Peter Vincent is a Senior Lecturer in the Department of Aeronautics at Imperial College London where he researches in the field of computational fluid dynamics, and teaches mathematics and numerical analysis. His lab can be found online at The Vincent Lab.
What do you see are the biggest challenges facing CFD in the next 3 years?
The nature of how HPC hardware is evolving (FLOPs >> memory bandwidth, and FLOPs massively parallelised) will make it hard to effectively leverage all the available FLOPs using ‘traditional’ CFD algorithms.
Dealing with large amounts of data generated by unsteady CFD simulations will be a challenge. There is a need to develop ‘in-situ’ visulisation/processing techniques, so that data can be visualized/processed in RAM without expensive IO to disk. The Catalyst extension to Paraview is an exciting effort in this direction. Frank Ham and the guys at Cascade Technologies also seem to be doing some interesting work in this area at the moment.
John: I find it interesting that the two challenges you cite above are also mentioned in NASA’s CFD Vision 2030 Study. Are you familiar with the study and if so what do you think of it overall?
Peter: Yes, I am familiar with it, and I am a fan! I think a lot of people are aware of the report now, and it is helping to focus thoughts, and motivate discussion.
John: I’m also glad you didn’t cite meshing as a challenge. Phew!
Peter: Yes – meshing is still a big challenge as well! One other topic I’ve recently got interested in is parallel-in-time. It’s worth checking out XBraid from LLNL.
John: I haven’t heard of XBraid. There’s always more to learn.
What are you currently working on?
We are trying to address the challenges detailed above. Specifically, we are developing/implementing high-order accurate numerical methods for next-generation hardware like NVIDIA GPUs etc. Our efforts are embodied in PyFR, an open-source implementation of high-order accurate Flux Reconstruction methods for a range of hardware platforms (including heterogeneous systems). We have begun to apply PyFR to a range of unsteady flow problems – and so far the results seem really promising!
Flow over a NACA 0021 airfoil in deep stall obtained using PyFR.
My group is also undertaking biological flow simulations using STAR-CCM+ from CD-adapco. In particular we are working to design more effective arterio-venous fistulae, which are used as access points for patients with kidney failure who need dialysis.
John: Ignoring the hardware, how does PyFR compare to “traditional” CFD solvers? Can you make any general conclusions regarding convergence rates, accuracy, or how certain physical models have to be implemented?
Peter: This is a good question. In fact, really this is the question. We are undertaking comparisons with “traditional” solvers at the moment – these should be published in the coming months, so watch this space!
John: I already asked this via Twitter but I’ll ask it again here. I pronounce PyFR as “pie-fire” and everyone at Pointwise shakes their head and says it’s “pie-F-R.” How do you say it?
Peter: The latter! But as I said on Twitter “pie-fire” sounds kind of cool! The FR stands for Flux Reconstruction.
How did you get to be where you are today?
I did my undergraduate degree in Physics at Imperial College, a Ph.D. in Aeronautics at Imperial College, and then a Postdoc in Aeronautics and Astronautics at Stanford University.
John: When did you first know that you wanted to do CFD? Was it the physics side or did you also have a programming interest?
Peter: I first got interested in fluid dynamics when I started my undergraduate degree – but we weren’t taught so much of it on my Physics course, so I did summer research placements in an experimental plasma group at Imperial College, and then a CFD group at Cambridge University – I guess I never really looked back after that. For me it was a great combination of maths, programming, and physics; and it has such wide applicability (from aircraft design to blood flow).
Who or what inspired you to get started in your career?
My Ph.D. supervisors Spencer Sherwin and Peter Weinberg played a major role. And working with Antony Jameson as a Postdoc at Stanford was an amazing experience – he works/thinks very differently to most other people. I learnt that just because everyone is doing something one way does not mean it is the correct way – don’t be scared to be different! Colin Caro at Imperial College got me involved with biological flow simulations, and these are now an important part of my research activities. When everything is going wrong I remember a conversation I had one evening at Hampton Inn Poway with Siegfried Zerweckh from General Atomics about the commitment/perseverance/vision required to make exciting things happen.
Flow in the wake of a SMC006 serrated nozzle obtained using PyFR.
John: Ain’t that the truth: perseverance and hard work. Do you ever have trouble selling students on those concepts?
Peter: I think a Ph.D. really helps to “teach” this – amongst a whole bunch of other skills/qualities.
What advice do you have for young people entering the field today?
Learn to code well! Ideally try to become a “software ninja”. You will be empowered, and able to articulate your ideas efficiently and effectively. Anyone reading this who is a “software ninja” and thinks they might be interested in CFD, should contact me – I’m hiring!
How do you know Pointwise?
We worked together to add a PyFR exporter to Pointwise.
Can you share with us your favorite tools and resources that help you get your job done?
- Software:
- Conferences:
- Websites:
If we were to come visit you where’s a good place to go out for dinner?
I like Suskan for Thai food, and Hare and Tortoise for sushi. If you are paying, then we can go to Restaurant Gordon Ramsey – I have never been, but I think it has three Michelin Stars!
John: Because I don’t think Gordon Ramsay needs our money, I’m keen on trying sushi from a restaurant named for a bunny and turtle. And the next time you’re in Fort Worth, I’ve got a Thai restaurant for you to try.
Thanks for taking the time for this interview.
RSS - Posts