JPH0665623A - Method for estimating carbon content in molten steel during blowing in converter - Google Patents

Abstract

PURPOSE:To constantly estimate the carbon content in the molten steel with high precision during the blowing operation even when the operational conditions or the like are changed with time in the converter operation. CONSTITUTION:A three-storied neural network is structured to which the solidifying temperature t1 of the molten steel to be measured during the blowing, and the operational information before the measurement (the amount of manganese ore t2, the manganese content t3 of the molten iron,... tn) are inputted and from which the carbon content [C] of the molten steel at the measuring time is outputted, and the carbon content in the molten steel during the blowing operation is estimated by using the neural network. The learning of the neural network is executed by using the measured values of the solidifying temperature in the actural operation, the operational information before the measurement, and the analytical value of the actually measured carbon content of the molten steel, and the coefficients of coupling Wij, Wj' (i=1-n, j=1-m) between the units constituting the neural network are amended.

Description

Detailed Description of the Invention

[0001]

The present invention relates to a method for estimating carbon concentration during converter blowing, and particularly for accurately estimating carbon concentration during converter blowing even when operating conditions change over time. The present invention relates to a method of estimating the carbon concentration in a converter blowable.

[0002]

2. Description of the Related Art In converter blowing (steel making), a sublance is used at the end of blowing in order to accurately control the end point so that the molten steel carbon concentration and the molten steel temperature at the end of blowing (end point) are adjusted to the target values. The molten steel temperature is measured and the carbon concentration is estimated by using it, and blowing control is performed by model calculation based on the measured value and the estimated value.

As a method for estimating the carbon concentration of molten steel as described above, Japanese Patent Laid-Open No. 2-57629 discloses a method for estimating the carbon content of molten steel during converter refining using dephosphorized hot metal.
The molten steel is sampled and its temperature and solidification temperature are measured, the primary estimated carbon amount in the molten steel is determined from the solidification temperature, and the primary estimated carbon amount and the molten steel temperature are used to form an equilibrium equation of Mn in the converter, A method of obtaining an estimated Mn amount in molten steel from a mass balance equation and obtaining a carbon amount using the primary estimated carbon amount and the estimated Mn amount is disclosed.

[0004]

However, the method disclosed in Japanese Patent Laid-Open No. 2-57629 has the following problems.
Since the carbon concentration is estimated from the solidification temperature using the multiple regression equation, the estimation accuracy deteriorates as the operating conditions change over time, and it has the function of correcting the estimation logic and coefficients. Therefore, there is a problem that the estimation accuracy cannot be maintained.

The present invention has been made to solve the above-mentioned conventional problems, and it is possible to accurately estimate the molten steel carbon concentration during converter blowing even when the converter operating conditions change with time. It is a first object to provide a method for estimating the molten steel carbon concentration during converter blowing that is capable of achieving the above.

[0006] The present invention also ensures that the concentration of carbon contained in molten steel during converter blowing is always at an appropriate level of accuracy required in actual operation, even when operating conditions and the like change over time. Moreover, it is a second object to provide a method for estimating the molten steel carbon concentration during converter blowing, which can be estimated in a short time.

[0007]

According to the first aspect of the present invention, the solidification temperature of molten steel measured during converter blowing and the operation information up to the time of measurement are input, and the molten steel carbon concentration at the time of measurement is output. When a neural network is constructed and the molten steel carbon concentration during converter blowing is estimated using the neural network, the measured value of solidification temperature during actual operation, operation information up to the time of measurement, and molten steel carbon The first object is achieved by learning the above-mentioned neural network using the measured value of the concentration and correcting the coupling coefficient between the units forming the neural network.

According to the first aspect of the present invention, in the method for estimating the molten steel carbon concentration in the converter blowing, as a pre-processing when learning the neural network, the molten steel carbon concentration in the converter blowing is set to a predetermined difference. Divide into multiple classes set by width, and if the analysis result of molten steel carbon concentration during converter blowing has a low actual frequency of the corresponding class, copy the operation result data corresponding to that class by copying. It is created and increased, and the frequency distribution equalization processing of the learning data is performed in advance so that the frequency distribution of the learning data for each class becomes substantially uniform.

