Refractory high-entropy alloy component optimization design method and device
By establishing a supercell model and a machine learning model, the composition of refractory high-entropy alloys was optimized, which solved the problem of low efficiency of traditional methods and achieved the effect of rapid screening of high-performance alloy compositions.
Patent Information
- Application Number
- CN202510960314.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-11
- Publication Date
- 2025-10-17
AI Technical Summary
Existing technologies make it difficult to efficiently screen out high-performance refractory high-entropy alloy components. Traditional methods are costly and inefficient, and require large first-principles computing resources and are slow.
A supercell model of the composition of refractory high-entropy alloys was established, the elastic matrix and mechanical properties data were calculated after structural relaxation optimization, a sample set was constructed and a machine learning model was trained, and the optimized alloy composition was determined through interpretable analysis.
Potential high-performance alloy components can be quickly and accurately screened without a large number of experiments, significantly improving the efficiency of optimizing the composition of refractory high-entropy alloys.
Smart Images
Figure CN120808939A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of metal material design, and particularly relates to a refractory high-entropy alloy component optimization design method and device. BACKGROUND
[0002] The refractory high-entropy alloy (RHEAs) material system breaks through the design concept of traditional high-temperature alloys. Compared with the 1200 DEG C temperature limit of traditional Ni / Co-based high-temperature alloys, through the component solid solution of multiple high-melting-point refractory metal components, the RHEAs can still maintain excellent strength at 1400 DEG C high-temperature environment; at the same time, the RHEAs have excellent high-temperature mechanical properties and heat resistance, and are an important development direction of the next generation of metal-based thermal protection materials.
[0003] However, due to the multi-principal-element design concept, the number of potential components of the refractory high-entropy alloy is much larger than that of the traditional alloy. Only considering the commonly used elements in the periodic table, the component space contains numerous alloy variants. So far, the explored refractory high-entropy alloy region is only a tip of the iceberg. Due to the huge component space region, there are great challenges in using the "trial and error method" to optimize the design of the high-performance refractory high-entropy alloy components.
[0004] The traditional alloy design method mainly relies on experimental research, which has high cost, low efficiency and is difficult to systematically explore a large-scale component space. In recent years, the first principle calculation can provide accurate prediction of material performance, but the calculation resource demand is large and the calculation speed is slow, and it is difficult to efficiently screen a large number of candidate materials. In this case, it is urgent to develop an accurate and efficient refractory high-entropy alloy component optimization design method and device. SUMMARY
[0005] The present application provides a refractory high-entropy alloy component optimization design method and device, which can quickly, accurately and efficiently predict the target performance of multi-principal-element refractory high-entropy alloys, and can screen potential high-performance alloy components without a large number of experiments.
[0006] In a first aspect, the present application provides a refractory high-entropy alloy component optimization design method, comprising:
[0007] establishing a supercell model of a refractory high-entropy alloy component;
[0008] calculating the elastic matrix and mechanical property data of the refractory high-entropy alloy component after structural relaxation optimization of the supercell model;
[0009] constructing a sample set based on the element content, physical characteristic information and mechanical property data of the refractory high-entropy alloy component;
[0010] The sample set is used to train a machine learning model to obtain a plurality of target prediction models; wherein the sample set includes element content, physical characteristic information as input and mechanical property data as output; different target prediction models output different types of mechanical property data;
[0011] The target prediction models are subjected to explainability analysis to obtain key elements of each target prediction model, so as to determine an optimized refractory high-entropy alloy composition according to the key elements.
[0012] Optionally, the refractory high-entropy alloy composition includes at least four elements among Al, Si, Ti, Cr, Zr, Nb, Mo, Hf, Ta and W.
[0013] Optionally, the supercell model of the refractory high-entropy alloy composition includes:
[0014] The supercell model is established by using a quasi-random approximation method; wherein the supercell model is a BCC structure solid solution.
[0015] Optionally, after the supercell model is subjected to structure relaxation optimization, the elastic matrix and mechanical property data of the refractory high-entropy alloy composition are calculated, including:
[0016] Parameters for structure relaxation optimization are determined, and structure relaxation optimization is performed based on the parameters to obtain a completely relaxed equilibrium lattice structure; wherein the parameters include a plane wave cutoff energy, a K-point grid, an energy convergence standard and a force convergence standard.
[0017] A stress tensor is obtained by applying a strain to the equilibrium lattice structure.
[0018] Based on the stress tensor and the strain, the elastic matrix is calculated.
[0019] According to the elastic matrix, the mechanical property data including bulk modulus, Young's modulus, shear modulus, Pugh ratio, Poisson's ratio and Cauchy pressure are calculated.
[0020] Optionally, the physical characteristic information includes atomic size difference, average electronegativity, electronegativity difference, valence electron concentration, melting point, thermal expansion coefficient, molar volume, mixing entropy, mixing enthalpy, entropy-enthalpy competition relationship and isochoric specific heat.
