From 734d1c63fd8a8805ae376c5d6374bf1448ede5cf Mon Sep 17 00:00:00 2001 From: Jae Hoon Seo Date: Sun, 8 Nov 2026 23:42:33 +0900 Subject: [PATCH 1/2] fix(dynamics): the step round-off floor takes the axial damping The force scale of a finite-EI dynamic step is bounded below by the round-off floor of the axial forces, 16 eps max(EA/L0) x_ref/tol. The axial damping force BA d|r'|/dt takes its strain rate from the Newmark velocity, whose rounding is gamma/(beta dt) times that of the coordinates, and the floor left it out. A held line damped near its first axial mode (BA = 2 zeta EA/omega_1, as mode1 axialDampingRatio gives, or a direct BA of that size) puts the damping term well above EA at a small dtM: its steps at rest stalled at a few times the tolerance and the run stopped (3 of 30 runs of a 400 m wire held at rest over BA 5e7 to 8e8 N s and dtM 0.002 to 0.01 s, at varying points of the sweep as the rounding falls). The floor now adds BA gamma/(beta dt) to the axial stiffness of each element, as the twist floor adds the torsional damping. Lines without axial damping take the same scale as before. The new hermite_damped_rest test runs the sweep and requires every run to complete with its end tensions at their static values to 1e-6. Signed-off-by: Jae Hoon Seo --- CHANGELOG.md | 7 ++ CMakeLists.txt | 11 +++ doc/theory.rst | 6 +- src/CableDyn_HermiteCableDynamic.f90 | 24 ++++--- tests/test_hermite_damped_rest.py | 101 +++++++++++++++++++++++++++ 5 files changed, 138 insertions(+), 11 deletions(-) create mode 100644 tests/test_hermite_damped_rest.py diff --git a/CHANGELOG.md b/CHANGELOG.md index e3549554..cc2d35c7 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -18,6 +18,13 @@ All notable changes to CableDyn are recorded in this file. The format follows The touchdown element now takes the reaction of its grounded stretch, from the grounded node to the touchdown point located from the two nodes' clearances: the error falls at the second order (2.1, 0.7, 0.2 N) and the outputs vary continuously. +- Finite-EI dynamic step at rest under a large axial damping coefficient: the round-off floor of + the step's force scale left out the axial damping, whose strain rate is formed from the Newmark + velocity and carries gamma/(beta dtM) times the rounding of the coordinates. A held line damped + near its first axial mode (`BA` = 2 zeta EA/omega_1, as `mode1 axialDampingRatio` gives, or a + direct `BA` of that size) stalled at a few times the tolerance and stopped (3 of 30 runs of a + 400 m wire held at rest, `BA` 5e7 to 8e8 N s, `dtM` 0.002 to 0.01 s). The floor now includes + `BA` gamma/(beta dtM), as the twist floor includes its damping. ## [0.4.1] - 2026-11-08 diff --git a/CMakeLists.txt b/CMakeLists.txt index 96dcf84c..8073e5a2 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -1794,6 +1794,17 @@ if(Python3_Interpreter_FOUND) TIMEOUT 600 ENVIRONMENT "${_cabledyn_python_test_environment}") + # A heavily damped finite-EI line held at rest completes every step and stays at rest over + # a sweep of the axial damping coefficient and dtM (the damping's residual round-off floor). + add_test(NAME hermite_damped_rest + COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/tests/test_hermite_damped_rest.py + --exe $ + --output ${CMAKE_BINARY_DIR}/hermite_damped_rest) + set_tests_properties(hermite_damped_rest PROPERTIES + LABELS "python;hermite;dynamic;robustness" + TIMEOUT 300 + ENVIRONMENT "${_cabledyn_python_test_environment}") + # Range graphs (LINES flag r) and touchdown channels on every standalone route: envelopes # against the node channels, the analytic catenary TDP, quasi-static tracking under a # moving fairlead, and agreement with cabledyn.touchdown_history (needs numpy). diff --git a/doc/theory.rst b/doc/theory.rst index a16eb76b..c3988d09 100644 --- a/doc/theory.rst +++ b/doc/theory.rst @@ -2317,8 +2317,10 @@ residual decreases; its convergence norm divides each translational residual by frozen at the predictor and each tangent-DOF residual additionally by the nodal tributary length. The force scale is the largest inertial or residual entry, bounded below by 1 N, by the peak element weight, and by the round-off floor -:math:`16\,\epsilon\,\max(EA/L_0)\,x_\text{ref}/\text{tol}` of the axial forces -(:math:`x_\text{ref}` the largest coordinate magnitude or