rec.autos.simulators

while reading Genta ( for the purpose of development of a vehicle sim)
I got a bit confused

let us consider Pacejka magic formula for longitudal force ( Genta p
64)

the maximum value of longitudal force might be equal to D

D = mu p *Fz

mu p =b1*Fz+b2

Fz - load and b1 and b2 are respective coeffs.

at pp 526  for large front wheel drive saloon  car the following
coeefs are given
b1=-7.6118 b2=1122.6

we take at front wheels Fz =3381 H ( or 3.381 kN for Pacejka)

so we have

D = 3708.49

Now we have that road adhesion might not exceed   (road friction
coeff) * Fz

even if I get road friction  ==1
anyway  seems that Pacejka max force is greater than road adhesion....

Does it mean I should limit output  force if it exceeds adhesion or
that
the force is actually physically OK?

the another question might be is there straight way to couple friction
coefficient with Pacejka coeffs?

or for each type of road/tire  there is a need in separate Pacejka
coeffs set?

Regards
Serge
http://www.racesimcentral.net/

What year is that book from anyway? Seems the Pacejka formula in that
book is from about '91. There was a major update in '96, with I think
combined long/lat forces.
Also, probably at that time as well, a lot of the coefficients changed
name.
Anyway, I should read Pacejka's book. ;)

...

I'm not sure about the difference between road adhesion and friction.
What b2 is doing is merely: mu=b2/1000+b1*Fz/1000, where Fz is in kN
to make things fuzzy. :)
So b2 is road friction coefficient, apart from the load sensitivity
(b1, which is the b2 change per kN if I write that up quickly, so
basically mu change per million Newton (that's an awkward way to say
it)).

No, the Pacejka result is quite ok as the road reaction force.
Whereever it derives from (friction) doesn't really matter; at the
given input variables (slip angles, slip ratio, camber) is gives you
the resulting force that the road puts back onto the wheel.

So there is no adhesion or friction difference here.

D is the peak force, and is just about mu (D/Fz = mu). The current
effective mu is just Fx/Fz longitudinally or Fy/Fz laterally.

Yes, unfortunately.
Still, the Pacejka '96 variant has lamba coefficients, which enable
you to tweak more physical aspects of the formula, instead of more
mysterious and subtle coefficients. Some lamba's affect more than one
Pacejka coefficient, that is.

Those lamba's might be useful in simulating multiple road variants
using a bit of lamba tweaking. Once you get on loose soil though,
there are completely different theories. Although I like to think you
can sum effects a bit.

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

There is a nice picture in Pacejka "Tire and Vehicle Dynamics" (Figure
2.11) that attempts to place all the different types of tire models into
one overall framework, and rank them on different aspects.  On the left
are pure curvefitting models like the Magic Formula scheme, next are what
he calls "similarity method" which include some simple tire theory in
addition to measurements on real tires.  This is where our Radt-MRA
Non-Dimensional approach fits in, see RCVD Chapter 14 and also a chapter
in Pacejka.  In this case, at least some of the numbers that describe the
tire have some physical significance, for example the cornering stiffness
and peak friction values (as functions of various operating variables).

I know that some sims use the Non-Dimensional scheme but can't tell you
off hand which ones (just that various people have contacted me over the
years and mentioned that they were using it).

We have a brief discussion here:
  http://www.millikenresearch.com/mrandim.html

For unknown reasons, Pacejka doesn't reference the early paper by Radt
and Milliken (my father), Motions of Skidding Automobiles, which also
discusses this approach.

-- Doug

No Pacejka here.  Just a work-in-progress Wasson model :-P

I actually did try using springs in the contact patch a long time ago as people
have talked about here often.  Using about 20-50 springs per contact patch
actually worked surprisingly well.  Sort of a poor-man's FEM tire model.  

The new one's better though ;-)

Todd Wasson
Racing Software
http://PerformanceSimulations.com
My car sim
http://performancesimulations.com/scnshot4.htm

Couple of things here; the fact that it is used professionally (but
I'm sure other models are). Second because I know people here that
deal with Pacejka a lot (even are his students), and they have fitting
programs and all that, so it's not really a dead end, but very
interesting in fact.

Where can I download the research papers? ;-)

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

I believe the main problem with the Pacejka formula is that in trying to
present a comprehensive tyre model it obscures the fact that the formula
itself (once all the constants have been evaluated for a given load)
probably represents the simplest mathematical expression having the
property of being everywhere analytical and at the same time capturing the
typical tyre  features (linear at small slip, reaching a peak and then
leveling off). Things then get *** once you try to describe the full
load dependent picture by introducing further constants, as some of the
constants do have a direct physical meaning while for others it's very
hard to distinguish between physics and simple fitting value. For a
*** racing simulation it then becomes hard to incorporate effects like
tyre wear, pressure dependence etc. in a meaningful way by altering the
coefficients directly.

