FLO Cycling – CFD Post Processing, Beyond the Numbers: A Matthew N. Godo Interview

It’s not often you get the opportunity to sit down with a true expert in a field, someone who is pioneering a movement in an industry.  We’re very grateful to have been given one of these opportunities.  Anyone who has spent time looking into cycling wheel aerodynamics has most likely come across the brilliant paper “An Aerodynamic Study of Bicycle Wheel Performance using CFD” by Matthew N. Godo.   When using CFD to design our wheels at FLO Cycling, Matts’s paper was a go to source of information.  CFD greatly impacted the final design of our wheels, ultimately delivering a faster more affordable product for our customers.  Matt is clearly one of the forward thinkers of our generation and his work in the cycling industry is shaking things up in a big way.  We had a chance to sit down with this expert to get his insight on cycling wheels aerodynamics and how Intelligent Light’s FieldView 13 software is moving the dial.
In this interview Matt discusses the differences between CFD and post processing software, the findings from his research for the AIIA technical paper, how software is affecting the cycling industry and a little about his own triathlon achievements.  
Here is the bio of today’s expert


Matthew N. Godo Ph.D. is the FieldView Product Manager at Intelligent Light, and has held this position for nearly 14 years.  He obtained his Ph.D. in biomedical engineering from the University of Toronto in 1992.  A 20 year practitioner of CFD, Matt has run numerical simulations on diverse applications such as the air flow patterns in a rat nose, to more recently, the aerodynamics of racing bicycle wheels and frames.  He and his wife are avid triathletes, and they try to find training time while parenting their five children.
FLO:  What exactly is FieldView software, and how does it differ from software like CD-Adapco’s Star CCM+?
Matt:  Computational Fluid Dynamics, or CFD, generally follows three steps. The first is to ‘mesh’ a model domain, breaking it into many small volumes called elements or cells. In the second step, a CFD solver, like CD-adapco’s STARCCM+ or Altair’s AcuSolve, is run, solving the basic equations of fluid motion on each element in the mesh generated from the first step. The third step is to then postprocess the output from the solver run, creating images, animations and calculating things like drag force and power requirements – this step is where we come in turning the solver data into ‘action-able’ information from which decisions can be made.  We specialize in this part of the CFD workflow, giving our users the fastest path to the engineering knowledge that they need.
FLO:  How did Intelligent Light get involved in the ground breaking research for your 2009 AIAA technical paper?
Matt:  One area of interest for us is to develop workflow strategies for CFD practitioners – it can be rocket science after all, and making CFD work productively for our customers is a key commitment for us. To understand the problems our customers face, we feel that the best way to learn is by doing.  I picked the study of bicycle wheel aerodynamics because we felt it could teach us the workflow lessons that we needed to understand.  And, as a triathlete, I have to concede that I’ve spend many hours during training rides contemplating the air flow around my front wheel. After a bit of research and some early CFD success comparing our numerical results to wind tunnel tests, I knew we were on to something solid.  With our background in postprocessing, I also realized that we could present these aerodynamic results in a way no one else had presented them before. That got us started, and the work was really well received by a lot of our customers (a surprising number of whom are also cyclists). Based on the feedback and encouragement from our customers, and many others in the cycling industry, we’ve not only been able to continue the work, but to significantly elevate the technical quality and scope of the original study.
FLO:  Can you give our followers a synopsis of your findings published in the AIAA technical paper?
Matt:  In our first paper in ’09, some really good trend agreement with wind tunnel data validated our CFD modeling methodology. We were pushing the CFD solver codes to do something that hadn’t been done before, and I was a little anxious that things wouldn’t work as well as they did. By doing a component breakdown, we were able to show that the main contributor to drag and side forces was coming from the tire and rim; the spokes and hub were small players. We were also able to identify some strong periodic flow structures coming off of the leading edge of the wheel, both from the tire and from the inner edge of the wheel rim. By plotting forces around the circumference of the wheel, we were able to see significant differences. When we extended the work in ’10 to look at several different wheels, we noted that the circumferential forces and periodic behavior were strongly dependent on the depth and shape of the wheel rim. We saw big differences in turning moments, comparing one wheel to another and that started us thinking about how to actually control stability and handling, as well as reducing drag, by coming up with better cross sections. Encouraged by some of our aerospace customers, we then looked at power requirements, not just drag forces, to help put the issue of the ‘negative drag claim’ into perspective. And, by including the front fork and downtube, we were able to see how these components could actually influence the forces and turning moments on the wheel. A great front wheel will only offer a benefit if it also works with the rest of your equipment. We just presented our latest work, examining the Scott Plasma III frame, fitted with Zipp 808 Firecrest wheels at the CD-adapco STAR GLOBAL conference in Amsterdam a few weeks ago.  In short, component interactions really do matter.
FLO:  What advances have high powered computers and software provided for the cycling/triathlon world, that would not have been possible prior to their existence?
Matt:  Wind tunnels do a great job of measuring overall performance and they can tell you pretty quickly whether one design is better than another. Where wind tunnels fall short is in helping designers to understand the whys and hows behind the performance differences. For instance, it would be impossible to be able to measure the drag on just the hub or to get a picture of the circumferential distribution of force around the wheel if all you have is wind tunnel data – this is where CFD can really benefit designers. For the cycling world, specifically, we are seeing cheaper, powerful, compute hardware, easier to use software, and new access models for High-Performance Computing (departmental HPC systems and cloud based on-demand HPC), coming together with dramatic improvements in the CFD solver technology (better performance, automated meshing, and the ability to handle more realistic physics).  Technology and accessibility are really coming together at just the right time.  In a few years when we look back, we’ll identify this time as the moment when things changed – where good quality CFD tools and methodologies started to become truly practical for smaller businesses.
