Operation condition determination device, operation condition determination method and program

The operating condition determining device addresses the challenge of predicting steel manufacturing parameters by using probability distributions and back-analysis to ensure material properties meet specifications, improving efficiency and accuracy in quality design work.

JP2024159169APending Publication Date: 2024-11-08NIPPON STEEL CORPORATION
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023074998
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-28
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing methods for determining operating conditions in steel manufacturing processes struggle with accurately predicting material properties due to multidimensional parameter variations and lack of consideration for uncertainty, leading to potential failure in meeting customer specifications and high workload in quality design work.

Method used

An operating condition determining device and method that uses a material property value prediction model to output probability distributions, allowing for back-analysis to ensure material property values meet or exceed specified thresholds, considering constraints and uncertainties in steel manufacturing processes.

Benefits of technology

Improves the efficiency of quality design work by accurately determining appropriate operating conditions, even under uncertain conditions, thereby enhancing the likelihood of meeting customer specifications and reducing the risk of material property value deviations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024159169000001_ABST
    Figure 2024159169000001_ABST
Patent Text Reader

Abstract

To provide an operation condition determination device and the like capable of determining an appropriate operation condition in a manufacturing process of a steel material according to order information.SOLUTION: An operation condition determination device 100 according to the present invention is the device which determines an operation condition in a manufacturing process of a steel material according to order information and includes a material property value prediction model which takes the operation condition in the manufacturing process of the steel material as inputs and takes the probability distribution of the material property values of the steel material as outputs. The operation condition in the manufacturing process is determined by performing inverse analysis of the material property value prediction model such that the acceptance probability of the material property value of the steel material calculated based on the probability distribution of the material property values of the steel material output when inputting the operation condition in the manufacturing process of the steel material to the material property value prediction model and the specifications of the material property value of the steel material included in the order information becomes equal to or greater than a prescribed threshold.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an operation condition determination device, an operation condition determination method, and a program for causing a computer to function as the operation condition determination device, which are capable of determining appropriate operation conditions in a steel manufacturing process in accordance with order information. [Background technology]

[0002] Steel products such as steel plates are often made to order, with different specifications for each customer. When making to order products, the task of predicting the quality of the steel product after production, such as its material properties (such as tensile strength), shape, and defects, and determining the appropriate operating conditions in the steel product manufacturing process so that the predicted quality meets the order specifications (quality design task) is an important task that affects manufacturing costs and quality.

[0003] Among quality design tasks, especially those related to material property values, proper operating conditions for new orders are usually searched for and determined by referring to past operational records and the material property value records obtained from the past operational records. However, since the parameters that make up the material property values ​​and operating conditions are multidimensional, it is difficult for quality designers who perform quality design tasks to visually check the results of all of these parameters at the same time and determine the operating conditions. Therefore, quality designers have traditionally determined the operating conditions by using a scatter diagram, as described below. Figure 1 is a schematic diagram of a scatter diagram used in conventional quality design work. In Figure 1, the plots of "◯", "△" and "□" represent past performance data. Quality designers draw multiple two-dimensional scatter diagrams, with operating conditions and material property values ​​on two orthogonal axes, as shown in Figure 1, according to the number of parameters, to estimate the relationship between the operating conditions and the material property values ​​and search for operating conditions that satisfy the customer's order specifications. In the example shown in Figure 1, for the higher grade steel material plotted with "◯" and the lower grade steel material plotted with "△", the past performance shows a tendency for the material property value (tensile strength) on the vertical axis to decrease as the operating condition (cooling temperature in the hot rolling process) on the horizontal axis increases. If the same tendency is considered for the grade (grade to be determined) corresponding to the customer's order specifications plotted with "□", it is considered that the operating conditions can be determined to satisfy the order specifications by determining the operating conditions to be within the predicted range surrounded by the dashed line in Figure 1.

[0004] However, if a scatter diagram is drawn using all past performance data, various operating conditions other than the two axes and parameter values ​​of material properties that differ are plotted on the same scatter diagram, which often results in a scatter diagram with large variations, making it difficult to properly estimate the relationship between operating conditions and material properties. Therefore, in practice, quality engineers specify the steel type and specifications they wish to refer to based on their experience and knowledge, extract performance data to be used in creating the scatter diagram, and then create the scatter diagram using this extracted performance data. In this way, even if the same performance data is used, the scatter diagrams created are different depending on the experience and knowledge of the quality designer, and the search range for the operating conditions is limited, so there are cases where appropriate operating conditions cannot be determined. Another problem is that the workload of the quality design work is very high.

[0005] In order to solve the above problems, for example, a steel product order receiving support system as described in Patent Document 1 has been proposed. The system described in Patent Document 1 is characterized by comprising a manufacturing information storage means for storing, as examples, the actual values ​​of material components and operating conditions, and the material actual values ​​for each product manufactured in the past; a material estimation means for estimating, using the data stored in the manufacturing information storage means, the material that will be obtained when a product is manufactured based on the input material components and specified values ​​of operating conditions; and a product quality design means for outputting, from the material estimated value output from the material estimation means and its estimation error, specified values ​​of material components and operating conditions that will enable the manufacture of a product that satisfies the required specifications (see claim 1 in Patent Document 1). However, in the system described in Patent Document 1, the material estimation value and its estimation error output from the material estimation means are respectively the average value and standard deviation of performance data near the input value or performance data related to the material, and the uncertainty of prediction due to the lack of past performance data cannot be taken into consideration. Therefore, even if steel is manufactured using the specified values ​​of the material components and operating conditions determined by the system described in Patent Document 1, there is a high risk that the material property values ​​of the manufactured steel (in Patent Document 1, material properties) will not satisfy the order specifications (the material property values ​​of the steel will fail).

[0006] Non-Patent Documents 1 and 2 describe heteroscedastic Gaussian processes (Gaussian process regression taking heteroscedasticity into account). [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent Publication No. 2022-14878 [Non-patent literature]

[0008] [Non-Patent Document 1] Alan D. Saul, James Hensman, Aki Vehtari, Neil D. Lawrence, "Chained Gaussian Processes", Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, PMLR 51:1431-1440, 2016 [Non-Patent Document 2] "Heteroskedastic Likelihood and Multi-Latent GP", [online], [Retrieved January 16, 2023], Internet<URL:https: / / gpflow.github.io / GPflow / develop / notebooks / advanced / heteroskedastic.html> Summary of the Invention [Problem to be solved by the invention]

[0009] The present invention has been made to solve the problems of the conventional technology, and has an object to provide an operating condition determination device, an operating condition determination method, and a program that can determine appropriate operating conditions in a steel manufacturing process in accordance with order information. [Means for solving the problem]

[0010] In order to solve the above-mentioned problems, the present invention provides an operating condition determination device that determines operating conditions in a steel manufacturing process in accordance with order information, the operating condition determination device having a material property value prediction model that receives the operating conditions in the steel manufacturing process as an input and outputs a probability distribution of the material property values ​​of the steel, and determines the operating conditions in the manufacturing process by inversely analyzing the material property value prediction model so that the probability of acceptance of the material property values ​​of the steel, calculated from the probability distribution of the material property values ​​of the steel output when the operating conditions in the steel manufacturing process are input into the material property value prediction model and the specifications of the material property values ​​of the steel included in the order information, is equal to or greater than a predetermined threshold value.

[0011] In the present invention, the "manufacturing process" refers to a process up to the process of manufacturing a steel material into a product by changing the thickness, surface condition, and internal metal structure of the steel material (such as a hot rolling process, cold rolling process, continuous annealing process, and continuous hot-dip galvanizing process), and does not necessarily include a finishing process in which the steel material is cut and inspected. In the present invention, examples of the "manufacturing process" for which the operating conditions are determined include a steelmaking process and a rolling process (hot rolling process and cold rolling process). In the present invention, the "operational conditions" refer to conditions that affect the quality of the steel material, such as its material property values, during the manufacturing process. Examples of such conditions include the component values ​​of carbon (C), manganese (Mn), silicon (Si), etc. during the steelmaking process, and the heating temperature, rolling temperature, and cooling temperature during the rolling process. In addition, in the present invention, "material characteristic value" means the mechanical properties of a steel material, and examples thereof include tensile strength (TS), yield point (YP), elongation (EL), surface hardness, bending strength, toughness value, and fatigue life. Furthermore, in the present invention, the "specification of the material property value of the steel material included in the order information" may include, for example, the required upper limit value or the required lower limit value of the material property value, or both.

