Rare earth magnesium alloy mechanical property prediction and process optimization method based on machine learning

Abstract

The invention relates to the technical field of metal processing, and particularly discloses a rare earth magnesium alloy mechanical property prediction and process optimization method based on machine learning, which mainly comprises the steps of alloy component design, ultrasonic pulping treatment, rheological extrusion forming, database establishment, machine learning model construction, multi-objective optimization algorithm and the like. The core of the method is that a machine learning model and a multi-objective optimization algorithm are combined to rapidly predict and optimize the mechanical properties of the rare earth magnesium alloy, and the method can significantly reduce the experimental cost, is suitable for an ultrasonic pulping semi-solid rheoforming process, and can rapidly design a high-strength and high-toughness rare earth magnesium alloy product.

Description

Technical Field

[0001] The present invention relates to the technical field of metal processing, and particularly designs a method for designing the composition of rare earth magnesium alloy, optimizing the process and predicting the mechanical properties based on machine learning, which is particularly applicable to the ultrasonic pulping semi-solid rheoforming process. Background Art

[0002] Magnesium and its alloys, as an innovative structural material, have been widely used in the fields of aerospace, electronic communication, military equipment, etc. due to their low density, high specific strength, large elastic modulus and excellent heat dissipation performance, and their market demand is rising sharply. Among the application scopes of magnesium alloys, cast magnesium alloy products occupy the dominant position in the market. However, compared with cast aluminum alloys, traditional cast magnesium alloys are slightly insufficient in mechanical properties and heat resistance. In view of the unique valence electron structure of rare earth elements and their significant strengthening effect in magnesium, rare earth (RE) alloying has become one of the key means to improve the strength and toughness of magnesium alloys and plays an important role in the research and development of new magnesium-based materials.

[0003] However, in terms of processing and forming, rare earth magnesium alloys face a series of severe challenges, which increase the difficulty of their casting and plastic processing, and thus hinder the further improvement of their performance. The rheoforming technology, as an advanced near-net-shape forming process, skillfully combines the advantages of liquid forming and plastic forming, and can produce parts with complex structures and excellent mechanical properties.

[0004] In addition, in terms of processing and forming, there are also some serious problems with rare earth magnesium alloys, which cause difficulties in casting and plastic processing and limit the further improvement of their performance. RE elements are unevenly distributed in the magnesium matrix and tend to segregate at grain boundaries; it is easy to form coarse rare earth compound phases enriched near grain boundaries, reducing the strength and toughness of magnesium alloys, and when the rare earth content is high, it is distributed in a network shape, with a greater harmful effect. The ultrasonic pulping process uses mechanical vibration to apply an external field to the molten metal in the semi-solid temperature range to obtain a semi-solid slurry with a certain volume fraction of primary phase, thereby improving the above problems.

[0005] When the prior art studies the above content, it often needs to design multiple groups of rare earth element compositions and continuously adjust the ultrasonic pulping and squeeze casting processes to obtain magnesium alloy castings with the best mechanical properties. Therefore, a large number of experiments need to be carried out to explore each variable.

[0006] In view of this, inventing a design method for the composition and process of ultrasonic pulping semi-solid rheoforming rare earth magnesium alloys, which is more efficient and accurate for high-performance rare earth magnesium alloy castings, is a problem worthy of research.

[0007] It should be noted that the information disclosed in this background art section is only intended to deepen the understanding of the overall background art of the present invention, and should not be regarded as an admission or any form of implication that this information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0008] The object of the present invention is to provide a method for predicting the mechanical properties and optimizing the process of rare earth magnesium alloys based on machine learning. By constructing a composition-process-property database and combining multi-objective optimization algorithms, a preparation plan for high-strength and tough rare earth magnesium alloys can be quickly generated, reducing experimental costs and achieving multi-objective collaborative optimization.

[0009] To achieve the above object, the present invention adopts the following technical solutions:

[0010] The present invention provides a method for predicting the mechanical properties and optimizing the process of rare earth magnesium alloys based on machine learning, including the following steps:

[0011] 1. Data collection: Design magnesium alloys with different rare earth element ratios, record ultrasonic pulping and rheoforming parameters, and establish a composition-process-property database.

[0012] 2. Model construction: Use a machine learning model (such as a BP neural network) to establish a prediction model, with the input being composition and process parameters, and the output being mechanical properties such as tensile strength and elongation.

[0013] 3. Multi-objective optimization and extension: Combine multi-objective optimization algorithms to optimize the prediction model, obtain combinations of alloy compositions and process parameters that meet the target mechanical properties, and use a generative adversarial network (GAN) to expand the data to improve the generalization ability of the model.

