Manufacturing condition determination method, manufacturing method of metal material, and manufacturing condition determination device

Abstract

The manufacturing condition determination method includes a defect prediction step of predicting the presence or absence of defects for each specified range by inputting manufacturing conditions for each specified range into a defect prediction model; a characteristic prediction step of predicting characteristics for each specified range by inputting manufacturing conditions for each specified range into a characteristic prediction model; and a manufacturing condition determination step of extracting manufacturing conditions that are predicted to be free of defects by the defect prediction model from among the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, and determining, from among the extracted manufacturing conditions, the manufacturing conditions under which the characteristics predicted by the characteristic prediction model are maximized or minimized as manufacturing conditions for the metal material.

Description

[Technical Field]

[0001] The present invention relates to a manufacturing condition determination method, a manufacturing method for a metal material, and a manufacturing condition determination device. [Background technology]

[0002] Patent Document 1 discloses a method for predicting the quality of a metal material for each predetermined range by inputting manufacturing conditions for each predetermined range of the metal material in each process into a quality prediction model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2021 / 014804 Summary of the Invention [Problem to be solved by the invention]

[0004] The method disclosed in Patent Document 1 does not take into consideration the occurrence of defects in the product and does not mention improving the properties of the metal material to a range where defects do not occur or where the occurrence of defects is tolerable, so there is room for improvement.

[0005] The present invention has been made in consideration of the above, and aims to provide a manufacturing condition determination method, a manufacturing method for metal material, and a manufacturing condition determination device that can determine manufacturing conditions that optimize the characteristics of a metal material while suppressing the occurrence of defects. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the object, the manufacturing condition determination method of the present invention is a manufacturing condition determination method for improving the quality of a metal material manufactured through one or more processes, and includes: a defect prediction step of predicting the presence or absence of defects for each predetermined range by inputting the manufacturing conditions for each predetermined range into a defect prediction model generated using the manufacturing conditions for each predetermined range as input variables and the presence or absence of defects for each predetermined range as output variables; a characteristic prediction step of predicting characteristics for each predetermined range by inputting the manufacturing conditions for each predetermined range into a characteristic prediction model generated using the manufacturing conditions for each predetermined range for each process as input variables and the characteristics for each predetermined range as output variables; and a manufacturing condition determination step of extracting manufacturing conditions that are predicted to result in no defects by the defect prediction model from the defect prediction model and the multiple manufacturing conditions input to the characteristic prediction model, and determining, from the extracted manufacturing conditions, the manufacturing conditions under which the characteristics predicted by the characteristic prediction model are maximized or minimized as manufacturing conditions for the metal material.

[0007] In addition, in the manufacturing condition determination method of the present invention, in the above invention, when the same manufacturing conditions are set for the specified range over the entire length of the metal material, the manufacturing condition determination step extracts, from the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, manufacturing conditions under which the proportion of the specified range predicted to have defects over the entire length of the metal material is within an acceptable range, and determines, from among the extracted manufacturing conditions, the manufacturing condition under which the statistical value of the characteristics predicted by the characteristic prediction model is maximum or minimum as the manufacturing condition for the metal material.

[0008] In addition, in the manufacturing condition determination method of the present invention, in the above invention, when the same manufacturing conditions are set for the specified range over the entire length of the metal material, the manufacturing condition determination step extracts, from the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, manufacturing conditions under which the proportion of the specified range predicted to have defects over the entire length of the metal material is within an acceptable range, extracts, from the extracted manufacturing conditions, manufacturing conditions within the specified range predicted to have no defects, and determines, from the extracted manufacturing conditions, the manufacturing condition under which the statistical value of the characteristics predicted by the characteristic prediction model is maximum or minimum as the manufacturing condition for the metal material.

[0009] In addition, in the manufacturing condition determination method of the present invention, in the above invention, when the same manufacturing conditions are set for the specified range over the entire length of the metal material, the manufacturing condition determination step extracts manufacturing conditions in the specified range that are predicted to be free of defects from the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, and determines, from among the extracted manufacturing conditions, the manufacturing conditions that result in the maximum or minimum statistical value of the characteristics predicted by the characteristic prediction model as the manufacturing condition for the metal material.

[0010] In addition, in the manufacturing condition determination method of the present invention, in the above invention, when the same manufacturing conditions are set for the specified range over the entire length of the metal material, the manufacturing condition determination step determines, as the manufacturing condition for the metal material, the manufacturing condition among the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model that maximizes the number of specified ranges that are predicted to be free of defects and whose characteristics are within the acceptable range.

[0011] In order to solve the above-mentioned problems and achieve the object, the method for manufacturing a metallic material according to the present invention manufactures a metallic material under manufacturing conditions determined by the above-mentioned method for determining manufacturing conditions.

[0012] In order to solve the above-mentioned problems and achieve the object, the manufacturing condition determination device of the present invention is a manufacturing condition determination device for improving the quality of a metal material manufactured through one or more processes, and comprises: a defect prediction unit that predicts the presence or absence of defects for each predetermined range by inputting the manufacturing conditions for each predetermined range into a defect prediction model generated with the manufacturing conditions for each predetermined range in each process as input variables and the presence or absence of defects for each predetermined range as output variables; a characteristic prediction unit that predicts the characteristics for each predetermined range by inputting the manufacturing conditions for each predetermined range into a characteristic prediction model generated with the manufacturing conditions for each predetermined range in each process as input variables and the characteristics for each predetermined range as output variables; and a manufacturing condition determination unit that extracts manufacturing conditions that are predicted to be free of defects by the defect prediction model from the defect prediction model and the multiple manufacturing conditions input to the characteristic prediction model, and determines, from the extracted manufacturing conditions, the manufacturing conditions under which the characteristics predicted by the characteristic prediction model are maximized or minimized as manufacturing conditions for the metal material. [Effects of the Invention]

