Pacejka 96 tyre model

[Silente]

Hi,

i know that probably this topic has been already discussed before, but i couldn't find where and, anyway, my question is more focused on alligning torque calculation rather than on cornering force.

I read that, somewhere in this forum, Mr. Kasprzak said that, in order to simulate a bad asfalt grip (like those we can find in parking lots, for example) on the pacejka 96 model calculation of Fy the right way is to set a value <1 for "lamda nuy".

This should also reproduce a little the fact that, with low grip surfaces, maximum force should be reached at lower slip angle than with surfaces with good grip.

My question is: what are the coefficients to work with (if there is any) to obtain the same result on alligning torque calculation (always with pacejka 96 model)?

Nuy is in the calculation of the B coeff, but there are also two coefficients (lambda t and lambda mr) that are not carried from Fy calculation. Should i work on them with the same magnitude change i did on "Lambda nuy" (if for example i put lambda nuy to 0.7 i have to do the same with them)?

Thanks

[Silente]

Hi,

i know that probably this topic has been already discussed before, but i couldn't find where and, anyway, my question is more focused on alligning torque calculation rather than on cornering force.

I read that, somewhere in this forum, Mr. Kasprzak said that, in order to simulate a bad asfalt grip (like those we can find in parking lots, for example) on the pacejka 96 model calculation of Fy the right way is to set a value &lt;1 for "lamda nuy".

This should also reproduce a little the fact that, with low grip surfaces, maximum force should be reached at lower slip angle than with surfaces with good grip.

My question is: what are the coefficients to work with (if there is any) to obtain the same result on alligning torque calculation (always with pacejka 96 model)?

Nuy is in the calculation of the B coeff, but there are also two coefficients (lambda t and lambda mr) that are not carried from Fy calculation. Should i work on them with the same magnitude change i did on "Lambda nuy" (if for example i put lambda nuy to 0.7 i have to do the same with them)?

Thanks

[BillCobb]

The lateral force scaling coefficient for friction surface variation in TNO code is LMUY.

This same variable is used in the evaluation of Mz via modification of Bt and Dr in the Mz formulation.

This means that adjustments to a Fy surface coefficient 'mu' are automatically factored into the Mz evaluation formula.

[Silente]

Thank you very much.

Since i saw in previous discussions you are inside Vehicle Dynamics simulations i would like to ask you another things connected to alligning torque.

I am trying to figure out the best weight bias for a given set of tyre from a pure cornering point of view (lateral acceleration maximization). Taking into account tires Mz for an example car which has infinite stiffness on suspension and is perfectly simetrical front to rear i found that the best performance is achieved with a rear biased configuration (how much wieght on the rear depends on the tire) and also that moving mass forward or rearward doesn't produce big changes in maximum performance.

This makes sense from your experience?

Thanks

[Charlie]

You need to quantify how much % mass movement you are experimenting with, and what you consider a 'big change' in maximum performance, for anyone to answer your question.

[Silente]

It depends on the tyres...

But to say a number: moving weight bias of steps of 1% from 40% to 45% and then to 50% at the front produce changes of less than 0.005 g for each step.

For example if at 47% i have 1.705 g than i have to go to 44 to have 1.71. At 40% i have 1.705 again...

This seems a little bit too less to me, but it's the first time i do this kind of analysis.

[exFSAE]

Not sure if static front percentage is what I'd be targeting for this analysis..

[Silente]

i wanted to use this kind of sim to have a target of where mass should be to use tires at their full potential.

But maybe there is something more interesting to look at. Do you have any suggestion?

[BillCobb]

Just because you have specified zero compliance in the suspensions and steering system does not mean that the net effect of Mz is also zero. The 4 aligning moments from the tires have a rigid body effect. This is also true for camber induced Mz.

Your results are not consistent with test or other sim results or theory. Check Units and signs !

Also, keep in mind that you are investigating a closed loop system. Chassis and mass settings based on open-loop steady state findings will probably be difficult to control (drive) in transient maneuvers. Or, you will be "slow" on the course.

The tendency for FSAE cars to be heavily rear weight biased causes its own set of tire specifications problems, not solved by pressure or compliance specifications.

[Silente]

Maybe i didn't explain corretcly what i mean.

I am considering the effects of the four Mz on car body. They have an understeering effetc on my model (with them the car has more understeer for a given lateral acceleration).

I am simulating increasing lateral g untill car and tyres cannot find equilibrium anymore or it has too much understeer or oversteer.

For now i am not investigating a course simulation or a lap around a circuit, only a turn of fixed radius negotiated by the car at a given Speed (lateral aceleration). Each time i increase a little bit centripetal acceleration and i wait to find an equilibrium solution till when this is not possible anymore.

I checked my signs and units and from what i read on Stackpole Eng. Paper on tyre pacejka models they should be correct. Not sure about MZ sign, but since it is giving an understeering tendency i guess it is correct.

Total MZ (sum of each wheel MZ) for car body is in the order of magnitude of 20-30 Nm at the limit.

Am i doing something wrong?

