1. Most pumps don’t fail. They’re murdered.
I’ve investigated maybe 40 pump failures. Actual manufacturing defects: two. Everything else: wrong selection, bad installation, or operations running the pump against a closed discharge valve for 20 minutes because “the pressure gauge looked fine.”
A pump operating at 20% of BEP (best efficiency point) is like driving a car in first gear on the highway. It’ll move, but you’re destroying it. The vibration data will tell you months in advance — if anyone’s looking at it.
2. The impeller is usually the wrong size.
Pump suppliers add safety factors. The process engineer adds safety factors. The project manager adds “just in case” margin. By the time the pump arrives on site, it’s sized for 140% of the actual required flow and 125% of the required head.
What happens: you throttle the discharge valve to control flow. The pump runs at a fraction of its design point. Efficiency drops from 78% to maybe 55%. The extra energy cost over 10 years is more than the pump itself. And nobody notices because the electricity bill is one line item buried in “utilities.”
Always check: what’s the actual operating point, not the datasheet point? If the impeller is more than 10% oversized, trim it before installation. A trimmed impeller at BEP is more reliable than a full-size impeller running at 40% flow.
3. NPSH margin is the cheapest insurance in engineering.
Cavitation eats impellers. The repair costs $5,000-15,000. The downtime costs $50,000-500,000. And 80% of cavitation problems I’ve seen were predictable from the NPSH calculation — someone used the pump curve NPSHR without adding the required margin.
NPSH available must exceed NPSH required by at least 1 meter, preferably 2. More if the fluid is hot, the suction comes from a vessel operating near its bubble point, or the suction piping has more than two fittings. The Hydraulic Institute recommends a margin ratio (NPSHA/NPSHR) of 1.1-2.0 depending on the application. Use the high end of that range. The extra 500 mm of vessel elevation costs almost nothing during design and saves everything during operation.
4. The suction piping is more important than the pump.
A pump is a dumb machine. It does exactly what the system tells it to do. If the suction piping is wrong — too small, too long, too many elbows, an air pocket at a high point — the best pump in the world will cavitate, vibrate, and fail.
Rules for suction piping that get ignored constantly:
– Eccentric reducer at the pump suction, flat side up. Flat side down traps air.
– Straight pipe for 5-10 pipe diameters upstream of the suction flange. Every elbow within that distance distorts the velocity profile into the impeller eye.
– No high points between the suction source and the pump. Air accumulates at high points. Air in the impeller = cavitation.
– Suction piping at least one size larger than the pump nozzle. The velocity at the pump flange shouldn’t exceed 1.5 m/s for most services, 1.0 m/s for fluids near their boiling point.
5. Vibration data is free. Not looking at it costs everything.
Every pump above 15 kW should have vibration monitoring. Not the once-a-month walkaround with a handheld meter — continuously, with trending. A pump that’s about to fail tells you weeks in advance: the 1x running speed vibration starts climbing, then harmonics appear, then the noise floor rises. By the time the operator hears “that pump sounds funny,” you’ve already missed three opportunities to fix it during a planned shutdown.
The cost of a wireless vibration sensor is about $500 per point. The cost of an unplanned pump failure is the pump repair plus the production downtime. Do the math once and you’ll never skip instrumentation again.
The common thread: pumps are predictable. They fail in predictable ways for predictable reasons. The information you need to prevent every one of these failures exists before the failure happens. The only question is whether anyone is paying attention.