Force-based roll centres vs geometrical roll centres 2

[TalonDG]

I've been noticing more and more mention of a procedure that calculates the roll centre of a vehicle from a force-based perspective, rather than a purely geometrical one.

Circle Track in particular has been running a series of suspension modelling articles, and the author is adamant that the geometric roll centre is NOT the actual centre that the sprung mass rolls around.

There is also supposed to be some benefit to tailoring the roll resistance at each end such that the ultimate roll angle (from a force-based analysis) is the same at both ends - the idea being that you don't want the car fighting itself.

This procedure is supposedly pretty new (last 10 years or so) and came as a result of comparing actual recordings of wheel travel from suspension position sensors to predicted values (for a given amount of roll) from a kinematic analysis.

I have WinGeo3 from Bill Mitchell, and while my suspension position sensors aren't on yet, visually the car does not seem to following the predictions from the software - it looks like the outside is compressing less, and the inside extending more, than would be expected. See

http://farnorthracing.com/nats2003/

for examples.

I bought SAE983033 which seems to explain WHY the geometric analysis isn't enough, but it glosses over a lot of steps involved in actually calculating the nonlinear analysis. And if the Circle track article series described it, I missed that issue.

Does anybody have a reference that describes the steps involved in a force-based roll centre analysis in detail?

DG

 

[mloew]

DG,

There likely are not too many (if any) references available in the public domain on how to calculate the roll centers classically. This is because of the inherent non-linearities involved. In the SAE International Technical Paper 983033 Short-Long Arm Suspension System Non-Linearities and Analysis, the author used ANSYS. Most any of the 3-d multi-body solvers that Greg Locock presents in his excellent FAQ Software for suspension analysis (faq800-768) will allow for the actual roll center to be determined. ADAMS, DADS, and visualNastran 4D are probably the most used general purpose solvers.



Best regards,

Matthew Ian Loew
"Luck is the residue of design."
Branch Rickey

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

 

[NormPeterson]

No, DG, you didn't miss the issue. Bob Bolles has been very cagey about discussing the details of his approach, most likely because he offers a software that apparently does the work. Not only is this info not in the CT series, neither does it show up in his softcover book "Stock Car Setup Secrets".

I've been wondering about much the same things, but can't offer much beyond conjecture that he may be considering roll center migration in the software. His text spends quite a bit of time discussing that point. No idea if any force-based approach is included, but I'm thinking that for top-level NASCAR and circle track cars that they'd be using rod ends rather than bushings anyway. IOW, I think the kinematic model is probably a better approximation in these cases than for suspensions that include compliant bushings (Greg, feel free to slap me upside the head if I'm off base here . . .).

Don't I know you as 'DG2' on another forum?

Norm

 

[GregLocock]

Norm, why would I do that? You have been on the money so far...

I am a very lazy person. I have a piece of software here that tells me what the driver is feeling (sort of) and what the tyres are doing (sort of). The horrible mathematics that connect those two are left to the computer to sort out. (and if you believe that...)

One of the reasons that I distrust even FBRC is that no one can actually tell me what the sprung body inertial roll axis is in practice - it certainly is NOT the axis about which the sprung body rolls! I am being duplicitous here, I know about principal moments of inertia, so I could work it out - the point is that generally people worry about suspension roll centres yet don't have the faintest idea where their principal axes are (and neither do we, in practice, until just before we launch the car).

ADAMS allows you to plot roll centre migration in Y and Z (and that's another thing, why should the true kinematic roll centre be confined to the axle plane?) for both geometric /AND/ force based roll centres. For real suspensions in real manouevres with stiction and everything else these curves are very spiky.

If you can get hold of the following paper (I don't know if it is available to the public) it discusses FBRCs and gRCs. It is the reason I am so cynical about them.

