18.1 Mechanical Liquid Pumps and NPSH

Liquid overfeed only works if a machine can take saturated ammonia from the low-pressure receiver (recirculator) and push it to evaporators against pipe friction, elevation, and control-valve drop. That machine is almost always a mechanical liquid pump—not a compressor. CIRO items in the liquid-overfeed area expect you to know how these pumps are built, why they cavitate, and which interlocks must stop them before they run dry.

A centrifugal refrigeration pump does not create NPSH. It consumes NPSH. The plant either supplies enough net positive suction head at the impeller eye or the liquid flashes there. Once vapor bubbles form and collapse, you get noise, falling flow, pitted metal, and underfed evaporators. The rest of this section is that sentence turned into operator practice.

Three pump families you will see on ammonia recirculators

Industrial plants use three mechanical arrangements. The hydraulic job is the same—move near-boiling liquid—but the leak path, the motor cooling path, and the round-check items are not.

Open-drive pumps look like a conventional centrifugal: a motor on a base, a coupling, and a pump casing. The shaft must leave the ammonia space, so a mechanical seal is the barrier. On ammonia that seal is a release waiting to happen if it runs dry, cavitates, or is starved of its flush. Open-drive machines are still in service, especially on older recirculators and some transfer pumps. Treat the seal as both a reliability item and a PSM leak source: oil film, frost pattern, and detector response around the pump matter.

Hermetic pumps put the motor in the same pressure boundary as the ammonia. There is no rotating seal to the room. The tradeoff is that motor losses become heat in the liquid. If flow collapses, that heat flashes the can or housing and the pump vapor-locks. Hermetic and canned designs therefore depend on flow even more than an open-drive pump does.

Canned-motor pumps are the sealless workhorses under vertical recirculators. A corrosion-resistant can separates the stator windings from the rotor. The rotor and impeller live in liquid ammonia. There is no dynamic seal to atmosphere, which is why plants like them under IIAR 2 machinery-room emission thinking. The same can that stops leaks also means you cannot service the motor like a NEMA frame on the floor: diagnostics are current, differential pressure, noise, and the equalizing line, not a flashlight on a seal face.

Do not mix these names up with compressor language. An “open” compressor and an “open-drive” pump both have a shaft seal. A “hermetic” compressor and a hermetic pump both keep the motor inside the refrigerant envelope. The CIRO distinction that matters operationally is seal to atmosphere versus no seal, plus how the motor is cooled.

NPSH available versus NPSH required

NPSH required (NPSHr) is a pump characteristic. The manufacturer publishes it versus flow, usually as the suction head at which developed head has already dropped about 3 percent. It rises at high flow (more inlet losses at the eye) and can also look ugly at very low flow if the impeller recirculates. You cannot “adjust” NPSHr from the control room except by changing speed or the operating point.

NPSH available (NPSHa) is a system characteristic:

NPSHa = pressure head at the liquid surface + static liquid height − vapor-pressure head − suction friction − velocity head.

On a water tank open to atmosphere, the first term is large. On an ammonia recirculator the liquid is at saturation. Surface pressure is the vapor pressure of that liquid. Those two terms cancel. For a boiling drum:

NPSHa ≈ liquid height from the free surface down to the impeller eye − friction and fitting losses in the drop, strainer, and suction valves.

That is why recirculators sit above the pumps, why suction piping is short and fat, and why a “the vessel is still showing 20 psig” comment does not save you. Gauge pressure on a saturated vessel is not subcooling. It is just the saturation pressure of whatever temperature is in the drum.

Work a plant-floor example (illustrative, not a nameplate). Liquid level is 6 ft above the impeller centerline. At design flow the strainer, isolation valve, elbow, and piping consume 1.5 ft of ammonia head. Velocity head is small. NPSHa ≈ 4.5 ft of liquid. If the curve shows NPSHr = 3.0 ft at that gallonage, you have 1.5 ft of margin. Drop the level to 3 ft with the same dirty strainer and NPSHa ≈ 1.5 ft. The pump is now in cavitation even though the high-level float never moved and the compressor is happily pulling vapor from the top of the vessel.

