Vibration analysis

[RPJ]

Hello!

I have two accelerometers for X and Y directions and I am trying to analyze the acquired signals.
I would like to determine the amplitude, frequency, Power spectral density, power spectrum, the displacement, and the actual dominant frequency and its amplitude. Unfortunately I do not have experience in what concerns vibrations. I was wondering if somebody out there has better knowledge about what I should do with the acquired signal. For now I took this signal through Tone measurements, Amplitude and level measurements and Spectral measurements. However when I run the vi I do not see but the original signal; the other charts (FFT - PD, PSD) are empty. It worth mentioning here that I did not have ' vibrations yet at the accelerometers and all I record is actually noise. The signal is more or less close to 0 (white noise) but I imagine the advanced vis should analyze the waves formed by this white noise?!?

Does anyone has some kind of example of how to do a vibration analysis starting from the signal captured from the accelerometer?
Please be aware that I do not have the "Sound and Vibration Analysis Toolkit" from NI (and I do not have money to buy it).

Thank you for your time and efforts.

Radu

[CoastalMaineBird]

Perhaps I can help a bit.

You are measuring acceleration.
Acceleration is rate of change of speed.
Inches per second (for example) is speed.
Inches per second per second would be acceleration.

Vibration is a position measurement (position vs. time).

If you integrate the accelerometer signal once, you would change IPSPS (accel) into IPS (Speed).
If you integrate the result again, you would change IPS (speed) into Inches (position).
THAT is the vibration signal you should analyze with FFTs, etc.

You will probably have to remove the DC offset from your signal before EACH integration. Otherwise the integration process generates a ramp function which is not real.

[RPJ]

Thank you for your time and answer.

I knew I have to integrate but the DC offset is a new one for me. How do I do that?
Do you have a simple example handy with such an integration and DC offset taken out?

RP

[Matt_F]

Hello RP,

The Application Note Signal Processing Eases Vibration Transducer Selection is a good resource on doing vibration measurements. It also discusses filtering and provides some example code.

Good luck with development.

-Matt F

[CoastalMaineBird]

You have to remove the DC, because, if you integrate it, the output continually climbs (or descends). Integration finds the area under-the-curve, which continually increases with a DC signal.

There are various methods, depending on the nature of your signal. Taking the simple average, and subtracting it out is the easiest, but may not be the most accurate (for long blocks). You can find the trend line, and subtract THAT out for a better fit in that case.

[WantToGolf]

"However when I run the vi I do not see but the original signal; the other charts (FFT - PD, PSD) are empty."

Just a note to keep in mind. The FFT vi will generate an ouput spectrum when a sufficient amount of data has been received. If you do not see any ouput from this function you have to increase the amount of data collected and processed by this vi. If you give it too large a chunk of data at a time it will error so I suggest you place the vi in a while loop and process small chunks of a large waveform acquisition until you see some output. Then you can experiment with averaging etc.

Hope this helps you.

[RPJ]

The example in the Application Note Signal Processing Eases Vibration Transducer Selection is a good resource on doing vibration measurements. Thank you Matt.
However, can I apply that example to any waveforms coming from accelerometer? Are there any values for dt or other parameters I should be aware of?

Radu

[joost]

In case you continue to see nothing more than noise from the sensors; are these ICP sensors? If so, you'll need to connect them to a constant current source. This is typically done using a 24V psu and a 1n5413 current diode. Use cap to remove DC to connect the sensor to your daq board.

Depending what kind of analysis you're trying to perform, going from acceleration to velocity (1 time integration) or displacement (2 times integration) may not what you really want. Notice that when you integrate the higher frequencies are suppressed severely. For e.g. bearing frequency analysis but also motor analysis one commonly looks at straight acceleration signals. If all you want to produce is a trend that you can compare to some other device, velocity is your best bet.

Good luck

[Cobus]

I am also trying the integration route to obtain displacement data from a accel signal at very low freq. (2 - 6 Hz @ 7mm p-p) I found that the example code works reasonably well with removing the drift and distortion to get a good second integral. But the amplitude of calculated displacement now changes with different filter settings. Is there any guideline for filtering at such low freq ??

[RPJ]

Hello! This is Radu again with couple more questions:

Besides Cobus, did anybody tried to use the example from the Application Note "Signal Processing Eases Vibration Transducer Selection"? If yes, do you have any particular notes or warnings to post?
I am trying to take the code as is and 'connect' it to the accelerometer channels from my setup. My sampling rate is approx. 10kHz and I am sampling for about 30 sec. up to 1 minute. The accelerometers are mounted on the bearing case of a grinding spindle and I am trying to capture forced vibrations or chatter in that spindle... Would the posted example be good?
Because I'm not so advanced using LabView or complex vibration analysis I'm still having hard time deciphering the example from the "Signal Processing Eases Vibration Transducer Selection". Are you aware of any other examples?

Thank you.
Radu