The design of BOF in Jiangyin Xingcheng Special Steel Co. Ltd is based on static and dynamic model. The BOF end-point carbon content and temperature are predicted by means of BP neural network. On basis of that, the consumption of material, output of static model, are regulated according to the prediction result. The shortcoming that the theory on static model is imperfect is overcome. The end-point hit rate is raised effectually in industrial practice.