Design and preparation method of hydrogen embrittlement-resistant high-toughness double-phase high-entropy alloy

Abstract

The invention discloses a design and preparation method of an anti-hydrogen embrittlement high-toughness double-phase high-entropy alloy. A double-phase system of an FCC base phase and an AxBy precipitated phase is screened through phase diagram calculation, fault energy and binding energy are calculated in combination with molecular dynamics, hydrogen trap binding energy is calculated according to a first principle, multi-scale data are integrated to construct a nonlinear expression, and a hydrogen embrittlement sensitivity factor is determined; according to the method, a multi-modal database containing experimental and multi-scale calculation data is established, an independent index prediction model is constructed by adopting a machine learning algorithm after preprocessing, a multi-layer data chain relation is analyzed through multi-modal fusion, comprehensive performance indexes are calculated, and finally a comprehensive performance synchronous optimization prediction model is established by utilizing Bayesian optimization. And through iterative screening and experimental verification, the hydrogen embrittlement-resistant high-toughness double-phase high-entropy alloy is prepared. Multi-scale collaborative analysis and performance collaborative improvement are achieved, the alloy design efficiency and precision are improved, and the prepared alloy has important application value in the fields of hydrogen energy, nuclear energy and the like.

Description

Technical Field

[0001] This invention relates to the field of metallic materials and computer-aided design technology, specifically to the design and preparation method of a high-strength, high-toughness, dual-phase high-entropy alloy resistant to hydrogen embrittlement. Background Technology

[0002] In key national strategic areas such as the hydrogen energy industry chain (hydrogen storage, transportation, and hydrogen fuel cells), nuclear energy engineering (e.g., fuel cladding, pressure vessels), and marine engineering equipment (e.g., offshore oil and gas platforms, subsea oil pipelines), currently used alloys (such as nickel-based alloys GH3536 and Ni718), while possessing excellent mechanical properties, suffer from insufficient resistance to hydrogen embrittlement, failing to meet practical needs. There is an urgent need to develop new alloy materials with superior resistance to hydrogen embrittlement. High-entropy alloys, with their unique multi-component design and outstanding mechanical properties, have become an important research direction for next-generation structural materials and are expected to be applied as novel alloy materials with high strength, toughness, and excellent resistance to hydrogen embrittlement in the aforementioned key areas.

[0003] In the field of materials science, researchers are dedicated to developing face-centered cubic (FCC) single-phase high-entropy alloys with good resistance to hydrogen embrittlement. However, the strength of these alloys often falls short of expectations. To improve the strength and toughness of high-entropy alloys, a second-phase precipitation strengthening strategy has been introduced, leading to the successful development of a series of two-phase high-entropy alloys, represented by high-performance second-phase precipitation-strengthened high-entropy alloys. Through various strengthening mechanisms of the precipitated second phase, such as solid solution strengthening, interface strengthening, strain distribution, and the synergistic effect of multiple deformation mechanisms, the traditional "inverted relationship" between strength and plasticity has been effectively overcome. Furthermore, precipitated two-phase high-entropy alloys exhibit unique advantages in terms of strengthening, corrosion resistance, and resistance to hydrogen embrittlement.

[0004] First-principles calculations, molecular dynamics, and phase diagram calculations can perform computational analysis under physical models at different scales, offering advantages such as cross-scale mechanistic analysis and optimization, extreme environment adaptability design, and multiphase / heterogeneous structure optimization. However, they also suffer from drawbacks such as high computational complexity and resource consumption, and difficulty in analyzing cross-scale models. Developing high-entropy alloys using artificial intelligence can demonstrate significant advantages in efficient composition design and screening, accurate performance prediction and mechanistic analysis, cross-scale correlation and multi-objective optimization, and reduced experimental costs and trial-and-error risks. However, it suffers from high dependence on sample data volume and insufficient model generalization ability.

[0005] In summary, while multi-scale computation and artificial intelligence methods each have their advantages in materials research, they also have limitations. Using either method alone is insufficient to effectively solve the design challenges of precipitated dual-phase high-entropy alloys, especially in achieving a synergistic improvement in strength and toughness versus resistance to hydrogen embrittlement. Therefore, it is necessary to develop a novel approach for designing high-strength, high-entropy dual-phase high-entropy alloys with resistance to hydrogen embrittlement, based on a new coupling of multi-scale computation and artificial intelligence methods. This approach aims to overcome the shortcomings of single methods, achieve a synergistic improvement in strength and toughness versus resistance to hydrogen embrittlement, and break through current research bottlenecks. Summary of the Invention

[0006] This invention aims to provide a design and preparation method for a high-strength, high-toughness, dual-phase high-entropy alloy resistant to hydrogen embrittlement. By establishing a multilayer structure prediction model, alloy compositions and process parameters that meet the requirements are gradually screened out, thereby achieving efficient and accurate design of high-performance high-entropy alloys.

