We presently have 2 centre hung, 250kW, single suction, 8000 l/s @ 60ºC, 15kPa dp fans, 2 pole DOL direct coupled (omega coupling) motors operating in a kiln scrubber Induced draft duty on our plant (duty and standby fans). The fans were recently commissioned and the vendor has basically stopped answering the phone. The problem is that after first commissioning the first fan only ran for 100 hours before catastrophic bearing failure. The second fan was brought online and only ran reach 40 hours before similar failure. Bearings are directly (no sleeves) mounted, 22217 C3 in SOFN housings (oil bath lubrication). The fans suffered a failure at the DE on one machine and the NDE on the other. The bearings that had not destroyed the cages and collapsed did show much cage and centre guide ring wear.

We reassembled the fans with the utmost care, and recommissioned and watched the bearing vibration and temperature with close attention. From the moment the fans were started I knew the root cause had not been addressed. 10 p-p spikes were present in the accel wave at 1x and 1x harmonics were prevalent in the accel and vel fft (both bearings)0-10khz accel is about 0.8 g's. Of more concern was the amount axial displacement that was present in the shaft (could be seen with the naked eye - about 0.5 to 1 mm). The bearing config is standard with the de locked and nde free to float with thermal elongation of the shaft. We shut down the fans and inspected the bearings to find very fine metallic wear partials in the bottom of the housings, some signs of rubbing on the ends of the rollers but clearances not greater than a brand new bearing. Production pressure dictated that the fans come back online with the same bearings. I was able to have strain measurements collected from one of the fan shafts during operation. This indicated transient spikes of compressive and tensile (reversing) load as high as 20kN (not good for bearings). The base loading is around 11kN. We are operating with the inlet dampers wide open to try and stabilise the transients after recommended by another fan vendor. Currently we have only 400 hours on one of the fans and I am seeing the same amount of wave form impacting. Velocity data now indicates more harmonics with the same 1x peak as commissioning (1.4mm/s). Overall velocity levels have gone from 4mm/s 0-1000hz to as high as 10mm/s due to the increase in 1x harmonic activity. I am trying to convince production to remove this fan and bring the new standby fan online to allow inspection of the near failed bearings in the machine that is presently online

While surfing the net in search of some other examples of fan problems I came across an old posting from this site in 2001 that fits the description of our problem almost exactly. Can anyone tell me what the final fix to this problem was? All the symptoms are similar although the fans are a little different. I am interested in what happen as apparently the fan had been running well until maintenance - where as our fan is brand new. Many are telling us that it is a rotor stability problem and we need completely new fans for satisfactory life. If I can find out what the fix was to the 2001 posting I may be able to try this on our fans. Presently my next option is to install a carb bearing at the DE and the sph roller at the nde with new housings. This will hopefully allow faster replacement of the sph roller and allow the de carb bearing to operate without any influence from the transient axial load. It is not addressing the root cause but we need the machines online while we come up with the exact root cause and address it by modifying the fans or replacing them.

Any help would be appreciated

Anthony

001331[1].html (50 Kb, 77 downloads) Old Rel Board Posting

Have you performed a bump test on the fan

I am wondering if your running speed is near natural frequency and resonance.

I am questioning shaft and fan stiffness but still if your near a resonance and if you could change speed some it might help

just a thought.

quote:

Originally posted by Anthony Powe:

Overall velocity levels have gone from 4mm/s 0-1000hz to as high as 10mm/s due to the increase in 1x harmonic activity.

Anthony,


The increase in harmonic activity of 1x in FFT, considering impactive pattern in the TWF, tells me that you have rather periodic then random impacting at 1x. (Previously you were mentioning transient pattern of impacting...).

Just an observation but it may lead you to a different avenue, such as shaft or fan bearings alignment, as an example.

Could you please post some plots? If you are using CSI take PeakVue TWF to see actual impact amplitude.

I'm also curious to know how have you performed strain measurements.

