METHOD FOR GENERATING A QUALITY PREDICTION MODEL FOR METALLIC MATERIALS, METHOD FOR PREDICTION OF QUALITY OF METALLIC MATERIALS, METHOD FOR ESTIMATING FACTORS AFFECTING QUALITY OF METALLIC MATERIALS, METHOD FOR MANUFACTURING METALLIC MATERIALS, APPARATUS FOR GENERATING A QUALITY PREDICTION MODEL FOR METALLIC MATERIALS, APPARATUS FOR PREDICTION OF QUALITY OF METALLIC MATERIALS, AND APPARATUS FOR ESTIMATING FACTORS AFFECTING QUALITY OF METALLIC MATERIALS

Abstract

The method for generating a quality prediction model for a metallic material includes a manufacturing data collection step of collecting manufacturing condition data for each process when manufacturing the metallic material and time data indicating the time when the manufacturing condition data was acquired; a quality data collection step of collecting quality data indicating the quality of the metallic material manufactured through each process; a data editing step of creating integrated process data by associating the manufacturing condition data, time data, and quality data; and a model generation step of generating a quality prediction model that predicts the quality of the metallic material from any manufacturing condition based on the integrated process data.

Description

[Technical Field]

[0001] The present invention relates to a method for generating a quality prediction model for a metallic material, a method for predicting the quality of a metallic material, a method for estimating factors influencing the quality of a metallic material, a manufacturing method for a metallic material, an apparatus for generating a quality prediction model for a metallic material, an apparatus for predicting the quality of a metallic material, and an apparatus for estimating factors influencing the quality of a metallic material. [Background technology]

[0002] Patent Document 1 discloses a method for predicting the quality of a material using a model that has been trained with data that links manufacturing condition data during the manufacturing of the material with quality data in the final process, taking into account the quality judgment position in the longitudinal direction of the material. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7207547 Summary of the Invention [Problem to be solved by the invention]

[0004] It is known that the quality of materials, particularly the mechanical properties of metallic materials such as tensile strength and yield stress, are significantly affected by manufacturing conditions such as the temperature of the metallic material and their changes over time. This is because the state of the metal structure varies depending on manufacturing conditions such as the temperature of the metallic material and is closely related to the mechanical properties of the metallic material. Therefore, in order to predict the quality of metallic materials with high accuracy, it is necessary to fully consider the manufacturing conditions such as the temperature of the metallic material and information on its changes over time.

[0005] However, the method disclosed in Patent Document 1 does not easily capture characteristic patterns of data that may arise due to physical relationships. Therefore, even if a uniform model is generated using all of the obtained manufacturing condition data as is, there is a problem that sufficient quality prediction performance cannot be achieved. Furthermore, it is difficult to reliably estimate factors that have a significant impact on quality based on such a model.

[0006] The present invention has been made in consideration of the above, and aims to provide a method for generating a quality prediction model for metallic materials, a method for predicting the quality of metallic materials, a method for estimating factors influencing the quality of metallic materials, a method for manufacturing metallic materials, an apparatus for generating a quality prediction model for metallic materials, an apparatus for predicting the quality of metallic materials, and an apparatus for estimating factors influencing the quality of metallic materials, which are capable of predicting the quality of metallic materials with high accuracy. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems and achieve the objectives, the method for generating a quality prediction model for a metallic material according to the present invention includes a manufacturing data collection step for collecting manufacturing condition data for each process when manufacturing a metallic material and time data indicating the time when the manufacturing condition data was acquired; a quality data collection step for collecting quality data indicating the quality of the metallic material manufactured through each process; a data editing step for creating integrated process data by associating the manufacturing condition data, the time data, and the quality data; and a model generation step for generating a quality prediction model that predicts the quality of the metallic material from any manufacturing condition based on the integrated process data.

[0008] Furthermore, in the method for generating a quality prediction model for metallic materials according to the present invention, in the above invention, the data editing step creates the integrated process data by associating the manufacturing condition data, the time data, and the quality data for each predetermined range of the metallic material, and creates historical data showing the history of changes in the manufacturing conditions of the metallic material for each predetermined range of the metallic material using multiple pieces of manufacturing condition data and multiple pieces of time data corresponding thereto, and the model generation step classifies the integrated process data into multiple groups according to the characteristics of the changes in the manufacturing conditions of the metallic material in the historical data, and generates the quality prediction model according to the number of groups by performing learning for each of the classified integrated process data.

[0009] In addition, in the method for generating a quality prediction model for metallic materials according to the present invention, in the above invention, the data editing step creates the history data using the temperature of the metallic material as the manufacturing condition data and the time at which the temperature of the metallic material was measured as the time data.

[0010] Furthermore, in the method for generating a quality prediction model for metallic materials according to the present invention, in the above invention, the model generation step uses a data clustering technique including hierarchical clustering and k-means to classify the integrated process data into a plurality of groups according to the characteristics of changes in the manufacturing conditions of the metallic material in the history data.

[0011] Further, in the method for generating a quality prediction model for a metallic material according to the present invention, in the above invention, the quality data is material quality data of the metallic material.

[0012] Further, in the method for generating a quality prediction model for a metallic material according to the present invention, the metallic material is a steel material.

[0013] Furthermore, in the method for generating a quality prediction model for metallic materials according to the present invention, in the above invention, the model generation step generates the quality prediction model using a statistical analysis method and a machine learning method, including linear regression, local regression, principal component regression, PLS regression, logistic regression, support vector machine, decision tree, regression tree, random forest, gradient boosting tree, and neural network.

