CN-122024933-A - Design method of high-temperature high-specific strength refractory high-entropy alloy component driven by calculation

CN122024933ACN 122024933 ACN122024933 ACN 122024933ACN-122024933-A

Abstract

The invention discloses a calculation-driven high-temperature high-specific-strength refractory high-entropy alloy component design method, which comprises the following steps of 1, collecting refractory high-entropy alloy sample data, 2, referring to element physicochemical quantities of refractory high-entropy alloy elements related to yield strength and phase composition, 3, constructing characteristics required by a machine learning model, 4, selecting the machine learning model and parameters with optimal performance to form a strength and phase composition prediction model, 5, obtaining a body-centered cubic single-phase alloy sample, 6, obtaining points to be marked, 7, updating a data set and a specific strength prediction model, and 8, repeatedly carrying out the steps 6 and 7 until the specific strength of the marked alloy sample in a potential high-temperature high-specific-strength alloy system in the step 5 is improved to meet convergence conditions. The method can effectively relieve the dependence of machine learning on experimental data, thereby optimizing refractory high-entropy alloy components faster and at lower cost.

Inventors

ZHANG TONG
GONG ZHANPENG
JIANG YIHUI
HUO WANGTU
CAO FEI
CHEN KAIYUN
DU XIN

Assignees

西安理工大学

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (9)

1. The design method of the high-temperature high-specific strength refractory high-entropy alloy component driven by calculation is characterized by comprising the following steps of: step 1, collecting refractory high-entropy alloy sample data; step 2, referring to the element physicochemical quantities of refractory high-entropy alloy elements related to yield strength and phase composition; Step 3, constructing the required characteristics of the machine learning model; Step 4, selecting a machine learning model with optimal performance and a parameter composition strength and phase composition prediction model; Step 5, obtaining a body-centered cubic single-phase alloy sample; Step 6, obtaining points to be marked; Step 7, updating the data set and the specific strength prediction model; and 8, repeating the step 6 and the step 7 until the specific strength of the marked alloy sample in the alloy system with the potential high-temperature high-specific strength in the step 5 is improved to meet the convergence condition.
2. The method for designing the high-temperature high-specific-strength refractory high-entropy alloy composition driven by calculation according to claim 1, wherein in the step 1, refractory high-entropy alloy sample data comprise specific alloy compositions, corresponding mechanical properties and phase structure compositions, wherein metal elements in refractory alloy in the sample data comprise Ti, V, zr, nb, mo, hf, ta, W, al, cr; in the step 1, alloy samples in the data of the collected refractory high-entropy alloy samples are prepared by arc melting, the strength is the alloy yield strength at 1000 ℃, and the phase composition of the alloy is obtained based on X-ray diffraction.
3. The method of claim 2, wherein the physical and chemical quantities of the refractory high-entropy alloy element include Barling electronegativity, armillarity, valence electron concentration, free electron concentration, electron work function, first ionization energy, second ionization energy/electron volt, number of filled valence orbitals, number of unfilled valence orbitals, number of filled s-orbitals, number of filled d-orbitals, number of unfilled d-orbitals, electron density of the surface of the Vignat-Siz cell, atomic number, family, period, bulk modulus of elasticity, shear modulus, young's modulus, relative atomic mass, molar volume, density, poisson's ratio, specific heat capacity, thermal conductivity, electrical conductivity, shear modulus at 0K, viscosity of liquid metal, metal radius, covalent radius, atomic volume, body-centered lattice constant, melting temperature, boiling point, heat of vaporization, heat of fusion, atomization enthalpy, cohesive energy, vacancy energy, and vacancy energy.
4. The computationally driven high temperature high specific strength refractory high entropy alloy composition design method according to claim 3, wherein in step 3, the features are used as input ends of a machine learning model, and the mechanical properties and phase structure composition of the alloy in step 1 are used as mapping ends; in step 3, the construction of the features is performed by the following formula: Wherein the method comprises the steps of For the weighted sum feature, For the square weighting characteristic of deviation, n is the number of element physical and chemical quantities, Is the atomic percent of the elements, Representing the value of i physicochemical quantities on the j-th element, A weighted average of the physical and chemical characteristic amounts of the elements is represented.
5. The method for designing the high-temperature high-specific strength refractory high-entropy alloy component driven by calculation according to claim 4, wherein the step 4 is specifically that importance sorting and screening are carried out among features in the step 3, parameter evaluation is carried out on a plurality of machine learning models, super-parameter space of five algorithm models is traversed, and a machine learning model with optimal performance and parameters for different mapping end evaluation indexes in the step 3 are selected to form a strength and phase composition prediction model; In step 4, the selected machine learning model comprises five models of extreme lifting tree, decision tree, random forest, gradient lifting tree and Gaussian regression, and the adjusted parameters comprise super parameters of various machine learning models; In the step 4, the process of sorting and screening the importance among the features in the step 3 is carried out according to the following modes that (1) the Person correlation coefficient among the features is calculated, the features with average correlation coefficient larger than 0.8 are deleted, (2) feature importance analysis is carried out on the undeleted feature set, and the features with the feature importance being ranked front 15 are reserved, (3) the 15 features are subjected to exhaustive analysis to obtain the feature combination with the best performance, wherein the combination comprises seven features of Yang Xing modulus deviation square E-2, melting point weighting Tm-1, relative atomic mass weighting Ar-1, specific heat capacity weighting C-1, armillarity deviation square χar-2, atomic volume deviation square Vat-2 and atomic radius weighting rate-1; In step 4, the index of the strength and phase composition prediction model evaluation is the classification accuracy of the model on the test set, and the index of the strength and phase composition prediction model evaluation is the determination coefficient between the predicted value and the true value of the model on the test set.
6. The method for designing the high-temperature high-specific strength refractory high-entropy alloy composition driven by calculation according to claim 5, wherein step 5 is specifically characterized in that based on SHAP analysis in a game theory, explanation analysis alloy elements are ordered based on importance of the features screened in step 4, a high-entropy alloy system with potential high-temperature high-specific strength is determined based on elements with the front importance order, alloy sample design is performed by setting element concentration range constraint, and a body-centered cubic single-phase alloy sample is obtained by performing body-centered cubic single-phase screening on the designed alloy sample by using a phase composition prediction model in step 4; In step 5, the process of determining the high-entropy alloy system with potential high-temperature high-specific strength through SHAP analysis comprises (1) performing SHAP analysis to give a relation between the characteristics of a machine learning model and performance indexes, (2) performing analysis to give a relation between the characteristic combination which is the best in step 4 and SHAP values, (3) determining the relation between the characteristic combination which is the best in step 4 and alloy related performance according to the analysis conclusion of (1) and (2), and determining the types of elements based on the relation, namely selecting the types of elements which are favorable for the high-temperature high-specific strength of the alloy according to the physicochemical characteristic values of the elements.
7. The method for designing the high-temperature high-specific strength refractory high-entropy alloy composition driven by calculation according to claim 6, wherein step 6 is specifically that the body-centered cubic single-phase alloy sample screened in step 5 is given out the next round of component points to be marked in active learning optimization based on a Bayesian optimization theory, so as to obtain a sample with a high potential value in function calculation as the point to be marked; in the step 6, after the alloy system to be optimized is determined, the body-centered cubic single-phase alloy sample is a component set obtained by screening the corresponding component space through the strength and phase composition prediction model; In step 6, the calculation formula of the function is as follows: Where EI is the desired improvement value for the alloy sample, For the deviation of the predicted mean of the specific intensities from the optimal specific intensity, To predict the standard deviation for the corrected specific intensity, For the purpose of the standard deviation of the light, Is a cumulative distribution function of a standard normal distribution, For a standard normal distribution probability density, 8 component points to be evaluated are selected each time in the implementation process.
8. The method for designing the high-temperature high-specific-strength refractory high-entropy alloy composition driven by calculation according to claim 7, wherein the step 7 is characterized in that the specific strength value of the composition point to be marked in the step 6 is given by using an alloy predicted strength method based on molecular dynamics calculation, and a data set and a specific strength prediction model are updated; In the step 7, the alloy predicted strength is obtained based on molecular dynamics calculation, and the critical slip stress of the edge dislocation and the screw dislocation on different slip planes is calculated based on the molecular dynamics calculation mode, so that the alloy critical cutting stress of the corresponding components is determined, and the macroscopic yield stress of the polycrystal is obtained through a Taylor factor; in step 7, updating the data set refers to adding the calculated intensity-performance data into the data set of the machine learning model, and updating the specific intensity prediction model refers to retraining the machine learning model based on the updated data set.
9. The method for designing the high-temperature high-specific-strength refractory high-entropy alloy composition driven by calculation according to claim 8, wherein step 8 is specifically that step 6 and step 7 are repeated until the specific strength of the marked alloy sample in the potential high-temperature high-specific-strength alloy system in step 5 is improved to meet the convergence condition, and the process is ended; in the step 8, the alloy specific strength improvement meets the convergence condition that EI values of two groups of optimal samples in two adjacent optimization round systems reach convergence.

