In this example, we look at the ability of twinLab to model the desorption (the physical process where a previously adsorbed substance is released from a surface) of tritium (a radioactive isotope of hydrogen) in the wall of a nuclear fusion reactor. Fusion generates almost no radioactive waste, and the little waste that it does produce has a short half-life. However, the interior of the reactor wall is bombarded by a high-neutron flux during fusion, far higher than any naturally-occurring radioactive process, and therefore the properties of materials under high neutron bombardment are unknown. Computer simulations are required to understand the material properties in such extreme circumstances, but simulations are expensive in terms of computational power, and cannot be run at every point in parameter space under consideration. twinLab can be used to train simulation surrogate models using data from a sparse array of simulations. This allows for meaningful interpolation and extrapolation to unexplored regions of parameter space, together with a calibrated uncertainty estimate on the accuracy of the simulation surrogate.
Clone the repository and change directory to the project root:
git clone https://github.com/digiLab-ai/FusionEnergy.git
cd FusionEnergyInstall the dependencies:
poetry installCopy the .env.example file to .env
cp .env.example .envand fill out your twinLab login details in .env
Run the demo notebook:
poetry run jupyter notebook demo.ipynb