CN-122003686-A - Quality prediction model generation method, quality prediction method for metal material, production condition presentation method for metal material, quality prediction model generation device, quality prediction device for metal material, production condition presentation device for metal material, and production system for metal material

CN122003686ACN 122003686 ACN122003686 ACN 122003686ACN-122003686-A

Abstract

The quality prediction model generation method includes an acquisition step (S1) of acquiring a specification variable selected from manufacturing conditions of each step and a target variable that is a state of a quality defect of a manufactured metal material, a storage step (S3) of storing the specification variable and the target variable in association with each other as learning data, a calculation step (S4-S6) of classifying the learning data into groups and checking whether or not there is a significant difference in the state of the quality defect, a search step (S7) of searching for a group having the highest significance, and a generation step (S8) of generating a quality prediction model by machine learning in the group having the highest significance according to the searched group.

Inventors

KIMURA YUKI

Assignees

杰富意钢铁株式会社

Dates

Publication Date: 20260508
Application Date: 20240809
Priority Date: 20231127

Claims (15)

1. A quality prediction model generation method for generating a quality prediction model of a metal material manufactured through one or more steps, comprising: An acquisition step of acquiring a specification variable selected from the manufacturing conditions of each step and a target variable that is a state of a quality defect of the manufactured metal material; a storage step of storing the explanatory variable and the target variable as learning data in association with each other; A calculation step of dividing the learning data into groups and checking whether or not there is a significant difference in the state of the quality defect; searching for the most significant group, and And a generation step of generating the quality prediction model by machine learning in the group of the group having the highest significance in accordance with the searched group.
2. The method for generating a quality prediction model according to claim 1, wherein, The explanatory variable is obtained as a numerical variable or a category variable.
3. The method for generating a quality prediction model according to claim 2, wherein, When the explanatory variable is acquired as a category variable, the calculating step performs numerical conversion in which the category variable is assigned different binary values to be classified, and calculates an index for evaluating the significant difference based on the assigned values and the state of the quality defect.
4. The method for generating a quality prediction model according to claim 3, wherein, The calculating step calculates the index for each of a plurality of packets.
5. The method for generating a quality prediction model according to any one of claims 1 to 4, characterized in that, After the searching step is performed, the calculating step further performs grouping and the checking on one of the groups grouped according to the searched highest significance.
6. The method for generating a quality prediction model according to any one of claims 1 to 5, characterized in that, The machine learning method at least comprises linear regression, local regression, principal component regression, PLS regression, neural networks, regression trees, random forests, lightGBM and XGBoost.
7. The method for generating a quality prediction model according to any one of claims 1 to 6, characterized in that, The quality prediction model is regenerated when the prediction accuracy of the quality prediction model deviates from a predetermined range or at predetermined intervals.
8. A quality prediction method for metal materials is characterized in that, The state of a quality defect of the metal material is predicted using the quality prediction model generated by the quality prediction model generation method according to any one of claims 1 to 7.
9. A method for producing a metal material, characterized in that, The method for predicting quality defects of a metal material according to claim 8, And changing at least a part of the manufacturing conditions when it is predicted that the quality defect of the metal material exists.
10. A method for prompting the manufacturing conditions of a metal material is characterized in that, The method for predicting quality defects of a metal material according to claim 8, wherein when the quality defects of the metal material are predicted to exist, the method searches for and presents a change in at least a part of the production conditions such as elimination of the quality defects.
11. A method for producing a metal material, characterized in that, A metal material is manufactured based on the change of the manufacturing conditions presented by the manufacturing condition presenting method of the metal material as set forth in claim 10.
12. A quality prediction model generation device for generating a quality prediction model of a metal material manufactured through one or more steps, comprising: An acquisition unit that acquires a target variable that is a state of a quality defect of the metal material to be manufactured and a description variable selected from manufacturing conditions of each step; a storage unit that stores the explanatory variable and the target variable in association with each other as learning data; A calculation unit configured to divide the learning data into groups and to perform a test as to whether or not there is a significant difference in the state of the quality defect; A search unit for searching for a packet having the highest significance, and The generation unit generates the quality prediction model by machine learning in the group of the group having the highest significance in accordance with the search.
13. A quality prediction device for metal materials is characterized in that, The state of the quality defect of the metal material is predicted using the quality prediction model generated by the quality prediction model generating device of claim 12.
14. A metal material manufacturing condition prompting device is characterized in that, Based on the presence or absence of the quality defect of the metal material predicted by the quality prediction device for metal material according to claim 13, a change of at least a part of the production conditions of the metal material is presented.
15. A system for manufacturing a metal material, characterized in that, The manufacturing apparatus for controlling each step of manufacturing a metal material according to the change of the manufacturing condition outputted from the output unit of the manufacturing condition presenting apparatus for a metal material according to claim 14.

