Please see the attached file.

Just looking at the spectrum most of the analysts would likely diagnose looseness.

Now take a look at the TWF. It is extremely non-sinusoidal. That explains multiple harmonics. But it is also very repeatable with every revolution (T=0.06901 sec)!!!! This pattern could be a result of non-linear vibration response, such as, in a case of looseness.

But could looseness produce a repeatable TWF? Or it is not looseness?

Thanks,
David

Kiln_fan2E.doc (37 Kb, 92 downloads)

Yes, looseness can produce a repeatable signal. Where's your acel data if you used an accelerometer? And if you used a CSI system you may not be looking at tit-for-tat IMHO. An FFT of 1000 Hz and 40 kHZ dosesn't totally relate and to complicate things integrating.

my vote would be looseness.

I think the idea of looseness being a little bit random is mostly taught to distinguish looseness from misalignment. This one doesn't look much like misalignment with so many harmonics. Also note there is some variability in the TWF.

I would also prefer to look at it as an acceleration TWF, particularly with the sharp jumps in the velocity twf (personal preference).

Doesn't really make sense, but I see looseness all the time that is repeatable, even from month to month. But a lot of what I see makes no sense, in that it doesn't fit the "diagnostic charts" that we so often see.

"I'm with Sam. A velocity waveform is not a reliable or accurate indicator. You are better off to keep your waveform in the native units in this case, acceleration. If you have it already or if you can take a new measurement, you'll see quite a difference."

Well, I did some checking and as a matter of fact, it is possible to see looseness quite accurately with the velocity waveform. See attached. 1x peaks.

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

Doc2.doc (32 Kb, 53 downloads)

quote:

Originally posted by vibbase:
"I'm with Sam. A velocity waveform is not a reliable or accurate indicator. You are better off to keep your waveform in the native units in this case, acceleration.

Since my concern was in the low frequency range (1x - 10x) where looseness will likely manifest itself I purpously took TWF in velocity units. After all, the spectrum is calculated based on an integrated from the acceleration TWF. As it is well known acceleration units will make higher frequencies to stand out, thus, possibly obscuring low frequencies. In my case acceleration TWF also was pretty much repeatable.

Yes, looseness (although not always) results in impacting where TWF in g's will give a better visible picture of a resonance ring down pattern, but, again, I was interested in bearing housing macro motion in order to detect looseness which I thought is rather random then repeatable.

As a side note, IMHO, integration errors in
1x-10x range are not that large for condemning the value of an integrated signal.

According to CSI, looseness will produce a periodic/repeatable TWF. (I'm not sure though on whether or not multiple harmonics pattern results from other then looseness phenomenon.) I guess, CSI's statement is not always correct (see Ron's example above).

David

Wouldn't the time waveform HAVE to be repeatable if the spectrum is showing discrete peaks/harmonics?

If the waveform was not periodic/repeatable wouldn't the spectrum just be noise?

One of the ways we confirm looseness is to look at the velocity waveform (integrated from an accelerometer). When once per rev impacting is present, and the the waveform shows a bias more to one side (either positive or negative), then we are pretty confident that looseness is the problem.

The attached document shows waveforms and spectra of a felt roll (on a paper machine) where the withdrawal sleeve has started to come loose, and has worsened.

loosenesswaveform.doc (108 Kb, 38 downloads)

quote:

Originally posted by IanS:
Wouldn't the time waveform HAVE to be repeatable if the spectrum is showing discrete peaks/harmonics?

Ian makes a good point: looseness severity differs. Nothing is black and white. At early stages the TWF is practiaclly repeatable every shaft revolution although the system is already non-linear. In more severe cases TWF becomes less and less repeatable and less linear, which in turn produces more noise and more harmonics in the spectrum (see Ian's attachment).

Ian,
How do you explane the impactive pattern in cases of looseness? Is it in your opinion something like bearing inner race slamming the shaft, or outer race slamming the housing, or a housing slamming the pedestal, or a pedestal slamming the sole plate?

Is TWF assymetry (in velocity units) has to do with directional constant gravity force?

As a side note, I guess, loosesness should be accompanied by possibly several component resonances.

David

quote:

In more severe cases TWF becomes less and less repeatable

I'm not sure that it becomes less repeatable; if it were not repeatable, the specturm would not show clear peaks, it would just be noise, wouldn't it?

quote:

How do you explane the impactive pattern in cases of looseness? Is it in your opinion something like bearing inner race slamming the shaft, or outer race slamming the housing, or a housing slamming the pedestal, or a pedestal slamming the sole plate?

Is TWF assymetry (in velocity units) has to do with directional constant gravity force?

Yes, in my opinion (and that's all it is!), looseness allows movement, which is abruptly stopped when the moving component comes into contact with something fixed. In the example I posted above, the sleeve had come loose on the shaft. So the shaft is effectively free to move radially, which it does. The impactive pattern is (I think) due to the shaft slamming the bore of the bearing (the bearing itself, in this case, is still intact)

I rationalised to myself that the TWF asymmetry is due to (in this case) the shaft being physically unrestrained, and so it can move wherever it likes within the confines of the bearing bore. Due to external loads, eg felt tension, gravity, any other external restraints, it will have more freedom to move in some given direction. If the roll was PERFECTLY balanced, PERFECTLY concentric, and all internal/external forces were PERFECTLY evenly applied, then even though the roll was free to move it wouldn't, as there would be no forcing function to cause it to move.
However, of course, the roll is not perfect, and it sure does move. And it will preferentially move in one plane, and in one direction in that plane (horizontally east, for example). So the roll effectively exerts an impactive force, that is very directional, and this results in a TWF that is asymmetrical.

The frequency of impacting is driven by the forcing function, which is generally roll speed (imbalance, eccentricity, bent, etc)

That's my theory, anyway!