Construction method of strength-plasticity analysis model for composite material

Abstract

The invention discloses a construction method of a strength-plasticity analysis model for a composite material. The construction method comprises the following steps: carrying out multi-dimensional parameter space sampling on a high-entropy alloy / graphene composite material to generate analog variables, carrying out molecular dynamics modeling on the analog variables, executing simulation, and extracting feature data and label data to construct a sample set; the feature data of the sample set comprises atomic basic attributes, chemical properties, physical properties and structural features; the label data comprises limited tensile strength and ductility; defining an ultimate tensile strength regression prediction model set, and training each model of the ultimate tensile strength regression prediction model set by using the sample set to form a strength-plasticity analysis model; and performing hyper-parameter optimization according to the strength-plasticity analysis model to obtain an optimal adaptive simulation variable. According to the technical scheme, the strength-plasticity balance limitation of a traditional high-entropy alloy can be broken through, a universal design normal form is provided for a high-entropy alloy-based composite material, and the universal design normal form is expanded to other two-dimensional material reinforcement systems.

Description

Technical Field

[0001] The present invention relates to the field of intelligent design of metal matrix composite materials in the field of materials science and engineering, and in particular to a method for constructing a strength-plasticity analysis model for composite materials. Background Art

[0002] As a new material system that breaks through the traditional alloy design paradigm, high entropy alloys form complex solid solutions by mixing multiple principal elements with equal atomic ratios. With core advantages such as thermodynamic high entropy effect and kinetic delayed diffusion effect, they exhibit excellent properties such as high strength, high temperature resistance and corrosion resistance, and have broad application prospects in aerospace, energy equipment and other fields. However, the inherent strength-plasticity trade-off of high entropy alloys is still the core bottleneck of engineering applications, as high-strength materials often lack plasticity, while high-plasticity materials have low strength, making it difficult to meet the stringent requirements of structural parts for comprehensive mechanical properties.

[0003] At present, the main strategies to improve the performance of high entropy alloys include alloy composition optimization and second phase enhancement. Among them, graphene is widely used to construct high entropy alloy / graphene composites due to its excellent mechanical properties and two-dimensional structure, and the strength and plasticity are synergistically improved through mechanisms such as interface strengthening and dislocation pinning. However, the analysis and regulation of the strength and plasticity of composite materials face multiple challenges: for example, since the traditional "melting-characterization" process consumes a lot of time and resources, and it is difficult to observe nanoscale deformation mechanisms such as dislocation evolution and interface failure behavior, the use of experimental trial and error for analysis and regulation is inefficient; although molecular dynamics simulation can reveal atomic-level behavior, the exploration of parameter space (such as composition, temperature, graphene orientation angle and volume fraction) is limited by computing resources, and a single group of simulations takes hundreds of hours, which shows that the simple simulation method has great limitations; the existing machine model is mostly based on a single scale feature (such as composition parameters), but because of the lack of comprehensive consideration of thermodynamics and structural characteristics, and poor interpretability, it is difficult to analyze the multi-factor coupling mechanism, and there is a problem of insufficient data to achieve satisfactory results.

[0004] Therefore, there is an urgent need for a technical solution that integrates multidisciplinary technologies to break through the bottleneck of traditional empirical design, improve the inherent inverse relationship between strength and plasticity of traditional high-entropy alloys, and facilitate the balanced analysis of strength-plasticity of alloy materials during the design process. Summary of the invention

[0006] To achieve the above object, the present application provides a method for constructing a strength-plasticity analysis model for composite materials, comprising the following steps: The high entropy alloy / graphene composite material is sampled in multi-dimensional parameter space to generate simulation variables, molecular dynamics modeling and simulation are performed on the simulation variables, and characteristic data and label data after simulation are extracted to construct a sample set; the simulation variables include: element concentration of high entropy alloy matrix, molecular dynamics simulation temperature, graphene volume fraction and orientation angle between graphene and high entropy alloy matrix; the characteristic data of the sample set is composed of major characteristics, and the major characteristics include: basic atomic properties, chemical properties, physical properties and structural characteristics; the label data of the sample set includes: ultimate tensile strength and ductility; defining an ultimate tensile strength regression prediction model set, using the sample set to train each model of the ultimate tensile strength regression prediction model set, selecting an optimal model, and forming a strength-plasticity analysis model; Hyperparameter optimization is performed according to the strength-plasticity analysis model to obtain the most suitable simulation variables of the strength-plasticity analysis model.

[0007] Among them, multidimensional parameter space sampling is achieved through Sobol quasi-random sequence; The element concentration of the high entropy alloy matrix includes the concentrations of Fe, Ni, Cr, Co and Cu; The molecular dynamics simulation temperature range is 0 to 1100K; The graphene volume fraction ranges from 0.4 to 3.4 vol%; The orientation angle between the graphene and the high entropy alloy matrix ranges from 0 to 45 degrees.

