Grinding process is a key link in modern industrial production, which directly affects the quality and various performance indicators of the products (LI Peiran et al. 2008).Taking automatic control for cement raw mill load not only can save large amount of electricity, also can increae the production. However, the grinding of the cement raw meal is a complex production process, which is nonlinear, big lag and difficult to establish mathematical model. Using some advanced or routine algorithm such as PID control algorithm can’t solve above problem completely.We summarized an expert control system according to experience through long-term practice and theoretical analysis, which is based on three parameters including mill sound, the back powder of the separator and the power load of the bucket elevator. This system has nine rules and divides the grinding process into three stages: “empty mill”, “normal” and “blocked mill”. Different control scheme will be implemented according to different stage and different control rule. Practice shows that the overall efficiency of this control algorithm can reach about 80% and it can avoid appearing “empty mill” and “blocked mill” phenomenon. In order to increase the efficiency, turn the unilateral control algorithm for each section into fusion intelligent control,the efficiency can be improved to about 85%. According to the different error value of the three parameters in grinding process, we take variable period sampling on the basis of the intelligent control, the operation

efficiency of the mill can be further improved to about 90%.