Lap simulation

[Kiril Varbanov]

Hi All,
I decided to share, just before the start of the season, the latest article and material which I was able to test - Simulating a lap @ F1 Framework.
I'm open to questions.
Again, I have not had the time to modify the car model and the track to suit F1, the idea was to present in general one of the tools that the teams are using.

[Greg Locock]

Interesting. Looks very similar to Lapsim. And completely unlike ADAMS /Drive or Smart Driver (that is a good thing).

To my mind one weakness with Lapsim is that you have to pick the apex of each corner, another that you don't have a circuit model as much as a defined path.

It would be very interesting to know if the speed vs position on track graph actually correlates very well, and whether the laptime improvements are directionally correct, or more accurate than that. I suspect that the correlation is not all that great since real drivers do not sit on the edge of the v-g-g envelope for very long, whereas that is almost exactly how this type of software works, it explores that boundary.

I had a go at writing a v-g-g based simulator, but ended up spending a lot of time with the tire models, which is just as well since that's my day job.

[Jersey Tom]

Interesting. Looks [...] completely unlike ADAMS

Well indeed given that it's quasi-static rather than dynamic.

(that is a good thing)

Debatable! But hey, right tool for the right job.

[Greg Locock]

One of the fish eating guys with a bulgy forehead in the USA has written a path optimiser I can use for double lane change (Moose test, or ISO B) for ADAMS. Even that rather trivial sequence takes up to 13 iterations to optimise, and that's in neutral, no braking, no ESC.

[Kiril Varbanov]

it explores that boundary.

If I got that correctly - yes, it's assuming 100% grip available anywhere on the track, which is, however, configurable.
In fact, almost every parameter in this software is configurable, which was OK for me to play with different setups and see how they correlate. For example, increasing the frontal area results in larger Cd and therefore deceased speed (unsurprisingly).
Altered gear ratios give better acceleration out of certain corners, as visible from the speed trace given the distance, but lower top speed on the straights, etc.

I was actually mostly interested in importing different aero maps, but have not had the chance to do so, so spent some time comparing the results post-simulation.

[Greg Locock]

If you still have access to it and have a reasonably good model of a known car on a known track, can you just dial in a 5% mass change, then a 5% torque curve boost, and then both together? If you are feeling really bored, you could do a 5% swing on Cd, CL, and mu as well.

Does the software allow you to set up massive numbers of runs with a matrix of changes(ie can you set up a taguchi style DoE)?

Incidentally in Lapsim I always set up a constant power engine so that gear changes didn't matter, otherwise you might be tracking local optima.

[Kiril Varbanov]

If you still have access to it and have a reasonably good model of a known car on a known track, can you just dial in a 5% mass change, then a 5% torque curve boost, and then both together? If you are feeling really bored, you could do a 5% swing on Cd, CL, and mu as well.

Does the software allow you to set up massive numbers of runs with a matrix of changes(ie can you set up a taguchi style DoE)?

Incidentally in Lapsim I always set up a constant power engine so that gear changes didn't matter, otherwise you might be tracking local optima.

I still have access, yes. I will do some tweaking around and let you know. The car model is precise Le Mans plus track.
I don't think batches of runs are possible, but I might be wrong.
# time is never enough recently ... vote for 30 hours per day?

[Greg Locock]

To answer my own question, having raided their website:

The simulation engine is provided as an ActiveX component that can be used in any environment that can host ActiveX, for example Microsoft Excel or Visual Basic.

This can be used to create a command line version, running exactly the same simulation code but allowing fully automated runs controlled by your own code or tools. This has been used:

-In Excel to set up parameter sweeps of, say, decreasing fuel load with decreasing tyre performance. Data can be wholly or partly sourced from the spreadsheet. For example most data could come from files saved from the standard version but gear ratios could be fed from a race engineering setup sheet.

-With a multi-objective parametric optimiser to automatically explore the trade-offs between different parameters.

-With MATLAB to drive a set of experiements.

Which is great, that is exactly how I'd approach it.

[Kiril Varbanov]

Which is great, that is exactly how I'd approach it.

Indeed, it's just I don't have the full rights to unlock that feature separately, instead, I have it simply embedded in the software.
Automation is literally everywhere, so command-line runs and batches are in fact a top priority for certain environments.

[Caito]

http://www.optimumg.com/software/optimumlap/

It's free. Check it out!

[silente]

Caito,

i was just posting the same link.

Optimumlap is anyway a point mass simulation, so no suspensions, no weight transfer etc. But up to a certain point can still be very useful in my opinion. The only thing it is not still great is the user interface in my opinion. I struggle a bit to go through data, but you can import them somewhere else if you need.

Honestly, i don´t know very well Aerolap, i had some experience with Lapsim (which was free and usable before, now it is still free but the downloadable version doesn´t allow to run simulations anymore) and they all use more or less the same approach (back - forward from corners).

One thing is not clear to me anyway: for a point mass simulator, it is straight forward to calculate maximum corner speed. For a simulator with a "complete vehicle model" you will have to at least investigate iteratively on front and slip angle to find the right combination to produce the right Fy and the right Yaw equilibrium. I wonder how can they do that so fast!

When i tried to write my own steady state simulation in matlab, it was very slow to even find the maximum speed of one corner! My programming skills are pure, so maybe it´s just a question of better programming. But still in a simulator like lapsim or aerolap they will have to increase gradually Ay or V and for each of them they need to find the Slip angles (or Beta and Delta) that give equilibrium. I still don´t understand how can they do that so quickly. In lapsim, it is really a matter of a second to find max corner speeds.

[Kiril Varbanov]

Great to spark such an interesting discussion.

I also have experience with OptimumLap - I ran into those guys last year prior to Austin GP.
Yes, the software is rather limited in some areas, but it's a free one, and in fact, unlike @silente, I liked the interface, although the creators were kind enough to listen to my feedback for some general usability improvements.

[Jersey Tom]

For a simulator with a "complete vehicle model" you will have to at least investigate iteratively on front and slip angle to find the right combination to produce the right Fy and the right Yaw equilibrium.

Not necessarily...

In any event, simulations - there's no magic to it. They're just a tool, different tools have different applications. There are times when I'd rather have a point mass predictive lap sim (which you can write over a weekend) than anything else, and there are times when nothing short of an ADAMS or other multi-body model will do.

[silente]

Not necessarily...

JT would you like to expand about this one?

My understanding of the problem is that, if you calculate max corner speed assuming a steady state condition, then net yaw moment need to be equal to zero. I know you could go other ways (for example with Milliken Moment method) but i don´t think it is what these softwares do...or am i wrong?

Ho do they find maximum corner speed in your opinion? I have also seen software that before to run the lap sim itself, create a v-g-g map (as Greg said) and then simply take from there the maximum speed. But to create this map they anyway need time...Lapsim (maybe also aerolap?) seems to have a different approach...

[Jersey Tom]

JT would you like to expand about this one?

Not much, no

Suffice to say it's much easier to ask a tire model, "Can you exist in this vehicle condition?" than, "What are the slip angles at this condition?"