[0014] In order to solve the above-mentioned problems and achieve the objectives, the quality prediction method for metal materials of the present invention includes a quality prediction step of predicting the quality of metal materials manufactured under any manufacturing conditions using a quality prediction model generated by the above-mentioned method for generating a quality prediction model for metal materials.

[0015] In order to solve the above-mentioned problems and achieve the objectives, the method for estimating factors affecting the quality of metal materials according to the present invention includes a quality influence factor estimation step of estimating factors that have a large influence on the quality of the metal material using a quality prediction model generated by the above-mentioned method for generating a quality prediction model for metal materials.

[0016] In order to solve the above-mentioned problems and achieve the objectives, the method for manufacturing a metallic material according to the present invention includes an importance index calculation step for calculating, for each manufacturing condition, an importance index that indicates the influence that the manufacturing conditions of the metallic material that are input to the quality prediction model have on the prediction of the quality of the metallic material that is output, using a quality prediction model generated by the above-mentioned method for generating a quality prediction model for a metallic material; a manufacturing condition estimation step for estimating manufacturing conditions with high importance indexes as manufacturing conditions that are factors that affect the quality of the metallic material; a quality prediction step for predicting the output quality using the quality prediction model that includes as input manufacturing conditions that are factors that affect the estimated quality of the metallic material; a manufacturing condition determination step for determining manufacturing conditions that are factors that affect the quality of the metallic material so that the predicted quality falls within a predetermined range; and a metallic material manufacturing step for manufacturing a metallic material based on the determined manufacturing conditions.

[0017] In order to solve the above-mentioned problems and achieve the objectives, the quality prediction model generation device for metallic materials of the present invention comprises a manufacturing data collection unit that collects manufacturing condition data for each process when manufacturing a metallic material and time data indicating the time when the manufacturing condition data was acquired; a quality data collection unit that collects quality data indicating the quality of the metallic material manufactured through each process; a data editing unit that creates integrated process data by associating the manufacturing condition data, the time data, and the quality data; and a model generation unit that generates a quality prediction model that predicts the quality of the metallic material from any manufacturing condition based on the integrated process data.

[0018] In order to solve the above-mentioned problems and achieve the objectives, the quality prediction device for metal materials of the present invention includes a quality prediction unit that predicts the quality of metal materials manufactured under any manufacturing conditions using a quality prediction model generated by the above-mentioned metal material quality prediction model generation device.

[0019] In order to solve the above-mentioned problems and achieve the objectives, the metal material quality influence factor estimation device of the present invention includes a quality influence factor estimation unit that uses the quality prediction model generated by the above-mentioned metal material quality prediction model generation device to estimate factors that have a large influence on the quality of the metal material. [Effects of the Invention]

[0020] According to the present invention, by using the time data at which the manufacturing condition data was acquired in addition to the manufacturing condition data, it is possible to generate a quality prediction model that is trained to reflect information on the temporal changes in manufacturing conditions, thereby making it possible to predict the quality of metal materials with high accuracy. [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of an information processing device that functions as a quality prediction model generation device, a quality prediction device, and a quality influence factor estimation device for metallic materials according to an embodiment of the present invention. [Figure 2]FIG. 2 is a flowchart showing the flow of a method for generating a quality prediction model for a metallic material according to an embodiment of the present invention. [Figure 3] Figure 3 is a diagram showing an example of manufacturing condition data and time data for each specified range in the longitudinal and width directions of a metal material collected by a manufacturing data collection unit in a method for generating a quality prediction model for a metal material according to an embodiment of the present invention. [Figure 4] FIG. 4 is a diagram showing an example of quality data for each predetermined range in the longitudinal and width directions of a metal material collected by a quality data collection unit in a method for generating a quality prediction model for a metal material according to an embodiment of the present invention. [Figure 5] FIG. 5 is a diagram showing an example of integrated process data (learning data) for each specified range in the longitudinal and width directions of a metal material, edited by a data editing unit in a method for generating a quality prediction model for a metal material according to an embodiment of the present invention. [Figure 6] FIG. 6 is a diagram showing the relationship between the temperature in a certain range of a metallic material and the time at which the temperature was measured, in the method for generating a quality prediction model for a metallic material according to the embodiment of the present invention. [Figure 7] FIG. 7 is a diagram showing an example of history data for each specified range in the longitudinal and width directions of a metal material, created by the data editing unit in the method for generating a quality prediction model for a metal material according to an embodiment of the present invention. [Figure 8] FIG. 8 is a diagram showing an image of how the model generation unit classifies history data into a plurality of groups in the method for generating a quality prediction model for a metallic material according to the embodiment of the present invention. [Figure 9] FIG. 9 is a flowchart showing the flow of the method for predicting the quality of a metallic material according to an embodiment of the present invention. [Figure 10] FIG. 10 is a flowchart showing the flow of a method for estimating factors affecting quality of a metallic material according to an embodiment of the present invention. [Figure 11] FIG. 11 is a graph showing the results of comparing the prediction accuracy of the quality prediction models of the example of the present invention, Comparative Example 1, and Comparative Example 2. DETAILED DESCRIPTION OF THE INVENTION

[0022] A method for generating a quality prediction model for metallic materials, a method for predicting the quality of metallic materials, a method for estimating factors affecting the quality of metallic materials, a method for manufacturing metallic materials, a device for generating a quality prediction model for metallic materials, a device for predicting the quality of metallic materials, and a device for estimating factors affecting the quality of metallic materials will be described with reference to the drawings.

