Jump to content


Photo

Simulating Rig Testing v2


  • Please log in to reply
32 replies to this topic

#1 Lukin

Lukin
  • Member

  • 1,930 posts
  • Joined: January 03

Posted 30 September 2011 - 14:18

To resurrect this with supporting roles played by this guy and that guy.

How did you go Ben? Has anyone else had any experience with simulation representation of a 7 post rig?

This paper covers the ‘construction’ of a model and there are a few programs (Car Sim, Chassis Sim, ADAMS) that lay claim to being able to do it (I’ve used one of the aforementioned programs but am not sure if I am even starting to use it correctly!).

What do people think of these sorts of programs? From what I have read the 7 post is good for determining chassis issues (flex, friction, poor construction etc. which the model would never do) as well as optimising (or at least providing options and combinations of) for the spring/damper/ARB/camber/tyre pressure combination. But is modelling a car in this environment a good start? Surely it would be better than nothing?

In general there doesn’t seem to be much written about the interpretation and analysis of 7 post testing. I have to admit that most of the information I have been able to uncover has come from Dave W in various forums. That’s no surprise given I would think he would be the most well known and respected 7 post operator going around?

The ARC paper proposes minimising the pitch response to a heave input (I’ve heard this referred to as pitch de-coupling). Given there are other inputs (roll, warp) is the thinking that if you have a car that is good (that is such a poor term for it) to a heave and pitch input that the warp and roll modes would (more or less) take care of themselves?

A note from one of the threads mentioned earlier from Dave W:
“- A digressive damping style for dampers attached to the steered axle will decrease steering time constant, may introduce a noticeable transient under-steer, and will increase the rate of heat input to the tyres.

- A digressive damping style for dampers attached to the driven axle may compromise traction, may introduce a noticeable transient over-steer, and will increase the rate of heat input to the tyres. The transient over-steer may reduce the magnitude of steering inputs (& hence may reduce turn in speed loss) on the entry to high speed corners.”

How do you measure heat input into the tyre? I can’t imagine that you would create that much heat on the rig that there is a noticeable gain in the tyre temperature or pressure over a run. Therefore I would assume you look at the tyre response?

If you were to create the model from the paper you would also be able to look at response of the sprung mass, each hub and each tyre for different inputs. From this you could see the tyre response. I would think changes in tyre stiffness against the spring stiffness would change the heat input to the tyre.

Would you want the response of the front and rear tyres (or hubs) to be at the same frequency? I remember reading once that if you do you risk having a pogo stick. Though another part of me thinks that if you have a range of frequencies seen on the track you could end up with wildly different car behaviour at different depending on the frequency range the car is subjected to.


Advertisement

#2 Fat Boy

Fat Boy
  • Member

  • 1,848 posts
  • Joined: January 04

Posted 03 October 2011 - 16:20

This is easily the best topic on the board right now, but it's tough to get any response because it's the questions are difficult.

I haven't been on a 7-post rig in a couple years. I guess it's a function of the economy and the teams I've been racing with. Our time/money seems to be better spent at the track.

I think that it would be very difficult to get good information from a simulation of a rig test (which itself is a simulation of the racetrack). It just seems like there are so many things going on there that are difficult to quantify that it would surprise me if you got a lot of applicable information. I guess you might be able to get some directional questions answered, but I really don't know. You would almost have to go to a rig and do a series of tests then work on the simulation until similar inputs get similar responses. Once you could match the actual rig data with your simulation, then maybe you could trust your simulation. It wouldn't be an easy task.

Incidentally, without going into too much detail, I've made some damping changes based off Dave W.'s philosophies that he's talked about with good results. This place has some good people roaming around on it.



#3 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 03 October 2011 - 22:18

Thanks, people, I'm flattered.

Time domain simulated rig tests are certainty able to help with car set-up. Adams (with a suitable integration algorithm) & RaceSim certainly work, & I imagine that the other routines you mention would also work. But it is advisable to validate the model using a multi-post rig test & K&C tests on a real vehicle (tyre characteristics, effective masses, installation stiffness are likely to be in error). In fact, a simulation of a rig test likely to be more valuable than a real one, simply because it is possible to explore the effect of non-vertical loads.

The organization of rig tests & processing of results is a dark & largely unpublished art. The reasons are partly commercial (race teams are unlikely to want to publish a winning formula), but also because a testing organization is unlikely to know in full the reason for a successful outcome. In short, success usually happens as a result of good cooperation between testing/modelling team, the race-team & the driver. There is, however, one analysis procedure that has been published. This entails processing in the frequency domain wheel platform, hub & body accelerations (only). I don't have a reference to that work, so I would be grateful if anybody can help to locate the reference (TC3000?).

