JP7544113B2 - Setting condition determination method, manufacturing method for product, manufacturing method for rolled material, and setting condition determination device - Google Patents

Description

The present disclosure relates to a setting condition determination method for determining setting conditions in manufacturing equipment, a manufacturing method for manufactured products, a manufacturing method for rolled material, and a setting condition determination device.

Regarding methods for determining set values in conventional manufacturing equipment, for example, there is a method for determining set values such as the initial roll gap by estimating the amount of rolling tool wear and material temperature in thickness control in a seamless hot rolling process (e.g., Patent Document 1). In addition, various methods for performing mill setup automatically have been proposed, and models have been created that can make predictions even when the operator manually modifies the set values (e.g., Patent Document 2).

In the piercing process of the seamless hot rolling process, the mill setup of the piercing rolling mill determines the outer diameter, length, and wall thickness of the rolled material (blank tube) after piercing rolling. In addition, if the outer diameter of the rolled material before piercing rolling is relatively large, an elongation process is performed after the piercing process in the seamless hot rolling process using a rolling mill (elongator) with a tool arrangement similar to that of the piercing rolling mill. The mill setup of this rolling mill also determines the outer diameter, length, and wall thickness of the rolled material (blank tube) after rolling. Optimizing these setting values (mill setup values) is important from the viewpoint of stable operation and quality assurance. Conventionally, prediction formulas for determining the outer diameter, length, and wall thickness of the blank tube from the mill setup have been created, for example, by the geometric relationship of the tools and rolled material during piercing rolling or by experimental regression formulas.

Japanese Unexamined Patent Publication No. 60-137516 JP 2020-134967 A

The set value determined by the method according to Patent Document 1 is an index of the set value in actual rolling. However, this set value does not necessarily result in a product thickness (or intermediate product thickness) within the target range, and pre-production verification of whether the set value is appropriate for achieving the target dimensions is not taken into consideration. In fact, when thickness control is performed using the set value determined by the method of Patent Document 1, for example, at the beginning of a lot, a considerable number of products (or intermediate products) outside the target range are manufactured. In addition, even with a prediction formula obtained by the geometric relationship of the tool and the rolling material during piercing and rolling or an experimental regression formula, it is difficult to accurately reproduce the piercing phenomenon, which has complex deformation and a high degree of freedom in the setting conditions. Thus, there is room for improvement in the technology related to the method of determining the setting conditions in manufacturing equipment.

In light of these circumstances, the purpose of this disclosure is to provide a setting condition determination method, a manufacturing method for manufactured products, a manufacturing method for rolled material, and a setting condition determination device that can improve the technology related to the method of determining setting values in manufacturing equipment.

(1) A setting condition determination method according to an embodiment of the present disclosure,
A method for determining setting conditions in a manufacturing facility, comprising:
Using a product dimension value prediction model trained based on first learning data including at least one of actual values of set values and actual values of operating conditions in the manufacturing equipment, a product dimension predicted value, which is a predicted value of the product dimension value, is obtained from prediction input data related to the initial setting values and operating conditions of the manufacturing equipment, and a setting value in the manufacturing equipment is determined based on the product dimension predicted value.

(2) As one embodiment of the present disclosure, in (1),
The method includes a setting value prediction step of determining the initial setting value from prediction input data related to the operating conditions by using a setting value prediction model trained on the basis of second learning data including actual values of the operating conditions.

(3) As one embodiment of the present disclosure, in (2),
The second learning data includes at least one of the actual values of the operating conditions, namely, the physical properties of the workpiece, the inlet/outlet dimensions, the inlet/outlet temperatures, the number of times the tool has been used, shape information, and the set reference values for the manufacturing equipment.

(4) As an embodiment of the present disclosure, in (2) or (3),
a product dimension value prediction step of determining the product dimension prediction value from the initial setting value obtained in the setting value prediction step and prediction input data related to operation conditions, using a product dimension value prediction model trained based on the first learning data;
a predicted set value correction step of determining whether the product dimension predicted value obtained in the product dimension value prediction step satisfies a predetermined condition, and if not, correcting the initially set value obtained in the set value prediction step based on the product dimension predicted value obtained in the product dimension value prediction step.