[0008] Furthermore, the molecular dynamics modeling includes: preliminarily constructing the atomic configuration of the composite material as a high entropy alloy model, and then inserting graphene into the high entropy alloy matrix to form a high entropy alloy / graphene composite model; The high entropy alloy / graphene composite material simulated by the high entropy alloy / graphene composite model is composed of FCC, and: graphene is embedded in the middle of the high entropy alloy matrix; the graphene volume is represented by the volume of graphene divided by the volume of the high entropy alloy / graphene composite material, expressed as: ,in, is the volume fraction of graphene, is the length of graphene, is the width of graphene, is the thickness of a single-layer graphene, is the volume of the high entropy alloy / graphene composite.

[0009] The simulation process includes: after performing zero pressure simulation and stretching simulation, obtaining the stress-strain curve corresponding to the simulation variable; during zero pressure simulation, xThe pressure in the z-axis direction is kept at 0 bar; during the stretching simulation, 1×10 9 s -1 The strain rate is kept constant; The simulation environment implementation includes: minimizing the energy of the high entropy alloy / graphene composite material system; After relaxing for 50 ps in an isothermal and isobaric ensemble, the system was further relaxed for 50 ps in a constant temperature and constant volume ensemble until the system energy reached thermodynamic equilibrium.

[0010] Further, the stress-strain curve is used to extract the ultimate tensile strength and ductility data corresponding to the simulation variable; the ultimate tensile strength and ductility data are used to construct the label data; A ductility type is marked according to the ductility, and the ductility type includes insufficient ductility and good ductility; if the sample sizes of insufficient ductility and good ductility are unbalanced, a SMOTE oversampling method is used to synthesize a balanced data set of minority class samples.

[0011] The basic properties of atoms include atomic number, atomic mass, metal radius, volume, electron affinity and valence electron concentration; the chemical properties include Pauling electronegativity, Mulliken electronegativity, ionization energy, mixing enthalpy, mixing entropy and mismatch entropy; The physical properties include density, melting point, specific heat capacity, thermal diffusivity, elastic modulus and Poisson's ratio; The structural features include atomic size differences, radius differences, electronegativity differences, and theoretical molar volumes; Integrate the basic properties of atoms, chemical properties, physical properties and structural characteristics, and use linear mixing rules, derivative mixing rules, deviation mixing rules and difference mixing rules to calculate and generate the first characteristic data; The first characteristic data is subjected to characteristic screening to form characteristic data; the characteristic data includes a ; used to predict ductility ;in, is a linear mixture of bulk moduli, is the derivative mixture of cohesive energy, is a linear mixture of vacancy formation energies, is the mixing enthalpy, T To simulate temperature 、 is the orientation angle of graphene, is the volume fraction of graphene.

[0012] Furthermore, the feature screening step includes: the first step: filtering out low-contribution feature data whose variance is lower than a set threshold in the first feature data by using a variance threshold, and retaining data with significant variability in the first feature data; Step 2: Use the random forest model to calculate the feature importance of each element of the first feature data and identify the features that affect the target variable; Step 3: quantify the linear correlation between features through Pearson correlation analysis, remove highly redundant feature data, and update the first feature data; Step 4: Use recursive feature elimination combined with model accuracy feedback to iteratively remove the secondary features in the first feature data and then update the first feature data; Step 5: Use the exhaustive method to search the first feature data and find the optimal combination to ensure the balance between the predictive performance and interpretability of the feature subset to form the feature data.

[0013] Further, the ultimate tensile strength regression prediction model set includes an ultimate tensile strength regression prediction model and a ductility classification prediction model; The selection of the optimal model is achieved by evaluating the prediction performance through the coefficient of determination and the root mean square error, expressed as: in, is the sample size, is the true value, for The average value, is the predicted value, yes and The number of values, is the determination coefficient, RMSE is the root mean square error; The prediction performance is expressed as: The value range of is [0,1], The closer the value is to 1, the better the model fits the data. In the case of complete fit, Close to 1; The smaller the RMSE value, the smaller the prediction error of the model; The value range is from 0 to 1. The closer the value is to 1, the better the classification effect of the model is and the higher the proportion of correct classification is. The value range of is also from 0 to 1. The closer the value is to 1, the better the model performance is.

[0014] Furthermore, hyperparameter optimization refers to reducing the simulation variable space of high entropy alloy / graphene composites through domain knowledge constraints and SHAP interpretability methods, and optimizing the simulation variables in combination with genetic algorithms to obtain a coupling parameter combination with higher ultimate tensile strength and ductility.

[0015] Furthermore, after obtaining the most suitable simulation variables of the strength-plasticity analysis model, a microstructure analysis is also performed to achieve closed-loop verification; wherein the closed-loop verification is used to verify whether the ultimate tensile strength and ductility corresponding to the most suitable simulation variables of the strength-plasticity analysis model are better than the original data set; The microstructure analysis refers to further verification of the deformation mechanism through common nearest neighbor analysis and dislocation density analysis.

