Views: 0 Author: Site Editor Publish Time: 2026-10-26 Origin: Site
A pump came apart at two years, to the day. The calendar called it. Faces still had a shine on them. Two bays down, an identical unit ran four years and wept its way through a planned stop. Same year, same part number, flipped result. The calendar explained zero of it.
Seals do not wait for a birthday. The 1978 United Airlines study that kicked off reliability centred maintenance, the one the whole planning model leans on, blamed age for maybe one in ten failures. The other nine were random or died young. That is why pulling a pump apart on a fixed clock can cost more than it returns. Break a seal, fit it back, the refit is a fresh chance to drop a clip or cross a gland bolt thread. The trade has a name for it, maintenance induced failure, and it owns a real slice of the stops blamed on seals.
Rubber runs its own clock though. O ring stock hardens, takes a set, leaks squeeze as heat and chemistry chew it year on year. That half of the seal truly ages on time. The one real argument for a calendar swap, even with faces that had years left. A shop that changes the whole cartridge on schedule is buying fresh faces just to reach the aged elastomer. Sometimes the cheaper move. Sometimes money down the drain. The pump duty decides, nothing else.
Signal | Normal | Act on it |
|---|---|---|
Weep rate | A few millilitres a day, steady | Doubles week to week, plan the changeout 30 to 90 days out |
Seal chamber temperature | Holds the baseline | About 20 degrees F, roughly 11 degrees C, above baseline, film starving |
Vibration | Steady pattern | Shifts and stays, something upstream moved |
API 682 lets one face pair weep up to 5.6 grams an hour. Treat that as the spine, not the bell. Under half, log it quarterly, watch the slope. Half to eighty, monthly, and go find why, pump out of line, flush pressure low, a face on the way out. Past eighty, order and swap at the next turnaround. Over, change at the first safe window. The slope beats the number every time. A weep up more than twenty percent between reads is coming apart, wherever it sits under the limit.
One PTA unit held a hard two year teardown rule and still dropped a reactor to an unplanned stop. The seal had crawled from three drops a minute to twenty, temperature flat, crew called it normal weep. Acid crystals choking the quill, not the faces. A cleanout would have bought months. Flip side, a pharma hydrogenator came apart at five thousand hours, catalyst powder scoring faces, spring shortened. Opposite fault, same question. Fluid calls it. Calendar does not.
The awkward truth, most plants run both, and run them sloppy. Cheap pumps get calendar swaps because watching them costs more than swapping them. Critical pumps get a weep check when someone happens to remember. The table above presumes a site that actually keeps the curve. Plenty never do. Plenty write the date and call it condition based.
Calendar wins where the law sets the date. Relief valves, fugitive emission kit, mandated intervals, no debate. It wins on the fractional horsepower pump where a sensor costs ten times the seal. Auditors want a date on paper, not a trend line. Step outside those three and the calendar serves the planner, not the pump.
Condition based pays on the pump with a sixteen week parts lead time, the one whose failure drains a tank or trips a unit. Early signal there is gold, months to stage spares, pick an outage. It suits dangerous service too, where a leak is a warning a site can work with, not a two am jolt. On the critical pump the early warning is the whole point. The calendar only hands a date and a hope.
The mistake that burns people is pulling good seals. A cartridge still holding gets pulled, moved, refitted, every step rolls the dice on a clip or a bolt. The install piece counted how many of those go wrong. Run it to its signal, not its date. Risk stays put that way. Quit asking does it leak. Ask what the rate is doing.
FBU builds cartridge seals to the pump and its duty, air tested before they leave, records riding with the unit. The cheap win is the seal changed on its own signal with the right parts already on the shelf. Spec the duty. Watch the weep. The curve decides the rest.
Is a fixed time interval ever the right call for a seal?
Three spots. Where the law sets the date. On the small cheap pump where a sensor costs more than the seal. And the elastomer, which ages on the clock no matter how fresh the faces look.
What leakage rate means the seal should be changed?
Hold it to the API 682 figure, 5.6 grams an hour per face pair. Under half, trend it. Past eighty percent, plan the swap. Over, change at the first window. And if it jumps more than twenty percent between checks, plan it regardless of where it sits.
Why not just change every seal on a schedule?
Most seal failures are random or early, not age related. Pull a good seal to refit it and the install becomes the next failure risk. Run it to its signal.