Thanks.

My quick vote on fault cause is loose bearing fit on shaft combined with excessive bearing clearance. By all means check for rotor critical, foundation resonance, and rotor balance; but you have to start with correctly installed bearings. Also make sure the inboard (DE) bearing is the thrust bearing and the outboard (NDE) floats axially.

Walt
w_f_strong {at} msn [dot] com

Have you checked the oil to see if it the correct type for these bearings and temperature?

The fan temperature may be causing a deflection to occur in the shaft and putting the bearing in an axial overloaded position. Are you sure the non held bearing has the proper amount of space to "grow" and is not in a "bind" and can not grow?

Are the bearings rated for the weight and speed of the fan?

I'm not familiar with fans like that or the resolution of that thread.

Similar along the lines of David's questions:
- Was the tensile/compression variation measured in the axial direction? Was the transient portion varying at 1x?
- What was the speed of visible movement of the shaft? Running speed? (that would be tough to see).

It might be interesting to take a strobe and confirm the frequency of that shaft movement if you haven't already.

While you've got the strobe handy, take a look at that omega coupling. Even though the evidence stronglhy suggests the problem comes from within the fan, it's easy to check... and you have to wonder whether that axial shaft movement will fatigue the coupling (especially if you have anti-friction bearings on the motor). Coupling failures can be ugly.

I suspect Walt's and Ralph comments are on the mark.

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

Anthony,

Has the rotor been balanced before putting into the machine.
Is there a balancing report available?

Just a thought

Greetings from Tim

To tell an opinion is neccesary to see a spectrum.
On my experience , once, after a fan was repaired due a reheating backflow in a furnace, the first start up yield visible movement of the shaft and the elastic supports (I was afraid while took data amplitude and phase faster possible), then balance. Was neccesary to use almost 1 kg of attached weight. It was in 1991.

Thanks for all the help so far, I am surprised that the response has been so great over a weekend.

I will try to answer all the questions asked in the following paragraphs.

Mike H:
We performed a bump test before balancing and found nothing around 1x. I did not delve any further into this subject as our vibration data did not have any elevated peaks that alerted me to resonance (hope I am right in saying this).

David_G
Impacting in the vibration twf is periodic, but the stain data indicates that transient spikes in axial compressive and tensile stress are being sustained. I have attached an acceleration wave form – initially I felt this may be due to lubrication as the original oil levels were on the high side. Lowering these to the SKF recommended level did not change the impacting.

Peakvue TWF is attached

I have also attached a prelim report from the contractor who performed the strain measurements. Please note that there are still a few errors in this report that are being fixed. Just focus on the trends that are displayed.

Walt Strong
Bearing config is DE thrust with the outer race located in the housing. NDE is free to move axially in the housing with 5mm axial clearance between the outer race and housing end covers.

The bearings are both shrunk fit onto their shafts with an m6 interference fit. The bearings are installed as follows – the inboard labyrinth sleeve is installed hard against the shaft shoulder, then the inner race is heated and installed, the outboard labyrinth is then installed and a lock washer and lock nut installed to hold everything in place. No evidence of looseness was found between the shaft and bearing inner race ways.

We confirmed that the NDE bearing was able to float by loosening the DE housing end covers and pushing the shaft back and forward axially within the clearances of the NDE housing.

Ralph Stewart
We are using mobilith SHC 629, originally it was Mobilith SHC 626 (10cst at 100ºC)but the bearing manufactures recommended increasing the viscosity to that of SHC 629 (20 cst @ 100ºC) for the operating temperatures that we see.

Refer to the comment above about how we confirmed that the thermal shaft growth is not restricted.

The bearings are rated for the weight and speed for the fan. They are actually a little on the large side for the loading that we have so skidding is a real possibility. I have tried to be vigilant by paying close attention to oil levels. Could skidding be the problem? I am not sure how to exactly rule this out.

