Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
Y7

Member
Registered:
Posts: 4
Reply with quote  #1 
I would like to ask you to share if you know different technique on how to analyze sleeve bearing at early stage of fault if only casing measurement is available? Is it really possible?

We are not able to effectively use casing measurement data (or at least lacking confidence) to detect sleeve bearing problem at its early stage.

Some techniques which may help:
- waveform pattern to detect abnormal G spike
- increase of multiple harmonics of TS
- look for raising noise floor
- using live view to detect possible oil whirl

Is there any other method which we can add to the list?

Is the casing vibration in this case credible trustworthy technique? Can We generate spectrum, polar plots, bode plots?
We have readonablr motors and the turbines with more than 800 kw which do not have proximters installed on their sleeve bearings and for which we do only casing vibration and rely on a reference value like a baseline.
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #2 

Yes you can. Sleeve bearing machines are no different than other machines. Problems creating forces in rotating parts are taken up by and creating loads in the bearings.
We have found that the similar methods used on anti friction bearings do give data also on sleeve bearings for touching as they only detect metal-metal contact anyway. What was used by accident was our simple way, a HP of 600Hz, velocity, RMS and touching was detected before irreparable damage on a new machine not seen by any other way part from the blue spot when taken apart.

Yes you can detect problems related to multiples of 1xRPM as on any machine.

Yes if you trend the 10-1000Hz RMS data you can see also seal touching and other touching coming and going if that is the case on your monitoring.

Yes oil whirl, instability pop up at the slightly less than 0.5xRPM area as it is supposed to etc.

All this is proven many times around here as we normally have both sensor types on larger turbines but also some that do not have eddy probes.
So collect the data and the problems may be seen. It may in some way be depending on design, huge blocks of steel do tend to make the levels on the bearings lower.... but the indications are there. Problems with shaft surface, runout etc. does not exist.

 


__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
vogel

Sr. Member
Registered:
Posts: 76
Reply with quote  #3 

Sleeve bearing machines are different than REB. Models for both type of bearings are different and dynamics too.

That said, you may have some success using casing measurements on sleeve bearing machines. But you would want to use an online system to capture transient data, as well as measuring phase to view plots in polar and bode formats.

I've seen instabilities diagnosed with casing sensors, but I've also seen operators confused because the prox probes trends of their pump had doubled but the route accelerometer trends were flat. On the other hand, I've seen some generators were the velocity trends showed a steady increase and pointed clearly to a developing fault while the prox probes data was not so conclussive.

There are two areas on which I've always wanted to do some testing with casing measurements and sleeve bearings. One is pointed by Oli, friction and HF vibration. Many say that it works, including me, but I have to see yet a protection system configured in such a way to protect a sleeve bearing machine from excessive rubbing. The other one is Blade/Vane Pass vibration in turbines and compressors. Should find some time one day.
John from PA

Sr. Member
Registered:
Posts: 963
Reply with quote  #4 
In my experience, casing measurements on sleeve bearings yield "too little, too late".  Assuming just a gravity loaded rotor, the vibration might be more or less central within the bearing clearance.  In such an instance the transmissiblity to the casing might be low, and what is measured on the casing may not be representative of any issue.  If we take the same amplitude of vibration but move the shaft to a high eccentricity ratio, i. e., much closer to the bearing wall, then the transmissibility ratio is likely much higher and the casing may give a better reading (with respect to diagnostics).  

Machine type may also be a factor.  A gearbox at or near nominal load is a good example where casing readings can give adequate readings for diagnostics.  Under nominal load the bearing eccentricity may approach 0.8 to 0.9.  The minimum oil film thickness is often only about 2 mils (50 um).  Under those conditions transmissibility to the casing is excellent. 
Shurafa

Sr. Member / Supporter
Registered:
Posts: 310
Reply with quote  #5 
Y7,

You have what you have until you have other tools.

In general, case measurement does not necessarily reflect the shaft measurement. It may however in some case, depending on how much of the force is transmitted from the shaft to the case, the amount of damping and other factors.

If the rotor's weight is significantly larger than the weight of the case, there is a chance your reading will help. A common example for this case is boiler forced draft fans mounted on fluid film bearings. Some of these machines have permanently mounted case vibration sensors.

