Unseen parts: The nightmare of aerodynamic correlation

Often there is discussion on “aerodynamic correlation issues” in the specialist press.  Correlation is a strategic aspect to deal with while developing a F1 car.  There then appears comments in the press like “we are suffering with bad correlation” or with horrific sentences like “our correlation is lost”.  Eventually, positive comments come out like “our correlation has improved” and this would magically reset the situation.

My opinion is that the main problem with all this is it leaves the reader with the impression of talking about a sort of “black magic”!  More seriously, I would not be surprised if the majority of you simply could not understand why, despite all the money invested and the level of the technology in F1, that this can still happen. Also, I’m sure that you would not understand why this seems to happen more often to the biggest team!

I will try to offer a concise and simplified list of the major reasons behind this dilemma. Like the famous ‘Sword of Damocles’, it undermines the serenity of all the aerodynamicists and all of their colleagues involved in the development process. But before, please let me clarify the meaning of word “correlation” that is, in some cases, confused if not abused.

Correlation means a mutual relationship or connection between two or more things. Ok, so: firstly we talk about relationship and not about equality (i.e. not a “perfect repeat” to be expected); secondly, for the matter of this article, the “two or more things” are a calculation, two or more experiments and a person. Sorry, say again please? Sure: the relationship/connection to find is between CFD results, measurements from wind-tunnels, track measurements, and the driver feedback.

Generally speaking however, the four environments are not put in a relationship at the same moment. As a “rule of thumb”, being perfectly aware of the crude limitation of any rule done by thumbs!, the first relationship to find and “secure” is between CFD outputs and wind-tunnel results.  That second tool is still in the majority of cases of aero development, the “final word” for what will be sent to production for track test and, hopefully, to be finally raced. But not always: there are, in fact specific and peculiar aerodynamic aspects (see for instance thermal effects like brake- and engine-cooling, or specific flow behavior like those related to flow-features involved in the rear wing behavior) for which CFD is nowadays considered the absolute reference, due to the constant improvement of the model and the intrinsic limitations related to a wind-tunnel experiment.

By far, when the tunnel-to-track correlation is suffering, the implications are vastly more serious and, generally, the solutions are expensive and time-consuming. This shouldn’t be big news for most of you (the braver readers should have noticed the usage of the plural when talking about wind-tunnels two paragraphs above): teams are now used to performing aero experiments in more than one facility. Do two tunnels “say” the same? Well, obviously no. In worst cases, with reasons that are behind the scope of this article, seriously conflicting results are the very first dilemma to solve if/when this happens. Result-conflicts would never become 100% agreement for all types of aero test, but experience and time (and patience…) generally helps to define a robust strategy to correctly manage the development in two or more tunnels and to elect the “master” facility for a specific item/area of the car to develop. In some cases the other two environments (CFD and track) can work as “referee” to support this decisional process.

The synergy, and not the competition, between CFD and tunnel activities is the key-factor for the aerodynamic development to be carried out as much successfully as possible. This item alone would merit a dedicated article, which I promise to think about…

With the above exceptions, the results from the tunnel, hopefully supported by the already-set correlation with CFD, are finally compared to the feedback arriving from the car: this step involves both measures (thanks to all the sophisticated sensors which a F1 car is now equipped with) and the driver “opinion”. The latter, as you should easily imagine, sometime can be “definitive”, despite not always as scientific as the engineers would expect. But please remember that there is not, and there will never be, any simulator/sensor, system, engineer’s ability in data analysis/data-interpretation, routine, etc… that is able to give feedback on the “overall car behavior” at the level of a top-driver.  Who is called upon to list the pros & cons of a specific modification to the car? Absolutely, his feedback deserves trust but, being subjective anyway, needs correct interpretation by experienced engineers. When the driver’s feedback agrees with the evidence of the data or, even better, is able to complete what the data cannot say (limited exercise, read after please)… bingo!

