Method for predicting stability of diboride high-entropy ceramic based on machine learning
Through machine learning-based methods, the formation energy prediction model of diboride high-entropy ceramics is established, which solves the problems of inefficiency and high cost of traditional design methods, and achieves efficient and accurate prediction effects of its stability.
Patent Information
- Application Number
- CN202411152608.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-21
- Publication Date
- 2025-01-07
AI Technical Summary
Traditional methods are inefficient and costly when designing high-entropy ceramic materials, making it difficult to fully cover all possible combinations and cannot effectively predict their stability.
Using machine learning-based methods, by collecting the physical and chemical parameters and thermodynamic parameters of diboride high-entropy ceramics, establishing a Lasso regression model, screening characteristic variables and building an energy prediction model, using decision trees, random forests, extreme random forests or gradients Improve the tree model for prediction.
It realizes efficient prediction of the stability of high-entropy ceramics of diboride, shortens the design cycle, reduces R&D costs, and the built model has strong generalization capabilities and high accuracy, and the 50% cross-verification score can reach 0.96.
Smart Images
Figure CN119274702A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of new material development, and in particular to a method for predicting the stability of diboride high-entropy ceramics based on machine learning. Background Art
[0002] As an emerging class of materials, high entropy ceramics have shown great application potential in aerospace, energy, electronics, biomedicine and other fields due to their unique high-temperature stability, oxidation resistance, high hardness and wear resistance. However, when faced with complex multi-component systems such as high entropy ceramics, traditional material design methods such as exploration based on physical intuition or trial and error experiments are often inefficient and costly, and it is difficult to fully cover all possible combinations.
[0003] In recent years, with the significant improvement of computing power and the rapid development of big data analysis technology, machine learning, as a powerful data-driven tool, has been increasingly introduced into the field of materials science. By learning the complex relationship between material properties and structure from limited experimental data, machine learning can achieve rapid prediction of the performance of new compositions, thereby greatly shortening the design cycle of new materials and reducing R&D costs. In materials science, a variety of machine learning models have been successfully applied to the prediction of material properties. These models have their own characteristics and show different advantages when processing different types of material data. Summary of the invention
[0004] The present invention provides a method for predicting the stability of diboride high-entropy ceramics based on machine learning, which provides an efficient and low-cost approach for the development of diboride high-entropy ceramics.
[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide a method for predicting the stability of diboride high entropy ceramics based on machine learning, comprising the following steps:
[0006] Obtaining characteristic variables of the diboride high entropy ceramic to be tested;
[0007] Inputting the characteristic variables into a formation energy prediction model to obtain a formation energy prediction value of the diboride high entropy ceramic to be tested;
[0008] The stability of the diboride high entropy ceramic to be tested is determined according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested.
[0009] The characteristic variables are determined in the following way:
[0010] The physicochemical parameters, thermodynamic parameters and kinetic parameters of diboride high entropy ceramics composed of various transition metal elements and boron elements were collected, and the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter were obtained by statistical methods;
[0011] Calculate the formation energy of diboride high entropy ceramics composed of various transition metal elements and boron;
[0012] A Lasso regression model was established with the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter as input features and the formation energy of diboride high entropy ceramics as output features, and characteristic variables were screened out from the input features according to the Lasso regression model.
[0013] The loss function of the Lasso regression model is: Among them, Loss(ω) is the loss function of the Lasso regression model, n is the number of samples, p is the number of input features, and y i is the true formation energy of the ith sample, is the predicted value of the formation energy of the i-th sample, ω j is the coefficient of the jth input feature, and α is the regularization parameter.
[0014] The formation energy of the diboride high entropy ceramic is calculated as follows: Among them, E form is the formation energy of diboride high entropy ceramics, E tot is the total energy of the metal diboride in the supercell, N k is the atomic number of the kth transition metal element in bulk state, is the single atom energy of the kth transition metal element in the bulk state, and K is the total number of transition metal elements in the diboride high entropy ceramic.
[0015] The formation energy prediction model is obtained by the following method:
[0016] Use the selected feature variables to construct a data set and divide the data set into a training set and a test set;
[0017] The machine learning model is trained using the training set, optimized by adjusting hyperparameters, and tested using R 2 Value and P-fold cross validation score to evaluate the performance of the optimized machine learning model;
[0018] When R 2 When the value and P-fold cross-validation score meet the preset requirements, the training of the machine learning model is completed and a prediction model is obtained.
[0019] The machine learning model is a decision tree model, a random forest model, an extreme random forest model or a gradient boosting tree model.
[0020] When determining the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the smaller the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the higher the stability of the diboride high entropy ceramic to be tested.