Regards- Ali M. Al-Shurafa
OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #6 
I have had the joy to follow 2pcs 38MW sleeve bearing motors on a 4 year journey in various places.
1. FAT on massive concrete foundation, north Europe, works to spec both absolute and relative reading.
2. Back to back test on 300 mm steel bar "shims" on large concrete on riverbed clay south Europe, due to coupling problem balancing was not possible as there was no space for 3Kg weight, 1Kg worked as expected but not enough. Absolute and relative was not really agreeing, hi absolute eg. above 3mm/s and high relative eg. above 100 microns. Also balancing solutions differ, a good run absolute do not give a good relative solution.
3. Second try same site modified mounting w/o "shims" solved coupling problems. Quick trim balancing apprx. 0.3Kg using relative probes getting good results both absolute and relative eg. less than 2mm/s absolute and less than 40 microns relative.
4. Full machine train full load test in Asia on a 13m hi pedestal out of wood, steel and rubber elements on clay river bed. Relative just within acceptance less than 60 microns, absolute hi levels, no data reported as regarded irrelevant but estimated 4-5mm/s or more.
5. On site on 13 m hi concrete foundation with earthquake isolating rubber element on bedrock northern Europe. Solo unloaded the absolute data is low less than 2mm/s or even 1. Relative it require a trim balancing of 3-400g that is school book balancing performed on relative data (compensated) reducing also absolute to less than 1mm/s. Only 1 is balanced solo, the other just about within limits.
6. Coupled, operating, full load. Solo balanced motor behave perfectly within spec both absolute compensated and uncompensated 30-40 microns and absolute less than 2mm/s.
7. Motor not balanced solo was coupled full load not within spec relative but ok absolute. Balancing was done using the calibration data from the other motor solo balancing. Trim runs was required and this had after the above travels, dropped on concrete while loading, returning to factory, standing on shore for 3 years w. a tarp a runnout about the same as acceptance level. So with finally some luck a very good balancing was possible after 3 adjustments like within ISO g 0.4 and it was good absolute and relative both with and w/o compensation as the uncompensated was the only accepted value to be considered. About 5-600 g was needed. Note that rotors have full balancing protocols within tolerance when they passed the factory at various times.
So why this long and winding road? My conclusion is: You may need both sets of data sometimes. Your foundation do also influence this.

__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
MJ

Member
Registered:
Posts: 5
Reply with quote  #7 
Eddy current probes (x,y,key) are the gold standard for sleeve bearing machinery diagnostics, however I've had good success using casing mount accelerometers on propulsion steam turbines, main reduction gears, boiler feed pumps, and forced draft blowers on Navy ships in a routine CBM program. In general, these machines have rotor weights that are significantly greater than the casing weight.

Shafts supported by journal bearings typically ride on an oil wedge. We locate the accelerometer adjacent the wedge to ensure best transmission of the high frequency excitation. Most of the documented narrowband analysis techniques apply when diagnosing machinery faults.

The accuracy of diagnostics with casing measurements on machines with sleeve bearings decreases with stiffer bearing foundations and when rotors are significantly lighter than the bearing support structure.

In the absence of Eddy current probes, case mount accelerometers can be a suitable substitute, as long as they are mounted in the load zone.

Shurafa

Sr. Member / Supporter
Registered:
Posts: 310
Reply with quote  #8 
OLi,

What is your key point from your last post?

I guess you wanted to say based on your experience, there is a strong and sufficient correlation between the vibration collected from the case and the condition of the machine, right?

Maybe most of us would agree on the concept but some would have a different opinion when it comes to the details. Maybe, a good example for this is when a light rotor installed inside a heavy case like the classic Elliot PAP compressors. Do you think a case vibration would be a good representation of the condition for this machine?

In my opinion, the case vibration might be too late/ inappropriate for this machine. But I could be totally wrong as in many situations.

Regards- Ali M. Al-Shurafa
 

OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #9 

One point was that the foundation of exactly the same machines do influence the results of the data.
Stiff foundation at FAT results can be considered to be about the same.
Soft foundation absolute data is higher, relative are in comparison lower.
Final foundation that can be seen as somewhere in between absolute data is low and relative are in comparison higher.
You should not have a runout in the same range as the acceptance level even if the standard says it is only to be good when leaving factory not on site.

So it is not only the rotor size in comparison to the bearing design but also on what the train is placed and how the foundation is designed.

I can also discuss the GBX monitoring where after 3 years storage absolute sensors on one GBX gave seriously higher readings
than the other one and the relaive probes are about the same (all implemented to API standards). Sensor was swapped, data remained the same. If it is a indication of
a different status btw. the GBX's that I think it is or as end user find it as a freak happening, time will tell. Within a year I think if they run them 24/7.

So I argue, it is not only design of the machine that influence the result of different monitoring but the design of the complete setup including foundation
and different machine types benefit more or less from different monitoring and in more than a few cases my opinion is that absolute monitoring give better and earlier
detection of specific problems than relative and in some cases it is the other way around.

I do try to play the devils advocate here to get some discussion. I tried to do that also with the end user here that also was responsible for another site by asking, "How many 
problems have the relative systems detected IRL on machines that have them on sites that you know of so a failure was prevented? Not any case right off. "How many false stops can the relative system be accounted for that was false, measurement errors or wrong interpretation that gave undue downtime at great expense? Quite a few.

