FSAE Tire Test Consortium -- Round 3

[Edward M. Kasprzak]

Hi all,

I've been watching the thread, and here are some answers to your questions:

1. How were tires selected? We're testing tires that are either new or are older but haven't been tested yet. The only repeat this time is the Hoosier 7.0, which will help compare against the previous tests since the test program has been revised.

2. Newbies? Anyone not familiar with the FSAE TTC should follow the link in my initial post (top of thread).

3. What Michelin? We've asked Michelin to supply tires, and they agreed to send us four. We're not sure what compound they're sending--tires should arrive at Calspan in the next few days. We'll know then. We're not testing any rain tires.

4. Rim width idea? We have had this idea in our pocket since the first round of tests. It was axed previously when it landed in the "you can't do everything" category. Previously we focused on putting as many tire constructions as possible through the machine. We hope that the round 3 data will show some general rim width effects that can be translated to tire constructions tested in the previous rounds.

5. Peaks? We try to get to the slip angle peaks, but if the peak is past 12 degrees you're probably not going to see that on the car anyway. One issue is that the surface at Calspan is very clean, unlike the surfaces most FSAE cars run on. We think this causes the peaks to be somewhat high on the test machine. We've had a LOT of discussions about what to do about this--there are many constraints at the test facility that limit our options. In round 3 we're using a different belt surface that we think will help this issue--another reason to re-run the Hoosier 7.0 as a "control" tire to compare against the data from the first two rounds.

We appreciate all the interest, comments and questions, not to mention the encouragement. Myself, Dr. Bob Woods and Doug Milliken put a fair amount of volunteer time into the project and it's nice to know it's appreciated. In the last week the 100th FSAE team paid their membership fee! We'll keep trying our best.

Let's hope for a good, uneventful test next week.

[Edward M. Kasprzak]

Hi all,

The schedule at Calspan has changed, owing to one of their major customers needing the time on the tire testing machine next week for a high priority project. We're tenatively rescheduled for June 28-29. I'll post again when I know more certain information.

[Mike Cook]

I hear hoosier is coming out with a new tire very soon. May want to wait to test that tire as well.

Mike

[ben]

Edward makes some very good points as to why obsessing about the peak on a flat trac test is possibly not worth the effort.

As a design judge what I'd like to see is someone clearly stating and accepting that the grip level on the flat-trac isn't the same as that on the track, using the flat trac to determine a model of the carcass (which is what it actually does best) and adapting this carcass model based on actual friction coefficients determind from a separate model or track test.

Bearing that in mind, the rim width test would be the priority for me because it will have the biggest effect on cornering stiffness and therefore vehicle balance.

The peak slip angle is simply a function of the torsional and lateral carcass stiffness and the grip level. Use the flat trac to determine the first two and accept that it may grossly overestimate the last one.

Ben

[Paul Garcia]

Ed,

Could you discuss the difference in peak friction coefficients from the test results to what we see on the track. Are there measures you are taking in this third round of testing to account for this or is it just the "way it is?" Any plots of normalized lateral force show the test results do not reflect the true grip capability of the tire. The data shows peak friction coefficients higher than 2.5.

I apologize if you've explained this before.I am sure you have at least heard the question as I raised this to Doug Milliken previously and he was familiar with the topic. His suggestion was to use road coefficients lower than 1, say 0.7, to attain more "useful" data. Judges this year were very interested in discussing this very topic.

Does this seem like a reasonable method of scaling the data?

Would the "line of peak" slip angles be changed significantly in a way that this method may not reflect?

Paul Garcia
Jayhawk Motorsports

[Jersey Tom]

One issue is that the surface at Calspan is very clean, unlike the surfaces most FSAE cars run on. We think this causes the peaks to be somewhat high on the test machine. We've had a LOT of discussions about what to do about this--there are many constraints at the test facility that limit our options. In round 3 we're using a different belt surface that we think will help this issue--another reason to re-run the Hoosier 7.0 as a "control" tire to compare against the data from the first two rounds.

Paul - is this what you're talking about? Sounds like theyre going to try to alleviate the issue a bit.

Scaling the road coefficient down.. I wondere if that just scales down the overall coefficients of friction, and doesn't really change the shape (thus not making any peaks stand out).

[Edward M. Kasprzak]

Thanks for your comments Ben. I generally agree with you, and if you're trying to build a model of the tire carcass then I completely agree. Since most students are not, however, I'd like to take Paul's request and comment a bit on the relationship between the peaks in the data and the peaks on the track. The data at the peaks is more applicable than Ben may have indicated.

Let's start with the obvious: the surface on the test machine at Calspan is not the same as the surfaces that FSAE cars drive on. Of particular concern to us is how clean the test facility surface is and how much dirtier real surfaces are. FSAE cars tend to run in parking lots which are often very dirty, and even at the Detroit event (Ford proving ground) the outdoor surface in the vehicle dynamics area gets dusty and rained on, etc. The surface would "come in" if enough cars did enough laps to work in a groove, but this might only start to happen toward the end of the endurance event.

Thus the real surfaces have less grip than the Calspan test machine. To close the gap we're running Round 3 on a test surface at Calspan that we think will have less grip, although the Calspan test data will probably still show higher grip levels.

