Damper sensors usefulness

[Belatti]

Hi,

I have been using linear damper sensors in a PI data acquisition system for a while.
The main purspose of this was to gather some data our consultant needed to perform a 4 post rig test (car with low aero downforce).

The rig test went very good, the laptimes decreased and the topics discussed here helped a lot.


I have used the PI Toolbox to plot histograms:





And tried to figure out downforce levels:





Taking into account the different motion ratio front and rear, I wonder how much can I trust these polynomials.



I have also processed the data in matlab, using FFT and some results were these:




You can see a peak at x:2,737 wich gets pretty close to the natural frequency of the car that I calculated using corner weight + spring rate at the wheel (2.75Hz)

Now I would like to know what are the other possible uses I can give to them.

[Jersey Tom]

Those don't look like histograms.. and somethin doesn't look right about em as a X/Y plot..

[Belatti]

Those don't look like histograms.. and somethin doesn't look right about em as a X/Y plot..

Sorry Tom, my bad. I uploaded the wrong pic, now I corrected.

[Jersey Tom]

Something still looks off with those X/Y charts. Should look like a scatter plot rather than a curve.

And with the FFT.. a spectrogram may be more telling of what's happening over the course of a lap.

http://www.mathworks.com/access/helpdes ... ogram.html

[gambler]

Could you please elaborate on the upper four grafts.
Is it representing one full stroke up and down ect,ect
Thanks

[Belatti]

The histogram are damper speeds, calculated as time derivates of the damper positions that the sensors read.

The other 4 X/Y plots are the 4 wheels damper position, from 150 to 210 kp, at a 200Hz sample rate. You can plot either points or lines, I prefer lines because you can see better.

[gambler]

What is the left hand vertical column representing. and
why are the number scales different on side by side comparisons?
Thanks for your patience.

[Jersey Tom]

The other 4 X/Y plots are the 4 wheels damper position, from 150 to 210 kp, at a 200Hz sample rate. You can plot either points or lines, I prefer lines because you can see better.

Still.. doesn't look right. If it's over the course of a lap it should cross back over itself many times over.

And that's an incredibly small range of damper travel.

[Belatti]

What is the left hand vertical column representing.

mm of damper displacement

and
why are the number scales different on side by side comparisons?

its a matter of screen space... just to see the graph better

Still.. doesn't look right. If it's over the course of a lap it should cross back over itself many times over.

I have said its from 150 to 210 kph, its a straight, not a complete lap.

And that's an incredibly small range of damper travel.

How can you say that? Do you know the car enough?

[DaveW]

Thanks for your original post, Belatti. I am pleased to hear that your rig test worked for you. A couple of questions on that, if you don't mind sharing. Was the rig test input a "swept sine"? If so, what was the platform input velocity?

To your posted data. It looks very sensible to me, but always be aware that potentiometers are not the most reliable race car transducers, & I think Pi, by default, still manipulates recordings to hide wiper "bounce" glitches caused by a dirty potentiometer track.

I think the "trend" lines of your example probably started too early - it looks as though the car was still in a turn up to 165 mph. (You could probably help matters by fitting the average of left & right displacements - which should also reduce slightly track input "noise").

The trend lines should reflect vertical loads reacted by the spring, bump rubbers with, possibly, a contribution from the dampers. The vertical loads should (ideally) depend only on sprung weight, longitudinal, lateral & vertical accelerations, and aero forces. The last will depend on dynamic pressure and on "lift" coefficient, which will normally be a function mainly of axle "ride heights". Unravelling that is possible (with faith & enough data), but it is not an easy task. Dampers can make the task even more difficult by introducing a time-dependent element (e.g. it looks as though your rear dampers are rebound-biased, which will mean that they will tend to "jack" the vehicle down on-track). Another potential problem is that suspension "geometry" can mean that the springs & dampers don't react all the vertical load carried by the suspension.

I think (with apologies) that your frequency domain plot is not very informative. I admit to being something of a pedant about such things, but I don't think that any race track (including Le Mans) can deliver enough useful (non-repeating) data to generate reliable frequency domain information about the vehicle in isolation. So, while the magnitude of your 2.737 Hz value will be affected by the vehicle heave mode, it will be dominated by the track characteristics. This is confirmed (for me) by the magnitudes of the 2.737 +/- one_frequency_step. JT's suggestion of averaging transforms will help, but not much (technically, the variance error will be proportional to the inverse of the square root of the number of independent transforms for a "stochastic" input).

