FRMS

Duty-time limits set the formal boundaries. They do not separate two duties of equal length that place sleep, wakefulness and work differently against the circadian rhythm.

One duty may end in daylight; another lands at the end of the night after a short rest. By the regulated hours the two can look the same.

Without a calculation that difference rests on the planner’s experience. After an occurrence fatigue may be named as a factor, and confirming its part is harder: no calculation existed at the moment the duty was released.

FRMS scores every planned, published and actual duty and returns a sleepiness forecast on the KSS scale, a profile over time and the score broken into its parts.

The planner has the result before the roster is published, can test an alternative in what-if mode and recalculate a duty when a flight is delayed.

Once a decision is taken the system carries the rest of the loop: the measure, its execution, the check on whether it worked, and the hazard factor and indicator handed to the SMS.

The planner, squadron commander, instructor, flight safety office, programme manager and auditor each work with their own slice of data and their own authority. Oversight roles have read access; the separation is enforced by the server.

• A score for every duty with the sleep and wake profile and an account of what the result is made of.
• What-if mode and recalculation on delay, leaving the source roster untouched.
• Crew fatigue reports; named access belongs to the flight safety office alone and carries its own log.
• Measures with proposal and approval split across roles; no single person moves a threshold.
• A reproducible evidence pack for the auditor, with a checksum and access limited in time.

Duties arrive from the planning system under a formal exchange contract: time in UTC, end-to-end identifiers for the duty, the pairing and the crew member, and the states planned, published, actual and cancelled.

Sending the same data twice creates no duplicate records. The result written back to the planning system is held apart from actual operational data.

The calculation runs in a separate biomathematical engine behind a replaceable software interface. The application holds no fatigue model of its own.

The forecast rests on the two-process model of sleep and wakefulness (Daan, Beersma, Borbély, 1984): homeostatic sleep pressure, a circadian component and a sleep-inertia process.

Time constants of 18.2 and 4.2 hours are used; the circadian component is the sum of a 24-hour and a 12-hour harmonic. The result is carried onto the KSS sleepiness scale by the Ingre (2014) regression.

Every constant carries its source, the exact place within that source and a confidence grade. Changing a value calls for a fresh run of the verification suite.

The KSS scale is left unclamped. On extreme rosters the calculation runs past its familiar range; a forced limit would hide exactly the cases that need attention.

Attention levels on a duty are a presentation rule laid over the engine’s result. They form no second risk model, and the interface shows that boundary plainly.

Likelihood and severity in the SMS risk matrix are set by a person. The model score is never placed into those fields automatically.

The link between fatigue and a given occurrence is likewise established by a specialist. The system does not compute it.

There is no summary score for a pilot. When replacement candidates are compared the order comes from the planning system, and FRMS adds a fatigue forecast for the duty in question.

A time-zone acclimatisation model is not provided by the engine.