What a Seal Flush Plan Is Actually For
Why this matters
A mechanical seal is two flat rings pressed together with a film of pumped liquid between them a few millionths of an inch thick, and that film is the whole product. Everything a seal does well or badly traces back to whether the film stayed liquid, stayed clean, and stayed cool. The flush plan is the piping that decides those three things, and it is the half of the specification shops skip. A seal that came back in eleven weeks and went back in with a harder face pairing and came back again in ten is almost never a face-material problem. It is a chamber that boils, or a chamber full of the solids the pump is moving, and no face material fixes either one.
Before you open a seal chamber
The seal chamber holds pumped liquid at pressure, and on a hot service it holds liquid above its atmospheric boiling point, which flashes to steam the moment you crack a fitting. Isolate the pump on both sides, lock and tag the valves and the driver disconnect under 29 CFR 1910.147, then prove the electrical side dead with a live-dead-live check against a known live source (29 CFR 1910.333(b)(2); NFPA 70E-2021, 120.5). Relieve chamber pressure through a vent to a drain routed away from where anyone stands, confirm zero on a gauge you trust rather than on the absence of a hiss, and let a hot casing cool below 140 F before bare hands touch it - read it with a non-contact infrared thermometer rather than by feel. Get the safety data sheet for the pumped fluid before you break the joint, because the flush line drains whatever the pump moves (29 CFR 1910.1200), and match the glove, eye and respiratory protection to what that sheet names. A fluid that gives off vapor at room temperature needs ventilation or a respirator selected under a 29 CFR 1910.134 program, not a face shield.
What the faces are doing, and the two ways it ends
The stationary and rotating faces run against each other with a thin liquid film carrying part of the load and the rest carried by asperity contact. That contact makes heat, continuously, at the one place in the pump with almost no thermal mass. The flush stream is the only thing carrying that heat away.
There are exactly two failure directions and they look nothing alike:
- The film vaporizes. The liquid between the faces flashes, the faces run dry for a fraction of a revolution, and they touch. You get puffing at the seal, a squeal that comes and goes, and on teardown a face with heat checks (fine radial cracks) or a hard face with a polished band and a discolored ring. This is the common one on hot water, on light hydrocarbons, and on any pump run near shutoff.
- The film carries solids. Particles drag through the interface and lap the faces flat and then past flat. You get a steady weep that grows, and on teardown a wide dull scored band with no heat marks at all.
The distinction is worth naming out loud on the teardown, because the first sends you to temperature and pressure and the second sends you to filtration or to a different flush source, and they are opposite fixes.
The three things a plan controls
Temperature at the faces. Flush flow removes seal heat. More flow, smaller rise. The chamber bulk temperature is what you can measure; the film is hotter than that, always.
Cleanliness. A flush taken from the pump's own discharge carries whatever the pump moves, at whatever concentration. Taking it from discharge does not clean it.
Pressure, and therefore phase. The margin that keeps the film liquid is the gap between the saturation temperature at the chamber pressure and the actual temperature at the faces. Both halves of that gap are set by the plan: the pressure by where the flush comes from and where it leaves to, the temperature by how much flow you push through.
The shape of a self flush, which is the plan you meet most
pump discharge
|
+-- flush line --> orifice --+
|
v
suction <-- along shaft -- seal chamber
|
seal faces
(heat made here)
Liquid is bled from discharge, throttled through an orifice, dropped into the chamber, and leaves along the shaft toward suction. Two consequences fall straight out of that shape. The driving pressure is discharge minus suction, so a pump running near shutoff has plenty of flush and a pump running far out on its curve has less, exactly backwards from where the heat is worst. And because the chamber outlet path runs to suction, the chamber sits near suction pressure, not near discharge pressure. That single fact is where most field margin calculations go wrong.
Who owns the plan number
Flush arrangements are numbered, and the numbering everyone uses comes from API 682, currently the 4th edition (2014), which is a purchaser's standard: it binds you only where your site's engineering specification or the purchase order adopts it, never on its own. The same is true of a pump built to ASME B73.1 for chemical service - the dimensional and sealing requirements reach you through the specification you bought against. So do not treat a plan number as a rule. Treat it as a shared vocabulary, and let the seal manufacturer's own selection guidance for your fluid, temperature and pressure own the actual answer, because they are the ones who publish a vapor margin for their faces.
