Nissan 370Z Forum - View Single Post

DLSTR · 10-31-2021, 02:18 PM

The secrets of preparing for a new F1 circuit
Autosport Plus
Formula 1 Special feature
The secrets of preparing for a new F1 circuit
By:
GP Racing
Oct 29, 2021, 7:07 AM
No Formula 1 team arrives at a ‘new’ circuit entirely unfamiliar with it. As PAT SYMONDS explains, simulation and artificial intelligence does the hard work even before the driver takes their turn

The 2020 and 2021 seasons will be remembered for many things, including the significant disruption caused by a calendar that often shifted in response to an ever-changing pandemic. While nothing good came out of the global crisis, F1 fans were at least able to enjoy their sport and see some variety compared with the formulaic calendars of the previous few years.

Imola and Istanbul were reintroduced in 2020, while F1 raced at Mugello and Portimao for the first time. This season, Jeddah in Saudi Arabia and Losail in Qatar have been added, as well as returning again to Imola, Zandvoort, Istanbul and Portimao to help achieve a 22-race championship.

One might argue that Imola and Istanbul are not new to F1, and cars have tested at Mugello and Portimao in recent years, but while historic data provides an indication of what may be expected of a track, cars and the tyres have changed so much in the intervening period that much of the data is comparative and anecdotal, not quantitative.

So how does a team go about preparing for a new circuits? The answer, of course, lies in simulation – but the initial work will not be done on a simulator, it will be done on a computer with a ‘virtual’ driver so that set-ups and driving lines can be optimised without the initial distraction of subjectivity.

In order to start this simulation, the standard car model will be used, generally with a known set-up that might have been used at another circuit with similar corner speeds and lengths of straights. This will provide a good starting point for the iterations that are needed to get close to an ideal set-up. The car model itself will be extremely sophisticated. Such is the computing power available these days, that is no problem. The first lap time simulation I used was in 1986 and, although it only optimised about five different aero levels and the gear ratios, it took all night to run. Today the simulation will analyse a lap in much less time than it takes to drive it.

This first simulation sweep will still concentrate on wing levels to get the right compromise between downforce and drag, but will now be able to adjust all the other settings on the car – such as rideheights and spring stiffnesses – to obtain an optimum. Unlike our early simulations, which always drove the car over a racing line that the engineer would determine by eye, these days the minimum lap time algorithms will seek the ideal racing line for each different set-up to ensure maximum performance.

Of course a good car model and an ‘artificial intelligence’ driver are of no use if we can’t accurately describe the track itself, and in these early runs we may not have particularly sophisticated track information. For a new circuit, the first information the teams will get is a 2D architectural map which they then have to digitise, converting the track limits to X-Y coordinates over the entire area. This early map may or may not have elevation information as well – often this comes in a later version. As the map will generally be based on what the architect intends for the circuit, long before it is actually built, it certainly won’t have all the details of the kerbs so the simulation will generally assume the car stays entirely within track limits.

The first stage of the investigation will generally look at a multi-factor optimisation. For this, the engineer will set certain bounds of a number of parameters that can be altered. For example, they may set the front rideheight to be investigated between 15mm and 20mm, the front roll stiffness between 1 and 1.2 Newton meters per radian, and the downforce to be in the entire range the wings designed for the car can obtain.

It would not be unusual to allow nine or 10 set-up parameters to be investigated this way. The simulation then automatically runs many combinations of the variables and presents the results in a specific type of diagram (below), where each of the input parameters is displayed in multiple axes alongside the output parameters such as lap time, maximum speed and end of straight rideheights. At first sight this might look like a spider's web, but a little inspection shows the trade-offs between lap time and end of straight speed, to name but two.

From this, a basic set-up is adopted and it’s time to move to the full simulator. This is sometimes known as the ‘driver-in-the-loop’ simulator, as the inputs and line seeking algorithms of the first simulations are replaced by a driver using visual and vestibular feedback to drive the car. At this stage a lot more detail is needed, and lidar scans of the circuit are used to give photo-realistic scenery and track markings. This adds a lot to the computing power needed, as does simulating the engine and transmission dynamic responses – which will be done by a control unit identical to that used in the real car.

The driver now works with the engineer to hone the set-up to his or her liking. Lap time is the ultimate goal, but using the same data analysis tools as they would use trackside, the engineering team will also examine factors such as stability, tyre energies and even, if the scenery detail is good enough, practice the pit entry for faster pitstops.

Simulation is a double-edged sword. Teams love it as they strive for perfection in a controlled environment, but that reduces jeopardy, which in turn reduces the appeal for spectators. However, the genie is out of the bottle. Simulations will only become more sophisticated in the years to come.