L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Design et création · SKILL.md

calculix-modal-analysis

calculix-modal-analysis

Workflow skill for CalculiX modal analysis (*FREQUENCY steps) — natural frequencies and mode shapes via the CalculiX MCP. read_results returns the frequency table; export_results with mode=N renders that mode shape in the browser viewer. Use when an agent must extract natural frequencies, check resonance margins, or visualize mode shapes on an open-source s…

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: calculix-modal-analysis description: Workflow skill for CalculiX modal analysis (*FREQUENCY steps) — natural frequencies and mode shapes via the CalculiX MCP. read_results returns the frequency table; export_results with mode=N renders that mode shape in the browser viewer. Use when an agent must extract natural frequencies, check resonance margins, or visualize mode shapes on an open-source solver.

CalculiX Modal Analysis

Free-vibration eigenanalysis on a CalculiX deck: a *FREQUENCY step requests N eigenpairs; ccx prints the eigenvalue table and per-mode eigenvectors to the .dat, so both frequencies and mode shapes parse from text (no .frd needed).

When to Use

Use when an agent must extract natural frequencies, check resonance/vibration margins, or visualize mode shapes on a CalculiX model. Driven by the same MCP tools as static runs (run_solver → read_results → export_results).

Not for: static stress/deflection (use calculix-fem), transient/dynamic response (not yet supported), or sizing optimization (calculix-sizing-optimization).

Workflow

  1. Confirm the deck has *DENSITY under its *MATERIAL — frequencies need mass; without density ccx fails the eigenvalue solve.
  2. Confirm the step is *FREQUENCY with the wanted mode count on its data line, and *NODE PRINT, NSET=<all> / U so eigenvectors reach the .dat. No load is applied (free vibration); keep the *BOUNDARY clamp set.
  3. run_solver_tool — submit the deck. Note: a *FREQUENCY run writes a header-only .sta (no increments in an eigenvalue solve); success accepts a non-empty .dat instead, so this is normal, not a failure.
  4. read_results_tool — returns frequencies as [{mode, eigenvalue, freq_rad_s, freq_hz}] and n_modes (a doubly symmetric section gives degenerate pairs — f1 = f2 — which is expected).
  5. export_results_tool with mode=N — writes result_mesh.json holding that mode's eigenvector as a stress-free displacement field; the viewer renders the mode shape with its usual auto-scaled deformation.

Rules

  • Units follow the .inp (commonly mm-t-s-MPa → frequencies in Hz).
  • Eigenvectors are mass-normalized; their magnitude carries no physical displacement meaning — only the shape does. The viewer auto-scales.
  • Fully-integrated C3D8 hexes shear-lock in bending: ccx frequencies run ~5-10% ABOVE the Euler-Bernoulli hand calc. For tight margins, refine the mesh through the thickness (or switch element type) and report the gap.
  • Hand-calc check for a clamped-free bar: f_n = (beta_n^2 / 2pi) * sqrt(E I / (rho A L^4)), beta_1 = 1.8751.

Example

MCP/CalculiX/examples/cantilever_modal.inp is the public cantilever benchmark with a 5-mode *FREQUENCY step. ccx gives f1 = f2 ~ 502 Hz (degenerate bending pair on the square section) vs the Euler-Bernoulli hand calc ~ 464 Hz (+8%, C3D8 shear locking). Mode 1 exports straight into the viewer as the classic half-sine bend.

Skills enregistrées