---
license: cc-by-4.0
pretty_name: digitalmodel Diffraction RAOs (multi-vessel, de-identified)
tags:
- analysis-results
configs:
- config_name: hydrostatics
  data_files: hydrostatics.parquet
- config_name: rao
  data_files: rao.parquet
---

# digitalmodel Diffraction RAOs (multi-vessel, de-identified)

Analysis/computation results published for query + visualization via the Hugging Face
datasets-server API. One row per (vessel, solver, damping, dof, frequency, heading).
Vessels are referenced only by de-identified vessel_db IDs.

## Tables

| config | rows | cols |
|---|---|---|
| `hydrostatics` | 5 | 7 |
| `rao` | 8448 | 11 |

## Schema 1.1.0 — `solver` column

The `rao` table now carries a `solver` column (`AQWA` \| `OrcaWave`), added when the first
vessel gained results from two solvers. Filter on `vessel_id` **and** `solver` (and
`damping_zeta_pct`); rows published under schema 1.0.0 were backfilled from the vessel
table below at the 1.1.0 migration and are otherwise byte-identical.

## Series in `rao`

| vessel_id | solver | damping_zeta_pct | rows | basis |
|---|---|---|---|---|
| `install_bokalift_2` | OrcaWave | 5, 10 | 600 | parametric hull from public particulars |
| `drill_samsung96k_1` | AQWA | 0 | 2700 | hull-class-level de-identified model, survival loading condition |
| `drill_samsung96k_1` | OrcaWave | 0 | 2574 | same hull class/condition; radiation-only (no external damping), open moonpool (no lid) |
| `drill_samsung96k_1` | OrcaWave | 5 | 2574 | as ζ=0 plus external linear roll damping B44 (total roll ζ ≈ 5 %) |

## Data-quality notes

**Screening-grade — faithful-to-solver ≠ correct for design.** No model here is class-approved.

### install_bokalift_2 (OrcaWave, parametric screening)

- **Low-frequency artifact (flagged, not omitted):** in `rao`, heave & sway at freq ≤ 0.35 rad/s (T ≥ 18 s) in non-beam headings are physically implausible (heave ≈ 7–9 m/m where beam seas correctly show ≈ 1.1). 40 rows carry `reliability = "low_freq_artifact"`. Use freq ≥ 0.5 rad/s. Tracked: digitalmodel#1548.
- **Withheld table — `added_mass_damping`:** the source matrices declare units kg / kg·m² but magnitudes are consistent with te / te·m² (~1000× off, e.g. heave added-mass 58,447 vs ~1e8 kg expected). Withheld pending unit reconciliation. Tracked: digitalmodel#1550.
- **Withheld — operation-envelope (Hs/Tp) limits:** derived limits reach implausible values (~17 m Hs heavy-lift) and would mislead as operational limits. A class panel mesh is required first.
- **Roll at resonance is high** (23.5 deg/m at ζ=5 %, 19.2 at ζ=10 %) — a real, expected consequence of light damping, not an artifact.

### drill_samsung96k_1 · AQWA (ζ = 0)

- **`damping_zeta_pct = 0` means NO external/viscous roll damping** — radiation damping only.
  The roll peak (15.6 deg/m at T = 19 s, beam seas) is therefore an undamped potential-flow
  upper bound, not an operability prediction; apply an external roll damping model before use.
- **Resonance band flagged (50 rows `reliability = "undamped_resonance"`):** roll rows at
  T = 18.48 and 18.98 s (all headings) are damping-controlled, matching the flag treatment of
  the OrcaWave ζ=0 series. Coupled sway/yaw are NOT flagged here — the AQWA grid straddles the
  resonance, so their contamination is negligible (sway ≤ 1.14 m/m, yaw ≤ 0.064 deg/m).
- **Frequency band is roll-focused:** 18 frequencies covering T ≈ 15.5–31.4 s (refined around
  the roll-response band). No rows outside this band — absence is by scope, not omission.
- **Moonpool:** the panel model includes moonpool walls with no damping lid; moonpool-resonance
  effects near the moonpool natural period are expected and not suppressed.
- Long-period sanity passed: beam-seas heave → 1.0 m/m and head-seas surge → 0.95 m/m at T = 31.4 s.
- Headings span −180°…180° at 15° (25 values).