What I'd suggest would be to use the 'bare' Pacejka formula (which
certainly has its merits) and link the few remaing parameters to physical
quantities like maximum grip, sliding force, stiffness, peak slip etc.,
all of which may change with load, wear, etc., which  would then bring it
closer to the nondimensinal scheme (but not quite).

-Gregor

Here we go.. I'm one. :-D

But he does reference you Doug.. he calls the RCVD tyre model the
"similarity model" and there's a reference to you in the first
paragraph.
Anyway... I wonder.. one thing I had to "tweak" myself in the RCVD
model is the stiffness value.. it is supposed to be function of Fz but
I couldnt find a way to express this this in RCVD.. have I missed
something?

Stefano Casillo
driving.multiplayer.it/netkar

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

Good to see you joining in. :)

On Mon, 31 Mar 2003 11:28:12 +0200, Gregor Veble

....

The later Pacejka formulae (1996+) allow this with lamba values, that
all have physical meaning. This is how I know some engineers tweak
tire data.
Each lamba can affect the formulae at more than one spot, leaving the
base Pacejka coefficients (that Racer uses for example) as more of a
fitting result. From there on, you only tweak the lamba values.

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

Only a minor point --Pacejka missed that one early paper, but as you say
he does reference two other papers.

Our ref 124:
 Radt, H. S. and D. A. Glemming.  "Normalization of Tire Force and Moment
 Data."  Tire Science and Technology, TSTCA, Vol. 21, No. 2, pp. 91-119.
 April-June, 1993.

is another place to look.  Not sure how to order a copy of this paper.

The description in RCVD is very brief (book is too long anyway...)

-- Doug Milliken

Can't comment because I don't have the book.. :-(

Todd Wasson
Racing Software
http://PerformanceSimulations.com
My car sim
http://performancesimulations.com/scnshot4.htm

Simple Physical Model might be, for example, a "brush" model of the tire
with stiffness to the "bristles".  There are a number of these types of
models in the literature with advantages and problems with each type.  The
4th and last category in Fig 2.11 is detailed physical models (finite
element) which calculate very slowly--it will be a few years<grin> before
these models can be used for anything approaching realtime.

Pacejka's new book is recommended to all you tire modelers out there.

Sorry for the delay.  I've been having a heck of a time getting online lately.

Indeed, my tire model is a physical model somewhere between the first and last
ones you described.  I did try a very, very basic FEM model at one point,
although it could be argued whether it was really an FEM model at all.

This was just a bunch of springs in the contact patch (I think I mentioned it
already).  It worked surprisingly well, as you don't have to worry about
combination slip effects; they work themselves out with no troubles.  But the
new approach is much faster, more precise, and more "accurate."  It lets me put
all sorts of dependencies with tire pressure, slip velocity, and so forth into
it, but of course it's all just theoretical, so loads of fun to play with to
see if my thinking of how tires work reflects reality :-)

Still though, I think it will automatically take care of load sensitive effects
to a large extent.  We'll see if what I think is the chief cause of load
sensitivity is really right after I can put that together and compare it with
some real data <rolling eyes>.  My gut tells me it'll be pretty good and let
you take a physical approach to tire design in the sim in a way that would make
sense to sim racers (not just programmers), and in a way not seen before.

Of course, it could completely suck too :-D

Todd Wasson
Racing Software
http://PerformanceSimulations.com
My car sim
http://performancesimulations.com/scnshot4.htm

Todd, funny about getting online, I've had troubles lately too.  When I
get on, things are often very slow.  Maybe all the people getting Iraq
news?  Or just surfing more because they are out of work<sigh>?

Your spring model fits the simple physical model category, I'll bet that a
little bit of searching will find something similar in the literature--
many of these have been proposed over the years and some of them work
quite well.

The way I heard it (at a tire company), to do finite element on a tire,
takes thousands of elements, and with tires the elements deform and
are nonlinear in load-deflection (unlike standard linear FEM on metal
structures).  It's still super-computer or cluster-computer stuff--if the
answer is supposed to model some real tire.

If your model works, you might consider publishing a paper on it
(eventually).

Hehe :-)  Fortunately today I haven't had any problems.  

The one I'm doing now was inspired by brush models, but goes futher into things
in some areas (and probably not as far into other things).  

Yes, it's certainly not something that you could do in real time on your home
PC (or any computer, probably).  I've got some basic FEM code, but haven't
looked at it much.  It's not well documented so it looks a bit too cryptic for
me to probably do much with.  It's still interesting stuff though.  

Quite some time back, I took a stab at making something like an FEM model.
It's quite hysterical :-)  This doesn't use any shearing action, which might
work fine if I remade it to do so.  For now it's just an interesting screen
saver ;-)

http://performancesimulations.com/files/funnytire.exe

Good for a nice laugh.  Actually, it's a completely accurate simulation of an
exploding tire ;-)

Interesting.  How would I go about doing that?  

Todd Wasson
Racing Software
http://PerformanceSimulations.com
My car sim
http://performancesimulations.com/scnshot4.htm