Data-driven Turbulence Modelling
Mondrian forests, trained on high fidelity CFD data, are used to augment turbulence models. Their uncertainty estimates provide crucial information on the suitability of the model, and the data used to train it.
Tackling the Curse of Dimensionality
Dimension reducing subspaces are shown to be a powerful tool for exploring high dimensional design spaces, providing important physical insights and allowing for in-depth analysis of uncertainties.
Machine Learning with Polynomials
Polynomial regression trees combine decision tree learning with polynomial regression, resulting in accurate yet interpretable models. The resulting models can be used for a wide range of tasks, from traditional supervised learning to uncertainty quantification.
Rapid Flowfield Predictions
Dimension reducing embedded ridge functions offer rapid flowfield predictions, with comparable accuracy to state-of-the-art deep learning methods, whilst being more interpretible and posessing baked-in uncertainty quantification.
Large Eddy Simulation for Turbomachinery
My PhD involved working with Rolls-Royce; examining the feasability of modifying an industrial CFD code to run high fidelity Large Eddy Simulations of the transitional flows in aero-engines.