[0013] According to the present invention, it is possible to determine manufacturing conditions that take into consideration the occurrence of defects in the product, suppress the occurrence of defects, and optimize the properties of the metal material. [Brief explanation of the drawings]

[0014] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of a manufacturing condition determination device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a flowchart showing an example of the flow of a manufacturing condition determination method according to an embodiment of the present invention. [Figure 3] FIG. 3 is a diagram showing an example of performance data collected by the manufacturing performance collection unit in the manufacturing condition determination method according to the embodiment of the present invention. [Figure 4] FIG. 4 is a diagram showing an example of performance data edited by a production performance editing unit in the manufacturing condition determining method according to the embodiment of the present invention. [Figure 5]FIG. 5 is a diagram showing an example of a case where a metal material is manufactured through a plurality of steps in the manufacturing condition determination method according to the embodiment of the present invention. [Figure 6] FIG. 6 is a diagram showing an example of performance data edited by a multi-process integrated performance editing unit in the manufacturing condition determining method according to the embodiment of the present invention. [Figure 7] FIG. 7 is a map showing the predicted results of the presence or absence of defects when multiple manufacturing conditions are input into a defect prediction model in a predetermined range over the entire length of a metal material in a manufacturing condition determination method according to an embodiment of the present invention. [Figure 8] Figure 8 is a map showing the predicted results of the presence or absence of defects when multiple manufacturing conditions are input into a defect prediction model in a specified range across the entire length of a metal material in a manufacturing condition determination method according to an embodiment of the present invention, and shows manufacturing conditions in which the proportion of the specified range predicted to have defects is within an acceptable range, indicated by a bold frame. [Figure 9] FIG. 9 is a diagram showing the learning results of defect prediction in the embodiment of the manufacturing condition determination method according to the present invention. [Figure 10] FIG. 10 is a diagram showing the verification results in the first embodiment of the manufacturing condition determining method according to the present invention. [Figure 11] FIG. 11 is a diagram showing the verification results in Example 2 of the manufacturing condition determining method according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0015] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A manufacturing condition determination method, a manufacturing method for a metal material, and a manufacturing condition determination device according to embodiments of the present invention will be described with reference to the drawings.

[0016] (Manufacturing condition determining device) The configuration of a manufacturing condition determination device according to this embodiment will be described with reference to Fig. 1. The manufacturing condition determination device is a device for determining manufacturing conditions for improving the quality of a metal material manufactured through one or more processes. Below, an example of determining manufacturing conditions for a metal material, particularly a steel material (e.g., a slab, a steel plate, etc.), will be described, but the manufacturing condition determination device is not limited to metal materials and can be used when determining manufacturing conditions for other materials in general.

[0017] Specifically, the manufacturing condition determination apparatus 1 is realized by a general-purpose information processing device such as a personal computer, a workstation, etc. The manufacturing condition determination apparatus 1 has as its main components a processor such as a CPU (Central Processing Unit) and a memory (main storage unit) such as a RAM (Random Access Memory) and a ROM (Read Only Memory).

[0018] As shown in Fig. 1, the manufacturing condition determination apparatus 1 includes a manufacturing record collection unit 11, a manufacturing record editing unit 12, a multi-process integrated record editing unit 13, a record database 14, a defect prediction model generation unit 15, and a defect prediction unit 16. Furthermore, as shown in Fig. 1, the manufacturing condition determination apparatus 1 includes a characteristic prediction model generation unit 17, a characteristic prediction unit 18, and a manufacturing condition determination unit 19. Note that the elements of the manufacturing condition determination apparatus 1 excluding the defect prediction unit 16, the characteristic prediction unit 18, and the manufacturing condition determination unit 19 constitute a manufacturing condition determination model generation apparatus. In the following description of the manufacturing condition determination apparatus 1, this manufacturing condition determination model generation apparatus will also be described.

[0019] A sensor (not shown) is connected to the manufacturing history collection unit 11. The manufacturing history collection unit 11 collects the manufacturing history of each process in accordance with the measurement cycle of the sensor and outputs the collected information to the manufacturing history editing unit 12.

[0020] The "manufacturing performance" mentioned above includes, for example, the manufacturing conditions of each process, the presence or absence of defects in the metal material manufactured through each process (hereinafter referred to as "presence or absence of defects"), the properties of the metal material manufactured through each process, etc. Furthermore, the "manufacturing conditions" mentioned above include, for example, the components of the metal material in each process, the temperature, pressure, plate thickness, plate threading speed, etc.

[0021] Furthermore, the aforementioned "presence or absence of defects in the metallic material" includes, for example, information regarding whether or not a defect is present in the relevant portion of the metallic material, the probability that a defect is present in the relevant portion of the metallic material (probability of defect occurrence), etc. Furthermore, the aforementioned "defects in the metallic material" include, for example, scratches on the surface of the metallic material, crystal defects, etc. Furthermore, the aforementioned "properties of the metallic material" refer to, for example, values ​​that indicate the material performance of the metallic material. Examples of the properties of the metallic material include the iron loss and tensile strength of the metallic material.

[0022] The manufacturing conditions for each process collected by the manufacturing record collection unit 11 include not only actual values ​​of the manufacturing conditions measured by sensors but also preset values ​​of the manufacturing conditions. That is, since there are cases where sensors are not installed in some processes, in such cases, set values ​​are collected as manufacturing records instead of actual values.

[0023] The manufacturing history collecting unit 11 collects the manufacturing conditions of each process for each predetermined range of the metal material. The manufacturing history collecting unit 11 also evaluates and collects the presence or absence of defects in the metal material manufactured through each process for each of the predetermined ranges. The manufacturing history collecting unit 11 also evaluates and collects the properties of the metal material manufactured through each of the processes for each of the predetermined ranges.