[0007] The technical solution to achieve the objective of this invention is:

[0008] A design method for a high-strength, high-toughness, dual-phase high-entropy alloy resistant to hydrogen embrittlement includes the following steps:

[0009] Phase diagram calculation and candidate system screening: The phase diagram calculation method was used to screen the "FCC matrix phase + A" of high-entropy alloys in different systems. x B y Precipitated phases (e.g., Ni3Al, TiAl, High-throughput thermodynamic calculations were performed on the composition and temperature range of the two-phase region (etc.) to obtain alloy phase information.

[0010] Molecular dynamics simulations: Applying molecular dynamics methods to different "FCC matrix phases + A" x B y "The stacking fault energy of the FCC matrix phase in the dual-phase high-entropy alloy was calculated; and the A-phase of different sizes and shapes was calculated using LAMMPS." x B y The binding energy of the phase reflects the bonding strength between the second phase and the base metal.

[0011] First-principles calculations: A for different structural types x B y For the second phase, first-principles density functional theory was used to optimize the matrix and interface, and the hydrogen trap binding energy of different types of second phase morphologies was calculated.

[0012] Multi-scale computational data integration: Integrating high-throughput computational data on alloy phase information, molecular dynamics stacking fault energy, binding energy, and first-principles hydrogen trap binding energy.

[0013] Nonlinear expression establishment: Using symbolic regression modeling, nonlinear expressions are established for stacking fault energy SFE, binding energy BE, and hydrogen trap binding energy HSE.

[0014] Determination of hydrogen embrittlement sensitivity factor: Iterative correction was performed based on the hydrogen embrittlement extension loss rate in actual experiments, and the "hydrogen embrittlement sensitivity factor" evaluation index was finally obtained.

[0015] Multimodal database construction: Integrating real experimental data of high-entropy alloys, thermodynamic phase diagrams, molecular dynamics stacking fault energy data, first-principles stacking fault energy / hydrogen trap binding energy data, and hydrogen embrittlement sensitivity factor data, a multimodal biphase high-entropy alloy genetic engineering database containing phase diagram calculations / first-principles / molecular dynamics / experimental reports is established.

[0016] Database preprocessing: The multimodal biphase high-entropy alloy genetic engineering database is preprocessed to map it to a unified dimension space and establish the target domain data.

[0017] Independent index prediction model establishment: Using a variety of artificial intelligence machine learning algorithms, an independent index prediction model for the microstructure, structure and properties of precipitated dual-phase high-entropy alloys was established.

[0018] Multi-layer data chain relationship analysis and comprehensive performance index calculation: The multi-modal model fusion method is used to analyze the multi-layer data chain structural relationships such as "composition-process-microstructure", "composition-process-structure", "composition-process-performance" and "composition-process-hydrogen embrittlement sensitivity factor", and calculate the comprehensive performance index of "microstructure-structure-performance-hydrogen embrittlement sensitivity factor" under different "composition-process" conditions.

[0019] Establishment of a comprehensive performance synchronous optimization prediction model: A comprehensive performance synchronous optimization prediction model for "composition-process-microstructure-performance-hydrogen embrittlement sensitivity factor" is further established using the Bayesian optimization algorithm.

[0020] Composition and process parameter optimization and screening: Construct a large set of composition and process parameters, set target values ​​for alloy performance, and use the constructed comprehensive performance synchronous optimization prediction model to perform multiple iterative cycles (prediction-experiment-feedback) on the screened composition and process parameter set to further optimize and screen, and finally determine the optimal range of comprehensive performance.

[0021] Experimental verification: The composition and process parameters within the range of optimal comprehensive performance were selected for experimental verification.

[0022] Alloy preparation: A dual-phase high-entropy alloy with composition-process parameters that meet the target performance is prepared by vacuum arc furnace melting and cold and hot working processes.

[0023] Performance characterization and database update: The high-entropy alloys obtained were characterized by various target performance tests, and the test data were updated to the multimodal dual-phase high-entropy alloy genetic engineering database.

[0024] The dual-phase high-entropy alloys obtained by the above method are "FCC+Ni3Al" dual-phase high-entropy alloy and "FCC+NiAl" dual-phase high-entropy alloy; their compositions by atomic percentage are as follows:

[0025] FCC+Ni3Al dual-phase high-entropy alloy: Fe=23.25%, Co=29.76%, Ni=39.99%, Al=2%, Ti=5%;

[0026] FCC+NiAl dual-phase high-entropy alloy: Fe=18.8%, Co=18.8%, Ni=28.6%, Cr=18.8%, Al=15%.

