| The development of iron and steel industry is one of the important reference standards for China’s GDP development.The blast furnace ironmaking process is an important process of iron and steel enterprises.The blast furnace ironmaking process plays an important role in the development of large iron and steel enterprises in China and in the control of resources,environment and energy.The temperature state of blast furnace is an important index reflecting the condition of blast furnace.The drastic change of furnace temperature will lead to abnormal state of blast furnace,which will affect the quality of pig iron production.Therefore,effective control of blast furnace temperature to keep it in a reasonable range is of great significance in the actual ironmaking process.In practical research,it is difficult to measure the temperature of blast furnace,and the silicon content in hot metal of blast furnace has long been an important factor for domestic and foreign scholars to study the temperature state of blast furnace.Therefore,the silicon content in hot metal of blast furnace system is usually measured to replace the temperature of blast furnace for research and analysis.Establishing an accurate mathematical model for predicting silicon content in molten iron is of great reference value and practical significance for guiding blast furnace operators to control furnace temperature,for the development of economic benefits of iron and steel enterprises,and for how to control fuel for energy saving and emission reduction.On the basis of previous research on the temperature state of blast furnace,this paper introduces two statistical methods,which are AdaBoost and Support Vector Machine(SVM),as the main tools of this study.The blast furnace production data of Baotou Iron Steel Group and Laiwu Iron Steel Group(in after referred to as Baogang and Laigang)are selected as the research objects,among which 840 are from Baogang and 800 from Laigang.These data are normalized and simulated by LIBSVM package in MATLAB.The weight of sample data is adjusted by AdaBoost algorithm.Based on this,an integrated algorithm based on AdaBoost and Support Vector Machine(SVM)is established to analyze and study the temperature state of blast furnace.The siliconcontent of hot metal in blast furnace is classified and predicted by regression.The classification of silicon in molten iron is evaluated by establishing G-means and classification accuracy.The regression problem is evaluated by establishing hit rate index,and the advantages and disadvantages of the integrated algorithm are analyzed.In this paper,three algorithms,original support vector machine algorithm(SVM),weighted support vector machine algorithm(W-SVM)and ensemble algorithm,are selected for empirical analysis and comparison.The experimental results show that the classification accuracy of ensemble algorithm in Baotou Steel’s blast furnace(79.3%)is higher than that of original SVM algorithm(78.6%)and weighted support vector machine(W-SVM).The accuracy(79.0%)has been improved slightly;the classification accuracy(79.2%)of the integrated algorithm in the blast furnace of Laiwu Iron and Steel Co.has been greatly improved compared with the original SVM algorithm(62.0%)and the classification accuracy(78.6%)of the W-SVM algorithm.It shows that the classification effect of the integrated algorithm is better and has certain research value in the classification of silicon content in molten iron.For the silicon content regression problem in hot metal,the hit rate of integrated algorithm in Baotou Steel’s blast furnace(61.0%)is slightly lower than that of original SVM algorithm(66.9%)and W-SVM algorithm(61.4%).The hit rate of integrated algorithm in Laigang’s blast furnace(63.6%)is slightly lower than that of original SVM algorithm(66.8%)and W-SVM algorithm(64.8%).。The results show that the integrated algorithm is ineffective in predicting silicon content in molten iron,and new algorithms need to be found to improve the prediction accuracy of the model. |
PDF Full Text Request