Method for predicting replacement time of furnace wall refractory bricks

Abstract

To provide a method for predicting a replacement time of a furnace wall refractory brick, which optimizes the number of explanatory variables of a multiple regression equation and improves prediction accuracy while reducing a calculation time.SOLUTION: When the replacement time of the refractory bricks 12 of the furnace wall of the argon-oxygen vacuum smelting furnace 1 is predicted by the multiple regression equation, the number of combinations of physical quantities selected as the explanatory variable x of the multiple regression equation from among a plurality of types of physical quantities related to the replacement time is sequentially increased from two, and calculation is performed by the multiple regression equation each time, and a combination of the physical quantities when the error between the predicted value and the measured value of the objective variable y of the calculated multiple regression equation falls within an allowable range is determined as the combination of the explanatory variable x for prediction of the replacement time SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for predicting the replacement time of furnace wall refractory bricks, and more particularly to a method for predicting the replacement time of furnace wall refractory bricks using multiple regression analysis. [Background technology]

[0002] Patent Document 1 discloses a method for predicting the time to replace refractory bricks in a coke oven by predicting, by simple regression analysis, the time when the amount of protrusion of the refractory bricks in the oven wall will exceed a threshold value.

[0003] Incidentally, to accurately predict the replacement timing of furnace wall refractory bricks in a vacuum argon-oxygen furnace (VCR), which has improved decarburization efficiency in low-oxygen regions by covering the top of the argon-oxygen furnace (AOD) with a vacuum lid to reduce the pressure inside the vessel, multiple regression analysis using multiple explanatory variables is required because there are many physical quantities that affect this. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 7493131 Summary of the Invention [Problem to be solved by the invention]

[0005] However, if the number of explanatory variables is increased unnecessarily, an excessively long calculation time is required, and the prediction accuracy is reduced due to the inclusion of physical quantities that do not have a significant effect on the replacement timing.

[0006] The present invention is devised to solve these problems, and aims to provide a method for predicting the replacement time of furnace wall refractory bricks, which optimizes the number of explanatory variables in a multiple regression equation, thereby reducing calculation time and improving prediction accuracy. [Means for solving the problem]

[0007] In order to achieve the above object, the present invention is characterized in that, when predicting the replacement time of furnace wall refractory bricks (12) by a multiple regression equation, the number of combinations selected as explanatory variables (x) of the multiple regression equation from among a plurality of physical quantities related to the replacement time is successively increased from two, and calculation by the multiple regression equation is performed each time, and the combination of the physical quantities when the error between the predicted value and the actually measured value of the calculated response variable (y) of the multiple regression equation falls within an allowable range is determined as the combination of explanatory variables (x) for predicting the replacement time.

[0008] The furnace wall refractory bricks (12) may be those used in an argon-oxygen vacuum smelting furnace (1), and whether the error is within the allowable range can be determined by the coefficient of determination (R2).

[0009] According to the present invention, it is possible to predict the replacement time of furnace wall refractory bricks with high accuracy in a short calculation time without employing an unnecessary large number of explanatory variables.

[0010] The symbols in parentheses above indicate, for reference, the correspondence with specific means described in the embodiments to be described later. [Effects of the Invention]

[0011] As described above, according to the method of predicting the replacement time of furnace wall refractory bricks of the present invention, it is possible to improve the prediction accuracy while reducing the calculation time for multiple regression analysis. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a schematic overall cross-sectional view of an argon-oxygen vacuum smelting furnace. [Figure 2] 10 is a box-and-whisker graph showing the distribution of coefficients of determination according to the number of selected explanatory variables. DETAILED DESCRIPTION OF THE INVENTION

[0013] The embodiments described below are merely examples, and various design improvements made by those skilled in the art without departing from the gist of the present invention are also included in the scope of the present invention.

[0014] Figure 1 shows a schematic overall cross-sectional view of an argon-oxygen vacuum smelting furnace (VCR) 1, which is the subject of the method of the present invention. The pot-shaped furnace body 11, which opens upward, has furnace wall refractory bricks 12 of a predetermined thickness arranged along the inner periphery of the metal furnace shell. A mixed gas of oxygen and argon is directly injected into the molten steel M charged into the furnace through tuyere 13 installed at the bottom of the furnace body 11, reducing the partial pressure of the CO gas generated and thereby promoting preferential decarburization. Additionally, a vacuum lid 2 is placed over the opening of the furnace body 11 to reduce the pressure inside the furnace body 11, thereby improving decarburization efficiency in the low-carbon range.

