Multimodal Data Fusion for Optimized Rock Fragmentation Simulation——Experimental and AI-Augmented Modeling ComparisonAbstract

(DEM)this work suggests a data-driven method that includes the use of 3D scanning, the discretional element method (DEM) simulation, and machine learning in the problem of insufficient accuracy of the traditional crushing simulation in the small particle size range (<5mm).a DEM model taking into account actual particle shapes was built by performing crushing experiments on typical limestone with a PE150 250 jaw crusher and combining 3D morphology reconstruction via Structure Sensor. To correct errors in the small particle size range, the XGBoost algorithm was also used. It was able to achieve a 41.3% improvement in prediction accuracy for the particle size fraction (error decreased from 58.7% to 17.4%) of the 0–4.75 mm particle size (error reduced from 58.