Automatic Parameterization for Aerodynamic Shape Optimization via Deep Geometric Learning

TL;DR

This paper introduces two deep geometric learning models that automate aerodynamic shape parameterization, embedding human knowledge and enabling efficient gradient-based optimization without manual feature crafting.

Contribution

The paper presents novel deep learning models for shape parameterization that incorporate human prior knowledge and facilitate end-to-end aerodynamic optimization workflows.

Findings

Both models successfully automate shape parameterization.

Models are fully differentiable and compatible with gradient-based optimization.

Effective shape optimization demonstrated on 2D airfoils.

Abstract

We propose two deep learning models that fully automate shape parameterization for aerodynamic shape optimization. Both models are optimized to parameterize via deep geometric learning to embed human prior knowledge into learned geometric patterns, eliminating the need for further handcrafting. The Latent Space Model (LSM) learns a low-dimensional latent representation of an object from a dataset of various geometries, while the Direct Mapping Model (DMM) builds parameterization on the fly using only one geometry of interest. We also devise a novel regularization loss that efficiently integrates volumetric mesh deformation into the parameterization model. The models directly manipulate the high-dimensional mesh data by moving vertices. LSM and DMM are fully differentiable, enabling gradient-based, end-to-end pipeline design and plug-and-play deployment of surrogate models or adjoint solvers. We perform shape optimization experiments on 2D airfoils and discuss the applicable scenarios for the two models.