[0021] Optionally, the sample set is used to train a machine learning model to obtain a plurality of target prediction models, including:
[0022] For each type of mechanical property data, the following is performed:
[0023] inputting element content and physical characteristic information of elements in the sample set into the machine learning model, and outputting the predicted mechanical property data of the type;
[0024] comparing the predicted mechanical property data of the type with the mechanical property data of the type included in the sample set to obtain a root mean square error and a determination coefficient;
[0025] when the root mean square error and the determination coefficient both satisfy a preset threshold, completing the training to obtain a target prediction model for the mechanical property data of the type.
[0026] Optionally, the target prediction model is subjected to explainability analysis to obtain a key element of each target prediction model, so as to determine an optimized refractory high-entropy alloy composition according to the key element, including:
[0027] for each target prediction model, the following is performed:
[0028] generating an interpreter and calculating a feature contribution value of each feature included in each sample by using the interpreter;
[0029] for each feature, the following is performed: calculating a mean value of the feature contribution value of the feature based on the feature contribution values of all samples including the feature;
[0030] determining a feature corresponding to the mean value of the feature contribution value greater than a preset mean value threshold as a first feature;
[0031] determining a contribution impact direction of the first feature;
[0032] determining a key feature as the first feature with a positive contribution impact direction, and determining an element corresponding to the key feature representing element content as a key element;
[0033] calculating an interaction intensity index between any two features;
[0034] retaining features corresponding to the interaction intensity index greater than a preset index threshold;
[0035] determining an optimized refractory high-entropy alloy composition satisfying a preset requirement according to the key elements and the feature pairs obtained from each target prediction model.
[0036] Optionally, when the preset requirement is to design a refractory high-entropy alloy composition with optimal ductility and containing Nb, the key elements include Nb, Hf, Mo, Ta, Ti and Zr; the second feature includes a mixing enthalpy feature; and the optimized refractory high-entropy alloy composition is NbTiZrMoTa with a molar ratio of 1:1:1:1:1, NbZrTiHfTa with a molar ratio of 1:1:1:1:1, or Nb1.5 ZrTiMoTa 0.5 .
[0037] In a second aspect, the present application further provides a refractory high-entropy alloy component optimization design device, comprising:
[0038] A construction module is configured to establish a supercell model of a refractory high-entropy alloy component.
[0039] A calculation module is configured to calculate an elastic matrix and mechanical property data of the refractory high-entropy alloy component after structural relaxation optimization of the supercell model.
[0040] A sample generation module is configured to construct a sample set based on element content, physical characteristic information and the mechanical property data of the refractory high-entropy alloy component.
[0041] A training module is configured to train a machine learning model using the sample set to obtain a plurality of target prediction models; wherein the sample set comprises element content and physical characteristic information as input and mechanical property data as output; and the target prediction models output different types of mechanical property data.
[0042] An optimization design module is configured to perform interpretability analysis on the target prediction models to obtain key elements of each target prediction model, and determine an optimized refractory high-entropy alloy component based on the key elements.
[0043] In a third aspect, the present application further provides a computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the refractory high-entropy alloy component optimization design method of any one of the above aspects.
[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed in a computer, causes the computer to perform the method of any one of the above aspects.
[0045] In a fifth aspect, the present application further provides a computer program product comprising computer instructions, wherein the computer instructions, when executed by a processor, implement the steps of the method of any one of the first aspects of the present application.
[0046] The application provides a refractory high-entropy alloy component optimization design method and device, which comprises the following steps: first, a supercell model of a refractory high-entropy alloy component is established, and after structural relaxation optimization of the model, the elastic matrix and mechanical property data of the refractory high-entropy alloy component are calculated; then, the element content, physical characteristic information and mechanical property data of each refractory high-entropy alloy component are taken as a sample to construct a sample set; then, the sample set is used to train a machine learning model to obtain a target prediction model for each type of mechanical property data, and the corresponding type of mechanical property data can be output by inputting the element content and physical characteristic information into the target prediction model; then, based on the target prediction model, the element content, the physical characteristic information and the mechanical property data, an interpretability analysis is performed to obtain key elements of each target prediction model, so that the refractory high-entropy alloy component is determined according to the key elements. In this way, by establishing a high-precision target prediction model of mechanical property data, potential high-performance alloy components can be screened out without a large number of experiments, and the key elements affecting the target mechanical property can be determined through the interpretability analysis, so that the efficiency of the refractory high-entropy alloy component optimization is significantly improved. BRIEF DESCRIPTION OF DRAWINGS
[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.
[0048] Figure 1 is a flow chart of a refractory high-entropy alloy component optimization design method provided by an embodiment of the present application;
[0049] Figure 2 is a feature contribution value importance relationship diagram of a Cauchy pressure target prediction model provided by an embodiment of the present application;
[0050] Figure 3 is a hardware architecture diagram of a computing device provided by an embodiment of the present application;
[0051] Figure 4 is a structural diagram of a refractory high-entropy alloy component optimization design device provided by an embodiment of the present application. DETAILED DESCRIPTION
[0052] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.