[0012] The operating condition determination device according to the present invention is equipped with a material property value prediction model that inputs operating conditions in a steel manufacturing process and outputs a probability distribution of material property values ​​of the steel. The material property value prediction model outputs the probability distribution of material property values ​​of the steel, in other words, the expected value and variation of the material property values, as prediction results, so that it is expected that the material property values ​​(especially the variation) can be predicted with high accuracy even for operating conditions of the input steel that have no past operating history. In addition, it is expected that the prediction accuracy of the material property values ​​can be improved even for operating conditions of steel where it is difficult to control the material property values ​​and variation in the material property values ​​is likely to occur. In the operation condition determination device according to the present invention, the pass probability of the material property value of the steel is calculated based on the probability distribution of the material property value of the steel output when the operation conditions in the steel manufacturing process are input to the material property value prediction model, and the specification of the material property value of the steel included in the order information. For example, when the specification of the material property value of the steel included in the order information is the required upper limit value and the required lower limit value of the material property value, the pass probability is calculated as the ratio of the area within the range from the required lower limit value to the entire area of ​​the probability distribution of the material property value. Therefore, by reverse-analyzing this material property value prediction model so that the probability of passing the material property value of the steel material is above a predetermined threshold, it is possible to determine appropriate operating conditions in the manufacturing process according to the order information, thereby making it possible to improve the efficiency of quality design work.

[0013] The material property value prediction model of the present invention is constructed by using known data in which, for example, the operational results of a manufacturing process of steel material manufactured in the past are input, and the material property values ​​measured on a test piece obtained from the steel material manufactured in the past are output. For example, the model may be constructed by machine learning using such known data as teacher data. In the present invention, the "operational results" are actual values ​​of operating conditions, and are data that affect the quality of the material property values ​​of the steel material. The operational results may include measured values ​​and set values ​​obtained in the manufacturing process of the steel material, as well as data obtained by processing these values ​​using physical formulas, etc. In addition, the material property value prediction model may be constructed by the operating condition determination device of the present invention, or a material property value prediction model constructed by another device may be stored in the material property prediction device of the present invention.

[0014] It is preferable that the operating condition determination device of the present invention determines operating conditions in the manufacturing process by performing inverse analysis of the material characteristic value prediction model under at least the constraints of the operating conditions resulting from the use of the steel material. The use of steel is a concept that includes the workability and weldability of steel, and constraints on operating conditions arising from the use of steel include, for example, the ranges of component values ​​such as carbon (C) and manganese (Mn) in the steelmaking process. In addition, it is preferable that the operating condition determination device of the present invention determines operating conditions in the manufacturing process by inversely analyzing the material characteristic value prediction model under the constraints of the operating conditions that take into account at least one of the productivity of the steel material and qualities other than the material characteristic values. An example of a constraint on the operating conditions taking into account the productivity of the steel material is the range of the rolling temperature in the rolling process. As an example of a constraint on the operating conditions that takes into consideration qualities (such as defects and shape) other than the material property values ​​of the steel material, the range of the heating temperature and cooling temperature in the rolling process can be mentioned.

[0015] In the operation condition determination device according to the present invention, the material characteristic value prediction model is, for example, a model using Bayesian estimation. Also, the material characteristic value prediction model is, for example, a model using a heteroscedastic Gaussian process.

[0016] In quality design work, when predicting the material property values ​​of steel, there is a high need to predict the material property values ​​by giving the parameters that make up the operating conditions as a range rather than a single value, or to determine the parameters that make up the appropriate operating conditions as a range rather than a single value so that the predicted material property values ​​meet the order specifications. In order to meet the above-mentioned needs, it is preferable that the operating condition determination device of the present invention further includes an operating condition prediction model that inputs the range of operating conditions in the steel manufacturing process and outputs a probability distribution of the operating conditions in the steel manufacturing process, and determines the range of operating conditions in the manufacturing process by inversely analyzing the operating condition prediction model and the material characteristic value prediction model so that the probability distribution of the operating conditions in the steel manufacturing process output when the range of operating conditions in the steel manufacturing process is input into the operating condition prediction model, the probability distribution of the material characteristic values ​​of the steel output when the range of operating conditions in the steel manufacturing process is input into the material characteristic value prediction model, and the acceptance probability of the material characteristic values ​​of the steel calculated from the specifications of the material characteristic values ​​of the steel included in the order information is equal to or greater than a predetermined threshold value.

[0017] In the above preferred configuration, "inversely analyzing the operating condition prediction model and the material characteristic value prediction model" means inversely analyzing a composite function of a function represented by the operating condition prediction model and a function represented by the material characteristic value prediction model. The above-mentioned preferred configuration includes, in addition to the material property value prediction model, an operating condition prediction model that inputs the range of operating conditions in the steel manufacturing process and outputs the probability distribution of the operating conditions in the steel manufacturing process. Therefore, if the range of operating conditions in the steel manufacturing process is input to the operating condition prediction model, the probability distribution of the operating conditions in the steel manufacturing process is output, and if this probability distribution of the operating conditions in the steel manufacturing process is input to the material property value prediction model, the probability distribution of the material property value of the steel according to the range of operating conditions in the steel manufacturing process is output. Therefore, by reverse-analyzing the operating condition prediction model and the material property value prediction model so that the probability of passing the material property value of the steel, calculated from the probability distribution of the material property value of the steel and the specifications of the material property value of the steel included in the order information, is above a predetermined threshold value, it is possible to determine the range of appropriate operating conditions in the manufacturing process according to the order information, and it is possible to improve the efficiency of quality design work in accordance with the above-mentioned needs.

[0018] The operating condition determination device according to the present invention is particularly effective when the order information for which the operating conditions are to be determined is new order information that has no past order record. The operating condition determination device according to the present invention preferably determines the initial operating conditions in the manufacturing process of the steel, which are predetermined at least by the specifications of the material property values ​​of the steel included in the order information, as the operating conditions in the manufacturing process of the steel when the pass probability of the material property values ​​of the steel calculated by the probability distribution of the material property values ​​of the steel output when the initial operating conditions in the manufacturing process of the steel, which are calculated by the specifications of the material property values ​​of the steel included in the order information, is equal to or greater than a predetermined threshold value, and determines the operating conditions in the manufacturing process by performing a reverse analysis of the material property value prediction model so that the pass probability of the material property values ​​of the steel becomes equal to or greater than a predetermined threshold value when the pass probability of the material property values ​​of the steel when the initial operating conditions are used is less than a predetermined threshold value.

[0019] According to the above-mentioned preferred configuration, when the order information for which the operating conditions are to be determined is new order information with no past order record, if the probability distribution of the material property value of the steel material output when the initial operating conditions in the steel manufacturing process determined in advance are input to the material property value prediction model and the specification of the material property value of the steel material included in the order information is equal to or greater than a predetermined threshold value, the initial operating conditions are directly determined as the operating conditions in the steel manufacturing process, so that there is no need to reverse-analyze the material property value prediction model, and the calculation load is reduced. On the other hand, if the probability of passing the material property value of the steel material when the initial operating conditions are used is less than the predetermined threshold value, the operating conditions in the manufacturing process may be determined by reverse-analyzing the material property value prediction model so that the probability of passing the material property value of the steel material is equal to or greater than the predetermined threshold value, as described above.

[0020] The above-described preferable configuration (configuration in which initial operating conditions are used for new order information with no past order record) can also be applied to a case in which the parameters constituting the appropriate operating conditions are determined as ranges rather than as single values. That is, in the present invention, the order information is new order information that has not been ordered in the past, and the probability distribution of the operating conditions in the manufacturing process of the steel, which is determined in advance at least by the specifications of the material property values ​​of the steel included in the order information, is output when the range of initial operating conditions in the manufacturing process of the steel is input into the operating condition prediction model. If the pass probability of the material property values ​​of the steel, calculated by the probability distribution of the material property values ​​of the steel output when input into the material property value prediction model and the specifications of the material property values ​​of the steel included in the order information, is equal to or greater than a predetermined threshold value, the range of the initial operating conditions is determined as the operating conditions in the manufacturing process of the steel, and if the pass probability of the material property values ​​of the steel when the range of the initial operating conditions is used is less than a predetermined threshold value, it is preferable to determine the range of operating conditions in the manufacturing process by reverse analysis of the operating condition prediction model and the material property value prediction model so that the pass probability of the material property values ​​of the steel is equal to or greater than a predetermined threshold value.

[0021] In the above-described preferred configuration, for example, the initial operating conditions may be predetermined based on at least the material property values ​​and thickness specifications of the steel material included in the order information. In the above-described preferred configuration, for example, the range of the initial operating conditions may be predetermined based on at least the material property values ​​and thickness specifications of the steel material included in the order information.

[0022] In the above-mentioned preferred configuration, when the manufacturing process includes a steelmaking process and a rolling process which is a process subsequent to the steelmaking process, for example, when the pass probability of the material characteristic value of the steel when the initial operating conditions are used is less than a predetermined threshold value, it is possible to fix the steel type of the steel and then determine the operating conditions for the processes in the manufacturing process after the rolling process by reverse analyzing the material characteristic value prediction model so that the pass probability of the material characteristic value of the steel is equal to or greater than a predetermined threshold value. Furthermore, in the above-mentioned preferred configuration, the manufacturing process includes a steelmaking process and a rolling process which is a process subsequent to the steelmaking process, and if the pass probability of the material characteristic value of the steel when the range of initial operating conditions is used is less than a predetermined threshold value, it is possible to determine the operating condition prediction model and the operating conditions for processes subsequent to the rolling process among the manufacturing process by inversely analyzing the material characteristic value prediction model so that the pass probability of the material characteristic value of the steel is equal to or greater than a predetermined threshold value, using the steel type of the steel as a constraint condition.

