Study On End-point Control Model Of BOF Steelmaking Based On Whale Optimization Algorithm And Twin Support Vector Machine

Basic oxygen furnace(BOF)steelmaking is one of the main methods of producing steel in the steel industry.For the steel industry,BOF steelmaking plays an important role.Based on the actual smelting situation of domestic BOF steelmaking,the strategic deployment goal of made in China2025 can be achieved with the rapid development of model and computer network technology.It is very meaningful to establish a BOF steelmaking end-point control model that conforms to the current state for improving the production efficiency,product quality,and automation level of BOF steelmaking.At present,the BOF steelmaking model established by intelligent methods such as neural networks,expert systems,and support vector machines has become one of the important ways to improve the automation level of BOF steelmaking.However,although these intelligent methods have achieved good results in the application of BOF steelmaking,there are still some shortcomings,such as low efficiency of model operation,easy falling into local extremum,and poor training ability of small samples.Therefore,new methods are needed to improve this problem.The methods of twin support vector regression(TSVR)and whale optimization algorithm(WOA)can solve the above problems,and these two methods are still in the vacancy stage in the application of the BOF steelmaking control model.Therefore,80 t BOF without sub-lance detection equipment and 260 t BOF with sub-lance detection equipment were researched in this thesis.Based on the TSVR and the WOA,a series of in-depth studies on the BOF steelmaking end-point control model were carried out.(1)Considering the traditional static control model had some problems in modeling and operation efficiency.To realize the end-point control of BOF without sub-lance detection equipment,a static end-point control model of 80t BOF steelmaking was established by using TSVR,WOA,and interval increment algorithms.The proposed control model has an MAE of 285 Nm~3for oxygen blowing volume and an MAE of 0.17 t for lime weight.The established static control model had good practical value for the BOF without sub-lance detection equipment.(2)At present,some of the domestic large-scale BOFs have installed sub-lance detection equipment,which has a certain degree of automation,but the control of the end-point can still be further improved.To realize the end-point control of BOF with sub-lance detection equipment,a static end-point control model of 260 t BOF steelmaking was established by using TSVR and WOA.Considering that the selection of model input variables may affect the accuracy of the model.Therefore,based on the input variables determined by mechanism analysis,the correlation between input and output variables and the independence between input variables and input variables were analyzed using grey relational analysis and partial correlation analysis.The proposed control model has an MAE of 9.3114 t for scrap weight,an MAE of 2.5791 t for lime weight,an MAE of 0.7919 t for dolomite weight,and an MAE of537.8215 Nm~3 for oxygen blowing volume.The proposed static control model can provide valuable reference for the early operation of BOF smelting with sub-lance detection equipment.(3)The BOF with sub-lance detection equipment entered the later stage of smelting,and the information in the furnace had been detected by sub-lance.At this time,the amount of oxygen added would affect whether the carbon content and temperature of molten steel could reach the end-point accurately.Based on this situation,the end-point dynamic control model of260 t BOF steelmaking was established by using the improved TSVR and the improved WOA.The proposed control model has MAE of 75.0201 Nm~3 for the end-blow oxygen volume.The established dynamic control model could guide the end-point control in the later stage of 260 t BOF smelting,and play a key role in realizing the automation development of BOF steelmaking in the future.