Conversion of acceleration to displacement for vibration

[RPJ]

I do NOT have vibration and sound package but I do have the "Full" LabView development system.

I've been trying for quite a while to determine a way to take an acceleration signal (coming in from an ...accelerometer) and convert it to displacement. Plain and simple; instead of looking at the amplitude of acceleration versus time I want to look at displacement amplitude over time. Then, I would like to be able to take this displacement signal and apply different transformations (FFT etc.).

Is anybody willing to share a simple example for such a vi: acceleration in --> displacement out?

I found some old examples but they have filters and there is quite an architecture and I don't understand at least half of it. I suppose there should be two integrations but as I am not very familiar with vibrations I 'get scarred' and everything becomes so complicated.

So, can it be done? Acceleration -> integration -> integration -> displacement

Anybody could help?

Thank you

[bul]

If you can measure the acceleration very accurate in time it easy to convert to displacement by integrating. But most probably you won’t be able to measure very clean acceleration signal. Best way is to have some trial measurements saved and work on these example signals to find out the best data processing method. If you can post some sample signals and your own trial processing vi, I can try to help you.

[Midlothian]

Hi,

If you try the integration approach you will find that, even for the best quality DC accelerometers, there is too much error.

The calculated velocity and displacement will drift from zero even with the accelerometer "static".

This is why more expensive sensors such as gyros. are necessary.

If you have a truly dynamic problem and the movement is really only transient you can introduce high pass filters on the velocity.

This will remove the low frequency content that causes the drift.

I hope that this helps.

[RPJ]

You are right about the drift.

Could you show or at least suggest what type of filter would be best to use ( IIR et)?

Right now I am not overly concened with precision; I just want to get to that RIGHT procedure to convert the acceleratiion coming from the accelerometers to displacement. I have posted a similar question  cuple of years ago but the solutions suggeed at that time left me confused. I just need a simple but correct procedure.

Next time I'll just buy capacitance probes...

[Midlothian]

Here is a screen snap of a filter arrangement that I used to show the principle.

[RPJ]

Thank you  - very good. However, this rises new questions:

1) There seems to be a bandwidth filter and then an integral, both express vis.

a) Are there any special settings that have to be used for the express vis or the defaults are good to use?

b) In general, what would be a "rule of thumb"?

2) How would I know what limits should I impose for the upper cut-off and lower cut-off?

3) Finally, if I want to build a waveform out of the displacement signal, would it be correct to connect the acceleration dt to displacement? That is, would this give me the actual displacement versus time 'as recorded at the accelerometer'?

Help!?

[Midlothian]

The band pass filter removes the "noise" that causes the drift.
It is only you and your environment that can separate what can be treated as noise and what is valid signal.

If you set up a small experiment as I did with a hand-held accelerometer, you will see this very clearly.
If you hold the accelerometer as still as you can with no filters the small accelerations average-out to zero with a small error.
This small error prevents the velocity form averaging to zero a high pass filter removes this near DC drift.
The same applies to the displacement calculation.

The band pass filter eliminates the nigh frequency noise that you "know" is not present in your acceleration or velocity and
therefore must be artifacts of the measurement and processing of the waveform.

If you move your accelerometer through steps and bob it up and down next to a measuring rule you will see that the
displacement is really bad for low frequency and slow steps but quite good for sinusoidal movements. You can then
"tune" your filters to suit the expected movement of your system. It may not be possible to separate signal/noise for
really slow movements and step and rest conditions. This is why I abandoned the approach and changed to gyros.

I hope that this is helpful.

Message Edited by Midlothian on 02-22-2007 09:30 AM

[RPJ]

Thank you. This helps indeed.

I will try what you suggested.

Any additional details are very welcome at this point.

If not, the matter is closed.

Thanks again ~ RPJ

[K_seeker]

Midlothian,

My understanding is

Gyros are great but they do not give you the tilt information at start up or initialization. In an INS system you need to get data from Gyros as well as Accelorometers. What do you think?

Thanks

[RPJ]

Just a comment: I am 'looking' at the vibrations of a machine tool spindle. At this point I feel the best would have been to use capacitance probes but I have no space to get near the spindle; this is why I put accelerometers on the spindle case. As for the gyros, what are these and when would you use them? (example of application please?)