Methods for generating quality prediction models for metallic materials, methods for predicting the quality of metallic materials, methods for estimating quality influencing factors of metallic materials, manufacturing methods for metallic materials, methods for generating quality prediction models for metallic materials, devices for generating quality prediction models for metallic materials, devices for predicting the quality of metallic materials, and devices for estimating quality influencing factors of metallic materials.
By collecting and adjusting the manufacturing conditions and quality data granularity of each process, a high-precision quality prediction model is generated, which solves the problem of insufficient accuracy of existing quality prediction models and realizes high-precision quality prediction and factor estimation.
Patent Information
- Application Number
- CN202580026390.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-04-09
- Filing Date
- 2025-01-24
- Publication Date
- 2026-11-10
AI Technical Summary
In the existing technology, the quality prediction model for metal materials does not fully consider the granularity of manufacturing conditions in intermediate processes, resulting in insufficient prediction accuracy and reliability of factor estimation. In particular, it is difficult to predict quality defects with high accuracy when manufacturing conditions change rapidly or nonlinearly.
By collecting manufacturing conditions data for each process and quality data for the final process, and determining the granularity of the manufacturing condition data and quality data based on specific indicators, a high-precision quality prediction model is generated using statistical analysis and machine learning methods, and the manufacturing condition data for each process is appropriately utilized.
It enables high-precision prediction of metal material quality under arbitrary manufacturing conditions and high-precision estimation of quality defects, thereby improving the accuracy of the quality prediction model and the reliability of factor estimation.
Smart Images

Figure CN123076448A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for generating a quality prediction model for metallic materials, a method for predicting the quality of metallic materials, a method for estimating the quality influencing factors of metallic materials, a method for manufacturing metallic materials, a method for generating a quality prediction model for metallic materials, a device for generating a quality prediction model for metallic materials, a device for predicting the quality of metallic materials, and a device for estimating the quality influencing factors of metallic materials. Background Technology
[0002] Patent Document 1 discloses a method for predicting the quality of a material by using a model that learns from data that correlates manufacturing condition data during the material’s manufacturing process with quality data in the final process, taking into account the quality determination position along the material’s length direction.
[0003] Existing technical documents
[0004] Patent documents
[0005] Patent Document 1: Japanese Patent No. 7207547 Summary of the Invention
[0006] The problem that the invention aims to solve
[0007] Material quality can vary depending on single or multiple manufacturing conditions, and defects can occur in different ways depending on the manufacturing conditions that contribute to them. Therefore, in order to accurately predict or estimate the factors contributing to defects, it is necessary to appropriately utilize information from the corresponding manufacturing conditions as factors, based on the type of defect.
[0008] However, in the prior art, because manufacturing condition data and quality data are correlated at a uniform granularity, information is not fully and appropriately utilized. Specifically, in the prior art, each step in the manufacturing of metal materials is aligned with the granularity (units of length of the metal material) of the quality data and manufacturing condition data in the final step for quality evaluation. Therefore, the utilization of the granularity of manufacturing condition data in intermediate steps, which could potentially be a major factor in quality defects, is not fully and appropriately considered.
[0009] For example, when each process is a rolling process, the metal material mainly increases in length after each process. Even when the length unit of the metal material in the final process is aligned with the final process, the length of the metal material in the corresponding intermediate process sometimes becomes shorter than the measurement cycle of the actual manufacturing conditions data for that intermediate process as the process is traced back.
[0010] Therefore, the manufacturing condition data of the metal materials in the intermediate processes is converted into data in units of length of the metal materials in the final processes by interpolation (e.g., linear interpolation) to align with the length units of the metal materials in the final processes used for quality evaluation. Then, a dataset is generated that associates the converted manufacturing condition data from the intermediate processes with the quality data and manufacturing condition data of the final processes.
[0011] However, when the measurement cycle for actual manufacturing conditions is shorter than that of intermediate processes, even with data interpolation, it may not accurately reflect changes in actual manufacturing conditions. Especially when manufacturing conditions change drastically or non-linearly, linear interpolation may sometimes fail to fully capture these changes, risking a decrease in the accuracy of the quality prediction model. Furthermore, if interpolation increases the number of data points for each process, the quality prediction model may overfit, leading to a decline in generalization performance.
[0012] Therefore, the accuracy of models (quality prediction models for metallic materials) that learn from datasets by interpolating manufacturing condition data from intermediate processes that might otherwise be major factors in quality defects, and then linking these datasets to quality defect data from the final process, sometimes cannot be sufficiently improved. Thus, in the prior art, there are limits to improving the accuracy of quality predictions for metallic materials and the reliability of inferring factors contributing to quality defects.
[0013] This invention was made in view of the above circumstances, and its purpose is to provide a method for generating a quality prediction model for metallic materials, a method for predicting the quality of metallic materials, a method for estimating the quality influencing factors of metallic materials, a method for manufacturing metallic materials, a method for generating a quality prediction model for metallic materials, an apparatus for generating a quality prediction model for metallic materials, an apparatus for predicting the quality of metallic materials, and an apparatus for estimating the quality influencing factors of metallic materials. Based on specific indicators including an index representing the prediction accuracy of the quality prediction model and an index representing the importance of the prediction of the quality of the final process, the roughness (granularity) of the data that aligns and correlates the manufacturing condition data and quality data of each manufacturing process is determined, and the granularity of the manufacturing conditions of each process is appropriately utilized. This enables high-precision prediction of the quality of the material relative to any manufacturing conditions and high-precision estimation of factors causing quality defects.
[0014] Methods for solving problems
[0015] To address the aforementioned issues and achieve the objective, the method for generating a quality prediction model for metallic materials according to the present invention includes: a first collection step, which collects the manufacturing conditions of metallic materials in each process according to each specified range of the metallic materials determined according to different processes; a second collection step, which evaluates and collects the quality of the metallic materials in the final process of each process according to each specified range determined for the final process; and a storage step, which, based on the width of a specified range in a specific process of each process, evaluates the manufacturing conditions of the metallic materials in other processes for the manufacturing conditions of the metallic materials collected in the first collection step and the quality of the metallic materials in the final process collected in the second collection step. The manufacturing conditions and the quality are edited, and the manufacturing conditions and the quality are repeatedly associated with and saved according to each uniform specified range corresponding to the number of each process, which is a specified range that supplements the density difference of the manufacturing conditions and the quality in each process; the model generation step generates one or more quality prediction models that predict the quality of each uniform specified range according to the manufacturing conditions of each uniform specified range saved in the saving step; and the model selection step selects the quality prediction model used to predict the quality of the metal material from the multiple quality prediction models based on the evaluation of specific indicators, if multiple quality prediction models are generated in the model generation step.
[0016] In the above invention, the saving step of the method for generating a quality prediction model for metallic materials of the present invention takes all processes of each process as the object, selects a specific process separately, takes the specified range in the selected specific process as the unified specified range, and repeatedly associates and saves the manufacturing conditions and the quality according to each range of one or more of the unified specified ranges.
