My research explores the use of machine learning and statistical methods for tackling aerospace engineering problems, with data primarily coming from computational simulations. To encourage industrial uptake, much of my work is focused on developing approaches with uncertainty quantification and interpretability in mind. I am funded by the EPSRC and the Lloyd’s Register Foundation, with previous projects including:
Uncertainty quantification for data-driven turbulence modelling, using Mondrian forests.
Rapid flowfield estimation via embedded ridge functions and deep neural networks.
Using dimension reduction techniques to tackle the curse of dimensionality in aerospace design tasks.
In 2017 I completed a PhD at the University of Cambridge, supervised by Prof. Paul Tucker. This involved working with Rolls-Royce to develop high fidelity turbulence resolving capability in their in-house CFD code. The resulting code was used to run Large Eddy Simulations of the complex three-dimensional and transitional flows present in gas-turbine compressors. Physical insights gained from these simulations helped inform compressor-specific turbulence modelling strategies.
I am also a member of the Effective Quadratures organisation, founded by Pranay Seshadri. Effective Quadratures provides a unique platform to build and deploy digital twins, fusing in sensor data with physics-based knowledge in a holistic and principled way. The foundation for this is our open-source python library equadratures, which provides tools for uncertainty quantification, machine learning, optimisation, numerical integration and dimension reduction – all using orthogonal polynomials.