[0016] According to the present invention, the optimal strength-plasticity analysis model can be constructed by extracting the feature set of the constituent elements, temperature conditions, orientation angle of graphene and its volume fraction and the correlation with the mechanical properties of the high entropy alloy / graphene material, introducing machine learning models, domain knowledge, SHAP interpretability analysis and optimization algorithms. Through the use of the strength-plasticity analysis model, the prediction of the ultimate tensile strength and ductility can be achieved, and multi-scale collaborative optimization can be achieved, breaking the traditional high entropy alloy strength-plasticity trade-off limitations. At the same time, a universal design paradigm is provided for high entropy alloy-based composite materials, which can be extended to other two-dimensional material reinforcement systems. The optimized materials are suitable for extreme environment components such as aeroengine blades, providing a new path for the development of high-performance composite materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a step diagram of a method for constructing a strength-plasticity analysis model for a composite material provided in an embodiment of the present invention; Figure 2 It is a schematic diagram of a closed-loop verification process of a strength-plasticity analysis model for composite materials provided according to an embodiment of the present invention; Figure 3 is a schematic diagram of sample distribution based on Sobol quasi-random sequence sampling according to an embodiment of the present invention; Figure 4 is a schematic diagram of molecular dynamics modeling with different element concentrations, graphene orientation angles and graphene volume fractions provided by an embodiment of the present invention; Figure 5 is a stress-strain curve obtained based on molecular dynamics simulation according to an embodiment of the present invention; Figure 6 is a distribution relationship diagram of ultimate tensile strength and ductility obtained from stress-strain curves according to an embodiment of the present invention; Figure 7is a schematic diagram of a five-step feature screening method for reducing feature dimensions provided according to an embodiment of the present invention; Figure 8 is the result of random forest feature importance, Pearson correlation analysis and exhaustive feature screening method in the five-step feature screening method provided by an embodiment of the present invention; Fig. 9 is a performance evaluation and comparison diagram of different classification models of the ultimate tensile strength regression prediction model set provided by an embodiment of the present invention; Fig.10 It is a schematic diagram of prediction results of the optimal regression model and the classification model in the ultimate tensile strength regression prediction model set provided by an embodiment of the present invention; Fig.11 is a parameter optimization strategy diagram provided according to an embodiment of the present invention; Fig.12 is a diagram comparing parameters and performances of a coupling parameter combination and a composite material after parameter optimization according to an embodiment of the present invention; Fig.13 is the distribution of the ultimate tensile strength and ductility in the original data set corresponding to the stress-strain curve obtained by molecular dynamics simulation according to an embodiment of the present invention; Fig.14 is a microstructure evolution analysis diagram implemented by a strength-plasticity analysis model provided in an embodiment of the present invention; Fig.15 Schematic diagram of dislocations in two optimized high entropy alloys / graphenes provided according to embodiments of the present invention. DETAILED DESCRIPTION

[0018] The present invention constructs sample characteristics for the correlation between the constituent elements, temperature conditions, orientation angle of graphene and its volume fraction and mechanical properties of high entropy alloy / graphene materials, combines the characteristics of machine learning models, DK domain knowledge, SHAP interpretability analysis and optimization algorithms, selects the optimal machine learning model structure, constructs the optimal model that can balance strength and plasticity analysis, and determines the most suitable material parameters for strength-plasticity analysis according to the characteristics of the optimal model. In the process of designing high entropy alloy / graphene composite materials, the comprehensive analysis of strength and plasticity can be accurately and efficiently achieved.

[0019] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings.

[0020] Figure 1 A step-by-step diagram of the construction method for the strength-plasticity analysis model of composite materials is provided, including: Step S100: performing multi-dimensional parameter space sampling on the high entropy alloy / graphene composite material to generate simulation variables, performing molecular dynamics modeling and simulation on the simulation variables, and extracting feature data and label data after simulation to construct a sample set; 1) First, the simulation variables are determined for the molecular dynamics simulation of the high entropy alloy / graphene composite material, and sample sampling is performed; the simulation variables proposed in the present invention include: the element concentration of the high entropy alloy matrix, the molecular dynamics simulation temperature, the graphene volume fraction and the orientation angle between the graphene and the high entropy alloy matrix; for example, the element concentration of the high entropy alloy matrix includes the concentrations of Fe, Ni, Cr, Co and Cu; the molecular dynamics simulation temperature range is 0 to 1100K; the graphene volume fraction range is 0.4-3.4 vol%; and the orientation angle range of the graphene and the high entropy alloy matrix is ​​0-45°.

[0021] The sample sampling method selects the Sobol quasi-random sequence to perform multi-dimensional parameter space sampling. In the embodiment of the present invention, 1000 groups of data samples are generated based on the Sobol quasi-random sequence. Figure 3 The results after sampling are shown. According to the results, compared with the traditional Monte Carlo method, this sequence algorithm can generate a uniformly distributed sample set with low diversity, which significantly improves the coverage and convergence speed of high-dimensional parameter space.

[0022] 2) The process of molecular dynamics modeling includes: preliminarily constructing the atomic configuration of the composite material into a high entropy alloy model, and then inserting graphene into the high entropy alloy matrix to form a high entropy alloy / graphene composite model.

[0023] This step can use open source software to implement a large-scale atomic / molecular parallel simulator (such as LAMMPS), and perform molecular dynamics modeling and molecular dynamics simulation calculations on 1000 groups of simulation variables containing different molecular dynamics characteristics in the embodiment of the present invention. The constructed molecular dynamics model has periodic boundary conditions and a size of 10.108nm³, such as Figure 4 As shown, the 1000 samples have the same size, but the simulation variables of molecular dynamics characteristics such as element concentration and graphene volume fraction are different.

