Dear all,

We have been taking vibration measures for around 6 months in 6 big fans(more than 100 kW), with roller bearings.
In all 6 machines we are seeing a strange behaviour:
- The vibration velocity values (mm/sec RMS) are stable and inside ISO standard limits (
around 2 mm/sec).
- The acceleration values (filtered 1-10 kHz) vary strongly between measures,
ranging (in the same point, different dates) from around 5 m/sec2 rms to more than 80 m/sec2.
- The envelope spectra do not show bearing fault frequencies, just noise.
- Some of the spectra present ski slope, probably due to the sensor overloaded
due to the high vibration (I don't have the waveforms to check that, but it's
very likely due to the high acceleration levels).

Do you know what can be causing this behaviour? We have asked to check the
greasing, and it seems to be working OK. The machines work at constant load, and
we cannot find anything which could be blamed for the increase in vibration.
We have also checked the cables, collector and accelerometer, and they work fine.
In fact,the measures taken at the fans motors don't exhibit that behaviour.

Please find attached a document with plots of one of the machines. This machine
started to show symptoms of bearing wear two months ago, but the levels went
down on last measure (don't think that the bearing is getting beter ).

Any help will be appreciated.

Plots.pdf (65 Kb, 72 downloads)

High acceleration overalls could be dampers banging, tow motor (locomotive) driving by, dirt buildup on the impellor rubbing on the inlet cone then falling off later. As far as the defective bearing goes, you could have had an overgrease situation whereby the balls stop spinning out of the load zone and then start again when they hit the load zone. This usually shows up as a skid mark on the outer race of the bearing in the load zone. My 2 cents.

AHG, I think you are losing something in all the "processing" that is being done to your data. I would mount an accelerometer firmly to the problem bearing, with a "solid" metal path between the accel and the bearing if possible. Stud mount is best, but a good flat pad epoxied on and a high-strength, flat magnet will also work OK.

Then look at the raw, unprocessed acceleration waveforms. If possible, look at these real-time so you can see what is happening over time.

Thanks for your responses.

Regarding the data taken, I don't have the acceleration available. However, the measures as taken as Rusty suggest (epoxied pad and flat magnet).

We will try to make the point suggested by Ron be checked. However, I do suspect that the problem us that bearings are not working properly, with the rollers skidding on the races. What I don't know is how the check that without dismounting the bearing (which right now is not feasible.

Best regards, and thanks again.

You mentioned roller bearings, are they cylindrical, or spherical roller bearings?, what is the calculated load on the bearings? I have seen bearings "skid" if they are too large for the load, or not installed properly. Skidding (friction)I guess in theory could cause high G's Accel. If you can obtain the bearing #s and discuss the application with the bearing manufacturer (not the fan manufacturer) you may make a discovery.

Just my 2 cents.

Alan

All goood responses guys. We don't have enough info here. My experience with enveloping is that skidding and over/under lub will send levels high and it will be up and down. At first there will be no defects, then later will develope. Skidding is a major problem with fans, especially overhung at the bearing next to the fan.

I have never really noticed accelleration readings going up until defects start to develope.

It seems like every new fan that I have seen lately have been over bearinged

I don't quite understand all that is showing here. Am I looking at, in the 2 top spectrums that are showing the "peaks", data taken between 1 and 10K Hz or is this normal data from 0 to 1K Hz and the bottom data is taken between 1 and 10K Hz?

If these peaks are bearing defects, what race are they assigned to, inner or outer?

Are you using acceleration enveloping with the third filter? If not try the third filter. I had a double roller that was very similiar. Had to change to a 10 mg/V accelerometer with a 500 G cap to moinitor. Turned out the bearing pedstal was flexing so as to effect a dual load on the rolling elements. Stiffened the pedstal and solved the problem.

Dear all,

Thanks for your replies. More data on this.
- The bearings are SKF 22230CC/W33, which is a spherical roller bearing, if I'm not wrong.
- The fault frequencies correspond to the Outer Race defect, although they are moved around 6 Hz from the exact value (maybe another symptom of skiding?).
- I don't know about the load calculations and so on, as we don't have the data yet. I hope to get it next week.
- The envelope spectra shown are taken using a 4 - 6,3 kHz band pass filter. We have also used 1-10 kHz and even shock pulse measures, all with similar results.
- In the data attached previously, the top graph is the trend of the acceleration value (1-10 kHz) and the botton graph is the envelope spectrum corresponding to the date selected on the trend (maybe the representation is a bit confusing).

Best Regards

Am I reading your first post correctly, are all six fans showing this same type of spectrums as the ones posted on this fan? If not....... on this posted fan is it possible that the outer race is crawling in the housing allowing the "defect" to be there today and not there tomorrow?