This opportunity is open to students willing to undertake an MSc by Research with the potential to advance to a PhD
1. Objectives:
The primary goal of this PhD project is to develop novel statistical and machine learning methods that combine high-fidelity and low-fidelity models of the nuclear fusion process. The aim is to optimize computational efficiency without sacrificing physical accuracy, thereby accelerating progress in fusion research.
In particular, we will use a multifidelity approach to target the specific problem of nonlocal thermal transport. Nonlocal electron thermal transport is critically important in determining energy transport from where it is absorbed to where the fusion fuel capsule is ablated in laser direct-drive inertial confinement fusion (ICF). It is also critically important in determining the heat load on the wall of the device in magnetic confinement fusion (MCF) devices. These methods will be critical for uncertainty quantification and design optimization, both of which are essential for advancing fusion technologies.
2. Background:
In the quest to make nuclear fusion a commercial reality, computational simulations allow researchers to try out new ideas without having to build a working model. While computational models have historically been used to support the fusion research, codes used to simulate the targets are complex, and design studies often incur substantial computational expense.
A promising solution lies in the development of surrogate models, which are machine learning models trained on simulation data to provide rapid, approximate solutions. However, surrogate models introduce their own challenges, such as inherent uncertainties, approximation errors, and insufficient uncertainty quantification. This calls for a more robust approach to integrating low-fidelity and high-fidelity models in a way that balances accuracy and efficiency.
3. Project description:
This project will explore multifidelity modelling, a technique that integrates data from both high-fidelity and low-fidelity models to create a single, cohesive framework. In this context, Gaussian Processes (GPs) are a promising approach, providing a flexible, non-parametric method to model uncertainty in fusion simulations. However, many GP-based methods assume a linear relationship between fidelity levels, which can be limiting in the nonlinear, multiscale, and multiphysics environment of nuclear fusion. To address this, the project will investigate the use of deep Gaussian Processes, which can model more complex, nonlinear relationships between fidelity levels by leveraging hierarchical, multi-layered structures.
Key aspects of the project include:
Bayesian Inference Approach: Casting the modelling as a Bayesian inference problem, enabling the integration of various sources of uncertainty in a coherent framework.
Gaussian Processes and Deep GPs: Using both traditional and deep GPs to capture complex, nonlinear relationships between different fidelity levels.
Kinetic and fluid modelling of plasmas. In particular developing surrogate models based on our novel, high-fidelity kinetic codes and applying these to large scale design calculations using fluid codes.
4. Expected Impact:
This project will significantly contribute to the development of computational tools for nuclear fusion, a field that holds the key to the future of sustainable energy. By improving the efficiency and accuracy of fusion simulations, the outcomes of this research could accelerate the path to commercially viable fusion power, offering a transformative solution to the world’s energy needs.
Students involved in this project will gain experience in cutting-edge statistical and machine learning techniques, with applications that extend beyond fusion energy to other fields of engineering, physics, and computational modelling.
The project will be mainly based in York, but there are opportunities to work with industry, travel to conferences and collaborations with other research groups.
How to apply
This opportunity is for students willing to undertake an MSc by Research with the potential to advance to a PhD.
Applicants should apply via the University’s online application system at https://www.york.ac.uk/study/postgraduate-research/apply/ and submit an MSc by Research application. Please read the application guidance first so that you understand the various steps in the application process.
This project is open-ended making it suitable for MSc by Research and PhD level
This project is offered by University of York. For further information please contact: Dr Jiannan Yang (jiannan.yang@york.ac.uk).
