Cross-cutting Simulations with Uncertainty for Fusion Applications

Fusion energy sources will test our understanding of materials and how they behave in the presence of high-temperature plasmas, radiation and fusion reaction byproducts. For sustained fusion to be feasible for energy production, fusion reactors will need to breed tritium for future use as fuel. Researchers need to know how tritium moves through materials and where it collects. Tritium migration is currently modeled mathematically, but there are few comparisons of these models to real world experimental results. We will demonstrate the fidelity of a tritium migration model both by comparing its predictions with the few real life sources of data, and also by using machine learning methods to model the uncertainty inherent in mathematical model predictions of possible tritium concentration values within materials. We will produce new validation examples of tritium migration models using data that has already been published in academic journals. We will also develop a machine learning algorithm to estimate uncertainties on tritium concentration values within the geometry of fusion reactor components. Finally, we will build an easy-to-use browser-based interface for setting up these calculations, in order to better serve fusion researchers. Fusion research and development companies are moving fast to create working reactors, and do not have time or manpower to dedicate to understanding nuanced simulation codes at the software developer level. We will sell these software tools to fusion developers that will save them time and money in their race to generate net-positive energy from fusion. This tool will help them to track down critical tritium within their system for a more efficient fuel cycle.