Partial Least Squares Prediction Of Silicon Content In Blast Furnace

Blast furnace process control in modern iron and steel metallurgy in an important research topic. As the blast furnace production process complexity and the diversity of parameters, to establish more appropriate and practical control model is very difficult. Blast furnace production process R & D is not only the integrated automation system control theory and control engineering technology frontier, but also iron and steel metallurgical industry, the forefront of science and technology topics. Blast furnace process is a very large, complex refining process, the operation process is often nonlinear, large time lags, spatial variable high dimensions, noise impact of a large, multi-state parameter correlation of various characteristics, which makes it difficult to establish from precise, effective silicon content in blast furnace temperature prediction model.Partial least squares method (Partial Least Squares short PLS), with the actual development of society and the needs arising from a broad applicability, the new multivariate regression analysis. It has solved many of the previous ordinary multiple linear regression with difficult problems - multiple correlation between variables. It is precisely this point, it is very suitable for the characteristics of blast furnace production, which covers the multiple linear regression, principal component analysis and canonical correlation analysis method for the integration of the basic functions of the integrated use of these types of data analysis methods to achieve development of the model accuracy.This paper selects Baotou 6 # BF (2500m3) online collection of data, first on the impact of each factor and the Temperature of hot metal silicon content [Si] the correlation into the analysis to prove that the influence factors and silicon content [Si ] Multiple sequence, there exists strong correlation. Then the correlation coefficient between them, the calculation and time series trends observed, a comprehensive analysis of the selected and obtained the blast furnace production process of State and control parameters and blast furnace hot metal [Si] the correlation between coefficients and their approximate time lag. The subject of the partial least squares regression method applied to blast furnace hot metal silicon content during temperature forecasting, the use of principal component analysis extraction furnace silicon content prediction model input parameters for subsequent modeling of the main components. The thesis made the following work: analysis of characteristics of blast furnace, blast furnace smelting process and are familiar with the use of partial least squares regression obtained to establish the validity of prediction model. Calculated silicon content in blast furnace and the furnace of the correlation coefficient of each factor and gives the main factors as independent variables, silicon content as the dependent variable of the forecast model, the basis of the direct prediction [Si], by [Si] and the relationship between the direct prediction of furnace temperature furnace. The model of the expected speed (LS), air temperature (FW), air flow (FQ), wind (FY), permeability index (FF), PCI (PM), iron poor, a comprehensive index and other factors consideration; using them as independent variables to establish prediction model of hot metal silicon content, and then forecast furnace temperature.