The families worth knowing by behaviour
| Arrangement | What it gives the film | What it costs you |
|---|---|---|
| Self flush from discharge | Flow and cooling, free of extra piping | Delivers the process solids straight to the faces; flow falls as the pump moves right on its curve |
| Self flush through a separator or filter | Cleaner liquid, same source | A device that fouls and needs service; separators need real differential to work |
| Self flush through a cooler | Raises the vapor margin directly | A cooler that fouls silently and a margin that decays with it |
| External clean flush from another source | Full control of temperature and cleanliness | Dilutes the process; you must be allowed to add that liquid |
| Quench on the atmospheric side | Washes away a fluid that crystallizes or cokes as it leaks | Only works with a drain that stays open; a plugged drain floods the bearing housing |
| Dual seal with a barrier fluid | A clean, cool, chosen liquid at the inner faces and containment of the process | A support system with its own level, pressure and failure modes to maintain |
Worked example: filling in the plan for a hot water transfer pump
A hot water pump keeps eating seals. The service is water at 200 F. Measured gauge suction is 10 psig, measured gauge discharge is 85 psig. The seal chamber is the common arrangement above, vented along the shaft to suction. The seal manufacturer's guidance for these faces asks for 25 F of margin below saturation at the chamber, so that is the criterion the sheet is filled in against.
Line 1, chamber pressure. Not discharge. The chamber vents to suction, so chamber pressure is suction pressure: 10 psig, which is 10 + 14.7 = 24.7 psia.
Line 2, the flattering version, printed so it can be rejected. Using discharge instead, 85 + 14.7 = 99.7 psia, saturation near 328 F, apparent margin over 120 F, seal "obviously fine." This is the number a hurried check produces and it is wrong by the whole vent path.
Line 3, saturation temperature at the real chamber pressure. At 24.7 psia water boils at about 240 F.
Line 4, raw margin before any correction. 240 - 200 = 40 F.
Line 5, the correction the general section demands: the faces are hotter than the fluid arriving. Measure it rather than assume it. With a surface probe on the chamber inlet and on the outlet line, this pump reads an 8 F rise across the chamber, so the bulk liquid leaving the faces is at 200 + 8 = 208 F.
Line 6, corrected margin. 240 - 208 = 32 F. That clears the 25 F criterion by 7 F, so the arrangement is defensible as installed.
Line 7, the term that is a bound and not an interval. The film between the faces is hotter than the 208 F bulk by an amount nobody on site can measure. So line 6 is an upper bound on the true margin: the real margin is < 32 F, and it is written with that one inequality sign rather than as 32 plus or minus anything.
What the sheet then says about this pump. A 32 F upper bound against a 25 F criterion is a thin file. Two things on the same job push it thinner: the pump runs part of the week throttled back toward shutoff, which raises discharge and therefore raises flush flow, which helps, but also raises the temperature of everything recirculating inside the pump, which hurts more. And the suction gauge reading of 10 psig was taken with the supply tank near full. Draw that tank down and chamber pressure falls with it, saturation temperature falls with it, and the margin goes with it. The fix that follows from the sheet is not a harder face. It is either raising chamber pressure (a throat bushing that lets the chamber sit closer to discharge, which the pump manufacturer has to approve) or putting the flush through a cooler so line 5's arriving temperature drops.
What getting it wrong looks like. Nothing dramatic. The seal works, then weeps, then puffs on start-up when the chamber is at its lowest pressure, then goes in eleven weeks. The teardown shows heat checks, somebody reads that as "the faces cannot take it," a harder pairing goes in, and the second seal fails the same way because the chamber is still boiling. The heat checks were the message and the material was never the fault.
How to verify you have this right
Take four readings on a running pump and write them down together, because any one alone is not evidence:
- Chamber pressure, from a gauge on the chamber itself if there is a port, not calculated from discharge. If there is no port, state in your notes that the chamber pressure is inferred from the vent path and say which path.
- Temperature into and out of the flush line, so you have the rise rather than an assumption about it.
- Suction pressure at the worst tank level or worst upstream condition the pump actually sees, not the condition on the day you happened to visit.
- Whether the flush line is flowing at all. An orifice can plug, a valve someone closed during a repair can stay closed, and a plan on the drawing is not a plan in the pipe. A flush line that is at room temperature on a hot pump is not flowing.
If the seal is failing and all four are inside the manufacturer's numbers, then and only then is it worth arguing about face materials - and at that point look at whether the pump is being run somewhere its curve does not want to be, because a pump held near shutoff heats its own contents and the seal chamber is downstream of that.
References
- API 682, 4th edition (2014), mechanical seals and piping plans for pumps, which binds through the purchase specification or site engineering standard that adopts it, not on its own
- ASME B73.1 chemical process pump requirements, as adopted by a purchase specification
- 29 CFR 1910.147 (control of hazardous energy for the mechanical and pressure isolation) and 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5 (electrical isolation and live-dead-live proving)
- 29 CFR 1910.1200 hazard communication, for the pumped fluid's safety data sheet before a joint is broken
- Seal manufacturer selection guidance for the specific fluid, temperature and pressure, which owns the vapor margin number
- See related: Why Seals Leak: The Physics; How Seals and Gaskets Fail; What Cavitation Is and How It Announces Itself