[0053] The concept of the present application is described below, please refer to Figure 1 The embodiments of the present application provide a refractory high-entropy alloy component optimization design method, comprising:
[0054] Step 100, a supercell model of the refractory high-entropy alloy component is established;
[0055] Step 102, after structural relaxation optimization of the supercell model, the elastic matrix and mechanical property data of the refractory high-entropy alloy component are calculated;
[0056] Step 104, based on the element content, physical characteristic information and mechanical property data of the refractory high-entropy alloy component, a sample set is constructed;
[0057] Step 106, the sample set is used to train a machine learning model to obtain a plurality of target prediction models; wherein the sample set includes element content and physical characteristic information as input and mechanical property data as output; the types of mechanical property data output by different target prediction models are different;
[0058] Step 108, the target prediction model is subjected to interpretability analysis to obtain key elements of each target prediction model, so as to determine the optimized refractory high-entropy alloy component according to the key elements.
[0059] In the embodiment of the present application, first, a supercell model of a refractory high-entropy alloy composition is established, and after structure relaxation optimization of the model, the elastic matrix and mechanical property data of the refractory high-entropy alloy composition are calculated, then the element content, physical characteristic information and mechanical property data of each refractory high-entropy alloy composition are taken as a sample to construct a sample set, then the machine learning model is trained by using the sample set to obtain a target prediction model for each type of mechanical property data, and by inputting the element content and physical characteristic information into the target prediction model, the corresponding type of mechanical property data can be output. Then, based on the target prediction model, the element content, the physical characteristic information and the mechanical property data, an interpretability analysis is performed to obtain the key elements of each target prediction model, so as to determine the optimized refractory high-entropy alloy composition according to the key elements. In this way, by establishing a high-precision target prediction model of mechanical property data, potential high-performance alloy compositions can be screened out without a large number of experiments, and the key elements affecting the target mechanical property can be determined through interpretability analysis, thereby significantly improving the efficiency of refractory high-entropy alloy composition optimization.
[0060] The following describes Figure 1 The execution manner of each step is shown.
[0061] In a preferred embodiment, the refractory high-entropy alloy composition includes at least four elements among Al, Si, Ti, Cr, Zr, Nb, Mo, Hf, Ta and W. For example, the refractory high-entropy alloy composition is composed of Nb, Ti, Zr, Mo and Ta; the refractory high-entropy alloy composition is composed of Al, Si, Ti and Cr, and the like.
[0062] In step 100, a supercell model of a refractory high-entropy alloy composition is established, including:
[0063] The supercell model is established by using a quasi-random approximation method; wherein the supercell model is a BCC structure solid solution.
[0064] In step 102, after structure relaxation optimization of the supercell model, the elastic matrix and mechanical property data of the refractory high-entropy alloy composition are calculated, including:
[0065] The parameters for structure relaxation optimization are determined, and based on the parameters, the structure relaxation optimization is performed to obtain a completely relaxed equilibrium lattice structure; wherein the parameters include a plane wave cutoff energy, a K-point grid, an energy convergence standard and a force convergence standard;
[0066] The equilibrium lattice structure is subjected to strain to obtain a stress tensor;
[0067] Based on the stress tensor and the strain, the elastic matrix is calculated;
[0068] According to the elastic matrix, the mechanical property data including bulk modulus, Young's modulus, shear modulus, Pugh ratio, Poisson's ratio and Cauchy pressure are calculated.
[0069] Specifically, the exchange correlation energy functional uses generalized gradient approximation GGA-PBE, the plane wave cutoff energy is 500 eV, and the K point is set to 15x15x7. The energy convergence criterion is 1x10 -6 -01. After structural relaxation optimization of the constructed supercell model, the elastic matrix C ij of the related refractory high-entropy alloy composition is calculated, and then the mechanical property data are calculated based on the elastic matrix; wherein,
[0070] The calculation formula of the bulk modulus B is:
[0071] The calculation formula of the Young's modulus E is:
[0072] The calculation formula of the shear modulus G is:
[0073] The calculation formula of the Poisson's ratio v is:
[0074] The calculation formula of the Cauchy pressure K cy is K cy =C 11 -C 44 .
[0075] The Pugh ratio is the ratio of the bulk modulus to the shear modulus;
[0076] wherein,
[0077] wherein, C 11 , C 12 , C 13 , C 22 , C 23 , C 33 , C 44 , C 55 , C 66 .
[0078] In a preferred embodiment, the physical characteristic information of step 104 includes: atomic size difference, average electronegativity, electronegativity difference, valence electron concentration, melting point, thermal expansion coefficient, molar volume, mixing entropy, mixing enthalpy, entropy-enthalpy competition relationship and isochoric specific heat.
