Why mistrust in development tools results from an inherent aerodynamic trait

The sophistication of aerodynamic research techniques has grown immensely in recent years. When I started wind tunnel testing in the 1980s, the models were crude, the tunnels small and the instrumentation limited to a simple balance to measure forces and moments as well as a manometer to measure a few pressures. CFD didn’t exist, at least not in the realms of motorsport. The aerodynamics department consisted of two full-time employees and two others (me included) who also performed other duties.

Today an aerodynamic department will be the largest single engineering group in a team: around 100 people with jobs ranging from developing CFD tools and methodologies, through wind tunnel model design and build, to those actively engaged in developing the aerodynamic shapes and keeping on top of actual performance.

With such expertise one might expect
that something near perfection would be commonplace.
After all, when a suspension component is designed and analysed for stress and reliability it rarely fails – what’s
so different with aerodynamics?

To answer this it’s worth looking at the methodologies and tools employed in developing aerodynamics. Within a given set of regulations, the aerodynamicists responsible for putting performance on the car will focus on shapes that interact with each other to produce consistently high downforce levels at minimum drag.

The operative word here is consistent. If we
go back to our suspension comparison, if a load is applied to
a suspension wishbone, we will ensure that the strain resulting from the load remains within the linear region of response. In other words, if we apply a load the wishbone will deflect a known amount and when the load is removed it will always return to its original condition.

Aerodynamics, unfortunately, don’t behave in such a predictable manner although our tools might have us believe they do. The starting point for development is normally the use of computational fluid dynamics, CFD.

PLUS: How F1's processing-intensive development tool really works

This is a powerful tool and the myriad of outputs from a simulation allow us to understand a great deal about the flow field around our components.

CFD is a powerful tool, but ultimately does not provide all the answers

Photo by: Shutterstock

Unfortunately, with the most common types of simulation, it gives a time-averaged solution to something that’s unsteady and therefore changes rapidly all the time. There are techniques that overcome this to some extent but they use a lot of computing time, which is limited under the F1 regulations. So these may only be used when there is some confidence to take the development through to the next stage.

Wind of change

This next stage will generally be wind tunnel testing. Many people think this is a better arbiter of good and bad and to some extent it is – but it’s still a simulation, albeit a physical rather than virtual one.

One of the problems with wind tunnel testing is that it assumes near-perfect conditions. The designer of the tunnel will spend a lot of effort to ensure that the flow in the ‘working section’ – where the model is placed – has very good flow characteristics, with low turbulence intensity, consistent flow velocity, and minimal ground boundary layer build-up.

Recently, the smooth steel belt has been
called into question since it may not represent the roughness
of the aggregate in a typical track surface. A number of teams are in the process of replacing them

Unfortunately, a real racing car rarely, if ever, runs in these conditions. It will normally be exposed to atmospheric wind, wall effects from barriers, and particularly turbulence from a car in front.

Ground conditions are particularly important with the current generation of cars due to the amount of load generated under the floor. In the tunnel, a large driven belt runs under the car to simulate the car moving over the track. For some years now this has been a flexible steel belt, which had a much
longer life than the old Ammeraal PTFE-backed polyamide belts.

Recently, however, the smooth steel belt has been
called into question since it may not represent the roughness
of the aggregate in a typical track surface. A number of teams are in the process of replacing these smooth belts with ones with a rougher surface.

In addition to these problems, there is the likelihood that aeroelasticity will affect the results. As far as bodywork deflections go, these can be understood and to some extent compensated for. The deflection of the tyres is more difficult. Loading devices on the wind tunnel model will compress the scale pneumatic tyres to approximate the contact patch size and the sidewall shape, but simulating the tyre shape when it’s deflected by lateral loads is far more difficult.

Cars never run in perfect conditions seen in the wind tunnel

Photo by: Andretti Autosport

So if these two primary methods of aerodynamic development have their limitations, one might expect that the real answer lies in what’s measured on the track. Red Bull team principal Christian Horner recently gave this credence, saying (in relation to his team’s well-publicised issues with development parts): “It’s not unusual that when something’s not working on the car, you end up with different readings from your simulation tools and they don’t converge, then you get three sets of data: you get CFD, you get wind tunnel and you get track. Obviously the one that counts is track…”

In my view, this is an aspirational simplification which doesn’t tell the full story. The problem with track measurements is that a racing car isn’t a laboratory instrument. Even with the sophistication of modern data acquisition it cannot measure to the standards of the simulation tools.

While a huge number of pressure tappings can be installed in the floor and wings, which give a pretty good idea of what’s happening, the actual load measurements are more difficult: measuring load directly through strain-gauged push- or pullrods only measures the loads on the body and misses the loads imposed by the wheels and hub assemblies. At the rear this is significant: the brake duct winglets and rear suspension contribute around 4.5% of the total load as well as aiding diffuser efficiency by interaction.

In addition, drag is hard to measure since this needs to be done by coast-down tests where the car is allowed to decelerate freely and the subsequent speed decay analysed to separate rolling resistance and aerodynamic drag – not a particularly easy thing to do during a race weekend. The use of aerodynamic pressure measuring rakes and, to some extent, flow-viz paint on the bodywork help understanding but neither actually provides a measure of load.

