History of the MFTire / Pac02 tire model implementation

[Thorsten Wittenator]

Hi all,

I am currently trying to simulate a racing car in Project Chrono. As such the tire simulation is very important to me. I fitted a magic formula v61 model to my experimental data and while browsing through the pull requests I found that there is/was a MFTire class for the versions 61 and 62 of the magic formula in Chrono. Since I didn't find this code in the current main branch I did a little deep dive and found out that the corresponding class was deleted in 4f7c0ea8f6b10896459c1867551858706f083632 on the 8th of June 2023 in the process of reworking the Pacejka2002 tire. I was a little confused by this since I thought that the MF version 61 and 62 are way newer than 2002 (imo 2008 atleast). Therefore I wanted to ask whether the Pac2002 tire contain all the formulas for the current MF versions or what the process was behind removing this class.

Sincerely,

Thorsten

[dr.ratz...@gmail.com]

Hi Thorsten,

the MFTire class was a c++ port of the Matlab tool. It contained a lot of features, nearly all proposed bei Hans Pacejka in his book. The original authors know, that many of these features or not used by anybody. They made analyses of the commercial MFTire solutions used e. g. in ADAMS for this and privided a switch to deactivate those features. We decided not to maintain code that nobody ever is going to use, that is the main reason to drop c++-MFTire as tire model in chrono. The c++ port was extremely buggy, that showed a test a test with a validated vehicle model (Fed-Alpha) equipped with validated Pacejka 2002 tire files (donated by Goodyear). Btw. our former Pac02 model worked well with it! We decided to rewrite it yet, to give it a more readable structure and to implement a parser for ADAMS/Car compatible *.tir files. It should be easier to use now.

We definitely don't have all possible functions implemented. Only steady state and combined forces can be used. Inflation pressure dependent data sets can be probably used but have not been tested, because we don't have validated example files for it. So if you have fitted tire parameters for your race car, you should Pac02 in chrono give a try.

Best,

Rainer

[Thorsten Wittenator]

Hi Rainer,

thank you very much for the explanation! I'll try my luck with the Pac2002 model. If I get around to validating the car and tire models after the season, I may come back here and report on our results.

Sincerely,

Thorsten

[Trevor Vidano]

Hi Rainer,

What happened to the transient extension of the Pacejka steady-state equations detailed in TR-2011-02? Is this what you are referring to as the features that no one uses? 

Also, is this the MATLAB tool you are referring to: 111375-magic-formula-tyre-tool. In vehicle modeling, tires are everything. If there are tools that are common to use to fit tire models to data, it would be very nice to have those same tire models supported in Chrono.

Relaxation length is one of the most used tire properties in tire modeling (same with pneumatic trail). The Fiala tire model supports this. It looks like the ChPacejkaTire supports this in Release/8.0.0, but it has since been removed in main. Also, CarSim supports Pacejka 5.2 and MFSwift tire models. This would indicate these models are standards and seem far more conventional than some of the other tire models supported by Chrono. It would be nice to have them supported in Chrono too.

[dr.ratz...@gmail.com]

Hi Thorsten,

TR-2011-02 is nice, but its implementation 'ChPacejkaTire' did not work for more than a few seconds before it crashed, so it has been removed and replaced by ChPac02Tire. TR-2011-02 describes very well the features most people would use if the had validated parameter sets for the features they want.

The MATLAB tool I mentioned is MFEVAL. It is okay, I also use it to test the parameters and to convert Pacejka 2002 into TMeasy or TMsimple. Other tools maybe okay too.

ChFialaTire is kept for historical reasons. It is not really useful, it shows what was known in the 1950s. The relaxation algorithm is simple and inaccurate. When used for Steady State Cornering it ends up with unrealistic results. It still has some fans yet.

It is important to mention, that we only want to support features which we can validate. For validation for tire algorithms we need a validated vehicle model and specific test results. We only have one validated vehicle model (FED-Alpha), its Pacejka 2002 tire files don't contain any of the needed parameters. There are no test results to identify relaxation parameters!

So, if you have a validated vehicle for chrono with a validated tire parameter set incl. relaxation parameters, donate it to Project::chrono, we will be pleased.

[Trevor Vidano]

Hi Rainer,

Thanks for the information on TR-2011-02. It is a shame that the implementation didn't work. I wonder how the results were generated... 

There are some validated vehicle models and tire models published in literature by the National Advanced Driving Simulator (NADS). I am currently working on a project to port one or more of those vehicles into chrono and create a user-friendly interface with Simulink (currently using TCP/IP cosimulation). One of the challenges is the tire model.

One publication ,Paper 2007-01-0816, presents MATLAB code and parameters for three different tires. Although this doesn't appear to be the tire model used by NADS. Instead, it appears they use the tire model developed in SAE Paper 970559. The tire equations are summarized in the appendix of SAE Paper 981190 (which also presents validated tire model parameters).

More tire parameters for the tire model developed in SAE Paper 981190 can be found in SAE Paper 2007-01-0818 and SAE Paper 1999-01-0121.

Thanks,

Trevor

[dr.ratz...@gmail.com]

Hi Trevor,

chrono is always under development and parts oft can get out of focus, like ChPacejka. Due to its complexity it is better to rewrite it than to refiddle with problems. Tire relaxation is not considered yet. Good to know that there is a direct need for this feature in chrono.

I will look at the papers you proposed. Tell us, if you see any weirdnesses with our actual tire model implentations.

Thanks!

Rainer