CAD-Based Geometry Representations for Monte Carlo Fusion Neutronics Methods and CSG vs. DAGMC Performance Tradeoffs in OpenMC

Abstract

Fusion reactors utilizing deuterium and tritium fuel produce high-energy 14.1 MeV neutrons, necessitating a thorough understanding of their behavior for effective reactor design. Neutron transport codes play a critical role in determining key parameters such as tritium breeding ratio, neutron wall loading, and heat deposition, vital for assessing operational considerations. Monte Carlo (MC) radiation transport methods have become standard in fusion neutronics due to their ability to handle energy and angular variables continuously. However, manual modeling of complex fusion geometries with traditional constructive solid geometry (CSG) methods remains labor-intensive, prompting the integration of computer-aided design (CAD) models into MC radiation transport. This thesis investigates the integration of CAD-based geometry representations into MC radiation transport, focusing on computational performance implications of the Direct Accelerated Geometry Monte Carlo (DAGMC) approach. This work examines different neutronics model representations, including CSG, Unstructured Mesh (UM), and DAGMC for the practical solutions they can provide for fusion neutronics needs. Tracking algorithms associated with each representation are explored, highlighting UM and DAGMC’s versatility in the way they integrate with CAD-based design processes. Performance comparison between CSG and DAGMC geometries in OpenMC is analyzed by evaluating particle simulation rates and memory usage across four progressively complex fusion-like models. Performance results reflect positively on DAGMC transport, but areas of future work are identified for more comprehensive results. From the lens of computational performance, this study contributes to determining the viability of CAD-based geometry representations for use in fusion-relevant MC radiation transport.