[0023] In order to solve the above-mentioned problems, the present invention is also provided as an operating condition determination method for determining operating conditions in a steel manufacturing process in accordance with order information, which uses a material property value prediction model that takes the operating conditions in the steel manufacturing process as input and outputs a probability distribution of the material property values ​​of the steel, and determines the operating conditions in the manufacturing process by inversely analyzing the material property value prediction model so that the pass probability of the material property values ​​of the steel, calculated from the probability distribution of the material property values ​​of the steel output when the operating conditions in the steel manufacturing process are input into the material property value prediction model and the specifications of the material property values ​​of the steel included in the order information, is equal to or greater than a predetermined threshold value.

[0024] Furthermore, in order to achieve the above object, the present invention also provides a program for causing a computer to function as the operation condition determination device. Effect of the Invention

[0025] According to the present invention, it is possible to determine appropriate operating conditions in the manufacturing process of steel materials according to order information, thereby making it possible to improve the efficiency of quality design work. [Brief description of the drawings]

[0026] [Figure 1] FIG. 1 is a schematic diagram showing a scatter diagram used in a conventional quality design operation. [Diagram 2]1 is a diagram for explaining an outline of an operation condition determination device according to an embodiment of the present invention; [Diagram 3] 1 is a block diagram showing a schematic configuration of an operation condition determination device according to a first embodiment. [Figure 4] 4 is a flowchart showing an outline of a procedure of an operation condition determining method executed by using the operation condition determining device shown in FIG. [Diagram 5] FIG. 2 is a diagram for explaining a construction example of a material characteristic value prediction model according to the first embodiment in comparison with a construction example of a conventional machine learning model. [Figure 6] FIG. 4 is a diagram showing an example of the contents output from an operation condition output unit 6 shown in FIG. [Figure 7] FIG. 6 is a block diagram showing a schematic configuration of an operation condition determination device according to a second embodiment. [Figure 8] 8 is a flowchart showing an outline of a procedure of an operation condition determining method executed by using the operation condition determining device shown in FIG. [Figure 9] This is an example of the results of comparing the operational performance of silicon (Si) component values ​​for a reference steel material in the steelmaking process, whose silicon component value indication range is 20 to 25, with the probability distribution of silicon component values ​​predicted by an operational condition prediction model. [Figure 10] FIG. 8 is a diagram showing an example of the contents output from an operation condition output unit 6 shown in FIG. [Figure 11] FIG. 11 is a block diagram showing a schematic configuration of an operation condition determination device according to a third embodiment. [Figure 12] 12 is a flowchart showing an outline of a procedure of an operation condition determining method executed by using the operation condition determining device shown in FIG. [Figure 13] 12 is a diagram showing an example of order information acquired by the order information acquiring unit 1 shown in FIG. [Figure 14] 12 is a diagram showing an example of a database used by the manufacturing system acquisition unit 10 shown in FIG. 11. FIG. [Figure 15] FIG. 13 is a diagram showing an example of initial operating conditions to be input to a material characteristic value model, and a probability distribution and pass probability of material characteristic values ​​obtained under the initial operating conditions. [Figure 16]FIG. 12 is a diagram showing an example of operation conditions searched for in cases 1 to 3 executed by the operation condition determination unit 5 shown in FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0027] [Outline of the operation condition determination device] Hereinafter, an operation condition determination device according to an embodiment (first to third embodiments) of the present invention will be described with reference to the accompanying drawings as appropriate, taking a case where the steel material is a steel plate as an example. First, an overview of the operation condition determination device according to the present embodiment will be described.

[0028] FIG. 2 is a diagram for explaining an outline of the operation condition determination device according to the present embodiment. The operation condition determination device according to the present embodiment is a device that determines operation conditions in a steel manufacturing process according to order information. Specifically, as shown in FIG. 2(a), the operation condition determination device according to the present embodiment is provided with a material property value prediction model that receives operation conditions in a steel manufacturing process as input and outputs a probability distribution of the material property value of the steel. Therefore, based on the probability distribution of the material property value of the steel output from the material property value prediction model and the specifications of the material property value of the steel included in the order information (for example, the required upper limit value and the required lower limit value of the material property value), the ratio of the area within the range from the required lower limit value to the required upper limit value (the area of ​​the hatched region shown in FIG. 2(a)) to the area of ​​the entire probability distribution of the material property value is calculated as the pass probability. As shown in Figure 2(b), the operating condition determination device of this embodiment determines the operating conditions in the manufacturing process by inversely analyzing the material property value prediction model so that the pass probability of the material property value of the steel, calculated from the probability distribution of the material property value of the steel output when the operating conditions in the manufacturing process of the steel are input into the material property value prediction model and the specifications of the material property value of the steel included in the order information (e.g., the required upper limit value and the required lower limit value of the material property value), (the area of ​​the hatched area shown in Figure 2(b) relative to the area of ​​the entire probability distribution) is equal to or greater than a predetermined threshold value.

[0029] In this embodiment, the manufacturing process for which the operating conditions are determined will be described as a steelmaking process and a rolling process (hot rolling process and cold rolling process). In this specification, the steel material for which the operating conditions are determined in the manufacturing process will be appropriately referred to as the "target steel material" in order to distinguish it from a reference steel material described later. Hereinafter, first to third embodiments, which differ from each other in the procedures for determining the operating conditions in the manufacturing process, will be described in order.

[0030] [First embodiment] Fig. 3 is a block diagram showing a schematic configuration of an operation condition determination device according to the first embodiment. Fig. 4 is a flow chart showing a schematic procedure of an operation condition determination method executed by using the operation condition determination device shown in Fig. 3. The operating condition determination device 100 according to the first embodiment is configured such that the parameters constituting the operating conditions in the manufacturing process of the steel material input to the material property value prediction model are each one value, and a probability distribution of the material property value of the steel material is output. Therefore, in the operating condition determination device 100 according to the first embodiment, the parameters constituting the operating conditions according to the order information are each determined as one value. As shown in FIG. 3, the operating condition determination device 100 according to the first embodiment includes an order information acquisition unit 1, a material property value prediction model storage unit 2, a material property value prediction model acquisition unit 3, a constraint condition setting unit 4, an operating condition determination unit 5, and an operating condition output unit 6.

[0031] The operating condition determination device 100 is composed of a computer having one or more hardware processors such as a CPU (Central Processing Unit), one or more memories such as a RAM (Random Access Memory) and a ROM (Read Only Memory), and one or more programs stored in the memory are executed by the one or more hardware processors to perform various calculations. As a result, the operating condition determination device 100 functions as an order information acquisition unit 1, a material characteristic value prediction model storage unit 2, a material characteristic value prediction model acquisition unit 3, a constraint condition setting unit 4, an operating condition determination unit 5, and an operating condition output unit 6. The operating condition determination device 100 may be a PLC (Programmable Logic Controller) or may be realized by dedicated hardware such as an ASIC (Application Specific Integrated Circuit). Each unit 1 to 6 included in the operating condition determination device 100 will be described in order below.

[0032] <Order Information Acquisition Unit 1> The order information acquisition unit 1 executes step ST11 shown in Fig. 4. Specifically, the order information acquisition unit 1 acquires, as order information, product sizes such as the plate thickness of the target steel material and specifications of material property values ​​such as the tensile strength (TS), yield point (YP), and elongation (EL) of the target steel material, as shown in Table 1 below, from a predetermined database (not shown). The order information acquisition unit 1 of the first embodiment acquires, as the specifications of the material property values ​​of the target steel material, the required upper limit value, the required lower limit value, or both of the material property values. [Table 1]

[0033] <Material characteristic value prediction model storage section 2> The material characteristic value prediction model storage unit 2 stores (memorizes) a material characteristic value prediction model constructed by a material characteristic value prediction model construction unit (not shown) included in the operation condition determination device 100 or by another device.

[0034] <Material characteristic value prediction model acquisition unit 3> The material characteristic value prediction model acquisition unit 3 executes step ST12 shown in Fig. 4. Specifically, the material characteristic value prediction model acquisition unit 3 acquires the material characteristic value prediction model stored in the material characteristic value prediction model storage unit 2.

[0035] Hereinafter, a method for constructing the material characteristic value prediction model stored in the material characteristic value prediction model storage unit 2 and acquired by the material characteristic value prediction model acquisition unit 3 will be described. The material property value prediction model is constructed by using known data in which the operation results in the manufacturing process of a steel material manufactured in the past (also referred to as "reference steel material" in this specification) are used as input and the material property values ​​measured on a test piece obtained from the reference steel material are used as output. For example, the material property value prediction model may be constructed by machine learning using the known data as teacher data. In the first embodiment, in order to quantify the uncertainty of prediction due to the lack of past operation results, a material property value prediction model is constructed using Gaussian process regression, which is a type of Bayesian estimation, but the present invention is not limited to this. For example, the material property value prediction model may be constructed by using a statistical method such as quantile regression using known data, or the material property value prediction model may be constructed by other machine learning such as a Bayesian neural network. Below, we will explain an overview of the Gaussian process regression used in the first embodiment. Note that in the following equations, variables written in bold represent vectors or matrices.

