A new free vehicle dynamics resource - Dan's Vehicle Dynamics Corner

[ChassisSim]

It's good to see some sensible points have been raised in this discussion.

Tim to your question on suspension configurations that aren't modelled in ChassisSim. Definitely the lookup table is certainly an option and I certainly haven't ruled out but given the past experience I have discussed I've been very gun shy about implementing about it. That being said in the code in ChassisSim it's really easy to add new suspension geometry configurations and I've done this for suspension configurations such as the F458 rear end and the BMW Z4 link. In this regards a K&C rig is worth it's weight in gold because it allows you to validate what you have come up with. This was really useful when I was developing the F458 rear end.

One thing I do need to touch upon is ChassisSim modelling process. Tim the technique you discussed is pretty much on the money except for the whole fudge factor thing. Yes you do enter parameters such as spring rates, damper curves, pick up points etc to construct the vehicle model. Once you have your data from a flying lap that's where the tyre force modelling comes into play.

The basis of the tyre force modelling toolbox is that using the multi body vehicle dynamics model in ChassisSim to do a whole bunch of track replays to minimise the difference between simulated and actual lateral acceleration. This is where the transient vehicle model comes into play because you can now do the bumps you can get a much better idea of what the loads are. This is where the reliance on fudge factors falls away.

A good question that was raised is how good does the data need to be? Obviously the better the driver the less work you need to do. However given the nature of the track replays if you do have a so so driver you do have something to work with. Let me give you a war story. A couple of years I was doing some work in a junior formula in the states. When I went through the tyre modelling process it was predicting lap times 1s a lap quicker then what the drivers where doing on the home circuit. I made a mistake and trusted the race car drivers. When they went to Sebring they get blown away. After finishing tearing my hair out I reset the grip factors to where they should have been and all of a sudden the car was doing lap times in comparison to the front runners. The moral of the tale is you can get significantly down the road with race data. Also here is your game plan on how to do it,

http://www.chassissim.com/blog/chass...-in-the-blanks

The senior members of the ChassisSim community have found this to be a really useful resource and I recommend it to you.

Also make no mistake this is an iterative process. It will take a couple of goes to get it right. Also everything we have discussed here has been used in formulas as diverse as GP2, Sportscars and V8 Supercars to name a few. Given the results my customers get I think the validity of this approach speaks for itself.

Also just a word on accuracy and lap time simulation. I totally concur with Claude that your goal in simulation is not the correlation game but ensuring what the car does the simulator does. That being said one of the things we have found in the ChassisSim community is that the better the tyre model the accuracy tends to look after itself.

Great food for thought.

BTW - For the Australians and New Zealanders reading this - Lest we forget have a good ANZAC day.

All the Best


Danny Nowlan
Director
ChassisSim Technologies

[Z]

Danny,

Given that this thread is an advertisement for your wares (namely racecar-consultant, software-guru, etc.), I suggest you think more carefully before pressing "Submit". Your last post on page 20 is doing you no favours. (Edit: You just posted (top p21) as I was writing this...)

For example:

To this end there are 4 input variables that will effect roll centres, cambers and other suspension geometry variables. These are,

*Roll angle
*Heave displacement of the front and rear end.
*Individual wheel movements dictated by tyre squish/movement of the hubs on the tyre springs.

Firstly, your explanation of all this is very poor (ie. NO CLEAR DEFINITIONS!), which is not what is expected from a good consultant.

Secondly, and more importantly, it has been explained several times now that your 4-D tables are NOT necessary. See exFSAE at top of p19, and again with Tim near bottom p19, and exFSAE's quite clear explanation near bottom p20. Again, your seeming lack of ability to grasp these things does not reflect well on your VD knowledge or software skills.

(BTW, to me, "tyre squish" is NOT so much an input variable, but more of an output, namely the result of computed Fz tyre-load. At most there might need to be a very small 1-D look-up table for the tyre's (non-linear) spring-rate curve. One such table for each different tyre. So only two small tables if different tyres F&R.)
~~~o0o~~~

... the roll centres ...
You might get away with a 2D table resolving wheel and chassis movement. However this gets destroyed by roll angle so know we have a 3D table. However where this runs into trouble is the effect the roll centre locations are fundamentally effected by what both the left and right hand sides are doing. So we are back to 4 variables.

Again, meaningless gibberish! And again, see exFSAE's most recent post.
~~~o0o~~~

... [programming language] C ... multi dimension lookup table algorithm...
... the finer your grid gets the more searching you have to do ... and you pay for it in run time.

