When we are asked to analyse an incident, we start with the parameter file, not the hardware. In a large share of cases the problem was already written there on commissioning day. These are the seven we meet again and again.
1. Protections left at factory defaults
A controller ships with generic default values. They are not wrong; leaving them is. Overcurrent must follow from the specific generator's characteristics and coordinate with the downstream protection; reverse power from the power the engine needs to be motored unloaded. Neither is known at the factory.
2. No breaker closing-time compensation
The synchronizer issues the command; the breaker closes tens or hundreds of milliseconds later. In that interval the phase angle has moved. If closing time has not been measured and entered, every closure is systematically out of phase — slightly, every time, for years.
3. Synchronizing windows opened too wide
When synchronizing becomes difficult, the quick 'fix' is to widen the dV, df and dφ windows until the breaker finally closes. The problem has not been solved; it has been transferred to the shaft and the windings.
4. Logic written without comments and without a backup
Modern controllers allow programmable logic for vessel-specific requirements. It is an excellent tool and simultaneously the easiest way to make a system opaque. Without comments and without an exported backup, the next team has no way to understand why the vessel does something that is written down nowhere.
5. Settings not migrated after a unit replacement
A unit is replaced under time pressure, the new one goes in with basic settings so the vessel can sail, and there it stays. Six months later nobody remembers the installation is running a temporary configuration.
6. Misconfigured inputs that are not used
Inputs declared but never wired — or the reverse — create latent block conditions. The system runs normally until the day one condition asserts and the controller refuses a start command with no visible cause.
7. Commissioning without fault-condition testing
The most expensive of the seven. A system tested only in normal operation has proven nothing about its behaviour in a fault. The tests skipped most often:
- Loss of one generator under load, verifying that load shedding acts in time.
- Blackout and automatic recovery, timed.
- Failure of the standby machine to start, and take-over by the next in priority.
- Loss of communication between controllers — what the system does when the bus drops.
- Operation on emergency power and clean return to the main network.
None of these is exotic. All of them follow from conditions the vessel will eventually meet — and the only choice is whether it meets them with an engineer alongside or in the middle of the night at sea.