Supercomputing the Future of Cycling

In this special guest feature from Scientific Computing World, Keely Portway describes how road bike manufacturers are using simulation software to deliver higher performance products in less time and at a lower cost than previously achievable.

Tour de France winner Geraint Thomas recently returned to his home town triumphant and rightly celebrated, but almost as much attention has been paid to the new bikes, components, clothing and other trends to have emerged during the tournament as the riders. Today’s competitive cyclist benefits from more lightweight, aerodynamic and rigid bike models – and, in a similar way, the prototyping process for manufacturers has become less of a heavy load – in terms of both time and cost – thanks to technological advances in simulation software.

Looking at aerodynamics, in particular, a recent simulation undertaken by professor Bert Blocken, from Eindhoven University of Technology in the Netherlands, alongside KU Leuven in Belgium, used some of these technological advances – including computational fluid dynamics (CFD), by Ansys; advanced supercomputers by Cray; alongside wind tunnel testing – to gain a greater understanding of the aerodynamic interactions between the 121 cyclists of a large racing group or ‘peloton’.

Researchers carrying out the study used this combination of computer simulations and wind tunnel measurements to examine two pelotons of 121 riders, where the distance between the rows differed slightly. Computer simulations amounted to three billion cells – which the university cites as a world record for a sports application – and required Cray’s American supercomputers and tens of thousands of software licenses from Ansys.

Center stage

Cyclists push air in front of them while riding, which creates an over pressure and a depression. This air resistance is known as the drag and, because of the aerodynamic interactions with the surrounding cyclists, the rider at the centre of the pack finds his or herself enclosed by the peloton-induced air motion.

Using the Ansys Fluent CFD software with validated physical modeling capabilities, running on a Cray, Professor Blocken was able to accurately predict the flow pattern between each cyclist, mapping the drag experienced by all riders. ‘These results were so surprising that we also set up a wind tunnel test,’ he explained. The reason being that, compared to the drag of an isolated cyclist, the results showed that the resistance in the core of the peloton is down to five per cent that of an isolated cyclist.

This demonstrated that it is approximately four times easier to cycle at the centre of the racing group. Whilst it was previously documented that the best position is in the core of a racing group – around row 12 to 14 – the computer models also calculated that the drag experienced by the athletes in this position is 10 to 20 times less than for an isolated cyclist. This is in contrast to the two or three times smaller that was previously believed. The results were validated using a wind tunnel test.

The future of competitive racing

The results could have a real impact on competitive racing, as Blocken explained: ‘The calculation models used by race teams to determine the best time to escape are, it turns out, based on the wrong assumptions. This may explain why so few escapes succeed and why the peloton hauls in the riders that do escape. Perhaps these results will lead to more successful escapes.’

Returning to the simulation itself, Thierry Marchal, global industry director for sports and healthcare at Ansys, agreed that there is much potential in competitive cycling using CFD and wind tunnel simulations in combination: “In a time when simulation is crucial to accelerate and amplify innovation for high-tech industries, the peloton project and its surprising results illustrate that this simulation technology is truly pervasive and can make a huge difference in a popular sport, such as cycling.”

At Ansys, we are developing and generating simulation, but typically these simulations have been used in the nuclear industry, aerospace, aeronautical and automotive industries to make sure that we can travel safely. Along this line we have seen that the automotive industry and the sport automotive industry, like Formula 1, has been in the driving seat since the 1990s. About 10 years ago the first one-billion cell simulation was done for a Formula 1 car. Then, later on, Red Bull was going all numerics, gearing up on this kind of wind tunnel testing. But what about the non-motorsport application? We can use wind tunnel analysis in order to understand what’s going on, so why move into simulation? Well, wind tunnel analyses are very important, but they can be expensive, time consuming and you need to have the athlete available. So, although they are extremely important for the validation of the model, a combination of wind analysis and CFD simulation might be extremely useful.”

A winning formula

The nod to Formula 1 is echoed by David Power, CTO at vScaler, which has been involved in CFD work. Power explained how the process transfers itself perfectly to non-motor sport: “CFD is absolutely used in cycling. It’s about how the air flows over the bike design, the stresses etc that you can see on it. You can understand where the metal is going to flex at which points and at which pressure etc.

The resistance that a cyclist encounters can be greater than 80 per cent, moving the riding positions alone can cut that by 40 per cent using basic maths, but to really limit the resistance needs you to understand the airflow across the bike and the rider in detail and model the most efficient position and components. Also in competitive racing, it helps to avoid any infractions with the rules, which again is so similar to Formula 1.”

vScaler, in partnership with Boston Ltd, installs optimized CFD software on its cloud platform (both private and public) and allows customers not only to run CFD in-house but also leverage public instances to handle excess or time-sensitive workloads as and when the business demands. ‘That means,’ said Power, ‘that it’s a very cost-effective way for small or less rich engineering firms to be able to access a similar suite of programs as the larger companies.

OpenFOAM is an industry standard open source software package. It’s very, very good. It can certainly compete with commercial software, as it allows the customer to set up an environment very quickly and very cost-effectively. The cloud environment is also open source, it’s based on open stack, and we tune for HPC. CFD is one of those programs so we create a framework in the cloud and that’s already optimized for the customer.”

