One fault, three weeks: why the same repair is a day ashore and a casualty at sea

The fault itself is ordinary

A mid-life vessel, four auxiliary generators, the plant arrangement you would find on several hundred ships in the same trade. One machine's automatic voltage regulator loses control of excitation. Field current collapses, the machine can no longer hold its share of reactive load, and the protection does exactly what it was designed to do: it takes the machine off the board.

In a shore power house that is a first-line job. A technician is on site the same shift, the spare regulator is in a cabinet down the corridor, the plant runs on the rest of the sets meanwhile, and the paperwork is a work order closed the same day. The engineering content of the failure is genuinely small.

At sea the same small fault removed the board. The remaining machines saw the load step, the sequence did not hold, and the ship went dark. Nothing exotic happened. What differed was not the fault. It was everything around it.

What actually costs the weeks

Nobody sends an invoice for the failure. They send invoices for the consequences of where it happened. The spare has to reach a specific hull in a specific place, which means a supplier, a freight agent, a customs entry and a berth that is already committed to cargo. The specialist has to reach the same hull, which adds a flight, a visa and a gate pass. The class surveyor has to be available for attendance, and class availability is a queue, not a phone call.

Then the parallel process starts. A blackout on a ship is not only an engineering event; it is a reportable one. Flag wants a report. Class wants the condition surveyed and, until it is, may attach a condition of class. The port state has its own view on whether the ship sails. Each of those is a different organisation with a different clock, and none of the clocks are yours.

But the number that actually hurts is not written by any of them. A liner ship runs a fixed proforma schedule with committed arrival windows down the whole string, and berth windows are first-come-first-served. A hold of a single day does not cost a day. It costs the window — and once the window is gone you are behind whichever vessel is already alongside, waiting for the terminal to clear, while the boxes behind you stack up and every port down the rotation re-shuffles around other people's committed slots. Recovery costs more again: a skipped call, or steaming hard and burning the fuel to claw the schedule back.

So the hold is only the trigger; the cascade is the cost. It is entirely possible for the authorities to release a ship inside twenty-four hours and for that day to detonate into a fortnight of rotation. The engineering was a morning. The commercial hole was weeks — and it was weeks because of berth-window scarcity, not because anybody detained anything for weeks.

The asymmetry nobody prices in

Ashore, a plant failure is contained by proximity: parts, people and authority are all within reach, and the plant can usually keep running while you fix it. At sea, every one of those is remote, and the failure removes several unrelated capabilities at once. Losing the board does not just stop a generator — it takes steering, it can take the deck machinery you would need to recover, and it takes the very services the emergency plan assumed would be there.

This is the part that gets flattened in a management summary. The report says the regulator failed. It does not say that the same failure ashore would have cost one shift and at sea cost a berth, a surveyor, a flight, an attendance and a fortnight of somebody's undivided attention. Both statements are true; only one of them is the reason budgets get missed.

The human factor sits before the fault, not after it

There is always a temptation to ask what the watchkeeper should have done differently. It is almost always the least useful question. By the time a board is lost, the crew are executing a recovery in the dark with the tools the plan left them, and any competent team performs about the same.

The place where attention actually mattered was weeks earlier and looked like nothing. An excitation-related alarm that came and cleared. A machine that took slightly longer to pick up load than its sisters. A regulator adjusted once, by somebody who has since gone on leave, with the reason living in a handover conversation rather than in a record. None of those stops a ship. All of them are visible — to somebody who is not simultaneously running three surveys, a bunker plan and a crew change.

That is what the human factor means in practice on a working ship. Not carelessness. Attention as a finite quantity, allocated under fatigue and interruption, spent on whatever is loudest rather than on the quiet thing that is drifting.

What a maintenance system owes you here

Three things, and none of them are dashboards.

It should notice the drift. A machine that behaves slightly unlike its sisters, an alarm that recurs and clears, a parameter that walks away from its own history — those are the only warnings this class of failure ever gives, and they are given early enough to be useful.

It should hold the evidence before anyone asks for it. When flag, class and the port state each open a file, the ship should be able to produce the maintenance history, the alarm record and the condition evidence as an existing artefact rather than as a week of somebody reconstructing it under pressure. The delay after a casualty is very often an evidence-assembly delay wearing an engineering costume.

And it should treat attention as a resource. If the plan puts three attention-heavy jobs on the same watch, that is a schedulable risk, and the system that built the plan is the one that should say so.

Why this is published this way

No company, no vessel, no port, no date, no maker, no report number. The plant arrangement and the age are described in bands broad enough to fit a large population of ships, and nothing here asserts a cause, a liability or a closure for any actual investigation — those belong to the authorities that hold them, and they are not ours to pre-empt.

What is left is the pattern, and the pattern is the whole point. Nothing above depends on knowing which ship it was. If you have run machinery at sea, you have met it.