[0017] In the above invention, the method for generating a quality prediction model for metallic materials of the present invention, in the model selection step, selects the quality prediction model to be used in predicting the quality of the metallic material from a plurality of quality prediction models, based on the specific index representing the prediction accuracy of the quality prediction model.
[0018] In the above invention, the model selection step of the method for generating a quality prediction model for metallic materials of the present invention selects the quality prediction model used in predicting the quality of the metallic material from a plurality of quality prediction models based on an importance index, wherein the importance index represents the influence of the manufacturing conditions that become the input of each quality prediction model on the prediction of the quality that becomes the output.
[0019] In the above invention, in the model selection step of the method for generating a quality prediction model for metallic materials of the present invention, a quality prediction model is generated based on actual data associated with the specified range of the final process as the unified specified range. For the generated quality prediction model, an importance index is calculated for each manufacturing condition, and the manufacturing condition with the highest importance index is extracted. The quality prediction model generated based on the actual data associated with the specified range of the process to which the extracted manufacturing condition belongs as the unified specified range is selected as the quality prediction model used for predicting the quality of the metallic material.
[0020] In the above invention, in the model selection step of the method for generating a quality prediction model for metallic materials, a first quality prediction model is generated based on actual data associated with the specified range of the final process as the unified specified range; for the first quality prediction model, a first importance index is calculated for each manufacturing condition, and the first importance index is extracted as the manufacturing condition with the highest importance; a second quality prediction model is generated based on actual data associated with the specified range of the process to which the extracted manufacturing condition belongs as the unified specified range; for the second quality prediction model, a second importance index is calculated for each manufacturing condition, and the second importance index is extracted as the manufacturing condition with the highest importance; the second quality prediction model when the manufacturing condition with the highest importance is consistent with the manufacturing condition with the highest importance is selected as the quality prediction model used for predicting the quality of the metallic material.
[0021] In the above invention, the model generation step of the method for generating a quality prediction model for metallic materials of the present invention uses statistical analysis methods and machine learning methods, including linear regression, local regression, principal component regression, PLS regression, logistic regression, support vector machine, decision tree, regression tree, random forest, gradient boosting tree, and neural network, to generate the quality prediction model.
[0022] To address the aforementioned issues and achieve the objectives, the metal material quality prediction method of the present invention includes a quality prediction step, wherein the quality prediction step uses a quality prediction model generated by the aforementioned metal material quality prediction model generation method to predict the quality of metal materials manufactured under arbitrary manufacturing conditions within each specified range.
[0023] To address the aforementioned issues and achieve the objectives, the method for estimating the quality influencing factors of metallic materials according to the present invention includes an influencing factor estimation step. This step uses a quality prediction model generated by the aforementioned method for generating quality prediction models for metallic materials to estimate the manufacturing conditions that influence the quality of the metallic materials.
[0024] In the above invention, in the influence factor estimation step of the method for estimating the quality influence factor of the metal material of the present invention, the importance index of the quality prediction model is calculated for each manufacturing condition, and the manufacturing conditions with high importance index are estimated to be manufacturing conditions that have an impact on the quality of the metal material.
[0025] To address the aforementioned issues and achieve the objectives, the present invention provides a method for manufacturing metallic materials comprising: a quality prediction step, using a quality prediction model to predict the quality of the output, wherein the quality prediction model includes, in its input, manufacturing conditions that influence the quality of the metallic materials, as estimated by the aforementioned method for estimating quality influencing factors of metallic materials; a manufacturing condition determination step, determining the manufacturing conditions that influence the quality of the metallic materials in a manner that ensures the predicted quality falls within a predetermined range; and a metallic material manufacturing step, manufacturing the metallic materials according to the determined manufacturing conditions.
[0026] To address the aforementioned issues and achieve the objective, the quality prediction model generation method of the present invention includes: a first collection step, collecting the manufacturing conditions of materials in each process according to each specified range of the materials determined according to different processes; a second collection step, evaluating and collecting the quality of the materials in the final process of each process according to each specified range determined for the final process; and a storage step, for the manufacturing conditions of the materials in each process collected in the first collection step and the quality of the materials in the final process collected in the second collection step, using the width of the specified range in a specific process as a benchmark, evaluating the manufacturing conditions and quality in other processes. The process involves editing and repeatedly associating and saving the manufacturing conditions and quality according to each uniform specified range corresponding to the number of each process, which serves as a supplement to the density differences of the manufacturing conditions and quality in each process; a model generation step, which generates one or more quality prediction models predicting the quality of each uniform specified range according to the manufacturing conditions of each uniform specified range saved in the saving step; and a model selection step, which, if multiple quality prediction models are generated in the model generation step, selects the quality prediction model to be used in predicting the quality of the material from the multiple quality prediction models based on the evaluation of specific indicators.
[0027] To address the aforementioned issues and achieve the objective, the metal material quality prediction model generation apparatus of the present invention comprises: a first collection unit that collects the manufacturing conditions of the metal material in each process according to each specified range of the metal material determined according to different processes; a second collection unit that evaluates and collects the quality of the metal material in the final process of each process according to each specified range determined for the final process; and a storage unit that, based on the width of a specified range in a specific process, evaluates the manufacturing conditions of the metal material in each process collected by the first collection unit and the quality of the metal material in the final process collected by the second collection unit, and then stores the manufacturing conditions in other processes. The conditions and the quality are edited, and the manufacturing conditions and the quality are repeatedly associated with and saved according to a unified specified range corresponding to the number of each process, which is a specified range that supplements the density difference of the manufacturing conditions and the quality in each process; the model generation unit generates one or more quality prediction models that predict the quality of each unified specified range according to the manufacturing conditions of each unified specified range saved by the saving unit; and the model selection unit selects the quality prediction model used for predicting the quality of the metal material from the multiple quality prediction models based on the evaluation of specific indicators when multiple quality prediction models are generated by the model generation unit.
[0028] In order to solve the above-mentioned problems and achieve the objectives, the metal material quality prediction device of the present invention includes a quality prediction unit, which uses a quality prediction model generated by the metal material quality prediction model generation device described above to predict the quality of metal materials manufactured under arbitrary manufacturing conditions within each specified range.
[0029] In order to solve the above-mentioned problems and achieve the objectives, the metal material quality influencing factor estimation device of the present invention includes an influencing factor estimation unit, which uses a quality prediction model generated by the metal material quality prediction model generation device to estimate the manufacturing conditions that affect the quality of the metal material.
[0030] Invention Effects
[0031] According to the present invention, the method for generating a quality prediction model for metallic materials, the method for predicting the quality of metallic materials, the method for estimating the quality influencing factors of metallic materials, the manufacturing method for metallic materials, the apparatus for generating a quality prediction model for metallic materials, the apparatus for predicting the quality of metallic materials, and the apparatus for estimating the quality influencing factors of metallic materials, by appropriately utilizing the granularity of manufacturing condition data from each process of manufacturing metallic materials, can reflect the maximum amount of information related to the quality of the object in the generated quality prediction model. Therefore, it is possible to predict quality with high accuracy and to estimate the factors causing quality defects with high accuracy. Attached Figure Description
[0032] [ Figure 1 ] Figure 1 This is a block diagram illustrating an example of the structure of an information processing device that functions as an embodiment of the present invention, comprising a metal material quality prediction model generation device, a quality prediction device, and a quality influence factor estimation device.
[0033] [ Figure 2 ] Figure 2 This is a flowchart illustrating the method for generating a quality prediction model for metallic materials and the process of the quality prediction method according to an embodiment of the present invention.
[0034] [ Figure 3 ] Figure 3 This is a diagram illustrating an example of manufacturing condition data for each predetermined range along the length direction of a metal material, collected by the manufacturing condition data collection unit in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus according to embodiments of the present invention.
[0035] [ Figure 4 ] Figure 4 This diagram illustrates an example of a situation in which, in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus of the embodiments of the present invention, the manufacturing condition data and quality data in the specified range of processes 2 and 3 are edited by the consistent process data editing unit based on the width of the specified range of process 1 (uniform specified range).
[0036] [ Figure 5 ] Figure 5 This figure illustrates an example of a situation in which, in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus of the embodiments of the present invention, the manufacturing condition data and quality data in the specified range of processes 1 and 3 are edited by the consistent process data editing unit based on the width of the specified range of process 2 (uniform specified range).
[0037] [ Figure 6 ] Figure 6This figure illustrates an example of a situation in which, in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus of the embodiments of the present invention, the manufacturing condition data and quality data in the specified range of processes 1 and 2 are edited by the consistent process data editing unit based on the width of the specified range of process 3 (uniform specified range).
[0038] [ Figure 7 ] Figure 7 This diagram illustrates an example of actual data in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus according to embodiments of the present invention, where the data is generated by a consistent process data editing unit, taking the width of the specified range of process 1 as a reference (uniform specified range), and combining the manufacturing condition data and quality data from the specified range of all processes.
[0039] [ Figure 8 ] Figure 8 This diagram illustrates an example of actual data in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus according to embodiments of the present invention, where the data is generated by a consistent process data editing unit, taking the width of the specified range of process 2 as a reference (uniform specified range), and combining manufacturing condition data and quality data from the specified range of all processes.
[0040] [ Figure 9 ] Figure 9 This diagram illustrates an example of actual data in the metal material quality prediction model generation apparatus, quality prediction apparatus, and quality influence factor estimation apparatus according to embodiments of the present invention, where the data is generated by a consistent process data editing unit, based on the width of the specified range of process 3 (uniform specified range), and combining manufacturing condition data and quality data from the specified range of all processes.
[0041] [ Figure 10 ] Figure 10 This is a flowchart illustrating the process of estimating the quality influence factor of metallic materials according to an embodiment of the present invention. Detailed Implementation
[0042] Referring to the accompanying drawings, the method for generating a quality prediction model for metallic materials, the method for predicting the quality of metallic materials, the method for estimating quality influencing factors, the apparatus for generating a quality prediction model for metallic materials, the apparatus for predicting the quality of metallic materials, and the apparatus for estimating quality influencing factors according to embodiments of the present invention will be described.
[0043] (Device Structure)
[0044] Reference Figure 1This embodiment describes the structure of the metal material quality prediction model generation apparatus, the quality prediction apparatus, and the quality influencing factor estimation apparatus. The quality prediction model generation apparatus is used to generate a quality prediction model that predicts the quality of metal materials manufactured through one or more processes. The quality prediction apparatus is used to predict the quality of metal materials manufactured through one or more processes using the quality prediction model. The quality influencing factor estimation apparatus is used to estimate the manufacturing conditions that influence the quality of the metal material using the quality prediction model. It should be noted that the metal material in this embodiment, for example, is a steel product, and examples include semi-finished products such as slabs and finished steel plates manufactured by rolling these slabs.
[0045] Furthermore, the following describes the quality prediction model generation device, the quality prediction device, and the quality influencing factor estimation device for metallic materials. However, in addition to metallic materials, it can also be applied to materials and other manufactured products that have undergone multiple processes in a continuous manufacturing production line.
[0046] The quality prediction model generation device, quality prediction device, and quality influencing factor estimation device, for example, can be used through... Figure 1 The information processing device 1 shown is used to implement this. Specifically, the information processing device 1 is composed of a personal computer, a workstation, etc. In addition, the information processing device 1 uses, for example, a processor composed of a CPU (Central Processing Unit) and a memory (main storage unit) composed of RAM (Random Access Memory) and ROM (Read Only Memory) as its main components.
[0047] The information processing device 1 includes a manufacturing condition data collection unit 11, a quality data collection unit 12, a continuous process data editing unit 13, a continuous process database 14, a model generation unit 15, a model selection unit 16, a quality prediction unit 17, and an influence factor estimation unit 18. Furthermore, the metal material quality prediction model generation device of this embodiment is composed of all the elements of the information processing device 1 except for the quality prediction unit 17 and the influence factor estimation unit 18. Additionally, the metal material quality prediction device of this embodiment is composed of all the elements of the information processing device 1 except for the influence factor estimation unit 18. Furthermore, the quality influence factor estimation device of this embodiment is composed of all the elements of the information processing device 1.
[0048] A sensor (not shown) is connected to the manufacturing condition data collection unit 11. The manufacturing condition data collection unit 11 collects manufacturing condition data (hereinafter referred to as "manufacturing condition data") for each process through this sensor and outputs it to the continuous process data editing unit 13. Examples of "manufacturing conditions for metal materials" include, for example, the composition of the metal material in each process, the temperature of the metal material, the pressure applied to the metal material, the tension applied to the metal material, the thickness of the metal material before and after rolling, and the throughput speed of the metal material.
[0049] The manufacturing condition data collected by the manufacturing condition data collection unit 11 for each process includes not only the measured values of the manufacturing conditions obtained by sensors, but also the preset values of the manufacturing conditions. That is, depending on the process, there may be cases where sensors are not installed. Therefore, in such cases, the preset values are collected as manufacturing condition data instead of the actual values.
[0050] The manufacturing condition data collection unit 11 collects manufacturing condition data for the metal materials in each process according to a specified range for each metal material, determined based on the specific process. A "specified range," for example, in the case of a slab or steel plate, refers to a predetermined range (position) along the length of the metal material, depending on the specific process. Furthermore, the specified range for each process may vary from process to process, and may also be the same. Additionally, the specified range along the length of the metal material is determined, for example, based on the distance the metal material travels in each process corresponding to the transport direction.
[0051] A sensor (not shown) or a data input terminal (not shown) is connected to the quality data collection unit 12. The quality data collection unit 12 collects quality data (hereinafter referred to as "quality data") of the metal material manufactured through each process via the sensor or data input terminal, and outputs it to the continuous process data editing unit 13. Examples of the "quality of the metal material" mentioned above include, for example, the tensile strength of the metal material in the final process, the defect inclusion rate of the metal material in the final process (the number of defects exposed per unit area), etc.
[0052] The quality data of the metal materials collected by the quality data collection unit 12 includes not only the measured values of quality measured by sensors, but also the quality judgment values input from the data input terminal, which are evaluated based on certain benchmarks or visual inspection by inspectors based on the measured values to determine whether the material is qualified.
[0053] The quality data collection unit 12 evaluates and collects the quality of the metal materials in the final process of each process according to the specified range determined for each final process. In addition, the specified range here is synonymous with the "specified range" in the manufacturing condition data collection unit 11.
[0054] The integrated process data editing unit 13 edits the actual data input from the manufacturing condition data collection unit 11 and the quality data collection unit 12. The integrated process data editing unit 13 associates the manufacturing condition data of each process collected by the manufacturing condition data collection unit 11 with the quality data of the metal material collected by the quality data collection unit 12 according to each specified range in a specific process and stores them in the integrated process database 14.
[0055] The integrated process data editing department 13 edits the manufacturing condition data of metal materials in each process and the quality data of metal materials in the final process of each process, using the width of a specified range in a specific process as a benchmark, and then edits the manufacturing condition data and quality data of other processes. At this time, the integrated process data editing department 13 repeatedly links the manufacturing condition data and quality data according to the specified range that supplements the density differences of the manufacturing condition data and quality data in each process, that is, to each uniform specified range corresponding to the quantity of each process, and saves them in the integrated process database 14.
[0056] In addition, the consistent process data editing department 13 takes all processes as objects, selects a specific process, takes the specified range in the selected specific process as a unified specified range, and repeatedly links the manufacturing condition data and quality data according to one or more unified specified ranges, and saves them in the consistent process database 14.
[0057] As will be described later, when there is only one process for manufacturing metal materials, only one uniform specification range is set. In this case, the consistent process data editing unit 13 creates only one dataset (e.g., a learning dataset) for editing manufacturing condition data and quality data, and saves it in the consistent process database 14. On the other hand, when there are multiple processes for manufacturing metal materials, multiple uniform specification ranges are set. In this case, the consistent process data editing unit 13 creates multiple datasets (e.g., a learning dataset) for editing manufacturing condition data and quality data, and saves them in the consistent process database 14. Furthermore, the details of the processing performed by the consistent process data editing unit 13 will be described later (see [reference]). Figures 4-9 ).
[0058] The model generation unit 15 generates one or more quality prediction models for each uniform specification range based on the manufacturing condition data and quality data stored in the consistent process database 14. These quality prediction models predict the quality of the metal material for each uniform specification range.
[0059] As will be described later, when there is only one process for manufacturing metal materials, only one uniform specification range is set. In this case, the model generation unit 15 generates only one quality prediction model. On the other hand, when there are multiple processes for manufacturing metal materials, multiple uniform specification ranges are set. In this case, the model generation unit 15 generates multiple quality prediction models.
[0060] The model generation unit 15 uses, for example, gradient boosting trees as statistical analysis methods and machine learning methods when generating quality prediction models. In addition to the above, various methods such as linear regression, local regression, principal component regression, PLS (Partial Least Squares Regression), logistic regression, support vector machines, decision trees, regression trees, random forests, and neural networks can also be used as statistical analysis methods and machine learning methods.
[0061] When the model selection unit 16 generates multiple quality prediction models from the model generation unit 15, it selects one quality prediction model from the multiple quality prediction models. "When multiple quality prediction models are generated" means, as described above, that there are multiple processes for manufacturing metal materials and multiple uniformly defined ranges are set.
[0062] The model selection unit 16 selects, for example, a quality prediction model used in the prediction of the quality of metallic materials (processing in the quality prediction unit 17) from multiple quality prediction models based on the evaluation of specific indicators. In this case, the model selection unit 16 selects, for example, a quality prediction model used in the prediction of the quality of metallic materials from multiple quality prediction models based on an indicator that represents the prediction accuracy of the quality prediction model as a specific indicator.
[0063] Furthermore, the model selection unit 16 can also select the quality prediction model to be used in predicting the quality of metallic materials from multiple quality prediction models based on an importance index representing the impact of manufacturing conditions, which serve as inputs to each quality prediction model, on the prediction of the output quality. Details of the processing by the model selection unit 16 will be described later.
[0064] The quality prediction unit 17 uses a quality prediction model generated by the model generation unit 15 and selected by the model selection unit 16 as needed to predict the quality of metal materials manufactured under arbitrary manufacturing conditions within each specified range.
[0065] The influence factor estimation unit 18 uses a quality prediction model generated by the model generation unit 15 and selected by the model selection unit 16 as needed to estimate the manufacturing conditions that affect the quality of the metallic material. Details of the processing performed by the influence factor estimation unit 18 will be described later.
[0066] (Quality Prediction Methods)
[0067] Reference Figures 2-9 This embodiment describes the quality prediction method and the quality prediction model generation method. The quality prediction method of this embodiment is performed... Figure 2 The processing steps S11 to S17 shown are described. In addition, the quality prediction model generation method of this embodiment performs the following besides... Figure 2 The processing of steps S11 to S16 other than step S17 shown.
[0068] In addition, Figure 2 In this model, step S11 corresponds to the first collection step, step S12 corresponds to the second collection step, step S13 corresponds to the saving step, and step S14 corresponds to the model generation step. Additionally, in... Figure 2 In this process, steps S15 and S16 are equivalent to the model selection steps, and step S17 is equivalent to the prediction steps.
[0069] First, the manufacturing condition data collection unit 11 collects manufacturing condition data for each process according to a predetermined range of metal materials, which is determined in advance based on different processes (step S11). For example, Figure 3 As shown, the manufacturing condition data collected by the manufacturing condition data collection unit 11 consists of data in each process, which are arranged according to the specified range of the metal material determined according to the process, including the actual values or set values of multiple manufacturing conditions.
[0070] Figure 3 The manufacturing condition data shown has the position l in the length direction of each process. 1 l 2 ... and multiple manufacturing conditions x1 measured by sensors at this location. 1 x1 2 ... x2 1 x2 2 The project consists of... In addition, Figure 3 The manufacturing condition data shown represents, for example, manufacturing condition data collected in a single process. Therefore, for example, in the case where there are a total of three manufacturing processes for a metal material, Figure 3 The manufacturing condition data shown is also collected in three parts, depending on the number of processes. At this time, in the manufacturing condition data collected for each process, sometimes the position l in the length direction is included. 1 l 2 The terms 、, … (=the specified range) are different, but sometimes they are the same.
[0071] Next, the quality data collection unit 12 evaluates and collects the quality data of the metal material manufactured through each process according to each predetermined range (step S12). The quality data collected by the quality data collection unit 12 is data collected in the final process of multiple processes, which has the evaluation results of the quality corresponding to the position of the metal material from the front end to the rear end in the length direction of each process.