[0021] The technical solution adopted by the present invention to solve the technical problem is: to provide a diboride high entropy ceramic stability prediction device based on machine learning, comprising:
[0022] An acquisition module, used for acquiring characteristic variables of the diboride high entropy ceramic to be tested;
[0023] A prediction module, used for inputting the characteristic variables into a formation energy prediction model to obtain a predicted value of the formation energy of the diboride high entropy ceramic to be tested;
[0024] The determination module is used to determine the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested.
[0025] The technical solution adopted by the present invention to solve its technical problem is: to provide an electronic device, including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the above-mentioned method for predicting the stability of diboride high entropy ceramics based on machine learning are implemented.
[0026] The technical solution adopted by the present invention to solve its technical problem is: providing a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned diboride high-entropy ceramic stability prediction method based on machine learning are implemented.
[0027] Beneficial Effects
[0028] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art: the present invention uses machine learning to predict the stability of diboride high entropy ceramics, and can learn the relationship between material properties and structure from limited experimental data, and then predict the performance of new compositions, greatly reducing experimental costs and R&D cycles. The prediction model constructed by the present invention has strong generalization ability and can better handle new data in the future. At the same time, the accuracy reaches 0.95, and the five-fold cross-validation score can reach 0.96, showing excellent data fitting ability and prediction effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 is a flow chart of a method for predicting stability of diboride high entropy ceramics based on machine learning according to a first embodiment of the present invention;
[0030] Figure 2is a feature screening result diagram under different regularization parameters in the first embodiment of the present invention;
[0031] Figure 3 is a heat map of the Pearson correlation coefficient matrix in the first embodiment of the present invention;
[0032] Figure 4 is a learning curve diagram of four machine learning models in the first embodiment of the present invention;
[0033] Figure 5 is a fitted regression graph of four machine learning models in the first embodiment of the present invention;
[0034] Figure 6 is the R of the four machine learning models in the first embodiment of the present invention. 2 Comparison chart of values and five-fold cross validation scores;
[0035] Figure 7 It is a mean Shapley value graph of the selected characteristic variables in the first embodiment of the present invention. DETAILED DESCRIPTION
[0036] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.
[0037] The first embodiment of the present invention relates to a method for predicting the stability of diboride high entropy ceramics based on machine learning. This embodiment is described by taking pentavalent diboride high entropy ceramics as an example. Figure 1 As shown, the following steps are included:
[0038] Step 1, obtaining characteristic variables of the five-element diboride high entropy ceramic to be tested. The characteristic variables in this step are determined by the following method:
[0039] Firstly, the physicochemical parameters, thermodynamic parameters and kinetic parameters of pentacyclic diboride high entropy ceramics composed of transition metal elements (Ti, Zr, Hf, V, Nb, Ta, Mo, Cr and W) and boron were collected, and then the collected data were processed by Python using common methods in descriptive statistics to obtain the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter;
[0040] Then, the formation energy of the five-element diboride high entropy ceramics composed of transition metal elements (Ti, Zr, Hf, V, Nb, Ta, Mo, Cr and W) and boron is calculated as follows: Among them, E form is the formation energy of diboride high entropy ceramics, E tot is the total energy of the metal diboride in the supercell, N k is the atomic number of the kth transition metal element in bulk state, is the single atom energy of the kth transition metal element in the bulk state, K is the total number of transition metal elements in the diboride high entropy ceramic, and since this embodiment predicts the five-element diboride high entropy ceramic, K=5.
[0041] Afterwards, a Lasso regression model was established with the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter as input features and the formation energy of pentacyclic diboride high entropy ceramics as output features, and characteristic variables were screened out from the input features according to the Lasso regression model.
[0042] In this implementation, the complexity of the model is penalized by introducing a loss function (least squares loss and L1 regularization term), which makes the model coefficients tend to be sparse and reduces the risk of overfitting. The loss function of the Lasso regression model is expressed as: Among them, Loss(ω) is the loss function of the Lasso regression model, n is the number of samples, p is the number of input features, and y i is the true formation energy of the ith sample, is the predicted value of the formation energy of the i-th sample, ω j is the coefficient of the jth input feature, and α is the regularization parameter.
[0043] Figure 2 The characteristic coefficients and five-fold cross-validation scores selected by the Lasso regression model under different regularization parameter α values (0.0009, 0.001, 0.002, 0.01). Among them, when α=0.001, the five-fold cross-validation score can reach 0.9794, which means that when using the Lasso regression model for feature selection, for the selected α=0.001, the model has a very good fitting effect on the data. Figure 3 is the correlation between the 12 feature variables included in the optimal descriptor combination.
