How to prevent cavitation in a centrifugal pump
Prevent cavitation in a centrifugal pump by giving it the suction conditions it asks for: NPSHa above NPSHr at the duty point, with a margin.
Almost every case that reaches us is a suction tank too small, or a suction head too small. The fixes run system side first and pump side last.
Cavitation is a suction-conditions fault, not an arithmetic one
Cavitation is vapour. Pressure at the impeller eye of a centrifugal pump falls below the vapour pressure of the medium, the medium boils, and the bubbles collapse again as the impeller raises the pressure back up. Each collapse takes a little metal with it. That mechanism is worth understanding once, and it is almost never the thing that was missing.
What was missing is the suction side of the installation. The pump needs a certain net positive suction head to keep the medium liquid at the impeller eye, and that figure is its NPSHr. What the installation delivers at the pump inlet is NPSHa. NPSHa has to sit above NPSHr with a margin, and when it does not, you cavitate. Two conditions account for nearly all of it: a suction tank that is too small, and a suction head that is too small.
Both are faults in the installation, not in the pump. Both get missed the same way. Suction pipework and the tank feeding it belong to somebody else's scope, or they were already on site when the pump was ordered, so nobody ever checks the pump against them. The pump carries the consequence and takes the blame for it.
Anything that pulls NPSHa down will do this to you: a long suction lift, a bore that is too small, a run carrying too many bends, an inlet valve nobody opened fully, a fouled strainer, or a medium running hotter than the one the pump was specified for. Anything that pushes NPSHr up does it from the other direction, and flow is the main one, because NPSHr climbs steeply as flow rises.
Make sure it is cavitation before you change anything
A cavitating centrifugal pump sounds like it is pumping gravel. That hard rattle from the wet end is the giveaway, and it arrives with vibration you can feel through the baseplate and with flow and head sitting below what the performance curve says you should be getting. Strip the pump and the pitting is concentrated on the leading edges of the impeller vanes, not spread across the casing.
Two other faults get mistaken for it, and each wants a different repair.
| Cavitation | Air entrainment | Bearing or seal | |
|---|---|---|---|
| Sound | Hard rattle, like gravel | Coarser, surges, comes and goes | Whine or grumble |
| Changes with flow | Yes, worse as flow rises | Follows the tank level | No, holds steady |
| Flow and head | Both below the curve | Both unsteady | Normal until it seizes |
| Where the damage shows | Pitting on the impeller vane leading edges | None inside the pump | Bearing housing or seal faces |
| What cures it | Raise NPSHa or lower NPSHr | Cure the leak or the vortex | Replace the part, then find out why |
Air entrainment sounds similar and is a different fault. Air is being drawn into the pump. Nothing is boiling inside it. The usual sources are a vortex in the tank because the level has dropped or there is no vortex breaker, a leaking gland or flange joint on the suction line, and a return line discharging above the liquid surface and whipping air in on its way down. Cure the leak or the vortex and it stops. Raising NPSHa will not touch it.
A worn bearing or a failing mechanical seal makes noise too, and by the time a pump has been cavitating for a few weeks it has probably damaged one or both. Noise that varies as you change the flow points at the hydraulics. Noise that holds steady whatever the duty is mechanical, and you are chasing the damage instead of the cause.
The quickest confirmation is the discharge valve. Close it in and the duty point moves back up the curve to a lower flow, where NPSHr is lower and the suction line loses less to friction. If the rattle fades as you throttle, you have proved the fault is NPSH margin. Leave it throttled only for as long as it takes to arrange the real fix, because you are burning the difference across a valve to do it.
Work the suction side first
Every fix here raises NPSHa or lowers NPSHr without touching the pump. They are in the order that pays, not the order that is easiest to arrange. On an installed pump, one of them is usually the whole answer.
Clear the restrictions in the suction line
Clean or remove the strainer, open the inlet valve fully, and check nothing has been left part shut. Somebody shutting a valve and forgetting it is shut is a real and unglamorous cause of pump failure, and it costs nothing to rule out. Then look at what the line is made of. Every bend, every reducer and every metre of length is head you do not get back.
Get the tank and the working level right
The tank has to sustain the flow the pump is running at, down to the lowest level the process ever takes it to. Raise the minimum working level, fit a vortex breaker where the outlet sits close to the floor, and check the fill rate keeps up with the pump. If the tank cannot deliver the flow, nothing you do to the pump will help.
Raise the source or lower the pump
Static suction head is the cheapest metre of NPSHa you will ever buy. Every metre the liquid level sits above the pump centreline is a metre added, and every metre of lift is one taken away. Dropping the pump into a pit, or raising the tank on a stand, often costs less than the pump you were about to order.
Shorten the run and open out the bore
Friction loss climbs with the square of velocity, so a suction line one size up costs far less head than the same run at the pump's inlet bore. Never go below that bore. Keep the line short and straight, use a long-radius bend at the inlet and not an elbow, and fit any reducer eccentric with the flat side uppermost so the line cannot trap an air pocket.