[0024] During the simulation, the atomic configuration of the composite material was first constructed as a high entropy alloy model, and then graphene was inserted into the high entropy alloy matrix to form a high entropy alloy / graphene composite model; the high entropy alloy / graphene composite material consists of a single-phase face-centered cubic (FCC) with a FCC lattice spacing of 3.61Å; in each model, graphene is embedded in the middle of the high entropy alloy matrix; different graphene volume fractions can be obtained by different edge lengths; in order to achieve the required volume fraction, the length of graphene along the stretching direction is fixed at 10.108nm, while its width varies from 1.25nm to 10.108nm, and the volume fraction varies from 0.4vol% to 3.4% accordingly; the orientation angle of graphene varies between 0-45°; the graphene volume fraction is represented by the volume of graphene divided by the volume of the high entropy alloy / graphene composite material, expressed as: in, is the volume fraction of graphene, is the length of graphene, is the width of graphene, is the thickness of a single-layer graphene, is the volume of the high entropy alloy / graphene composite.

[0025] On the other hand, appropriate potential functions are selected to improve the accuracy of the simulation results: the adaptive intermolecular reaction empirical bond order (AIREBO) potential is used to describe the interaction between CC in graphene, the embedded atom method (EAM) is used to describe the interaction between Fe, Ni, Cr, Co, and Cu atoms in the HEAs matrix, and the LJ potential is used to describe the interaction between HEAs and carbon atoms.

[0026] 3) In the present invention, LAMMPS is used to perform all simulation processes; OVITO visualization software is used to post-process the atomic trajectories, and common neighbor analysis (CNA) and dislocation analysis (DXA) are combined for structural analysis and visualization.

[0027] The process of executing the simulation includes: after executing the zero pressure simulation and the stretching simulation, obtaining the stress-strain curve corresponding to the simulation variable; wherein, during the zero pressure simulation, x The pressure in the z-axis direction is kept at 0 bar; during the stretching simulation, 1×10 9 s -1 The strain rate is kept constant; The simulation environment implementation includes the following steps: Minimize the energy of high entropy alloy / graphene composite systems; After relaxing for 50 ps in an isothermal and isobaric ensemble, the system was further relaxed for 50 ps in a constant temperature and constant volume ensemble until the system energy reached thermodynamic equilibrium.

[0028] After the tensile simulation, the stress-strain curves of 1000 samples can be obtained, as shown in Figure 5.

[0029] The stress-strain curves can be used to extract the ultimate tensile strength and ductility data corresponding to the simulation variables (e.g. Figure 6 The ultimate tensile strength and ductility data are used to construct the label data (i.e., the response variable); Ductility types are labeled according to ductility, including insufficient ductility and good ductility; for example, samples with ductility greater than 13.4% are labeled as "good ductility" and samples with ductility less than 13.4% are labeled as "inadequate ductility".

[0030] If the number of samples with insufficient ductility and good ductility is unbalanced, for example, samples with "insufficient ductility" only occupy a small part, that is, serious class imbalance occurs, the present invention provides a SMOTE oversampling method to synthesize a balanced data set of minority class samples to solve the problem of ductility class imbalance.

[0031] SMOTE is a widely used oversampling method that balances the dataset by synthesizing minority class samples, thereby alleviating the class imbalance problem. Specifically, SMOTE generates new samples by interpolating between minority class samples. These newly generated samples can increase the diversity of the minority class while avoiding overfitting. By applying the SMOTE technique, more "ductile" high entropy alloy / graphene composite samples are generated to balance the class distribution in the dataset.

[0032] 4) After the present invention determines the label data (limited tensile strength and ductility) of the sample set through the simulation process, it is also necessary to extract feature data from the sample set, i.e., 1000 groups of samples: The characteristic data of the sample set in the present invention is composed of major characteristics, which include: basic atomic properties, chemical properties, physical properties and structural characteristics.

[0033] First, determine the basic atomic properties of the sample, including atomic number, atomic mass, metal radius, volume, electron affinity and valence electron concentration; chemical properties include Pauling electronegativity, Mulliken electronegativity, ionization energy, mixing enthalpy, mixing entropy and mismatch entropy; physical properties include density, melting point, specific heat capacity, thermal diffusion coefficient, elastic modulus and Poisson's ratio; structural characteristics include atomic size difference, radius difference, electronegativity difference and theoretical molar volume; among them, the formulas of some parameters are expressed as follows: , , , in, =8.314.

[0034] In this step, the basic properties of atoms, chemical properties, physical properties and structural characteristics are integrated, 19 elements are selected, and the first characteristic data are calculated and generated using the linear mixing rule, derivative mixing rule, deviation mixing rule and difference mixing rule, which are expressed as: , , , , in, and are the concentration of each component and the corresponding parameter value, is the linear mixing result, is the derivative mixing result, The mixed result is the deviation. is the difference mixing result.

[0035] The first characteristic data includes 93 initial characteristic factors including thermodynamic parameters (mixing entropy) and structural characteristics (atomic size difference). Combined with the angular molecular dynamics simulation temperature, graphene volume fraction and the orientation angle between graphene and the high entropy alloy matrix, a characteristic parameter space of size 96 is constructed.

