How to model car launching

[Spetsnazos]

I'm looking for insight on how to properly model car launching via lap simulation.

Can anyone help me get started on how to simulate the vehicle that is starting from 0 mph.

So far my program takes into affect the vehicle torque, shifting, rpm curve, tire size, vehicle weight, etc. I'm just having a difficult time modeling the acceleration from a stop.

Cheers

[murpia]

Other things I suggest you add are longitudinal weight transfer and clutch slip.

Regards, Ian

[ffrgtm]

clutch slip being torque multiplication

[Spetsnazos]

I am intending to implement longitudinal weight transfer in the near future, thanks. I just really want to get a general idea of how to model launching so I can get the ball rolling.

I know when launching we want to rev to a certain RPM and the drop the clutch or whichever strategy works best for everyone...The problem I am facing is how to model this acceleration.

Thanks guys!

[thewoundedsoldier]

what program are you using?

[Spetsnazos]

Originally posted by thewoundedsoldier:
what program are you using?

I'm writing the program and it is in Matlab at the moment.


I'm probably going to be moving toward C# in the near future(for my sake) but I don't really have the time to rework it at this time. I find C# to be vastly superior, the only problem is future FSAE students at our university wont be able to use/compile it as it is not part of the regular engineering curriculum. Anyways, I digress...

Have any links to a sensible way to model vehicle launching?

[murpia]

Originally posted by ffrgtm:
clutch slip being torque multiplication
No, that only happens with a torque converter.

Regards, Ian

[murpia]

Originally posted by Spetsnazos:
I know when launching we want to rev to a certain RPM and the drop the clutch or whichever strategy works best for everyone...
Try holding constant engine revs, slipping the clutch and modulating engine / clutch torque (they are the same) according to available tyre grip.

That's the essence of the 'launch control' strategies within the commonly used ECUs.

When the clutch slip reaches zero, assume engagement and allow the engine revs to rise.

I've built this exact model in Excel. However, I would stick to MATLAB rather than go to C#, or you will have to re-implement all the hundreds of handy MATLAB functions by yourself.

Regards, Ian

[Spetsnazos]

Originally posted by murpia:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by ffrgtm:
clutch slip being torque multiplication
No, that only happens with a torque converter.

Regards, Ian </div></BLOCKQUOTE>

Nah I think what he was saying is to model clutch slipping as torque multiplication, in that a certain percentage of clutch slip implies only a percentage of the maximum torque at that RPM.

[Spetsnazos]

Originally posted by murpia:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Spetsnazos:
I know when launching we want to rev to a certain RPM and the drop the clutch or whichever strategy works best for everyone...
Try holding constant engine revs, slipping the clutch and modulating engine / clutch torque (they are the same) according to available tyre grip.

That's the essence of the 'launch control' strategies within the commonly used ECUs.

When the clutch slip reaches zero, assume engagement and allow the engine revs to rise.

I've built this exact model in Excel. However, I would stick to MATLAB rather than go to C#, or you will have to re-implement all the hundreds of handy MATLAB functions by yourself.

Regards, Ian </div></BLOCKQUOTE>

Going to try this and read about it a little more as well. Thanks for the start!

Cheers

[Numchuks]

Don't forget about forces like rolling resistance and drag. Easily expressible through the right equations, difficult get good real-world magnitudes to without verification against physical test data. Even though the car is overcoming these forces during acceleration they're present as long as vehicle speed is greater than zero, and the only forces present while changing gear (while driving in a straight line on level ground).

For a keen eye and mind that is wondering about toe and camber as deceleration forces...I've always absorbed them into a single rolling resistance equation in my acceleration simulations (in Matlab and Excel).

Tire data is another keystone of a good acceleration simulation. As long as friction coefficients aren't infinite, assuming that it is isn't valid and won't yield accurate simulation results. The vehicle isn't always moving as fast as the wheels are spinning. The engine makes the force that moves a vehicle. But tire slip decides how much of it gets used. http://fsae.com/groupee_common/emoticons/icon_wink.gif

Other parameters that I know I've used:
-Torque curve
-Gear ratios
-Final drive ratio
-Shift duration
-Shift point
-Weight distribution for tire data (including driver)
-Inertia losses

Search the forums too! This topic has been discussed at length before!

