Learning Moisture-Induced Damage From Vision: Diffusion Models for Real-Time Monitoring of Additive Manufacturing...

Area of Science:

• Materials Science
• Polymer Science
• Additive Manufacturing

Background:

• Moisture absorption in hygroscopic polymers like thermoplastic polyurethane causes defects (stringing, pores, bubbles) in additive manufacturing (AM).
• These defects degrade print quality and mechanical properties, challenging AM's reliability.
• Real-time monitoring and integrity estimation are crucial for quality control in AM.

Purpose of the Study:

• To develop an in situ visual monitoring system for fused filament fabrication (FFF) to detect moisture-induced degradation.
• To evaluate the quality and mechanical performance of 3D printed parts using this system.
• To establish a foundation for enhancing AM reliability and sustainability through defect detection.

Main Methods:

• An optical camera-based setup was created for in situ visual monitoring during FFF.
• A diffusion model-based anomaly detection framework was employed to identify moisture-induced degradation.
• The correlation between anomaly scores and mechanical performance was analyzed for nondestructive evaluation.

Main Results:

• The visual monitoring system successfully identified filaments affected by moisture.
• The system assessed the extent of moisture-induced degradation from captured images.
• Anomaly scores from the system showed a strong correlation with the mechanical performance of printed parts.

Conclusions:

• An integrated visual monitoring system with generative AI offers a robust method for enhancing AM reliability.
• This approach enables early, nondestructive detection of defects, supporting resource-efficient sustainability.
• The system provides a pathway for improved quality control and material integrity in polymer-based additive manufacturing.