Study On The Optimum Model Of Intelligent Smelting Process In Electric Arc Furnace

Electric arc furnace,using electricity as heat source,can produce different kinds of steel which have restrict requirements on chemical composition and mechanical performances,such as special tool steel,and high alloy steel.Compared with converter steelmaking,arc furnace steelmaking has several advantages,such as controllable atmosphere in furnace,flexible temperature control of molten steel,high thermal efficiency,simple equipment and short technological process.In addition,the use of scrap as raw material for arc furnace steelmaking,effectively solve the problem of pollution and steel ore tight scrap.With the increasing awareness of environmental protection,EAF steelmaking is more competitive.The research background is an arc furnace smelting process dalipal company.Using BP neural network algorithm,the simulation platform is designed by Visual Basic 6.0 programming language.Optimal blending model,temperature prediction model of molten steel,and the end composition prediction model intelligent control model are established,which realised man-machine integration.The establishment of the element yield dynamic database,real-time and accurately obtain the elements of different kinds of steel yield,and ensuring the premise of tapping component,with the minimum cost principle,established the optimized model of ingredients.Energy balance model is established to provide theoretical basis for later temperature prediction model.The process of the forecasting model,through the analysis and the improvement of the algorithm,determine the model algorithm,and the parameters of the model are analyzed.based on the actual situation,the model of the network structure,the input and output of parameters to determine the final.data screening module is added,to test the model and improve the accuracy of model prediction.This thesis took 2000 heats data dalipal company as research data prediction model,through the data pretreatment stage,the final 1550 heats training samples and 50 testing samples of heats.The end point prediction model and the molten steel temperature prediction model are verified.Through the training and testing of the model,the following conclusions are drawn:(1)Using the batch optimization model can make the furnace charge per ton cost decreased by 2.46%to 7.65%,with an average reduction of 5.31%,and effectively guide the rational distributing field staff,control the shape and size of charge,reduce pressure feeding time.(2)To verify the terminal composition prediction model,the mean square error that mass fraction of C at the end of the element is 0.0418%,the minimum error and the actual value of the forecast value is 0.0045%,the maximum error is less than 0.079%,the error value is less than 0.02%of the hit rate is 66%,the error value is less than 0.05%hit rate to 94%.The mean square error of the mass fraction of P at the end of the element is 0.0033%,the minimum error and the actual value of the forecast value is 0.0009%,the maximum error is less than 0.0057%,the error value is less than 0.003%of the hit rate is 84%,the error value is less than 0.004%hit rate to 94%.(3)To verify the molten steel temperature prediction model,the mean square error is 0.0362%,the error absolute value of the furnace is less than 2?,and the error absolute value of the 96%furnace times is less than 12?.Finally,the causes of the error of the model are analyzed and discussed.Through the verification of the model,this model has some rationality and guiding significance for practical application.