9 replies to this topic

[dancin stu]

I know a few people on this board have written their own lap time simulations, so am seeking advise on the subject of tyre models, and how they implement them.

At present I am using one based on the magic formula model as in Chapter 14 of Milliken with a couple of other bits and pieces thrown in, and so far it seems to correlate ok using Avon's data on their website.

After several different experiments, it seems that the handling model will be based using a constrained vehicle model (MMM) using a non linear model I already have.

[Greg Locock]

I'm idly thinking about writing a lap time simulator as well. I see no compelling reason to go to a Pacejka model, was more inclined to use the PHOR one, and fit that to the Avon tyre data. The biggest issue with Pacejka in this context is combined slip, isn't it?

What's MMM? I was just going to use a rigid body with 6 dof, and then a 2dof model for each front wheel (steer and bump), and bump only for the rears. Is that too crude?

My intention is to hook it up to an evolutionary setup generator (for which I have just finished writing the proof of concept code) as discussed in another thread.

[Ben]

Pacejka's book has a non-dimensional model with longitudinal force as an input which may be simpler. The book is very good for understanding tyre modelling in general as well.

Ben

[remmosffej]

I came across an article along these lines recently. In the book:

Automatic differentiation of algorithms : from simulation to optimazation / George Corliss ... [et al.], editors

LOC (US library of Congress) call#: QA304.A78 2002

On page 117 there's an article called "Application of Automatic Differentiation to Race Car Performance Optimisation" (one of the named authors is Pat Symonds). It doesn't go into detail about car models, but it talks about a particular method, given a model, to optimize lap time. One of the references listed is "Application of optimal control theory to inverse simulation of car handling" in the journal "Vehicle System Dynamics" issue 26(6) pp449-462, from 1996.

[desmo]

Here's a paper coauthored by Symonds that kind of approaches lap simulation from the other direction if I remember correctly.

http://www.ee.ic.ac..../circ_recon.pdf

[Paolo]

Great paper, Desmo, but it seems many pics were not saved. Any hope of getting the missing bits ?

[Lukin]

I second that. Cheers.

[dancin stu]

Originally posted by Greg Locock
I see no compelling reason to go to a Pacejka model, was more inclined to use the PHOR one, and fit that to the Avon tyre data. The biggest issue with Pacejka in this context is combined slip, isn't it?

Is Phor the Physics of Racing one? Havnt looked into that much, have come across it briefly. Not sure about the combined slip, I'm currently having issues with MUy and MUx which I believe to be road mu. However, I have found by curve fitting that this seems to change as a function of vertical load, certainly gives more accurate results if it does. Have so far managed to get a pure slip model to correlate very well with the AVON Data, although it still takes a lot of pre processing of the actual data.

What's MMM? I was just going to use a rigid body with 6 dof, and then a 2dof model for each front wheel (steer and bump), and bump only for the rears. Is that too crude?

MMM is Milliken Moment Method, solving every tyre and body slip and steer angle. THe way I have thought to do it is to create a program that generates a plot for a certain vehicle configuration, from which selective post processing will occur. I already have a fully non linear weight transfer model, so it incorprates track change, rc migration etc, so can get accurate normal loadings already.

Originally posted by Ben
Pacejka's book has a non-dimensional model with longitudinal force as an input which may be simpler. The book is very good for understanding tyre modelling in general as well.

Ben

Yes it is a very good book, had it all year from our library, advantage of being the only person on our course interested in vehcile dynamics............. but have had to give it back cos schools out. Tried a couple of the models out of there, not the longitudinal one though, but couldnt get them to work very well, a legacy of not having enough time to go through it in more detail.

Ultimately I think the best bet would be to fork out for a copy of tyregene, tried the demo and its very good, just no money=no toys

[Ben]

Yeah, but Tyregene's only a GA solver to fit for the coefficients you still have to enter the tyre model and get a vehicle dynamics model that works with it. Tyregene won't help with that.

Ben

[Greg Locock]

PHOR=physics of racing. In part 24 he comes up with an empirical model (well, for that matter Pacejka is empirical) that fits the data reasonably well and is easy to calculate. Since I'm going to have a plug-in tyre model I'll run with the easiest one at first.

the effect of vertical load on friction seems to be roughly

max(Fy)=Kyz*(Fz^0.7)

for a wide variety of tyres.