Skip to main content
Log in

Integrating deep learning into CAD/CAE system: generative design and evaluation of 3D conceptual wheel

  • Industrial Application Paper
  • Published:
Structural and Multidisciplinary Optimization Aims and scope Submit manuscript


Engineering design research integrating artificial intelligence (AI) into computer-aided design (CAD) and computer-aided engineering (CAE) is actively being conducted. This study proposes a deep learning-based CAD/CAE framework in the conceptual design phase that automatically generates 3D CAD designs and evaluates their engineering performance. The proposed framework comprises seven stages: (1) 2D generative design, (2) dimensionality reduction, (3) design of experiment in latent space, (4) CAD automation, (5) CAE automation, (6) transfer learning, and (7) visualization and analysis. The proposed framework is demonstrated through a road wheel design case study and indicates that AI can be practically incorporated into an end-use product design project. Engineers and industrial designers can jointly review a large number of generated 3D CAD models by using this framework along with the engineering performance results estimated by AI and find conceptual design candidates for the subsequent detailed design stage.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12
Fig. 13
Fig. 14
Fig. 15
Fig. 16
Fig. 17
Fig. 18
Fig. 19
Fig. 20
Fig. 21
Fig. 22
Fig. 23
Fig. 24
Fig. 25
Fig. 26

Similar content being viewed by others


  • Ahmed F, Deb K, Bhattacharya B (2016) Structural topology optimization using multi-objective genetic algorithm with constructive solid geometry representation. Appl Soft Comput 39:240–250

    Article  Google Scholar 

  • Al Shalabi L, Shaaban Z (2006) Normalization as a preprocessing engine for data mining and the approach of preference matrix. In 2006 International conference on dependability of computer systems, IEEE, pp 207-214

  • Altair (2019) SimLab. Retrieved from

  • Andreassen E, Clausen A, Schevenels M, Lazarov BS, Sigmund O (2011) Efficient topology optimization in MATLAB using 88 lines of code. Struct Multidiscip Optim 43(1):1–16

    Article  Google Scholar 

  • Autodesk (2020a) Generative design. Retrieved from

  • Autodesk (2020b) Retrieved from

  • Bendsoe MP, Kikuchi N (1988) Generating optimal topologies in structural design using a homogenization method

  • Bourdin B (2001) Filters in topology optimization. Int J Numer Methods Eng 50(9):2143–2158

    Article  MathSciNet  Google Scholar 

  • Bruns TE, Tortorelli DA (2001) Topology optimization of nonlinear elastic structures and compliant mechanisms. Comput Methods Appl Mech Eng 190(26–27):3443–3459

    Article  Google Scholar 

  • Burnap A, Liu Y, Pan Y, Lee H, Gonzalez R, Papalambros PY (2016) Estimating and exploring the product form design space using deep generative models. In ASME 2016

  • Catmull E (1978) A hidden-surface algorithm with anti-aliasing. In Proceedings of the 5th annual conference on Computer graphics and interactive techniques (SIGGRAPH ’78)

  • Chen W, Ahmed F (2021) PaDGAN: learning to generate high-quality novel designs. J Mech Des 143(3)

  • Cunningham JD, Simpson TW, Tucker CS (2019) An investigation of surrogate models for efficient performance-based decoding of 3D point clouds. J Mech Des 141(12)

  • Du X, Sun C, Zheng Y, Feng X, Li N (2020a) Evaluation of vehicle vibration comfort using deep learning. Measurement 108634

  • Du X, Xu H, Zhu F (2020b) Understanding the effect of hyperparameter optimization on machine learning models for structure design problems arXiv preprint arXiv: 2007.04431

  • Feng Y, Feng Y, You H, Zhao X, Gao Y (2019) MeshNet: mesh neural network for 3D shape representation. In Proceedings of the AAAI Conference on Artificial Intelligence, vol 33, pp 8279-8286)

  • Geurts P, Ernst D, Wehenkel L (2006) Extremely randomized trees. Mach Learn 63(1):3–42

    Article  Google Scholar 

  • Guest JK, Prévost JH, Belytschko T (2004) Achieving minimum length scale in topology optimization using nodal design variables and projection functions. Int J Numer Methods Eng 61(2):238–254

    Article  MathSciNet  Google Scholar 

  • Guo X, Li W, Iorio F (2016) Convolutional neural networks for steady flow approximation. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, ACM, pp 481-490

  • Hinton GE, Salakhutdinov RR (2006) Reducing the dimensionality of data with neural networks. Science 313(5786):504–507

    Article  MathSciNet  Google Scholar 

  • International Design Engineering Technical Conferences and Computers and Information in Engineering Conference (n.d.) American Society of Mechanical Engineers, pp V02AT03A013-V02AT03A013

  • Jang S, Yoo S, Kang N (2021) Generative design by reinforcement learning: enhancing the diversity of topology optimization designs. arXiv preprint arXiv:2008.07119

  • Kallioras NA, Lagaros ND (2020). DzAIℕ: Deep learning based generative design. Procedia Manufacturing, 44, 591–598.

