# LW33 — readiness, handoff and remaining work

Updated 11 September 2026. This is a development handoff, not a validated performance report.

## Load, CG and wave sensitivity — 11 September

51 paired cases completed at 240/480 Hz: 15 load/speed cases (3,600/4,100/4,500 kg, 20–40 kn), 18 CG cases (±300 mm longitudinal offset and assumed VCG 0.6/0.8/1.0 m at 25/35 kn), and 18 wave cases (regular/irregular/crossing, 0.6/2 m, headings 0/90/180° at 30 kn). All 33 calm cases settled; all 18 wave cases met the internal time-step comparison screen. The model was not changed for this sweep. These finite cases do not establish accuracy, stability across all conditions or safe operating limits. See sensitivity.md/json and the interactive sensitivity.html review.

## Latest calm-water check — 11 September

The previous formulation failed to settle at 25–30 kn. A controlled correction includes relative contact vertical velocity in dynamic lift incidence. Repeating the 60 s runs at 0, 10, 15, 20, 25 and 30 kn, at both 1/240 s and 1/480 s, now yields numerically settled results under the same assumed inputs. Mass, CG, inertia, lift coefficient and damping constants were unchanged. **This resolves the reproduced default-model oscillation, not real-boat validation.** See **calm-water-check.md/json** for current results and **calm-water-check-r01.md/json** for the preserved failures.

No new CFD run or real-browser/device visual test was completed. The supported preview environment does not support this app's current buildless Worker setup. The earlier CFD findings below remain historical, not superseded by the browser motion model.

## Measurement and comparison workflow

Open measurements.html from the cockpit. It provides a blank, printable sheet for mass inventory, CG reference/direction/height, drafts, hull closures and drive/propeller evidence. Filled sheet data can be exported/imported as JSON; there is no automatic saving or upload. Print the sheet to PDF for the engineers and attach the marked drawing/photos separately.

Import a single observation JSON or the sea-trial log backup into the comparison view. Measured running trim is compared with the existing assumed-CG calm-water model, using bilinear interpolation only inside 3,600–4,500 kg and 20–40 kn. No engine RPM or fuel-flow prediction is offered. Unknown CG/geometry matching is explicit; all deltas are provisional benchmark differences, not validation errors. Calibration and held-out validation roles are recorded separately without automatic fitting. Export comparison records before closing the page. A held-out record used in tuning must no longer be treated as independent validation.

## Current boat reference

- Loaded mass: **4,100 kg**, user estimate (3,600 kg dry + 200 kg fuel + 300 kg people/load); measured weight and detailed inventory pending.
- Engine: **one Volvo Penta D6-340**.
- Reported propeller: **Volvo Penta H6**. H8 is an unverified candidate, not a confirmed match.
- Reference CG: **2,750 mm**. Datum, measurement direction, vertical CG and its datum are unconfirmed.
- Drive model, gear ratio, trim-reference convention and pitch inertia remain unspecified.

## Completed work

- STEP-derived hull model and replay cockpit, with separate visual-water animation and recorded CFD histories.
- Three controlled R19–R21 mesh investigations; rejected all-layer removal and documented wall-resolution tradeoffs.
- R18 continuation completed normally at **0.025016747 s**, with final air/water maxima **25.47 / 15.97 m/s**. This remains fixed-hull startup at imposed 30 kn, on the historical 3,500 kg reference.
- Independent determinant calculations explain why 14 R18 boundary cells pass the builder metric but fail the final connectivity metric.
- An additional R22 tetrahedron-quality control; its actual metrics are appended below when complete.
- Provisional zero-trim mass sensitivity for 4,100 kg, reproducing the previous 3,500 kg result first.
- Public shared sea-trial log, CSV export, JSON draft export/import and JSON backup of loaded current observations.
- Durable revision history from this update onward; old revisions can be loaded into the form and saved as a new revision. Existing observations receive a starting snapshot of their current state; previously overwritten history cannot be reconstructed.
- Standalone HTML embeds hull, histories, report downloads and boat reference. Public writes and revision history require the live site.

