Abstract: The grinding process is an operation process with high production cost in the concentrator. Its intelligent control has the characteristics of nonlinearity，multiple variables and strong coupling. In order to realize the intelligent control of grinding process，the following methods are proposed for the control objectives and control difficulties of tower grinding classification process. The data-driven model-free adaptive control method has the characteristics of strong robustness，few parameters，and is suitable for the design of control loops in industrial sites. The feed concentration and feed pressure of hydrocyclones are divided into two simple control loops. Based on model-free adaptive control，two single-loop controllers for tower mill classification process control are designed，achieving the goal of accurate control of feed concentration and feed pressure.

Key words: classification process of tower mill, feed concentration, feed pressure, model-free adaptive control