Title:	Performance tests of tools for new nuclear fuels and reactor cores
DNr:	NAISS 2025/22-1015
Project Type:	NAISS Small Compute
Principal Investigator:	Erik Bergbäck Knudsen <erik.bergback.knudsen@gu.se>
Affiliation:	Göteborgs universitet
Duration:	2025-09-20 – 2026-04-01
Classification:	10301
Keywords:

Abstract

The latest advancement of nuclear reactors and fuels form new boundary conditions for tools required to assess performance, radiation dose, shielding requirements etc. in the various proposed geometries. This project aims at in its 1st phase on model systems: 1. Evaluating our new tools for Monte Carlo neutronics and dose calculations on cluster hardware available 2. Find bottlenecks where new development is necessary. 3. Evaluate various variance reduction techniques available to us at the time of writing. It is expected that the main part of computations will be allocated for Monte Carlo neutronics, generally well suited for parallelism, yet often requires tweaks tailored to the platform to run optimally. Such things we hope to evaluate during this project. Typically, an initial core Monte Carlo run for a reactor project takes about 8 core-hours, subsequent runs which are the core of this application aimed at actual dose computations in a surrounding reactor hall (depending on spatial resolution) typically take ~200 core-hours. To accommodate benchmarks at some scale we would need to run 2-5 such computations / month, leading to the ask for 5 x 1000 per system, to allow for some headroom. In a 2nd phase - we will attempt to fix (some of) the bottlenecks identified, and in a later proposal ask for resources for high fidelity runs to expand model systems to fulls scale reactor plant simulations. In the context of nuclear reactor performance Monte Carlo techniques are often coupled with CFD and differential equation solvers - all tend to be computationally heavy. Hence the dual application to different hardware platforms to evaluate performance differences.