[0023] (Device configuration) The configurations of a quality prediction model generation device, a quality prediction device, and a quality influence factor estimation device for metallic materials according to this embodiment will be described with reference to FIG. 1. The quality prediction model generation device is a device for generating a quality prediction model for predicting the quality of metallic materials manufactured through one or more processes. The quality prediction device is a device for predicting the quality of metallic materials manufactured through one or more processes, using the quality prediction model generated by the quality prediction model generation device. The quality influence factor estimation device is a device for estimating factors that have a significant impact on the quality of metallic materials, using the quality prediction model generated by the quality prediction model generation device.

[0024] The metallic material in this embodiment is, for example, a product made of steel material, such as a semi-finished product such as a slab, or a product such as a steel plate produced by rolling the slab. In the following, a quality prediction model generation device, a quality prediction device, and a quality influence factor estimation device will be described that are targeted at metallic materials, but the devices can also be applied to materials other than metallic materials, such as materials produced through multiple processes in a continuous production line, and other manufactured products.

[0025] The quality prediction model generation device, the quality prediction device, and the quality influence factor estimation device can be realized by, for example, an information processing device 1 as shown in Fig. 1. Specifically, the information processing device 1 is configured by a personal computer, a workstation, or the like. The information processing device 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).

[0026] The information processing device 1 includes a manufacturing data collection unit 11, a quality data collection unit 12, a data editing unit 13, a data storage unit 14, a database 15, a model generation unit 16, a quality prediction unit 17, and a quality influence factor estimation unit 18.

[0027] The quality prediction model generation device is configured, for example, by elements of the information processing device 1 excluding the quality prediction unit 17 and the quality affecting factor estimation unit 18. The quality prediction device is configured, for example, by elements of the information processing device 1 excluding the model generation unit 16 and the quality affecting factor estimation unit 18. The quality affecting factor estimation device is configured, for example, by elements of the information processing device 1 excluding the model generation unit 16 and the quality prediction unit 17.

[0028] Sensors (not shown) are connected to the manufacturing data collection unit 11. Examples of sensors connected to the manufacturing data collection unit 11 include a radiation temperature sensor, a color temperature sensor, a load sensor, a tension sensor, and a speed sensor. Using these sensors, the manufacturing data collection unit 11 collects manufacturing condition data for each process when manufacturing a metal material and time data indicating the time when the manufacturing condition data was acquired, and outputs the data to the data editing unit 13.

[0029] Examples of "metallic material manufacturing conditions" include the temperature of the metallic material in each process, as well as other factors such as the components of the metallic material, the pressure applied to the metallic material, the tension applied to the metallic material, the thickness of the metallic material before and after rolling, and the speed at which the metallic material is passed through.

[0030] The manufacturing condition data for each process collected by the manufacturing data collection unit 11 includes not only actual values ​​of the manufacturing conditions measured by sensors, but also preset values ​​of the manufacturing conditions. That is, since sensors may not be installed in some processes, in such cases, set values ​​are collected as manufacturing condition data instead of actual values. Furthermore, the time data collected by the manufacturing data collection unit 11 includes not only the actual measurement time but also a time based on the time calculated from the moving distance (conveying distance) and moving speed (conveying speed) of the metal material. That is, the time data may be data related to the time associated with each manufacturing condition data, and the method of acquiring the time data is not particularly limited.

[0031] The manufacturing data collection unit 11 collects manufacturing condition data and time data for the metal material in each process, which is a predetermined range in the longitudinal and width directions of the metal material, for each predetermined range of each process. For example, if the metal material is a slab or steel plate, the "predetermined range of each process" refers to a certain range (position) in the longitudinal and width directions of the metal material in each process. Furthermore, the predetermined range of each process may be different for each process, or may be the same for each process.

[0032] Here, in the configuration shown in FIG. 1, only one manufacturing data collection unit 11 is provided, and it is assumed that this single manufacturing data collection unit 11 collects manufacturing condition data and time data for each process. However, for example, multiple manufacturing data collection units 11 may be provided in accordance with the number of processes, and the manufacturing condition data and time data for each process may be collected by separate manufacturing data collection units 11.

[0033] Sensors and data input terminals (not shown) are connected to the quality data collection unit 12. Examples of sensors connected to the quality data collection unit 12 include material sensors and surface defect sensors. Examples of data input terminals connected to the quality data collection unit 12 include personal computers for data input connected to a display, keyboard, mouse, etc.

[0034] The quality data collection unit 12 collects data on the quality of the metal material manufactured through each process (hereinafter referred to as "quality data") using these sensors or data input terminals, and outputs the data to the data editing unit 13. The "quality of the metal material" mentioned above is, for example, material data on the metal material, such as the mechanical properties of the metal material in the final process (tensile strength, yield stress, elongation, etc.). Other examples of the "quality of the metal material" include the defect contamination rate (the number of defects appearing per unit area) of the metal material in the final process.

[0035] The quality data of metal materials collected by the quality data collection unit 12 includes not only actual quality values ​​measured by sensors, but also quality judgment values ​​entered from a data input terminal, where pass / fail is evaluated based on certain standards based on the actual measurement values ​​or by visual inspection by an inspector, etc.

[0036] The quality data collection unit 12 evaluates and collects quality data of the metal material in the final process of each process for each predetermined range of the final process. For example, if the metal material is a slab or steel plate, the "predetermined range of the final process" refers to a certain range (position) in the longitudinal and width directions of the metal material in the final process.

