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
70 changes: 68 additions & 2 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -123,8 +123,12 @@ All notable changes to CableDyn are recorded in this file. The format follows
velocities, with `A`, `B`, `C` and `D` by central differences of the accelerations solved at the
perturbed state. On the UMaine chains the stiffness `D - C A^-1 B` matches MoorDyn-F's to 0.1 %
and the lowest nine line-mode frequencies to 0.3 %; with a lazy-wave power cable every coupled
point matches the deck's own static solutions to 3e-4, also with a hang-off spring. A cable
end connection with a numeric stiffness now linearises. The viscoelastic slow-branch stretch
point matches the deck's own static solutions to 3e-4, also with a hang-off spring. The
reduction is the static stiffness at an operating point whose coupled points do not
accelerate: the coupled load includes the line's inertia, so at the accelerating `t = 0`
platform the lazy wave's force per unit rotation reads 1.2 % off, and 2.7e-4 with the platform
DOFs disabled (`doc/openfast.rst`). A cable end connection with a numeric stiffness now
linearises. The viscoelastic slow-branch stretch
and the Syrope slow strain are first-order states, so the reduction gives these ropes' relaxed
static stiffness (each matches its deck's static solutions to 7e-5; held, they would read the
chain 0.6 % and the polyester 2.9 times too stiff). Free/Connect points, a `Rigid` cable end
Expand Down Expand Up @@ -357,6 +361,68 @@ All notable changes to CableDyn are recorded in this file. The format follows
a pinned arm rod numbered 1, its tether on an explicit `Rod1B` POINT listed first, now writes
the outputs of the same deck with the arm as rod 3 bit for bit; before, the arm turned 172 deg
in pitch, the body 1.06 deg, and a leg tension moved 3.0 kN in 1 s.
- The static start of a neutral finite-EI line whose length equals its span (straight at zero
strain, so with neither weight nor tension) failed with neutral `ATTACHMENTS` or with
attachments that only carry a drag area: the attachment polish stalled at the round-off of the
axial forces, 7.6e-8 N, against a tolerance of 1e-8 times the solver's 1 N force floor, at
every load step. For a line whose chord is at least its length the force scale of the
convergence test now includes that round-off floor over the tolerance
(`16 eps max(EA/L0) |x| / tol`, as on the `EI = 0` path), so these lines start straight and
unloaded, also between two moving ends, restrained in torsion at one or both ends, and with a
chord 1e-8 longer than the line (its 0.1 N tension to 2.2e-7 N); slack lines and lines that
carry weight or tension above the floor are unchanged.
- The Newton tangent of the finite-EI static solve in a depth-profile current or a current grid
carries the current's gradient: the drag samples the current at the node positions, so it moves
with them. The drag part of the tangent now matches central differences of the residual to
8e-10 (before off by 5e-2 to 0.2 in sheared profiles). Newton takes up to a third fewer
iterations on a light cable in a strong shear (16 instead of 23); the equilibria are unchanged
within the solver tolerance.
- The static start of an `EI = 0` line held above the still-water level kept the buoyancy of its
dry part, while the march (with `WtrDpth` or a bathymetry) gives that part its dry weight. The
start was not the march's equilibrium: a chain hung from 5 m above the surface onto the
seabed moved by 1e-3 of its fairlead load in the first step, and the coupled start differed
from the standalone one by 2.4e-4. The static solve now carries the dry weight (a fixed point
on the wetted lengths): the start holds to 1e-10, the two routes agree to 1.4e-8, and a
vertical chain's end loads sum to its weight to 1.5e-8 (before 1.4 % low). The force balance
of `Connect` and `Free` points and free bodies and rods takes the same dry weight: a 50 t
`Connect` point hung by two chains from 10 m above the surface holds its position to the
output's 8 digits over 2 s (before it sank 1.1e-2 m). There a finite-EI line, carried without
its bending stiffness, takes its dry weight with or without a seabed, as its cubic-Hermite
cable does: a stiff body with a light cable pinned 5 m above the surface on a deck without a
seabed holds its heave to 7e-6 m over 3 s (before it moved 2.8e-3 m), and an `EI = 0` chain
between that body 5 m above the surface and a `Connect` clump holds both to the output's 8
digits (before 5.5e-3 m). Lines below the surface are unchanged.
