Operation condition determination method, operation guidance method, and operation condition determination device

Abstract

To provide an operation condition determination method, an operation guidance method, and an operation condition determination device capable of determining an appropriate operation condition capable of improving product characteristics throughout an annealing process.SOLUTION: An operation condition determination method includes a data division step of dividing an annealing process of heating an object to be heated in an annealing furnace into a plurality of sections and dividing an operation condition in the annealing process in accordance with the divided sections, and an operation condition search step of searching for an operation condition in which a value indicating a product characteristic is optimum for each section using a learned model created for each section and learned using the operation condition in the section as input data and the product characteristic as output data. In the operation condition search step, when an operation condition in an arbitrary section other than a first section of the plurality of sections is searched for, an operation condition in an immediately preceding section is input as input data to the learned model.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to an operation condition determination method, an operation guidance method, and an operation condition determination device. [Background technology]

[0002] In annealing furnaces, thermocouples are installed to measure the furnace temperature, which is the atmospheric temperature inside the furnace, and the furnace temperature is often monitored and controlled through the thermocouples during operation. It is possible to install a thermocouple on the heated object, but in some cases the heated object itself is too hot before entering the annealing furnace to install a thermocouple.

[0003] In addition, in a continuous annealing furnace, the objects to be heated move within the furnace, and it is difficult to install thermocouples inside the furnace so that they can follow that movement. Furthermore, when a large number of objects to be heated are annealed frequently, installing a thermocouple on each object would result in a huge workload. For this reason, in the annealing process, the temperature of the objects to be heated is not measured directly, and only the furnace temperature is monitored using thermocouples installed inside the furnace.

[0004] Heat sources inside an annealing furnace include burners, heaters, induction heating (IH), and other heating devices, and heat penetrates primarily from the outer surface of the object being heated. Therefore, the larger the object being heated, the longer it takes for the heat to penetrate into the interior. For example, when annealing a coil, the thermal conductivity in the lamination direction of the coil is low, making it difficult for heat to penetrate into the interior, so it may take more than five days for the center of the coil cross section to reach the specified temperature.

[0005] Furthermore, if the heated object is large, a temperature distribution occurs within the heated object, which causes uneven heating history depending on the part of the heated object, resulting in variations in the characteristics of the final product. For example, when annealing a coil, the temperature rise near the center of the coil cross section is slow, resulting in insufficient heating and making it impossible to achieve the desired product characteristics.

[0006] In recent years, the speed of data science analysis has improved, and operational analysis using machine learning with big data has been actively studied. For example, Patent Document 1 proposes a method for predicting product characteristics using a neural network, a machine learning method, with manufacturing conditions including temperature as input data in product manufacturing.

[0007] Furthermore, Patent Document 2 proposes a method of mist cooling rather than heating, but of predicting the coil state using machine learning to adjust the cooling time, etc. Furthermore, Patent Document 3 proposes a method of using a neural network to feedforward control the operating conditions during operation in order to optimize the product characteristics, which are the objective variables. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Japanese Patent Application Publication No. 2023-39965 [Patent Document 2] Japanese Patent Application Publication No. 2023-141880 [Patent Document 3] Patent No. 7056592 Summary of the Invention [Problem to be solved by the invention]

[0009] In conventional techniques, in order to stabilize product characteristics, an optimal furnace temperature is determined, for example, by performing a correlation analysis between product characteristics and furnace temperature. However, depending on the type of object to be heated and the purpose of annealing, it may not be possible to uniquely determine the furnace temperature. In other words, it is not sufficient to determine a single target temperature through analysis and simply raise the temperature of the object to that target temperature. Since the appropriate temperature changes even during the annealing process, there are cases where strict management is required, such as controlling the temperature at each stage. This leaves room for improvement in conventional techniques.

[0010] The present invention has been made in view of the above, and an object of the present invention is to provide an operating condition determination method, an operation guidance method, and an operating condition determination device that are capable of determining appropriate operating conditions that can improve product characteristics throughout the entire annealing process. [Means for solving the problem]

[0011] (1) The operating condition determination method according to the present invention comprises: a data division step of dividing an annealing process in which a heated body is heated in an annealing furnace into a plurality of sections and dividing operating conditions in the annealing process according to the divided sections; an operating condition search step of searching for the operating conditions for each section that will maximize the value indicating the product characteristics, using a trained model created for each section and trained using the operating conditions for the section as input data and the product characteristics as output data; Including, In the operating condition search step, when searching for the operating conditions of any section other than the first section of the plurality of sections, the operating conditions of the previous section are input as the input data to the trained model.

[0012] (2) The operating condition determination method according to the present invention is the operating condition determination method described in (1) above, the operating conditions include a size of the object to be heated, a furnace temperature of the annealing furnace, and an operating time of the annealing furnace; The product characteristics include mechanical properties of the product, shape of the product, and magnetic properties of the product; The input data includes, in addition to the operating conditions, a temperature of the heated body analyzed by a physical simulation using the operating conditions, The operating condition searching step searches for the furnace temperature for each of the intervals using an optimization method.

[0013] (3) The method for determining operating conditions according to the present invention is the method for determining operating conditions according to the above (1) or (2), The number of the plurality of sections is three or more.

