Search

CN-122024970-A - High-entropy alloy design method based on machine learning and high-entropy alloy for nuclear use

CN122024970ACN 122024970 ACN122024970 ACN 122024970ACN-122024970-A

Abstract

The application discloses a machine learning-based high-entropy alloy design method and a nuclear high-entropy alloy, and belongs to the technical field of metal material design. The design method comprises the following steps of collecting historical high-entropy alloy data, establishing an initial data set of composition-process-phase structure-mechanical property, establishing an initial characteristic variable composed of element characteristics and experience characteristics, screening to obtain key characteristic variables, respectively establishing a classification model of a predicted phase structure and a regression model of the predicted mechanical property, screening alloy compositions according to a target phase structure and the target mechanical property based on the key characteristic variable, the classification model and the regression model to obtain design compositions, and verifying the design compositions through experiments. The design method realizes the cooperative optimization of the strength and the toughness on the premise of meeting the phase structure, thereby being capable of designing the high-entropy alloy meeting various performance requirements, in particular to the high-entropy alloy which is applied to an advanced nuclear energy system and has FCC phase structure, high strength and toughness and high temperature resistance.

Inventors

  • ZHENG MINGJIE
  • PAN HAO
  • LIU XIN
  • DING WENYI

Assignees

  • 中国科学院合肥物质科学研究院

Dates

Publication Date
20260512
Application Date
20260214

Claims (10)

  1. 1. The high-entropy alloy design method based on machine learning is characterized by comprising the following steps of: collecting historical high-entropy alloy data, establishing an initial data set of composition-process-phase structure-mechanical property, and cleaning the data in the initial data set to obtain a cleaning data set of the high-entropy alloy; constructing an initial characteristic variable composed of element characteristics and experience characteristics, and screening the initial characteristic variable to obtain a key characteristic variable; based on the phase structure information and the mechanical property information in the cleaning data set, respectively constructing a classification model for predicting a phase structure and a regression model for predicting mechanical property; setting a search space, and screening alloy components according to a target phase structure and target mechanical properties based on the key feature variables, the classification model and the regression model to obtain design components; and if the experimental result is consistent with the predicted result, ending the design, otherwise, adding the experimental result into the training set, and re-designing until the high-entropy alloy meeting the nuclear requirement is designed.
  2. 2. The high entropy alloy design method according to claim 1, wherein the initial data set includes at least the following data information: the data information 1 comprises the atomic percentages of elements of the high-entropy alloy and corresponding elements; Data information 2, phase structure information of the high-entropy alloy; data information 3, mechanical property data of the high-entropy alloy; Preferably, the phase structure information includes whether it is composed of a solid solution phase, whether it is a single phase, and whether it includes a face-centered cubic structure phase; Preferably, the mechanical property data is tensile test data, including yield strength, tensile strength and total elongation; preferably, the initial data set further comprises data information 4, the preparation and/or heat treatment process of the high entropy alloy corresponding to the data information 1.
  3. 3. The method of designing a high-entropy alloy according to claim 1, wherein the cleaning of the data in the initial dataset includes the steps of: (1) Confirming whether the atomic percentage addition in the initial data set is 100%, and if not, carrying out normalization processing; (2) Calculating the quartile of the mechanical property data column in the initial data set, and carrying out preliminary evaluation on the composition, process and phase structure information of outlier data points, calculating N sample points of each alloy and the neighbor of each alloy in a composition space based on data information 1, wherein N is an integer not less than 3, calculating the relative performance difference between each alloy and the surrounding alloy components and constructing difference distribution, and considering that the recorded alloy performance data is abnormal and deleting the data points if the relative difference between the mechanical property of the similar components and the performance of the point is greater than 3 sigma; (3) And (3) manually reviewing the rest data points, and removing sample points which are different by more than 2 sigma and cannot be interpreted by a citation document or are insufficient to be supported by expert knowledge by combining phase formation criteria, mechanical properties and phase structures of alloys composed of the same elements.
  4. 4. The method of designing a high-entropy alloy according to claim 1, wherein the element features are formed by subjecting the atomic percentage composition of the high-entropy alloy in the cleaning dataset to data processing with elements according to formulas (1) to (4): , Wherein M p 、S p 、F p and D p are respectively weighted average, weighted standard deviation, range and difference values of element parameters and components, c i represents the atomic percent of the ith element, p i is the element parameter of the ith element, and corresponding c k and p k are the atomic percent of the kth element and the element parameter; The empirical characteristic is from a characteristic that has been demonstrated in the past literature to be relevant to the target performance.
  5. 5. The method of designing a high-entropy alloy according to claim 1, wherein the screening of the initial characteristic variables includes the steps of: Based on the initial characteristic variable, screening the mechanical properties in the cleaning data set based on correlation to obtain a characteristic subset X; and constructing a correlation coefficient matrix for the feature subset X according to the correlation among the features, and removing redundant features with high correlation to obtain a feature subset X' composed of key feature variables.
  6. 6. The method of claim 1, wherein the machine learning algorithm of the classification model and the regression model is selected from any one of decision trees, gradient-lifting decision trees, support vector machines, k-nearest neighbors, random forests, and feed-forward neural networks.
  7. 7. The high-entropy alloy design method according to claim 1, wherein the search space is set as: selecting at least 4 metal elements meeting target requirements, wherein the atomic percentage of each element is 0-100%; Preferably, the target requirement is a core requirement, and the metal elements satisfying the core requirement are at least four of Al, cr, fe, mn, ni, ta, ti, V, W and Zr.
  8. 8. The method for designing a high-entropy alloy according to claim 1, wherein the design components are screened as follows: Setting a search space, sampling components of the complete space, constructing characteristics of the collected virtual alloy components according to the requirement of initial characteristic variables, then sampling in the component space by using a pseudo-random number method, ensuring the total sum of the extracted components to be 100%, constructing the characteristics according to the construction method of the initial characteristic variables according to the extracted samples, predicting a phase structure based on the classification model, and screening the high-entropy alloy meeting FCC, SP, SS; And uniformly sampling the component range given after screening, inputting the key characteristic variable into the regression model to obtain a mechanical property prediction structure, and selecting components meeting the mechanical property requirement as design components.
  9. 9. A machine learning based high entropy alloy design system/device, characterized in that the system/device, when running, performs the design method according to any of claims 1-8.
  10. 10. A high-entropy alloy for a core is characterized in that the high-entropy alloy for the core has an expression of Al u Cr v Fe w Mn x Ni y Ti z , wherein u, v, w, x, y and z in the expression respectively represent the atomic percentages of corresponding elements, and the following conditions are satisfied that u is more than or equal to 1 and less than or equal to 5, v is more than or equal to 10 and less than or equal to 15, w is more than or equal to 45 and less than or equal to 55, x is more than or equal to 5 and less than or equal to 15, y is more than or equal to 20 and less than or equal to 30, z is more than or equal to 2 and less than or equal to 8, and u+v+w+x+y+z=100.

Description

High-entropy alloy design method based on machine learning and high-entropy alloy for nuclear use Technical Field The application belongs to the technical field of metal material design, and particularly relates to a high-entropy alloy design method based on machine learning, and a high-entropy alloy obtained by the design method, in particular to a nuclear high-entropy alloy. Background The high-entropy alloy is taken as an emerging metal material design concept, and a simple solid solution phase structure (such as face-centered cubic FCC or body-centered cubic BCC) is formed by mixing multi-principal-element equal atomic ratios or near-principal-atomic ratios, so that various excellent performances which are not possessed by the traditional alloy are displayed, including high strength, high toughness, excellent high-temperature resistance and excellent radiation damage resistance. In view of the excellent performance of the high-entropy alloy, the high-entropy alloy has great application prospect in advanced nuclear energy systems represented by a fourth-generation fission reactor and a fusion reactor, and particularly the high-entropy alloy with a single-phase FCC structure has great potential in the aspect of being used as a next-generation nuclear structural material due to good structural stability and strong plastic deformation capability. At present, a material informatics method represented by machine learning provides a brand new paradigm for the design of high-entropy alloys. By constructing the relation between the components, the structure and the performance, the machine learning model can efficiently perform virtual screening and performance prediction in a wide component space, so as to guide experimental design, realize the design of the multi-performance collaborative optimization high-entropy alloy and remarkably accelerate the development process of new materials. However, for the multi-objective and multi-constraint collaborative design such as "high strength and toughness" and "specific phase structure" (such as single-phase FCC) required by the current advanced nuclear energy system, a system and efficient design strategy is still lacking, and particularly under the constraint of the specific phase structure, collaborative optimization of various mechanical properties such as strength (such as yield strength and tensile strength) and plasticity (such as elongation) is not available at present. Disclosure of Invention In view of the above, the primary objective of the present application is to provide a machine learning-based high-entropy alloy design method, which realizes the collaborative optimization of strength and toughness on the premise of meeting the phase structure, so as to design a high-entropy alloy meeting various performance requirements, in particular to a high-entropy alloy with FCC phase structure, high strength and toughness and high temperature resistance for advanced nuclear energy system application. In order to achieve the above purpose, the present application adopts the following technical scheme: the application discloses a high-entropy alloy design method based on machine learning, which comprises the following steps of: collecting historical high-entropy alloy data, establishing an initial data set of composition-process-phase structure-mechanical property, and cleaning the data in the initial data set to obtain a cleaning data set of the high-entropy alloy; constructing an initial characteristic variable composed of element characteristics and experience characteristics, and screening the initial characteristic variable to obtain a key characteristic variable; based on the phase structure information and the mechanical property information in the cleaning data set, constructing a classification model for predicting a phase structure and a regression model for predicting mechanical properties; setting a search space, and screening alloy components according to a target phase structure and target mechanical properties based on the key feature variables, the classification model and the regression model to obtain design components; and if the experimental result is consistent with the predicted result, ending the design, otherwise, adding the experimental result into the training set, and re-designing until the high-entropy alloy meeting the nuclear requirement is designed. Another aspect of the application discloses a machine learning based high entropy alloy design system/apparatus that, when operated, performs the design method described in the present application. In another aspect of the application, the high-entropy alloy for the core has an expression of Al uCrvFewMnxNiyTiz, wherein u, v, w, x, y and z respectively represent the atomic percentages of corresponding elements, and the conditions that u is more than or equal to 1 and less than or equal to 5, v is more than or equal to 10 and less than or equal to 15, w is more than or equal to 45 and less t