[0044] Step 2: Input the characteristic variables into the formation energy prediction model to obtain the formation energy prediction value of the tested pentacyclic diboride high entropy ceramic. The prediction model in this step can be obtained by the following method:
[0045] First, using the data collected when determining the characteristic variables, the data set was constructed using the selected characteristic variables, and the data set was divided into a training set and a test set in a ratio of 7:3;
[0046] Next, the machine learning model is trained using the training set, optimized by adjusting the hyperparameters, and tested using R 2 The performance of the optimized machine learning model is evaluated by using the P-fold cross validation score. In this embodiment, the machine learning model can adopt a decision tree model, a random forest model, an extreme random forest model and a gradient boosting tree model. Figure 4 The learning curves of the above four machine learning models on the data set are shown in Figure 2. By observing the R of the machine learning model training set and cross-validation set under different numbers of training samples, 2 Score trend analysis to evaluate the performance of machine learning models and diagnose overfitting or underfitting. Figure 5 As the fitted regression scatter plot of the machine learning model, we can get Figure 5 The point distribution of D (i.e., the gradient boosting tree model) is closer to the oblique line, indicating that the gradient boosting tree model fits the data set better than the other three models.
[0047] The gradient boosting tree model obtained by adjusting the hyperparameters (n_estimators represents the number of decision trees to be used, max_depth represents the maximum depth of each decision tree, learning_rate represents the learning rate, min_samples_split represents the minimum number of samples required for node splitting, and min_samples_leaf represents the minimum number of samples required for leaf nodes) has an R score of 0. 2 The value is 0.95, and the five-fold cross validation can reach 0.96. Figure 6 R for four machine learning models 2 The R value of the gradient boosting tree model can be obtained by comparing the values and the five-fold cross validation scores. 2 The R value and five-fold cross-validation score are higher than those of the other three models, indicating that the gradient boosting tree model performs best on this dataset, fits the data best, and can predict the changes of the target variable more accurately than the other three models. 2 The value and the five-fold cross-validation score meet the preset requirements, and the machine learning model training is completed. In this implementation, the trained gradient boosting tree model is used as the formation energy prediction model.
[0048] In this step, the 12 selected characteristic variables are input into the formation energy prediction model, which can predict the formation energy of the five-element high entropy diboride combination to be tested.
[0049] Step 3, determining the stability of the five-component diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the five-component diboride high entropy ceramic to be tested. In this step, the smaller the predicted value of the formation energy of the five-component diboride high entropy ceramic to be tested, the higher the stability of the five-component diboride high entropy ceramic to be tested, that is, the greater the possibility of the composition being formed, the easier it is for the five-component high entropy diboride to be synthesized into a single phase.
[0050] Figure 7 The mean Shapley value diagram of the selected feature variables shows the overall impact of the selected feature variables on the prediction of the gradient boosting tree model, and more intuitively shows the contribution of each feature variable in a single sample to the prediction result, as well as the interaction between the feature variables.
[0051] It is not difficult to find that the present invention uses machine learning to predict the stability of diboride high-entropy ceramics, which can learn the relationship between material properties and structure from limited experimental data, and then predict the performance of new compositions, greatly reducing experimental costs and R&D cycles.
[0052] The second embodiment of the present invention relates to a device for predicting the stability of diboride high entropy ceramics based on machine learning, comprising:
[0053] An acquisition module, used for acquiring characteristic variables of the diboride high entropy ceramic to be tested;
[0054] A prediction module, used for inputting the characteristic variables into a formation energy prediction model to obtain a predicted value of the formation energy of the diboride high entropy ceramic to be tested;
[0055] The determination module is used to determine the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested.
[0056] The characteristic variables are determined in the following way:
[0057] The physicochemical parameters, thermodynamic parameters and kinetic parameters of diboride high entropy ceramics composed of various transition metal elements and boron elements were collected, and the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter were obtained by statistical methods;
[0058] Calculate the formation energy of diboride high entropy ceramics composed of various transition metal elements and boron;
[0059] A Lasso regression model was established with the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter as input features and the formation energy of diboride high entropy ceramics as output features, and characteristic variables were screened out from the input features according to the Lasso regression model.
[0060] The loss function of the Lasso regression model is: Among them, Loss(ω) is the loss function of the Lasso regression model, n is the number of samples, p is the number of input features, and y i is the true formation energy of the ith sample, is the predicted value of the formation energy of the i-th sample, ω j is the coefficient of the jth input feature, and α is the regularization parameter.
