CN-122003602-A - Method for evaluating steel, program, and method for generating model

Abstract

The present invention provides a method for evaluating steel, etc., which can evaluate steel satisfying specifications related to composition in more detail. A method for evaluating steel includes analyzing a composition of steel, inputting the composition data obtained by the analysis into a model machine-learned to output the same data as the input data when the composition data satisfies a predetermined specification, and calculating an error loss between the output data output from the model and the composition data, and determining whether the error loss exceeds a predetermined reference value.

Inventors

• FUKUMOTO SHIHO

Assignees

• 株式会社博迈立铖

Dates

Publication Date

20260508

Application Date

20240924

Priority Date

20230926

Claims (6)

1. 1. A method for evaluating steel, characterized in that, Analyzing the composition of the steel; Inputting the composition data obtained by the analysis into a model machine-learned to output the same data as the input data when the composition data satisfies a predetermined specification, and calculating an error loss between the output data output from the model and the composition data, and It is determined whether the error loss exceeds a predetermined reference value.
2. 2.A method for evaluating steel according to claim 1, wherein, And when the error loss exceeds a preset reference value, additionally checking the steel.
3. 3. A method for evaluating steel according to claim 1, wherein, The model is an automatic encoder model comprising an encoder and a decoder.
4. 4. A method for evaluating steel according to any one of claim 1 to 3, The model is machine-learned using composition data of steel manufactured after the change manufacturing process as training data, The reference value is set to a value that is larger than a maximum value of error loss between the composition data and output data that are input into the model after the end of machine learning and output from the model, and is smaller than a maximum value of error loss between the composition data and output data that are input into the model after the end of machine learning and output from the model, the composition data of steel manufactured before the manufacturing process is changed.
5. 5. A program for causing a computer to execute: Acquiring composition data of steel; Inputting the composition data into a model machine-learned to output the same data as the input data when the composition data satisfies a predetermined specification, and calculating an error loss between the output data output from the model and the composition data, and It is determined whether the error loss exceeds a predetermined reference value.
6. 6. A method for generating a model is characterized in that, Acquiring training data recording a plurality of sets of composition data representing the content of each component of the steel, and Based on the acquired training data, a model is generated by machine learning, and when the composition data is input, the model outputs the same data as the input composition data.

Description

Method for evaluating steel, program, and method for generating model Technical Field The present invention relates to a method for evaluating steel, a program, and a method for generating a model. Background Steel is an alloy in which small amounts of carbon, silicon, manganese, phosphorus, and other additives are added to iron. The properties of steel vary greatly depending on the proportions, i.e., compositions, of the individual components, and thus various steels, such as carbon steel, tool steel, spring steel, etc., are manufactured and used. For each steel, specifications relating to composition are specified (patent document 1). Prior art literature Patent literature Patent document 1 Japanese patent laid-open No. 2006-152356 Disclosure of Invention Problems to be solved by the invention In the above specification, the ratio of each component is determined in terms of width. In manufacturing steel, the content of each component is adjusted so as to be within a determined width, that is, to satisfy specifications concerning the component. Various shipment checks including composition analysis are performed for the manufactured steel, and the composition check is used to confirm that the specifications related to the composition are actually satisfied. The steel that is checked for pass will be shipped. Experience has shown, however, that when comparing batches of steel manufactured to meet specifications associated with the same composition, there are sometimes cases where certain steel properties are superior to other batches of steel. By detecting the occurrence of such excellent steel characteristics and reflecting the same to the management of the manufacturing process, it is expected to improve the quality of the steel. In one aspect, an object is to provide a steel evaluation method and the like capable of evaluating steel satisfying specifications relating to composition in more detail. Solution for solving the problem The method includes analyzing a composition of steel, inputting the composition data obtained by the analysis into a model machine-learned to output the same data as the input data when the composition data satisfies a predetermined specification, calculating an error loss between the output data output from the model and the composition data, and determining whether the error loss exceeds a predetermined reference value. Effects of the invention In one aspect, a method of evaluating steel or the like capable of evaluating steel satisfying specifications relating to composition in more detail can be provided. Drawings Fig. 1 is an explanatory diagram illustrating determination and application processes of an evaluation method of steel. Fig. 2 is an explanatory diagram illustrating an application stage of the steel evaluation method. FIG. 3 is a table showing the composition of samples used in the preliminary study stage. Fig. 4 is a graph showing the heat treatment dimensional change rate of steels having different molybdenum contents in parallel. Fig. 5 is a graph showing the heat treatment dimensional change rate of steels having different chromium contents in parallel. FIG. 6 is a graph showing the heat treatment dimensional change rate of steels having different vanadium contents in parallel. Fig. 7 is a graph showing the heat treatment dimensional change rate of steels having different silicon contents in parallel. Fig. 8 is a graph showing the heat treatment dimensional change rate of steels having different manganese contents in parallel. Fig. 9 is an explanatory diagram illustrating the structure of an information processing apparatus for generating a model. Fig. 10 is an explanatory diagram illustrating the model. Fig. 11 is a flowchart illustrating a processing flow of a program for generating a model. Fig. 12 is a graph showing the variation of the average absolute error loss generated at the time of manufacture. Fig. 13 is an explanatory diagram illustrating the structure of the steel evaluation system. Fig. 14 is a flowchart illustrating a processing flow of the program in the application phase. Detailed Description Embodiment 1 Fig. 1 is an explanatory diagram illustrating determination and application processes of an evaluation method of steel. In the present embodiment, description will be made taking as an example a case where an evaluation method for fine-tuning a composition for steel that has been mass-produced so as to stably reproduce excellent characteristics and stably manufacturing the fine-tuned steel is determined. A summary of the process of determining the evaluation method of steel will be described with reference to fig. 1. First, a preliminary research phase is carried out. In the preliminary stage of the study, a plurality of steel samples having slightly different compositions were prepared. Samples include steels of standard characteristics among those in mass production, as well as steels in which the content of various components is intention