Component design method of metastable beta titanium alloy with low Young modulus

Abstract

The invention discloses a component design method of a metastable beta titanium alloy with low Young modulus, relates to the technical field of metal material design, and is used for solving the technical problems that the existing titanium alloy component design is relatively weak in target orientation, relatively large in deviation between the performance of the designed titanium alloy and the actual demand, and low in efficiency. The component design method comprises the steps that the titanium alloy is obtained; performing molecular dynamics simulation on the titanium alloy to obtain components of the titanium alloy and an original data set of Young modulus corresponding to the components; characteristic parameters corresponding to the multiple characteristics of the titanium alloy are determined; determining target features according to the feature parameters; firstly, a virtual component space is preset, multiple trained machine learning models are used for predicting Young modulus corresponding to different component contents under virtual components, predicted values are obtained, and the metastable beta titanium alloy component corresponding to the minimum Young modulus is obtained. According to the technical scheme, the component design method of the metastable beta titanium alloy with the low Young modulus is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material design, and more specifically, to a composition design method for a low Young's modulus metastable β titanium alloy. Background Art

[0002] In the field of human bone transplantation, the biocompatibility of materials is a very important indicator, and its potential genotoxicity, cytotoxicity, carcinogenicity, mutagenicity, sensitization, neurological effects and Young's modulus must be considered. When the Young's modulus of the material used is higher than the surrounding human bone tissue, stress shielding is a potential problem and increases the risk of related complications.

[0003] Compared with traditional 316L stainless steel and Co-Cr alloy, the low modulus characteristics of titanium alloy make it play a key role in the field of bone transplantation. Among titanium alloys, (α+β)Ti-6Al-4V is still the most commonly used one in orthopedic applications, although its Young's modulus is about 110GPa, almost 4 times that of cortical bone tissue (about 27GPa), and it contains a large amount of toxic vanadium and aluminum related to neurological side effects. In the medical field, the ideal biomedical titanium alloy should have the characteristics of low elastic modulus, low toxicity, and high ductility. In the existing technology, the composition design of biomedical titanium alloys is mainly based on empirical trial and error. The purpose orientation of the empirical trial and error method is weak, the performance of the designed titanium alloy is greatly deviated from the actual demand, and the efficiency is low, which cannot meet the composition design requirements of biomedical titanium alloys. Summary of the invention

[0004] The purpose of the present invention is to provide a composition design method for a low Young's modulus metastable β titanium alloy, which is used to solve the technical problems that the existing titanium alloy composition design has a weak purpose orientation, a large deviation between the designed titanium alloy performance and the actual demand, and low efficiency. In view of this, the present invention is implemented through the following technical solutions.

[0005] The present invention provides a composition design method for a low Young's modulus metastable β titanium alloy, comprising: Get titanium alloy; A molecular dynamics simulation is performed on the titanium alloy using a MEAM potential function to obtain an original data set of the composition of the titanium alloy and the Young's modulus corresponding to the composition; Determining characteristic parameters corresponding to a plurality of characteristics of the titanium alloy according to the composition of the titanium alloy; According to the characteristic parameters, multiple characteristics with the highest correlation with the Young's modulus of the titanium alloy are selected as target characteristics; Using a plurality of machine learning models respectively, according to the original data set and the characteristic parameters corresponding to the target features, determine the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy, and complete the training of each machine learning model; Preset a virtual component, and use multiple machine learning models that have been trained to predict the Young's modulus corresponding to different component contents under the virtual component to obtain a predicted value; The metastable β titanium alloy composition corresponding to the minimum Young's modulus is obtained in the predicted value.

[0006] Compared with the prior art, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, according to the characteristic parameters corresponding to the above original data set and the target feature, a plurality of machine learning models are trained. After the training is completed, the virtual composition of the titanium alloy is preset, and then the Young's modulus corresponding to the different component contents under the virtual composition is predicted by using the trained plurality of machine learning models to obtain the predicted value. According to the predicted value, the metastable β titanium alloy composition corresponding to the minimum Young's modulus can be obtained to complete the composition design. In the above process, the original data set is obtained by molecular dynamics simulation of the titanium alloy, which means the data set of the relationship between the Young's modulus corresponding to different components of the titanium alloy and any component. In the process of training the machine learning model, the nonlinear relationship between the composition and the Young's modulus of the titanium alloy cannot be accurately obtained by using the above original data set alone. Therefore, the present invention further introduces the characteristic parameters corresponding to the above target feature, and the target feature is determined by the composition of the titanium alloy, which can be either a chemical feature of the titanium alloy or a physical feature. In the process of training the machine learning model, the input is the characteristic parameters corresponding to the original data set and the target feature, and the output is the Young's modulus. That is to say, by taking the characteristic parameters corresponding to the original data set and the target feature as input at the same time, the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy is accurately obtained. In the process of determining the predicted value, the virtual component is used as input, and the Young's modulus corresponding to the different component contents under the virtual component is output, that is, the output is the predicted value; it has been verified that in this process, the predicted value, that is, the Young's modulus corresponding to the different component contents under the virtual component is consistent with the experimental trend. Since the predicted value is the Young's modulus under the ideal state, the predicted value is larger than the experimental value. Since the predicted value is consistent with the experimental value trend, the metastable β titanium alloy composition corresponding to the minimum Young's modulus can be obtained from the predicted value, and then the corresponding metastable β titanium alloy can be manufactured according to the metastable β titanium alloy composition to meet the requirements of low Young's modulus. The composition design method of the present invention has high accuracy and strong purpose orientation, and significantly improves the design efficiency by introducing the machine learning model, which solves the technical problems that the existing titanium alloy composition design has weak purpose orientation, the designed titanium alloy performance has a large deviation from the actual demand, and the design efficiency is low.

[0007] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, after the virtual composition is preset and before the Young's modulus corresponding to different component contents under the virtual composition is predicted, it also includes: The metastable β titanium alloy is screened out from titanium alloys with different component contents under the virtual composition.

[0008] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, the step of selecting the metastable β titanium alloy from the titanium alloys with different composition contents under the virtual composition comprises: The metastable β titanium alloy is screened out from titanium alloys with different component contents under the virtual composition using screening factors; the screening factors include one or more of d orbital binding energy, average electron orbital parameter, average valence electron concentration, Mo equivalent method and binding force.

[0009] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, when the titanium alloys with different component contents under the virtual composition meet all the screening factors of the metastable β titanium alloy, the corresponding titanium alloy is determined to be a metastable β titanium alloy.

[0010] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, in the process of determining characteristic parameters corresponding to multiple characteristics of the titanium alloy according to the composition of the titanium alloy, the characteristics include one or more of Pauling electronegativity, density, atomic weight, number of valence electrons, melting point, specific heat capacity, metal radius, total number of electrons, atomic radius and first ionization energy.

[0011] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, the characteristic parameters include parameters corresponding to the characteristics and the variance of the parameters corresponding to each characteristic.

[0012] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, in the process of determining the nonlinear relationship among the composition, characteristic parameters and Young's modulus of the titanium alloy based on the original data set and the characteristic parameters corresponding to the target features, a variety of machine learning models are utilized, and the input of each machine learning model is the characteristic parameters corresponding to the target features and the original data set, and the output is the Young's modulus corresponding to the original data set.

[0013] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, in the process of predicting the Young's modulus corresponding to different component contents under the virtual composition through the virtual composition and the characteristic parameters corresponding to the virtual composition, a plurality of machine learning models that have been trained are utilized, with the input being the virtual composition and the output being the Young's modulus corresponding to different component contents under the virtual composition.

[0014] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, after obtaining the original data set of the composition of the titanium alloy and the Young's modulus corresponding to the composition, it also includes: The original data corresponding to the Young's modulus lower than 50 GPa in the original data set are removed.

[0015] Furthermore, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, the multiple features with the highest correlation with the Young's modulus of the titanium alloy are screened according to the characteristic parameters, including: Using Pearson correlation analysis, feature parameters with correlation greater than 0.9 were removed; The random forest algorithm is then used to perform importance analysis on the remaining feature parameters to screen out multiple features that have the greatest impact on Young's modulus. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 This is a performance evaluation diagram of the machine learning model for predicting the Young's modulus of the β titanium alloy in Example 3 of the present invention; Figure 2 It is a Pearson correlation analysis diagram of all physical and thermodynamic characteristics in Example 3 of the present invention; Figure 3 This is a feature importance ranking diagram obtained by using the random forest algorithm in Example 3 of the present invention; Figure 4 A comparison chart of the predicted value and the simulated value of the support vector machine regression model used in Example 3 of the present invention on the training set; Figure 5 A comparison chart of the predicted value and the simulated value of the support vector machine regression model used in Example 3 of the present invention on the test set; Figure 6 A comparison chart of the predicted value and the simulated value of the multilayer perceptron regression model used in Example 3 of the present invention on the training set; Figure 7 A comparison chart of the predicted value and the simulated value of the multilayer perceptron regression model used in Example 3 of the present invention on the test set; Figure 8 A comparison chart of the predicted value and the simulated value of the XGBoost Regression model used in Example 3 of the present invention on the training set; Fig. 9 A comparison chart of the predicted value and the simulated value of the XGBoost Regression model used in Example 3 of the present invention on the test set; Fig.10 It is a trend comparison chart of the Young's modulus model prediction value of some TiZrNb ternary alloys in Example 3 of the present invention and the experimental value in the literature; Fig.11 It is a trend comparison chart of the Young's modulus model prediction value of some TiZrNbTa quaternary alloys in Example 3 of the present invention and the experimental value in the literature; Fig.12 This is a trend comparison chart of the Young's modulus model predicted values ​​of some TiZrHfNbTa quinary system alloys in Example 3 of the present invention and the experimental values ​​in the literature. DETAILED DESCRIPTION

[0017] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or indirectly on the other element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or indirectly connected to the other element.

[0019] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. The meaning of "several" is one or more, unless otherwise clearly and specifically defined.

[0020] Compared with traditional 316L stainless steel and Co-Cr alloy, the low modulus characteristics of titanium alloy make it play a key role in the field of bone transplantation. Among titanium alloys, (α+β)Ti-6Al-4V is still the most commonly used one in orthopedic applications, although its Young's modulus is about 110GPa, almost 4 times that of cortical bone tissue (about 27GPa), and it contains a large amount of toxic vanadium and aluminum related to neurological side effects. In the medical field, the ideal biomedical titanium alloy should have the characteristics of low elastic modulus, low toxicity, and high ductility. In the existing technology, the composition design of biomedical titanium alloys is mainly based on empirical trial and error. The purpose orientation of the empirical trial and error method is weak, the performance of the designed titanium alloy is greatly deviated from the actual demand, and the efficiency is low, which cannot meet the composition design requirements of biomedical titanium alloys.

[0021] In order to solve the above technical problems, the present invention provides a composition design method for a low Young's modulus metastable β titanium alloy, comprising: Get titanium alloy; A molecular dynamics simulation is performed on the titanium alloy using a MEAM potential function to obtain an original data set of the composition of the titanium alloy and the Young's modulus corresponding to the composition; Determining characteristic parameters corresponding to a plurality of characteristics of the titanium alloy according to the composition of the titanium alloy; According to the characteristic parameters, multiple characteristics with the highest correlation with the Young's modulus of the titanium alloy are selected as target characteristics; Using a plurality of machine learning models respectively, according to the original data set and the characteristic parameters corresponding to the target features, determine the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy, and complete the training of each machine learning model; Preset a virtual component, and use multiple machine learning models that have been trained to predict the Young's modulus corresponding to different component contents under the virtual component to obtain a predicted value; The metastable β titanium alloy composition corresponding to the minimum Young's modulus is obtained in the predicted value.

[0022] In the case of adopting the above technical solution, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, according to the characteristic parameters corresponding to the above original data set and the target feature, multiple machine learning models are trained. After the training is completed, the virtual composition of the titanium alloy is preset, and then the Young's modulus corresponding to the different component contents under the virtual composition is predicted by using the trained multiple machine learning models to obtain the predicted value. According to the predicted value, the metastable β titanium alloy composition corresponding to the minimum Young's modulus can be obtained to complete the composition design. In the above process, the original data set is obtained by molecular dynamics simulation of the titanium alloy, which means the data set of the relationship between the Young's modulus corresponding to any component of the different components of the titanium alloy. In the process of training the machine learning model, the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy cannot be accurately obtained by using the above original data set alone. Therefore, the present invention further introduces the characteristic parameters corresponding to the above target feature, and the target feature is determined by the composition of the titanium alloy, which can be either a chemical feature of the titanium alloy or a physical feature. In the process of training the machine learning model, the input is the characteristic parameters corresponding to the original data set and the target feature, and the output is the Young's modulus. That is to say, by taking the characteristic parameters corresponding to the original data set and the target feature as input at the same time, the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy is accurately obtained. In the process of determining the predicted value, the virtual component is used as input, and the Young's modulus corresponding to the different component contents under the virtual component is output, that is, the output is the predicted value; it has been verified that in this process, the predicted value, that is, the Young's modulus corresponding to the different component contents under the virtual component is consistent with the experimental trend. Since the predicted value is the Young's modulus under the ideal state, the predicted value is larger than the experimental value. Since the predicted value is consistent with the experimental value trend, the metastable β titanium alloy composition corresponding to the minimum Young's modulus can be obtained from the predicted value, and then the corresponding metastable β titanium alloy can be manufactured according to the metastable β titanium alloy composition to meet the requirements of low Young's modulus. The composition design method of the present invention has high accuracy and strong purpose orientation, and significantly improves the design efficiency by introducing a machine learning model, thereby solving the technical problems of weak purpose orientation in the existing titanium alloy composition design, large deviation between the designed titanium alloy performance and actual demand, and low design efficiency; illustratively, the above-mentioned machine learning models include support vector machine regression model (Support Vector Regression), multilayer perceptron regression model (MLP regression), extreme gradient boosting regression model (XGBoost Regression), AdaBoost regression model (AdaBoost Regressor), decision tree regression model (DecisionTreeRegressor) and gradient boosting regression model (GradientBoostingRegressor).

