Tunnel Vision: wind tunnels and CFD in Formula One

With Formula One now entering a phase of restricted engines, specification ECUs and single tyre supply, one of the greatest areas of development and research by the teams has been left largely unchecked.

Aerodynamics is critical to a modern F1 car. Despite frequent regulations to reduce downforce, the teams claw the deficit back. The tools in this battle are wind tunnels and Computational fluid dynamics (CFD) - two very different solutions, each with its roots in different eras, now working side by side. So how do these solutions work, and how are they going to develop in the future? And will CFD trounce wind tunnels as the key tool for the teams?

Aerodynamics has always played a part in racing car design, even if the intention was not there to maximise its use. In the early days, aerodynamics was not understood and generally ignored, often to the detriment of a car's stability or performance. Any car travelling at speed is subject to lift and drag.

In the early days, streamlining and reduction in frontal area were the aims to reduce drag. Towards the late sixties, teams realised lift was creating a handling problem at high speed. The addition of spoilers and then wings reversed the lift and created downforce.

The sudden realisation that adding downforce would then improve lap times lead to a steep change in the design of the car. Early attempts both before and after the banning of suspension-mounted aerofoils were simple and crude. Few teams backed up their designs with any sort of testing to validate the effect they were having.

Around that time, wind tunnels were in use in aerospace, plus a few educational establishments had their own small and simple tunnels. The first use of F1 cars in Wind tunnels with small wooden models did not provide the accuracy required.

In those days, teams only had a few engineering staff, who were only trained in mechanical design. Additionally, at this stage there was no research material on race car aerodynamics, the subject had only just been created. What knowledge existed was aimed at aircraft or the basics of road car aerodynamics.

A few individuals persisted with the development, and through the 70s and 80s aerodynamics came to the fore, and suddenly F1 was not a simple chassis and engine formula any more. Aerodynamic knowledge was on such a steep learning curve, that large gains in lap times were being found constantly.

With the advent of ground effect in the 80s, suddenly a car's mechanical and engine designs were being severely compromised to accommodate better aerodynamics. With the increase in performance from aerodynamics also came rule changes, and the cat and mouse game between the teams and governing body began.

Throughout this time, teams were slowly gaining knowledge in how to embrace aerodynamics and how the wind tunnel could be used to evaluate their ideas.

Wind Tunnels

In essence, a wind tunnel is a very simple solution - basically a duct with a fan drawing air through a test section where the model sits. If the model is accurate, then as a result the force measurements taken from the model should also be accurate.

But there is a lot more subtlety to how a tunnel works and how it should relate to reality. At first, wind tunnels simply had the car sat on the floor of a tunnel that would more normally be used for aircraft models, mounted in the centre of the test section.

This provided a fairly accurate representation of flow over the top of the car, but F1 cars sit very low to the ground and, what's more, the car and ground are static relative to each other. In real life, the car moves and the ground stays still.

So wind tunnels adopted a moving belt to sit the car on, and the movement of the belt had the same relative effect as the car moving. Suddenly this bore a closer relationship to the real world; the car and wheels moved relative to the ground, and the complex airflow near the ground, which is not relevant to aircraft, was being recreated.

As test models grew larger, there became a need to account for the blockage the model creates within the test section. Although the actual model might fit neatly within the test section, the car's wake spiralling from the rear wing take up much more space. If this isn't allowed to expand, then the results will not be accurate.

One cheap method to retain a relatively small fan and tunnel section was to use an open jet. This is where the tunnel is split open around the test section, where the model sits, and allows the wake to form in a much larger area. But some corrections need to be made to the results on account of the open section.

Nowadays, with F1 teams having their own tunnels and are the sole users of the facility, they can tailor its size to suit their needs. Much larger tunnels and, accordingly, much larger test sections are created; this eases the blockage effect, although some teams sport a slotted wall that still allows for larger wakes to be managed.

The accuracy of the model and its attitude to the track has also been considered. The real car rolls, yaws, goes up and down through ride height changes and even steers its wheels. These were duly reproduced, at first manually and latterly by automated actuators within the car model.

Computational Fluid Dynamics

By the end of the 20th century, wind tunnels were a very accurate method of aero development, but a new option came on the scene: computer simulations.

With the explosion and complexity possible with modern software, the effect of the air flow over a virtual car was possible. Technically called "computational fluid dynamics" - or CFD for short - this embryonic technology was adopted quickly by the teams and in just a few years has matured into a very reliable solution.

CFD is actually not a single application, but a suite of tools. There is a basic three-stage process: meshing, solving, and analysis.

First, a 2D or 3D model of the part is created - with F1 cars now completely designed with computer-aided design (CAD) systems, the conversion is an easier and accurate process. The model is a mesh of cells including not only the car but also the area around it. This area represents the test section of the wind tunnel, and the airflow calculations are carried out for every cell with in the area. The more cells, the more complex the model. First tests were only carried out in 2D, and only simple 3D models were tested.

The limitation in model size and complexity lay in the need to run the model through the second phase, where the effect of the airflow over model is resolved. This takes a huge amount of computing power and time. Typically, runs were set off in the afternoon and carried out over night, with the results ready for analysis the next day.