I know a similar comparison should be discussed also for absolute monitoring. You should maybe not discuss and compare apples and oranges but the question was sort of put forward and maybe it is worth thinking and considering that machines should be monitored in the best way possible to give the opportunity to detect and plan ahead and all data should be considered when evaluating the status of a machine IMHO. Only my 1SEK and in this specific case and machine you may be correct but I need some data over some time to evaluate. In general I would say that if you look at trends you should see a problem emerging on any of these systems if the problem was technically able to detect it and I am happy to maybe exaggerate slightly by saying that absolute systems can detect more problems on a wider set of machines with better accuracy and repeatability.  

 


__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Shurafa

Sr. Member / Supporter
Registered:
Posts: 310
Reply with quote  #10 

Sorry but I’m a slow learner.

Maybe I would tend to agree with you on that case measurement is very good, better than relative (shaft) vibration and even best option if we consider all machines. The majority of the machines in my world are mounted on rolling element bearings. So, if I would select only one type of measurements, I would defiantly select case vibration.

 

Assume we limit the population to only machines of fluid film bearings. Now, I have to select the single common method of measurement for all of these machines, what would you advise me?

 

Regards- Ali M. Al-Shurafa

OLi

Sr. Member
Registered:
Posts: 1,917
Reply with quote  #11 
In theory it may be relative if all things are perfect but looking at the reality it is another story. Comments from end user above was mainly a comparatively large population of large sleeve bearing machines.
Considering the runout problems and the in my view less than optimal way of mounting relative probes in the nodes of the machines, the majority of monitoring are optimized to minimize the signal to noise ratio and get the worst result. I have worked with measurements all my life and these systems are one of the few industrial systems where the applicable standards define such procedures. My general argument is that a measurement system where the noise level regardless from what is in the same level as the acceptance and or alarm levels are not that good and prone to give bad data. I also argue that vector monitoring of large machines that are stable enough to use it are the best way to have a defined set of data to trend to indicate most main problems. I know, some machines misbehave and have vectors moving all the time but most larger machines in steady state are reasonably stable for both absolute and relative data in vector form.
So what do we want to achieve part from filling a standard spec.? A stable monitoring that give trend information and where we can set warning and alarm levels as far as possible from noise levels and other interference.
We used to have a local turbine manufacturer up to nuke size turbines here, it is now split and sold but machines remain, they made a rule based "expert" system for larger machines above 500MW or so and they found 10-15 fault cases that was possible to extract and define based on cases found with data world wide, these use both absolute and relative systems but originally not to API definition, so look at the cases I can think of:

No.        Fault             Absolute detection      Relative detection
1.          Unbalance     Yes                            Yes
2.          Alignment      Yes                            Yes
3.          Touching        Yes                            Yes
4.          Instability      Yes                            Yes
5.          Loss of part see No. 1
6.          Various flow    Yes                           Yes
7.          Stuck casing    Yes                           Yes
etc.

Pls. fill in the rest more exotic ones....... and also the one I can't find where relative systems are the only and where absolute fail.
So I argue all problems found in rotating parts create forces that are taken by the bearings and if they are vibration in nature they give varying forces and give absolute vibration of the bearing. There are cases where severe absolute vibration are not detected by relative probes due to soft mounting and or flimsy bearings or sensor mounts so the probe move, vibrate in the same way as the shaft and no relative movement exist and that is bad, see full load test in above example.
In the opposite way, when you have large relative movements and small absolute, will a fault not be possible to trend in the absolute data even if the levels are low. I believe they can be monitored taking in account the lower levels. Will you be able to handle and fix a problem before touching the bearings and seals, maybe not. Would you really as was the case with the earlier example motors need to balance one down to the limit of what was possible likely near the perfect G 0.4 of ISO balancing 1940 or whatever the number is nowadays to "record player" acceptance levels (vinyl that is and a 30-40T rotor)? I argue that absolute sensor data with good sensors... do have a higher signal to noise ratio compared to the relative even in the worst and stiffest and heaviest machinery. And bad signal to noise ratio do kill all your endeavors as you look at a measurement error not at the machinery fault signal and so absolute can be as good or better as relative in most cases. Pls. present a specific IRL case where it was not so. I do however agree that on large machines you should be able to pay for both systems for best monitoring available but eddy probes give better signal to noise data mounted as far as possible from bearings IMHO. It will also make balancing more fun as it is not easy to optimize balancing weights both for absolute and relative data.... even H/V or X/Y are not always that fun, considering 4 signals per bearing is a pain. I also argue that most fluid film bearings have about the same oilfilm thickness as in one extreme you will have instability and in the other you will get metal-metal contact and designer aim to end up in between having the best margins they can design for. So in my view it is the sensor placement, measurement principle and design and mounting of the machinery that create problems for various measurement systems.



__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.