Machine Learning‐Based Optimization of Pixel Light Intensities for Improving Polymerization Accuracy in Digital Light Processing 3D Printing

Abstract Digital light processing (DLP) 3D printing shows a significant increase in industrial applications due to its ability to fabricate complex parts with fine features. However, the printed sample suffers from oversizing and loss of small features because the solidification of photocuring‐resin beyond the targeted regions. To improve printing accuracy, there are significant efforts dedicated to modeling and optimizing the printing process. Nevertheless, most works do not accurately model the chemical reaction throughout the DLP, as doing so requires expensive computational cost. As a result, they are unable to capture the root cause of overcuring. In this work, a machine learning (ML) framework is presented to optimize light patterns to improve the printing accuracy. A reduced‐order model that can accurately track the evolution of the degree of conversion (DoC) of photocuring as a function of light dose in 3D space at low computational cost is introduced. Then, it is included in an ML to optimize light patterns for producing the DoCs that yield targeted shapes. The ML is applied to two resin formulations and complex geometries, demonstrating its robustness. The speed and accuracy of this approach pave the way for advanced applications of 3D printing, such as real‐time digital twinning.