Simultaneous Pore Detection and Morphological Features Extraction in Laser Powder Bed Fusion with Image Processing

Area of Science:

• Materials Science
• Additive Manufacturing
• Quality Control

Background:

• Internal pore defects are common in laser powder bed fusion (LPBF) parts, negatively impacting mechanical properties.
• Current supervised models for defect detection are costly due to extensive data, labeling, and computational requirements.

Purpose of the Study:

• To develop a cost-effective method for classifying and detecting pore defects in LPBF parts using optical microscopy (OM) images.
• To improve the accuracy of pore defect detection, especially with limited sample data.

Main Methods:

• Collected extensive LPBF sample data.
• Developed a novel algorithm to detect pore defects and classify them based on morphological features in OM images.
• Evaluated the method's performance against advanced models.

Main Results:

• The proposed method demonstrated superior detection accuracy on pore defect datasets with limited data compared to other advanced models.
• Achieved average accuracy scores exceeding 85% in image segmentation for pore defect detection.
• The algorithm is suitable for rapid and accurate identification of pore defects in OM images.

Conclusions:

• The developed algorithm offers a viable and efficient solution for identifying pore defects in LPBF parts.
• This method can significantly reduce the cost and complexity associated with defect detection in additive manufacturing.
• The findings contribute to advancing deep learning applications in quality control for manufactured parts.