Pumps, Mechanical Seals and Packing

A pump has one hole it cannot avoid: the place where the drive shaft passes out of a casing full of liquid. Everything else about a pump can be made tight with a gasket and bolts. That one opening has to stay sealed while a shaft turns through it, and it is where most pumps actually fail.

Why it matters on the job

Pumps are the most numerous piece of rotating equipment in a process plant, and the seal is the component that decides how often you visit them. A pump whose seals last four years and a pump whose seals last four months are usually the same pump, treated differently.

Seal failures are also the classic misdiagnosis. The seal is the part that leaks, so the seal gets replaced, and it fails again in a month because the actual fault (misalignment, a bent shaft, cavitation, a dry-running condition) never left.

Two families of pump

Centrifugal pumps spin an impeller that throws liquid outward, adding velocity that the casing converts into pressure. Flow and pressure trade against each other along the pump’s curve: throttle the discharge and the flow drops while the pressure rises. Most plant water, chemical and general transfer duty is centrifugal, and it dominates because the design is simple, cheap and forgiving.

Positive displacement pumps trap a fixed volume and push it along: gear, screw, lobe, vane, piston, diaphragm and progressive cavity. Flow is roughly constant regardless of pressure, and the pressure rises to whatever the system demands.

That difference has a consequence you have to know before you touch either one. Close the discharge valve on a positive displacement pump and it does not stop pumping. It keeps delivering the same volume into a space that cannot accept it, until something yields (a relief valve, a gasket, a shaft or the casing). Close the discharge on a centrifugal pump and flow stops, but the impeller keeps churning the same trapped liquid and heating it, which will destroy the seal and can flash the contents. Neither is a safe place to leave a pump.

Two ways to seal a rotating shaft

The two answers are different in kind, not in quality, and the difference is which surface does the sealing.

Packing sits in the stuffing box: rings of braided material compressed around the outside of the shaft or shaft sleeve by a gland follower. The seal is made on the shaft’s outer surface, along its length. Packing is cheap, tolerant of a bit of shaft movement, and rebuildable with hand tools.

A mechanical seal makes its seal on a pair of flat faces standing perpendicular to the shaft. One face rotates with the shaft, the other is fixed in the housing, and they are pressed together by a spring and by system pressure. The sealing happens between those two faces, across a film of liquid thinner than a hair. The shaft’s surface is not the sealing surface at all.

Packing seals along the shaft; a mechanical seal seals across it, on two flat faces

That geometric difference produces the rule you actually use: packing is supposed to drip and a mechanical seal is not. Packing needs a small controlled leak to lubricate and cool the rings, so a stuffing box run bone dry burns the packing and scores the shaft. A mechanical seal’s faces are lubricated by a film so thin that the leakage is normally invisible, so any visible drip from a mechanical seal is a failure in progress rather than an adjustment opportunity.

Cranking down the gland on a mechanical seal to stop a drip achieves nothing except a hotter seal. There is no adjustment.

Worked example: reading shaft runout

Before a mechanical seal goes on, the shaft or sleeve it runs on gets checked for runout with a dial indicator, and it gets misread constantly.

Set the indicator against the shaft in the seal area and turn the shaft one full revolution. The needle sweeps from its lowest to its highest reading, and the difference is the total indicator reading. Suppose the TIR is 0.004 in.

That does not mean the shaft centerline is 0.004 in off. As the shaft turns, the high spot swings from one side to the other, so the indicator sees the eccentricity twice in a revolution: once going out and once coming back. The actual eccentricity is half the TIR, or 0.004 / 2 = 0.002 in.

Now think about what that means in service. At 1,750 rpm the shaft turns 1,750 / 60 = 29.2 times a second, so the seal faces are worked through that 0.002 in of wobble roughly 29 times every second, hour after hour. A number that sounds trivial on a bench is an enormous number of cycles by the end of a shift.

Compare the measurement against the seal manufacturer’s stated limit, which is on the seal drawing and varies by seal type and speed. Do not carry a remembered limit between jobs.

Rebuilding one properly

• Clean is not a preference here. Seal faces are lapped flat to within a few wavelengths of light. A fingerprint on a face is a contour, and grit is a gouge. Handle faces by the edges, unwrap them last, and assemble in a clean space.
• Check what the seal will run on before fitting: shaft or sleeve runout, surface finish, the condition of the bore, and the squareness of the gland face to the shaft.
• Set the working length from the drawing. Mechanical seals are compressed to a specified dimension to develop the right face load. Too little and the faces open, too much and they run hot. Guessing this is why seals fail early.
• Lubricate the elastomers as the manufacturer specifies, and confirm the material suits the process fluid and temperature. An elastomer that swells or hardens in service takes the seal with it.
• Align the pump and driver after the rebuild. A seal that is perfect on the bench will fail in a machine whose shafts are misaligned, and the alignment lesson in this trade covers how.
• Fill and vent before starting. Every mechanical seal start on a dry casing costs seal life, and some cost the whole seal.

Where it bites

• A repeat seal failure is a message about something else. Three seals in a year means the problem is misalignment, a bent shaft, pipe strain on the casing, cavitation, or a dry-running condition. Replacing the seal a fourth time is the expensive way to not answer the question.
• Suction problems look like seal problems. A pump starved on the suction side cavitates, and cavitation shakes the machine, hammers the bearings and destroys seals. The evidence is at the suction (a blocked strainer, a throttled valve, too little liquid above the pump) rather than at the leak.
• Pipe strain is invisible and common. Piping pulled into place with flange bolts distorts the casing, moves the bearing housings, and misaligns the shaft before the pump has run a minute. Check that flanges meet freely before you bolt them.
• Do not convert packing to a mechanical seal casually. The two want different things from the shaft, the box and the flush arrangement. It is a good upgrade done properly and a fast repeat failure done as a parts swap.
• Flush and quench plans are part of the seal. The piping that brings clean, cool liquid to the seal faces is what makes the seal survive its duty. A seal replaced without restoring its flush line is a seal on a countdown.