[0027] The preparation methods of the "FCC+Ni3Al" dual-phase high-entropy alloy and the "FCC+NiAl" dual-phase high-entropy alloy include the following steps:

[0028] (1) Weigh each alloying element raw material according to the atomic percentage converted to the weight percentage;

[0029] (2) The alloy was prepared by using the HVAX-3 simple non-consumable vacuum arc melting equipment. The alloy was melted under an argon protective atmosphere and the melting was repeated 5 times. Before each melting, the ingot was turned over. The arc current during the melting process was 400mA and the electromagnetic stirring was turned on.

[0030] (3) The prepared high-entropy alloy ingots were subjected to homogenization heat treatment in a tubular atmosphere furnace. The homogenization heat treatment process for the two types of hydrogen embrittlement resistant, high-strength and tough dual-phase high-entropy alloys with precipitated second phases, namely "FCC+Ni3Al" and "FCC+NiAl", is as follows:

[0031] FCC+Ni3Al dual-phase high-entropy alloy: 1150℃ / 12 hours / air cooling;

[0032] FCC+NiAl dual-phase high-entropy alloy: 1200℃ / 6 hours / air cooling;

[0033] (4) Using a twin-roll mill, the high-entropy alloy after homogenization heat treatment is subjected to deformation treatment. The rolling processes are as follows:

[0034] FCC+Ni3Al dual-phase high-entropy alloy: cold-rolled, with a reduction of 70%;

[0035] FCC+NiAl dual-phase high-entropy alloy: hot rolled at 1150℃ with a reduction of 30%;

[0036] (5) The rolled high-entropy alloy samples were subjected to recrystallization annealing heat treatment. The heat treatment processes were as follows:

[0037] FCC+Ni3Al dual-phase high-entropy alloy: 1200℃ / 2 minutes / water cooling;

[0038] FCC+NiAl dual-phase high-entropy alloy: 1100℃ / 5 minutes / water cooling;

[0039] (6) For the high-entropy alloy after recrystallization annealing heat treatment, a second-phase precipitation aging heat treatment is performed. The heat treatment processes are as follows:

[0040] FCC+Ni3Al dual-phase high-entropy alloy: 800℃ / 6 hours / water cooling;

[0041] FCC+NiAl dual-phase high-entropy alloy: 650℃ / 4 hours / water cooling;

[0042] (7) After the above steps, the preparation of the hydrogen embrittlement resistant, high strength and toughness dual-phase high entropy alloy is completed.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. Multi-scale collaborative analysis: It can perform rapid calculations and searches from three dimensions of the alloy: macroscopic, mesoscopic and atomic scales, to fully understand the physical characteristics and properties of the material at different scales.

[0045] 2. Synergistic performance improvement: A method for synergistically improving the hydrogen embrittlement resistance and toughness of high-entropy alloys was developed, and a high-strength and high-toughness dual-phase high-entropy alloy with hydrogen embrittlement resistance was successfully designed, breaking through the limitation that traditional alloys cannot simultaneously achieve both strength and hydrogen embrittlement resistance.

[0046] 3. Highly efficient and precise design: Through the collaboration of multi-scale calculation and artificial intelligence, a quantitative relationship of "composition-process-microstructure-performance" is established, breaking through the limitations of traditional trial and error methods, greatly improving the efficiency and accuracy of alloy design, and reducing experimental costs and trial and error risks.

[0047] 4. Continuous optimization of the database: The multimodal database is continuously updated through experimental verification, which continuously improves the model's predictive ability and provides strong support for the design of more high-performance alloys in the future.

[0048] 5. Performance tests show that the FCC+Ni3Al dual-phase high-entropy alloy has a yield strength >1100MPa, elongation >25%, and hydrogen embrittlement elongation loss rate <10%; the FCC+NiAl dual-phase high-entropy alloy has a yield strength >1350MPa, elongation >25%, and hydrogen embrittlement elongation loss rate <10%. Attached Figure Description

[0049] Figure 1 A flowchart for the design of a high-strength, high-toughness, dual-phase high-entropy alloy resistant to hydrogen embrittlement;

[0050] Figure 2 This is a schematic diagram of a multimodal model fusion method.

[0051] Figure 3 This is a schematic diagram of the prediction effect of the test set and training set of the high-entropy alloy comprehensive performance (hydrogen embrittlement resistance, high strength and toughness) prediction model based on multimodal model fusion under different algorithms in the embodiment.