[0023] As a possible implementation manner, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, after the virtual composition is preset and before the Young's modulus corresponding to different component contents under the virtual composition is predicted, it also includes: The metastable β titanium alloy is screened out from titanium alloys with different component contents under the virtual composition.

[0024] In the case of adopting the above technical solution, only a part of the titanium alloy is a metastable β titanium alloy under the virtual composition. In order to improve the accuracy of the output prediction value and improve the efficiency of the machine learning model, the technical solution further screens out all the required metastable β titanium alloys from the titanium alloys with different component contents under the virtual composition. Further, for the above process, the metastable β titanium alloy can be screened out from the titanium alloys with different component contents under the virtual composition using screening factors; the screening factors include one or more of d orbital binding energy (Md), average electron orbital parameter (Bo), average valence electron concentration (VEC), Mo equivalent method (Mo) and bonding force (Bonding Force). For the screening requirements, when the titanium alloys with different component contents under the virtual composition meet all the screening factors of the metastable β titanium alloy, the corresponding titanium alloy is determined to be a metastable β titanium alloy. By way of example, the screening factors may be d orbital binding energy (Md), average electron orbital parameter (Bo), average valence electron concentration (VEC), Mo equivalent method (Mo) and bonding force. When titanium alloys with different composition contents under the virtual composition meet the requirements of d orbital binding energy (Md), average electron orbital parameter (Bo), average valence electron concentration (VEC), Mo equivalent method (Mo) and bonding force (Bonding Force) of metastable β titanium alloys, the corresponding titanium alloy is determined to be a metastable β titanium alloy.

[0025] As described above, in the process of determining the characteristic parameters corresponding to multiple characteristics of the titanium alloy according to the composition of the titanium alloy, the characteristics can be chemical characteristics or physical characteristics, and the characteristics are determined by the composition of the titanium alloy. For example, the characteristics include one or more of Pauling electronegativity, density, atomic weight, valence electron number, melting point, specific heat, metal radius, total electron number, atomic radius and first ionization energy. Furthermore, the characteristic parameters include parameters corresponding to the characteristics and the variance of the parameters corresponding to each characteristic.

[0026] It should be understood that after the molecular dynamics simulation of the titanium alloy is performed using the MEAM potential function to obtain the original data set of the composition of the titanium alloy and the Young's modulus corresponding to the composition, since the Young's modulus in the original data set is obtained through simulation and is greater than the actual value, in order to make the Young's modulus in the original data set not too low and to improve the accuracy of the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy, the present invention further removes the original data corresponding to the Young's modulus lower than 50 GPa in the original data set.

[0027] As a possible implementation mode, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, the multiple features with the highest correlation with the Young's modulus of the titanium alloy are screened according to the characteristic parameters, including: Using Pearson correlation analysis, feature parameters with correlation greater than 0.9 were removed; The random forest algorithm is then used to perform importance analysis on the remaining feature parameters to screen out multiple features that have the greatest impact on Young's modulus.

[0028] In the case of adopting the above technical solution, in the composition design method of the low Young's modulus metastable β titanium alloy of the present invention, multiple features with the greatest impact on Young's modulus are screened from multiple features through Pearson correlation analysis and random forest algorithm, and the multiple screened features are used as target features, input into the machine learning model to train the machine learning model, and a more accurate nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy is obtained. Using Pearson correlation analysis, feature parameters with a correlation greater than 0.9 are removed. When a feature is highly correlated with the remaining multiple features (>0.9), it is deleted. Further, the process of using the random forest algorithm to perform importance analysis on the remaining feature parameters can be, according to the feature importance, the multiple features are sorted, and the importance of the features corresponding to the order from front to back is added, and all features before the sum of importance is greater than 0.98 are retained, and the retained features are used as target features.

