The invention discloses a decision-making composition method for a blast furnace iron smelting expert system. Taking individual decision-making for multiple types of blast furnace expert systems as input, the invention establishes four types of decision-making composition operator modules, in which linear and non-linear regress technologies are used for performing parameter identification for the decision-making composition operator module; in selection reference of characteristic vectors of blast furnace conditions and history behaviors of different types of blast furnace expert systems, the invention converts the selection problem of optimization in decision-making composition operators into a classification device construction problem; a non-linear multi-classification supports the vector machine module to perform modeling for classification devices, so that an optimized classification device can be gained through a quick classification device training arithmetic; furthermore, the classification device is used for performingl increment secondary correction for decision-making composition operator module parameters; the optimized classification device can carry forth dynamic selection for optimized decision-making composition operators adapting to present furnace conditions, so as to optimize decision-making in blast furnace operation. The intelligent control on blast furnace smelting processes, together with the decision-making composition have common universality, so as to improve decision-making accuracy and system robustness of blast furnace expert systems.