VehicleSDF: A 3D generative model for constrained engineering design via surrogate modeling

TL;DR

This work generates diverse 3D models of cars that meet a given set of geometric specifications, while also obtaining quick estimates of performance parameters such as aerodynamic drag.

Abstract

A main challenge in mechanical design is to efficiently explore the design space while satisfying engineering constraints. This work explores the use of 3D generative models to explore the design space in the context of vehicle development, while estimating and enforcing engineering constraints. Specifically, we generate diverse 3D models of cars that meet a given set of geometric specifications, while also obtaining quick estimates of performance parameters such as aerodynamic drag. For this, we employ a data-driven approach (using the ShapeNet dataset) to train VehicleSDF, a DeepSDF based model that represents potential designs in a latent space witch can be decoded into a 3D model. We then train surrogate models to estimate engineering parameters from this latent space representation, enabling us to efficiently optimize latent vectors to match specifications. Our experiments show that we can generate diverse 3D models while matching the specified geometric parameters. Finally, we demonstrate that other performance parameters such as aerodynamic drag can be estimated in a differentiable pipeline.

Figures (14)

• Figure 1: An illustration of our proposed pipeline and components. The latent vector $\bm{z}_i$ in (\ref{['fig: Training of DeepSDF using 3D shapes']}) is initialized randomly from $\mathcal{N}(0, \sigma^2)$ and optimized. The optimized and augmented latent vectors then are used in (\ref{['fig: Training of surrogate parameter estimator']}). The orange-outlines mark the components we trained.
• Figure 2: Vehicle geometric parameters to be specified during design
• Figure 3: Drag estimation pipeline
• Figure 4: Optimization of 3D shape to match specified parameters
• Figure 5: Results of optimization starting from four different initial shapes. The arrow indicate the points of change from the leftmost image. Figure \ref{['fig: Anather results of optimization starting from four different initial shapes. The arrow indicate the points of change from the leftmost image']} in Appendix \ref{['appendix: VehicleSDF generation']} show another examples.