Repository navigation
feat: OrcaFlex "as laid" seabed friction in statics - #17
Conversation
Seabed friction had no effect on the static solve: the friction springs are anchored to the frictionless still-water layout, so a grounded run on a slope with tan(theta) < mu, or laid with surplus length, still slid or folded (an EI = 0 line on a 5 deg slope with mu = 1 folded 24 m past its anchor, exactly as with mu = 0). The new option staticsFriction asLaid follows OrcaFlex's "As laid" policy: the line is assumed laid from its lower end along a lay azimuth (default from anchor to fairlead, layAzimuth / layAzimuth<ID>) at an as-laid tension (asLaidTension / asLaidTension<ID>, default 0), and friction resists displacement from those positions in the static solve, with the regularised law of the dynamics (isotropic and anisotropic). The dynamics continue from the static friction anchors. Both line families, the coupled aggregate and the C API share it. With asLaid, the 5 deg EI = 0 line holds where it is laid (FairTen 714.2 -> 622.9 kN, friction within mu N at every grounded node, held runs drift by 1.5e-10 of FairTen), and a 650 m finite-EI cable on a 10 deg slope, which failed, solves and holds. On a flat seabed laid at its run tension the result equals the frictionless one (FairTen within 6e-6). As an illustration, the IEA-15 mixed example with asLaid has a chain AnchTen of 368.2 kN instead of 1351.6 kN: friction holds back a run laid slack and pulled taut. The default is staticsFriction none, the previous behaviour and that of MoorDyn, which is conservative for anchor loads; every example and test deck is byte-identical under it. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
… options - A finite-EI asLaid line whose lay azimuth leaves the vertical plane of the chord (e.g. 45 layAzimuth) was solved in that plane, because the 3D requirement was set after the planar constraints were built, so a permanent lateral friction force remained (a held run drifted 1.2e-3). The requirement is now decided before the constraints; an azimuth in the plane (180) keeps the planar solve. - The finite-EI reference is laid from the lower end, as on the EI = 0 route, and only the geometric cross-plane offset of an in-plane line is zeroed, keeping the bathymetry sink offset (a cross-slope lateral anchor force of +269 N now -64 N against -59 N on the EI = 0 route). - The fallback anchor walk and the planar-run diagnosis test for a real current, not for the friction carrier of a still-water asLaid line. - asLaid is rejected by name on the multibody output and single-line routes that cannot apply it; an overflowing layAzimuth<ID> suffix names no LINE. - The Python deck reader, writer and builder accept and validate staticsFriction, layAzimuth[<ID>] and asLaidTension[<ID>]. - Tests: off-plane and flipped azimuths, cross-slope, per-line precedence, anisotropic friction, a multi-section line, mobilised friction on the finite-EI slope, input rejection; Python tests. With the default option every example and test deck is byte-identical. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f69cd75312
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
On a bathymetry whose slope varies, a positive asLaidTension turns at every node of the laid path and presses the line into a crest or lifts it from a hollow. The sink of the as-laid reference counted only the node's weight, so a line laid at its as-laid tension did not rest at its reference and friction was mobilised spuriously. The reference now adds -T (t_out - t_in).n / k to each interior node's sink, evaluated on the laid path before the sink, on the EI = 0 and the finite-EI routes. With no laid tension the reference is bit-identical. A unit test on a crest checks the correction against an independent evaluation. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1f729383bf
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
layAzimuth<ID> and asLaidTension<ID> rows were range-checked only as the final per-line value, so an invalid row (a negative tension, a NaN azimuth) followed by a valid row for the same line passed silently. The OPTIONS row check now applies the family rule (finite azimuth, non-negative tension) to every suffixed row, as for the global rows and as the Python reader does. Tests: shadowed negative tension and NaN azimuth rows are rejected. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: aa8846d862
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
The laid tension's normal load P deepens a node's normal penetration by P/k, which moves it horizontally by (P/k) grad(z)/s with s = sqrt(1 + |grad z|^2); the sink, the factor of the gradient (w/k for the weight), therefore gains P/(k s), not P/k. The correction now carries the 1/s factor, taken from the node's own normal, and the unit test and the theory page state it. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1981277808
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
A finite but huge asLaidTension stretched the laid elements without limit, so the seabed path integration of the as-laid reference could overflow its sub-step count or loop for billions of steps. A laid tension above an element's EA (more than 100 % strain) now fails closed with a named message, and a laid length beyond 1e7 m is refused, which bounds the sub-step count. Test: a 1e30 N asLaidTension is rejected. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 16314565ad
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
The static-friction label marks a paragraph, not a section title, so the bare :ref: references to it had no text and the Sphinx build (run with warnings as errors) failed on four of them. Each reference now carries its link text. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