[0037] The data editing unit 13 edits the performance data input from the manufacturing data collection unit 11 and the quality data collection unit 12. Specifically, the data editing unit 13 associates the manufacturing condition data and time data of each process collected by the manufacturing data collection unit 11 with the quality data of the metal material collected by the quality data collection unit 12 for each predetermined range of the final process to create integrated process data.

[0038] Furthermore, the data editing unit 13 creates historical data showing the history of changes in the manufacturing conditions of the metal material for each predetermined range of the metal material, using the multiple manufacturing condition data in the integrated process data and the multiple corresponding time data. This historical data shows the temporal changes and characteristics of the manufacturing condition data included in the integrated process data, and is used as an index for classifying the integrated process data into multiple groups, as will be described later.

[0039] The data editing unit 13 creates historical data when generating a quality prediction model (learning phase) and when predicting the quality of metal materials using the quality prediction model (inference phase) (see step S4 in Figure 2 and step S11 in Figure 9).

[0040] Furthermore, the data editing unit 13 can create history data (temperature history data) using, for example, the temperature of the metallic material, which has a large effect on the metal structure of the metallic material and a large effect on the mechanical properties of the metallic material, and the time when the temperature of the metallic material was measured, among the manufacturing condition data. Examples of mechanical properties include tensile strength (TS) and yield strength (YS). The temperature history data also includes information that affects the crystal structure and phase transformation of the metallic material, such as the heating temperature, cooling rate, holding time during heating, and temperature change rate. Details of the creation of history data by the data editing unit 13 will be described later (see FIGS. 6 and 7).

[0041] Furthermore, the data editing unit 13 can create history data (temperature history data) using, for example, the temperature of the metallic material, which has a large effect on the mechanical properties of the metallic material, and the time at which the temperature of the metallic material was measured, among the manufacturing condition data. Details of the creation of history data by the data editing unit 13 will be described later (see FIGS. 6 and 7).

[0042] The data storage unit 14 stores the integrated process data created by the data editing unit 13 in the database 15 for each metal material. The data storage unit 14 also stores the history data created by the data editing unit 13 in the database 15 for each metal material and for each predetermined range. The database 15 stores the integrated process data and history data edited for each predetermined range of the metal material for each metal material.

[0043] The model generation unit 16 generates a quality prediction model that predicts the quality of a metal material from any manufacturing condition, based on the integrated process data and history data stored in the database 15. The model generation unit 16 generates the quality prediction model using the manufacturing condition data and quality data included in the integrated process data as learning data.

[0044] Furthermore, the model generation unit 16 may generate a quality prediction model using, as learning data, time-related data based on time data in addition to the manufacturing condition data and quality data included in the integrated process data. For example, consider a case where temperatures are measured at the entry and exit of a certain manufacturing line for a predetermined range (e.g., the tip) of a metal material. In this case, the quality prediction model is generated using, as learning data, data on the temperatures measured at the entry and exit of the manufacturing line, data on the time between each measurement point calculated from the time each temperature was measured, and quality data.

[0045] The model generation unit 16 uses, for example, random forest as a statistical analysis method and machine learning method when generating a quality prediction model. Note that, in addition to the above, various other statistical analysis methods and machine learning methods can be used, such as linear regression, local regression, principal component regression, PLS regression, logistic regression, support vector machine, decision tree, regression tree, gradient boosting tree, and neural network.

[0046] Specifically, the model generation unit 16 classifies the integrated process data into a plurality of groups according to the temporal transitions and characteristics of the manufacturing conditions indicated by the history data created by the data editing unit 13. In this case, the model generation unit 16 can classify the integrated process data into a plurality of groups according to the temporal transitions and characteristics of the manufacturing conditions of the metal material in the history data, using a data clustering method such as hierarchical clustering or k-means. Details of the classification of the integrated process data by the model generation unit 16 will be described later (see FIG. 8).

[0047] Next, the model generation unit 16 generates a quality prediction model according to the number of groups by performing learning for each of the integrated process data classified by the history data. For example, if the integrated process data is classified into p groups, the model generation unit 16 generates a total of p quality prediction models.

[0048] The quality prediction unit 17 uses the quality prediction model generated by the model generation unit 16 to predict the quality of a metal material manufactured under any manufacturing conditions for each predetermined range of the final process. The quality prediction unit 17 first classifies the manufacturing condition data and time data collected from the metal material to be predicted into multiple groups based on the history data created by the data editing unit 13. The method for classifying the manufacturing condition data and time data is the same as the method used by the model generation unit 16 described above.

[0049] Next, the quality prediction unit 17 predicts the quality of the metal material using a quality prediction model corresponding to the group classified as described above, from among the multiple quality prediction models generated by the model generation unit 16.

[0050] The quality influence factor estimation unit 18 uses the quality prediction model generated by the model generation unit 16 to estimate factors that have a significant impact on the quality of the metal material. From the quality prediction model generated by the model generation unit 16, the quality influence factor estimation unit 18 calculates, for each manufacturing condition, the variable importance of the quality prediction model, i.e., an importance index that represents the impact that the manufacturing condition data of the metal material that is input to the quality prediction model has on the prediction of the quality of the metal material. In particular, the manufacturing conditions of the metal material may be arranged in descending order of variable importance value, and any number of the manufacturing conditions from the top may be output as estimation results. Note that the quality influence factor estimation unit 18 calculates an index corresponding to the influence of each factor on the entire quality prediction model as the variable importance.

