OpenMC Fusion Benchmarks is a platform for V&V of fusion neutronics. It focuses on fusion-relevant integral benchmarks. It relies on an automated workflow for model simulation, data postprocessing, visualization and analysis. It embeds a database of experimental and numerical results for quick comparisons to which users can contribute to. The database contribution workflow is fully automated.
pip install -e .
The src/ folder provides all the classes and functions of the package. It also has additional data in the data/ subfolder (e.g. IRDFF-2 cross sections) and some neutron source characteristics, importable as openmc.Source object, in the neutron_sources/ subfolder (so far there is only the Frascati Neutron Generator - FNG energy and angular characteristics).
The notebooks/ folder contains some useful tutorial notebooks.
The models/ folder has all the models and benchmarks covered by this repo (e.g. the fng-str and fns-duct benchmarks). Each bechmark provides an openmc_model.py file that has the Python API input file for the relative openmc model. It can be run via command line and often has some different options for running it. There should be also a postprocessing.ipynb jupyter notebook for quick result visualization and confrontation against other experimental or other codes' results. The run_and_store.py script is an automated way to run the openmc model, postprocess the results and store them in a hdf5 file in the results_database/ in just one click. It sometimes runs the model more than once in order match all the results provided by the experiments (e.g. the fng-str experiment has three different configurations and thus requires three different simulations).
Each model has a results_database/ subfolder that contains the experimental results as well as previous simulation results of the benchmark in form of hdf files. Tutorial notebooks in the notebooks/ folder show how to navigate those files. All the hdf files belonging to the same model have same structure (same tallies with the same name, postprocessed in the same way and extractable in the same way). Each results_database/ subfolder should have at least the experiment.h5 file. Results from openmc and other codes are more than welcome, especially if built consistently with the files already present. The result hdf5 files from codes should have the name of codename-codeversion_xsnameandversion.h5.
The contribution workflow is automated for the openmc code. Users can run any openmc model and then the postprocessing notebook for a direct validation of the openmc version they use but they can also contribute to the repository. To contribute it is sufficient to:
- Run the
run_and_store.pyscript of a benchmark of choice according to therun_and_store.ipynbtutorial notebook (this would generate a result hdf5 file in the respectiveresults_database/folder) - Review the result hdf5 and open a pull request to propose the new file to the database