[0014] (4) The operating condition determination method according to the present invention is the operating condition determination method according to any one of (1) to (3) above, The object to be heated is a steel material.

[0015] (5) The operation guidance method according to the present invention comprises: The operating conditions determined by the method for determining operating conditions according to any one of (1) to (4) above are provided to an operator as guidance.

[0016] (6) The operating condition determination device according to the present invention is a data division unit that divides an annealing process in which a heated body is heated in an annealing furnace into a plurality of sections and divides operating conditions in the annealing process according to the divided sections; an operating condition search unit that searches for the operating conditions for each section that will maximize the value indicating the product characteristics using a trained model that is created for each section and trained using the operating conditions for the section as input data and the product characteristics as output data; Equipped with When searching for the operating conditions of any section other than the first section of the plurality of sections, the operating condition search unit inputs the operating conditions of the previous section as the input data to the trained model. [Effects of the Invention]

[0017] According to the present invention, by dividing the annealing process into a plurality of sections and analyzing in detail the relationship between the operating conditions and the product characteristics, it is possible to determine appropriate operating conditions that can improve the product characteristics throughout the entire annealing process. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of an information processing device for realizing an operation condition determination device, an operation guidance device, and a model creation device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a flowchart showing an example of a model creation method according to an embodiment of the present invention. [Figure 3]FIG. 3 is a diagram showing an example of training data for a certain section divided in the data division step of the model creation method according to the embodiment of the present invention. [Figure 4] FIG. 4 is a diagram for explaining the relationship between explanatory variables and response variables in training data used in the model creation step of the model creation method according to the embodiment of the present invention. [Figure 5] FIG. 5 is a flowchart showing an example of an operating condition determination method according to an embodiment of the present invention. [Figure 6] FIG. 6 is a diagram for explaining that the furnace temperature at which iron loss is minimized is searched for by random search in the operating condition search step of the operating condition determination method according to the embodiment of the present invention. [Figure 7] FIG. 7 is a diagram for explaining the point of fixing the coil width and the coil weight in the operation condition searching step of the operation condition determining method according to the embodiment of the present invention. [Figure 8] FIG. 8 is a diagram for explaining the point of fixing the coil width, coil weight, and thermocouple temperature in the operation condition searching step of the operation condition determining method according to the embodiment of the present invention. [Figure 9] FIG. 9 is a diagram for explaining that in the operating condition search step of the operating condition determination method according to the embodiment of the present invention, the temperature at the end of the second section (thermocouple temperature, coil temperatures 1 and 2) is used as the temperature for the third section to search for the furnace temperature. [Figure 10] FIG. 10 is a diagram for explaining the step of searching for operating conditions in the method for determining operating conditions according to the embodiment of the present invention, in which all required points of the random search are plotted to search for the furnace temperature at which iron loss is minimized. [Figure 11] FIG. 11 is an example of the method for determining operating conditions according to the embodiment of the present invention, and is a graph showing the relationship between the furnace temperature and the coil temperature determined in a comparative example. [Figure 12] FIG. 12 is an example of the method for determining operating conditions according to the embodiment of the present invention, and is a graph showing the relationship between the furnace temperature determined in the comparative example and the furnace temperature determined in the example. DETAILED DESCRIPTION OF THE INVENTION

[0019] An operation condition determination method, an operation guidance method, and an operation condition determination device according to an embodiment of the present invention will be described with reference to the drawings.

[0020] (Device configuration) The configuration of an information processing device 1 for realizing an operation condition determination device, an operation guidance device, and a model creation device according to the embodiment will be described with reference to Fig. 1. The information processing device 1 is realized by, for example, a general-purpose computer such as a workstation or a personal computer, or a server located on a cloud. The information processing device 1 also includes an input unit 10, a storage unit 20, a calculation unit 30, and an output unit 40.

[0021] The operation condition determination device according to the embodiment is realized by the components of the information processing device 1, excluding at least the guidance unit 36 ​​of the calculation unit 30. The operation guidance device according to the embodiment is realized by all the components of the information processing device 1. The model creation device according to the embodiment is realized by the components of the information processing device 1, excluding at least the operation condition search unit 35 and the guidance unit 36 ​​of the calculation unit 30.

[0022] The input unit 10 is an input means for the calculation unit 30, and is realized by an input device such as a keyboard, a mouse pointer, a numeric keypad, etc. The input unit 10 also inputs information required for various calculations in the calculation unit 30.

[0023] The storage unit 20 is configured with storage devices such as an EPROM (Erasable Programmable ROM), a hard disk drive (HDD), a solid state drive (SSD), etc. The storage unit 20 stores, for example, an operation DB (database) that collects operational records of annealing furnaces, etc. The storage unit 20 may also store, as necessary, analysis results from the numerical analysis unit 32, training data after division from the data division unit 33, trained models created by the model creation unit 34, search results for operational conditions from the operational condition search unit 35, etc.

[0024] The calculation unit 30 is configured by a processor (arithmetic processing device) such as a CPU (Central Processing Unit), etc. The calculation unit 30 also functions as a data acquisition unit 31, a numerical analysis unit 32, a data division unit 33, a model creation unit 34, an operating condition search unit 35, and a guidance unit 36.