[0052] Figure 4 The example uses a multimodal model fusion model to predict the comprehensive performance (hydrogen embrittlement resistance, high strength and toughness) of high-entropy alloys under different algorithms. 2 Schematic diagram;

[0053] Figure 5 This is a schematic diagram of the RMSE prediction model for the comprehensive performance (hydrogen embrittlement resistance, high strength and toughness) of high-entropy alloys based on multimodal model fusion under different algorithms in the embodiment.

[0054] Figure 6 The graph shows the strength, plasticity, and electrochemical hydrogen-charged tensile hydrogen embrittlement elongation loss rate of the high-strength and tough FCC+Ni3Al high-entropy alloy in the examples.

[0055] Figure 7 The graph shows the strength, plasticity, and electrochemical hydrogen-charged tensile hydrogen embrittlement elongation loss rate of the high-strength and tough FCC+NiAl high-entropy alloy in the examples.

[0056] Figure 8 The image shows the microstructure of the hydrogen embrittlement resistant, high-strength and tough FCC+Ni3Al high-entropy alloy in the examples.

[0057] Figure 9 The image shows the microstructure of the hydrogen embrittlement resistant, high-strength and tough FCC+NiAl high-entropy alloy in the examples. Detailed Implementation

[0058] This invention provides an evaluation method for high-strength, high-entropy, hydrogen-embrittlement-resistant dual-phase high-entropy alloys, such as... Figure 1 As shown, the specific steps include:

[0059] Step 1: Phase diagram calculation and candidate system screening

[0060] Using the commercial Pandat phase diagram calculation software and the PanHEA2022_TH+M database, we designed enhanced phase stability criteria by leveraging Pandat's high-throughput calculation module in conjunction with thermodynamic parameters such as mixing enthalpy, atomic size difference, and valence electron concentration. High-throughput calculations of various isothermal cross-sectional phase diagrams, longitudinal cross-sectional phase diagrams, and alloy thermodynamic properties were performed on multiple series of high-entropy alloys.

[0061] For high-entropy alloy systems containing elements such as Ni-Fe-Co-Cr-Al-Ti-Zr-V-Nb-Zr-Mn-Hf-La-Ta-WY-Cu, the "FCC matrix phase + A" method was developed. x B ySecond phase (e.g., Ni3Al, NiAl, TiAl, , ,FeAl, , TiFe , , Thermodynamic calculations of the composition and temperature range of the two-phase region were performed to obtain the ratio of the FCC phase to the second phase under different alloying elements and their addition amounts.

[0062] Based on the ratio of FCC phase to the second phase, combined with information such as alloy melting point, solid solubility lines, temperature changes, and equilibrium and non-equilibrium solidification processes, the output is characterized by "FCC+A". x B y High-throughput computational thermodynamic data for alloy systems with a "precipitated phase" two-phase structure, including: ① Effective composition points: recording the composition coordinates that satisfy the two-phase condition, such as... _zAlTi; ②Temperature window: Marks the temperature range in which the two phases are stable, such as the TiAl precipitate being stable at 800–1100℃.

[0063] Step 2: Molecular dynamics simulation calculations

[0064] The Atomsk and LAMMPS software were used to target different "FCC base phase + A" x B y Molecular dynamics calculations were performed on the stacking fault energy of the FCC matrix phase in a dual-phase high-entropy alloy.

[0065] A polycrystalline structure of FCC matrix was created, generating grains with different crystal orientations and grain boundaries. The two-phase system was optimized to the lowest energy state and the bonding interface was ensured to be stable. Based on the optimized structure, strain or shear was applied to induce slip in the two-phase contact region to simulate stacking fault formation. The stacking fault energy was calculated by comparing the total energy change of the system with that of the initial state without stacking faults and after the introduction of stacking faults.

[0066] For the polycrystalline FCC matrix constructed above, LAMMPS was used to perform high-throughput calculations of A-values ​​for different sizes and shapes by selecting appropriate potential functions (such as EAM potential and machine learning potential functions). x B y The binding energy of the phase reflects the bonding strength between the second phase and the base metal.

[0067] Step 3: First-principles calculations

[0068] For different structural types of A x B y For the second phase, matrix and interface optimization were performed using first-principles density functional theory, and the hydrogen trap binding energy of different types of second phase morphologies was calculated using high-throughput methods.

[0069] We used software such as VESTA to construct second-phase models with different structures, sizes, and shapes and docked them with the matrix metal model. For each type of second-phase structure, we used GGA-PBE functional optimization to optimize the interface structure between the matrix metal and the second phase to obtain a stable interface. We focused on interface lattice matching, atomic positions, and stress distribution to ensure interface stability and reliability.

[0070] After optimizing the interface between the matrix and the second phase, hydrogen atoms are introduced and the hydrogen trapping binding energy is calculated. This binding energy reflects the strength of the binding between hydrogen atoms and the material interface and the material's ability to trap hydrogen. By calculating the hydrogen trapping binding energy of different second phases, their hydrogen adsorption and capture capabilities are evaluated, and the hydrogen trapping binding energies of different second phases are compared and analyzed to select the optimal type.