[0036] The input x expressed by the following formula (1) i When the output vector f, which is expressed by the following equation (3) and corresponds to the input x, which is expressed by the following equation (2), follows a Gaussian distribution N(0,K) with a mean of 0 and a covariance matrix K, which is expressed by the following equation (4), the output f is said to follow a Gaussian process, and can be written as shown in the following equation (5).

number

number

number

[0037] Although it is possible to use normal Gaussian process regression as shown in the above formula (7) as a material property value prediction model, normal Gaussian process regression cannot take into account changes in the variation in material property values ​​caused by differences in operational performance. For this reason, in the first embodiment, Gaussian processes gP(0,K1) and gP(0,K2) are defined that individually estimate the expected value (average value) and variance (standard deviation) represented by the following formulas (8) to (12), and a non-uniform variance Gaussian process (a Gaussian process considering non-uniform variance) is used in which a Gaussian distribution with the output of each Gaussian process as a parameter is the final prediction result.

number

[0038] By developing the above method, it is possible to estimate the covariance matrix Σ(x) between output variables, for example, as described below. Let D be the number of explanatory variables (input variables to the material property value prediction model), M be the number of objective variables (output variables from the material property value prediction model), and N be the number of records of actual data. For example, the past output actual data Y expressed by the following formula (13) is decorrelated based on formula (15) using the load matrix W expressed by the following formula (14) to obtain T expressed by the following formula (16).

number

number

[0039] That is, the output mean value vector μ corresponding to the new input x is expressed by the following equation (19). g (x) and the output standard deviation vector σ gBased on these prediction vectors, the output mean vector μ(x) and covariance matrix Σ(x) can be calculated by the following equations (21) and (22), respectively.

number

[0040] The material property value prediction model of the first embodiment is constructed by the method described above. The material property value prediction model is a model that takes operating conditions in a steel manufacturing process as input and outputs a probability distribution of the material property values ​​of the steel. When correlation between output variables is not taken into consideration, the material property value prediction model is a model expressed by the above-mentioned formula (12). When correlation between output variables is taken into consideration, the material property value prediction model is a model expressed by the above-mentioned formulas (18), (21), and (22).

[0041] FIG. 5 is a diagram for explaining an example of construction of a material characteristic value prediction model of the first embodiment in comparison with an example of construction of a conventional machine learning model. Specifically, for convenience of illustration, FIG. 5 shows an example of construction of a model that predicts tensile strength (TS) as a material characteristic value by plotting only the cooling temperature of a steel material in a hot rolling process among a plurality of parameters constituting the operating conditions. FIG. 5(a) shows an example of construction of a conventional machine learning model, and FIG. 5(b) shows an example of construction of a material characteristic value prediction model of the first embodiment. Data plotted with "●" in FIG. 5 is known data. In the conventional machine learning model shown in Fig. 5(a), the average value of the material characteristic value (TS) is output as the prediction result for the input operating condition (cooling temperature). Therefore, in the area surrounded by the dashed line in Fig. 5(a) where the material characteristic value varies widely even under the same operating condition, there is a problem that the prediction accuracy of the material characteristic value deteriorates. In contrast, the material characteristic value prediction model of the first embodiment shown in FIG. 5(b) outputs a probability distribution of the material characteristic value as a prediction result. Specifically, the material characteristic value prediction model of the present embodiment is a model using a non-uniform variance Gaussian process, and outputs a mean value and a standard deviation as a prediction result. The hatched area shown in FIG. 5(b) is a 95% prediction range (95% confidence interval) of a normal distribution obtained from the mean value and the standard deviation. As shown in FIG. 5(b), in the material characteristic value prediction model, an area with a large variation in the material characteristic value (an area corresponding to the area surrounded by a dashed line in FIG. 5(a)) is included in the probability distribution (95% prediction range) and is reflected as an output. In addition, in the material characteristic value prediction model, a probability distribution (95% prediction range) is output even in an area with no operating record (an area corresponding to the hatched area in FIG. 5(a)), so that the uncertainty of the prediction is taken into consideration.

[0042] <Constraint Condition Setting Section 4> The constraint condition setting unit 4 executes step ST13 shown in FIG. 4. Specifically, the constraint condition setting unit 4 sets the constraint conditions for the inverse analysis executed by the operation condition determination unit 5 described later as necessary. In the first embodiment, as shown in Table 2 below, the constraint conditions are set on the component values ​​of carbon (C) and manganese (Mn) in the steelmaking process in consideration of the application (including workability and weldability) of the target steel material, and the constraint conditions are set on the rolling temperature in the rolling process in consideration of the productivity of the target steel material. In addition, the constraint condition setting unit 4 may set a constraint condition between parameters constituting the operation conditions (for example, heating temperature in the rolling process ≧ rolling temperature + 100°C, etc.). The reason why the constraint conditions are necessary is that the specifications of the component values ​​are specified in the order information, or it is necessary to impose constraints on the operation conditions in order to create detailed material characteristic values ​​according to the application. [Table 2]

[0043] <Operation condition determination section 5> The operating condition determination unit 5 executes step ST14 shown in Fig. 4. Specifically, as outlined with reference to Fig. 2, the operating condition determination unit 5 determines operating conditions in the manufacturing process (in the first embodiment, component values ​​of carbon (C), manganese (Mn), and silicon (Si) in the steelmaking process, and heating temperature, rolling temperature, and cooling temperature in the rolling process) by performing inverse analysis of the material characteristic value prediction model acquired by the material characteristic value prediction model acquisition unit 3 so that the pass probability of the material characteristic value of the target steel material is equal to or higher than a predetermined threshold value (maximum in the first embodiment). The content of the inverse analysis executed by the operating condition determination unit 5 will be described below by taking as an example a case in which tensile strength (TS), yield point (YP), and elongation (EL) are considered as material characteristic values.

[0044] As described above, by using the material property value prediction model, when the operating condition x expressed by the following formula (23) is given, the material property value y (tensile strength y ts , yield point y yp , elongation y el ) can be output (using equation (12) or using equations (18), (21) and (22)).

number

number

[0045] In the above formula (25), y el upper , y el lower are the upper and lower limits of elongation (EL) determined by the order information. yp upper , y yplower are the upper and lower limits of the yield point (YP) determined by the order information. ts upper , y ts lower are the upper and lower limits of the tensile strength (TS) determined by the order information. If the order information does not include either the upper or lower limit (in the example shown in Table 1, there is no upper limit for EL), it is sufficient to set the upper limit to a sufficiently large value and the lower limit to a sufficiently small value. In addition, in the above equation (25), μ(x) and Σ(x) are the mean value vector and covariance matrix of the probability distribution p(y|x) of the material characteristic value y output from the material characteristic value prediction model.

[0046] Alternatively, each material characteristic value (tensile strength y ts , yield point y yp , elongation y el ) are assumed to be mutually independent, the operating conditions x may be searched for and determined such that the pass probability P(x), expressed as the geometric mean of the pass probability of each material characteristic value, is equal to or greater than a predetermined threshold (maximum), as shown in the following equation (26), without using the covariance matrix Σ(x).

number

[0047] In addition, the search for the operating conditions x for which the pass probability P(x) is equal to or greater than a predetermined threshold (maximum) may be performed by sequentially changing the operating conditions x under conditions that satisfy the constraints shown in Table 2 until the pass probability P(x) is equal to or greater than the predetermined threshold, or by using an optimization algorithm such as the constrained trust region method.

[0048] <Operation condition output section 6> The operating condition output unit 6 executes step ST15 shown in Fig. 4. Specifically, the operating condition output unit 6 outputs the operating conditions in the manufacturing process determined by the operating condition determination unit 5 (component values ​​of carbon (C), manganese (Mn) and silicon (Si) in the steelmaking process, and the heating temperature, rolling temperature and cooling temperature in the rolling process). It is preferable that the operating condition output unit 6 not only outputs the operating conditions in the determined manufacturing process, but also outputs a probability distribution (average value, standard deviation) of the material characteristic value output from the material characteristic value prediction model when the determined operating conditions are input to the material characteristic value prediction model, and an acceptance probability of the material characteristic value calculated from this probability distribution and order information. Fig. 6 is a diagram showing an example of the contents output from the operational condition output unit 6. The operational conditions (composition, rolling conditions) shown in Fig. 6 satisfy the constraints shown in Table 2, and also show that when the target steel material is manufactured under these operational conditions, the pass probability of the tensile strength (TS), yield point (YP) and elongation (EL) is as high as 0.99.

[0049] As described above, the operating condition determination device 100 according to the first embodiment can determine appropriate operating conditions for the manufacturing process in accordance with the order information so that the pass probability of the material characteristic value of the target steel material is equal to or greater than a predetermined threshold value (in the first embodiment, the parameters constituting the operating conditions are determined by a single value), thereby making it possible to improve the efficiency of quality design work.