(5) As an embodiment of the present disclosure, in (4),
The product dimension value prediction step includes a guidance function for predicting product dimension values not only from the input data for prediction relating to the initial setting values and operating conditions obtained in the setting value prediction step, but also from the input data for prediction relating to the setting values and operating conditions determined by an operator.

(6) As an embodiment of the present disclosure, in (4) or (5),
The predetermined condition includes a condition that the product dimension predicted value obtained in the product dimension value predicting step is within a range of a product dimension target value.

(7) As an embodiment of the present disclosure, in any one of (4) to (6),
The correction of the initial setting value in the predicted setting value correction step is performed using a correction value obtained from a geometric model of the manufacturing equipment or a correction value obtained using a product dimension correction value prediction model,
The product dimension correction value prediction model is created by learning based on third learning data including at least one of actual values of set values and actual values of operating conditions in the manufacturing equipment.

(8) As an embodiment of the present disclosure, in (7),
The third learning data includes at least one of the physical properties of the workpiece, the inlet and outlet dimensions, the inlet and outlet temperatures, the number of times a tool is used, shape information, and the actual values of the set values in the manufacturing equipment, among the actual values of the set values and the actual values of the operating conditions in the manufacturing equipment, and further includes the actual set value difference between the workpiece in the manufacturing equipment and the outlet dimension value difference between the workpiece and the actual set value difference between the workpiece in the manufacturing equipment and the outlet dimension value difference.

(9) As an embodiment of the present disclosure, in any one of (1) to (8),
The first learning data includes at least one of the physical properties of the workpiece, the inlet/outlet dimensions, the inlet/outlet temperatures, the number of times a tool is used, shape information, and the actual values of the set values in the manufacturing equipment, among the actual values of the set values in the manufacturing equipment and the actual values of the operating conditions.

(10) As an embodiment of the present disclosure, in any one of (1) to (9),
The settings include mill setup values.

(11) A method for producing a product according to an embodiment of the present disclosure includes the steps of:
The manufacturing equipment is controlled in accordance with the set values determined by any one of the setting condition determination methods (1) to (9) to manufacture a product.

(12) A method for producing a rolled material according to an embodiment of the present disclosure includes:
The rolling mill is controlled in accordance with the mill setup values determined by the setting condition determination method (10) to produce rolled material.

(13) A setting condition determination device according to an embodiment of the present disclosure,
The control unit executes any one of the setting condition determination methods (1) to (9).

(14) As an embodiment of the present disclosure, in (2),
The second learning data is obtained by calculating the difference between the actual values of the set values in the manufacturing equipment and the actual values of the product dimensions and the standard data within a production lot of the product, and includes the difference values of the set values in the manufacturing equipment and the difference values of the product dimensions, and includes at least one of the physical properties of the workpiece, the inlet and outlet temperatures, the number of times a tool is used, shape information, and the set standard value in the manufacturing equipment among the actual values of the operating conditions.

The present disclosure provides a setting condition determination method, a manufacturing method for manufactured products, a manufacturing method for rolled material, and a setting condition determination device that can improve technology related to the method of determining setting values in manufacturing equipment.

