I. Introduction
In mineral processing, crushing and grinding always account for over 60% of electricity consumption in concentration plants. At the same time, variations in mill feed properties such as composition, size distribution, and grindability are mainly caused by the heterogeneous nature of the ore bodies, which leads to a changeable operational strategy, especially in the SAG feeding plan[l]. Nowadays, rock lithological analysis mainly consists of two methods: elemental composition analysis and the drop-weight test [2]. However, these analysis results still lag behind SAG process control. Operators judge the rock lithology from observations on on-site monitors, based on rock color and texture, to control the feeding. A real-time rock lithological detection method needs to be developed for SAG circuits to improve their energy utilization efficiency.