Electricpete
Tensile/compression was measured axially, refer to the attached prelim report for more info – remember that this is a prelim report so don’t worry about the words to much. Just concentrate on the data.

Axial velocity has not been measured, I am still tempted to set up a prox. Basically it was easy to see the shaft bounce back and forward randomly. No strobe was required to see this. I will take the strobe out and inspect the housings/shaft/structure for any displacement.

Like many of you I have seen the result of omega coupling failures. There were no signs of bearing fatigue when the fans were last stripped down a few weeks ago.


Tim EPP
The rotors were both balanced to G1.0. Initial max 1x vibration was 3mm/s at the NDE vertical of one fan – all other readings on both fans were less than 1.5mm/s.

I have shut down the fan from which the attached vibration data was collected, we are now running on the second fan and I am awiaiting delivery of the spectral data for this machine - presently online readings are 3mm/s at both bearings. Hopefully I can get some good information from the fan taken out of service. A quick check by levering the shaft axially has shown there is quite a bit of clearance - more than would be expected for a well located sph roller bearing.

Thanks

Anthony

Acceleration_Data_from_DE_bearing_of_one_fan.doc (172 Kb, 34 downloads) Some vibration Data

Strain testing report is attached. Please temper this with the knowledge that it is a draft report. The shaft diameter requires correction (affects the load amplitudes, the report shows lower than actual values as a 85 dia was used instead of 96), both bearings are sph roller (report indicates one is cyc roller) and the fan was run on clean air and flow controlled by varying the flow of water to the upstream venturi scrubber.

Enjoy!

Anthony

APJ-013-505A_report.pdf (1,397 Kb, 32 downloads) Strain testing report

OK, I feel like a wild boar looking at a wrist watch, no idea of what I see, so help me out here.

What does this strain test conclude, too much, too little or just the right amount of "tensial/compression" on the shaft?

A single suction setup, Hmmm. Strange that both fans are failing in basically the same way. Looks like there would have to be a design flaw somewhere. Is the suction on the DE or NDE bearing?

I have never seen a strain gage used on a rotating shaft. High tech stuff.

Out of curiosity, did you ask those if strain gages are affected by rotation (centrifugal force), and what kind of frequency response they get (can they see variations at 1x).

Also a little disappointing they connected the two strain gages as shown in the report.... any bending stresses of the shaft gives a 0 output. All that’s left is axial thrust. But thrust couldn’t have killed the outboard floating bearing which you confirmed was floating. So I don’t understand why the test was set up to surpress any bending stress from the output of the strain bridge.

You said you don’t need a strobe to look at the shaft movement. Can you characterize the movement. Was it a lot slower than 1x?

There is a method to check for loss of axial and/or radial clearance during changes of operation, especially thermal changes, using a combination of envelope spectra and normal spectra. To do this, you make repeated sets of envelope and normal spectra measurements, measuring the first several sets of running speed harmonics of rotation frequency, especially 1x and 2x and 0x, the overall average level. Usually the best way to do this is to do a cold start and then trend continuously until all the operating parameters including temperatures stabilize. I know that you have checked for float but sometimes there are surprises.

This message has been edited. Last edited by: Duncan Carter,

Anthony,

It looks to me that there is excessive clearence (axial and radial) in the bearing.
Take a look at the attached data you have originally provided. This "innocent" very low frequency hay stack (circled)is caused by the shaft motion you see with the naked eye.
There are actually two types of motion here: one at a very low random frequency ( don't think it is a trivial ski slope caused by noise, it is real motion) and another one with periodic impacting at 1x, also a result of internal looseness.

Could be chronic bearing installation problem.

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

Low_random.doc (24 Kb, 28 downloads)

My concern is still why would 2 seperate fans fail in almost the same way if it were not a design flaw or installation error?

Anthony,

Do you have any pictures of the bearings, not just off one fan but off both?