FIG. 1 is a flowchart showing a setting condition determination method according to an embodiment. FIG. 2 is a diagram showing a result of a comparison evaluation, by RMSE, between conventional control and a mother tube dimensional accuracy of the mother tube rolled with set values determined using the method for determining set conditions according to one embodiment in the production of an intermediate product (mother tube) by seamless hot rolling. FIG. 3 is a diagram showing a seamless hot rolling process line in which a setting condition determination method according to an embodiment is implemented. FIG. 4 is a diagram showing explanatory variables and specific data examples thereof when creating learning data used in a setting condition determination method according to an embodiment. FIG. 5 is a scatter plot showing the relationship between the outer diameter of a mother tube (predicted) predicted using predicted set values determined by a setting condition determination method according to an embodiment and the outer diameter of a mother tube (actually measured), which is an actual value. FIG. 6 is a scatter plot showing the relationship between the outer diameter of a blank pipe (predicted) predicted using the set values determined by the operator and the outer diameter of a blank pipe (actually measured), which is an actual value. FIG. 7 is a scatter plot showing the relationship between the mother pipe length (prediction) predicted using the predicted set values determined by the setting condition determination method according to one embodiment and the mother pipe length (actual measurement), which is an actual value. FIG. 8 is a scatter diagram showing the relationship between the mother pipe length (prediction) predicted using the set values determined by the operator and the mother pipe length (actual measurement), which is an actual value. FIG. 9 is a diagram showing a seamless hot rolling process line different from that shown in FIG. 3 , in which a setting condition determination method according to an embodiment is implemented. FIG. 10 is a diagram showing explanatory variables and examples of specific data thereof when learning data is created that utilizes the difference between actual values of setting values in a manufacturing facility and actual values of product dimensions and reference data within a production lot of the product, and when learning data is created that does not utilize the difference, which are used in a setting condition determination method according to one embodiment. 11 is a scatter diagram showing the relationship between a predicted value of a difference value (predicted difference value) determined by a setting condition determination method using learning data utilizing a difference from within-lot reference data according to an embodiment, and a difference value (actual difference value) between an actual value and within-lot reference data. Here, the subject of the predicted setting value is the plug advance amount, as an example. 12 is a scatter diagram showing the relationship between the difference value between the predicted value and the lot reference data (difference evaluation of predicted value) determined by a setting condition determination method using learning data that does not utilize the difference from the lot reference data according to an embodiment, and the difference value between the actual value and the lot reference data (difference actual value). Here, the target of the predicted setting value is the plug advance amount, as an example.

Below, a setting condition determination method, a manufacturing method for a product, a manufacturing method for rolled material, and a setting condition determination device according to an embodiment of the present disclosure will be described with reference to the drawings. Here, the setting condition is one or more setting values in the manufacturing equipment.

In summary, the setting condition determination method according to this embodiment uses a product dimension value prediction model to obtain predicted values of product dimension values (predicted product dimension values) from the initial setting values of the manufacturing equipment and prediction input data related to the operating conditions, and determines the setting values for the manufacturing equipment based on the predicted product dimension values.

Figure 1 is a flowchart showing the setting condition determination method according to this embodiment. At the start of Figure 1 (step S1), the product dimension value prediction model, the setting value prediction model, and the product dimension correction value prediction model have already been created, so we will first explain how to create each learning model.

Each learning model is created using a past production database.

In creating a product dimension value prediction model, learning is performed based on first learning data that includes at least one of the actual values of the set values in the manufacturing equipment and the actual values of the operating conditions. The first learning data includes at least one of the following explanatory variables among the actual values of the set values in the manufacturing equipment and the actual values of the operating conditions: the physical properties of the workpiece, the inlet/outlet dimensions, the inlet/outlet temperatures, the number of times the tool is used, the shape information, and the actual values of the set values in the manufacturing equipment. Representative examples of the items in Figure 4 include the material steel type classification for the physical properties of the workpiece, the outer diameter of the cylindrical material for the inlet/outlet dimensions, the furnace outlet temperature for the inlet/outlet temperatures, the number of times the tool is used, the number of times it is rolled, the shape information, the plug shape symbol, and the roll spacing for the actual values of the set values in the manufacturing equipment.

When creating a set value prediction model, learning is performed based on second learning data that includes actual values of the operating conditions. The second learning data includes, among the actual values of the operating conditions, at least one of the following as explanatory variables: the physical properties of the workpiece, the inlet/outlet dimensions, the inlet/outlet temperatures, the number of times the tool is used, shape information, and the set reference value for the manufacturing equipment.

