RACE CAR TETHERED TESTING

[BillyShope]

Section 8.2 of Race Car Vehicle Dynamics describes the constrained testing which was done before the development of the sophisticated fixturing now common to the major manufacturers. The car was tethered to a much heavier vehicle and observations and measurements were made as the two moved...side-by-side...at essentially walking speed and the car was steered away from the heavier vehicle.

For most race car developers, both the sophisticated fixturing and the constrained testing might be impractical. But, it recently occurred to me that some valuable information can be gained within the confines of a small shop with a very simple test setup.

The motion involved in the constrained testing is necessary to develop the tire slip angles, but, if all that is needed is information on the roll couple distribution, the motion is not necessary. Suppose that a horizontal chain is attached to the right side of an oval track car, exactly at the midpoint of the wheelbase and at a height equal to, or greater than, that of the center of gravity. It would be necessary, of course, to fully scale the car before this test begins. Now, with the scales removed, jack up the left side of the car, shorten the chain, and then ease the left side tires down onto wheel scales. After reading the scales and then calculating the right side loads, it is possible to determine roll couple distribution.

By attaching the chain at two different heights, it would also be possible to determine the roll axis height at the wheelbase midpoint.

 

[GregLocock]

A more sophisticated version of that, using chains at both axles, is used to find roll centre heights and so on at one of our facilities. Works well, it is quick, and tells you the force based roll centre height, not the GRC.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Warpspeed]

Yet another way is to place the vehicle sideways across a steeply inclined flat surface, perhaps at 30 or 45 degrees.

Gravity then produces a sideways force, simulating centrifugal cornering forces. The individual tyre loadings, wheel geometry changes, and body roll will respond appropriately.

With the steering rigidly fixed in the straight ahead position, roll the vehicle forward. Tyre slip angles can then easily be measured by noting how the actual wheel paths at front and back deviate from the theoretical straight ahead direction.

While not perfect, it is an interesting test.

 

[BillyShope]

Greg, I realized, after I posted, that an extension of this idea will give you the roll center heights.

But, before I get to that, I must correct something. In order to determine the roll couple distribution, it is necessary to first know how much of the overturning couple is roll couple. So, you must either first know the location of the roll axis OR you can, as I mentioned in the first post, locate the chain at two heights and determine the roll center height at the wheelbase midpoint.

Now, going on to roll center heights: If you rotate this concept 90 degrees in the XY plane and have the horizontal chain extending out the rear, you can determine the roll stiffness distribution. It is necessary, of course, that the rear tires be free to provide torque to the driveshaft and that the engine be restrained from rotation.

Once the roll stiffness and roll couple distributions and the roll axis location are known, the roll center heights can be calculated.

 

[BillyShope]

That's "determine the roll AXIS height at the wheelbase midpoint."

 

[BillyShope]

And, as usual for me, I'm assuming a RWD beam axle car.

 

[NormPeterson]

One of the magazines that caters to oval track racing ran an article describing a similar sort of testing arrangement, so I'd guess that at least a few teams are experimenting with it. I don't remember what they were specifically looking for, but I do recall there being some issues with lateral grip.

Norm

 

[BillyShope]

Further clarification: If only the first test is run, it is with the assumption that the roll center heights are equal. If the second test is run (chain out the back) and roll center heights calculated, it is necessary to iterate the calculations until the differences are acceptably small. This is an iteration of the calculations only, using the data from the single side and back chain tests. This iteration will, of course, give you a more accurate evaluation of the roll couple distribution.

 

[MovingChicane]

I build scaling platforms (surface plates), and I am also a programmer.

The approach that is being taken is flawed in the context that unsprung weight, lateral anti's, differential of roll and ride rates, and understanding that the roll center is a conceptualization, is being ignored. There a many more issues -- however I don't want to become long winded.

Raising or compressing a chassis on a set of scales will reveal the wheel rates. The roll rates are something different. Olley was very clear, and he did mentor Milliken on this very issue.

RC theory should be viewed as lateral anti's. Mr Shope, I know you have a very rich backround in this arena. In my humble view; RC theory and terminology should be changed to Direct Lateral (Linkage) Loading. That being said; those actual forces cannot be replicated on a platform.

Olley produced formula's for roll. They were different than ride. I use them in my programming. Milliken was kind enough to publish Olley's work recently. His solutions are correct, probably the best kept secrets of the century.

Live axles have an interesting mix of problems and are graced with another set of attributes. Especially concerning roll couple. Is the springing in roll to be viewed as a one wheel bump or will we view spinging to be reciprical. Roll couple distribution will vary considerably. Olley was here long before we.

