Using tire data in simulation programs

[J.Pedro]

Hi guys,

I’m using simulation programs to predict the behaviour of the car on track. As you know, the tire data is a great deal in the feasibility of these programs, as you must already know, crap in crap out!

Basically my problem is this: I have two different tire‘s data: one from AVON (the tire manufacturer) and the other from TTC. The tire data from Avon looks more realistic because it doesn’t have max. lateral accel. of 2g’s like TTC has. Although, I don’t know in which tread the tire is rolling neither what is the machine they used. From TTC, as you know, the data needs correction because the tire is rolling in a high friction surface, and that’s why the lateral accel. that the tire can achieve are so high. Furthermore, in avon the tire is rolling at 20.2 km/h and in TTC at 38km/h.

First of all, I don’t know if it’s correct to compare tire data from different sources? In theory, if the tire is the same, the data should be the same. But as the machines and the treads probably are different, I don’t know if this is true.

I can apply the tire data from TTC, but I have to somehow make that data more “realistic”. I have figure out two options:

1) I multiply the friction coefficient by factor inferior to one, to try simulate real conditions on track, instead of the high friction surface they run tests on. This way I don’t change the cornering stiffness.
2) Multiply the FY by a factor inferior to one, for the same reasons I stated before. This way I change the cornering stiffness. Although, this way the TTC data fits almost “perfectly “ the data from AVON.

From what I stated previously:

Which tire data should I use?

If you’re answer is the TTC. Which option should I use to try to simulate track conditions? If none, please advice in any way you can.

I look forward to hearing from you soon.

Best Regards,

João Pedro
Projecto FST

[J.Pedro]

Hi guys,

I’m using simulation programs to predict the behaviour of the car on track. As you know, the tire data is a great deal in the feasibility of these programs, as you must already know, crap in crap out!

Basically my problem is this: I have two different tire‘s data: one from AVON (the tire manufacturer) and the other from TTC. The tire data from Avon looks more realistic because it doesn’t have max. lateral accel. of 2g’s like TTC has. Although, I don’t know in which tread the tire is rolling neither what is the machine they used. From TTC, as you know, the data needs correction because the tire is rolling in a high friction surface, and that’s why the lateral accel. that the tire can achieve are so high. Furthermore, in avon the tire is rolling at 20.2 km/h and in TTC at 38km/h.

First of all, I don’t know if it’s correct to compare tire data from different sources? In theory, if the tire is the same, the data should be the same. But as the machines and the treads probably are different, I don’t know if this is true.

I can apply the tire data from TTC, but I have to somehow make that data more “realistic”. I have figure out two options:

1) I multiply the friction coefficient by factor inferior to one, to try simulate real conditions on track, instead of the high friction surface they run tests on. This way I don’t change the cornering stiffness.
2) Multiply the FY by a factor inferior to one, for the same reasons I stated before. This way I change the cornering stiffness. Although, this way the TTC data fits almost “perfectly “ the data from AVON.

From what I stated previously:

Which tire data should I use?

If you’re answer is the TTC. Which option should I use to try to simulate track conditions? If none, please advice in any way you can.

I look forward to hearing from you soon.

Best Regards,

João Pedro
Projecto FST

[Dash]

I'd use the TTC data. That way you can run simulation on two or more tracks. You could start by multiplying FY to make it match the AVON data. Then tinker with it to simulate a track that matches the max lateral acceleration you measured on your car. That would probably give very good and realistic results.

[ed_pratt]

João,

If you have a look around I think you'll find that it is suggested that the friction coefficient in the TTC data is multiplied by 0.6 to more accurately simulate the "real" track surface.

Ed

[exFSAE]

Other thing to consider is that there are other good ways of using tue data which do not entail track simulation

[Pennyman]

It's my understanding that the TTC data includes all the equations needed to put together the Pacejka 96 tire model. There's a scaling factor that is used to change the coefficient of friction to something more realistic. I think you'll find that the curve changes differently than simply multiplying by a scaling factor.

Sounds like a fun project!

[BillCobb]

What is the basis for the 2g claim in your question? Simulation's I've done (as posted on the TTC forum) show about a 1.5 g max lat steady state achievment with most tires. Please don't say the normalized TY data shows values over 2. A car uses more than 1 wheel, so it can't optimize all tires to these values. I've also posted a machine correction formulation function there so that when you get out on a real track, you can tune in the vehicle and tire model. How many surfaces did you get from Avon? FY, MZ, and MX should be your minimum daily requirement. What test machinery did Avon use to produce your data? Belt, inside drum, outside drum, TIRF, MTS, honed surface, steel, 3Mite, Safety Walk, ??? Have they ever run the SAE standard test tire to compare with other industry tire test equipment? Get the facts, get the answers.