[0036] 5) Further, the first feature data is subjected to feature screening to form feature data; the feature screening is implemented by adopting a five-step feature screening strategy, as follows: Figure 7 shown.

[0037] The feature screening step comprises: Step 1: filter out low-contribution feature data whose variance is lower than the set threshold in the first feature data through the variance threshold, and retain data with significant variability in the first feature data. After the first step, 47 features for predicting ultimate tensile strength and 47 features for predicting ductility remain. Step 2: The feature importance of each element of the first feature data, i.e., the importance score, is calculated using the random forest model to identify the features that have an impact on the target variable, such as features with a large impact. After the second step, 20 features remain for predicting the ultimate tensile strength and ductility, respectively; Step 3: perform correlation analysis, i.e., quantify the linear correlation between features through Pearson correlation analysis, remove highly redundant feature data (e.g., |r|>0.95), and update the first feature data. At this time, there are 16 features left for predicting ultimate tensile strength and ductility respectively; Step 4: Use recursive feature elimination combined with model accuracy feedback to iteratively remove the secondary features in the first feature data and update the first feature data. At this time, there are 8 and 7 features left for predicting the ultimate tensile strength and ductility, respectively. The fifth step is to use exhaustive feature elimination, that is, to search the first feature data through exhaustive method, find the optimal combination to ensure the balance between the predictive performance and interpretability of the feature subset, and construct the feature data. At this time, there are 7 and 6 features remaining for predicting the ultimate tensile strength and ductility, respectively.

[0038] Figure 8 The random forest model calculates feature importance scores, Pearson correlation analysis quantifies linear correlations between features, and the final results of the exhaustive search are shown.

[0039] Through the feature screening in this step, 7 key features can be effectively screened out from the 93 initial features (i.e., the first feature data) ( ) constitutes the characteristic data to predict the ultimate tensile strength, and selects 6 key features ( , , , T、 , ) constitutes characteristic data to predict ductility. Among them, is a linear mixture of bulk moduli, is the derivative mixture of cohesive energy, is a linear mixture of vacancy formation energies, is the mixing enthalpy, T To simulate temperature 、 is the orientation angle of graphene, is the volume fraction of graphene.

[0040] Through step S100, the key features and their corresponding labels can be obtained to form a data set to realize the training of the ultimate tensile strength regression prediction model set. Further, 70% of the data in the data set is used as a training set, and 30% of the data is used as a test set.

[0041] Step S110: defining an ultimate tensile strength regression prediction model set; The ultimate tensile strength regression prediction model set includes multiple ultimate tensile strength regression prediction models and multiple ductility classification prediction models.

[0042] Eleven different machine learning models were selected for the ultimate tensile strength regression prediction model, including linear regression (LR_R), ridge regression (RR_R), Bayesian ridge regression (BR_R), support vector regression (SV_R), K nearest neighbor (KNN_R), light gradient boosting machine regression (LightGBM_R), extreme gradient boosted tree regression (XGBoost_R), decision tree (DT_R), extreme random tree regression (ET_R), gradient boosted decision tree (GBDT_R), and random forest regression (RF_R); Twelve different machine learning models were selected for the scalable classification prediction model, including logistic regression (LR_C), linear discriminant analysis (LDA_C), decision tree (DT_C), random forest (RF_C), gradient boosting (GBDT_C), extreme gradient boosted tree classifier (XGBoost_C), light gradient boosting machine classifier (LightGBM_C), classification boosting algorithm (CatBoost_C), Gaussian naive Bayes (GNB_C), K nearest neighbor (KNN_C), multilayer perceptron classifier (MLP_C), and support vector machine (SVM_C).

[0043] Step S120: using the sample set to train each model of the ultimate tensile strength regression prediction model set, selecting the optimal model, and forming a strength-plasticity analysis model; In the present invention, the optimal model is selected by judging the predictive performance of the regression model, specifically by the determination coefficient (R 2 ) and root mean square error (RMSE), expressed as: in, is the sample size, is the true value, for The average value of is the predicted value, yes and The number of values, is the determination coefficient, RMSE is the root mean square error; The prediction performance is expressed as: The value range of is [0,1], The closer the value is to 1, the better the model fits the data. In the case of complete fit, Close to 1; The smaller the RMSE value, the smaller the prediction error of the model; The value range is from 0 to 1. The closer the value is to 1, the better the classification effect of the model is and the higher the proportion of correct classification is. The value range of is also from 0 to 1. The closer the value is to 1, the better the model performance is.

[0044] Fig. 9 The performance evaluation results of various machine learning models are shown. Fig. 9 (a)-(d) and (e)-(h) show the performance of regression model and classification model respectively. It can be found that SV_R, LGBM_R, XGBoost_R and other models all show high prediction accuracy. 2 >0.9, which indicates that they can fit the data well and have a high accuracy in predicting the target variable. Among the classification models, RF_C, XGBoost_C, LGBM_C, and CatBoost_C also show an Accuracy higher than 0.9, indicating that they can effectively distinguish different categories in the classification task and have a high classification reliability.

