JPH05339617A - Converter blowing method - Google Patents

Abstract

PURPOSE:To determine a decarburization locus and to improve the hit rate of an end point carbon quantity by inputting operation data to a neural network in which the coefft. of a decarburization reaction equation is learned as teacher data, and determining this coefft. CONSTITUTION:The molten metal 12 in a converter 10 is subjected to decarburization blowing by using a lance 20. The carbon concn. of the molten metal 12 is measured via a sub-lance 21 at the adequate time of the blowing, for example, the end period of blowing and is inputted together with the information on the blowing and others to a coefft. determining computer 50 via a computer 40. The neural network in which the coefft. of the decarburization reaction equation determined from the actual data of the converter operation learned as the teacher data is previously constituted in this computer. The inputted information mentioned above is applied to this neural network and is subjected to computer processing, by which the coefft. is determined. The decarburization locus of the converter 10 is determined by a process computer 40 using this coefft. and the dynamic control thereof is executed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION This invention relates to converter blowing
The present invention relates to a converter blowing method for improving the hit rate of the target carbon content of a steel bath.

[0002]

2. Description of the Related Art In dynamic control of converter blowing, molten steel temperature and composition at the end of converter blowing are determined from data obtained by dipping a sublance into molten steel during blowing and sampling the molten steel. So that the composition matches the target value,
The blowing operation amount (blowing oxygen amount and cooling agent amount) and the like from the time of sublance measurement to the blowing end point are determined, and then blowing is performed with this operation amount.

Recently, from the viewpoint of preventing melting damage of refractory in the converter furnace and improving the time consistency with the ladle refining and casting processes, the need for speeding up the steel removal from the converter has increased, Therefore, further improvement in control accuracy in dynamic control of the converter is required.

[0004]

However, the conventional formula for determining the decarburization orbit used for dynamic control, that is, the decarburization reaction formula, has the following problems. Various model formulas are adopted as the formulas for expressing the decarburization reaction in the final stage of blowing, but in any case, the desired accuracy cannot be obtained unless the coefficients in the model formulas are properly determined.

As a method of determining the coefficient of the decarburization reaction formula, there is a method by multiple regression analysis as disclosed in Japanese Patent Publication No. 54-32117. This method is based on "estimating the decarburization rate from the flow rate and composition of the exhaust gas generated from the steel bath during oxygen converter blowing, and controlling the blowing end point with the exhaust gas system in advance as a function of the converter blowing conditions. Using the empirical value C0 of the steel bath C concentration at the time of the exhaust gas decarburization rate drop determined by multiple regression of the operating data and the steel bath C concentration at the time when decarburization begins to stall, the target of the end of blowing The decarburization rate Af at the end of blowing is calculated according to the following formula (1) according to the C concentration Cf,
This is a method for dynamically controlling the blowing of an oxygen converter, which is characterized in that the blowing end point is set when the decarburization rate Af is reached. Af = (Cf-C0) .Amax / (Cz-C0) ... (1) "Also," The decarburization rate after Cz above is accurately approximated by a linear expression of the C concentration of the steel bath, and sublance insertion at the end of blowing is performed. That is, a certain allowable time required for the correction of the blowing pattern based on the measurement and actual measurement of the steel bath C concentration represented by the following formula (2) is ensured, and the steel bath C concentration becomes the target C concentration Cf. The timing at which the concentration Cx estimated to come closest is realized is judged by the exhaust gas decarburization rate or the estimated steel bath C concentration, and a sublance is inserted at the predetermined time, and the measured C concentration is set as Csl.
To required Cf up to Cf

[0006]

[Equation 1] Is determined by the integration method of the following formula (3), and the blowing end point is set at the time when the residual oxygen amount is blown in, which is a method for dynamically controlling the blowing of the oxygen converter.