You asked about "tyre heat rate".

Here is a plot of CPL against reconstructed tyre velocity (deflection rate) for both axles of a race vehicle. The regression slopes ("Slope") are estimates of tyre rate, whilst the areas enclosed by the trajectories represent the work done by the tyre per cycle. Two successive trajectories are plotted in each case to give subjective confidence in the reliability of the results. Note that both trajectories are "rising rate".

Frequency response diagrams of the same variables yield more information. Above 2 Hz, they shown that the stiffness varies slightly (probably with amplitude), whilst the damping remains mainly hysteretic - the work done is Mag*sin(phase).

The work done can also be computed, & then plotted, for the various elements of the vehicle - Tyres, dampers & D/F actuators, as well as the overall power input to the rig, as shown here The magnitudes are scaled for an arbitrary 10 mm/sec, & the legend shows both peak & average values over a frequency range 3-30 Hz. It is the average values for the tyres that is used as a (relative) measure of work done.

Finally, the total work done can be used to scale the separate functions, as shown here. It provides an interesting view of the suspension set-up, which I prefer not to comment upon at present.

Edited by DaveW, 04 October 2011 - 03:56.


#4 Lukin

Lukin
  • Member

  • 1,930 posts
  • Joined: January 03

Posted 04 October 2011 - 11:05

Thanks for the reply guys. It’s so far out of my ballpark at the moment but if it didn’t help no one would use it!

FB, I agree on the damping directions for sure. The digressive vs. linear vs. hybrid is something I've touched on but probably not understood well enough.

TC3000, if you know where to look that would be great!

Some of my observations from Dave’s pics.
First question is what is the input? I would assume sinusoidal heave?

Here is a plot of CPL against reconstructed tyre velocity (deflection rate) for both axles of a race vehicle. The regression slopes ("Slope") are estimates of tyre rate, whilst the areas enclosed by the trajectories represent the work done by the tyre per cycle. Two successive trajectories are plotted in each case to give subjective confidence in the reliability of the results. Note that both trajectories are "rising rate".

The tyre spring rate ratio is ~ 1.1. I read somewhere that is a good starting point.
The front tyre has a bigger area (it’s hysteresis like?) On a later pic it shows more work done by the front tyre which matches up with that. That would imply that the car setup has an effect on the effective tyre stiffness, not just camber, pressure etc?
The repeatability is excellent!

Frequency response diagrams of the same variables yield more information. Above 2 Hz, they shown that the stiffness varies slightly (probably with amplitude), whilst the damping remains mainly hysteretic - the work done is Mag*sin(phase).

I looked at that for 5 minutes before I got any semblance of what was going on. The phase difference being bigger for the front tyre gives the ‘hysteresis’ look of the tyre; the load application (input) and motion (output) are out of phase which explains the difference between the loading and unloading phase.

The work done can also be computed, & then plotted, for the various elements of the vehicle - Tyres, dampers & D/F actuators, as well as the overall power input to the rig, as shown here The magnitudes are scaled for an arbitrary 10 mm/sec, & the legend shows both peak & average values over a frequency range 3-30 Hz. It is the average values for the tyres that is used as a (relative) measure of work done.

The peak Frequency of rear dampers is a bit smaller (~4.8 Hz) than the front (~5.2 Hz) (Aero car?)
Front and rear tyres are reasonably well matched in terms of frequency. Is this desirable? The average is skewed however, with the front tyres working harder. The opposite ratio is found in the dampers; with the front dampers working harder than the rear dampers on average. Is that a conscious decision to look after rear tyres (rear drive?) or something to work on?

I'd be interested to see this for the dampers front and rear (that's not a request). I read someone that the phase angle is a tell tale on what you work on, 0 phase angle implies all spring and 90 degrees implies all damper.

"The phase angle between the push rod loads and the CPL at its maximum response is a measurement used to distinguish damping forces from spring forces at the respective corners of the car. A force, in phase with displacement, is considered a spring force (displacement sensitive), and a force, 90 degrees out of phase, is considered a damping force (velocity sensitive). For example, a value greater than 45 deg. indicates that the damper is a larger component of the total force. Conversely, values less than 45 deg. indicates that the spring is a larger component of the total force. A 45 deg. value would indicate equal force from damper and spring."

Finally, the total work done can be used to scale the separate functions, as shown here. It provides an interesting view of the suspension set-up, which I prefer not to comment upon at present.

The first question is where does the extra energy go? From 2-4 Hz it's close to 100% but then drops, comes back slightly and drops again.
The front and rear work dampers work differently depending on the input frequency (I assume that’s what it's saying)? Would that would be hard to tune at a track if the car sees a range of input frequencies.
Assume the low work rate at 1-2 Hz for the front tyre come from the low amplitude at those frequencies seen in picture 100212010263?


