Shutdown maintenance on a continuous line: the window is short, so the sequence matters

The work list should be fixed a month before the shutdown. It draws from three sources: hazards found in daily checks, periodic tests now due (accuracy, vibration, oil samples, insulation), and items left over from the last shutdown. Anything added late needs approval, or the window will not stretch.

Write each item out in full: scope, trades needed, spares and tooling, estimated hours, acceptance criteria. An item with no acceptance criteria leaves you unable to say whether the work actually passed.

Sequence by risk, not by convenience. Do first the items that cannot be inspected without stripping and cost dearly to open (gears inside a gearbox, bearing housing internals); leave externally serviceable items for later.

Stop, isolate, tag out, prove dead, relieve pressure, cool down — this order cannot be shortened or reversed. When trades overlap, name one person in charge, and write down who isolates which machine. This procedure was bought with blood.

Photograph the teardown: how things fitted before they came apart, where the shims sat, how the wires were marked. Not being able to match things back up is the number-one cause of shutdown overruns.

Measure and record key dimensions on the spot: bearing clearances, gear backlash, coupling alignment values, bolt torques. These are the basis for reassembly quality and the comparison baseline for the next shutdown.

Reassemble in reverse of the teardown record, and fit new seals, gaskets and split pins — these single-use parts must not be reused. Do not scrimp on that.

Commission in three steps: turn it by hand to confirm nothing binds → jog it to confirm direction and limits → run it idle for 2–4 hours while checking temperature and vibration. Only then load it. Skipping the idle run and going straight to load is the main cause of early failures after maintenance.

File the maintenance record the same day: what was replaced, what data was measured, what is left over. Carry-over items go onto the next work list — do not rely on memory.

The maintenance window is never long enough - that is the norm on continuous production lines. The first trick for gaining window time is "tier and bundle": split every maintenance item into three tiers - those requiring a full stop (chain or belt replacement, teardown inspection, structural repair), those doable at reduced speed or no load (some testing, retorque checks), and those doable fully online (vibration and temperature readings, oil sampling, visual inspection). Only tier one goes into the window; tiers two and three move to running time, cutting window demand by thirty to forty percent outright.

The second trick is "front-loading the window": all preparation is completed before the window opens - spares received and inspected, tooling ready, teardown plan for old components, crew assignment sheet, and safety isolation measures. Window time is then spent only on hands-on work, with not one minute lost to hunting tools, waiting for parts or holding coordination meetings. Our field data: well-prepared maintenance achieves window utilization above eighty percent; poorly prepared work spends half its time waiting and searching.

The third trick is "parallel scheduling": assign non-interfering work fronts to different crews running simultaneously - mechanical crew changing chain, electrical crew checking drives, lubrication crew doing service - staggered in space and time. Draw the schedule as a timeline, marking start, finish and owner for every task, and coordinate the crossing points in advance. Use all three tricks well and compressing a three-day shutdown into a day and a half becomes routine - the whole secret lies in what happens before the window.

The value of a maintenance record lies not in having written it, but in being able to find it later and compare it. Most plants keep only a completion slip: on such a date, such equipment was serviced, such a part replaced, signed by so-and-so. That satisfies an audit but does almost nothing for management - you cannot find the measured data from the last similar job, you do not know how often that component is replaced, and no degradation trend is visible. A record worth keeping has at least four layers: the work performed (what was done, to what standard), measured data (clearances, alignment values, insulation readings, wear), replaced-part details (model, batch, quantity), and anomalies and recommendations (problems found, points to watch next time).

The measured-data layer is the one most often skipped - and the one worth the most. A concrete example: when changing a chain, record the old chain's measured elongation and cumulative running time. After three cycles you have a chain life curve for your own duty, and everything follows from it - how far ahead to place the next order, how many spares to hold. Conversely, if all you record is "one chain replaced", ten years later you still will not know how long a chain lasts under your conditions, and purchasing will always be guesswork.

The medium of the record does not matter; the standard does. Paper forms, spreadsheets or maintenance software all work - what matters is uniform fields, filled in as the work happens, and searchable by equipment number. When we perform maintenance, the record sheet is one of the standard deliverables: fields designed on the four layers above, completed and signed on the spot at handover. Filling records in afterwards is the main cause of data distortion, because the details are no longer remembered.