A method and system for predicting and designing mechanical properties of Mg-TM-RE alloy
By constructing multiple machine learning regression models and SHAP analysis, the problem of screening high-strength and high-ductility alloys in the Mg-TM-RE alloy system was solved, enabling efficient and accurate alloy design and processing, and reducing R&D costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LANZHOU UNIV
- Filing Date
- 2026-04-17
- Publication Date
- 2026-07-10
AI Technical Summary
Existing technologies make it difficult to simultaneously screen for candidate alloys with high strength and high ductility in the Mg-TM-RE alloy system. Traditional methods have long development cycles and high costs, and single machine learning models are difficult to achieve optimal performance across multiple mechanical properties.
Multiple machine learning regression models of different types were constructed to predict yield strength, tensile strength and elongation respectively. The SHAP method was used to analyze the feature contribution and screen out alloy composition and processing parameters that meet the preset conditions.
This enables the high-precision screening of Mg-TM-RE alloys that simultaneously meet the optimal balance of various mechanical properties within a large-scale composition-process space, shortening the research and development cycle and reducing experimental costs.
Smart Images
Figure CN122369742A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of materials design and calculation technology, and in particular to a method, system, equipment and medium for predicting and designing the mechanical properties of Mg-TM-RE alloys. Background Technology
[0002] Magnesium alloys have broad application prospects in the field of lightweight structural materials due to their low density, high specific strength, and good damping properties. With the deepening research on the role mechanism of rare earth elements in magnesium alloys, rare earth magnesium alloys have shown significant advantages in terms of strength, plasticity, and microstructure stability, and have gradually become an important research direction for high-performance magnesium alloy systems. At the same time, by introducing transition metal elements such as Zn, Zr, Al, and Mn, a multi-element Mg-TM-RE alloy system can be further constructed, thereby improving the comprehensive performance of the alloy.
[0003] However, the Mg-TM-RE alloy system is characterized by complex composition, narrow process window, and diverse microstructure evolution paths. Its mechanical properties are highly dependent on the synergistic regulation between alloy composition, heat treatment regime, and plastic processing technology. Traditional alloy development methods mainly rely on empirical design and repeated trial and error experiments, which result in long research and development cycles, high experimental costs, and difficulty in quickly screening candidate alloys with both high strength and high ductility in a large-scale composition-process space.
[0004] In recent years, machine learning methods have been gradually introduced into the field of materials science. In existing research, it is generally assumed that a "universal" machine learning model (such as random forest or support vector machine) can accurately predict mechanical properties such as strength and plasticity at the same time, and output key composition parameters and key processing parameters under the predicted mechanical properties to screen alloys and design and process them. However, in the Mg-TM-RE alloy system, the dominant factors affecting each mechanical property are very different. For example, the dominant factor affecting yield strength is the volume fraction and distribution of the second phase, while the dominant factor affecting elongation is the grain size and texture of the matrix. This essential difference in physical mechanism makes it difficult for any single model to achieve the optimal performance on multiple prediction tasks at the same time. Ultimately, it is difficult for a single model to achieve the optimal performance on all mechanical properties at the same time. Therefore, it is difficult to simultaneously screen the composition parameters and processing parameters required for the optimal balance of each mechanical property in the candidate Mg-TM-RE alloy system space, so as to design and process Mg-TM-RE alloys with high precision. Summary of the Invention
[0005] This invention provides a method and system for predicting and designing the mechanical properties of Mg-TM-RE alloys, which can solve the problems existing in the prior art.
[0006] This invention provides a method for predicting and designing the mechanical properties of Mg-TM-RE alloys, comprising the following steps: Obtain the alloy composition parameters, processing parameters, and mechanical property parameters of the Mg-TM-RE alloy from publicly available experimental data; Using alloy composition parameters and processing parameters as input features, and each mechanical performance index within the mechanical performance parameters as an independent output target feature; for each output target feature, multiple different types of machine learning regression models are constructed and trained in parallel, and based on preset evaluation indicators, the model with the best prediction accuracy for each mechanical performance index is selected. Under preset composition and processing constraints, a candidate Mg-TM-RE alloy design space is constructed. Based on the optimal model selected from various mechanical property indicators, predictions are made in this design space to select the composition parameters and processing parameters of candidate Mg-TM-RE alloys that simultaneously meet the preset thresholds of various mechanical property indicators, so as to design the candidate Mg-TM-RE alloys.