[0007]

[Equation 2]

Further, Japanese Patent Publication No. 1-247526 discloses a method of feeding back the results of this heat to the next heat. That is, this method is a "converter refining method using dephosphorized hot metal, in a converter end point control method for adjusting the carbon content and temperature of a steel bath to a target value, the carbon content of the steel bath at the end of blowing. And the temperature is measured, the required blown oxygen amount is obtained by the formula of the decarburization model shown below, and further, the decarburization transition point of this heat is calculated from the blowing result data, and the above-described heat is transferred to the next heat. A converter end point control method characterized by feeding back a decarburization transition point.

[0009]

[Equation 3] However, C: carbon content of steel bath, Q: amount of oxygen blown, CT: transition point of decarburization, CL: critical point of decarburization, α, β: constants
"

In the method of formulating the coefficient by the above multiple regression analysis, it is necessary to standardize the operating conditions as much as possible. Therefore, the data is analyzed every time the operating standard changes, and the It is necessary to perform regression analysis. Therefore, it is often difficult to maintain the desired system. Also, in the method of automatically adjusting the coefficient by feedback, it can follow moderate changes in equipment such as furnace age, but can follow abrupt changes such as changes in operating conditions such as the number of bottom blowing nozzles. Can not. Therefore, in reality, the coefficient must be constantly monitored and adjusted.

The present invention has been made to solve such a problem, and by appropriately determining the coefficient of the decarburization reaction formula corresponding to the operation, the target carbon amount of the converter blowing can be controlled. It is an object of the present invention to provide a converter blowing method capable of improving the medium ratio.

[0012]

As a means for solving the above-mentioned problems of the present invention, the present invention is to decarburize in this charge from the amount of carbon measured at an appropriate time immediately before converter blowing. In determining the trajectory and controlling the blowing end point, a neural network was constructed in which the coefficient of the decarburization reaction formula obtained from the operation record data was learned in advance as teacher data, and blowing was performed. Of the decarburization reaction formula and the decarburization trajectory by determining the coefficient of the decarburization reaction formula by applying the operating data to the input layer of the neural network for each heat Is the way.

[0013]

OPERATION As a method of expressing the decarburization reaction in the final stage of the blowing of the converter, the case of the formula based on the following formula (4) will be described, but the same argument holds when other formulas are used. It is generally accepted that the following equation (4) holds between the decarbonation efficiency (dC / dO ₂ ) and the carbon concentration (C) at the end of the converter blowing.

[0014]

[Equation 4]

Here, a is a decarboxylation efficiency dC / dO ₂ value when all of the supplied oxygen is consumed in the decarburization reaction, and is given as a theoretical value (constant = 1.07 (kg / Nm ³ ). CL
Means the lower limit of the carbon concentration in the converter decarburization reaction, and may generally be fixed at about 0.025%. The coefficient b is
It is a coefficient that depends on the stirring power of the molten metal. FIG. 3 shows a diagram in which the equation (4) is plotted with the value of b as a parameter. Figure 3
Indicates that b has a strong influence on the decarburization reaction formula.
Operating conditions that are closely related to the stirring force include, for example, the amount of slag, the number of bottom blowing nozzles, the amount of bottom blowing gas, the lance height, the top blowing oxygen flow rate, and the furnace age.

As an example, FIG. 4 and FIG. 5 show how the decarboxylation efficiency changes when the amount of slag and the number of bottom blowing nozzles change. The decarboxylation efficiency and the carbon concentration in FIGS. 4 and 5 are calculated based on the amount and composition of converter exhaust gas. In the present invention, paying attention to this point, the operation data closely related to the stirring force is given to the input layer, and it is calculated back from the operation result data and the equation (4) as the teacher data. Taking a coefficient b, a neural network in which many cases have been learned is configured in advance, and the learned network is learned.
The decarburization reaction formula is determined using a Lar network. The coefficient b is determined by applying the operating conditions given for each blowing heat to the input layer of the neural network.

[0017]

1 is a block diagram showing an embodiment of the present invention. In FIG. 1, 10 is a converter, 12 is a molten metal, 14 is a bottom blowing nozzle, 16 is a pipe, 18 is a flow meter, 20 is a lance, 2
1 is a sublance, 24 is a flow meter, 26 is a hood, 28 is a duct, 30 is a printer, 32 is a scale, 40 is a process computer, 50 is a coefficient determination computer, 51 Indicates a storage device.

