Skip to main content

Replies sorted oldest to newest

@Registered Member posted:

its a oil pump from with variable speed 0 - 1900 rpm

Centrifugal pump or gear type pump?

Any idea why you have a 330400 accelerometer for use on a pump?  Typically one measures in velocity units, mm/sec RMS for Europe.

You probably should see what if any applicable ISO specification covers the pump, or go back to the OEM and see what the recommendations might be.  But lacking all that I'd suggest you set something up, take a measurement where you intend to mount the transducer and then set accordingly.  Your 50 m/sec^2 would be equivalent to about 5 g's.

Any similar machines that have already been set up with a BN3500 rack?

Last edited by Registered Member

Thanks John,

axial flow pump

as for ISO 13373-3 for a Bearing this type 2 330400 in 90°offset should be good to see if the bearing is good/nedds to be checked/bad.

we just want to check this one bearing as we had it fail one time, no recommendation from MFG

never mounted a 330400 for bearing condition monitoring - we just setup 330400 for gearbox vibration monitoring 50 mm/s pk recommended by MFG

Use of a 330400 is a bit unusual for monitoring a bearing, hence the reason I asked.  It is a perfectly acceptable transducer for a gearbox where the monitoring  may be directed to gear condition, hence mesh frequency.  Typically the units of measurements are velocity units, mm/sec, for rolling element bearings as opposed to acceleration units.

Unfortunately I don’t have 3500 rack software (I retired BN in 2011) to check to see if you can integrate the signal to velocity.  But you should be able to check that and perhaps use something better than acceleration.  If you know the bearing, check the anticipated bearing fault frequencies over the speed range because you may need to set up some band pass filters.

So if no other alternative try my proven L-method.
Filter HP 600Hz to avoid most signals not from bearing.
Measure velocity, RMS, mm/s.
Good, less than 0,5 mm/s
Warning above 1 mm/s
Alarm above 2 mm/s

It is free and used for 15 years on everything from papermachines to turbines by now nearly 500 possible users.

So if you don't have anything better, try it. Levels are rules of thumb and may be adjusted for your specific case and trend is your friend and as a last addition, do you have temperature monitoring? Hope it may be useful. Method was developed in cooperation with University of Linköping Sweden and extensive practical testing.
I don't think I seen anything above L 12-15 mm/s surviving any longer time, my rear wheel hub unit aka wheel bearing on my Volvo XC90 was L 7 mm/s when I got it back from the workshop and used a lathe to spin in, just as a reference. At that stage my kids complained about the noise in the back of the car.

@Registered Member posted:

So if no other alternative try my proven L-method.
Filter HP 600Hz to avoid most signals not from bearing.
Measure velocity, RMS, mm/s.
Good, less than 0,5 mm/s
Warning above 1 mm/s
Alarm above 2 mm/s



I'd basically be in agreement for a fixed speed machine but the OP has stated "its a oil pump from with variable speed 0 - 1900 rpm."

Again, as I said earlier, "If you know the bearing, check the anticipated bearing fault frequencies over the speed range because you may need to set up some band pass filters."

I may not have the latest versions in this series of ISO documents.  Their intent are for condition monitoring and diagnostics.  As far as the range for input amplitude in condition monitoring, make sure you do not exceed the range and distort the signal. 

Part 3 gives logic, flow charts, faults/symptoms, etc.  It does talk about both rms and peak on the section for rolling element bearings (informative, not normative).  If you are going to follow this standard some of the advice here may not be relevant.

Other parts of the standard may help, like part 1.  For me personally, there is some good information in this series, but there are things with which I do not agree as being good diagnostics - use your judgement.  Much is advice compared to normative (again my copies may be old - I do not use these much.).

The frequency range for you measurements is not the clearest in part 3.  Part 1 gives some guidance.  Sometime  you may want to set an alarm level; this may not be where you get that.  Although, there is a figure and some text in acceleration for rolling element bearings.

There are many techniques mentioned to do diagnostics in this series. You are not likely to have them all available, but there should be enough for a good CM program.