[0071] Step 4: Multi-scale computational data integration

[0072] A physical characteristic attribute data matrix was constructed, encompassing mixing enthalpy ΔHmix, mixing entropy ΔSmix, atomic radius difference δ, valence electron concentration VEC, electronegativity difference Δχ, phase structure recombination parameter Λ, thermodynamic stability, stacking fault energy, binding energy, hydrogen trap binding energy recombination parameter, etc.

[0073] Step 5: Establishing the nonlinear expression

[0074] Using symbolic regression modeling, a three-layer genetic algorithm was constructed based on the PySR toolkit. The initial function basis covered polynomial, exponential, and logarithmic forms, and the fitness function adopted the Bayesian information criterion (BIC). Nonlinear expressions were established for the stacking fault energy SFE, binding energy BE, and hydrogen trap binding energy HSE.

[0075] Basic operators: binary_operators = ["+", "-", "*", " / ", "^"]

[0076] Extended function form: unary_operators = ["exp", "log", "abs", "sqrt"]

[0077] Bayesian Information Criterion: BIC = nln +k lnn

[0078] Preliminary expression for hydrogen embrittlement sensitivity factor:

[0079]

[0080] Step 6: Determination of hydrogen embrittlement sensitivity factors

[0081] By iteratively correcting the hydrogen embrittlement extension loss rate (δ) from actual experiments, the final evaluation index, the "Hydrogen Embrittlement Sensitivity Factor (HESF)," is obtained.

[0082] Initial hydrogen embrittlement sensitivity factor:

[0083] Ultimate hydrogen embrittlement sensitivity factor:

[0084] Corrective measures:

[0085] Corrected hydrogen embrittlement sensitivity factor:

[0086]

[0087] Specifically, when the high-entropy alloy HESF < 1.5 (δ < 5%), the alloy has low hydrogen embrittlement sensitivity (δ < 10%); when 1.5 < HESF < 2.8, the alloy has medium hydrogen embrittlement sensitivity (10% < δ < 15%); and when the alloy HESF > 2.8, the alloy has high hydrogen embrittlement sensitivity and high risk (δ > 15%).

[0088] Step 7: Building a Multimodal Database

[0089] By integrating real experimental data on high-entropy alloys [composition-process-microstructure (grain size, second phase fraction, size, etc.)-structure-mechanical properties-hydrogen embrittlement elongation loss rate], thermodynamic phase diagrams, molecular dynamics stacking fault energy data, first-principles stacking fault energy / hydrogen trap binding energy data, and hydrogen embrittlement sensitivity factor data, a multimodal two-phase high-entropy alloy genetic engineering database containing phase diagram calculations / first-principles / molecular dynamics / experimental reports is established.

[0090] Step 8: Database Preprocessing

[0091] The multimodal dual-phase high-entropy alloy genetic engineering database is preprocessed using data standardization / normalization methods, mapping it to a unified dimension space to establish target domain data: the database is dimensionless to obtain standardized index variables (i is the amount of data, m is the number of material performance indicators).

[0092] ,in , A standardized data matrix is ​​obtained. ,in ;

[0093] Step 9: Establishing an Independent Indicator Prediction Model

[0094] A variety of machine learning algorithms, including ridge regression, least squares method, decision tree, support vector machine, random forest, boosting tree, extreme gradient boosting, artificial neural network and K-neighbor, as well as multi-model fusion algorithm, were used to establish independent index prediction models for the microstructure, structure and properties of precipitated biphase high-entropy alloys.

[0095] Step 10: Multi-layer data link relationship analysis and comprehensive performance index calculation

[0096] The multimodal model fusion method is used to analyze the multi-layer data chain structure relationships such as “composition-process-microstructure”, “composition-process-structure”, “composition-process-performance”, and “composition-process-hydrogen embrittlement sensitivity factor”, and calculate the comprehensive performance index of “microstructure-structure-performance-hydrogen embrittlement sensitivity factor” under different “composition-process” conditions.

[0097] Step 11: Establishment of a comprehensive performance synchronous optimization prediction model

[0098] A Bayesian optimization algorithm is used to proxy the fusion model of comprehensive performance indicators. Further iterative optimization is then performed to establish a comprehensive performance synchronous optimization and prediction model encompassing "composition, process, microstructure, performance, and hydrogen embrittlement sensitivity factor." The specific steps are as follows:

[0099] (1) Initialize the fusion model, use the rbf kernel function, and analyze x. i The corresponding value f(x) i The surrogate model is constructed based on the multivariate Gaussian distribution of .

