Measurement Adequacy and Parameter Identifiability Before Calibration of a Physics-Informed Ship-Performance Model: A Methodological Case Study on Four Tanker Voyages

Calibrating ship-performance models on operational data presupposes traceable, independent and sufficiently excited onboard channels. This methodological case study audits 207 records (0.5–5 h each) from four voyages of one tanker before calibration; the fuel and power instruments are undocumented. A five-parameter fuel observation is structurally limited to two composite directions, of which the available records support one: a two-composite fit returns inadmissible added-power multipliers. The scalar estimated instead κ is the ratio of reconstructed main-engine fuel to reported power at a nominal specific consumption, not a physical efficiency. Frozen after the opening quarter of each voyage, κ predicts remaining-route fuel totals to within −5.67% to +0.70%, against −24.95% to +0.16% uncorrected; a scalar fitted once on the same window performs equally, so the gain is scaling, not recursion. Across two fuel-conversion baselines, two fuel-field readings and seven freeze-point definitions, the totals stay within −8.5% to +2.1% and the dispersion of κ within 25–31%; per-voyage RMSE changes and segment-level speed profiles are not stable. Reported and modelled power are different observation operators: on matched references, the V2 cost reduction is +1.15% with κ and +0.06% with the scale that modelled power requires. Optimisation results are conditional sensitivity experiments, not realised savings.