| Vose Software

Industry: Defense
Product: ModelRisk
Application: Optimizing Naval Operations Under Uncertainty


A Five-Ship Task Group Holds the Station Barely Half the Time

A standing maritime commitment requires keeping at least two surface combatants effectively on station, continuously, for a year. The arithmetic looks reassuring: each ship cycles through roughly 45 days on station, a week of transit each way, and about five weeks of in-port maintenance, so a ship spends a little under half its time on task — and five ships should therefore deliver about two on station at any moment. Size the task group to that average and you commit five ships. The simulation says that a five-ship group meets the two-ship presence requirement in only 53% of campaign-years — barely better than a coin flip — because maintenance overruns, unplanned breakdowns and a shared sea-state state pull ships off station in clusters the average never shows.

A fleet operations staff rebuilt its force-generation case in ModelRisk after a rotation in which scheduled upkeep slipped and a breakdown coincided, briefly dropping the station below its mandated presence. The mandate was to size the task group against the distribution of delivered presence — with the assurance probability and the cost of a presence gap attached — not against the steady-state duty-cycle number.

On-station presence distribution by task-group size

The three task-group sizes produce three distinct presence distributions. The four-ship group centres at 1.6 ships and sits entirely left of the requirement; the five-ship group straddles the line, with nearly half its mass below two; only the six-ship group clears the requirement with margin. The width of each distribution — set by how the cycle phases line up across ships in a given year — is the planning signal the duty-cycle average discards.

Presence as a rotating cycle, not a static count

This is a sortie-generation and on-station-availability model, not an inventory or deployment-lift problem: each ship rotates through a maintenance-transit-station cycle, and presence is what the rotation delivers, not what a duty factor promises. The simulation runs 8,000 campaign-years, advancing every ship through its cycle day by day:

  • On-station endurance — LogNormal, median 45 days, right-skewed by fuel, stores and crew limits.
  • Transit each way — Triangular(4, 7, 14) days, reflecting routing and the best/likely/worst sea passage.
  • Maintenance availability — LogNormal, median 35 days, right-skewed: most upkeep periods run to plan, but the tail of overrunning availabilities is where presence quietly erodes.
  • Unplanned breakdowns — Poisson, ~0.35 per cycle, each adding about 12 days to the next maintenance period.
  • Weather / sea-state availability — a shared seasonal factor (Beta on [0.6, 1.0], mean ~0.91) that gates every on-station ship the same day. One heavy-weather state drops the whole task group's effective presence together, so presence is not a sum of independent ships. That common factor is what keeps the presence distribution genuinely wide instead of collapsing toward its mean.

The deterministic duty-cycle calculation uses mean phase lengths and a mean weather factor; it never sees the year when two maintenance overruns and a rough-weather stretch coincide — the conjunction that opens a presence gap.

Why a point estimate fails here

The duty-cycle math is not wrong about the average — five ships really do deliver about 2.01 on station on average. It is wrong about the thing the commitment actually demands: reliability. Presence is driven by how the cycles phase against each other and against the shared weather state, and those align badly often enough that:

  • 4 ships: mean presence 1.61, and the requirement is met in 0% of years — the average alone already condemns it.
  • 5 ships: mean 2.01 — dead on the requirement — yet met in only 53% of years, with a P10 of just 1.78.
  • 6 ships: mean 2.41, P10 2.15, requirement met in 97% of years.

A force sized to "average equals requirement" is a force that fails the requirement nearly half the time. Only the distribution exposes that the mean sitting exactly on the target is the worst place to be.

Sizing the task group: assurance against cost

Task-group sizing presence assurance vs cost

Sweeping the rotation from 3 to 10 ships traces the trade-off. Fleet sustainment cost rises linearly at $14M per ship; the presence-shortfall penalty — the cost of buying alternative coverage when the station falls short, here $0.3M per ship-presence-day below requirement — falls steeply as ships are added. Total cost bottoms out at a five-ship group ($77M), the cost optimum. But that optimum holds the station only 53% of the time. The six-ship group is the first to reach 90%+ assurance (98%), for about $7M more — and from six ships up, the shortfall penalty has effectively vanished, so every further ship is pure cost with no readiness return. The cost-optimal force and the assured force differ by a single hull, and only the simulation shows that the cheaper option buys a coin flip.

What drives sustained presence

What drives sustained on-station presence

Around the five-ship group — baseline of 2.01 ships on station — the tornado ranks the levers. On-station endurance dominates: a ±25% change in how long a ship can hold station before cycling home moves mean presence by about 0.56 ships, more than any other factor. Maintenance/upkeep duration is second (~0.42 ships) — every day of upkeep overrun is a day off station. The shared weather/sea-state factor is third (~0.32 ships); unplanned breakdowns and transit time follow well behind. The reading: extending on-station endurance (longer logistics legs, at-sea replenishment) and tightening upkeep schedules buy presence more cheaply than adding a hull — exactly where a duty-cycle spreadsheet offers no guidance.

How reliably each task group holds the station

How reliably each task-group size holds the station

The cumulative view makes the assurance gap explicit. The five-ship group falls below the two-ship requirement in 47% of campaign-years; the six-ship group, in just 3%. Reading the curves the other way, the six-ship group delivers a P10 presence of 2.15 ships — even a bad year clears the requirement — which is the property a standing commitment actually needs and a single mean-presence figure can never demonstrate.

What the model changed

  • The task group was set at six ships, not the cost-minimizing five — the assurance-optimal size — because the marginal hull buys the jump from a 53% coin flip to 97% presence assurance.
  • An at-sea-replenishment and endurance-extension program was prioritized ahead of further hull commitments once the tornado showed on-station endurance as the top presence driver.
  • Maintenance scheduling was tightened to attack upkeep overruns, the second-ranked driver, recovering presence without adding ships.
  • Expected presence shortfall fell from 143 ship-days a year (four ships) to about one (six ships) — the alternative-coverage cost saved many times over the added sustainment bill.

ModelRisk Functionality Used

  • Rotating-cycle presence simulation across 8,000 campaign-years — LogNormal endurance and maintenance, Triangular transit, Poisson breakdowns, and a shared sea-state common factor — producing the delivered-presence distribution the duty-cycle formula cannot.
  • Fleet-size sweep tracing the cost U-curve and the assurance curve from one engine, exposing the single-hull gap between the cost-optimal (5) and assurance-optimal (6) task group.
  • Tornado on mean presence ranking on-station endurance and maintenance duration above weather and breakdowns — pointing investment toward endurance and upkeep discipline rather than hull count.
  • Presence CDF quantifying P(maintain ≥ 2 ships) at 0%, 53% and 97% for the three sizes, with the six-ship P10 of 2.15 proving even a bad year holds the station.

A presence plan written against a mean duty cycle is a plan for an average year that the rotation rarely delivers; the task group sized to that average is the one that opens a gap in the campaign that matters. ModelRisk turns "how many ships hold the station?" into "what is the distribution of presence delivered, and what is the smallest force that holds the line even in a bad year?"