JP2014018844A - Heat transfer coefficient predictor for steel material and cooling control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To highly accurately predict a heat transfer coefficient between steel material and a cooling medium in advance before the steel material reaches a cooling process.SOLUTION: A similarity degree calculation part 10a calculates similarity degree Wbetween a cooling condition x to be predicted and each of multiple cooling conditions xstored in an actual result database 4. A prediction formula creation part 10b creates a prediction model indicating the relation between the cooling condition x and a heat transfer coefficient y, using data of the cooling conditions xstored in the actual result database 4, and determines a model parameter of the prediction model, by solving an optimization problem with an evaluation function for evaluating a prediction error of the prediction model, having, as weights, the similarity degree Wcalculated by the similarity degree calculation part 10a. A heat transfer coefficient prediction part 10c predicts the heat transfer coefficient y when steel material is cooled under the cooling condition x to be predicted, by inputting the cooling condition x to be predicted into the prediction model created by the prediction formula creation part 10b.

Description

  The present invention is based on a heat transfer coefficient prediction device for a steel material that predicts a heat transfer coefficient between the steel material and the cooling medium, and a heat transfer coefficient between the steel material and the cooling medium predicted by the heat transfer coefficient prediction device. The present invention relates to a cooling control method for a steel material for controlling the cooling condition of the steel material.

  Generally, the product quality of a steel material is made by adjusting chemical components in a steelmaking process and performing a heating process, a rolling process, and a cooling process in the rolling process. Material characteristic values such as hardness are the most important quality indicators in the product quality of steel materials. In order to make the product quality of the steel material constant, the chemical composition and the conditions of the heating process, rolling process, and cooling process of the rolling process may be controlled according to the target values at all times. Among these conditions, the steel material temperature at the cooling process delivery side, which is the final stage of the rolling process, greatly affects the product quality of the steel material. However, the temperature of the steel material on the exit side of the cooling process often does not become the target value due to disturbance. The largest disturbance with respect to the steel material temperature on the outgoing side of the cooling process is a change in the heat transfer coefficient between the steel surface and the cooling medium in the cooling process. For this reason, accurately predicting the heat transfer coefficient is very important for quality control and quality control.

  From such a background, a technique for online estimation of a heat transfer coefficient between a steel surface and a cooling medium has been proposed. Specifically, Patent Document 1 discloses a calculated value of the steel material temperature on the cooling process outlet side calculated by a heat transfer calculation model by a search method such as a simulated annealing method using an actually measured temperature and a heat transfer calculation model. And a method for obtaining an optimum heat transfer coefficient that matches the actual measured value of the steel temperature measured by a thermometer. Further, Patent Document 2 uses the measured temperature and the heat transfer calculation model to repeatedly correct two or more parameter coefficients of the heat transfer calculation model, and the steel material at the cooling process outlet side calculated by the heat transfer calculation model. A method for obtaining an optimum heat transfer coefficient in which the calculated value of the temperature and the actual value of the steel material temperature actually measured by the thermometer coincide is described.

JP 2004-244721 A JP 2006-281258 A

  In general, in order to properly control the steel temperature at the cooling process delivery side, the heat transfer coefficient between the steel and the cooling medium is accurately predicted in advance before the steel reaches the cooling process, and the predicted heat transfer. It is necessary to obtain a cooling condition using a coefficient so that the steel temperature at the delivery side of the cooling process becomes a target value. However, the techniques described in Patent Documents 1 and 2 estimate the heat transfer coefficient between the steel material and the cooling medium after the steel material has passed through the cooling process, and the steel material is preliminarily used before the steel material reaches the cooling process. It does not predict the heat transfer coefficient between the air and the cooling medium.