[0050] [Second embodiment] The operating condition determination device 100 according to the first embodiment described above is configured to determine the parameters constituting the appropriate operating conditions according to the order information with one value. However, in quality design work, when predicting the material property value of the target steel material, there is a high need to predict the material property value by giving the parameters constituting the operating conditions as a range rather than as a single value, or to determine the parameters constituting the appropriate operating conditions as a range rather than as a single value so that the predicted material property value satisfies the order specification. Therefore, the operating condition determination device of the second embodiment further includes an operating condition prediction model that inputs the range of operating conditions in the steel manufacturing process and outputs the probability distribution of the operating conditions in the steel manufacturing process, and determines the range of operating conditions in the manufacturing process for the target steel by inversely analyzing the operating condition prediction model and the material characteristic value prediction model (inversely analyzing a composite function of a function represented by the operating condition prediction model and a function represented by the material characteristic value prediction model) so that the probability distribution of the operating conditions in the manufacturing process of the target steel that is output when the range of operating conditions in the manufacturing process of the target steel is input into the operating condition prediction model, the probability distribution of the material characteristic values ​​of the target steel that is output when the range of operating conditions in the manufacturing process of the target steel is input into the material characteristic value prediction model, and the probability of acceptance of the material characteristic values ​​of the target steel calculated using the specifications of the material characteristic values ​​of the target steel included in the order information is equal to or greater than a predetermined threshold value.

[0051] Fig. 7 is a block diagram showing a schematic configuration of an operation condition determination device according to the second embodiment. Fig. 8 is a flow chart showing a schematic procedure of an operation condition determination method executed by using the operation condition determination device shown in Fig. 7. As shown in FIG. 7, the operation condition determination device 100A according to the second embodiment is composed of, for example, a computer, similarly to the operation condition determination device 100 according to the first embodiment, and includes an operation condition prediction model storage unit 7 and an operation condition prediction model acquisition unit 8 in addition to the respective units similar to those of the operation condition determination device 100 according to the first embodiment. Among the units 1 to 8 included in the operating condition determination device 100A, the operations of the order information acquisition unit 1, the material characteristic value prediction model acquisition unit 3, and the constraint condition setting unit 4 (steps ST21, ST22, and ST24 shown in FIG. 8) are the same as those in the first embodiment, and therefore will not be described here. The operating condition prediction model storage unit 7, the operating condition prediction model acquisition unit 8, the operating condition determination unit 5, and the operating condition output unit 6 will be described below in order.

[0052] <Operation condition prediction model storage section 7> The operating condition prediction model storage unit 7 stores (memorizes) an operating condition prediction model constructed by an operating condition prediction model construction unit (not shown) included in the operating condition determination device 100A or by another device.

[0053] <Operation condition prediction model acquisition unit 8> The operating condition prediction model acquisition unit 8 executes step ST23 shown in Fig. 8. Specifically, the operating condition prediction model acquisition unit 8 acquires the operating condition prediction model stored in the operating condition prediction model storage unit 7.

[0054] Hereinafter, a method for constructing the operating condition prediction model stored in the operating condition prediction model storage unit 7 and acquired by the operating condition prediction model acquisition unit 8 will be described. The operating condition to be input to the material property value prediction model is represented by the following formula (27) x, the material property value to be predicted is represented by y, and the lower limit of the range of the operating condition x to be input to the operating condition prediction model is represented by the following formula (28) x lo The upper limit of the range of the operating conditions that are input to the operating condition prediction model is x up (i.e., x lo ≦x≦x up ), the probability distribution p(y|x lo ,x up ) is expressed by the following equation (30).

number

[0055] Considering the correlation between the variables of the operating condition prediction model, the operating condition prediction model p(x|x lo ,x upWhen constructing a probability model that spans a multidimensional space, with output variables in M ​​dimension and input variables in 2M dimension, the model is constructed, for example, by machine learning using performance data as training data, which leads to a deterioration in prediction accuracy. Therefore, in the second embodiment, it is assumed that the variables of the operating condition prediction model are independent, that is, the following formula (31) holds, and the probability distribution is estimated independently for each variable.

number

[0056] Fig. 9 shows an example of a comparison between the operational performance of silicon component values ​​and the probability distribution of silicon component values ​​predicted by the operational condition prediction model for a reference steel material in a steelmaking process, the silicon (Si) component value range of which is 20 to 25. In Fig. 9, the operational performance of silicon component values ​​is shown as a frequency ratio for comparison with the probability distribution of silicon component values ​​predicted by the operational condition prediction model. As shown in Fig. 9, it can be seen that the probability distribution of the operating conditions can be estimated with high accuracy by the operating condition prediction model.

[0057] On the other hand, for the steel temperature after the hot rolling process, the assumption of independence between processes does not hold, so it is necessary to build an operating condition prediction model taking into account the correlation between variables. Therefore, we consider sampling only data that is close to the correlation between variables calculated based on past performance data. The temperature-related variables are expressed by the following equation (32): x temp The lower limit of the temperature variable is expressed by the following equation (33): x temp,lo The upper limit of the temperature variable is expressed by the following equation (34): x temp,up Then, the probability distribution p(x temp |x temp,lo ,x temp,up ) is sampled based on the following equation (36): smpl Let us assume that.

number

[0058] As a method for evaluating the degree of deviation in correlation between variables, for example, a method using principal component analysis can be used, which will be described below. The actual temperature data is expressed as D in the following equation (37), and the eigenvector W i If the weighting matrix in which the eigenvalues ​​are arranged in descending order is W, which is expressed by the following equation (38), then x, which is expressed by the following equation (39), smpl The principal component T of and the restoration vector x smplrv can be calculated by the following equations (41) and (42), respectively.

number

number

[0059] In addition, since the integral calculation in the above-mentioned formula (30) cannot be calculated analytically, it is calculated by numerical integration based on samples in the operation condition determination unit 5 described later. For this reason, the operation condition prediction model storage unit 7 actually stores the probability distribution p(xi |x i lo ,x i up ) and the weight matrix W are stored, and the operating condition prediction model acquisition unit 8 acquires these.

[0060] <Operation condition determination section 5> The operating condition determination unit 5 executes step ST25 shown in Fig. 8. Specifically, the operating condition determination unit 5 determines the range of operating conditions in the manufacturing process by performing reverse analysis on the operating condition prediction model and the material property value prediction model so that the probability distribution of the operating conditions in the manufacturing process of the target steel material output when the range of operating conditions in the manufacturing process of the target steel material is input to the operating condition prediction model acquired by the operating condition prediction model acquisition unit 8 is calculated based on the probability distribution of the material property value of the target steel material output when the range of operating conditions in the manufacturing process of the target steel material is input to the material property value prediction model acquired by the material property value prediction model acquisition unit 3 and the specification of the material property value of the target steel material included in the order information, and the probability distribution of the material property value of the target steel material output when the range of operating conditions in the manufacturing process of the target steel material is input to the material property value prediction model acquired by the material property value prediction model acquisition unit 3 is equal to or higher than a predetermined threshold value.

[0061] Specifically, the contents executed by the operation condition determination unit 5 of the second embodiment are almost the same as the contents executed by the operation condition determination unit 5 of the first embodiment. However, the operation condition determination unit 5 of the second embodiment differs from the first embodiment in that the operation condition determination unit 5 searches for and determines the range of parameters constituting the operation conditions, whereas the operation condition determination unit 5 of the second embodiment searches for and determines the value of one of the parameters constituting the operation conditions. In the second embodiment, x represented by the above formula (28) lo , x represented by the above formula (29) up When the material characteristic value y (tensile strength y ts , yield point y yp , elongation y el ) conditional probability distribution p(y|x lo ,x up ), the operating condition determination unit 5 can estimate the lower limit x of the range of the operating condition x at which the probability of passing the target steel product is equal to or greater than a predetermined threshold value (maximum in the second embodiment). lo and upper limit x upThat is, the operation condition determination unit 5 searches for and determines the pass probability P(x lo ,x up The range of operating conditions (x lo ,x up ) is explored and determined.

number

[0062] Hereinafter, N(y S ;μ S (x lo ,x up ),σ S (x lo ,x up ) is calculated as follows. Operating condition prediction model (range of operating condition x (x lo ≦x≦x up ) is the probability distribution of operating condition x given p(x|x lo ,x up ) is sampled by the following equation (46): smpl In this case, N(y S ;μ S (x lo ,x up ),σ S (x lo ,x up ) can be approximated by the following equation (47).

number

[0063] The above equation (47) may be used as a nonparametric probability distribution as it is, but if we approximate the above equation (47) with a normal distribution, N(y S ;μ S (x lo ,x up ),σ S (x lo ,x up ) average value μ S (x lo ,x up ) and standard deviation σ S (x lo ,x up ) can be calculated using the following equations (48) and (49).

number

[0064] The pass probability P(x lo ,x up The range of operating conditions (x lo ,x up ) is searched for the pass probability P(x lo ,x upThe range of operating conditions (x) is set under the condition that satisfies the constraint conditions set by the constraint condition setting unit 4 until the value of x lo ,x up ) may be changed sequentially, or an optimization algorithm such as the constrained trust region method may be used. In addition, since it is difficult to specify a too narrow range of operating conditions in an actual manufacturing site, for example, lo +100 <x1 up It is desirable to explore the range of operating conditions with constraints such as the above.