[0079] Specifically, the atomic size difference (δ r ):
[0080] Electronegativity difference (δ χ ) :
[0081] Average electronegativity (χ a ) : χ a =∑c i ×χ i
[0082] Valence electron concentration (VEC) : VEC =∑c i ×(VEC) i
[0083] Mixing enthalpy (ΔHmix) :
[0084] Mixing entropy (ΔSmix) : ΔSmix =-R×∑c i ×ln(c i )
[0085] Ω parameter : Ω = Tm×ΔSmix / |ΔHmix|
[0086] Molar volume (Vm) : Vm =∑c i ×V i
[0087] Melting point (Tm) : Tm =∑c i ×(Tm) i
[0088] Specific heat capacity (Cv) : Cv =∑c i ×(Cv) i
[0089] Thermal expansion coefficient (Tep) : Tep =∑c i ×(Tep) i
[0090] wherein c i is the concentration (i.e. atomic percentage) of element i, r i is the atomic radius of element i, is the average value of atomic radius in the refractory high-entropy alloy composition (i.e. the sum of all atomic radii in the refractory high-entropy alloy composition divided by the total number of elements contained), χ i is the electronegativity of element i, is the average value of electronegativity in the refractory high-entropy alloy composition; is the enthalpy between element i and element j; R is the gas constant; (VEC) i is the valence electron concentration of element i; (Tm) i is the melting point of element i; (Tep) i is the thermal expansion coefficient of element i; (Cv) iSpecific heat capacity of element i.
[0091] In step 106, the sample set is used to train the machine learning model to obtain a plurality of target prediction models, including:
[0092] For each type of mechanical property data, the following is performed:
[0093] The element content and physical feature information in the sample set are input into the machine learning model, and the prediction mechanical property data of this type is output;
[0094] The prediction mechanical property data of this type is compared with the mechanical property data of this type included in the sample set to obtain the root mean square error and the determination coefficient;
[0095] When the root mean square error and the determination coefficient both satisfy the preset threshold, the training is completed, and the target prediction model for the mechanical property data of this type is obtained.
[0096] In the present application, each prediction model only outputs one type of mechanical property data, for example, the Cauchy pressure target prediction model is used to output the Cauchy pressure; the Pugh ratio target prediction model is used to output the Pugh ratio; and the Poisson ratio target prediction model is used to output the Poisson ratio. In this way, by training the machine learning model with the sample set to obtain the target training model, the mechanical property data of any refractory high-entropy alloy composition can be determined without experiments, and subsequent explainable analysis can be performed on various types of mechanical property data to obtain an optimized refractory high-entropy alloy composition.
[0097] It should be noted that, for example, the present application calculates the mechanical property data of 152 refractory high-entropy alloys composed of Al, Si, Ti, Cr, Zr, Nb, Mo, Hf, Ta and W using the method of step 102, and constructs an original data set. The original data set is used as a sample set to train the machine learning model, and the machine learning model is optimized. The 5-fold cross-validation method is used to evaluate the model performance. By comparing existing regression machine learning algorithms, the best CatBoost model is selected to predict the Cauchy pressure, Pugh ratio and Poisson ratio, respectively, to obtain the optimized Cauchy pressure target prediction model, Pugh ratio target prediction model and Poisson ratio target prediction model. The root mean square error of the Cauchy pressure target prediction model is 4.172 GPa, and the determination coefficient is 0.9746; the root mean square error of the Pugh ratio target prediction model is 0.355, and the determination coefficient is 0.9640; the root mean square error of the Poisson ratio target prediction model is 0.009, and the determination coefficient is 0.9441, which shows that the optimized target prediction model can accurately predict the Pugh ratio, Poisson ratio and Cauchy pressure of the refractory high-entropy alloy.
[0098] In step 108, the target prediction model is subjected to an explainability analysis to obtain key elements of each target prediction model, so as to determine an optimized refractory high-entropy alloy composition according to the key elements, including:
[0099] For each target prediction model, the following is performed:
[0100] An explainer is generated, and the explainer is used to calculate a feature contribution value of each feature included in each sample;
[0101] For each feature, the following is performed: a feature contribution value mean of the feature is calculated based on the feature contribution values of all samples including the feature;
[0102] A feature corresponding to a feature contribution value mean greater than a preset mean threshold value is determined as a first feature;
[0103] A contribution impact direction of the first feature is determined;
[0104] The first feature with a positive contribution impact direction is determined as a key feature, and an element corresponding to the key feature representing an element content is determined as a key element;
[0105] An interaction intensity index between any two features is calculated;
[0106] Features corresponding to an interaction intensity index greater than a preset index threshold value are retained;
[0107] An optimized refractory high-entropy alloy composition satisfying a preset requirement is determined according to the key elements and feature pairs obtained from each target prediction model.
[0108] It should be noted that the other features are features other than the first feature.
[0109] Specifically, the features include element and element content and physical characteristic information. The feature contribution value mean is determined by the following formula:
[0110]
[0111] wherein, I j is the feature contribution value mean of feature j; n is the number of samples in the sample set; is the feature contribution value of feature j in the i th sample;
[0112] The feature contribution value means are sorted from high to low, features with a feature contribution value mean greater than a preset mean threshold value are retained, and the features are determined as first features;
[0113] Then the slope of the first feature is calculated to determine the contribution impact direction of the first feature; wherein the slope is greater than 0, then the contribution impact direction of the first feature is positive; otherwise, it is negative; the slope is determined by the following formula:
[0114]
[0115] Wherein, β j is the slope of the first feature j; x j is the numerical value (i.e. feature value) of the first feature j; cov(φ j , x j ) is the covariance of the feature contribution value of the first feature j and the numerical value of the first feature j; var(x j ) is the variance of the numerical value of the first feature j;
[0116] The first feature with a positive contribution impact direction is determined as a key feature, and the element corresponding to the key feature representing the element content is determined as a key element;
[0117] Then the interaction intensity index between any two features is calculated by detecting the joint effect of the two features; the interaction intensity index is determined by the following formula:
[0118]
[0119] Wherein, Int jk is the interaction intensity index between feature j and feature k; n is the number of samples in the sample set; is the feature contribution value of feature j in the i th sample; is the feature contribution value of feature k in the i th sample; is the feature contribution value of feature j and feature k in the i th sample; j≠k;
[0120] The feature pair with an interaction intensity index greater than a preset index threshold is screened, and the feature pair contains two features with synergistic effect.
