Measuring Displacement With Accelerometer

[uglyfatjoe]

Background:
I have a "rigid" metal frame attached to a vibration stand. Housed in the "rigid" frame is a plastic tank with hard contact points on the top and two steel bands holding the tank on the bottom. The two bands fasten to the sides of the "rigid" frame and when tightened compress the plastic tank into the top contact points assuring a tight fit.

I'm shaking the tank 3G peak-to-peak at 13Hz.

I am noticing that the fit of the tank is loose when shaking and tight again once I stop. When vibrating the steel bands seem to be flexing controbuting to the looseness condition.

Question:
If I put an accelerometer on the middle point of each band (or whereever I feel I am getting maximum bending) can I use those to measure the displacement of the bands relative to the control accelerometer. My thought is that if the band is rigid and does not bend then they should read the same acceleration as the control. If the bands are bending then the accelerometers should read a higher acceleration than the control and I can calculate the displacement from the delta between them. I'm guessing that the peak accelerations wouldn't be in phase with each other but it shoudln't matter since I only care about the maximum between the two at one cycle.

I would like a little sanity check before I waste time pursuing.

 

[IRstuff]

How are you, "noticing that the fit of the tank is loose when shaking?" How much "loose?"

What are your accels telling you currently? If they're really loose, then you should already be seeing a resonance when that occurs. If you don't adding another accel probably won't help.

TTFN

FAQ731-376

 

[uglyfatjoe]

>>>How are you, "noticing that the fit of the tank is loose when shaking?" How much "loose?"

Just by watching it shake and seeing daylight between the top of the tank and the hard mounting points. I would say it's roughly 5 mm. It's apparent that the design needs to change, however, I would like to quantify the deflection so I can correlate it to an FEA model and prove out the new design in FEA first. My reason to try and quantify it is that the ~5 mm I am seeing may not be all from the bands.

>>>What are your accels telling you currently? If they're really loose, then you should already be seeing a resonance when that occurs. If you don't adding another accel probably won't help.

Up until this point I have not ran with any accelerometers on the points described. I just watch the test earlier today and was trying to think of a way to quantify the amount of bending. I could easily put some accelerometers on the bands but if my general thought process is way off then I'm just wasting time. In this current environment wasting resources to get garbage just doesn't fly.

 

[BobM3]

In general it is difficult to get displacment from accels. I assume you plan on double integrating the accel signals? That's tough because any offset in the accel data can swamp out the integration. Plus if there is any harmonic distortion it may be tough to know what frequency components you are dealing with.

 

[Bribyk]

I'd bet a photo with long shutter duration (couple of cycles of the strap frequency) and a decent visual scale would give you as good of an estimate of the deflection and probably cheaper/quicker.

 

[GregLocock]

Or use a high speed video camera. Or a stroboscope.

The accelerometer may help you but while it may provide data it may not provide information.

Cheers

Greg Locock

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

 

[IRstuff]

But, what are the accels that you do have tell you? Or are you saying you had no accels anywhere? Surely, you had at least on the unit under test.

If the bands are separating from the unit, then the unit will be doing something other than what's being driven by the shaker, and that should show up in the response as a resonance, or at least, an anomaly.

In any case, while displacement, as others have mentioned can not be readily detected, decoupling of the bands from the frame/box, or decoupling of the box from the frame, will should show up as resonances.

TTFN

FAQ731-376

 

[Ron]

Greg...I like your comment that the accelerometer may provide data but no information...how true!

Displacement transducers might help, but the timed optical approach is good for this application

Ron

 

[SomptingGuy]

If there is a single excitation frequency, you could calculate a running mode shape using a roving accelerometer and a reference. It seems like overkill though. It'll explain the problem but not fix it.

- Steve

 

[uglyfatjoe]

Thanks all, I'm not a NVH guy so I was just stabbing in the dark. I'll opt for the high speed camera approach.

 

[izax1]

Either you have a "static" displacemnt of the tank at 3G's (inertia of the tank) or you are hitting a natural freuqency in the system (Tank or band)(Not unlikely at 13 Hz) The displacemnet (x) will be x=a/(2*pi*f)^2

(pi=3,14)

 

[blacktalon]

They do sell displacement sensors....

 

[izax1]

Yes, but why bother. You can handle it perfectly OK with accel and convert via the frequency.

d=a/(2*pi*f)^2

With a=3g at f=13 Hz:

d=3*9810/(2*3,14*13)^2

d=4,42 mm

 

[uglyfatjoe]

Yes this is why I asked the question but based on the various responces above I have been steered away...

 

[SomptingGuy]

... and then there's always the problem of convincing yourself that the "3g" was peak. Or was it RMS?? Hmm, let's see if I can tell from the calibrator. Nope, still no clearer. And on it (often) goes.

- Steve

 

[IRstuff]

The displacement that's calculated with d=a/(2*pi*f)^2 doesn't tell you what the separation of the band is relative to the tank, because the equation applies to both objects. You need to have the phase information to get anything meaningful.

Moreover, you apparently have no other accel information, so you have no idea what the actual acceleration is on the tank, only what's driving the frame. You could have 10 g or 20 g of acceleration on the tank because of a resonance.

TTFN

FAQ731-376