Search

CN-121983164-A - Tissue prediction model suitable for alloy steel and establishment method and application thereof

CN121983164ACN 121983164 ACN121983164 ACN 121983164ACN-121983164-A

Abstract

The invention provides a tissue prediction model suitable for alloy steel and an establishing method and application thereof, comprising the following steps of S1, acquiring component and tissue type data of which component ranges cover various steel types, calculating the statistical characteristics of each element by taking the mass fraction of alloy elements as a characteristic variable, S2, carrying out data processing to obtain a decision boundary, dividing the tissue types into single-phase austenite, austenite plus ferrite or martensite, ferrite and/or martensite, S3, respectively adopting the linear discriminant analysis of the acquired data in the step S2 to reduce the dimension and combining with softmax regression, obtaining 2 LDA expressions after the linear discriminant analysis of the dimension reduction, and obtaining a shaeffler-like tissue prediction map according to decision boundary lines of the corresponding 3 tissue types. Most of the organization of the materials can be correctly classified by the invention, and the accuracy is higher, thereby being convenient for popularization and use in new materials and design components.

Inventors

  • HUANG DONG
  • ZHANG YUXIANG
  • WANG RENFU
  • YANG CHAOFEI
  • ZHOU XIAOFENG

Assignees

  • 洛阳船舶材料研究所(中国船舶集团有限公司第七二五研究所)

Dates

Publication Date
20260505
Application Date
20260126

Claims (10)

  1. 1. The method for establishing the structure prediction model suitable for the alloy steel is characterized by comprising the following steps of: S1, acquiring component and tissue type data of which component ranges cover various steel types, and calculating the statistical characteristics of each element by taking the mass fraction of alloy elements as characteristic variables; S2, data processing is carried out, a decision boundary is obtained, and the tissue type is divided into single-phase austenite, austenite plus ferrite or martensite, ferrite and/or martensite; S3, a prediction model: And (3) performing linear discriminant analysis and dimension reduction on the obtained data in the step (S2) and combining softmax regression, obtaining 2 LDA expressions after performing linear discriminant analysis and dimension reduction, and obtaining a tissue-like shaeffler prediction map according to decision boundary lines of the corresponding 3 tissue types.
  2. 2. The method for building a tissue prediction model according to claim 1, wherein step S2 specifically comprises the following steps: s21, carrying out standardization processing on all samples containing d components to standardize d-dimension data; S22, obtaining a conversion matrix W d×2 by adopting an LDA algorithm to the standardized data; s23, multiplying the standard deviation after the standardization processing by a conversion matrix W d×2 , and reducing d-dimension data to two dimensions; s24, classifying by adopting a multi-element softmax classification method, and obtaining a decision boundary.
  3. 3. The method of building a tissue prediction model according to claim 2, wherein the probability that sample x belongs to a class is calculated by first calculating the score s k (x) that sample x belongs to class k, and then using a softmax function Then, cross entropy is adopted as a loss function, and a final parameter matrix is solved through a gradient descent method 。
  4. 4. The method for building a tissue prediction model according to claim 1, wherein the score s k (x) is calculated by the formula (1), and the probability is calculated by the formula (2) The parameter matrix The form of (C) is shown as a formula (3), 。
  5. 5. The method for building a tissue prediction model according to claim 1, wherein in step S3, the 2 LDA expressions obtained after the linear discriminant analysis and dimension reduction are represented by the following formulas (4) and (5), 。
  6. 6. The method for building a tissue prediction model according to claim 1, wherein in the step S3, decision boundary lines of the corresponding 3 tissue types are shown in formulas (6) to (8), 。
  7. 7. The method according to claim 1, wherein in step S1, composition and structure type data of a plurality of steel types such as medium manganese steel, high manganese steel, light steel, stainless steel, nitrogen-containing stainless steel, etc. with a composition range covering are obtained as samples, and it is ensured that the corresponding structure covers a complete austenite structure, an austenite+ferrite/martensite, a ferrite, a martensite, a ferrite+martensite structure.
  8. 8. A structure prediction model suitable for alloy steel, characterized in that it is obtained by the method for establishing a structure prediction model according to any one of claims 1 to 7.
  9. 9. The application of the structure prediction model suitable for alloy steel in structure prediction is characterized in that the method for establishing the structure prediction model according to any one of claims 1-7 and/or the structure prediction model according to claim 8 are adopted.
  10. 10. The use according to claim 9, wherein C, si, mn, cr, ni, mo, al, cu major components of new materials or design components are marked in the class shaeffler tissue prediction map according to 2 LDA expressions to visually determine the tissue type based on the region of the new components in the class shaeffler tissue prediction map.