[0065] <Operation condition output section 6> The operating condition output unit 6 executes step ST26 shown in Fig. 8. Specifically, the operating condition output unit 6 outputs the ranges (upper and lower limits) of the operating conditions in the manufacturing process determined by the operating condition determination unit 5 (component values ​​of carbon (C), manganese (Mn) and silicon (Si) in the steelmaking process, and the heating temperature, rolling temperature and cooling temperature in the rolling process). It is preferable that the operating condition output unit 6 not only outputs the ranges of the operating conditions in the manufacturing process determined, but also inputs the determined ranges of the operating conditions into an operating condition prediction model, and outputs a probability distribution (average value, standard deviation) of the material characteristic value output from the material characteristic value prediction model when the output of the operating condition prediction model is input into the material characteristic value prediction model, and an acceptance probability of the material characteristic value calculated from this probability distribution and order information. Fig. 10 is a diagram showing an example of the contents output from the operational condition output unit 6. The operational conditions (components, rolling conditions) shown in Fig. 10 satisfy the constraint conditions shown in Table 2 described in the first embodiment, and show that when the target steel material is manufactured within the range of these operational conditions, the pass probability of the tensile strength (TS), yield point (YP), and elongation (EL) is a high value of 0.95. However, unlike the first embodiment, in the second embodiment, the parameters constituting the operational conditions are determined not as one value but as a range, and the pass probability is calculated taking into account the variation in the operational conditions. Therefore, the pass probability is calculated to be smaller than the result shown in Fig. 6 described above.

[0066] In the second embodiment, even if the material characteristic value prediction model is a deterministic model (a model that outputs only expected values, not probability distributions), it is possible to search for and determine the range of appropriate operating conditions. In this case, the above-mentioned formula (30) can be rewritten as the following formula (50).

number

[0067] [Third embodiment] The operating condition determination device according to the third embodiment judges whether the order information for which the operating conditions are to be determined is order information that has been ordered in the past (hereinafter, this is appropriately referred to as "repeat order information") or new order information that has not been ordered in the past (hereinafter, this is appropriately referred to as "non-repeat order information"). When the order information for which the operating conditions are to be determined is non-repeat order information, the operating condition determination device according to the third embodiment judges whether the pass probability of the material characteristic value of the target steel, which is calculated based on the probability distribution of the material characteristic value of the target steel output when the initial operating conditions in the manufacturing process of the target steel, which are determined in advance based on at least the specifications of the material characteristic value of the target steel included in the non-repeat order information, are input to the material characteristic value prediction model, and the specifications of the material characteristic value of the target steel included in the non-repeat order information, is equal to or greater than a predetermined threshold value. When the pass probability when the initial operating conditions are used is equal to or greater than a predetermined threshold value, the operating condition determination device according to the third embodiment determines the initial operating conditions as the operating conditions in the manufacturing process of the target steel. On the other hand, if the pass probability of the material characteristic value of the target steel when the initial operating conditions are used is less than a predetermined threshold value, the operating condition determination device of the third embodiment is configured to determine the operating conditions in the manufacturing process for the target steel by inversely analyzing the material characteristic value prediction model so that the pass probability of the material characteristic value of the target steel is equal to or greater than the predetermined threshold value.

[0068] Fig. 11 is a block diagram showing a schematic configuration of an operation condition determination device according to the third embodiment. Fig. 12 is a flow chart showing a schematic procedure of an operation condition determination method executed by using the operation condition determination device shown in Fig. 11. As shown in FIG. 11, the operation condition determination device 100B according to the third embodiment is composed of, for example, a computer, similar to the operation condition determination device 100 according to the first embodiment, and includes a repeat order determination unit 9 and a manufacturing system acquisition unit 10 in addition to the respective units similar to those of the operation condition determination device 100 according to the first embodiment. Among the units 1 to 6, 9, and 10 included in the operating condition determination device 100B, the operations of the material characteristic value prediction model acquisition unit 3 and the constraint condition setting unit 4 (steps ST33 and ST37 shown in FIG. 12) are the same as those in the first embodiment, so that the description thereof will be omitted here. The order information acquisition unit 1, the repeat order determination unit 9, the manufacturing system acquisition unit 10, the operation condition determination unit 5, and the operation condition output unit 6 will be described below in order.

[0069] <Order Information Acquisition Unit 1> The order information acquisition unit 1 executes step ST31 shown in Fig. 12. Specifically, the order information acquisition unit 1 acquires, as order information, an order ID, a manufacturing specification code of the target steel material, product size such as plate thickness of the target steel material, standards, processes (water-cooled material, heat-treated material, etc.), specifications of material characteristic values ​​such as tensile strength (TS), and a manufacturing system of the target steel material from a predetermined database (not shown), as shown in Fig. 13. The manufacturing specification code and manufacturing system of the target steel material may or may not be included in the order information. The manufacturing specification code shown in Figure 13 is a code that is assigned the same value to order information with the same specifications and is set only for repeat order information. The relationship is manufacturing specification code: order ID = 1: many, and order information with the same manufacturing specification code has completely matching order specifications. Moreover, the manufacturing system shown in FIG. 13 is a value created from a part of the order specifications, and in the example shown in FIG. 13, it is a value obtained by concatenating the top two digits of the lower limit value of TS, "50", and the top two digits of the lower limit value of YP, "37". A relationship is formed in which multiple manufacturing specification codes are linked to one manufacturing system. The manufacturing system is created by classifying the order specifications based on information on a part of the order specifications (the lower limit value of TS and the upper limit value of YP in the example shown in FIG. 13), and is used for the purpose of assigning the initial operating conditions described below.

[0070] <Repeat order determination unit 9> The repeat order determination unit 9 executes step ST32 shown in FIG. 12. Specifically, the repeat order determination unit 9 determines whether the order information acquired by the order information acquisition unit 1 is repeat order information or non-repeat order information. More specifically, the operation condition determination device 100B stores a database in which past manufacturing specification codes are linked to operation conditions corresponding to the manufacturing specification codes, as shown in Table 4 below. The repeat order determination unit 9 searches the database to determine whether the manufacturing specification code (in the example shown in FIG. 13, "S100") included in the order information acquired by the order information acquisition unit 1 exists, and if the manufacturing specification code exists, it determines that the order information is a repeat order. If the repeat order determination unit 9 determines that the information is a repeat order (if "Yes" in step ST32 shown in Figure 12), the operating condition determination unit 5 determines the operating conditions corresponding to the manufacturing specification code present in the above database as the operating conditions for the target steel product, and the operating condition output unit 6 outputs the operating conditions (step ST39 in Figure 12). [Table 4] On the other hand, if the repeat order determination unit 9 determines that the order information acquired by the order information acquisition unit 1 does not contain a manufacturing specification code, or that the order information contains a manufacturing specification code but does not exist in the above database, and therefore it is not repeat order information (i.e., it is non-repeat order information) (if the answer is "No" in step ST32 shown in Figure 12), the manufacturing system acquisition unit 10, which will be described later, will operate.

[0071] <Manufacturing system acquisition department 10> The manufacturing system acquisition unit 10 executes step ST34 shown in FIG. 12. Specifically, the manufacturing system acquisition unit 10 acquires the operating conditions corresponding to the manufacturing system included in the order information acquired by the order information acquisition unit 1 as the initial operating conditions. More specifically, the operating condition determination device 100B stores a database in which a manufacturing system with a proven track record is linked to the operating conditions corresponding to the manufacturing system as shown in FIG. 14. FIG. 14(a) is the database itself, FIG. 14(b) is the database in which the contents of the steel type shown in FIG. 14(a) are stored, and FIG. 14(c) is the database in which the contents of the rolling condition pattern shown in FIG. 14(a) are stored. The manufacturing system acquisition unit 10 searches whether or not the operating conditions corresponding to the manufacturing system included in the order information acquired by the order information acquisition unit 1 exist in the above database, and if they exist (if "Yes" in step ST34 shown in FIG. 12), the operating conditions corresponding to the manufacturing system are acquired as the initial operating conditions, and the operating condition determination unit 5 operates using these initial operating conditions as described later. In addition, when there are multiple operating conditions corresponding to the manufacturing system, the manufacturing system acquisition unit 10 is configured to acquire, from among the multiple operating conditions, an operating condition that also corresponds to the plate thickness of the target steel material included in the order information acquired by the order information acquisition unit 1 as an initial operating condition. On the other hand, if the operating conditions corresponding to the manufacturing system are not present in the database ("No" in step ST34 shown in FIG. 12), the operating condition determination unit 5 operates without using the initial operating conditions, as described below. In addition, when the order information acquired by the order information acquisition unit 1 is as shown in FIG. 13 described above, since the manufacturing system is "5037" and the plate thickness is 12 mm, the operating conditions surrounded by a thick frame in FIG. 14 will be acquired as the initial operating conditions.