In creating the set value prediction model, the difference between the actual set value and the actual product dimension in the manufacturing equipment and the reference data in the manufacturing lot of the product is obtained. Data including the difference value of the set value and the difference value of the product dimension in the manufacturing equipment and the actual value of the operating conditions may be used as the second learning data, and learning based on this data may be performed. The actual value of the operating conditions used in the second learning data at this time includes at least one of the physical properties of the workpiece, the entry side dimension, the entry and exit side temperature, the number of times the tool is used, the shape information, and the set reference value in the manufacturing equipment as explanatory variables among the actual values of the operating conditions. Here, the reference data in the manufacturing lot of the product is data that does not include the actual value of the set value and the actual value of the product dimension when the set value is changed frequently in the same manufacturing lot. Here, the actual value of the set value and the actual value of the product dimension are, for example, the actual value of the set value and the actual value of the product dimension at the end of the same manufacturing lot, or, for example, the average value of the actual value of the set value and the actual value of the product dimension in the stable range of the manufacturing conditions in the same manufacturing lot. This set value prediction model can be used when there are actual values for the set values and product dimensions in a manufacturing facility that has produced at least one production run within the production lot in question.

In creating a product dimension correction value prediction model, learning is performed based on third learning data including at least one of the actual values of the set values in the manufacturing equipment and the actual values of the operating conditions. The third learning data includes at least one of the physical properties of the workpiece, the inlet and outlet dimensions, the inlet and outlet temperatures, the number of times the tool is used, shape information, and the actual values of the set values in the manufacturing equipment as explanatory variables among the actual values of the set values in the manufacturing equipment and the actual values of the operating conditions. The third learning data further includes the actual set value difference with the workpiece in the manufacturing equipment and the outlet dimension value difference with the workpiece as explanatory variables. Here, the actual set value difference indicates the difference before and after the change when the actual set value (actual value of the set value) is changed. Also, the outlet dimension value difference indicates the difference in the outlet dimension value before and after the change of the actual set value.

In creating the learning model, the above learning data based on actual performance values was used, and machine learning was performed using software and hardware tools for algorithm selection, hyperparameter tuning, etc. Any algorithm may be used for machine learning, such as Random Forest or deep learning.

In step S2 (setting value prediction step), an initial setting value is determined. The initial setting value is determined based on the operating conditions (input data for prediction) at the time of setting prediction using a setting value prediction model. In another embodiment, the initial setting value may be a setting value determined by an operator.

In step S3 (product dimension value prediction step), a product dimension value prediction model is used to determine a product dimension value prediction value (product dimension prediction value) based on the prediction setting value (initial setting value) determined in step S2 and the operating conditions (input data for prediction) at the time of product dimension value prediction. Here, if the initial setting value is a setting value determined by the operator, a guidance function may be executed to predict the product dimension value in step S3 and present the product dimension prediction value as output to the operator.

In step S4, it is determined whether the product dimension predicted value determined in step S3 satisfies a predetermined condition. If the product dimension predicted value satisfies the predetermined condition (for example, if the product dimension predicted value is within the target range), the process proceeds to step S5. If the product dimension predicted value does not satisfy the predetermined condition (for example, if the product dimension predicted value is not within the target range), the process proceeds to step S6.

In step S5, mill setup is performed using the predicted setting values (initial settings) determined in step S2, and the process ends (step S8).

In step S6, a correction calculation is performed to obtain a correction value for the predicted set value (initial setting value) determined in step S2 so as to satisfy a predetermined condition. In the correction calculation, a correction value is calculated from the difference between the product dimension predicted value and the product dimension target value determined in step S3, using a geometric model of the manufacturing equipment or a product dimension correction value prediction model. Then, the initial setting value is corrected by the correction value, and the correction set value is determined (calculated).

In step S7, mill setup is performed using the correction set value determined in step S6, and the series of processes ends (step S8). Steps S4, S6, and S7 correspond to the predicted set value correction step.

In accordance with this flowchart, the predicted set value (initial setting value or corrected set value) that satisfies the specified conditions is determined as the set value in the manufacturing equipment, and then the manufacturing equipment is controlled according to the determined set value to manufacture the product. Specifically, for example, a rolling mill can be controlled according to a mill setup value determined as the set value to manufacture rolled material. In this embodiment, the set value includes the mill setup value, but is not limited to the mill setup value.