[0061] The formation energy of the diboride high entropy ceramic is calculated as follows: Among them, E form is the formation energy of diboride high entropy ceramics, E tot is the total energy of the metal diboride in the supercell, N k is the atomic number of the kth transition metal element in bulk state, is the single atom energy of the kth transition metal element in the bulk state, and K is the total number of transition metal elements in the diboride high entropy ceramic.
[0062] The formation energy prediction model is obtained by the following method:
[0063] Use the selected feature variables to construct a data set and divide the data set into a training set and a test set;
[0064] The machine learning model is trained using the training set, optimized by adjusting hyperparameters, and tested using R 2 Value and P-fold cross validation score to evaluate the performance of the optimized machine learning model;
[0065] When R 2 When the value and P-fold cross-validation score meet the preset requirements, the training of the machine learning model is completed and a prediction model is obtained.
[0066] The machine learning model is a decision tree model, a random forest model, an extreme random forest model or a gradient boosting tree model.
[0067] When the determination module determines the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the smaller the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the higher the stability of the diboride high entropy ceramic to be tested.
[0068] A third embodiment of the present invention relates to an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the first embodiment of the method for predicting the stability of diboride high-entropy ceramics based on machine learning when executing the computer program.
[0069] A fourth embodiment of the present invention relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting the stability of diboride high-entropy ceramics based on machine learning of the first embodiment.
[0070] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.
[0071] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0072] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction method, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0073] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0074] 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 method for predicting the stability of diboride high entropy ceramics based on machine learning, characterized in that: The following steps are involved: Obtaining characteristic variables of the diboride high entropy ceramic to be tested; Inputting the characteristic variables into a formation energy prediction model to obtain a formation energy prediction value of the diboride high entropy ceramic to be tested; The stability of the diboride high entropy ceramic to be tested is determined according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested.
2. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 1, characterized in that: The characteristic variables are determined in the following way: The physicochemical parameters, thermodynamic parameters and kinetic parameters of diboride high entropy ceramics composed of various transition metal elements and boron elements were collected, and the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter were obtained by statistical methods; Calculate the formation energy of diboride high entropy ceramics composed of various transition metal elements and boron; A Lasso regression model was established with the mean, mean absolute deviation, range, maximum value, minimum value and mode of each parameter as input features and the formation energy of diboride high entropy ceramics as output features, and characteristic variables were screened out from the input features according to the Lasso regression model.
3. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 2, characterized in that: The loss function of the Lasso regression model is: Among them, Loss(ω) is the loss function of the Lasso regression model, n is the number of samples, p is the number of input features, and y i is the true formation energy of the ith sample, is the predicted value of the formation energy of the i-th sample, ω j is the coefficient of the jth input feature, and α is the regularization parameter.
4. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 2, characterized in that: The formation energy of the diboride high entropy ceramic is calculated as follows: Among them, E form is the formation energy of diboride high entropy ceramics, E tot is the total energy of the metal diboride in the supercell, N k is the atomic number of the kth transition metal element in bulk state, is the single atom energy of the kth transition metal element in the bulk state, and K is the total number of transition metal elements in the diboride high entropy ceramic.
5. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 2, characterized in that: The formation energy prediction model is obtained by the following method: Use the selected feature variables to construct a data set and divide the data set into a training set and a test set; The machine learning model is trained using the training set, optimized by adjusting hyperparameters, and tested using R 2 Value and P-fold cross validation score to evaluate the performance of the optimized machine learning model; When R 2 When the value and P-fold cross-validation score meet the preset requirements, the training of the machine learning model is completed and a prediction model is obtained.
6. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 5, characterized in that: The machine learning model is a decision tree model, a random forest model, an extreme random forest model or a gradient boosting tree model.
7. The method for predicting the stability of diboride high entropy ceramics based on machine learning according to claim 1, characterized in that: When determining the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the smaller the predicted value of the formation energy of the diboride high entropy ceramic to be tested, the higher the stability of the diboride high entropy ceramic to be tested.
8. A device for predicting the stability of diboride high entropy ceramics based on machine learning, characterized in that: include: An acquisition module, used for acquiring characteristic variables of the diboride high entropy ceramic to be tested; A prediction module, used for inputting the characteristic variables into a formation energy prediction model to obtain a predicted value of the formation energy of the diboride high entropy ceramic to be tested; The determination module is used to determine the stability of the diboride high entropy ceramic to be tested according to the predicted value of the formation energy of the diboride high entropy ceramic to be tested.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method for predicting the stability of diboride high entropy ceramics based on machine learning are implemented as described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting the stability of diboride high entropy ceramics based on machine learning as described in any one of claims 1 to 7 are implemented.