  • Kanezaki A, Matsushita Y, Nishida Y (2018) Rotationnet: joint object categorization and pose estimation using multiviews from unsupervised viewpoints. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 5010-5019

  • Kang N, Ren Y, Feinberg F, Papalambros P (2019) Form+ function: optimizing aesthetic product design via adaptive, geometrized preference elicitation. arXiv preprint arXiv:1912.05047

  • Kanopoulos N, Vasanthavada N, Baker RL (1988) Design of an image edge detection filter using the Sobel operator. IEEE J Solid State Circuits 23(2):358–367

    Article  Google Scholar 

  • Khadilkar A, Wang J, Rai R (2019) Deep learning–based stress prediction for bottom-up SLA 3D printing process. Int J Adv Manuf Technol 102(5–8):2555–2569

    Article  Google Scholar 

  • Krish S. (2011). A practical generative design method. Computer-Aided Design, 43(1):88–100

  • Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems, pp 1097-1105

  • Kunakote T, Bureerat S (2011). Multi-objective topology optimization using evolutionary algorithms. Engineering Optimization, 43(5):541–557

  • LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521(7553):436–444

    Article  Google Scholar 

  • Maaten LVD, Hinton G (2008) Visualizing data using t-SNE. J Mach Learn Res 9:2579–2605

    MATH  Google Scholar 

  • Masci J, Meier U, Cireşan D, Schmidhuber J (2011, June) Stacked convolutional auto-encoders for hierarchical feature extraction. In: International Conference on Artificial Neural Networks. Springer, Berlin, Heidelberg, pp 52–59

    Google Scholar 

  • Matejka J, Glueck M, Bradner E, Hashemi A, Grossman T, Fitzmaurice G (2018) Dream lens: exploration and visualization of large-scale generative design datasets. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems, pp 1-12

  • Mathworks [Computer software] (2020) Retrieved from

  • Maturana D, Scherer S (2015, September) Voxnet: a 3d convolutional neural network for real-time object recognition. In 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, pp 922-928

  • Napac (2020) Light alloy wheel categorization by design. Retrieved from

  • Nie Z, Jiang H, Kara LB (2020) Stress field prediction in cantilevered structures using convolutional neural networks. J Comput Inf Sci Eng 20(1):011002

    Article  Google Scholar 

  • Oh S, Jung Y, Lee I, Kang N (2018) Design automation by integrating generative adversarial networks and topology optimization. In ASME 2018 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference. American Society of Mechanical Engineers Digital Collection

  • Oh S, Jung Y, Kim S, Lee I, Kang N (2019) Deep generative design: integration of topology optimization and generative models. J Mech Des 141(11)

  • Pan SJ, Yang Q (2009) A survey on transfer learning. IEEE Trans Knowl Data Eng 22(10):1345–1359

    Article  Google Scholar 

  • Qi CR, Su H, Mo K, Guibas LJ (2017) Pointnet: deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp 652-660

  • Rawat W, Wang Z (2017) Deep convolutional neural networks for image classification: a comprehensive review. Neural Comput 29(9):2352–2449

    Article  MathSciNet  Google Scholar 

  • Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D (2017) Grad-cam: visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE international conference on computer vision, pp 618-626

  • Shea K, Aish R, Gourtovaia M (2005). Towards integrated performance-driven generative design tools. Automation in Construction, 14(2):253–264.

  • Sigmund O (2007) Morphology-based black and white filters for topology optimization. Struct Multidiscip Optim 33(4–5):401–424

    Article  Google Scholar 

  • Simonyan K, Zisserman A (2014) Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556

  • Singh V.Gu, N. (2012). Towards an integrated generative design framework. Design studies, 33(2):185–207

  • Su H, Maji S, Kalogerakis E, Learned-Miller E (2015) Multi-view convolutional neural networks for 3d shape recognition. In Proceedings of the IEEE international conference on computer vision, pp 945-953

  • Sun H, Ma L (2020). Generative Design by Using Exploration Approaches of Reinforcement Learning in Density-Based Structural Topology Optimization. Designs, 4(2)10.