[0121] In a more preferred embodiment, step 108 determines the optimized refractory high-entropy alloy composition according to the key element, comprising:
[0122] According to the key element, the mean value of the feature contribution value of each feature, the interaction intensity index, the feature pair and the slope of each feature, the multidimensional evaluation value of any refractory high-entropy alloy is calculated; the multidimensional evaluation value is determined by the following formula:
[0123]
[0124] Wherein, C is the multidimensional evaluation value; σ() represents the Sigmoid function; β jis the slope of feature j; p is the total number of features; λ is the magnification factor, λ>1; P represents the set of feature pairs; Int jk is the interaction strength index between feature j and feature k in a feature pair; ΔI jk is the synergistic effect amount between feature j and feature k; I j , I k , I jk are the average values of feature contribution values of feature j, feature k, and feature j and feature k existing simultaneously, respectively; q is the number of key elements; c m is the current content of element m; is the key content threshold of element m; δ m is the tolerance coefficient of element m, that is, the window width of the selected content range interval of element m; it should be noted that the absolute value of the change rate of the feature contribution value is maximum at
[0125] The refractory high-entropy alloy composition corresponding to the multi-dimensional evaluation value greater than the preset evaluation threshold is input into the target prediction model to obtain predicted mechanical property data.
[0126] An optimized refractory high-entropy alloy composition is determined based on the predicted mechanical property data.
[0127] In the embodiments of the present application, the first term of the above formula strengthens the features that have a positive directional contribution to the mechanical property data, and suppresses the features that have a negative directional contribution; the second term is used to reward synergistic effect combinations and punish antagonistic effect combinations; and the third term focuses on giving element compositions deviating from the optimal concentration. In this way, through the interaction of each term, the complex material relationship is converted into an operable optimization path, thereby greatly improving the design efficiency of the refractory high-entropy alloy. The higher the multi-dimensional evaluation value, the more excellent the mechanical properties of the corresponding refractory high-entropy alloy; and the refractory high-entropy alloy composition with the highest predicted mechanical property data is determined as the optimized refractory high-entropy alloy composition.
[0128] As shown in the foregoing example, taking the Cauchy pressure target prediction model as an example, Figure 2 The characteristic contribution value honeycomb chart and feature importance of the model are shown, in which the element contents of Hf, Nb, Mo and Ta have greater influence on the model output. At the same time, the key feature analysis results of the Cauchy pressure target prediction model and the Pugh ratio target prediction model and the Poisson ratio target prediction model described in the prior art show that the mixing enthalpy, valence electron concentration and melting point all have important influence in the three prediction models. In particular, as the mixing enthalpy increases, the corresponding Pugh ratio, Poisson ratio and Cauchy pressure all increase, indicating that a larger mixing enthalpy is conducive to improving the ductility of the alloy. In addition, in the Pugh ratio target prediction model, the element contents of Nb, Mo, Ti and Zr significantly affect the model output; in the Poisson ratio target prediction model, the element contents of Nb, Hf, Zr and Mo significantly affect the model output. At the same time, the specific influence of the element content on the mechanical properties is as follows: 1) Hf element: as the Hf content increases, the Cauchy pressure and Poisson ratio of the refractory high-entropy alloy (hereinafter referred to as alloy) slightly decrease, indicating that the relationship between the Hf content and the mechanical properties of the alloy is relatively complex; 2) Nb element: as the Nb content increases, the Cauchy pressure, Pugh ratio and Poisson ratio of the alloy all increase, indicating that the Nb element can significantly improve the ductility of the alloy; 3) Zr element: as the Zr content increases, the Pugh ratio and Poisson ratio of the alloy increase, indicating that the Zr element can effectively improve the ductility of the alloy; 4)
[0129] Mo element: as the Mo content increases, the Cauchy pressure increases but the Poisson ratio decreases, and the influence on the mechanical properties is relatively complex; 5) Ti element: as the Ti content increases, the Pugh ratio of the alloy increases, which is conducive to improving the ductility of the alloy.
