Abstract: To address the issue in the grinding process where the core disturbance—ore grindability—cannot be measured
online,leading to production control that relies on subjective experience,delayed adjustments,and high energy consumption,
this paper proposes a novel online identification method based on the inversion of a coupled mechanistic model. The core
innovation of this method lies in reducing the dimensionality of the high-dimensional selection function matrix,which describes
ore breakage characteristics,into a dimensionless " online grindability index" (Kg ) with clear physical significance through
parametric modeling. This transforms a complex ill-posed inverse problem into a robust single-variable parameter optimization
problem. By utilizing easily accessible real-time measurements from the Distributed Control System (DCS)—specifically,mill
power and circulating load—as observations,the method dynamically estimates the optimal Kg value by minimizing the deviation
between the mechanistic model predictions and actual measurements online. Furthermore,standardized operating procedures
(SOPs) are established based on the identified Kg to form a human-machine collaborative operation strategy. Dynamic
simulation verification demonstrates that the proposed method can accurately and rapidly track changes in ore grindability.
Compared to the experience-dependent operation mode,applying the proposed strategy can reduce the maximum deviation of
product particle size (P80) by 73. 4%,shorten the adjustment time by 66. 7%,increase the particle size qualification rate from
68% to 92%,and improve system throughput by approximately 3. 7% when the ore is easily grindable,achieving significant
quality improvement,production stabilization,and efficiency enhancement. This study provides an effective theoretical solution
and technical pathway for transforming the grinding process from passive,lagging control to active,predictive optimization control.