Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 17 additions & 6 deletions VALIDATION.md
Original file line number Diff line number Diff line change
Expand Up @@ -743,17 +743,28 @@ cannot restart the buoyancy ramp. AutoMesh warm-start refinement is tested by
buoyant fraction. Each is solved by the production mesh-sequenced solver from a
library-built trivial seed (`CD_HermiteCable_Trivial_Seed`: endpoints, rest
lengths, and seabed in; isometric seed out; no catenary, seed files, or per-case
tuning). Result: **117 pass, 3 known failures, 2 cases annotated as known failures
that now pass, 0 unexpected failures, and 0 silent wrong answers**. Known-failure
annotations are checked both ways: an annotated case must either fail cleanly or be
reported when it passes.
tuning). Result: **118 pass, 4 known failures, 0 unexpected failures, and 0 silent
wrong answers**, the same on every OpenBLAS kernel. Known-failure annotations are checked
both ways: an annotated case must either fail cleanly or be reported when it passes.

The deck route solves all 122 cases but the one beyond the L-shape bound that it refuses by
name: a deck seeds a line with a buoyant section as a lazy wave, starts it from the
`EI = 0` catenary and refines a kinked state locally, never this kernel from the trivial seed.
Run through the standalone driver on both `cable_statics` routes, the 121 equilibria agree
between the routes to 1.3e-7 in tension and 1e-5 m in position, with `max(h*kappa)` at most
0.56. The four known failures of the kernel and the former fifth are cases of the
`cable_statics_matrix` test (entries `SIS_*`, both routes): where the arch stays submerged
their fairlead tension is within 2.4e-4 of an independent elastic multi-segment `EI = 0`
catenary; two float their arch at the surface.

- On a frictionless bed, a smooth planar equilibrium exists only while the rest
length is below the L-shape bound `L < h_fairlead + x_anchor`; beyond it the
surplus must fold on the bed. The slack axis runs to 0.97× the bound. A separate
family crosses it and must return either a clean error or a finite
bounded-curvature solution, never NaN or a silent cusp. The three known failures
are hard contact cases at the lowest EI of the sweep (5e2 N·m²).
bounded-curvature solution, never NaN or a silent cusp. The four known failures
are lazy waves at the lowest EI of the sweep (5e2 N·m²): from the trivial seed the
kernel reaches a residual-converged kink of a few elements in mid-water, which the
check below rejects.
- A residual-converged state is not sufficient: the solver rejects an
element-localised branch when `max(h*kappa) > 0.7`, so a folded equilibrium
cannot enter dynamics.
Expand Down
Loading
Loading