Repository navigation
fix(multibody): couple a finite-EI cable pinned to a moving object implicitly - #78
Conversation
…plicitly The monolithic multibody step stepped a finite-EI cable pinned at a Rigid6 body point or a free or pinned rod end (its other end held) once per step with the Newmark predictor's attachment motion and held its end force through the outer iterations. That explicit coupling has a step amplification above 1 once the cable's condensed dynamic end stiffness times dt^2 exceeds about 0.85 of the object's effective mass along the cable, so a light object on a stiff cable left a balanced static start: a 3 t free rod spar with a 55 m cable (EA 5e8 N, 40 elements) pinned at its upper end, dtM 0.01 s, had the step change of the cable tension alternate in sign and grow 2.3 times per step; the tension rose from 11 kN past 1 MN in 0.7 s and the rod end moved 6.3 m in 2 s. The static pose is a stable equilibrium of the march: the same deck with an EI = 0 cable, or with the cable clamped (already coupled implicitly), stays at rest to 1e-7 m, and the staggered scheme holds it too. Every cable with an end, pinned or connected, on an object the march moves is now re-stepped at every outer iteration with its end stiffness in the Newton matrix, as clamped cables and cables between two objects already were. New CTest pinned_cable_light_object (on dev e760bc6, 14 of its gates fail): - free rod spar, cable pinned at the upper end, in a 0.5 m/s current, and at the lower end: rod end moves 6.3, 5.4, 6.5 m and tension changes of 161, 472, 5596 times the start before; now 1e-10, 6.3e-4 and 1.8e-10 m and 2.7e-8, 1.2e-4 and 3.4e-8 (the current case swings slowly about its start, as on the staggered scheme); - the cable from a fixed rod's end to the free rod (End B on the object): 6.1 m before, now 2e-7 m; written from its anchor: the same output; - a 4 t Rigid6 buoy with a pinned keel cable: 0.22 m and 38 times before, now 1e-9 m and 3.3e-8; - monolithic vs staggered schemes: 6.0 and 6.5 m apart before, now 1e-10 and 1.7e-10 m; - controls (unchanged): EI = 0 cable at rest to 1e-10 m, clamped cable at the upper and lower rod ends to 1e-7 m. line_attachments 11 and 12 (the spar rod with the pinned cable, with and without a 1500 kg clump) march at dtM 0.01 s again, over 1 s: the rod holds to 1e-7 and 0 m, leg tensions to 2.9e-7 and 3.8e-8 (on dev at dtM 0.01 s: 0.82 and 1.3 m, 13 and 4.4). Over 3 s the clump deck holds to 1.1e-6 m (15 m on dev). Examples: 52 of 53 identical; mixed_body_rods_points (150 t buoy with a pinned cable) changes by 3.5e-5 m in body motion and 0.6 N of 8.1 kN in cable top tension, its body error against dtM / 4 falling from 3.7e-5 to 1.6e-6 m; it runs about 45 % longer (the cable is stepped once per outer iteration). Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
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. |
|
Codex Review: Didn't find any major issues. Breezy! 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". |
A finite-EI cable written from a seabed termination up to a floater (End A on the bed below End B, e.g. a fixed rod's end on the seabed to a free spar rod) was refused by the catenary seed (the fairlead, End A, must sit above the bed), so the mesh-sequenced route alone solved it, and its branch depended on platform round-off: the pinned_cable_light_object deck pcl_fixrod solved on Windows and reached a kinked state (max h kappa 1.96) on the Linux gfortran Debug CI job, which failed the deck. The seed is now laid from End A, the same catenary seen from its other end, with its nodes put back in order; no anchor walk follows a reversed seed. The deck starts at the same equilibrium as before on Windows, and new gate R moves the fixed rod by 1 ulp either way in x or 1e-12 m up: the rod's start is unchanged (0 m). pcl_fixrod holds to 2.0e-7 m and 4.6e-8 in tension. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
|
Codex Review: Didn't find any major issues. Keep them coming! 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". |
What changed
The monolithic multibody step (
bodyScheme monolithic, the default) coupled a finite-EI cable pinned to a moving object (a Rigid6Body<N>point or a free or pinnedRod<N>end, the other end held) at the Newmark predictor: the cable was stepped once per step with the predicted attachment motion and its end force was held through the outer iterations. That is an explicit, partitioned coupling. Its step amplification exceeds 1 onceK_e dt^2 / m_eis above about 0.85 (1.5 at 1.3, 2.5 at 2), withK_ethe cable's condensed dynamic end stiffness andm_ethe object's effective mass along the cable. A light object on a stiff cable therefore left a balanced static start.A 3 t free rod spar with a 55 m cable (
EA5e8 N, 40 elements) pinned at its upper end, atdtM0.01 s:The start was not at fault. It is a stable equilibrium of the march:
EI = 0cable stays at rest to 1e-10 m;Rigid, already coupled implicitly) it stays at rest to 1e-7 m;The fix. In
src/CableDyn_DeckDriver.f90, every cable with an end on an object the march moves now setscab_imp. This covers pinned or connected ends, End A or End B, bodies and free or pinned rods. Such a cable is re-stepped at every outer iteration with its end stiffness in the Newton matrix, as clamped cables and cables between two objects already were. A cable with no moving end is still stepped once. The staggered scheme and the OpenFAST and C-API routes are unchanged.Docs.
doc/driver_format.md: the multibody route bullet.doc/theory.rst: the monolithic scheme paragraph, now with the stability criterion of the old coupling, and the clamped-cable paragraph.CHANGELOG.md: an[Unreleased]Fixed entry. The feat(cable): finite-EI lines with two moving ends, between two floaters or two objects, TURBINES rows turning both end frames #75 Added entry no longer says a pinned cable is "coupled at the predictor".Gates
New CTest
pinned_cable_light_object. On dev e760bc6, 14 of its gates fail.EI = 0cable; cable clamped at upper / lower end)Gate tolerances:
line_attachments11 and 12 had to march atdtM0.002 s on #77 because of this coupling. They are back atdtM0.01 s, over 1 s instead of 0.3 s:dtM0.01 sOver 3 s the clump deck holds to 1.1e-6 m; dev moves it 15 m.
Example A/B (reference driver built at e760bc6)
52 of 53 example decks are identical.
mixed_body_rods_pointschanges; it is a 150 t buoy with a pinned power cable, the one example with a pinned cable on a moving object:Rod2Rychanges by 2.3e-4 deg;FairTen8changes by 0.6 N of 8.1 kN.The new run is the more accurate one. Against the same deck at
dtM/ 4, the body surge error falls from 3.7e-5 m to 1.6e-6 m and the pitch error from 1.1e-4 to 2.2e-5 deg. The cable tension error, 17 N, comes from the cable's own time step and does not change.The deck runs about 45 % longer (6.8 s to 9.8 s), because the cable is now stepped once per outer iteration instead of once per step.
Checks
-fcheck=all,--timeout 5400):pinned_cable_light_object,line_attachments,static_dry_weight,cable_between_bodies,monolithic_gates,clamped_cable_ends,body_rod_physics,bodies_rodsanddriver_clamped_cable_pinned_rodpass, 9 of 9.python/testsandtests: with the native parity tests on (CABLEDYN_TEST_DRIVER), 2179 passed, 75 skipped.-Ware clean.dup_public: 0 duplicates.Notes
EI = 0cable in the same current holds to 1e-10 m. So it is a small static-versus-dynamic difference of the finite-EI cable in a current, not part of this coupling, and is left for the open list.K_e dt^2 / m_eis safely below 0.85. That needs estimates ofK_eandm_e; this change takes the uniformly implicit coupling for robustness.