“problem formulation, hierarchical decomposition, and solution generation”

Sahadevan et al., Advanced Engineering Informatics, 2025
DESIGN SPACE / 05

A berth is only useful if you can reach it.

Explore an irregular quay, five yacht classes and a changing tide. Move the piers yourself, or search for a layout that makes better use of depth and manoeuvring space.

SELECT A YACHT TO WATCH ITS APPROACH

TIDE EXPLORER / RELATIVE TO MEAN SEA LEVEL

Contours show fixed seabed elevations. Stippled water is too shallow for the selected yacht’s draft and clearance at this tide. Berths are rechecked without moving your piers.

YACHTS ALLOCATED—
BERTH VALUE INDEX—
VS. BASELINE VALUE—

Search, movement & assumptions

Synthetic bathymetry, not a navigation chart. Seabed elevations are negative relative to mean sea level; water depth equals sea level minus seabed elevation. The low/mean/high tide presets are illustrative. The five yacht classes are 8 / 12 / 16 / 22 / 30 m long, with beams 2.8 / 3.8 / 4.8 / 6 / 8 m and drafts 1.2 / 1.8 / 2.4 / 3 / 3.7 m. Buttons show allocated / requested yachts.

Sliders directly set the layout. Search compares pier counts, spacing, position and staggering at the chosen reach and tide. A descending-size packing heuristic serves a fixed fleet. The search objective is either yacht count or a berth value index: length × (beam + 1.5) × (1 + draft / 3) + 0.35 × turning radius². This weights space, depth and manoeuvring room; it is a design proxy, not a price estimate. The baseline is always three piers, 75 m spacing and 100 m reach, checked at the same tide. Best means best among the tested candidates.

Approaches combine forward straight runs with tangent circular turns, ending with the bow facing the fairway. The yacht stops, changes gear and reverses straight into its berth, keeping the same heading. Minimum radii are 5 / 8 / 11 / 9 / 11 m. The two larger classes assume low-speed thruster assistance; no yacht makes an instantaneous point turn. Rotated hull envelopes are sampled along the route and checked against other moored yachts, piers, shoreline and water depth. Berths without a valid approach are excluded.

Selecting a yacht vacates its berth and animates the hull along its verified path. It accelerates in open water, slows into turns, stops for 0.8 seconds and reverses at up to 0.75 m/s before decelerating to dock; playback is 4× real time. Reduced-motion visitors can start this explicitly with Replay. This geometric study excludes real steering dynamics, currents, wind, waves, structural design, mooring lines and traffic scheduling.

The first model is the question

Design teams often automate evaluation after they have already fixed the problem representation. Yet the earliest choices — what counts as a variable, how functions relate, where boundaries sit — can exclude better possibilities before the search begins.

The Knowledge Augmented Generalizer Specializer research explores explicit functional, behavioural and structural reasoning for early concept generation. The design-by-analogy paper looks at retrieving and mapping knowledge across domains. Both point to the same opportunity for practice: useful creative support starts with a well-structured question and a way to revise it.

I treat a design model as a conversation among people, constraints and evidence. If a model cannot explain its assumptions, polishing its final answer will not repair the missing reasoning.

REFERENCES & FURTHER READING

Short paper excerpts above are attributed to their authors. Commentary is Kane Borg’s interpretation.