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

Selective laser melting (SLM) is a key technology in metal additive manufacturing (AM), but the widespread presence of porosity defects in fabricated parts significantly degrades mechanical performance and limits practical applications. Machine learning (ML) and deep learning (DL) have shown great potential in porosity prediction, defect detection, and performance modeling. However, their application remains constrained by the lack of systematic "processes-images-properties" datasets and the high cost of experimental data acquisition. To address this challenge, this study proposes an innovative normalization method for multi-source SLM porosity images (NMI). The method integrates scale bar detection and removal, physical size normalization, and resolution harmonization to ensure dimensional consistency while preserving critical pore features. Systematic validation using both literature-derived and experimental datasets demonstrates that NMI effectively integrates heterogeneous image data, enhances dataset consistency, and promotes the reuse of existing imaging resources. This framework provides a scalable and resource-efficient pathway for DL-based defect prediction and process optimization, and establishes a solid foundation for constructing standardized and extensible materials datasets.