This project will develop uncertainty quantification methods for nuclear fusion, motivated by yield prediction in tritium fuel cycles. The lack of scalable tools necessitates large engineering tolerances, increasing reactor cost. Empirical tests are expensive, while simulations are subject to error due to uncertainty in nuclear data and unresolved physical processes e.g. thermal expansion and fine-scale inhomogeneities. Generating independent simulation replicates accounting for all relevant uncertainties is currently infeasible.
This project will deliver a rigorous understanding of industry-standard simulation and variance reduction methods, typically variants of importance sampling. We will connect these methods to modern formulations of Monte Carlo algorithms to improve their accuracy, scalability, and overall computational cost. The methodology so delivered is fundamental for the sustainable delivery of fusion power to the electricity grid.
This project is based at Newcastle University, but will be co-supervised by the UK Atomic Energy Authority. The successful applicant will also be part of the MathRad research network on mathematics of radiation transport, which has groups at the Universities of Warwick, Bath, and Cambridge.