Skip to main content

This topic has had discussion in the past but:
Collecting vibration data - system integrity check to ensure valid data (transducers, cables, instrument) - goes without saying. But placing the transducer solidly and time delay plus no cable strain against the transducer; how careful are we?

Spirit vial level: calibrate before each use? Calibrate as necessary. Allow thermal stability? Are all taking the time to ensure valid data?

Optical alignment: calibrate before each reading or each new transit mount - if you move it any, recalibrate!

Dial indicators: calibrate each time or have established stamped known repetive values in dedicated brackets.

Laser alignment: check calibration each use at every alignment job? Or assume it's OK.

Do you think data integrity is addressed adequately in the work place? Interested and curious.
Original Post

Replies sorted oldest to newest

Is it possible that analyzers are really just "smoke-and-mirrors" devices, that they sort of make stuff up? I ask this, because there will be several collection points each day that will repeat the previous month's value, to 5 decimal places... and this is not necessarily low amplitude or high amplitude readings, sometimes the machine will be very noisy, and sometimes not. How is this possible? I mean the reading will exactly duplicate the previous month's reading, for example 0.04536 in/sec, peak. This seems almost impossible to me. I mean the odds are 1 in 10,000 that it would randomly repeat. And this is not using stud mounted accels, attached pads, or even a magnet.
I have noticed a similar phenomenon with my Entek data collectors. I have them set up to use "most recent" as a previous reading display. I believe the collector will use this value as a starting point for auto ranging. If I go to off route, the data can be slightly different and if I jiggle the transducer it can change the value somewhat. We are dealing with very sensitive instruments and if you, Rusty, like myself collect data at the same point at the same interval over many years, I don't see why a stable piece of equipment shouldn't read the same every time. Although I'm not aware of the reading being the same to 5 decimal places!
I've had an overall level repest but never all frequencies having the same magnitude or amplitude of vibration. Never! And I also think never on 5 dp on an OA.

If I had one identical I would suppose it as an old db that hadn't been written over or added to!

Again: checking calibration. I like to set two transducers side-by-side and compare amplitude/phase. Having a new one for a comparator is good.

I will say I have full confidence in my data and rely on it when personal feeling or opinions are different. Example: machine feels smooth as the baby's rear but data says it's rough - I go with the analyzer. It always proves correct. No smoke and/or mirrors.
We had a bad once with this problem. A summer student running data into an FFT analyzer and even listening to the data as he was doing it. When I checked up on him a while later, there was a nice stack of IDENTICAL graphs. 10 different machines, but absolutely identical graphs.

A little head scratching and looking at some live data revealed the problem. The analyzer had a "hold" button. It was taking a 16 ensemble average of the same data every time the average button was pressed!

Didn't each measurement sound different I asked. "Yep." was the reply. Don't you think the grpahs should look different if the data sounds different? "Well, I wondered about that."

In the submarine world, there are 1/3 octave criteria for each type of machine. These criteria, developed from average data, usually should higher levels in the bands containing shaft rate, vane rate and other parameters. The simplistic approach is to look at levels and if they're below the criteria, call the machine good. One of my co-worker's method was better. The data should be below the curve, but it should look something like the curve - there should be higher levels at shaft rate, vane rate and other forcing frequencies. If the data doesn't have a similar shape, it might not be a valid measurement...perhaps the machine was at the wrong speed or even the wrong machine was tested.

I'd be suspicious of data that never changed in level even in the decimal places! Even measurements taken one after the other show some variation. You might try intentionally taking a "bad measurement", maybe the wrong location or the accelerometer hanging in air. If it still looks the same....well, you can be sure something's not right!

Spintelligent Labs

Add Reply

Link copied to your clipboard.