So is the data junk? Absolutely not! Rather, this is an issue that has been around for as long as there's been laboratory tire testing. And even if you managed to get good tire data recorded from on-track driving the surface changes with temperature, amount of rubber put down, clouds/sun, etc., so even then you'd be faced with the same issue. This is where modeling of the tire data begins to play a big role.

What you would like is for the peaks in the tire data to have the same magnitude as the peaks the tire gives on the car when it's operating on your particular surface. Thus, some scaling of the data is required. By far the easiest (and incorrect) way to do this is to take all the lateral force data and reduce it by some ratio so the height of the peaks match. The problem with this approach is two-fold. First, while the height of the peak may now match your on-track experience the slip angle at which it occurs has not been altered from the test data--this is not consistent with what really happens. Lower grip surfaces tend to make peaks at lower slip angles (think of it roughly as not having enough grip to twist the footprint of the tire around an axis normal to the ground). Second, by simply ratio-ing the lateral force down you have also changed all the cornering stiffness slopes. These slopes are, in reality, largely independent of the surface the tire operates on.

Allow me to put my Milliken Research hat on for a minute (I've worked there since '96) and talk about the MRA Nondimensional Tire Model. This model is included on the DVDs and is generally described by Chapter 14 of the Milliken's "Race Car Vehicle Dynamics". The Nondimensional scheme models tires on a more fundamental level than just trying to match the measured curve shape. One of the parameters that is modeled is the friction coefficient, defined as the peak of the lateral force curve divided by the normal load. This value can be multiplied by a surface friction value to scale the data up or down. If scaled down, the resulting tire data has the same cornering stiffness as the original data but lower peaks occuring at new (lower) peak slip angles. This is representative of how real tires behave. Done this way, the Calspan surface friction coefficient is defined as "1.0" and your surface may be "0.7" or something like that. If you try this with the FSAE TTC data you may find peaks occuring for reduced surface friction coefficients on curves where the raw data doesn't show a peak on the test machine.

MRA customers who use the Nondimensional model use this surface friction modifier to match the tire data to their surface (although it's always tricky to get it really close). This value is an input to the model supplied on the DVDs. While tuning this parameter requires a vehicle simulation, FSAE teams can do a reasonably good job by running your car on a skidpad and doing a simple four-wheel model (with load transfer, roll stiffness distribution, etc. to get the wheel loads right) and then adjust the surface friciton coefficient to match lap time. It won't be exact but it will be close (and you'll learn a lot in the process).

Now I'll take my MRA hat off and put my FSAE TTC co-director's hat back on....

I'm not sure how the Stackpole Engineering Services model handles surface friction coefficient variations. We may want to ask SES about that.

As far as Calspan goes, the surface we used in Rounds 1 and 2 is their standard surface. For most tires (passenger and racing) it does a very good job of replicating the grip levels on a normal road or track. It will always be a bit different, but it's usually close. Apparently the FSAE tires are so much softer and liked the clean surface a lot more than most other tires they've tested. Thus, the larger than expected mismatch, and thus the attempt in Round 3 to use a lower-grip surface.

This is a good thread with good questions and comments. One of the FSAE TTC goals is to expose students to tire testing--warts and all. You're identifying good issues and asking good questions. We threw everyone in on the deep end when this consortium was founded in the hopes that discussions like this would develop. Keep thinking and asking questions!

Edward

[Paul Garcia]

That was a great answer Edward, thanks so much for all the help.

For what its worth...an interesting point I learned while in Detroit was how load variation would effect all of these calculations. In calculating theoretical steer angles (so ackermann) and lateral force potential of the car we did typical load calculations that resulted in 50 lbs (or even less) on an unladen tire.

The comment on this load variation was that actual track performance may see variation in load as high as 50 lbs meaning that trying to interpret optimum slip angles and lateral force for that tire may not be very useful. I would imagine that in his applications (RCR = Nascar) this was true but our much lighter cars see less load variation?

We spent a great deal of time comparing the "line of peaks" to determine steering geometry (which is why we ended up just making it highly adjustable); the drivers said they prefered ackermann while the tire data along tends to indicate the reverse. The steering geometry is best left for another discussion but what did seem important was that good curve fits (like the MRA tire model) remove ALOT of scatter in slip angle and load maybe making this data very "theoretical" at best.

Just throwin those comments out there. You can keep um' or send um' right back...

Paul
Jayhawk Motorsports

[Jersey Tom]

There's been some looonng discussions on Ackermann, not too far back. Think the big thing was.. the data can indicate that a heavily laden tire likes being at higher slip angle.. but since the force vectors are no longer perpendicular to the direction of the car and you dont know where they are exactl you cant really draw any conclusions beforehand.

[ben]

Edward - great reply.

I wasn't suggesting that the team's would want to do a carcass model. Obviously having a peak on the flat-trac data and using the MRA tyre model would be the most convinient.

My point as a judge was that the current data doesn't have these peaks and not all the tyres will be retested. This gives the students two options:

1. Believe the data and appear puzzled that you're not achieving the artificially high friction levels on track.

2. Approximate the lateral and torsional carcass stiffness from the data and measure a friction coefficient on track. Use a stretched string model to approximate the true cornering stiffness and then modify the coefficients of the Pacejka model to fit your estimates, then use this tyre model in your simulations.

Of course just being given the peaks from the test would be easier and nice to have. But the student that does what I've suggested above would learn more about tyres and score more design points.

Ben