Cooling-Water Pump Cavitates Only in the Afternoon
A cement plant's cooling-water pump draws 45 m3/h from an open sump through 4 m of 100 mm ID suction pipe fitted with two 90-degree bends and a strainer. The pump centreline stands 3.5 m above the normal sump level. In summer the water reaches 35 degC, where the vapour pressure is 5.6 kPa; site atmospheric pressure is 101.3 kPa and density is 994 kg/m3. At 45 m3/h the suction-line friction loss, bends and clean strainer included, is 1.1 m, and the pump curve gives a required NPSH of 3.8 m. Operators report a gravel-like rumble and a 12 percent drop in discharge pressure that appear only in the afternoon, when the sump falls about 1.2 m below the normal line, and vanish at night. Friction loss is unchanged at that level; static lift becomes 4.7 m. Maintenance proposes three remedies: replace the suction line with 150 mm ID pipe, throttle the discharge valve to 35 m3/h (where required NPSH is about 2.9 m), or lower the pump by 1.5 m.
1. Compute the available NPSH at the normal and at the low sump level, and use the numbers to explain why the fault is time-of-day dependent.
Read the worked answer
Head available from atmospheric pressure above vapour pressure:
(101,300 - 5,600) / (994 x 9.81) = 95,700 / 9,751 = 9.81 m
Available NPSH = 9.81 - static lift - suction friction loss.
| Condition | Lift (m) | Friction (m) | NPSHa (m) | Margin over 3.8 m |
|---|---|---|---|---|
| Normal level | 3.5 | 1.1 | 5.21 | 1.41 m |
| Afternoon low level | 4.7 | 1.1 | 4.01 | 0.21 m |
At normal level the margin is a comfortable 1.4 m. At the low level the margin collapses to 0.21 m, well inside the 0.5 to 1.0 m cushion that a cooling-water installation should keep. That is why the rumble is time-of-day dependent: it is not a machine fault that drifts, it is a sump-level fault that recurs. Note also the seasonal amplifier - vapour pressure at 35 degC (5.6 kPa) has already cost about 0.6 m of NPSHa compared with cold water, so the same sump level that was safe in winter is not safe in summer. The 12 percent discharge-pressure drop is consistent with the pump operating on a cavitating, partially broken head curve.
2. Evaluate all three proposed remedies quantitatively against a target margin of at least 1.0 m at the low sump level, including their operating and process consequences.
Read the worked answer
(1) 150 mm ID suction line. Velocity falls by (100/150)^2 = 0.44, and friction scales with velocity squared: 1.1 x 0.198 = 0.22 m. NPSHa at low level = 9.81 - 4.7 - 0.22 = 4.89 m, margin 1.09 m. Meets target, no loss of flow, no energy penalty, but needs a shutdown, new fittings and a new strainer size.
(2) Throttle to 35 m3/h. Friction = 1.1 x (35/45)^2 = 0.67 m. NPSHa = 9.81 - 4.7 - 0.67 = 4.45 m, margin against 2.9 m required is 1.55 m. Hydraulically the best of the three, but it delivers only 78 percent of design cooling water and deliberately burns head across a valve - an energy penalty paid every hour to hide a suction-side problem. Acceptable only as a temporary measure.
(3) Lower the pump 1.5 m. Low-level lift becomes 3.2 m, NPSHa = 9.81 - 3.2 - 1.1 = 5.51 m, margin 1.71 m. Best hydraulic outcome, but check sump flooding risk, foundation, floor drainage and coupling realignment.
Recommendation. Option 3 if civil works and floor level permit; otherwise option 1. Do not accept option 2 as the permanent fix.
3. The maintenance superintendent wants to replace the impeller, arguing that the pump is worn. Using the evidence in the scenario, say whether you agree, and describe a single measurement that would settle the argument on site.
Read the worked answer
Disagree. The decisive evidence is the pattern, not the noise. A worn impeller or eroded wear ring degrades performance continuously and progressively; it cannot restore itself overnight. This fault appears in the afternoon with the sump low and disappears at night when the level recovers, which is a signature of suction-side NPSH, not of impeller wear. The arithmetic supports the same conclusion: NPSHa falls to 4.01 m against a required 3.8 m, a 0.21 m cushion, precisely where cavitation begins.
Replacing the impeller would consume a shutdown, cost money, and leave the root cause untouched - the pump would cavitate again the next hot afternoon, and the new impeller would be damaged faster than the old one.
The settling measurement: fit a calibrated compound (vacuum/pressure) gauge on the suction flange and read it at the same flow, first at night with a full sump and then in the afternoon at low level. Expected suction pressure is -(lift + friction), i.e. about -5.8 m water gauge (roughly -57 kPa) at low level. If that reading is reproduced, the pump is being starved and the suction system is at fault, not the hydraulics of the impeller. Repeat the reading after the chosen fix to confirm the margin has been restored.