Gravure Printing Defect Prediction via Machine Learning
Gravure printing is a high-throughput, precision process drawing renewed interest for the production of low-cost, large-area, flexible electronics systems. It is compatible with a wide range of materials, including colloidal inks and low-molecular-weight polymers and capable of sub-5µm features at speeds greater than 1m/s. Like most coating and printing methods, defects manifest in the form of particle aggregation, pinholes, striations, etc., that must be overcome with fine-tuning of the ink/substrate properties and process parameters. This study focuses on applying machine learning algorithms to predict the onset of such defects using an extensive in-house database of thirty-one properties and parameters on over 800 print results. Support Vector Machine and Recursive Feature Elimination approaches are trained with the database and subsequently tested to predict the presence and severity of defects produced. Classification and significance of various properties and parameters yield insight into the development of ideal ink, substrate, and printing parameter combinations. Equally important, this approach helps validate the root-cause intuition of printing process domain specialists.