#5 NeilR

NeilR
  • Member

  • 459 posts
  • Joined: October 09

Posted 04 October 2011 - 12:23

Where's GregL....he knows Adams

#6 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 04 October 2011 - 14:43

First question is what is the input? I would assume sinusoidal heave?

The tyre spring rate ratio is ~ 1.1. I read somewhere that is a good starting point.
The front tyre has a bigger area (it’s hysteresis like?) On a later pic it shows more work done by the front tyre which matches up with that. That would imply that the car setup has an effect on the effective tyre stiffness, not just camber, pressure etc?

The peak Frequency of rear dampers is a bit smaller (~4.8 Hz) than the front (~5.2 Hz) (Aero car?)
Front and rear tyres are reasonably well matched in terms of frequency. Is this desirable? The average is skewed however, with the front tyres working harder. The opposite ratio is found in the dampers; with the front dampers working harder than the rear dampers on average. Is that a conscious decision to look after rear tyres (rear drive?) or something to work on?

The first question is where does the extra energy go? From 2-4 Hz it's close to 100% but then drops, comes back slightly and drops again.
The front and rear work dampers work differently depending on the input frequency (I assume that’s what it's saying)? Would that would be hard to tune at a track if the car sees a range of input frequencies.
Assume the low work rate at 1-2 Hz for the front tyre come from the low amplitude at those frequencies seen in picture 100212010263?


The input (at the wheels) was a sinusoidal with a peak velocity just shy of 160 mm/sec. What happens above the platforms depends very much on the vehicle & tyre characteristics & suspension set-up.

Tyre spring rate was ~ 1.1 (or, as I prefer rear/front, ~0.9). The data was from an old F1 rig test from 1997, so tyres were single supply Bridgestone's. A "good" ratio for a neutrally ballasted car of the time would have been around 1.3, so the tyres were not a good match to the vehicle (They would have been closer to a good split for an OzV8, as it happens). It is true that tyre stiffness is affected by car set-up - It is also true (importantly) that car set-up is affected by tyre stiffness.

The work done plot shows that, for a simple heave input, most disturbance energy occurs at the heave mode (4.8 Hz) and, in this case, the rear damper(s) take care of this. The front dampers do (relatively) more work to control the pitch mode (7.6 Hz). This a typical for a mid-engined, rear drive, aero vehicle.

The "missing" energy indicated in the last plot in the 3.5 - 20 Hz region is probably caused by the sprung mass not acting as a monolith (& therefore absorbing energy), for example "sprung" structure (probably radiators) and the driver (ballast). In the 3.5 - 8 Hz region, the peak vertical acceleration probably exceeds 1.0 gn (relatively). Other things can go wrong, however, in one previous case the "lost" energy exceeded 30 percent at the heave mode - which turned out to be an engine (stiffness) problem. An Indy engine (when such things were free) had a similar problem.

I normally ignore what happens at frequencies below 2-3 Hz, because here the displacement is computed from not much (acceleration) minus not much at all. (i.e. the calculation is poorly conditioned).

Edited by DaveW, 04 October 2011 - 20:59.


#7 TC3000

TC3000
  • Member

  • 1,021 posts
  • Joined: September 09

Posted 05 October 2011 - 17:04

Sorry, I don´t think, that I can contribute in terms of the paper mentioned by Dave, but here some other info, touching on the subject.

more or less the same as Dave allready said, with some different data shown
Multimatic presentation VD expo

Öhlins approach to Rig test analysis


some papers/dissertation´s, which some may (or may not) find interesting, on the wider subject of the thread

Track Simulation and Vehicle Characterization with 7 Post Testing

Vibration analysis of an experimental suspension system

Road profile excitation on a vehicle measurements and indoor testing using a four-post rig

CONTINUOUS ACTION REINFORCEMENT LEARNING APPLIED TO VEHICLE SUSPENSION CONTROL

KINETO-DYNAMIC ANALYSES OF VEHICLE SUSPENSION FOR OPTIMAL SYNTHESIS Dissertation

HIGH PERFORMANCE DAMPER OPTIMIZATION USING COMPUTER SIMULATION AND DESIGN OF EXPERIMENTS Dissertation [may take a while to load]

Development and Evaluation of Vehicle Suspension Tuning Metrics Dissertation




#8 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 06 October 2011 - 10:09

Thanks TC.

I have on more to add to your list Henri Kowalczyk. He uses the process I had in mind in his paper (but doesn't reference the source).

#9 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 08 October 2011 - 22:02

I have used the 4 poster rig in ADAMS. I've not had much luck although at 5-20 Hz it does produce usable results. Directionally the results are ok, but the correlation back to the test rig is not fantastic.


