Ice Yacht 2026 · Preliminary engineering study

Rebuilding a 1999 record-attempt ice yacht as a verified 3-D vehicle — and simulating it properly for the first time

A quarter-century-old paper design for a 240 km/h single-seat ice yacht existed only as three dimensioned 2-D views and a set of hand-calculated force balances. This study rebuilds it as a parametric 3-D model, checks that a real pilot fits inside it, and puts the complete vehicle through full 3-D viscous computational fluid dynamics — the first time this concept has been simulated as a whole.

The headline finding is not a faster vehicle. It is a trustworthy one: a lower predicted speed than the original claimed, arrived at with methods, checks and stated uncertainty the original did not have — and a safety margin problem the original had no way to see.

Above: conceptual visualisation. The vehicle’s proportions and layout follow the engineering model; the carbon finish and setting are presentational.

What changed: from a drawing to a vehicle

The original design survives as a dimensioned general-arrangement drawing. Everything about its three-dimensional form — how surfaces meet, whether a pilot fits, how air actually moves around it — had to be reconstructed and then built as a real solid model.

The vehicle, orthographically

Does a pilot actually fit?

The original design did not resolve pilot accommodation. In a vehicle whose entire performance case rests on a small frontal area, this matters: every millimetre of cabin height is paid for in drag. The 2026 study places a 176 cm pilot inside the hull in three dimensions, with seat, inclined footrest, steering yoke, five-point harness, mast-step bulkhead and a checked forward sight line.

Clearances are tight but positive throughout, the governing point being the pilot’s thigh with 2.9 mm to spare. Cockpit structure and fittings total 4.4 kg.

Surface quality

A body shaped for 240 km/h cannot have subtle waviness in its surfaces. The hull is checked with zebra striping — a standard surfacing technique in which reflected stripes reveal curvature discontinuities the eye would otherwise miss. Smooth, evenly-flowing stripes mean a fair surface.

The engineering toolchain

Every result in this study is produced by open, scriptable, reproducible tools — no manual modelling steps and no unrecorded hand calculations.

Simulating the complete vehicle

The vehicle is enclosed in a virtual wind tunnel and the space around it divided into 618,030 cells, finest at the surfaces and in the wake. The solver then computes the steady airflow, pressure and wall friction over the whole machine at racing speed.

Operating condition: 239.1 km/h with an 66.9 m/s apparent wind at 7.12° of yaw — a beam reach at racing speed. Steady RANS, k-ω SST turbulence, 4,500 iterations to convergence.

The flow field

What the simulation found

Averaged over the final 500 of 4,500 solver iterations. Forces act in vehicle axes: x along the hull, y across it, z vertical.

Two limitations, measured rather than assumed

Both were found by the review process, and both are quantified rather than merely flagged.

Progressively excluding the sharp junction corners barely moves the dominant sail force, but moves the small along-hull force substantially — which is why this study reports the sail force with confidence and the drag figure as indicative.

What it means for the vehicle

Feeding the simulated aerodynamic drag into the project’s vehicle model — which also accounts for sail trim, runner friction, weight transfer, roll stability and runner loading — gives the speed the vehicle could actually sustain while remaining safe and controllable.

What would close the gap

Roll margin must reach 1.25 and runner load must stay above 200 N for the vehicle to be considered controllable. No single change reaches the original 239 km/h target; the strongest reaches 223 km/h.

Compared with the original study

Limitations of this study

• One mesh and one operating condition. Grid-independence has not yet been demonstrated, so these figures cannot close a formal design gate on their own.
• The first-pass mesh omits boundary layers, leaving near-wall resolution coarser than ideal over much of the vehicle.
• Junction fillets present in the approved CAD are absent from this simulation surface, a deliberate simplification that introduces measurable local errors at those corners.
• Steady-state simulation: gusts, manoeuvres, suspension motion and ice-surface roughness are not modelled.
• The vehicle-level speed figures come from a screening model intended for design comparison, not certification.
• This is a research and design study. It is not a construction release, and a vehicle of this type would require independent aerodynamic, structural and safety review before any physical build.

Next steps

• Refine the near-wall mesh specifically to bring surface resolution into the turbulence model’s valid range, which would tighten the drag figure.
• Reinstate fillets at the two junctions shown to matter, rather than across the whole vehicle.
• Evaluate crossbeam downforce as the primary route to closing the stability gap, within the structure’s measured strength envelope.