# LW33 R04 — airflow isolation, 10 September 2026

No validated calm-water performance baseline is available. The original STEP is unchanged. Reference loading is 3,500 kg with inferred CG; 30 kn is prescribed towing speed, not a D6-380 speed prediction.

| Control | Saved time (s) | Max air U (m/s) | Max water U (m/s) |
|---|---:|---:|---:|
| Layered, air only | 0.00049449766 | 951.59 | Not applicable |
| Layered, zero surface tension | 0.00049529981 | 939.93 | 16.82 |
| Layered, one outer corrector | 0.00064443980 | 917.04 | 16.81 |
| Same refinement, wall layers disabled | 0.01000000000 | 28.98 | 16.45 |

The first three cases retained the 465,010-cell mesh and stopped after the timestep fell below 5 microseconds. The no-layer mesh has 307,421 cells, passes standard checkMesh, reached 0.005 s in 289 wall seconds and restarted to 0.010 s without timestep collapse. Final timestep was 0.0002 s. The air maximum was 31.06 m/s at 0.005 s. Air/water maxima use alpha.water <0.01 / >0.99 at saved cell centres; these are not volume averages.

Interpretation: the instability is reproducible without a water interface, so surface tension is not necessary to trigger it. Removing layers suppresses the early collapse. This strongly implicates the layered mesh or its interaction with the discretisation. It does not identify one uniquely faulty cell or prove a stable wall-resolved replacement. The meshing change also changes cell topology/count.

The no-layer test is a diagnostic, not an acceptable friction/resistance mesh. The fixed hull cannot establish running trim or heave. At 30 kn one hull-length transit is about 0.64 seconds; 0.01 seconds is only 1.6% of one transit. Startup forces must not be used as steady resistance, lift or engine/propeller performance.

## Remaining engineering gates

1. Rebuild wall layers with controlled transition, thickness and quality, especially near the above-water hotspot (approximately X5.616, Y-1.234, Z1.173 m). Compare initial potential-flow field, continuity and local cell quality against the no-layer control.
2. Repeat the fixed-hull 30 kn startup with the corrected wall-resolved mesh; require bounded air velocities, no timestep collapse and consistent water fields before extending.
3. Re-establish initial equilibrium with confirmed loading/CG, then run calm-water heave/pitch long enough for force and attitude statistics to settle. Assess averages across consecutive physical-time windows, not wall-clock duration.
4. Quantify mesh/timestep/domain sensitivity and compare resistance/trim with sea-trial observations. No wave-response or engine-performance prediction until this gate is met.

## Contents and restart

Controls ZIP contains the three layered cases. Mesh-Control ZIP contains the no-layer case, its mesh-generation dictionaries, and both startup/restart logs. Each case includes initial fields, mesh, system dictionaries and saved checkpoints. Extract both alongside prior lw33-cfd-update archives. Scripts refuse to overwrite generated cases. Existing no-layer case can be restarted with OpenFOAM v2412 by setting startFrom latestTime and a reviewed endTime in system/controlDict, then running interFoam -case <case-directory>. This is an explicit diagnostic continuation, not a certified simulation workflow.

Raw summaries: audit/r04-trials.json. Force histories: data/r04-records.json and case postProcessing files. The cockpit displays these separately from measured sea-trial observations.