# LW33 R11 — correction-limiter control

## Result

R08 restart at 0.003 s; non-orthogonal Laplacian and surface-normal gradient corrections limited to 0.5. Geometry, surface tension and R08 advection retained. Final saved air/water maxima 31.53/128.78 m/s at 0.003941077 s. Startup diagnostic only, not a reliable calm-water baseline.

Stop reason: global velocity exceeds 100 m/s. Solver exit code: 0.
The maximum saved velocity occurs in cell 55864, with water fraction 0.99999988.

## Controlled change

Restarted from the same pre-failure R08 fields used for R09/R10, at 0.003 s.
Only the non-orthogonal correction family changed: both Laplacian and snGrad
schemes use `limited 0.5` instead of full correction. The installed OpenFOAM
v2412 `limitedSnGrad.H` describes 0.5 as limiting the non-orthogonal component
to the orthogonal component. This is not an assertion that the local limiter
activated at every failing cell.

The hull geometry and fluid mesh were not changed. Prescribed towing speed is
30 kn with a fixed hull; this is NOT a predicted attainable speed. Surface
tension remains 0.072 N/m. Controls request a stop above 100 m/s; the next saved
step may exceed that threshold. Maximum-velocity histories are global across
both phases and begin at the restart, not at zero.

## Interpretation and next work

This correction-limiter change alone did not establish stable flow. It does
not prove the mesh is the sole cause. The next numerical work remains local
layer-termination/transition quality and a controlled remesh, followed by
longer runs and mesh/time-step sensitivity checks. No engine, resistance,
running-trim or wave-response predictions are validated by this diagnostic.

## Other workstream status

The cockpit now implements a server-side allowlist for the private log; an
empty allowlist denies every account. The owner must sign in and provide the
site-scoped account ID before authorized access can be configured. Public
hull viewing stays separate. Tests use synthetic observations in a local
test database; they are not sea-trial evidence.

Browser/device verification remains incomplete: the supported Sites managed
preview does not support this custom buildless Worker setup. No real-device
pass is claimed. Real RPM/GPS/fuel observations, measured CG and exact
Duoprop/drive details were not found in the available project material.

## Reproduction

The archive contains the R11 case, the 0.003 s starting fields, final saved
fields, logs, results and preparation script. The unchanged mesh is included.
The run script refuses to overwrite an existing case. For a fresh rerun,
prepare a separate case directory from the included inputs; do not overwrite
the evidence. OpenFOAM v2412 and Python/VTK are not bundled.