[Spetsnazos]

Originally posted by Numchuks:
Don't forget about forces like rolling resistance and drag. Easily expressible through the right equations, difficult get good real-world magnitudes to without verification against physical test data. Even though the car is overcoming these forces during acceleration they're present as long as vehicle speed is greater than zero, and the only forces present while changing gear (while driving in a straight line on level ground).

For a keen eye and mind that is wondering about toe and camber as deceleration forces...I've always absorbed them into a single rolling resistance equation in my acceleration simulations (in Matlab and Excel).

Tire data is another keystone of a good acceleration simulation. As long as friction coefficients aren't infinite, assuming that it is isn't valid and won't yield accurate simulation results. The vehicle isn't always moving as fast as the wheels are spinning. The engine makes the force that moves a vehicle. But tire slip decides how much of it gets used. http://fsae.com/groupee_common/emoticons/icon_wink.gif

Other parameters that I know I've used:
-Torque curve
-Gear ratios
-Final drive ratio
-Shift duration
-Shift point
-Weight distribution for tire data (including driver)
-Inertia losses

Search the forums too! This topic has been discussed at length before!

I'm accounting for everything you've mentioned except inertia losses and weight distribution(for now).

My model already accounts for rolling resistance and drag(both of which are minimal to say the least...). I read a few threads on it and I am working myself up to the end but I need to overcome launching and a few other minor obstacles.

PS: I don't dislike MatLab or Excel, I just think they all have their uses and I prefer C# to both of them. I hardly use any of the MatLab features and just program them myself. Dumb? Maybe, maybe not. Whenever you use a built-in function you just read the tooltip of how it works and what parameters it is asking for, right? I prefer to set the parameters myself and know EXACTLY what that function does. Just a matter of preference. To me, this is how you can move from one language to the next and not fail because that special function isn't built-in.



Cheers

[Numchuks]

That's a good way at looking at it. Full control is always best....hence my extreme hatred towards ABS and traction control systems in passenger vehicles.

In both Matlab and Excel I built mine from the ground up, accounting for one factor and refining it at a time. My first sims assumed zero shift time duration until I could figure out a way to implement it for a user specified duration. It's now grown into a table of fully independent shift durations for every gear change.

I've stuck with Excel. It's tricky, but doable. The file is a couple megs in size, but I've been developing it for nearly 2 years now accounting for more and more factors (wind speed, wind direction, slip-limitations...etc).

[JasperC]

I think it all starts with tyre data, doesn't it. What you want is some data for friction coefficient vs. slip ratio. I assume you'll be able to figure out the slip ratio (as a function of RPM, clutch slip and car speed); then you can use the corresponding friction coefficient to find the momentary acceleration of the car and use that to go to the next moment in time.

If you get that to work properly, you've done a very nice job if you ask me. http://fsae.com/groupee_common/emoticons/icon_smile.gif



Originally posted by Numchuks:
The vehicle isn't always moving as fast as the wheels are spinning.
I'd say: the vehicle is NEVER moving as fast as the wheels are spinning (and if it is, it means you're neither accelerating nor braking or cornering which means the driver is doing a bad job :P). We found from data last year that we always have a slip ratio of 3 to 5% even when doing 100kph. As long as you put torque on it, it'll slip.


Kind regards,
Jasper Coosemans

Chief Drivetrain 2009-2010
DUT Racing Team
Delft University of Technology

[exFSAE]

TIRE DATA!! Fx as a function of slip ratio. How on earth is this going to be representative without it?

[Spetsnazos]

Originally posted by JasperC:
I think it all starts with tyre data, doesn't it. What you want is some data for friction coefficient vs. slip ratio. I assume you'll be able to figure out the slip ratio (as a function of RPM, clutch slip and car speed); then you can use the corresponding friction coefficient to find the momentary acceleration of the car and use that to go to the next moment in time.

