FFT's, Waterfalls, Order plots from waveform data

[Olde_Tom]

Well, I've done the searches both here and elsewhere and I'm stumped.  New DIAdem user, but not at all new to vibration analysis, so I'm hoping this is just a "vocabulary issue. I was doing this work in MATLAB, but it's just too slow and cumbersome; we don't have the Signal Processing Toolbox.  I was hoping I'd figure out the right buttons myself & have made some progress, but I have to get this moving.

What I have is two sets of time waveforms; unfortunately my customer had cards that sampled at two slightly different frequencies (6250 Hz and 6450 Hz). Then there is a speed-time curve, but the speeds are only every 10 sec.  Shaft bearing X and Y probes are also included at the 1st frequency.

The plan was to make cascade (waterfall, or order plots).  The data was sent to me in TDMS, came in great, and again all the various vibration channels came in as waveforms, not numeric.

But I'm having trouble with the single-value FFT.  With the given sampling frequency and a Nyquist factor of 2.56, I'd expect a bandwidth of 2441 and 2519 Hz respectively.  When I'm taking data in the field I would specify no. of lines (6400 here, 2^14 = 16384 samples), window (Hanning), %overlap (50%) and no. of averages (let's just say 5, I can fiddle later as need be).  Here there is no entry for number of averages and I end up with 1 spectrum for the whole 3000 sec run  - It's a variable speed run looking for evidence of resonances).  As well, there is no input for the Nyquist factor so I end up with spectra of 3200 and 3125 Hz, but no big deal, the most interesting frequency content is toward the bottom of the spectrum anyway.

I may have another option for speed that would be far better than interpolating from the speed-time curve with the 10 sec intervals - there is a tachometer channel in the 6250 Hz set, but I'd have to get from my customer how to convert it to speed-time, and then from y'all how to get this waterfall/order plot thing going.  The speed does vary slowly, this is a big hunk of rotating machine.

Much thanks to anyone who can help me out a bit here.  DIAdem is like MATLAB or ANSYS - you can do all sorts of wonderful analysis, but you've got to get with its methods.

Tom

edited to add:  currently downloading the Time-frequency-order-analysis tool (2015) - looks like it might make this easier.  I'd bypassed a reference indicating it because it was dated 2010 and figured the features would be incorporated in a later software version.  But any help would still be appreciated.

[Olde_Tom]

The edit window expired. With the FFT, I was able to choose "No averaging" and generate the whole series of FFT's for one channel.  I also made up a time channel having the same number of entries as the number of FFT's.  But still no way to create waterfall plots.

I tried using the Time-Frequency Order Analysis (2015) but am stymied by errors when I run the script.

This is right out of the gate in the script, I bolded lines 15 & 18:

OPTION EXPLICIT
Const VASfile = "Time Frequency Order Analysis.VAS"
Const FileDlg = 0 ' 0 = no file dialog;
' 1 = pop up a file dialog first thing
' Load DIAdem Variables and launch Order Analysis dialog
IF ProgramRevision < 1110 THEN Call AutoQuit("This application will only work with DIAdem 11.1 and later")
AutoDrvUser = AutoActPath
IF NOT VSA(VASfile) THEN Call UserVarCompile(CurrentScriptPath & VASfile)
InitFileDlg_ = FileDlg
ExtFilter_ = "TDMS; TDM; DAT, *.TDMS; *.TDM; *.DAT"
IF DataDir_ = "" THEN DataDir_ = AutoActPath
DataDrvUser = DataDir_
Call Add_TFOA_Icons(Array("NAVMain", "VIEWMain", "ANAMain", "REPMain", "SCRMain"))
Call ScriptStart(CurrentScriptPath & "Create Group Time Channels.VBS")

So I'm off to find out about groups and collections and DataDir and things like that.

I actually had waterfalls going in MATLAB - I did the FFT, calculate time & interpolated speed at the midpoint of each FFT window, and had both speed and straight time waterfalls, but it was taking too long.  We're looking to possibly replace our IOTech hardware & software & NI is a natural first choice, but I guess I have a bit of learning to do here.  Like I said, any help would be appreciated!

[Olde_Tom]

Well, I posted this once and it went away, can you not reply to your own posts?  The edit window expired.  I was able to get the FFT's for my waterfall plots by choosing "no averaging" in the dialog.  But I'm getting errors right out of the gate with the 2015 Joint Time Frequency Analysis Tool:

The code, with the error lines bolded:

OPTION EXPLICIT
Const VASfile = "Time Frequency Order Analysis.VAS"
Const FileDlg = 0 ' 0 = no file dialog;
' 1 = pop up a file dialog first thing
' Load DIAdem Variables and launch Order Analysis dialog
IF ProgramRevision < 1110 THEN Call AutoQuit("This application will only work with DIAdem 11.1 and later")
AutoDrvUser = AutoActPath
IF NOT VSA(VASfile) THEN Call UserVarCompile(CurrentScriptPath & VASfile)
InitFileDlg_ = FileDlg
ExtFilter_ = "TDMS; TDM; DAT, *.TDMS; *.TDM; *.DAT"
IF DataDir_ = "" THEN DataDir_ = AutoActPath
DataDrvUser = DataDir_
Call Add_TFOA_Icons(Array("NAVMain", "VIEWMain", "ANAMain", "REPMain", "SCRMain"))
Call ScriptStart(CurrentScriptPath & "Create Group Time Channels.VBS")

I had this going in MATLAB - had the FFT's, calculated the time & speed at the center of each FFT frame, and was generating both time and speed ordered waterfall plots.  But it was just so, so slow.  We are evaluating hardware & software to replace our IOTECH products, and NI seems like a natural choice, but either I'm missing something or ...

Anyway, if anyone can help I sure appreciate it.

Tom

[Brad_Turpin]

Hi Olde_Tom,

Are you able to post the input data to these calculations?

Have you checked out the "Joint Time Frequency and Order Analysis Tool for DIAdem"?

https://knowledge.ni.com/KnowledgeArticleDetails?id=kA03q000000x14vCAA&l=en-US

Brad Turpin

Senior Technical Support Engineer

National Instruments