2 shows the result of determining the set value using the setting condition determination method according to the present embodiment and the result of determining the set value by conventional control (for example, the method of Patent Document 1) in the case of manufacturing an intermediate product by seamless hot rolling. Here, the intermediate product is a blank tube formed by a piercer, which is a piercing process, in the seamless hot rolling process line shown in FIG. 3. Also, the elongator, which is a stretching process in the seamless hot rolling process line shown in FIG. 9, is a mill with the same tool arrangement as the piercer. Therefore, the set value can be determined using the setting condition determination method according to the present embodiment. Therefore, the intermediate product also includes a blank tube formed by an elongator, which is a stretching process, in the seamless hot rolling process line shown in FIG. 9. As shown in FIG. 3, the seamless hot rolling process line includes the steps of heating by a heating furnace, piercing rolling by a piercer, stretching rolling by a mandrel mill, reheating by a reheating furnace, sizing rolling by a reducer, and heat treatment. Also, the seamless hot rolling process line when the outer diameter of the rolled material is relatively large includes the steps shown in FIG. 9. That is, the process includes heating in a heating furnace, piercing rolling in a piercer, elongation rolling in an elongator and plug mill, tube rolling in a reeler, reheating in a reheating furnace, sizing rolling in a sizer, and heat treatment.

As shown in Figure 2, an evaluation was performed using the root mean squared error (RMSE) for predicting the outer diameter and length dimensions of the intermediate product, the raw tube. It was found that the setting condition determination method according to this embodiment reduces the variation by about 20% on average compared to conventional control.

In the setting condition determination method according to this embodiment, the validity of the predicted setting value can be confirmed by determining the predicted product dimension value after determining the predicted setting value (initial setting value). Furthermore, if the predicted setting value does not fall within the range of the target value, the prediction accuracy can be improved compared to conventional control.

Here, in the setting condition determination method according to this embodiment, the steel type of the workpiece, the entry side dimensions, the entry side temperature, the number of times the tool is used, and the shape information are learned as explanatory variables from among the actual values of the operating conditions. Figure 4 is a diagram showing explanatory variables and examples of specific data when learning data is created. The specific data examples are actually shown as specific numerical ranges and specific symbol names according to each explanatory variable, but are represented as xxxx-yyyy in Figure 4. In this embodiment, the amount of data used for learning is approximately 300,000. The algorithm used is gradient boosting.

As shown in Figs. 5 to 8, the prediction accuracy was compared and evaluated when manufacturing a blank tube. The prediction accuracy of the blank tube dimensions predicted using the predicted setting values (initial settings) determined using the setting condition determination method according to this embodiment and the prediction accuracy of the blank tube dimensions predicted using the setting values determined by the operator were compared and evaluated using the mean square error. In the scatter diagrams of Figs. 5 to 8, when a plot is on a line with a slope of 1 shown in each diagram, it indicates that the actual measurement value and the predicted value are equal, and the closer the distance between this line with a slope of 1 and each plot is, the smaller the mean square error is, indicating a higher prediction accuracy. Figs. 5 and 6 compare the prediction accuracy for the blank tube outer diameter. Figs. 7 and 8 compare the prediction accuracy for the blank tube length.

As is clear from these comparisons, the prediction accuracy of the blank pipe dimensions predicted using the setting values determined by the operator is equivalent to the prediction accuracy of the blank pipe dimensions predicted using the predicted setting values determined using the setting value prediction model. In other words, in the setting condition determination method according to this embodiment, good prediction accuracy can be obtained in step S2 whether the setting values determined by the operator or the initial setting values determined using the setting value prediction model are used.

In addition, in the setting condition determination method according to the present embodiment, the difference between the actual values of the set values in the manufacturing equipment and the actual values of the product dimensions and the reference data in the manufacturing lot of the product may be obtained. Then, learning may be performed using the steel type of the workpiece, the entry side dimensions, the entry side temperature, the number of times the tool is used, and the shape information as explanatory variables among the actual values of the operating conditions, including the difference values of the set values in the manufacturing equipment and the difference values of the product dimensions. FIG. 10 is a diagram showing explanatory variables and specific data examples in the seamless hot rolling process line shown in FIG. 9 as an example different from the examples shown in FIGS. 2 to 8. The specific data examples are actually shown with specific numerical ranges and specific symbol names according to each explanatory variable, but are written as xxxx-yyyy in FIG. 10. FIG. 10 shows examples of learning data creation that utilizes the difference between the actual values of the set values and the actual values of the product dimensions in the elongation process and the reference data in the manufacturing lot of the product, and learning data creation that does not utilize the difference. Here, the reference data in a production lot of a product is data that does not include the actual values of the setting values and the actual values of the product dimensions when the setting values are changed frequently in the early part of the same production lot. The actual values of the setting values and the actual values of the product dimensions are, for example, the actual values of the setting values and the actual values of the product dimensions at the end of the same production lot, or, for example, the average values of the actual values of the setting values and the actual values of the product dimensions within the stable range of the production conditions in the same production lot. In this embodiment, the number of data used for learning is approximately 30,000. The algorithm used is gradient boosting.

As shown in Figs. 11 and 12, when manufacturing a blank pipe, a prediction evaluation of the set value determined by the learning model was performed using the setting condition determination method according to this embodiment. In the prediction evaluation, the set values determined by each learning model in the learning data creation utilizing the difference with the reference data in the production lot of the product (Fig. 11) and the learning data creation without utilizing the difference (Fig. 12) were evaluated. In the learning model utilizing the difference, the difference is output as a predicted value, so the output predicted value is the vertical axis of the graph shown in Fig. 11, and the difference of the actual value is the horizontal axis of the graph shown in Fig. 11, and a scatter diagram is created and evaluated. In addition, in the learning model not utilizing the difference, a predicted value of the set value that is not a difference is output. Therefore, for the output predicted value, the value calculated by calculating the difference between the reference data in the production lot of the product and each predicted value (predicted value subjected to the same difference evaluation as the difference evaluation of the actual value) is set as the vertical axis of the graph shown in Fig. 12. In addition, the evaluation was performed by creating a scatter diagram with the difference of the actual value, which is the same value as the horizontal axis of the graph shown in Fig. 11, as the horizontal axis of the graph shown in Fig. 12. In this example, the plug advance amount, which is a set value, is shown as a representative. Based on the relationship that when the prediction accuracy of the set value improves, the prediction accuracy of the dimension value also improves, the prediction accuracy using the set value was evaluated. As a result, when the prediction accuracy of a learning model that utilizes the difference and a learning model that does not utilize the difference were compared using the mean squared error, the learning model that utilizes the difference had better prediction accuracy because the mean squared error value was smaller.

Here, the setting condition determination method according to the present embodiment is executed by a setting condition determination device. The setting condition determination device may be realized by a computer such as a process computer that controls the manufacture of a product (e.g., rolled material). The computer includes, for example, a memory and a hard disk drive (storage device), a control unit (processing device) such as a CPU, and a display device such as a display. The operating system (OS) and application programs for performing various processes can be stored in the hard disk drive, and when executed by the CPU, they are read from the hard disk drive to the memory. In addition, data during processing is stored in the memory, and if necessary, is stored in the HDD. Various functions are realized by organically cooperating hardware such as the CPU and memory with the OS and necessary application programs. For example, the series of processes in FIG. 1 may be executed by, for example, a control unit such as a CPU reading and executing a program. In addition, for example, the product dimension prediction value presented to the operator by the above guidance function may be displayed on a display device.

As described above, the setting condition determination method, product manufacturing method, rolled material manufacturing method, and setting condition determination device according to the present embodiment can perform highly accurate predictions using the above-mentioned configuration and process. This improves the technology related to the method of determining setting values in manufacturing equipment.

Although the embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or amendments based on the present disclosure. Therefore, it should be noted that these modifications or amendments are included in the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined into one or divided. The embodiments of the present disclosure can also be realized as a program executed by a processor included in the device or a storage medium having a program recorded thereon. It should be understood that these are also included in the scope of the present disclosure.

