The characteristics of the knowledge discovery for basic oxygen furnace(BOF) steelmaking are analyzed by using the Rough Set Theory. The production data of BOF steelmaking are preprocessed by using the methods of data withdrawal, standardization, discretization and so on. The main influencing factors of steelmaking production are set as the knowledge discovery property. The endpoint control objectives of BOF steelmaking are used as the decision attribute of knowledge discovery. Then the knowledge discovery model of BOF steelmaking based on rough set theory is established, which makes the automation of the production knowledge discovery, access and rule extraction come true. The model is tested by using the production data of 210 t BOF, and takes the temperature variation of smelting endpoint as the decision attribute. The results show that the influencing factors, such as silicon content, iron ore weight, oxygen consumption and so on, are of very importance to the endpoint temperature of molten steel. Besides, the rules of molten steel temperature extracted by the model vary with current converter steelmaking process, which proves the validity of the model.