Exascale combustion CFD for industry relevant combustor geometries

Project Description

Combustors are a crucial part of many technologies including gas turbines, and Computational Fluid Dynamics (CFD) is a crucial modelling technique used to better understand these components. This PhD will explore the use of number representations, alternative to the dominant FP64, for this workload.

Primary Supervisor: Dr Paul Bartholomew

Project Overview

The aim of this PhD project is to push the boundaries of combustion CFD as applied to realistic combustor geometries. The code used for this research will likely be, though does not have to be limited to, ASiMoV-CCS (see first further information link below). Areas of research include the use and impact of mixed precision, in particular when using complex geometries, the implications of using lower-precision hardware, and different numerical methods. The focus will be on numerical accuracy, scalability and performance to support industry use cases.

Overview of the research area

The VECTA EPSRC Strategic Prosperity Partnership continues the work began in the ASiMoV project to develop the modelling capabilities to virtually design and certify the next generation of aeroengines. A significant challenge faced by these designs will be addressing their contribution to climate change. This will require engines that are more efficient than ever before, combined with alternative low-carbon and carbon-neutral fuels. Such fundamental changes will require unprecedented simulation capabilities to achieve high-resolution predictive capabilities for a new regime where decades of institutional knowledge cannot be relied upon. To do so, the entire simulation workflow, including the combustion modelling that is the focus of this project, will need to be Exascale-capable. For this project to succeed, the computational capability must be demonstrated in industry relevant geometries. The student will investigate and assess how to optimise an unstructured “CFD+combustion” code for Exascale computing, maintaining the required prediction accuracy.

Potential research questions

  • Can we use reduced precision computing in complex industrial cases and still maintain acceptable accuracy?
    • How do different modules/subroutines (e.g. flow solver, chemistry module) benefit (or not) from reduced precision?
    • Are different modules subroutines sensitive to reduced precision when considering accuracy of the result?
  • How can we exploit specialised hardware acceleration in an unstructured CFD+combustion problem?
    • For example, matrix accelerators (such as tensor cores) often operate on low-precision data only.
  • Are the standard numerical methods suitable for low-precision computation or do we need to adapt the algorithm to the numerical precision?

How much do we need to decouple the physics implementation from the parallelisation to be able to flexibly adapt the code for heterogeneous hardware?

Student Requirements

A UK Masters degree, 1st in undergraduate integrated Masters, or its international equivalent, in a relevant subject such as computer science and informatics, physics, mathematics, engineering.

The student must be a strong programmer in at least one of C, C++, or Fortran with experience of developing or contributing to scientific applications. The student must be familiar with mathematical concepts such as algebra, linear algebra, probability and statistics 

English Language requirements as set by University of Edinburgh

Important: Please note that there are restrictions on nationalities for the PhD studentship. If you are a national (or dual-national) of one of the countries on the US Commerce Control List (https://www.bis.gov/regulations/ear/746) you are not eligible to apply for this funding.

Recommended/Desirable Skills

Experience with numerical methods, scientific programming and HPC are highly desirable, as are an understanding of the fundamentals of CFD, combustion and/or numerical analysis.

How to apply

Applications should be made via the University application form, available via the degree finder. Please note the proposed supervisor and project title from this page and include this in your application. You may also find this page is an useful starting point for a research proposal and we would strongly recommend discussing this further with the potential supervisor.

Further Information

  • P.Bartholomew, A.Borissov, C.Goddard, S.Lakshminarasimha, S.Lemaire, J.Zarins & M.Weiland, “ASiMoV-CCS – A new solver for scalable and extensible CFD & combustion simulations”, PASC25 (2025), https://dl.acm.org/doi/10.1145/3732775.3733577
  • A.Haidar, H.Bayraktar, S.Tomov, J.Dongarra, N.J.Higham, “Mixed-precision iterative refinement using tensor cores on GPUs to accelerate solution of linear systems”, Proc. R. Soc. A (2020), https://doi.org/10.1098/rspa.2020.0110