[0130] In a preferred embodiment, as described in the prior art, when the preset requirement is to design a refractory high-entropy alloy composition with optimal ductility and containing Nb, the key elements include Nb, Hf, Mo, Ta, Ti and Zr; the second feature includes the mixing enthalpy feature; therefore, the following optimization design strategy for the refractory high-entropy alloy composition is proposed:
[0131] 1) Preferentially consider adjusting the alloy system with a larger mixing enthalpy, which is conducive to improving the ductility of the alloy;
[0132] 2) In terms of element selection, appropriately increasing the contents of Nb and Zr helps to significantly improve the ductility of the alloy;
[0133] 3) The adjustment of the Hf element content needs to be analyzed on a case-by-case basis, and the relationship between the Hf element content and the mechanical properties of the alloy is relatively complex;
[0134] 4) The adjustment of the Mo element content needs to consider its differential influence on different mechanical performance indicators;
[0135] 5) Under the premise of ensuring the alloy formability, appropriately increasing the Ti element content can improve the Pugh ratio of the alloy, thereby improving the ductility;
[0136] Therefore, based on the above-mentioned refractory high-entropy alloy composition optimization design strategy, the refractory high-entropy alloy composition set is NbTiZrMoTa with a molar ratio of 1:1:1:1:1, which is expected to have good ductility by increasing the contents of Nb, Zr and Ti; NbZrTiHfTa with a molar ratio of 1:1:1:1:1, which is expected to have good strength and ductility by increasing the contents of Nb, Zr and Ti while controlling the content of Hf; and Nb 1.5 ZrTiMoTa 0.5 , by non-equi-molar ratio design, further increasing the content of Nb and reducing the content of Ta, which is expected to have better ductility;
[0137] Finally, the above-mentioned Cauchy pressure target prediction model, Pugh ratio target prediction model and Poisson ratio target prediction model are used to predict the mechanical properties of the three refractory high-entropy alloy compositions, and the results are shown in Table 1:
[0138] Table 1
[0139] Alloy composition Predicted Pugh ratio Predicted Poisson ratio Predicted Cauchy pressure NbTiZrMoTa 2.07 0.332 42.8 NbZrTiHfTa 2.15 0.338 47.5 Nb 1.5 ZrTiMoTa 0.5 ]]> 2.23 0.345 51.3
[0140] As shown in Table 1, the designed refractory high-entropy alloys all meet the requirements of Pugh ratio > 1.75, Poisson ratio > 0.26 and positive Cauchy pressure, indicating good ductility. In particular, the Nb 1.5 ZrTiMoTa 0.5 alloy has the highest predicted values of Pugh ratio, Poisson ratio and Cauchy pressure, indicating the best ductility.
[0141] As shown in Figure 3 , Figure 4 , the present application provides a refractory high-entropy alloy composition optimization design device. The device embodiment can be realized by software, or by hardware or a combination of software and hardware. From the hardware layer, as shown in Figure 3 , a hardware architecture diagram of a computing device where the refractory high-entropy alloy composition optimization design device of the present application is located, in addition to the processor, memory, network interface and non-volatile memory shown in Figure 3 , the computing device where the device is located in the embodiment can also include other hardware, such as a forwarding chip responsible for processing packets, etc. Taking the software implementation as an example, as shown in Figure 4As shown, as a logical device, it is formed by the CPU of the computing device where it is located to read the corresponding computer program in the non-volatile memory into the memory for running. The embodiment provides a refractory high-entropy alloy component optimization design device, which comprises:
[0142] The construction module 400 is configured to establish a supercell model of the refractory high-entropy alloy component.
[0143] The calculation module 402 is configured to calculate the elastic matrix and mechanical property data of the refractory high-entropy alloy component after structural relaxation optimization of the supercell model.
[0144] The sample generation module 404 is configured to construct a sample set based on the element content, physical characteristic information and mechanical property data of the refractory high-entropy alloy component.
[0145] The training module 406 is configured to train a machine learning model using the sample set to obtain a plurality of target prediction models; wherein the sample set comprises element content and physical characteristic information as input and mechanical property data as output; the types of mechanical property data output by different target prediction models are different.
[0146] The optimization design module 408 is configured to perform interpretability analysis on the target prediction model to obtain key elements of each target prediction model, so as to determine an optimized refractory high-entropy alloy component according to the key elements.
[0147] In some specific embodiments, the construction module 400 can be configured to perform the above step 100, the calculation module 402 can be configured to perform the above step 102, the sample generation module 404 can be configured to perform the above step 104, the training module 406 can be configured to perform the above step 106, and the optimization design module 408 can be configured to perform the above step 108.
[0148] In some specific embodiments, the refractory high-entropy alloy component comprises at least four elements selected from Al, Si, Ti, Cr, Zr, Nb, Mo, Hf, Ta and W.
[0149] In some specific embodiments, the construction module 400 is further configured to perform the following operations:
[0150] The supercell model is established by using a quasi-random approximation method; wherein the supercell model is a BCC structure solid solution.
[0151] In some specific embodiments, the calculation module 402 is further configured to perform the following operations:
[0152] Determine parameters for structure relaxation optimization, and perform structure relaxation optimization based on the parameters to obtain a fully relaxed equilibrium lattice structure; wherein the parameters include a plane wave cutoff energy, a K-point grid, an energy convergence criterion, and a force convergence criterion;
[0153] Apply a strain to the equilibrium lattice structure to obtain a stress tensor;
[0154] Based on the stress tensor and the strain, calculate an elastic matrix;
[0155] According to the elastic matrix, calculate mechanical property data including bulk modulus, Young's modulus, shear modulus, Pugh ratio, Poisson's ratio, and Cauchy pressure.