[0024] The "predetermined range" mentioned above refers to a certain range in the longitudinal direction of the metal material, for example, when the metal material is a slab or a steel plate. This predetermined range is determined based on the moving distance (threading speed) of the metal material according to the conveying direction in each process. Specific processing contents of the manufacturing performance collection unit 11 will be described later (see FIG. 2).

[0025] In this embodiment, a case will be described in which a metal material is evaluated for each predetermined range to create a model (a defect prediction model, a property prediction model), and the presence or absence of defects and properties are predicted, thereby determining manufacturing conditions for each predetermined range, but the entire metal material may also be evaluated as a whole. That is, the manufacturing condition determination device 1 may collect manufacturing results for each metal material (for example, for each slab), create a model, and determine manufacturing conditions for each metal material by predicting the presence or absence of defects and properties.

[0026] 1, it is assumed that only one manufacturing performance collection unit 11 is provided, and data on manufacturing performance (hereinafter referred to as "performance data") for each process is collected by this single manufacturing performance collection unit 11. However, for example, multiple manufacturing performance collection units 11 may be provided in accordance with the number of processes, and performance data for each process may be collected by a separate manufacturing performance collection unit 11.

[0027] The manufacturing performance editing unit 12 edits the performance data for each process input from the manufacturing performance collecting unit 11. That is, the manufacturing performance editing unit 12 edits the performance data collected by the manufacturing performance collecting unit 11 in units of time into performance data in units of length of the metal material, and outputs the data to the multi-process integrated performance editing unit 13. The specific processing contents of the manufacturing performance editing unit 12 will be described later (see FIG. 2).

[0028] The multi-process integrated performance editing unit 13 edits the performance data input from the manufacturing performance editing unit 12. The multi-process integrated performance editing unit 13 stores the manufacturing conditions of each process, the presence or absence of defects in the metal material manufactured under these manufacturing conditions, and the properties of the metal material manufactured under these manufacturing conditions in a performance database 14, associating them for each predetermined range.

[0029] The defect prediction model generation unit 15 generates a defect prediction model that predicts the presence or absence of defects in a metal material manufactured under manufacturing conditions based on the manufacturing conditions for each predetermined range in each process stored in the performance database 14. For example, if the output for the manufacturing conditions in each process input to the defect prediction model is a defect occurrence probability, and this defect occurrence probability exceeds a preset threshold, the defect prediction unit 16 predicts the presence of a defect. The defect prediction model generation unit 15 generates the defect prediction model by setting a threshold so that no defects go undetected.

[0030] The defect prediction model generation unit 15 uses a combination of linear regression that emphasizes generalization (PLS regression when multicollinearity is also taken into consideration) and random forest that can take nonlinearity into consideration as a machine learning method. In addition, various other methods such as linear regression, local regression, principal component regression, PLS regression, neural network, regression tree, and random forest can be used.

[0031] The defect prediction unit 16 predicts the presence or absence of defects in a metal material manufactured under given manufacturing conditions using the defect prediction model generated by the defect prediction model generation unit 15. The defect prediction unit 16 predicts the presence or absence of defects for each predetermined range by inputting the manufacturing conditions for each predetermined range into a defect prediction model generated using the manufacturing conditions for each predetermined range of the metal material in each process as input variables and the presence or absence of defects for each predetermined range as output variables. For example, the output for the manufacturing conditions for each process input into the defect prediction model is a defect occurrence probability, and if this defect occurrence probability exceeds a preset threshold, it is predicted that there is a defect, and if the defect occurrence probability does not exceed the threshold, it is predicted that there is no defect.

[0032] The property prediction model generating unit 17 generates a property prediction model that predicts the properties of the metal material produced under the manufacturing conditions stored in the performance database 14 for each predetermined range in each process.

[0033] The characteristic prediction model generation unit 17 uses a combination of linear regression that emphasizes generalization (PLS regression when multicollinearity is also taken into consideration) and random forest that can take nonlinearity into consideration as a machine learning method. In addition, various other methods such as linear regression, local regression, principal component regression, PLS regression, neural network, regression tree, and random forest can be used.

[0034] The property prediction unit 18 predicts the properties of a metal material manufactured under arbitrary manufacturing conditions using the property prediction model generated by the property prediction model generation unit 17. The property prediction unit 18 predicts the properties for each predetermined range by inputting the manufacturing conditions for each predetermined range into the property prediction model generated using the manufacturing conditions for each predetermined range in each process as input variables and the properties for each predetermined range as output variables.

[0035] The manufacturing condition determination unit 19 determines the manufacturing conditions for the metal material based on the prediction results from the defect prediction unit 16 and the characteristic prediction unit 18. Specific processing contents by the manufacturing condition determination unit 19 will be described later (see FIG. 2).

[0036] (Manufacturing condition determination method) The manufacturing condition determination method according to this embodiment will be described with reference to Figures 2 to 8. The manufacturing condition determination method according to this embodiment performs the processes of steps S1 to S8 shown in Figure 2.

[0037] First, the manufacturing performance collection unit 11 collects performance data relating to the manufacturing conditions of the metal material in each process, the presence or absence of defects in the metal material, and the characteristics of the metal material (step S1). The manufacturing performance collection unit 11 collects performance data relating to the manufacturing conditions, the presence or absence of defects, and the characteristics of each process for each metal material.

[0038] The performance data collected by the manufacturing performance collection unit 11 is data in which performance values ​​(or set values) of a plurality of manufacturing conditions are arranged for each time, as shown in the table of Fig. 3. The performance data shown in Fig. 3 is data in which performance values ​​(or set values) of a plurality of manufacturing conditions are arranged for each time, as shown in the table of Fig. 3. 1 ,t 2 ...and the speed of the metal material at that time (threading speed) v1 ,v 2 ...and multiple manufacturing conditions measured by the sensor at that time x1 1 ,x1 2 …,x2 1 ,x2 2 ...and has items consisting of

[0039] Of the multiple processes, the performance data collected in the final process includes items related to the presence or absence of defects and characteristics of the metal material in addition to the items shown in Figure 3. In this case, the presence or absence of defects and characteristics of the metal material may include information on the presence or absence of multiple different defects and information on multiple different characteristics.