[0015] In such a VCR 1, the furnace wall refractory bricks 12 are subject to wear and tear through repeated use, and in severe cases, they fall into the furnace, causing interruptions in operation and reduced productivity. On the other hand, if the furnace wall refractory bricks 12 are replaced early enough to avoid this, it will result in increased costs due to the waste of furnace wall refractory bricks.

[0016] Therefore, it is necessary to accurately predict the replacement timing of furnace wall refractory bricks to prevent a decline in productivity and an increase in costs, and multiple regression analysis can be considered for this prediction. However, if the objective variable y in the multiple regression equation shown below (1) is the number of times the furnace wall refractory bricks 12 can be used (replacement timing), there are approximately 10 possible physical quantities that could affect this and become the explanatory variables xi in the multiple regression equation, as shown in Table 1 below. y=α0+Σαi·xi (i=1~10) … (1) where α0 is the bias and αi is the regression coefficient.

[0017] [Table 1]

[0018] Therefore, in this embodiment, for the above 10 explanatory variables x1 to x10, multiple regression equations (i=2 in the above equation (1)) are created for 10C2 (=45) combinations of two selected from these, and 39 pieces of regression model creation data (learning data) are given to each of these to calculate the regression coefficients αi (and bias α0) of the above multiple regression equations. Then, for each of the 45 multiple regression equations, 33 pieces of evaluation data are given to calculate the coefficient of determination R2.

[0019] Next, for the above 10 explanatory variables x1 to x10, three of these are selected to create 10C3 (=120) multiple regression equations (i = 3 in the above equation (1)), and 39 pieces of data (learning data) are used to create a regression model, and the regression coefficients αi (and bias α0) of the multiple regression equations are calculated. Then, the coefficient of determination R2 is calculated for each of the 120 multiple regression equations above, using 33 pieces of evaluation data.

[0020] Furthermore, for the above 10 explanatory variables x1 to x10, multiple regression equations are created for each of the combinations, increasing sequentially from 4 to 10 types, with the number of combinations being 10C4 to 10C10, and the coefficient of determination R2 is calculated for each of the multiple regression equations using the same procedure as above.

[0021] In this way, the coefficient of determination R2 is calculated for each of the multiple regression equations for the number of combinations of 2 to 10 explanatory variables selected from x1 to x10. Figure 2 shows the distribution of the coefficient of determination R2 of the multiple regression equation according to the number of explanatory variables selected (2 to 10) in a box-and-whisker graph. According to this, the coefficient of determination is maximized when five explanatory variables x are selected (arrow in Figure 2), and the combination of explanatory variables that shows this maximum value is x1, x2, x3, x4, and x5.

[0022] As a result, it can be seen that the number of service lives (replacement time) of the furnace wall refractory bricks 12 of the VCR1 shown in this embodiment can be predicted with high accuracy in a short calculation time by a multiple regression equation that uses a combination of x1, x2, x3, x4, and x5 as the explanatory variable x, without using too many explanatory variables unnecessarily.

[0023] Although the above embodiment has been described with respect to predicting the replacement time of refractory bricks in the furnace wall of a VCR, the method of the present invention is not limited to VCRs and can also be applied to predicting the replacement time of refractory bricks in the furnace wall of other industrial furnaces. Furthermore, the combination of explanatory variables to be used is not limited to the one that maximizes the coefficient of determination, but the number of combinations of explanatory variables may be increased sequentially, and a combination may be adopted when the coefficient of determination falls within an acceptable range. Furthermore, it is not necessary to use the coefficient of determination, and other indices that indicate the error between the predicted value and the actual measured value of the dependent variable may be used. [Explanation of symbols]

[0024] 1...argon-oxygen vacuum smelting furnace, 12...furnace wall refractory brick, 13...tuyeres, 2...vacuum cover, M...molten steel.

Claims

1. a method for predicting the replacement time of furnace wall refractory bricks using a multiple regression equation, comprising: selecting from a plurality of physical quantities related to the replacement time combinations as explanatory variables of the multiple regression equation, starting from two, successively increasing the number of combinations, and performing calculations using the multiple regression equation each time; and determining, as the combination of explanatory variables for predicting the replacement time, the combination of physical quantities when an error between a predicted value and an actually measured value of a response variable of the calculated multiple regression equation falls within an allowable range.

2. 2. The method for predicting the replacement time of furnace wall refractory bricks according to claim 1, wherein the furnace wall refractory bricks are used in an argon-oxygen vacuum smelting furnace.

3. 3. The method for predicting the replacement time of furnace wall refractory bricks according to claim 1, wherein whether or not the error falls within an allowable range is determined by a coefficient of determination.

Images