  • Sutradhar A, Park J, Haghighi P, Kresslein J, Detwiler D, Shah JJ (2017, August) Incorporating manufacturing constraints in topology optimization methods: a survey. International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, vol 58110. American Society of Mechanical Engineers, p V001T02A073

  • Umetani N (2017, November) Exploring generative 3D shapes using autoencoder networks. In SIGGRAPH Asia 2017 Technical Briefs. ACM, p 24

  • Umetani N, Bickel B (2018) Learning three-dimensional flow for interactive aerodynamic design. ACM Transactions on Graphics (TOG) 37(4):89

    Article  Google Scholar 

  • Viana FA. (2016). A tutorial on Latin hypercube design of experiments. Quality and reliability engineering international, 32(5):1975–1985.

  • Wang, G. G., & Shan, S. (2007). Review of metamodeling techniques in support of engineering design optimization

    Book  Google Scholar 

  • Williams G, Meisel NA, Simpson TW, McComb C (2019) Design repository effectiveness for 3D convolutional neural networks: application to additive manufacturing (DETC2019-97535). J Mech Des:1–44

  • Zhang Z, Jaiswal P, Rai R (2018) FeatureNet: machining feature recognition based on 3D convolution neural network. Comput Aided Des 101:12–22

    Article  Google Scholar 

  • Zhou B, Khosla A, Lapedriza A, Oliva A, Torralba A (2016) Learning deep features for discriminative localization. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp 2921–2929

Download references


The authors would like to thank Hyundai Motor Company’s Jiun Lee, Sangmin Lee, Min Kyoo Kang, ChangGon Kim, and ChulWoo Jung for their valuable feedback and ideas on our research. We would also like to thank Altair Korea’s Jeongsun Lee and Seung-hoon Lee for their help in automating the CAE process.


This work was supported by Hyundai Motor Company and the National Research Foundation of Korea (NRF) grants funded by the Korean government (grant numbers 2017R1C1B2005266, 2018R1A5A7025409).

Author information

Authors and Affiliations


Corresponding author

Correspondence to Namwoo Kang.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Replication of results

The code used in this paper cannot be disclosed as it is the property of Hyundai Motor Company that funded this study.

Additional information

Responsible Editor: Helder C. Rodrigues

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


Appendix 1.

1.1 Data augmentation effect on autoencoder

The effect of data augmentation was confirmed. We compared the learning curve without data augmentation and the case with data augmentation, as shown in Fig. 27. The loss value for data augmentation is relatively small. We checked the example of reconstruction without data augmentation, as shown in Fig. 28. These images are the same in Fig. 6 but with relatively poor quality.

Fig. 27
figure 27

Comparison of loss value between data augmentation and without data augmentation

Fig. 28
figure 28

Reconstructed wheel image without data augmentation

Appendix 2.

1.1 Detailed algorithm for sorting and grouping points

We introduce a detailed algorithm to select and group each point. The following initial preparations are required: The first coordinate (x0, y0) of array A, where the whole point was stored, is declared as the initial value. The coordinates (x0, y0) used as the initial value (x _ init, y_init) are deleted from array A. An array is created to store the points to be grouped. The initial value (x _ init, y _ init) is stored in the i-th array of the group array. At this time, i is zero. At the end of the initial operation, the for loop is executed as follows.

  1. 1)

    The initial value is declared as a fixed point, and the nearest fixed point is obtained in array A.

  2. 2)

    The closest point is declared as a new initial value and deleted from array A.

  3. 3)

    The distance between the fixed point and the initial value (the closest point to the fixed point) is calculated. If the distance between the fixed point and the initial value does not exceed the threshold, the initial value is stored in the i-th group array. Otherwise, a new (i + 1)-th group array is created, and the initial value is stored in the (i + 1)-th group array. The for loop process is repeated until array A is empty. Any point cannot belong to another group at the same time because it is declared as the initial value and deleted from array A.

For determining the distance threshold, an initial test was conducted at equal intervals of 10 steps from 10 to 100. Thus, all points were completely separated into each spoke-shaped group when the threshold reached 10. A second test was conducted at equal intervals of five steps from 1 to 10 to confirm the precise distance threshold. In the second test, all points were completely separated from the threshold of three. Therefore, the final threshold was chosen as five, which is the next value of the lowest threshold. Figure 29 shows the group’s separation results for each threshold.

Fig. 29
figure 29

Tests for determining the distance threshold

At the end of the above grouping process, all points are sorted in order. We can then take one point of the group array along the same interval and save it as a new array to store fewer points than the existing array. The spline curve becomes a closed curve when the first coordinates of the stored new array are inserted at the end. Closed curves are recognized as surfaces in CAD software, enabling “body” generation. Figure 30 shows examples of spline, reduced point spline, and closed curves.

Fig. 30
figure 30

Example of spline curve

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yoo, S., Lee, S., Kim, S. et al. Integrating deep learning into CAD/CAE system: generative design and evaluation of 3D conceptual wheel. Struct Multidisc Optim 64, 2725–2747 (2021).

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: