|Funding for tuition fees and a living stipend are available on a competitive basis. Funding will be awarded on a rolling basis, so apply early for the best opportunity to be considered.
|1st December 2023
|31st August 2024
Project title: Using physics-informed neural networks and data assimilation to predict and improve urban air quality
Supervisory Team: Sean Symon and Christina Vanderwel
The aim of this project is to use physics-informed neural networks (PINN’s) and data assimilation techniques to convert limited measurements of urban fluid dynamics into highly accurate models of pollutant dispersion in urban settings. These models are crucial for environmental management, public health and urban planning so that authorities can assess and mitigate the impact of pollutants on air quality and the well-being of city residents.
The project will initially tackle low Reynolds number problems around simplified models of urban environments which include buildings, streets or uneven terrain. The PINN’s will be trained on high-fidelity simulation data and will identify unknown closure terms in the governing equations, most notably the scalar flux which is typically modelled in simulations. Furthermore, the optimal training points and network hyperparameters will be identified.
The second phase of the project will consider experimental velocity and concentration fields at higher Reynolds numbers that are obtained using particle image velocimetry and planar laser-induce fluorescence. The PINN will use either existing experimental data or new measurements from the water tunnel depending on the problem of interest. Since the passive scalar can be measured at very high resolutions compared to the velocity, the PINN can predict the velocity field more accurately than possible without passive scalar measurements. This offers an exciting prospect for being able to improve state-of-the-art experimental techniques without the need for more hardware.
The final phase of the project will aim to predict unsteady flow structures and their frequencies using the reconstructed flow fields from PINN’s. The latter are inputs to the linearised equations of motion which reveal the most amplified structures in the flow.
We aim to build a diverse and inclusive team to tackle challenging problems where we develop new skills and expertise in our team members. You will have a unique opportunity to work alongside other team members (PhD students and postdoctoral researchers) with different backgrounds and experience. You will the unique opportunity to be trained in using state-of-the-art machine learning algorithms and advanced data-analysis methods that will enable you to pursue a career in academia or industry. Finally, you will be able to travel to international conferences to present your work and develop new collaborations with research groups around the world.
Further information on the type of projects carried out in our lab as well as information on current lab members can be found on our website (https://sites.google.com/view/seansymon/home).
A very good undergraduate degree (at least a UK 2:1 honours degree, or its international equivalent) in Engineering with the motivation to do experimental work
Closing date: 31 August 2024.
Funding: Funding for tuition fees and a living stipend are available on a competitive basis. Funding will be awarded on a rolling basis, so apply early for the best opportunity to be considered.
How To Apply
Apply online: Search for a Postgraduate Programme of Study (soton.ac.uk). Select programme type (Research), 2024/25, Faculty of Engineering and Physical Sciences, next page select “PhD Engineering & Environment (Full time)”. In Section 2 of the application form you should insert the name of the supervisor Bharathram Ganapathisubramani
Applications should include:
Two reference letters
Degree Transcripts/Certificates to date
Type / Role: