Abstract: Addressing issues such as reliance on manual experience，control lag，and poor stability in traditional grinding processes，this study focuses on the grinding system of Zhongguan Iron Mine.Based on an in-depth analysis of the current process conditions，including uneven ore feeding，load fluctua⁃ tions，and the lack of key parameter measurements，a technical framework for a grinding expert system in ⁃ tegrating "perception-decision-execution"is proposed，leveraging the core theories of expert systems in knowledge representation，reasoning mechanisms，and self-learning.The solution establishes a multi source，high-precision process parameter perception system by introducing advanced detection equipment such as ore AI blockage analyzers，grinding sound spectrum analyzers，Na-22 concentration meters，and laser-based particle size analyzers.Simultaneously，the basic automation system is upgraded to achieve functions such as one-click start-stop and automatic material diversion，providing reliable support for in ⁃ telligent control.For critical processes such as semi-autogenous mill load and concentration，as well as cy⁃ clone classification efficiency，a multi-variable coordinated control model and targeted control strategies are developed.A phased implementation roadmap is designed，encompassing data collection，model con⁃ struction，offline testing，and online deployment.The project is expected to achieve significant results in stabilizing production processes，optimizing process indicators，and reducing energy consumption and la⁃ bor costs，while delivering considerable economic and management benefits.This research holds exempla⁃ry significance and promotional value for advancing the transformation and upgrading of traditional mining industries toward intelligence and digitalization.

Key words: expert system, grinding process, intelligent control, multi-source data fusion, process opti? mization