Automated defects detection in extrusion 3D printing using YOLO models

Fused Deposition Modelling (FDM) three-dimensional (3D) printing technique has experienced significant advancement, enabling the design and creation of complex geometrical objects with a variety of materials. However, it still faces critical challenges from extrusion defects such as under-extrusion, over-extrusion, stringing, and spaghetti, which compromise the quality and efficiency of 3D-printed products. This study aims to develop a robust, intelligent defect detection system to enhance quality control during FDM printing. Based on artificial intelligence-augmented additive manufacturing (AI2AM) and using advanced YOLO models: YOLOv11, YOLOv10, YOLOv9, YOLOv8, and YOLOv5, we evaluated their performance based on precision, recall, F1-score, and mean Average Precision (mAP) across Intersection over Union (IoU) thresholds. Results revealed that YOLOv11s outperforms other models, achieving mAP@0.5 of 0.8308 and mAP@0.5:0.95 of 0.5361, while YOLOv11m balanced precision 0.9128 and recall 0.8990 effectively, demonstrating their suitability for real-time applications. The findings demonstrate the superiority of YOLO models in improving detection reliability, minimizing material waste, and streamlining FDM workflows, with YOLOv11 models setting new benchmarks for defect detection in additive manufacturing.