[0007] Preferably, the alloy composition parameters include one or more of Mg, Al, Mn, Cu, Zn, Zr, Y, Ce, Nd, Gd, Dy, and Er; The processing parameters include solution temperature, solution time, homogenization temperature, homogenization time, extrusion temperature, extrusion ratio, aging temperature and aging time. The mechanical properties include yield strength YS, tensile strength UTS, and elongation EL.
[0008] Preferably, the step of selecting the model with the best prediction accuracy for each mechanical performance index includes: The alloy composition parameters and processing parameters are used as shared input features, and the yield strength, tensile strength and elongation are used as three independent output target features respectively. For each output target, multiple different types of machine learning regression models are built and trained in parallel, and the performance of the machine learning regression models is compared based on preset evaluation indicators to determine the optimal machine learning regression model for yield strength, tensile strength and elongation. Among them, machine learning regression models include support vector regression (SVR), multilayer perceptron (MLP), random forest regression (RFR), gradient boosting regression (GBR), and extreme gradient boosting (XGBoost). Among them, the preset evaluation indicators include the coefficient of determination R. 2 Root mean square error (RMSE) and mean absolute error (MAE).
[0009] Preferably, after screening the model with the best prediction accuracy for each mechanical property index, the SHAP method is used to perform feature contribution analysis on the model with the best prediction accuracy for each mechanical property index, and the alloy composition parameters and processing parameters with the strongest influence on each mechanical property index are identified.
[0010] Preferably, the design of candidate Mg-TM-RE alloys includes: Preset compositional and processing constraints that conform to the actual industrial production conditions, and under these compositional and processing constraints, construct a Mg-TM-RE alloy candidate design space containing a large number of candidate alloy compositional parameters and their corresponding processing parameters through combination enumeration. Using the model with the best prediction accuracy for each mechanical property index, the yield strength, tensile strength and elongation of each combination of composition parameters and processing parameters in the candidate design space of Mg-TM-RE alloys are predicted in batches. Then, based on a preset threshold that simultaneously meets each mechanical property index, the composition parameters and processing parameters of all candidate Mg-TM-RE alloys that meet the threshold of each mechanical property index are obtained, so as to design the candidate Mg-TM-RE alloys.
[0011] Preferably, the compositional constraints include: Mg 85-95 wt.%; Zn 0.5-5.5 wt.%; Zr 0.3-1 wt.%; Y 0-10 wt.%; Nd 1-5 wt.%; Gd 1-10 wt.%; Er 0-5 wt.%; and rare earth elements selected from one or two, with the total mass percentage of each element being 100 wt.%. The processing constraints include: solution temperature selected from 0, 350, 400, 450, 500, 550, or 600℃; solution time selected from 0, 6, 12, 24, or 48 h; homogenization temperature selected from 0, 350, 400, 450, 500, or 550℃; homogenization time selected from 0, 6, 12, 24, or 48 h; extrusion temperature selected from 0, 350, 400, or 450℃; extrusion ratio selected from 0, 15, 20, 25, or 30; aging temperature selected from 0, 180, 200, or 250℃; aging time selected from 0, 12, 24, 36, 48, or 96 h; and the corresponding parameter is 0 when no corresponding processing step is performed.
[0012] Preferably, after obtaining the alloy composition parameters, processing parameters, and mechanical property parameters, the composition parameters, processing parameters, and mechanical property parameters are preprocessed. During preprocessing, the process parameters corresponding to specific process steps that have not been implemented are uniformly assigned a value of 0, and abnormal samples that are missing key input features and cannot be completed by rules are screened out in order to maintain the consistency of feature dimensions of each parameter sample.
[0013] This invention also provides a system for predicting and designing the mechanical properties of Mg-TM-RE alloys, comprising: The data construction module is used to obtain the alloy composition parameters, processing parameters, and mechanical property parameters of Mg-TM-RE alloys from publicly available experimental data; The model evaluation module is used to take alloy composition parameters and processing parameters as input features and each mechanical performance index within the mechanical performance parameters as an independent output target feature. For each output target feature, multiple different types of machine learning regression models are constructed and trained in parallel, and the model with the best prediction accuracy for each mechanical performance index is selected based on the preset evaluation index. The high-throughput screening module is used to construct a candidate Mg-TM-RE alloy design space under preset composition and processing constraints. In this design space, the optimal model selected based on various mechanical property indicators is used to predict and screen out the composition parameters and processing parameters of candidate Mg-TM-RE alloys that simultaneously meet the preset thresholds of various mechanical property indicators, so as to design the candidate Mg-TM-RE alloys.
