Anomaly detection in 3D printers using image processing technology on high-resolution camera images

3D printing technology is rapidly expanding from industrial manufacturing to individual use, transforming production processes. Despite its advantages, such as prototyping and personalized manufacturing, this technology is susceptible to errors during the printing process and often requires human supervision. Errors encountered during production lead to material, energy, and time loss. The primary aim of this study is to detect errors that may occur during the 3D printing process in real time using computer vision and deep learning techniques. In this context, a two-stage fault detection model has been developed for FDM-type 3D printers. In the first stage, the YOLOv11 model is used to segment the printing area, and in the second stage, an EfficientNet-based classification model is employed to detect and classify printing defects. The proposed system aims to improve the printing process by detecting errors at an early stage, thereby reducing material, energy, and time waste while enhancing production efficiency. Real-time experimental results demonstrate that the proposed model achieves high accuracy rates. This study aims to contribute to the literature on anomaly detection in 3D printing and provide a new perspective on optimizing additive manufacturing processes. The developed model can be further trained with larger datasets in future studies, enabling adaptation to different printer types and material variations, thus facilitating its integration into industrial manufacturing processes.