In real life we have some measured road profiles, and hub and body accelerations.

So in theory we can put the time history of the road into each actuator, and then check the hub and body response.

This can be made to work albeit it takes a while. But trying to do the same in ADAMS just doesn't work. The tire to post interface is too weird.

#10 cheapracer

cheapracer
  • Member

  • 10,388 posts
  • Joined: May 07

Posted 09 October 2011 - 02:20

Welcome back Greg.

#11 Tony Matthews

Tony Matthews
  • Member

  • 17,499 posts
  • Joined: September 08

Posted 09 October 2011 - 08:19

Welcome back Greg.

:up: I was going to post that last night, but it seemed a bit presumptuous, coming from me...

Edited by Tony Matthews, 09 October 2011 - 08:20.


#12 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 09 October 2011 - 09:35

I have used the 4 poster rig in ADAMS. I've not had much luck although at 5-20 Hz it does produce usable results. Directionally the results are ok, but the correlation back to the test rig is not fantastic.


I presume you are talking about a road vehicle. In which case your comment suggests that you didn't have good correlation of heave & pitch mode response. Possibilities are, I suppose:

1. Integration algorithm. Adams offers several options, not all of which give "good" results, I am told. I believe that DASSL can be made to work and produce reasonable correlations when compared with a parameter based model & a rig test (in racing application, to be fair).
2. Incorrect "dynamic" mass estimates. The major items of a sprung mass computed by Adams are (again I believe) normally assumed to be monolithic. The reality (that this is not the case) implies that the "dynamic" sprung mass should be set between 3 & 10 % higher than the "corner weights". "My" local Adams modellers normally make this adjustment.
3. Inadequate damper modelling. An ideal load-velocity trajectory rarely models dampers accurately. The differences can be significant.

With a bit of work (fiddling if you like) an Adams model can reproduce rig test results with a repeatably of a few percent - for a race vehicle - certainly accurate enough to optimize damper settings.

The tire to post interface is too weird.


Sorry, Greg. Be grateful for for more detail....



#13 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 09 October 2011 - 17:37

on the four poster our tire sits in a wooden dish. In response to the vertical road inputs the car jump around in all 3 directions. ADAMS models the tire as a linear spring vertically, and doesn't understand wooden dishes at all. Consequently the reaction to significant bumps and so on is rather different in ADAMS than on the rig.

Presumably we could replace the tire/dish with a direct drive to the hub (and this is what is done for durability tests), and could then drive the hubs directly with acceleration data. However that means that we'd need hub accel data for a given suspension tune, defeating the object.

The reason I am not stressed about this is that the dynamic stiffness of a rolling tire over typical road profiles is different in many ways to that of a stationary tire bouncing around in a wooden dish.

The interaction between the engine and the wheels is very great for secondary ride, so any serious attempt at secondary ride modelling needs a representative engine mount model- and a good model of the driver and seat, which I have yet to see.

I do have a proprietary method for ride tuning in ADAMS, it is neither rigorous nor bulletproof, but it does get us in the ballpark for shock cals/engine mount selection. To be honest it is a moot point whether it is really worth using as you'd get the same results in a couple of days on the car, if you've got a car. There again I'm a bit of a cynic when it comes to modelling.

#14 cheapracer

cheapracer
  • Member

  • 10,388 posts
  • Joined: May 07

Posted 09 October 2011 - 19:40

The reason I am not stressed about this is that the dynamic stiffness of a rolling tire over typical road profiles is different in many ways to that of a stationary tire bouncing around in a wooden dish.


Porcelain maybe?


#15 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 09 October 2011 - 21:44

on the four poster our tire sits in a wooden dish. In response to the vertical road inputs the car jump around in all 3 directions. ADAMS models the tire as a linear spring vertically, and doesn't understand wooden dishes at all. Consequently the reaction to significant bumps and so on is rather different in ADAMS than on the rig.


Ah.. I understand, I think. Make the wheel pans flat with a low friction surface & insert a "poor mans" slip plane between the pans & the tyres (actually, if the pans are PTFC lined, talcum powder between the pan & the tyres appears to work well). And leave the transmission in neutral. If that worries you, then I have been doing that for 15 years or so without losing a car. Alternatively, horizontal restraining wires should work (as per the Renault rig).

To be honest it is a moot point whether it is really worth using as you'd get the same results in a couple of days on the car, if you've got a car. There again I'm a bit of a cynic when it comes to modelling.


I agree with you, on both propositions. It you have a good vehicle model, then it just makes the development process more efficient (my development friends claim to be able to iterate heuristically to a satisfactory set-up in around 100 damper builds...).