Hold on, and the lap-time? Well, despite this appearing illogical to you Gents, lap-time is important, but quite rarely is it the first figure to look at when deciding if a new aero-package is “better” or not. The lack of time for testing a new configuration in different moments of the day coupled with the fast-changing nature of the track and tires (there they are again!!!), make lap time based judgment quite difficult, if not risky in some circumstances. Please be aware that, in the majority of cases, the claimed improvement in terms of lap-time is in the order of few tenths.

Before proceeding, please let me digress a bit? The current lack of testing imposed by the Sporting Regulations, has a consequence that should be evident to lots of you and is a topic that is seriously debated. That is a different story; hence I will not offer a deep comment here. Just let me say that in any competitive exercise, what is limited somewhere is compensated for somewhere else. Those who have bigger resources can react faster. In F1, that is an extremely competitive exercise, the progressive reduction of track testing was surely responsible of the proliferation of all the off-track jigs and simulators, whose level and sophistication has improved massively in the last few years. It’s a very interesting challenge for any engineer, but, I’m not so sure if the final result on an economical basis was the best answer ever. Think about the ultimate spirit of motorsport, where the word “sport” is still at the same level as “motor”, correct? More testing opportunity, even if regulated and optimized (on a Monday following the GP, for instance) should not to be too expensive and should be given back to Teams. While writing Teams and Federation seem to find an agreement for a wise re-introduction of track test starting from 2014. Definitively, this is a positive new.

Back to correlation now! As you already can imagine dear readers, the correlation-exercise is quite complex to control. And while correlating tunnels to track, that is the core-business of this exercise, all should reckon that two experiments are involved here, each of them with their own experimental errors, limitations and levels of uncertainty. These aspects have to be taken into account carefully; otherwise the comparison will not be fair (i.e. a fundamental error of approach would be made). But there is another important aspect when dealing with aerodynamic correlation in F1, this not always being easy to remember, even for some of the insiders: CFD, tunnel(s), and track are environments with strong specificities.

A CFD run would tell us everything about the complex features of the flow around the car (flow path lines, vorticity levels, wakes, flow recirculation’s and separations, internal aerodynamic features, single-part contributions, effect of unsteadiness, etc… ) around a CAD-originated “virtual car” in a fixed set-up, perfectly “rigid and stable”, immersed in a empty space and running on a cold smooth-floor which speed is synchronized with the free-stream, with controlled level of turbulence managed by a mathematical model; and when run-time and cluster-size is not an issue, flow unsteadiness and thermal effects can also be turned-on to improve the realism. Finally, due to the specific requirements of the dev-program in CFD, the calculated configuration is never 100% equal to those from either tunnel or track: some differences are tolerated by experience, when experience does not become a high risk decision to take.

At the same time, the wind-tunnel experiment (using a hand-made scaled-down model replicating the reference configuration as perfectly as possible) would tell you about the evolution of the aerodynamic forces, internal flows and surface-pressure while changing the car’ “attitude”, i.e. varying front and rear rides, steer angle, roll and yaw angles; five variables properly combined during a tunnel run that can be the result of a sequence of fixed combinations or obtained by synchronized slow-motion movements of the above five parameters. But the experiment in the tunnel implies that the real physics is reversed (model stationary inside an “moving-air pipe”, hence in a confined space, where the air and the “ground”, this with limited dimensions…, runs at same speed, fixed to a specific value) and, normally, no thermal effects are involved. Also, the model is inevitably less accurate than the real car and not “perfectly rigid”, hence local deformations (not all of them under control, unfortunately!) are inevitably present. And what about “tires realism” and related limitations? Well, this is such of a fundamental aspect that a specific article was devoted on the matter by the writer, please refer to it. Last but not least in this environment, and for similar reasons as for CFD, the model configuration would normally differ “a bit” from those from both CFD and track. Again, the experience should (should!) indicate which level of difference can be accepted not to have results in conflicts with those from the other environments…

Finally we arrive at the track, where all is “real”. Real speed variation, real lateral and longitudinal accelerations, real wear of components (tires above all, here they again!!!) and speed-dependant part deformation, real details and split-lines between bodywork-components, real heat-exchanges phenomena to cool the internal components, real track surface and its status-changing including induced effect on the air-layer close to ground due to tarmac temperature, real variability of weather conditions (wind above all…), real driver ability/mistakes, real helmet motion, etc, etc… Last but not least, “real” is the limited capability of measuring both the car’ downforce (the contributions from the four wheel-groups are lost because directly bridged by the ground) and its aero-drag, the latter being indirectly derived by assuming engine-power and rolling-resistance of tires (stop talking about tires!!!).