In the second aspect of the present invention, a neural network is constructed in which the solidification temperature of molten steel measured during converter blowing and the operation information up to the time of measurement are input, and the molten steel carbon concentration at the time of measurement is output. When estimating the molten steel carbon concentration in the converter blowing using the neural network, the molten steel carbon concentration in the converter blowing is divided into a plurality of classes set with a predetermined difference width, While configuring the output layer with the same number of units as the number of classes, from each unit that constitutes the output layer, the priority selection index number, which is the reference for selecting the corresponding class, is output,
When actually operating, the operation information until the time of the measurement, the actual measurement value of the molten steel solidification temperature during converter blowing, and the analysis result of the molten steel carbon concentration at the time of the learning of the neural network. By modifying the coupling coefficient between the units that make up the neural network,
The second object is achieved.

According to the second aspect of the present invention, in the method for estimating the molten steel carbon concentration during converter blowing, the learning results of the molten steel carbon concentration during converter blowing are output when the neural network is learned. The unit has priority selection index 1 and output units of other classes have priority selection index 0.
Each is input as teacher data, and each output unit is 0
The number of priority selection indexes from 1 to 1 is output.

In the second aspect of the present invention, in the method for estimating the molten steel carbon concentration during the converter blowing, the analysis result of the molten steel carbon concentration during the converter blowing corresponds to the pre-processing when learning the neural network. If the frequency performance of a class is low, copy the operating performance data to increase it and perform a homogenization process of the learning data frequency distribution beforehand so that the frequency distribution of the learning data for each class becomes almost uniform. It is something to do.

[0012]

In the first aspect of the present invention, since the molten steel carbon concentration during blowing is estimated by using the neural network, human empirical rules and subjectivity are not included, and linear,
Since it is possible to model regardless of nonlinearity, it is superior in accuracy to statistical methods such as multiple regression, and therefore extremely highly accurate estimation is possible.

Further, since the neural network is learned based on the actual operation results and the function is corrected, it is possible to follow the change of the prediction target, and always with high accuracy and in a short time. It is possible to estimate.

Further, in the first aspect of the present invention, in the learning of the neural network, as a pretreatment, the molten steel carbon concentration in the converter blowing is divided into a plurality of classes set with a predetermined difference width, and the blowing is performed. If the frequency of analysis of the molten steel carbon concentration in the corresponding class is low, the operation result data corresponding to that class is duplicated by copying to increase the frequency distribution of the learning data for each class. So that
When performing the frequency distribution equalization processing of learning data in advance,
It is extremely effective in improving the accuracy of estimation for operation patterns that belong to a class with a low track record.

According to the second aspect of the present invention, when estimating the carbon concentration during converter blowing using a neural network, the molten steel carbon concentration during blowing is divided into a plurality of classes set with a predetermined difference width. Then, the output layer of the neural network is configured with the same number of units as the number of the classes, and each unit constituting the output layer outputs the number of priority selection indexes which is a criterion for selecting the corresponding class. As described above, learning is performed in the same manner as in the case of the first aspect of the invention, and the coupling coefficient between the units is corrected. Therefore, by setting the difference between classes according to the required accuracy of actual operation,
It is possible to execute the estimation of the molten steel carbon concentration during blowing with a desired accuracy that is actually commensurate and in a short time while following the change of the prediction target.

In the conventional method for estimating the molten steel carbon concentration during converter blowing based on the multiple regression equation, since the molten steel carbon concentration during blowing was calculated by numerical calculation, there is an upper and lower limit to the obtained value. In any case, the answer for the number of digits that can be calculated within that range can be output as a continuous quantity (value). However, when using the multiple regression equation, as described above, there is a problem in accuracy because it cannot follow the change in operating conditions over time, and even if an accurate numerical value is obtained, in actual operation, When the required accuracy used in the blowing calculation is not required, for example, 0.01% or more, or even if it is required, it is not feasible from the measurement accuracy of the solidification temperature, or it is meaningless considering the accuracy of the subsequent blowing control. There is.