[0072] Next, the continuous process data editing unit 13 associates and saves the manufacturing condition data of each process with the quality data of the metal material manufactured under those manufacturing conditions (step S13). Based on the manufacturing condition data and quality data of the metal material's position along its length collected in steps S11 and S12, the continuous process data editing unit 13 edits the manufacturing condition data and quality data of other processes, using the width of a specified range in a specific process as a reference. Then, the continuous process data editing unit 13 aligns and associates each uniform specified range after interpolation to compensate for differences in the granularity of the specified ranges of the manufacturing condition data and quality data for each process.
[0073] Furthermore, "granularity of the specified range" refers, for example, to the length of the acquisition (collection) cycle for manufacturing condition data and quality data. When the acquisition cycle for manufacturing condition data and quality data is short, the granularity of the specified range becomes finer. Conversely, when the acquisition cycle for manufacturing condition data and quality data is long, the granularity of the specified range becomes coarser. Additionally, "uniform specified range" refers to the width of the specified range in a specific process within each process.
[0074] In this way, the continuous process data editing department 13 associates and saves the manufacturing condition data of each process with the quality data of the metal material manufactured under those manufacturing conditions, according to each uniformly defined range along the length of the metal material. (See below for reference.) Figures 4-9 An example of the processing performed by the consistent process data editing department 13 will be explained.
[0075] For example, Figures 4-6 As shown, consider the case of manufacturing a metallic material through processes 1, 2, and 3. Figures 4-6 In this context, the longitudinal position of the metal material is represented by aligning its relative positions to the front and rear ends. Processes 1, 2, and 3 each have different defined ranges. Additionally, in process 1, materials are collected based on manufacturing conditions x. a To represent the manufacturing condition data within a specified range, in process 2, data is collected based on manufacturing condition x. b To represent the manufacturing condition data within a specified range, in process 3, data is collected based on manufacturing condition x. cThis refers to manufacturing condition data within a specified range and quality data not shown in the figure.
[0076] Additionally, in processes 1, 2, and 3, for example... Figure 3 As shown, there are also cases where multiple manufacturing condition data are collected within each process, but the data acquisition cycle is basically the same for each. Therefore, the defined ranges for process 1, process 2, and process 3 are each the same. That is, one defined range is defined for each process.
[0077] As the first sub-step in the save process, such as Figure 4 As shown, the consistent process data editing unit 13 uses the width of the specified range of process 1 as a reference (as a unified specified range) to edit the manufacturing condition data and quality data in the specified ranges of process 2 and process 3. Thus, relative to the granularity of the specified range of process 1, the differences in granularity between the specified ranges of process 2 and process 3 are compensated by interpolation.
[0078] Specifically, the consistent process data editing unit 13 summarizes the manufacturing condition data for two specified ranges of process 2, which are equivalent in width to the specified range of process 1, and uses this data as representative data for the two specified ranges in process 2, linking it to each specified range of process 1. Similarly, the consistent process data editing unit 13 summarizes the manufacturing condition data and quality data for four specified ranges of process 3, which are equivalent in width to the specified range of process 1, and uses this data as representative data for the four specified ranges in process 3, creating a dataset linked to each specified range of process 1.
[0079] As the second sub-step in the save step, such as Figure 5 As shown, the consistent process data editing unit 13 uses the width of the specified range of process 2 as a reference (as a unified specified range) to edit the manufacturing condition data and quality data in the specified ranges of process 1 and process 3. As a result, the differences in the density of the specified ranges of process 1 and process 3 are compensated by interpolation relative to the granularity of the specified range of process 2.
[0080] Specifically, the consistent process data editing unit 13 uses the manufacturing condition data of the specified range of process 1, which is equivalent in width to the specified range of process 2, as representative data and associates them with each specified range of process 2. At this time, the representative data of process 1 becomes the same data in two adjacent specified ranges of process 2. In addition, the consistent process data editing unit 13 summarizes the manufacturing condition data and quality data of the two specified ranges of process 3, which are equivalent in width to the specified range of process 2, and uses them as representative data for the two specified ranges in process 3, creating a dataset associated with each specified range of process 2.
[0081] As the third sub-step in the save process, such as Figure 6 As shown, the consistent process data editing unit 13 edits the manufacturing condition data within the specified ranges of process 1 and process 2 based on the width of the specified range of process 3 (as a unified specified range). Thus, relative to the granularity of the specified range of process 3, the differences in granularity between the specified ranges of process 1 and process 2 are compensated by interpolation.
[0082] Specifically, the consistent process data editing unit 13 uses the manufacturing condition data of the specified range of process 1, which is equivalent in width to the specified range of process 3, as representative data, and creates a dataset associated with each specified range of process 3. At this time, the representative data of process 1 becomes identical data in the four adjacent specified ranges of process 3. Furthermore, the consistent process data editing unit 13 uses the actual manufacturing condition data of the specified range of process 2, which is equivalent in width to the specified range of process 3, as representative data, and associates it with each specified range of process 3. At this time, the representative data of process 2 becomes identical data in the two adjacent specified ranges of process 3.
[0083] Furthermore, each sub-step is actually repeated K times, corresponding to the number K processes. That is, the continuous process data editing unit 13 creates K datasets corresponding to the number K processes. Additionally, when editing the actual data in the continuous process data editing unit 13, generally known statistical measures such as average, maximum, and minimum values can be used in the decision based on the summarized representative data. Alternatively, actual data selected from multiple specified ranges of actual data based on arbitrary conditions can also be used. (The following is a return to the previous section.) Figure 2 Let's continue with the explanation.
[0084] The model generation unit 15 generates one or more quality prediction models (step S14) based on the manufacturing condition data (dataset) for each uniformly defined range in each process, according to the granularity of the uniformly defined range that associates the manufacturing condition data with the quality data. When there is only one process for manufacturing the metal material and only one uniformly defined range is set, the model generation unit 15 generates only one quality prediction model. On the other hand, when there are multiple processes for manufacturing the metal material and multiple uniformly defined ranges are set, the model generation unit 15 generates multiple quality prediction models corresponding to the number of datasets produced.
[0085] Next, the model selection unit 16 determines whether multiple quality prediction models were generated in the model generation step (step S15). If multiple quality prediction models were generated in the model generation step (yes in step S15), the process proceeds to step S16. On the other hand, if multiple quality prediction models were not generated in the model generation step (no in step S15), the process proceeds to step S17.
[0086] Next, the model selection unit 16 selects the quality prediction model to be used from multiple quality prediction models based on the evaluation of indicators representing the prediction accuracy of the quality prediction model (step S16). The model selection unit 16 evaluates the indicators representing the prediction accuracy of the quality prediction model, for example, by leave-out validation or cross-validation, and selects a quality prediction model with granularity based on a uniformly defined range of minimum or maximum values represented by the indicator, depending on the indicator used.
[0087] When the quality prediction model is a regression model, metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²) can be used as the aforementioned metrics. Conversely, when the quality prediction model is a classification model, metrics such as Accuracy, Precision, Recall, and F-score can be used as the aforementioned metrics.