Institut für Kraftfahrwesen der RWTH Aachen
Prof. Dr.-Ing. H. Wallentowitz
Diplomarbeit / Diploma Thesis
CAE investigation of the influence of suspension characteristics on the vehicle roll behavior for passenger cars
Cand.-Ing. Eggord Thomaschky
Matr.-Nr. 188 587
Betreuer: Dipl.-Ing. Peter Holdmann
November 1995

Eggord was the author, and does not have seemed to have published anything since, sadly.

I admit, roll centres are a nice idea, as a broad brush concept, but the nitty gritty seems to me to be so full of caveats that you might as well go the whole hog and work out the vertical loads at the contact patch, which when all is said and done is what you are really interested in, and the forces on the body.

As to the car fighting itself being a bad thing - I don't quite agree. If the car was not fighting itself then you couldn't get any rear axle steer happening, and rear axle steer is a very useful concept. The ONLY way you have to communicate what the front axle is doing, to the rear axle, is via the rigid body motions of the sprung body, particularly in roll and yaw.

Too many cups of coffee?


Cheers

Greg Locock

 

[NormPeterson]

I was in a bit of a hurry above, and I completely neglected to include my intended mention of said bushings being bonded and capable of some rotational spring constants of their own, thus generating end moments on what kinematics assumes to be frictionless pin connections . . .

Which kind of got me thinking that what you determine statically may not be quite what happens in reality, since suspension motion is, by definition, a dynamic event.

I think that Bob Bolles is primarily looking at avoiding twist over the length of the chassis. At least that's how the passage(s) in his book that refer to equalizing front and rear roll angles reads to these eyes. Maybe I'm just coming at this from a different point of view - more of a civil/structural analysis, with a vague idea of mass centroid axis and its offset from some roll axis as a loading basis. Anyway, it doesn't seem to me like the phrase "the chassis fighting itself" is a particularly good synonym for "chassis twist" or even "net chassis twist over the wheelbase"; consequently it may serve mostly to confuse people not raised on circle track racers.

FYI, and probably the reason that hard discussion of details is at least not readily available. In the section on "Roll Angle Analysis" (page 23, for those who may have this reference), he notes that "A method has been invented and patented that will accurately predict the front and rear roll angles in a racecar . . . The patent title is "Method of Land Vehicle Suspension Evaluation and Design Through Roll Angle Analysis", U.S. Patent number 5,723,782.

The horrible mathematics that connect those two are left to the computer to sort out. (and if you believe that...)

Actually I do. Just attempting to write a few spreadsheet applications gives me a hint as to the sheer volume of math that's potentially involved. And as avoidable as I can make it (I'm kind of lazy too, not to mention entirely capable of occasionally dropping the odd 'minus' sign).

Sounds like the Eggord paper was his doctoral thesis, no? Now I need to go back to that other thread where 'E' was mentioned, and the FBRC threads that came up on the search . . .

Too many cups of coffee?

Ummm, just what time of day (night?) were you posting that?

Norm

 

[yawpower]

I think the circle track article is referring to aligning the roll axis of the front and rear suspension, with the idea that if they are not aligned, something will have to flex for the body to roll.

What actually happens is roll steer, not flex of the body/mounting points. The end result of front and rear roll axis alignment may be a more predictable car, and that may be the benefit he is seeing.

I think the end result may be good, but the proposed reasoning is flawed.

I know I asked this in another thread that I started, but how do we quantify the effects of lateral roll center movement. To put it more simply, if two cars have the same vertical roll center location, let's say at the front of the car, but one has the RC on the vehicle centerline, and the other has it displaced 12" towards the inside of the turn, what would be the end result? How would the weight transfer calculations differ?

Paul Yaw
Yaw Power Products

 

[BillyShope]

"Yawpower," as the roll axis is displaced, angularly, in the X-Y plane, the effect of front-rear roll stiffness on roll couple distribution diminishes. If the roll axis were aligned with the Y Axis, as an extreme, roll stiffness would have no effect at all. So, I would multiply those terms involving the roll stiffness, in the equations on pages 682 and 683 of Milliken, by the cosine of the angular displacement.

 

[pmwltd]

As a point of interest, Bill Mitchell's software rolls the car around a point at ground level, rather than the indicated geometric roll center indicated in the graphic.

