Checked a large extruder drive motor yesterday that had a heck of a racket at the drive end motor bearing. Sounded like a bearing in very poor condition, but the motor looked brand new. Lots of broadband noise in the spectrum, fairly low g-levels in the acceleration time waveform. This has got to be "looseness" but how? Maint. Supt. said it was replaced last Saturday and was a spare and it was reported to be very quiet when started up. So a newly rebuild motor goes from quiet to bad in 3 days?

Maint. Supervisor pulls me aside a bit later and confides that "We sort of screwed up one of the mounting bolts on the coupling taper-lock bushing.... think the coupling might be loose on the shaft?"

The circular plot of the time waveform shows this looseness very clearly. Acceleration plot looks about the same but with fewer revs.

Looseness.pdf (322 Kb, 213 downloads)

Cool. Anybody remember "Spirograph"?

Rusty
What exactly are you refering to as the looseness in the circular plot.I have been looking at these for some time and do not understand them as much as I would like to so you could probably help quite a few people to get a better grasp of these plots if you could explain this more please.

You're talking about acceleration and TWF in acceleration but I don't see any. And the circular plot --- it's value?

What am I missing?

Rusty,

Do you usually look at twf's in velocity or just hit the wrong key?

Is this a Taperlock (hubless with a set screw) or a QD bushing (with a hub and 3 through bolts)?

The attached file is an "apples-to-apples" comparison of identical motors running the same speed/load. The spectrum of the "East" motor (the bad one) shows high broadband vibration. When I took a high-resolution spectrum some of the noise resolved to distinct peaks, but there was still a lot of noise there. This tells me I have a lot of random movement. You can see this in the normal acceleration waveform, but visually, it's a little hard to interpret. But when you look at the circular waveform plot it's easy to see. Notice on the first plot, for the West motor, that each revolution lays pretty much atop the others. But on the 2nd plot, the East motor, the plots are very non-symetric. To me the value of the circular waveform is it lets you more easily visualize the "periodicity" of the vibration, or the lack of it.

But you have to have the rpm entered very accurately. Otherwise you will have drift in the revolutions on the plot, and it will appear more random than it is.

Danny, I use velocity to diagnose lower-order mechanical faults, but my default waveform collection is acceleration waveform. If I suspect a problem, I take a higher resolution velocity spectrum which give me a velocity waveform with lots of revolutions. The "spirograph" effect of the first plot I posted is due to random vibration in a large number of revolutions. Were this a simple balance problems, the revolutions would overlay almost exactly and look like a single, thick waveform.

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

TWF_CircPlot_Example.pdf (219 Kb, 120 downloads)

Rusty,
I cannot open this last attachment you posted. The first opens with no problem. It sounds like you are showing each revolution this time and that would be interesting to see how the different revolutions compare. You would expect with looseness it would be random and each revolution would be different. We have seen some cracked inner races and that will show in the same location in your circular plot if your speed is set right. Its like you are looking at the load zone. Will you repost this last one?

Thanks,
Ronnie

Ronnie, I redid the small with smaller JPG's... try it now. This was created in Adobe and is a 7.0 file. Do you have the latest version of the reader?

Rusty,
I can view it now. I see what you mean about the individual rotations not lining up in the velocity waveforms.

Thank You,
Ronnie

Rusty,

Of course it is a good visual representation of non-repeatability or randomness in the signal but it is kind of redundant since the spectrum tells it all. Circular representation may be very usefull though in cases when one has to identify bad spot in a gear or a roll.

david

David, actually the spectrum in this case does not tell the whole story. You will often see this spectrum with a defective bearing where there really is very little "looseness" and the associated time waveform will also look noisy (viewed normally), but the circular TWF plot will be much less random than the TWF in this example.

In this case, just listening to the bearing (with both headphones and a "yellow-handled-analyzer") it was impossible to rule out a "bad" bearing. Someone less experienced, or who did not know the motor had just been replaced, would have called this bearing bad and had it replaced. Had I relied on just the spectrum, or not taken a high resolution spectrum (that confirms it is looseness), I'd probably have called this one wrong.

The circular plots are from my routine acceleration waveforms and would have been available had I collected no extra data at all. And they are in fact very distinctive.

Contrary to what is often taught, and to what the wall charts present, hardly any spectrum is specific to a particular fault. Show me any spectrum, and I bet I have one that looks the same, but is a different problem. You should use all the confirmation you can get to make any call, but especially on a machine that has just been replaced, and most especially when the maint. supt. is already a little "hostile" towards you..... He doesn't care that I've been checking these machines for 12 years, or that I'm #3 on the maintenance seniority list (even if I'm a contractor).... he's still "the boss" and I'd darn well better be right!

quote:

Originally posted by rustythevibeguy:
... the spectrum in this case does not tell the whole story. You will often see this spectrum with a defective bearing where there really is very little "looseness" and the associated time waveform will also look noisy (viewed normally), but the circular TWF plot will be much less random than the TWF in this example.

I think the whole story gets told best by comparison ( between E and W in this case), which has been demonstrated in the second attachment. Talking of randommnes and TWF, the circular plot is IMO a better tool then the normal one. Also, randomness/looseness assessment is rather an art then a sceince and trending or comparison is the most reliable method regarless of whether it is a spectrum or circular TWF. Hard to do it from vibration data in absolute terms, since there is always some present.

I wonder what did you find wrong with the bearing?

David

We don't believe there is anything wrong with the bearing. The random looseness is (we think) due to a loose taper lock on the motor-side coupling half (see the original post). If I ever hear back, I'll follow up on this (or in a couple months when I do the next survey).

Rusty,

First, looseness and its vibration signature always been an intricate part of vibration analysis and excites me in a sense that sometimes it is hard to assess its severity. In the above case it is easy, though, since you have something to compare with.

You know as good as I do that in extreme cases, such as loose tapered sleeve, looseness manifests itself as a distinctive series of fractional harmonics, may be up to 5-6 orders (in my experience). BTW, I am not sure if structural ( pedestal, plate, bearing housing to plate looseness, etc.) looseness will generate fractionals.

In your case there is rather a raised floor then distinctive peaks. This floor is protruding to 30x frequency range, and I am not sure if it can be attributed to a shaft impacting the sleeve.

Refer to this for more on looseness

David

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

If you want to filter out any random events and just look at periodic data in your time waveform you should autocorrelate your waveform.