If you get that to work properly, you've done a very nice job if you ask me. http://fsae.com/groupee_common/emoticons/icon_smile.gif


<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Numchuks:
The vehicle isn't always moving as fast as the wheels are spinning.
I'd say: the vehicle is NEVER moving as fast as the wheels are spinning (and if it is, it means you're neither accelerating nor braking or cornering which means the driver is doing a bad job :P). We found from data last year that we always have a slip ratio of 3 to 5% even when doing 100kph. As long as you put torque on it, it'll slip.


Kind regards,
Jasper Coosemans

Chief Drivetrain 2009-2010
DUT Racing Team
Delft University of Technology </div></BLOCKQUOTE>

Good call mate.

Going to start working on this now!

Thanks again!

[Gruntguru]

Originally posted by Numchuks:
That's a good way at looking at it. Full control is always best....hence my extreme hatred towards ABS and traction control systems in passenger vehicles. LOL. Traction control and ABS both introduce compromise but they have also useful capabilities beyond human drivers.

[Numchuks]

Originally posted by Gruntguru:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Numchuks:
That's a good way at looking at it. Full control is always best....hence my extreme hatred towards ABS and traction control systems in passenger vehicles. LOL. Traction control and ABS both introduce compromise but they have also useful capabilities beyond human drivers. </div></BLOCKQUOTE>

Agreed, they do have their place and attempt to keep you in better control of your vehicle. But when you live in a city where it is -30 celsius or colder (before windchill) on a regular basis at this time of year, those safety features become incredibly annoying!

[Bus_Lengths]

Clutch slip will provide higher torque to the wheels than the engine can. Unless you actually have a good data acquisition system, you will need to guess/approximate this effect. From drag racing data I've collected myself, clutch lockup (50 lb flywheel @ 7000 RPM clutch drop)seems to occur at about 0.1 seconds give or take.

Knowing the dimensions (and material) of the clutch system and simple clutch equations you can work out the maximum holding torque - which will be the torque applied during slipping - kinda. I've found that a 70% factor seems to work well to convert the static holding torque into a kinetic slipping torque. Supposedly there is also a speed difference (between flywheel and clutch disk) effect on friction torque, but I've never found any hard numbers on this so I haven't included it.

But yeah, you basically use the slip torque as the torque being applied to the drivetrain while the clutch is slipping (and this torque is subtracting from the intial launch RPM kinetic energy the cltuch system had).

Using this method (disregarding tires for a moment) I've gotten my model to predict 60' times quite accurately. The model will match front and rear wheel speed during the launch extremely well - which is ultimately what you are trying to do.

Now tires provides an interesting problem. Most longitudinal tire data I've seen talks about 10% slip ratios. There are about 800 definitions for slip ratio, but it's ultimately just a relation between forward speed of the vehicle and the 'distance/time' the rear tires are covering. With this in mind, for a low pressure, bias-ply, these sissy slip ratio's mean nothing. Where's the data for 800% slip ratio? The slick's I've run produce highest longitudinal thrust when they are spinning ~10 - 20 times faster than the front wheels. So, again, some guessing/approximating comes into play. In order to match real life performance, I had to create a coefficient of friction table that varied with slip - and was far more extreme than what you see for radial street tires.

One thing you MUST add is inertial losses. 1st gear torque will be far less than what you think it is and thus you will be vastly over predicting accelerative performance if you assume the engines power curve is the same in each gear.

Second, for added fun, throw in a simple spring/mass/damper system to your weight transfer model. Quite hilarious to see what can be done with damper settings. For such a simple change, they can have relatively noticeable effect on 60' times just due to dynamically loading up the tire beyond the 'assumed' weight transfer limits.

I've played around a lot with determining what would be required to lift the wheels at an FSAE event. I'm convinced that it could be done without grossly affecting performance in the 'twisties' (just setup changes). However, I could never convince my team to front the money for some 13" Mickey Thompson's to test with.

