Cavitation After Feed Pre-heating on a 60 m³/h Transfer Pump
A chemical plant transfers demineralised water from an open atmospheric tank to a reactor through a centrifugal pump rated at 60 m³/h. The pump centreline sits 0.50 m above the tank's minimum operating level. The suction line is 8.0 m of 100 mm internal-diameter pipe (Darcy friction factor 0.020) fitted with one 90° elbow (K = 0.30) and a foot valve with strainer (K = 4.5). At rated flow the manufacturer specifies NPSH required = 7.5 m. In winter the tank sits at 25 °C (vapour pressure 3.17 kPa) and the unit runs quietly. Since the plant began pre-heating the feed to 45 °C (vapour pressure 9.59 kPa) to shorten reactor cycle time, operators report rumbling noise, a fluctuating discharge pressure gauge, and pitting on a replacement impeller after only five months. Atmospheric pressure is 101.3 kPa and water density is 990 kg/m³. Maintenance proposes two fixes: throttling the suction isolation valve to smooth out the flow, and raising the tank on a 1.5 m plinth.
1. Calculate the net positive suction head available at 25 °C and at 45 °C, and decide whether the pump is cavitating.
Read the worked answer
Flow area = π(0.100)²/4 = 7.854 × 10⁻³ m², so v = 0.01667/7.854 × 10⁻³ = 2.12 m/s and v²/2g = 0.230 m.
Major loss = f(L/D)(v²/2g) = 0.020 × 80 × 0.230 = 0.367 m. Minor loss = ΣK (v²/2g) = (0.30 + 4.5) × 0.230 = 1.102 m. Total suction loss h_L = 1.469 m.
With ρg = 990 × 9.81 = 9712 Pa/m and the pump 0.50 m above the minimum level (a suction lift): NPSH_a = (p_atm − p_v)/ρg − 0.50 − h_L
At 25 °C: (101300 − 3170)/9712 = 10.10 m → NPSH_a = 10.10 − 0.50 − 1.469 = 8.14 m. Margin over 7.5 m is 0.64 m, which is why winter operation is quiet.
At 45 °C: (101300 − 9590)/9712 = 9.44 m → NPSH_a = 9.44 − 0.50 − 1.469 = 7.47 m. This is 0.03 m below the required 7.5 m, i.e. zero margin.
The coincidence of noise, gauge fluctuation and inlet-side impeller pitting with the temperature change, plus a computed deficit, identifies cavitation. Strictly, the pump is at the threshold at rated flow; any level drop or strainer blockage pushes it clearly into cavitation, and the five-month impeller life is consistent with that.
2. Evaluate the two proposed remedies quantitatively and recommend a course of action.
Read the worked answer
Throttling the suction valve adds a minor loss upstream of the impeller, so NPSH_a falls and cavitation worsens. It also risks damaging the valve and cannot be justified at any setting; reject it.
Raising the tank 1.5 m adds 1.5 m of static head: NPSH_a at 45 °C becomes 7.47 + 1.50 = 8.97 m, a margin of 1.47 m (about 20 % of NPSH_r). That is a sound fix, but civil work on a live tank is disruptive.
A cheaper option worth costing is upsizing the suction line to 150 mm. Then v = 0.943 m/s, v²/2g = 0.0453 m; major loss = 0.020 × 53.3 × 0.0453 = 0.048 m and minor loss = 4.8 × 0.0453 = 0.218 m, total 0.266 m. NPSH_a at 45 °C = 9.44 − 0.50 − 0.266 = 8.67 m, a margin of 1.17 m at a fraction of the civil cost.
Recommendation: eliminate the suction lift and reduce suction losses (upsize the line, or lower the pump, or raise the tank), keep the feed below about 35 °C until the modification is complete, and add a low-NPSH trip. Do not throttle suction under any circumstances.
3. The failed impeller shows pitting concentrated on the suction side of the blades near the inlet, while the discharge pipework shows no corrosion. Explain the location and the gauge fluctuation.
Read the worked answer
Vapour bubbles form where the local static pressure is lowest, which is the suction side of the blade leading edge, just downstream of the eye where the relative velocity peaks and the flow turns sharply. Pitting there and not in the discharge line is the signature of hydrodynamic cavitation rather than chemical corrosion; the discharge region is at high pressure, so bubbles collapse upstream of it. Collapse occurs slightly downstream of the low-pressure zone, which is why the damage band sits a short distance back from the leading edge.
The discharge gauge fluctuation follows from the same physics: each bubble cloud collapses over a timescale of microseconds, so the flow through the impeller passages is intermittently blocked, and the delivery head oscillates. Because the phenomenon is stochastic, the gauge needle wanders rather than settling at a steady reduced head, and pump efficiency and flow both drop.
This matters for the diagnosis: replacing the impeller in a harder material would only lengthen the failure interval. The operating point, not the metallurgy, is the root cause.