Thankyou for all of the feedback provided so far:

Ralph –
The equivalent dynamic load derived for an L10h life of only 4380 hours (1/2 year – minimum life acceptable) I found that for a radial load of only 1Kn per bearing (basically the dead load of the rotor) our max axial load is only around 8kN – the report indicated nominal axial loads of around 15kN – no wonder we are having trouble. Rough theoretical calculation of the thrust produced by the single suction radial flow impeller conducted before the strain testing indicated that the thrust bearing was large enough. The suction is on the NDE side of the fan.

I also feel the reversal of loading is causing accelerated wear in the bearing cages and inner guide rings in both the DE and NDE bearings – these are the components that must are loaded by the inertia of the roller complement when load is reversed.

Electricpete –
Minimum data sampling rate is 32768 samples/sec – more than enough to capture a 1x cycle. I do not know the freq response characteristics of these gauges as they are mainly used for static measurements – I will ask the question though.

As for the bridging – I only asked for axial strain – and that’s all I got basically. I would have been handy to see the bending moment in the shaft as this would have made it possible to see the effect of the rotors radial load. It may have only complicated the subject though.

Movement is definitely slower that 1x – I will try to capture it with our digital camera – it just looks like someone is hitting the other end of the fan shaft with a hammer – basically the shaft randomly bounces out and then back in again.

Duncan –
Thanks for the tip – I will use this at the next cold start up with new bearings

David-
Spot on with your comment about the low freq vibration – some of the fan boffins I have spoken to have indicated this may in fact be the frequency which recirculation or another fluid dynamic problem is producing. I have attached a vel spectra collected when the fans were operated with low damper openings of only 20% - you can actually see 6hz sidebands around 1x and the harmonics – we no longer use the dampers to control air flow – they are only used to start the fans.

Ralph –

The only aspect of the installation I am not comfortable with is that the bearings and housings were not aligned to the shaft centreline. Most installations with sph roller bearings do not require detailed alignment (correct me if I am wrong of course). For the next bearing installation I will request that the housings and bearings are aligned to the shaft by using a shaft mounted dial indicator measuring the axial runout of the outer race and bearing housing faces. This may not be the silver bullet that fixes our problem but it will give any bearing a fighting chance.

With the knowledge in hand at the moment it certainly starting to look like we have all aspects of installation and operation covered – now I need to put my design hat on to keep these things running while a high level decision is made on funding for new fans.

Thanks all for your help so far - I will keep you posted.

AP

6Hz_1x_sidebands.doc (25 Kb, 7 downloads) Vel Spectum with roughly 6hz sidebands around 1x harmonics

Final draft strain testing report attached. Loads have increased since the shaft diameter was corrected.

Results - Mean axial thrust was typically up to +15 kN (tensile). However, during
transients, thrust values ranging between +34 kN (tensile) to -30 kN
(compressive) were found in the data. These transients are of short
duration, most being about 4 mS in total duration. However, some
extended periods of unstable thrust were found in the data, up to 0.5
second.

APJ-013-505A_report.pdf (1,397 Kb, 4 downloads) Final draft of report

Anthony - if you do what I suggest, take enough averages for each envelope spectrum to get a fairly smooth noise floor in each spectrum. In our diagnostic systems, we use 12 averages as a default for this. You don't need to use more than 1 or 2 for the normal spectrum.

In view of all the excellent analysis you've done, I almost hate to ask this, but have you physically verified that you don't have a 'rub' in this fan? The best way to do this is to slow roll the fan (preferrably by hand) while 'listening' to the housing with a screwdriver or stethoscope. A rub can look like looseness, and can also cause excessive axial loading/thrusting. If you have inlet cones, this is especially an area of concern.

No inlet cone rusty - just a flat baffle seperates the fan volute from the inlet duct.

Your are right though, the 1x impact could be a rub - the guys went to the trouble of moving the bearing housings slightly to allow the impellers to rotate with out touching in the casing - maybe we have had some movement.

I will check as you have described

Thanks

AP