Optimization Of The Refining Slag And Temperature Model Prediction Of LF

With the technology progress in the steelmaking and the high requirement of the properties and clearity of the steel, optimization of refining slag has been become a focused technique in the secondary refining process. Researching on the proper composition of refining slag in accordance to different facility of second refining bears important engineering and exoterical values. Meanwhile, the taping temperature of the steel is an important factor. It is an important to predict accurately the temperature of molten steel in LF.According to the desulphurization dynamics and thermodynamics in LF, based on the predecessor research work, orthogonal design method was usd to investigated the melting point and desulphurization ability of the slag. The temperature prediction model of molten steel in Ladle Furnace (LF) was established with back-propagation neural network, six factors which influence the temperature of molten steel in LF were established.The results can be summarized as:(1) The orthogonal design experiment revealed that the significant effect on the melting point is slag basicity, CaF2 contents. The melting point as the only inspection target, the best slag composition is slag basicity 2.5, MgO 10%wt, CaF2 contents 8%wt in slag.(2) Desulfurization experiment in the laboratory indicated that the maximum desulfurization rate of 91.6%wt. The desulfurization rate as the only inspection target, the major factor affecting the desulfurization rate is slag basicity (R), the sub major factor Al2O3 and CaF2 contents, while MgO contents had few influence. (3) BP neural network was used to learn nonlinear system, and obtained a good result. In the LF refining process of physical and chemical reactions and heat transfer is very complicated. So, it's difficult to accurately describe the temperature with the mathematical equations. In addition, the predicted targets is nonlinear parameters in the steelmaking process, it is very difficult to establish an accurate mechanism model.(4) Considering the accuracy of calculation, the additional momentum algorithm and Levenberg-Marquardit were used to train BP network. The main factors affecting the temperature of molten steel in LF were molten steel weight, heating power, the amount of added alloy, slag thickness, argon flow rate and initial temperature of molten steel. And the program of the temperature forecast was raised.