Maybe my mistake could be tha reason why i don't find big differences spitting weight more to the front or to the rear.

[Zac]

Double check your assumptions. How are you handling lateral load transfer?

[Silente]

I followed milliken equations and then i put a very high stiffness on elastic members just to start with something similar to a rigid suspension car.

I am not considering things like caster effects for now.

I am quite sure about load trasnfer.

I am not sure about the handling model. I already discussed here about this in previous topics.

And to me it's strange that i don't see nearly no difference in maximum lateral acceleration changing so much weight distribution...

[exFSAE]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Silente:
i wanted to use this kind of sim to have a target of where mass should be to use tires at their full potential.

But maybe there is something more interesting to look at. Do you have any suggestion? </div></BLOCKQUOTE>

Well my first question would be... how are you handling TLLTD? Don't see how looking at static load split by itself really says much without varying TLLTD as well (2 variable sweep).

If you're using the same tire on all corners of the car... 50% nose weight and 50% nose TLLTD is going to give you the most raw cornering grip.

[BillCobb]

Make sure the signs and units of your wheel loads are consistent with the signs and units of your tire model. From what you have been describing, I could believe your tire model is expecting different units or sign than you are delivering. On the FSAE/TTC forum, I've posted a graphic showing the 'trajectories' of 4 tires from a hypothetical FSAE car under an increasing cornering situation projected onto a 'carpet plot'. That should be done by you to acertain that the tire model is living in the tire data world. I posted a simple handling model on there too, and it could be used to compare your results with mine.

[Silente]

I checked it again. Both signs and unit seem to be correct.

at the beginning of my work i also checked that calculated Fy were similar to raw Fy measuared during tyre testing.

My sintax is not so efficient, so maybe i have something wrong on programming. I suspect i didn't do correctly the growing of lateral acceleration to find the maximum. At a fixed lateral g the sim seems to work fine and the results are quite consistent with theory.

Can you give me any suggestion on this? Now I am doing it creating a vector of lateral accelerations (from 1.5 to 1.75) and then looking for front and rear slip angles to find force and moments equilibrium for each value of ay.
The sim stops when force or moments produce by the four tires cannot create equilibrium anymore.

Moreover i have no access to TTC forum, since tyre data were taken by another guy of our school and i don't know the details to login.

[thewoundedsoldier]

You can register on the forums with a university email address. If the contact is lost, you should just email them to help you get it straightened out.

[BillCobb]

Is your proram developed in Matlab and/or do you have access to a recent version of Matlab? If so, it may be possible for you to run my 'demo' model for comparison purposes. I'll have to fire up a computer which has just been upgraded to Windows 7. I'm not sure yet whether Matlab is happy with that. I know I am, though...

[ben]

Slightly off topic - but having compared a number of our tyres using Pacjeka 96 and 2002 (model fitted in OptimumT) the way Pac96 fits the Mz peaks gives completely the wrong trend vs. load, 2002 is genrally a lot better in my experience.

Therefore I would consider fitting to 2002 before using a sim to answer a question at that level of detail.

Ben

[Zac]

I'd actually recommend that you go the other way and fit using pac94 if you want a better read on aligning moment. pac94 is still an empirical model. ever since pac96/mf5.0, TNO has been strapping more of an analytical tire model to that empirical base. I'm not too keen on the results.

[Silente]

Bill,

I think i know understad what you mean.

Does anybody can direct me on any paper or document where correct signs and units for pacejka 96 model are explained?

Till now i am using signs and conventions shown in SES papers, but i have seen somebody using negative FZ and Slip Angles in Degrees.

[Silente]

Back on the topic.

i probably fixed the problem.

To be precise, the only problem i found was that, with the suspensions configuration i modeled, there was too little TLLTD change when moving mass to the front or to the rear.

I now tried with a different modelling of suspensions (really simulating a no rolling car) and the results are changing more with mass distribution movements.

Does this make sense to you guys?

I also would like to ask which kind of analysis you normally do with steady state simulations.

Last question:
Is it worth to include in this kind of sim a separation between longitudinal and lateral forces from tyres? for what i now Fy data from TTC are measured in a direction perpendicular to the rim, not to the tire. So, when the tire is twisted (Slip angle) the measured Fy should be splitted in "Fy*cos(SA)" (Real Fy) and "Fy*sen(SA)" (Drag components). Is this correct in your opinion?

Thanks

[RollingCamel]

After the Claude Rouelle seminar, i'm having a look at tyre modeling. We still didn't buy the TTC data i had a look at what Avon provide which is pure lateral model only. In Lotus Suspension the coefficient designation is different as it uses 'A' coefficients.

Anyone knows the equivalent of these coefficients to 'pCy1,pDy1,..etc' type?

[exFSAE]

They're the same.

[RollingCamel]

I understand that they are the same but i don't know if the A0 is for the shape factor or maximum cornering stiffness for example.

[exFSAE]

When they're listed, they typically just match up in order.

a0 = Shape factor C
a1 = pDy1

etc

[RollingCamel]

Okay then. Thanks for the help.