[0029] In order to better understand the present invention, the content of the present invention is further explained below in conjunction with specific embodiments, but the content of the present invention is not limited to the following embodiments. Example 1

[0030] This embodiment provides a composition design method for a low Young's modulus metastable β titanium alloy, comprising: S100, obtain titanium alloy; S200, performing molecular dynamics simulation on the titanium alloy using a MEAM potential function to obtain a composition of the titanium alloy and an original data set of Young's modulus corresponding to the composition, and removing original data corresponding to a Young's modulus lower than 50 GPa in the original data set; S300, determining characteristic parameters corresponding to 10 characteristics of the titanium alloy according to the composition of the titanium alloy, the 10 characteristics are Pauling electronegativity (Pauling EN), density (Density), atomic weight (Atomic Weight), valence electron number (Valence e), melting point (Melting Point), specific heat (Specific Heat), metal radius (Metallic Radius), total electron number (Total Electrons), atomic radius (Atomic Radii) and first ionization energy (Firstionization energy), and the characteristic parameters are parameter values ​​corresponding to the above characteristics and the variance of the parameter value of each characteristic; S400, selecting three features with the highest correlation with the Young's modulus of the titanium alloy according to the feature parameters as target features; the screening process is: using Pearson correlation analysis to remove feature parameters with a correlation greater than 0.9; then using a random forest algorithm to perform importance analysis on the remaining feature parameters, selecting three features with the greatest impact on the Young's modulus, and using the three features as target features; S500, using a plurality of machine learning models respectively, taking the characteristic parameters corresponding to the target features in step S400 and the original data set remaining after removing the original data corresponding to the Young's modulus lower than 50 GPa in the original data set in step S200 as the input of each machine learning model, and outputting the Young's modulus corresponding to the original data set, thereby determining the nonlinear relationship among the composition, characteristic parameters and Young's modulus of the titanium alloy, and completing the training of each machine learning model; S600, presetting a virtual component, using a plurality of machine learning models that have been trained, taking the virtual component as input, and outputting the Young's modulus corresponding to different component contents under the virtual component, that is, outputting a predicted value; S700, obtaining the metastable β titanium alloy composition corresponding to the minimum Young's modulus from the predicted value in step S600, and completing the composition design. Example 2

[0031] This embodiment provides a composition design method for a low Young's modulus metastable β titanium alloy, comprising: S100, obtain titanium alloy; S200, performing molecular dynamics simulation on the titanium alloy using a MEAM potential function to obtain a composition of the titanium alloy and an original data set of Young's modulus corresponding to the composition, and removing original data corresponding to a Young's modulus lower than 50 GPa in the original data set; S300, determining characteristic parameters corresponding to 10 characteristics of the titanium alloy according to the composition of the titanium alloy, the 10 characteristics are Pauling electronegativity (Pauling EN), density (Density), atomic weight (Atomic Weight), valence electron number (Valence e), melting point (Melting Point), specific heat (Specific Heat), metal radius (Metallic Radius), total electron number (Total Electrons), atomic radius (Atomic Radii) and first ionization energy (Firstionization energy), and the characteristic parameters are parameter values ​​corresponding to the above characteristics and the variance of the parameter value of each characteristic; S400, selecting three features with the highest correlation with the Young's modulus of the titanium alloy according to the feature parameters as target features; the screening process is: using Pearson correlation analysis to remove feature parameters with a correlation greater than 0.9; then using a random forest algorithm to perform importance analysis on the remaining feature parameters, selecting three features with the greatest impact on the Young's modulus, and using the three features as target features; S500, using a plurality of machine learning models respectively, taking the characteristic parameters corresponding to the target features in step S400 and the original data set remaining after removing the original data corresponding to the Young's modulus lower than 50 GPa in the original data set in step S200 as the input of each machine learning model, and outputting the Young's modulus corresponding to the original data set, thereby determining the nonlinear relationship among the composition, characteristic parameters and Young's modulus of the titanium alloy, and completing the training of each machine learning model; S600, presetting a virtual composition, and using screening factors to screen out a metastable β titanium alloy from titanium alloys with different composition contents under the virtual composition; the screening process is to select screening factors as d orbital binding energy (Md), average electron orbital parameter (Bo), average valence electron concentration (VEC), Mo equivalent method (Mo) and bonding force (Bonding Force), and when the titanium alloys with different composition contents under the virtual composition meet all the screening factors of the metastable β titanium alloy, determine that the corresponding titanium alloy is a metastable β titanium alloy; S700 uses a variety of machine learning models that have been trained, takes the composition of the metastable β titanium alloy in the virtual composition as input, and outputs the Young's modulus corresponding to different component contents under the composition of the metastable β titanium alloy, that is, the output obtains the predicted value; S800, obtaining the metastable β titanium alloy composition corresponding to the minimum Young's modulus from the predicted value in step S700, and completing the composition design. Example 3