[0040] Next, the manufacturing history editing unit 12 converts the performance data collected by the manufacturing history collecting unit 11 into length units of the metal material (step S2). That is, the manufacturing history editing unit 12 converts the performance data collected in time units as shown in Fig. 3 into performance data in length units of the metal material as shown in Fig. 4. A method for converting the performance data in Fig. 3 into the performance data in Fig. 4 will be described below.

[0041] First, the production history editing unit 12 calculates the position (position coordinates) of the metal material at each time in Figure 3, utilizing the property that multiplying time by speed (threading speed) results in distance. Next, the production history editing unit 12 detects the leading and trailing ends of the metal material, utilizing the property that performance data is recorded when the metal material passes through sensors installed in each process, and missing values ​​are recorded when the metal material does not pass through. Next, the production history editing unit 12 creates performance data corresponding to the positions from the leading end to the trailing end of the metal material, except when the metal material does not pass through the sensor.

[0042] Although the data is in units of the length of the metal material, it is not periodic data. Therefore, the manufacturing history editing unit 12 converts the data into length-based, periodic performance data for the metal material by, for example, linear interpolation. That is, in each process, if the threading speed of the metal material is slow, the performance data that can be collected will be finer, and if the threading speed of the metal material is fast, the performance data that can be collected will be coarser. Therefore, the manufacturing history editing unit 12 performs the interpolation described above to make the performance data uniform. By performing the above processing, the manufacturing history editing unit 12 creates performance data in units of the length of the metal material, as shown in FIG. 4.

[0043] Next, the multi-process integrated performance editing unit 13 aligns and combines the performance data of all processes in units of the length of the metal material (step S3). The multi-process integrated performance editing unit 13 aligns and combines the performance data in units of the length of the metal material created by the manufacturing performance editing unit 12 in units of the length of the metal material at the outlet side of the final process.

[0044] In this way, the multi-process integrated performance editing unit 13 associates the manufacturing conditions of the metal material in each process with the presence or absence of defects and characteristics of the metal material manufactured under these manufacturing conditions for each predetermined range in the length direction of the metal material, and stores the information in the performance database 14. An example of processing by the multi-process integrated performance editing unit 13 will be described below.

[0045] For example, consider the case where a metal material (material) is manufactured through steps 1, 2, and 3, as shown in Figure 5. Steps 1 to 3 are, for example, rolling steps, and the longitudinal length of the material increases with each step. Also, as shown in Figure 5, material A is divided into material A1 and material A2 when moving from step 1 to step 2, and material A1 is divided into material A11 and material A12 when moving from step 2 to step 3.

[0046] The multi-process integrated performance compilation unit 13 considers the performance data for each process and combines the performance data for all processes, which are collected in detail in the longitudinal direction by a sensor (not shown), into the length unit of the metal material in the final process. At that time, as shown in Figure 5, the material lengths of processes 2 and 1 are scaled to match the material length of process 3, which is the final process (see the dashed line in Figure 5).

[0047] The multi-process integrated performance compilation unit 13 then identifies the location where each metal material was taken, and for each predetermined range of the metal material in the final process, associates the presence or absence of defects in the predetermined range, the characteristics in the predetermined range, and the manufacturing conditions for all processes in the predetermined range, and stores these in the performance database 14. For example, in Figure 5, the shaded area where material A11 was taken in process 3, the final process, is identified by tracing back to material A1 in process 2 and material A in process 1. By repeating this process for all metal materials, performance data is created in which multiple manufacturing conditions, presence or absence of defects, and characteristics of the metal material in all processes are aligned and combined in units of the length of the metal material, as shown in Figure 6. Returning to Figure 2, the explanation of step S4 and subsequent steps will continue below.

[0048] The defect prediction model generation unit 15 generates a defect prediction model that predicts the presence or absence of defects for each predetermined range of the metal material from the manufacturing conditions for each predetermined range of the metal material in each process (step S4). Subsequently, the defect prediction unit 16 predicts the presence or absence of defects for each predetermined range of the metal material manufactured under any manufacturing conditions using the defect prediction model generated by the defect prediction model generation unit 15 (step S5).

[0049] In step S5, the defect prediction unit 16 predicts the presence or absence of defects for each manufacturing condition by inputting, for example, multiple manufacturing conditions of the same type into a defect prediction model. For example, if the manufacturing condition is temperature, multiple manufacturing conditions (e.g., α°C, α+10°C, α+20°C, etc.) in which a predetermined value is added to (or subtracted from) a reference temperature are input into the defect prediction model. Note that, when multiple types of manufacturing conditions (e.g., temperature, pressure, etc.) are input, multiple manufacturing conditions in which a predetermined value is added to (or subtracted from) each manufacturing condition are similarly input into the defect prediction model.

[0050] Next, the property prediction model generation unit 17 generates a property prediction model that predicts the property of each predetermined range of the metal material from the manufacturing conditions for each predetermined range of the metal material in each process (step S6). Next, the property prediction unit 18 predicts the property of each predetermined range of the metal material manufactured under any manufacturing conditions using the property prediction model generated by the property prediction model generation unit 17 (step S7).

[0051] In step S7, the characteristic prediction unit 18 predicts the characteristics for each manufacturing condition by inputting, for example, a plurality of manufacturing conditions of the same type into the characteristic prediction model. For example, if the manufacturing condition is temperature, a plurality of manufacturing conditions (e.g., α°C, α+10°C, α+20°C, etc.) in which a predetermined value is added to (or subtracted from) a reference temperature are input into the characteristic prediction model. Note that, when inputting a plurality of types of manufacturing conditions (e.g., temperature, pressure, etc.), a plurality of manufacturing conditions in which a predetermined value is added to (or subtracted from) each manufacturing condition are similarly input into the characteristic prediction model.