The last phase is where the results are analysed. While a wind tunnel has a limited set of forces to describe drag and downforce, the post-processing results from CFD give millions of data points. These can describe not only the total forces created on the car, but also the pressures and velocities in the air around the car.

These can be delivered in a bewildering array of visual formats and sliced in any direction to get to see the flow at any given point.

The ability to visualise these results has aided aerodynamicists, by being able to go back and look in detail at what the air is doing - something that is not possible with wind tunnels.

CFD relies on a lot of assumptions, as the mathematics isn't fully understood. The more complex the model and the flows around it, the less reliable the results were. This has improved, but still CFD is not yet an exact science.

Modern Wind Tunnels

In recent years, the incessant pressure for teams to produce more gains in aerodynamic performance has driven the use of wind tunnels in two directions: firstly capacity, and secondly accuracy.

Car performance will nowadays see improvements in smaller increments, and to make these smaller gains, more parts need to pass through the wind tunnel - which in turn means the tunnel itself needs to be able to measure these smaller differences in drag and downforce.

As each run in the wind tunnel takes around half an hour, there is already a theoretical limit in the parts you can try per day. But a team can maximise what they achieve in each run, complete more runs through longer shifts, and then increase capacity with a second tunnel.

A modern wind tunnel model is commonly a 60% scale model; this large scale has benefits in the correction factors needed to convert the loads measured in the tunnel into the actual life-size loads. The size also allows for more instrumentation to be fitted inside the model.

Suspended by the vertical stinger mounted inside the roof above the tunnel, the model can now have the mechanics to allow for attitude changes to be made automatically. So in a single run, the car can adopt over 30 points of ride height, roll and yaw, as well as steer its wheels.

These different attitudes create a map of the car's downforce and drag for a given configuration. These maps are used to decide what set-ups are too sensitive to attitude changes, and the maps of successful set-ups are handed to the race engineers to plan for race set-ups on the real car.

To achieve more runs in a day, the teams need more aerodynamics staff. These will work in shifts to achieve the maximum 24/7 operation of the tunnel. The staff are split into groups who will work as teams to set up their model in advance and then have a block of tunnel time to complete their runs.

For teams with sufficient budget, a second tunnel is the next step in boosting tunnel throughput. In simple terms, the second tunnel doubles capacity.

But adopting a second tunnel is littered with issues. Teams with older or simpler first tunnels will normally opt to commission a second tunnel with greater model sizes and different internal layouts and mechanics. This then creates problems in comparing results from one wind tunnel to the other. Williams were particularly afflicted by this problem with their second tunnel.

As the law of diminishing returns matures in F1, the aerodynamicists who were once happy to take improvements in single steps, are now happy just to make improvements in tenths or hundredths of a point.

This has lead to most of team's aero upgrades to be minute, unnoticeable changes. This emphasis on ever finer details also has forced the staff working with the tunnels and models to be even more diligent in recreating a realistic testing environment.

The move to larger models has aided this process as well. The models are accurate to the same scale as the model, so in physical terms the models are even more accurate than the real thing. Every part of the car is repeated faithfully, from the bodywork to the internal detail in the sidepods and brakes.

This fastidiousness on detail now extends to pneumatic rubber tyres. These are provided by the team tyre partner (Bridgestone alone in 2007). They allow the tyre to form the squash and deformation seen at the contact patch; this makes the interface between the front wing endplate and wheel much more accurate.

Another area that is starting to be examined is unsteady flows. You might believe the air around the car always flows in the same controlled manner and that the loads measured in the wind tunnel represent the only loads the car could see around a lap. But, in reality, an F1 car going up and down on its suspension is akin to a bird flapping its wings.

In a wind tunnel, the car could be measured at ten different ride heights, measured static, one at a time, giving ten results. A bird's wing in motion may pass through ten different positions, but this is a dynamic motion - the upstroke and downstroke of the wing will be significantly different at each of the ten different positions. This transient response is also seen on a real car.

Attitude changes on a real car are rapid. As the car brakes, the rear goes from static ride height to full extension and back again in a few seconds. This equates to the bird-like flapping motion, and the car's diffuser will react very differently to the wind tunnel predictions.

Therefore, teams developed even more complicated actuators within the model, which can move the car dynamically to recreate the movement in real time.

As the model needs to be suspended either by the vertical stinger or optionally by the wheel supports, there is always an interference that is not seen on a real car.

Durham University has started a prototype set-up that uses magnets to make the model actually float on the rolling road; this still allows the forces to be measured through the wheels. This is far from being ready for adoption in F1, but it highlights the continued pressure for accuracy in the tunnel.

Modern CFD

After a basic start, CFD has developed massively in the past ten years. The expansion into 3D models with ever greater resolutions has made the results from the simulations far more accurate. Every F1 team has a CFD programme, with the larger teams investing even more resources into the programmes.

Just like a wind tunnel programme, CFD programmes require large investments, though not into buildings and industrial equipment. Instead, the investment is in the computing power, software development and staff.

