Vibration increase at only a narrow temperature band 1

[Nick22]

I have a very unusual pumping system. It is a closed loop hot oil system, (Syltherm 800) being pumped by a Gould's 3620 API 610 between the bearing pump at 2975 rpm, 2250 m3/hr. The pump starts from ambient and is heated to 300 deg C. The flow is maintained constant. Is part of a heating/cooling system for a reactor so the manual valve are set and the flow is stable. The pump runs very well, with in API guidelines. The vibrations increase from 2 mm/s to 3.5 mm/s as the pump heat up to operating temperature. The problem is that as the temperature approaches 250 deg C the vibration starts increasing dramatically/exponentially peaking at over 10 mm/s at 255 deg C. As we move away from 255 degrees to a higher temperature to 265 deg the vibration falls back to the 3 to 4 mm/s range and stays lower up to 300 degC actually even improving as getting closer to 300 deg C. This is the first case I have ever seen a pump having such a dramatic increase in vibration at a narrow temperature band within the operating temperature range. I and many colleagues have been analyzing this problem for years and have eliminated may possible root causes. I am interested in others feedback, theories or similar experiences. The vibration is at vane pass frequency and the impeller is a six vane impeller. This being a double volute pumps with the slitter and cut water being 180 degrees apart can be problematic but not the root cause of this phenomenon.
Thanks,

 

[GPRnD]

Since the vibration band of interest is vane pass and temperature dependent the two things that come to mind are:

1. Changes in support stiffness as the structure heats up. It may be that the thermal expansion provisions on the pump/baseplate and/or pipe loadings is changing with temperature. This might cause the overall structural stiffness to change such that the modal response coincides with vane pass frequency. You could verify this either through bump testing or through an ODS.

2. The fluid properties may be changing sufficiently such that the speed of sound changes and allows an acoustic resonance to occur. This can be checked by measuring the pressure pulsation profile in the suction and discharge with a high frequency (piezo) pressure transducer.

Perhaps you could list some of the root causes you've examined and discounted ?

 

[Nick22]

Root cause we have worked on or eliminated.
1. NPSHa too low. Calculation work fine and the vibration improves even as the temperature and vapor temperatuer increases.
2. Impeller six vanes and cut water splitter. Have modifiec the cutwater splitter profile and reduced impeller size to 410mm from 416 mm. Max impeller is 432 mm.
3. Piping changes. Increased the piping inlet straight diameters to 5 diameters from 2. Outlet is 1.5 diameters.
4. We have done some analysis of the support and have plans to "beef this up" if you will.
5. A bump test for natural frequency was done early on and eliminated.
6 A CFD anlaysis on the piping flow was done.

It is interesting that your number 2 is still on our list of possible root causes but very difficult to analyze. Are you aware of anyone who does such an analysis? I would think this is envasive and would require connections/nozzles to be made in the piping? What can be done to remedy if proven so? Antinodes or pulsation dampeners? Who designs these? As you can understand this is why this still remains but is looking more like the reason.

Thank you very much for your response.

Regards,

 

[GPRnD]

We've used MSI (Mechanical Solutions Inc) in the past. They are pretty good with tricky pump problems. Since it looks like you are outside the US, they may not be your lowest cost source however.

If acoustic resonance is found to be the cause, the solution generally involves either changing piping and/or putting in a pulsation damper. Because you have high temperatures a pulsation damper wouldn't be my first choice. Changing pipework typically involves playing with the locations of pipe section changes which act as a reflective surface.

As a last resort you could always change the excitation magnitude and frequency. I'm sure the pump supplier could provide a split and staggered impeller with 5 or 7 vanes (most big pump suppliers have alternate impeller designs in their back pocket). You could also consider cutting back one volute lip to give you a 168 deg volute angle. I've found this can work in some instances.

 

[Nick22]

This pretty much puts us on the same page. We did make a modification to the volute lips cutting them back to the 168 deg angle with some improvement on the peak vibration seen. 20 mm/s before vs 9 mm/s now but not sure if it was this modification or increasing the piping suction diameters. We have explored the 5 and 7 vane option but that is pretty much a completely new pump from the ground up, (from the manufacturer). Seems like we need to continue down our six sigma path and continue to validate the remaining root causes. Down to the hard ones now. I was hoping someone would have the magic bullet or at least some other options.

I am currently in the USA but the problem still resides elsewhere.

Thanks,

 

[1gibson]

Pretty much everything else has been mentioned, so have you considered that it could be a rotordynamic issue with the oil viscosity at that specific temperature? Bearing and wear ring clearance (due to thermal expansion, and due to wear) would modify the rotordynamic characteristics of the pump. Do you notice different problem temperature ranges with new bearings/wear rings vs old? You say you've been looking at it for years, so I assume at least one refurb?

I suppose one (expensive, and almost certainly impractical) way to check would be to change the heat transfer fluid to a different grade, and see if the problem tracks temperature, viscosity, or specific gravity. You'd need a lot of before and after data to pin that down.

