Why a Condensate Pump Cavitates When Nothing Is Wrong With It

Why this matters

A condensate pump that knocks like it is passing gravel gets a new impeller, then a new pump, then a second new pump, and the noise comes back each time within a season. It comes back because the pump was never the fault. A condensate pump is the one pump in the building drawing from liquid that is already at its own boiling point, which means its entire suction margin can be consumed by a twenty-degree change somewhere else in the plant, by people who were doing something sensible and had no idea they were touching the pump at all.

Before the diagnosis, the part that burns people

Every measurement below is taken with the set running and the vessel hot. A vented receiver holds water within a few degrees of 212 F and it flashes the moment it is opened, so nothing here involves opening the tank, cracking a suction fitting, or bleeding a line to see whether vapour comes out. If a strainer genuinely has to come apart, stop the return, isolate and lock out the pump circuit under 29 CFR 1910.333(b)(2), prove it dead using the live-dead-live sequence of NFPA 70E-2021, 120.5 as adopted by your employer's electrical safety program, isolate the stored pressure in the return piping under 29 CFR 1910.147, drain to a floor point nobody is standing at, and confirm the contents are below 120 F with a non-contact reading from arm's length before a bolt moves. Contact readings on hot tank shells are taken with a probe and heat-rated gloves, not a bare hand, and not while leaning across the vent.

The two numbers that decide whether any pump cavitates

Cavitation is not a mechanical defect. It is the suction pressure at the impeller eye falling below the vapour pressure of the liquid it is handling, so the liquid boils in the eye and the bubbles collapse a few thousandths of a second later against metal. The whole question is whether the pressure available at the suction exceeds the pressure required by that pump at that flow.

Available net positive suction head, in feet of the liquid being pumped:

NPSH available = static submergence + (surface pressure - vapour pressure, converted to head) - suction friction - acceleration losses

Required net positive suction head comes off the pump curve at the duty point and belongs to the manufacturer. It is a property of the impeller geometry, and it does not change when the plant changes.

The middle term is where condensate is different from every other service. Convert psi to feet with head in feet = 144 x psi divided by the liquid density in lb per cubic foot, using the density at the actual temperature, around 60 lb per cubic foot near 200 F rather than the 62.4 you would use cold. Water's vapour pressure runs about 7.5 psia at 180 F, 9.3 at 190 F, 11.5 at 200 F, 14.1 at 210 F and 14.7 at 212 F, and that curve is steep at the top: the last ten degrees costs more than the twenty before it.

The call

A vented receiver and duplex pump set on a 100 psig distribution had run without complaint for eleven years. In September it began knocking, always in the afternoon, never first thing. The site had replaced the mechanical seal in July after a normal-wear leak.

The knocking was loudest after the pump had been running for a stretch and stopped entirely overnight. Discharge pressure was holding. Flow was adequate; the boiler was not short of feedwater. The pump was noisy and doing its job.

What was checked and eliminated

The seal replacement. Timing pointed at it and it was the wrong suspect. A seal that leaks pulls air in only when the suction is below atmospheric, and this suction sits under a vented tank with the pump below it, so the suction is above atmospheric at all times. The seal weeps outward here, not inward. Eliminated on the geometry, not on the calendar.

The suction strainer. Cleaned under full isolation and cooldown as above. Debris was light. The knocking returned within an hour of restart, which also killed the theory: a strainer that has just been cleaned cannot reload in an hour, and friction losses would have had to grow by feet, not inches, to matter.

The vent. Clear, with a steady plume, checked from the floor against a dark background rather than by hand. A blocked vent was the best remaining candidate because it would pressurise the tank, and it was not blocked.

Low level and vortexing. Watched through several cycles. The level never approached the pump-off mark during the noise, and the noise was continuous rather than the gulping surge a vortex produces.

The number nobody had measured

The receiver temperature. An immersion reading in the tank well came back at 210 F. The maintenance log from the previous winter, taken with the same probe in the same well, read 190 F.

Between those two readings the plant had insulated the return mains as an energy project. It worked. Less heat left the returns, so condensate arrived hotter, and the receiver ran twenty degrees warmer than the pump had ever seen.

One note on trusting those two numbers. A probe reading has a systematic component and a random one, and they behave differently. A contact probe strapped to a tank shell reads low by a fixed amount set by the insulation and the contact quality, and a fixed offset from one instrument largely cancels when you subtract two readings taken the same way at the same point, leaving the twenty-degree change trustworthy even if neither absolute is. The absolute value is what mattered here, because vapour pressure is non-linear across that band, so the immersion well reading is the one that entered the arithmetic and the shell readings were used only to confirm the direction.

