LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

Sound and Vibration pallet is only for ...

The VIs you'll need are all in the same palette we've been discussing.  Look at Amplitude and Level, Peak Detection, and Limit Test VIs.  All are in Signal Processing > Waveform Measure.

They should help you monitor your waveform for absolute levels.  Don't forget the Distortion, and FFT tools previously mentioned if you want a more advanced analysis.

LabVIEW Pro Dev & Measurement Studio Pro (VS Pro) 2019 - Unfortunately now moving back to C#, .NET, Python due to forced change to subscription model by NI. 8^{
0 Kudos
Message 21 of 25
(424 Views)

I thought will be easy to procedd with MIN and MAX VI by difrentiating the noise images but now i use the FFT to detect the noise frequencies and which gave me good result. Problem is, some of the motors are good when they are below 500Hz and more than this frequnecy is noise but some of the motors is still working good but having less noise over 500Hz. How can i make it in labview to take it as clear signal and no need adjust more. See the references down here.

Download All
0 Kudos
Message 22 of 25
(422 Views)

It looks like the spectrum of the "slightly" noisy pump is not much different from the good ones, but the amplitude discrimination still seems to work.  I would stay with the amplitude discrimination for now, but continue to monitor the spectrum to see if it may be useful later.

 

You will not find a VI or function in the palettes which does exactly what you want.  LV is a programming language and to do anything beyond the simple and fundamental things, you need to write your own programs.

 

You are considering different behavior for three or more amplitude ranges. When you determine which range you have, then send the value to be written to the next state via a shift register. The value you send could be 30, 10, 5, or 0. When you get 0 two times in a row, then you decide the pump has been adjusted properly and end the test.

 

Lynn

0 Kudos
Message 23 of 25
(415 Views)

 to Johnsold 

                   

It looks like the spectrum of the "slightly" noisy pump is not much different from the good ones, but the amplitude discrimination still seems to work.  I would stay with the amplitude discrimination for now, but continue to monitor the spectrum to see if it may be useful later.

 

I did not understand your intention, can you make me clear on which way is better to go ahead. I am new to Labview programming and spending most of the time on it. I am confused to proceed because I have the spectrum and normal time domain waveform. How can I differentiate this noise and intimate to serial. I just want to know about differentiating the noise and rest of the program I can make by myself. Please help me about these differentiation that when ever noise came the serial device should adjust. I thought that more than 500hz should be written a value to serial but how can I read the frequency from FFT waveform to a comparison Vi or is there any other way to do it by normal time domain waveform.

0 Kudos
Message 24 of 25
(412 Views)

It is difficult for me to tell you what to do because I really do not know enough about your pump and the mechanism which generates the noise.

 

One of the issues is that the criteria for defining which ones are good and which ones are bad do not seem to be clear to you.  It is easy to have a human operator adjust the stroke until the noise stops.  After some experience the oprators probablby learn how much to change the stroke and how much margin to add bsed on how it sounds and how large the previous adjustment was.  Human minds are extraordinarily good at detecting patterns and certain kinds of changes in sounds, but not very good at describing in a quantitative way what they did.  Unfortunately, that quantitative description is needed to write a deterministic computer program.

 

What I am trying to say is that you will most likely need to write a program, try it in comparison with the manual methods which have been used in the past, and then make adjustments in the program. Some of the "adjustments" might require rewriting much of the program if the algorithm turns out to be unreliable.

 

I once built a similar system (before LabVIEW) to analyze the performance of a motor/gearbox assembly.  The original inspection system involved the operator pressing the blade of a screwdriver against the machine and holding the handle against his ear.  We built an instrument with filters and meters to measure the energy in two frequency bands.  We set the threshold for rejection so that the same percentage failed as with manual inspection.  We found that even though the same number were rejected, about half of the inidvidual units were different. That is the inspector rejected about half of the units our tester passed and our tester rejected about half of the units the inspector passed.  Six months later they were using our instrument exclusively and the failure rate had dropped to about 25 % of the previous value.  The Quality Control Manager attriuted the improvement to the accurate feedback our instrumetn gave to the inspectors and the production crews.  When they were told what the problem was and it was always what they were told, they started making them better.

 

The image you posted in Message 22 called "Untitled.jpg" shows a noisy signal but little high frequency energy.  That is the kind of situation which is hard to decide in a program.  Is this a good unit or a bad one?  Does the result change if the stroke value is changed slightly?  My guess is that this unit is marginal.  It probably had very light impacts on the first few cycles and then had a section where it did not impact.

 

After looking at the spectra more carefully, I have a suggestion for a slight modification to your decision algorithm. The fundamental frequency is around 170 Hz.  Because the signal is not sinusoidal you have some harmonics present even on the good pumps.  You called this "Noise but OK" on one of your graphs. The spectrum of the very bad unit had higher harmonics which were much larger than the fundamental.  Here is my suggestion. Normalize all of the harmonics out to about 2 kHz (9th or 10th harmonic) to the value of the fundamental (~170 Hz) component.  Learn what the typical ratios of those harmonics are for a set of good pumps and for a set of noisy pumps.  Then look at some of the marginal ones.  I suspect that you may find that the 5th through 10th harmonics will increase significantly in the bad pumps compared to the good ones. You will need to run tests on several pumps to find out what works.

 

Lynn

0 Kudos
Message 25 of 25
(394 Views)