Scoping analysis of the Advanced Test Reactor using SN2ND

Abstract

A detailed set of calculations was carried out for the Advanced Test Reactor (ATR) using the SN2ND solver of the UNIC code which is part of the SHARP multi-physics code being developed under the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program in DOE-NE. The primary motivation of this work is to assess whether high fidelity deterministic transport codes can tackle coupled dynamics simulations of the ATR. The successful use of such codes in a coupled dynamics simulation can impact what experiments are performed and what power levels are permitted during those experiments at the ATR.The advantages of the SN2ND solver over comparable neutronics tools are its superior parallel performance and demonstrated accuracy on large scale homogeneous and heterogeneous reactor geometries. However, it should be noted that virtually no effort from this project was spent constructing a proper cross section generation methodology for the ATR usable in the SN2ND solver. While attempts were made to use cross section data derived from SCALE, the minimal number of compositional cross section sets were generated to be consistent with the reference Monte Carlo input specification. The accuracy of any deterministic transport solver is impacted by such an approach and clearly it causes substantial errors in this work. The reasoning behind this decision is justified given the overall funding dedicated to the task (two months) and the real focus of the work: can modern deterministic tools actually treat complex facilities like the ATR with heterogeneous geometry modeling.SN2ND has been demonstrated to solve problems with upwards of one trillion degrees of freedom which translates to tens of millions of finite elements, hundreds of angles, and hundreds of energy groups, resulting in a very high-fidelity model of the system unachievable by most deterministic transport codes today. A space-angle convergence study was conducted to determine the meshing and angular cubature requirements for the ATR, and also to demonstrate the feasibility of performing this analysis with a deterministic transport code capable of modeling heterogeneous geometries. The work performed indicates that a minimum of 260,000 linear finite elements combined with a L3T11 cubature (96 angles on the sphere) is required for both eigenvalue and flux convergence of the ATR. A critical finding was that the fuel meat and water channels must each be meshed with at least 3 "radial zones" for accurate flux convergence. A small number of 3D calculations were also performed to show axial mesh and eigenvalue convergence for a full core problem. Finally, a brief analysis was performed with different cross sections sets generated from DRAGON and SCALE, and the findings show that more effort will be required to improve the multigroup cross section generation process.The total number of degrees of freedom for a converged 27 group, 2D ATR problem is ~340 million. This number increases to ~25 billion for a 3D ATR problem. This scoping study shows that both 2D and 3D calculati...

Development status of PROTEUS-MOC

“…Assuming a radial mesh containing E elements can be decomposed into a structured grid with E=T 2 total elements, the size of the solution vector can be estimated to be roughly 2 2 2 KT NG KTNG + where we assume an axial mesh with K planes, an angular cubature with N directions, and G energy groups. For 10 6 elements, 20 planes, 200 angles, and 100 energy groups this defines a problem with ~10 12 DOFs noting that PROTEUS-SN typically defines 100-200 axial meshes.…”

“…The first step in that process is to take advantage of the DeCART geometry concept of building an extruded geometry, but use the unstructured mesh treatment from PROTEUS-SN to define the planer geometry. This allows complex geometries such as the Advanced Test Reactor [12] shown in Figure 1 to be modeled with PROTEUS-MOC. With regard to the transport equation, the focus of this fiscal year was to develop and demonstrate a methodology that does not suffer the algorithmic problems of DeCART.…”

Continued Research for Improvement of PROTEUS-SN

“…The general goal is to reduce the uncertainties and biases in various areas of reactor design activities by providing enhanced prediction capabilities. As a fast reactor component of SHARP, a high-fidelity deterministic neutron transport code named PROTEUS-SN [1][2][3][4][5][6] was developed specially for analysis and design of sodium fast reactors (SFRs). The application scope of PROTEUS-SN was to range from conventional homogenized assembly approaches to explicit geometry, time-dependent transport calculations coupled with thermal-hydraulics and structural mechanics for reactor accident simulations.…”

FY2012 status report on subgroup library development

“…While PROTEUS-SN2ND has demonstrated good accuracy for homogeneous fast reactor problems [4,5] and partially heterogeneous fast reactor problems [6], when applied on fully heterogeneous problems such as the ATR [7], the results were not as exemplary. Of course it is not the solver of the transport equation itself that introduces the large errors, but the multi-group cross section data the user feeds into the solver that is [7].…”

“…Of course it is not the solver of the transport equation itself that introduces the large errors, but the multi-group cross section data the user feeds into the solver that is [7]. The fundamental problem in generating such data arises from the equivalence that is assumed to exist between the cross section generation step (i.e.…”