Research Studentships

Numerical simulation plays a central role in modern engineering design, enabling the analysis of complex fluid-dynamic phenomena in aerospace, automotive, and energy systems. In industrial computational fluid dynamics (CFD), Reynolds-Averaged Navier–Stokes (RANS) and Unsteady RANS (URANS) approaches remain widely used because of their relatively low computational cost. However, these approaches rely heavily on turbulence models and often struggle to accurately predict complex unsteady flow phenomena. Large-Eddy Simulation (LES) represents a higher-fidelity alternative, resolving the large energy-containing turbulent structures while modelling only the smaller subgrid-scale motions. As a result, LES provides significantly improved predictive capability for many turbulent flows. However, the computational cost of LES remains extremely high. High-Reynolds-number industrial simulations may require billions of grid points and millions of timesteps, meaning that practical simulations can still require weeks of runtime even on modern high-performance computing systems.

Hybrid URANS-LES approaches seek to blend the efficiency of URANS with the accuracy of LES, making them eminently suitable for simulating complex turbulent flows in practical engineering applications. A common approach used for industrial, high Re flows is Detached Eddy Simulation (DES). This utilises URANS for the near wall region, and LES for separated, free shear regions. A further development is Delayed Detached Eddy Simulation (DDES), which determines a switching criterion based upon flow features (rather than grid spacing) to infer whether to use a URANS or LES resolution approach.

A potential avenue for work in this field would be to develop an AI informed selection criterion, to allow for faster convergence of these flows during computation and also optimisation of the computational workflows to allow rapid multi-fidelity LES / URANs simulations. A key research direction will be the development of machine-learning approaches for improved selection of these sub-models, choosing the best combination of complex turbulent through to RANs models, as well as the LES AI models to drive this. This may include choice of advanced meshing schemes, interrogation of flow behaviour time histories and code compilation enhancement to take advantage of the latest supercomputer chipsets.

Since LES explicitly resolves only the largest turbulent structures, the accuracy of the simulation depends strongly on the models used to represent the unresolved scales. At a detailed level the project will investigate how data-driven models trained from high-fidelity simulation data can learn improved subgrid-scale representations while maintaining physical consistency with the governing equations.

The developed approaches will be evaluated using representative turbulent flow configurations relevant to industrial applications, including external aerodynamic flows and internal turbomachinery flows. These test cases will allow the investigation of how AI-enhanced methods influence the accuracy, efficiency, and robustness of LES simulations compared with conventional approaches.

The expected outcome of this research is the development of new hybrid AI–LES methodologies capable of improving the efficiency and predictive capability of high-fidelity turbulence simulations. Such capabilities align directly with VECTA’s ambition to achieve transformational performance improvements—potentially exceeding 100× speedups—that are unlikely to be realised through conventional hardware or algorithmic advances alone. Beyond aerospace applications, the methods developed in this research will be broadly applicable to other scientific domains involving complex multiscale fluid flows.

Applications are invited for a University of Warwick PhD Studentship in The Department of Computer Science in collaboration with the Department of Psychology. The PhD will start October 2026 on Machine Learning and Psychophysiological Deception Detection. The studentship is part sponsored by GCHQ and funded for up to 3.5 years with fees and a stipend at the standard UKRI rate. The position is only open to Home fee status students.

PhD positions are available at the Theory and Foundations group in the Department of Computer Science, University of Warwick, UK. The group works on various aspects of theoretical computer science including:

* algorithmic game theory

* approximation algorithms

* automata and formal languages

* combinatorics and graph algorithms

* computational complexity

* logic and games

* online and dynamic algorithms

* parallel algorithms and distributed computing

* parameterized complexity and structural graph theory

* random structures and randomized algorithms

* sublinear and streaming algorithms

* theoretical foundations of machine learning