  In a process in which similar steel materials pass through the cooling process one after another, it is considered that the heat transfer coefficient between the steel material and the cooling medium does not change greatly. If the cooling control is performed assuming the same, the temperature of the steel material can be controlled appropriately to some extent. However, in recent years, products with various dimensions, materials, and components have been manufactured in response to diversifying customer requirements. Generally, when dimensions, materials, and components change, the cooling conditions change accordingly, and the heat transfer coefficient between the steel material and the cooling medium changes. For this reason, it has become difficult to control cooling well with the conventional method that assumes the same heat transfer coefficient as that of the immediately preceding steel material.

  The present invention has been made in view of the above-described problems, and the object of the present invention is to heat the steel that can accurately predict the heat transfer coefficient between the steel and the cooling medium in advance before the steel reaches the cooling step. An object is to provide a transfer coefficient prediction apparatus.

  Another object of the present invention is to provide a steel material cooling control method capable of appropriately controlling the steel sheet temperature on the cooling process delivery side to a target value.

  In order to solve the above-described problems and achieve the object, a heat transfer coefficient prediction apparatus for steel according to the present invention is a heat transfer between a steel cooling condition and a steel and a cooling medium in a cooling process performed in the past. A performance database that stores coefficients in association with each other, a similarity calculation unit that calculates a similarity to a cooling condition to be predicted for a plurality of cooling conditions stored in the performance database, and a database that is stored in the performance database. A prediction model representing the relationship between the cooling condition and the heat transfer coefficient is created using the information on the cooling condition, and an evaluation function weighted by the similarity calculated by the similarity calculation unit is By solving an optimization problem as an evaluation function for evaluating a prediction error, a prediction formula creation unit that determines parameters of the prediction model and a prediction formula creation unit The heat transfer coefficient prediction for predicting the heat transfer coefficient between the steel material and the cooling medium when the cooling process is performed under the cooling condition of the prediction target by inputting the cooling condition of the prediction target into the prediction formula created in the above And a section.

  The heat transfer coefficient prediction apparatus for steel according to the present invention is characterized in that, in the above invention, the prediction formula creation unit solves the optimization problem using physical characteristics of a steel material to be predicted as a constraint.

  In the heat transfer coefficient prediction device for steel according to the present invention, in the above invention, the similarity calculation unit calculates a product of a similarity between the prediction target cooling condition and a temporal similarity between the prediction targets as the similarity. It is characterized by doing.

  The steel material heat transfer coefficient prediction apparatus according to the present invention is the above-described invention, wherein the cooling condition used by the performance database, the similarity calculation unit, the prediction formula creation unit, and the heat transfer coefficient prediction unit is a principal component. It is characterized by linear transformation and dimension compression by analysis.

  In order to solve the above-described problems and achieve the object, the steel material cooling control method according to the present invention includes a heat transfer coefficient between the steel material and the cooling medium predicted by the heat transfer coefficient prediction device for steel material according to the present invention. And a step of controlling the cooling conditions of the cooling process based on the above.

  According to the heat transfer coefficient predicting apparatus for steel materials according to the present invention, the heat transfer coefficient between the steel materials and the cooling medium can be predicted with high accuracy before the steel materials reach the cooling step.

  According to the steel material cooling control method of the present invention, it is possible to appropriately control the steel sheet temperature on the outlet side of the cooling process to the target value.

FIG. 1 is a block diagram showing a configuration of a cooling control system according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of performance data stored in the performance database illustrated in FIG. 1. FIG. 3 is a flowchart showing a flow of heat transfer coefficient prediction processing according to an embodiment of the present invention. FIG. 4 is a histogram showing a prediction error of the heat transfer coefficient predicted using the conventional heat transfer coefficient prediction method and the heat transfer coefficient prediction method of the present invention, the actual value of the heat transfer coefficient, and the conventional heat transfer coefficient prediction. It is a figure which shows the relationship with the predicted value of the heat transfer coefficient estimated using the method and the heat transfer coefficient prediction method of this invention. FIG. 5 shows a heat transfer coefficient prediction error histogram and a heat transfer coefficient actual value in the heat transfer coefficient prediction method of the present invention when the time similarity is not considered and when the time similarity is considered. It is a scatter diagram with a predicted value. FIG. 6 shows a heat transfer coefficient prediction error histogram and a heat transfer coefficient in the heat transfer coefficient prediction method of the present invention when the cooling condition is used as it is and when the cooling condition linearly transformed by the principal component analysis is used. It is a scatter diagram of the actual value and the predicted value.

