Roll Center Position & Migration 2

[traindriver]

Any thoughts on roll center position and migration during a roll/dive event. I am working with a non-parallel 4 link suspension typically found on front engined, rear wheel drive US autos. I would think that the front roll center position should stay as "calm" (centered) as possible during a dynamic roll/dive event. This should stabilize the dynamic wheel loading and not cause any strange driver feel. If the R/C did migrate to the outside or inside, say 400 mm, what effect would this have on the dynamic wheel loads at the inside and outside tires? The rear R/C position is assumed to be fixed because of the live axle/panhard bar arrangement. This is a high speed maximium cornering grip application.

 

[hpfreak]

I think of horizontal roll center movement like an imaginary number. It's significance is questionable and is mainly used to calculate the vertical position. The vertical movement is the real concern and the one that shows up in all the loading equations. Once you get into roll/dive (turn-in?) the horizontal movement of the rc may be hard to control if your paying attention to the camber and jacking action at the tire due to steering axis, castor, etc.

 

[traindriver]

hpfreak

Thanks for the reply. Sounds like you and I may be in the same business.

Would not horizontal movement towards the outer wheel reduce the lever arm length from the roll axis to the outer contact patch. This would increase the loading to the outer wheel and increase roll. Belief in my business is that this is what is happening. I have seen no data to prove this but in theory is seems reasonable. Asimple free-body diagram confirms the actions.

I am just seeking varification.

Thanks

 

[hpfreak]

I think a diagram should prove there is no difference at the tires. The inputs dont change(cg*cornering acceleration) and the outputs dont change(tire reactions)It may change the motion of the chassis from pure roll to combination roll and bump. Because of the asymmetric roll axis.

I dont think the horizontal position will effect tire loadings. I can't say for sure. Less movement is considered good b/c what are the chances it moves horizontally without moving vertically?

 

[SWB]

Move the roll center 2 feet to the left of the left front tire in a left turn and both front springs will be in compression.

Conversely, if we move the roll center 2 feet to the right of the right front tire in a left turn, we will have both springs in extension as the car attempts to trip over itself.

IF it is centered, the right front will compress and the left front extend.

At least in theory...

Since we know how we would like the situation to be (both springs holding equal amounts of weight), we can think about it and go from there.

Remember that banking will play a part in how the car rolls (or doesn't). The more banking, the less roll. You need to use some method of finding the compression/extension of the springs to actually know what the car needs for a given racetrack.

Sean

 

[GregLocock]

OK: I've tried to write this three times now, and failed miserably. v4 follows:

In a bounce or pitch event you can see where the lateral location of the roll centre is. Essentially it is the single point at which all the loads from the axles suspension can be reacted by a pure force, rather than a force and moment (thinking in front view only). On the two front and two rear suspensions that I've checked the lateral motion of this point during a +/- 100 mm bounce test was less than 6 mm for three of them, and one of them had a spike to 50 mm, in a ride height change of 10mm, that I don't understand, but was probably due to a jounce bumper engaging.

Its location laterally is controlled by different kinematics side to side, due to non-square arms etc, and by different compliances, such as jounce bumpers.

This point still exists during a roll type event, but its location will also be affected by the ratio of sta bar torsional stiffness to roadspring torsional stiffness. I don't have an easy way way of working the effect of all these things out on the instantaneous roll centre.

The practical effect of it moving about is that the body roll angle vs latacc line will have a curve in it, or if you prefer the wheel load will not vary linearly with lat acc. This is a bad thing for the driver as it reduces the linearity of the vehicle.

Well, (a) is that right and (b) does that help? I've looked in Gillespie and he says nothing. The rest of the library is out at the proving ground where it will be more use!

Cheers

Greg Locock

 

[harrisj]

Interested in your correspondence re roll centres. I did quite a lot of work on this back in the 70s on F2 / F3 race cars here in the UK. (Everyone knows European cars go round corners better than US ones!)

For what it's worth, the conclusions I reached following theoretical and empirical experimentation were:

Vertical position of RC is immaterial as long as it doesn't move about unpredictably.

Lateral (transverse) position is largely unimportant in most conventional suspensions

Matching camber change to roll angle so that tyre contact remains maximum is more important than RC and best compromise should concentrate on this rather than 'ideal' RC position.

Non-linearity of latacc / roll angle is not a problem as long as it is a continuous function. Bump (jounce) stops produce a discontinuity which most drivers discover just before they fly off into the green bits.

Hope you don't mind a Brit muscling in on your discussion!

Regards - John H

 

[GregLocock]

Muscle in. Well that is 3 different nationalities (at least) and 3 different suspension cultures.

On an F2/F3 how much jounce/rebound travel did you have? What was the track (roughly)? how long were the suspension arms (roughly)?

