Anomaly detection of 3D printing process using machine learning

• Main Author: Nur Muizzu Hamzah
• Other Authors: Yeong Wai Yee
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Manufacturing
• Online Access: https://hdl.handle.net/10356/148934
• Institution: Nanyang Technological University

Additive Manufacturing processes are used in various industries and the utilisation of the technologies are growing sharply. Ongoing studies focus on the improvement and advancement of AM processes. However, AM processes have several drawbacks with regard to quality parts and printing repeatability. The occurrence of defects often leads to these drawbacks. This paper aims to develop and implement an in-situ monitoring system on a Fused Filament Fabrication (FFF) 3D printer to detect defects and perform corrections in real-time with the use of an Object Detection model and Computer Vision. Image data of two classes of defects were collected for model training. An object detection model was selected, trained and evaluated against several metrics. The selected model was further optimised to improve the inference speed. Classification accuracy of 89.8% and an inference speed of 70 FPS were obtained. Prior to the implementation of the in-situ monitoring system, a correction algorithm was developed to perform simple corrective actions based on the classification of defects. The implementation successfully demonstrated real-time monitoring and autonomous corrections in a FFF 3D printing process. This implementation will path the way for in-situ monitoring and correction system through closed-loop feedback for other AM processes.