Description

Tissue prediction model suitable for alloy steel and establishment method and application thereof Technical Field The invention relates to the technical field of metal material structure prediction, in particular to a structure prediction model suitable for alloy steel, and a building method and application thereof. Background Alloy steels generally include medium manganese steels, high manganese steels, light weight steels, stainless steels, nitrogen containing stainless steels, and the like. For example, high manganese steel and fe—mn—al—c low density steel have been attracting attention in terms of weight reduction in structures such as automobiles and ships because of their high strength, high plasticity and low density. The high-nitrogen stainless steel has the advantages that the nickel nitride is adopted, the nickel content is greatly reduced, meanwhile, the toughness is obviously improved, the single austenite structure is maintained, and the biocompatibility can be obviously improved due to the lower nickel content. Therefore, the method is increasingly applied to nuclear power and biological medicine. In order to ensure the solubility of nitrogen, the alloy composition also contains a large amount of manganese. The novel steel material is characterized by comprising manganese element in a certain mass fraction. The structure often contains ferrite, martensite, austenite or a mixed structure thereof, and the difference of the structures significantly influences the mechanical property, corrosion resistance, magnetism and other application properties. In order to obtain the corresponding mechanical properties and application properties, the components need to be designed to obtain the corresponding tissue type. Generally, cr-Ni based stainless steel is constituted by a composition and structure according to SCHAEFFLER, WRC-2000, delong, or the like, and SCHAEFFLER is the most widely used. The schiff diagram (SCHAEFFLER DIAGRAM) is a common method for predicting the tissue type of the stainless steel through chemical components, and compared with an equilibrium phase diagram established by adopting a CALPHAD method, the SCHAEFFLER diagram establishes a relationship between the tissue type and the chemical components by adopting actual test data, so that a prediction result is more reliable, and the tissue type in an unbalanced state can be predicted. However, as the alloy composition deviates from Cr-Ni based stainless steel, the accuracy of SCHAEFFLER diagrams is greatly reduced. Kleuh and Lee et Al set up a SCHAEFFLER-like diagram suitable for medium-, high-manganese-, and Fe-Mn-Al-C-based low density steels by adjusting the weight coefficient of each component in Cr equivalent and Ni equivalent in SCHAEFFLER diagram and performing operations such as translation and rotation on decision boundaries, but it cannot be applied to high nitrogen stainless steel. The WRC-2000 drawing and Delong drawing are not suitable for tissue prediction of medium manganese steel, high manganese steel, light steel, nitrogen-containing stainless steel and the like. Disclosure of Invention In view of the above, the invention aims to provide a structure prediction model suitable for alloy steel and an establishment method thereof, so as to solve the problems that shaeffler diagrams, WRC-2000 diagrams, delong diagrams and the like in the prior art cannot be uniformly suitable for structure prediction of alloy steel such as medium manganese steel, high manganese steel, light steel, nitrogen-containing stainless steel and the like, and the accuracy of structure prediction is low. In order to achieve the above purpose, the technical scheme of the invention is realized as follows: In a first aspect, the present invention provides a method for establishing a structure prediction model suitable for alloy steel, including the following steps: S1, acquiring component and tissue type data of which component ranges cover various steel types, and calculating the statistical characteristics of each element by taking the mass fraction of alloy elements as characteristic variables; S2, data processing is carried out, a decision boundary is obtained, and the tissue type is divided into single-phase austenite, austenite plus ferrite or martensite, ferrite and/or martensite; S3, a prediction model: And (3) performing linear discriminant analysis and dimension reduction on the obtained data in the step (S2) and combining softmax regression, obtaining 2 LDA expressions after performing linear discriminant analysis and dimension reduction, and obtaining a tissue-like shaeffler prediction map according to decision boundary lines of the corresponding 3 tissue types. Further, the step S2 specifically includes the following steps: s21, carrying out standardization processing on all samples containing d components to standardize d-dimension data; S22, obtaining a conversion matrix W d×2 by adopting an LDA algorithm to the standardized data; s23, m