A Data-Driven Approach to Complex Voxel Predictions in Grayscale Digital Light Processing Additive Manufacturing Using U-nets and Generative Adversarial Networks

Published

July 6, 2023

Author(s)

Jason Killgore, Thomas Kolibaba, Benjamin Caplins, Callie Higgins, Jake Rezac

Abstract

Machine learning models such as U-nets like the pix2pix conditional generative adversarial network (cGAN) are shown to predict 3D printed voxel geometry in digital light processing (DLP) additive manufacturing. The models are trained on microscopic voxel geometry data from thousands of pixel interactions, acquired with a high-throughput workflow. The input photomasks have randomly assigned grayscale intensity values for each pixel, providing considerable variation in the print outcomes. Pix2pix considers the spatial interactions between reacting voxels allowing for complex cure depth maps to be predicted. More traditional U-nets without a discriminator block predict lateral extent of cure, but they blur the height predictions and ignore longer range interactions. The trained cGAN performs virtual DLP experiments such as feature size dependent cure depth, anti-aliasing, and sub-pixel geometry control by gray bordering. The pix2pix model is also applicable to larger masks than it is trained on. This allows for over- and under-polymerization of print outputs from larger photomasks to be predicted layer-by-layer. To this end, the model can predict layer-scale and voxel-scale print failures in real 3D printed parts. Overall, machine learning models, exemplified by U-nets and cGANs, show considerable promise for predicting and correcting photomasks to achieve increased precision in DLP additive manufacturing.

Citation

Small

Pub Type

Journals

Download Paper

https://doi.org/10.1002/smll.202301987

Local Download

Keywords

DLP, machine learning, neural network, pix2pix, 3D printing, vat photopolymerization, Stereolithography

Citation

Killgore, J. , Kolibaba, T. , Caplins, B. , Higgins, C. and Rezac, J. (2023), A Data-Driven Approach to Complex Voxel Predictions in Grayscale Digital Light Processing Additive Manufacturing Using U-nets and Generative Adversarial Networks, Small, [online], https://doi.org/10.1002/smll.202301987, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=936540 (Accessed December 3, 2024)