Scalable multi-physics for fusion reactors with AURORA. (1st February 2023)

• Record Type: Journal Article
• Authors: Brooks, H; Davis, A
• Is Part Of: Plasma physics and controlled fusion. Volume 65:Number 2(2023)
• Publication Date: 2023-02-01
• Subjects: multi-physics -- neutronics -- HPC; Plasma (Ionized gases) -- Periodicals; Controlled fusion -- Periodicals; 530.44
• DOI: 10.1088/1361-6587/aca998
• Languages: English
• ISSNs: 0741-3335

Abstract

Abstract: To support the design of fusion reactors for energy production, there is an urgent need to develop multi-physics software tools capable of modelling critical tokamak components as a single cohesive whole. Although some loosely-coupled physics tools do exist, these lack fidelity and will fundamentally fail to capture any closed-loop feedback effects between separate physics modules. To address this need, we introduce a new open-source code: AURORA: A Unified Resource for OpenMC (fusion) Reactor Applications. Anticipating the need for high-fidelity simulation of complex physics and geometry mandates the need to target high performance computing from the outset. AURORA has been built upon two demonstrably scalable open-source codes: MOOSE for the provision of finite element analysis and OpenMC for Monte Carlo neutron transport. In this application, the heat deposited by neutrons is calculated by OpenMC and tallied upon an unstructured mesh, providing a source term for transient heat conduction and thermal expansion. MOOSE calculates the corresponding change in temperature and density on the same mesh, whereafter local temperature and density regions are defined via binning in these variables. Finally, these regions are updated within OpenMC as new materials having modified nuclear cross sections. The procedure is subsequently iterated until a desired stopping condition is reached. We present some qualitative results as a proof-of-concept demonstration of AURORA. We further demonstrate that there is no measurable degradation in shared-memory performance arising from wrapping OpenMC in a MOOSE application. While AURORA should provide utility as a standalone tool for coupled thermo-mechanical and neutronics analyses, we view it as a first step towards our ultimate goal of having a single suite capable of capturing non-trivial couplings between the many disciplines—encompassing in addition fluid dynamics, electromagnetism, materials science and chemistry—involved in the simulation of a tokamak.