Therefore, it is obtained by dividing the carbon concentration in the range expected at the time of measuring the molten steel solidification temperature during blowing by a predetermined difference width, as compared with the accuracy of less than 0.01% in the model that outputs continuous values. It may be efficient to correctly estimate the class corresponding to the numerical range so that the accuracy does not deteriorate by 0.01% or more. According to the second aspect of the present invention, it is possible to realize the actual estimation of the molten steel carbon concentration during blowing by using the computer by setting the class with an appropriate numerical difference and selecting the class. It will be possible.

Further, by using the neural network as described above, the human empirical rule and the subjectivity are not included and the modeling can be performed regardless of linear or non-linear as in the first invention. It is superior in accuracy to statistical methods such as multiple regression, and from this point also, highly accurate estimation is possible.

In the second aspect of the present invention, the number of units in the output layer, that is, the number of classes, is determined by, for example, dividing the carbon concentration in the range expected at the time of measuring the molten steel solidification temperature during blowing by equally dividing it. The width (class difference) is determined from the tolerance of the molten steel carbon concentration used in the subsequent blowing calculation, the control accuracy and measurement accuracy of the oxygen gas injection control device and auxiliary raw material metering and charging equipment, and the operator's empirical knowledge. It is decided by comprehensive judgment.

Further, in the second aspect of the present invention, when learning the neural network, the priority selection index number 1 is given to the unit of the class corresponding to the analysis result of the molten steel carbon concentration during blowing, and the unit of the other classes is given priority. When the selection index number 0 is input as the teacher data and each output unit outputs the priority selection index numbers 0 to 1,
A priority selection index number from 0 to 1 is output to each of the output layer units, and the class that outputs the maximum value (the value closest to 1) of the output priority selection indexes is the corresponding answer. It becomes possible to Therefore, it is possible to select the representative value, for example, the median value, from the range and determine it as the estimated carbon concentration during blowing.

Further, in the second aspect of the present invention, when the above learning is performed, as a pretreatment, if the analysis result of the molten steel carbon concentration during converter blowing has a low record frequency of the applicable class, the data of the operation record is obtained. Make duplicates by copying and increase, so that the frequency distribution of the learning data for each class becomes almost uniform,
When performing the frequency distribution equalization processing of learning data in advance,
It is extremely effective in improving the accuracy of estimation for operation patterns that belong to a class with a low track record.

In the present invention, the operation information up to the time of measurement is various information that influences the molten steel carbon concentration during blowing, and, for example, the composition and composition of the main and auxiliary raw materials initially charged in the converter. Amount, composition composition of main and auxiliary raw materials such as slag forming agent, iron ore, scrap iron, etc. to be introduced into the converter from the start of blowing to the time of measurement, input amount, input amount, oxygen gas injection amount, etc. is there.

[0023]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a schematic diagram showing the configuration of a neural network applicable to the first embodiment according to the first invention. The neural network is composed of three layers, an input layer, an intermediate layer and an output layer. Has been done.

FIG. 2 is an explanatory view showing the outline of the apparatus configuration applied to this embodiment. Reference numeral 10 in FIG. 2 is a converter for blowing molten steel M, and process information in the converter 10 is input to the process computer 12. A neural network, which will be described in detail below, is built in this process computer 12, and the molten steel carbon concentration during blowing, which is an estimated value by inputting process information input online in the process computer 12 to the neural network. [C] is output.

The process computer 12 is connected to a learning server 14. The learning server 14 collects process information from the process computer 12 and uses the information to perform self-learning of a neural network. The correction information is calculated, and the weighting coefficient of the neural network in the process computer 12 is corrected based on the correction information.