[0025] The data acquisition unit 31 acquires operational records, for example, from an annealing furnace, an operation DB, etc. The operational records indicate explanatory variables (input data) and objective variables (output data) of a trained model created by the model creation unit 34, which will be described later.

[0026] The explanatory variables correspond to the operating conditions (annealing conditions) when the object to be heated is treated in an annealing furnace. The object to be heated is, for example, a steel material. Examples of the steel material include semi-finished products such as slabs, steel sheets produced by rolling slabs, and coils.

[0027] The explanatory variables include, for example, information about the size of the object to be heated, information about the heat source of the annealing furnace, information about the atmospheric gas in the annealing furnace, and information about the operating time of the annealing furnace. Information about the size of the object to be heated includes, for example, the width, height, inner diameter, outer diameter, plate thickness, weight, and components of the object to be heated. Information about the heat source of the annealing furnace includes, for example, the temperature of a thermocouple installed in the annealing furnace (hereinafter referred to as "thermocouple temperature") and the flow rate of fuel in a burner installed in the annealing furnace (hereinafter referred to as "burner fuel flow rate"). Information about the atmospheric gas in the annealing furnace includes the composition and flow rate of the atmospheric gas.

[0028] Here, in order to search for an optimal furnace temperature capable of improving product characteristics in the operation condition search unit 35 described below, it is preferable to use at least the size of the heated body (width, inner diameter, outer diameter in the case of a coil), the thermocouple temperature, and the operation time of the annealing furnace as explanatory variables. Furthermore, if the temperature of the heated body (temperature at multiple locations on the heated body) can be measured directly, the temperature of the heated body may be included in the explanatory variables. On the other hand, if the temperature of the heated body cannot be measured directly, the temperature of the heated body is calculated based on the operation record in the numerical analysis unit 32 described below and used as an explanatory variable.

[0029] The objective variables are variables corresponding to the characteristics of the object to be heated when it is made into a product (hereinafter referred to as "product characteristics"). The objective variables include, for example, the mechanical characteristics of the product, the shape of the product, and the magnetic characteristics of the product. The magnetic characteristics of the product include, for example, iron loss, magnetic flux density, and coating characteristics. The objective variables may also include an index that quantifies, in some way, the degree of variation in the product characteristics over the entire length of the object to be heated. An example of this index is the difference between the maximum and minimum values ​​of the magnetic characteristics over the entire length of the object to be heated.

[0030] The data acquisition unit 31 acquires operational results (explanatory variables and target variables) when the model creation unit 34 creates a trained model (learning phase) as described above. The data acquisition unit 31 also acquires operational conditions (explanatory variables) when the operational condition search unit 35 searches for operational conditions using the trained model (inference phase). The data acquisition unit 31 also acquires all necessary data from an operation DB or the like during the learning phase. Meanwhile, the data acquisition unit 31 may acquire some of the data required for searching for operational conditions (for example, the thermocouple temperature measured at the end of the previous interval) from an annealing furnace in operation during the inference phase. Details will be described later.

[0031] When the temperature of the heated object cannot be measured directly, the numerical analysis unit 32 performs numerical analysis based on operational records to determine the temperature of the heated object. Specifically, the numerical analysis unit 32 determines the temperature of the heated object through a physical simulation using a physical model. That is, the numerical analysis unit 32 determines the temperatures at multiple locations on the heated object by solving a non-steady heat conduction equation (differential equation) based on the size of the heated object and the thermocouple temperatures.

[0032] In numerical analysis, the shape of the annealing furnace itself is modeled using CAD, and the shape of the heated object that is actually in operation is changed for each analysis opportunity.In this case, the temperature history (temperature history curve) of the heated object is obtained by either providing the thermocouple temperature measured with a thermocouple inside the furnace as the temperature boundary condition, which is the heat source, or by modeling the burner itself inside the furnace and analyzing the flow of combustion gases as well.

[0033] In addition, variations occur in the operation of an annealing furnace each time, and even if the furnace temperature is set to the same value every time, it is not possible to control it accurately every time due to the influence of subtle differences in the position of the heated object, fluctuations in the burner fuel flow rate, seasonal changes in temperature, etc.

[0034] After completing the analysis of the temperature history of the heated object, the numerical analysis unit 32 extracts evaluation indices from the temperature history. Since the evaluation indices vary depending on the purpose, such as the temperature at a specific timing during annealing, the rate of temperature rise in a certain section, and the time during which a predetermined temperature can be maintained, the analyst must arbitrarily determine and extract evaluation indices that are considered important.

[0035] Furthermore, since the numerical analysis can determine the temperature distribution throughout the entire interior of an annealing furnace, in some cases, evaluation indexes may be extracted from history curves of not only the temperature of the heated body but also the furnace temperature, furnace temperature deviation, furnace wall refractory temperature, etc.

[0036] The numerical analysis unit 32 performs numerical analysis not only when the model creation unit 34 creates a trained model (learning phase), but also when the operating condition search unit 35 searches for operating conditions using the trained model (inference phase). In the inference phase, the numerical analysis unit 32 may instantly determine the temperature of the heated body during operation based on data acquired by the data acquisition unit 31 from the operating annealing furnace (for example, the thermocouple temperature measured at the end of the previous interval). Details will be described later.