As shown in FIG. 1, when decarburization of the molten metal 12 in the furnace progresses in the final stage of blowing of the converter 10, the sublance 2
1 is dropped to measure the carbon concentration, and the carbon concentration is input to the process computer 40. Top blowing oxygen flow rate, height of main lance 20, number of bottom blowing nozzles 14, bottom blowing gas flow rate,
The components of the auxiliary raw material and the input amount are respectively detected, and these detected values are also input to the process computer 40. Prior to blowing, the hot metal composition and amount, and the furnace age are given to the process computer 40. These data are supplied from the process computer 40 to the decarburization reaction coefficient determination computer 50 at the timing when the carbon concentration is measured by the sublance 21. The decarburization reaction coefficient determination computer 50 performs processing and obtains the coefficient b, which is provided to the process computer 40. Then, the process computer 4
0 determines the decarburization orbit of the converter 10 and performs dynamic control. In this embodiment, the operating data that is closely related to the stirring force is used.
The amount of slag, the number of bottom blowing nozzles, the bottom blowing gas flow rate,
Six types of data were used: lance height, top blowing oxygen flow rate, and furnace age. The slag amount is calculated from the hot metal component, the amount of auxiliary raw material input, and the component.

FIG. 2 shows the structure of the neural network in this embodiment. In the figure, the input layer of the neural network is composed of six neurons each of which inputs the above-mentioned six kinds of data. As the intermediate layer, one layer is composed of three neurons. The output layer is composed of one neuron whose coefficient b is the output. As a whole, the neural network has a three-layer structure.

On the other hand, learning is performed as follows. When the actual carbon concentration value measured immediately after blowing the converter 10 is input, the coefficient b is calculated by back-calculating the equation (4). The coefficient b, which is calculated back from the operation data, is transferred to the coefficient determination computer 50 and stored in the storage device 51. When learning becomes necessary, the coefficient determination computer 50
Reactivate the weight of each neuron by activating.

Next, an example in which the present invention is applied to a 300 ton top-bottom blowing converter will be shown. The carbon concentration of the charged hot metal is 4.0 to 4.
5% and blowout carbon concentration is 0.035 to 0.150%
, The accuracy of estimating the carbon concentration by the conventional method is σ =
What was 0.015% was σ = 0.00 in the present invention.
It improved to 7%. Furthermore, it is possible to easily re-learn about changes in operation.

[0022]

As described above, according to the converter blowing method of the present invention, it is possible to improve the hit rate of the blowing carbon amount at the end of the converter, and as a result, the blowing carbon amount of the blowing carbon amount can be improved. It was possible to obtain the effect of reducing the re-blow ratio by 3.0% from 1.0% to 1.0%, and to improve the productivity of the converter operation.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a converter blowing device according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of a neural network in the present embodiment.

FIG. 3 is a graph showing the relationship between decarboxylation efficiency and carbon concentration.

FIG. 4 is a relationship graph between decarbonation efficiency and slag amount.

FIG. 5 is a graph showing the relationship between decarbonation efficiency and the number of bottom blowing nozzles.

[Explanation of symbols]

 10 Converter 12 Molten Metal 14 Bottom Blowing Nozzle 16 Tube 18 Flowmeter 20 Lance 21 Sublance 24 Flowmeter 26 Hood 28 Duct 30 Shooter 32 Weigher 40 Process Computer 50 Coefficient Determination Computer 51 Storage Device

Claims (1)

[Claims] 1. The converter operation results are previously determined in order to determine the decarburization trajectory in this charge from the carbon content of the steel bath measured at an appropriate time immediately before the converter blowing and to control the blowing end point. A neural network is constructed in which the coefficient of the decarburization reaction formula obtained from the data is learned as teacher data, and one heat of blowing is added to the input layer of the neural network. A converter blowing method characterized in that a coefficient of a decarburization reaction formula is determined by giving operation data for each of them to determine the decarburization orbit.