An effective detection model based on YOLO for pore defects in additive manufacturing.

Area of Science:

• Materials Science and Engineering
• Additive Manufacturing
• Quality Control

Background:

• Microscopic imaging is vital for assessing selective laser melting (SLM) quality.
• Manual inspection methods for SLM defects are inefficient and lack reproducibility.
• Automated defect detection is needed for improved SLM quality control.

Purpose of the Study:

• To develop a synergistic method for analyzing pore defects in SLM microscopic images.
• To integrate advanced image segmentation with polynomial fitting for defect quantification.
• To establish a correlation between SLM process parameters and pore defect formation.

Main Methods:

• A high-performance image segmentation model with adaptive learning rates and attention mechanisms was designed.
• The model automatically quantifies pore defect pixel proportions in micrographs.
• Chebyshev polynomials were used to fit the relationship between process parameters and porosity.

Main Results:

• The proposed method demonstrated excellent performance in defect detection and quantification on an SLM pore defect dataset.
• The study successfully calibrated and quantified defect information from microscopic images.
• An in-depth analysis revealed the intrinsic correlation between process parameters and pore defect formation.

Conclusions:

• The developed automated tool effectively detects and analyzes pore defects in SLM.
• This method enhances the quality assessment and control of the selective laser melting process.
• Understanding the process-defect relationship aids in optimizing SLM parameters to minimize porosity.