[Wesley]

Originally posted by Bus_Lengths:
I've played around a lot with determining what would be required to lift the wheels at an FSAE event. I'm convinced that it could be done without grossly affecting performance in the 'twisties' (just setup changes). However, I could never convince my team to front the money for some 13" Mickey Thompson's to test with.

Good luck running wrinklewalls through a slalom. Ever put taco bell trays under your rear wheels? http://fsae.com/groupee_common/emoticons/icon_biggrin.gif

If you tech with MT slicks, you run all events with MT slicks.

Try putting two grande meals on top of those trays before you drive on top of them. That's what it will feel like in the corners.

[murpia]

Originally posted by Spetsnazos:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by murpia:
<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by ffrgtm:
clutch slip being torque multiplication
No, that only happens with a torque converter.

Regards, Ian </div></BLOCKQUOTE>

Nah I think what he was saying is to model clutch slipping as torque multiplication, in that a certain percentage of clutch slip implies only a percentage of the maximum torque at that RPM. </div></BLOCKQUOTE>
Well, that's not quite right either.

A generic clutch can transfer the maximum engine torque while slipping. The slipping dissipates energy, so the power at the wheel is lower than the power at the engine, but the torque is the same.

Of course, your actual clutch may have limitations that a generic clutch doesn't, so you'll need data on that too.

Regards, Ian

[murpia]

Originally posted by Bus_Lengths:
Clutch slip will provide higher torque to the wheels than the engine can. Unless you actually have a good data acquisition system, you will need to guess/approximate this effect. From drag racing data I've collected myself, clutch lockup (50 lb flywheel @ 7000 RPM clutch drop)seems to occur at about 0.1 seconds give or take.

Knowing the dimensions (and material) of the clutch system and simple clutch equations you can work out the maximum holding torque - which will be the torque applied during slipping - kinda. I've found that a 70% factor seems to work well to convert the static holding torque into a kinetic slipping torque. Supposedly there is also a speed difference (between flywheel and clutch disk) effect on friction torque, but I've never found any hard numbers on this so I haven't included it.

But yeah, you basically use the slip torque as the torque being applied to the drivetrain while the clutch is slipping (and this torque is subtracting from the intial launch RPM kinetic energy the cltuch system had).
I agree that transiently you could increase clutch output torque beyond the maximum engine torque by decelerating the engine inertia down to a lower rpm.

I think that strategy may not work too well for a low inertia / low bottom end torque engine like a 600cc bike engine. Hence my suggestion to copy the 'launch control' strategies which essentially hold the engine rpm constant.

Also, I think when modelling the car, you need to add inertia terms for the rotating inertia of the drivetrain, and once clutch slip = 0 that off the engine too. In the low gears the amount of torque required to accelerate the engine is a significant percentage of the maximum.

Regards, Ian

PS a 50lb flywheel would be the first place I'd look when trying to save weight!

[exFSAE]

Originally posted by Bus_Lengths:
There are about 800 definitions for slip ratio

Interesting. I'm familiar with 2 - slip ratio (SR, from Calspan) and longitudinal slip (SL, from Calspan).


With this in mind, for a low pressure, bias-ply, these sissy slip ratio's mean nothing. Where's the data for 800% slip ratio? The slick's I've run produce highest longitudinal thrust when they are spinning ~10 - 20 times faster than the front wheels.

I'll admit that on some tires, there is something to be said for extra wheelspin in generating heat - perhaps as the first run for a FSAE car before making a 'real' pass the second go.

I'll also say that at 0.0 forward speed, definitions become poorly defined.

In general though, I find it very hard to believe that 800%, 100%, or even 50% slip ratio ("real" slip ratio) is required for peak drive force. Both from lab data, as well as more anecdotal data from watching the FSAE acceleration events. Quickest cars do not leave the line in a blaze of smoke, fire, and glory (which you would certainly have going down the lane at 100% slip). Almost deceivingly small amounts of initial wheel slip, and away they go to run a 4.0-ish time.

[SNasello]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content"> With this in mind, for a low pressure, bias-ply, these sissy slip ratio's mean nothing. Where's the data for 800% slip ratio? The slick's I've run produce highest longitudinal thrust when they are spinning ~10 - 20 times faster than the front wheels.



