Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DeepWheel: Generating a 3D Synthetic Wheel Dataset for Design and Performance Evaluation

Soyoung Yoo, Namwoo Kang

TL;DR

DeepWheel tackles the shortage of large-scale, high-fidelity 3D wheel datasets with a fully automated, multi-modal data-generation pipeline. It combines 2D rendering via Stable Diffusion guided by topology-optimized designs, domain-adapted depth prediction for 2.5D geometry, a 2D-to-3D reconstruction pipeline, and CAE-based performance simulations to produce over 6,000 renderings and 904 CAD-ready 3D meshes with coupling to modal analysis data. The work demonstrates that depth-map–based embeddings enable balanced design-space sampling and that topology optimization expands both design and performance spaces, enabling efficient surrogate-model training and data-driven design exploration. By releasing a CC BY-NC 4.0 multi-modal wheel dataset, DeepWheel provides a scalable framework for data-driven engineering in wheel design and is extensible to other complex design domains.

Abstract

Data-driven design is emerging as a powerful strategy to accelerate engineering innovation. However, its application to vehicle wheel design remains limited due to the lack of large-scale, high-quality datasets that include 3D geometry and physical performance metrics. To address this gap, this study proposes a synthetic design-performance dataset generation framework using generative AI. The proposed framework first generates 2D rendered images using Stable Diffusion, and then reconstructs the 3D geometry through 2.5D depth estimation. Structural simulations are subsequently performed to extract engineering performance data. To further expand the design and performance space, topology optimization is applied, enabling the generation of a more diverse set of wheel designs. The final dataset, named DeepWheel, consists of over 6,000 photo-realistic images and 900 structurally analyzed 3D models. This multi-modal dataset serves as a valuable resource for surrogate model training, data-driven inverse design, and design space exploration. The proposed methodology is also applicable to other complex design domains. The dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International(CC BY-NC 4.0) and is available on the https://www.smartdesignlab.org/datasets

DeepWheel: Generating a 3D Synthetic Wheel Dataset for Design and Performance Evaluation

TL;DR

DeepWheel tackles the shortage of large-scale, high-fidelity 3D wheel datasets with a fully automated, multi-modal data-generation pipeline. It combines 2D rendering via Stable Diffusion guided by topology-optimized designs, domain-adapted depth prediction for 2.5D geometry, a 2D-to-3D reconstruction pipeline, and CAE-based performance simulations to produce over 6,000 renderings and 904 CAD-ready 3D meshes with coupling to modal analysis data. The work demonstrates that depth-map–based embeddings enable balanced design-space sampling and that topology optimization expands both design and performance spaces, enabling efficient surrogate-model training and data-driven design exploration. By releasing a CC BY-NC 4.0 multi-modal wheel dataset, DeepWheel provides a scalable framework for data-driven engineering in wheel design and is extensible to other complex design domains.

Abstract

Data-driven design is emerging as a powerful strategy to accelerate engineering innovation. However, its application to vehicle wheel design remains limited due to the lack of large-scale, high-quality datasets that include 3D geometry and physical performance metrics. To address this gap, this study proposes a synthetic design-performance dataset generation framework using generative AI. The proposed framework first generates 2D rendered images using Stable Diffusion, and then reconstructs the 3D geometry through 2.5D depth estimation. Structural simulations are subsequently performed to extract engineering performance data. To further expand the design and performance space, topology optimization is applied, enabling the generation of a more diverse set of wheel designs. The final dataset, named DeepWheel, consists of over 6,000 photo-realistic images and 900 structurally analyzed 3D models. This multi-modal dataset serves as a valuable resource for surrogate model training, data-driven inverse design, and design space exploration. The proposed methodology is also applicable to other complex design domains. The dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International(CC BY-NC 4.0) and is available on the https://www.smartdesignlab.org/datasets

Paper Structure

This paper contains 50 sections, 11 equations, 18 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Research Background
  3. Research Objective
  4. Research Contributions
  5. Related works
  6. 3D Dataset for Engineering Design
  7. Generative Model for Automotive Wheel Design
  8. 3D Reconstruction from Image
  9. Research Gap & Summary
  10. Methodology
  11. Overall Framework
  12. Stage1: Image to Image Translation for 2D Generation
  13. Design Variations using Topology Optimization
  14. Photorealistic Rendering based on Stable Diffusion
  15. Overall 2D Design Workflow
  16. ...and 35 more sections

Figures (18)

  • Figure 1: Overview of public 3D object and engineering datasets
  • Figure 2: Research framework
  • Figure 3: Overall workflow of generating diverse 2D wheel designs and photorealistic renderings.
  • Figure 4: Architecture of depth estimator(Marigold)
  • Figure 5: RGB-D dataset collection
  • ...and 13 more figures