Proving a Device Works

Three documents about the same pre-swirl duct reach one superintendent's desk in a month. The vendor's brochure says 8 per cent. A model test at a reputable tank says 6.4 per cent, quoted to one decimal place. Eleven months of the ship's own data, differenced against a sister that did not get the duct, says about 1 per cent, and the analyst who produced it will not commit to the sign. Nobody has lied. The three numbers answer three different questions, and the superintendent has no procedure for telling them apart.

That is a measurement problem before it is an engineering one, and it is hard for a reason that has nothing to do with instrument quality. A saving is not measured; it is the difference between two measured quantities of almost the same size. On the Panamax bulker that carries this course, about 75,000 dwt, a device sold on 4 per cent is worth about 253 kW, sitting inside a day-to-day scatter of normalised consumption worth 500 to 950 kW. The thing you are trying to see is between a quarter and a half the size of the thing you cannot control, and a better torsionmeter does not touch that ratio. (She is a teaching hull with round numbers, and her power is light for her size: a real Panamax at that speed needs nearer 680 kN of resistance than her 577 kN, so every kilowatt quoted for her runs about 15 per cent below a real ship's. The ratios the course turns on are the same.)

What is inside: the four counterfactuals a claim can rest on and the specific confound that attacks each; the ladder of evidence from brochure to long in-service study, with the uncertainty band each rung can honestly carry and the awkward asymmetry that cheap evidence is good at refutation and bad at measurement; CFD read properly, which means verification and validation kept apart, a three-grid convergence index worked so you can repeat it, and a frank account of the flow regimes where a RANS answer cannot support a device claim; model testing, the analysis that turns a resistance and self-propulsion pair into a power difference, the scale trap and its direction, and the one device family a tank cannot help with at all; the sea trial taught through the ITTC wind-propulsion procedure, its run pairing, its environmental window and the measurements that decide the answer; the three in-service comparisons and how to repair the confound of the one you chose; the power calculation that tells you how many usable days a campaign needs before it can resolve the effect you are paying to detect; and the drafting of a performance guarantee with a measurand, a baseline, a method by edition and an acceptance rule that an arbitrator could apply.

The opening lesson, The Counterfactual Problem, is free to read as a sample. The worked ship is our reference Panamax bulker, the same ship that runs through most of this specialisation (Air Lubrication works on a Newcastlemax), and every number in the course is computed in front of you rather than asserted.

By the end you should be able to take any saving claim, name the comparison it rests on and the confound that comparison invites, say what its evidence legitimately establishes and where it stops, size the monitoring campaign that would settle the question and price the delay, and write or mark up a guarantee clause so that the two sides can tell afterwards whether it was met.