Repository navigation
fix(output): report the section under a bend stiffener - #89
Conversation
The node outputs divided the section moment by the element stiffness on the side giving the larger curvature. Under a taper that stiffness is one value per element, so it steps at every node although the section does not: the softer element always won, and the curvature and the cable's moment were high by about h |EI'| / (2 EI), first order in the element length. The line now keeps EI(s) at each node from the stiffener rows, one limit per side (CD_Stiffener_Node_EI, on the final mesh), and node_values reads those. The limits differ only at a real step of the section: the bare tip, rows that do not match and two line types, where the larger-side rule still decides. No threshold on the stiffness ratio is needed. Curv, CurvX, CurvY, BendMom, StfMom, the static profile and the range graphs all take the values from node_values; a line without a stiffener is unchanged. driver_bend_stiffener gains a local gate over a sweep of element sizes, the identities between the channels and the real steps. The error on 0.1 m elements falls from 3.3 % to 0.24 % of the local curvature. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
… outputs The peaks of .elements.out and the node values of OutputRecovery element read the element's interpolation. On an element with a sleeve that interpolation bends under one stiffness, so its curvature is no section's: it differs from the section's by h |EI'| / (2 EI), high where the curvature falls toward the tip and low where it rises, and EI_e times it is the linear fit of the moment, low at the element ends by h^2 M''/12. An element with a sleeve now reports the section. Its peak is the largest |M(u)|/EI(u) along it, with M(u) the quadratic through the recovered moments of the end nodes that keeps the mean of the fit, and a row end or tip inside the element kept as a step of EI(u). OutputRecovery element takes the recovered moment over EI(s) on such an element, with the exact trace on every other element. bend_check and element_range_graph read the table and follow. The geometric traces stay available as CD_HermiteCable_Dyn_Curvature_Exact and in the mesh diagnostics. A line without a stiffener is unchanged bit for bit. driver_bend_stiffener gates both outputs under the taper over the same element sizes: on 0.1 m elements the error falls from 3.1 % to 0.24 % of the local curvature. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
Codex Review Summary: ✅ CompletedThis 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: 72f2908d1f
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 peak of an element under a bend stiffener divided the recovered moment by EI(s) everywhere. That moment carries an error of order w L0^2/12 that does not depend on EI, so on a sleeve below the stiffness that resolves it the peak was noise: 24 times the curvature of the element's nodes on a 100 m soft sleeve of the coarse lazy wave. The nodes already fall back on the interpolation's curvature there, and the peak now does the same, point by point, with the same check on EI(s) and the same EI_e/EI(s) scale. The peak path also takes the tangent from the element's midpoint, which the curvature call has checked, in place of the chord, which vanishes on a folded element, and stops on moments that are not finite instead of reporting a zero peak. Signed-off-by: Jae Hoon Seo <jaehoon.seo@inha.ac.kr>
|
@codex review |
|
Codex Review: Didn't find any major issues. Hooray! 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
Under a tapered bend stiffener the bending stiffness is one value per element, so every node sits on a small stiffness step that the section does not have. The node outputs took the softer side, which overstated the curvature by about
h |EI'| / (2 EI): first order in element length, 3.3 % at 0.1 m elements. The element outputs carried a bias of the same size, of either sign depending on the deck.EI(s)at every node, one limit per side. The larger-side rule compares those limits, so a real step (the bare tip, a step between stiffener rows, a junction of line types) stays a step, and no tolerance on the stiffness ratio is needed.Curv,CurvX,CurvY,BendMom,StfMom,.static.outand the range graphs follow..elements.outpeaks, andOutputRecovery elementon sleeved elements, report the section moment overEI(s).Result changes
lazy_wave_bend_stiffener.dat, 0.152 m elements, static: nodeCurvatureandBendMomentchange at nodes inside the stiffener by -0.6 to -4.4 %;.elements.outpeaks by -1.1 to +4.4 %. Positions, tensions and everything outside the sleeve are bit-identical.Checks
Local error under the taper against the planar elastica, floored at 0.1 % of the root curvature:
.elements.outpeaks afterdriver_bend_stiffenergains this gate, swept over element size; it fails without the change.OPENBLAS_CORETYPE=Prescott; pytest passes; no new compiler warnings.The CHANGELOG entry lands with the release batch.