The input layer of the neural network is
In n input units Ii (i = 1 to n), the intermediate layer is m
Each of the intermediate units Hj (j = 1 to m) has an output layer composed of one output unit O.

In this neural network, each of the n input units Ii has m intermediate units Hj.
Of the weighting factors Wij (i = 1 to n, j
= 1 to m). That is, the input unit I1
In the case of, the intermediate units H1 to Hm are respectively coupled with weighting factors W11 to W1m, and although not shown, the input units I2 to In-1 are similarly coupled,
The last input unit In is connected with weighting factors Wn1 to Wnm.

The m intermediate units Hj forming the intermediate layer are connected to the output units O forming the output layer by weighting factors Wj '(j = 1 to m).

In the above neural network,
Signals corresponding to n input units Ii
t _i (i = 1 to n) is input. The signal t _i is input as process information that is input online from the converter 10 to the process computer 12. When signals are input to the input units Ii in this way, for example, from the unit I1, the intermediate units H1 to H1.
W11, in which the input signal t ₁ is multiplied by a weighting coefficient for m
t _1, W12 · t _1, being adapted to W13 · t ₁ ~W1m · t ₁ are input, similarly input unit I2
Each input signal t ₂ ~ t signal weighting factor is multiplied by each of the _n with respect to the intermediate unit H1 ～Hm is adapted to be inputted from -In. Therefore, each intermediate unit H
A signal u _j represented by the following equation (1) for calculating the sum of i = 1 to n is input to each _j .

U _j = ΣW ij · t _i (1)

As described above, the signal u is sent to each intermediate unit Hj.
_{When j} is input, the intermediate unit Hj outputs a signal u _j ′ (j = 1 to m) to the output unit O. The signal u _j ′ is given as a value obtained by multiplying the function f (u _j ) incorporating the input signal u _j by the weighting coefficient W _j ′, as shown in the following equation (2). The function f (u _j ) can be given by, for example, the sigmoid function of the following expression (3).

U _j ′ = W _j ′ · f (u _j ) ... (2) f (u _j ) = 1 / (1 + exp (−u _j )) (3)

[0034] As described above, when each intermediate unit H1, respectively from ~Hm u ₁ '~ u _m' is input to the output unit O, the output unit O is molten steel carbon concentration in blowing as the final output [C] Is output. This output [C] is given by the following equation (4). In this equation, the sigmoid function can be used as the function f.

[C] = f (u ₁ ′ + u ₂ ′ + u ₃ ′ ... + u _m ′) (4)

In this embodiment, the input unit I
Signal as process information for each of 1 to In
t ₁ = molten steel solidification temperature (℃) during measurement during blowing, t
By inputting ₂ = Mn ore usage (Kg / t), t ₃ = hot metal Mn concentration (%), etc., the output unit O outputs the molten steel carbon concentration [C] during blowing as an estimated value. . Here, the signals t ₂ to t _n correspond to operation information up to the time of measurement during blowing of the converter.

As described above, for the neural network, the actual results of actual operation, that is, the solidification temperature t _{1 of the} molten steel actually measured, the operating conditions t ₂ to t up to the time of measurement
Learning to modify the weighting factor of the coupling between the units is performed by using _n and the molten steel carbon concentration [C] r during blowing that is obtained by actual analysis. When learning, the initial value of the weighting coefficient between units can be set using a random number table.

As this learning method, a back propagation (error back propagation) method can be used. This back-propagation method is a weighting coefficient between units forming the network so as to reduce the value of the square of the error δ between the estimated molten steel carbon concentration [C] by the neural network and the measured molten steel carbon concentration [C] r. And is achieved by repeating forward and backward error backpropagation calculations.

As described above, by repeating the operation and learning the neural network, for example, periodically and using the latest data, even if the operating condition changes, the latest operation can be performed by the neural network. Since the estimation that reflects the tendency of the condition can be performed, the correct estimation can always be performed and the estimation accuracy can be maintained.