[0037] The data dividing unit 33 divides the annealing process in which the object to be heated is heated in an annealing furnace into a plurality of sections, and divides the operating conditions in the annealing process according to the divided sections. Furthermore, when the temperature of the object to be heated is calculated by the numerical analysis unit 32, the data dividing unit 33 also divides the temperature of the object to be heated according to the above-mentioned sections.

[0038] As described above, in the annealing process, if the furnace temperature is not precisely controlled during the process, the product characteristics may deteriorate. This means that there is an optimal furnace temperature setting for each time period during operation, and therefore, when the optimal furnace temperature is determined by the operating condition search unit 35 (described later), it is necessary to determine the temperature for each of the multiple time periods.

[0039] The method of dividing the sections varies greatly depending on the type of annealing furnace. For example, if the time from inserting a coil into an annealing furnace until heating and cooling are complete is 120 hours, it can be divided into 12 sections of 10 hours each. In this case, the target furnace temperature is determined for each of the 12 sections. On the other hand, some continuous annealing furnaces for thin sheets have short annealing times, and the process from heating to cooling can be completed in just a few minutes. In this case, the continuous annealing furnace is divided into three zones: the heating zone, the soaking zone, and the cooling zone, so it is sufficient to divide it into three sections accordingly.

[0040] Furthermore, the method of dividing the sections may be spatial division as well as time division. For example, if the heated object is processed in multiple annealing furnaces of different shapes, the section is divided for each annealing furnace, that is, at the locations where the shape of the annealing furnace changes. In this case, information about the shape of the annealing furnace is added to the explanatory variables when creating the trained model, and the shape of the annealing furnace is also trained.

[0041] Specifically, the data dividing unit 33 divides the operational performance acquired by the data acquiring unit 31 and the numerical analysis results (temperature of the heated body) by the numerical analysis unit 32 into the determined number of sections. Then, using the divided data as training data, the model creating unit 34, which will be described later, creates a trained model for each section.

[0042] Here, the number of sections divided by the data dividing unit 33 is preferably at least 3. Since continuous annealing furnaces, rotary annealing furnaces, and the like are often divided into multiple zones or rooms such as a heating zone, a soaking zone, and a cooling zone, at least one section is required for each zone.

[0043] Furthermore, in a batch annealing furnace, the heated object does not move between chambers. However, if the temperature is not divided during the heating process (heating + soaking) and is set to the soaking temperature from the beginning, only one section will be left for the heating process, and the benefits of applying this embodiment will not be realized. Therefore, it is necessary to divide the process into at least three sections: heating, soaking, and cooling. However, three sections is merely the minimum number. Setting a larger number of sections naturally results in a more precise furnace temperature in each section (hereinafter referred to as the "furnace temperature pattern"), increasing the likelihood of improving product characteristics. Furthermore, while it is difficult to determine a maximum number of sections, the greater the number of sections, the longer the time required for the operational condition search step by the operational condition search unit 35.

[0044] Here, the determination of the operational conditions by the operational condition determination device according to the embodiment is assumed to be performed before the operation of the annealing furnace (hereinafter referred to as "pre-operational implementation") and during the operation of the annealing furnace (hereinafter referred to as "in-operational implementation"). In pre-operational implementation, for example, the furnace temperatures for all sections from the start of annealing to the end of annealing are determined in advance, and then the operation is started. In pre-operational implementation, the furnace temperature pattern once determined is not changed during operation, and the operational condition search step is not performed again during operation. Therefore, in the case of pre-operational implementation, time constraints are relaxed, so the number of sections divided by the data division unit 33 may be increased as much as possible, and the furnace temperature pattern may be determined over a long period of time in the operational condition search step.

[0045] On the other hand, in in-operation implementation, the furnace temperature for each section is determined and changed sequentially during operation. Therefore, the input data input to the trained model when searching for the furnace temperature includes the operating conditions measured during operation (for example, the thermocouple temperature measured at the end of the previous section) and the results of numerical analysis (temperature of the heated object) obtained immediately during operation. Therefore, in-operation implementation, there are strict time constraints, and if the numerical analysis takes too long, it becomes impossible to determine the furnace temperature before the heated object enters the next section.

[0046] When the search is performed during operation, for example, if the annealing time of a continuous annealing furnace is 3 minutes and the furnace is divided into 10 sections, the furnace temperature for each section must be determined within 18 seconds in the operational condition search unit 35. Whether the furnace temperature for each section can be determined within 18 seconds depends on the parameter settings of the trained model used to search for the operational conditions. In order to shorten the search time for the operational conditions in the operational condition search unit 35, measures such as reducing the number of explanatory variables or reducing the number of layers of the neural network used as the trained model can be considered.

[0047] The model creation unit 34 creates multiple trained models using the operating conditions for each interval divided by the data division unit 33 as input data and the product characteristics as output data. For example, if the data division unit 33 divides the data into 20 intervals, a total of 20 trained models are created for each interval. The machine learning method used by the model creation unit 34 is not particularly limited, but because both the explanatory variables and the target variables are numerical, it is preferable to use, for example, a multiple regression model, a neural network including a convolutional neural network, or the like.

