Machine learning method and device for predicting giant magnetostrictive property of RFe2 type material
By constructing the composition-performance-magnetic scaling dataset of RFe2 type materials and using machine learning models, the problem of discovering RFe2 type magnetostrictive materials in the prior art is solved, and the problem of discovering RFe2 type magnetostrictive materials with both large strain, narrow hysteresis and wide working temperature zones is realized. Fast and effective material discovery reduces R&D costs.
Patent Information
- Application Number
- CN202411007212.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-11-22
AI Technical Summary
It is difficult to quickly discover RFe2 magnetostrictive materials with large strain, narrow hysteresis and wide working temperature zones, resulting in long experimental cycles and high cost.
By constructing the composition-performance-magnetic scaling dataset of RFe2 type materials, using machine learning models to search for materials with these characteristics in a huge unknown component space, including obtaining elemental composition and physical parameter data, performing weighted calculations and model training, The training set and the test set are divided to verify the model accuracy and are ultimately used to predict the giant magnetostrictive performance of the material.
It has achieved rapid screening of RFe2 magnetostrictive materials with large strain, narrow hysteresis and wide temperature domains, reducing R&D costs and significantly improving material discovery efficiency and accuracy.
Smart Images

Figure CN119005366A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of magnetostrictive materials and machine learning material prediction technology, and in particular to a machine learning method and device for predicting the giant magnetostrictive performance of RFe2 type materials. Background Art
[0002] As a typical rare earth magnetostrictive material, RFe2 compounds are widely used in high-tech fields such as high-precision sensors, micro-displacement actuators and ultra-precision transducers. Generally, properties such as large strain, narrow hysteresis and wide operating temperature range are crucial for the application of magnetostrictive materials in devices. At present, obtaining materials with these properties mainly relies on the traditional "trial and error method". This research and development model greatly prolongs the experimental cycle, which in turn causes a surge in experimental costs. Therefore, the preparation of RFe2 magnetostrictive materials that have large strain, narrow hysteresis and wide operating temperature range faces considerable challenges. In view of this, there is an urgent need to find new technologies and solutions to accelerate the discovery of high-performance RFe2 materials.
[0003] Currently, machine learning has become a strategy to efficiently predict material properties and accelerate material design. However, the complex magnetic-crystal structure of the RFe2 magnetostrictive material system makes it more difficult to build a machine learning database, which in turn hinders the development of machine learning in this field.
[0004] Based on this, for the RFe2 magnetostrictive material system, it is urgently necessary to establish a set of machine learning solutions or paths that can accurately predict the performance of large strain, narrow hysteresis and wide operating temperature range, so as to quickly and effectively realize the discovery of new materials and reduce R&D costs. Summary of the invention
[0005] (I) Purpose of the invention
[0006] The purpose of the present invention is to provide a machine learning method and device for predicting the giant magnetostrictive properties of RFe2 type materials. The method establishes a composition-performance-magnetostriction data set of the RFe2 type material system, and searches for magnetostrictive materials with large strain, narrow hysteresis and wide temperature range in a huge unknown composition space based on machine learning, thereby facilitating the development of smart devices.
[0007] (II) Technical solution
[0008] To solve the above problems, the first aspect of the present invention provides a machine learning method for predicting the giant magnetostrictive properties of RFe2 type materials, comprising the following steps:
[0009] Construct a composition-properties-magnetostriction dataset for RFe2-type materials;
[0010] Dividing the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determining a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set;
[0011] Using the final data set to train a machine learning model to obtain a trained machine learning model, wherein the training set is used to train the machine learning model, and the test set is used to verify the prediction accuracy of the machine learning model;
[0012] The giant magnetostrictive properties of the RFe2 type material are predicted by the trained machine learning model.
[0013] Further, the construction of the composition-performance-magnetostriction data set of the RFe2 type material comprises the following steps:
[0014] Obtaining the content of each element in a known RFe2 type material compound and the corresponding magnetostriction value of the RFe2 type material compound;
[0015] Obtain the physical parameter data of each element in the RFe2 type material compound according to the preset element physical parameter performance table;
[0016] Based on the stoichiometric ratio of the RFe2 type material compound and the physical parameter data of each element, the average value of the corresponding physical parameters of the RFe2 type material compound is obtained by weighted calculation;
[0017] Calculating the physical parameter performance data of the corresponding RFe2 type material compound respectively through the average values of the corresponding physical parameters of the RFe2 type material compound;
[0018] Taking the composition content of each element in the RFe2 type material compound and the calculated physical parameter performance data of the corresponding RFe2 type material compound as the characteristic values, and the corresponding magnetostriction value of the RFe2 type material compound as the target value, a composition-performance-magnetostriction data set of RFe2 type materials is constructed.
[0019] Furthermore, based on the stoichiometric ratio of the RFe2 type material compound and the physical parameter data of each element, the average value of the physical parameters corresponding to the RFe2 type material compound is obtained by weighted calculation, using the following calculation formula:
[0020]
[0021] in, represents the average value of the physical parameters of the RFe2 type material compound obtained by weighted average calculation; n is the total number of atoms in the RFe2 type material compound; C i is the stoichiometric ratio of the i-th atom in the RFe2-type material compound; fi It is the physical property data of the i-th atom in the RFe2 type material compound.
[0022] Furthermore, the physical parameter performance data of the corresponding RFe2 type material compound are calculated respectively by the average values of the corresponding physical parameters of the RFe2 type material compound, and the calculation process is as follows:
[0023]
[0024] or
[0025] Among them, f i,1 Represents the physical parameter performance values of RFe2 type material compounds.
[0026] Further, the composition-performance-magnetostriction data set is divided into training sets and test sets of different proportions, and the final data set is determined according to the calculation results of the training sets and the test sets at different proportions, including:
[0027] The training set and the test set are divided into two parts according to preset ratios:
[0028] The mean absolute error (MAE) and root mean square error (RMSE) of the test set under different division ratios are calculated, and the set with the smallest MAE and RMSE of the test set is selected as the final division ratio of the training set and the test set to determine the final data set.
[0029] Furthermore, the using the final data set to train the machine learning model to obtain the trained machine learning model includes:
[0030] During the training process of the machine learning model,
[0031] The coefficient of determination R 2 evaluating the predictive performance of the machine learning model;
[0032] The model hyperparameters are adjusted to optimize the machine learning model, and the optimal hyperparameters of the model are determined by using a ten-fold cross validation method and a grid search method;
[0033] Combine the optimal hyperparameters of the model and the determination coefficient R 2 The predictive performance of the machine learning models is jointly evaluated.
[0034] Furthermore, the determination coefficient R 2 The calculation formula is as follows:
[0035]
[0036] Among them, y i and prerepresent the true value and the predicted value respectively. Represents y i The average value of .
[0037] Furthermore, the method further comprises:
[0038] A target RFe2 type material corresponding to the predicted giant magnetostrictive performance of the RFe2 type material is prepared, and the magnetostrictive performance of the target RFe2 type material is tested.
[0039] Further, the preparation of a target RFe2 type material corresponding to the predicted giant magnetostrictive performance of the RFe2 type material, and testing the magnetostrictive performance of the target RFe2 type material, comprises the following steps:
[0040] Selecting components with excellent magnetostrictive properties from the target RFe2 type material for experimental preparation to obtain RFe2 type prepared material;
[0041] The magnetostriction values of the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field are tested respectively, and the difference in magnetostriction between the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field is calculated.
[0042] A second aspect of the present invention provides a machine learning system for predicting giant magnetostrictive properties of RFe2-type materials, the system comprising:
[0043] Building module for constructing composition-properties-magnetostriction datasets for RFe2-type materials;
[0044] A partitioning module, used to divide the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determine a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set;
[0045] A training module, used to train a machine learning model using the final data set to obtain a trained machine learning model, wherein the training set is used to train the machine learning model, and the test set is used to verify the prediction accuracy of the machine learning model;
[0046] A prediction module is used to predict the giant magnetostrictive performance of the RFe2 type material through the trained machine learning model.
[0047] The machine learning method and device for predicting the giant magnetostrictive performance of RFe2 type materials provided by the present invention have the following beneficial technical effects: the method establishes a composition-performance-magnetostrictive data set of the RFe2 type material system, divides the composition-performance-magnetostrictive data set into training sets and test sets of different proportions, and determines the optimal division ratio of the training set and the test set by calculation; then the data set with the optimal division ratio is used to train the machine learning model, and the accuracy of the machine learning model is verified on the test set, and finally used to predict the magnetostrictive performance of RFe2 unknown materials. The present invention predicts the magnetostrictive performance of materials based on machine learning, and can quickly screen out RFe2 type magnetostrictive materials with large strain, narrow hysteresis and wide temperature range, truly realizing the rapid and effective discovery of new materials and greatly reducing research and development costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 is a flow chart of a machine learning method for predicting giant magnetostrictive properties of RFe2 type materials provided by an embodiment of the present invention;
[0049] Figure 2 is a schematic diagram of the regression prediction results of the machine learning model according to an embodiment of the present invention;
[0050] Figure 3 is a room temperature magnetostriction performance diagram of an embodiment of the present invention, and the gray illustration in the lower right corner is a local enlarged diagram of the magnetostriction curve under a specific magnetic field;
[0051] Figure 4 It is a comparison curve of the magnetostriction value of the embodiment of the present invention and the comparative example in the direction parallel to the magnetic field as a function of temperature. The illustration in the lower right corner is a comparison diagram of the working temperature range of the embodiment and the comparative example.
