# LW33 — layer-transition diagnostics R12 / R13

## Current boat reference

User-corrected CG: **2,750 mm (2.75 m)**. Datum and measurement direction remain unconfirmed. Loaded mass: **4,100 kg**. One **Volvo Penta D6-340**. Reported propeller: **Volvo Penta H6**; **H8 is a proposed candidate, not a verified match**. Drive model, ratio and vertical CG remain unspecified.

These corrections are saved in `LW33-reference-inputs-v2.json`. They have not yet been applied to hull equilibrium, the flow cases, or the published cockpit defaults. The new CG must not be equated to the earlier aft-keel datum without confirmation.

## What was tested

OpenFOAM v2412, fixed hull, two-phase flow, prescribed towing speed 30 kn. Historical 3,500 kg reference, initial waterline and inferred LCG 2.9405231959611386 m forward of aft keel were retained to isolate the mesh investigation. This is not a simulation of the updated loaded boat or a prediction that it can attain 30 kn.

R12 changes the original six-layer termination buffer from 0 to 3. R13 is a separate repair candidate: three relative-thickness layers below the waterline, expansion ratio 1.2, final-layer thickness 0.3, minimum thickness 0.05 and termination buffer 1. R13 changes several mesh parameters together, so it is not a single-parameter causality test. The hull triangulation is identical in both: SHA-256 `e7fada1f1334c7998e43d9067d4e9a4d511f39464fe5ca6659a95d8d2f5554ef`. The source STEP is unchanged.

## Recorded results

| Case | Saved time (s) | Maximum air speed (m/s) | Maximum water speed (m/s) | Wet area with ≥3 layers | Poor determinant cells |
|---|---:|---:|---:|---:|---:|
| LW33-r12-buffered-termination | 0.004962643 | 31.08 | 16.63 | 4.61% | 28 |
| LW33-r13-three-layer-transition | 0.004962644 | 31.08 | 16.63 | 35.46% | 42 |

Air is sampled where water fraction <0.01; water where water fraction >0.99. These are phase velocity maxima, not resistance or boat-speed predictions. Layer coverage uses face-centroid selection below the historical initial waterline. R08 previously reached a maximum water speed of 615.70 m/s at 0.0041105473 s with 332 poor-determinant cells.

- LW33-r12-buffered-termination: solver exit 0; stop reason: normal solver completion; wall time 262.4 s; all saved velocities finite: True.
- LW33-r13-three-layer-transition: solver exit 0; stop reason: normal solver completion; wall time 276.3 s; all saved velocities finite: True.

## Status of requested instability fix

Short startup survival is useful evidence, but **step 2 is not closed**. Both candidates still fail three detailed mesh checks: small cell determinants, concave cells and small face interpolation weights. A zero checkMesh process exit code does not override these failures. R12 also collapses most nominal six-layer stacks to only one or two layers. R13 retains more three-layer coverage but has not established adequate wall resolution.

The requested run endpoint is 0.005 s; actual saved endpoints are shown above. Adaptive stepping can finish near the requested endpoint. These intervals are far shorter than one hull transit time (~0.64 s at 30 kn). They cannot demonstrate developed calm-water flow, converged resistance, adequate y+, or stability over a useful simulation duration. Force histories are retained for diagnostics only.

Next numerical work: repair remaining layer-transition and mesh-quality defects; confirm adequate wall resolution; extend the selected mesh beyond startup; then perform mesh/time-step sensitivity and force convergence checks. Apply the confirmed 4,100 kg loading and CG datum to hydrostatic equilibrium before producing a new boat baseline. Independent sea-trial comparison is still required. No validated top speed, fuel consumption, engine/propeller match, running trim or wave response is supplied.

## Reproduction and archive layout

Each checkpoint includes its mesh inputs and generated mesh, initial and final two saved flow fields, complete post-processing, logs, audits and scripts. OpenFOAM v2412 and Python with VTK/NumPy are not bundled. The fluid mesh is stored once under `meshes`; after extraction run `python code/restore_transition_mesh.py r12` or `r13` to copy it into the corresponding case. Run that from within the extracted `lw33-cfd-update` directory. Existing meshes are never overwritten.

Inspect the included system dictionaries before rerunning. Preserve the evidence by making a separate working copy. The archived flow case is already initialized and can be used with OpenFOAM directly. The mesh creation and initial run scripts are provenance scripts that reference historical source cases; fresh mesh generation can instead use the complete archived mesh inputs with blockMesh, snappyHexMesh -overwrite and checkMesh -allGeometry -allTopology in a separate copy. The finalizer can re-extract phase maxima from an extracted case after restoring its mesh.

The published site has not been changed by this numerical checkpoint.