How to Prime a Pump Without Damaging It

Why this matters

Almost nobody fails to prime a pump because there is not enough water in the casing. They fail because the suction path is drawing air in from a joint that looks perfectly dry, or because the air already in the casing has nowhere to go. Both are invisible to inspection and both survive a pressure test, which is why the same pump gets filled and cranked five times across two visits and then primes on the sixth attempt with a seal that is already damaged. On the suction side a leak does not leak - it draws - and the only instrument that finds it is a vacuum test that takes about ten minutes.

This card is organised around the one part of the job you cannot take back, and then around the conditions that have to be true before the shaft turns.

The one irreversible minute

Every other mistake in this procedure is recoverable. Running the seal dry is not.

A mechanical seal has no lubrication and no cooling except the liquid it is pumping, and its faces heat-check in seconds to a couple of minutes when run dry. Get the manufacturer's permitted dry-run time for the specific seal before the first attempt; if no figure is available, treat it as under one minute and hold to that. Set a clock, out loud, before anyone touches the starter, and agree who calls stop. On a packed pump, the same rule applies to the packing and the shaft sleeve, and the flush water must be flowing before the shaft turns, not after.

Do not fill a hot casing with cold water. Thermal shock cracks casings and distorts seal faces. Let the pump cool below about 120 F, or fill with liquid at a similar temperature.

Open a casing vent only after confirming the casing is not under pressure, and on a hot system route the vent to a drain, stand to the side, and wear a face shield and heat-rated gloves; water above 212 F held under pressure flashes at the opening.

Between attempts, lock and tag the disconnect under 29 CFR 1910.147 before your hand goes into a strainer basket, a suction bell or a foot valve. A level control or an auto-restart makes "I only pressed stop" the wrong isolation.

Repeated start attempts overheat the motor. Respect the starts-per-hour figure on the motor nameplate; four attempts in five minutes on a large motor is a rewind, not a diagnosis.

Where the liquid is a fuel, a solvent or anything with a flash point, vapour is displaced out of the vent as the casing fills. Bond and ground the container and the pump before transferring, keep ignition sources out of the area, and provide ventilation or respiratory protection under a written program per 29 CFR 1910.134 for the vapour route - gloves address the skin route and do nothing about the air you are standing in. The safety data sheet for that liquid owns the exposure limit and the control.

If the source is a wet well, a pit or a tank, entry is a permit-required confined space under 29 CFR 1910.146 in general industry or 29 CFR 1926 Subpart AA on a construction site. Read levels from outside.

On the suction side, a leak does not leak

A suction lift line runs below atmospheric pressure along its whole length. A defect in that line therefore admits air rather than releasing liquid. There is nothing to see, nothing to feel and nothing to hear over a running motor, and the joint will pass a hydrostatic test at any pressure you like, because pressure pushes the gasket into its seat while vacuum pulls it away.

That single fact explains most of the field behaviour techs describe as a pump that "loses its prime." A pump does not lose its prime spontaneously. Air accumulated at the high point of the casing until the impeller eye was in a gas pocket, and the air came from somewhere: a gland, a gasket, a threaded joint sealed with a compound that bridged a gap under pressure, a cracked plastic union, a suction gauge tapping, or a vortex at the inlet drawing air off the surface because the submergence is too shallow.

The conditions that have to be true before the shaft turns

Treat this as a checklist you fill in rather than a sequence you walk, because they can be established in any order and skipping any one of them puts you back at the dry-run clock.

1. Casing and suction chamber full of liquid, verified at the vent. Liquid, not spray, out of the open vent. Re-verify after every failed attempt, because the attempt itself can push liquid out.
2. The suction path holds vacuum, established by test rather than by inspection. Method and acceptance number below.
3. Air has an exit on the discharge side. A closed discharge valve or a seated check valve traps the air the pump is trying to push out, and the impeller then churns the same gas pocket indefinitely. Crack the discharge or open a high-point vent.
4. The suction lift is inside the arithmetic. Net positive suction head available has to exceed what the pump requires, with margin. The sibling card on repeat pump failure runs that calculation line by line; do not guess it from a catalog lift figure.
5. Foot valve or suction check in the right orientation and free. A foot valve held open by a stone drains the line back every time the pump stops, and it presents exactly as "loses its prime overnight."
6. Rotation confirmed against the casing arrow, with the guard in place. Many self-priming designs will not prime at all backwards, and a reversed three-phase pump that has been re-landed after panel work is a common cause of a first-time-ever priming failure on a pump that has always worked.
7. Seal or packing has its liquid. Flush line open, seal chamber vented, gland leak-off present on a packed pump.
8. A dry-run clock agreed before the first start, with one person watching it.

The vacuum hold test, and a number to accept it against

Isolate the suction line at the source, apply vacuum with a hand vacuum pump at a gauge tapping, close the valve on the vacuum source, and watch the gauge.