- `ATTACHMENTS` above the still-water level kept their buoyancy in the static start of a
finite-EI line, while the march loads them with their weight alone. A 500 kg, 0.2 m^3 module
3 m above the surface on a cable hung from 5 m above it moved the fairlead load by up to
6.1 % in the first second, and as much on the same cable restrained in torsion. The
attachment polish and the torsion stage after it now give each attachment the load of its side
of the surface (a fixed point that repeats the solve until none changes side); both starts
hold to the output's 8 digits. An attachment that its weight in air takes below the surface
and its weight in water above it (a net-buoyant module that would float) has no static
position and fails closed, naming the attachment that crossed the surface.
- A bending-pinned finite-EI line with `ATTACHMENTS` from a Rigid6 body or a rod end left its
attachments out of the body static balance, which carries such a line as a cable without
bending stiffness: on a moored body with a 2 t clump 50 m along a pinned cable the body started
at the pose of the line without the clump and drifted by up to 2.9e-2 m and 0.20 deg in 5 s.
Such a line now takes the static passes of a connected end, which balance the body on the
cubic-Hermite cable's end loads; the body holds its pose to the output's 8 digits. Their
passes are Anderson-mixed, as those of a line between two objects, and the core cable of the
passes carries the line's `ATTACHMENTS` smeared, so the passes correct only the lumped,
side-of-the-surface and bending differences: on a free spar rod with a 1500 kg clump on a cable
pinned at its upper end the plain passes on a core cable without the clump diverged (the deck
stopped), and Anderson-mixed ones converged in 26 passes or stopped by round-off (a 1-ulp
change of an anchor coordinate; the Linux build stopped). They now converge in 6, the same
for those 1-ulp changes, and the rod holds its pose to 8 digits. The mixing drops history
columns that add less than 1e-6 of a new direction (1e-10 before), which had let round-off
columns drive the mixed loads to 1e6 N. Across a sweep of the rod deck (clump 0 to 4 t, cable
50 to 70 m, rod mass, clump position, start offsets, 1-ulp changes) 42 of 45 decks converge
(34 before), the three left with no stable equilibrium found on either core cable.
- A bending-pinned finite-EI line on a Rigid6 body or a rod end without `ATTACHMENTS` was left to
the body static solve's cable without bending stiffness, whose end force lacks the bending's
share: a free spar rod with a 55 m pinned cable (EI 2e4 N m^2) started 5.7e-2 m off its
equilibrium, drifted 7e-2 m in 3 s and moved a leg tension by 2.3 %. Such a line now takes the
static passes too (4 here); the rod holds its pose to the output's 8 digits.
- In OpenFAST, Free Rigid6 bodies and Free, Pinned and CoupledPinned rods started their first
step from zero acceleration: the module restores its state mirror after the output call at
`t = 0`, which marked the objects' accelerations as committed before the start accelerations
Expand Down
11 changes: 11 additions & 0 deletions CMakeLists.txt
Original file line number Diff line number Diff line change
Expand Up @@ -663,6 +663,13 @@ target_link_libraries(test_hermite_cable_static PRIVATE cabledyn_core)
add_test(NAME hermite_cable_static COMMAND test_hermite_cable_static)
set_tests_properties(hermite_cable_static PROPERTIES LABELS "fortran;unit;static")

# Finite-EI static tangent of a depth-profile current: the profile's vertical gradient in the drag
# tangent, against central differences of the residual
add_executable(test_hermite_static_current_gradient tests/test_hermite_static_current_gradient.f90)
target_link_libraries(test_hermite_static_current_gradient PRIVATE cabledyn_core)
add_test(NAME hermite_static_current_gradient COMMAND test_hermite_static_current_gradient)
set_tests_properties(hermite_static_current_gradient PROPERTIES LABELS "fortran;unit;static;hermite")

