Prediction Model For Influence Of Sinter Properties On Vanadium Extraction Rate In Blast Furnace

With the application of big data technology in the metallurgical industry,the study on vanadium-titanium magnetite in blast furnaces smelting process has made new developments.The historical actual production data of CHENG STEEL 4#blast furnace were collected,and the quality of the data were optimized by preprocessing the data.The pre-processed data was edited and stored to establish a scientific and universal database.When the state of the blast furnace was relatively stable,the distribution curves of the vanadium content in the molten iron and the vanadium extraction rate of the blast furnace were plotted.The influence relation among data parameters was analyzed by different trend change charts.The large-data technologies such as Pearson and Random forest were used to perform feature selection on target parameters,and feature parameters with strong correlations were retained for constructing prediction models.Among them,16 characteristic parameters such as TFe of sinter were selected,they had a strong correlation with the target parameter vanadium content,and 19 characteristic parameters such as Ti O₂of sinter and Al₂O₃of sinter were selected,they had a strong correlation with the vanadium extraction rate of the blast furnace.The vanadium content of the molten iron was analyzed by the time series of the input parameter and the target parameter,and the correlation coefficients between the input parameter and the target parameter at different lag times were obtained.The Ada Boost and Random forest algorithms were selected to construct the prediction model of vanadium content in molten iron of blast furnace,and the precision of the Ada Boost model was better.The model was optimized by adjusting the number of base classifiers in the Ada Boost model,and the accuracy of the prediction model（±0.015）is 90.84%,the root mean square error is 0.008687,and the fit goodness is0.859.The BP Neural Network algorithm was selected to construct a blast furnace vanadium extraction rate prediction model,and the L-M algorithm was used to optimize the model.It was found that the optimized prediction model had a higher precision and it was more suitable for prediction of blast furnace vanadium extraction rate.The root mean square error of the prediction model is 0.03204.Figure 41;Table 10;Reference 79...