element length). The +:math:`16\,\epsilon\,\max\big((EA + BA\,\gamma/(\beta\,\Delta t))/L_0\big)\,x_\text{ref}/\text{tol}` +of the axial forces (:math:`x_\text{ref}` the largest coordinate magnitude or element length; +the axial damping takes the strain rate from the Newmark velocity, whose rounding is +:math:`\gamma/(\beta\,\Delta t)` times that of the coordinates). The deck defaults are relative tolerance :math:`10^{-8}`, absolute tolerance :math:`10^{-14}`, 30 iterations, and 12 backtracks (option ``dynamic_solver``). Option ``modified_newton`` reuses the factorised tangent within a step while each iterate reduces the residual to at most a quarter diff --git a/src/CableDyn_HermiteCableDynamic.f90 b/src/CableDyn_HermiteCableDynamic.f90 index 2a209e54..388856e4 100644 --- a/src/CableDyn_HermiteCableDynamic.f90 +++ b/src/CableDyn_HermiteCableDynamic.f90 @@ -4828,7 +4828,7 @@ SUBROUTINE eval_res(qcur, use_trial, rn, esx, emx) REAL(wp), POINTER :: Gout(:) INTEGER :: ii, nd INTEGER(INT64) :: ec0, ec1 - REAL(wp) :: xref, xd, xc, ea_r, kt, tw_round + REAL(wp) :: xref, xd, xc, ea_r, ka, kt, tw_round REAL(wp), PARAMETER :: RESID_ROUNDOFF = 16.0_wp REAL(wp), PARAMETER :: STATE_DIVERGED = 1.0e12_wp ec0 = 0_INT64 ! defined even with profiling off (see the step-entry note) @@ -4948,18 +4948,24 @@ SUBROUTINE eval_res(qcur, use_trial, rn, esx, emx) ! whose rounding is (K^2/GJ) eps |r|/l0, so the floor takes the whole axial stiffness ! EA - K^2/GJ + K^2/GJ of the line (without it the floor fell by 1 - K^2/(EA GJ): at ! K = 0.99 sqrt(EA GJ) to 2 % of the rounding it bounds, and a held line failed its step). + ! The axial damping force BA d(|r'|)/dt takes the strain rate from the Newmark velocity, + ! whose rounding is gamma/(beta dt) times that of the coordinates, so the floor adds + ! BA gamma/(beta dt) to the axial stiffness, as the twist floor below adds Bt: without it a + ! held line damped near its first axial mode (BA = 2 zeta EA/omega_1, which at small dt + ! puts the damping term well above EA) stagnated at a residual of a few times the + ! tolerance and failed its step. xref = MAXVAL(model%l0) DO ii = 1, ndof IF (MOD(ii - 1, 6) < 3) xref = MAX(xref, ABS(qcur(ii))) END DO - IF (model%torsion%active .AND. ALLOCATED(model%torsion%couple)) THEN - ea_r = CD_ZERO - DO ii = 1, SIZE(model%l0) - ea_r = MAX(ea_r, (ABS(model%EA(ii)) + model%torsion%couple(ii)**2*model%torsion%gj(ii))/model%l0(ii)) - END DO - ELSE - ea_r = MAXVAL(model%EA/model%l0) - END IF + ea_r = CD_ZERO + DO ii = 1, SIZE(model%l0) + ka = ABS(model%EA(ii)) + IF (model%torsion%active .AND. ALLOCATED(model%torsion%couple)) & + ka = ka + model%torsion%couple(ii)**2*model%torsion%gj(ii) + IF (model%has_axial_damping) ka = ka + ABS(model%BA(ii))*gm/(bt*dt) + ea_r = MAX(ea_r, ka/model%l0(ii)) + END DO fscale = MAX(fscale, RESID_ROUNDOFF*EPSILON(CD_ONE)*ea_r*xref/tol) ! Nodal twist: the torque scale of the angle rows, frozen with fscale and formed like it ! from the nodal loads (a node's torque imbalance, not the torque the line carries, diff --git a/tests/test_hermite_damped_rest.py b/tests/test_hermite_damped_rest.py new file mode 100644 index 00000000..3cd972cc --- /dev/null +++ b/tests/test_hermite_damped_rest.py @@ -0,0 +1,101 @@ +#!/usr/bin/env python3 +# SPDX-License-Identifier: Apache-2.0 +# Copyright (c) 2026 Jae Hoon Seo, SMI Lab, Inha University +"""A heavily damped finite-EI line held at rest stays at rest on every time step. + +A taut 400 m wire (EA = 2.19e9 N, 111 elements) is held at both ends and started from its +static equilibrium. The axial damping coefficient BA is swept over 5e7 to 8e8 N s (2 zeta +EA/omega_1 with zeta from about 0.4 to 6 for its first axial mode, the damping that +``mode1 axialDampingRatio`` gives) and dtM over 0.002, 0.005 and 0.01 s. The damping force +takes the strain rate from the Newmark velocity, whose rounding is gamma/(beta dt) times +that of the coordinates, so at these BA and dtM the residual round-off of a step at rest +lies well above the tolerance on the undamped force scale. Every run must complete (exit +code 0) and keep its end tensions at their static values to 1e-6. + +usage: test_hermite_damped_rest.py --exe --output +""" + +from __future__ import annotations + +import argparse +import os +import subprocess +import sys +from pathlib import Path + +DECK = """--------------------- CableDyn Input File ------------------------------------ +Taut finite-EI line held at rest with a large axial damping coefficient +--------------------- LINE TYPES --------------------------------------- +TypeName Diam Mass EA BA EI Cd_n Cd_t Ca_n Ca_t +(-) (m) (kg/m) (N) (N-s) (N-m2) (-) (-) (-) (-) +wire 0.156 121.7 2.19e9 {ba} 963.0 1.2 0.008 1.0 0.0 +--------------------- POINTS ------------------------------------------- +ID Type X Y Z +(-) (-) (m) (m) (m) +1 Coupled 82.52 80.42 -8.60 +2 Fixed -117.94 395.14 -139.38 +--------------------- LINES -------------------------------------------- +ID NodeA NodeB Outputs +(-) (-) (-) (-) +1 1 2 - +--------------------- SECTIONS ----------------------------------------- +LineID LineType Length NumSegs +(-) (-) (m) (-) +1 wire 400.52 111 +--------------------- OPTIONS ------------------------------------------ +139.38 WtrDpth +2.9e6 kBot +{dt} dtM +{tmax} TMax +--------------------- OUTPUTS ------------------------------------------ +FairTen1 +AnchTen1 +--------------------- need this line ----------------------------------- +""" + +BA = ["5.0e7", "7.0e7", "1.0e8", "1.3e8", "1.6e8", "2.0e8", "3.0e8", "4.0e8", "6.0e8", "8.0e8"] +DT = ["0.002", "0.005", "0.01"] + + +def main() -> int: + ap = argparse.ArgumentParser() + ap.add_argument("--exe", required=True) + ap.add_argument("--output", required=True) + a = ap.parse_args() + out = Path(a.output) + out.mkdir(parents=True, exist_ok=True) + env = dict(os.environ) + env.setdefault("OPENBLAS_NUM_THREADS", "1") + failures = [] + for ba in BA: + for dt in DT: + name = "held_%s_%s" % (ba, dt) + deck = out / (name + ".dat") + deck.write_text(DECK.format(ba=ba, dt=dt, tmax=repr(round(40 * float(dt), 6))), encoding="utf-8", + newline="\n") + root = out / name + p = subprocess.run([a.exe, str(deck), str(root)], capture_output=True, text=True, env=env, + cwd=str(out)) + if p.returncode != 0: + failures.append("BA=%s dtM=%s: exit %d: %s" % (ba, dt, p.returncode, (p.stdout + p.stderr)[-300:])) + continue + rows = Path(str(root) + ".out").read_text(encoding="utf-8").splitlines() + names = rows[1].split() + vals = [[float(v) for v in r.split()] for r in rows[2:] if r.strip()] + if len(vals) != 41: + failures.append("BA=%s dtM=%s: %d output rows, expected 41" % (ba, dt, len(vals))) + continue + for ch in ("FairTen1", "AnchTen1"): + k = names.index(ch) + t0 = vals[0][k] + drift = max(abs(v[k] - t0) for v in vals) / abs(t0) + if drift > 1.0e-6: + failures.append("BA=%s dtM=%s: %s drifts by %.3e of its static value" % (ba, dt, ch, drift)) + print("held damped line: %d runs, %d failures" % (len(BA) * len(DT), len(failures))) + for f in failures: + print("FAIL:", f) + return 1 if failures else 0 + + +if __name__ == "__main__": + sys.exit(main()) From 33d9c9bbfb55c971164958651667fd0682fceb85 Mon Sep 17 00:00:00 2001 From: Jae Hoon Seo Date: Mon, 9 Nov 2026 00:13:33 +0900 Subject: [PATCH 2/2] chore(changelog): the entry moves to the batch changelog change The [Unreleased] entries of the open changes are added together in one change, so that they do not conflict with each other. Signed-off-by: Jae Hoon Seo --- CHANGELOG.md | 7 ------- 1 file changed, 7 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index cc2d35c7..e3549554 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -18,13 +18,6 @@ All notable changes to CableDyn are recorded in this file. The format follows The touchdown element now takes the reaction of its grounded stretch, from the grounded node to the touchdown point located from the two nodes' clearances: the error falls at the second order (2.1, 0.7, 0.2 N) and the outputs vary continuously. -- Finite-EI dynamic step at rest under a large axial damping coefficient: the round-off floor of - the step's force scale left out the axial damping, whose strain rate is formed from the Newmark - velocity and carries gamma/(beta dtM) times the rounding of the coordinates. A held line damped - near its first axial mode (`BA` = 2 zeta EA/omega_1, as `mode1 axialDampingRatio` gives, or a - direct `BA` of that size) stalled at a few times the tolerance and stopped (3 of 30 runs of a - 400 m wire held at rest, `BA` 5e7 to 8e8 N s, `dtM` 0.002 to 0.01 s). The floor now includes - `BA` gamma/(beta dtM), as the twist floor includes its damping. ## [0.4.1] - 2026-11-08