Research And Prediction Of Si Content In Blast Furnace Based On Fuzzy Model

The process of blast furnace iron-making is a very complicated and highly couplednonlinear system, and its operation mechanism has the characteristics of nonlinearity,time-varying, high dimension and big noise etc. Furnace temperature is the key factor toguarantee the stable and efficiency of blast furnace production. However, there aremany parameters that can influence the blast furnace temperature, and each parameter ismutual influence. It is the main research direction of blast furnace temperatureprediction that mining the potential factor influencing furnace temperature, and thenestablishing blast furnace model to improve the accuracy of blast furnace temperatureprediction. Data driven modeling technology, which aims to achieve the complexsmelting process modeling and optimization of the blast furnace, is still at the stage ofdevelopment. To establish the prediction model of silicon content in hot metal of blastfurnace from the perspective of data driven can better reflect the changes of siliconcontent in hot metal, and can achieve a higher hit rate.In this paper, the model identification algorithm based on data driven wasresearched, which was used in the modeling of blast furnace iron-making process. Thespecific content are as follows:1. In order to establish a more accurate model, the original production datacollected from worksite in blast furnace of BaoGang are mostly missing values,abnormal values, and different magnitude. According to the expert experience, most ofthe parameters that affect the blast furnace temperature have a certain time lag situation.After taking pretreatment and calculation analysis to those data, correlation analysismethod was utilized to select the main process parameters which affect furnacetemperature, and the lag time of each parameter affecting the blast furnace temperaturewas analyzed and calculated.2. Taking the selected5process parameters as input and taking silicon content asoutput, the T-S model was established. In this paper, the optimized fuzzy cluster method(FCM) was adopted to classify all the data, and then to identify the parameter of IF partmembership function. According to the clustered data and the membership value, therecursive least square method was utilized to identify the parameter of THEN part linearequation. In this way, the T-S fuzzy model of blast furnace was established. Combiningthe actual data and situation of blast furnace, the fuzzy rules of system was established. Then the extracted fuzzy rules were adjusted through deleting the similarity andconflict fuzzy rule.3. The fuzzy rules were optimized by ANFIS fuzzy neural network optimizingmethod. According to the network after training, the IF part parameter and THEN partparameter were once determined. Through the verification of the optimized rules, the hitrate of silicon content prediction in blast furnace was improved.Finally, the hit rate of the identification model based on the measured data wasverified. It also shows that the modeling identification algorithm based on data drivencan better satisfy the expected results of modeling. Focused on the efficient modelingproblem of multivariable complicated system, this paper gave effective theory basis andit set the foundation for practical application.