[0045] In this step, the accuracy is higher than 0.9 and R 2 >0.9, each model is selected as the optimal model, and the optimal model after training is constructed into a strength-plasticity analysis model.

[0046] Step S130: performing hyperparameter optimization according to the strength-plasticity analysis model to obtain the most suitable simulation variables of the strength-plasticity analysis model.

[0047] In this step, the size of the high entropy alloy / graphene composite coupling parameter space is reduced by combining domain knowledge constraints and SHAP interpretability analysis, and parameter optimization is performed in combination with genetic algorithms. In order to concentrate resources, in this step, models that have shown high performance potential are optimized, that is, models with evaluation scores greater than 0.9 in step S120 are optimized by hyperparameter optimization to further improve the prediction performance of the model.

[0048] Hyperparameter optimization uses the Bayesian optimization algorithm to implement a coupled search strategy, which is based on genetic optimization. The coupled search strategy includes four different search experiments to comprehensively evaluate the impact of different constraints on the optimization effect. The four different conditions are: the entire parameter space without constraints, constraints based on domain knowledge, constraints based on domain knowledge and SHAP interpretability analysis, and output judgment based on domain knowledge constraints and SHAP interpretability analysis combined with classification model.

[0049] Hyperparameter optimization refers to reducing the simulation variable space of high entropy alloy / graphene composite materials through domain knowledge constraints and SHAP interpretability method, optimizing the simulation variables with genetic algorithm, and obtaining coupling parameter combinations with higher ultimate tensile strength and ductility. The specific steps are as follows: Fig.11 As shown: First, the original data is preprocessed to calculate the characteristic factors required by the model (i.e., the first characteristic data before feature screening); then, the two models XGBoost_R and GBDT_C are trained to provide a basis for subsequent optimization operations; among them, XGBoost_R has the highest prediction accuracy compared to other regression models, so it is used to predict tensile strength and named ML-UTS (machine learning - Ultimate tensile strength), and GBDT_C is used to predict ductility and named ML-EL (machine learning -elongation); then, the parameters of the genetic algorithm are initialized, including key parameters such as population size, number of iterations, and crossover mutation rate; during the optimization process, the output value of the XGBoost_R model is used as the fitness function of the genetic algorithm to evaluate the quality of individuals. The algorithm continuously optimizes the population in an iterative manner, and each iteration includes operations such as selection, crossover, and mutation to gradually improve the overall fitness of the population. After calculating the fitness of the initial population, optimization experiments were carried out under four different constraints; each constraint imposes different restrictions on the optimization process, thus affecting the final optimization results. When the preset iteration stop condition is met, the algorithm outputs the optimal coupling parameter result and its corresponding fitness.

[0050] Among them, domain knowledge mainly refers to the reduction of parameter space through valence electron concentration (VEC>8.225); SHAP interpretability method ( <21.5°, T <580K and <1.81vol%∪ >3 vol%) refers to the range recommended by the SHAP method for parameter space reduction.

[0051] Specifically, the population size of the genetic algorithm is 100, the crossover mutation rate is 0.1, and the number of iterations is 100.

[0052] In the process of hyperparameter optimization, the coupling parameter space of high entropy alloy / graphene composite materials is reduced through domain knowledge constraints and SHAP interpretability method, and the coupling parameters are optimized in combination with genetic algorithm to obtain coupling parameter combinations with higher ultimate tensile strength and ductility.

[0053] After hyperparameter adjustment, XGBoost_R and GBDT_C showed the highest prediction performance in the regression model and classification model. Their prediction results on the training set and test set are shown in the following figure. Fig.10 Therefore, they are used to predict the ultimate tensile strength and ductility, respectively.

[0054] Further, such as Fig.12 As shown in Figure 2, two optimal high entropy alloy / graphene composite Fe 10.7 Ni 29.6 Cr 22.8 Co 21.5 Cu 15.4 / Gr (3.26vol%, 0.04°) and Fe 24.3 Ni 12.3 Cr 23.6 Co 11.3 Cu 25.6 / Gr (0.62vol%, 1.8°) .

[0055] In general, in the whole process of the present invention, data sampling and molecular dynamics simulation, feature engineering and machine learning modeling, coupled parameter space optimization and search are performed to obtain the most suitable simulation variables for the strength-plasticity analysis model; and microstructure analysis is performed to achieve closed-loop verification of each link, as shown in the following example. Figure 2 shown.

[0056] Closed-loop verification refers to molecular dynamics simulation based on the coupling parameters of two optimal high-entropy alloy / graphene composite materials. Fig.13 As shown in the figure, the stress-strain curves of the two optimal high entropy alloy / graphene composites were obtained by molecular dynamics simulation, and the ultimate tensile strength and ductility were obtained by statistics. As can be seen from the figure, the mechanical properties of the two composites are better than the original data set (ultimate tensile strength>14.5Gpa, ductility>13.4%). This shows that by combining DK domain knowledge and SHAP interpretability method for optimization, the mechanical properties of high entropy alloy / graphene composites can be significantly improved, bringing them to a new level in terms of high strength and high ductility.

[0057] Microstructure analysis refers to further verification of the deformation mechanism through common nearest neighbor analysis and dislocation density analysis.

