monte carlo – Advanced Reactors Group

Adjoint-based uncertainty quantification in multiphysics reactor modeling

Manuele Aufiero, Michael Martin, Massimiliano Fratoni

Coupled neutronics-thermal/hydraulics simulations are of great interest for the analysis and design of nuclear reactors. Ongoing studies of advanced and GEN-IV reactors call for the adoption of accurate modeling tools that are based on Monte Carlo neutron transport and CFD-based T/H solutions. In this framework, the capability to propagate uncertainties in the input data through the coupled simulation is highly desirable.

Recently, Generalized Perturbation Theory (GPT) methods have been implemented in continuous energy Monte Carlo codes, broadly expanding their capabilities. Some of these methods (e.g., available in the Serpent code) are suitable to be adopted in combination with Open Source finite-volume libraries for continuum mechanics solvers (e.g., the OpenFOAM C++ multiphysics toolkit).

The present project involves the projection of the input uncertainties and the reactor generalized responses onto sets of orthogonal basis functions, along with the adoption of extended GPT methods for the calculation of sensitivities in the coupled problems. The comparison of nuclear data uncertainty propagation results against standard methods in simple benchmark cases shows that the new approach might provide a reliable and efficient option for Uncertainty Quantification in multiphysics problems.

Sensitivity and uncertainty analysis in Monte Carlo transport and burnup calculations

Yishu Qiu, Manuele Aufiero, Kan Wang (Tsinghua University), Massimiliano Fratoni

There is an increasing interest to couple Monte Carlo (MC) transport calculations to depletion/burnup codes since Monte Carlo codes can provide exact flux distributions or cross sections. One of the main concerns about using a MC transport-depletion method is how uncertainties from Monte Carlo statistical uncertainties as well as nuclear data uncertainties are be propagated between the Monte Carlo codes and burnup codes. This project is going to develop sensitivity and uncertainty analysis capabilities in RMC-Depth which is an in-coupling Monte Carlo transport-depletion code developed by Tsinghua University, China. To be more specific, the goals of this project are:
1. Study methods suitable for computing k-eigenvalue sensitivity coefficients with regard to the continuous-energy cross sections and implement them in RMC; conduct sensitivity and uncertainty analysis of the effective multiplication factor to nuclear data uncertainties in the transport calculations.
2. Study methods appropriate for computing general response sensitivity coefficients with regard to the continuous-energy cross sections and implement them in RMC; conduct sensitivity and uncertainty analysis of general responses in the form of linear response functions, such as relative powers, isotope conversion ratios, multi-group cross sections, and bilinear response functions, such as adjoint-weighted kinetic parameters, to nuclear data in the transport calculations.
3. Study the methods suitable for analysis and uncertainty propagation in Monte Carlo transport-burnup calculations. With the proposed methods, propagate uncertainties in the Monte Carlo transport-burnup calculations that come from nuclear data, the Monte Carlo statistics, the isotope number densities, and the cross-correlations between the nuclear data and the number densities. These effects should be analyzed separately in each burnup step of the burnup calculations.
4. Study the methods suitable for uncertainty qualifications for other parameters such as temperature and system dimensions.

Angle-Informed Hybrid Methods

Madicken Munk, Garrett Baltz, Rachel Slaybaugh, Richard Vasques

Hybrid methods for radiation transport aim to use the speed and uniform uncertainty distribution obtained from deterministic transport to accelerate and improve performance in Monte Carlo transport. An effective use of this type of transport hybridization can lead to a reduced uncertainty in the solution and/or a faster time to a solution. However, not all hybrid methods work for all types of radiation transport problems. In problems where the method is not well-suited for the problem physics, a hybrid method may perform more poorly than analog Monte Carlo, leading to wasted computer time and energy, or even no acceptable solution.

This project builds on existing software infrastructure (ORNL’s Denovo and ADVANTG) to generate hybrid methods for deep-penetration radiation transport problems. Specifically, we are developing variance reduction parameters for problems with strong angular anisotropy without explicitly including angular biasing parameters. No existing, highly accessible, automated hybrid method has incorporated angular-dependence of the flux in generating variance reduction parameters, which has led to difficulty in the analysis of highly anisotropic problems. Our method should improve the computational performance of hybrid methods for anisotropic problems while also maintaining similar space and processing metrics as energy- and space- exclusive hybrid methods.

Multi-physics modeling of fluoride-cooled high-temperature reactors (FHRs)

Xin Wang, Dan Shen, Katy Huff, Manuele Aufiero, Massimiliano Fratoni, April Novak

To improve understanding of coupled physics in FHRs, this work involves the development of tools and methods for coupling at thermal hydraulics and neutronics within the context of FHRs. Low-dimensional models relying on simplified neutron kinetics and heat transfer have been implemented in a python package, PyRK. Higher dimensional models that couple these physics in finite element frameworks (including both MOOSE and COMSOL) are also being developed. Finally, models which coupled monte carlo simulation with CFD tools are also being iterated upon.

WARP (“Weaving All the Random Particles”)

Ryan Bergmann (alumnus), Kelly Rowland, Rachel Slaybaugh, Jasmina Vujic

To improve reactor design and operation, fast and accurate neutron transport calculations are needed. Today’s supercomputers are comprised of heterogeneous architectures designed to reduce power consumption, and new algorithms are required to use these hardwares. WARP, which can stand for “Weaving All the Random Particles”, is a three-dimensional (3D), continuous energy, Monte Carlo neutron transport code developed to efficiently execute on a CPU/GPU platform. WARP is able to calculate multiplication factors, flux tallies, and fission source distributions for time-independent problems and can run in both criticality or fixed-source modes. WARP currently transports neutrons in unrestricted arrangements of spheres, cylinders, parallelpipeds, and hexagonal prisms and is able to entertain both vacuum and reflecting (specular) boundary conditions.

What sets WARP apart from previous, somewhat similar endeavors is its breadth of scope and novel adaptation of the event-based Monte Carlo algorithm. Previous codes have been limited to restricted nuclear data or simplified geometry models, where WARP instead loads standard data files and uses a flexible, scalable, optimized geometry representation. WARP uses a suite of highly-parallelized algorithms and employs a modified version of the original event-based algorithm that is better suited to GPU execution.