FLO:  In your research, what level of accuracy is achieved when comparing CFD data and real world/wind tunnel results?
Matt:  To get accurate results, you need to start with accurate physics. To be able to compare CFD results directly with wind tunnel data, we would have needed to base our model on the actual wind tunnel geometry, the sting (the part in the tunnel that holds the wheel in place), and the drive motor (if present). If we had done that, I would have expected the agreement to be good, certainly accurate enough to differentiate wheel designs. Since the wind tunnel data I was able to find came from several different sources, and since we couldn’t get the wind tunnel geometry, we chose instead to model a more realistic physical domain (rotating wheel, matching the forward velocity of the bike, in contact with the ground, using a realistic ground contact patch). Despite these differences between our model and the wind tunnel(s), our comparisons with the wind tunnel trends were extremely good, particularly when looking at changes in drag force as a function of yaw angle. This told us that we’d at least gotten the large scale considerations right. Because we modeled a different domain than the wind tunnel, we didn’t expect the results to compare so favorably.
When you talk about CFD accuracy in general, one of the key concerns is that your starting ‘mesh’ is good enough to resolve the physics you’re trying to get to. A higher resolution mesh, while more accurate does add cost, so what you try to do is to start with a coarse, cheaper mesh, and keep refining it gradually until the change in the results from one mesh to the next falls below a certain tolerance. We did these ‘grid independence’ checks in our studies to satisfy ourselves and our technical reviewers that the underlying CFD methodologies we used were sound.
FLO:  Our CFD results suggest that cycling wheel aerodynamics are more about a combination of rim width and depth, as opposed to rim depth alone. What is your opinion on this topic?
Matt:  The ‘shape’ of a particular rim & tire, includes width, depth and cross sectional profile. Increase the rim depth, and the cross sectional profile has to change.  There is still a tremendous amount of opportunity for optimization here.   Since we still see a considerable range of commercial products, it strikes me that no one has gotten it completely right yet. And, the ‘best wheel’ is going to depend on what race you are doing…  A deep rim might have some aerodynamic superiority over a shallower one, but, if you know you’re going to be running into cross winds on a given course, it might pay to go with the shallower wheel.  Ultimately, as a competitor, you need to know your equipment and accept your limits.
FLO:  Where is CFD/FieldView software headed in the future and how will these advances in technology affect the cycling industry?
Matt:  I feel that the biggest challenge we need to address is one of data management. In the work I’ve done over the past few years, I have generated many! terabytes of data. And, cheaper compute with faster CFD tools is going to make it even easier to generate even more data! It is imperative that the CFD software tools keep up with these trends.  That aside, in our view, a designer should be an aerodynamicist first, using CFD as one of many tools to guide their product development process and validate their ideas.  Furthermore, they must be able to easily develop a highly productive CFD workflow that can continue to evolve as their knowledge develops.  We’re highly committed to making CFD practical and effective, and it is our hope that the cycling industry will embrace (and for some, advance) the use of CFD in their production efforts.
FLO:  What advice would you give your peers when it comes to utilizing software when designing aerodynamic cycling products?
Matt:  Be critical of the results. If something from a CFD simulation doesn’t look right, question it. This is part of why we feel that CFD postprocessing is so important. It can be pretty easy to get caught up in the ‘cool factor’ of CFD, and I think that there is a tendency to believe the results because a complex piece of software generated a colorful, appealing picture or animation. But, keep in mind, we are talking about non-linear fluid dynamics and more often than not, it takes some consideration to understand why a big change might have a small overall effect, or why a seemingly small change might have a big overall effect.
Be realistic. If you’ve had no previous experience, its going to take time to come up to speed. As with cycling components, there are a lot of choices in CFD tools, and starting off with a consultant might be a good way to get things started.
Consider whether the problem is amenable to CFD analysis. When all you have is a hammer, everything tends to look like a nail. Some things might not be a good fit for CFD. For example, lets say you wanted to look at a rear bottle cage. Unless you are willing to model everything in front of the cage, including the rider, there’s probably not much point to using CFD to optimize the shape here.
FLO:  As a Ph.D who loves research, where would you focus your attention if you had an unlimited budget to improve cycling wheel design?
Matt:  Whew. Loaded question! Let’s say that I have a few ideas…
FLO:  You’re a triathlete yourself. Tell us about your most memorable experience as a triathlete.
Matt:  Chesapeake Man ’05 (in the time of my life before the twins!)  It was one of those races where everything went right. I just found my rhythm on the point-to-point swim, and for that hour, I’ve never been more relaxed in the water. I had a great bike, finishing with the 2nd best split for the field that day, which was particularly nice. I was able to hang on during the run, and grab 5th overall. My wife also raced – it was her first full distance triathlon and she finished 4th overall on the Women’s side. Now that the twins are a little older, I’m looking forward to getting back to racing full-time, and of course, being in a new age group gives me a lot of renewed hope that I might someday get to Kona…
We hope you found this interview as educational as we did.  Special thanks goes out to Matt and the Intelligent Light team.  They’re definitely an awesome group.  
For more great content, please register for our free monthly newsletter at the top of the column on the right.  We send links to all the articles we post during the month.  Please add your comments below.  We’d love to hear from you.
Take care,

Leave a comment

All comments are moderated before being published