[0051] Furthermore, the quality influencing factor estimation unit 18 may predict the output quality using a quality prediction model generated by the model generation unit 16, which includes as input factors that affect the quality of the metal material and estimated manufacturing conditions.The quality influencing factor estimation unit 18 may then determine a control range for the factors that affect the quality of the metal material and estimated manufacturing conditions so that the predicted quality falls within a predetermined range.The control range for the manufacturing conditions can be obtained, for example, by a mathematical programming method such as the branch and bound method.

[0052] The quality affecting factor estimation unit 18 may use a method for calculating a prediction importance factor specific to the algorithm that created the quality prediction model, such as importance based on Gini impurity in a decision tree algorithm, or may use a general-purpose method for calculating a prediction importance factor that is independent of the algorithm that created the quality prediction model, such as "Permutation Importance."

[0053] The information processing device 1 may further include an output unit that outputs, as an output signal, the control range of the manufacturing conditions estimated as factors affecting the quality of the metal material, as determined by the quality influence factor estimation unit 18. If the output signal is output to an equipment control device, the output signal may be a control signal for the equipment that sets the manufacturing conditions. If the output signal is output to an operator's operation terminal, the output signal may be operator guidance information that is displayed as guidance on a screen. Then, the manufacturing conditions are adjusted so that the manufacturing conditions estimated as factors affecting the quality of the metal material fall within the determined control range, and the metal material is manufactured.

[0054] (Quality prediction model generation method) The method for generating a quality prediction model for a metallic material according to the embodiment will be described with reference to FIGS.

[0055] First, the manufacturing data collection unit 11 collects manufacturing condition data and time data for each process (step S1). In step S1, the manufacturing data collection unit 11 collects the manufacturing condition data and time data for each process for each predetermined range of the metal material.

[0056] The manufacturing condition data and time data collected in step S1 have a structure, for example, as shown in Fig. 3. In the data shown in Fig. 3, actual values ​​or set values ​​of a plurality of manufacturing conditions and time data corresponding to each manufacturing condition are arranged for each position from the leading end to the trailing end of the metal material in the longitudinal direction and for each position from one end to the other in the width direction in each process.

[0057] The data shown in FIG. 3 includes the following items: (1) The longitudinal position of the metal material in each process 1 ,l 2 ,… (2) The widthwise position of the metal material in each process 1 ,w 2 ,… (3) Multiple manufacturing conditions measured by sensors at the positions (1) and (2) above x1 11 ,x112 ,…,x2 21 ,x2 22 ,… (4) The time when the manufacturing conditions in (3) above were acquired x1 11 ,x1 12 ,…,x2 21 ,x2 22 ,…

[0058] Next, the quality data collection unit 12 collects quality data of the metal material manufactured through each process (step S2). The quality data collected in step S2 has a structure, for example, as shown in Fig. 4. In the data shown in Fig. 4, actual measured values ​​or judged values ​​of quality are listed for each position from the leading end to the trailing end in the longitudinal direction and for each position from one end to the other in the width direction of the metal material manufactured through each process.

[0059] The data shown in FIG. 4 includes the following items: (1) The longitudinal position of the metal material in each process 1 ,l 2 ,… (2) The widthwise position of the metal material in each process 1 ,w 2 ,… (3) Multiple quality y measured at the locations (1) and (2) above 11 ,y 12 ,…

[0060] Next, the data editing unit 13 edits the manufacturing condition data, time data, and quality data to create integrated process data (step S3).

[0061] The data editing unit 13 associates multiple pieces of manufacturing condition data and time data for the metal material in all processes with the corresponding quality data for each predetermined range of the final process of the metal material, and edits them as integrated process data for the metal material in units of a predetermined range, as shown in Figure 5, for example. The integrated process data shown in Figure 5 is data for one metal material, with quality data, manufacturing condition data, and time data arranged in the columns. In the integrated process data shown in Figure 5, each row of data represents a predetermined range of data for the metal material.

[0062] Although the present embodiment shows an example of integrated process data edited in units of a predetermined range of metal materials, it may also be edited in units of metal materials. That is, the data editing unit 13 may associate multiple pieces of manufacturing condition data and time data for metal materials in all processes with the corresponding quality data for each metal material, and compile them as integrated process data for each metal material.

[0063] Next, the data editing unit 13 uses the plurality of manufacturing condition data and time data to create history data for each metal material that indicates the history of changes in the manufacturing conditions of the metal material (step S4).

[0064] In step S4, the data editing unit 13 creates historical data such as that shown in Figure 6 using, for example, the temperature data of the metal material and the time data when the temperature data was measured from the manufacturing condition data and time data of the metal material included in the integrated process data.

[0065] The historical data (temperature history data) shown in Figure 6 is created by interpolating data between measurement times for temperature data measured at different times during the manufacturing process of a metal material. In this embodiment, data between measurement times is interpolated using linear interpolation, but the data interpolation method is not limited to linear interpolation. For example, it is possible to assume a rough transition in temperature between specific measurement times in advance, and perform interpolation such that the temperature changes by a rate of u minutes to v (u>v>0) over a time period up to a rate of s minutes t (s>t>0) between measurement times, and then changes by the remaining rate over the remaining time period.

[0066] The history data created in step S4 has a structure such as that shown in Fig. 7. In the data shown in Fig. 7, for each predetermined range of metal material, the elapsed time from the measurement time of any temperature data used as the starting reference and the measured temperature value are listed.