[0048] The operating condition search unit 35 uses the trained model created for each interval by the model creation unit 34 to search for operating conditions for each interval that will provide the best value indicating the product characteristics (that can improve the product characteristics). The operating condition search unit 35 creates, for each interval, a combination of multiple operating conditions in which operable operating conditions (for example, furnace temperature) are randomly changed from among the operating conditions acquired by the data acquisition unit 31. Then, each combination is input into the trained model, and the operating conditions for which the value indicating the product characteristics output is minimized or maximized are searched for for each interval.

[0049] When searching for the operating conditions of any section except the first section of the multiple sections, the operating condition search unit 35 inputs the operating conditions of the previous section as input data to the trained model. In addition, the operating condition search unit 35 uses an optimization method to search for the furnace temperature, which is one of the operating conditions, for each section. The "operating conditions of the previous section" include, for example, the furnace temperature, the thermocouple temperature, the temperature of the heated body, etc.

[0050] Examples of optimization techniques include random search and genetic algorithms. In this embodiment, the description will be given assuming the use of random search. The operating condition search unit 35 first determines the values ​​of all explanatory variables using a random function and creates a large number of combinations of explanatory variable values ​​(random operating condition patterns). The larger the number of combination sets, the higher the possibility of finding a combination that can improve the value of the objective variable. The operating condition search unit 35 then predicts the value of the objective variable by inputting the values ​​of the explanatory variables for each combination into the trained model. Note that the "value of the explanatory variable" refers to, for example, the size of the heated body, the thermocouple temperature, the temperature of the heated body, etc. Furthermore, the "value of the objective variable" refers to a product characteristic.

[0051] Next, the operating condition search unit 35 extracts a combination of explanatory variables that improves the predicted value of the dependent variable. In this case, for example, only the combination of explanatory variables that most improved the dependent variable may be extracted, or the average of multiple combinations that belong to a top group may be extracted.

[0052] In the random search, the explanatory variables whose values ​​are determined by the random function are only those that can be controlled by the operation, and here, furnace temperature is mainly assumed. Other explanatory variables, such as the size of the heated object, are known at the time of determining which heated object's furnace temperature is to be determined, so they do not need to be determined randomly; known values ​​can be entered for all combinations of explanatory variable values. Furthermore, when randomly allocating values, it is preferable to set the range of the furnace temperature within a range that, for example, changes the product characteristics.

[0053] As described above, the optimal furnace temperature setting for a certain section of the annealing process varies depending on the temperature information (furnace temperature, thermocouple temperature, and temperature of the heated object) from the previous section. Therefore, by including the temperature information from the previous section as an explanatory variable when predicting the target variable in a certain section, the value of the target variable is more likely to be improved. In this case, since the temperature information from the previous section is known (already measured or analyzed), there is no need to randomly assign values ​​using a random search; known values ​​can simply be entered. Note that, as described above, when performing the annealing process during operation, the furnace temperature for the next section must be determined one after another during operation. In this case, the temperature information from the previous section is obtained and those values ​​are immediately used to determine the furnace temperature for the next section.

[0054] The guidance unit 36 ​​provides guidance to the operator on the operating conditions found and determined by the operating condition search unit 35, for example, via the output unit 40. The guidance unit 36 ​​may provide guidance to the operator on the operating conditions (e.g., furnace temperature) that can be manipulated in the operation of the annealing furnace, among the operating conditions. The operator can then improve the product characteristics by changing the operating conditions of the annealing furnace in accordance with the guided operating conditions.

[0055] The output unit 40 is an output means for outputting the calculation results by the calculation unit 30, and is realized by an output device such as a display, a printer, etc. The output unit 40 outputs, for example, the operating conditions (e.g., furnace temperature) searched and determined by the operating condition search unit 35.

[0056] The operation condition determination device and the operation guidance device according to the embodiment can be applied to all devices that heat-treat objects, regardless of the type or shape of the objects.

[0057] (Model creation method) The model creation method according to the embodiment will be described with reference to Figures 2 to 4. As shown in Figure 2, the model creation method according to the embodiment includes a data acquisition step (step S1), a numerical analysis step (step S2), a data division step (step S3), and a model creation step (step S4).

[0058] In the following explanation, it is assumed that the object to be heated is a coil and that the coil is annealed in a rotary annealing furnace. A rotary annealing furnace is divided into multiple chambers, and the temperature of the chambers changes sequentially. In addition, in the rotary annealing furnace, the floor on which the coil is placed rotates, for example, clockwise, and moves to the next chamber every few hours, completing one rotation of the floor, thereby annealing the coil in a specified furnace temperature pattern. After annealing is completed, the coil is shaken off in a separate process and magnetic properties such as iron loss and magnetic flux density are measured along its entire length to determine the coil's magnetic properties.

[0059] A thermocouple is installed on the coil stand on which the coil is placed. Because this thermocouple is installed close to the coil, the temperature measured by this thermocouple is considered to be the coil temperature (hereinafter referred to as "coil temperature") and is managed accordingly. However, the temperature measured by the thermocouple is not actually the temperature inside the coil, but rather a local measurement on the underside of the coil, so the temperature of the entire coil cannot be determined. Therefore, if the magnetic properties of the upper surface of the coil deteriorate, for example, the temperature of the upper surface is unknown, making it impossible to determine the relationship between temperature and magnetic properties, making it difficult to set an optimal furnace temperature. Therefore, in the model creation method according to the embodiment, the coil temperature (temperature of the entire coil) is calculated in a numerical analysis step described below. Each step will be described in detail below.

[0060] <Data acquisition steps> In the data acquisition step, operational records are acquired, for example, from an annealing furnace or an operation database (Step S1). The operational records include at least the coil size (width, inner diameter, outer diameter), the thermocouple temperature on the underside of the coil, the operation time of the annealing furnace, and the magnetic properties of the coil (for example, iron loss).

[0061] <Numerical analysis step> In the numerical analysis step, a numerical analysis is performed based on the operational record to determine the coil temperature (temperature of the entire coil) (Step S2). In the numerical analysis step, the temperatures at multiple locations on the coil (temperature of the entire coil) are determined by solving a non-steady heat conduction equation based on, for example, the coil size (width, inner diameter, outer diameter) and the thermocouple temperature.

[0062] <Data division step> In the data division step, the annealing process is divided into multiple sections (step S3). For example, in the case of a rotary annealing furnace, each chamber is divided into multiple sections. Then, the training data, i.e., the operational results obtained in step S1 and the coil temperature calculated in step S2, are divided according to the divided sections.

[0063] Figure 3 shows an example of training data for a certain section divided in the data division step. In Figure 3, furnace temperature indicates the temperature in each chamber of the rotary annealing furnace. Thermocouple temperature indicates the temperature on the underside of the coil measured by a thermocouple installed on the coil stand. Coil temperatures 1 and 2 indicate the temperature at each location on the coil calculated in the numerical analysis step. In addition, training data for the same section (e.g., all in section 2) in different operations are arranged vertically in Figure 3.

[0064] The training data shown in Figure 3 is prepared for each section divided in the data division step. Furthermore, among the training data shown in Figure 3, the furnace temperature, thermocouple temperature, and coil temperatures 1 and 2 change during annealing, so their values ​​vary depending on the section. On the other hand, among the training data shown in Figure 3, the iron loss is the value after annealing is completed, and the coil width and coil weight do not change during annealing. Therefore, these values ​​do not change depending on the section, and are the same in every section.

[0065] <Model creation steps> In the model creation step, a trained model is created for each section based on the training data created in the data division step, with the operating conditions and numerical analysis results as input data and the product characteristics as output data (step S4).

[0066] For example, in the case of a multiple regression model, the model creation step determines parameters a1, a2, a3, and a4 of the multiple regression equation shown in Fig. 4. In the trained model created in the model creation step, iron loss y is calculated by specifying one value for explanatory variables x1, x2, x3, and x4, as shown in Fig. 4. In this embodiment, the operating conditions for which iron loss y is to be predicted (e.g., furnace temperature, coil width, coil weight, thermocouple temperature, and coil temperatures 1 and 2), that is, a set of specifically specified values ​​for explanatory variables x1, x2, x3, and x4, are referred to as a "request point."

[0067] The training of the training data in the model creation step may be performed only once, or may be retrained as necessary. When training of the training data is performed only once, a trained model is created in the model creation step, and then the trained model continues to be used in subsequent operations. On the other hand, when retraining of the training data is performed, retraining is performed each time using the operational results obtained in subsequent operations. When a trained model is created and applied to operations, new operational results that can be used for training increase as operations are repeated, and therefore, by adding these operational results and numerical analysis results and updating the trained model, prediction accuracy can be improved.

[0068] (Method of determining operating conditions) The method for determining operating conditions according to the embodiment will be described with reference to Figures 5 to 10. As shown in Figure 5, the method for determining operating conditions according to the embodiment includes a data acquisition step (step S11), a numerical analysis step (step S12), and an operating condition search step (step S13).

[0069] <Data acquisition steps> In the data acquisition step, the operating conditions are acquired, for example, from an annealing furnace or an operation database (step S11). The operating conditions include at least the coil size (width, inner diameter, outer diameter), the thermocouple temperature on the underside of the coil, the operating time of the annealing furnace, etc. In addition, in the data acquisition step, the operating conditions are acquired for each section (see step S3 in FIG. 2) that was previously divided when the trained model was created.

[0070] In addition, when the operating conditions are determined during operation of the rotary annealing furnace (when the determination is carried out during operation), the data acquisition step acquires the operating conditions measured during operation (for example, the thermocouple temperature measured at the end of the previous section) from the rotary annealing furnace.

[0071] <Numerical analysis step> In the numerical analysis step, a numerical analysis is performed based on the operating conditions to determine the coil temperature (temperature of the entire coil) (step S12). In the numerical analysis step, the temperatures at multiple locations on the coil (temperature of the entire coil) are determined by solving a non-steady heat conduction equation (differential equation) based on, for example, the coil size (width, inner diameter, outer diameter) and the thermocouple temperature.

[0072] In addition, when the operating conditions are determined while the rotary annealing furnace is in operation (when the determination is performed during operation), in the numerical analysis step, the data acquisition unit 31 instantly determines the coil temperature of the current section during operation based on data acquired from the annealing furnace (for example, the thermocouple temperature measured at the end of the previous section).