The cars I am talking about are ordinary mass market large sedans, for which we have 100 mm jounce, 135 mm rebound (aspirational goal) the lateral arms in the lower triangle are roughly 460 mm long and the track is 1626 mm

Obviously from this we get large deflections and so we'd expect to see more roll centre shift than on a modern circuit car with a very limited suspension travel.

What interested me was that we don't deliberately optimise the roll centre lateral position, but our suspensions seem to end up with small lateral displacements, so I suspect that it effectively shows up in another curve.

I can also see why the migration of a point with relatively small moment arm is of lesser concern than camber control, if you have big fat tyres.

Cheers

Greg Locock

 

[GregLocock]

Well I think I can guess who you work for - do you use ADAMS as well? Which module? We use a customised version of Chassis, called Pre.

The reason I ask is that I have had no luck getting a measure of the roll centre location during full vehicle simulations, and wondered how you did it. I haven't asked anyone else.

I can see the Saab guy's point, but without a way of looking at the position during an event like that I can't tell if the cars that are nice at the limit actually do this or not. Having written that I'm not even sure which cars are 'nice' when lifting wheels - it's not a manouevre I can rememeber assessing. It would be very interesting to plot the roll centre location during a fishhook or J turn, as we have one variant that lifts a front wheel, and another that lifts a rear wheel, and I can't rememeber what the other one does.

Why did you think it was rubbish? because it sounded like a idealised theory ? Or is it that you think that roll centre locations are not especially important?

Cheers

Greg Locock

 

[GregLocock]

SVSSA, I've actually been shown an idiot proof solution, which I can build into my templates, but thanks for the lead. I try not to mess about with the .adm file directly, as that is a lot of extra work. Our standard templates are much easier to use and less error prone than hacking the adm.

Moving the roll centre outboard seems to me to be easy (at first glance), since you are rolling the axle anyway, so if there is a tendency for it to move out in jounce and in in rebound then both sides of the car will conspire to move it outboard together. To take the Saab theory one step forward you might even want to move it down at the same time.







Cheers

Greg Locock

 

[mloew]

All of the methods discussed so far to determine the roll center are kinematics approximations assuming that the supporting linkages are two force members. This method works fine for SLA and struts as a first-order analysis tool. As soon as there is a spring/damper on a control arm the assumption of a two-force member breaks down. Busing compliance also disturbs the assumption of two-force members.

The following SAE Technical Paper describes the situation:

983033: Short-Long Arm Suspension System Non-Linearities and Analysis . This paper is available form

http://www.sae.org.

The actual roll center is determined by the force vectors. This is evident in comparing the roll center from dynamic simulations and K&C data to the original kinematics estimates.
Best regards,

Matthew Ian Loew

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

 

[GregLocock]

Well, actually when I measure the roll centre ADAMS uses the force based method to find it. I did find a moderately alarming paper that compared results for the force based method against the geometric method, and found that even for common architectures the values /and the trends/ derived by the two methods disagreed, particluarly as the vehicle rolled.

I think this is another argument for finding better ways of describing force transfer across and along the car than roll centre heights. Cheers

Greg Locock

 

[c2s]

Greg,
That's an interesting idea. The roll center ht. is a usefull tool for initial layout and packaging. For handling quantification, ADAMS inclusion of non-linearities clearly yields the best solution. However, perhaps a definition of roll center variance (other than just height) would be usefull. One might measure the the roll center as a vector & angle variance with lateral acceleration. It would one way to try and quantify the nonlinearity of the system.
The question with such a scheme is would this dynamic roll center parameter provide meaningful trend data.
Kevin

 

[GregLocock]

We measure it sometimes, it's a standard output for k&c roll analysis (I think). I'm going to start including it in my J turns analysis because I run hundreds of those things and they should build up to give me a better feel for it. In J turns we look directly at the vertical wheel loads and the time relationship between the four wheel loads. I haven't done it yet as I'm embroiled in other things, bu Cheers

Greg Locock

 

[GregLocock]

I have set up an SLA so that at 5 degrees of roll the geometric roll centre has moved right over to the outboard contact patch, as 2SlowJoe suggested. This may lead to better control as the inner wheel lifts off. The pay-off is that the track change in bounce and roll is excessive. This is partly due to the high camber gain, which I wanted to get the tyre square to the road at 5 degrees roll.

Rolling radius of tyre 300 mm
static camber 0 deg
Y Z
centre of wheel -770.0 550.0
hence contact patch -770.0 250.0
Lower control arm to body -350.0 460.0
Lower control arm to spindle -750.0 484.2
Upper control arm to body -460.6 755.0
Upper control arm to spindle -575.6 800.8

The roll centre moves on a fairly straight line with increasing roll, so it should give a linear feel. Has anybody got a good way of evaluating this, analytically?

My thought was to drop this into a constant radius test and compare the SWA and SWT vs latacc curves. Hopefully this will be less non linear. But, apart from the tiny minority of people who drive around on two wheels, why would you bother? Cheers

Greg Locock