According to the present embodiment, as described above, the operation is repeated and the learning for the neural network is performed based on the operation result, so that the molten steel carbon concentration during blowing can be estimated with high accuracy at all times. can do.

In the learning of the neural network, it is effective to improve the estimation accuracy by performing, for example, the following learning data frequency distribution equalization processing as preprocessing.

First, the molten steel carbon concentration during blowing is divided into a plurality of grades set with a predetermined difference width, and an actual frequency distribution is obtained for the grade to which the analysis result of the molten steel carbon concentration at the time of solidification temperature measurement corresponds.

Next, letting F _{MAX be} the frequency number of the class with the highest record frequency, and F _x the frequency number of the class with the record record frequency lower than that, the data of each operation record belonging to the class with the record record frequency is small. _Are duplicated by copying so that the frequency becomes F _MAX / F _x times. At that time, if the above F _MAX / F _x does not become an integer, it is converted into an integer by a predetermined constant integer conversion rule of rounding down, rounding off, or rounding up.

By the above processing, the number of learning data belonging to each class becomes substantially uniform.
By performing processing in this way, it is possible to effectively improve the estimation accuracy for operation patterns that belong to a class with a low actual record frequency.

Next, a specific example to which this embodiment is applied will be described. A 230 ton bottom blowing converter is used as the converter 10, and 230 tons of hot metal is put into the converter 10 to carry out blowing,
The molten steel carbon concentration during blowing is estimated according to this embodiment by using the molten steel solidification temperature measured using the sublance during the blowing, and the Mn ore usage amount, the hot metal Mn concentration, etc. as operation information up to the time of the measurement. The results obtained are shown in the graph of FIG.
-1) and (1-2).

In FIG. 3, the blank bar graph indicates the frequency distribution equalization process in advance when the frequency distribution uniformization process is not performed during learning (invention method (1-1)). The case (the present invention method (1-2)) corresponds to each case. In this graph, the horizontal axis shows the error (× 10 ^-2 %) between the measured value (analytical value) of the molten steel carbon concentration during blowing and the estimated value according to the present embodiment, and the vertical axis shows the frequency.

Further, FIG. 4 shows the results estimated by the conventional method for the same converter operation in order to clarify the effect of this embodiment. In this conventional method, the molten steel carbon concentration [C] during blowing is estimated by the following multiple regression equation (5). Further, this multiple regression equation is obtained by the least square method from the same learning data as in this embodiment. In the formula, A _{0 to} A ₃ are constants.

[C] = A ₀ + A ₁ × solidification temperature (° C.) + A ₂ × Mn ore usage (Kg / t) + A ₃ × hot metal Mn concentration (%) (5)

As is clear from FIGS. 3 and 4, the estimation errors of the methods (1-1) and (1-2) of the present invention are significantly smaller than those of the conventional method, and the frequency distribution is previously calculated. The method (1-2) of the present invention that was subjected to the homogenization treatment was better than the method (1-1) of the present invention that was not subjected to the homogenization treatment of the frequency distribution.
It can be seen that the estimation error is slightly smaller.

Further, when the neural network is trained at a rate of once a day using the molten steel carbon concentration measured at the time of tapping as teacher data, the change of the estimation error due to the change with time (the method of the present invention (1 )) Is shown in FIG. 5 together with the transition (conventional method) when estimated using the conventional multiple regression equation (5). It can be seen from FIG. 5 that the present embodiment (neural network) has a smaller estimation error over time and is more stable.

According to this embodiment described in detail above, based on the solidification temperature measured during blowing and the operation information up to the time of measurement, the molten steel carbon concentration at the time of the measurement is estimated by the neural network, and By modifying the neural network by self-learning,
Since the molten steel carbon concentration during blowing can be estimated with high accuracy and the accuracy can be maintained, it is possible to improve the controllability of the end point of blowing. as a result,
It has become possible to reduce costs by improving the production steel yield and reducing the basic unit of deoxidizing aluminum.

FIG. 6 is a schematic diagram showing the configuration of a neural network for estimating the molten steel carbon concentration during blowing applied to the second embodiment according to the second invention.