Take the temperature out of it
Vapour pressure rises with temperature, so an installation that runs happily on a cold medium can cavitate on the same medium hot. Check the temperature the process actually runs at, not the one on the original datasheet, and check the cleaning temperature separately on a hygienic line. Where the process temperature has crept up since the pump was specified, that alone can account for the failure.
Then move the duty point, and change the pump last
Change the duty before you change the pump. If the suction side is as good as the installation will allow and the margin still is not there, the duty point is the next thing to move.
Slowing the pump down is the strongest lever you have, because NPSHr falls with speed. A variable-speed drive that brings a pump back to the flow the system actually wants will usually pull the duty point inside the curve and cut the absorbed power at the same time. It pays twice on the media that want pumping slowly anyway. An abrasive medium wears the wet end in proportion to how hard you drive it, and a shear-sensitive product breaks up at speed. Neither constraint appears anywhere on a performance curve, which is exactly why both get missed.
Trimming the impeller pulls head and flow out of the pump, and it is the right move where the pump is simply making more head than the system needs. It does less for the suction side than slowing the machine down, so take the speed first where you have the choice.
Only after all of that does a different pump make sense, and it is worth saying what one costs. A pump with a lower NPSHr is usually a larger, slower machine, or one carrying an inducer ahead of the impeller. You pay for it in capital, often in efficiency, and it does nothing at all about a tank that cannot supply the flow.
Moving off centrifugal entirely is an honest answer where the medium argues for it anyway. A positive displacement pump tolerates a degree of cavitation with the right seal selection, and you get vibration rather than a destroyed wet end. That is a change of pump family, though, and it is only the right call if the medium justifies it: above roughly 3–4% solids as a trade rule of thumb, or where the medium is genuinely viscous. Do not buy your way out of a suction problem with a pump family you did not need.
If the pump was oversized, the fault is upstream of the suction line
Some cavitation was designed in. A centrifugal pump sized for more flow than the
system can deliver runs off the end of its performance curve, where it needs back
pressure it is not getting. Size for 10 m³/h when the tank will only ever give
you 5 m³/h and you are pumping air.
That pump will cavitate however good the suction pipework is, because the duty point was never on the curve to begin with. Speed control or a smaller impeller will bring it back, and that is the usual field fix. The reason it happened is a selection fault, and the chain of stacked margins that produces it is worth following before you specify the next one. See how a duty gets worked out for that, and reading a performance curve for where the NPSHr line sits and how to keep a duty point off the right-hand end.
On a new installation, design the suction side before you pick the pump
On a new installation you design the suction side first, and that costs a fraction of correcting it later. Fix the duty honestly, design the suction side against that duty, then choose the pump against what the suction side will actually give it.
Work out NPSHa at the worst case rather than the typical one: the lowest tank level the process reaches, the hottest the medium gets, and the highest flow the system could ever pull the pump out to. NPSHr climbs steeply to the right, so a margin that looks comfortable at the duty point can be gone by run-out. Then keep margin over it. A bare pass is not a margin.
Treat the pump and the system it sits in as one thing, because most of what presents as a pump fault is not one. A food factory came to us with repeated suction inlet blockages across a bank of pumps and asked us to replace the lot with something bigger. The pumps were sized correctly. Changing the inlet size fixed it, at the cost of a housing rather than a bank of pumps, and there has been no trouble since.
Tell us your duty and we will size it. Send the flow, the head, the medium and the connection, and send the suction side with them: the tank, the lift, the pipe run and the temperature. Where the answer is a change to the pipework rather than a pump, we will say so.
Frequently asked
What is cavitation in a centrifugal pump?
Cavitation is vapour forming and collapsing inside the pump. Pressure at the impeller eye falls below the medium's vapour pressure so the medium boils, and the bubbles collapse again as the impeller raises the pressure, taking metal off the vane edges every time.
How do you stop a pump cavitating?
Raise NPSHa or lower NPSHr until there is a margin between them. On an installed pump that means clearing the suction line, raising the tank level, lowering the pump or shortening the run first, and slowing the pump down with a variable-speed drive second.
What does a cavitating pump sound like?
Like it is pumping gravel, with a hard rattle from the wet end, vibration through the baseplate, and flow and head below what the performance curve promises. Noise that varies as you change the flow points at cavitation; noise that holds steady is usually a bearing or a seal.
Will throttling the discharge valve stop cavitation?
It usually stops the noise, because closing the valve in moves the duty to a lower flow where NPSHr is lower and the suction line loses less to friction. Use it to confirm the diagnosis, not as the fix, because you are then wasting head across a valve.
Does cavitation damage a pump permanently?
Yes, and it takes the mechanical seal with it. Collapsing vapour pits the impeller vane edges and the casing, and a seal running in a vibrating, part-vaporised medium fails early. Correct the suction conditions before you refit, or the new wet end goes the same way.