[0052] Figure 5 A flow chart of a machine learning device for predicting the giant magnetostrictive properties of RFe2 type materials provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0053] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.
[0054] like Figure 1 As shown, the present invention provides a machine learning method for predicting the giant magnetostrictive properties of RFe2 type materials, comprising the following steps:
[0055] S101, constructing a composition-property-magnetostriction dataset for RFe2-type materials;
[0056] S102, dividing the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determining a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set;
[0057] S103, training a machine learning model using the final data set to obtain a trained machine learning model, wherein the training set is used for training the machine learning model, and the test set is used for verifying the prediction accuracy of the machine learning model;
[0058] S104, predicting the giant magnetostrictive properties of the RFe2 type material through the trained machine learning model.
[0059] Furthermore, S101 specifically includes:
[0060] Step 1011, obtaining the content of each element component in a known RFe2 type material compound and the corresponding magnetostriction value of the RFe2 type material compound. The content of each element component in a known RFe2 type material compound and the corresponding magnetostriction value of the RFe2 type material compound can be obtained from published literature, for example;
[0061] Step 1012, obtaining physical parameter data of each element in the RFe2 type material compound according to preset element physical parameter properties; wherein the preset element physical parameter properties are, for example, a publicly available element physical parameter property table, wherein the physical parameters include atomic radius difference, melting point temperature, specific heat, thermal conductivity, electronegativity, valence electron concentration, bulk modulus, shear modulus, Poisson's ratio, and volume susceptibility related to electronic structure and microstructure, and elastic deformation;
[0062] Step 1013, based on the stoichiometric ratio of the RFe2 type material compound and the physical parameter data of each element, obtain the average value of the physical parameters corresponding to the RFe2 type material compound by weighted calculation, and determine it using the following calculation formula (1):
[0063]
[0064] In formula (1), represents the average value of the physical parameters of the RFe2 type material compound obtained by weighted average calculation; n is the total number of atoms in the RFe2 type material compound; C i is the stoichiometric ratio of the i-th atom in the RFe2-type material compound; f i It is the physical property data of the i-th atom in the RFe2 type material compound.
[0065] Step 1014, respectively calculating the physical parameter performance data of the corresponding RFe2 type material compound by using the average values of the corresponding physical parameters of the RFe2 type material compound, and the calculation process is shown in formula (2) or formula (3):
[0066]
[0067] Among them, f i,1 Represents the physical parameter performance values of RFe2 type material compounds.
[0068] Step 1015, using the composition content of each element in the RFe2 type material compound and the calculated physical parameter performance data of the corresponding RFe2 type material compound as characteristic values, and the corresponding magnetostriction value of the RFe2 type material compound as the target value, constructing a composition-performance-magnetostriction data set of the RFe2 type material system. The present invention can quickly screen out RFe2 type magnetostrictive materials with large strain, narrow hysteresis and wide temperature range by establishing a composition-performance-magnetostriction data set of the RFe2 type material system and predicting the magnetostriction properties of the material based on machine learning.
[0069] Furthermore, S102 specifically includes:
[0070] Step 1021, dividing the training set and the test set into different ratios such as 9:1, 8:2, 7:3, 6:4, 5:5, and 4:6;
[0071] Step 1022, calculate the mean absolute error MAE and root mean square error RMSE of the test set under different division ratios, and select the set with the smallest MAE and RMSE of the test set as the final division ratio of the training set and the test set, and determine the final data set. The MAE and RMSE are calculated as shown in calculation formulas (4) and (5) respectively:
[0072]
[0073]
[0074] In the above calculation formula, m is the total number of samples, y is i and pre are the true value and the predicted value respectively.