[0156] In some specific embodiments, the physical characteristic information includes: atomic size difference, average electronegativity, electronegativity difference, valence electron concentration, melting point, thermal expansion coefficient, molar volume, mixing entropy, mixing enthalpy, entropy-enthalpy competition relationship, and isochoric specific heat.
[0157] In some specific embodiments, the training module 406 is further configured to perform the following operations:
[0158] For each type of mechanical property data, the following operations are performed:
[0159] Input the element content and the physical characteristic information in the sample set into the machine learning model to output predicted mechanical property data of the type;
[0160] Compare the predicted mechanical property data of the type with the mechanical property data of the type included in the sample set to obtain a root mean square error and a determination coefficient;
[0161] When the root mean square error and the determination coefficient both satisfy a preset threshold, training is completed to obtain a target prediction model for the type of mechanical property data.
[0162] In some specific embodiments, the optimization design module 408 is further configured to perform the following operations:
[0163] For each target prediction model, the following operations are performed:
[0164] Generate an interpreter and use the interpreter to calculate a feature contribution value of each feature included in each sample;
[0165] For each feature, the following operation is performed: calculate a mean value of the feature contribution value of the feature based on the feature contribution values of all samples including the feature;
[0166] Determine a feature corresponding to a mean value of the feature contribution value greater than a preset mean value threshold as a first feature;
[0167] Determine a contribution impact direction of the first feature;
[0168] determining the first feature with a positive contribution influence direction as a key feature, and determining an element corresponding to the key feature representing the element content as a key element;
[0169] calculating an interaction intensity index between any two features;
[0170] retaining a feature pair corresponding to the interaction intensity index greater than a preset index threshold;
[0171] determining an optimized refractory high-entropy alloy composition satisfying a preset requirement according to the key elements and feature pairs obtained by each target prediction model.
[0172] It can be understood that the structure of the embodiment of the present application does not constitute a specific limitation on the refractory high-entropy alloy composition optimization design device. In other embodiments of the present application, a refractory high-entropy alloy composition optimization design device can include more or fewer components than the illustration, or combine certain components, or split certain components, or different component arrangement. The illustrated components can be implemented in hardware, software, or a combination of software and hardware.
[0173] The information interaction, execution process, and the like between the modules in the above device are based on the same concept as the method embodiments of the present application, and the specific content can be referred to the description in the method embodiments of the present application, which will not be described here.
[0174] The embodiment of the present application also provides a computing device including a memory and a processor, the memory stores a computer program, and the processor implements the refractory high-entropy alloy composition optimization design method in any of the embodiments of the present application when executing the computer program.
[0175] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program makes the processor execute the refractory high-entropy alloy composition optimization design method in any of the embodiments of the present application when being executed by the processor.
[0176] The embodiment of the present application also provides a computer program product, which includes a computer program, and the processor of the computer device reads the computer program from the computer readable storage medium, and the processor executes the computer program, so that the computer device executes the refractory high-entropy alloy composition optimization design method in any of the above embodiments.
[0177] Specifically, a system or device equipped with a storage medium can be provided, and the storage medium stores a software program code for implementing the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device reads and executes the program code stored in the storage medium.
[0178] In this case, the program code itself read from the storage medium can implement the functions of any of the above-described embodiments, and thus the program code and the storage medium which stores the program code constitute a part of the present application.
[0179] Embodiments of storage media for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk such as a CD-ROM, CD-R, CD-RW, a DVD-ROM, a DVD-RAM, a DVD- RW, a DVD+RW, a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded to a server computer from an external package via a communication network, and is stored in the server computer, and then is downloaded to a user computer which is connected to the server computer via the network.
[0180] Further, it should be understood that, not only the program code read out from the computer, but also the operating system or the like operating on the computer based on the instructions of the program code can perform part or all of the actual operations to realize the functions of any of the above-described embodiments.
[0181] Further, it should be understood that, the program code read out from the storage medium can be written into the memory provided in the expansion board inserted into the computer or the memory provided in the expansion module connected to the computer, and then part or all of the actual operations can be performed based on the instructions of the program code by the CPU or the like mounted on the expansion board or the expansion module to realize the functions of any of the above-described embodiments.
[0182] It should be noted that the terms such as first and second, which are used herein merely to distinguish one entity or operation from another, do not necessarily require or imply that these entities or operations are different in nature or order. Also, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such process, method, article or apparatus. Without more limitations, the element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus including the element.
[0183] It should be understood by those of ordinary skill in the art that all or part of the steps of the above-described method embodiments can be completed by program instruction-related hardware, and the aforementioned program can be stored in a computer-readable storage medium, and the program performs the steps of the above-described method embodiments when executed; and the aforementioned storage medium includes ROM, RAM, magnetic disk or optical disk, and various media which can store program code.