[0032] This embodiment provides a composition design method for a low Young's modulus metastable β titanium alloy, comprising: S100, obtaining a titanium alloy, wherein the composition of the titanium alloy is TiZrHfNbTa; S200, performing molecular dynamics simulation on the titanium alloy using a MEAM potential function to obtain a composition of the titanium alloy and an original data set of Young's modulus corresponding to the composition, and removing original data corresponding to a Young's modulus lower than 50 GPa in the original data set; The raw data in this step include Ti, Zr, Hf, Nb, Ta alloy elements and Young's modulus data. The data contains 1780 sets of samples, and there are 20 thermodynamic and physical characteristic parameters established according to the chemical composition of the alloy; S300, determining characteristic parameters corresponding to 10 characteristics of the titanium alloy according to the composition of the titanium alloy, the 10 characteristics are Pauling electronegativity (Pauling EN), density (Density), atomic weight (Atomic Weight), valence electron number (Valence e), melting point (Melting Point), specific heat (Specific Heat), metal radius (Metallic Radius), total electron number (Total Electrons), atomic radius (Atomic Radii) and first ionization energy (Firstionization energy), and the characteristic parameters are parameter values ​​corresponding to the above characteristics and the variance of the parameter value of each characteristic; The high entropy alloy model was built with 2000 atoms and subjected to relaxation treatment. The alloy model was simulated using molecular dynamics (TiZrHfNbTa-MEAM potential function), and the elastic constants of the alloy were calculated to obtain the Young's modulus of the alloy. Since the original data was simulated data, when the Young's modulus was lower than 50 GPa, it was deleted. S400, according to the characteristic parameters, three features with the highest correlation with the Young's modulus of the titanium alloy are selected as target features; the screening process is: using Pearson correlation analysis, when a feature is highly correlated with the other multiple features (>0.9), it is deleted, and the importance of the remaining feature parameters is analyzed by random forest algorithm, multiple features are sorted according to feature importance, and the importance of the features corresponding to the order from front to back is added, and all features before the sum of importance is greater than 0.98 are retained, and the retained features are used as target features. According to the above process, three target features are determined in this embodiment; the three target features are Pauling electronegativity (Pauling EN), metal radius (Metallic Radius) and atomic weight (Atomic Weight). The Pearson correlation analysis results of specific features are detailed in Figure 2 ; S500, using 6 machine learning models respectively, taking the characteristic parameters corresponding to the target features in step S400 and the original data set remaining after removing the original data corresponding to the Young's modulus lower than 50 GPa in the original data set in step S200 as the input of each machine learning model, and outputting the Young's modulus corresponding to the original data set, thereby determining the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy, and completing the training of each machine learning model; Among them, the six machine learning models are support vector machine regression model (Support VectorRegression), multilayer perceptron regression model (MLP regression), extreme gradient boosting regression model (XGBoostRegression), AdaBoost regression model (AdaBoost Regressor), decision tree regression model (DecisionTreeRegressor) and gradient boosting regression model (GradientBoostingRegressor); Figure 1 This is the performance evaluation diagram of the machine learning model for predicting Young's modulus of β titanium alloy. Figure 1 From Table 1 and Table 2, see Figure 1 , and combined with the following Tables 1 and 2, the performance of the above machine learning model is excellent on both the training set (10-CV) and the test set (Test); Table 1 Performance indicators of machine learning models under 10-fold cross-validation strategy

[0033] Table 2 Performance index test composition of machine learning models on the test set