[0100] (2) Calculate f(x) using the conditional distribution property of the Gaussian process. i The posterior mean and variance of the target function are used to predict new points and complete the probabilistic model of the objective function.

[0101] (3) The acquisition function is the expected improvement function (EI), which measures the expected improvement of f(x) over the current best value f(x+) (smaller, for minimization problems), and selects the next evaluation point.

[0102] (4) Use the x with the largest acquisition function value as the next observation data x. n +, and obtain the true function value f(x) n +1), then repeat the above steps; if the budget (such as the number of iterations) or convergence condition (such as the function value changing very little) is reached, stop; otherwise, update the model and repeat the operation.

[0103] A schematic diagram of the multimodal model fusion method is shown below. Figure 2 As shown, the prediction results of the test set and training set of the high-entropy alloy comprehensive performance (hydrogen embrittlement resistance, high strength and toughness) prediction model based on multimodal model fusion under different algorithms are as follows: Figure 3 As shown.

[0104] Step 12: Optimization and screening of components and process parameters

[0105] Based on the phase diagram, the composition range of different second phases is calculated. The Random algorithm tool is used to construct a candidate composition and process parameter set containing tens of millions of data points within the data range of composition and process. The target value of alloy performance is set. The comprehensive performance synchronous optimization and prediction model of "composition-process-microstructure-performance" is used to perform multiple iterative calculations on the candidate composition and process parameter set, further optimize and screen, and finally determine the region with the best comprehensive performance.

[0106] Step 13: Experimental Verification and Alloy Preparation

[0107] I. Model Construction and Validation

[0108] Based on the multimodal biphase high-entropy alloy genetic engineering database, four machine learning algorithms—support vector machine, random forest, extreme gradient boosting, and deep neural network—were employed. A multimodal model fusion method was used to establish a fusion model for the comprehensive performance indicators of biphase high-entropy alloys. The specific process is as follows:

[0109] ① Perform heterogeneous modal feature mapping based on the standardized data matrix:

[0110] ,in Let be the projection matrix, and d be the uniform dimension.

[0111] ② Cross-Attention Settings:

[0112] Query / Key / Value Transformation:

[0113] Attention weight calculation (scaled dot product): ,in , , Normalize along the bond dimension;

[0114] ③ Joint mapping of shared weight matrices:

[0115] ④ Calculate the comprehensive performance index based on the joint mapping data matrix: Establish a fusion model.

[0116] like Figure 4 and Figure 5As shown, the experimental results demonstrate that the multimodal fusion limit gradient enhancement model performs excellently in predicting the comprehensive properties (hydrogen embrittlement resistance, high strength and toughness) of high-entropy alloys, with a coefficient of determination R² of 0.9489 and a root mean square error RMSE of 0.3485.

[0117] II. Alloy Preparation and Performance Testing

[0118] Based on the phase diagram, the compositional range with the presence of the FCC+ second phase is calculated. A comprehensive performance optimization and prediction model integrating composition, process, microstructure, and properties is used to perform multiple iterative calculations on the candidate composition and process parameter set. This yields the optimal composition and process for the high-entropy alloy that meets the comprehensive performance requirements. The steps for synthesizing the predicted target alloy are as follows:

[0119] Step (1) Select the elemental raw materials to be used, such as Fe (block), Co (block), Ni (granular), Cr (flaky), Al (granular), Ti (granular), all with a purity higher than 99.9%; before weighing the ingredients, ultrasonically clean the raw materials in anhydrous ethanol for 20 minutes.

[0120] Step (2) According to the selected alloy composition, after drying, use an electronic balance with an accuracy of 0.0001g to prepare the alloy according to the predicted atomic percentage. After batching, put the raw materials into a 45℃ vacuum oven for storage to prevent oxidation.

[0121] Step (3) Alloy melting preparation is carried out using HVAX-3 simple non-consumable vacuum arc melting equipment: the copper crucible is wiped clean with 1000-mesh sandpaper and a dust-free cloth, and then the weighed metal raw material is placed in the copper crucible.

[0122] Step (4) Turn on the power to the mechanical pump and open the baffle valve to evacuate the furnace cavity; turn on the diffusion pump preheating switch; when the vacuum level reaches... At this time, close the mechanical pump baffle valve and turn on the diffusion pump solenoid plate to continue pumping high vacuum until... Turn off the vacuum system.

[0123] Step (5) Introduce high-purity argon gas and melt the alloy under the protective atmosphere of argon gas. To ensure the quality of the alloy, repeat the melting process 5 times. Before each melting, the ingot is turned over. The arc current during the melting process is 400mA, and the electromagnetic stirring is turned on.