Subcooling would add NPSH (each degree of subcooling is extra pressure above vapor pressure). Recirculated liquid leaving the drum is not subcooled. Any heat into the drop or any pressure drop in the drop makes vapor, which is the opposite of NPSH.

What actually causes cavitation on ammonia pumps

Treat these as a round checklist, not a theory list.

• Low drum (recirculator) level. You spent the static head. This is the classic low-level cutout reason. A false-high glass (oil in the column, frost hiding the meniscus) can let the pump run after the real ammonia inventory is gone.
• High liquid temperature relative to local pressure. Heat gain on an uninsulated or poorly insulated suction drop, a pump sitting in a hot room, or liquid returning warmer than the drum saturation temperature flashes the suction. On a canned pump, motor heat at low flow does the same job from the inside.
• High pressure drop on the suction side. A plugged strainer, a suction isolation valve left throttled after service, undersized drop legs, or a partly closed equalization that should not be in the liquid path all steal feet of NPSH. Operators often say “high ΔP” for this: the suction pressure at the pump flange is lower than drum pressure by more than the liquid height can explain.
• High pump ΔP / high flow. If evaporators, a missing balancing orifice, or a wide-open bypass pulls the pump out to the right of the curve, NPSHr increases just as NPSHa is falling from extra friction. Cavitation and a loud pump at “high differential” can be a flow problem, not a full drum.
• Flashing in the suction. Two-phase flow into the eye is cavitation’s cousin. Causes include the items above plus a vapor equalizer that is shut, a can full of vapor after a shutdown, or starting a pump against a vapor-bound suction.

Cavitation sounds like gravel in the casing, reads as bouncing amps and unstable discharge pressure, and shows as starving coils (high superheat or poor TD on overfeed evaporators) while the recirculator still has a vapor space. Prolonged cavitation erodes the impeller and, on canned pumps, can overheat the can. After a cavitation event, inspect the strainer; it may now be full of metal as well as rust and gasket crumbs.

Equalizing lines

A canned or hermetic pump generates vapor from motor losses even when the drum is healthy. That vapor must leave the hydraulic space. The usual arrangement is a vapor equalizing line from the high point of the can or casing back to the vapor space of the recirculator—not into the liquid nozzle, and not into a wet return that can push liquid the wrong way.

If that equalizer is isolated, undersized, full of oil, or valved off “because it frosts,” vapor collects at the eye. The pump then loses prime, amps fall, and the no-flow or low-differential switch should trip. Dual-pump packages also need equalizing and isolation so the idle pump stays flooded and does not trap liquid without a relief path. After service, verifying equalizer open, suction open, discharge check free, vents back to the drum is as important as bumping the motor.

Pump protection you must not defeat

IIAR 2 and competent package design treat the pump as a liquid-full machine that can be isolated by its own discharge check and by downstream solenoid valves. That is why a hydrostatic or differential-pressure relief device (or a non-closeable vent) belongs on the pump and its associated piping: liquid ammonia trapped between the pump and closed automatic valves will expand. The relief should go to another part of the system (typically the vessel), not casually to atmosphere.

Electrical and instrument protection is the other half:

• Low-level cutout on the recirculator stops the pump before static head disappears. This is a safety interlock, not an operating-level control. Operating level may be a modulating feed valve; the low cutout is a dedicated float or transmitter trip.
• No-flow / low differential proves liquid is actually moving. Canned motors need that flow to carry heat away. A discharge pressure switch, differential-pressure switch across the pump, or a flow switch is typical. Confirm the setpoint against the pump curve; a switch set below deadhead will never see a real fail.
• Motor overload (and often overtemperature on canned units) stops a locked rotor, a seized bearing, or a pump grinding vapor. Overload is not a substitute for NPSH, but it is the last brake.

Start permissives should include vessel level proven, equalizer aligned, and typically a time delay so the pump is not jogging on a bouncing float. If you find a jumper on a low-level or no-flow circuit during a CIRO-style “what is wrong with this ladder” item, the correct operational answer is that the pump is unprotected, not that someone was being practical.