  Hereinafter, the configuration and operation of a cooling control system according to an embodiment of the present invention will be described with reference to the drawings.

[Configuration of cooling control system]
First, the configuration of a cooling control system according to an embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing a configuration of a cooling control system according to an embodiment of the present invention. FIG. 2 is a diagram illustrating an example of performance data stored in the performance database illustrated in FIG. 1.

  As shown in FIG. 1, a cooling control system 1 according to an embodiment of the present invention includes an input device 2, an output device 3, a performance database 4, a heat transfer coefficient prediction device 5, and a cooling control device 6 as main components. As prepared. The input device 2 is configured by an information input device such as a keyboard, a mouse pointer, and a numeric keypad, and is operated when an operator inputs various information to the heat transfer coefficient prediction device 5. The output device 3 is configured by an information output device such as a display device or a printing device, and outputs various processing information of the heat transfer coefficient prediction device 5.

As shown in FIG. 2, each time the steel material passes through the cooling process, the performance database 4 has a cooling condition data of the cooling process and a heat transfer coefficient between the steel material and the cooling water (hereinafter simply referred to as a heat transfer coefficient). Data may be associated with each other and stored as performance data. Specifically, the actual value database 4 includes actual values y ⁿ (where n = 1, 2,..., N) and actual values x ⁿ (= [x ₁ ⁿ , x ₂ ⁿ , .., X _M ⁿ ] ^T ) (where n = 1, 2,..., N, M are the number of input variables) and are stored in association with each other. In this case, the output variable is the heat transfer coefficient, and the input variable is the size of the steel material that is physically and causally related to the heat transfer coefficient, the chemical composition adjusted in the steelmaking process, and the input and output sides of the cooling process. These are the temperature of the steel material, the water density of the cooling facility, and the conveying speed of the steel material (the speed of the steel material passing through the cooling process). The performance database 4 is configured such that old performance data is removed by a method such as a first-in first-out method so that a prediction model can be constructed based on the latest performance data.

  Returning to FIG. The heat transfer coefficient prediction device 5 is configured by an information processing device such as a workstation or a personal computer, and includes a CPU 10, a RAM 11, and a ROM 12 as main components. The CPU 10 controls the overall operation of the heat transfer coefficient prediction device 5. The CPU 10 functions as a similarity calculation unit 10a, a prediction formula creation unit 10b, and a heat transfer coefficient prediction unit 10c by executing a heat transfer coefficient prediction program 12a stored in advance in the ROM 12. The functions of these units will be described later. The cooling control device 6 operates the cooling condition so that the steel temperature on the cooling process delivery side becomes a target value by operating the cooling condition based on the heat transfer coefficient predicted by the heat transfer coefficient prediction device 5.

[Heat transfer coefficient prediction process]
In the cooling control system having such a configuration, the heat transfer coefficient prediction device 5 predicts the heat transfer coefficient of the steel material in the cooling process by executing the heat transfer coefficient prediction process described below. Hereinafter, the operation of the heat transfer coefficient prediction device 5 when the heat transfer coefficient prediction process is executed will be described with reference to the flowchart shown in FIG.

  FIG. 3 is a flowchart showing a flow of heat transfer coefficient prediction processing according to an embodiment of the present invention. The flowchart shown in FIG. 3 starts at the timing when the data of the cooling condition is input by giving the cooling condition of the steel material that passes the cooling process to the input device 2 to the input device 2, and the heat transfer coefficient prediction process is a step. The process proceeds to S1.

