Iterative Methods for Neutron and Thermal Radiation Transport Problems
Abstract
We develop, analyze, and test iterative methods for three kinds of multigroup transport problems: (1) k-eigenvalue neutronics, (2) thermal radiation transport, and (3) problems with “upscattering,” in which particles can gain energy from collisions.
For k-eigenvalue problems, many widely used methods to accelerate power iteration use “low-order” equations that contain nonlinear functionals of the transport solution. The nonlinear functionals require that the transport discretization produce strictly positive solutions, and the low-order problems are often more difficult to solve than simple diffusion problems. Similar iterative methods have been proposed that avoid nonlinearities and employ simple diffusion operators in their low-order problems. However, due partly to theoretical concerns, such methods have been largely overlooked by the reactor analysis community. To address theoretical questions, we present analyses showing that a power-like iteration process applied to the linear low-order problem (which looks like a k-eigenvalue problem with a fixed source) provides rapid acceleration and produces the correct transport eigenvalue and eigenvector. We also provide numerical results that support the existing body of evidence that these methods give rapid iterative convergence, similar to methods that use nonlinear functionals.
Thermal-radiation problems solve for radiation intensity and material temperature using coupled equations that are nonlinear in temperature. Some of the most powerful iterative methods in use today solve the coupled equations using a low-order equation in place of the transport equation, where the low-order equation contains nonlinear functionals of the transport solution. The nonlinear functionals need to be updated only a few times before the system converges. We develop, analyze, and test a new method that works in the same way but employs a simple diffusion low-order operator without nonlinear functionals. Our analysis and results show rapid iterative convergence, comparable to methods that use nonlinear functionals in more complicated low-order equations.
For problems with upscattering, we have investigated the importance of linearly anisotropic scattering for problems dominated by scattering in Graphite. Our results show that the linearly anisotropic scattering encountered in problems of practical interest does not degrade the effec-tiveness of the iterative acceleration method. Additionally, we have tested a method devised by Hanuš and Ragusa using the semi-consistent Continuous/Discontinuous Finite Element Method (CDFEM) diffusion discretization we have devised, in place of the Modified Interior Penalty (MIP) discretization they employed. Our results with CDFEM show an increased number of transport iterations compared to MIP when there are cells with high-aspect ratio, but a reduction in overall runtime due to reduced degrees of freedom of the CDFEM operator compared to the MIP operator.
Subject
transportdiffusion
acceleration
preconditioner
k-eigenvalue
radiative transfer
upscattering
low-order
Citation
Barbu, Anthony Petru (2022). Iterative Methods for Neutron and Thermal Radiation Transport Problems. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /197351.