In additive manufacturing (AM, or 3D printing), part geometry can differ from the designed dimensions depending on the part size and shape, as well as manufacturing parameters such as the machine used or location of the part within the printer. The relationships between part design, manufacturing parameters, and geometric accuracy are not well understood for AM, and there is a need to develop methods that effectively predict these defects. This tutorial introduces and demonstrates the use of machine learning (ML) to address this need. Using data collected from an AM factory, you will train a support vector regression (SVR) model to predict the dimensions of AM parts based on the design geometry and manufacturing parameters. You will learn to combat ML bias using grid search hyperparameter tuning and nested cross validation, such as k-fold and Monte Carlo subsampling. Finally, you will compare SVR to other ML algorithms, such as k-nearest neighbors (KNN), and evaluate their computational cost and predictive accuracy.

Davis McGregor is a Senior Manufacturing Scientist at Fast Radius Inc., a cloud manufacturing company. He received his Ph.D. in Mechanical Engineering from the University of Illinois Urbana-Champaign in January 2022, co-advised by William King and Sameh Tawfick. His graduate research investigated the quality of additively manufactured parts using advanced computational and statistical tools, including computer vision and machine learning.

Davis J. McGregor, Miles V. Bimrose, Chenhui Shao, Sameh Tawfick, and William P. King (2022), “Using machine learning to predict dimensions and qualify diverse part designs across multiple additive machines and materials,” Additive Manufacturing, 55, 102848, doi: 10.1016/j.addma.2022.102848

Davis J. McGregor, Miles V. Bimrose, Chenhui Shao, Sameh Tawfick, and William P. King (2022), “Code: Using machine learning to predict dimensions and qualify diverse part designs across multiple additive machines and materials,” Mendeley Data, V1, doi: 10.17632/h8tzpxkvdc.1

Researchers should cite this work as follows:

• Davis J. McGregor, Miles V. Bimrose, Chenhui Shao, Sameh Tawfick, and William P. King (2022), "SVR Machine Learning Workshop," https://nanohub.org/resources/svr

• Davis McGregor (2022), "SVR Machine Learning Workshop," https://nanohub.org/resources/svr. (DOI: 10.21981/SSXB-QY97).

  BibTex | EndNote

1. additive manufacturing
2. design
3. Jupyter notebooks
4. machine learning
5. manufacturing
6. NCN Group - Machine Learning Hands-on
7. python
8. regression
9. support vector regression (SVR)
10. SVR Machine Learning Workshop