In the process of step S <b> 1, the similarity calculation unit 10 a calculates the similarity between the cooling condition data input from the input device 2 and the cooling condition data stored in the record database 4. Specifically, first, the similarity calculation unit 10a selects a point in the input variable space corresponding to the cooling condition input from the input device 2 as the requested point x (≡ [x ₁ , x ₂ ,..., X _M as] ^T), the actual value x ⁿ for each input variable that is stored in the result database 4, to calculate the distance L ⁿ from the request point x using equation (1) below.

In Equation (1), the parameter λ _m is a weighting factor for scaling input variables measured on different scales such as chemical components and temperature. Then, the similarity calculation unit 10a uses the following formula (2) for the actual value x ⁿ of each input variable stored in the actual result database 4 to determine the similarity of a point at a distance L ⁿ from the request point x to calculate the W ^n. In Equation (2), parameter σ _L is a standard deviation of distance L ⁿ represented by Equation (1) with respect to the actual data, and parameter p is an adjustment parameter.

Also, the similarity calculation unit 10a calculates the product of the similarity with respect to the cooling condition of the prediction target and the temporal similarity with the prediction target as the similarity W ⁿ as shown in the following formula (3). Also good. Note that the parameter λ in Equation (3) is a forgetting factor, and is an adjustment parameter having a value greater than 0 and less than 1. By including this forgetting factor, the similarity of the new actual value ^xn is increased, and the similarity of the old actual value ^xn is decreased. By considering the temporal similarity with the prediction target, the heat transfer coefficient can be predicted with higher accuracy. Thereby, the process of step S1 is completed and the heat transfer coefficient prediction process proceeds to the process of step S2.

In the processing of step S2, the prediction formula creation unit 10b uses the N pieces of actual data (actual values x ^{n of} input variables) stored in the actual database 4 and the similarity W ⁿ between the request points x and the results. Then, a local prediction model of the heat transfer coefficient that emphasizes past performance data similar to the request point x is created. Specifically, the prediction formula creation unit 10b creates a heat transfer coefficient prediction model represented by the following formula (4). The model parameter θ represented by the following formula (5) constituting the formula (4) is represented by the following formulas (6) to (9), and the heat transfer coefficient weighted by the similarity W ⁿ Can be calculated by solving an optimization problem that minimizes the value of the evaluation function J, which is the sum of squares of errors between the actual measurement value and the predicted value.

Here, in the formula (7), the parameter y ⁿ (where n = 1, 2,..., N) is the value of the output variable corresponding to the n-th actual data, and in the formula (8), the parameter diag (s) represents a diagonal matrix with the elements of s as main diagonal elements. A local prediction model that better fits actual data with high similarity, i.e., close to the required point x, by calculating a model parameter that minimizes the weighted sum of squares between the predicted value of the heat transfer coefficient and the measured value. Can be created.

  When solving the optimization problem, the optimization problem may be solved by giving the following constraint conditions. Specifically, the constraint condition is expressed by the following formulas (11) to (13) with respect to the range of the partial regression coefficient φ of the input variable in the model parameter represented by the following formula (10). Restrictions may be provided. Here, it is assumed that physical foresight information between the input and output variables is given to the lower limit value and the upper limit value represented by Expression (12) and Expression (13).

  Specifically, the prediction of the heat transfer coefficient in the cooling process of a thick steel plate, which is a kind of steel material, will be described as an example. If the steel material temperature on the cooling process entry side given as an input variable increases, the heat transfer coefficient decreases. Therefore, for the model parameter corresponding to the steel material temperature on the cooling process entry side, the lower limit value and the upper limit value are set to −∞ and 0, respectively. Moreover, if the conveyance speed of the cooling process given as an input variable increases, the heat transfer coefficient decreases. Therefore, for the model parameters corresponding to the conveyance speed of the cooling process, the lower limit value and the upper limit value are set to −∞ and 0, respectively. Furthermore, the hardness increases as the carbon concentration of one chemical component given as an input variable increases. Therefore, for the model parameter corresponding to the carbon concentration, the lower limit value and the upper limit value are set to 0 and ∞, respectively. By adding constraints on the foresight information obtained from the physical model, it is possible to better fit the actual data close to the requested point and create a local prediction model with partial regression coefficients that match the physical characteristics of the prediction target can do. Thereby, the process of step S2 is completed and the heat transfer coefficient prediction process proceeds to the process of step S3.

