My group and I model flows, investigate their physics, and optimize their behaviour.
We usually start from physics-based models of a flow inside or around a device. We infer the models' parameters from experimental data using Bayesian Inference accelerated with adjoint methods. This turns qualitatively-accurate models into quantitatively-accurate models and ranks different models by their evidence (their marginal likelihood), given the data.
We then sometimes set targets and constraints and, with adjoint methods, determine how the targets are affected by all the model parameters. We then use gradient-based algorithms to optimize the flow through or around the device.
If we do not know the physics, or cannot model it, then we use data-driven approaches.
Click on the links above to find out more.