However, modelling never revealed any thing about a vehicle that was unknown. A rig, however, can. As an example, a former employer of yours tried unsuccessfully for several months to set up a vehicle heuristically. An hour on my rig revealed the reason - it had nothing to do with springs & dampers, & everything to do with top mounts (which, strangely, were not part of the remit).

Edited by DaveW, 09 October 2011 - 22:39.


#16 Fat Boy

Fat Boy
  • Member

  • 1,848 posts
  • Joined: January 04

Posted 10 October 2011 - 03:38

Porcelain maybe?


For the dish or the model?

#17 pugfan

pugfan
  • Member

  • 174 posts
  • Joined: August 09

Posted 10 October 2011 - 23:18

This is easily the best topic on the board right now, but it's tough to get any response because it's the questions are difficult.


That's what makes them the best topics.

#18 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 11 October 2011 - 12:29

The reason I am not stressed about this is that the dynamic stiffness of a rolling tire over typical road profiles is different in many ways to that of a stationary tire bouncing around in a wooden dish.


I can't disagree with your sentiments, largely because there is a clear logical difference between the way a tyre reacts to wheel platform and road inputs. I do doubt, however, that the difference can be held to explain, for example, the discrepancy between rig derived & manufactures published vertical tyre stiffness. I guess that much is down to test technique & measurement philosophy. I think you are (mildly) guilty of that in your conclusion above - which would, I think, be slightly less dismissive if you were to remove the constraints imposed by your "wooden dish".

Whatever your thoughts on tyre behaviour on a rig, It seems, in my experience, that rig testing can help to chose the "right" construction & pressure settings for both race & road tyres. In the case of road vehicles, it would appear that the measured transmissibility of the vehicle & its tyres reflects fairly accurately subjective assessments of ride, & has the adding benefit of indicating what needs to be attacked to improve matters.


The interaction between the engine and the wheels is very great for secondary ride, so any serious attempt at secondary ride modelling needs a representative engine mount model- and a good model of the driver and seat, which I have yet to see.


You are absolutely correct, and a rig test is a good way to put numbers to the vertical model of a suspended engine. I spent some time helping road vehicle manufacturers select hydraulic engine mounts. I find it interesting that the response of an engine/mount combination does not model accurately - typically, I have to increase the engine mass by around 10 percent over weighbridge values in order to obtain the observed natural frequencies. I think this is (yet again) because engines don't respond as monoliths. (One that subject, I once worked with an F1 team rig testing its vehicles. One test, out of several, clearly did not belong to the set - it turned out that a "slave" engine had been used in that case.)

I'm not sure what needs to be done about seats. Currently I take the view that it helps to limit test analysis to the seat rails, life becomes more complex above that. Certainly I know that some manufacturers improve ride with seat design (note I didn't say good), whilst others create difficulties for themselves (this is a **** seat - we must work around it).

Certainly (in my biased view) rig testing, for all its limitations, is an invaluable tool when it comes to understanding & modelling vehicle dynamics.

Edited by DaveW, 11 October 2011 - 12:31.


#19 RDV

RDV
  • Member

  • 6,731 posts
  • Joined: March 02

Posted 11 October 2011 - 21:16

Yowza! This would come in the middle of travelling...keep it up, keep it up!

Advertisement

#20 pugfan

pugfan
  • Member

  • 174 posts
  • Joined: August 09

Posted 11 October 2011 - 22:32

So it sounds like the driver could be considered something of a mass damper in Formula 1 (and passenger cars) and should therefore be banned.

What would be a ballpark estimate for the average power the driver experiences in Formula 1?

Edited by pugfan, 11 October 2011 - 22:32.


#21 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 11 October 2011 - 23:12

oddly enough one measure of driver comfort uses 'vibration dose values'. This is somewhat related to the enegry aborbed by the body during an event, it is frequency dependent



#22 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 12 October 2011 - 00:23

I will talk to the rig people about replacing the pie dishes with something more appropriate. Trouble is it is effectively a production facility operated by our sworn enemies so we have to fit in with them.

#23 Ben

Ben
  • Member

  • 3,181 posts
  • Joined: May 01

Posted 12 October 2011 - 07:29

I've worked with a number of cars running very different damper philosophies and worked with Dave's rig on many of them. My experience has been that the nature of the tyre has a role to play on what the optimum energy dissipation through the tyre needs to be.

Some teams assume that minimising the energy going into the tyre is a good thing, but if you're working with a very hard control tyre this might not be the right approach. Similarly I've had two identical chassis/engine packages running the same tyres. On Dave's rig a lower performance index is "better" - and don't get me wrong Dave it invariably is - but for these two cars the higher PI car was generally better on track and the lower PI car struggled to get any tyre temp. The tyres had been developed with the car with the higher PI. To be clear, the more aggressive car was still better with a lower PI compared to where it started, but the tyre had been developed in that ballpark so lowering the PI a lot further was a negative direction.

