CNERG research lies at the middle of a continuum of computational work, with pure methods development at one end and pure systems analysis at the other. CNERG’s goal is to help improve capability for real analysis of complex systems by investigating and implementing new methods in production software tools. The primary product of our work is a suite of technology to enable these analyses, although much of our funding comes from analysis efforts.
The cornerstone of the CNERG neutronics group’s work is the Direct Accelerated Geometry Monte Carlo (DAGMC) toolkit. This ray-tracing interface to the MOAB software library allows Monte Carlo radiation transport software to perform ray-tracing and related operations directly on CAD-based representations of complex geometry. DAGMC has been successfully integrated with MCNP5 from Los Alamos National Laboratory to produce the DAG-MCNP5 package.
Current research based on the DAGMC technology includes:
DAGMC performance improvements using novel surface meshing algorithms: The benefit of the OBB-tree accelerations that are vital to DAGMC’s performance can be undermined by a faceting of the CAD surfaces that produces vertices that are shared by too many facets: so-called high valence vertices. Different faceting algorithms will be compared with some mesh cleanup algorithms to identify schemes to arrive at a faceting that is better suited to the OBB-tree acceleration scheme for challenging problems.
Unstructured mesh tallies with alternative estimators: With complex geometry and 3-D source definitions now common-place, the focus turns to the ability to tally results in Monte Carlo simulations with better fidelity. The Cartesian overlay mesh common to most solutions introduces physical approximations for complex geometry that does not align conveniently with such a grid. This work uses conformal unstructured mesh to define the tally mesh for a traditional track-length estimator, and then explores alternative estimators such as the kernel density estimator applied at mesh nodes rather than across mesh elements.
High-fidelity shutdown dose workflow for activated systems: Known as the Rigorous-2-Step method, or R2S, in fusion neutronics, this workflow performs a multi-group spatially resolved neutron transport calculation to determine neutron fluxes throughout the system, uses those fluxes to determine the isotopic change in the activated materials, and then samples the photon source that arises from that activation to perform a detailed photon dose calculation at various times after the shutdown of the neutron source. The critical research & development steps are the generation of material definitions for Cartesian grid elements superimposed on a complex geometry, and the efficient sampling of a grid-based multi-group photon source.
Hybrid deterministic-Monte Carlo workflow: For nearly 2 decades, the use of deterministic hybrid approaches to Monte Carlo variance reduction have been explored and refined. Extending this capability to the CAD-based DAGMC transport problems requires the generation of a mesh for use in the deterministic solution. Leveraging the improvements made for the R2S workflow above, this process should be straightforward and allow most of the normal hybrid technology to be applied directly to DAGMC-based problems.
In addition, students in this area often pursue spin-off projects through collaborations with national laboratories.
The CNERG fuel cycle group’s Cyclus project was recently selected as the base infrastructure for the Department of Energy’s Next-Generation Fuel Cycle Simulator. This project aims to address gaps in the available simulator tools by:
- modeling discrete facilities and material transactions
- using an agent-based approach with dynamically loadable plugins to encapsulate behavior
- open source development pratices to lower the barrier to adoption
- common physics infrastructure with alternative output layers for different user communities.
In addition to developing the core technology as a team, the following research areas are also being pursued by individual CNERG members:
- geologic repository model (Cyder) with plugin architecture for repository physics models
- application of social-behavioral models of individual agents to study nuclear material diversion and proliferation issues.