Non synchronous dominant peak at ammonia booster pump

We are experiencing high vibration at ammonia booster pump. The spectrum shows dominant non-synchronous peak at 1.19 order. This pump has an John crane mechanical seal (109B) in tandem arrangement. Is it something related with mechanial seal? Has anybody experienced any similar phenomenon?

[Attached is the vibration report along with spectrums]

Thanks in advance.

Attachments

Original Post

Hi,

I did't suspect that this peak is a symptom for mechanical seal problem!! From previous history  the amount of droplets come out from mechanical seal is good indication of it is condition.

I suppose that some how change if fluid condition can cause this, Is there any change in flow parameters?

I face a case ones that due to clearance failure inside pump (the fluid was chemical and impeller material did't suitable for it) similar peak appear.

 

VIB.Welt posted:

Hi,

I did't suspect that this peak is a symptom for mechanical seal problem!! From previous history  the amount of droplets come out from mechanical seal is good indication of it is condition.

I suppose that some how change if fluid condition can cause this, Is there any change in flow parameters?

I face a case ones that due to clearance failure inside pump (the fluid was chemical and impeller material did't suitable for it) similar peak appear.

 

Dear VIB.WELT,

This peak is not new. In last 5 years data the peak was pesent in the spectrum, however its amplitude was not significant.

Also, this pump was just taken in sevrice after its maintenance (to attend leakage from its mechanical seal). The impeller was fine and in good condition during physical checks.

We haven't experienced anychange in flow paramaters. The motor ampere also remained same. 

Insp Engr posted:
Raul Rios posted:

Hi !!

Please collet a FFT 60 to 30kcpm @ 1600 line.

 

Dear Roul, 

What do you expect by this action?

Do you suspect bearing fault frequencies in this range?

I would like to see more clearly what is occurring in this frequency range.
They may be bearing frequencies, but you must know that bearing is installed.

Ralph Stewart posted:

Can you post a new 500 Hz Fmax and 6400 Lines Resolution, without any peaks positions marked on the spectrum and the actual  running speed  located and corrected and the cursor sitting on top of your frequency of concern?

Dear Ralph,

The pump is driven by fixed speed motor 3000 rpm (50Hz). Also we have taken data using 12800 Lines of resolution but in our scenario 1X (order) or running speed peak is totally absent from spectrum. That is really something we cannot understand

Raul Rios posted:
Insp Engr posted:
Raul Rios posted:

Hi !!

Please collet a FFT 60 to 30kcpm @ 1600 line.

 

Dear Roul, 

What do you expect by this action?

Do you suspect bearing fault frequencies in this range?

I would like to see more clearly what is occurring in this frequency range.
They may be bearing frequencies, but you must know that bearing is installed.

Dear Raul,

The spectrum is clean and doesnot show any peak in this region.  

nsp Engr quote:

"The pump is driven by fixed speed motor 3000 rpm (50Hz). Also we have taken data using 12800 Lines of resolution but in our scenario 1X (order) or running speed peak is totally absent from spectrum. That is really something we cannot understand"

I am only wanting to see the low end of the spectrum, with a fixed amplitude scale of 1.0 mm/sec rather than a 4 mm/sec vertical scale. If there is a 1x in the data, this should show the 1x if the data is  displayed at this 1.0 mm/sec scale with 6400 line in 10 orders.

I am not sure if you have looked at the spectrum with a vertical scale less than 4.0 mm/sec and expanded out to a 2 or 3 or 4 or 5 orders Fmax or not, in search of the 1x frequency.

Also a 500 Fmax, actually taken at this frequency, will show no more than 10 orders with 6400 lines, making things clearer at the 1 order range. A 500 Hz Fmax (10x) at 12800 lines would or should really separate the frequencies out in a more clear fashion.

I am guessing that the 2970 rpm is the assigned 1x to the measurement point and not the "real" 1x under running conditions.

If there is no 1x in the data, this machine is running extremely good , other than for the 1.19x frequency being there.  My guess is 1x and 1.19x are either so close they are not separating with your lines in your data or something is wrong.

I have seen equipment that had VF Drives and the analyst would not even know it. I am not saying this is the case but IMO, it is worth confirming.Plot 2

 

 

Attachments

Photos (1)
Ralph Stewart posted:

nsp Engr quote:

"The pump is driven by fixed speed motor 3000 rpm (50Hz). Also we have taken data using 12800 Lines of resolution but in our scenario 1X (order) or running speed peak is totally absent from spectrum. That is really something we cannot understand"

I am only wanting to see the low end of the spectrum, with a fixed amplitude scale of 1.0 mm/sec rather than a 4 mm/sec vertical scale. If there is a 1x in the data, this should show the 1x if the data is  displayed at this 1.0 mm/sec scale with 6400 line in 10 orders.

I am not sure if you have looked at the spectrum with a vertical scale less than 4.0 mm/sec and expanded out to a 2 or 3 or 4 or 5 orders Fmax or not, in search of the 1x frequency.

Also a 500 Fmax, actually taken at this frequency, will show no more than 10 orders with 6400 lines, making things clearer at the 1 order range. A 500 Hz Fmax (10x) at 12800 lines would or should really separate the frequencies out in a more clear fashion.

I am guessing that the 2970 rpm is the assigned 1x to the measurement point and not the "real" 1x under running conditions.

If there is no 1x in the data, this machine is running extremely good , other than for the 1.19x frequency being there.  My guess is 1x and 1.19x are either so close they are not separating with your lines in your data or something is wrong.

I have seen equipment that had VF Drives and the analyst would not even know it. I am not saying this is the case but IMO, it is worth confirming.Plot 2

 

Dear Ralph,

Thanks for your valuable inputs.

we had already taken data with Fmax restricted to 500 Hz and 12800 lines. 1X amplitude with respect to dominant peak (1.18 X) in spectrum is almost negligible. please see attached file for further information.

Also, there in no VFD connected to this specific motor (Just for Info)

Attachments

Files (1)
Fmax 500 Hz and 12800 lines

Thanks Insp Engr.

I see that 1x appears to either have a harmonic or it is sideband modulating the 1.18x peak.

The 1.18 peak has a broad floor noise similar to fluid noise.

My guess now is that, you possible have excessive "wear-ring" seal clearance, either on the front side of the impeller and/or on the back side, (depending upon the design.)

See the attached image.

Just my opinion and I could be totally wrong.

Attachments

Photos (1)

When I see a 1X peak with well defined sidebands that I have no explanation for, I just think something besides the shaft is rotating and modulating the turning speed. I see this a lot with bearings that have tapered sleeves that are loose on the shaft and turn a little slower than the shaft.  But I think I saw this once a long time ago on a pump impeller with an attached wear ring that was loose.  Somebody else said maybe a pump bushing (rubber bearing) could also cause this and that is probably more likely.  A rub tends to excite a lot of harmonics.  I bet it is a simple problem.  Maybe best not to overthink it.

"A lot is possible. But most of the time a little is sufficient."

Is there another machine in close proximity that is running at your 1.18 orders? The 2.36 order would then be the 2nd harmonic, both peaks show "natural" frequency pattern, broadband skirting at the base of the peaks.

Typically, sidebands around blade pass frequency are an indication of wear ring issues, and the time waveform would identify if there is modulation which would mean there is sidebands.

Dave

Ralph Stewart posted:
My guess now is that, you possible have excessive "wear-ring" seal clearance, either on the front side of the impeller and/or on the back side, (depending upon the design.)

See the attached image.

Just my opinion and I could be totally wrong.

YOU ARE RIGHT,1.nX=RING WEAR

Add Reply

×
×
×
×