Piffle! The whole point of look-up tables is that while they might take up more memory, their time-cost is tiny and constant. Regardless of how many table entries you have, each one is accessed by the same simple "address call". Computing the address is likewise a trivially simple and fast operation, and equal for all entries. (Note that some "high-level" languages like Matlab might make this whole process ridiculously slow, but not so in the lower-level languages like C.)
~~~o0o~~~

The ultimate payoff of this is the results. Check out the image in this link.
...
... there are no grip or fudge factors here. This is what happens when you develop tyre models from real data...

Huh? NO FUDGE FACTORS???

As pointed out by Tim (middle p20), your method starts with the "real data" (ie. the 100% correct answer!), then you "develop tyre models" which are fed back into your Simulation, which finally gives you a result which is not quite the same as the "real data" that you started with...

(Edit: Danny's latest post (top p21) seems to confirm that it is all a huge fudge!)
~~~o0o~~~

Now, even though your Simulations do not output the same correct answers that are being fed into them, they still have some potential to be useful predictive tools.

But, IMO, a Simulation is ONLY USEFUL IF you very clearly state all the simplifying ASSUMPTIONS you make, and also give a very clear estimate of the size of ALL POTENTIAL ERRORS. (Claude, I suggest Optimum-G also do this. There is much more to this than just "deltas".)

To paraphrase Prof. Walter Lewin's much repeated comment on these matters,

"Any measurement, or Simulation, is MEANINGLESS (!!!!!), without knowledge of its UNCERTAINTY!!!"

(And, again, apologies to Prof. Lewy if I didn't quite catch the right tone of his shouting here. His "meaningless" is very high-pitched, and much repeated! )
~~~o0o~~~

Here are just two potential sources of error that should be clearly accounted for, beyond the most obvious one of (track-surface NOT-EQUAL test-belt-surface), so (road-tyre-Mu = roll-the-dice).

1. Does your (or Optimum-G's, ++) Dynamic-Simulation include information about the directions of the mass-distribution "Principal-Axes" of the various bodies? That is, do you realise that a car in very simplified Steady-State cornering (ie. flat-road, constant radius/speed/beta, zero-aero-forces, etc.) has Rearward-Load-Transfer when its longitudinal PA slopes down-to-front, but Forward-LT when the PA slopes up-to-front? Do you know the size of these potential errors (ie. changes to F/R tyre Fzs, and hence also Fxs and Fys)?

2. Does your Dynamic-Sim include the effects of all the gyroscopic-couples acting on the car from all its rotating components (ie. 4 x wheels, the engine and driveline components, energy-storage-flywheels, etc.)? Again, how big are these potential errors (ie. changes to ALL tyre Fzs/Fxs/Fys)?

I have calculated the potential size of these errors. Which is why I have not bothered buying any of the Sims out there!
~~~o0o~~~

exFSAE,

The "Principle of Virtual Work" is indeed a powerful thing. As a historic note, it was used at least as far back as the third century BC, by Archimedes in his proof of "The Law of The Levers" (where it amounted to a purely mechanical version of the First Law of Thermodynamics, namely "there is no free lunch").

Classical Mechanics has deep roots!

Z

[DougMilliken]

..."tyre squish" is NOT so much an input variable, but more of an output, namely the result of computed Fz tyre-load. ...

It is considerably more complex if the goal is to predict ride height with some accuracy (part of simulating ground effects aero). Tires are rarely standing up perfectly straight, and loaded radius is also a function of (to name a few):
+ Inclination angle, or camber in car axes
+ Fy distorts and typically shortens the tire sidewalls
+ Fx has a small effect for normal tires but is a major factor for drag race tires with "wrinkle walls"
+ Tire pressure

[dynatune]

Since we are talking about Software & Dynamics let me add my 1p.

I have spent a long time of my professional life working with Multi-Body-Analysis Tools (like ADAMS, DADS etc.) an developing cars. In these complex tools all of the
above described "problems" in the discussions are "automatically" implemented. The only - but huge - disadvantage that it is fairly difficult to filter out any
specific root causes for any specific effects. This "desire" of coming to a better understanding has created many tools - also commercially available - that each have
their claim to fame. Yet, all do fail to some extend more or less with respect to the "big" ones. To me it seems also quite clear that it is difficult for David to
beat Goliath, especially since Goliath is not sitting on his butt doing nothing at all. Even so, these models are still in many ways not as close to reality as I had
expected or would like to have.

