Bayesian shape design improved using example-based parameter choices
Example-driven Parametrisations for Bayesian Shape Optimisation
Machine Learning
Summary
Designing shapes like airplane wings or radio parts is hard because testing new shapes can take a lot of time and resources. Bayesian optimisation helps find good shapes, but it needs a simple way to describe all possible shapes to search effectively. The authors show how to create this description automatically by learning from a collection of existing shapes, making the search more efficient and flexible. This method works for various shapes in two and three dimensions and lets designers explore better shapes faster.
What this means in practice
- •For aerospace engineers: Accelerate the design of wings and aerofoils by automatically generating efficient search spaces from existing shape designs.
- •For radio-frequency device designers: Improve tuning of radio-frequency cavities by learning shape parameterisations directly from prior designs to enable faster design iterations.
Authors
Gabriel Diaz-Aylwin, Joseph Neighbor, Abiel Malkani Talwar, Rui-Yang Zhang, Henry B. Moss
Abstract
Bayesian optimisation is the natural tool for shape design when objectives are expensive and non-differentiable, but it needs a compact yet expressive parameterisation of the search space. Hand-crafting one is a complex endeavour requiring domain expertise, and often yields implicit infeasible regions, artificial bounds, and coupled, unordered coordinates. We instead learn the parameterisation from a collection of existing designs, applying principal component analysis to the deformations between shapes. The result is a linear, interpretable search space in which the number of components explicitly trades expressivity against dimensionality. Across aerofoils, wings, and radio-frequency cavities, spanning 2D geometry to 3D aerodynamics and electromagnetics, we show improved sample efficiency and the ability to explore beyond the confines of hand-crafted baselines.