Running the arithmetic on the set

Measured on site: static submergence from the pump-off level to the suction centreline, 2.5 ft. Suction friction at duty, 0.8 ft, from the pipe run and fittings. Vented, so surface pressure is atmospheric at 14.7 psia. Take the required figure off this set's curve as 4.0 ft at the duty point; that number is illustrative here and yours comes off your own curve.

At the old 190 F: vapour pressure 9.3 psia, so the pressure term is 14.7 - 9.3 = 5.4 psi, which at about 60.3 lb per cubic foot is 144 x 5.4 / 60.3 = 12.9 ft.

NPSH available = 2.5 + 12.9 - 0.8 = 14.6 ft against 4.0 required. Margin of more than ten feet. No pump on that curve would have complained.

At the new 210 F: vapour pressure 14.1 psia, so the pressure term is 14.7 - 14.1 = 0.6 psi, which at about 59.9 lb per cubic foot is 144 x 0.6 / 59.9 = 1.4 ft.

NPSH available = 2.5 + 1.4 - 0.8 = 3.1 ft against 4.0 required. The set is 0.9 ft short.

Nothing about the pump changed. Twenty degrees of return temperature removed 11.5 ft of suction margin, which is more than four times the entire static submergence the installer provided, and the set crossed from comfortable to deficient somewhere in the last few degrees. That is why it knocks in the afternoon and not at seven in the morning: overnight the returns cool, the tank sits well under 200 F, the margin is back, and the first hour of the day is quiet. If you cannot name what the symptom does at the other end of the range, you have not established the mechanism, and here the other end is audible every morning.

Why pressure cannot buy the margin back

The obvious move is to raise the pressure over the liquid. It does not work on a receiver that is being fed flash.

Raise the tank pressure and the surface pressure term goes up, but the liquid keeps receiving flash steam, so it heats until it boils at the new pressure, and its vapour pressure rises to meet the surface pressure. Both terms move together and their difference collapses back toward zero. A receiver taking a steady flash load pins itself: whatever the tank pressure is, the water sits at the saturation temperature for it. Capping or throttling a vent to "get some pressure on it" gains nothing at the pump, pushes back on every trap upstream, and turns an atmospheric vessel into a pressure vessel that was never built or protected as one.

Only three things move the number, and they are the three fixes:

• Elevation. More static submergence. Raise the receiver or drop the pump. This is real, permanent, and usually a structural job.
• Subcooling. Deliberately take heat out of the return before the receiver, most often by routing it through feedwater preheat or by mixing in cold makeup upstream. This trades recovered heat for suction margin, and on the case above it would have partly undone the project that caused the problem.
• A machine that does not need the margin. A pressure-powered pump uses motive steam and a float mechanism to displace condensate, has no impeller eye to cavitate, and is indifferent to the liquid being at saturation. Where elevation is not available, this is usually the honest answer rather than a fourth electric pump.

What actually got done

The set was replaced with a pump of lower required suction head at the same duty, which bought back about two feet of the deficit, and the pump-off float was raised to add nine inches of submergence. That closed the gap without touching the insulation project. The plant kept the recovered heat and the knocking stopped.

The failure mode worth naming is the one that nearly happened: the third pump had already been quoted. Two pumps had been replaced on a symptom that was never in a pump, and a fourth would have followed, because the only measurement that would have ended it was a temperature nobody thought belonged in a pump complaint.

Telling this apart from the two things it gets mistaken for

Air entrainment or vortexing makes a surging, gulping noise that tracks the water level and stops when the level rises. Cavitation from saturation is continuous through the whole run and tracks temperature instead. Watch the sight glass while you listen.

Mechanical noise - a dry bearing, a rubbing wear ring, a bent shaft - does not care about temperature or level at all and will make the same sound on a cold receiver first thing in the morning. If the noise is present at start-up on a cool tank, stop reading suction numbers and look at the machine.

References

• ASME Steam Tables, or an equivalent saturated-steam table, for vapour pressure and liquid density at the temperatures in your return system
• Pump manufacturer curve and installation literature, which owns the required net positive suction head and the minimum submergence for the set
• 29 CFR 1910.147, control of hazardous energy, and 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5 as adopted by your employer's electrical safety program, for isolating and proving the set before any component is opened
• See related: articles in this library on what a condensate receiver and pump set is doing, and on what condensate tells you about the system that made it