I'm not trying to be a wet blanket but, it just isn't that simple. I myself am pursuing a solution. Much time and money questing for the grail. I think Olley will help those interested in this thread the most. SAE sells his book.

 

[BillyShope]

Moving, I agree that the terms we use might not be totally descriptive, but they do allow us to quantify the problem and provide useful working tools.

I have the book to which you refer and am truly honored that my very small contribution is mentioned in the same book that honors this giant of the field. That's as close as I'll ever get to that level of significance.

You have hit upon a limitation of tethered testing which can be significant and that is the distinction between sprung and unsprung masses. When, after my last post, I got around to working a "for instance" of the problem I posed, I found that I was forced to include both tests in order to determine all the unknowns. So, with the chain on the X axis and scales under the front wheels, it is possible to calculate the roll stiffness distribution. But, as you point out, there really should be proportional chain loads on the rear axle assembly and on the rest of the car. Fortunately, if you know the weight of the rear axle assembly, you can calculate your way around this small problem. In most cases, however, I don't believe the error is large enough to be of concern.

So, with the roll stiffness ratio known, there still must be two pulls, at different chain heights, in the Y direction. An iterative solution, however, is not required.

 

[BillyShope]

I've simplified the procedure CONSIDERABLY by eliminating the X pull and adding a third in Y. The three chain locations must be at different points along the wheelbase and no more than two of the chain heights may be equal.

I've written a little JavaScript program and have worked "back and forth" through it to verify the results.

Haven't done a thorough error analysis, but changing the scale readings, in the JavaScript form, to the nearest ten pounds makes very little difference. (Chain tension is 2000 pounds with a 3000 pound car.) Of course, accuracy benefits from widely separated chain locations. Doug mentioned chain locations at the extremes of the wheelbase. For my test calculations, I pulled at the midpoint and at 1/4 and 3/4 of the wheelbase with two pulls at 20 inches height and the third at 30.

 

[MovingChicane]

After re-reading your original post, it seems you are desiring roll couple distribution approximations to enhance your published (ThX + QRL) / (ThX - QRL) formula. Am I correct?

 

[BillyShope]

No, Moving, not at all. This is, instead, an extension of my interest in obtaining the maximum information with a minimum of expense and time. When we were little kids and ran across something new, like a big beetle or a toad, we'd poke it with a stick to see what happens. Well, that's sort of what I'm doing here. There are three unknowns, so I'm "poking" the car three times and observing the change in wheel loadings.

Presently, the NASCAR teams go to outfits like Goodyear and pay the big bucks to learn this sort of thing. (And a lot more, of course.) The simple test I'm proposing can be done by any Saturday night racer. Now, as to whether it is of value to him is another matter. Certainly, he is then able to calculate the wheel loads in a 1G corner, but, again, is that of any value?

 

[GregLocock]

More to the point- you aren't blowing your budget on a 7 point rig to learn the easy stuff.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[MovingChicane]

All good points, -- I would defer however as referring to it as "easy stuff". So many have got it wrong.

An independent suspensions conceptual RC, is really derived from track change. Chassis movement and a trammel bar will tell you that. Small track change means low conceptual RC (low anti), and vice-versa.

A live axle naturally has a higher conceptual RC (there are many ways of manipulating that value with linkages).
All well known; and reinforce the existance of a higher anti.

We are left then with an inclined axis of sorts (biased linkage anti's), that are then compensated for by biasing the front roll couple (wheel rates). This gets messier however with different CG heights (dependant on application), as different CG heights in relation to the above, influence the bias requirements, to obtain that steady state term -- "balance".

None of this requires any equipment other than the trammel bar. We are pretty close to a roll couple percentage. That CG value is the biggest unknown. Difficult to measure by the Saturday nighter, that however could be deduced in an appropriate fashion.

Would we need to even allocate a budget for this? As pointed out though; the thinner the air, the easier it is to defend "apirations".

Most Saturday night racers however are very content to pursue trial and error, and ignore any and all engineering approachs.

I do hope however, that my post has helped you in achieving the answer you initially desired, at the top of the page.

 

[BillyShope]

The three lateral chain "pulls" have provided the information quite nicely, Moving. And, since the chain tension simulates the inertial force with the CG located at three different positions, the results are quite "real." My Internet supplier provides me with a small free site, which I have devoted to helpful [?] tips for the amateur non-engineer racer. The aforesaid program is featured on the first page:

http://home.earthlink.net.~whshope