… drag - The EI = 0 helper solve that seeds a finite-EI asLaid line did not mark its friction load as as-laid, so CD_Static_Line_Equilibrium tried the ordinary catenary/fold seeds instead of the laid-run seed; it now sets friction_as_laid. - The finite-EI static solver ignored the carrier flag drag = .FALSE.: a friction-only current object with a nonzero velocity would still have loaded the line with drag and required drag properties. The element drag and its section properties are now zero for such a carrier, as on the EI = 0 route; the still-water carriers of the as-laid policy, which have zero velocity, are unaffected. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: cc0436be0c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
… slope - A finite-EI asLaid line with ATTACHMENTS on a bathymetry had its friction reference built from the smeared weights, while the attachment polish lumps each module's net weight at one node, so the normal sink showed up as a tangential deflection that mobilised friction. The reference is rebuilt on the final mesh from the lumped nodal loads (bare weight share plus the module's net weight) before the polish: on a 5 deg cross slope the kink at the module nodes falls from 4.65 mm to 0.16 mm and the held run stays at rest. - The as-laid seed of a line whose last node lies below node 1 (lines without a Fixed End B) is now built on the reversed line from that grounded end and reversed back; it had fallen back to the suspended catenary seed. - A finite-EI asLaid line over a bathymetry that slopes across its chord is solved in 3D, as with an off-plane lay azimuth; the planar solve kept the seed's lateral positions and the held run drifted. Tests: att_cross (modules on a cross slope), case_as_laid_seed_from_last (both numberings on a flat floor and a slope give mirrored equilibria), and an at-rest march for the cross-slope cable. Default outputs are byte-identical to dev. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 5de7b1765a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
… fixes - The as-laid friction reference was laid once on the deck mesh and interpolated onto refined meshes, so on a seabed curved along the line with a section change on the grounded run the refined solve differed from the same line declared on the fine mesh (0.95 mm, an 8 N anchor tension change and a reversed lateral anchor force). The friction type now keeps the lay itself (origin, azimuth, tension, lower end, frame) and every static solve, and the dynamic hand-off, lays the reference on its own mesh with that mesh's lengths, EA, lumped weights and seabed stiffnesses (CD_HermiteCable_As_Laid_Offsets); the refined line now equals its fine-mesh twin to 1e-12 m. - The finite-EI compatibility route (a line whose End B is not Fixed) ignored staticsFriction asLaid; it now refuses it by name. - A still-water asLaid line keeps the tensile-polish fallback of a line without current, and its fixed-point and load-step failures name the friction instead of a current. - A laid path longer than the cross-slope sample cap takes the 3D solve conservatively; the sample count is computed without integer overflow. - Docs: the spring of a line laid on a slope carries the slope and turning load only (the normal sink adds no friction deflection); the none policy's anchors are described for still water and a current; the lay is from the lower end (the anchor, or the fairlead when that is lower); the asLaid rejection scope and its message are corrected. Tests: a refined two-section line against its fine-mesh twin, the compatibility-route rejection. Default outputs remain byte-identical. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
|
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Summary
Seabed friction in the static solve, following OrcaFlex's "As laid" policy, as an opt-in option. The default is unchanged.
Problem. Seabed friction had no effect on the static solve. The friction springs are anchored to the frictionless still-water layout, so a grounded run on a slope with tan θ < μ, or one laid with surplus length, still slid or folded. An
EI = 0line on a 5° slope with μ = 1 folded 24 m past its anchor, exactly as with μ = 0, and a finite-EI cable on a 10° slope failed outright.What OrcaFlex does (sources cited in
doc/theory.rst):New options.
staticsFriction asLaid|none. The default isnone: the previous behaviour, and that of MoorDyn, which is conservative for anchor loads. Every example and test deck is byte-identical under it.layAzimuth/layAzimuth<ID>: the lay direction. The default runs from anchor to fairlead.asLaidTension/asLaidTension<ID>: default 0.Design.
CD_Static_As_Laid_Referencelays the reference from the line's lower end on a flat seabed or bathymetry. On a slope each target carries the node's static penetration, so a line resting where it was laid carries no friction.EI = 0: an as-laid seed is tried first.EI = 0equilibrium.asLaidby name.Results (with
asLaid)EI = 0, 5° slope, μ = 1As an illustration of the effect, the IEA-15 mixed example with
asLaidhas a chain AnchTen of 368 kN instead of 1352 kN, because friction holds back a run laid slack and pulled taut. This is why the default staysnone.Known limit. A 700 m finite-EI cable laid past the point under its fairlead must form a hook at touchdown, which the seeds do not reach. Another
layAzimuthsolves it (documented in troubleshooting).Tests
tests/test_static_friction_as_laid.py(fails on the base):EI = 0and finite-EI) with mobilised friction;EI = 0route;pre-commit run --all-filespasses.Checklist
pre-commit run --all-filespassesctest --test-dir build --output-on-failurepassespython -m pytest tests python/testspasses (if Python or validation records changed)[Unreleased](heading added if missing) for user-visible changesdoc/updated for changed inputs, options, or outputs (options, theory, modelling, troubleshooting, driver format)feat:,fix:,docs:, ...)git commit -s) under theDeveloper Certificate of Origin