Abstract
Internal defects commonly occur during the 3D printing process of Polylactic Acid (PLA), and significant challenges remain in detecting and extracting these defects, as well as understanding the relationship between defects and material fatigue life. This research proposes the Chroma-YOLO Enhanced Integrated Framework, an improved YOLOv11n-based model that integrates HSV defect extraction module and a random forest prediction model. Comprehensive ablation experiments demonstrate that the Chroma-YOLO model achieves significant improvements of 6.9% and 7.3% for mAP50 and mAP50-95 metrics, respectively, compared to the baseline YOLOv11n model, confirming substantial enhancements in feature extraction capability and target localization accuracy. Furthermore, this framework establishes a comprehensive model from defect detection to fatigue life prediction by combining the HSV color space-based defect detection technique with the random forest machine learning algorithm. The random forest-based predictive model achieves a remarkable accuracy of 96.25% and 99.09%for the test and validation set, respectively, for fatigue life prediction of 3D-printed PLA, which shows significant improvement compared to the conventional prediction methodologies.