[0088] Next, the quality prediction unit 17 uses the quality prediction model generated by the model generation unit 15 and selected by the model selection unit 16 as needed to predict the quality of the metal material manufactured under arbitrary manufacturing conditions within each specified range (step S17).
[0089] Furthermore, in this embodiment, after the quality prediction model is selected in step S16, quality prediction is performed in step S17. However, it is not necessary to perform all the processes up to step S16 every time step S17 is performed. For example, steps S11 to S16 can be performed in advance to select the quality prediction model. Then, whenever step S17 is performed using the selected quality prediction model, only steps S11, S13, and S17 are performed.
[0090] (Method for estimating quality impact factors)
[0091] Reference Figure 10 This embodiment describes the method for estimating the quality influencing factors. Furthermore, Figure 10 The processing of steps S21 to S25 shown is... Figure 2 The processes in steps S11 to S15 are the same, so the following explanation is omitted.
[0092] In addition, Figure 10 In this model, step S21 corresponds to the first collection step, step S22 corresponds to the second collection step, step S23 corresponds to the saving step, and step S24 corresponds to the model generation step. Additionally, in... Figure 10In this process, steps S25 and S26 are equivalent to the model selection steps, and step S27 is equivalent to the quality influence factor estimation steps.
[0093] The model selection unit 16 selects the quality prediction model to be used from multiple quality prediction models based on an evaluation of an index representing the prediction accuracy of the quality prediction model (step S26). The model selection unit 16 selects the quality prediction model to be used in predicting the quality of the metallic material based on an importance index, which represents the influence of the manufacturing condition data input to the multiple quality prediction models generated by the model generation unit 15 on the prediction of the output quality data. Furthermore, there are roughly two methods for model selection performed by the model selection unit 16, which will be described separately below.
[0094] <First Model Selection Method>
[0095] First, the model selection unit 16 calculates an importance index (variable importance) for each manufacturing condition based on the actual data generated by linking the specified range of the final process where quality data is collected in each process with a unified specified range. Then, the model selection unit 16 extracts the manufacturing condition with the highest importance index.
[0096] Importance metrics are indices that correlate with the overall influence of factors within the prediction model. For example, algorithms specific to the algorithm that created the model, such as importance based on Gini impurity in decision tree algorithms like random forests, can be used. Alternatively, well-known methods, such as "permutation importance," can be used to calculate importance metrics universally, regardless of the algorithm used to create the model.
[0097] Next, the model selection unit 16 selects a quality prediction model generated based on actual data that is linked to the specified range of the manufacturing conditions with the highest importance index as a unified specified range, and uses this model as the quality prediction model for quality prediction.
[0098] For example, as mentioned above Figures 4-6 As shown, this considers the case where metal materials are manufactured through processes 1, 2, and 3, i.e., the case where three quality prediction models are generated based on three actual data points. In this case, the actual data (refer to...) are linked based on the specified range of process 3, which is the final process, as a unified specified range. Figure 6 The generated quality prediction model calculates an importance index for each manufacturing condition. Then, for example, in manufacturing condition x... a When the importance index is at its maximum, select the manufacturing condition x. aThe actual data related to the specified scope of process 1 as a unified specified scope (refer to) Figure 4 The generated quality prediction model is used as the quality prediction model in quality prediction.
[0099] <Second Model Selection Method>
[0100] First, the model selection unit 16 calculates a first importance index (variable importance) for each manufacturing condition based on the first quality prediction model generated by linking the specified range of the final process where quality data collected in each process is used as a unified specified range. Then, the model selection unit 16 extracts the manufacturing condition with the highest importance index.
[0101] Similar to the method for selecting the first model, the importance metric is an indicator that calculates and predicts the influence of each factor in the overall model. For example, an algorithm specific to the algorithm used to build the model can be used, such as the importance metric based on Gini impurity in decision tree algorithms, such as random forests. Alternatively, a well-known method, such as "permutation importance," can be used to calculate the importance metric generally, regardless of the algorithm used to build the model.
[0102] Next, the model selection unit 16 calculates the second importance index for each manufacturing condition using the second quality prediction model generated based on actual data that is linked to the specified range of the process to which the manufacturing condition with the highest first importance index belongs as a unified specified range. Then, the model selection unit 16 extracts the manufacturing condition with the highest second importance index.
[0103] At this time, if the process to which the manufacturing condition in which the second importance index reaches its maximum value belongs is the same as the process to which the manufacturing condition in which the first importance index reaches its maximum value belongs, the model selection unit 16 selects the second quality prediction model as the model to be used.
[0104] On the other hand, when the process to which the manufacturing condition with the highest second importance index belongs is inconsistent with the process to which the manufacturing condition with the highest first importance index belongs, the model selection unit 16 extracts a third importance index for each manufacturing condition, based on the third quality prediction model generated using actual data that is linked according to a defined range of the process to which the manufacturing condition with the highest second importance index belongs as a unified defined range. Then, the model selection unit 16 extracts the third importance index as the manufacturing condition with the highest importance.
[0105] Then, the model selection unit 16 extracts the kth importance index for each manufacturing condition of the kth quality prediction model (k≥1), and repeats the above process until the process to which the manufacturing condition with the highest kth importance index belongs is the same as the process to which the manufacturing condition with the highest (k-1)th importance index belongs is the same. If the process to which the manufacturing condition belongs is the same in the Nth repetition, the Nth quality prediction model is selected as the model to be used. Alternatively, an arbitrary upper limit can be set for the number of repetitions.
[0106] For example, as mentioned above Figures 4-6 As shown, this considers the case where metal materials are manufactured through processes 1, 2, and 3, i.e., the case where three quality prediction models are generated based on three actual data points. In this case, the actual data (refer to...) are linked based on the specified range of process 3, which is the final process, as a unified specified range. Figure 6 The generated first quality prediction model calculates the first importance index for each manufacturing condition. Then, for example, in manufacturing condition x... a When the importance index is at its maximum, for the manufacturing condition x a The actual data related to the specified scope of process 1 as a unified specified scope (refer to) Figure 4 The generated second quality prediction model calculates a second importance index for each manufacturing condition. Then, for example, in manufacturing condition x... b When the importance index is at its maximum, for the manufacturing condition x b The actual data related to the specified scope of process 2 as a unified specified scope (refer to) Figure 5 The generated third quality prediction model calculates the third importance index for each manufacturing condition. Then, for example, in manufacturing condition x... b When the importance index is at its maximum, the manufacturing process to which the third importance index is at its maximum corresponds to the manufacturing process to which the second importance index is at its maximum. Therefore, the third quality prediction model is selected as the quality prediction model to be used in quality prediction. (The following is a return to the previous section.) Figure 10 Let's continue with the explanation.