I'll admit that on some tires, there is something to be said for extra wheelspin in generating heat - perhaps as the first run for a FSAE car before making a 'real' pass the second go.

I'll also say that at 0.0 forward speed, definitions become poorly defined.

In general though, I find it very hard to believe that 800%, 100%, or even 50% slip ratio ("real" slip ratio) is required for peak drive force. Both from lab data, as well as more anecdotal data from watching the FSAE acceleration events. Quickest cars do not leave the line in a blaze of smoke, fire, and glory (which you would certainly have going down the lane at 100% slip). Almost deceivingly small amounts of initial wheel slip, and away they go to run a 4.0-ish time. </div></BLOCKQUOTE>

While i agree that common thinking would suggest launching at or near the slip ratio where peak longitudinal force is generated would produce the quickest over-all acceleration run, experiences from our past car would suggest otherwise.

With our 'fastest' acceleration run (3.85 or so seconds in Italy) we saw slip ratios well upwards of 800% (no smoke or fire however). Developing a longitudinal force model of a tire for use in an acceleration simulation is somewhat difficult when you only have data up to 20% slip. I currently have 2 models that I fit to the same set of data, that both match the data quite well, but behave very differently beyond the limit of the data.

The first model suggests that (sliding) coefficient of friction drops off relatively quickly and drastically beyond the peak, and the other saturates at some peak sliding friction coefficient much closer to the overall peak.

Would be really neat to see data from some Kistler wheels in a 1600% slip ratio acceleration launch.

[Bus_Lengths]

Originally posted by Wesley:

Good luck running wrinklewalls through a slalom. Ever put taco bell trays under your rear wheels? http://fsae.com/groupee_common/emoticons/icon_biggrin.gif

If you tech with MT slicks, you run all events with MT slicks.

Try putting two grande meals on top of those trays before you drive on top of them. That's what it will feel like in the corners.

Haha, our team was never going to seriously be competitive in every event. Thus, I always thought it would be interesting to just devote everything towards one event and make a name for yourself.
40/60 front/rear weight distribution, upped ride height, correctly setup dampers, MT slicks, 15,000 RPM clutch drop. I've always thought that seeing an FSAE car lift the front wheels and destroy everyone in the 0-75m would be hilarious.

On top of that, the design event would have been great. "Wait, hold on, you PURPOSELY made your CG 19" off the ground?"




Originally posted by murpia:
I agree that transiently you could increase clutch output torque beyond the maximum engine torque by decelerating the engine inertia down to a lower rpm.

I think that strategy may not work too well for a low inertia / low bottom end torque engine like a 600cc bike engine. Hence my suggestion to copy the 'launch control' strategies which essentially hold the engine rpm constant.

Also, I think when modelling the car, you need to add inertia terms for the rotating inertia of the drivetrain, and once clutch slip = 0 that off the engine too. In the low gears the amount of torque required to accelerate the engine is a significant percentage of the maximum.

Regards, Ian

PS a 50lb flywheel would be the first place I'd look when trying to save weight!


You have a point that for an FSAE style car, the inertia stored by the clutch system is nowhere near what it's like in a street car. Though I think you could still see some effects. From data I've logged on a street car, the slipping clutch can deliver about 600 lb-ft into the transmission (about 50% more than engine torque) - resulting in extremely high initial accelerations.

So, while diameter is squared when looking at inertia (meaning very small inertia for 600cc motor), speed is also squared when looking at kinetic energy (meaning higher KE potential due to very high redline). So being able to rev twice as high as a normal street car will have some benefit. However, this is somewhat pointless as I never remember being able to hook on an FSAE tire with more than about 7000 - 9000 RPM. But, if the conditions were right...




Originally posted by exFSAE:

Interesting. I'm familiar with 2 - slip ratio (SR, from Calspan) and longitudinal slip (SL, from Calspan).


I'll admit that on some tires, there is something to be said for extra wheelspin in generating heat - perhaps as the first run for a FSAE car before making a 'real' pass the second go.

