Abstract: Multi-model integration method combining the mechanism knowledge with the intelligent error compensation technology is used to build a global model to predict the alumina raw meal slurry quality.Blending expert database is established according to the quality index influence coefficient and classification rules.Based on this,a multi-objective hierarchical reasoning strategy is proposed and the ratio optimization setting is realized.This optimization control method has been successfully applied to the raw meal slurry blending process in an alumina plant,and the results verify that the proposed method can resolve the optimization problem of a kind of industrial processes characterized by time delay and multi-constraints.