Published: Mar 25, 2025 by Liam Pattinson
Project info
FusionDLS is a Python project that solves for the position of plasma detachment fronts in tokamak divertors. By solving a reduced model, FusionDLS is capable of quickly producing parameter scans over various divertor designs and magnetic geometries to optimise the design of divertors for next-generation devices.
When we began work on FusionDLS, the project was comprised of a collection of Python modules and Jupyter notebooks in the top level of its GitHub repository. As our first step, we reorgansed the project into an installable package, as this would serve as a basis for all of our future upgrades. We then set up formatting and linting with ruff, converted some of the existing Jupyter notebooks into integrated tests, and created some HTML docs for the project using Sphinx. All of this was automated using GitHub actions.
The Jupyter notebooks were reworked to serve as examples for users of the project, and we were able to integrate them into the online docs using the [myst-nb][mystnb] Sphinx plugin. To ensure the notebooks were kept up-to-date with the rest of the project, we used the nbmake pytest plugin to run the notebooks in our testing CI jobs.
We also make some substantial upgrades and feature additions to FusionDLS. Previously, much of the data within FusionDLS was passed around using plain dictionaries, and by converting these to dataclasses we were able to provide stronger guarantees about data types and availability at each stage of the simulation. The conversion to an object-oriented design also allowed us to define a number of new utility methods and properties to clean up repetitive code.
Finally, we added the ability to trace scrape-off layer field lines from G-EQDSK files – a commonly used format for tokamak equilibria. This was made much easier using a combination of FreeQDSK to read the files and FreeGS to trace the field lines.