The infernal triangle

So, we have three measurements – from CFD, the windtunnel and the car – but each has some deficiency. I term these measurements the ‘infernal triangle’ since, if we plot lift and drag from each method, we end up with a triangle of points.
We hope the correct figure lies somewhere within the bounds
of that triangle – but where?

Even if we knew the answer to this vexing question we still wouldn’t have a complete picture of what was going on. The reason for this is that in simplistic terms, if we consider load measurement, we will tend to do this on track at a constant speed.

We would then compare
the results with a CFD run done at the same ride heights and an extraction from the wind tunnel map, again at the same ride heights. We would hope there isn’t a side wind but, if there was, this can be measured on-car and corrections made. Drag measurements are a bit more complex because the coast-down test will involve changes of ride height as the speed decreases.

The infernal triangle is an important if vexing area of focus

Photo by: Ferrari

For this reason, teams are often more interested in pressure measurements made at multiple positions on the floor. Integrating these values over the area of the floor can give a better understanding of the load the floor is producing – which, after all, is the majority of the aerodynamic load. Pressure taps on the wings can be used in a similar way and both give a dynamic reading throughout the lap. Pressure measurements can also give an idea of when the aerodynamics start to behave in an unexpected manner.

To appreciate the significance of this, it’s important to understand that the aerodynamics of a racing car run much closer to stall conditions than one would ever find on an aircraft. That’s because this is where maximum performance is found.

Also, unlike an aircraft, Formula 1 cars rely a lot on vortices to enhance downforce. The strakes at the entry to the floor, just below the radiator inlets for example, are critical to setting up vortical structures that encourage flow under the floor in a coherent manner.

Correlation means that if a positive incremental change is found in the aerodynamic tools, then that positive increment also shows up on the car

The trouble with both these factors, particularly the use of vortices, is that they can be very critical to operating conditions. A vortex that’s encouraging strong flow under the floor, and behaves perfectly in the refined conditions of the wind tunnel, may burst in the rapidly changing dynamic conditions caused by movement of the chassis relative to the ground as the car negotiates small bumps.

Appearance vs reality

This season it appears that the lack of correlation between the directions suggested by the development tools and the behaviour of the real car is more common than in previous years. But is this a fact or a combination of circumstances?

The season started with expectations that Red Bull was again going to dominate the championship and, apart from the problems in Australia, it maintained great performance even when introducing a floor upgrade in Japan. The turning point was Miami but no changes were made to the Red Bull car for this race.

Instead, McLaren produced one of the biggest upgrades seen for a long time: a completely new front wing; altered front suspension geometry; revised brake ducts front and rear; and a heavily revised floor which required changes to bodywork, engine cover, sidepod inlets and rear suspension.

Since then the balance has swung in McLaren’s favour and it’s evident that subsequent upgrades have worked well. The real question is whether Red Bull has lost performance or just stayed still.

PLUS: Why Haas is a crunch barometer of F1's biggest improvers in 2024

McLaren has made giant strides this year, but the extent to which Red Bull has stuttered is tough to quantify

Photo by: Lubomir Asenov / Motorsport Images

On the surface, one might think the team has gone backwards since Max Verstappen has been vocal in his complaints – something we hadn’t heard in the previous season. However, in competition all things are relative. If you have a car noticeably faster than the opposition then, once out front, in relatively clean air, the driver can dictate the pace, looking after his tyres and maybe driving just a few tenths off the car’s maximum. This makes a world of difference since he never has to explore the limits.

As soon as the car is pushed towards those limits, maybe now in dirty air since it’s no longer leading, the inadequacies of the handling become apparent. Is this simply what Red Bull is now experiencing?

We can’t ignore correlation problems – but what do we mean by correlation? To many it means that forces measured on the car replicate wind tunnel forces, but we’ve already explained that is not possible. To me, correlation means that if a positive incremental change is found in the aerodynamic tools, then that positive increment also shows up on the car.

Equally, we need to consider the full aerodynamic map. For convenience, we may express aerodynamic performance as single, weighted sets of numbers, but the reality is that the car undergoes not just a trajectory through a corner but also a trajectory through the aero map as it traverses that corner.

The aerodynamic map is a multi-dimensional representation of the load developed front and rear at any given condition of ride heights, roll, steer and yaw. A secret of good aerodynamics is understanding that path and ensuring the car maintains a similar aerodynamic balance of front and rear load as it enters the corner at high speed, drops speed to the apex, and then accelerates out of the corner. If the balance is changing too much, particularly in the wrong direction through the corner, then the driver may feel the front and rear are disconnected –
a phrase we’ve heard a lot from Horner recently.

So, have our trusted tools suddenly let us down – or is the closer and more intense competition this season demanding a more exacting set of solutions to the inevitable questions simulation is required to provide? While there’s an element of both, I think the answer leans more toward the latter.

True, we’ve seen indecision about the fidelity of upgrades from RB, Ferrari, Red Bull and Aston Martin, but this isn’t
a new problem. At Renault, we won the championship in
2006 but were less competitive at the end of the season than at the beginning. On reviewing the situation post-season,
we established that many of our upgrades had in fact
detracted from performance.

This year the extreme rivalry leaves nowhere to hide both on the track and in the engineering office. Just as the drivers need to be on the limit and will occasionally exceed it, so too will the aerodynamicists and their extremely sophisticated tools.

Engineers have access to huge amounts of data, but it doesn't make their jobs easy

Photo by: Williams F1