The present embodiment is substantially the same as the first embodiment except that the process computer 12 shown in FIG. 2 is constructed with the neural network shown in FIG. 6 described in detail below. is there.

In the neural network applied to this embodiment, m intermediate units Hj constituting the intermediate layer are
K output units O _x (x
= 1 to k) and the weighting coefficient W _jx (j = 1 to m, x = 1 to k)
The neural network is the same as the neural network of the first embodiment except that the neural network is connected with.

The number of units in this output layer, that is, the number of classes k
Is determined by equally dividing the molten steel carbon concentration in the range expected when measuring the molten steel solidification temperature during blowing, and the width (class difference width) at that time is used in the subsequent blowing calculation as described above. It is decided by making a comprehensive judgment based on the tolerance of the molten steel carbon concentration, the control accuracy and measurement accuracy of each of the oxygen gas injection control device, the auxiliary raw material metering and charging equipment, and the empirical knowledge of the operator.

Therefore, in the above-mentioned neural network, as in the case of the first embodiment, the signals t _i (i = 1) corresponding to the n input units Ii, respectively.
When ~n) is input, for example from unit I1, the weighting factor is multiplied by the input signal t ₁ to the intermediate unit _{H1 ~Hm W11 · t 1 ~W1m ·} t 1 is inputted, similarly input unit I2 input signals t ₂ ~ t _n the signal is multiplied by the weighting coefficients of binding to the intermediate unit H1 ～Hm from ~In is adapted to be input.

Therefore, for each intermediate unit Hj
Is the signal expressed by the above equation (1) for calculating the sum of i = 1 to n
Issue u_jWill be input. Thus, each intermediate user
Signal u to knit Hj_jIs input, the intermediate unit
Hj is k output units O _xFor (x = 1 to k)
Signal u_jxOutput (j = 1 to m, x = 1 to k) respectively
It This signal u_jxIs expressed by the above equation (6).
Input signal u_jThe function f (u_j) To the weighting factor W
_jxIt is given as a value multiplied by. Also, the above function f (u
_j) Is, for example, the sigmoid function of the following equation (7)
Can be given.

U _jx = W _jx · f (u _j ) (6) f (u _j ) = 1 / (1 + exp ( _{-u j} )) (7)

As described above, k output units O _x (x
= 1 to k) from each intermediate unit H1 to Hm
_{When 1x} ~ u _mx is input, the output unit O _x has a rank
A molten steel carbon concentration within the range defined by x is output as a priority selection index number αx of 0 to 1 that indicates the possibility of conforming to the estimated value of the molten steel carbon concentration when measuring the molten steel solidification temperature during blowing. That is, the closer the number of priority selection indexes αx is to 1, the more appropriate the class is as an estimated value, and conversely 0
The closer it is to, the less suitable the class is. Finally, a class in which αx is closest to 1 is selected, a representative value (for example, median value) of the class is output, and the estimated value of the molten steel carbon concentration during blowing is selected.

In this embodiment, as in the case of the first embodiment, a signal as process information for each of the input units I1 to In: t ₁ = molten steel solidification temperature (° C.), t ₂ = Mn By inputting ore consumption (Kg / t), t ₃ = concentration (%) of hot metal Mn, etc., priority is given to the class of molten steel carbon concentration during blowing from k output units O _x of the neural network. Selection index number α
1 to αk is output as a value in the range of 1 to 0 and used as an estimated value.

As described above, for the neural network, the actual results of operation, that is, the molten steel solidification temperature t ₁ during the smelting, the amount of Mn ore used t ₂ , the concentration of the hot metal Mn,
The weighting factor of the bond between the units is corrected using t ₃ , the operating conditions t ₄ to t _n up to the solidification temperature measurement, and the molten steel carbon concentration [C] r at the solidification temperature measurement obtained by actual analysis. Learn to do.

