Research On Recognition And Modeling Of Blast Furnace Gas Flow Based On Multi-source Information Fusion

Steel is the blood of a country's industrial development.At present,my country's steel output ranks among the top in the world.With the rapid development of my country's economy,steel output has continued to increase.At the same time,my country's iron and steel smelting industry is facing a severe situation of higher costs and higher energy consumption.Regardless of economic considerations or from the perspective of energy conservation and environmental protection,the steel smelting industry needs to use today's advanced technology to reduce costs and energy consumption.Modeling of blast furnace gas flow is an important part of blast furnace ironmaking,and it provides the main basis for the stable operation of blast furnace conditions.In traditional blast furnace gas flow modeling,the operating state of the furnace is judged mainly based on mechanism modeling and manual experience.However,the blast furnace ironmaking process is a non-linear,strongly coupled black box reaction with large lag.The operation of the blast furnace is carried out in harsh environments such as high temperature and high pressure,involving three complex chemical and physical reactions of solid and liquid.Due to the difference in blast furnace structure and the complexity of the furnace operating environment,traditional methods cannot effectively improve the smoothness,energy conservation and environmental protection of blast furnace ironmaking.With the development of sensor technology,technicians can collect a large number of comprehensive blast furnace operating data,which can reflect the operating status of the blast furnace in a timely manner.The development of artificial intelligence technology,data analysis and other algorithms provides strong support for processing data.In this paper,the blast furnace gas flow image,gas utilization rate and cloth operation data are multi-source fusion to model the blast furnace gas flow.1.Use machine algorithms to model the time series of blast furnace gas utilization.Aiming at the difficulty of modeling due to the nonlinearity,strong coupling and large hysteresis characteristics of the blast furnace ironmaking process.Based on the CEEMDAN-LSTM-SVM,this paper first realizes the decoupling of the gas utilization rate,and further uses the long and short-term memory LSTM and the nonlinear model SVM to model the decoupled information,and finally combines the models to obtain the gas utilization prediction model.2.Feature extraction for blast furnace gas flow images,cloth cycle extraction,classification and division of gas flow images.The gas flow image collected in the blast furnace is rich in a large amount of blast furnace reaction information.The two-norm and self-encoding are used to extract the deep and shallow features of the gas flow,the shallow feature is used to extract the gas flow cycle,and the K-means classification method is used to Classification of gas flow image patterns.3.Carry out the correlation modeling of the classified gas flow image and the gas utilization rate,analyze the development state of the blast furnace condition through the gas flow image,and analyze the correlation between the gas flow development and the blast furnace condition.