Damper set up, analysis and shaker rig work

[simonR]

I’ve read some interesting threads on damping and shaker rigs and this got me thinking about damper set up and performance analysis where budgets simply can’t stretch to proper shaker rig work.

I know that when I’ve really put some thought into the damping, spent some time on the dyno/revalving, done datalogged testing and then been able (or lucky enough) to get the damping really well set, the improvements in performance have been significant.

So for me setting dampers initially with respect to critical damping on the dyno, logging on track and reviewing as raw displacement data in the time domain, damper velocity histograms and in the frequency domain gives a fairly well-rounded look at things with a pretty small budget.

I’m really interested in the merit of building a single wheel “dyno” or ¼ car shaker rig to look at the complete “tyre-suspended mass-spring/damper” system of just one corner (clearly this would involve doing front and rear). My thought is that if I can go through this process and identify any resonance issues before track testing that’s considerable time saved.

I was thinking of a scotch yoke thing putting in say 10mm amplitude and going from say 1 to 15hz with a load cell between the input and the tyre. So it would record tyre contact patch load, position and also damper displacement (pot on the damper). Its not too hard then to do a wealth of analysis with this raw data.

I do realise this only gives a constant amplitude for all frequencies and cannot factor in proper heave, roll or pitch but it is worth it for say the look at high speed damping alone?

I’d be really interested in any experiences, thoughts and opinions. I know there’s a wealth of knowledge on this forum!

[Jersey Tom]

Does it need to be a physical rig, as opposed to a multi-body or other simulation?

[simonR]

No, i'm open to that but i do worry about the accuracy of inputs to a simulation.

i'm thinking that if i dont get tyre spring rates, stiction etc factored in right then i'm just going to kid myself with the outputs.

i guess i had just felt more comfortable with a physical test with this in my mind.

do you think a simple physical test would yield wortwhil results?

[Jersey Tom]

No, i'm open to that but i do worry about the accuracy of inputs to a simulation.

i'm thinking that if i dont get tyre spring rates, stiction etc factored in right then i'm just going to kid myself with the outputs

All models are wrong. Some models are useful. By your own argument, I could make the case that what's the use of a 1/4 car rig without the rest of the car, pitch and roll inertias, ARB's, etc etc... and "you're just kidding yourself with the outputs." Or that you're using a stationary tire rather than a rolling one, etc etc.

Models and simulations are useful, you can learn a lot. Can be much more insightful than a pure physical test, and often cheaper and easier. You can make a quarter car sim model complete with tire rates and friction literally in minutes in something like Simulink.. or in straight code in MATLAB over lunch with a beer or two for that matter. With everything parameterized you can then learn what tire rate does - what's the effect of having it vs not having it, or how your response changes with say a 10% tire rate change, etc. You can rapidly sweep through linear and nonlinear dampers. You can include or not include friction.

[simonR]

yes, i do accept the limitations of a 1/4 car rig you mention and i certainly didnt expect it would give a complete picture.

my thought was that it would provide an useful insight into tyre contact patch load variations in at higher frequencies where the body is effectively isolated from the road input.

[DaveW]

A very interesting question, one that probably requires a book to answer. I would make three observations:

If you are doing a "hardware in the loop" test, then use real hardware (something that F1 teams generally don't do, I afraid).

A simulation never supplies new information, if you are lucky it simply presents what you know in a more convenient form (with apologies to JT). To be fair, I do a lot a simulation, but mostly using model parameters that I identify using a hardware in the loop test.

A hardware in the loop test is cheap (relatively).

[Jersey Tom]

A simulation never supplies new information

I'd like to hear you elaborate on this point.. as it stands there I can't agree with it. Couldn't disagree more, really. Simulations - especially the simplest ones most focused on fundamentals - can be fantastic learning tools.

You could maybe make the argument that to make a simulation you already have to know the physics of what you're making. But I'd make the counterpoint that you may understand a spring, mass, and damper all individually quite well.. but combining them in a system level simulation gives much more insight than any individual piece can by itself.

[Greg Locock]

I'm currently building theoretical tires (ie modding the pacejka coeffficients) to derive specs to give to suppliers to stop SUVs rolling over while they will still handle and stop satisfactorily. I can't imagine a cost effective approach other than modelling for that.

[DaveW]

Simulations .... can be fantastic learning tools.

Absolutely. The fact remains that junk in = junk out. It is very easy to forget that.

I can't imagine a cost effective approach other than modelling for that.

Hi Greg, your right, of course. I hope, for your sake, that you are working with a validated model. Since many Pacejka coefficients are "abstract" terms, I trust you will also valitate your work with full hardware in the loop tests before releasing the results to the public.