L value can be used with variable speed, like the old ISO 10-1000Hz can as it is broadband with only hardware limiting upper freq. but 12kHz is suggested if possible lower may be used as a compromise. Bearing must rotate thus. Very low speeds do give lower levels like a few 10ths of RPM so you need to adapt to the case. Thumb rules are for pumps, fans 1500-3600 RPM approx. The trend will be there anyway.
Worst case you may need to select normal operating speed ranges for measurement to get good trends if you can. That would be required with most methods anyway due to the nature of the signal I think.

@Registered Member posted:

So if no other alternative try my proven L-method.
Filter HP 600Hz to avoid most signals not from bearing.

OLi, wouldn't this setup miss running speed faults like coupling faults?

Though the OP specifies the interest in the bearing monitoring, I would be careful to assume it is safe to simply drop lower frequencies. If my math is correct, HP 600 Hz means removing a range around 0-20X shaft speed.

Maybe I'm missing something.



Regards- Ali M. Al-Shurafa

That is the principle OP wanted a bearing method and it is that only. You can't have the cake and eat it. If you like machine fault monitoring, config another 10-1000Hz velocity RMS parameter and monitor that as usual. You can't have both in one, then you have no ability to separate machine/bearing faults w/o FFT.
We have a converter doing this in hatdware and giving 2x4-20mA outputs operating in a P&G diaper plant in Saudi for the last 10 years in that case also with bearing temp that is a good compliment.

@Registered Member posted:

That is the principle OP wanted a bearing method and it is that only. You can't have the cake and eat it. If you like machine fault monitoring, config another 10-1000Hz velocity RMS parameter and monitor that as usual. You can't have both in one, then you have no ability to separate machine/bearing faults w/o FFT.
We have a converter doing this in hatdware and giving 2x4-20mA outputs operating in a P&G diaper plant in Saudi for the last 10 years in that case also with bearing temp that is a good compliment.

I'm with Ali on this.  The OP very specifically stated he wanted "...to monitor a deep groove ball bearing Qj214 with 2 BN Sensors 330400" and later stated "...its a oil pump from with variable speed 0 - 1900 rpm", then further added its an axial flow pump.  So he clearly wanted to monitor a machine and the use of a high pass filter at 600 Hz would be a questionable practice.

Was not the OP problem repeated bearing fails w/o warning? Why not monitor relevant parameters for both purposes compared to a parameter that again warn rather late in my experience or what are your suggestions to solve this then?

wow, thanks @all for that feedback i did not expect that and it is very interesting to read!!!

well, we really just wana monitor the one bearing - not the whole pump - as this is the only roller bearing in the pump and the rest are sleeve bearings-   not haveing any problems at all - also the coupling is not an issue as it is driven directly by a gear from the gearbox.

The mauals i have from BN also telling me that 330400 "can" be used to monitor roller bearings frequenzies of harm.

i know that there might be better systems/sensors/solutions out there but as we use BN3500 and have 330400 sensors in stock i hope this would be a "good" way for monitoring and will give us an indication when the bearing will start failing?!"

thanks in advance,

BR

@Registered Member posted:

Was not the OP problem repeated bearing fails w/o warning? Why not monitor relevant parameters for both purposes compared to a parameter that again warn rather late in my experience or what are your suggestions to solve this then?

OLI, relative to your statement "Was not the OP problem repeated bearing fails w/o warning?"  That is correct, and I could be wrong, but I think that Simon was in a situation where he had nothing on the bearing, and now desires to move to monitoring its condition in hopes of getting some warning.

@Registered Member posted:

The mauals i have from BN also telling me that 330400 "can" be used to monitor roller bearings frequenzies of harm.

i know that there might be better systems/sensors/solutions out there but as we use BN3500 and have 330400 sensors in stock i hope this would be a "good" way for monitoring and will give us an indication when the bearing will start failing?!"

thanks in advance,

BR

The 330400 can definitely be used, but is a bit unusual for monitoring a bearing.  Normally a velomitor would be suggested for a rolling element bearing.

I'm not expert on bn3500 system, but a a 100mv/g accelerometer, with +-3db 10-15kHz, it good for bearing monitoring, the challenge is how to trend a parameter with the speed variations, as Oli is suggesting a high pass filter will do fine for bearing monitoring if you have a walk around data collection that will take care of coupling/unbalance/alignment problems

I think the real issue here is OLI’s statement “Filter HP 600Hz to avoid most signals not from bearing”.