So I came to the conclusion - and some others with me - that you either go for the full blown multibody car model and accept it's limitations (and with all risks
included if "driven" by a novice) or one goes for a as simple as possible model that allows a quick verification of most basic principles of Vehicle Dynamics. All other
software "in between" is just an attempt - each of them valid for some degree and probably very well suited for a specific objective but overall still an attempt that
comes only close to a cigar.

Furthermore reading the discussion on what parameters / effects should be considered in the software - and what effects they would/could have on the results - I have
been consistently surprised that the most important of them all are just not considered in many software.

The last 25 years of suspension design have not really brought mind blowing new insights in kinematics but certainly did teach us a lot on suspension compliance. Yet
many of the commercially available tools for suspension design do not consider them nor provide a way of estimating them at least to some degree. Whereas moving
pickup-points on the axis of rotation on for instance a lower wishbone does not affect kinematics it can most definitely change your compliant behavior of the
suspension quite drastically. Often the effects of compliance can be more than the effects of kinematics. Of course arguments are made that racing suspensions are
"stiff" meaning that there would be no need for looking at it, but at the end of the day even the most "rigid" suspension link is to some degree flexible. The huge
progress that has been made in the last 25 years is to start making these ever presented compliance work in favor for you (for instance classic example create Toe-In
under braking on a rear suspension).

With respect to Suspension Stiffness if for instance we take a chain of compliance in a non-bushed solid track-rod link with:

Upright Bracket Stiffness = 25000 N/mm
Rose Joint Stiffness = 25000 N/mm
Link Stiffness = 100000 N/mm
Rose Joint Stiffness = 25000 N/mm
Chassis Bracket = 25000 N/m

The total stiffness adds up - as springs in series - to 5882 N/mm. I might say nly 5882 N/mm. Getting to a number above 10000 N/mm is already quite challenging as one
can see quite easily.

Based on my experience with compliance I think that any vehicle dynamics tool that is not considering the effect of suspension compliance (in one way or another) on
the tire angles is simply not sufficiently correct. Especially on race cars with slick tires, that work with small slip angles, compliance can have a huge impact.
And as we know the tire slip angle is the mother of all forces on the car. So, if the mother of all forces is calculated by not considering the major part of all
suspension parameters that lead to it's creation ....

Cheers,
dynatune, www.dynatune-xl.com

[DougMilliken]

... The last 25 years of suspension design have not really brought mind blowing new insights in kinematics but certainly did teach us a lot on suspension compliance. Yet
many of the commercially available tools for suspension design do not consider them nor provide a way of estimating them at least to some degree.

Nice example, but your timing is off by a couple of generations... Olley and his co-workers recognized that compliance was important in the 1950's or earlier. They called it "deflection steer", so you might not find it in the literature under compliance.

Tom Bundorf gave a nice overview of the work at GM in his presentation at the SAE Automotive Dynamics and Stability Conference in 2000. A pdf version can be downloaded from the bottom of http://www.millikenresearch.com/olley.html or a direct link http://www.millikenresearch.com/Vehi...yRTBundorf.pdf

Deflection steer and rigs to measure it are noted on pages 15 and 35 and page 36 features a rendering of the Vehicle Handling Facility (VHF), perhaps the first unified rig for measuring suspension K&C parameters. It's described in Nedley, A. L. and W. J. Wilson, "A New Laboratory Facility for Measuring Vehicle Parameters Affecting Understeer and Brake Steer." SAE Paper 720473, 1972.

Summary -- understanding compliance (deflection steer) is one area where passenger car development was well ahead of race car development. I'm sure there are a few exceptions--for example, Chaparral may have tested on a compliance rig at GM in the 1960s? It seems possible although I do not know for sure either way.

[Tim.Wright]

I think he meant it (MBS simulation) give you a better understanding on how the compliance occurs within a given suspension - not the effects on vehicle dynamics. Prior to full 3D MBS models, I'd assume any tuning using compliance was based largely on trial and error.

Regarding the importance of compliance in racecars. I did a hand calc on the axial stiffness of a 300mm long steel tube, 19mm diameter x 1mm wall thickness (something FSAE sized) and found its the same stiffness as the rubber bushings used on the road car I have been working with in the last few years.

