Feed Pump Cavitation from a Deaerator with a Fouled Suction Strainer
A deaerator at a textile mill supplies boiler feedwater at 1.2 bar absolute, where the saturation temperature is 105 °C, to a centrifugal feed pump rated 120 m³/h at 1470 rpm. The manufacturer's curve gives a required NPSH of 4.5 m at the duty flow. The deaerator storage tank normal water level is 6.0 m above the pump centreline and falls to 4.0 m at the low-level alarm. The suction line is DN200 seamless steel with an equivalent length of 45 m and a Darcy friction factor of 0.018. A basket strainer in the line is rated at 0.4 m pressure drop when clean. Commissioning was nine months ago. Operators now report a rattling noise from the pump casing, erratic discharge pressure and a slowly rising bearing temperature. The strainer differential gauge reads 2.6 m and the strainer has not been cleaned since commissioning. Flow is held at 120 m³/h and the discharge valve is throttled to 60% for drum level control.
1. Calculate the NPSH available at the pump for the four combinations of normal/low tank level and clean/dirty strainer. State whether cavitation is expected and where the margin is thinnest.
Read the worked answer
Key principle. The deaerator water surface is at saturation, so the surface pressure equals the vapour pressure and the term (p_s − p_v)/ρg is zero. NPSHa is then simply the static head above the pump centreline minus the suction-line losses. This is the single most important point: you cannot buy margin by pressurising the tank, because the water temperature rises with it.
Hydraulics.
- Q = 120/3600 = 0.0333 m³/s; A = π(0.2)²/4 = 0.0314 m²; v = 1.06 m/s
- Pipe loss h_f = f L v²/(2gD) = 0.018 × 45 × 1.06²/(2 × 9.81 × 0.2) = 0.23 m
- Clean strainer 0.40 m → total 0.63 m; dirty strainer 2.60 m → total 2.83 m
Results.
| Level | Strainer | NPSHa |
|---|---|---|
| 6.0 m | clean | 5.37 m |
| 4.0 m | clean | 3.37 m |
| 6.0 m | dirty | 3.17 m |
| 4.0 m | dirty | 1.17 m |
Judgement. A workable margin is NPSHr plus 0.5–1.0 m, i.e. about 5.0–5.5 m available. Only the full-tank/clean-strainer case qualifies. Three of the four operating states are in cavitation, and the low-level/dirty-strainer case is 3.3 m short - deep cavitation. Note also that NPSHr rises with flow, so any future uprating of the pump makes this worse. The real design fault is that the plant has no operating margin: even with a clean strainer, the low-level alarm condition is below NPSHr.
2. Explain how the reported symptoms are consistent with cavitation, and why throttling the discharge valve does not cure the problem. How would you distinguish cavitation from air ingress and from flashing?
Read the worked answer
Symptoms. Vapour bubbles form where the local pressure at the impeller eye drops below the vapour pressure and collapse a short distance downstream, on the blade surfaces. The collapse is violent and produces the characteristic gravel-like rattle at 100–1000 Hz, plus random flow and head oscillation that the operator sees as an erratic discharge pressure. The asymmetric collapse pattern also produces radial hydraulic forces that load the bearings, which explains the rising bearing temperature. If left running, the impeller eye and shroud will show spongy erosion with a bright, smooth appearance, quite different from the sharp-edged damage of solid-particle erosion.
Why throttling does not help. The discharge valve is being used for drum level control, not for the noise. Closing it reduces flow and head, which does lower NPSHr somewhat, but it also raises the risk of discharge and suction recirculation, which is a separate cavitation mechanism. The defect is on the suction side: a 2.6 m strainer loss plus only 4.0 m of static head at the low-level condition. No downstream action can restore that.
Diagnosis.
- Air ingress gives a slow, cyclic surging of flow and ammeter, hissing at the gland or flange, and no dependence on water temperature. Partly closing the suction isolating valve makes air ingress better (less air drawn in) but cavitation worse - a quick and decisive field test.
- Flashing occurs when saturated water crosses a restriction and the pressure drops below p_v; the noise appears downstream of the restriction, not at the impeller eye, and pump head is largely unaffected.
- Check the deaerator pressure against the storage-tank temperature. At 1.2 bar absolute the water should be at 105 °C; any subcooling or superheat reading tells you whether the tank is running saturated and how much margin the temperature alone gives.
3. Rank remedial measures by effectiveness, quantify each, and recommend a package. Include the effect of subcooling and of changing the suction line size.
Read the worked answer
1. Clean the strainer (immediate, zero cost). Restores the loss from 2.83 m to 0.63 m but gives only 3.37 m at low level - still below NPSHr. Necessary, not sufficient. Fit a differential pressure switch with an alarm at 1.0 m so the condition cannot recur unobserved.
2. Raise the tank or lower the pump. Each metre of extra static head is a metre of NPSHa. To reach 5.0 m at low level with a clean strainer you need about 1.6 m more height; that is a structural change to the deaerator support and should be costed against option 4.
3. Increase suction line to DN250. Velocity falls from 1.06 to 0.68 m/s, so friction and strainer losses fall roughly with v² (factor 0.41): pipe 0.09 m, dirty strainer 1.07 m, total 1.16 m. At low level NPSHa becomes 2.84 m - improved but still short of 4.5 m. It is the right long-term correction but cannot stand alone.
4. Subcool the storage water by about 10 K. Cooling from 105 °C to 95 °C takes p_v from 121 kPa to about 84.5 kPa, so (p_s − p_v)/ρg = (121 − 84.5)/9.81 ≈ 3.7 m of recovered margin. This is by far the most powerful lever, because it works on the vapour pressure rather than on losses. A small cooler in the deaerator downcomer, or a feedwater recirculation cooler, achieves it.
Recommendation. Clean the strainer and add the differential alarm now; fit the 10 K subcooler and raise the low-level alarm setpoint to 4.8 m as the permanent fix; schedule the DN250 suction upgrade at the next shutdown. As an alternative, if the subcooler is unacceptable to the boiler chemistry regime, specify a replacement pump with NPSHr ≤ 2.5 m at 120 m³/h, or a vertical can-type pump, and verify against the full curve rather than the single duty point.