Signal Analysis Results not OK?

[samurai9]

Hello everybody

I'm new to Labview and I'm facing a problem with measuring vibration. I'm trying to measure vibration on eccentric press with accelerometer. I'm sampling at frequency 51,2k, and I have 409,6k samples to read. Because the working speed of eccentric press is 1,25 Hz, this means that I sample vibration from 10 iterations (10 iterations in 8 seconds, 51,2k x 8 = 409,6k), because I want to get steady results which is not possible if I only sample one working cycle. The results I get are shown on the link bellow.

http://shrani.si/f/3V/Fv/22PjN67N/vibracije.jpg

When I sample the vibrations with DAQ Assistant I firstly show results on the left graph. Then I use Lowpass filter with max frequency of 200 Hz. This signal then goes through Spectral measurements and results are shown on the right graph. Same signal goes to Amplitude and Level Measurements and results are shown in the boxes above the graphs.

The problem that bothers me is that on the first graph it's seen that peak to peak value is around 50 g, but when I use Amplitude and Level Measurements Express VI I get result that peak to peak value is 6.8g. Does anybody know where the problem is?

Thanks for help 🙂

[gonczolp]

Hoi,

Your question is not quite clear for me. So the measured raw data visualized on the screenshot is going through a low pass filter, then the resulted signal goes to Spectral Measurements and Amplitude and Level Measurements Express VI parallelly? If so, then the difference you are asking for is between the raw displayed signal’s peak-to-peak and the express VI’s output right?

If the above said is right, try troubleshooting with the low pass filter. Place an indicator before and after it, to see the actual difference it makes, you could also try to analyze the unfiltered data, to see whether you filter out the previously seen 50g value, or the problem is with some settings of the Express VI.

If your project is not confidential, you could also upload the required VI’s here, for further help.

Regards,

Peter

[samurai9]

Thank you for your suggestions. Currently I'm waiting for new accelerometer, so sadly I can't try different things right away.

I'm posting some pictures of block diagram, maybe someone can identify the problem.

Aquiring data: http://shrani.si/f/2l/Ro/1t1n6DAQ/1.png

DAQ Assistant configuration: http://shrani.si/f/3S/tA/IJJllYU/2.png

Further data analysis (in While loop): http://shrani.si/f/2L/pI/2vK2x8me/3.png

Lowpass filter configuration: http://shrani.si/f/2S/9l/4H19nvVE/4.png

Amplitude and Level configuration: http://shrani.si/f/1T/OO/3yaVNSPt/5.png

Power spectrum configuration: http://shrani.si/f/1F/4I/1B5VhIBv/6.png

By the way, which results do you think are "the real ones"? Around 7g or aroung 50g? 🙂

[gonczolp]

Hoi,

In my opinion, both of them are ‘real’, but have a different purpose.  If you apply a low pass filter, you will get rid of all of those high frequency components of the signal, which would describe the vibrations of the machine. This is mainly, why you get a different peak to peak value before and after filtering. It is not a programming mistake. You just need to decide whether you want to track the caused vibrations or the movement of the press (I would not recommend this because of the huge error it accumulates) etc.

Also I would recommend doing the calculation in the same while loop as the DAQ itself, simply because it does not require that much time, and the loop period time is also around 8 seconds. This way you don’t need to use local variables, which also eliminates race condition (they might be written and read in the same time).

Lastly, if you are willing to dig deeper in this topic, take a closer look at the Help/Example Finder. It has some well documented code (Vibration Analysis (DAQmx) for example), which uses the DAQmx APIs instead of Express Vis. This would make your program even more efficient.

Regards,

Peter