The Legendre Polynomial Axial Expansion Method
Nicholas Herring · Deep Blue (University of Michigan) · 2022
Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.
Deep Blue (University of Michigan) · University of Michigan · dissertation · 2022
Abstract
Designing and analyzing nuclear reactors requires the development of software tools for the simulation of reactor physics. One of the most important aspects of reactor simulation is the determination of the distribution of neutrons in the system. This distribution is described by a Partial Differential Equation (PDE) called the Neutron Transport Equation (NTE). Typical reactor systems are too complex to analytically solve the NTE, so instead computational methods of solving the equation are developed and used. A variety of methods to efficiently solve the 3D NTE for reactor systems have been devised. Prominent among these methods are the 2D/1D methods.
Traditional 2D/1D methods generate a 3D solution to the NTE by iteratively solving 2D radial transport equations coupled with 1D axial PN equations. The standard iterative methods to solve the traditional 2D/1D equations have inherent instabilities. These instabilities can lead to divergence of the problem solution. The instabilities are of particular concern in problems with neutronically important void-like regions adjacent fuel. This work proposes to fill in these limitations of traditional 2D/1D methods in Michigan PArallel Characteristics-based Transport (MPACT) for such problems using the axial expansion method.
Proposed by Argonne National Laboratory (ANL) and first implemented in Proteus-MOC, axial expansion methods directly solve the full 3D NTE, but use Method of Characteristics (MOC) for the discretization of the radial direction, while using discontinuous finite element methods for the axial direction treatment. This work develops a new discretization using an explicit Legendre polynomial form for the axial expansion, and introduces an alternative axial coupling treatment to allow for non-extruded axial meshes. One of the most computationally expensive portions of axial expansion methods involves the calculation of matrix exponentials.
While the matrix exponential is well defined, general explicit forms are only known for up to 3 by 3 matrices. The explicit Legendre Polynomial Axial Expansion Method (LPAEM) formulation allowed investigation into the precise form of the matrix exponentials present in the calculation. This investigation revealed a form that lent itself to matrix exponential tables, which showed substantial speedup compared to analytic or iterative methods of computing the matrix exponential. This work implements the LPAEM with non-extruded axial mesh coupling and matrix exponential tables in the reactor physics code MPACT.
Using this implementation, numerical investigations into the stability of the method in comparison to traditional 2D/1D are performed. The investigations performed in this work demonstrate the stability of the axial expansion method in systems with neutronically important void-like regions adjacent to fuel, that fail to converge using traditional 2D/1D.
Open the resolved locator · 10.7302/4752
Metadata licence: OpenAlex metadata (CC0)