AG88 Engineeringnotes
← all notes

Reading cavitation from the pump curve

2026-07-08 · rotating equipment

A pump that is cavitating sounds like it is passing gravel. By the time you can hear it across the room, the impeller has been losing material for weeks. The useful skill is seeing it coming on paper, before it becomes audible.

Two numbers that are easy to confuse

NPSHrequired is a property of the pump, measured on a test rig, and it rises steeply with flow. NPSHavailable is a property of your system:

NPSHa = (P_suction_absolute − P_vapour) / (ρ g) + static head − friction losses

Every term on the right drifts in service. Liquid temperature rises through the summer and P_vapour rises with it, non-linearly. A suction strainer fouls and friction losses climb. Someone throttles the suction valve to "balance" two parallel pumps. Tank level runs lower than design because the level controller was retuned. None of these appear on the datasheet, and together they can eat a metre of margin without anyone touching the pump.

Where on the curve you are running

Run to the right of best efficiency point and NPSHr climbs while NPSHa falls — the two curves converge, which is why cavitation shows up first on the pump that is doing the most work, not the oldest one. Run far to the left and you get a different failure: recirculation at the impeller eye, which sounds similar and damages the pressure side of the vanes rather than the suction side. The damage pattern tells you which one you had.

A margin worth holding

Vendors quote NPSHr at 3 % head drop, which is the point where cavitation is already established, not the point where it begins. For water-like fluids I want NPSHa to exceed NPSHr by at least a metre, and for hot or light hydrocarbons rather more. If the margin at design flow is 0.3 m, the pump is not "fine" — it is one warm afternoon away from not being fine.

Cheap instrumentation that pays for itself

A pressure transmitter on the suction line and a temperature element in the same neighbourhood let you compute NPSHa continuously and trend it. On one plant this turned an annual impeller replacement into a strainer-cleaning schedule, which is the sort of trade that makes instrumentation budgets easy to defend.