Some time ago I was called in to perform acceptance testing on 7 new pumps in a wastewater application. These are vertical centrifugal solids-handling pumps with a jack shaft going upstairs one elevation to the drive motor. All of them operate at or near 900 RPM, with 5 of them on VFDs (the 2 largest on soft starts). There were lots of problems with air locking and venting of the lines to get the pumps to start pumping. Once we did individual runs of the pumps, I found that pumps 1/2 and 3/4, which are different sizes, were exhibiting some impacting on the thrust bearing, which is a Timken tapered roller on the top side of the bearing housing. The impacting was most evident in PeakVue (was also visible as 2xBSF in normal high frequency data) and was highest on pumps 3/4 (16-25 g-s TWF). All frequencies matched ball spin of the associated bearing with sidebands of cage and harmonics of cage. I called this out and recommended checking lube and preload on the bearings as this was a common pattern to all 4 pumps.
The OEM became involved and got a second opinion, which indicated no faults. Not long after the #4 pump had an over-temperature issue with the pump upper/thrust bearing reaching 400F. The bearings were pulled and inspected and found to visually look fine. The bearing manufacturer identified excessive preload, or lack of endplay, as the cause. I did not see the actual report. The preload was supposedly fixed and the bearings were replaced. Not long after, the same thing happened to the #3 pump with similar results. Nothing has happened with pumps 1/2, which are smaller but have exhibited similar BSF/FTF patterns albeit at lower amplitudes (8 g-s).
I have recently gone back in to the facility and collected data again and found the same pattern, but with #4 exhibiting levels of impacting in PV at around 42 g-s. I also could hear cavitation in the pumps and recorded data on the pump impeller casing and found significant broadband high frequency energy. This was most obvious on pumps 3/4, the biggest pumps (15 MGD each). IR thermal profiles on the pump bearings were very normal.
I have discussed with the customer my concerns and also my hesitation to recommend pulling these bearings given the condition the last set appeared to be in when they were pulled. I find it unusual that the common fault frequency is BSF and cage. Normally PeakVue has been a great backup to normal data collection, especially once you look at slower speed shafts or have to sift through lots of flow turbulence for fault frequencies. Could this be a case where the endplay is still improperly set and the end result are these frequencies? Note that the pumps in question have about 10 run hours on them to date. They have not been greased since the OEM replaced the bearings.
I appreciate any thoughts.