Another example, which I can't compare with rig data was a car that insisted on running very low spring rates with very large damping forces and very high compression:rebound bias. The thing really hurt tyres that we would have expected to be in the right area for this type of car, but running tyres that were much harder all of a sudden it had decent mechanical grip and very good tyre appearance - good to know if you're in a series with bullet control tyres. I was thinking a lot about this situation when I watched Alonso get lapped on the Pirelli hard at Barcelona - the one thing a very pitch sensitive aero car tends not to have is compression biased damping.

The final part of the previous example though, is what happens in terms of dynamic loads generated by those damper characteristics vis a vis carcass integrity - take care :-)

Ben

#24 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 18 October 2011 - 14:03

Vertically stiff tyres (or, at least, a high ratio of tyre/suspension stiffness) allow greater control over contact patch load variation though a disturbance, & hence better "mechanical" control. However, a race tyre will generally have a significantly greater adhesion "chemical grip" at an elevated carcass temperature, which is introduced & sustained when the tyres are dissipating energy. In general, less energy will dissipated by the tyre when the ratio of tyre/suspension stiffness is high, which implies that, for optimum performance, a compromise might well be necessary - trading "mechanical" grip (load control) with "chemical" grip (tyre temperature).

The correct compromise is difficult to find on a rig (although there are statistical indicators). In reality it depends on the tyre construction, tyre state, the level of down-force, driver abilities & preferences, and other set-up characteristics, and, of course the track. The best that can be done is to arrive at "optimized" damper set-ups for a range of possible spring selections. I suspect that this would also be the case for simulations, ranging from simple rig tests to DIL.

For interest, what does a rig test say about tyre energy dissipation? For one example, hot tyres, at hot pressures, dissipated between 30 & 40 percent less energy than new, temperature adjusted, tyres. The same tyres, after running for a time on the rig (i.e. light running in), dissipated between 10 & 20 percent less energy than when new.

Edited by DaveW, 18 October 2011 - 14:27.


#25 CWard

CWard
  • New Member

  • 17 posts
  • Joined: August 06

Posted 18 October 2011 - 14:35

I have just seen this post and read it with great interest having been involved with the subject from its birth in the 1960’s. I have been retired for 10 years and confess I am not up to date with recent developments and have not been involved with motorsport. However, I try to keep up with progress where possible and earlier this year attended an Engineering Integrity Society (www.e-i-s.org.uk) forum at the Silverstone Circuit that included a subject, “7posters – is that three too many”. The forum was chaired by Dr. Colin Dodds of Dodds and Associates. Short presentations were given by the chairman and representatives from Lotus-Renault, Cranfield University, Lola and ex. Ford on the history and the various applications to which both 4 and 7 post systems are being used. The presentations were followed by questions and general discussion on simulator testing. The consensus of the meeting was that 7 post systems were suited to refining dynamic response in the racing environment but had little application elsewhere. On the other hand 4-post systems were shown to be capable of a wide range of applications for all types of vehicles covering such areas as durability, ride and dynamic responses. I have found out that EIS are planning another forum next March, again at Silverstone at which the following subjects will be discussed:-
Kinetic Energy Recovery Systems
CAE Predictions vs. Physical Testing
Vision and Laser Systems
Application of Electric Actuators.

I question what benefits are gained by matching computer models with rig testing. Rig test usage implies that a vehicle is already in existence and by the use of rigs and other testing techniques can be used to sort out problems. The use of computers will always end up with having to run a vehicle on the road. I agree that computing can help to reduce the development time of a vehicle but it will never completely eliminate road testing.
Computer analysis can be thought of as theory and road testing as reality with rig testing as a step between them. Matching the two seems to be a case of “running after the horse after it has bolted”. If one is successful in a matching for one specific vehicle I would question that the same “tweeks” would be transferrable to a vehicle of different design or in fact to another vehicle of the same design when manufacturing tolerances are taken into account.
The following are some comments relative to this post based on my experience that may be relevant.
1. Having been involved with test and development activities on a wide range of vehicles from cars to heavy commercial vehicles for several companies, which at times involved both computer modelling and lab simulation, experience has led me to conclude that both computer modelling and rig testing should be considered as tools that approximate reality. The level of approximation or fidelity is very dependent on the methods used combined with an understanding of the limitations of both tools.
There are both technical and financial limitations when trying to reproduce real time conditions in a laboratory. And it is a case of “cutting the cloth” to fit specific requirements.
Rig simulation using inputs through the tyres scratches the surface of what occurs on the road. The input on both 4 and 7 post rigs is, as previously stated, only in the vertical direction into non- rolling tyres of a static vehicle, when in reality there are 3 directions and 3 moments acting on the spindle through rolling tyres of a moving vehicle.
1. Vertical force though the spindle
2. Longitudinal force through the spindle
3. Lateral force at the tyre contact patch
4. Brake moment about the spindle– caused by a force at the tyre contact patch
5. Camber moment about the spindle
6. Steer moment about the spindle