Path of least resistance

The bike manufacturer then simply needs to begin testing their models, as Power continued: ‘They create a model which is a design of their bike and they will run it through the algorithms, computational part of the fluid dynamics and that gives them the capability to view the results and make changes to the bike to optimize its performance. If you’re talking about a racing bike, they’re almost certainly looking at pregeometries of the riders, so they can see that when the rider is cycling, the position that he has, the amount of resistance or drag, and then looking at the bike design to make sure that it has the minimum amount of friction resistance speed through the air. So that’s essentially what they’re looking for, to optimize how that bike moves through the air.”

vScaler uses its OpenFOAM expertise to tune models for customers as a service, having already gone through the process of all of the optimizations in the background. “That will include,” explained Power, ‘things like TPU pinning, optimizing the meshing, lots of different optimizations that are done to ensure that when the customer starts to run that program they are running it on optimized systems, so essentially we are taking away a lot of the requirement for a CFD software specialist at the customer’s end, to allow them to actually run their models and get the best performance from the computer.”

Looking to the future of modeling and simulation, Power referred back to the Formula 1 model comparatively, and concurred with Marcha’s view that developments are moving beyond the traditional wind tunnel alone. “You’re able to compute a lot larger models now many times,’ he said. “You used to have to do a lot of modeling in the wind tunnel, and that is really being moved away from. There are only very specific parts that you now need to model within the wind tunnel. For example, within Formula 1, there is a much larger move towards computational process, and then proving on the track, than the amount of time that was spent building models, putting them into a wind tunnel, running it, analyzing it, whatever. It’s significantly faster in terms of the time that they can get their results. This will absolutely apply in the racing bike arena as well.”

Tunnel vision

Simulation technology specialist Altair has an entire suite of simulation products available under the HyperWorks brand. This was utilized by the research and development team at Specialized Bicycle Components when they developed a new bike for the 2015 Tour de France.

This bike is our go-fast bike,’ explained Chris Yu, aerodynamics and research and development lead at Specialized. ‘It’s a race bike, so it was designed for aerodynamics, first and foremost, but also a very close partner to that, in terms of target, was structural and weight, so keeping those things balanced, in terms of the design process, was one of the things we did throughout.”

Yu believes that, when it comes to designing a new product and trying to optimize for aerodynamics, the foundation and key first step is simulation. “It’s really cost- and time-prohibitive to go and make 1,000 prototypes,’ he said. ‘Whereas, in simulation, you can really explore the boundaries of that envelope to figure out which design to use. When we have a road race like the Tour de France, there are many stages where aero is the biggest factor to overcome, but there are a lot of stages that are basically hill climbs and a sense that you’re fighting gravity, and in those cases, minimizing weight up to the minimum allowed in the rules is a big deal. Having the tools to be able to optimize the system to have minimum weight for the aero and structural targets we’ve set is super critical.”

The team utilized the Hyperworks’ CFD capability using CFD solver AcuSolve for the fluids optimization, and Optistruct to optimize the structural and carbon ply by ply analysis to ensure that there is no excess material anywhere in the frame. Yu continued: “When we started transitioning into using powerful tools like the Altair Hyperworks suite, we were able to realize a lot of those small changes that maybe you can’t physically measure individually, if you stack them up in simulation and then build a prototype, you’re able to then measure a really large gain in the tunnel.

By using the Hyperworks tools and the CFD analysis, on this frame, especially in terms of the amount of prototypes, we’re able to put through the simulation in the time block that we had to design this thing, I feel is at least a two- to four-fold increase in what we had before, in terms of the number of prototypes that we’re able to run virtually.’

Virtual wind testing

Returning to the wind tunnel testing debate, Altair also has available for customers a virtual wind tunnel tool under the Hyperworks umbrella. VWT is a vertical solution providing an efficient environment for external aerodynamic studies. With its automated and streamlined workflow based on the company’s CFD solver AcuSolve, the virtual wind tunnel performs simulations of flow around objects, delivering transient or steady state solutions.

With a focus on drag and lift prediction of vehicles in the automotive sector, other uses include aerodynamics of buildings, motor bikes and bicycles, as Montreal-based cycle manufacturer
Argon 18 discovered.

Argon 18 partnered with the ÉTS Research Chair on engineering of processing, materials and structures for additive manufacturing to make a track bike for Lasse Norman Hansen, a Danish cyclist competing at the 2016 Rio Olympic Games. The aim was to develop a bike that was stiffer, highly integrated and more aerodynamic, providing greater efficiency.

Martin Faubert, research and development manager at Argon 18, explained: “Our main goal is to enhance the performance of the rider by providing the best bike possible. Improving the structural design and aerodynamic performance by using Altair HyperWorks’ greatly streamlined our product development process.”

A key aspect of the project was the development of a new aluminium stem to be used by Hansen, and the company used OptiStruct for structural analysis, AcuSolve for CFD, and VWT.

Drag reduction

Finite element analysis (FEA) was employed to understand the structure of the design, and then to improve and optimize it. CFD analyses and virtual wind tunnel simulation helped to improve the aerodynamics.

Several iterations between the FEA and CFD processes followed, trying different configurations of the components, making the blade wider, thinner, and taking it far from the wheel, bringing it closer, while keeping a close watch on the CFD and FEA data.

The design improvements resulted in a significant reduction (six per cent) of the aerodynamic drag (CdA), a critical parameter in making faster bikes. The linear stress analysis was conducted using Altair OptiStruct for validation of the stem body and clamp design.

The stress analysis demonstrated a greater stiffness, about nine per cent, than the typical carbon fibres stem. It also identified several dimensions to be adjusted in order to preserve the integrity of the parts, such as the thickness of the tubular section and the handlebar clamping section.

To ensure its reliability, a fatigue test was performed on the final design, which showed no significant loss of rigidity or cracking, as well as good correlation with the stress analysis. The stiffness proved to be greater than the fibre-reinforced composite bracket.

This story appears here as part of a cross-publishing agreement with Scientific Computing World.

Sign up for our insideHPC Newsletter