calculix-sizing-optimization

CalculiX Sizing Optimization

Two-stage sizing/parameter optimization: minimize mass subject to stress, displacement, or natural-frequency constraints by editing scalar design variables in place (shell thickness, beam section, material E/nu/density, load magnitude). The mesh and geometry never change — only scalar cards.

This is sizing optimization, not topology optimization. It thins sections; it does not redistribute material in space.

When to Use

Use when an agent must lighten a CalculiX shell or beam model while keeping von Mises stress and displacement within limits (static deck), or must lighten it while keeping a natural frequency above a resonance floor (modal deck). Driven by the optimize_structure_tool MCP tool.

Do NOT use for:

• Solid (C3D8 / C3D8R) models. Solids expose no scalar geometry card — their mass is set by node-defined volume x density, so there is no thickness to thin. Material/load variables on a solid are degenerate for mass minimization (density changes mass but not stiffness; E changes stiffness but not mass). Solid lightweighting needs shape or topology optimization, which is a different problem and is not covered here.
• Topology optimization (material distribution over a fixed mesh) — separate, future work.

Workflow

1. parse_inp / list_design_vars_tool — confirm the deck and find the shell.<elset>.thickness (or beam section) var_id and its current value.
2. Choose bounds {var_id: [lower, upper]} to bracket the search. Mass falls monotonically with shell/beam thickness.
3. optimize_structure_tool — run the two-stage loop (LHS sweep, then coordinate descent). Each evaluation is a real ccx solve, so set max_solves to bound wall time.
4. Inspect the result: best (vars, mass_kg, stress_vm, disp, feasible, mass_reduction_pct), converged / termination_reason, and history.
5. Optional: export_results_tool on the persisted <stem>.optimized.inp to render the optimized design in the viewer.