# Finite-EI mesh-resolution diagnosis: h*kappa peak / curvature-jump / contact-island metrics
# and the power-of-two refinement recommendation -- the preflight trigger for the adaptive mesh
# policy. Synthetic controlled profiles + fail-closed validation.
Expand Down Expand Up @@ -1201,6 +1208,10 @@ target_link_libraries(test_flagship_compression_note PRIVATE cabledyn_core)
add_test(NAME flagship_compression_note
COMMAND test_flagship_compression_note ${CMAKE_SOURCE_DIR}/examples/iea15mw_umaine_mixed_cabledyn.dat)
set_tests_properties(flagship_compression_note PROPERTIES LABELS "fortran;integration;static;contact")
add_executable(test_static_dry_weight tests/test_static_dry_weight.f90)
target_link_libraries(test_static_dry_weight PRIVATE cabledyn_core)
add_test(NAME static_dry_weight COMMAND test_static_dry_weight WORKING_DIRECTORY ${CMAKE_BINARY_DIR})
set_tests_properties(static_dry_weight PROPERTIES LABELS "fortran;integration;static;openfast")

add_executable(test_viv_kernels tests/test_viv_kernels.f90)
target_link_libraries(test_viv_kernels PRIVATE cabledyn_core)
Expand Down
17 changes: 17 additions & 0 deletions VALIDATION.md
Original file line number Diff line number Diff line change
Expand Up @@ -392,6 +392,23 @@ scripts, results, and source provenance are in
lazy-wave power cable (positions and tangents of its free nodes, and their rates). Every coupled
point's linearised stiffness equals central differences of the deck's own static solutions to
≤3.3e-4, the cable's 36.0 / 2.62 / 38.5 N/m included, and also with a 5e4 N·m/rad hang-off spring.
With that spring, the t = 0 linearisation's force per unit rotation reads 1.2 % below the static
solutions out of plane (Fy per x-rotation 2.583 against 2.614 N/rad). In plane it is 0.08 % and
0.32 % off (Fx and Fz per y-rotation). The cause is the operating point, not the `.lin`
matrices. A probe that differentiates the core's acceleration recompute as `lin_eval` does
reproduces `A`, `B`, `C` and `D` to 1e-8. The platform starts out of balance, so at t = 0 the
fairlead accelerates at (−0.019, 0, −0.016) m/s². The cable's coupled load is its support
reaction, which includes the inertia `−M(q) a` of the line, and turning the end tangent changes
that reaction while the fairlead accelerates. With this acceleration prescribed, the probe
reproduces the rotation columns to 2e-5 (2.5826 N/rad). With the platform DOFs disabled in
ElastoDyn, the coupled point does not accelerate. The OpenFAST reduction is then symmetric,
and it matches the static solutions, Richardson-extrapolated over 0.125° and 0.25° rotations,
in plane to 1.1e-5 (Fx per y-rotation 33.8954 / 33.8952 N/rad) and out of plane to 2.7e-4
(2.6149 / 2.6142 N/rad). The out-of-plane residue is the O(h²) truncation of the 1 µm state
step on the soft sideways mode of the frictionless laid tail. In the probe it falls to 2.6e-5
at a 0.3 µm step and grows to 2.9e-2 at 10 µm, and the sway translation column shows the same
residue. The moments per unit translation carry no inertia, so they match the static solutions
at either operating point.
Two step sizes had to be small. A millimetre position step lifted the cable's grounded nodes out
of their 0.3 mm seabed penetration on one side of the difference, and a 5 cm coupled-point step
slackened the chain's last element. Each halved a stiffness (the cable's sway stiffness read 242
Expand Down
18 changes: 11 additions & 7 deletions doc/driver_format.md
Original file line number Diff line number Diff line change
Expand Up @@ -421,9 +421,12 @@ connection moment at every connected end, and at both ends of a cable between tw

Static initial condition: the static solve moves the free bodies, the free rods, the pinned rods
(two rotations about the pin, the pin force on the parent body) and the `Free`/`Connect` points
together. A finite-EI line enters it with its axial stiffness and weight, without its bending
stiffness (a note names the line); its own shape is then solved with the bending stiffness at
the equilibrium end positions.
together. A finite-EI line enters it with its axial stiffness and weight (its part above the
surface with its dry weight), without its bending stiffness (a note names the line); its own
shape is then solved with the bending stiffness at the equilibrium end positions. A line
connected (clamped, elastic or restrained in torsion) at a body or rod end, a line between two
objects, and a bending-pinned line on a Rigid6 body or a rod end are balanced on their
cubic-Hermite end loads in static passes.

