I have 2 hi pressure pumps in a HP spray system. They are the same age, shaft diameter are within 0.1 mm the same. One has run with normal service approx 15 years, the other have had fatal shaft fails 2-3 times in that period. Always under the last stage with signs of fretting etc. "Only" (obvious) difference is the connection to the piping, bad one have a bad connection closer to 90 deg pipe turn. Would you believe in hydraulic noise exciting shaft resonance and giving fatigue fail? There are signs during operation on higher levels at 2xRPM and knocktest confirm complete shaft resonance close to 2xRPM, at some used operation point. They are now speed controlled and valve controlled so you can make them run in umpteen combinations. Other suggestions? Mechanics pray on their alignment book that booth are perfectly and equal aligned and fitted as equal as twins. Olov

I would be suspect of the piping configuration and its support. Why? Improperly supported and/or poor pipe alignment can mean thermals will grow into the pump and distort the case.

If hydraulic forces; one would expect to see some honeycomb effect or something to indicate wear associated with those type forces.

I assume these are 25 Hz machines rps for resonance to be 2X RPM?

They are in a steel mill for descaling so the water do circulate but I don´t think it get that hot and they have some hefty performance, test points were: 35 bar/50 m3/h, 90 bar/200m3/h, 128 bar/230m3/h, 128 bar/250m3/h och 128bar/278m3/h speeds are 2250- 2700 RPM. Shaft on the pump that break do crack first, it takes like 3 years to fail, if you look after 2,5 years you have cracks and indications that either the last wheel in the pump is able to move or the shaft is bending so heavily that it leave indications of movement inside the wheel fitting and that is where it snap off if left running. Twin pump have no indications, still running on original shaft. Olov

With a short turn and "no offset reducer"???? one can expect a vortex that can deflect due to unequal pressures thus causing 2X at the impeller leading to fatigue.

It appears you're associating failures with piping configuration? Can this be a possible cause? Exactly what is the configuration.

Oli,

I really can't offer much as far as pump operation, but have you considered that another difference may be in the fluid itself in addition to the piping? Maybe more scale in one system than the other?

Nope it´s only one running at a time and the other is spare at least it´s the same stuff in both they end up in the same spray and suck from the same sump, but there may be some difference in the suction side also, will check that. Olov

Natural critical for pumps are much higher then 2x because added stiffness from wear rings and bushings. You can impact the rotor not running and get the natural frequency which is closer to a worn pump condition.

I don't think resonance is the issue although resonance will contribut to vibration. It's the deflection of the shaft to cause fatigue. What causes this bending motion? I listed one possible cause dependant on unequal pressures on the impeller from one end / one side to the other which will deflect the shaft at 2X. As Oli iterated, it's 2X along with resonance contributing to vibration. And without the 2X; I don't think the resonance near 2X will cause fatigue although I could be wrong. Just my thought.

A client had a pair of centrifugal blowers, single stage, cast alloy impeller, beautiful. One had a very slight warp along one edge that caused a flow induced shove against the impeller along the edge parallel to the shaft w/ each rotation. It used bearings each 3 months, shaft each 6 months and the "twin" gave no problems.

The warp was very slight, not noticable unless you were looking for it. Until the warp was found, the client was insisting on a balance job, but the vibs did not show an unbalanced condition.

Possibly a flaw in the impeller?

OLI, how did you impact the shaft, in the case, in a cradle, running not running? I have a hard believing a rotor resonance exists at 2x on a pump. If OLI impacted the rotor while the machine is down and a resonance appeared near 2x, that is more realistic. Once the machine is running, you can add small bearings at each wear ring and bushing equal to ~12,000 to 25,000 lb's of stiffness each. Something else is going on. As Sam points out something is forcing the rotor at 2x. I have seen pumps break shafts when run way off thier curve.

OLI, please post vibration data and pump cross section drawing to get an idea what this machine looks like. Thanks

Knock tests are done on assembled machine.
Pump 2 is the one that crack repeatedly every 6 years, data is from the new reassembled pump.
Currently all operational points are outside defined pumpcurve, "it hasen´t always been like that" how much bending will it take to crack a pumpshaft? Olov

HPpumps.doc (1,190 Kb, 43 downloads) Pump meas

The vibration signatures presented are taken on bearing housing, if Iam correct. Since the failure is shaft oriented, I would suggest to see the shaft vibration behaviour between these two pumps, which can give more idea about what is happening internally. If you do not have any shaft probe, try to install on both ends (Drive & Non drive ends)on a temporary fixture and collect the data. Shaft vibration is the good indicator of any premature failure on rotors.

OLI bently nevada cover this problem. Read this case study http://www.bently.com/articles/4Q06orbit.htm
and if you dig depth they have a good write up on crack shafts and inlet/discharge at 90 degrees

Oli,
First, I would like to comment on the impact spectrums. All your impact spectrums have a discrete peak at ~20 Hz. Check the waveform data and see if this frequency is present in the pre-trigger data. It could be vibration from a near by machine. The 125Hz frequency has a lot of dampening, which is not typical for a ring test on a machine that is not running. Does the pump have packing or a mechanical seal?
Also, how big is this pump, shaft diameter, bearing to bearing centerline distance. These dimensions will give us a better perspective.

Steady state vibration doesn't seem to be that bad. There's a little looseness but with low amplitudes, doesn't seem that bad.

Erik,

You are correct, 20Hz is from external disturbance we did note that in our report, just didn´t tell to see if you are awake, and you are :-)
Pump has 11 stages and 7 blades it is of a type HGC 4/11 if it give some reference I would estimate the size as a average feedpump. I have requested to get some more detailed size data but everybode is on vacation now. I agree, vibrations are not so hi concidering they run way outside pumpcurve. Historically there has been cases with feedpumps that nice and w/o dramatic vibration pop off the shaft w/o any more drama than that they stop turning due to crit speed excitation especially when extra stiffening/dampening items as seals etc. are worn but what irks me is that they should behave in some way similar. I can´t see the difference that make them behave so different part from piping connection and that is also strange as it´s the pump with a nearby bend that survive and the pump with the "best" path that cracks.
As may be known I am not that a fan of Eddy probes and the casing data is enough to shut it down before it goes too bad and there may be some problem to get a free shaft to measure against. I will still suggest some run up/down tests, maybe also with eddy probes to see if the resonances when running differ. Maybe they have different material in the shafts, such things do happen.... Olov

Some pic´s, bad pump got much higher stack of shims, maybe that decrease damping?
Just happen to be the same site as the OH Crane.
Have not had chance to do run down test, have been playing Tarzan in the crane, not really. Olov

pump_pics.doc (315 Kb, 12 downloads) Pumps