parameter-optimization

Design of experiments (DOE) structures parameter sweeps systematically using sampling plans like Latin Hypercube Sampling (LHS) or Sobol sequences. This Skill generates DOE samples, recommends optimizers based on your budget and noise level, and ranks parameter influence to guide model calibration efficiently without exhaustive trial-and-error.

Sensitivity analysis ranks which parameters most influence model outputs using Sobol analysis and factorial designs. This Skill automates sampling and sensitivity calculations to identify high-impact parameters, reducing the dimensionality of your calibration problem and focusing optimization effort where it matters most.

Yes. Surrogate modeling fits a fast approximation to expensive simulation outputs. This Skill generates initial DOE samples, fits surrogates quickly, and recommends Bayesian optimization or other strategies to explore the parameter space efficiently with your evaluation budget.

Optimizer selection depends on problem size, noise tolerance, and available evaluations. This Skill recommends the right optimizer based on your parameter bounds, evaluation budget, and noise level, then outputs recommendations and decision guidance for your calibration workflow.

No. This Skill uses Python standard library only, with no external dependencies. It generates LHS, Sobol, and factorial DOE samples, performs sensitivity analysis, and supports optimizer selection for materials, FEM, and thermal simulations out of the box.

You specify noise level as an input parameter. This Skill adjusts sampling strategies and optimizer recommendations accordingly—recommending robust methods like CMA-ES for noisy problems and guiding your decision on when to invest in surrogate modeling versus direct optimization.