I'll also say that at 0.0 forward speed, definitions become poorly defined.

In general though, I find it very hard to believe that 800%, 100%, or even 50% slip ratio ("real" slip ratio) is required for peak drive force. Both from lab data, as well as more anecdotal data from watching the FSAE acceleration events. Quickest cars do not leave the line in a blaze of smoke, fire, and glory (which you would certainly have going down the lane at 100% slip). Almost deceivingly small amounts of initial wheel slip, and away they go to run a 4.0-ish time.

Regarding Slip ratio. I'm just saying I've seen it defined multiple ways in different text books and papers. And not all of them are 'trivially' different. So, I'll be honest I couldn't even tell you offhand what the true definition is other than the two factors it relates.

I guess where I'm going with this is that in any 'launch modeling' consisting of a high RPM clutch drop, you are going to see high levels of rear wheel slip at very low vehicle speeds. (like 15 mph vs. 0.4 mph) or (25 mph vs. 1.5 mph). I couldn't tell you offhand what these slip ratios come out to be but I'm of the impression that these are beyond most 'slip ratio' graphs. And slip ratio graphs would have you believe that these high, 'off-the-chart', slip ratios result in minimal forward tractive force. However this cannot be true as very high accelerations can be seen in this situation. I will say, my experience with this is not with FSAE tires (as we could never afford any usable data aq), and obviously the tire changes things drastically.

[Tim.Wright]

Simulating a launch from stand-still is a ball ache of a simulation problem. I have spent some months working with a 10k€ simulation package that could never get a launch to be stable. So the first question you need to ask yourself is how badly do you NEED the launch-from-0 functionaliy. Would a simulation from 5kph not be adequate?

And what exactly are you wanting to measure? Are you comparing different mass properties? different clutch techniques? different tyres? different kinematics?

In my experiance, if you want to measure the maximum launch capability of the vehicle there are two problems already touched on in terms of modelling

1. Tyre model
2. Clutch model

The biggest problem being the tyre model, since they are not valid for such low speeds and the pacejka model actually doesnt work at zero speed (slip angle is undefined)

Also, at very low speeds (ie the first rotaion of the wheel) any slight disturbances in the wheel speed cause huge variations in slip ratio making your Fx calculations useless and you end up in a huge oscillating mess.

With the clutch model, this is a piecewise function incorporating static and dynamic friction and is very hard to get right without adding damping and a torsional spring into the model (unless you use a simplifed first order model).

Then you have the problem of controlling the clutch and throttle. This is a huge job because if you want reliable comparisons you need a repeatable control system that does not need re tuning if you change the tyres or something else.

So, my path was as follows;

1. Tyres: I still have no answer for this. I used a "low speed" model in my software but I really don't trust the results, and at the switch to the high speed model, it created chaos.

2. Clutch: Ditch the clutch and write a traction control algorithm (which monitors long slip) that cuts (multiplies x <1) engine torque directly while the driver just gives full gas all the time.

This will give you the amount that the vehicle is traction limited. I.e. your traction control signal is at 0.6? it means you can only put 60% of the engine power to the ground. I found this a very useful tool.

Then you simply choose your desired slip ratios, put it into the traction control and then you have a launch simulation operating in a repeatable environment.

Its not perfect, there are still other things to look at, but you need to decide where you want accuracy in your model.

Tim

[Spetsnazos]

Thank you all for replying. I am working with the suspension guys to try to understand the tire model that we have. Will report back as I get more progress.

THanks!!!!

[Shebert]

Its a pretty interesting and feasable project.

For tire, what you need is a relaxation length model, its an inline model that takes the slip at the wheel and modulates it in function of the tire response caracteristic and return the contact patch slip.

It may sound fancy but with only the longitudinal(in your case) rigidity you can figure out a good model.

Read chapter 7 of Pacejka's Book.

Cheer,

Shebert

[Edward M. Kasprzak]

The biggest problem being the tyre model, since they are not valid for such low speeds and the pacejka model actually doesnt work at zero speed (slip angle is undefined)


Be careful how you generalize. Some tire models work down to zero speed. Many don't. We supply both types for our MRA customers, depending on the application. As Shebert mentioned, "relaxation length" is a key concept for low speed tire behavior.