[0067] The data shown in FIG. 7 includes the following items: (1) The elapsed time t from the starting reference point 1 ,t 2 ,…,t i ,…(i is a positive integer) (2) Temperature T after each elapsed time 1 ,T 2 ,…,T i ,…(i is a positive integer)

[0068] Furthermore, the history data is created for each predetermined range of metal material, i.e., the history data is created for each row in Fig. 5 and is linked to the manufacturing condition data, time data, and quality data of each row.

[0069] Next, the data storage unit 14 stores the integrated process data and history data created by the data editing unit 13 in the database 15 (step S5). Next, the model generation unit 16 classifies the integrated process data into a plurality of groups based on the history data created in step S4 (step S6).

[0070] In step S6, the model generation unit 16 classifies the integrated process data into a plurality of groups (for example, p groups) as shown in Fig. 8 according to the temporal transition and characteristics of the manufacturing conditions indicated in the history data. Note that in step S6, the manufacturing condition data included in the integrated process data is classified into a plurality of groups based on the history data, but since time data is associated with each manufacturing condition data (see Fig. 5), the time data is also classified into a plurality of groups as a result.

[0071] In step S6, the integrated process data is classified using a data clustering method that automatically classifies data into multiple groups based on the temporal progression and characteristics of individual manufacturing conditions. For example, k-means, a type of non-hierarchical clustering, can be used as the data clustering method. In k-means, the number of groups to be classified is defined as p (p is a positive integer), and data with similar characteristics are classified into the same group. Note that, in addition to k-means, various hierarchical clustering and non-hierarchical clustering methods may also be used as the data clustering method. In step S6, the integrated process data may be classified into multiple groups based on any arbitrary condition, such as the temperature at a specific measurement time or the time between specific measurement times.

[0072] The model generation unit 16 generates a quality prediction model for each of the integrated process data classified in step S6 (step S7). In step S7, a total of p quality prediction models are generated to match the number of groups, each of which predicts the quality of the metal material for each predetermined range in the final process from the manufacturing condition data and quality data for each predetermined range in the final process.

[0073] In this embodiment, historical data is created from one type of manufacturing condition (temperature), and the integrated process data is classified into p items using the historical data, and p quality prediction models are generated accordingly, but historical data may also be created from two or more types of manufacturing conditions.

[0074] For example, consider a case where first historical data is created from temperature data of the metallic material and the integrated process data is classified into p items, and second historical data is created from pressure data of the metallic material and the integrated process data is classified into q items. In this case, in step S7, a total of p+q quality prediction models are generated to match the number of groups, based on the manufacturing condition data and quality data for each specified range of the final process.

[0075] (Quality prediction method) The quality prediction method for a metallic material according to the embodiment will be described with reference to FIG.

[0076] First, the data editing unit 13 uses a plurality of pieces of manufacturing condition data and time data collected from the metal material to be predicted to create, for each metal material, history data indicating the history of changes in the manufacturing conditions of the metal material (step S11). The method for creating the history data is the same as that of step S4 in FIG. 2. The manufacturing condition data used when creating the history data in step S11 is the same as the manufacturing condition data used when creating the history data in step S4 in FIG. 2. For example, if the temperature data of the metal material is used to create the history data in step S4 in FIG. 2, the temperature data of the metal material is also used to create the history data in step S11.

[0077] Next, the quality prediction unit 17 classifies the manufacturing condition data and the time data into a plurality of groups based on the history data created in step S11 (step S12). The method of classifying the manufacturing condition data and the time data is the same as step S6 in FIG. 2.

[0078] Next, the quality prediction unit 17 predicts the quality of the metal material manufactured under any manufacturing conditions for each predetermined range of the final process using a quality prediction model corresponding to the group classified in step S12 (step S13). For example, consider a case where the manufacturing condition data and time data are classified into "group 1" in step S12. In this case, in step S13, the quality prediction unit 17 selects a quality prediction model generated from the integrated process data also classified into "group 1" from among the multiple quality prediction models generated by the model generation unit 16. Then, the quality prediction unit 17 predicts the quality of the metal material using the selected quality prediction model.

[0079] (Quality Influence Factor Estimation Method) The method for estimating factors affecting quality of a metallic material according to the embodiment will be described with reference to FIG.

[0080] The quality influencing factor estimation unit 18 estimates factors that have a large influence on the quality of the metal material based on the quality prediction model generated by the above-mentioned quality prediction model generation method (step S21). In step S21, factors that have a large influence on the quality of the metal material are estimated for each group based on p quality prediction models corresponding to the classified groups.

[0081] (Metal material manufacturing method) The method for manufacturing a metallic material according to this embodiment includes an importance index calculation step, a manufacturing condition estimation step, a quality prediction step, a manufacturing condition determination step, and a metallic material manufacturing step.

[0082] In the importance index calculation step, the quality influence factor estimation unit 18 uses the quality prediction model generated by the model generation unit 16 to calculate an importance index (variable importance) that indicates the influence that each manufacturing condition of the metal material that is the input of the quality prediction model has on the prediction of the output quality.

[0083] In the manufacturing condition estimation step, the quality influencing factor estimation unit 18 estimates manufacturing conditions with high importance indices as manufacturing conditions that are factors that affect the quality of the metal material.

[0084] In the quality prediction step, the quality influence factor estimation unit 18 predicts the output quality using the quality prediction model generated by the model generation unit 16, which includes as input manufacturing conditions that are factors that affect the quality of the estimated metal material.

[0085] In the manufacturing condition determination step, the quality influence factor estimation unit 18 determines manufacturing conditions that are factors that affect the quality of the metallic material so that the predicted quality falls within a predetermined range. That is, the quality influence factor estimation unit 18 determines the control range of the manufacturing conditions that are estimated as factors that affect the quality of the metallic material so that the predicted quality falls within the predetermined range. The control range of the manufacturing conditions can be determined, for example, by a mathematical programming method such as the branch and bound method.

