447 Figure 22. Descriptive statistical analysis of operational responses by cluster in milling Phase 1. 5. CONCLUSIONS Based on the results obtained, the following conclusions are drawn: The proposed Deep Clustering framework (CVAE + GMM) effectively captures the nonlinear, multivariate, and temporally dependent behavior of SAG mill feed, providing a robust representation of complex ore–process interactions. The integration of geological, geochemical, geomechanical, and operational variables enables a unified and interpretable characterization of the material, overcoming the limitations of traditional linear approaches. The 4-cluster configuration successfully identifies distinct and persistent operating regimes, each associated with specific mineralogical composition and comminution response, demonstrating temporal stability and physical consistency. A critical operating condition (Cluster 1) was identified, characterized by highly competent and abrasive material, reduced throughput, lower fines generation, and increased pebble production, representing a low-frequency (3.2%) but high-impact regime. The methodology provides a solid basis for the implementation of adaptive and proactive control strategies, enabling improved circuit stability, energy efficiency, and throughput, and establishing a scalable tool for advanced geometallurgical decision-making.
RkJQdWJsaXNoZXIy MTM0Mzk2