[0014] This invention also provides an electronic device, including a memory and a processor; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the Mg-TM-RE alloy mechanical property prediction and design method as described above.
[0015] This invention also provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the steps of the Mg-TM-RE alloy mechanical property prediction and design method as described above.
[0016] This invention provides a method and system for predicting and designing the mechanical properties of Mg-TM-RE alloys. Compared with the prior art, its advantages are as follows: This invention constructs multiple regression models of different types and independently selects the optimal model for each mechanical performance index. This allows the prediction task for each mechanical performance index to select the model with the strongest prediction ability for that index. For example, the prediction task for yield strength and tensile strength can select the model with the strongest fitting ability to nonlinear relationships, while the prediction task for elongation can select the model with stronger robustness to overfitting. This "targeted" strategy ensures that the prediction accuracy of each mechanical performance index reaches its optimal level, avoiding the predicament of a single model having poor prediction ability for a certain mechanical performance index. Thus, in the preset candidate Mg-TM-RE alloy design space, by simultaneously using the model with the strongest prediction ability for each mechanical performance index, the composition parameters and processing parameters that simultaneously meet the optimal balance of each mechanical performance index can be selected, and finally, the Mg-TM-RE alloy can be designed and processed with high precision. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the overall process provided for an embodiment of the present invention; Figure 2 This is a schematic diagram of dataset pre-analysis provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the evaluation performance of different machine learning models for yield strength prediction provided in embodiments of the present invention. Figure 4 This is a schematic diagram illustrating the evaluation performance of different machine learning models for tensile strength prediction provided in embodiments of the present invention. Figure 5 This is a schematic diagram illustrating the evaluation of the elongation prediction performance of different machine learning models provided in the embodiments of the present invention; Figure 6 A schematic diagram illustrating the SHAP feature importance analysis of the optimal model provided in this embodiment of the invention; Figure 7 This is a schematic diagram comparing experimental and predicted results provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of alloy reverse design based on an optimal model, provided for an embodiment of the present invention. Detailed Implementation
[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0019] In recent years, machine learning methods have been gradually introduced into the field of materials science. Compared with traditional physical models, machine learning can extract nonlinear relationships between composition, processing, and properties from large amounts of experimental data, and is particularly suitable for complex alloy systems with multivariate coupling. Existing research has shown that machine learning has high potential in predicting the properties of rare earth magnesium alloys, but existing solutions generally have the following shortcomings: first, there is insufficient research on Mg-TM-RE alloy systems with multiple rare earth elements and multiple processing parameters; second, the ability to predict and interpret plasticity-related indicators such as elongation is limited; and third, there is a lack of an integrated solution that combines high-precision performance prediction with high-throughput reverse design under practical constraints.
[0020] Therefore, it is necessary to provide a high-precision mechanical property prediction and rapid design method for Mg-TM-RE alloy systems to improve the development efficiency of high-performance rare-earth magnesium alloys and reduce experimental screening costs. Based on this, this invention provides a machine learning-based method for predicting and rapidly designing the mechanical properties of Mg-TM-RE alloys, such as... Figure 1 As shown, it includes the following steps: Step S1: Collect publicly available experimental data on Mg-TM-RE alloys and construct a dataset containing alloy composition parameters, processing parameters, and mechanical property parameters.
[0021] The dataset is derived from 104 publicly published papers from 2003 to 2025, containing a total of 571 experimental samples.
[0022] The alloy composition parameters include one or more of Mg, Al, Mn, Cu, Zn, Zr, Y, Ce, Nd, Gd, Dy, and Er; the processing parameters include solution temperature, solution time, homogenization temperature, homogenization time, extrusion temperature, extrusion ratio, aging temperature, and aging time; the mechanical property parameters include yield strength YS, tensile strength UTS, and elongation EL.
[0023] Step S2: Preprocess the dataset to obtain standardized input and output data for model training, such as... Figure 2 As shown; where, to maintain consistency in sample feature dimensions, the parameters corresponding to the steps in which no process steps were implemented are assigned a value of 0; abnormal samples lacking key input features can be screened out.
