Development and application of an instrument for microstructure matrix inclusion distribution analysis in oversized metallic materials.

To address the urgent need for inclusion analysis in clean steel production, this study develops an automated detection system for large metallic samples. Integrating a high-precision CNC stage, multi-unit microscopic imaging, laser spectroscopy, and a YOLOv11-based deep learning model, the system enables full-area rapid scanning of meter-scale samples. It automatically identifies and classifies inclusions (types A-D) with significantly improved efficiency-over 20 times faster than conventional methods. Experimental validation on automotive sheet samples successfully characterized 533,041 inclusions in size, distribution, and composition, while directly locating the largest inclusions, overcoming the limitations of small-area extrapolation.