Abstract: The coal blending and coking process ,in which much uncertainty exists, is too complicated to describe with mathematical models ,and it can not be controlled properly by traditional methods. According to the coking theory and using statistical data acquired from industry processes ,a mathematical model is designed and a rule model is proposed with qualitative knowledge derived from experts' experience. The rule model is combined with the mathematical model, and qualitative and quantitative methods are integrated to build up a model for pre- diction of coke quality and to compute the blending ratio of coal. Furthermore the flow of each kind of coal is controlled. Since 1994 the system has been running in a plant with accuracy of coal blending and prediction precision of coke quality up to 97% and 95% ,respectively.