Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
electricpete

Sr. Member
Registered:
Posts: 651
Reply with quote  #1 
We have a consultant analysing piping vibration in our plant,  and they requested us to collect vibration measurement in velocity Power Spectral Density (PSD).... units (ips)^2 / hz

Apparently PSD is useful for analysis of broadband signals as discussed here:
https://community.plm.automation.siemens.com/t5/Testing-Knowledge-Base/What-is-a-Power-Spectral-Density-PSD/ta-p/360969
... Specifically the "floor" level is constant regardless of resolution for PSD (picture 7), unlike our "normal" spectra (picture 3) where the floor level of a given broadband signal changes when we change the frequency resolution. 

But we have no idea how to collect data in this format. 

Questions:

  1. Can PSD be collected with CSI equipment (2140)? 
  2. Alternatively, is there any way to export the raw TWF data (other than wav format). 


JuddJones

Sr. Member
Registered:
Posts: 384
Reply with quote  #2 
I can't be of any assistance on your 2 questions, but I can tell you that Power Spectral Density is one of the FFT types native in SPM Condmaster software. You can choose from Linear, Amplitude, Power, Power Spectral Density, and Cepstrum in the FFT Type pull down in the measuring point data window. I have also seen some great motion amplification videos on piping surveys. It would be a very quick way to see whats moving, how much, and even measure the phase between different sections of pipe.
Walt Strong

Sr. Member
Registered:
Posts: 889
Reply with quote  #3 
Pete,

I have never seen PSD offered for spectrum settings by CSI, but I am not up on current analyzer and PC software features.
Have you checked 2140 Spectrum settings?
Have you checked AMS Machinery Manager spectrum settings?
Have you contacted Emerson-CSI Support?
Does anyone in Customer Support know what PSD is?

You may be stuck post-processing the waveform (WAV) files with third party software like LabView or DasyLab (requires file conversion). Otherwise a different measurement-analysis system may be needed, depending on the number of measurement points and other issues. I can confirm that the DasyLab software that I have has both power density and power spectral density choices for scaling the FFT spectrum.

Walt

ivibr8

Sr. Member
Registered:
Posts: 163
Reply with quote  #4 
Pete
Wish I could help.
I used to use a Rockland 5810 analyzer that had a specific PSD function.
Just as you described, it was used to measure overall broadband noise. I only used it several times to provide a single overall noise level.

I have not seen that type of request since then in the past 20 years.

Jim P 
Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,398
Reply with quote  #5 
Cynical me thinks the consultant has no idea.... Meaning, is this really a meaningful measure of what is trying to be measured.
Beatnik

Sr. Member / Supporter
Registered:
Posts: 183
Reply with quote  #6 
It's just another overall unit it seems.

http://vibrationdata.com/tutorials2/psd.pdf

"The power spectral density (PSD) is simply the (overall level)^2 divided by the bandwidth."

electricpete

Sr. Member
Registered:
Posts: 651
Reply with quote  #7 

Thanks for the comments.

 

BigJ - Good point.  Motion Amplification is something we will consider.

 

Walt - it is not discussed in AMS manual or 2140 manual based on my word searching of the pdf.  I talked to our local distributor / support guy (Puffer Sweiven) who's pretty proficient with the software and he didn't know.   I am close to concluding that it can't be done, but I figured I'd ask here since I've been surprised before on this forum with people telling me how to do things I didn't think could be done.

 

ivibr8  - I'm not sure I get your meaning. You are doubting the value of PSD for this application?  To me it seems very logical that our typical spectra are limited for characterizing broadband vibration because the magnitude depends on the resolution (I notice this most often when looking at the "noise floor"... it's much lower in high resolution spectra than lower resolution spectra on the same machine).   PSD solves this problem.

Beatnik

Sr. Member / Supporter
Registered:
Posts: 183
Reply with quote  #8 
I think Electricpete has me on ignore!! [tongue]

http://vibrationdata.com/tutorials2/psd.pdf 

Note that the amplitude is actually [ GRMS^2 / Hz ], where RMS is root-mean-square.

The RMS notation is typically omitted for brevity.

The Hz value in [G^2/Hz] refers to a bandwidth rather than to the frequency in Hz along

the X-axis.

[...]

The power spectral density (PSD) is simply the (overall level)^2 divided by the

bandwidth
[...]

The overall GRMS value can be obtained by integrating the area under the power spectral

density curve. The GRMS value is then equal to the square root of the area.

This is equivalent to calculating the "square root of the sum of the squares,"



If you know the TWF settings (number of line and fmax) that they want you to base the PSD on, you can get the Overall in inch/s with the 2140. you just have to use PSD =  "(overall level)^2 divided by the bandwidth"

bandwidth = Fmax/lines * Windows function

Windows function = 1.5 for hanning window.

electricpete

Sr. Member
Registered:
Posts: 651
Reply with quote  #9 

Quote:
I think Electricpete has me on ignore!!

Hah, never.  I only used that for Sam P on the old board.  I apologize but I missed your post.  I read the first three responses earlier and came back later to compose my reply.  Didn't notice another response in the meantime. 

Quote:
"The power spectral density (PSD) is simply the (overall level)^2 divided by the bandwidth."

