The Research On Sulfur Content Prediction Model For LF

The final component needs to be managed accurately in order to save energy and increase productivity in the process of steel-making. However, the on-line measurement of oxygen and sulfur content in the final component is very difficulty and the cost is also very expensive. Therefore, it is necessary to establish the predictive model for oxygen and sulfur content.The desulphurization predictive model is established based on analyzing the principium of desulfurization in LF refining process. The main tasks are as follows:Firstly, the thesis introduces the principle on deoxygenating technology. Because the mechanism equations contain many uncertain and immeasurable parameters, the thesis selects the method of support vector machine (SVM) for modeling. The paper determines the inputs through mechanism analysis and uses Particle Swarm Optimization to optimize the parameters of SVM.Secondly, thermodynamic principium and dynamic principium desulphurization reaction are introduced. The mechanism model for prediction is established by calculating the values of sulfur capacity. Besides, the thesis establishes a hybrid model which synthesizes mechanism model and intelligence model in order to improve the accuracy. The intelligence model uses Radial Basis Function which is used to provide compensation for mechanism model.Finally, hybrid prediction model is tested by actual data. It takes conclusion from simulation results that the hybrid prediction model is suitable for the prediction of terminal content of sulfur. Therefore, the research in the thesis is of practical value.