[0052] Next, the manufacturing condition determination unit 19 determines, from among the multiple manufacturing conditions, manufacturing conditions that result in no defects and in the best characteristics (or within an acceptable range), as manufacturing conditions for the metal material (step S8). In step S8, the manufacturing condition determination unit 19 determines, from among the multiple manufacturing conditions input to the defect prediction model and the characteristic prediction model, manufacturing conditions that are predicted by the defect prediction model to be defect-free and that are predicted by the characteristic prediction model to have predetermined characteristics, as manufacturing conditions for the metal material. In step S8, the manufacturing conditions for the metal material can be determined using multiple methods. Each method will be described below.

[0053] <First Method> The first method assumes that different manufacturing conditions are set for a given range of metal materials. In this method, first, from among the multiple manufacturing conditions input to the defect prediction model and the characteristic prediction model, manufacturing conditions predicted by the defect prediction model as being defect-free are extracted. Next, from among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the characteristics predicted by the characteristic prediction model are determined as the manufacturing conditions for the metal material. Note that the above "maximum or minimum" also includes values ​​quasi-maximum and minimum, respectively.

[0054] For a property such as tensile strength, where a larger value is considered better, the manufacturing conditions for the metallic material are determined to be those that maximize the property predicted by the property prediction model.For a property such as iron loss, where a smaller value is considered better, the manufacturing conditions for the metallic material are determined to be those that minimize the property predicted by the property prediction model.

[0055] Figure 7 shows an example of a map of defect prediction results, where the horizontal axis represents a predetermined range (20 divisions) across the entire length of a metal material, and the vertical axis represents five levels of manufacturing conditions. Fig. 7 shows the results of predicting the presence or absence of defects when five manufacturing conditions (α°C, α+10°C, α+20°C, α+30°C, and α+40°C) are input into a defect prediction model. In Fig. 7, the white squares (hereinafter referred to as "white squares") represent areas (predetermined range) predicted to be free of defects, and the black squares (dotted) represent areas predicted to be defective.

[0056] In this method, as shown in Fig. 7, within a specific predetermined range (e.g., predetermined range number 1) of a metal material, manufacturing conditions (e.g., α°C, α + 10°C, α + 20°C, α + 30°C) predicted to be free of defects are extracted. Next, the predicted characteristics for the specific predetermined range are compared, and the manufacturing conditions corresponding to the predetermined range predicted to have the best characteristics are determined as the manufacturing conditions for predetermined range number 1. Then, the same process is performed for the remaining predetermined range numbers 2 to 20, and manufacturing conditions for each predetermined range are determined.

[0057] In this method, the characteristics are predicted only for a predetermined range predicted to be free of defects. That is, only the manufacturing conditions corresponding to the predetermined range predicted to be free of defects by the defect prediction model (e.g., "α°C," "α+10°C," "α+20°C," and "α+30°C" in FIG. 7) are input to the characteristic prediction model.

[0058] <Second Method> The second method assumes that the same manufacturing conditions are set for a predetermined range across the entire length of the metal material. In this method, first, from among the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, manufacturing conditions are extracted that result in an acceptable range of predicted defects across the entire length of the metal material (e.g., within 20%). Next, from among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the statistical value of the characteristic predicted by the characteristic prediction model are determined as the manufacturing conditions for the metal material. Examples of "statistical value" include the average, maximum value, minimum value, and mode. Furthermore, the above "maximum or minimum" also includes values ​​approximating the maximum and minimum, respectively.

[0059] For a property such as tensile strength, where a larger value is considered better, the manufacturing conditions for the metallic material are determined to be those that maximize the statistical value of the property predicted by the property prediction model.For a property such as iron loss, where a smaller value is considered better, the manufacturing conditions for the metallic material are determined to be those that minimize the statistical value of the property predicted by the property prediction model.

[0060] Figure 8 shows an example of a map of defect prediction results, where the horizontal axis represents a predetermined range (20 divisions) across the entire length of a metal material, and the vertical axis represents five levels of manufacturing conditions. That is, Figure 8 shows the results of predicting the presence or absence of defects when five manufacturing conditions (α°C, α+10°C, α+20°C, α+30°C, and α+40°C) are input into a defect prediction model. In Figure 8, the white squares (hereinafter referred to as "white squares") represent areas (predetermined range) predicted to be free of defects, and the black squares (dotted) represent areas predicted to be defective.

[0061] In this method, as shown in Figure 8, manufacturing conditions are extracted that result in a defect rate within an acceptable range (for example, within 20%) across the entire length of the metal material. In Figure 8, two manufacturing conditions, "α°C" and "α + 10°C," are extracted, as indicated by the bold frame. Next, statistical values ​​of predicted properties are calculated for each specified range (all cells) of "α°C." Similarly, statistical values ​​of predicted properties are calculated for each specified range (all cells) of "α + 10°C." The statistical values ​​of the properties for "α°C" and "α + 10°C" are then compared, and the manufacturing conditions corresponding to the best statistical value of the properties are determined as the manufacturing conditions across the entire length of the metal material.

[0062] In this method, the characteristics are predicted using only manufacturing conditions under which the percentage of the predetermined range predicted to have defects falls within an acceptable range (for example, within 20%). That is, only manufacturing conditions under which the percentage of the predetermined range predicted to have defects by the defect prediction model falls within an acceptable range (for example, all the cells of "α°C" and all the cells of "α+10°C" in FIG. 8) are input to the characteristic prediction model.

[0063] <Third Method> The third method assumes that the same manufacturing conditions are set for a predetermined range across the entire length of the metal material. In this method, first, from among the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, manufacturing conditions are extracted that result in a percentage of the predetermined range predicted to have defects across the entire length of the metal material that is within an acceptable range (e.g., within 20%). Next, from among the extracted manufacturing conditions, manufacturing conditions within the predetermined range predicted to be free of defects are extracted. Next, from among the extracted manufacturing conditions, the manufacturing conditions that result in the maximum or minimum statistical value of the characteristic predicted by the characteristic prediction model (e.g., average, maximum, minimum, mode, etc.) are determined as the manufacturing conditions for the metal material. Note that the above "maximum or minimum" also includes values ​​quasi-maximum and minimum, respectively.

[0064] For a property such as tensile strength, where a larger value is considered better, the manufacturing conditions for the metallic material are determined to be those that maximize the statistical value of the property predicted by the property prediction model.For a property such as iron loss, where a smaller value is considered better, the manufacturing conditions for the metallic material are determined to be those that minimize the statistical value of the property predicted by the property prediction model.

[0065] In this method, as shown in Figure 8, manufacturing conditions are extracted that result in a defect rate within an acceptable range (e.g., within 20%) across the entire length of the metal material. In Figure 8, two manufacturing conditions, "α°C" and "α + 10°C," are extracted, as indicated by the bold frame. Next, from each predetermined range (all cells) of "α°C," manufacturing conditions for the predetermined range (white cells) predicted to be defect-free are extracted, and the statistical values ​​of predicted characteristics for the extracted predetermined range are calculated. Similarly, from each predetermined range (all cells) of "α + 10°C," manufacturing conditions for the predetermined range (white cells) predicted to be defect-free are extracted, and the statistical values ​​of predicted characteristics for the extracted predetermined range are calculated. The statistical values ​​of the characteristics for "α°C" and "α + 10°C" are then compared, and the manufacturing conditions corresponding to the best statistical value of the characteristics are determined as the manufacturing conditions for the entire length of the metal material.

[0066] In this method, the characteristics are predicted using only manufacturing conditions within a predetermined range where the proportion of a predetermined range predicted to have defects is within an acceptable range (for example, within 20%) and where the proportion of a predetermined range predicted to have defects is within an acceptable range and where the proportion of a predetermined range predicted to have defects by the defect prediction model is within an acceptable range and where the proportion of a predetermined range predicted to have defects ... (for example, the white box for "α°C" and the white box for "α+10°C" in FIG. 8) are input to the characteristic prediction model.

[0067] <Fourth Method> The fourth method assumes that the same manufacturing conditions are set for a predetermined range across the entire length of the metal material. In this method, first, from among the multiple manufacturing conditions input into the defect prediction model and the characteristic prediction model, manufacturing conditions within a predetermined range that are predicted to be defect-free are extracted. Next, among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the statistical values ​​(e.g., average, maximum, minimum, mode, etc.) of the characteristics predicted by the characteristic prediction model are determined as the manufacturing conditions for the metal material. Note that the above "maximum or minimum" also includes values ​​approximating the maximum and minimum, respectively.

[0068] In this method, as shown in Figure 7, from among the predetermined ranges (all cells) of "α°C," the manufacturing conditions for the predetermined ranges (white cells) predicted to be defect-free are extracted, and the statistical values ​​of predicted characteristics for the extracted predetermined ranges are calculated. Similarly, for "α+10°C," "α+20°C," "α+30°C," and "α+40°C," the manufacturing conditions for the predetermined ranges (white cells) predicted to be defect-free are extracted, and the statistical values ​​of predicted characteristics for the extracted predetermined ranges are calculated. Next, the statistical values ​​of the characteristics for "α°C," "α+10°C," "α+20°C," "α+30°C," and "α+40°C" are compared. The manufacturing conditions corresponding to the best statistical value of the characteristics are then determined as the manufacturing conditions for the entire length of the metal material.

[0069] <Fifth Method> The fifth method assumes that the same manufacturing conditions are set for a predetermined range along the entire length of the metal material. In this method, the manufacturing conditions that maximize the number of predetermined ranges predicted to be defect-free and whose properties are within the allowable range are determined as the manufacturing conditions for the metal material from among the multiple manufacturing conditions input to the defect prediction model and the property prediction model.

[0070] It is also possible that the defect prediction model will predict that all of the specified ranges will be "defective." In this case, the characteristics predicted by the characteristic prediction model will not be taken into consideration, and evaluation will be based solely on the defect occurrence probability calculated by the defect prediction model. In other words, the manufacturing conditions for the metal material will be determined to be those that minimize the defect occurrence probability in each specified range predicted to be defective.

[0071] (Metal material manufacturing method) The manufacturing condition determination method according to the embodiment can also be applied to a manufacturing method of a metallic material, in which manufacturing conditions are determined by the manufacturing condition determination method according to the embodiment, and the metallic material is manufactured under the determined manufacturing conditions.

[0072] (Example) An example of the manufacturing condition determination method according to the embodiment will be described with reference to Figures 9 to 11. In this example, offline verification was carried out using the manufacturing condition determination method according to the embodiment when the metal material was a grain-oriented electrical steel sheet.

[0073] In this example, the objective variable for defect prediction is the presence or absence of secondary recrystallization defects in the grain-oriented electrical steel sheet after production. The explanatory variables (manufacturing conditions) for defect prediction include the chemical composition of the metallic material in the steelmaking process, the temperature of the slab cut after casting in the casting process, and the extraction temperature of the slab in the slab heating process in the heating furnace. The explanatory variables (manufacturing conditions) for defect prediction also include the temperature of the steel sheet in the hot rolling process (the sheet surface temperature at the entry side of the finishing rolling mill, the intermediate temperature within the mill, and the exit side of the rolling mill), the temperature of the steel sheet in the cooling process, the temperature of the steel sheet in the cold rolling process, and the temperature of the metallic material in the annealing process. It is said that controlling secondary recrystallization is important for stable production of grain-oriented electrical steel sheet.

