Speaker
Description
The development of an automated, parametric workflow to generate high-quality neutronics meshes from complex source geometries remains a key practical challenge in fusion neutronics. These workflows are particularly challenging given the need for element quality, metadata tracking, and mesh conformity throughout a multi-stage pipeline.
Stellarmesh is an open-source library developed at Thea Energy to address these challenges. It provides a unified API for the creation of MOAB, libMesh, and DAGMC meshes from source geometry in CadQuery, build123d, STEP, and BREP formats. In addition, Stellarmesh supports imprinting and merging of conformal geometry, native surface and volume meshing using Gmsh and OpenCASCADE, and the programmatic manipulation of .h5m files.
A key feature of Stellarmesh is its extensible backend architecture. While it ships by default with a fully open-source toolchain to mesh OpenCASCADE geometry, its unified backend API enables users to additionally integrate custom geometry and meshing backends. Given an API-compatible mesh file, it offers the same visualization, manipulation, and export functionality as the built-in backends.
This functionality has enabled Thea Energy to develop a separate backend to the Stellarmesh workflow that supports complex engineering models. These geometries, which are modeled by the engineering team in Siemens NX, are automatically exported to Parasolid files, meshed using the Simmetrix Simulation Modeling Suite, and imported into Stellarmesh using the API.
In this presentation, two workflows are showcased that demonstrate Thea Energy’s end-to-end process to create OpenMC models from CAD geometry. The first shows a sample blanket study using the open-source toolchain to illustrate how analysts can drive geometry generation, meshing, and neutron transport from a single Python file or environment. The second shows an analysis of a high-fidelity NX model, demonstrating how component metadata and geometric fidelity are preserved from CAD to OpenMC.
Stellarmesh is MIT licensed and available on both PyPI and conda-forge. This work shows that a well-designed and open-source meshing layer can substantially reduce the complexity of traditional CAD to neutronics pipelines and enable fully parametric, automated analysis.
| Formatted abstract uploaded? | Done. |
|---|