K-means clustering for activity-based classification of metallurgical slag

The efficient utilization of metallurgical slag plays a crucial role in promoting a circular economy. However, conventional activity evaluation methods are constrained by single-indicator assessments, long testing cycles, and limited adaptability to multi-source heterogeneous samples. To address these challenges, this study proposes an intelligent activity classification framework that integrates K-means clustering with principal component analysis (PCA) for automatic grouping and quantitative activity identification of metallurgical slag. A total of 83 steel slag and blast furnace slag samples were collected from domestic and international steel mills through field sampling, literature review, and statistical compilation. An unsupervised clustering algorithm was applied in conjunction with principal component dimensionality reduction and weighted fusion strategies to construct a Comprehensive Activity Index (CAI). The clustering performance was further improved through optimization of key chemical features. Kruskal–Wallis tests revealed statistically significant differences in CAI distributions among clusters, confirming the reliability of the proposed classification approach. Based on the statistical distribution of CAI, quantitative classification criteria were established, where steel slag can be categorized as low activity (CAI ≤ 0.1), medium activity (0.1 < CAI ≤ 1.0), and high activity (CAI > 1.0); and blast furnace slag categorized as low activity (CAI ≤ −0.8), medium activity (−0.8 < CAI ≤ 0.4), and high activity (CAI > 0.4). Experimental verification further demonstrated the robustness and generalizability of the intelligent classification algorithm. The proposed method provides a scientific and efficient approach for the rapid screening and targeted utilization of multi-source metallurgical slags, thereby offering significant practical value for advancing solid waste resource recovery in the steel industry.