[0074] In this example, the objective variable for property prediction is the iron loss of the grain-oriented electrical steel sheet after production. The explanatory variables (manufacturing conditions) for property prediction include the chemical composition of the metallic material in the steelmaking process, the temperature of the slab cut after casting in the casting process, and the temperature of the slab extracted from the heating furnace during the slab heating process. The explanatory variables (manufacturing conditions) for property prediction also include the temperature of the steel sheet in the hot rolling process (the sheet surface temperature at the entry side of the finishing rolling mill, the intermediate temperature within the mill, and the exit side of the rolling mill), the temperature of the steel sheet in the cooling process, the temperature of the steel sheet in the cold rolling process, and the temperature of the metallic material in the annealing process. The iron loss is defined as the power loss per unit weight (W17 / 50) when excited to a maximum magnetic flux density of 1.7 T at a commercial frequency of 50 Hz.

[0075] The defect and property prediction method used a combination of linear regression and random forest. Only the temperature of the annealing process, which is the manipulated variable determined in this example, was used as a linear term, and the other terms were used as nonlinear terms. The defect prediction model and property prediction model had 2,000 training data sets and 700 validation data sets.

[0076] FIG. 9 shows the results of defect prediction learning. FIG. 9 shows the frequency distribution of defect occurrence probability when there is a defect and when there is no defect. In this example, the threshold value set for determining the presence or absence of a defect is set to "0.35" to avoid non-detection. Under this premise, the temperature of the annealing process was varied in 10°C increments from α°C to α+40°C, and the temperature at which no defects occurred and the best characteristics were obtained was selected from these five temperatures.

[0077] In Example 1, the first technique of the manufacturing condition determination method according to the embodiment was used to verify manufacturing conditions under which no defects occur within each predetermined range of the electromagnetic steel sheet and iron loss is minimized (700 verifications were performed). FIG. 10 shows the verification results of Example 1, illustrating the frequency distribution of the actual iron loss values ​​and calculated iron loss values ​​of the electromagnetic steel sheet. Note that the iron loss values ​​are normalized by the average actual iron loss values. As shown in FIG. 10, the average normalized actual iron loss corresponding to the conventional technology that does not consider the occurrence of defects is 100, while the average normalized calculated iron loss corresponding to the present invention is 93. Therefore, by applying the present invention, an improvement of "7" in the iron loss value is expected.

[0078] In Example 2, using the second method of the manufacturing condition determination method according to the embodiment, the percentage of defects along the entire length of an electromagnetic steel sheet that is determined to be within the acceptable range was set to "within 5%," and manufacturing conditions that minimized iron loss were verified (700 verifications). FIG. 11 shows the verification results of Example 2, illustrating the frequency distribution of the actual iron loss values ​​and calculated iron loss values ​​of the electromagnetic steel sheet. Note that the iron loss values ​​were normalized by the average actual iron loss values. As shown in FIG. 11, the average normalized actual iron loss corresponding to the prior art is 100, and the average normalized calculated iron loss corresponding to the present invention is 95. Therefore, by applying the present invention, an improvement of "5" in the iron loss value is expected.

[0079] In the manufacturing condition determination method, manufacturing method for a metallic material, and manufacturing condition determination device according to the above-described embodiments, a defect prediction model and a characteristic prediction model are generated that associate, for each predetermined range, the manufacturing conditions of each process with the presence or absence of defects and the characteristics of the metallic material manufactured under those manufacturing conditions. Then, by determining the manufacturing conditions of the metallic material based on the prediction results of the generated defect prediction model and characteristic prediction model, it is possible to take into account the occurrence of defects in the product and improve the characteristics to the extent that defects do not occur.

[0080] Furthermore, in the manufacturing condition determination method, manufacturing method for a metallic material, and manufacturing condition determination device according to the embodiments, the presence or absence of defects and the characteristics of a product manufactured under a plurality of manufacturing conditions within a specific, predetermined range of the metallic material are predicted. Then, by comparing the characteristics (predicted characteristics values) corresponding to the manufacturing conditions predicted to be defect-free, it is possible to determine manufacturing conditions within a specific, predetermined range that will result in no defects and the best characteristics.

[0081] In addition, in the manufacturing condition determination method, manufacturing method for a metallic material, and manufacturing condition determination device according to the embodiments, the presence or absence of defects and the characteristics of a product manufactured under a plurality of manufacturing conditions within each predetermined range over the entire length of the metallic material are predicted. Then, manufacturing conditions under which the defect rate is within an acceptable range over the entire length of the metallic material are extracted, and statistical values ​​of the predicted characteristic values ​​for each corresponding predetermined range are calculated and compared. This makes it possible to determine manufacturing conditions that result in the defect rate being within an acceptable range and the best characteristics over the entire length of the metallic material.

[0082] In addition, in the manufacturing condition determination method, metal material manufacturing method, and manufacturing condition determination device according to the embodiments, it is possible to determine manufacturing conditions that maximize the number of predetermined ranges in which the predicted characteristic values ​​for each predetermined range without a corresponding defect are within the acceptable range, among the presence or absence of defects and characteristics for which multiple manufacturing conditions for each predetermined range are input, over a predetermined range spanning the entire length of the metal material.

[0083] Here, in the manufacturing condition determination method according to the embodiment, the metal material is divided into predetermined ranges to create models (defect prediction models, characteristic prediction models), and manufacturing conditions are determined for each predetermined range, but it is also possible to determine manufacturing conditions without dividing into predetermined ranges.