  In the process of step S3, the heat transfer coefficient prediction unit 10c calculates the predicted value of the heat transfer coefficient of the steel material by substituting the value of the request point x into the prediction model created by the process of step S2. Thereby, the process of step S3 is completed and a series of heat transfer coefficient prediction processes are complete | finished.

Before executing the processes of steps S1 to S3, principal component analysis is performed on the data relating to the cooling conditions used for the processes of the results database 4, the similarity calculation unit 10a, the prediction formula creation unit 10b, and the heat transfer coefficient prediction unit 10c. May be used for linear transformation and dimension compression. By using linear transformation and dimensionally compressed cooling conditions, the heat transfer coefficient can be predicted more accurately. Specifically, the actual value of the cooling condition as an input variable is x ⁿ (= [x ₁ ⁿ , x ₂ ⁿ ,..., X _L ⁿ ] ^T ) (where n = 1, 2,..., N, L Is the number of input variables), first, each actual value ^xn is standardized using Equation (14) below so that the average is 0 and the standard deviation is 1. In Equation (14) _{, xL, av} is an average value of the actual value ^xn , and the denominator value indicates a standard deviation. The actual value x ⁿ of the cooling condition after standardization is ^expressed as z ⁿ (= [z ₁ ⁿ , z ₂ ⁿ ,..., Z _L ⁿ ] ^T ) or the following formula (15).

Next, a covariance matrix V defined by the following equation (16) is calculated, and an eigenvalue of the covariance matrix V and an eigenvector corresponding thereto are calculated. The covariance matrix V has a plurality of non-negative eigenvalues and a plurality of eigenvectors corresponding to them. Therefore, the eigenvectors are rearranged in descending order of the corresponding eigenvalues, and the M eigenvectors extracted in descending order of the corresponding eigenvalues are represented as a matrix P (= [w ₁ , w ₂ ,..., W _M ] ^T ). However, M is a natural number smaller than the number L of input variables, and the matrix P is called a loading matrix. Then, a result obtained by linearly converting the actual value z of the cooling condition using the loading matrix P as shown in the following formula (17) is stored in the actual result database 4.

Similarly, each element of the cooling condition x (= [x ₁ , x ₂ ,..., X _L ] ^T ) to be predicted is _first standardized using the following formula (18), and after the standardization, The cooling condition z (= [z ₁ , z ₂ ,..., Z _L ] ^T ) to be predicted is calculated. Then, a linear transformation using the loading matrix P as shown in the following equation (19) is used as a required point.

[Cooling control processing]
The cooling control device 6 operates the cooling condition based on the heat transfer coefficient between the steel material and the cooling water predicted by the cooling prediction device 5 so that the steel material temperature at the cooling process outlet side becomes a target value. Manipulate conditions. Specifically, when the surface temperature of the steel material is Ts [K], the cooling water temperature is Tw [K], and the heat transfer coefficient is α [W / (m ² · K)], the distance between the steel plate and the cooling water is The heat flux density J [W / m ² ] is expressed as the following mathematical formula (20).

  Given heat transfer coefficient, steel thickness and width, steel cooling process entry temperature, conveyance speed, and area where cooling water is applied to steel, by solving partial differential equation using equation (20) as a boundary condition, The steel material temperature at the delivery side of the steel material can be calculated. Compare the calculated value of the steel temperature of the cooling process delivery side with the target value given in advance, manipulate the area where the cooling water is applied until these values match, solve the partial differential equation and exit the cooling process Calculate the steel material temperature. The number of cooling banks of the cooling facility, that is, the number of nozzles that inject cooling water is determined based on the area that the cooling water takes when the calculated value and the target value finally match. By setting the cooling equipment based on this, it is possible to appropriately control the steel material temperature on the cooling process delivery side.