**OpenFAST `CompMooring = 5` (aggregate route):** finite-EI cables run on the cubic-Hermite path:
lazy-wave statics and generalised-α dynamics driven by the coupled fairlead, including platform
Expand Down Expand Up @@ -1027,12 +1030,13 @@ Behaviour:
A deck current (a `current` row, uniform or profile, or a MoorDyn-C `Currents` field at
`t = 0`, sampled at the attachment node) loads the attachments in the static initial condition
with the same drag law at rest, so a run in a current starts in equilibrium.
An attachment above the free surface carries its weight only. Attachments have no seabed
contact of their own; the node's contact acts on the line.
An attachment above the free surface carries its weight only, in the static initial condition
too. Attachments have no seabed contact of their own; the node's contact acts on the line.
- The static equilibrium is solved first with each attachment's net weight spread over the two
elements at its node, then carried to the discrete loads by continuation. The current drag
depends on the node's position and line direction, so the final solve is repeated until the
node positions change by less than 1e-9 m. On a line restrained in torsion the torsion stage
depends on the node's position and line direction, and the weight on the side of the surface
an attachment is on, so the final solve is repeated until the node positions change by less
than 1e-9 m and no attachment changes side. On a line restrained in torsion the torsion stage
(the imposed twist) follows on the discrete equilibrium, with the attachments' net weights
and drag as nodal loads (the drag again to a fixed point), so a run starts in the twisted
equilibrium. An attachment adds no polar inertia to the line's twist.
Expand Down
16 changes: 14 additions & 2 deletions doc/openfast.rst
Original file line number Diff line number Diff line change
Expand Up @@ -484,8 +484,20 @@ turn of an end frame or connection direction is nonlinear); a run without linear
operating point, such as the still-water start, they lie on the linearised surface, while at a
moving one the accelerations and the end-frame rates of the last step enter them.

The zero-frequency reduction ``D − C A⁻¹ B`` is the static mooring stiffness. It agrees with
MoorDyn-F and with CableDyn's own static solutions (:doc:`validation`). Quadratic drag has no
The zero-frequency reduction ``D − C A⁻¹ B`` is the static mooring stiffness at an operating point
whose coupled points do not accelerate. It agrees with MoorDyn-F and with CableDyn's own static
solutions (:doc:`validation`). A cable's coupled load is its support reaction, which includes the
inertia ``−M(q) a`` of the line, and the mass moves with the configuration: the added mass turns
with each element, and the end tangent's mass couples to the fairlead. At a coupled-point
acceleration ``a`` the force rows therefore also carry the configuration derivative of that
inertial reaction. The rotation columns, which turn the end tangent, show it most. An OpenFAST run
generally starts out of balance, so its platform accelerates at ``t = 0``. On the lazy-wave cable of
the mixed example, held by a 5e4 N·m/rad hang-off spring, a fairlead acceleration of 0.025 m/s²
moves the force per unit rotation by 1.2 % out of plane and by 0.08–0.32 % in plane. The translation
columns and the moments per unit translation stay within 3e-4 of the static solutions. At zero
acceleration the rotation columns match them too, to 2.7e-4 out of plane and 1.1e-5 in plane.
Compare a reduction with static solutions only at an operating point whose coupled points do not
accelerate, for example with the platform degrees of freedom disabled. Quadratic drag has no
slope at rest, so the 0.1 m/s step gives its equivalent linear damping at that velocity amplitude.
CableDyn integrates the drag over the element, and a one-node step then gives about half of
MoorDyn-F's lumped value.
Expand Down
Loading
Loading