Two groups, at our facility, collect casing vibration information on a particular motor-pump combo using mag-mounted accelerometers.

- The Operations group collects data we use for our ASME Boiler Code testing. It is collected quarterly. They record simple overall vibration via a Dytran 4177 vibration meter. This box performs a single analog integration, and displays amplitude in ips-p. It's filter range is 0 - 20kHz. Spectral data is not available from this box.
- Our PdM group collects bi-weekly data with an SKF Microlog. Overall and spectral data is trended and evaluated for maintenance recommendations. Fmax for this data is 60,000 cpm. My understanding is that data from this box is first digitized, then run through an FFT, then finally integrated by mathematically converting the amplitude at each frequency bin from acceleration to velocity. The overall amplitude is finally determined by calculating a "square-root of the sum of the squares" from each bin.

We've noticed a gradual increase, over time, in data collected at the OB motor with the analog-integrated Dytran instrument. This increase has not been observed with the 60,000cpm Fmax Microlog. Opening up the Microlog Fmax, we see a significant peak at about 212,000 cpm (119X of motor shaft speed). Curious...I attached a battery powered analog-integrator to the casing mounted accel, and used the Microlog to perform an FFT (without digital integration) to see the spectral content of this signal. The 212,000 cpm peak was there...but at twice the magnitude of that created by digital integration??? Digital vs analog integration seems to play a substantial role in the amplitude displayed by this specific higher frequency peak.

Damn...that was long winded. Sorry...and thanks to anybody who suffered through my explanation. To the point...
-What pitfalls occur when trying to compare data that is integrated by a digital vs analog process?
-Is one of the two more accurate or trendable?
-Are some frequencies more susceptible to error (eg higher frequencies)?
-Please add any other illuminating facts...

Thanks in Advance,
George

George,

The old comparison of digital to analog used to be the difference between CSI and IRD. Ird always used an analog overall, claiming you wouldn't miss any vibration, whether generated by the machine or something else that was affecting the machine. This is true, and will catch anything depending on how high your transducer is capable of seeing, and how high your box is capable of capturing.
The downside of the analog is it takes more time to capture that high frequency data.
The digital overall is quicker, and you have already seen the downfall of it, maybe. If all you are trending is the sum of the squares from 60 to 60,000, you will never see anything that occurs above or below that. I said maybe because whatever is occuring may or may not mean squat to your machine being monitored.
However, being the curious type, I like to try and find out where that 212,000 CPM is coming from. Is my transducer resonanting? Has something broken and now in resonance?
Or is that 2X rotor bar freq? If it is, and it has been gradually getting worse, then it is certainly worth watching.
I don't know anymore what the different manufacturers are doing. I know Rockwell/Entek/IRD gives you the option of using either the RSS or Analog overall filter.
I use analog everwhere because that is what I have always done, and would rather not miss a transient. There are six small pumps that I have been monitoring for years, and have never seen change. I use the digital on those just for speed sake.
I don't know whether one is more trendable than the other or not. I would hazard a guess and say the digital is probably more straight in trend, but I can't prove that since I don't use it enough to tell. The 6 pumps I do use it on have a very straight line in comparison to the analog overall trends of other pumps and machines.
I'm getting too long winded too. I am sure there are others much more knowledgable about the inner workings and hidden mechanisms of these boxes, and I'm sure they will be here shortly.
Good question, you have possibly opened Pandoras box.

Dave

The first step I'd take is to look at the peak in acceleration and velocity on the Microlog and see if the levels in the spectra make sense. This will show if the integration process is working correctly.

Jon
Spintelligent Labs

Hi guys,

"Analog Integration is superior to the equivalent digital method because it produces a better estimate of low-frequency components in the vibration spectrum and improves dynamic range."

I never use the digital integration for taking regular data.Can anybody tell me where i can use the digital method instead of analog?

Have a nice day!!

This is one issue that I think causes confusion because we are not always comparing apples to apples, some who get reports go by numbers, but dont understand the difference in analog and digital are left to wonder why?

With my CSI box I am able to determine how I want to measure both my "overall level" and "signal integration mode".

Digital overall - will include frequencies between my lower and uppper cutoff. (FMAX)
Analog overall - will include all frequencies from 1hz to 80khz, or the capability of the meter.

For example, I am collecting data at 50 orders on a 1800rpm(30hz) motor, with a lower cutoff at 1hz.
Digital overall setup will measure and display an overall level based on data in the frequency range of 1hz-1500hz.
Analog overall setup will measure and display an overall level based on data in the frequency range of 1hz to 80khz.

So lets say Joe and John work for 2 companies, both are called in to check the same motor for company "A" and report to manager the condition of equipment.
Lets say this motor has a bearing defect showing up in the 2000hz-3000hz range.

Joe uses Digital overall - reports overall level at .150in/sec.
John uses Analog overall- reports overall level at .500in/sec

Now.. why shouldnt the manager be confused?

I realize any knowledgable analyst is going to do further analysis and both should find and report the bearing defect, but based on overall levels alone.. this is confusing to many that only want to know the numbers and we are getting lots of these in the field today.

Should overall vibration level calculations be standardized?

Is there a ISO standard for Overall Level Calculations?

If there is, should manufacturers of our magic black boxes take that into consideration?


have a good day

Mike

This message has been edited. Last edited by: MikeH,

There are ISO standards for overall calculations. Unfortunately, the upper frequency limit for these standards is 1 kHz, which doesn't allow for proper bearing measurements. At lower frequencies where the standards can be applied, a machine's movement can mostly be described as rigid body movement; at the higher frequenies useful for bearing measuremment, the movement is more of a local deformation by rapid attenuation of the signal with distance, increased frequency, and component juntion reflection with the result that it's not practical to make useful standards for bearing condition.

For an illustration of the attenuation effects, see:

http://vibrotek.com/article.php?article=articles/factors/index.htm

The link explains analog/digital Overall and analog/digital integration.

The analyzer I use is a CSI 2130 with Digital Overall and Analog Integration. For bearings I also trend VHFD (variable high frequency detection) from 300 Hz to 5000Hz. The VHFD value is in G's rms.


Regards

http://www.compsys.com/drknow/pubtnot.nsf/9f0e142090062...06ec8c0?OpenDocument

[EMAIL]erik.a.concha at shell dot com[/EMAIL]