Description

Design method of high-temperature high-specific strength refractory high-entropy alloy component driven by calculation Technical Field The invention belongs to the technical field of refractory high-entropy alloy component design, and particularly relates to a calculation-driven high-temperature high-specific-strength refractory high-entropy alloy component design method. Background The higher the working temperature of the high-temperature section of the aero-engine is, the higher the operation efficiency is. Reviewing the history, breakthroughs in engine performance are based on the continued development of high specific strength structural materials. The materials not only have high-temperature stability and excellent mechanical properties, but also have low-density core requirements to realize the comprehensive targets of high reliability, long service life and light weight. Refractory high-entropy alloy (Refractory High-entopy alloy, RHEA) is a high-entropy alloy taking IV-VI group refractory metal elements as main elements, has excellent high-temperature mechanical properties and stability, and is a high-specific strength material system expected to be applied to high-temperature and high-load working conditions after the conventional high-temperature alloy. However, empirically, the improvement of the high-temperature performance depends on the addition of a large amount of elements such as W, nb, ta and the like, so that the high-temperature mechanical performance of RHEA has an inverse relation with the density, and the development of the high-specific strength RHEA is restricted. In general, the addition of different elements and adjustment of the element proportions allows for a wide range of adjustments REHEA to properties (including mechanical properties and density). However RHEA is complex in composition, large in component space, and lacks accurate models and guidelines to guide the rational design of RHEA, so optimizing its performance by the traditional means of expensive experimental test-feedback-redesign is greatly hindered. Although machine learning methods have achieved a series of success in rapidly optimizing the design of high-entropy alloy compositions in recent years, building accurate machine learning models requires a large amount of expensive alloy composition-performance data, and high-temperature test conditions further increase the difficulty in acquiring these data. Even selective marking performance by active learning is time consuming and economical. The resource requirement limits the rapid and low-cost improvement RHEA of the high-temperature specific strength by the experiment auxiliary active learning. Therefore, how to effectively relieve the dependence of machine learning on experimental data, thereby optimizing refractory high-entropy alloy components faster and at lower cost, and having important significance for developing refractory high-entropy alloy with high-temperature-aspect-ratio strength. Disclosure of Invention The invention aims to provide a calculation-driven high-temperature high-specific-strength refractory high-entropy alloy component design method, which can effectively relieve the dependence of machine learning on experimental data, so that the refractory high-entropy alloy component is optimized faster and at lower cost. The technical scheme adopted by the invention is that the design method of the high-temperature high-specific strength refractory high-entropy alloy component driven by calculation specifically comprises the following steps: step 1, collecting refractory high-entropy alloy sample data; step 2, referring to the element physicochemical quantities of refractory high-entropy alloy elements related to yield strength and phase composition; Step 3, constructing the required characteristics of the machine learning model; Step 4, selecting a machine learning model with optimal performance and a parameter composition strength and phase composition prediction model; Step 5, obtaining a body-centered cubic single-phase alloy sample; Step 6, obtaining points to be marked; Step 7, updating the data set and the specific strength prediction model; and 8, repeating the step 6 and the step 7 until the specific strength of the marked alloy sample in the alloy system with the potential high-temperature high-specific strength in the step 5 is improved to meet the convergence condition. The invention is also characterized in that: in the step 1, refractory high-entropy alloy sample data comprise specific alloy components, corresponding mechanical properties and phase structure components, wherein metal elements in refractory alloy in the sample data comprise Ti, V, zr, nb, mo, hf, ta, W, al, cr; in the step 1, alloy samples in the data of the collected refractory high-entropy alloy samples are prepared by arc melting, the strength is the alloy yield strength at 1000 ℃, and the phase composition of the alloy is obtained based on X-ray diffraction. In