[0072] <Operation condition determination section 5> The operation condition determination unit 5 of the third embodiment executes different operations depending on the conditions. (If it is a repeat order) As described above, when the repeat order determination unit 9 determines that the order information acquired by the order information acquisition unit 1 is repeat order information ("Yes" at step ST32 shown in Figure 12), the operating condition determination unit 5 determines the operating conditions corresponding to the manufacturing specification code present in the database as the operating conditions for the target steel product.

[0073] (If it is non-repeated order information) When the repeat order determination unit 9 determines that the order information acquired by the order information acquisition unit 1 is non-repeat order information (in the case of "No" in step ST32 shown in Fig. 12) and the manufacturing system acquisition unit 10 can acquire the operating conditions corresponding to the manufacturing system as the initial operating conditions (in the case of "Yes" in step ST34 shown in Fig. 12), the operating condition determination unit 5 executes step ST35 shown in Fig. 12. Specifically, the operating condition determination unit 5 calculates the pass probability of the material characteristic value of the target steel material based on the probability distribution of the material characteristic value of the target steel material output when the initial operating conditions are input to the material characteristic value prediction model acquired by the material characteristic value prediction model acquisition unit 3 in step ST33 shown in Fig. 12, and the specification of the material characteristic value of the target steel material included in the order information.

[0074] FIG. 15 is a diagram showing an example of initial operating conditions input to a material characteristic value model, and a probability distribution and pass probability of material characteristic values ​​obtained under the initial operating conditions. The material characteristic value y (tensile strength y ts , yield point y yp ) Probability of passing p S (x)(S∈{ts,yp}) is the probability distribution N(y S ;μ S (x),σ S (x)) and the upper limit y of each material property value of the target steel material included in the order information S uppper and lower limit y S lower Therefore, it can be calculated by the following formula (53). For material characteristic values ​​for which no upper limit value exists in the order information, the upper limit value y S uppperFor material property values ​​that do not have a lower limit in the order information, the lower limit y S lower Then, we set it to -∞.

number

[0075] The above-described calculation of the pass probability is based on the assumption that the material characteristic values ​​are independent of each other, and may be lower than the actual pass probability. Therefore, as a method of calculating the pass probability when the correlation between the material characteristic values ​​is strong, the pass probability (total pass probability) may be calculated by the following formula (54) by estimating the covariance matrix Σ(x).

number

[0076] Next, the operating condition determination unit 5 executes step ST36 shown in Fig. 12. Specifically, the operating condition determination unit 5 judges whether or not the calculated pass probability of the material characteristic value of the target steel material is equal to or greater than a predetermined threshold value. For example, if the predetermined threshold value is set to 0.95, in the example shown in Fig. 15, the pass probability (total pass probability) is less than the threshold value ("No" in step ST36 shown in Fig. 12). For this reason, the operating condition determination unit 5 judges that the initial manufacturing conditions corresponding to the manufacturing system are inappropriate, and executes the contents of cases 1 to 3 described below. On the other hand, when the calculated pass probability of the material characteristic value of the target steel is equal to or higher than a predetermined threshold value (when "Yes" is selected in step ST36 shown in FIG. 12), the operating condition determination unit 5 determines the initial operating conditions as the operating conditions in the manufacturing process of the target steel. In this case, it is preferable to adopt an operation in which, when the initial operating conditions are used, the operator judges whether there are any problems with the quality (defects, shape, etc.) and production indexes (manufacturing cost, productivity, yield, etc.) other than the material characteristic value of the target steel in addition to the pass probability of the material characteristic value, and when it is judged that there are no problems, the initial operating conditions are finally determined as the operating conditions.

[0077] When the manufacturing system acquisition unit 10 cannot acquire the operating conditions corresponding to the manufacturing system as the initial operating conditions (when the answer is "No" in step ST34 shown in FIG. 12) or when the pass probability of the material characteristic value of the target steel calculated by the operating condition determination unit 5 is less than a predetermined threshold value (when the answer is "No" in step ST36 shown in FIG. 12), the constraint condition setting unit 4 sets the constraint conditions as necessary (step ST37 shown in FIG. 12), and the operating condition determination unit 5 searches and determines the operating conditions in the manufacturing process of the target steel by performing a reverse analysis of the material characteristic value prediction model (step ST38 in FIG. 12). Specifically, the operating condition determination unit 5 searches and determines the operating conditions in the manufacturing process of the target steel in any one of cases 1 to 3 described below, or in multiple cases. FIG. 16 is a diagram showing an example of the operating conditions searched for in cases 1 to 3. In FIG.

[0078] (Case 1) In case 1, the operation condition determination unit 5 inputs the operation conditions obtained by arbitrarily combining the steel type stored in the database as shown in FIG. 14(b) and the rolling condition pattern stored in the database as shown in FIG. 14(c) into the material property value prediction model to calculate the pass probability of the material property value of the target steel. Then, the operation conditions under which the calculated pass probability of the material property value is equal to or greater than a predetermined threshold value are determined as the operation conditions in the manufacturing process of the target steel. In the example shown in FIG. 16, the operation conditions surrounded by a thick frame in case 1 are equal to or greater than the threshold value of 0.95, and are determined as the operation conditions in the manufacturing process of the target steel. At this time, it is preferable to adopt an operation in which, when the determined operation conditions are used, the operator judges whether there is any problem in the quality (defects, shape, etc.) and production index (production cost, productivity, yield, etc.) other than the material property value of the target steel in addition to the pass probability of the material property value, and if it is judged that there is no problem, the operation conditions are finally determined as the operation conditions.

[0079] (Case 2) Case 2 is effective, but not limited to, for example, when suitable operating conditions cannot be obtained in case 1. In case 2, the operating condition determination unit 5 uses the steel type stored in the database as shown in Fig. 14(b) as a constraint (without using the rolling condition pattern as shown in Fig. 14(c) as a constraint) to perform inverse analysis of the material characteristic value prediction model, in other words, to solve an optimization problem with the pass probability as an evaluation function, in the same manner as in the first embodiment described above, and searches for and determines operating conditions in the manufacturing process of the target steel material that make the pass probability of the material characteristic value of the target steel material equal to or greater than a predetermined threshold value. In the example shown in Fig. 16, the operating conditions enclosed by a thick frame in case 2 are equal to or greater than the threshold value of 0.95, and are determined as the operating conditions in the manufacturing process of the target steel material.

[0080] (Case 3) Case 3 is effective, but not limited to, when suitable operating conditions cannot be obtained in cases 1 and 2, for example. In case 3, the operating condition determination unit 5 searches for and determines operating conditions in the manufacturing process of the target steel material that will result in a pass probability of the material property value of the target steel material equal to or greater than a predetermined threshold value by performing a reverse analysis of the material property value prediction model, i.e., solving an optimization problem with the pass probability as an evaluation function, in the same manner as in the first embodiment described above, without using the steel type stored in the database as shown in Fig. 14(b) or the rolling condition pattern as shown in Fig. 14(c) as constraint conditions. In the example shown in Fig. 16, the operating conditions enclosed by a thick frame in case 3 are equal to or greater than the threshold value of 0.95, and are determined as the operating conditions in the manufacturing process of the target steel material.

[0081] As described above, the operating condition determination unit 5 executes one of cases 1 to 3, or a plurality of cases. The case to be executed may be displayed on a selection screen on a monitor of a computer constituting the operating condition determination device 100B, so that an operator can manually select a case to be executed. Alternatively, it is also possible to adopt a mode in which the priority order of cases to be executed is preset, such as executing case 1 first, executing case 2 if appropriate operating conditions are not obtained in case 1, and executing case 3 if appropriate operating conditions are not obtained in case 2.

[0082] <Operation condition output section 6> The operational condition output unit 6 executes step ST39 shown in Fig. 12. Specifically, the operational condition output unit 6 outputs the operational conditions in the manufacturing process determined by the operational condition determination unit 5 (component values ​​such as carbon (C) in the steelmaking process, heating temperature in the rolling process, rolling temperature, etc.). When the order information acquired by the order information acquisition unit 1 is repeat order information, the operation condition output unit 6 outputs the operation conditions corresponding to the manufacturing specification code as the operation conditions for the determined manufacturing process. If the order information acquired by the order information acquisition unit 1 is non-repeat order information and the pass probability of the material characteristic value obtained under the initial operating conditions is equal to or greater than a predetermined threshold value, the initial operating conditions are output as the operating conditions for the determined manufacturing process. If the order information acquired by the order information acquisition unit 1 is non-repeat order information and the initial operating conditions cannot be acquired, or if the pass probability of the material characteristic value obtained under the initial operating conditions is less than a predetermined threshold value, the operating conditions obtained in any one of the above-mentioned cases 1 to 3, or in multiple cases, are output as the operating conditions for the determined manufacturing process. The operating condition output unit 6 preferably not only outputs the operating conditions in the determined manufacturing process, but also outputs the probability distribution (average, standard deviation) of the material characteristic value output from the material characteristic value prediction model when the determined operating conditions are input to the material characteristic value prediction model, and the pass probability of the material characteristic value calculated from this probability distribution and the order information. When all of the above-mentioned cases 1 to 3 are executed, all of the data surrounded by a thick frame in FIG. 16 may be output, and when any of the above-mentioned cases 1 to 3 is executed, the data surrounded by a thick frame for the corresponding case in FIG. 16 may be output. It is also possible to adopt a mode in which all data including data whose pass probability is less than a predetermined threshold value is output (i.e., all of the data shown in FIG. 16 is output).

