Design and Implementation of a Machine Learning-Based Layer Uniformity Management System for Dual-Layer Coextruded Protective Films

Dual-layer coextruded protective (CP) films are widely used in manufacturing, processing, and transportation to protect products, such as backlight modules, polarizing plates, and consumer electronic device casings from contamination and impact damage. These films consist of a base layer (BL) and an adhesive layer (AL) made from polymers, and a major quality control (QC) challenge is maintaining a uniform thickness across these layers to prevent uneven adhesion or breakage. This article proposes a layer uniformity management (LUM) system, which serves as a machine learning-enhanced solution designed to streamline the inspection, recording, and tracking of the layer thicknesses at various positions across dual-layer CP films. A film cross-sectional sampling (FCS) device, a digital microscope sensor, and a computing and information management (CIM) platform are integrated into the LUM system. The FCS device produced with 3-D printing technology ensures consistent sample preparation and image quality during microscopy. The CIM platform features a layer thickness ratio measurement (LTRM) module, a data visualization (DV) module, and a database. The LTRM module utilizes image processing techniques combined with a line segment clustering (LSC) algorithm to determine the relative thicknesses of the base and ALs in the sampled images. The DV module presents these data visually and stores them in the database. According to experimental results, this system achieves an overall 0.97 accuracy in estimating the layer thickness ratio of films, thus enhancing the standardization and digitization of quality inspection processes for CP films, reducing the inspection time, and facilitating digital transformation in product quality management.