[0124] Step (6) involves homogenizing the prepared high-entropy alloy ingot using a tubular atmosphere furnace.

[0125] Step (7) involves recrystallizing and aging annealing the rolled sample to obtain a dual-phase high-entropy alloy that meets the set requirements.

[0126] Based on the above implementation steps, the information on the two types of second-phase alloys with high strength and toughness against hydrogen embrittlement, designed and prepared, is as follows:

[0127] FCC+Ni3Al dual-phase high-entropy alloy:

[0128] FCC+NiAl dual-phase high-entropy alloy:

[0129] Performance test results: The hydrogen embrittlement elongation loss rate of the FCC+Ni3Al high-entropy alloy is 8.1%, such as... Figure 6 As shown; the hydrogen embrittlement elongation loss rate of FCC+NiAl is 9.3%, as... Figure 7 As shown in the figure; the results show that both types of high-entropy alloys exhibit excellent resistance to hydrogen embrittlement and high toughness, demonstrating the high efficiency of the synergistic optimization design of the comprehensive performance of high-entropy alloys with high strength and toughness against hydrogen embrittlement in this invention. The SEM images of the microstructure of the FCC+Ni;Al and FCC+NiAl high-entropy alloys are shown in the figure. Figure 8 and Figure 9 As shown.

[0130] Step 14: Performance Characterization and Database Update

[0131] The high-entropy alloys obtained were characterized by their mechanical properties, hydrogen embrittlement resistance, and other target properties. The test data were then updated to the multimodal dual-phase high-entropy alloy genetic engineering database to improve the model's prediction accuracy and generalization ability.

Claims

1. A design method for a high-strength, high-toughness, dual-phase high-entropy alloy resistant to hydrogen embrittlement, characterized in that, It includes the following steps: 1) Using phase diagram calculations, the "FCC matrix phase + A" of high-entropy alloys in different systems were analyzed. x B y High-throughput thermodynamic calculations were performed on the composition and temperature range of the two-phase region of the precipitated phase to obtain alloy phase information. 2) Applying molecular dynamics methods to calculate different "FCC matrix phases + A x B y The stacking fault energy of the FCC matrix phase in a two-phase high-entropy alloy with precipitated phases was determined, and the stacking fault energies of different sizes and shapes of A phases were calculated using LAMMPS. x B y Phase binding energy; 3) For different structural types of A x B y For the second phase, first-principles density functional theory was used to optimize the matrix and interface, and the hydrogen trap binding energy of different types of second phases was calculated. 4) Integrate the high-throughput calculation data of alloy phase information, molecular dynamics stacking fault energy, binding energy, and first-principles hydrogen trap binding energy; 5) Use symbolic regression modeling to establish non-linear expressions for stacking fault energy SFE, binding energy BE, and hydrogen trap binding energy HSE; 6) Combine the actual test hydrogen embrittlement elongation loss rate for iterative correction to obtain the "hydrogen embrittlement sensitivity factor" evaluation index; 7) Integrate the real test data of high-entropy alloys, thermodynamic phase diagrams, molecular dynamics stacking fault energy data, first-principles stacking fault energy / hydrogen trap binding energy data, and hydrogen embrittlement sensitivity factor data to establish a multi-modal dual-phase high-entropy alloy genetic engineering database; 8) Pretreat the multi-modal dual-phase high-entropy alloy genetic engineering database, map it to a unified dimension space, and establish target domain data; 9) Adopt a variety of artificial intelligence machine learning algorithms to establish an independent index prediction model for the microstructure, structure, and properties of the precipitation-type dual-phase high-entropy alloy; 10) Use the multi-modal model fusion method to analyze the structural relationship of multi-layer data chains and calculate the comprehensive performance index; 11) Adopt the Bayesian optimization algorithm to establish a comprehensive performance synchronous optimization prediction model of "composition-process-microstructure-structure-property-hydrogen embrittlement sensitivity factor"; 12) Construct a set of composition and process parameters, and use the comprehensive performance synchronous optimization prediction model for cyclic iterative optimization screening to determine the optimal interval of comprehensive performance; 13) Select the composition and process parameters in the optimal interval for experimental verification, prepare the dual-phase high-entropy alloy and conduct performance characterization, and update the test data to the multi-modal database.

2. The method according to claim 1, characterized in that, The phase diagram calculation in step 1) uses Pandat phase diagram calculation software and the PanHEA2022_TH+M database. It combines mixing enthalpy, atomic size difference, and valence electron concentration thermodynamic parameters to design criteria for enhancing phase stability. The A... x B y The precipitated phase includes at least one of Ni3Al, NiAl, TiAl, ZrV2, Ni3Ti, FeAl, Zr2Fe, LaNi5, TiFe, Co3Ti, Fe3Al, and Zr3Al.

