact design and safety analysis without a suitable multiscale framework. An ongoing IRP project led by Penn State is pursuing novel methods for scale bridging. We discuss details of the project, and, in particular, we review some recent progress centered on four industry-driven challenge problems.