NEMOH v3 full-QTF vs WAMIT for IEA-15-MW UMaineSemi: good mean drift but severe off-diagonal difference/sum QTF disagreement

Wind & Water Computer-Aided Engineering Software Tools
1

Hello,

I am validating second-order hydrodynamic coefficients generated with NEMOH v3.0.3-2 for the IEA-15-MW UMaineSemi platform before using them in OpenFAST/HydroDyn, and I would appreciate guidance on a persistent discrepancy between NEMOH full-QTF results and the distributed/reference WAMIT QTF data.

The first-order NEMOH-to-WAMIT validation is satisfactory. After accounting for the NEMOH/WAMIT excitation phase convention, added mass, radiation damping and first-order excitation agree reasonably well.

The problem appears specifically in the second-order QTF calculation.

For a controlled reference-depth case I used:

  • IEA-15-MW UMaineSemi
  • water depth = 200 m
  • β = 0°
  • 46 frequencies, ω = 0.25:0.05:2.50 rad/s
  • difference- and sum-frequency QTFs
  • all 6 DOFs
  • NEMOH contrib = 3
  • DUOK = 1
  • HASBO = 1
  • HASFS+ASYMP = 1
  • explicit free-surface mesh
  • comparison against the corresponding WAMIT .12d/.12s data.

The NEMOH QTF components were converted to the WAMIT convention without applying any empirical scaling. I also tested the complex-conjugation convention; conjugation clearly gives better phase agreement than using the raw NEMOH complex values, so a simple phase-convention error does not appear to explain the remaining discrepancy.

The interesting result is that the diagonal difference-frequency / mean-drift terms are relatively close to WAMIT, particularly in the original mesh calculation. Approximate median magnitude errors for surge/heave/pitch were:

  • surge: 13.4%
  • heave: 16.8%
  • pitch: 8.8%

However, the off-diagonal difference-frequency QTFs are very different. Median magnitude errors for surge/heave/pitch were approximately 59/73/77%, and the complex correlation of the total NEMOH QTF against WAMIT was essentially zero (|CC| <= ~0.04). The resonance structure visible in the WAMIT off-diagonal QTF is not reproduced by NEMOH.

The sum-frequency comparison is also poor, with approximately 90–100% median magnitude errors and ~90–125° median phase differences.

Another suspicious feature is the β=0 result for the nominally symmetric sway/roll/yaw components. WAMIT values are approximately 1e-5–1e-3, while the NEMOH full-QTF ASYMP contribution produces values approximately:

  • sway: ~3
  • roll: ~37
  • yaw: ~122

i.e. several orders of magnitude larger.

I investigated whether the body’s internal/free-surface lid panels were responsible. The original NEMOH mesh contained 228 flat z=0 lid panels. I therefore performed one controlled no-lid experiment:

  • exactly 228 lid panels removed;
  • wetted geometry otherwise unchanged;
  • 2766 → 2538 panels;
  • first-order NEMOH solution regenerated consistently for the no-lid mesh;
  • identical depth, frequencies, heading, free-surface mesh and QTF settings;
  • full QTF chain run once without tuning.

The no-lid experiment completed normally but did not improve the comparison. Therefore, removing the lid did not recover the WAMIT off-diagonal structure and actually degraded part of the diagonal mean-drift agreement.

My questions are:

  1. Is full off-diagonal QTF agreement between NEMOH v3 and the WAMIT QTFs supplied for the IEA-15-MW UMaineSemi expected for this configuration?
  2. Are there known differences in the second-order formulation, free-surface treatment, irregular-frequency treatment, ASYMP contribution, or body/free-surface meshing requirements that could explain why diagonal mean drift can be reasonably close while off-diagonal difference- and sum-frequency QTFs are substantially different?
  3. For β=0 and a symmetric platform, should the NEMOH ASYMP sway/roll/yaw QTF components be close to zero? The several-orders-of-magnitude difference relative to WAMIT is particularly concerning.
  4. Is there a recommended NEMOH configuration/example that is considered appropriate for generating .12d/.12s data for use with OpenFAST HydroDyn, especially when MnDrift = 2 and/or second-order potential-flow loads are required?
  5. Given satisfactory first-order validation but unresolved full-QTF validation, would you recommend not using the NEMOH-generated full QTFs in HydroDyn until this discrepancy is understood? Would a first-order-only or separately validated mean-drift approach be the more defensible interim choice?

I have preserved the original and no-lid NEMOH cases, solver logs, QTF component files (DUOK, HASBO, HASFS, ASYMP), converted .12d/.12s files, and detailed NEMOH-vs-WAMIT comparison tables, and I can provide any of these if useful.

Thank you for any guidance on whether this behavior is expected, a configuration issue, or indicative of a limitation/incompatibility between the NEMOH and WAMIT second-order formulations.