Data driven framework for defect detection in polymer additive manufacturing

Although additive manufacturing (AM) has enabled the production of intricate and customized polymer components, ensuring part quality and consistency remains a major challenge. Traditional defect detection methods rely on isolated sensing approaches and lack the ability to operate adaptively in real time. While several studies apply machine learning (ML) or sensor-based monitoring, the existing literature remains fragmented, with no unified framework that connects defect mechanisms, data acquisition strategies, ML model selection, and feedback-based process control. The novelty of this study lies in proposing a structured, data-driven conceptual framework specifically tailored for defect detection in polymer AM. This framework synthesizes insights from sensor fusion, signal processing, and machine learning into a coherent, step-by-step structure that has not been formalized in prior reviews. In addition, the paper critically evaluates existing detection methods, identifies unresolved gaps across sensing, data pipelines, and model robustness, and outlines research opportunities such as multi-sensor integration, digital-twin-enabled monitoring, and real-time adaptive control. By positioning defect detection as a closed-loop data-driven system, this work provides a clear roadmap for progressing polymer AM toward more autonomous and reliable manufacturing.