[0073] <Operating condition search step> In the operating condition search step, an optimization method is applied to the trained model created for each interval to search for the furnace temperature for each interval (step S13). When random search is used as the optimization method, as shown in Figure 6, for example, the values ​​of each explanatory variable are randomly specified for each interval to create a large number of requirement points (combinations of operating conditions), which are then input into the trained model to predict iron loss for each requirement point. Then, in the operating condition search step, the requirement point at which iron loss is minimized, for example, as shown in a bold frame, is determined as the optimal operating condition.

[0074] Figure 6 shows multiple combinations of operating conditions created by random search in a certain section and the iron loss predicted from them. The vertical direction of Figure 6 lists multiple combinations of operating conditions and iron loss in the same section (for example, all in the second section).

[0075] In the operational condition search step, some operational conditions that do not change during annealing may be fixed. For example, as shown in Figure 7, the coil size (coil width, coil weight) does not change during annealing. Therefore, in the random search, the coil size is fixed and the other operational conditions are randomly specified.

[0076] Furthermore, when determining the operating conditions during operation of the rotary annealing furnace (when the operation is performed during operation), the operating conditions that have become known (confirmed) during operation may be fixed in the operating condition search step. For example, when searching for operating conditions for the third section in the operating condition search step, as shown in Fig. 8, the thermocouple temperatures that have already been measured and become known in the second section are fixed, and the other operating conditions are randomly specified.

[0077] Let's consider the case where operation has been completed up to Section 2 and is about to enter Section 3 during operation. In this case, the operating condition search step searches for the furnace temperature in Section 3, as shown in the graph in Figure 9. As mentioned above, the coil width and coil weight are fixed values. Furthermore, since the thermocouple temperature is constantly measured during operation, the value at the end of Section 2 is used. Furthermore, since coil temperatures 1 and 2 can only be determined by numerical analysis, the value determined at the end of Section 2 is used. Note that coil temperatures 1 and 2, which are the results of numerical analysis, are included as explanatory variables to increase the number of data items and improve analysis accuracy. However, furnace temperature search is possible even if coil temperatures 1 and 2 are not included as explanatory variables.

[0078] Next, as shown in Figure 10, a furnace temperature is randomly specified using a random search, and the operating conditions, including that furnace temperature, are input into the trained model for the third section to predict each iron loss. The furnace temperature at which iron loss is minimized is then determined as the furnace temperature for the third section. For example, in Figure 10, the furnace temperature "504°C," at which iron loss is "0.10," is determined as the furnace temperature for the third section. This process is repeated for all remaining sections to determine the furnace temperature pattern.

[0079] In Figure 10, all operating conditions except for the furnace temperature are fixed values. The trained model created in advance is trained using all operating records that do not have restrictions on coil size, etc. Therefore, if the coil size is fixed in the operating condition search step, data with coil sizes that are far removed will not be given importance, improving the prediction accuracy of the target variable (iron loss).

[0080] As shown in Figure 10, whether or not the optimum furnace temperature is searched for sequentially for each section during operation (whether or not it is possible to carry out the search during operation) depends on whether or not the explanatory variables include operational conditions that can be collected in real time from the annealing furnace during operation. For example, if operational conditions that can be collected in real time, such as the thermocouple temperature on the underside of the coil, are not included, there is no point in determining the furnace temperature for the next section in real time, so the furnace temperatures for all sections are determined in advance (pre-operational implementation is carried out).

[0081] There are countless reasons for variations in operating conditions, such as a malfunctioning heating device, a slightly misplaced coil, low air temperature, etc. These effects are reflected in the thermocouple temperature on the underside of the coil. Therefore, by collecting operating conditions in real time from an annealing furnace during operation and determining the furnace temperature, it is possible to reduce day-to-day variations in operating conditions.

[0082] In FIG. 10, the thermocouple temperature measured in the previous section, Section 2, is included as an explanatory variable. However, the furnace temperature determined in Section 2 may also be included as an explanatory variable. The "furnace temperature determined in Section 2" refers to the furnace temperature in Section 2 at which iron loss is minimized. In this way, in the operating condition search step, when searching for operating conditions for a certain section, by including multiple operating conditions in the previous section as explanatory variables, it is possible to more effectively suppress day-to-day variations in operating conditions.

[0083] (Operation Guidance Method) In the operation guidance method according to the embodiment, the guidance unit 36 ​​provides guidance to an operator on the operating conditions (e.g., furnace temperature) determined by the operating condition determination method. The operator can improve product characteristics by changing the operating conditions of the annealing furnace in accordance with the guided operating conditions.

[0084] According to the operation condition determination device, operation guidance device, and model creation device of the above-described embodiments, the annealing process is divided into a plurality of sections, and the relationship between the operation conditions and the product characteristics is analyzed in detail, thereby making it possible to determine appropriate operation conditions that can improve the product characteristics throughout the entire annealing process.

[0085] (Example) An embodiment of the method for determining operating conditions according to the present invention will be described with reference to Figures 11 and 12. In the following, an example in which the method for determining operating conditions according to the present invention is applied to a batch annealing furnace for annealing coils will be described.

[0086] Fig. 11 shows the furnace temperature pattern and coil temperature determined by the current method, which is a comparative example. In Fig. 11, the horizontal axis represents annealing time, and the vertical axis represents temperature. In Fig. 11, the solid line represents the furnace temperature, and the dashed line represents the coil temperature.