[Experimental example]
Using the heat transfer coefficient prediction method of the present invention and the conventional heat transfer coefficient prediction method, an experiment in which the heat transfer coefficient of the steel material is predicted in advance before the steel material reaches the cooling process for the accelerated coolant of the thick steel plate The results will be described. Here, the conventional heat transfer coefficient prediction method is a method of prediction assuming that the heat transfer coefficient of the target material is the same as the heat transfer coefficient of the steel material that has just passed through the cooling step.

4 (a) and 4 (b) are histograms showing prediction errors (predicted value-actual value) of the heat transfer coefficient predicted using the conventional heat transfer coefficient prediction method and the heat transfer coefficient prediction method of the present invention, respectively. It is. 4C and 4D show the relationship between the actual value of the heat transfer coefficient and the predicted value of the heat transfer coefficient predicted by using the conventional heat transfer coefficient prediction method and the heat transfer coefficient prediction method of the present invention. FIG. The values of the heat transfer coefficients shown on the horizontal axis of the graphs shown in FIGS. 4A and 4B and the vertical axis and horizontal axis of the graphs shown in FIGS. 4C and 4D are standardized and dimensionless. Value That is, when the heat transfer coefficient before standardization (unit [W / (m ² · K)]) is expressed as y and the heat transfer coefficient after standardization is expressed as Ψ, the relationship is expressed by the following formula (21). There is. Note that the denominator on the right side of Equation (21) and the second term of the numerator indicate the standard deviation and the average value of the actual values y ⁿ (n = 1, 2,..., N) of the heat transfer coefficient, respectively. It is expressed as (22) and (23). That is, the standardization is linearly converted so that the actual value of the heat transfer coefficient is 0 on average and 1 on standard deviation. Since the heat transfer coefficient after standardization is made dimensionless, there is no unit. In FIG. 5 and subsequent figures, the heat transfer coefficient values thus standardized are used for illustration.

  As shown in FIGS. 4A and 4C, the RMSE (Root Mean Square Error) of the prediction error of the heat transfer coefficient predicted by using the conventional heat transfer coefficient prediction method is 0.245. (After standardization). On the other hand, as shown in FIGS. 4B and 4D, the RMSE of the prediction error of the heat transfer coefficient predicted by using the heat transfer coefficient prediction method of the present invention is 0.078 (after standardization). there were. From this, it became clear that according to the heat transfer coefficient prediction method of the present invention, the heat transfer coefficient can be predicted with high accuracy.

  5 (a) and 5 (b) show the heat transfer coefficient prediction error (prediction) when the similarity in time is not considered and when the similarity in time is considered in the heat transfer coefficient prediction method of the present invention. (Value-actual value) histogram. FIGS. 5C and 5D are scatter diagrams of the actual value and the predicted value of the heat transfer coefficient when the similarity in time is not considered and when the similarity in time is considered. Here, 0.995 is selected as the value of the forgetting factor for calculating the similarity in time. In addition, the horizontal axis of the graph shown to Fig.5 (a), (b) is a prediction error (after standardization) of a heat transfer coefficient. In addition, the horizontal axis and the vertical axis of the graphs shown in FIGS. 5C and 5D respectively show the actual value (after standardization) and the predicted value (after standardization) of the heat transfer coefficient. As shown in FIG. 5C, the RMSE of the heat transfer coefficient prediction error when the similarity in time was not taken into account was 0.138 (after standardization). On the other hand, as shown in FIG. 5D, the prediction error of the heat transfer coefficient when considering the similarity in time was 0.085 (after standardization). When the similarity in time is not considered, as shown in FIG. 5C, it can be seen that the absolute value of the error average is large. This can be considered that the learning that sufficiently responds to the secular change in the characteristics of the manufacturing process is not performed sufficiently, and that there is a stationary deviation in the prediction of the heat transfer coefficient. If the similarity in time is not taken into consideration, it is considered that the old data and the new data accumulated in the performance database are handled in the same manner. Therefore, the forgetting factor is introduced so that the similarity in time is taken into account, the similarity of old data is small, and the similarity of new data is large. Thereby, as shown in FIG.5 (d), it turns out that the absolute value of an error average is significantly small. From this, it was found that the heat transfer coefficient can be predicted with higher accuracy by considering the similarity in time.

