What a Mechanical Seal Is Actually Sealing Against

Why this matters

A mechanical seal does not stop leakage. It relocates it, from a shaft moving through a hole to a thin film between two very flat rings, and the film that does the sealing is the same film that does all the cooling those rings will ever get. That single fact decides almost every seal argument in the field, because it means the condition governing the seal is what the liquid is doing at the faces, not what the pump is doing at the discharge. Two identical seals, on the same pump, with the same liquid at the same temperature, can have 77 F of margin above boiling or none at all, depending only on how the seal chamber is piped. The worked case below is exactly that pair, and the failing one fails by 20 F.

Before you open a seal chamber

A pump that has been running can hold pressure after it stops. If the discharge check valve seats, the casing and the seal chamber stay at discharge pressure with the motor off. Lock and tag the disconnecting means under 29 CFR 1910.147 rather than switching it off, then relieve pressure at the casing vent and confirm it on a gauge before a gland bolt moves.

On hot service, let it cool. Liquid at 220 F under pressure is stable until you open it, and then it flashes. Isolate, wait, drain to a routed hose rather than to the floor, break the gland from the side rather than over it, and wear a face shield rather than glasses.

Do not put a torch or a heat gun on a gland to free a seized seal until you know what the secondary seal is made of. PTFE and PTFE-based components give off decomposition products when overheated, and that route is inhalation, so the control is respiratory under the employer's program at 29 CFR 1910.134 or, better, not applying the heat at all.

Do not dress or grind a seal face in the field. Hard faces are commonly silicon carbide or tungsten carbide, and dry grinding either one produces respirable dust; on a cobalt-bound tungsten carbide, cobalt is the health concern and the control is respiratory, not a glove. Lapping is done wet by a shop set up for it. For the same reason, do not blow a face clean with compressed air: it is both an airborne-particulate route and a good way to drive grit across the surface you are trying to protect.

A dual seal's barrier reservoir is a pressure vessel. Relieve it before opening it, and check the SDS for the barrier fluid before it goes anywhere.

What it replaced, and what it did with the leak

Packing, which is a sibling article's subject, seals on the shaft: at a surface that slides past it and wears. A mechanical seal moves the sliding interface off the shaft entirely and puts it between two flat rings facing each other along the shaft axis, one rotating with the shaft and one held still in the gland. A spring or bellows pushes them together, and system pressure pushes them together harder. The shaft no longer slides against anything that seals.

The leak did not go away. It got smaller and it changed phase. A film of the pumped liquid, a fraction of a micron thick, sits between the two faces; it lubricates them and it carries away the heat they generate. At the outside diameter, that film reaches atmosphere and flashes off. A correctly operating seal leaks continuously and you cannot see it, because the leak leaves as vapor. The actual rate depends on size, pressure, speed and the face pair, and the seal manufacturer publishes it; what you need to know in the field is that "zero leakage" is a marketing word, not a design condition.

The three paths out

Only one of them is the faces, and misreading which path is leaking sends a seal back to the shelf that had nothing wrong with it.

1. Across the faces. The dynamic path, the designed one. When this path fails it is because the film failed: it went dry, it flashed, or something got between the faces.
2. Past the secondary seal on the shaft or sleeve. An O-ring, a wedge or a bellows seals the rotating face to the shaft and must accommodate small axial movement. This path is chemistry and temperature: elastomer attacked by the fluid, hardened by heat, or swollen by a lubricant it was never rated for.
3. Past the static joints. The stationary face's seat in the gland plate, the gland plate to the casing, and on a sleeve arrangement the sleeve-to-shaft gasket. These do not move, so when they leak it is installation: a pinched gasket, a scratched register, a sleeve that was not seated.

A leak that runs down the shaft and slings off is usually the faces or path two. A leak that weeps at the gland face and does not sling is usually path three, and it is fixed with a gasket and a clean surface rather than a new seal.

What the faces actually see

The seal does not live at discharge pressure and it does not live at suction pressure. It lives at seal chamber pressure, which is whatever the chamber arrangement makes it, and on a standard pump that is set by the throat bushing, the chamber geometry, and how the chamber is piped. Some arrangements vent the chamber back toward suction. Some feed it from discharge through a restriction. The two are not close to each other, and the pump manufacturer's chamber arrangement owns which one you have.

Chamber pressure matters because it sets the temperature at which the liquid boils. Boiling in the chamber is survivable. Boiling between the faces is not: the film is what separates them, and when it flashes the faces run on vapor, which lubricates nothing and carries almost no heat. Seconds of that produce heat checking on the hard face.