At that time, actual molten steel obtained by actual analysis
Output unit O to which the carbon concentration corresponds _x(X = 1 to k)
1 for one of the output units, other not applicable
All 0s are input to the output unit as teacher data.

As a pretreatment, for a class with a low actual frequency corresponding to the actual molten steel carbon concentration, the operation actual data is increased by copying so that the frequency distribution of the learning data for each class becomes almost uniform. The frequency distribution of the learning data is made uniform. The method of equalizing the frequency distribution of the learning data may be the same as the method described in the first embodiment of the present invention.

When learning, as in the case of the first embodiment, the initial value of the weighting coefficient between units can be set by using a random number table. , A back propagation (error back propagation) method can be used.

Next, a specific example to which this embodiment is applied will be described. As in the case of the first embodiment, a 230 ton bottom blowing converter is used as the converter 10, 230 tons of hot metal is put into the converter 10 to carry out blowing, and molten steel during blowing at that time Using the Mn ore usage (Kg / t), hot metal Mn concentration (%), etc. as solidification temperature and operation information until the solidification temperature is measured,
The molten steel carbon concentration during blowing was estimated according to this example.

Here, for the output layer of the neural network, the carbon concentration class width is set to 0.01% for the above-mentioned reason, and as shown in parentheses in FIG.
The output layer was composed of a number of units corresponding to 30 classes from 0.10 to 0.40%.

500 pieces of blowing data of low carbon steel are prepared, and 1 is input to the output unit of the class to which the actual data of the molten steel carbon concentration of the neural network is applied by the initial learning, and the output units of other classes are not applicable. Entered 0 for. In addition, the neural network having a small number of data of the corresponding class was increased by copying, and the learning data was homogenized so that the number of data of all classes became almost uniform.

After the above pre-processing, the number of units in the intermediate layer (hidden layer) is adjusted and verified using the operation data, and the neural network having the highest estimation accuracy is selected to optimize the configuration. Was made.

After the configuration of the neural network was determined, the latest operation data of 500 charges was input and re-learning was performed, and the neural network was used in the actual operation thereafter.

The result of estimation using the neural network constructed as described above is shown in the graph of FIG. In this graph, as in the case of FIG. 3, the error (×) between the measured value (analytical value) of the molten steel carbon concentration during blowing on the horizontal axis and the estimated value according to the present embodiment.
10 ^-2 %) and the vertical axis shows the frequency.

In order to clarify the effect of this embodiment, the results estimated by the conventional method for the same converter operation are also shown in FIG. This conventional method is described in (5) above.
The molten steel carbon concentration [C] during blowing is estimated by the multiple regression equation. Further, this multiple regression equation is obtained by the least square method from the same learning data as in this embodiment.
It can be seen from FIG. 7 that the estimation error of this embodiment (invention method (2)) is considerably smaller than that of the conventional method.

Further, with respect to the neural network of the present embodiment, the transition of the estimation error due to the change with time when learning is performed at a rate of once a day using the molten steel carbon concentration actually measured during the blowing as teacher data. FIG. 5 showing the transition of the first embodiment is also shown as the method (2) of the present invention. It can be seen from FIG. 5 that the present embodiment has a small estimation error with time and is stable as in the first embodiment.

According to the present embodiment described in detail above, the molten steel carbon concentration is estimated by the neural network based on the molten steel solidification temperature measured during blowing and the operation information up to that time, and self-learning is performed. By performing the correction of the neural network according to the above, it is possible to estimate the molten steel carbon concentration during blowing with high accuracy and maintain the accuracy, and also to obtain the appropriate accuracy required in the actual operation, and to shorten it. It can be estimated in time.

As a result, the controllability of the blowing end point can be improved as in the first embodiment, and the cost can be reduced by improving the production yield of steel and reducing the basic unit of aluminum for deoxidation. became.

The present invention has been specifically described above, but the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention.

For example, in the above embodiment, the case of using the neural network having the three-layer structure has been described, but the present invention is not limited to this, and the number of layers may be four or more.