  6 (a) and 6 (b) respectively show the heat transfer coefficient in the heat transfer coefficient prediction method of the present invention when the cooling condition is used as it is and when the cooling condition linearly transformed by the principal component analysis is used. It is a histogram of a prediction error (predicted value-actual value). 6 (c) and 6 (d) respectively show the distribution of the actual value and the predicted value of the heat transfer coefficient when using the cooling condition as it is and when using the cooling condition linearly transformed by the principal component analysis. FIG. Note that the horizontal axis of the graphs shown in FIGS. 6A and 6B is the heat transfer coefficient prediction error (after standardization). In addition, the horizontal axis and the vertical axis of the graphs shown in FIGS. 6C and 6D respectively indicate the actual value (after standardization) and the predicted value (after standardization) of the heat transfer coefficient.

  As shown in FIG. 6C, the RMSE of the hardness prediction error in the case of using the cooling condition as it is was 0.105 (after standardization). On the other hand, as shown in FIG. 6D, the RMSE of the prediction error of the heat transfer coefficient when using the cooling condition linearly converted by the principal component analysis was 0.084 (after standardization). When using the cooling conditions as they are, it can be seen that the RMSE of the prediction error is large as shown in FIG. This is a state called multicollinearity, in which the actual data of cooling conditions includes a very high correlation, so the cooling conditions to be predicted are put in the model made based on such data. The calculated heat transfer coefficient prediction value tends to have a large error. Therefore, we tried to avoid the problem of multicollinearity by compressing dimensions by principal component analysis. Thereby, as shown in FIG. 6D, it can be seen that the RMSE of the prediction error is significantly reduced. From this, it was found that the heat transfer coefficient can be predicted with higher accuracy by using the cooling condition linearly transformed by the principal component analysis.

  Furthermore, the control error was compared about the steel material temperature of the cooling process delivery side about the accelerated cooling material of a thick steel plate with the conventional cooling control method and the control method of this invention. The conventional cooling control method is a method of controlling the steel material temperature on the cooling process exit side using the heat transfer coefficient predicted using the conventional heat transfer coefficient prediction method, and the control method of the present invention is the method of the present invention. This is a method of controlling the steel temperature at the cooling process delivery side using the heat transfer coefficient predicted using the heat transfer coefficient prediction method. About each method, the RMSE of the control error (actual value-target value) of the steel material temperature at the cooling process delivery side was compared. The RMSE of the control error in the conventional cooling control method was 17.0 [° C.]. On the other hand, the RMSE of the control error in the material control method of the present invention is 12.7 [° C.], and the control error of the steel temperature at the cooling process outlet side can be greatly reduced compared to the conventional cooling control method. It was. From this, it became clear that according to the cooling control method of the present invention, the steel temperature at the cooling process outlet side is brought close to the target value with high accuracy.

As is clear from the above description, according to the heat transfer coefficient prediction process that is an embodiment of the present invention, the similarity calculation unit 10a is configured to obtain a plurality of cooling conditions x ⁿ stored in the performance database 4. The degree of similarity W ⁿ with respect to the cooling condition x to be predicted is calculated, and the prediction formula creation unit 10b uses the data of the cooling condition x ⁿ stored in the result database 4 to calculate the cooling condition x and the heat transfer coefficient y. Prediction models are created by creating a prediction model representing a relationship and solving an optimization problem using an evaluation function weighted by the similarity W ⁿ calculated by the similarity calculation unit 10a as an evaluation function for evaluating the prediction error of the prediction model. By determining model parameters of the model, the heat transfer coefficient prediction unit 10c inputs the cooling condition x to be predicted into the prediction model created by the prediction formula creation unit 10b. The heat transfer coefficient y when the steel material is cooled under the cooling condition x to be predicted is predicted. According to such a configuration, since the prediction model can be automatically adjusted based on the actual value stored in the actual result database 4, the heat transfer coefficient is set in advance before the steel material reaches the cooling process. Predict with high accuracy.