All this is to say that, while correlating the factory-based models (CFD and/or tunnel) with track, a “common language” has to be established.  Otherwise the exercise would risk becoming a sort of guessing game and quite often a cause of painful confrontation…. In other words, similar conditions have to be ensured if at all possible, and the un-removable differences need to be recognized and known in terms of effects, for them to be taken into account very, very carefully.

This is why, for instance, teams are used to perform peculiar aero-test going up and down on a straight-line, at fixed speed and controlled ride-heights, with the aim to be “closer” to CFD and wind-tunnel to verify their claims. Probes, with different dimensions and aims, are quite often installed around the car to “interrogate” the surrounding  flow, this to extend the figures to look at while correlating (see the picture here above!) .  Due to the already-commented limitations imposed by the sporting rule, this peculiar test is sometimes also carried out during the GP weekend (track lay-out, general condition and “traffic” permitting). Problem is that this sort of “full-scale wind-tunnel” cannot simulate, by definition, the aerodynamics involved in braking and cornering conditions, which is even more important than the aero-performance during the acceleration-phase along the straight. In the majority of F1 tracks in fact, due to their modern lay-out, the lap-time primarily comes from these two key-maneuvers, for which the correlation is still not at the level achieved for speed&rides controlled straight-line testing.  No problem, the braver drivers can compensate by giving us useful feedback!

It’s a complex exercise, isn’t it? Yes, but not impossible to solve or, better, “control”. The secret is methodology, organization and innovation. Robust methods and procedures should ensure that the data-comparison between the environments involved is constant and well under control.  This should feed the process, thanks to knowing which methodology has a stronger correlation to which aerodynamic area of the car. Well-defined and well-followed procedures should help maintaining testing environments in order to avoid undesired effects on the data (issues that, unfortunately, if happen will not be immediately evident). Secondly, organization means that a specific team of people has to constantly deal with these aspects, devoting their work-hours to looking after the whole data-chain and the tools for post-processing, ensuring pro-active interactions with colleagues fully immersed in CFD, tunnel and track works. Finally, innovation means promoting any technology helping to improve the realism of both CFD and tunnel(s), to elevate the precision of all of the measures and the maths behind them, to maintain the testing quality at the highest level and to extend the domain of what is possible to measure at the track. These technologies need to be constantly considered, validated and adopted.

Correlation is not something that can be mysteriously lost or magically improved, as sometimes seems to emerge from the press. F1 engineers are well aware of the implications of the peculiar approach and the compressed-timing dictating by high-level motor racing, hence they are prepared to accept, despite no perfect reasons can always be offered, periods where the relationships with track-results is not as satisfactory as it was. Personally, I’ve learned more from poor-correlation periods than from the issues-free ones. It’s the perfect occasion to identify which elements of the global process are still “lacking”, i.e. where directing your resources, and/or if a specific aerodynamic concept is still not adequately supported by pre-track analysis: in other words, not fully-understood!

And please, do not be surprised if the biggest teams seem to suffer more by this issue than the minors. Two main reasons behind: firstly, they “are in news” more than the others; secondly, the main aspect, their mission is to push harder, delivering aero-updates with greater frequency and nature (very often “on the edge”) and that makes them more exposed to correlation “fluctuations”. 

But, believe me; if they are always able to start from the first rows of the grid, it is because the overall level of their correlation is, on average, much better than what it sometimes seems to be!