[0058] The present invention uses CNA to statistically analyze the changes in the atomic number fractions of FCC, body-centered cubic (BCC), hexagonal close-packed (HCP) and disordered structure (OTHER) in two optimal high-entropy alloy / graphene composites. Fig.14The evolution of the numerical scores of the four structures with strain is shown separately: during the deformation process, the atomic proportion of the FCC structure shows a gradual downward trend. This shows that with the increase of strain, the atoms that originally existed in the FCC structure gradually underwent structural transformation; at the same time, the atomic proportions of HCP and OTHER showed an upward trend. This is related to the stress state and energy changes experienced by the material during the deformation process. Specifically, with the increase of strain, the stress distribution inside the material changes, causing some atoms to transform from the higher energy FCC structure to the lower energy HCP structure, and also producing some disordered atomic arrangements.

[0059] The present invention also uses DXA to count the dislocations in two optimized high entropy alloys / graphenes, as shown in FIG. Fig.15 As shown in Figure 2, dislocations are intrinsic defects, and their dynamic evolution has a profound impact on the mechanical properties and deformation behavior of materials.

[0060] Specifically, the two composite materials are shown to be <110> Perfect dislocation, 1 / 3 <100> (Hirth dislocation), 1 / 3 <111> (Frank dislocation), 1 / 6 <110> (Stair-rod dislocation), 1 / 6 <112> The evolution of (Shockley dislocation) and OTHER dislocations over time, and the inset shows the evolution of dislocation density. It can be seen that Shockley dislocations, Stair-rod dislocations and OTHER dislocations are the dominant types in the two systems, among which Shockley dislocations account for the highest proportion and show a trend of first increasing and then decreasing, and its peak value is closely related to the yield point of the material. It is worth noting that Fe 10.7 Ni 29.6 Cr 22.8 Co 21.5 Cu 15.4 / Gr (3.26vol%, 0.04°) The overall dislocation density of the system is significantly lower than that of Fe 24.3 Ni 12.3 Cr 23.6 Co 11.3 Cu 25.6 / Gr (0.62vol%, 1.8°) , which is in sharp contrast to its higher strength. This phenomenon can be explained by the dislocation strengthening mechanism, that is, although high dislocation density can improve strength, excessive dislocation entanglement leads to premature strain localization.

[0061] Microstructural analysis shows that a smaller graphite training orientation angle corresponds to better strength. A larger graphene volume fraction hinders the propagation of dislocations. When dislocations move above and below the graphene layer respectively, there is a high probability of collision with other dislocations, thus forming dislocation tangles and enhancing the strength of the material. In contrast, smaller graphene fails to fully occupy a plane, resulting in less interfacial stress concentration, thereby reducing the initiation and propagation of dislocations, making the material more prone to plastic deformation during the tensile process, and thus improving ductility.

[0062] The method for constructing a strength-plasticity analysis model for composites provided by the present invention can construct an optimal model applicable to the design of high-entropy alloy / graphene composites. By integrating multi-scale data and multi-technical means, it achieves a significant improvement in calculation speed while ensuring high prediction accuracy. Compared with the construction of traditional network models, the method provided by the present invention has lower costs and a shorter cycle. It can not only quickly and accurately predict the ultimate tensile strength and ductility before experimentally preparing high-entropy alloy / graphene composites, but also synergistically improve strength and plasticity. At the same time, the model exhibits excellent performance on the dataset, which not only verifies the effectiveness of the method but also lays a foundation for its extension in other different high-entropy alloy systems.

[0063] The above are only several specific embodiments of the present invention disclosed. However, the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for constructing a strength-plasticity analysis model for composite materials, characterized in that: The following steps are involved: The high entropy alloy / graphene composite material is sampled in multi-dimensional parameter space to generate simulation variables, molecular dynamics modeling and simulation are performed on the simulation variables, and characteristic data and label data after simulation are extracted to construct a sample set; the simulation variables include: element concentration of high entropy alloy matrix, molecular dynamics simulation temperature, graphene volume fraction and orientation angle between graphene and high entropy alloy matrix; the characteristic data of the sample set is composed of major characteristics, and the major characteristics include: basic atomic properties, chemical properties, physical properties and structural characteristics; the label data of the sample set includes: ultimate tensile strength and ductility; defining an ultimate tensile strength regression prediction model set, using the sample set to train each model of the ultimate tensile strength regression prediction model set, selecting an optimal model, and forming a strength-plasticity analysis model; Hyperparameter optimization is performed according to the strength-plasticity analysis model to obtain the most suitable simulation variables of the strength-plasticity analysis model.

2. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The multidimensional parameter space sampling is achieved by a Sobol quasi-random sequence; The element concentration of the high entropy alloy matrix includes the concentrations of Fe, Ni, Cr, Co and Cu; The molecular dynamics simulation temperature range is 0 to 1100K; The graphene volume fraction ranges from 0.4 to 3.4 vol%; The orientation angle between the graphene and the high entropy alloy matrix ranges from 0 to 45 degrees.

3. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The molecular dynamics modeling includes: preliminarily constructing the atomic configuration of the composite material into a high entropy alloy model, and then inserting graphene into the high entropy alloy matrix to form a high entropy alloy / graphene composite model; The high entropy alloy / graphene composite material simulated by the high entropy alloy / graphene composite model is composed of FCC, and: graphene is embedded in the middle of the high entropy alloy matrix; the graphene volume is represented by the volume of graphene divided by the volume of the high entropy alloy / graphene composite material, expressed as: in, is the volume fraction of graphene, is the length of graphene, is the width of graphene, is the thickness of a single-layer graphene, is the volume of the high entropy alloy / graphene composite.

4. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The process of executing the simulation includes: after executing the zero pressure simulation and the stretching simulation, obtaining the stress-strain curve corresponding to the simulation variable; wherein, during the zero pressure simulation, x The pressure in the z-axis direction is kept at 0 bar; during the stretching simulation, 1×10 9 s -1 The strain rate is kept constant; The environmental implementation for performing the simulation includes: minimizing the energy of the high entropy alloy / graphene composite system; After relaxing for 50 ps in an isothermal and isobaric ensemble, the system was further relaxed for 50 ps in a constant temperature and constant volume ensemble until the system energy reached thermodynamic equilibrium.

5. The method for constructing a strength-plasticity analysis model for composite materials according to claim 4, characterized in that: The stress-strain curve is used to extract the ultimate tensile strength and ductility data corresponding to the simulation variable; the ultimate tensile strength and ductility data are used to construct label data; A ductility type is marked according to the ductility, and the ductility type includes insufficient ductility and good ductility; if the sample sizes of insufficient ductility and good ductility are unbalanced, a SMOTE oversampling method is used to synthesize a balanced data set of minority class samples.

6. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The basic atomic properties include atomic number, atomic mass, metal radius, volume, electron affinity and valence electron concentration; the chemical properties include Pauling electronegativity, Mulliken electronegativity, ionization energy, mixing enthalpy, mixing entropy and mismatch entropy; The physical properties include density, melting point, specific heat capacity, thermal diffusivity, elastic modulus and Poisson's ratio; The structural features include atomic size differences, radius differences, electronegativity differences, and theoretical molar volumes; Integrate the basic properties of atoms, chemical properties, physical properties and structural characteristics, and use linear mixing rules, derivative mixing rules, deviation mixing rules and difference mixing rules to calculate and generate the first characteristic data; The first characteristic data is subjected to characteristic screening to form characteristic data; the characteristic data includes a To predict ductility in, is a linear mixture of bulk moduli, is the derivative mixture of cohesive energy, is a linear mixture of vacancy formation energies, is the mixing enthalpy, T To simulate temperature 、 is the orientation angle of graphene, is the volume fraction of graphene.

7. The method for constructing a strength-plasticity analysis model for composite materials according to claim 6, characterized in that: The feature screening step comprises: a first step: filtering out low-contribution feature data whose variance is lower than a set threshold in the first feature data by using a variance threshold, and retaining data with significant variability in the first feature data; Step 2: Use the random forest model to calculate the feature importance of each element of the first feature data and identify the features that affect the target variable; Step 3: quantify the linear correlation between features through Pearson correlation analysis, remove highly redundant feature data, and update the first feature data; Step 4: Use recursive feature elimination combined with model accuracy feedback to iteratively remove the secondary features in the first feature data and then update the first feature data; Step 5: Use the exhaustive method to search the first feature data and find the optimal combination to ensure the balance between the predictive performance and interpretability of the feature subset to form the feature data.

8. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The ultimate tensile strength regression prediction model set includes an ultimate tensile strength regression prediction model and a ductility classification prediction model; The selection of the optimal model is achieved by evaluating the prediction performance through the coefficient of determination and the root mean square error, expressed as: in, is the sample size, is the true value, for The average value, is the predicted value, yes and The number of values, is the determination coefficient, RMSE is the root mean square error; The prediction performance is expressed as: The value range of is [0,1], The closer the value is to 1, the better the model fits the data. In the case of complete fit, Close to 1; The smaller the RMSE value, the smaller the prediction error of the model; The value range is from 0 to 1. The closer the value is to 1, the better the classification effect of the model is and the higher the proportion of correct classification is. The value range of is also from 0 to 1. The closer the value is to 1, the better the model performance is.

9. The method for constructing a strength-plasticity analysis model for composite materials according to claim 1, characterized in that: The hyperparameter optimization refers to reducing the simulation variable space of high entropy alloy / graphene composite materials through domain knowledge constraints and SHAP interpretability method, optimizing the simulation variables in combination with genetic algorithm, and obtaining a coupling parameter combination with higher ultimate tensile strength and ductility.

10. The method for constructing a strength-plasticity analysis model for composite materials according to claim 9, characterized in that: After obtaining the most suitable simulation variables of the strength-plasticity analysis model, a microstructure analysis is further performed to achieve closed-loop verification; wherein the closed-loop verification is used to verify whether the ultimate tensile strength and ductility corresponding to the most suitable simulation variables of the strength-plasticity analysis model are better than the original data set; The microstructure analysis refers to further verification of the deformation mechanism through common nearest neighbor analysis and dislocation density analysis.

Images