[0024] Step S3: Using alloy composition parameters and processing parameters as input features, and yield strength, tensile strength and elongation as output features, construct multiple machine learning regression models and train them.
[0025] Machine learning regression models include Support Vector Regression (SVR), Multilayer Perceptron (MLP), Random Forest Regression (RFR), Gradient Boosting Regression (GBR), and Extreme Gradient Boosting (XGBoost).
[0026] Step S4: Compare the performance of the machine learning regression model based on preset evaluation indicators to determine the optimal prediction model for different mechanical performance indicators, such as... Figure 3 The graph shows the evaluation performance of different machine learning models in predicting yield strength. Figure 4 Evaluation graphs of the tensile strength prediction performance of different machine learning models, such as... Figure 5 This is a graph evaluating the performance of different machine learning models in predicting elongation.
[0027] The model was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE).
[0028] Step S5: Perform feature contribution analysis on the optimal prediction model to identify key component parameters and key processing parameters that affect the mechanical performance indicators.
[0029] The SHAP method was used for feature contribution analysis to identify the direction and magnitude of the influence of different input features on the prediction results of yield strength, tensile strength, and elongation. Figure 6 This is a plot showing the importance of SHAP features for the optimal model.
[0030] Analysis shows that the extrusion ratio and extrusion temperature are key process factors affecting strength-related properties, while Mg content and Zr content are important component factors affecting elongation.
[0031] Step S6: Construct a design space for candidate Mg-TM-RE alloys under preset composition and process constraints, and perform batch prediction and screening of candidate alloys based on the optimal prediction model.
[0032] The compositional constraints include: Mg 85-95 wt.%; Zn 0.5-5.5 wt.%; Zr 0.3-1 wt.%; Y 0-10 wt.%; Nd 1-5 wt.%; Gd 1-10 wt.%; Er 0-5 wt.%; rare earth elements are selected from one or two, and the total mass percentage of each element is 100 wt.%.
[0033] The process constraints include: solution temperature selected from 0, 350, 400, 450, 500, 550, or 600℃; solution time selected from 0, 6, 12, 24, or 48 h; homogenization temperature selected from 0, 350, 400, 450, 500, or 550℃; homogenization time selected from 0, 6, 12, 24, or 48 h; extrusion temperature selected from 0, 350, 400, or 450℃; extrusion ratio selected from 0, 15, 20, 25, or 30; aging temperature selected from 0, 180, 200, or 250℃; aging time selected from 0, 12, 24, 36, 48, or 96 h; and the corresponding parameter is 0 when the corresponding process step is not performed.
[0034] Under constrained conditions, 1,841 candidate alloy composition combinations and 91,728 process parameter combinations were constructed, resulting in a total of 168,871,248 Mg-TM-RE alloys in the entire space.
[0035] Step S7: Output the composition of the candidate Mg-TM-RE alloy and the corresponding processing parameters that meet the target mechanical property threshold.
[0036] The target mechanical property thresholds are a yield strength of not less than 300 MPa, a tensile strength of not less than 400 MPa, and an elongation of more than 15%.
[0037] The candidate Mg-TM-RE alloy composition of the output is: (86-90)Mg-(1-3.5)Zn-0.5Zr-(0-5)Y-(0-1)Nd-(8-9)Gd.
[0038] The specific implementation steps include: Step 1: Database construction.
[0039] Experimental data on Mg-TM-RE alloys published between 2003 and 2025 were collected, yielding 104 articles and 571 sets of experimental samples. Twelve categories of compositional variables and eight categories of process variables were extracted as input features. Compositional variables included Mg, Al, Mn, Cu, Zn, Zr, Y, Ce, Nd, Gd, Dy, and Er. Process variables included solution temperature, solution time, homogenization temperature, homogenization time, extrusion temperature, extrusion ratio, aging temperature, and aging time. Output variables were yield strength (YS), tensile strength (UTS), and elongation (EL).
[0040] Step 2: Data preprocessing.