Though I would concede - it practically impossible to have a decent estimate on the compliance of a racecar without a full assembly FEM study (including compliant joints) or by actually building and testing the car (by which point its too late to simulate it).

[DougMilliken]

... Prior to full 3D MBS models, I'd assume any tuning using compliance was based largely on trial and error. ...

Olley and his associates started out using roll steer (and later, I think they decided it was a bad idea) , but were also using deflection steer.

"Deflection Steer Due to Lateral Forces" and "Deflection Steer Due to Aligning Torques" are included in the simple understeer budget calculations in Bundorf, R. T. "The Effect of Vehicle Design Parameters on Characteristic Speed and Understeer", SAE paper 670078, 1967. While this presents a very simple analysis method, it does hit the important points and is still a nice quick way to get a feel for a car.

The pioneers had to think...because they didn't have powerful computers!

[dynatune]

And another paper from Bundorf & Leffert in 1976 was called "The Cornering Compliance Concept for Description of Vehicle Directional Control Properties" but I can equally well live
with deflection steer. The only thing I was and I am still surprised about is that - since we all seem agree to agree that compliance/deflection is a very important contributor to vehicle
behavior - many of current commercially available software tools either for Suspension Design or for Vehicle Dynamics Analysis do not consider this phenomenon. I fully subscribe to Tim's
experience that a full steel tube/joint suspension link is hardly any stiffer than a bushing from a road-car. That by itself should make people wonder.

Cheers,
dynatune, www.dynatune-xl.com

[BillCobb]

In case you didn't know, there was this paper: SAE(840561) "Typical Vehicle Parameters for Dynamic Studies: Revised for the 1980's" that gives good information about kinematic and compliance suspension parameters (among other things necessary for good simulation). In case you need it for the 21st Century (not that much has changed), take a look at these pictures. Sign convention: + is understeer effect regardless of axle. Minus is the "O" effect word.

These days, these compliances (and 3 others) are managed by design, specification or audit. It's part of a chassis design 'recipe' that originates from competitive vehicle assessment or from packaging or suspension synthesis tools (synthesis meaning I want this value, what does it take to get it). In pawing thru these and other suspension design factors (like roll steer or stiffness or roll axis heights [determined from FY side-loading and FZ constraint loads) in cloud form, you can/could see the results from some perhaps prestigious car makers that would or should bother you, especially if you have measurements from multiple samples of the very same model. I might also point out that when you see a car with very different tire and or rim sizes and pressures front to rear, you know something has gone hay wire in their handling objectives. (I use plastic twine myself in my John Deere baler for this reason).

Anyways, when I hired in as a junior co-op student, the K&C machine was in Areaway 5 building, GM Milford Proving Grounds with air bearings under each when and with 2 air cylinders either in phase or out of phase supplying FY or MZ inputs. Ride and roll inputs were formed by pumping water into tanks hung from a support frame under the car. Outputs from HP XY pen recorders ! (on pre printed and labeled paper, thank-you). This was early 60's technology and could easily be duplicated by a creative FSAE team and stationed on a trailer at a track to make some money with. Shark Tank candidate.

Still, today, SAE and Vehicle Dynamics International papers are being published which completely ignore the roll, camber, and deflection steer wheel reactions. As a result, their predictions for vehicle understeer levels are totally ridiculous.

I also go along with the race car having the same or worse compliances. Measurements clearly show where things go wrong. Yeah, that's a splendid looking control arm you have there Laddy, too bad you mounted it to and undercarriage softer than yesterday's spaghetti. Power steering gears are the worst performers here, but their softness may be part of the 'insufficient understeer' gimmick. Surprizing amounts of MZ and MX deflection from wheel bearings and and camber deflections from strut rods.

Just about all of these compliances are linear with some hysteresis this my comments elsewhere that gradients are acceptable. However, the steering system can provide an ugly nonlinear plot that sometimes defies even the most sophisticated equation fitted model.

And THAT is the rest of the story. No reason not to apply this type of design recognition to a FSAE car. Cleanup in aisle 3 !!!

[dynatune]

You mean something ugly non-linear like this ? LATA.jpg
I could not follow completely the "where have you been" reasoning since I think we are fully agreeing that compliance induced steer effects should be analyzed on an FSAE car too.
Yet almost no one seems to offer the tools.

Cheers,
dynatune, www.dynatune-xl.com

PS: the graph explanation is meant for educational purposes to teach to "novice" students the principal difference between on-center behavior &
off-center handling. So don't even start ...