3. The method according to claim 1, characterized in that, Step 2) involves performing molecular dynamics calculations using Atomsk and LAMMPS software, calculating A by selecting the EAM potential or a machine learning potential function. x B y Phase binding energy; In step 3), the VESTA software is used to construct the second-phase model, the GGA-PBE functional is used to optimize the interface structure, and hydrogen atoms are introduced to calculate the hydrogen trap binding energy.

4. The method according to claim 1, characterized in that, In step 5), a three-layer genetic algorithm is constructed based on the PySR toolkit, the initial function basis covers polynomial, exponential, and logarithmic forms, and the Bayesian Information Criterion (BIC) is selected as the fitness function.

5. The method according to claim 1, characterized in that, The corrected expression of the hydrogen embrittlement sensitivity factor (HESF) in step 6) is: where when HESF < 1.5, the alloy has low hydrogen embrittlement sensitivity; when 1.5 < HESF < 2.8, the alloy has medium hydrogen embrittlement sensitivity; when HESF > 2.8, the alloy has high hydrogen embrittlement sensitivity.

6. The method according to claim 1, characterized in that, The machine learning algorithms in step 9) include at least one of ridge regression, least squares method, decision tree, support vector machine, random forest, boosting tree, extreme gradient boosting, artificial neural network, and K-nearest neighbors.

7. The method according to claim 1, characterized in that, The Bayesian optimization algorithm in step 11) constructs a surrogate model using the rbf kernel function, calculates the posterior mean and variance through the conditional distribution of the Gaussian process, and selects the next evaluation point with the Expected Improvement function (EI).

8. A dual-phase high-entropy alloy, characterized in that, Obtained by the method according to any one of claims 1-7.

9. The dual-phase high-entropy alloy according to claim 8, characterized in that, The dual-phase high-entropy alloy is the "FCC + Ni3Al" dual-phase high-entropy alloy and the "FCC + NiAl" dual-phase high-entropy alloy; by atomic percentage, its composition is as follows: FCC + Ni3Al dual-phase high-entropy alloy: Fe = 23.25%, Co = 29.76%, Ni = 39.99%, Al = 2%, Ti = 5%; FCC+NiAl dual-phase high-entropy alloy: Fe=18.8%, Co=18.8%, Ni=28.6%, Cr=18.8%, Al=15%.

10. A dual-phase high-entropy alloy according to claim 9, characterized in that, The preparation methods of the "FCC+Ni3Al" dual-phase high-entropy alloy and the "FCC+NiAl" dual-phase high-entropy alloy include the following steps: (1) Weigh each alloying element raw material according to the atomic percentage converted to the weight percentage; (2) The alloy was prepared by using the HVAX-3 simple non-consumable vacuum arc melting equipment. The alloy was melted under an argon protective atmosphere and the melting was repeated 5 times. Before each melting, the ingot was turned over. The arc current during the melting process was 400mA and the electromagnetic stirring was turned on. (3) The prepared high-entropy alloy ingots were subjected to homogenization heat treatment in a tubular atmosphere furnace. The homogenization heat treatment process for the two types of hydrogen embrittlement resistant, high-strength and tough dual-phase high-entropy alloys with precipitated second phases, namely "FCC+Ni3Al" and "FCC+NiAl", is as follows: FCC+Ni3Al dual-phase high-entropy alloy: 1150℃ / 12 hours / air cooling; FCC+NiAl dual-phase high-entropy alloy: 1200℃ / 6 hours / air cooling; (4) Using a twin-roll mill, the high-entropy alloy after homogenization heat treatment is subjected to deformation treatment. The rolling processes are as follows: FCC+Ni3Al dual-phase high-entropy alloy: cold-rolled, with a reduction of 70%; FCC+NiAl dual-phase high-entropy alloy: hot rolled at 1150℃ with a reduction of 30%; (5) The rolled high-entropy alloy samples were subjected to recrystallization annealing heat treatment. The heat treatment processes were as follows: FCC+Ni3Al dual-phase high-entropy alloy: 1200℃ / 2 minutes / water cooling; FCC+NiAl dual-phase high-entropy alloy: 1100℃ / 5 minutes / water cooling; (6) For the high-entropy alloy after recrystallization annealing heat treatment, a second-phase precipitation aging heat treatment is performed. The heat treatment processes are as follows: FCC+Ni3Al dual-phase high-entropy alloy: 800℃ / 6 hours / water cooling; FCC+NiAl dual-phase high-entropy alloy: 650℃ / 4 hours / water cooling; (7) After the above steps, the preparation of the hydrogen embrittlement resistant, high strength and toughness dual-phase high entropy alloy is completed.

Images