In the not too distant past, I was asked to carry out a rig test on an F1 vehicle to produce raw data, which would be used to validate a mathematic model of the vehicle. I was also asked to provide tyre estimates (stiffness & damping coefficient). Some time later I started to receive questions about the performance of my rig. Eventually, I asked them to send me a sample output from their model. I processed this & discovered that they were using the wrong tyre properties. When I querying this, I was told they had measured the properties themselves & were confident in their estimates. The problem turned out be "secant" as opposed to "tangent" estimates - junk in = junk out..... (The rig must be wrong!)

[Jersey Tom]

I'm currently building theoretical tires (ie modding the pacejka coeffficients) to derive specs to give to suppliers to stop SUVs rolling over while they will still handle and stop satisfactorily. I can't imagine a cost effective approach other than modelling for that.

Set LMX = 0 but that's kinda cheating That's a fun project though having been down that road before (so to speak!)

Simulations .... can be fantastic learning tools.

Absolutely. The fact remains that junk in = junk out. It is very easy to forget that.

Applies just as much to track or physical rig testing as it does to math modeling though.

[Tim.Wright]

Well I'd echo what JT alluded to and say a rig is just a big lifesize model/simulation. The fact that you don't have any lat/long tyre forces, no rotating wheels, no inertial effects and a whole lot of other simplifications means that you are not actually observing what happens on the track but a simplified representation. Rubbish in - rubbish out applies to physical tests too. I've seen some pretty bad testing techniques that have made the resulting data practically unuseable.

Simulation models and rig tests are good at somethings and bad at others. They are not interchangeable. You can't solve everything with simulation, you can't solve everything with rig tests.

Regarding the Pacejka comments, I'm sure Greg will elaborate, but the good old Pacejka model isn't as abstract as its made out to be. While its often used purely as a curve fitting tool, it is what I'd call a mildly physical model in that it creates its forces and moments by modelling deformations in the contact patch. The coeeficients actually can have a physical relationship to the compound and construction of the tyre. See http://members.xoom.virgilio.it/adiaforo/epcjk.htm

[GSpeedR]

Quarter car test rigs have been built for both academic and commercial purposes, but usually the overall goal of the test is to validate a model of a component, or a control scheme, or durability test components, etc, rather than tuning. I agree that simulations are really the best tool to further your understanding as complexity can be added as needed, but can begin with the basics.

Most of the important low frequency behavior can't be reproduced without the appropriate degrees of freedom, which you won't have with a quarter car suspension, so I don't think it'll be as useful to translate back to your car. Where you may find some insight by building such a rig is if your suspension components cannot be easily modeled (the benefit of 'hardware-in-loop'). Building a rig is a large amount of work, but if you have no interest in modeling then it can be a good way to apply theory.

[Greg Locock]

Quick question - is damping a circuit car easier than damping a road car? Or is it just that relatively less time is spent on it due to lack of resources?

The reason I ask is that i have only seen one quarter car rig in the various companies I have worked at, all shock absorber tuning seems to be done done on full vehicles. Now that may be because they are fine tuning to the n'th degree, or it may be that pure vertical inputs aren't enough, or it may be that quarter car doesn't catch body motions caused by the other three wheels.

[Belatti]

It happened to me that I thought I learnt a lot with computer simulations. Then I was involved in the design and development of a 2 post shaker rig and began contrasting my computer results with "hardware in the loop" results.

All I can say is
1) postrig setups/results (in my case) are always better
2) "unlearn" something you thought you knew is a very difficult thing to do

Besides, the cost of buying/building a damper dyno is very similar to the cost of building a 1 post shaker. (not saying buying here because theres nothing in the south american market I know)

[DaveW]

Well I'd echo what JT alluded to and say a rig is just a big lifesize model/simulation. The fact that you don't have any lat/long tyre forces, no rotating wheels, no inertial effects and a whole lot of other simplifications means that you are not actually observing what happens on the track but a simplified representation. Rubbish in - rubbish out applies to physical tests too.

I agree with most of that. But there is a big difference between a mathematical model and a rig test. In the first case you are working with the limitations of the model you create and the parameters you adopt. A poor decision often goes undetected. In the second case you are working with (hopefully) a real vehicle, and you have the opportunity to improve your understanding of the vehicle, accepting always the limitations of the test.

I've seen some pretty bad testing techniques that have made the resulting data practically unuseable.

So have I. 7 post rig testing fell out of favour in F1 a few years ago, largely because rigs tended to be poorly calibrated and tests ill-managed. That is slowly changing, however. For example, I Currently "see" around 70 race vehicles a year, many of them returning customers. Incidentally, most also have their own mathematical modelling tools.

Quarter car test rigs have been built for both academic and commercial purposes, but usually the overall goal of the test is to validate a model of a component, or a control scheme, or durability test components, etc, rather than tuning.

Agree with that, too. We built a quarter car rig in 1980 and used it to understand & measure the properties of dampers, isolators, active control algorithms, etc. I should say that making it work properly (suppressing local resonances) proved very difficult. Building a four post rig was far easier. We used one of those extensively once an active car became available.