The Pump That Lost Flow Every Afternoon With Nothing Changed

Why this matters

The first diagnosis on this job was correct. The strainer really was loading, the seal really was weeping, both were really replaced, and flow really did come back. It came back for four days. A correct repair that restores suction margin does not fix a system that consumes that margin again every afternoon, and the reason the second visit went differently is that it stopped recording what the pump was doing and started recording what the liquid was doing. A fault that keeps a clock is driven by a variable that keeps the same clock, and on this system the variable everybody suspected was worth 0.9 ft while the one nobody was logging was worth 7.3 ft.

The safety calls this job forced first

The pump work on visit one. Pulling the strainer basket and the seal meant isolating the pump, locking and tagging the disconnecting means under 29 CFR 1910.147 rather than switching it off, relieving trapped pressure at the vent, and draining to a routed hose. The suction section sits under vacuum at afternoon conditions, so the vacuum was broken deliberately at the high-point vent before any joint moved, rather than by cracking the strainer cover and letting the line pull air and debris inward.

Reading gauges on visit two. Every reading below was taken with the pump running in normal service, coupling guard on, sleeves buttoned, one hand on a gauge isolation valve and nothing else near the machine.

Getting inside the tank later. Fitting the vortex breaker meant entering the tank, which is a permit-required confined space under 29 CFR 1910.146 in general industry, or 29 CFR 1926 Subpart AA where the work is construction. Every line into the tank was positively isolated rather than valved shut, the tank was drained, the atmosphere was tested for oxygen, flammables and toxics before entry and monitored continuously, and an attendant stayed outside with retrieval. Reaching the roof manway on an outdoor tank without a guardrail is a fall-protection duty under 29 CFR 1910.28 in general industry, or 29 CFR 1926 Subpart M in construction, and the anchorage was arranged before anyone climbed.

Visit one, and why being right did not hold

Complaint: a transfer pump on an outdoor break tank loses flow every afternoon and recovers overnight. The tech found the basket strainer about a third loaded and the mechanical seal weeping steadily. Both were replaced. Flow measured 262 gpm on completion against a design 260, and the customer signed it off.

Four days later the same complaint came back, on the same clock.

Nothing about visit one was wrong. The basket was loaded and would have been a problem eventually; the seal was weeping and was going to fail. But both findings were downstream of the real fault, which is why fixing them bought four days. Free gas at the impeller eye takes the liquid film out from between a mechanical seal's faces, and a seal run dry weeps and then fails. The seal was a symptom the same way the flow was.

The log that replaced the symptom

Visit two recorded nothing about the pump for the first day. It recorded four things about the liquid, every one to three hours, tightening the interval as the shift went on, from start of shift to end of shift: tank level above the outlet nozzle, liquid temperature, suction gauge reading, and measured flow.

Read the flow column literally: it held through 14:00, then fell. It did not decline steadily all day. Level and temperature moved steadily and monotonically all day; the flow did not. A steadily moving cause with a sudden effect means a threshold was crossed, not that a term ran out. That distinction is the whole diagnosis and it is visible in the table before any arithmetic is done.

The gauge column carries the other half. From 15:30 the reading is not a value, it is a value plus a behaviour, and the behaviour is the evidence. A swinging suction gauge on a pump whose conditions are changing slowly is free gas arriving in slugs.

The term everybody suspected, and the term that did it

Two things moved between 08:00 and 16:30, and it is worth putting both in the same unit before assigning blame.

Temperature. Vapor pressure of water at 74 F is about 0.42 psia; at 95 F it is about 0.82 psia. That is 0.40 psi, and at a specific gravity of 0.994 the conversion is 2.324 ft per psi, so the temperature swing cost 0.93 ft of available suction head.

Level. The surface fell 7.3 ft, from 9.2 ft over the nozzle to 1.9 ft. That is 7.3 ft of static head, straight off the top of the budget.

The level term is about eight times the temperature term. Everyone on this job had been talking about the afternoon heat, because it is what you notice standing next to an outdoor tank at 16:00, and it is the smaller of the two by a wide margin.

