Study On The Model Of Alloy Optimization Control System For LF Furnace Based On Data

Ladle furnace(LF)is the key equipment of secondary refining in iron and steel smelting,which plays an important role in improving the quality and production efficiency of the enterprise.As the core step of LF refining technology,the narrow composition control of alloy elements should not only ensure the alloy content in molten steel meet the expectation,but also control the content of impurity elements within the standard range.According to the batching scheme with the minimum cost and the minimum impurity elements in the added materials,the narrow composition control effect of alloy elements in refined steel can be effectively achieved,After improving the proportioning scheme,the cost of steelmaking enterprises can be reduced and the production efficiency of metallurgical enterprises can be improved.Based on the research background of LF furnace feeding system in an iron and steel enterprise,the actual smelting data are obtained by combining the ERP system and PLC control system of the company.Meanwhile,according to the production process,the key and difficult points of alloy narrow composition control in LF refining process are analyzed,and the alloy feeding model of LF furnace is developed and established1?In this paper,the prediction model of alloy element yield is established by using support vector regression(SVM),and the model is compared with the model established by BP neural network.The results show that the prediction results of SVM algorithm are more consistent with the expected results,and the prediction effect is better.However,with the decrease of sample data,the prediction stability of the model fluctuates greatly.To solve this problem,this paper selects grey wolf optimization algorithm(GWO)to optimize the two important parameters of SVM model,namely penalty factor and kernel function parameter g,to improve the prediction effect of the model.2?Grey wolf optimization algorithm(GWO)is used to optimize the selection of penalty factor C and kernel function parameter g of SVM model.In order to compare the optimization effect of GWO algorithm,cross validation(CV)method is used to optimize the selection of SVM model parameters.Gwo-svm model and cv-svm model are established respectively,and the two optimized models are compared and analyzed through many simulation experiments.The experimental results show that the performance indexes(root mean square error RMSE,mean percentage error MAPE,coefficient of determination)of the optimized model(gwo-svm)are better than cv-svm model.Gwo-svm model has the highest prediction accuracy of alloy element yield,the best stability at the same time,and the best generalization ability and fitting ability.3?The alloy feeding model was established.Based on the method of multi-objective programming and gwo-svm model,an alloy feeding model for LF furnace with the minimum cost and the minimum impurity element content in the added materials is established.The model can automatically complete the setting of alloy feeding and improve the accuracy and efficiency of feeding.Finally,the feasibility of the system is verified by running the alloy charging model of the model LF furnace with the actual smelting system.Compared with the charging combination in the field,the model proposed in this paper is better than the actual situation in minimizing the charging ratio.