Multi-Source Porosity Image Normalization (NMI) in Selective Laser Melting for Reliable Reuse of Heterogeneous...

Area of Science:

• Materials Science
• Additive Manufacturing
• Data Science

Background:

• Selective laser melting (SLM) is crucial for metal additive manufacturing (AM), but porosity defects hinder mechanical properties.
• Machine learning (ML) and deep learning (DL) show promise for porosity analysis, yet lack standardized datasets.
• High costs of experimental data acquisition limit the development of ML/DL models for AM.

Purpose of the Study:

• To develop an innovative normalization method for multi-source SLM porosity images (NMI).
• To address the challenge of heterogeneous image data and high experimental costs in AM.
• To facilitate the creation of robust datasets for ML/DL-based defect prediction and process optimization.

Main Methods:

• Proposed an innovative normalization method for multi-source SLM porosity images (NMI).
• Integrated scale bar detection/removal, physical size normalization, and resolution harmonization.
• Validated the method using literature-derived and experimental datasets.

Main Results:

• NMI effectively integrates heterogeneous image data from various sources.
• The method enhances dataset consistency and preserves critical pore features.
• Demonstrated improved data integration and resource reuse for ML/DL applications.

Conclusions:

• The NMI framework offers a scalable and resource-efficient pathway for DL-based defect prediction in SLM.
• Established a foundation for standardized and extensible materials datasets in AM.
• Enables more effective utilization of existing imaging resources for process optimization.