State recognition of light radiation of BOF end-point based on fuzzy support vector machine

In view of the end point of BOF smelting, there are many uncertain and unavoidable errors in the traditional judgment of flame by human eye. A method of BOF endpoint estimation was studied by recognition of light radiation with fuzzy support vector machine. A non-contact system was designed for light radiation acquisition of furnace mouth. Based on the analysis of the radiation, three parameters characterizing the overall spectral fitted by Gauss function and two parameters corresponding to emission peaks were extracted respectively and then used as inputs of support vector machines. Oxygen consumption, oxygen gun vibration amplitude, oxygen gun vibration time and feeding quantity of the process of production were chosen to construct subsample, the membership factors were calculated and the prediction model was built by using fuzzy support vector machine. The experimental results show that the proposed method has better recognition accuracy than the manual method and the traditional SVM method, and can provide reference for converter operator to determine the end point accurately.