Establishing And Optimizing Of Static Control Models For Vanadium Extraction By Converter Process

The vanadium-titanium magnetite is the main resource containing vanadium whichis extracted along with iron. The vanadium containing hot metal is oxidized to producesemi-steel and vanadium slag in the converter. Through the processes of roasting,leaching and precipitation, the vanadium can be extracted from vanadium slag as V₂O₅.Actually, vanadium extraction in converter is a key step in the whole process. Thequalities of vanadium slag and semi-steel directly affect the follow-up process ofvanadium extraction and steelmaking process. Controlling the smelting temperature ofconverter and the oxidation degree of elements, improving the extraction rate ofvanadium, and ensuring the better quality of semi-steel, are the main purposes of theend point control. Therefore, besed on the research on the cooling effect of the coolants,the influence of anthracite on endpoint and the balances of material, heat and vanadiumduring vanadium extraction, a static control model for vanadium extraction-by converterprocess was constructed and about to be used in Pan-steel.In this work, the cooling effects of the coolants under different temperatures havebeen studied by both theoretical calculation and DSC experiment. The result showed，that the capacity of the pellet was3.6,1.8,1.7times larger than the pig iron, vanadiumslag and the residual slag respectively. Taking the cost and the cooling capacity intoconsideration, the pellet is suggested to be used as the main coolant in vanadiumextraction by converter.The experimental results showed that, adding1kg anthracite in1ton hot metal,0.02%carbon content would be increased, and6.9℃temperature would be enhanced. Ithad no effect on TFe content in the slag by adding anthracite, but might increase thecontent of V in the semi-steel.The oxidation rate of vanadium in the converter was66.10%, which was still lowfor the process. A better process was thus needed to reduce the residual vanadium in thesemi-steel, and which would be beneficial to improve the oxidation rate. The lost of ironin the process is⁵.24%. The thermal efficiency of the converter is⁹3.90%. Tappingthe vanadium slag for several heats in converter would contribute to improve thethermal efficiency of the converter.Combined with the reaction mechanism, the Coolant Control Model, OxygenContol Model and Forecast Model have been established using BP neural networks with the MATLAB software. In order to simplify the parameters and make the model moreaccuratecy, the reaction mechanism has been added into the model. According to thecharacteristics of the process, the data are systematically analyzed using the clusteringalgorithm, then the model has been trained and tested. The result showed that the testerror of the Coolent Control Model is13.7%, and that of the Oxygen Contol Model is7.04%. The two models were proven to be easy to control, and have good stability. Theprediction error of the endpoint temperature and C content in the Forecast Model were0.97%and2.61%respectively, and which was lower than V content prediction error.The model parameters were optimized, to achieve the optimized with the geneticalgorithm. The intermediate layer neuron number of the models has been optimized bytraining. The result indicated that the intermediate layer neuron number of the CoolantControl Model and Oxygen Control Model were15and24respectively, while25wasthe best choice in Forecast Model.