There is no published universal acceptance rate for this, so set your own and write it in your procedure. A workable shop rule is that a suction line losing more than about 1 in. Hg in 5 minutes has a leak worth finding; tune it to your typical line lengths and joint counts. What matters more than the exact figure is that it is a rate, not a demand for zero decay. Zero is unachievable on a real line with a valve stem and a gauge tapping in it, and a crew chasing zero gives up and calls the test useless.

If it fails, find the leak by soaping joints while the line is still under vacuum. Bubbles will be drawn in, so watch for a soap film disappearing into a joint rather than for a bubble forming on it. That is the opposite of what everyone's hands have learned on pressure testing, and it is the reason the test gets misread.

Worked case: a self-priming pump on a 14 ft lift that would not prime

A transfer pump at a site roughly 1,000 ft above sea level, barometric pressure 14.2 psia, drawing water at 75 F. Suction is a 14 ft static lift. The pump requires 10 ft of NPSH at its duty flow. Two techs had filled and cranked it five times over two visits.

First, was the lift ever possible? Every term printed, with the correction on its own line.

• Surface pressure in feet of this fluid: 14.2 x 2.31 / specific gravity 0.994 = 33.0 ft
• Vapor pressure at 75 F, 0.43 psia, in feet of this fluid: 0.43 x 2.31 / 0.994 = -1.0 ft
• Static lift: -14.0 ft
• Suction friction at duty flow: -1.5 ft
• NPSH available = 16.5 ft, against 10 ft required, a margin ratio of 1.65

So the lift is comfortably inside the arithmetic and is not the problem. Worth noting what the same arithmetic says about the practical ceiling here: 33.0 - 1.0 - 1.5 - 10.0 = 20.5 ft is the most this pump can lift at this site with this water, which is well short of the sea-level cold-water figure of about 34 ft that gets quoted from catalogs and that ignores vapour pressure, friction and the pump's own requirement entirely.

Then the vacuum test. Pulled to 20 in. Hg. The gauge fell to 11 in. Hg in 4 minutes, a decay of 9 in. Hg, or 2.25 in. Hg per minute against the shop's acceptance of 0.2 in. Hg per minute. More than eleven times the acceptance rate, which is not a marginal result requiring judgment.

Finding the leak. Soap at each joint under vacuum. The film at the strainer lid union pulled inward and vanished. The gasket had taken a set and was sealing perfectly against pressure, which is why it had never dripped and why the visual inspection on both prior visits had passed it.

After the repair. New gasket, retested: 20 in. Hg falling to 19.6 in. Hg in 5 minutes, a decay of 0.08 in. Hg per minute, comfortably inside the acceptance. Casing filled and vented until liquid ran clear, discharge cracked open so displaced air could leave, dry-run clock set at 45 seconds. It primed in about 20 seconds on the first attempt.

What the failed approach cost. Five prior start attempts against an unrepaired air leak, each one running a mechanical seal on gas. The seal was replaced at the same visit as a matter of course, because five dry cranks is past any manufacturer's tolerance and a heat-checked face does not announce itself until it weeps a month later. That is the concrete failure mode of skipping the vacuum test: you do not fail to prime, you succeed on the sixth attempt with a seal that has already been damaged, and the callback arrives with no obvious link to the visit that caused it.

What changes the method

Flooded suction. With liquid standing above the pump there is nothing to prime, and a pump that will not pump on a flooded suction has trapped air rather than a priming problem. Vent the casing at the high point and check for a closed suction valve, which is the more likely fault by a wide margin.

Submersible and vertical turbine. Priming is not a step; the bowls sit in the liquid. What replaces it is submergence, since too little liquid above the inlet lets a vortex form and draws air down the same way a shallow suction bell does.

Positive displacement. Many are self-priming by design and some, particularly progressing cavity and vane types, are damaged very fast by a dry run because the running elements rely on the fluid for lubrication rather than just for cooling. The dry-run clock gets shorter here, not longer, and the sibling card on why these machines cannot be throttled covers the family behaviour.

How to verify you got this right

Prime is confirmed when the discharge pressure is steady, not when the pump is loud. Watch the suction gauge: a primed pump on a lift holds a steady vacuum, and a needle that swings a few inches of mercury rhythmically is drawing air or vortexing at the inlet, whatever the discharge is doing.

Then leave it and come back. Stop the pump, wait an hour, and read the suction gauge again. A line that holds its column has a sound foot valve; a line that has drained back will fail to restart, and finding that on your visit is far cheaper than finding it on the customer's Monday morning.

References

• Seal manufacturer documentation for the specific seal, which owns the permitted dry-run time, and pump manufacturer documentation for NPSH required, minimum submergence and self-priming lift
• 29 CFR 1910.147 for isolation before reaching into a strainer or suction inlet, 29 CFR 1910.134 for respiratory protection where a volatile liquid is being transferred, and 29 CFR 1910.146 in general industry with 29 CFR 1926 Subpart AA on construction sites for any pit or wet well entry
• Safety data sheet for the pumped liquid, which owns the exposure limit and the required ventilation control
• See related: The Airlock That Looked Like a Dead Pump; The Pump That Was Replaced Twice and Failed the Same Way Each Time; What Cavitation Is and How It Announces Itself