At the moment, the bottleneck in CFD programmes is in the computing power. A typical modern model still takes 20 hours to resolve, before the aerodynamicists can even look at the results, let alone analyse them.

With this restriction, the aerodynamicists need to find ways to make the use of the CFD tool more effective. This can be achieved by prioritising what scenarios are run through the tool, what to do with an increase in computer power, and what scenarios aren't possible with a wind tunnel. To decide what models are run through the tool, the workload from the wind tunnel needs to be taken into account.

As running a full car takes too long, running just smaller components is very effective. Front and rear wings are often carried out largely in CFD - particularly the front wing, which runs in air uninterrupted by other parts of the car.

Some things can't be monitored in the wind tunnel, such as areas hidden within body work whose trailing flows need to be examined. CFD excels in this area, and mapping the flow inside the sidepod or through brake ducts can be clearly visualised in CFD. This is apparent by the explosion of brake duct solutions seen in the past few years.

Additionally, wind tunnels are good at measuring multiple attitudes from a single configuration. CFD can only run a single attitude and model configuration per run. With this in mind, it would be inefficient to use CFD to map the details of a single configuration. So with the bread and butter work of aero mapping the car ruled out, CFD focuses on development work.

When creating a new wind tunnel model isn't economic in terms of budget or time, CFD becomes the rough cut for new ideas. New mesh models can be created rapidly either by cutting and pasting new elements into the model or by morphing surfaces on the current model. The ideas that prove the most promising then go to the wind tunnel programme.

Equally, the model used is important. CFD so far works best with attached flows; an F1 car has a cleaner airflow when running straight ahead. When a car is at an angle to the airflow (yaw), the flow becomes detached from the bodywork and modelling becomes more difficult.

This problem can be turned to an advantage, however, as CFD models need to be run in a straight-ahead attitude. Then, only half the car needs to be run (split along the car's centreline) and symmetry assumes the flow over the other half of the car. As the model is only half the size, it takes less time to resolve the run.

Once the mesh has been run through the CFD programme, various tools can be used to interpret the results and can even alter the mesh slightly to morph a surface, providing slightly different results. This offsets some of the inefficiencies.

With computer power increasing all the time, the speed at which a CFD programme can resolve a given umber of cells also increases.

This increase in computing efficiency can be used either to speed up the run, or allow even more detailed meshes to be achieved. As CFD has yet to mimic a wind tunnel's accuracy on full car models, the latter option to enhance the model is chosen.

The computer that the teams run is not in fact a single computer at all. Clustering technology links together lots of processors on separate computers to effectively make one 'super-computer'. The use of over 150 processors clustered to complete a run that still takes over half a day shows that CFD is still very much technology-dependant. However, technical partnerships with computer manufacturers, as well as the reducing costs of hardware, mean that the power available to the engineers will keep on increasing.

A restriction in what is achievable in a wind tunnel is what the model can be made to recreate and how the wind through the tunnel behaves. A real F1 car has hot surfaces, blowing exhausts, and goes around corners. While it is possible but impractical to create the first two in a model, the latter is clearly not possible.

CFD, on the other hand, can recreate the thermal effect and present the oncoming flow in a curve to replicate the oncoming air the car actually sees on the track. These and other possibilities open the potential of CFD to go far beyond what a wind tunnel can achieve.

Now and the Future

There's always the question whether CFD will replace the wind tunnel. Right now the answer is no - CFD is neither fast nor accurate enough for the rapid pace of development in F1. Wind tunnels are not perfect, but their results are close enough to the real world to be the main tool for many years to come.

But CFD is improving. It already provides the engineers with views of flows not seen before, and it is already the main method used to largely develop major aerodynamic pieces, such as the front wing. But the results are always referred to the wind tunnel for correlation, and the subsequent mapping of the part's effectiveness at all attitudes.

So tunnels and CFD are complementary; the development carried out in CFD is taken to the wind tunnel to cross-check the results, then to the track to finally confirm the parts.

But looking into the future, with the increase in computing power and the size of the clusters used by the teams, CFD run times could be reduced to the same if not less than a wind tunnel run. If each run could incorporate the different attitudes, and results are dropped out into visualisations that are fed back into the team's designers and simulation tools, then CFD has the ability to revolutionise the way aerodynamics are used.

Imagine this scenario: an aero configuration is tested in CFD around a virtual lap on a specific track, and the results are logged. For a given track, one aero set-up could prove faster than one with the usual 30-point aero map from a tunnel, as the set-up provides more downforce in certain corners relevant to the track.

This virtual lap aero data could be collated with engine power curves, set-up information and feedback from the seven-post rig. All this data is assembled together, and then the race engineer would have an unlimited set of options to base his race set-up on - even, perhaps, running a few CFD iterations from his laptop in the pit garage to refine the set-up.

It might be ten years before CFD can beat a wind tunnel on speed and accuracy, but thereafter it has the possibility to go far beyond what we know today.

We want to hear from you!

Let us know what you would like to see from us in the future.

Take our survey

- The Autosport.com Team

Read and post comments