[0107] Next, the impact factor estimation unit 18 uses the quality prediction model selected by the model selection unit 16 to estimate the manufacturing conditions that are factors affecting the quality of the metallic material (step S27). For example, based on the quality prediction model, the impact factor estimation unit 18 calculates the importance of the variables in the quality prediction model for each manufacturing condition, and extracts an arbitrary number of variable importance values in descending order, starting from the manufacturing conditions with the largest variable importance values. Then, the impact factor estimation unit 18 displays the extracted manufacturing conditions as a list of items whose names are estimated as impact factors, or displays them as a descending bar chart or other graphical representation.
[0108] (Manufacturing methods for metallic materials)
[0109] The method for manufacturing metallic materials in this embodiment includes a quality prediction step, a manufacturing condition determination step, and a metallic material manufacturing step.
[0110] In the quality prediction step, the influencing factor estimation unit 18 uses the quality prediction model selected by the model selection unit 16, which includes manufacturing conditions in the input that are estimated to affect the quality of the metal material, to predict the quality of the output.
[0111] In the manufacturing condition determination step, the influencing factor estimation unit 18 can also determine the management range of manufacturing conditions for factors estimated to affect the quality of the metallic material in a way that makes the predicted quality fall within a predetermined range. The management range of manufacturing conditions can be determined, for example, by mathematical programming methods such as the branch-limit method.
[0112] In the metal material manufacturing process, metal materials are manufactured using manufacturing conditions adjusted based on factors presumed to affect the quality of the metal materials, within a determined management scope.
[0113] The information processing device 1 may also include an output unit that outputs a management range of manufacturing conditions, determined by the influencing factor estimation unit 18, which is estimated to affect the quality of the metal material, as an output signal. When the output signal's destination is a control device for the equipment, the output signal may also be a control signal for the equipment that sets the manufacturing conditions. Furthermore, when the output signal's destination is an operator's terminal, the output signal may also be operator guidance information displayed on a screen.
[0114] (Example)
[0115] An embodiment of the method for predicting the quality of metallic materials according to the embodiments will be described. In this embodiment, the method for predicting the quality of metallic materials according to the embodiments is applied to the prediction of the number of surface defects in surface-treated steel sheets (molten galvanized steel sheets).
[0116] The specified particle sizes for each of the casting, hot rolling / rough rolling, hot rolling / finish rolling, cold rolling, and surface treatment processes are 0.2m spacing, 0.2m spacing, 0.5m spacing, 1m spacing, and 1m spacing, respectively.
[0117] In this invention example, a quality prediction model is created based on a dataset of actual data from each process, linked together by a defined range (1m spacing) for the surface treatment process, which is the final process. This defined range is then used as a uniform defined range. The model outputs the number of surface defects within each uniform defined range. Next, for the created quality prediction model, an importance index (variable importance) for each manufacturing condition is calculated.
[0118] Next, casting speed, as the most important manufacturing condition, was extracted from the casting process. Then, the specified range (0.2m interval) of the casting process to which the corresponding manufacturing condition (casting speed) belonged was used as a unified specified range, and a quality prediction model generated from the dataset of manufacturing condition data for each related process was selected. If the coil length of the molten galvanized steel sheet is 2000m and the slab length after the casting process is 8m, then with the specified range (0.2m interval) of the casting process set as a unified specified range, the corresponding coil length of the molten galvanized steel sheet is 50m.
[0119] In the comparative example, a quality prediction model was selected from the dataset of actual data of each process that is linked together with the specified range (1m spacing) of the surface treatment process as the final process as a uniform specified range. The number of surface defects in each uniform specified range was selected as the output.
[0120] Table 1 shows the comparison results of the reproduction rates of defect-free and defective examples. As shown in Table 1, it can be seen that, compared with the comparative example, although the reproduction rate of defect-free examples is slightly lower, the reproduction rate of defective examples is significantly improved.
[0121] [Table 1]
[0122] The metal material quality prediction model generation method, metal material quality prediction method, metal material quality influencing factor estimation method, metal material manufacturing method, quality prediction model generation device, metal material quality prediction device, and metal material quality influencing factor estimation device described above, by appropriately utilizing the granularity of manufacturing condition data from each process of manufacturing the metal material, can reflect the maximum amount of information related to the quality of the object in the generated quality prediction model. Therefore, it is possible to predict quality with high accuracy and to estimate factors causing quality defects with high accuracy.
[0123] Furthermore, according to the embodiments of the metal material quality prediction model generation method, metal material quality prediction method, metal material quality influencing factor estimation method, metal material manufacturing method, quality prediction model generation method, metal material quality prediction model generation apparatus, metal material quality prediction apparatus, and metal material quality influencing factor estimation apparatus, one or more quality prediction models are generated that associate the manufacturing conditions of each process with the quality of the metal material manufactured under those manufacturing conditions according to each specified range of each process, and the quality prediction model to be used is selected based on specific indicators. Therefore, the quality of the metal material under any manufacturing conditions can be predicted with higher accuracy than before, and the manufacturing conditions that influence the quality of the metal material can be estimated with high accuracy.
[0124] The foregoing has provided a detailed description of the method for generating a quality prediction model for metallic materials, a method for predicting the quality of metallic materials, a method for estimating the quality influencing factors of metallic materials, a method for manufacturing metallic materials, a method for generating a quality prediction model for metallic materials, an apparatus for generating a quality prediction model for metallic materials, an apparatus for predicting the quality of metallic materials, and an apparatus for estimating the quality influencing factors of metallic materials, through methods and embodiments for carrying out the invention. However, the scope of the invention is not limited to these descriptions and must be broadly interpreted based on the claims. Furthermore, various modifications and alterations based on these descriptions are also included within the scope of the invention, which is self-evident.
[0125] Explanation of reference numerals in the attached figures
[0126] 1. Information processing device
[0127] 11 Manufacturing Conditions Data Collection Department
[0128] 12 Quality Data Collection Department
[0129] 13. Integrated Process Data Editing Department
[0130] 14. Integrated Process Database
[0131] 15. Model Generation Department
[0132] 16 Model Selection Section
[0133] 17 Quality Forecasting Department
[0134] 18. Impact Factor Estimation Section
Claims
1. A method for generating a quality prediction model for metallic materials, wherein, The method for generating the quality prediction model for the metallic material includes: The first collection step involves collecting the manufacturing conditions of the metal materials in each process according to each specified range of the metal materials determined based on the different processes. The second collection step involves evaluating and collecting the quality of the metal materials in the final stage of each process according to a specified range determined for each final stage. In the saving step, for the manufacturing conditions of the metal materials in each process collected in the first collection step and the quality of the metal materials in the final process collected in the second collection step, the manufacturing conditions and quality in other processes are edited based on the width of a specified range in a specific process. The manufacturing conditions and quality are repeatedly associated with and saved according to each uniform specified range corresponding to the number of processes, which supplements the density differences of the manufacturing conditions and quality in each process. The model generation step, based on the manufacturing conditions for each of the uniformly specified ranges saved in the saving step, generates one or more quality prediction models for each of the uniformly specified ranges to predict the quality of each of the uniformly specified ranges; and In the model selection step, if multiple quality prediction models are generated in the model generation step, the quality prediction model to be used in predicting the quality of the metallic material is selected from the multiple quality prediction models based on the evaluation of specific indicators.
