ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

Research of process of the oxygen converting by the methods of mathematical statistics

Journal: Reporter of the Priazovskyi State Technical University. Section: Technical sciences (Vol.33, No. 1)

Publication Date:

Authors : ;

Page : 21-32

Keywords : regressive analysis; normal distribution; form of dependence; authenticity of model parameters;

Source : Downloadexternal Find it from : Google Scholarexternal

Abstract

An array of 177 meltings of the oxygen converter shop of Azovstal plant was formed for the analysis. The dependent variable is the carbon content [C] in the molten metal and independent variables being manganese [Mn], phosphorus [P] and sulphur [S] content, oxygen activity a[o] and the melt temperature (T, ºC). After excluding the variables that do not affect the value of [C], (T, ºC), [Mn] and a[o] were found to be significant. The constructed model [C] = f([Mn], a[o] (T, °C)) is adequate: the value of the Fisher criterion is – 42,06 with the significance level much less than 0,05. The introduction of the additional variable «converter» did not result in a significant improvement of the regression model – the value of the specified regression coefficient increased from 0,41 to 0,45. Histogram of the distribution of the remainders for the obtained model indicate the lack of a normal distribution. Analysis of the data sets separately for the first and second converters has significantly increased the value of the specified regression coefficient up to 0,85 for the first converter, but reduced to 0,20 for the second convertor. In this case the variable (T, ºC) in the regression equation for the second converter turned out to be insignificant. Graphs of the dependence of the remainders for each independent variable on the values of the variable are characterized by a nonlinear trend, indicating lack of normal distribution. Graphical illustrations of the remainders distribution for the first and second converters are alike. However, the correlation coefficient of the remainders and variables values is at the level of 10-7…10-8, indicating the adequacy of the obtained models. Thus, the results of the analysis indicate that kinetic regularities of physico-chemical processes in the first converter are substantially different from the second. The prediction probability values of the variable [C] in accordance with the obtained regression equations confirmed for the first Converter and no for the second

Last modified: 2018-04-11 18:52:18