Description

Quality prediction model generation method, quality prediction method for metal material, production condition presentation method for metal material, quality prediction model generation device, quality prediction device for metal material, production condition presentation device for metal material, and production system for metal material Technical Field The present disclosure relates to a quality prediction model generation method, a quality prediction method for a metal material, a manufacturing condition presentation method for a metal material, a quality prediction model generation device, a quality prediction device for a metal material, a manufacturing condition presentation device for a metal material, and a manufacturing system for a metal material. Background As a method for predicting quality for an arbitrary requirement, for example, a method is known in which distances between a plurality of past observation conditions stored in a performance database and a desired requirement are calculated, and quality is predicted using the calculated distances. For example, patent document 1 discloses a method of calculating a weight of observation data (actual result data) from a calculated distance, creating a function for fitting the vicinity of a request condition from the calculated weight, and predicting the quality for the request condition using the created function. Patent document 1 Japanese patent No. 7207547 The quality prediction method of patent document 1 generates a quality prediction model that correlates the manufacturing conditions of each step and the quality of the metal material manufactured under the manufacturing conditions to store actual result data for each predetermined range, and predicts the quality of each predetermined range of the metal material. The predetermined range is determined in consideration of, for example, a cutting position of the metal material. Here, although the quality prediction method of patent document 1 can accurately predict quality for any manufacturing condition, it is required to further improve prediction accuracy. For example, it is expected to further improve the prediction accuracy by objectively dividing data so as to correspond to a change in quality of a metal material and generating a quality prediction model in each division. Disclosure of Invention An object of the present disclosure, which has been made in view of the above circumstances, is to provide a quality prediction model generation method, a quality prediction method for a metal material, a manufacturing method for a metal material, a quality prediction model generation device, and a quality prediction device for a metal material, which can accurately predict the quality of a metal material. (1) A quality prediction model generation method according to an embodiment of the present disclosure is a quality prediction model generation method for generating a quality prediction model of a metal material manufactured through one or more steps, and includes: An acquisition step of acquiring a specification variable selected from the manufacturing conditions of each step and a target variable, which is a state of a quality defect of the manufactured metal material; A storage step of storing the explanatory variable and the target variable as learning data in association with each other; a calculation step of dividing the learning data into groups and checking whether or not there is a significant difference in the state of the quality defect; searching for the most significant group, and And a generation step of generating the quality prediction model by machine learning in the group of the group having the highest significance searched for. (2) As one embodiment of the present disclosure, on the basis of (1), The above explanatory variables are obtained as numerical variables or category variables. (3) As one embodiment of the present disclosure, on the basis of (2), When the explanatory variable is obtained as a category variable, the calculating step performs numerical conversion in which the category variable is classified by assigning different binary values, and calculates an index for evaluating the significant difference based on the assigned values and the state of the quality defect. (4) As one embodiment of the present disclosure, on the basis of (3), The calculating step calculates the index for each of the plurality of packets. (5) As one embodiment of the present disclosure, based on any one of (1) to (4), After the searching step is performed, the calculating step further performs grouping and the checking on one of the groups grouped according to the searched group having the highest significance. (6) As one embodiment of the present disclosure, based on any one of (1) to (5), The above machine learning method at least includes linear regression, local regression, principal component regression, PLS regression, neural networks, regression trees, random forests, lightGBM, a