2. The argument relating to the effect of the effect of rolling or non-rolling tyres has been a matter of contention from the start over 40 years ago, but it is clear that it is possible to obtain close approximations of real time characteristics sufficient to provide close approximations to reality. The striving for absolute accuracy has to my mind been conditioned by the use of computers that produce results that are perceived to be accurate to an nth order of decimal places. The results depend on the level of accuracy of the inputs and the calculation methodology. However I would suggest that given two or more examples of the same vehicle will not necessarily produce identical test results and therefore absolute accuracy and complete matching will be unachievable. In reality one would be lucky to achieve accuracy of one decimal place on tests between two apparently identical vehicles.
As an illustration of the problem of the use of computer modelling, the following is the program content list of the major components of a US commercial vehicle with which I was involved:-
o Gross vehicle weights from 5000 to 30,000kg
o 19 different wheel-bases,
o 38 different chassis frame configurations,
o 42 different engine configurations,
o 19 different transmissions,
o 17 different rear axles.
o basic cab configurations – Cab over engine (tilt), cab behind engine (conventional), and crew cab(extended cab)
o Engine and transmission assembly weight variations of 650 to 1200kg

This list does not include an almost infinite number of vehicle bodies that are possible and which are not the responsibility of the vehicle OEM. In this situation, computer simulation and simple rig testing can only be a guide, as it is impossible to test all possible variations.

3. The lack of the fore aft component generated at the tyre road interface producing a vectored force input at the hub can hide some characteristics of a vehicle design. This aspect can be very significant if poor suspension geometry is used that produces unacceptable harsh ride characteristics.

As an example:-
A commercial vehicle I worked on had excellent ride performance on smooth surfaces but was almost un-driveable on rough surfaces. The reason was identified as the steep angle used for the installation of the rear springs (front eye lower than the rear) that generated major fore aft forces into the rear axle. This condition was not picked up using a rig test or on a computer model. The problem was simply solved using road tests by flattening the angle of the rear spring installation, in effect, pulling rather than pushing the wheel over the surface.

4. When it comes to matching computer modelling to rig performance, there has been no mention of the effect of heat. Comments have been made of the mismatch of responses particularly in relation to engine mounts that require fiddle factors to make a close match. A computer model invariable uses a linear rate model of a mount but as far as I know does not include the possibility of rate change as a result of heat build-up in the mount. In a real vehicle, an engine mount is subject to both, heat from the engine and heat build-up in the rubber as a result of vibration. The latter can produce melting of the rubber in either, excess levels of vibration, high levels of damping in the rubber material used, or a combination of both. The effect of heat will soften the rubber and reduce the rate of the mount and thus affect the system frequencies. The change in rate and hence frequency would give rise to an apparent increase in dynamic mass between the model and rig test even though the engine is not running. The effect of vibration testing on a rig is known to noticeably affect heat build-up in components as there is no cooling effect that would occur with a moving vehicle.

An example of the effect of heat build-up:-
I was given the job of identifying a catastrophic failure of an engine mount bracket attached to a commercial vehicle transmission. I had the test people instrument the bracket with strain gauges and asked them to run the vehicle over a rough track test course. They came back with the results of the test showing no problem with the stress levels after a couple of passes through the course. I then asked them to repeat the test but to monitor the increase the temperature in the mounts until the increase to a stable maximum level by repeated runs through the course. The resulting stress levels went sky high sufficient to lead to bracket failure.

5. Another aspect I found trying to match modelling and road and rig testing results was the difficulty of identifying significant frequencies and modes between modal data obtained from a computer model and road data. The computer modelling produced a plethora of information that could not be easily identified with road data. One example was a computer model of a commercial vehicle that identified 41 modes in the 1 to 20 Hz frequency range when an Operational Deflection Shape (ODS) analysis of road data only identified 13 relevant modes. To make comparisons more difficult was the fact that frequencies differed sufficient to question the computer analysis. As a result of this experience I relied on rig tests and ODS for the majority of N&V work.
I concur with the comment that just shaking a vehicle on a rig, even without instrumentation, was more effective and quicker for diagnosing problems than waiting for computer modelling results. In any case computer modelling can be a lost cause in a vehicle programme if there are multiple variants and not all details of the makeup of a vehicle are known. (See item 2 above).
6. A point to be borne in mind, when determining rig input from data obtained from a specific vehicle, is that the input levels cannot always be used with confidence with other vehicles, or the same vehicle if there are changes that alter the dynamics of the test vehicle from the original setup. This was brought home to me while trying to rectify a vibration failure of an engine bracket in a rig test. The input for the rig test was based on road data from a vehicle. After various changes to the mounting system using the same input data it dawned on us that the changes we were making were changing the dynamics of the system being tested and using the same initial input data was in most cases over testing the components. From this experience it is important to consider the dynamics of what is being tested in relation to the level of input used in the test. This may mean going back to the road to redo input levels if there is a major change in the system dynamics.


