Skip to content

Choosing variables and running the optimizer

The Design tab offers fin span, root chord, tip chord, sweep distance, thickness, axial fin position, plus optional ballast variables. The source exposes variable-specific bound controls and geometry preview; it requires at least one enabled optimization variable.

Practical workflow

  1. Confirm the baseline CDX1 and model compatibility.
  2. Enable only the parameters you wish to optimize. For a focused initial study, consider span and sweep; check the rendered extremes.
  3. Keep geometric constraints coherent. The app checks that root chord does not extend beyond the permitted axial location.
  4. In Optimization, choose Maximum evaluations, Starting points, output folder, and optional random seed.
  5. Configure static and optional flutter screens, then press Run.
  6. Examine the best feasible evaluated candidate and other shortlisted configurations. Budget exhaustion is not convergence and does not imply a global optimum.

The code incorporates initial Latin-hypercube sampling and Powell local refinement with candidate-vector caching. Assignment 14 demonstrated three distinct initial samples plus three distinct refinement candidates during a nine-call budget, with some cached repeat calls. More evaluations generally give better coverage but increase runtime.

Suggested use of budgets

Begin with a modest pilot run to verify CDX1 compatibility and automation. Then increase the budget until you have enough actual feasible candidates to compare. There is no validated universal evaluation count for a given rocket; choose a budget based on dimensions, bounds, and available compute time.

Important: the program automates desktop RASAero and may capture mouse/keyboard focus. Pause and resume is available but was not actively tested for Assignment 14.

Source basis: Assignment 14 review materials supplied with this documentation. Internal archive paths and raw evidence are not published.