 

[Nick22]

Three different pumps with all the same results and the vibration pretty much in the same temperature band within a few degrees C. We are very vigilant about our measurements and document them. I believe different heat transfer fluid would be last for expense and envirnmental reasons. Found a very cool case study that fits very close. Link below.
Thanks,

http://turbolab.tamu.edu/proc/pumpproc/P24/Case Study 01.pdf

 

[GPRnD]

Here is my favourite acoustic/vane pass study. I remember being astounded when it was presented.

Basically it is a checklist of what not to do...

[URL unfurl="true"]http://turbolab.tamu.edu/proc/pumpproc/P27/Rotor%20Retrofits%20Improve%20Pump%20Station%20Vibration.pdf[/url]

 

[Nick22]

Thanks much for sharing. This has been quit an enlightening experience the last couple years. I am feeling more confident that we moved down the logical path and now just need to get to the end.

 

[JJPellin]

I have seen this two times. The first time we attempted to prove accoustic resonance. We had a contractor attempt to duplicate the problem with air in the pump using a sound wave generator. It did not work. We tried to prove it using many of the methods you have tried. We failed. We replaced the 4th stage impeller that had five vanes with a six vane impeller with the same performance. That worked perfectly.

The second time it happened the temperature range was narrower (175 to 185 F). We have not resolved this second instance. We are getting proposals for different vane counts. The first time taught us that it is much easier to implememnt the solution than it was to try and prove it.

Our accoustic resonance was ocurring inside the pump and was not detected in pressure pulsations within the piping.

Johnny Pellin

 

[Nick22]

Thanks can you give more detail on the fluid pumped and type of pump? Sounds very similar to our situation except we are single stage double suction. And your vibration went away once above the slot temperature too?

 

[GPRnD]

JJPellin does bring up a good point. Where you measure the pulsation when you have a resonance does matter. The pipe tappings need to coincide with the expected location of the pulsation and not be at a node. Having 2 sets of tappings spaced apart is a definite advantage. MSI claim they can obtain information even by mounting sensors on the outside of the pipe to monitor acceleration, although I've not tried that myself.

I'd presume that that the "inside the pump" acoustic resonance was for a BB3 long crossover ? These can be problematic on high stage counts with fluids other than water as the length is sufficient to allow standing waves in the vane pass frequency.

 

[TenPenny]

I am a bit surprised that changing the number of vanes hasn't been explored in more detail. In your shoes, I would go back to them and push for that idea to be explored in more depth. They have, in their system, the ability to do new patterns for these impellers, I don't see why it would involve a complete new pump, unless the person you are dealing with is focussed on 'catalog' offerings only. if this is a problem that you would like fixed, there's no reason they can't explore it in depth. If you think that it might help solve your issue, I would go back to your factory guy, and push him/her harder. If you are located in the US, the person you need to have involved is in Houston, your factory rep knows how to contact him.

 

[Nick22]

Thanks everyone. The impeller redesign was discussed with the vendor early on and is still on our possible root cause list. It is one of the more difficult items to act upon. The impeller would have to be designed, a casting made and machined. The pump would have to be sent back for API 610 performance testing. Not an easy task shipping a used pump from China to Korea. There is not a shelf impeller of this design available. It would be a brand new design and untested design. At least that is what the Gould's cheif engineer says. 16 x 18 x 17SQ and 7000 lbs.

 

[JJPellin]

The two pumps I mentioned are both horizontal multi-stage API pumps. We also have a Gould 3640 with high vibe at twice vane pass. This pump has five vanes in the first stage with two coke crusher vanes. We had an FCC charge pump of the same configuration as your pump with excessive vane pass vibration. In every instance we had to change the vane count. I would perform a ODS with many, many points to look for a tricky structural resonance. Bearing brackets on this style of pump are often resonant at vane pass. Consider converting to 360 degree brackets. I would suspect suction recirculation cavitation exciting a bearing bracket. What is the suction specific speed?

Johnny Pellin

 

[Nick22]

I feel the ODS would be a good next step. I will look into the bearing bracket. Sounds like you really work with some awesome pumping systems. Specific suction speed is 12000. Really appreciate your input.

 

[JJPellin]

Your suction specific speed is quite high. I would take additional vibration data. I would look at acceleration rather than velocity and I would take data out to very high frequency. Ring each bearing housing in the vertical and horizontal to look for bearing bracket natural frequencies.

Johnny Pellin

 

[Nick22]

Understood. We have a great vibration technician/engineer. We can collect all the peak view data. I will pass this on to him. I was a little concerned about changing the vanes as a solution. Could it just shift the problem to a different temperature band, frequency and see the same problem but just at a differnt place. Could be more tolerable or worse?
Thanks,

 

[Nick22]

What I am saying is our reactor turn around is every 30 hours and we move through this temperature band fairly quickly. If we shift the problem to the 280 degree C area it would be a larger problem because that is where we normally operate at for the thirty hours. 280 to 300.

 

[JJPellin]

You may not have to change vane count. If the ODS show something is resonating, detune that part to eliminate the resonance. If the high frequency data reveals suction recirculation, consider reducing suction specific speed. I have done this with an Ingersoll Rand pump of the same configuration as yours. We increased the NPSHr to reduce the Nss and the vibration was below our limits for the first time in 10 years.

We have changed vane count many times, but have never stumbled upon another resonance. It always solved the problem. You could be unlucky. But it is not likely.

Johnny Pellin