Identification Pacejka tyre model

[pietro82]

Hi all,

I need to derive the Pacejka data of a tyre installed on a commercial vehicle to perform a multibody simulation of the vehicle. I performed the ISO steady state test. The commercial vehicle doesn't have any rear suspension and also it's top speed is of 60 km/h. The steady state behaviour is almost linear in the whole range. The maximum lateral acceleration is 0.4g. I try to derive Pacejka's coefficient through the bycicle model. Is it a good idea?

thanks

Pietro

 

[GregLocock]

It's a great idea, in some respects as it will give you a usable tire model, in the linear range, and a better idea of how Pac models work.

However, you probably need to set all the non linear scaling functions to sensible values, as the data you have won't help much with them.

If your only concern is linear range then a simpler tire model might be easier to work with.

As an example of the problems you face, you probably don't have a good idea of load sensitivity or camber effects from your test.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pietro82]

Hi Greg,

thanks for your reply. I performed the steady state test in different loading condtions and trends are linear only for the unladen vehicle. For the full loaded one, front tyres are almost saturated. Yesterday I tried and the fit is pretty good in all conditions. I will try with the full vehicle model and check the differences in both methods and try to identificate tyres parameters. Moreover if the parameteres are normalized to the vertical load, should I get the same parameters for each wheel in different vehicle conditions, right?

Thanks

Pietro

 

[GregLocock]

No, because the Pac model includes a load sensitivity effect - although I'd argue it oversimplifies the load sensitivity, at least with Pac 94.

The normal way of fitting a Pac model is to use ALL your data in one almighty matrix, not to try and fit at just one load and adjust it.

What procedure or program are you using to fit the data?


Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pietro82]

Hi Greg,

thanks for your reply. I'm sorry for the silly question, I know it, I wanted to edit the post immedialy but it wasn't possible. I used matlab curve fitting toolbox. I fitted data for each loading condition due to the load sensitivity effect, but I normalized each coefficient with the tractor static load. Is it the right way?

Thanks

Pietro

 

[GregLocock]

Doesn't that give you a rather large number of load sensitive coefficients? As I said, I do it by fitting every datapoint to the Pac model in one go, rather than splitting them up. I have the great advantage that i am working from test rig data so I have regularly spaced datapoints in the (Fz, alpha, gamma) domain.

So, roughly, I have a spreadsheet of Fz alpha gamma Fy Mx Mz PacFy PacMx PacMz where the pac columns are calculated from my current Pac coefficients. (I don't actually do this, but conceptually this is the approach).



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pietro82]

Hi Greg,

thanks for your reply. Why isn't it better to fit the data for each loading condition separately? Since they are load sensitive, I thought it would be better. I'm sorry if I miss something.

Cheers

Michele

 

[GregLocock]

The Pac model aready accounts for load sensitivity (although not in a particularly accurate way) so you are introducing more noise into the process. bear in mind that as you manoeuvre the wheel load varies instantaneously, so you won't just be sitting on one of your measured fz curves.

It would be interesting to do it both ways and compare the results.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[pietro82]

Thanks for your reply and for the interesting tip to compare the result. Maybe in the next week I will have some time and I'll do it. I'll show the result.

Thanks

Cheers

Pietro