Dave

 

[c2s]

I've gone thru the equations, looked at Mr. Bolles patent and it appears that he has sucessfully patented the concept of a roll couple distribution of 1.0
Kevin

 

[NormPeterson]

TalonDG - Have you considered the effects of any sources of rising rate in that a given amount of weight transfer will compress the outside by a lesser amount than the inside is extended? In addition to (perhaps?) progressive rate springs this would also include consideration of the bump stops being encountered.

Norm

 

[mloew]

BTW, here is a link to the US patent by Robert C. Bolles, Jr., Method of land vehicle suspension evaluation and design through roll angle analysis:

http://patft.uspto.gov/netacgi/nph-...d=ptxt&s1=5,723,782&OS=5,723,782&RS=5,723,782

Best regards,

Matthew Ian Loew
"Luck is the residue of design."
Branch Rickey

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

 

[TalonDG]

The springs on the car are fixed-rate Hypercoils.

The motion ratio stays pretty much constant in the normal range of suspension motion.

The roll centres move laterally in roll, but on the order of a couple of inches.

And there are no bumpstops - the tire will contact the fender before shock travel is exhausted, and the ride height is set such that there is only light contact with the tire under the harshest bumps (the car cannot roll onto the tire)

My one experiment with Caroll Smith style silastoes was an unmitigated disaster. Instant terminal understeer.

Interesting that Mitchell's software doesn't roll around the roll centre! I wonder why he does that?

DG

 

[GregLocock]

"Interesting that Mitchell's software doesn't roll around the roll centre! I wonder why he does that?
"

Perhaps because cars DON'T roll around the roll centre.

Or, more accurately, the instantaneous axis of rotation of the sprung mass rarely coincides with the "roll centre".

Cheers

Greg Locock

 

[c2s]

Greg,
You said it in your first post, and you said it in this last post - and I agree - cars usually do not roll around the roll centers. The roll centers are defined by the suspension geometry (and hence the forces by Kennedy's theorem). Therefore, these are, by definition, force based roll centers. However, when you add an antiroll bar this classical analysis breaks down because the forces are no longer axial on the member that the bar is attached to (Kennedy's theorem no longer applies). Agree, disagree?
Kevin

 

[funnelguy]

Greg,
Mitchell's software does not make an attempt to roll the sprung mass around either the force based center or the geometric center, AFAIK. The software, as I understand it, restrains the lateral center to the vehicle's centerline. I was under the impression that the software did roll the chassis around the correct geometric height for the RC. I have seen other people post that the chassis is rolled around a point on centerline and at ground level. I have never verified this, perhaps others can offer their experiences?

There exists an option to download data from onboard aquisition systems into Mitchell's software so as to actually view the true real world displacements through his software. Sorry, I have no experience with this option.

I really just wanted to clarify my understanding of Mitchell's quirk. Sorry to muddy this interesting thread.

 

[mloew]

Kevin,

The two-force member assumption is not really valid for any real automotive suspension. Rarely are the suspension linkages loaded in pure tension and compression. Bushings will introduce moments about the bushing axis (rotational) and off axis (conical). The location of a spring, damper, and/or anti-roll bar on one of the linkages (other than the upright) will also introduce bending loads on the suspension linkages.

The vehicle will roll instantaneously around an axis that has only rotational velocity. The actual calculation of this axis is best left to a non-linear Multi-Body Dynamics analysis. There is no point in trying to use simple kinematics assumptions to determine the instantaneous roll centers and axis for anything other than a first-order simulation.

Best regards,

Matthew Ian Loew
"Luck is the residue of design."
Branch Rickey

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

 

[GregLocock]

Matthew, I'm 99.7% sure that Kevin is well aware of all that, evil grin.

