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
32 changes: 32 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -451,6 +451,38 @@ All notable changes to CableDyn are recorded in this file. The format follows
the IEA-15 MW mixed example in a 60 s platform motion changes by 1.6e-6 of its cable tension.
Cost: a coupled cable step takes about 2.0 tangent factorisations instead of 1.1, and the
aggregate step of that example (chains included) 38 % more CPU time (the D03 deck: 19 %).
- A held line under strong torque-tension coupling stays still in its dynamics. With
`K >= 0.85 sqrt(EA GJ)` (and at `-0.8` with one end free in torsion) a line in exact static
equilibrium failed its step within 0.05-0.9 s ("Newton did not converge, ||G|| = 2.4e-8 >
tol = 1e-8"): the round-off floor of the step's convergence norm took only the kernel's
`EA - K^2/GJ`, not the coupling's share of the axial force and of the torque. It now takes
both; at `K` up to `0.99 sqrt(EA GJ)` in either sense the held line holds at the default `dtM`
and tolerance (`torsion_coupling_still`), and lines with `|K| <= 0.8 sqrt(EA GJ)` step as before.
- The static passes of a finite-EI cable restrained in torsion between two Rigid6 bodies
converge at every pretwist. They compared the change of the cable's end loads with 1e-10 of
their scale, below the resolution of the core's body balance (1e-9 of the bodies' net weight),
so at 360 deg (Rigid ends) and 300 deg (Pinned) they stalled near 2e-9 and the deck stopped,
with advice (`bodyIC deck`) that was not the remedy. They now stop at the resolution the core
leaves: the Rigid 360 deg case converges in 6 passes (torque 4886.24 N m), and the
pretwist sweep 120-720 deg of both end kinds converges (`torsion_body_passes`).
- A lazy wave restrained in torsion starts in an in-plane current. In a 0.5 m/s current its
untwisted static state, solved in its vertical plane, is unstable out of that plane on its mesh
(one negative stiffness mode), which the deck without torsion accepts with a note; the torsion
stage refused the twisted state as unstable although the twist added no negative mode ("torsion
continuation stalled at Phi = 0 (target 0)" at pretwist 0, an unstable equilibrium after a
failed Newton solve at 360 deg). A twisted state with no more negative modes than its untwisted
one is now kept, and the note names the instability as the untwisted state's
(`torsion_current_start`). Its start also depended on the pretwist and on the round-off of the
BLAS kernel: the coarse branch-selection build of a line restrained in torsion twisted it, and
that twisted state seeded the untwisted solve on the deck mesh (pretwist 180 deg stopped in the
attachment polish, 360 deg needed the polish's fallbacks and stopped with
`OPENBLAS_CORETYPE=Prescott`); and each pass of the attachment fixed point after the first
applied the drag of the last twisted shape in full at the untwisted start (270 deg and more
stopped before any twist). The static helper builds now solve the untwisted line, and the
change of the attachment loads is ramped with the imposed twist: pretwists from 0 to 720 deg
start with the same torques on every kernel, proportional to the pretwist.
- The "unresolved/kinked equilibrium" refusal names the element and the factor to refine its
section's `NumSegs` by.
- A rod numbered as a body no longer joins the body's attachment points. A `Rod<N>A/B` end point
keeps its rod's id where a `Body<N>` point keeps its body's, and the body's point list took
every point with its id: a rod end POINT row listed before the body's points became the body's
Expand Down
25 changes: 25 additions & 0 deletions CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -1803,6 +1803,18 @@ if(Python3_Interpreter_FOUND)

# A rod numbered as a body is not one of the body's points: a Rod1B end POINT listed before
# the Body1 points gives the run of the same deck with the rod renumbered, bit for bit.
# A lazy wave restrained in torsion starts in an in-plane current (its untwisted state's
# out-of-plane instability is kept, not refused by the torsion stage).
add_test(NAME torsion_current_start
COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/tests/test_torsion_current_start.py
--exe $<TARGET_FILE:cabledyn>
--data ${CMAKE_CURRENT_SOURCE_DIR}/tests/data
--output ${CMAKE_BINARY_DIR}/torsion_current_start)
set_tests_properties(torsion_current_start PROPERTIES
LABELS "python;driver;integration;static"
TIMEOUT 900
ENVIRONMENT "${_cabledyn_python_test_environment}")

add_test(NAME body_rod_numbering
COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/tests/test_body_rod_numbering.py
--exe $<TARGET_FILE:cabledyn>
Expand Down Expand Up @@ -2321,6 +2333,12 @@ add_executable(test_cable_between_bodies tests/test_cable_between_bodies.f90)
target_link_libraries(test_cable_between_bodies PRIVATE cabledyn_core)
add_test(NAME cable_between_bodies COMMAND test_cable_between_bodies)
set_tests_properties(cable_between_bodies PROPERTIES LABELS "fortran;driver;integration;dynamic;static")
# The static passes of a cable restrained in torsion between two moored bodies converge at every
# pretwist (they stop at the resolution of the core's balance): the mirror-symmetric start, held.
add_executable(test_torsion_body_passes tests/test_torsion_body_passes.f90)
target_link_libraries(test_torsion_body_passes PRIVATE cabledyn_core)
add_test(NAME torsion_body_passes COMMAND test_torsion_body_passes)
set_tests_properties(torsion_body_passes PROPERTIES LABELS "fortran;driver;integration;dynamic;static")