[murpia]

I don't wish to promote 'dumbing down' of this thread, especially as it contains a lot of good info.

But, following the 80/20 rule where you can get 80% of the results for 20% of the effort has served me well in many cases like this.

As ever, the key is correlation data. I have in the past used a single 'mu' value to represent tyre grip and fitted the resulting acceleration profile to real data by tweaking that single value of 'mu'. Just playing with that model, in terms of understanding the effect of final drive ratio, gear ratios, shift strategies, engine torque curves etc. was very useful.

Regards, Ian

[Tim.Wright]

Originally posted by Edward M. Kasprzak:
Be careful how you generalize. Some tire models work down to zero speed. Many don't. We supply both types for our MRA customers, depending on the application. As Shebert mentioned, "relaxation length" is a key concept for low speed tire behavior.

Yes sorry, I forgot the word Pacejka in that sentance. Like I said I have used a low speed tyre model before but in my opinion unless you know what its doing you are really shooting in the dark using it. Especially if you have a transition from low speed to high speed in you simulation.

Also, on the relaxation length. I have always been suspicious of this number in a tyre data set, especially if I know the tests done were steady state tests like the TTC data (at least the first couple of sets)

Am I right to be suspicious about this or am I missing something?

Tim

[jrickert]

Getting the launch right is very tricky. I made a very detailed model factoring in tires, weight transfer inertia of all components, clutch modulation, throttle, torsional compliance, power curves etc. It turned out to work well. It worked too well in that it was sensitive to how you launched it just like the real thing. The point of the model was to understand the impact of broad things like gearing, shifting, power curves, weight distribution, launch strategies etc.

Initially i modeled clutch torque and throttle as open loop functions of time. It was very tricky to set up these function to launch the car well. Most of the time it would either bog or do a big nasty burnout unless they were balanced perfectly. A driver has a feel and judgement that an open loop function doesn't have. After some tweaking you could get things dialed in well but any tweak to other parameters such as power or gear ratio would throw things off again and screw up the launch parameters.

To combat this problem i made a closed loop traction controller/driver that reduced engine torque based on on slip error. This stabilized things GREATLY but not to my satisfaction.

I think i may try the hold engine rpm constant and modulate clutch to achieve goals until lockup strategy.

In the end the model was pretty damn close to real times and useful for understanding big picture changes despite "annoying realism". Currently the only car we have with good data acquisition is not operation and will not likely be for some time. I'm very eager to get a car with some good data aq on it to calibrate and compare my model. Thats where the real fun starts. To me real testing is king. I make it a rule to not believe any model that has not been validated. It is easy to make a model what matches existing data. The real test of a model is whether or not it can make effective predictions.

I like this thread in that people are sharing actual experiences. Usually FSAE is steeped in negativity, cockyness and vague assertions of why people are wrong without proposing the right answer.

[exFSAE]

Originally posted by Timo:
Also, on the relaxation length. I have always been suspicious of this number in a tyre data set, especially if I know the tests done were steady state tests like the TTC data (at least the first couple of sets)

TTC data doesn't look steady state to me.

[Edward M. Kasprzak]

Timo: While the TTC tests were not designed to specifically look at transient tire behavior, they aren't "steady-state" either. Most of the tests involve sweeping slip angle or slip ratio at constant rates. The warm-up cycles at the start of each run provide something closer to step changes in slip angle--might be a good place to look if you're trying to estimate a relaxation length.

Relaxation length can also be estimated as some percentage of the tire's circumference. I'll leave it to you look up the standard range of values in the literature.

Having the relaxation length effect in your model will stabilize it at low speeds. Even if you don't have the exact value, having a value and the corresponding equations will allow you to run down to zero speed. I suggest trying a wide range of relaxation length values--see how sensitive your results are to this value and then decide how accurate a value you need for your application.

[Shebert]

There is a lot of good SAE papers on tire model for ABS system development that can be use for reference.