Lap-Time Simulation

[dug]

Hi all,
I'm in the process of rewriting the laptime sim I did last year, (hmmm, that was more work than I expected...) and I'm interested as to the capability other teams have in this area. Do you use commercially available software or custom? Quasi-Steady-State or full transient? What Tyre model etc. I haven't seen much on the forums regarding this and thought it might be interesting...

[dug]

Hi all,
I'm in the process of rewriting the laptime sim I did last year, (hmmm, that was more work than I expected...) and I'm interested as to the capability other teams have in this area. Do you use commercially available software or custom? Quasi-Steady-State or full transient? What Tyre model etc. I haven't seen much on the forums regarding this and thought it might be interesting...

[KevinD]

i use matlab/simulink with two toolboxes, the simmechanics box and the simdriveline box. the simmechanics box is great for modeling the chassis and the driveline box has some premade models for a slipping clutch, pajecka tire model (longitudinal only), transmissions, diffs, and a really really basic engine model. i think you will find you can put together a basic model very quickly and modify for fidelity from there.

[dug]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by KevinD:
i use matlab/simulink with two toolboxes, the simmechanics box and the simdriveline box. the simmechanics box is great for modeling the chassis and the driveline box has some premade models for a slipping clutch, pajecka tire model (longitudinal only), transmissions, diffs, and a really really basic engine model. </div></BLOCKQUOTE>

I've recently "discovered" MatLab, and found it to be an encredibly powerful and intuitive tool. My previous sim was written in C# using Visual Studio. And I spent most of my time learning to program, rather than actually writing useful code.
I am particularily interested in the tyre models used. This is arguably the most important part of the sim, and the hardest to get right. It seems a lot of teams are using one of Pacejka's Magic Formulae, however are any teams using transient (brush or other) models. Pacejka is a very nice model, however it (the older 94/ADAMS version anyway) doesn't allow for transient loads or temperature. Allowing for temp isn't too hard, but driving the slip-angles for transient conditions is giving me headaches.
Any ideas?

[ben]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by dug:
I am particularily interested in the tyre models used. This is arguably the most important part of the sim, and the hardest to get right. It seems a lot of teams are using one of Pacejka's Magic Formulae, however are any teams using transient (brush or other) models. Pacejka is a very nice model, however it (the older 94/ADAMS version anyway) doesn't allow for transient loads or temperature. Allowing for temp isn't too hard, but driving the slip-angles for transient conditions is giving me headaches.
Any ideas? </div></BLOCKQUOTE>

How are you accounting for temp? Are you using slip velocity multiplied by force as an energy input and then using a specific heat capacity approach to model the temp rise. What are you using to model the rate of cooling?

If you doing this seriously, the TTC data will be invaluable, I assume RMIT are getting the data? We use Pacejka 96 for car and I'm using the MF-MC model from Pacejka's book for motorcycle.

As for tyre lags. I've seen enough reports to be fairly sure that a first order lag would be fine for transients. Unless you're doing something like Hallum's model I don't think going overboard with transients within a Pacejka model setting is worth the effort.

Ben

[Jeff Curtis]

To do a basic parameter sensitivity based laptime simulation, there is almost no need for tire data. In 2001, I used a simple model that used basic parameters like tread width, CG height, engine power curve, gear ratios and a simple tire model represented by a coefficient of friction. You can learn a lot more practically applicable knowledge from a basic simulation. In other words, what is the trade off in tread width vs. the ability to slice through the slalom portions of the course? What are the optimum gear ratios given a certain shift time? These type of analyses can be done without a sophisticated tire model. When getting to more complicated handling models, most individuals on this "college"� level of experience are going to have a tendency to make a lot of wrong assumptions and make errors in judgment in analyzing simulation outputs. I think it admirable that you are interested in learning more about simulation and trying to apply it to make your car faster. I just issue these words of caution for you to always realize the large amounts of assumptions that are made even on the highest levels of racing simulations.
JAC

[Z]

I agree with Jeff.

Pick any reasonable numbers for "the tyre" and you will have an accurate model of some tyre, somewhere. But trying to get an accurate model of your particular tyre is like trying to nail jelly to the ceiling. The tyre parameters keep changing (temperature, wear, age, etc.). And the road surface, which is the other half of the "equation", also changes (different tracks, bumps, rubber laid down, dust, oil, water...). Small bumps (ripples) and a stiff suspension can make a huge difference to tyre performance through transient unloading. I don't know how easily this can be extracted from tyre data - it probably depends more on the bumps (track profile) and suspension characteristics.

Lap-time simulations are very useful tools, but mainly to get a big picture idea of what sort of car you need. An accurate model of the track (bumps and all) is just as important as an accurate model of the car.

Z

[dug]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">How are you accounting for temp? Are you using slip velocity multiplied by force as an energy input and then using a specific heat capacity approach to model the temp rise. What are you using to model the rate of cooling? </div></BLOCKQUOTE>
What I meant to say was that it easy to alter the Pacejka formula to account for a temperature characteristic. The calculation of the temperature should take into account heat inputs (tyre deformation and scrubbing) and outputs (conduction, convection etc). Obviously there will be some tuning of empirical parameters to fit data... Check out paper by Mizuno for ideas. I do agree with Jeff and Z to some extent as to the accuracy of environment needing to match the accuracy of the car, however a tyre model that is too simple can lead to erroneous assumptions. Tyres are in fact load sensitive, and this has too be taken into account if trying to back-to-back model the effects of aero devices for example. A lot can be done quasi-steady-state on excell when looking for broad trends to decide gear ratios etc, but as with everything, the last 10% takes 90% of the time.
I understand the force available from tyres change with wear, surface, temperature, pressure... but <span class="ev_code_PURPLE">how much</span>? Considering the cost of tyres, it would be nice to not have to wear out as many of them trying a miriad of potential setup options for different conditions etc.
The other thing is a lot of teams will be getting a lot of tyre data soon... http://fsae.com/groupee_common/emoticons/icon_eek.gif

I think the main thing is the better the sim, the more issues can be found early on, rather then when the car is built.

Cheers

[Z]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by dug:
... but as with everything, the last 10% takes 90% of the time. </div></BLOCKQUOTE>
So the first 90% (where you get the bulk of the useful information) only takes 10% of the time http://fsae.com/groupee_common/emoticons/icon_smile.gif.

So any teams who are undecided about doing simulations - start simulating!

Z

[John Bucknell]

Ten years ago, I made a quasi steady excel model of a FSAE car and decided CVTs and wings were the way to go! Lots of bad assumptions led to a rather optimistic lap time, but even those without Matlab (US$8k a license for professionals) can use a the iterative functions to achieve a feel for performance with changes to global parameters. I use a derivative of that original program (well-correlated and physics-based now) professionally today - even though I have Matlab vehicle performance models available to me. Granted, I came to Matlab later in my training so I am biased - but both can give excellent results for simple cases....