[0034] S600, presetting a virtual component, and using screening factors to screen out a metastable β titanium alloy from titanium alloys with different component contents under the virtual component; the screening process is to select screening factors as d orbital binding energy (Md), average electron orbital parameter (Bo), average valence electron concentration (VEC), Mo equivalent method (Mo) and bonding force (Bonding Force), and when the titanium alloys with different component contents under the virtual component meet all the screening factors of the metastable β titanium alloy, it is determined that the corresponding titanium alloy is a metastable β titanium alloy. After the screening is completed, the amount of data to be predicted in this embodiment is reduced from 1,000,000 groups to 62,749 groups; S700, selecting three machine learning models with the highest accuracy from the multiple machine learning models that have completed training, using the three machine learning models, taking the composition of the metastable β titanium alloy in the virtual composition as input, and outputting the Young's modulus corresponding to different component contents under the composition of the metastable β titanium alloy, that is, outputting the predicted value; please refer to Table 1, in this embodiment, the three machine learning models with the highest accuracy, namely, the support vector machine regression model (Support Vector Regression), the multilayer perceptron regression model (MLP regression) and the extreme gradient boosting regression model (XGBoost Regression) are selected from Table 1 to output the predicted value; S800, obtaining the metastable β titanium alloy composition corresponding to the minimum Young's modulus from the predicted value in step S700, and completing the composition design; the process may further include: Each machine learning model has a prediction value under different contents of virtual components. This embodiment uses the support vector machine regression model (Support Vector Regression), the multilayer perceptron regression model (MLP regression) and the extreme gradient boosting regression model (XGBoost Regression) in step S700 for prediction, which is equivalent to having three prediction values ​​under different contents of virtual components; Furthermore, the predicted values ​​under each virtual component of the content can be scored, and finally the virtual component with the lowest score corresponding to the content is selected as the target component obtained by screening in this embodiment, and the target component is the metastable β titanium alloy component corresponding to the above-mentioned minimum Young's modulus; For the above score, it can be expressed as: , where Mean is the weighted mean and Std is the standard deviation; according to this equation, when using three machine learning models, is the average of the predicted values ​​obtained by the three machine learning models under any content of the virtual component, and Std is the variance of the predicted values ​​obtained by the three machine learning models under any content of the virtual component. From this, the score Score of Young's modulus under each content of the virtual component can be obtained. The virtual component with the lowest score corresponding to the content is selected from all the scores as the target component to complete the composition design. After completing the composition design, the corresponding titanium alloy can be further manufactured according to the target composition.

[0035] It should be noted that, in order to verify the accuracy of the above-mentioned component design process, after obtaining the predicted value in step S700, the predicted value (Young's modulus) can be compared with the Young's modulus of the experimental value, and the correlation between the predicted value and the experimental value can be compared. The experimental value can be obtained by querying existing literature. The verification process can also specifically include the following steps: S900, after obtaining the predicted value in step S700, the predicted value (Young's modulus) is compared with the Young's modulus of the experimental value. The experimental value can be obtained by the following process: consulting existing literature to search for relevant β titanium alloy data. β titanium alloys include ternary, quaternary and quinary alloys. The theoretical screening method of metastable β titanium alloys is used to roughly screen the data collected in the literature data; For the rough screening process, it can be as follows: in the experimental data of the searched literature, it must be ensured that the titanium alloy is β titanium alloy, the literature must have carried out a tensile test on the titanium alloy and measured its Young's modulus, the Young's modulus data of β titanium alloy measured by nanoindentation is not collected in this step, and for conflicting literature data, the high-quality journal data shall prevail.

[0036] See also Figures 4 to 9 , Figures 4 to 9 In the figure, the horizontal axis is the simulated value, and the vertical axis is the predicted value. The simulated value and the predicted value are distributed near the 45-degree line, indicating that the support vector machine regression model (Support Vector Regression), the multilayer perceptron regression model (MLP regression) and the extreme gradient boosting regression model (XGBoost Regression) of this embodiment have a good fitting effect on the original data of molecular dynamics calculation. Please refer to Fig.10 , Fig.11 and Fig.12 ,from Figures 10 to 12 The change trend of the predicted value is observed in the experiment. The results show that the change trend of the predicted value is consistent with the change trend of the experimental value. The difference between the predicted value and the experimental value is shown in Table 3 below. In Table 3 below, the data corresponding to Young modulus is the Young modulus of the experimental value. SVR Prediction, XGB Prediction and MLPR Prediction are the predicted values ​​predicted by three machine learning models, namely, Support Vector Regression, XGBoost Regression and MLP regression. It can be seen from Table 3 that for titanium alloys with different component contents, there is a large difference between the predicted value of Young's modulus and the experimental value (greater than the experimental value). This is because the above process is based on an ideal environment where the crystal is completely defect-free, so there is such a difference. The purpose of this embodiment is to screen out candidate alloy components with low Young's modulus, not just considering the difference between the model predicted value and the experimental value. As long as the change trend of the predicted value is consistent with the change trend of the experimental value, the minimum Young's modulus can be obtained.