[0075] The optimal division ratio of the training set and the test set determined in the embodiment of the present invention is 8:2. Under this ratio, Figure 2 As shown in the figure, the determination coefficient R of the constructed support vector machine regression radial basis kernel function machine learning model 2It reached 0.96, and the MAE was only 16.50ppm, which proved that the machine learning model established according to the technical solution of the present invention has excellent predictive and generalization capabilities. Figure 2 In the figure, the units of the test value on the horizontal axis and the predicted value on the vertical axis are both ppm, the test value is obtained from published literature, and the predicted value is predicted by the method of the present invention.
[0076] Further, in S103, the using the final data set to train the machine learning model to obtain the trained machine learning model specifically includes:
[0077] During the training process of the machine learning model,
[0078] Step 1031, according to the final data set determined in step 1022 (the ratio of the training set to the test set is 8:2), the training set is used for training the machine learning model, and the test set is used for verifying the prediction accuracy of the machine learning model;
[0079] Step 1032, using the coefficient of determination R 2 Evaluate the predictive performance of the machine learning model; the coefficient of determination R 2 The calculation formula is as follows:
[0080]
[0081] Among them, y i and pre represent the true value and the predicted value respectively. Represents y i The average value of .
[0082] Step 1033, adjusting the model hyperparameters to optimize the machine learning model, and using a ten-fold cross validation method and a grid search method to determine the optimal model hyperparameters;
[0083] Step 1034, combining the optimal hyperparameters of the model and the determination coefficient R 2 The predictive performance of the machine learning models is jointly evaluated.
[0084] Furthermore, the method further comprises:
[0085] S105, preparing a target RFe2 type material corresponding to the predicted giant magnetostrictive performance of the RFe2 type material, and testing the magnetostrictive performance of the target RFe2 type material, specifically comprising:
[0086] Step 1051, based on a series of target RFe2 type materials predicted by the machine learning model in step S104, select components with excellent magnetostrictive properties for experimental preparation to obtain RFe2 type prepared materials;
[0087] The magnetostriction values of the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field are tested respectively, and the difference in magnetostriction between the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field is calculated.
[0088] Figure 3 and Figure 4 The actual performance of the predicted material (predicted embodiment) is tested and verified from aspects such as room temperature magnetostriction value, hysteresis and working temperature range; wherein Figure 3 This is a magnetostrictive performance diagram of the predicted embodiment of the present invention. The predicted embodiment selects Tb 0.34 Dy 0.66 Fe 1.95 V 0.05 The magnetostriction value (λ ∥ ) and the magnetostriction value (λ ⊥ ), and the difference between the two can be calculated to obtain the magnetostriction value λ of the embodiment. a ,Right now:
[0089] λ a =λ || -λ ⊥ ; (7)
[0090] Figure 3 The gray illustration in the lower right corner of FIG. 1 is a local enlarged view of the magnetostriction curve under a specific magnetic field, which is used to calculate the maximum hysteresis of the predicted embodiment. The calculation formula of the maximum hysteresis is as follows:
[0091] (Δλ a ) max / (λ a,max ); (8)
[0092] Among them, (Δλ a ) max Indicates the maximum difference of the magnetostriction curve under a specific magnetic field during the process of raising and lowering the magnetic field, λ a,max is the maximum value of magnetostriction of the embodiment. Figure 3 It can be seen that the predicted embodiment (alloy) obtained by the support vector machine regression radial basis kernel function model constructed based on the technical solution of the present invention has the following test performance: under a 2T magnetic field, the room temperature magnetostriction can reach 1310.39ppm. In addition, the magnetostriction value of the embodiment of the present invention shows an almost completely recoverable feature during the process of raising and lowering the magnetic field (recoverable specifically refers to Figure 3The two curves in the process of increasing and decreasing the magnetic field are basically completely overlapped, which also shows that the embodiment has almost no hysteresis), especially when the magnetic field is removed, there is only 1.7% hysteresis, which shows that the RFe2 type magnetostrictive material predicted by the method of the present invention is expected to become a key component of micro-intelligent devices such as high-precision sensors and micro-displacement drivers. Figure 3 The horizontal axis is the magnetic field (H), the unit is Tesla T, and the vertical axis is the magnetostriction value λ of the embodiment a ,λ a The unit is ppm.
