A Vision Based System for Anomaly Detection and Classification in Additive Manufacturing

This paper presents a method to install a vision-based system on 3D printers for the purpose of detection and classification of anomalies during the printing process. Recently, 3D printing has becoming popular in manufacturing industry. Hence, quality monitoring is introduced to inspect the printing process to ensure minimal defects occurred and reduce wastage of printing materials if defects occur. This project is performed using Polylactic Acid (PLA) material as the printing material. The printing process is segregated into several checkpoints to capture the top surface of the product. Several popular image processing techniques were reviewed and employed in this paper. Images taken using Raspberry Pi are processed and classified into defective and non-defective groups using Python OpenCV. Hu moments analysis is carried out to extract the information on binary features of captured images and template images. Several shape distance formulae and also combination of different Hu moments factors is employed to enhance the rate of successful detection. The detection results are presented in detection classification matrices, while detection success rate and false alarm rate are calculated.