[0041] To maintain the consistency of features across the entire data sample, parameters corresponding to process steps that were not implemented were assigned a value of 0; for example, when no solution treatment was performed, the solution temperature was 0 and the solution time was 0; when no extrusion process was performed, the extrusion temperature was 0 and the extrusion ratio was 0; abnormal samples that lacked key input features and could not be expressed by a unified rule were removed; after preprocessing, statistical distribution analysis and correlation analysis were performed on the dataset to confirm that each input feature was independent in the experimental design and that there was no obvious coupling phenomenon requiring forced dimensionality reduction.
[0042] Step 3: Model training and evaluation.
[0043] The preprocessed dataset was divided into training and testing sets, with 90% of the original data used for model training and 10% for model testing. 10-fold cross-validation was performed during training. Support vector regression, multilayer perceptron, random forest regression, gradient boosting regression, and extreme gradient boosting models were constructed. R², RMSE, and MAE were used to evaluate the performance of each model. The results showed that the XGBoost model performed best overall in the YS and UTS prediction tasks, while the GBR model performed best overall in the EL prediction task.
[0044] Step 4: Feature Contribution Analysis Based on the optimal model obtained in step three, the Tree SHAP method is used to analyze the interpretability of the prediction results; such as Figure 7 As shown in the analysis results, for YS and UTS, the extrusion ratio and extrusion temperature are the most important process factors; for EL, Mg and Zr are the more significant component factors; when the extrusion ratio is increased, it has a significant positive contribution to the strength prediction results; when Mg is in a specific effective range and is combined with an appropriate Zr content, it is beneficial to improve the elongation.
[0045] Step 5: High-throughput design screening.
[0046] Based on the analysis of experimental data distribution and characteristic contribution, the following compositional constraints were established: Mg 85-95 wt.%; Zn 0.5-5.5 wt.%; Zr 0.3-1 wt.%; Y 0-10 wt.%; Nd 1-5 wt.%; Gd 1-10 wt.%; Er 0-5 wt.%; one or two rare earth elements can be selected, and the sum of the contents of all elements is 100 wt.%; The following process constraints were also set: solution temperature selected from 0, 350, 400, 450, 500, 550, or 600℃; solution time selected from 0, 6, 12, 24, or 48 h; homogenization temperature selected from 0, 350, 400, 450, 500, or 550℃; homogenization time selected from 0, 6, 12, 24, or 48 h. h; extrusion temperature is selected from 0, 350, 400 or 450℃, extrusion ratio is selected from 0, 15, 20, 25 or 30; aging temperature is selected from 0, 180, 200 or 250℃, aging time is selected from 0, 12, 24, 36, 48 or 96 h; and the corresponding parameter is 0 when the corresponding process step is not performed.
[0047] Under the aforementioned boundary constraints, 1841 candidate alloy compositions and 91728 process parameter combinations were constructed, resulting in 168,871,248 full-space combinations of Mg-TM-RE alloys; for example... Figure 8 As shown, the above combinations are input into the optimal prediction model for batch screening. The screening criteria are set as YS≥300 MPa, UTS≥400 MPa and EL>15%. Finally, a high-performance rare earth magnesium alloy window that meets the conditions is obtained, among which the typical candidate composition is: (86-90)Mg-(1-3.5)Zn-0.5Zr-(0-5)Y-(0-1)Nd-(8-9)Gd.
[0048] Step Six: Compare predicted values with experimental values.
[0049] Implementation 1: Magnesium alloys containing rare earth elements Gd, Dy, and transition metal elements Zr were selected as the research object. The input parameters of the prediction module are shown in Table 1.
[0050] Table 1. Input parameters for the prediction module Model output: Yield strength 128 MPa, tensile strength 211 MPa.
[0051] Experimental results: Yield strength 131 MPa, tensile strength 210 MPa.
[0052] Compared with actual experiments, the prediction error is less than 3%.
[0053] Implementation 2: Magnesium alloys containing rare earth elements Gd, Dy, and transition metal elements Zr were selected as the research object. The input parameters of the prediction module are shown in Table 2.
[0054] Table 2 Input parameters for the prediction module Model output: Yield strength 137 MPa, tensile strength 225 MPa.
[0055] Experimental results: Yield strength 135 MPa, tensile strength 226 MPa.
[0056] Compared with actual experiments, the prediction error is less than 2%.
[0057] Implementation Three: Magnesium alloys containing rare earth elements Y, Nd, and Gd, and transition metal element Zr were selected as the research object. The input parameters of the prediction module are shown in Table 3.
