Three-Dimensional Printing Quality Inspection Based on Transfer Learning with Convolutional Neural Networks.

Area of Science:

• Additive Manufacturing
• Computer Vision
• Machine Learning

Background:

• Fused Deposition Modeling (FDM) is a widely used additive manufacturing technique for creating 3D objects layer by layer.
• Despite its maturity, FDM printing is susceptible to various defects that can compromise part quality and functionality.
• Automated quality inspection is crucial for ensuring the reliability and efficiency of FDM processes.

Purpose of the Study:

• To develop an automated, image-based quality inspection system for FDM-printed 3D objects.
• To evaluate the effectiveness of combining transfer learning and ensemble learning for defect detection in FDM parts.
• To assess the impact of object geometry and color variations on the inspection accuracy.

Main Methods:

• Utilized transfer learning with pre-trained models (VGG16, VGG19) as feature extractors.
• Implemented ensemble learning by combining multiple models to enhance classification accuracy.
• Applied the developed model to inspect the quality of FDM-printed objects with varying geometries.

Main Results:

• The combination of transfer learning and ensemble learning demonstrated high accuracy in quality inspection.
• Model combinations using VGG16 and VGG19 achieved the best performance across most scenarios.
• Classification accuracy was largely unaffected by variations in print color.

Conclusions:

• Transfer learning combined with ensemble learning provides an effective and robust method for FDM 3D print quality inspection.
• This approach significantly reduces material waste and inspection time.
• The developed method contributes to improving the overall quality and reliability of 3D-printed components.