EPSRC NFFDy Summer Programme 2024
In July and August, LIFD was the host for the EPSRC 2024 NFFDy Summer Programme, which is a training component of the National Fellows in Fluid Dynamics postdoctoral fellow scheme. The programme at Leeds explored applications of data driven fluid dynamics and included lectures on methodologies and examples of applications, including industry challenges and policy. We were delighted to welcome many guest speakers (e.g. Taraneh Sayadi, Sorbonne University and Jean-Christophe Loiseau, Ecole Nationale Supérieure d’Arts et Métiers) who are leaders in their field, and delivered some fantastic talks to the participating fellows. Whilst at Leeds key LIFD fluids research groups presented an overview of their research areas and the fellows also toured some of the fluids labs at Leeds. A key feature of the programme were the group projects undertaken by fellows, the group project topics were:
- Continuum modelling of droplet impact on granular materials using experimental data
- Mixing Processes in Geophysical Flows
- Simulation and data-driven improvement of electrolysis cells for green hydrogen
- Data driven modelling of lava flow crust
- Nonlinear Internal Waves in Geophysical Flows
- Reduced order modelling of impinging jets for corneal material characterisation
- Learning RANS turbulence model parameters from magnetic resonance velocimetry data
Two participants of the programme have written a piece about their experience Dr Muting Hao and Dr Osama Maklad.
NFFDy Summer Programme Reflection
Written Dr Osama Maklad, University of Greenwich
Lecturer in Mechanical Engineering
I really enjoyed working with different mindsets in the summer programme and researchers from around the UK and also abroad. I enjoyed the lectures given at the start by well known researchers in the field of using data in fluid dynamics. The hands-on sessions were useful in practicing some of the codes. I was part of a project to produce a reduced order model of the human eye under the air puff test for corneal material characterisation. We used machine learning and neural networks to analyse the corneal deformation parameters and air pressure distribution to produce a model that takes a fraction of the time to produce the same results. We got very good primary results that we will build on to potentially apply for a research grant or publish a journal article.
My Experience in the EPSRC National Fellowship Programme in Fluid Dynamics
Written by Dr. Muting Hao, University of Oxford
Career Development Research Fellow at St John’s College
Associate member of faculty in Engineering Science Department
This summer, I had the amazing opportunity to join the EPSRC National Fellowship in Fluid Dynamics (NFFDy) summer programme, hosted by the University of Leeds. Organized by Prof. Steven and Dr. Claire on behalf of EPSRC, the programme offered intense lectures and training in fluid dynamics and machine learning, chances to apply these concepts into our proposed projects and a supportive community of fellow mathematicians. This experience not only advanced my technical skills but also enriched my personal and professional life.
A World-Class Training Experience
A highlight of the programme was a series of specialized courses led by leading experts in machine learning and fluid dynamics from institutions like Imperial College London, the University of Cambridge, Defence Science and Technology Laboratory, University of Chicago, and RWTH Aachen University. The training covered advanced topics such as the Adjoint Method, Spectral Method, Bayesian Methods, Cindy, and other machine learning techniques, each offering valuable potentials for tackling complex problems in fluid dynamics. These courses offered daily engagement with experts, fostering discussions on new ideas and advancements.
Research Talks, Lab Visits, and Collaborative Learning
The programme also featured frequent research talks by Leeds faculty, covering the broad applications of fluid dynamics and machine learning in areas like oceanography, weather forecasting, flood prediction, blood flow simulation, and thermofluid dynamics. These talks expanded our und
erstanding of how machine learning is revolutionizing diverse fields, and we were encouraged to engage, ask questions, and discuss ideas with the speakers. We also visited the fluid dynamics labs at Leeds, where we saw facilities for testing water and air flow and learned about high speed cameras and sensors.
Hands-On Projects in Machine Learning for Fluid Dynamics
I had the chance to work on two other projects together with other fellows.
Project 1: Mixing Process in Stratified Flow
Together with Dr. Edward and Dr. Xiaodong, we investigated the mixing process in stratified flows, analyzing patterns such as Holmboe waves, turbulent flow, and laminar flow in different aspects. We used a Classifier Neural Network to differentiated flow schemes and other Neural Networks to explore Reynolds stress scaling across stratified layers. Physics-Informed Neural Networks (PINN) was tried to predict the flow with integrating physical laws. SPOD helped us analyse frequency components in the mixing process.
Project 2: Reduced-Order Modelling of Air Puff Dynamics on the Cornea
This study focuses on improving intraocular pressure (IOP) measurement accuracy in patients with keratoconus, a degenerative eye disease. The challenge is that the IOP alters corneal thickness, shape, and biomechanical properties, leading to misleadingly low IOP readings. Traditional fluid-structure interaction (FSI) simulations of the air puff-cornea interaction are precise but computationally intensive. I worked with Dr. Osama to develop a machine learning model that efficiently captures the correlation between air puff pressure and corneal properties, aiming for a faster, patient-specific IOP assessment.
These projects were a fantastic opportunity to apply diverse machine learning techniques to a real-world fluid dynamics challenge and work collaboratively with a team of talented peers.
Building Community and Camaraderie
Living together at Leeds fostered a vibrant community among other national fellows. Prof. Steven and Dr. Claire made the experience memorable with both academic and social events. A highlight was our trip to Maths City, where we bonded as a group while enjoying interactive math-based games that blended fun with learning.
We also have self-organised events within the national fellows in fluid dynamics and other participants in this programme. Participating hiking, cooking, or sharing meals together, I valued our time together discussing research topics and career development.
Reflections
I am sincerely grateful for this EPSRC NFFDy summer experience. My thanks go to EPSRC, the University of Leeds, Prof. Steven, Dr. Claire, Kate, all the speakers and everyone involved for making this summer unforgettable. I look forward to applying the insights gained here to my own fellowship research on fluid dynamics in gas turbines and hope to contribute to this vibrant community in the future.