The things I puzzle over are:

a)in the past people have found GRCs to be a useful concept, much quoted and discussed.

b) even now GRC heights for front and rear axles for most production cars are in the rough ballpark that tradition has defined (ie 50-150 for the front and 100-200 at the rear)

Now I can cynically explain (b) - if 'everybody' knows that the RCH has to be at such and such a height, then every suspension will be designed that way, EVEN IF an alternative layout might better meet the actual requirements. Or even more cynically - if it doesn't really affect things then it doesn't matter where it ends up.

I don't believe that latter point - Milliken includes some very helpful diagrams of the force build-up in the suspension during a corner, showing how forces in the arms are the quickest way of feeding forces into the body, hence setting up roll(followed by the springs, shocks and then a/r bar, from memory), which can be used to tell the rear suspension where to go, via the paths in the same order.

Cheers

Greg Locock

 

[c2s]

Greg,
I agree with your point about roll center height being bracketed by traditional designs (packaging constraits, etc) - which minimize vehicle performance variability via RCH concept.
Except for development engineers, lay folks and even a lot of auto enthusiast didn't really notice differences in RCH.
In lumped parameter math models, with 1st & 2nd order parameters, the transient (J-turn & Slalom) model's do fit the aquisitioned data fairly well up thru ~ 0.5g's. These models do tend to exhibit a little more damping. The parameter variability required for a lay person to notice the difference is typically beyond the model error computed from aquisitioned data.
Soooo ... I believe the concept of roll centers is useful, if for nothing else to tell you that the suspension geometry is in the ball park - lest you wind up with a boat
Finally, lumped parameter model vs ADAMS - I'll take ADAMS
ADAMS vs full vehicle NASTRAN model - I'll take NASTRAN
Kevin

 

So what about the argument that roll centers shouldn't move through the ground plane in transient conditions? How does that add into this?

 

[bhart]

I have also been following Mr. Bolles' articles in Circle Track and here are my interpretations of what he has written: Using his method of matching the "front" and "rear" roll angles will magically give you a "balanced" setup b/c the front end of the car will want to roll the exact same amount as the rear... he doesn't really go on to assert that this will reduce/minimize chassis flex, he just states that this is the way to do things, and if you believe him you can go faster.
After reading his patent, I am still just as skeptical about his method as when I read the articles. Not that I'm saying that his method is any better/worse than any other method of "optimizing" a chassis setup on paper, but I can't seem to find in any of the literature how he is calculating the CG height for a specific end of the vehicle (and if it was encompassed in the figures of his patent, niether of my browsers seem to want to let me see those). To me, leaving out this detail is a crucial flaw, b/c the method hinges on being able to look at a vehicle as two separate entities (front and rear), but how do you figure out the CG heights for a given end of the vehicle that will give you two sub-systems that are equivalent to the total system?
I am familiar with the method of determining an effective CG height as described in Milliken, however, they ultimately resort back to the standard rigid body (roll axis) equations to determine the equivalent CG height, "h_e," and they do not imply that this "h_e" is a magical equivalent CG height that can be used to calculate axle loads for any combination of roll center and roll stiffness at an axle, rather it is just a more convenient way of representing the roll moment at the axle for the purpose of analyzing the lateral force potential at the axle.
My belief (which I hope is actually just physical reality) is that there is no way to calculate a truly "equivalent" CG height for an axle that applies to all roll center/roll stiffness combinations, otherwise, there would have been all kinds of papers and text books teaching us this method as opposed the the seemingly more laborious roll axis model.
Mr. Bolles is supposed to include "computer examples" in upcoming issues of Circle Track, perhaps he will shed some light on how to determine the equivalent front and/or rear CG height.
If I were you TalonDG, I would stick to the more traditional methods(physically based) of setting up your racecar rather than completely selling yourself on "matching roll angles." Many of the people that I have spoken with who struggle with the same issue you have (force effects of the roll center) have resorted to keeping track of the jacking force coefficients (F_y/F_z) in each side of the suspension, and virtually, but not quite, forgetting about the roll center all together. Keeping track of the jacking coefficient is just like turning the front view situation into two separate "anti" situations typically used to describe the side view characteristics (anti-dive, anti-squat, etc).
Looking forward to that next Circle Track,
bhart