[0058] Table 3 Input parameters for the prediction module Model output: Yield strength 122 MPa, tensile strength 187 MPa; Experimental results: Yield strength 130 MPa, tensile strength 186 MPa; Compared with actual experiments, the prediction error is less than 7%.
[0059] Implementation Four: Magnesium alloys containing rare earth elements Y, Nd, and Gd, and transition metal elements Zn and Zr were selected as the research object. The input parameters of the prediction module are shown in Table 4.
[0060] Table 4 Input parameters for the prediction module Model output: Yield strength 197 MPa, tensile strength 303 MPa; Experimental results: Yield strength 198 MPa, tensile strength 276 MPa; Compared with actual experiments, the prediction error is less than 10%.
[0061] Implementation 5: Magnesium alloys containing rare earth element Y and transition metal element Zr were selected as the research object. The input parameters of the prediction module are shown in Table 5.
[0062] Table 5 Input parameters for the prediction module Model output: Yield strength 161 MPa, tensile strength 219 MPa; Experimental results: Yield strength 153 MPa, tensile strength 231 MPa; Compared with actual experiments, the prediction error is less than 10%.
[0063] Implementation Six: Magnesium alloys containing rare earth elements Y, Nd, and Gd, and transition metal elements Zn and Zr were selected as the research object. The input parameters of the prediction module are shown in Table 6.
[0064] Table 6 Input parameters for the prediction module Model output: elongation rates of 7.9%, 6.6%, and 26.2%; Experimental results: elongation rates of 9.7%, 7.6%, and 29.1%; Compared with actual experiments, the prediction error is less than 20%.
[0065] This invention establishes a unified database covering composition, process, and mechanical properties by integrating experimental data from publicly available literature, enabling systematic modeling of the composition-process-performance relationship of Mg-TM-RE alloys. By comparing various machine learning models, this invention determines the optimal prediction model applicable to different mechanical property indices, improving the accuracy and generalization ability of performance prediction. Combined with SHAP interpretability analysis, this invention can identify key compositional and process factors, providing interpretable evidence for alloy optimization. Furthermore, this invention enables high-throughput screening within a large-scale composition-process design space, rapidly outputting candidate Mg-TM-RE alloys that meet the requirements of high strength and high plasticity synergy, thereby significantly shortening the R&D cycle and reducing experimental screening costs.
[0066] This invention further identifies the "effective range" of key parameters based on SHAP analysis. This design rule, extracted from the black box and with clear physical meaning, enables materials scientists to understand the decision-making logic of the model, thereby more confidently adopting the candidate alloys recommended by the model, and conducting further mechanism exploration or process optimization on this basis. This invention adopts a full combinatorial enumeration (within the constraint range) or intelligent sampling strategy, which can cover the entire preset design space and ensure that the selected candidate alloys are the global optimal solution or Pareto front solution under the constraint conditions.
[0067] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.
Claims
1. A method for predicting and designing the mechanical properties of Mg-TM-RE alloys, characterized in that, Includes the following steps: Obtain the alloy composition parameters, processing parameters, and mechanical property parameters of the Mg-TM-RE alloy from publicly available experimental data; Using alloy composition parameters and processing parameters as input features, and each mechanical property index within the mechanical property parameters as an independent output target feature; For each output target feature, multiple different types of machine learning regression models are constructed and trained in parallel, and the model with the best prediction accuracy for each mechanical performance index is selected based on the preset evaluation index. Under preset composition and processing constraints, a candidate Mg-TM-RE alloy design space is constructed. Based on the optimal model selected from various mechanical property indicators, predictions are made in this design space to select the composition parameters and processing parameters of candidate Mg-TM-RE alloys that simultaneously meet the preset thresholds of various mechanical property indicators, so as to design the candidate Mg-TM-RE alloys.
2. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 1, characterized in that, The alloy composition parameters include one or more of Mg, Al, Mn, Cu, Zn, Zr, Y, Ce, Nd, Gd, Dy, and Er; The processing parameters include solution temperature, solution time, homogenization temperature, homogenization time, extrusion temperature, extrusion ratio, aging temperature and aging time. The mechanical properties include yield strength YS, tensile strength UTS, and elongation EL.
3. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 2, characterized in that, The methods for selecting the models with the best prediction accuracy for each mechanical performance index include: The alloy composition parameters and processing parameters are used as shared input features, and the yield strength, tensile strength and elongation are used as three independent output target features respectively. For each output target, multiple different types of machine learning regression models are built and trained in parallel, and the performance of the machine learning regression models is compared based on preset evaluation indicators to determine the optimal machine learning regression model for yield strength, tensile strength and elongation. Among them, machine learning regression models include support vector regression (SVR), multilayer perceptron (MLP), random forest regression (RFR), gradient boosting regression (GBR), and extreme gradient boosting (XGBoost). Among them, the preset evaluation indicators include the coefficient of determination R. 2 Root mean square error (RMSE) and mean absolute error (MAE).
4. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 3, characterized in that, After selecting the model with the best prediction accuracy for each mechanical property index, the SHAP method is used to perform feature contribution analysis on the model with the best prediction accuracy for each mechanical property index, and to identify the alloy composition parameters and processing parameters that have the strongest influence on each mechanical property index.
5. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 3, characterized in that, The design of candidate Mg-TM-RE alloys includes: Preset compositional and processing constraints that conform to the actual industrial production conditions, and under these compositional and processing constraints, construct a Mg-TM-RE alloy candidate design space containing a large number of candidate alloy compositional parameters and their corresponding processing parameters through combination enumeration. Using the model with the best prediction accuracy for each mechanical property index, the yield strength, tensile strength and elongation of each combination of composition parameters and processing parameters in the candidate design space of Mg-TM-RE alloys are predicted in batches. Then, based on a preset threshold that simultaneously meets each mechanical property index, the composition parameters and processing parameters of all candidate Mg-TM-RE alloys that meet the threshold of each mechanical property index are obtained, so as to design the candidate Mg-TM-RE alloys.
6. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 5, characterized in that, The compositional constraints include: Mg 85-95 wt.%; Zn 0.5-5.5 wt.%; Zr 0.3-1 wt.%; Y 0-10 wt.%; Nd 1-5 wt.%; Gd 1-10 wt.%; Er 0-5 wt.%; rare earth elements are selected from one or two, and the total mass percentage of each element is 100 wt.%. The processing constraints include: solution temperature selected from 0, 350, 400, 450, 500, 550, or 600℃; solution time selected from 0, 6, 12, 24, or 48 h; homogenization temperature selected from 0, 350, 400, 450, 500, or 550℃; homogenization time selected from 0, 6, 12, 24, or 48 h; extrusion temperature selected from 0, 350, 400, or 450℃; extrusion ratio selected from 0, 15, 20, 25, or 30; aging temperature selected from 0, 180, 200, or 250℃; aging time selected from 0, 12, 24, 36, 48, or 96 h; and the corresponding parameter is 0 when no corresponding processing step is performed.
7. The method for predicting and designing the mechanical properties of Mg-TM-RE alloys according to claim 1, characterized in that, After obtaining the alloy composition parameters, processing parameters, and mechanical property parameters, preprocessing is performed on the composition parameters, processing parameters, and mechanical property parameters. During preprocessing, the process parameters corresponding to specific process steps that have not been implemented are uniformly assigned a value of 0, and abnormal samples that are missing key input features and cannot be completed by rules are screened out in order to maintain the consistency of feature dimensions of each parameter sample.
8. A system for predicting and designing the mechanical properties of Mg-TM-RE alloys, characterized in that, include: The data construction module is used to obtain the alloy composition parameters, processing parameters, and mechanical property parameters of Mg-TM-RE alloys from publicly available experimental data; The model evaluation module is used to take alloy composition parameters and processing parameters as input features, and each mechanical property index within the mechanical property parameters as an independent output target feature. For each output target feature, multiple different types of machine learning regression models are constructed and trained in parallel, and the model with the best prediction accuracy for each mechanical performance index is selected based on the preset evaluation index. The high-throughput screening module is used to construct a candidate Mg-TM-RE alloy design space under preset composition and processing constraints. In this design space, the optimal model selected based on various mechanical property indicators is used to predict and screen out the composition parameters and processing parameters of candidate Mg-TM-RE alloys that simultaneously meet the preset thresholds of various mechanical property indicators, so as to design the candidate Mg-TM-RE alloys.
9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the method for predicting and designing the mechanical properties of Mg-TM-RE alloy as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a processor, implements the steps of a method for predicting and designing the mechanical properties of a Mg-TM-RE alloy as described in any one of claims 1 to 7.