add_executable(test_pinned_cable_light_object tests/test_pinned_cable_light_object.f90)
target_link_libraries(test_pinned_cable_light_object PRIVATE cabledyn_core)
Expand Down Expand Up @@ -2355,6 +2373,13 @@ add_executable(test_torsion_coupling tests/test_torsion_coupling.f90)
target_link_libraries(test_torsion_coupling PRIVATE cabledyn_core)
add_test(NAME torsion_coupling COMMAND test_torsion_coupling)
set_tests_properties(torsion_coupling PROPERTIES LABELS "fortran;unit;driver;integration;static")
# A held line under strong torque-tension coupling (K up to 0.99 sqrt(EA GJ), either sense, one
# end free in torsion) stays still at the default dtM and dynamic tolerance: the round-off floor of
# the step's convergence norm carries the coupling's share of the axial force and of the torque.
add_executable(test_torsion_coupling_still tests/test_torsion_coupling_still.f90)
target_link_libraries(test_torsion_coupling_still PRIVATE cabledyn_core)
add_test(NAME torsion_coupling_still COMMAND test_torsion_coupling_still)
set_tests_properties(torsion_coupling_still PROPERTIES LABELS "fortran;driver;integration;dynamic")

# Torsion and prescribed motion on the standalone mixed EI = 0 + finite-EI route (the driver
# executable marches the aggregate): route equivalence with the cubic-Hermite route under vessel
Expand Down
30 changes: 23 additions & 7 deletions doc/troubleshooting.rst
Original file line number Diff line number Diff line change
Expand Up @@ -281,11 +281,17 @@ Model data
supported in a deck marched as one multibody system. End the cable on a Fixed point, a
body or a rod end, or run the cable and the lines it joins as separate decks.
* - ``static body/rod equilibrium with the bending end loads of finite-EI LINE <L> (its End A|B,
on a body or rod) did not converge after 60 passes``
on a body or rod) did not converge after 60 passes (relative end-load change ...,
tolerance ...)``
- the static passes that balance a finite-EI line between two objects on its cubic-Hermite
end loads (Anderson-accelerated) did not settle; the line's bending couples the objects
more strongly than their own restoring holds them. Check the connection directions and
the mooring, or start from the deck pose (``deck bodyIC``).
more strongly than their own restoring holds them. The passes stop at a change of 1e-7 of
the end-load scale (1e-10 with torsion), or at the resolution each pass's balance of the
objects leaves (four times its residual load) where that is larger, so a change above the
reported tolerance is not round-off. Restrain the object more stiffly (its moorings or
``C44``/``C55``) or soften the connection (a finite ``Stiffness``, a ``Pinned`` end or a
finite ``TorsStiffness``); ``OPTION bodyIC deck`` skips the balance and starts the objects
out of equilibrium at their deck poses.
* - ``finite-EI LINE <L>: a Coupled/Vessel End B (POINT <n>) would need motionFile coupling to
drive it``
- the multibody march takes no ``motionFile``, so it has nothing to drive a ``Coupled``/
Expand Down Expand Up @@ -546,9 +552,14 @@ but a few geometries are genuinely hard:
- **A residual-converged folded branch.** CableDyn checks the element rotation measure after every
finite-EI sequence attempt, every final deck-built equilibrium, and the public automatic-mesh
route. ``max(h*kappa) > 0.7`` is rejected by name even if Newton's residual is small. This is a
mesh/branch safety failure, not permission to relax the tolerance. Increase resolution in the
reported high-curvature section and inspect the static profile. The check samples element
interiors adaptively, so a cubic-element fold cannot hide between output nodes.
mesh/branch safety failure, not permission to relax the tolerance. The refusal names the
element and the factor to refine ``NumSegs`` by there (``refine NumSegs at least <n>x about
element ...``): on a resolved bend the curvature does not change with the mesh, so ``h*kappa``
falls with the element length and ``n`` is ``max(h*kappa)/0.7`` rounded up, times the
refinement the solve already applied to the deck element. Refine the reported section by it
and inspect the static profile; a fold that stays kinked on the finer mesh is a non-physical
branch, not a resolution limit. The check samples element interiors adaptively, so a
cubic-element fold cannot hide between output nodes.
- **Odd, prime, or section-wise nonuniform meshes.** These are supported; do not change
``NumSegs`` merely to make the total divisible by two. The nested hierarchy retains physical
section interfaces and interpolates at the true cumulative rest-length coordinate. Its
Expand Down Expand Up @@ -595,7 +606,12 @@ but a few geometries are genuinely hard:
equilibrium on the path, typically because a loop is forming (hockling), which needs a
dynamic analysis and is not resolved as self-contact. ``finite-EI torsion static solve ended on
an unstable equilibrium`` means the descent from a buckled (unstable) twisted state did not
find a stable one. ``the twist moved more than pi/2 from its committed value in one step``
find a stable one. The twist stages refuse, or descend from, only an instability the twist
adds: a twisted state with no more negative stiffness modes than its untwisted state (a
planar state that is unstable out of its plane, which a line without torsion starts from with
a note) starts with the note ``the twisted static state keeps the instability of its untwisted
state``. Where such a state cannot be kept, ``torsion stage failed: the untwisted equilibrium
it starts from is itself unstable`` names the untwisted state, not the twist, as the cause. ``the twist moved more than pi/2 from its committed value in one step``
stops a dynamic step that would change the twist by more than a quarter turn (the unwrapping
of the twist cannot be trusted beyond it): reduce ``dtM`` or slow the imposed roll.
``tangent turns by more than 120 degrees`` names an element or node pair where the twist of
Expand Down
Loading
Loading