[speedsense]

Hi,

I have been using linear damper sensors in a PI data acquisition system for a while.
The main purspose of this was to gather some data our consultant needed to perform a 4 post rig test (car with low aero downforce).

The rig test went very good, the laptimes decreased and the topics discussed here helped a lot.


I have used the PI Toolbox to plot histograms:





And tried to figure out downforce levels:





Taking into account the different motion ratio front and rear, I wonder how much can I trust these polynomials.



I have also processed the data in matlab, using FFT and some results were these:




You can see a peak at x:2,737 wich gets pretty close to the natural frequency of the car that I calculated using corner weight + spring rate at the wheel (2.75Hz)

Now I would like to know what are the other possible uses I can give to them.

All the graphs are "full" lap graphs. This a problem when looking for shock trends and aero trends.
With shocks, the histograms can be "gated" according to lateral G. You minus the straightway inputs and include only lateral G generated data. You can also apply an long G/brake pressure gate to see the pitch control moments (front and rear) either separately or inclusive of the cornering forces. You don't care how the car rides down a straightway, just what they are doing in the corners. Including a straight away in a histogram will only cloud the analysis.

Don't forget that Shock speeds by way of math using DERV functions, is by nature of the math, by a small percentage, a dirty signal.

With Aero downforce, the same technique can be applied, with perticular speeds chosen for gates. It is best to analyze, with the car straight (no lateral, or braking long g) and keep the range within a few KM's for each target speed.
Without a pitot tube however, you could be spitting into the wind... (pun intended)

[Belatti]

Thanks Dave, Tom, Speedsense

You have gave me some more ideas to the way I can process the data I have got. However I need some more track time for various reasons, one being that I have discovered that the way I did calibration on the sensors was not the ideal.

After re-thinking carefully the meaning of all those plots I did and with your help I can figure out that they dont worth much.

So, I must put my head down and keep workig!

[Jersey Tom]

Looking at it again.. that your dampers are working almost entirely at less than 2 in/s seems really odd.

[Belatti]

Looking at it again.. that your dampers are working almost entirely at less than 2 in/s seems really odd.

Sorry, when I said that I had calibration issues I was referring to the track vs. post rig tests. What I question now its those graphs practical usefulness.

The damper speed graphs are not wrong, though. I re-plotted them for you to see that higher speeds are reached but they still dont contributed much to the histogram:



After continuing judging them, you should know more about the cars motion ratio, and maybe masses and rigidity, because remember they are damper and not wheel movement.

All the graphs are "full" lap graphs. This a problem when looking for shock trends and aero trends.

The FFT graph is for a braking + curve + beginning of acceleration (arround 8 seconds) for a single wheel.
The damper speed histogram is for a whole lap.
The DF level graph is only for a straight, I have re-plotted to graph erasing the first part as DaveW suggested and calculated the 2 front damper average displacement:

With Aero downforce, the same technique can be applied, with perticular speeds chosen for gates. It is best to analyze, with the car straight (no lateral, or braking long g) and keep the range within a few KM's for each target speed.
Without a pitot tube however, you could be spitting into the wind... (pun intended)

Right.

[Jersey Tom]

Even without knowing anything about the rest of the car, < 2 in/s for probably 95% of that histogram is a little funky to me.

In my experience most dampers have the knee between low-speed and high-speed rates at around 1-2 in/s.. which would indicate almost no high speed damper action. If that's the case the MR of the car would seemingly be way off... then you get to mechanical grip issues with damper friction, yatta yatta. The Time/Distance plot of velocity does look a good bit different though, not sure it wouldn't be reflected in the histogram. As was mentioned, how you gate the data plays a big factor. Easy to do in Pi.

In any event, damper histograms and displacement plots can certainly be useful.. particularly if you can boil the information down into metrics / performance indicators.

For example... % Rebound vs Compression... % Highspeed vs Lowspeed.. etc, at each corner of the car.