Does mechanical vibration have an offset ? 1

Hi !
I'm new in vibration mesureing. So it might be a simple question.
Can mechanical vibration have an offset ? Is it right that if this is true the method of getting the velocity through integration doesn't work properly ?
How can i calculate the amplitude of the vibration (in mm) ? Through integration of the velocity i get the displacement but this is dependent of the sample duration. How can i get this right ?

Alot of questions ... please be patient.
Thanks alot.
Alex

 

[tomirvine]

There are several problems performing integration of a signal in the time domain.

The main problem is that the original signal may have a spurious offset or trend. Integration will greatly amplify this error. Thus, highpass filtering should usually be performed on the signal prior to integration.

Preferably, the highpass filtering should be performed on the analog signal prior to digitization.

Sincerely, Tom Irvine

http://www.vibrationdata.com

 

[VibrationSpecialist]

Dear Alex,
Vibration parameters i.e. acceleration, velocity and displacement are mathematically related for sinusoidal wave signals. From instrumentation point of view, signals produced by the accelerometers and velocity transducers are in terms of AC while the proximity probe produce AC and DC (offset).
To get velocity through integration means that you have acceleration input signal produced from an accelerometer, which produce only AC signal. Most data collector/dynamic analyzers have this integration feature that you can select to have analog or digital integration. The obtained integrated spectrum is considered correct.
To calculate amplitude of vibration in displacement from velocity signal in time domain, you can select the dominant frequency component and use it in the equation D=V/w where D is displacement and V is the velocity and w is the frequency in rad/sec. Pay attention to the unit conversion and the type of parameter (Peak to Peak or Zero to Peak, etc)

 

If you are looking to integrate a time waveform from acceleration to velocity or double integrate to get displacement, you should use a high pass filter before each integration step. If you want to integrate a spectrum, use the simple formula given by Alex (disp = V/omega). Remember to use frequency in rad/s not Hz.

 

building a high performance gas engine

 

[viblab]

According to Dave Smallwood removing spurious DC componenets via wavelet decomposition instead of high pass filtering is preferable. A wavelet analysis packet is available from Matlab. Dave is at Sandia in New Mexico and has published papers on the subject. My wavelet packet is on order, so I haven't had a chance to try it, but Dave's reputation speaks for itself. [sig][/sig]