Two things add up against that margin. The liquid arrives at whatever temperature the pump is handling, and the faces themselves generate heat, which rises with the load pushing them together and with sliding speed, so it rises with pressure, with shaft speed and with seal diameter. The seal manufacturer's selection data gives both the flush flow and the temperature rise that flow will hold.

Seal specifications set a required margin between the liquid's temperature at the faces and its saturation temperature at chamber pressure. API 682, in the edition the purchaser's specification calls out and binding through that specification rather than on its own, sets such a margin, and a figure on the order of 20 F for water service is commonly used. Your seal manufacturer's data governs where it differs.

The flush is chosen for one of three jobs

A flush plan is a small piece of piping around the seal, and every one of them exists to fix exactly one of three problems. Naming which one you have prevents fitting the wrong plan.

• Temperature. Get heat out of the chamber. A line from discharge into the chamber, or a line from the chamber back to suction, moves liquid through and carries face heat away. Where the pumped liquid is already too hot, the same line runs through a cooler first.
• Cleanliness. Keep solids out. A line from discharge through a separator, or clean liquid injected from an outside source at a pressure above chamber pressure so the flow is always outward, keeps abrasives away from the faces. This is the plan that saves seals on any service with grit.
• Margin above boiling. Raise chamber pressure so saturation temperature climbs away from the liquid temperature. This is the discharge-referenced arrangement in the worked case below, and it is why hot-water and light-hydrocarbon services get piped differently from cold water.

A dual seal with a barrier fluid solves a fourth problem that is not on this list: it keeps the pumped liquid off the faces entirely, which is what you fit when the liquid cannot be allowed to reach atmosphere at all.

Worked example: same seal, same liquid, two chamber arrangements

A boiler feed pump handling water at 220 F, taking suction from a deaerator. Suction gauge reads +6.0 psig; station barometric is 14.5 psia. The seal manufacturer's selection data for this size and speed calls for a flush flow that holds the chamber within 10 F of the pumped liquid. Required margin above saturation, per the specification, is 20 F.

Arrangement A: chamber vented back toward suction.

The seal is 20 F short. The gross margin looked like 10 F before the face heat was added, which is already inside the required 20 F, and the rise took the rest. The film flashes at the faces, and the hard face heat checks. Nothing about the seal is defective and no better seal fixes it.

Arrangement B: discharge-referenced flush through a restriction, chamber held at 60 psig.

Same seal, same pump, same liquid at the same temperature, and 77 F of margin instead of none, because 54.0 psi of extra chamber pressure moved the boiling point up by 77 F.

What would flip this. If the flush cannot hold the chamber to 10 F, everything moves with it: a 25 F rise on arrangement B still leaves 62 F and passes, while on arrangement A it makes the deficit worse rather than different. And if the liquid were 250 F rather than 220 F, arrangement B would sit at 307 minus 260, or 47 F, still passing, while arrangement A would not be a seal application at all without a cooler in the flush line.

The failure mode, and it is a common one: the seal on arrangement A is replaced, fails again in weeks, is replaced with a different face pair, and fails again. Each replacement is a real repair of a real symptom. The chamber piping is never the thing anyone looks at, because it is not a part with a number on it.

How to verify you got this right

Measure chamber pressure, do not assume it. Many seal chambers have a tapping, and a gauge on it settles in one reading whether you are on a suction-referenced or a discharge-referenced arrangement. If there is no tapping, trace the flush piping by hand and say out loud where it comes from and where it goes.

Then take a surface temperature at the gland with an infrared thermometer, from arm's length, with the guard on, and read the change from a known-good baseline at the same marked spot rather than the absolute value. Emissivity error is systematic for a given instrument on a given surface, so it is common to both readings and largely cancels in the difference; it does not cancel between two different spots.

Finally, when a seal comes out, look at the hard face before anything else. Fine radial cracking is heat, and heat means the film went. That points at this article's arithmetic rather than at the seal, and the sibling card on reading a failed seal takes it from there.

References

• 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before opening a seal chamber
• 29 CFR 1910.134, respiratory protection, where a face is machined, a fluoropolymer component is overheated, or an SDS names an inhalation route
• API 682, in the edition the purchaser's specification calls out and binding through that specification rather than on its own, for seal chamber vapor-pressure margin, and the seal manufacturer's own selection data where it differs
• Published steam tables or the fluid supplier's property data, for saturation temperature at the measured chamber pressure
• See related: How a Mechanical Seal Fails and What You See; What Packing Does That a Mechanical Seal Cannot; Why Seals Leak: The Physics