Ruling out cavitation with arithmetic

The obvious hypothesis at this point is that the suction budget simply ran out. It did not, and the check takes four minutes. Station barometric at this site is 14.5 psia. Suction line is 4 in schedule 40. The gauge tapping is 0.6 ft above the shaft centerline. Take the 16:30 row, using the mean of the swinging gauge:

A ratio of nearly four to one is not a suction budget in trouble. The liquid arriving at the pump is nowhere near boiling. Whatever gas is in the eye did not come out of the liquid, so it came from outside it, and the sibling card on that distinction has the field key.

The geometry underneath it

The tank outlet is a 4 in nozzle in the floor. As the surface drops toward it, the liquid approaching the nozzle keeps whatever rotation it has and, below a certain depth, the rotation organizes into a dimple and then into a full core that reaches the nozzle and swallows air.

   tank wall tank wall
   |                                                |
   |   liquid surface, 08:00                        |
   |   ===========================================  |
   |                                                |
   |   liquid surface, 16:30                        |
   |   =================\      /=================   |
   |                     \    /  surface dimple     |
   |                      \  /   over the outlet    |
   |                       ||    air drawn down     |
   +-----------------------++-----------------------+
                     outlet nozzle
      submergence is measured surface to nozzle

There is a published screening relationship for the submergence needed to prevent this, of the form submergence over diameter equals one plus 2.3 times the Froude number at the opening, and it appears in Hydraulic Institute intake design guidance, which binds through the specifying engineer's design or through contract rather than on its own. It was derived for a vertical suction bell in a rectangular sump with a defined approach, and a floor nozzle in a cylindrical tank with a swirling approach is a different geometry, so what follows is an approximation used as a screen, not a calculation.

At 260 gpm through a 4.026 in bore, velocity is 6.55 ft/s and the Froude number is 1.99, giving a required submergence of about 5.6 diameters, or 1.87 ft. The observed onset was at or above 2.6 ft. The screen under-predicted the requirement, which is the direction to expect when the real geometry is worse than the one the coefficient came from, and it is why the number was used to confirm the mechanism rather than to set the cutout.

One more thing the arithmetic explains: as flow falls the Froude number falls with it, so the required submergence falls too. At 178 gpm the screen gives about 1.39 ft. That is why the pump degraded into a stable, noisy, low-flow state instead of stopping dead.

The repair, and how they chose it

Three options were on the table.

Raise the low-level cutout from its existing 1.5 ft to 3.5 ft over the nozzle. A setpoint change, minutes of work, available the same afternoon. It costs working volume: on this 10 ft diameter tank, one foot of depth is about 587 gallons, so raising the cutout by 2.0 ft removes about 1,174 gallons, which at the afternoon draw of 260 gpm is roughly 4.5 minutes of buffer. Whether that matters depends entirely on the fill system's recovery cycle.

Fit a vortex breaker over the nozzle. The real fix, and it costs a permit-required tank entry with an attendant, which is a multi-person job and the largest labour item on the list by a wide margin.

Stagger the afternoon draw across the three process lines. Costs no parts and no entry, and requires a scheduling change the plant would not accept.

What they did: raised the cutout the same afternoon as containment, then scheduled the vortex breaker onto the next planned tank cleanout, when the tank would already be drained and the entry permit already written. That turned the entry from its own job into an add-on to work already scheduled, which is the difference between two attendance days and about an extra hour.

How they proved it held

They kept the same four-column log running for a full week, with no other change. The acceptance criteria were written before the repair, in the behaviour language the fault was actually speaking: the suction gauge holds steady through the end of shift with no swing, and flow stays within a few gpm of 260 through the afternoon peak. Both held on all five working days, including the two hottest, which also confirmed the temperature term was never the driver.

Those four columns are now permanent fields on the pump's PM form. The strainer and the seal are still on that form too, because both findings from visit one were real. They are just not the ones that keep the clock.

References

• 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before opening the pump or the strainer
• 29 CFR 1910.146, permit-required confined spaces, in general industry, and 29 CFR 1926 Subpart AA for construction, covering the tank entry to fit the vortex breaker
• 29 CFR 1910.28, walking-working surfaces and fall protection, in general industry, and 29 CFR 1926 Subpart M in construction, for reaching an unguarded tank roof manway
• Hydraulic Institute intake design guidance for the submergence screening relationship, as adopted by the specifying engineer or by contract, and the tank or pump manufacturer for the actual requirement at this geometry
• See related: What Air Entrainment Does That Cavitation Does Not; How to Work Out NPSH Available at the Pump You Are Standing At; How Suction Conditions Fail in Practice