Claims (11)

A method for determining setting conditions in a manufacturing facility, comprising:
a setting value prediction step of determining an initial setting value of the manufacturing equipment from prediction input data related to the operating conditions by using a setting value prediction model trained on the basis of second learning data including actual values of the operating conditions;
a product dimension value prediction step of determining a product dimension prediction value, which is a predicted value of the product dimension value, from the initial setting value and the prediction input data by using a product dimension value prediction model trained on the basis of first learning data including at least one of an actual value of a setting value and an actual value of an operating condition in the manufacturing equipment;
a predicted set value correcting step of determining whether the product dimension predicted value obtained in the product dimension value predicting step satisfies a predetermined condition, and if not, correcting the initial setting value obtained in the set value predicting step based on the product dimension predicted value obtained in the product dimension value predicting step to calculate a corrected set value,
determining to use the initial setting value as a setting value in the manufacturing equipment when the predetermined condition is satisfied, and determining to use the corrected setting value when the predetermined condition is not satisfied;
the correction of the initial set value in the predicted set value correction step is performed using a correction value obtained by using a product dimension correction value prediction model,
A setting condition determination method , wherein the product dimension correction value prediction model is created by learning based on third learning data including at least one of actual values of setting values and actual values of operating conditions in the manufacturing equipment . 2. The method for determining setting conditions according to claim 1, wherein the second learning data includes at least one of the following actual values of the operating conditions: physical properties of the workpiece, dimensions of the material , temperature at the outlet side of the heating furnace , number of times a tool is used, shape information, and a setting reference value which is a range or representative value of specific data based on actual results in the manufacturing equipment . 2. The setting condition determination method according to claim 1 , wherein the product dimension value prediction step includes a guidance function for predicting the product dimension value not only from the input data for prediction relating to the initial setting values and operating conditions obtained in the setting value prediction step, but also from the input data for prediction relating to the setting values and operating conditions determined by an operator. 2. The set condition determining method according to claim 1 , wherein the predetermined condition includes a condition that the product dimension predicted value obtained in the product dimension value predicting step is within a range of a product dimension target value. 2. The setting condition determination method according to claim 1, wherein the third learning data includes at least one of the physical properties of the workpiece, the dimensions of the material , the temperature at the outlet side of the heating furnace , the number of times a tool is used, shape information, and the actual values of the setting values in the manufacturing equipment, among the actual values of the setting values and the actual values of the operating conditions in the manufacturing equipment, and further includes a difference between before and after the change of the actual setting value for the workpiece in the manufacturing equipment and a difference in the dimensions of the material before and after the change of the actual setting value for the workpiece. 2. The setting condition determination method according to claim 1, wherein the first learning data includes at least one of the following: physical properties of the workpiece, dimensions of the material , temperature at the outlet side of the heating furnace , number of times a tool is used, shape information, and actual values of the setting values in the manufacturing equipment, among actual values of the setting values and actual values of the operating conditions in the manufacturing equipment. The method of claim 1 , wherein the set value includes a mill setup value. A method for manufacturing a product, comprising the steps of: controlling the manufacturing equipment in accordance with a set value determined by the method for determining set conditions according to any one of claims 1 to 6 to manufacture the product. A method for producing a rolled material, comprising controlling a rolling mill in accordance with a mill setup value determined by the method for determining setting conditions according to claim 7 to produce the rolled material. A setting condition determination device comprising a control unit that executes the setting condition determination method according to any one of claims 1 to 6 . 2. The method for determining setting conditions according to claim 1, wherein the second learning data includes a difference between an actual value of a setting value in the manufacturing equipment and an actual value of a product dimension value and a reference data within a production lot of the product, the difference between the setting value in the manufacturing equipment and the difference between the product dimension values , and the actual values of the operating conditions include at least one of the physical properties of the workpiece, the temperature on the outlet side of the heating furnace , the number of times a tool is used, shape information, and a setting reference value which is a range or representative value of specific data based on actual values in the manufacturing equipment.

Images