[0086] The quality influence factor estimation unit 18 may output the control range of the manufacturing conditions estimated as factors affecting the quality of the metal material as an output signal via the output unit of the information processing device 1. If the output signal is output to a control device of the equipment, the output signal may be a control signal of the equipment that sets the manufacturing conditions. Also, if the output signal is output to an operation terminal of an operator, the output signal may be operator guidance information to be displayed as guidance on the screen.

[0087] In the metal material manufacturing step, the metal material is manufactured based on manufacturing conditions that are adjusted so that the manufacturing conditions estimated to be factors that affect the quality of the metal material are within the determined control range.

[0088] (Example) An example of the quality prediction method for metallic materials according to the embodiment will be described. In this example, the quality prediction method for metallic materials according to the embodiment was applied to predict the tensile strength TS and yield strength YS of a surface-treated steel sheet (hot-dip galvanized steel sheet).

[0089] The manufacturing condition data for metal materials included the metal material's composition, the metal material's temperatures such as the rolling temperature and cooling temperature in the hot rolling process, the metal material's temperatures in each section in the surface treatment process, and the line speed.For the manufacturing condition data for the hot rolling and surface treatment processes, the acquired time data and quality data for the tensile strength TS and yield strength YS in the surface treatment process, which is the final process, were linked for each specified range (1m pitch) of the surface-treated steel plate to create integrated process data.

[0090] In the examples, historical data showing the history of changes in manufacturing conditions in the hot rolling and surface treatment processes was created for each specified range based on the integrated process data. Furthermore, the integrated process data was classified into three patterns—Pattern A, Pattern B, and Pattern C—according to the temporal changes and characteristics of the manufacturing conditions using a data clustering technique. Then, a quality prediction model predicting tensile strength (TS) and yield strength (YS) was generated for each of the classified patterns (Pattern A, Pattern B, and Pattern C) using a random forest. The explanatory variables of the quality prediction model were the composition of the metal material, the temperature of the metal material in the hot rolling process, the temperature of the metal material in the surface treatment process, and the line speed of the surface treatment process. Furthermore, the holdout method was applied to verify the accuracy of the generated quality prediction model, and the most recent 10% of the data was used for evaluation.

[0091] In Comparative Example 1, classification of the integrated process data by the data clustering method performed in the Example was not performed, and one quality prediction model was generated.

[0092] In Comparative Example 2, the temperatures of the metal material in the hot rolling process and the surface treatment process used as explanatory variables in the quality prediction model of Comparative Example 1 were determined by using the average values ​​of each temperature over the entire length as representative values.

[0093] Fig. 11 shows the results of comparing the prediction accuracy of the quality prediction models of Example, Comparative Example 1, and Comparative Example 2. As prediction accuracy, the standard deviation of the prediction error is shown for Example and Comparative Example 1, with Comparative Example 2 as the reference (100%). As shown in Fig. 11, it can be seen that the Example of the present invention has improved tensile strength TS and yield strength YS compared to Comparative Examples 1 and 2.

[0094] According to the above-described embodiments of the method for generating a quality prediction model for metallic materials, the method for predicting the quality of metallic materials, the method for estimating factors influencing the quality of metallic materials, the manufacturing method for metallic materials, the device for generating a quality prediction model for metallic materials, the device for predicting the quality of metallic materials, and the device for estimating factors influencing the quality of metallic materials, by using the time data at which the manufacturing condition data was acquired in addition to the manufacturing condition data, it is possible to generate a quality prediction model that has been trained to reflect information on the changes in manufacturing conditions over time, thereby making it possible to predict the quality of metallic materials with high accuracy.

[0095] Furthermore, in the metallic material quality prediction model generation method, metallic material quality prediction method, metallic material quality influencing factor estimation method, metallic material manufacturing method, metallic material quality prediction model generation device, metallic material quality prediction device, and metallic material quality influencing factor estimation device according to the embodiments, historical data showing the history of changes in the metallic material manufacturing conditions is used to classify integrated process data into multiple groups, and a quality prediction model that predicts the quality of the metallic material from the manufacturing conditions is created based on the integrated process data classified into each group. This makes it possible to predict with high accuracy the quality that can change depending on characteristic patterns in the history of the metallic material manufacturing conditions, and to estimate factors that have a significant impact on the quality of the metallic material.

[0096] Furthermore, in the metallic material quality prediction model generation method, metallic material quality prediction method, metallic material quality influence factor estimation method, metallic material manufacturing method, metallic material quality prediction model generation device, metallic material quality prediction device, and metallic material quality influence factor estimation device according to the embodiments, history data is created using the metallic material temperature as manufacturing condition data and the time at which the metallic material temperature was measured as time data. This makes it possible to accurately predict the quality, particularly mechanical properties, which may change depending on the characteristic pattern of the metallic material's temperature history, which has a significant impact on the metallic structure of the metallic material, among the manufacturing condition data. Then, by determining the control range of the metallic material's temperature conditions, it is possible to manufacture a metallic material whose mechanical properties are controlled within a predetermined range.

