Before You Change the Pump Bearings, Check the Speed

How resonance works

Why one speed can make the structure shake

Every machine pushes on its structure as it turns, because no rotor is perfectly balanced and no shaft line is perfectly aligned. A pump rotordynamics tutorial from the Texas A&M Turbomachinery and Pump Symposium names the forcing frequencies seen most often in pumps: running speed (1×), twice running speed (2×) and vane pass, which is the number of impeller vanes times running speed.

The structure has natural frequencies of its own, the rates at which it rings if you rap it. When a forcing frequency lands on one, each push arrives just as the motion from the last one comes round. The motion builds until damping soaks up as much energy as each push adds. The tutorial puts the usual amplification at 2 to 25 times the movement the same force would cause if it were steady.

A fixed-speed pump can be designed to keep clear of its natural frequencies. On a VFD, every forcing frequency sweeps across a range. The DOE guide to variable speed pumping, written with the Hydraulic Institute and Europump, says this greatly increases the chance of resonance within the operating range, and that vertical pumps are more prone to it than horizontal ones because their natural frequencies are lower.

On a vertical pump, look first at the tall discharge head and motor above the base, which can sway together like a reed: the Hydraulic Institute's pump dynamics guideline calls this the reed frequency. The access windows cut into the discharge head can make it stiffer in one direction than the other, so the frequency in line with the discharge pipe can differ from the one across it, as a Hydraulic Institute pump FAQ explains.

The force has not grown. The structure's response has, and the tutorial warns that resonant motion often damages bearings, seals and running clearances. New bearings fitted into the same structure would meet the same motion.

Same readings, new order

Same readings, a different pattern

Plot the same twelve readings against drive speed. The calculation assumes a natural frequency of 1,320 rpm (22 Hz), 5% damping and unchanged unbalance. It plots vibration velocity at running speed; nothing wears. The four-pole motor stays below 1,500 rpm at the drive's 50 Hz maximum.

The readings now follow one curve, and similar speeds months apart agree. Some rise with speed is expected: SKF's spectrum analysis guide lists that among the signs of unbalance. Here, the sharp climb near 1,320 rpm deserves a closer look.

The model also leaves out the scatter of real readings, and the curve is dashed where no reading supports it. No visit fell above 1,335 rpm, so the descending tail beyond that speed is calculated, not measured. A planned test has to supply it.

This works only if speed is in the record, ideally from a tachometer. The drive's output frequency is second best: an induction motor runs slightly below its synchronous speed, and further below it under load, as the US Department of Energy's motor fact sheet explains.

Record direction too. Take both horizontal directions at the motor top and mark the one in line with the discharge; a reading much higher in one direction fits the directional stiffness described above.

Speed also moves a pump's flow and head, so a hydraulic problem can change with speed too. Our field note on where the pump is running covers those checks.

Confirming the band

Give the suspected resonance a proper test

Test the suspected band under an agreed plan. ISO 13373-2 includes refining a diagnosis by changing operating conditions. Both tests below need a qualified analyst (ISO 18436-2), site permits and operating-team approval.

A slow speed sweep. In a low-demand window, the operating team moves the pump slowly through its approved speed range, or the analyst records through a planned stop or start. At each speed the analyst logs the amplitude and phase of the running-speed (1×) component in both directions at the motor top. Keep the sweep moving through the band rather than holding the pump in it.

Phase needs a once-per-turn reference from the shaft, as Prosig explains; a common choice is an optical or laser tachometer reading a reflective mark on the shaft, as PCB Piezotronics describes.

The lower chart shows displacement phase lag, not velocity phase. Its 90-degree crossing marks the natural frequency. A velocity signal is shifted by a constant 90 degrees relative to displacement, so the size of the phase change across speeds stays the same.

Sensor position and the shaft mark add an offset to a field phase reading, another reason to compare the change rather than an absolute value. The same SKF guide gives the phase values in its fault rules a tolerance of about 30 degrees either way, for mechanical variance. The change of about 146 degrees in the calculated sweep is far outside that.

An impact test. With the pump stopped and isolated under the site's permit, the analyst strikes the structure, often with an instrumented hammer, and records how it rings. Natural frequencies show up as peaks in the response. ISO 7626-5 notes that even a quick impact test gives approximate natural frequencies.

Test the installed pump, because water in the column, the pipework and the foundation all affect its natural frequencies. A clean vibration test at a single speed is no proof either: the DOE guide notes that a pump meant for variable speed is often tested at only one speed.

A sweep shows how the structure behaves while running; an impact test finds its natural frequencies at rest. A practitioner article summarised on Emerson's blog advises at least two tests before anything is changed.

What each pattern points to

Getting out of the band

Choose the fix after confirming the band

Once the band is confirmed, there are two common ways out of it. Neither is a bearing change.

Change the structure. Added stiffness raises a natural frequency; added mass lowers it. For vertical pumps the Hydraulic Institute FAQ lists changes to the discharge head windows or shell, ribs or gussets, the structural height, the flanged joints or the driver.

Keep the natural frequency a clear margin from the whole operating speed range, not just the usual speed. Check 2× and vane pass too, because a natural frequency moved away from 1× can land on another forcing frequency. This is work for the pump maker or a structural specialist, followed by a repeat sweep.

Keep the drive out of the band. ABB's ACS880 firmware manual describes a critical speeds function for avoiding mechanical resonance, and SEW-EURODRIVE's documentation explains how its skip frequency band behaves: if the requested speed falls inside the band, the drive holds at its upper or lower edge, depending on the previous setpoint, and crosses the band on its normal ramp.

The pressure control then has to live with a gap in the pump's range. Agree the band with operations, keep a margin either side of the measured peak and confirm that the remaining range still meets demand. A skip band can be a permanent setting, or a stopgap while a structural change is designed.

If the bearing evidence also points to damage, change the bearings and deal with the band as well. Otherwise the new set goes straight back into the resonance.