#26 DaveW

DaveW
  • Member

  • 393 posts
  • Joined: January 09

Posted 18 October 2011 - 16:37

A long, detailed, and interesting post. If a may, I will choose a couple of extracts from it.

The consensus of the meeting was that 7 post systems were suited to refining dynamic response in the racing environment but had little application elsewhere. On the other hand 4-post systems were shown to be capable of a wide range of applications for all types of vehicles covering such areas as durability, ride and dynamic responses.


I will agree with that... However, I would suggest that a seven post rig is simply an augmented four post rig so, logically, anything that can be achieved the second can also be accomplished on the first.


I question what benefits are gained by matching computer models with rig testing. Rig test usage implies that a vehicle is already in existence and by the use of rigs and other testing techniques can be used to sort out problems. The use of computers will always end up with having to run a vehicle on the road. I agree that computing can help to reduce the development time of a vehicle but it will never completely eliminate road testing.


You are correct, of course. I would simply add that you appear to have answered your opening question in the remainder of the paragraph. I would add that a good computer model is a convenient & efficient way of sorting out problems. Validating the model is a precursor to using a model with some degree of confidence, arguably demonstrating that the vehicle is adequately understood.

Edited by DaveW, 18 October 2011 - 16:38.


#27 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 19 October 2011 - 00:33

Things we do with correlated models of existing vehicles

- tests that are too dangerous for test drivers eg high speed roll over.
- statistical studies, especially interactions
- hardware in loop testing of new systems
- detailed analysis of failures - for example strut top loads can be measured by cutting the structure up and putting a load cell in. But by doing that you've changed the impedance of the structure and the loads you measure are different to the real ones. In a model we can estimate the loads quite accurately throughout the car.
- use them as baselines for the next design, for example we now generate the first pass at ABS/ESC calibration on a model, not the car.

Having said that it is easy to devote far too much time to modelling, there again modelling time is cheap, proto time is expensive. Typically a proto costs 100-1000 k and might get 100 days of testing before it is superceded, so each day is worth 1000 to 10000 , plus everybody's time. You can go to a third party modeller and they'll charge the bottom end of that, and generate far more data, albeit of course it is almost certainly not as useful.







#28 Magoo

Magoo
  • Member

  • 2,481 posts
  • Joined: October 10

Posted 19 October 2011 - 12:56

- use them as baselines for the next design, for example we now generate the first pass at ABS/ESC calibration on a model, not the car.


That's fairly noteworthy, I think. How has it worked out?

#29 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 19 October 2011 - 20:34

The first time it'll be used in anger is early next year, up til now we have been checking out each piece of the puzzle.

#30 Magoo

Magoo
  • Member

  • 2,481 posts
  • Joined: October 10

Posted 20 October 2011 - 04:08

That's very impressive to me. Not many years ago I was thinking never.

#31 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 20 October 2011 - 05:56

That's very impressive to me. Not many years ago I was thinking never.

10 years back, at least in Oz, we built our first serious ADAMS model...roughly speaking it is about 1/10 the complexity of what we build now-last year we took a big step forward with the steering column, and as we gain experience worldwide our ability to correlate improves. Other things we've got now - better K&C rigs, better moment of inertia rigs and tire models (or maybe the tires) seem to have improved a lot.

The original driver for all this was the SUV rollover debacle, of course. Suddenly multi body dynamics went from a minority interest to an essential part of many vehicle programs.



#32 NeilR

NeilR
  • Member

  • 459 posts
  • Joined: October 09

Posted 20 October 2011 - 22:45

I will talk to the rig people about replacing the pie dishes with something more appropriate. Trouble is it is effectively a production facility operated by our sworn enemies so we have to fit in with them.



OMG...you mean *GASP*...the accountants!

#33 Greg Locock

Greg Locock
  • Member

  • 4,512 posts
  • Joined: March 03

Posted 20 October 2011 - 23:40

OMG...you mean *GASP*...the accountants!

Nah, NVH