The terminology is a little strange here. The way we typically use the term "overall" is a single number to describe the entire signal (or at least up to a certain Fmax).   In contrast, the PSD has a value for each frequency bin (just like our normal spectra has a value for each frequency bin).

In the examples in your link (table 1 and table 2),  they're using the term overall a little differently. Table 2 gives a different "overall" for each of the frequency bands listed.

I agree we could (if we wanted) calculate the PSD from the FFT on a bin by bin basis: for the kth bin: PSDk = FFTk^2/BW, possibly with  your window factor. 

ivibr8

Sr. Member
Registered:
Posts: 163
Reply with quote  #10 
EPete
I apologize for the poor wording.
I don't doubt the value at all.
In fact, PSD was the correct measurement for me to perform (20 years ago) during an estimation of broadband
energy coming from the discharge of (critical) valves from our nuclear plant.... (a long story not worthy of mentioning here.)

I simply meant to imply I have not had a need nor a request to do similar measurement since then

Jim P 
Beatnik

Sr. Member / Supporter
Registered:
Posts: 183
Reply with quote  #11 
Quote:
Originally Posted by electricpete


Hah, never.  I only used that for Sam P on the old board.  I apologize but I missed your post.  I read the first three responses earlier and came back later to compose my reply.  Didn't notice another response in the meantime. 

The terminology is a little strange here. The way we typically use the term "overall" is a single number to describe the entire signal (or at least up to a certain Fmax).   In contrast, the PSD has a value for each frequency bin (just like our normal spectra has a value for each frequency bin).

In the examples in your link (table 1 and table 2),  they're using the term overall a little differently. Table 2 gives a different "overall" for each of the frequency bands listed.

I agree we could (if we wanted) calculate the PSD from the FFT on a bin by bin basis: for the kth bin: PSDk = FFTk^2/BW, possibly with  your window factor. 




They say in your link: "By convention, the amplitude of the data in a Power Spectral Density is squared. For example, if one were measuring a 5 g amplitude (rms) sine wave, the amplitude shown in a PSD would be 25 g2/Hz"

If I understand correctly, if your client ask for PSD at a specific frequency, it's just the rms value squared.

If they ask a PSD of a range of frequency, it's probably like my link explained.
Walt Strong

Sr. Member
Registered:
Posts: 889
Reply with quote  #12 
"If I understand correctly, if your client ask for PSD at a specific frequency, it's just the rms value squared.

If they ask a PSD of a range of frequency, it's probably like my link explained."

The PSD value in a linear amplitude spectrum is the RMS value squared divided by the filter bandwidth. The filter bandwidth depends on lines of resolution, F-max, and filter type. There is a difference between PS power spectrum and PSD power spectral density! As Pete says, the PSD would have to be calculated on a Bin by Bin based from spectrum. I don't think Emerson-CSI can export a spectrum to use in Excel (such as an ASCII file).

Walt
electricpete

Sr. Member
Registered:
Posts: 651
Reply with quote  #13 

To recap, the PSD is not an overall, it has a different value for each frequency bin (similar to spectrum). 

Quote:
They say in your link: "By convention, the amplitude of the data in a Power Spectral Density is squared. For example, if one were measuring a 5 g amplitude (rms) sine wave, the amplitude shown in a PSD would be 25 g2/Hz"

The example they give is imo misleading. The example appears geared toward illustrating the units for a single bin containing a sinusoid (not as an overall representing the whole spectrum).  If we look at the example blindly as a recipe for calculating PSD of anything, we find it cannot be a correct recipe because the bin width is not considered in the example (they should have stated 1 hz bin width and perhaps made a comment about bin leakage if they were trying to be precise).    The example was just a way to attempt to illustrate the units described in the sentence preceding the example imo. 

Walt Strong

Sr. Member
Registered:
Posts: 889
Reply with quote  #14 
I agree with Pete. It might be better to discuss random vibrations (white noise) instead of sine waves. If you had a PSD spectrum, then you could get an overall value from the entire frequency range, or from a selected sub-band, by RMS energy summation. I am not sure if the overall PSD value would have any real value! What are you going to do, have your staff do the measurements or require that the Consultant do it?

Walt
electricpete

Sr. Member
Registered:
Posts: 651
Reply with quote  #15 

Quote:
What are you going to do, have your staff do the measurements or require that the Consultant do it?

We sent the consultants our normal spectra (velocity and displacement is what they were interested in) in graphic format, along with time waveforms and the setup (lines and Fmax). So they were able to guesstimate what the PSD would be at certain points along the graph (square it and then multiply by bin width).   

 

In the end, the most prominent vibration parameter listed in the consultant's analysis report was the true peak velocity that they took from our velocity time waveforms.    As I understand it, they used that true peak velocity to support a conclusion that our piping is below any fatigue endurance limit.   I didn't look at the report closely since I am not much involved in this particular issue (following from a distance). 

 

At any rate, my original question is obsolete / overcome-by-events (we don't have to send any more data).  As far as I can tell, there is no way to get PSD directly from CSI equipment (other than calculation on our part) and no way to export raw data other than wav file ( which can be converted to excel and other numeric formats with a little work). 

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.