2. The method for generating a quality prediction model for metallic materials according to claim 1, wherein, In the saving step, all processes of each process are taken as objects, a specific process is selected individually, the specified range of the selected specific process is taken as the unified specified range, and the manufacturing conditions and the quality are repeatedly associated and saved according to each range of one or more of the unified specified ranges.
3. The method for generating a quality prediction model for metallic materials according to claim 1, wherein, In the model selection step, the quality prediction model used for predicting the quality of the metallic material is selected from a plurality of quality prediction models as a specific indicator based on an indicator representing the prediction accuracy of the quality prediction model.
4. The method for generating a quality prediction model for metallic materials according to claim 1, wherein, In the model selection step, the quality prediction model used for predicting the quality of the metallic material is selected from a plurality of quality prediction models based on an importance index, wherein the importance index represents the influence of the manufacturing conditions that are input to each quality prediction model on the predicted quality that becomes the output.
5. The method for generating a quality prediction model for metallic materials according to claim 4, wherein, In the model selection step, A quality prediction model is generated based on actual data that associates the specified range of the final process with the unified specified range. For the generated quality prediction model, an importance index is calculated for each manufacturing condition, and the manufacturing condition with the highest importance index is extracted. A quality prediction model generated based on actual data associated with the specified range of the process to which the extracted manufacturing conditions belong is selected as the quality prediction model used for predicting the quality of the metal material.
6. The method for generating a quality prediction model for metallic materials according to claim 4, wherein, In the model selection step, A first quality prediction model is generated based on actual data that associates the specified range of the final process with the unified specified range. For the first quality prediction model, a first importance index is calculated for each manufacturing condition, and the first importance index is extracted as the manufacturing condition with the highest importance. A second quality prediction model is generated based on actual data that is associated with the specified range of the process to which the extracted manufacturing conditions belong, using the unified specified range as the reference. For the second quality prediction model, a second importance index is calculated for each manufacturing condition, and this second importance index is extracted as the manufacturing condition with the highest importance. The second quality prediction model, which is selected when the manufacturing condition under which the second importance index is maximized coincides with the manufacturing condition under which the first importance index is maximized, is the quality prediction model used for predicting the quality of the metal material.
7. The method for generating a quality prediction model for metallic materials according to any one of claims 1 to 6, wherein, The model generation step uses statistical analysis methods and machine learning methods, including linear regression, local regression, principal component regression, PLS regression, logistic regression, support vector machine, decision tree, regression tree, random forest, gradient boosting tree, and neural network, to generate the quality prediction model.
8. A method for predicting the quality of metallic materials, wherein, The method for predicting the quality of metallic materials includes a quality prediction step, wherein the quality prediction step uses a quality prediction model generated by the method for generating a quality prediction model for metallic materials according to any one of claims 1 to 7 to predict the quality of metallic materials manufactured under arbitrary manufacturing conditions within each specified range.
9. A method for estimating the quality influencing factors of a metallic material, wherein, The method for estimating the quality influencing factors of the metallic material includes an influencing factor estimation step, wherein the influencing factor estimation step uses a quality prediction model generated by the method for generating a quality prediction model of metallic materials according to any one of claims 1 to 7 to estimate the manufacturing conditions that affect the quality of the metallic material.
10. The method for estimating the quality influencing factors of metallic materials according to claim 9, wherein, In the influence factor estimation step, the importance index of the quality prediction model is calculated for each manufacturing condition, and the manufacturing conditions with high importance indices are estimated to be manufacturing conditions that have an impact on the quality of the metal material.
11. A method for manufacturing a metallic material, wherein, The method for manufacturing the metallic material includes: The quality prediction step uses a quality prediction model to predict the quality of the output, wherein the quality prediction model includes manufacturing conditions that affect the quality of the metal material, which are estimated by the method for estimating the quality influence factors of the metal material as described in claim 9. The manufacturing conditions determination step determines the manufacturing conditions that influence the quality of the metal material in a way that ensures the predicted quality falls within a predetermined range; and The manufacturing process of metal materials involves producing metal materials according to the determined manufacturing conditions.
12. A method for generating a quality prediction model, wherein, The method for generating the quality prediction model includes: The first collection step involves collecting the manufacturing conditions of the materials in each process according to each specified range of the materials, as determined by the different processes. The second collection step involves evaluating and collecting the quality of the materials in the final stage of each process according to each specified range determined for the final stage. In the saving step, for the manufacturing conditions of the materials in each process collected in the first collection step and the quality of the materials in the final process collected in the second collection step, the manufacturing conditions and quality in other processes are edited based on the width of a specified range in a specific process. The manufacturing conditions and quality are repeatedly associated with and saved according to each uniform specified range corresponding to the number of processes, which supplements the density differences of the manufacturing conditions and quality in each process. The model generation step, based on the manufacturing conditions for each of the uniformly specified ranges saved in the saving step, generates one or more quality prediction models for each of the uniformly specified ranges to predict the quality of each of the uniformly specified ranges; and In the model selection step, if multiple quality prediction models are generated in the model generation step, the quality prediction model to be used in predicting the quality of the material is selected from the multiple quality prediction models based on the evaluation of specific indicators.
13. A device for generating a quality prediction model for metallic materials, wherein, The device for generating a quality prediction model for metallic materials includes: The first collection unit collects the manufacturing conditions of the metal materials in each process according to each specified range of the metal materials determined according to different processes; The second collection unit evaluates and collects the quality of the metal materials in the final stage of each process according to each specified range determined for the final stage; The storage unit, for the manufacturing conditions of the metal materials in each process collected by the first collection unit and the quality of the metal materials in the final process collected by the second collection unit, edits the manufacturing conditions and quality in other processes based on the width of a specified range in a specific process, and repeatedly associates and saves the manufacturing conditions and quality according to a unified specified range corresponding to the quantity of each process, which supplements the density differences of the manufacturing conditions and quality in each process. The model generation unit generates one or more quality prediction models for each uniform specification range based on the manufacturing conditions for each uniform specification range stored by the storage unit. as well as The model selection unit, when multiple quality prediction models are generated by the model generation unit, selects the quality prediction model to be used in predicting the quality of the metallic material from the multiple quality prediction models based on the evaluation of specific indicators.
14. A device for predicting the quality of metallic materials, wherein, The metal material quality prediction device includes a quality prediction unit that uses a quality prediction model generated by the metal material quality prediction model generation device according to claim 13 to predict the quality of the metal material manufactured under arbitrary manufacturing conditions within each specified range.
15. A device for estimating the quality influencing factors of metallic materials, wherein, The device for estimating the quality influencing factors of the metal material includes an influencing factor estimation unit, which uses a quality prediction model generated by the metal material quality prediction model generation device according to claim 13 to estimate the manufacturing conditions that affect the quality of the metal material.