I’m assuming the use of “HP” by OLI means high pass; the reality is that would exclude most if not all of the bearing fault frequencies.

Last edited by Registered Member

John, with roller bearings it is to late to wait for the fundamental fault freq's when they pop up your time to failure is short. If you like to follow the bearing failure you need to look at the gathering of the vast number of multiples that come with a new damage, (normally), there are always exceptions. Anything else is a misunderstanding. Reason that I follow a broad band parameter is that I sum all signals I find from any kind of hi freq fault that normally are bearing related (exceptions are normal) and even if you have a gearbox that give a hi value naturally, it will be increased so you can follow a trend if a bearing fails unless the gear related signals go up or down at the same time but that is kind of a different story.

fburgos, that is not correct. A HP of 600Hz takes away all the things you list. It put forward the higher freq's where bearing fault freq's exist and their multiples and in this case it  is those I like to trend as that was the OP problem all other problems on your list are (mostly) filtered out.
As I written above, if you like to monitor other problems like those on your list use another filter like 10-1000Hz or 1500Hz and you see no bearing problems until very late.
Only my view of this that I used the last 40+ years on-line as well as route. How you collect the data does not necessarily relate to how you process and interpret the data. I basically (with some exceptions) use the same principles not depending on how I acquire the data.
So I can just say even if contradicts your faith, this works and I look forward to you presenting a different approach that do the same I can suggest at least one other way that give less stable trend but earlier detection if you like that or use both as I often do,

I can't find the defect frequencies for the QJ214 ball bearing.  The "Q" by the way is for a 4-point contact bearing as opposed to the normal 2-point contact.

The 7214 BEP bearing is of similar physical size (ID & OD) so the defect frequencies might be similar to what would be expected from the QJ214 ball bearing.  I used a shaft speed of 1900 RPM to generate the anticipated frequencies of the 7214 BEP bearing (see below).  You should follow up directly with SKF for deterining the defect frequencies of the QJ214 bearing.  They readily share that information to help the end user in the identification of any faults.

The BN3500/330400 combination should work well for the anticipated issue.  As you can readily see, a high pass frequency of 600 Hz will exclude the frequencies of interest that are commonly used for monitoring and diagnostics.

7214 BEP

Attachments

Images (1)
  • 7214 BEP

I normally use 10x the basic fault frequency of a bearing fault for monitoring and that is common praxis here. Waiting for the basic fault freq. is normally waiting to long if you like to have some time to plan activities.
When a standard fault is new it create what is close to a Dirac pulse since it is so sharp and signal include in theory infinite or in real life  very hi frequencies like the prehistoric SKF parameter at 340kHz or SPM and various shock pulse method at approx 23-32kHz etc. etc. However when damage get older it flat out and the energy maximum will lower in freq. eventually lower than the std filter LP 1000Hz but it will be late, very late so maintenance is required directly. So a HP of 600Hz does enhance the signal over the most time the bearing fault signal exist. It will however take away signal from other mechanical problems as previously listed and they would be monitored by another parameter like 10-1000Hz.

I can describe a great variation of parameters that are based on similar filtering like the old CSI paramter g value or what you like to call it that is a HP of 2kHz so that would be even worse according to your theory then?
All Envelop, Demodultaion, PeakView or whatever you call it are based on various filtering principles most HP 2000Hz, various BP filtering etc. etc. most higher than 600Hz so I go low in freq. to cater for reasonably low RPM machines. It is all a compromise and my flavor is normally simple to implement and can be used in most systems of any brand and I have used it in at least 5 different brand system.
So I think you misunderstand the aim for this theory or are all bearing fault monitoring principles of every brand in the world that monitor bearings based on a misunderstanding and why do we all on regular basis find bearing faults that are verified by the bearings found to be faulty?
So you don't need to know the specific fault freq's unless you specific benefit from knowing what part of the bearing is faulty and how the fault spread.
Basically the trend is your friend.

If you don't monitor anything else for a roller bearing check temperature.
Can your problem be a low bearing load by mistake? Roller bearings have a minimum load specified and if the load are lower the rollers may skid and friction increase temperature. It may not increase vibration unless a damage is created but in may be having a increased noise level.

Add Reply

Post
×
×
×
×
Link copied to your clipboard.
×