[Edward M. Kasprzak]

To compare the two I suggest learning more about the data from Avon first. Are they the same tire (construction/compound) used by the TTC, and where was it tested? Test procedure? Then compare with the TTC. We (TTC directors) answer as many questions as best we can on the secure forum: h t t p : / / sae.wsu.edu/ttc/

As you noted, scaling the whole FY channel changes the cornering stiffness--this is usually undesireable as cornering stiffness is affected much less by changes in surface friction than the height of the peak.

I disagree when you say "if the tire is the same, the data should be the same". Tire data is strongly influenced by the test procedure and test machine. Nothing is easy when it comes to tire testing.

Translating laboratory tire data to on-track tire performance is not a fully solved problem, even at the top levels of motorsports. I suggest using your first option (friction coefficient reduction) and then stop messing with the tire data/model unless your on-track testing strongly suggests otherwise. It's pretty easy to improve simulation correlation by changing the tire model, but this can mask important vehicle model elements which are missing or incorrect.

[Charlie]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Edward M. Kasprzak:
I suggest using your first option (friction coefficient reduction) and then stop messing with the tire data/model unless your on-track testing strongly suggests otherwise. </div></BLOCKQUOTE>

I agree, change track surface only until you have exhausted all other options.

[Zac]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">I disagree when you say "if the tire is the same, the data should be the same". Tire data is strongly influenced by the test procedure and test machine. Nothing is easy when it comes to tire testing. </div></BLOCKQUOTE>

Agreed. Even running the same procedure on the same tire (or as close to the same tire as you can realistically get) you can see big differences between different machines.

[Yunlong Xu]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by ed_pratt:
João,

If you have a look around I think you'll find that it is suggested that the friction coefficient in the TTC data is multiplied by 0.6 to more accurately simulate the "real" track surface.

Ed </div></BLOCKQUOTE>

Hi, I was looking for the correct scaling factor at about the same time last year. The impression that I've got was about 0.7, but it wasn't very conclusive. Could you let me know from where you found this 0.6 correciton factor?

cheers

[exFSAE]

They're both arbitrary dude. Hell, you might have to scale grip to 62% in the morning and 71% in the afternoon... or 67% on one autocross run and 69% the second. It's a moving target.

[js10coastr]

WHAT?!?!?! The theory doesn't match up with the practical!?!? Well... I never would have thought!

[ed_pratt]

Sorry, I can't remember where exactly I saw it - either in the TTC stuff or in these forums somewhere. But as exFSAE said, it doesn't really matter as they're both arbitrary.

If this is your first design and you have no data from a previous car then pick some value and use your simulations to help guide the design a little rather than pin down an absolute laptime - which you'll never get anyway.
If you're lucky enough to have some data from a previous years use of those tyres then you can apply a factor which equates the peak friction coefficients.

ed

[Big Bird]

If you want to ground your estimations in some real data, take a look at skid pad data from various comps (for non-aero cars of course). You know radius, you know time, you can calc average lateral acceleration in g's. Direct conversion to average coefficient of friction. That is a good start, (or at least better than throwing your arms in the air in despair when you don't get the perfect answer straight away).

Notes:
* Make sure you pick the good cars, who you would believe to be operating at somewhere near full potential. Say, top 3 over a number of comps.
* Pick the non-aero cars, otherwise the calcs don't work out.
* If you know the surface, (e.g. concrete, asphalt), then that will help your interpretation too.

The TTC data helps you with trends. Does the coefficient of friction drop off with increasing vertical load? If so, at what sort of rate (in say, % per kg weight of vertical load). Does the tyre favor Ackermann or anti Ackermann? That sort of stuff is the best starting point, not trying to optimize the perfect laptime straight up.

Cheers,

[ben]

I'd like to back up Edward's comments regarding test procedure. The same tyre on different test rigs will look very different. In fact the same tyre on the same rig with two different test procedures will look like two different tyres.

Fundamentally it will be impossible for you to objectively compare your Avon data with the TTC data.

The TTC data will allow you to compare the tyres run, because they were on the same rig with the same procedure. This is why it's such a fantastic project.

Ben