[0037] Table 3 Comparison of machine learning model prediction values ​​and literature experimental values

[0038] In the description of the above embodiments, specific features, structures, materials or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0039] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A composition design method for a low Young's modulus metastable β titanium alloy, characterized in that: include: Get titanium alloy; A molecular dynamics simulation is performed on the titanium alloy using a MEAM potential function to obtain an original data set of the composition of the titanium alloy and the Young's modulus corresponding to the composition; Determining characteristic parameters corresponding to a plurality of characteristics of the titanium alloy according to the composition of the titanium alloy; According to the characteristic parameters, multiple features with the highest correlation with the Young's modulus of the titanium alloy are selected as target features; Using a plurality of machine learning models respectively, according to the original data set and the characteristic parameters corresponding to the target features, determine the nonlinear relationship between the composition, characteristic parameters and Young's modulus of the titanium alloy, and complete the training of each machine learning model; Preset a virtual component, and use multiple machine learning models that have been trained to predict the Young's modulus corresponding to different component contents under the virtual component to obtain a predicted value; The metastable β titanium alloy composition corresponding to the minimum Young's modulus is obtained in the predicted value.

2. The composition design method of low Young's modulus metastable β titanium alloy according to claim 1, characterized in that: After the virtual component is preset, and before the Young's modulus corresponding to different component contents under the virtual component is predicted, the method further includes: The metastable β titanium alloy is screened out from titanium alloys with different component contents under the virtual composition.

3. The composition design method of low Young's modulus metastable β titanium alloy according to claim 2, characterized in that: The step of selecting a metastable β titanium alloy from titanium alloys with different composition contents under the virtual composition comprises: The metastable β titanium alloy is screened out from titanium alloys with different component contents under the virtual composition using screening factors; the screening factors include one or more of d orbital binding energy, average electron orbital parameter, average valence electron concentration, Mo equivalent method and binding force.

4. The composition design method of low Young's modulus metastable β titanium alloy according to claim 3, characterized in that: When titanium alloys with different component contents under the virtual composition satisfy all the screening factors of the metastable β titanium alloy, it is determined that the corresponding titanium alloy is a metastable β titanium alloy.

5. The composition design method of low Young's modulus metastable β titanium alloy according to claim 4, characterized in that: In the process of determining characteristic parameters corresponding to multiple characteristics of the titanium alloy according to the composition of the titanium alloy, the characteristics include one or more of Pauling electronegativity, density, atomic weight, number of valence electrons, melting point, specific heat capacity, metal radius, total number of electrons, atomic radius and first ionization energy.

6. The composition design method of low Young's modulus metastable β titanium alloy according to claim 5, characterized in that: The feature parameters include parameters corresponding to the features and variances of the parameters corresponding to each feature.

7. The composition design method of low Young's modulus metastable β titanium alloy according to claim 6, characterized in that: In the process of determining the nonlinear relationship among the composition, characteristic parameters and Young's modulus of the titanium alloy based on the original data set and the characteristic parameters corresponding to the target features, multiple machine learning models are used. The input of each machine learning model is the characteristic parameters corresponding to the target features and the original data set, and the output is the Young's modulus corresponding to the original data set.

8. The composition design method of low Young's modulus metastable β titanium alloy according to claim 7, characterized in that: In the process of predicting the Young's modulus corresponding to different component contents under the virtual component through the virtual component and the characteristic parameters corresponding to the virtual component, a plurality of machine learning models that have been trained are used, with the input being the virtual component and the output being the Young's modulus corresponding to different component contents under the virtual component.

9. The composition design method of low Young's modulus metastable β titanium alloy according to claim 8, characterized in that: After obtaining the composition of the titanium alloy and the original data set of the Young's modulus corresponding to the composition, the method further includes: The original data corresponding to the Young's modulus lower than 50 GPa in the original data set are removed.

10. The composition design method of low Young's modulus metastable β titanium alloy according to claim 9, characterized in that: The method of screening multiple features with the highest correlation with the Young's modulus of the titanium alloy according to the feature parameters includes: Using Pearson correlation analysis, feature parameters with correlation greater than 0.9 were removed; The random forest algorithm is then used to perform importance analysis on the remaining feature parameters to screen out multiple features that have the greatest impact on Young's modulus.

Images