## Provisional flotation result — not loaded equilibrium

| Mass | Assumed density | Displaced volume | Zero-trim waterline above historical aft keel |
|---|---:|---:|---:|
| 3,500 kg | 1,025 kg/m³ | 3.4146 m³ | 0.47481 m |
| 4,100 kg | 1,025 kg/m³ | 4.0000 m³ | 0.51041 m |
| 4,100 kg | 1,000 kg/m³ | 4.1000 m³ | 0.51630 m |

At imposed zero trim, the seawater assumption raises the waterline about **35.6 mm**. This solves displaced volume only. It does not solve longitudinal moment balance, static trim, stability or running attitude, and has not been applied to CFD. The historical hull coordinate datum must not be equated with the user's CG datum.

**Geometry limitation:** the computational hull includes artificial closure patches with boundaries beginning around 0.492 m above the historical aft keel. The heavier-load waterline reaches some of these regions. Their correspondence to the real hull/closed openings must be checked before accepting the loaded hydrostatics. The original STEP has not been modified.

## Inputs to obtain from Kris / LW engineers

| Required input | What to provide | Why it matters |
|---|---|---|
| Longitudinal CG | Mark the reference point on a side-view drawing; state whether 2,750 mm is forward/aft and horizontal/along another line | Places the supplied CG in CAD coordinates |
| Vertical CG | Height, reference plane and measurement/calculation method; uncertainty if estimated | Static trim balance, stability and motion |
| Loaded weight breakdown | Hull/equipment, engine/drive, fuel litres, water litres, people/locations and cargo; explain whether these total 4,100 kg | Reproduces the operating condition |
| Drive and ratio | Model/serial-plate photo or specification confirming reduction ratio | Converts engine RPM to propeller RPM |
| Propeller | Markings/photos of both H6 propellers; confirm what is actually fitted | Separates current hardware from H8 alternatives |
| Openings and closures | Confirm whether the capped hull regions are sealed external surfaces or real openings; supply updated geometry if needed | Establishes valid submerged volume and flow boundary |
| Hydrostatic check | Measured bow/stern waterline or drafts at known loading, with measurement positions | Checks the CAD flotation result |
| Motion properties | Pitch inertia or mass inventory sufficient to calculate it | Required before accepting free-pitch and wave response |

A fillable data structure is supplied in **LW33-Engineering-Inputs.json**. Null values mean unresolved; they are not defaults. A marked-up drawing is preferable to a guessed reference name.

## Engineering and computational work still to perform

| Order | Work | Responsible workstream | Acceptance evidence |
|---|---|---|---|
| 1 | Confirm physical closure geometry at the new waterline | LW engineers + CAD analysis | Submerged boundary matches the real boat |
| 2 | Resolve remaining cell shape/connectivity defects with sufficient wall resolution | CFD development | Complete mesh audit; justified treatment of every retained warning |
| 3 | Solve the 4,100 kg loaded equilibrium after CG mapping | Hydrostatic analysis | Displaced weight and moments balance; draft/trim checked against measurements |
| 4 | Establish turbulence/wall treatment and assess y+ during flow | CFD development | Near-wall resolution consistent with the chosen model |
| 5 | Run longer calm-water cases, first fixed and then free heave/pitch | CFD development + compute environment | Stable phase fields, continuity and settled force/motion histories |
| 6 | Test mesh, time-step and domain/boundary sensitivity | CFD verification | Changes meet an agreed accuracy target; uncertainty reported |
| 7 | Build resistance/trim/sinkage curves over selected speeds and loads | CFD development | Accepted cases with recorded inputs and convergence evidence |
| 8 | Obtain engine/drive/propeller performance data and model losses | Propulsion analysis | Traceable torque/power/thrust/efficiency inputs, not guessed H-series curves |
| 9 | Compare with measured reciprocal sea trials | Boat team + analysis | Error bands across RPM/speed/load; hold back independent runs for validation |
| 10 | Add regular waves, then irregular seas and headings | Seakeeping analysis | Wave propagation verified; motions and added resistance checked |

### Compute requirement

The latest continuation advanced about 0.01503 simulated seconds in 653 seconds of serial computation. A linear extrapolation is roughly **12 hours per simulated second**, before sensitivity studies; this is a rough planning estimate, not a runtime guarantee. One hull transit at imposed 30 kn is about 0.64 s, and a transit alone does not imply convergence. Solver efficiency and a suitable longer-running compute environment need attention before a broad performance sweep. MPI execution was unavailable in the current environment; a production parallel run has not been demonstrated.

