Point2FFD: Learning Shape Representations of Simulation-Ready 3D Models for Engineering Design Optimization

Methods for learning on 3D point clouds became ubiquitous due to the popularization of 3D scanning technology and advances of machine learning techniques. Among these methods, point-based deep neural networks have been utilized to explore 3D designs in optimization tasks. However, engineering computer simulations require high-quality meshed models, which are challenging to automatically generate from unordered point clouds. In this work, we propose Point2FFD: A novel deep neural network for learning compact geometric representations and generating simulation-ready meshed models. Built upon an autoencoder architecture, Point2FFD learns to compress 3D point clouds into a latent design space, from which the network generates 3D polygonal meshes by selecting and deforming simulation-ready mesh templates. Through benchmark experiments, we show that our proposed network achieves comparable shape-generative performance than existing state-of-the-art point-based generative models. In real world-inspired vehicle aerodynamic optimizations, we demonstrate that Point2FFD generates simulation-ready meshes of realistic car shapes and leads to better optimized designs than the benchmarked networks.

1. Introduction

Geometric data are ubiquitous in engineering design processes. During product development, engineers define different representations of 3D shapes to explore solutions, analyze performance and verify the compliance to manufacturing standards. However, handcrafted design features often bias the design exploration and constrain the solutions [42]. Furthermore, since these representations are often product-specific, this approach hinders the exploitation of engineering expertise embedded in similar previous products by transferring design properties.