[0083] In the above-described third embodiment, as in the first embodiment, the parameters constituting the operating conditions according to the order information are determined as one value each, but the present invention is not limited to this. Although detailed description is omitted, it is also possible to adopt a configuration in which, while having the features of the third embodiment (a configuration in which initial operating conditions are used for non-repeat order information), an operating condition prediction model storage unit 7 and an operating condition prediction model acquisition unit 8 are provided as in the second embodiment, and parameters constituting appropriate operating conditions are determined as a range rather than as a single value. That is, in the above configuration, the probability distribution of the operating conditions in the manufacturing process of the target steel, which is output when the range of initial operating conditions in the manufacturing process of the target steel, which is determined in advance by at least the specifications of the material property values ​​of the target steel included in the non-repeat order information, is input to the operating condition prediction model, is determined whether the pass probability of the material property values ​​of the target steel, which is calculated by the probability distribution of the material property values ​​of the target steel output when input to the material property value prediction model and the specifications of the material property values ​​of the target steel included in the non-repeat order information, is equal to or greater than a predetermined threshold value. If the pass probability when the range of initial operating conditions is used is equal to or greater than the predetermined threshold value, the above configuration determines the range of initial operating conditions as the operating conditions in the manufacturing process of the target steel. On the other hand, if the pass probability of the material property values ​​of the target steel when the range of initial operating conditions is used is less than the predetermined threshold value, the above configuration determines the range of operating conditions in the manufacturing process for the target steel by reverse analysis of the operating condition prediction model and the material property value prediction model so that the pass probability of the material property values ​​of the target steel is equal to or greater than the predetermined threshold value.

[0084] In the present embodiment described above, the operation condition determination apparatus 100 (or the operation condition determination apparatus 100A or 100B) may be implemented using a plurality of computers. For example, the operation condition determination apparatus 100 may be implemented using an apparatus such as a cloud server. Furthermore, the hardware processor and memory of the operation condition determination apparatus 100 may be distributed and implemented in a plurality of computers.

[0085] Although one embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and designs that do not deviate from the gist of the present invention are also included. [Explanation of symbols]

[0086] 1. Order information acquisition unit 2. Material property value prediction model storage section 3. Material property value prediction model acquisition section 4. Constraint Condition Setting Section 5. Operation condition determination section 6. Operation condition output section 7. Operation condition prediction model storage section 8. Operation condition prediction model acquisition section 9. Repeat order determination section 10...Manufacturing System Acquisition Department 100, 100A, 100B...Operation condition determination device

Claims

1. An operating condition determination device that determines operating conditions in a steel manufacturing process according to order information, A material characteristic value prediction model is provided that inputs operational conditions in a manufacturing process of the steel material and outputs a probability distribution of material characteristic values ​​of the steel material, determining operation conditions in the manufacturing process by performing a reverse analysis on the material property value prediction model so that the pass probability of the material property value of the steel, calculated from a probability distribution of the material property value of the steel output when the operation conditions in the manufacturing process of the steel are input into the material property value prediction model and the specifications of the material property value of the steel included in the order information, is equal to or greater than a predetermined threshold value; Operation condition determination device.

2. determining operational conditions in the manufacturing process by performing a reverse analysis of the material characteristic value prediction model under constraints of the operational conditions at least due to the application of the steel material; The operating condition determining device according to claim 1 .

3. determining operation conditions in the manufacturing process by performing a reverse analysis of the material characteristic value prediction model under constraints of the operation conditions taking into account at least one of the productivity of the steel material and qualities other than the material characteristic values; 3. The operating condition determining device according to claim 1 or 2.

4. The material characteristic value prediction model is a model using Bayesian estimation.

3. The operating condition determining device according to claim 1 or 2.

5. The material characteristic value prediction model is a model using a heteroscedastic Gaussian process.

3. The operating condition determining device according to claim 1 or 2.

6. An operating condition prediction model that inputs a range of operating conditions in the steel manufacturing process and outputs a probability distribution of operating conditions in the steel manufacturing process, a probability distribution of operating conditions in the steel manufacturing process output when the range of operating conditions in the steel manufacturing process is input to the operating condition prediction model, and the probability distribution of the material characteristic values ​​of the steel output when the range of operating conditions in the steel manufacturing process is input to the material characteristic value prediction model is calculated based on the probability distribution of the material characteristic values ​​of the steel output when the range of operating conditions in the steel manufacturing process is input to the material characteristic value prediction model and the specifications of the material characteristic values ​​of the steel included in the order information, and the probability of acceptance of the material characteristic values ​​of the steel is determined by reverse analysis of the operating condition prediction model and the material characteristic value prediction model, so that the range of operating conditions in the manufacturing process is equal to or greater than a predetermined threshold value; 3. The operating condition determining device according to claim 1 or 2.

7. The order information is new order information that has not been ordered in the past, When an initial operating condition in a manufacturing process of the steel material, which is predetermined at least by the specification of the material characteristic value of the steel material included in the order information, is input to the material characteristic value prediction model, a probability distribution of the material characteristic value of the steel material output when the initial operating condition in the manufacturing process of the steel material is input, and a probability distribution of the material characteristic value of the steel material included in the order information is calculated based on the specification of the material characteristic value of the steel material included in the order information, is equal to or greater than a predetermined threshold value, the initial operating condition is determined as an operating condition in the manufacturing process of the steel material, When the probability of passing the material characteristic value of the steel material when the initial operating conditions are used is less than a predetermined threshold value, the operating conditions in the manufacturing process are determined by performing a reverse analysis of the material characteristic value prediction model so that the probability of passing the material characteristic value of the steel material is equal to or greater than a predetermined threshold value.

3. The operating condition determining device according to claim 1 or 2.

8. The order information is new order information that has not been ordered in the past, a probability distribution of operating conditions in the manufacturing process of the steel that is output when a range of initial operating conditions in the manufacturing process of the steel, which is predetermined at least by the specifications of the material property values ​​of the steel included in the order information, is input into the operating condition prediction model, and if a probability of acceptance of the material property values ​​of the steel calculated from the probability distribution of the material property values ​​of the steel that is output when the range is input into the material property value prediction model and the specifications of the material property values ​​of the steel included in the order information is equal to or greater than a predetermined threshold value, the range of initial operating conditions is determined as the operating conditions in the manufacturing process of the steel; When the pass probability of the material characteristic value of the steel material when the range of the initial operating conditions is used is less than a predetermined threshold value, the range of operating conditions in the manufacturing process is determined by performing a reverse analysis on the operating condition prediction model and the material characteristic value prediction model so that the pass probability of the material characteristic value of the steel material is equal to or greater than a predetermined threshold value. The operating condition determining device according to claim 6.

9. The initial operating conditions are predetermined based on at least the material property values ​​and thickness specifications of the steel material included in the order information. The operating condition determining device according to claim 7.

10. The range of the initial operating conditions is predetermined at least by the material property values ​​and thickness specifications of the steel material included in the order information. The operating condition determining device according to claim 7.

11. The manufacturing process includes a steelmaking process and a rolling process which is a process subsequent to the steelmaking process, When the probability of passing the material characteristic value of the steel material when the initial operating conditions are used is less than a predetermined threshold value, the material characteristic value prediction model is reverse-analyzed using the steel type of the steel material as a constraint condition so that the probability of passing the material characteristic value of the steel material is equal to or greater than a predetermined threshold value, thereby determining operating conditions for the processes subsequent to the rolling process in the manufacturing process. The operating condition determining device according to claim 7.

12. The manufacturing process includes a steelmaking process and a rolling process which is a process subsequent to the steelmaking process, When the pass probability of the material characteristic value of the steel material when the range of the initial operating conditions is used is less than a predetermined threshold value, the material characteristic value prediction model is reverse-analyzed so that the pass probability of the material characteristic value of the steel material is equal to or greater than a predetermined threshold value, using the steel type of the steel material as a constraint condition, thereby determining the operating condition prediction model and the operating conditions for the processes after the rolling process among the manufacturing processes. The operating condition determining device according to claim 8.

13. An operating condition determination method for determining operating conditions in a steel manufacturing process in accordance with order information, comprising: A material characteristic value prediction model is used, which inputs the operating conditions in the manufacturing process of the steel material and outputs a probability distribution of the material characteristic value of the steel material, determining operation conditions in the manufacturing process by performing a reverse analysis on the material property value prediction model so that the pass probability of the material property value of the steel, calculated from a probability distribution of the material property value of the steel output when the operation conditions in the manufacturing process of the steel are input into the material property value prediction model and the specifications of the material property value of the steel included in the order information, is equal to or greater than a predetermined threshold value; How to determine operating conditions.

14. A program for causing a computer to function as the operation condition determination device according to claim 1 or 2.

Citation Information

Patent Citations

  • Manufacturing specification determination support device, manufacturing specification determination support method, computer program, and computer readable recording medium

    JP2022014878A