## Sea-trial collection procedure

1. Confirm the boat is fit for testing and choose suitable calm conditions with the boat team. Use normal operating limits and an appropriate test area.
2. Record date/time, load inventory, engine/drive/propeller, water conditions, wind/current and instrument sources. Record trim/tab settings consistently; a drive-gauge number is not automatically an angle in degrees.
3. At each chosen operating point, let readings settle and record engine RPM, GPS speed, fuel flow when available, heading and observations. Do not treat a transient peak as a settled reading.
4. Repeat in the reciprocal direction close in time at comparable loading/settings. Pair the records in notes and repeat enough to assess variability. Opposite-direction averaging reduces some current influence; it does not remove all wind, wave or changing-current effects.
5. Preserve raw instrument logs when available. Use the cockpit for observations; export the log after each session. Do not fill unknown fuel flow or load with invented values.
6. Provide a separate validation set after fitting or adjusting the model. A fitted match alone is not independent validation.

## Website work remaining

- Real Chrome desktop and mobile Safari/Chrome rendering and interaction QA. Automated logic checks are not a device/browser pass. The existing buildless Worker has no compatible managed preview in this environment; no live Sites cloud-browser test was performed.
- Hosted CFD job execution, resource limits, queueing and durable result storage. The public cockpit currently prepares requests and replays saved tests; public access does not create solver infrastructure.
- Complete-log pagination and a tested bulk restore workflow if the log grows beyond the current 5,000 loaded observations. JSON backup is explicitly limited to loaded current rows; revision history remains on the live site. Independent backups and a restore drill remain required.
- An accepted-result report format with uncertainty once validated results exist. Current reports correctly label startup diagnostics.

## What “good” means for the next milestone

One reproducible **4,100 kg calm-water operating point**, using the confirmed CG and physical hull, with acceptable mesh/wall treatment, stable settled forces, numerical sensitivity evidence and a comparison to real boat measurements. Engine/propeller recommendations and wave predictions remain unvalidated until their own steps are complete.

## R22 mesh result

Only minTetQuality changed from 1e-15 to 1e-4 relative to R18. The source STL hash remains e7fada1f1334c7998e43d9067d4e9a4d511f39464fe5ca6659a95d8d2f5554ef. The mesh build completed normally; the full checkMesh audit is authoritative.

| Metric | R18 reference | R22 control |
|---|---:|---:|
| Cells | 330,151 | 319,210 |
| Poor determinant cells | 14 | 2 |
| Concavity flags | 2,823 | 2,198 |
| Low face-interpolation weights | 0 | 2 |
| Detailed failed checks | 2 | 3 |
| Wet area with ≥1 wall layer | 80.42% | 44.05% |
| Wet area with ≥3 wall layers | 34.63% | 11.72% |

**R22 is not promoted.** It reduces some defects but loses substantial wall-layer coverage and adds a failed interpolation-weight check. No R22 flow test or performance result is claimed. R18 remains the historical flow reference, still unvalidated. Next mesh work should target local connectivity and layer termination while preserving wall resolution.

The checkpoint contains R22 inputs, generated mesh, audit/log evidence, the mass-only hydrostatic script and source geometry, and this handoff. OpenFOAM v2412 and Python dependencies are not bundled. The R22 archived case can be rebuilt with its own settings; the controlled preparation script also references the previously saved R18 source case. The loaded hydrostatic calculation does not require or infer the missing CG.

## Verification of this update

Ten automated checks pass, covering public access, database save/edit/history, migration of existing records, stale-write rejection, visible-form restoration/import, and control logic. The separate portable smoke check covers embedded assets and application behavior; neither is a real browser or device rendering test. The site remains public with no sign-in requirement. Revision snapshots are retained prospectively; the update does not erase existing observations.