  In addition, according to the heat transfer coefficient prediction process according to an embodiment of the present invention, the prediction formula creation unit 10b solves the optimization problem using the physical characteristics of the steel material to be predicted as a constraint, so that the prediction is contrary to the physical phenomenon. It can suppress that a model is created and can further improve the prediction accuracy of a heat transfer coefficient.

  Furthermore, according to the cooling control process that is one embodiment of the present invention, the cooling process delivery side is based on the heat transfer coefficient prediction value that is more accurate than the prior art by the heat transfer coefficient prediction process that is one embodiment of the present invention. Therefore, the steel plate temperature on the cooling process delivery side can be appropriately controlled to the target value. Here, as the cooling condition to be operated, select the number of cooling banks of the cooling facility, that is, the number of nozzles that inject cooling water, calculate an appropriate value to reach the target steel temperature, and set the cooling facility I did it.

  Although the embodiment to which the invention made by the present inventor is applied has been described above, the present invention is not limited by the description and the drawings that form a part of the disclosure of the present invention according to this embodiment. That is, other embodiments, examples, operational techniques, and the like made by those skilled in the art based on the present embodiment are all included in the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Cooling control system 2 Input device 3 Output device 4 Results database 5 Heat transfer coefficient prediction device 6 Cooling control device 10 CPU
10a Similarity calculation unit 10b Prediction formula creation unit 10c Heat transfer coefficient prediction unit 11 RAM
12 ROM
12a Heat transfer coefficient prediction program

Claims (5)

A performance database that associates and stores the cooling conditions of the steel material and the heat transfer coefficient between the steel material and the cooling medium in the cooling process performed in the past;
For a plurality of cooling conditions stored in the results database, a similarity calculation unit that calculates the similarity to the cooling condition to be predicted; and
A prediction model representing the relationship between the cooling condition and the heat transfer coefficient is created using information on the cooling condition stored in the performance database, and the similarity calculated by the similarity calculation unit is weighted A prediction formula creation unit that determines parameters of the prediction model by solving an optimization problem using the evaluation function as an evaluation function for evaluating a prediction error of the prediction model;
By inputting the cooling condition of the prediction target in the prediction formula created by the prediction formula creation unit, the heat transfer coefficient between the steel material and the cooling medium when the cooling process is performed under the cooling condition of the prediction target is predicted. A heat transfer coefficient predicting unit,
An apparatus for predicting a heat transfer coefficient of a steel material.   The said prediction formula preparation part solves the said optimization problem on the basis of the physical characteristic of the steel material of prediction object, The heat transfer coefficient prediction apparatus of the steel material of Claim 1 characterized by the above-mentioned.   The said similarity calculation part calculates the product of the similarity with respect to the cooling condition of prediction object, and the temporal similarity of prediction object as a similarity, The heat | fever of the steel materials of Claim 1 or 2 characterized by the above-mentioned. Transfer coefficient prediction device.   The cooling conditions used for processing by the results database, the similarity calculation unit, the prediction formula creation unit, and the heat transfer coefficient prediction unit are linearly converted and dimensionally compressed by principal component analysis. The heat transfer coefficient prediction apparatus for steel materials according to any one of claims 1 to 3.   The step of controlling the cooling condition of the cooling process based on the heat transfer coefficient between the steel material and the cooling medium predicted by the heat transfer coefficient prediction device for steel material according to any one of claims 1 to 4. A cooling control method for steel, comprising:

Images