Abstract: In order to solve the problems of traditional low manual control accuracy，lagging parame⁃ ter detection and insufficient automation level in the process of grinding and classification，a molybdenum mine intelligently upgraded the 1# grinding and classification system of 10 000 tons series in the ore dress⁃ ing workshop，and constructed an intelligent grinding control system integrating intelligent sensing instru⁃ ment，multi-modal data fusion and expert system control.The system collects key parameters such as mill load and cyclone overflow particle size in real time through perception equipment such as mill load on-line detection system and pulp particle size on-line analyzer，and integrates soft measurement model to accu⁃ rately correct overflow particle size.Based on the expert system algorithm，the ore feeding amount of the mill，the water feeding amount of the pump tank and the frequency of the classification slurry pump are dy⁃ namically adjusted to realize the adaptive optimization of the whole process of grinding and classification. The application results show that the system reduces the fluctuation of overflow particle size of -74 μm cy⁃ clone by 50%，and improves the efficiency of ball mill by 2.3%. At the same time，the liquid level stabili⁃ ty of the classification pump pool is increased by 37.5%，and the classification pressure stability of the cy⁃ clone is increased by 33%.The system significantly improves the intelligent level and product qualification rate of grinding classification，which can provide practical reference for the intelligent upgrading of the whole process of mine.

Key words: ball mill, grinding classification, mill load, intelligent control, expert system