Further, when the output layer is composed of the same number of units as the class, the class width is not limited to 0.01% shown in the embodiment and can be arbitrarily changed.

In the above embodiment, the case of the bottom blowing converter has been described, but other blowing furnaces such as a top blowing converter or a bottom blowing converter may be used.

[0079]

As described above, according to the first invention,
Even when the operating conditions and the like change with time in the converter operation, the concentration of carbon contained in the molten steel during blowing can always be estimated with high accuracy and in a short time.

Further, according to the second aspect of the present invention, it is possible to properly estimate the molten steel carbon concentration according to the accuracy required in the actual operation.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram showing a neural network applied to a first embodiment according to the present invention.

FIG. 2 is an explanatory diagram showing an outline of a device configuration applied to the embodiment.

FIG. 3 is a diagram showing estimation accuracy according to the embodiment.

FIG. 4 is a diagram showing a conventional estimation accuracy.

FIG. 5 is a diagram showing changes over time in the estimation error between the example method and the conventional method.

FIG. 6 is a schematic configuration diagram showing a neural network applied to a second embodiment according to the present invention.

FIG. 7 is a diagram showing estimation accuracy according to the same embodiment and a conventional method.

[Explanation of symbols]

 10 ... Converter 12 ... Process computer 14 ... Learning server 16 ... Main lance 18 ... Sublance M ... Molten steel

Claims (5)

[Claims] 1. A solidification temperature of molten steel measured during converter blowing,
A neural network that inputs the operation information up to the time of measurement and outputs the molten steel carbon concentration at the time of measurement is constructed, and when the neural network is used to estimate the molten steel carbon concentration during blowing of the converter, the actual operation is performed. In this case, the neural network is learned by using the measured value of the solidification temperature, the operation information up to the time of the measurement, and the measured value of the molten steel carbon concentration, and the coupling coefficient between the units forming the neural network. A method for estimating the carbon concentration of molten steel during converter blowing, characterized in that 2. The method according to claim 1, wherein the learning process of the neural network is performed by dividing the molten steel carbon concentration in the converter blowing into a plurality of classes set with a predetermined difference width as a pretreatment. If the frequency of analysis of the molten steel carbon concentration in the corresponding class is low, the operation result data corresponding to that class is duplicated by copying to increase the frequency distribution of the learning data for each class. As described above, the method for estimating the carbon concentration of the steel output from a converter is characterized in that the frequency distribution of the learning data is uniformized in advance. 3. A solidification temperature of molten steel measured during converter blowing,
When the operation information up to the time of measurement is input, a neural network that outputs the molten steel carbon concentration at the time of measurement is constructed, and when using this neural network to estimate the molten steel carbon concentration during converter blowing, The molten steel carbon concentration during smelting is divided into a plurality of classes set with a predetermined difference width, and the output layer of the neural network is composed of the same number of units as the classes and each of the output layers is composed. From the unit, so that the number of priority selection indicators, which is the standard for selecting the corresponding class, is output, when actually operating, the operation information up to the time of the measurement, and the molten steel solidification temperature during the blowing of the converter The neural network is learned by using the measured value and the analysis result of the molten steel carbon concentration at the time of the measurement, and the coupling coefficient between the units forming the neural network. Estimation method of molten steel carbon concentration in the converter blowing, characterized in that to fix. 4. The learning unit according to claim 3, wherein, when learning the neural network, the output unit of the class to which the analysis result of the molten steel carbon concentration in the converter blowing corresponds is the priority selection index number 1 and the output units of the other classes. The number of priority selection indexes of 0 is input as teacher data, and each output unit outputs the number of priority selection indexes of 0 to 1. Estimation method. 5. The method according to claim 3, wherein, when learning the neural network, as a pre-processing, when the analysis result of the molten steel carbon concentration in the converter blowing has a small frequency of the corresponding class, the operation result data is Make duplicates by copying and increase the frequency distribution of the learning data so that the frequency distribution of the learning data for each class is almost uniform. Estimation method.