Sean Hanrahan
- shanrahan@student.unimelb.edu.au
- Desk 4.054, Level 4, Melbourne Connect
Thesis Title
Improving Reynolds-averaged Navier-Stokes Turbulence Models with Physics-Informed Neural Networks
Research overview
As Physics-Informed Neural Networks (PINNs) only require sparse data to model inverse problems, this technology can model complex flows from experimental data and known physics. The intent of my project is to apply PINNs with embedded Reynolds-averaged Navier-Stokes equations to experimental flows that are challenging to simulate with existing turbulence models. PINNs do not require an explicit turbulence model, rather, closure is inferred during training. It is anticipated that trained PINN models can be used with turbulence modelling tools such as Gene Expression Programming to develop novel turbulence models from experimental data, without the need for high fidelity numerical data.
Supervisors
Qualifications
M.Eng. Mechanical Engineering, The University of Melbourne, Australia (2020)
B.Sc. Mechanical Systems, The University of Melbourne, Australia (2017)