[0084] In this case, for example, the defect prediction unit 16 predicts the presence or absence of defects in the metal material by inputting the manufacturing conditions into a defect prediction model generated using the manufacturing conditions of the metal material in each process as input variables and the presence or absence of defects in the metal material as output variables. The characteristic prediction unit 18 predicts the characteristics of the metal material by inputting the manufacturing conditions into a characteristic prediction model generated using the manufacturing conditions of the metal material in each process as input variables and the characteristics of the metal material as output variables. The manufacturing condition determination unit 19 extracts, from the manufacturing conditions input to the defect prediction model and the characteristic prediction model, manufacturing conditions predicted by the defect prediction model as being free of defects. The manufacturing condition determination unit 19 then determines, from the extracted manufacturing conditions, the manufacturing conditions under which the characteristics predicted by the characteristic prediction model are maximized or minimized as the manufacturing conditions for the metal material. Note that the above "maximum or minimum" also includes values ​​quasi-maximum and quasi-minimum, respectively.

[0085] In this way, in the manufacturing condition determination method according to the embodiment, even when the manufacturing conditions are determined without dividing the metal material into predetermined ranges, it is possible to take into account the occurrence of defects in the product and improve characteristics within a range in which defects will not occur.

[0086] The manufacturing condition determination method, manufacturing method for metal material, and manufacturing condition determination device according to the present invention have been specifically described above using the detailed description and examples for carrying out the invention, but the gist of the present invention is not limited to these descriptions and must be broadly interpreted based on the claims. Furthermore, it goes without saying that various changes and modifications based on these descriptions are also included in the gist of the present invention. [Explanation of symbols]

[0087] 1 Manufacturing condition determining device 11 Manufacturing Performance Collection Department 12 Manufacturing Performance Editorial Department 13 Multi-Process Integrated Track Record Editorial Department 14 Performance database 15 Defect prediction model generation unit 16 Defect Prediction Unit 17. Characteristic prediction model generation unit 18 Characteristics prediction section 19 Manufacturing Condition Determination Department

Claims

1. A manufacturing condition determination method for improving the quality of a metal material manufactured through one or more processes, comprising: a defect prediction step of predicting the presence or absence of defects for each predetermined range by inputting the manufacturing conditions for each predetermined range into a defect prediction model generated using manufacturing conditions for each predetermined range of the metal material in each process as input variables and the presence or absence of defects for each predetermined range as an output variable; a characteristic prediction step of predicting the characteristics for each of the predetermined ranges by inputting the manufacturing conditions for each of the predetermined ranges into a characteristic prediction model generated using the manufacturing conditions for each of the predetermined ranges in each of the processes as input variables and the characteristics for each of the predetermined ranges as output variables; a manufacturing condition determination step of extracting manufacturing conditions predicted by the defect prediction model as being free of defects from among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model, and determining, from among the extracted manufacturing conditions, manufacturing conditions under which the characteristic predicted by the characteristic prediction model is maximized or minimized as manufacturing conditions for the metallic material; A manufacturing condition determination method including:

2. In the manufacturing condition determination step, when the same manufacturing conditions are set for the predetermined range over the entire length of the metal material, extracting, from among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model, manufacturing conditions in which the proportion of the predetermined range predicted to have defects over the entire length of the metallic material falls within an allowable range; Among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the statistical value of the properties predicted by the property prediction model are determined as the manufacturing conditions for the metallic material. The manufacturing condition determination method according to claim 1 .

3. In the manufacturing condition determination step, when the same manufacturing conditions are set for the predetermined range over the entire length of the metal material, extracting, from among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model, manufacturing conditions in which the proportion of the predetermined range predicted to have defects over the entire length of the metallic material falls within an allowable range; extracting, from the extracted manufacturing conditions, manufacturing conditions within the predetermined range that are predicted to be free of defects; Among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the statistical value of the properties predicted by the property prediction model are determined as the manufacturing conditions for the metallic material. The manufacturing condition determination method according to claim 1 .

4. In the manufacturing condition determination step, when the same manufacturing conditions are set for the predetermined range over the entire length of the metal material, extracting the manufacturing conditions within the predetermined range that are predicted to be free of defects from among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model; Among the extracted manufacturing conditions, the manufacturing conditions that maximize or minimize the statistical value of the properties predicted by the property prediction model are determined as the manufacturing conditions for the metallic material. The manufacturing condition determination method according to claim 1 .

5. In the manufacturing condition determination step, when the same manufacturing conditions are set for the predetermined range over the entire length of the metal material, determining, as the manufacturing conditions for the metallic material, the manufacturing conditions that are predicted to be free of defects and that maximize the number of predetermined ranges in which the characteristics are within the allowable ranges, among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model; The manufacturing condition determination method according to claim 1 .

6. A method for manufacturing a metallic material, which manufactures a metallic material under manufacturing conditions determined by the manufacturing condition determination method according to any one of claims 1 to 5.

7. A manufacturing condition determination device for improving the quality of a metal material manufactured through one or more processes, a defect prediction unit that predicts the presence or absence of defects for each predetermined range by inputting the manufacturing conditions for each predetermined range into a defect prediction model that is generated using manufacturing conditions for each predetermined range of the metal material in each process as input variables and the presence or absence of defects for each predetermined range as an output variable; a characteristic prediction unit that predicts the characteristics for each of the predetermined ranges by inputting the manufacturing conditions for each of the predetermined ranges into a characteristic prediction model that has been generated using the manufacturing conditions for each of the predetermined ranges in each of the processes as input variables and the characteristics for each of the predetermined ranges as output variables; a manufacturing condition determination unit that extracts manufacturing conditions predicted by the defect prediction model as being free of defects from among the plurality of manufacturing conditions input to the defect prediction model and the characteristic prediction model, and determines, from among the extracted manufacturing conditions, the manufacturing conditions under which the characteristic predicted by the characteristic prediction model is maximized or minimized as the manufacturing conditions for the metallic material; A manufacturing condition determination device comprising:

Images