Is there a minimum RPM when using OAT Easy Resample Order Tracking.vi

[jmajkowski]

Trying to use Order Analysis Toolkit with the OAT Easy Resample Order Tracking.vi for fan wheel vibration analysis. It needs to work from 400-4200 RPM. I have it working good("stable" readings) at 1600+ RPM with 1 pulse per revolution, 20000Hz sample rate and 15000 samples. At 400 RPM both magnitude and phase angles are all over the place. Anyone know if there is a minimum RPM OAT needs to sample properly or do I need to change sample rates and number of samples taken?

Thanks in advance,

[bspears]

Hi,
The number of samples you will need will depend upon which order you are tracking. The information you need is found in the LabVIEW Order Analysis Toolset User Manual which should have been installed when you installed the toolset. In Windows, it's found in Start>>All Programs>>National Instruments>>Order Analysis. On page 4-2 and 4-3 the manual addresses your concern.

Let me know if you need more assistance!
Brian Spears

[jmajkowski]

Thanks, I found the section you suggested. We are only concerned with 1st order analysis at this time, so I guess I need 2.5 samples per rev. I'm a little confused by the "multiply the number of pulses generated by the tachometer..." Would you know if I simply mathmatically do that to the tach input signal or would I have to add more strips of reflective tape to the fan wheel? The latter not being an ideal solution once(if) I ever get this into our factory for their use...
Thanks again for the help,

[bspears]

Hi,
When the manual says: "you need to multiply the number of pulses generated by the tachometer" it is just a simple multiplication of the number of tach pulses by the interpolation factor "L" that's shown in the equation right after that paragraph. You're just adding more points for interpolation, not creating more digital inputs.

Let me know if you need more help!
Brian Spears

[Jim@SONGS]

Howdy
You know, the slower you go the bigger sample block size you need. How many revs do yo want, too. If the sample rate is too slow and not enough revs (sample block size) the resolution will suffer too. So if you figure your sample rate and "samples to acquire" for full speed, you may not be waiting long enough for the buffer to get full or alternately you may be waiting too long at slow speed and get a buffer overflow. Do you get any errors?
All this has to be figured depending on the application, I think. Also, if you use accel or velocity probes, there will be so little signal from 1X the angle calculator won't be able to catch the high spot. With prox probes and runout there is usually signal all the way down, but the 1Xvector is slow roll.
jim

[jmajkowski]

OK, I did some digging through the OAT Easy Resample Order Tracking.vi and I am pretty sure it already multiplies the tach signal by 2.5 of the order, and it also has the CIC filter, so that I don't think that is the problem. Maybe Jim, you are on to something though, as I've played around with sample rates and samples to get, I noticed I can get a lot more resamples out, i.e. rate 20KHz / samples to acquire 60K. But that really slows things down. I am on continous sampling before I head into a loop and use the "DAQmx Read(Analog 1D Wfm NChan NSample).vi" I then select the signals I want with the express select signals vi and go from there.

If anyone knows of any kind of "rule of thumb" about the sampling rate and samples to acquire please let me know, because I have no idea about where to set it. I am only using 20KHz and 15K samples to acquire because that's what it was set at when the project was turned over to me. And it seemed to work pretty well at 1800rpm.

Anyway, I went back and looked at the raw accelerometer signals(well, I bandpassed it right around the Hz of the fan wheel for a nice clean waveform)and the peaks will not stay at a constant time interval. So, I've got problems before I even send it into the resampler. Also, the waveforms from the horz. and vert. accelerometer(using 1 tri-axial) should always be 90° apart, and I am watching as they seem to have a mind of their own, overlaping each other and whatnot.

By the way I am collecting the accel. signals and analog tach signal on a BNC-2110 to a PCI-6034E.
The accel. signals are on AI0-AI2, AI6 is grounded to the breakout board AI ground, and AI7 is the analog tach.
AI3-AI5 will be used for the other tri-axial accel. once it is up and working...

[Jim@SONGS]

Howdy Again
Let's start on 1X cause typically the most useful order tracker is for a balancer.
The OAT manual rule is "2.5*K" where K is the maximum order number you want. So, minimum sample rate seems it would be 1X speed (4200RPM/60=70Hz), 70Hz*2.5=175Hz for 1X order. Not much good for anything else but balance, so lets open up to 10X and get some ball bearing frequencies, too.
70Hz*10X*2.5=1750Hz per rev sample rate.
Now, how many samples per block?
How about we want 5revs to see how consistent and stable our vibration and phase are. If we are on a critical speed, the phase will be wacky and not appear synched to the phase probe.
1750samp per sec/70revs per sec=25 samp per rev*5 revs= 125 samples should be all we need to analyze 10X for 5 revs at 4200RPM.

That does not seem like very much data to me, but we are not looking for 1/4Hz FFT resolution either.

Now, lets slow down RPM to get a bode or polar plot and see where our critical speed is. You picked 400RPM minimum.

400/60=6.66Hz
1750 samp/sec/ 7Hz =250 samp/rev *5revs= 1250 samples

But look, 125samples/1750=71mSEC per 5revs at speed and
1250samples/1750=710mSec per 5 revs at 400RPM

Since the OAT does not adjust rates on the fly, it looks like we have to set the sample rate to 1750Hz for our max RPM and the block size to at least 1250 samples for our min RPM and it will take a minimum of 710mSEC to acquire a block with the processing time on top of that. So real time blocks will be a bout 1 second apart.

I assume OAT will handle the different number of revs at different speeds and that is why the data comes out as an array of multiple mags and phases in an array for the block.

I hope this makes sense, I did it before lunch, blood sugar you know.

[jmajkowski]

Thanks Jim,
I'll give these formulas a try and see what I can accomplish...

[Jim@SONGS]

Great. I would appreciate a peer check to make sure I am on the right track. Write a direct mail if you wish. I have some other ideas too. henderjr@songs.sce.com

[jmajkowski]

One of the outputs of OAT easy resample order tracking is samples/rev. No matter what I do with sample rate and samples per block, this is always 1. Is that right? I watch the arrays coming out and they will usually have about 8 final values in them per time through the loop for mag. and phase, so I'm thinking it is doing more than 1 sample/rev. Just like to verify...I also set "reset" on the vi on the first run through the loop to true, and then leave it to false to use previous states. Don't know if that is correct either.