This article introduces an arithmetic approach to establish an optical system aiming to eliminate the shortcomings in the steel making technique. The current basic oxygen furnace technique is flawed because the level of carbon content in the molten iron is estimated by the workers observation through experience. It is hard to compare, measure and control. The proposed model, on the other hand, is much more computable: relevant data is collected from spectrum distribution during the process of basic oxygen furnace. It predicts the end-point of BOF relatively accurate because spectrum is quantifiable, and the changing process of the furnace flame is essentially the changing process of the spectrum. In this model, we can measure the temperature of the molten steel by the Flame emission spectroscopy theory principle. Further more, the result of the experiment conducted based on the model shows that the model meets the requirements of endpoint judgment online.