[0093] Figure 4 The magnetostriction value (λ) of the embodiment and the comparative example predicted by the present invention when the magnetic field is parallel to the direction of the strain gauge || ) versus temperature, the comparative example is a commercial Terfenol-D alloy. Compared with the comparative example, the predicted embodiment has a more excellent magnetostrictive property. Figure 4 The illustration in the lower right corner is a comparison of the working temperature ranges of the embodiment and the comparative example under a 2T magnetic field, where ΔT1 and ΔT2 represent the working temperature range widths of the predicted embodiment and the comparative example, respectively. Compared with the comparative example, the working temperature range of the predicted embodiment of the present invention is increased by 18.7%, and 18.7% is the ratio of the working temperature range widened by the predicted embodiment compared with the comparative example, i.e., ΔT1-ΔT2; in addition, 34.5% and 18.1% in the illustration are the values for measuring the temperature stability of the comparative example and the predicted embodiment, respectively. The smaller the value, the smaller the temperature fluctuation, and correspondingly, it also indicates a larger temperature range. Compared with the comparative example, the temperature stability of the predicted embodiment increased by 16.4%, showing good temperature stability, which indicates that the materials predicted based on the machine learning model of the present invention have great potential in practical applications. Figure 4 The horizontal axis is the temperature (T), in Kelvin K, and the vertical axis is the magnetostriction value (λ || ).
[0094] Obviously, Figure 4 It is more intuitively explained that the predicted embodiment has a wider working temperature range than the comparative example, and it is also proved that the RFe2 type magnetostrictive material predicted by the method of the present invention has more excellent magnetostrictive performance, and the present invention shows good temperature stability compared with the comparative example, which indicates that the RFe2 type magnetostrictive material predicted by the machine learning model designed based on the technical solution of the present invention has great application potential in intelligent devices such as high-precision sensors, micro-displacement actuators and ultra-precision transducers.
[0095] Another aspect of the present invention provides a machine learning device for predicting the giant magnetostrictive properties of RFe2 type materials, such as Figure 5 As shown, including:
[0096] A construction module 201 is used to construct a composition-property-magnetostriction data set of RFe2 type materials;
[0097] A division module 202 is used to divide the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determine a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set;
[0098] A training module 203 is used to train a machine learning model using the final data set to obtain a trained machine learning model, wherein the training set is used to train the machine learning model, and the test set is used to verify the prediction accuracy of the machine learning model;
[0099] The prediction module 204 is used to predict the giant magnetostrictive performance of the RFe2 type material through the trained machine learning model.
[0100] The present invention provides a machine learning method and device for predicting the giant magnetostrictive performance of RFe2 type materials. The method establishes a composition-performance-magnetostrictive data set of an RFe2 type material system, and then divides the data set into training sets and test sets of different proportions. The final data set is determined by calculating the mean absolute error (MAE) and the root mean square error (RMSE) of the test set under different division ratios, and the final data set is used to train a machine learning model. The optimal division ratio of the training set and the test set determined in the present invention is 8:2. Under this ratio, the determination coefficient R of the constructed support vector machine regression radial basis kernel function machine learning model is 8:2. 2 It reached 0.96, and the MAE was only 16.50ppm, showing excellent predictive and generalization capabilities. Among the predicted materials, the embodiment selected by the present invention can reach a room temperature magnetostriction value of 1310.39ppm under an applied 2T magnetic field. In particular, the magnetostriction value of the embodiment of the present invention shows an almost completely recoverable feature in the process of raising and lowering the magnetic field, and when the magnetic field is removed, there is only 1.7% hysteresis. In addition, compared with the comparative example, the embodiment of the present invention has a significantly broadened operating temperature range, which further shows that the RFe2-type magnetostrictive material predicted based on the technical solution of the present invention is expected to assist in the creation and development of smart devices.
[0101] It should be understood that the above specific embodiments of the present invention are only used to illustrate or explain the principles of the present invention, and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.
Claims
1. A machine learning method for predicting the giant magnetostrictive properties of RFe2-type materials, characterized in that: The following steps are involved: Construct a composition-properties-magnetostriction dataset for RFe2-type materials; Dividing the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determining a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set; Using the final data set to train a machine learning model to obtain a trained machine learning model, wherein the training set is used to train the machine learning model, and the test set is used to verify the prediction accuracy of the machine learning model; The giant magnetostrictive properties of the RFe2 type material are predicted by the trained machine learning model.
2. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 1, characterized in that: The method of constructing a composition-performance-magnetostriction data set of RFe2 type materials comprises the following steps: Obtaining the content of each element in a known RFe2 type material compound and the corresponding magnetostriction value of the RFe2 type material compound; Obtain the physical parameter data of each element in the RFe2 type material compound according to the preset element physical parameter performance table; Based on the stoichiometric ratio of the RFe2 type material compound and the physical parameter data of each element, the average value of the physical parameters corresponding to the RFe2 type material compound is obtained by weighted calculation; Calculating the physical parameter performance data of the corresponding RFe2 type material compound respectively through the average values of the corresponding physical parameters of the RFe2 type material compound; Taking the composition content of each element in the RFe2 type material compound and the calculated physical parameter performance data of the corresponding RFe2 type material compound as the characteristic values, and the corresponding magnetostriction value of the RFe2 type material compound as the target value, a composition-performance-magnetostriction data set of RFe2 type materials is constructed.
3. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 2, characterized in that: Based on the stoichiometric ratio of the RFe2 type material compound and the physical parameter data of each element, the average value of the physical parameters corresponding to the RFe2 type material compound is obtained by weighted calculation, using the following calculation formula: in, represents the average value of the physical parameters of the RFe2 type material compound obtained by weighted average calculation; n is the total number of atoms in the RFe2 type material compound; C i is the stoichiometric ratio of the i-th atom in the RFe2-type material compound; f i It is the physical property data of the i-th atom in the RFe2 type material compound.
4. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 2, characterized in that: The physical parameter performance data of the corresponding RFe2 type material compound are calculated respectively by the average values of the corresponding physical parameters of the RFe2 type material compound, and the calculation process is as follows: or Among them, f i,1 Represents the physical parameter performance values of RFe2 type material compounds.
5. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 1, characterized in that: Dividing the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determining a final data set according to calculation results of the training sets and the test sets of different proportions, including: The training set and the test set are divided into two parts according to preset ratios: The mean absolute error (MAE) and root mean square error (RMSE) of the test set under different division ratios are calculated, and the set with the smallest MAE and RMSE of the test set is selected as the final division ratio of the training set and the test set to determine the final data set.
6. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 1, characterized in that: The method of training the machine learning model using the final data set to obtain the trained machine learning model includes: During the training process of the machine learning model, The coefficient of determination R 2 evaluating the predictive performance of the machine learning model; The model hyperparameters are adjusted to optimize the machine learning model, and the optimal hyperparameters of the model are determined by using a ten-fold cross validation method and a grid search method; Combine the optimal hyperparameters of the model and the determination coefficient R 2 The predictive performance of the machine learning models is jointly evaluated.
7. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 6, characterized in that: The coefficient of determination R 2 The calculation formula is as follows: Among them, y i and pre represent the true value and the predicted value respectively. Represents y i The average value of .
8. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 1, characterized in that: The method further comprises: A target RFe2 type material corresponding to the predicted giant magnetostrictive performance of the RFe2 type material is prepared, and the magnetostrictive performance of the target RFe2 type material is tested.
9. The machine learning method for predicting giant magnetostrictive properties of RFe2 type materials according to claim 8, characterized in that: The step of preparing a target RFe2 type material corresponding to the predicted giant magnetostrictive performance of the RFe2 type material and testing the magnetostrictive performance of the target RFe2 type material comprises the following steps: Selecting components with excellent magnetostrictive properties from the target RFe2 type material for experimental preparation to obtain RFe2 type prepared material; The magnetostriction values of the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field are tested respectively, and the difference in magnetostriction between the RFe2 type prepared material in the directions parallel to the magnetic field and perpendicular to the magnetic field is calculated.
10. A machine learning device for predicting the giant magnetostrictive properties of RFe2-type materials, characterized in that: include: Building module for constructing composition-properties-magnetostriction datasets for RFe2-type materials; A partitioning module, used to divide the composition-performance-magnetostriction data set into training sets and test sets of different proportions, and determine a final data set according to calculation results of the training sets and the test sets at different proportions, wherein the data set includes the training set and the test set; A training module, used to train a machine learning model using the final data set to obtain a trained machine learning model, wherein the training set is used to train the machine learning model, and the test set is used to verify the prediction accuracy of the machine learning model; A prediction module is used to predict the giant magnetostrictive performance of the RFe2 type material through the trained machine learning model.
Citation Information
Patent Citations
Rare earth giant magnetostrictive material screening method based on machine learning
CN115527638A
Titanium alloy mechanical property prediction method and device based on machine learning
CN116230136A
Prediction method of iron-based nanocrystalline alloy based on machine learning
CN116364205A
Sintered neodymium-iron-boron material performance prediction model and training method and application thereof
CN116844672A
Novel wide temperature domain giant magnetostriction material and its preparing method
CN1492522A
