Machine Direction Registration Modelling in Roll-to-Roll Screen Printing by Deep Learning

• Published in: 2024 IEEE 18th International Conference on Advanced Motion Control (AMC) pp. 1 - 6
• Main Authors: Gafurov, Anton Nailevich, Kim, Daehyeon, Choi, Yuchang, Park, Hyejin, Kim, Inyoung, Oh, Dongho, Lee, Taik-Min
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28.02.2024
• Subjects: Artificial neural networks; deep neural network; Predictive models; printed electronics; Printing; Process control; Production; registration; roll-to-roll; Root mean square; screen printing; Training
• ISSN: 1943-6580
• DOI: 10.1109/AMC58169.2024.10505681

This study explores roll-to-roll (R2R) screen printing, a foundational process in the large-scale production of flexible electronics. By employing a deep neural network (DNN), we investigate the factors influencing layer registration, focusing on web transporting and screen printing parameters. We analyze various screen printing factors, such as printing gap, pressure, and speed, as well as web transporting factors, such as feeding speed, rewinder roll diameter, and active compensation gain, and their contributions to registration error. The registration was investigated in the machine direction as point accuracy and precision, namely mean and standard deviation, respectively. Leveraging machine learning, we develop a predictive model capable of approximating the relationship between printing, web handling parameters, and registration. The optimized DNN model demonstrates the performance of root mean square error (RMSE) of 2.69 and 2.63 for training and validation sets, respectively. The DNN model showed a highly nonlinear approximation compared to a simple linear regression model. These results highlight the model's potential for optimizing critical processing parameters for enhanced registration in R2R screen printing systems.