[0184] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for optimizing the composition of a refractory high entropy alloy, characterized in that: include: Establish a supercell model of the composition of refractory high-entropy alloys; After performing structural relaxation optimization on the supercell model, the elastic matrix and mechanical property data of the refractory high entropy alloy composition are calculated; Constructing a sample set based on the element content, physical characteristic information and mechanical property data included in the composition of the refractory high entropy alloy; The sample set is used to train a machine learning model to obtain a plurality of target prediction models; wherein the sample set includes element content and physical characteristic information as input and mechanical property data as output; different target prediction models output different types of mechanical property data; An interpretability analysis is performed on the target prediction model to obtain key elements of each target prediction model, so as to determine the optimized refractory high entropy alloy composition based on the key elements.
2. The method according to claim 1, characterized in that The refractory high entropy alloy composition includes at least quaternary elements of Al, Si, Ti, Cr, Zr, Nb, Mo, Hf, Ta and W; and / or, The establishing of a supercell model of a refractory high entropy alloy composition comprises: The supercell model is established by using a quasi-random approximation method; wherein the supercell model is a BCC structure solid solution.
3. The method according to claim 1, characterized in that After performing structural relaxation optimization on the supercell model, calculating the elastic matrix and mechanical property data of the refractory high entropy alloy composition includes: Determining parameters for structural relaxation optimization, and performing structural relaxation optimization based on the parameters to obtain a fully relaxed equilibrium lattice structure; wherein the parameters include a plane wave cutoff energy, a K-point grid, an energy convergence criterion, and a force convergence criterion; applying strain to the equilibrium lattice structure to obtain a stress tensor; Calculating the elastic matrix based on the stress tensor and the strain; The mechanical property data including bulk modulus, Young's modulus, shear modulus, Pugh ratio, Poisson's ratio and Cauchy pressure are calculated based on the elastic matrix.
4. The method according to claim 1, wherein The physical characteristic information includes: atomic size difference, average electronegativity, electronegativity difference, valence electron concentration, melting point, thermal expansion coefficient, molar volume, mixing entropy, mixing enthalpy, entropy-enthalpy competition relationship and isochoric specific heat; and / or, The sample set is used to train the machine learning model to obtain several target prediction models, including: For each type of mechanical properties data, the following are performed: Inputting the element content and physical characteristic information in the sample set into the machine learning model, and outputting the predicted mechanical property data of this type; Comparing the predicted mechanical property data of the type with the mechanical property data of the type included in the sample set to obtain a root mean square error and a coefficient of determination; When both the root mean square error and the determination coefficient meet preset thresholds, the training is completed and a target prediction model for this type of mechanical property data is obtained.
5. The method according to any one of claims 1 to 4, characterized in that The interpretability analysis of the target prediction model is performed to obtain key elements of each target prediction model, so as to determine and optimize the composition of the refractory high entropy alloy according to the key elements, including: For each target prediction model, execute: Generate an interpreter, and use the interpreter to calculate the feature contribution value of each feature included in each sample; For each feature, the following steps are performed: calculating the mean feature contribution value of the feature based on the feature contribution values of all samples including the feature; Determine the feature corresponding to the feature contribution value mean greater than a preset mean threshold as the first feature; Determining the contribution influence direction of the first feature; The first feature with a positive contribution influence direction is determined as a key feature, and the element corresponding to the key feature representing the element content is determined as a key element; Calculate the interaction strength index between any two features; The feature pairs corresponding to the interaction strength index greater than the preset index threshold are retained; According to the key elements and the characteristic pairs obtained by each target prediction model, an optimized refractory high entropy alloy composition that meets preset requirements is determined.
6. The method according to claim 5, characterized in that When the preset requirement is to design a refractory high entropy alloy composition with optimal ductility and containing Nb, the key elements include Nb, Hf, Mo, Ta, Ti and Zr; the second characteristic includes a mixing enthalpy characteristic; the optimized refractory high entropy alloy composition is NbTiZrMoTa with a molar ratio of 1:1:1:1:1, NbZrTiHfTa with a molar ratio of 1:1:1:1:1 or Nb with a molar ratio of 1.5:1:1:1:0.
5. 1.5 ZrTiMoTa 0.5 .
7. A device for optimizing the composition of refractory high entropy alloys, characterized in that: include: Building blocks for constructing supercell models of refractory high-entropy alloy compositions; A calculation module, configured to calculate the elastic matrix and mechanical property data of the refractory high entropy alloy composition after performing structural relaxation optimization on the supercell model; A sample generation module, configured to construct a sample set based on the element content, physical characteristic information, and mechanical property data of the refractory high entropy alloy composition; a training module for training a machine learning model using the sample set to obtain a plurality of target prediction models; wherein the sample set includes element content and physical characteristic information as input and mechanical property data as output; different target prediction models output different types of mechanical property data; The optimization design module is used to perform interpretability analysis on the target prediction model to obtain the key elements of each target prediction model, so as to determine the optimized refractory high entropy alloy composition based on the key elements.
8. A computing device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to execute the method according to any one of claims 1 to 6.
10. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 6.