### drill_samsung96k_1 · OrcaWave (ζ = 0 and ζ = 5)

- **Grid:** 33 frequencies (T ≈ 4–31.4 s, refined 10–22 s at 0.5 s) × 13 headings (0°–180° at
  15°; port/starboard symmetry assumed) × 6 DOFs = 2,574 rows per series. Phases in the
  **Orcina lag** convention — align conventions before phase-level comparison with AQWA.
- **ζ = 0 series — resonance band flagged (117 rows `reliability = "undamped_resonance"`):**
  radiation-only roll damping for this hull is ζ ≈ 0.2 % critical, so the roll RAO at
  resonance (358 deg/m at T = 18.5 s, beam seas) and the coupled sway/yaw peaks are
  **damping-controlled numerical upper bounds, not physical response predictions**. Roll,
  sway and yaw rows with T = 17.9–19.1 s carry the flag. Off-resonance rows are usable
  screening values. The open moonpool (no damping lid) is also unsuppressed in this run.
- **ζ = 5 series:** adds external linear roll damping B44 equivalent to ≈ 5 % total critical;
  beam-seas roll peak drops to 4.48 deg/m at T = 18.5 s (~80× below the ζ=0 value) — use this
  series as the damping-sensitivity anchor; match your own damping budget before design use.
- **Cross-solver agreement (same hull, both ζ=0):** off-resonance beam-seas roll RAOs agree
  with the AQWA series within ~12–18 % — consistent for undamped potential-flow codes on
  different grids. Resonance-peak *magnitudes* are NOT comparable between the two ζ=0 series:
  the AQWA grid straddles the peak while the OrcaWave grid samples it exactly (grid sampling,
  not physics). Roll natural period agrees at 18.5–19.0 s in both codes.
- **Withheld — added mass & damping matrices:** present in the source results but not yet
  published pending the same unit reconciliation as digitalmodel#1550.

## Provenance

- Source repo: `digitalmodel` · solvers: `orcawave-diffraction@1537` (install_bokalift_2),
  AQWA v261 licensed-run lane (drill_samsung96k_1 AQWA, run `lr_acma_bb24945c4b8b`),
  OrcaWave 11.x licensed-run lane (drill_samsung96k_1 OrcaWave: ζ=0 run `lr_acma_2a4a51f24387`,
  ζ=5 run `lr_acma_36fb1ffbc5e0`)
- Schema version: `1.1.0` (adds `solver`; 1.0.0 rows backfilled per the series table)
- Source snapshots (sha256):

- `hydrostatics.parquet` — sha256 `cb0141548b6bf07f3892560f4a564147bb76648fcc7ea1c3383577ff62357f53`
  (adds 3 drill_samsung96k_1 rows: class-level analysis mass for the OrcaWave series; the AQWA
  row's `mass_te` is null — genuinely absent from its de-identified source. CoG/CoB/stiffness
  remain withheld: GM-recoverable.)
- `rao.parquet` — sha256 `d855c015efb7bd13e05429da2e54613cab213421fde4f4eb57c967b53f5a5347`
  (2026-07-16b: AQWA resonance-band flags added; values byte-identical to the prior revision)
- drill_samsung96k_1 AQWA de-identified source JSON — sha256 `ef6e7d90b52b4bf730c94e7aebcc1b295008ccb9b9c80ad3e0ef1a5d26c05dec`
- drill_samsung96k_1 OrcaWave ζ=0 de-identified source JSON — sha256 `af4b37cb0acde7d7f68b7c2f60ac2cc34b8928caeb5e7aa3e06ade13ad81a081`
- drill_samsung96k_1 OrcaWave ζ=5 de-identified source JSON — sha256 `36acd2d4943a97a44d2b31fa7bebda7623ef56b6bc0d9bb304af6b3f05bf5fa5`

Vessels are de-identified: `install_bokalift_2` from public particulars; `drill_samsung96k_1`
identified only at hull-class level. Not a run-identity ledger (that is the separate
`aceengineer/*-runs` contract).

Nulls are genuine missing values (non-finite floats were coerced to null). Load a table:
`pandas.read_parquet(hf_hub_download("aceengineer/digitalmodel-diffraction-rao", "<config>.parquet", repo_type="dataset"))`.

Live visualizations of this dataset: https://www.aceengineer.com/reports/diffraction/