Development of AI-Based Monitoring System for Stratified Quality Assessment of 3D Printed Parts

Abstract

The composite material layering process has attracted considerable attention due to its production advantages, including high scalability and compatibility with a wide range of raw materials. However, changes in process conditions can lead to degradation in layer quality and non-uniformity, highlighting the need for real-time monitoring to improve overall quality and efficiency. In this study, an AI-based monitoring system was developed to evaluate layer width and assess quality in real time. Three deep learning models Faster Region-based Convolutional Neural Network (R-CNN), You Only Look Once version 8 (YOLOv8), and Single Shot MultiBox Detector (SSD) were compared, and YOLOv8 was ultimately selected for its superior speed, flexibility, and scalability. The selected model was integrated into a user-friendly interface. To verify the reliability of the system, bead width control experiments were conducted, which identified feed speed and extrusion speed as the key process parameters. Accordingly, a Central Composite Design (CCD) experimental plan with 13 conditions was applied to evaluate layer width and validate the system's reliability. Finally, the proposed system was applied to the additive manufacturing of an aerospace component, where it successfully detected bead width deviations during printing and enabled stable fabrication with a maximum geometric deviation of approximately 6 mm. These findings demonstrate the critical role of real-time monitoring of layer width and quality in improving process stability and final product quality in composite material additive manufacturing.