Hot-Weather Cavitation on a Clarified-Water Pump
A 250 mm horizontal centrifugal pump at a water-treatment works lifts clarified water from an open sump to the filter gallery. Nameplate duty is 45 L/s at 32 m head and 1,470 rpm. The pump centreline sits 2.5 m above the sump's minimum water level, and the suction line, strainer and fittings lose 1.2 m of head at 45 L/s. Site barometric pressure is 101.3 kPa. The manufacturer's published curve gives NPSHr of 3.4 m at 45 L/s, rising to 5.6 m at 58 L/s. Since a filter backwash duty was added last summer, operators run the pump at about 58 L/s for six hours a day. In hot weather the water reaches 42 °C (vapour pressure 8.2 kPa, density 991 kg/m³, against 5.6 kPa and 994 kg/m³ at the 35 °C design condition). Maintenance reports gravel-like rumbling from the casing, discharge pressure fluctuating by ±0.4 bar, and a slight loss of flow, but motor current is steady and the bearings run cool. The pump was overhauled four months ago and impeller clearances are within tolerance.
1. Using the site data, determine whether cavitation explains the reported symptoms. Show the net positive suction head available (NPSHa) at both the design duty and the current operating condition, and state what else you would check to distinguish cavitation from other causes.
Read the worked answer
Method. NPSHa = (p_atm − p_v)/(ρg) + z − h_f, with z negative because the pump is above the sump surface, and h_f scaling with the square of flow.
At design duty (45 L/s, 35 °C):
- Pressure head: (101,300 − 5,600)/(994 × 9.81) = 95,700/9,751 = 9.81 m
- NPSHa = 9.81 − 2.5 − 1.2 = 6.11 m
- NPSHr = 3.4 m, so margin = 2.71 m. Adequate.
At current duty (58 L/s, 42 °C):
- Friction loss: 1.2 × (58/45)² = 1.2 × 1.66 = 1.99 m
- Pressure head: (101,300 − 8,200)/(991 × 9.81) = 93,100/9,722 = 9.58 m
- NPSHa = 9.58 − 2.5 − 1.99 = 5.09 m
- NPSHr at 58 L/s = 5.6 m, so the margin is minus 0.5 m.
Interpretation. Two changes combined to remove a comfortable margin: the pump is running at the right-hand end of its curve where NPSHr climbs steeply, and the added Q² friction plus the higher vapour pressure cost roughly 1.0 m of available head. The negative margin matches the rumbling, the pressure fluctuation and the loss of flow. The steady current and cool bearings argue against bearing damage, misalignment or a mechanical rub, and the recent overhaul with correct clearances argues against worn wear rings.
Discriminating checks. Air ingress through the gland or a vortex would also give noise and pulsation, but would usually show entrained air in a sight glass and erratic rather than steady current. Suction recirculation occurs at low flow, not high. Confirm by fitting a calibrated pressure transmitter at the suction nozzle and recording at 58 L/s: NPSHa = (p_gauge + p_atm − p_v)/(ρg) + v²/2g. If that agrees with about 5.1 m, the diagnosis is settled, and inlet-side impeller erosion found at the next strip-down would corroborate it.
2. Recommend corrective measures in priority order, distinguishing between immediate operating limits and permanent fixes. Give one field measurement that would confirm the diagnosis within a week.
Read the worked answer
Immediate (days):
- Limit flow to the qualified range. Reduce maximum flow to 45 L/s, achieved by throttling on the discharge side, and revise the operating procedure and the backwash sequence accordingly. Saving 0.8 m of suction loss and dropping back to NPSHr 3.4 m restores a margin of about 1.5 m.
- Keep the sump high and the water cool. Any operation near minimum level or at peak summer temperature should be prohibited; a level interlock that starts the standby pump or alarms at low level is cheap insurance.
- Clean or replace the suction strainer. The design h_f of 1.2 m assumes a clean strainer; a half-blocked basket could easily add 0.5–1.0 m and is the most common hidden cause of a marginal installation going into cavitation.
Permanent (next outage or planned project):
- Increase the suction line to DN 300. For the same flow, loss scales roughly with D⁻⁵, so 1.99 × (250/300)⁵ ≈ 0.8 m, recovering about 1.2 m and giving a margin above 1 m at 58 L/s.
- Alternatively or additionally, install a variable-speed drive with a maximum-frequency limit, or fit a low-NPSH impeller, and ask the OEM to confirm the NPSHr curve at the actual duty.
- Remove the elbow immediately at the suction nozzle if one is present, and check that the sump geometry avoids vortexing at minimum level.
Confirmation measurement. Install a calibrated pressure tap and transmitter at the suction nozzle, log suction pressure and flow together for a full backwash cycle at peak temperature, and compute NPSHa directly. A sustained NPSHa below NPSHr exactly when the noise occurs confirms the mechanism; a normal NPSHa with the noise present redirects the investigation to air ingress or recirculation.
3. The operations superintendent proposes throttling the suction isolation valve to bring flow down to 45 L/s, arguing that 'the noise stops when we do it'. Evaluate this proposal.
Read the worked answer
Reject it, and explain why in writing.
Throttling the suction valve does reduce flow, so the symptom may appear to improve, but it does so by adding a loss exactly where the pump cannot afford one. NPSHa = (p_atm − p_v)/(ρg) + z − h_f, and the throttled valve increases h_f. To bring 58 L/s down to 45 L/s through a part-closed suction valve, the extra loss will be at least of the order of the 1.2 m design friction - and it is being added upstream of a pump that is already 0.5 m short. The pump may still cavitate, and it will do so at a lower flow where the noise may simply be masked.
Two further problems:
- A part-closed valve creates a low-pressure zone immediately upstream of the gland and flange joints, which encourages air to be drawn in. That produces a similar rumbling and makes the true cavitation margin impossible to diagnose.
- Suction throttling is inherently unstable as a flow-control method; small changes in sump level or strainer condition move the operating point unpredictably, which is unacceptable for a backwash sequence that must deliver a repeatable volume.
Correct alternative. Throttle the discharge valve or, better, cap flow with a VFD or a trimmed impeller. Discharge throttling moves the operating point up the head curve to lower flow, which simultaneously lowers h_f and lowers NPSHr - the margin improves on both sides of the equation. The cost is wasted energy: a discharge-throttled pump runs at a higher head than necessary. That is a reason to plan the VFD, not a reason to accept suction throttling.
Also address the precedent. The fact that the noise 'stops' should be treated as a false negative, not evidence. The superintendent needs the mechanism explained with the numbers above, and the operating procedure should state explicitly that the suction isolation valve is to be either fully open or fully closed, never used for control.