[0087] As shown in Figure 11, the coil annealing involves two temperature rises (first temperature rise, second temperature rise) and two soaking times (first soaking, second soaking), and the furnace temperature is overshot before the first soaking in order to quickly raise the coil temperature to the first soaking temperature. With the method of the comparative example, long-term operation can result in poor magnetic properties, with iron loss (a magnetic property) falling below the value expected from the material's actual capabilities, and defects occur particularly in the outermost coil part of the final product.

[0088] Therefore, the present invention was applied to optimize the operating conditions. The target of optimization was the furnace temperature pattern with respect to time, i.e., the relationship between annealing time and furnace temperature. First, operational records for the past several months were acquired, and the coil temperature was calculated for each operation by numerical analysis, thereby preparing training data including both the operating conditions and the results of the numerical analysis.

[0089] Next, the furnace temperature pattern was divided into 20 intervals, and the variable items of the training data were also divided into 20 intervals. Next, the training data for each interval was trained into a neural network model, and 20 trained models were created. In this case, the operating results of the furnace temperature, coil size, and the coil temperature resulting from the numerical analysis were used as explanatory variables, and the measured value of iron loss was used as the objective variable.

[0090] Next, a random search was performed for each section to determine the furnace temperature at which iron loss was minimized for each section. The furnace temperatures for each section were then connected to create a furnace temperature pattern, which was compared with the current pattern (comparison example) as shown in Fig. 12. In Fig. 12, the horizontal axis represents annealing time and the vertical axis represents furnace temperature. In Fig. 12, the difference in furnace temperature when the reference temperature is 600°C is plotted.

[0091] As shown in Figure 12, the furnace temperature pattern determined in the example is significantly different from the furnace temperature pattern determined in the comparative example in terms of the temperature during the overshoot period of the primary heating. The reason for this is thought to be that the furnace temperature was temporarily overshot to increase the temperature rise rate during the primary heating, but because the outermost coil is the part of the coil where the temperature rises most easily, the temperature rose faster than expected, exceeding the primary soaking temperature for a short period of time.

[0092] In the comparative example, it is believed that the above problem occurred because the coil temperature distribution could not be measured. On the other hand, in the example, when the temperature of the outermost part of the coil was extracted from the coil temperature distribution obtained by numerical analysis, it was confirmed that coils with poor magnetic properties tended to have their outermost part temporarily exceed the primary soaking temperature.

[0093] Therefore, when the furnace temperature pattern determined by the analysis was followed and the temperature during overshoot was slightly lowered, the magnetic property defects were successfully reduced. As described above, it was confirmed that the method for determining operating conditions according to the present invention is effective in improving the product properties of coils.

[0094] The operation condition determination method, operation guidance method, and operation condition determination device according to the present invention have been specifically described above using the preferred embodiment and examples, but the scope of the present invention is not limited to these descriptions and should be broadly interpreted based on the claims. It goes without saying that various changes and modifications based on these descriptions are also included in the scope of the present invention. [Explanation of symbols]

[0095] 1. Information processing equipment 10 Input section 20 Memory section 30 Arithmetic section 31 Data Acquisition Section 32 Numerical Analysis Department 33 Data division section 34 Model Creation Department 35 Operational Conditions Search Department 36 Guidance Department 40 Output section

Claims

1. a data division step of dividing an annealing process in which a heated body is heated in an annealing furnace into a plurality of sections and dividing operating conditions in the annealing process according to the divided sections; an operating condition search step of searching for the operating conditions for each section that will maximize the value indicating the product characteristics, using a trained model created for each section and trained using the operating conditions for the section as input data and the product characteristics as output data; Including, In the operating condition search step, when searching for the operating conditions of any section other than the first section of the plurality of sections, the operating conditions of the previous section are input as the input data to the trained model. Method for determining operating conditions.

2. the operating conditions include a size of the object to be heated, a furnace temperature of the annealing furnace, and an operating time of the annealing furnace; The product characteristics include mechanical properties of the product, shape of the product, and magnetic properties of the product; The input data includes, in addition to the operating conditions, a temperature of the heated body analyzed by a physical simulation using the operating conditions, the operating condition searching step searches for the furnace temperature for each of the sections using an optimization method; The method for determining operating conditions according to claim 1.

3. 2. The method for determining operating conditions according to claim 1, wherein the number of the plurality of sections is three or more.

4. 2. The method for determining operating conditions according to claim 1, wherein the object to be heated is a steel material.

5. An operation guidance method for providing an operator with guidance on the operation conditions determined by the operation condition determination method according to any one of claims 1 to 4.

6. a data division unit that divides an annealing process in which a heated body is heated in an annealing furnace into a plurality of sections and divides operating conditions in the annealing process according to the divided sections; an operating condition search unit that searches for the operating conditions for each section that will maximize the value indicating the product characteristics using a trained model that is created for each section and trained using the operating conditions for the section as input data and the product characteristics as output data; Equipped with When searching for the operating conditions of any section other than the first section of the plurality of sections, the operating condition search unit inputs the operating conditions of the previous section as the input data to the trained model. Operating condition determination device.

Images