In the attached file there is a classic looseness vibration pattern for a bearing on a shaft - multiple 1/2x harmonics as vibration was measured on the casing. Losseness has been verified - about 0.012" gap was present between the shaft and the sleeve. This is classic.

What is possibly abnormal is development of looseness. On April 10 big 1x with harmonics and practically no 1/2x is present. On May 21 - already fully developed 1/2x pattern, but with 1x being 4 times lower then on April 10. Overall vibration has also dropped by a factor of about 2.

Why did 1x go down? It is clear that relative to the pillow displacement has likely increased due to bearing-shaft looseness, but why seismic vibration at 1x and overall dropped so much?

David

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

Looseness.doc (32 Kb, 94 downloads)

David,

I have some theories (SWAGS?). But first, can you send me a small database with this machine? Belt drive or direct? How is it mounted? What type of bearing? Waveforms? PeakVue? Any pictures? Nice food for thought right after dinner! Dessert I guess..

You have to look at the waveform to understand the dynamics of the signal. Also, vibration is force/dynamic stiffness. Either the force increased or the dynamic stiffness decreased. Looseness in a sleeve bearing will drop the dynamic stiffness, which is oil wedge stiffness and pedistal stiffness.

The waveform will show 1/2x as the primary frequency and will look like impacting. This impacting resembles a square wave, which produce harmonics in the fft. Please post the waveforms. Thanks

How do you get 1/2X harmonics with a square wave? I thought that you only get 1X harmonics. At least that is what I remember from Fourier Series analysis in college.

Bill, Erik, Steve,

Attached are detailed "before" and "after" plots. As expected, the TWF shows a developed impacting modulated at 1/2x. The TWF by itself can be interpreted that looseness (likely in the shaft-bearing fit) is present.

The question though is not how a 1/2x spectrum pattern has been produced. Instead, the question is why 1x and overall has dropped so drastically? If there where conventional alarms set up, this serious fault would have never been detected.

Of course, I made an assumption that neither pedestal stiffness nor imbalance have not changed, but why would stiffness go up and balance improve on its own?

BTW, this is an overhung belt driven fan. The one which has developed the fault is IB, fixed bearing.

My guess is that my original question can be answered only by modeling vibration with non-linear stiffness - not an easy task.

David

Full_plot_looseness.doc (104 Kb, 76 downloads)

David:
I think of it in these terms:
The energy that was previously primarily expressed in your 1x signal is distributed into all of your additional 1/2 time peaks. As your looseness developed clearance and your 1/2 times signal came up, also 1/2 time harmonics came up.
This has sucked energy from your previously dominant 1x peak.

WARNING: Major tanjent
Have you seen this discovery channel show about the mathmatical theories on rogue (ocean) wave generation? Anyway one of the theories describes how in a field of roughly equal amplitude waves moving in the same direction, one wave somehow begins to take energy from the trailing and preceding wave and becomes a wave of 3x normal amplitude. Interesting stuff.


What is your overall trend?

Nevermind, I see your overall trend in the plot. Despite the fact that your 1x plummented, your overall went up slightly.

So your shaft is no longer impacting equally every rotation as seen in the first waveform, but dominant every other rotation as seen in the second waveform. Note also that your acceleration tripled.

Thanks for sharing your data.

Steve, the 1/2X becomes the predominent frequency as is shown in the waveform.

David
Imbalance stays the same but Stiffness actually drops. less stiffness due to more clearance allows the shaft to move more.

The overall drops because the energy is no longer at one discrete frequency. The impacting frequency will be averaged down and spread out due to nonlinearity as David points out. The spectrum overall V-DG( digital overall from 10-24000cpm) is calculated as the sqrt(pk1^2 + pk2^2 ....)

The crest factor gave a slight indication the waveform shape was changing. You could also create a sub-synchronous parameter and set the alarm limit low to give you warning.

Also, change the waveform to velocity. The 1/2x frequency will be easier to see.

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

Dear David, If the fan has a dynamis unblance, it will be increasing by bearing looseness at 1X with clear phase, but for this case, I see no bias to a heavy spot in tha fan, so fan impeller is out of problem, focus on bearing looseness, you may use 2 perpind. proximitors to check orbit or may by oil pressure drop throgh the bearing, replce bearing if you find these levels harmful to your machines.

Regards