[0097] The method for generating a quality prediction model for metallic materials, the method for predicting the quality of metallic materials, the method for estimating factors influencing the quality of metallic materials, the method for manufacturing metallic materials, the device for generating a quality prediction model for metallic materials, the device for predicting the quality of metallic materials, and the device for estimating factors influencing the quality of metallic materials according to the present invention have been specifically described above using a description of the preferred embodiment and examples, but the scope of the present invention is not limited to these descriptions and should be broadly interpreted based on the claims. Furthermore, it goes without saying that various changes and modifications based on these descriptions are also included in the scope of the present invention. [Explanation of symbols]

[0098] 1. Information processing equipment 11 Manufacturing Data Collection Department 12 Quality Data Collection Department 13 Data Editorial Department 14 Data storage section 15 Databases 16 Model generation unit 17 Quality Prediction Department 18 Quality Influence Factor Estimation Department

Claims

1. a manufacturing data collection step of collecting manufacturing condition data for each process in manufacturing a metal material and time data indicating the time when the manufacturing condition data was acquired; a quality data collection step of collecting quality data indicating the quality of the metal material manufactured through each process; a data editing step of creating integrated process data by associating the manufacturing condition data, the time data, and the quality data; a model generation step of generating a quality prediction model for predicting quality from any manufacturing condition of the metallic material based on the integrated process data; Including, The data editing step includes: creating the integrated process data by associating the manufacturing condition data, the time data, and the quality data for each predetermined range of the metal material; creating history data showing a history of changes in the manufacturing conditions of the metallic material for each predetermined range of the metallic material using a plurality of the manufacturing condition data and a plurality of the time data corresponding thereto; The model generation step includes: classifying the integrated process data into a plurality of groups according to characteristics of changes in the manufacturing conditions of the metal material in the history data; generating the quality prediction model according to the number of the plurality of groups by performing learning for each of the classified integrated process data; A method for generating quality prediction models for metallic materials.

2. 2. The method for generating a quality prediction model for a metallic material according to claim 1, wherein the data editing step creates the history data using the temperature of the metallic material as the manufacturing condition data and the time at which the temperature of the metallic material was measured as the time data.

3. 3. The method for generating a quality prediction model for metallic materials according to claim 2, wherein the model generation step uses a data clustering method including hierarchical clustering and k-means to classify the integrated process data into a plurality of groups according to characteristics of changes in the manufacturing conditions of the metallic material in the history data.

4. The method for generating a quality prediction model for a metallic material according to claim 1 , wherein the quality data is material quality data for the metallic material.

5. The method for generating a quality prediction model for a metallic material according to any one of claims 1 to 3, wherein the metallic material is a steel material.

6. 4. The method for generating a quality prediction model for a metallic material according to claim 1, wherein the model generation step generates the quality prediction model using a statistical analysis method and a machine learning method, including linear regression, local regression, principal component regression, PLS regression, logistic regression, support vector machine, decision tree, regression tree, random forest, gradient boosting tree, and neural network.

7. A quality prediction method for a metallic material, comprising a quality prediction step of predicting the quality of a metallic material manufactured under any manufacturing conditions using a quality prediction model generated by the quality prediction model generation method for a metallic material according to any one of claims 1 to 3.

8. A method for estimating factors affecting the quality of a metallic material, comprising a quality influence factor estimation step of estimating factors that have a large influence on the quality of the metallic material using a quality prediction model generated by the method for generating a quality prediction model for a metallic material described in any one of claims 1 to 3.

9. an importance index calculation step of calculating, for each manufacturing condition, an importance index that indicates the influence that a manufacturing condition of the metallic material that is an input of the quality prediction model has on the prediction of the quality of the metallic material that is an output, using the quality prediction model generated by the method for generating a quality prediction model of a metallic material according to any one of claims 1 to 3; a manufacturing condition estimation step of estimating the manufacturing conditions with high importance indices as manufacturing conditions that will affect the quality of the metal material; a quality prediction step of predicting output quality using the quality prediction model, which includes, as input, manufacturing conditions that are factors that affect the estimated quality of the metal material; a manufacturing condition determination step of determining manufacturing conditions that are factors that affect the quality of the metallic material so that the predicted quality falls within a predetermined range; a metal material manufacturing step of manufacturing a metal material based on the determined manufacturing conditions; A method for producing a metal material comprising:

10. a manufacturing data collection unit that collects manufacturing condition data for each process when manufacturing a metal material and time data indicating the time when the manufacturing condition data was acquired; a quality data collection unit that collects quality data indicating the quality of the metal material manufactured through each process; a data editing unit that associates the manufacturing condition data, the time data, and the quality data to create integrated process data; a model generation unit that generates a quality prediction model that predicts quality from any manufacturing condition of the metallic material based on the integrated process data; Equipped with The data editing unit creating the integrated process data by associating the manufacturing condition data, the time data, and the quality data for each predetermined range of the metal material; creating history data showing a history of changes in the manufacturing conditions of the metallic material for each predetermined range of the metallic material using a plurality of the manufacturing condition data and a plurality of the time data corresponding thereto; The model generation unit classifying the integrated process data into a plurality of groups according to characteristics of changes in the manufacturing conditions of the metal material in the history data; generating the quality prediction model according to the number of the plurality of groups by performing learning for each of the classified integrated process data; A quality prediction model generator for metal materials.

11. A quality prediction device for metallic materials comprising a quality prediction unit that predicts the quality of metallic materials manufactured under any manufacturing conditions using a quality prediction model generated by the metallic material quality prediction model generation device described in claim 10.

12. A quality influence factor estimation device for metallic materials, comprising a quality influence factor estimation unit that estimates factors that have a large influence on the quality of the metallic material using a quality prediction model generated by the quality prediction model generation device for metallic materials described in claim 10.

Images