[0014] The beneficial effects of the present invention are:

[0015] 1. Reduce the number of experiments and shorten the R & D cycle through the prediction model;

[0016] 2. Multi-objective optimization avoids the decline of other performances caused by the improvement of a single performance;

[0017] 3. Generate synthetic data to solve the small sample problem and improve the robustness of the model. Description of the Drawings

[0018] Figure 1 It is a flowchart of a method for predicting the mechanical properties and optimizing the process of rare earth magnesium alloys based on machine learning provided by an embodiment of the present invention, showing the complete process of data collection, modeling, and optimization;

[0019] Figure 2Schematic diagram of the ultrasonic pulping device used in the embodiments of the present invention, where: 11. Ultrasonic generation controller, 12. Lifting bracket, 13. Transducer, 14. Amplitude transformer, 15. Thermocouple, 16. Magnesium alloy melt, 17. Insulation furnace;

[0020] Figure 3 Schematic diagram of the rheological extrusion forming equipment used in the embodiments of the present invention, showing the pressure and speed control module, where: 21. Moving mold, 22. Semi-solid slurry, 23. Fixed mold, 24. Ejector rod, 25. Part;

[0021] Figure 4 Curve of the comparison between the tensile strength and the actual value in the ultrasonic pulping semi-solid rheological forming rare earth magnesium alloy database by the neural network model in the embodiments of the present invention;

[0022] Figure 5 Prediction result of the yield strength by the neural network model in the ultrasonic pulping semi-solid rheological forming rare earth magnesium alloy database in the embodiments of the present invention;

[0023] Figure 6 Prediction result of the elongation by the neural network model in the ultrasonic pulping semi-solid rheological forming rare earth magnesium alloy database in the embodiments of the present invention. Specific embodiments

[0024] The following further elaborates on the technical features and advantages of the present invention in conjunction with the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making the protection scope of the present invention more clearly defined.

[0025] The embodiments of the present invention provide a model that uses machine learning algorithms to establish the composition and process of rare earth magnesium alloys based on ultrasonic pulping semi-solid rheological forming, specifically including parameters such as rare earth element content, ultrasonic process, and forming process, to predict their mechanical properties. The inventive concept of this method is summarized as follows:

[0026] 1. Data collection: Design magnesium alloys with different rare earth element ratios, perform ultrasonic pulping treatment and rheological forming processing, record the parameters of the ultrasonic pulping and rheological forming processes, and test the corresponding mechanical properties to establish a composition-process-performance database.

[0027] 2. Model construction: Construct a machine learning model, where the input layer includes alloy composition and process parameters, and the output layer is the mechanical properties.

[0028] 3. Multi-objective optimization: Use at least two mechanical properties as optimization objectives, and adopt a multi-objective optimization algorithm to iteratively optimize the above machine learning model to generate an optimal parameter combination.

[0029] As an example, the above method will be described in detail below using a neural network model as the machine learning model and a genetic algorithm as the multi-objective optimization algorithm. Please refer to Figure 1 , which is the flowchart of the method, showing the complete process of data collection, modeling, and optimization. Figure 2 Fig. shows the structural schematic diagram of the ultrasonic pulping device used in the embodiment of the present invention. The device includes: 11. Ultrasonic generation controller: used to control the action range and power of ultrasonic waves; 12. Lifting bracket: used to control the lifting of the magnesium alloy melt in the holding furnace; 13. Transducer: converts electrical energy into ultrasonic waves; 14. Horn: used to control the frequency of ultrasonic waves; 15. Thermocouple: measures the melt temperature during ultrasonic treatment; 16. Magnesium alloy melt: the molten metal obtained after melting magnesium alloy ingots and master alloys; 17. Holding furnace: used to control the temperature of the melt during ultrasonic treatment. Figure 3 Fig. shows the structural schematic diagram of the pressure and speed regulation module in the rheological extrusion forming equipment used in the embodiment of the present invention, including: 21. Moving die: used to apply extrusion pressure; 12. Semi-solid slurry: the semi-solid structure obtained after ultrasonic pulping of the magnesium alloy melt; 23. Fixed die: used to support the semi-solid slurry and control forming; 24. Ejector rod: used for demolding and ejecting the casting; 25. Part: that is, the final product. The specific steps of the method are as follows:

[0030] Step S1: Design multiple rare earth magnesium alloy compositions and carry out melting. The rare earth elements include at least two of Zn, Y, Ce, La, Zr, and Ni;

[0031] Step S2: Carry out ultrasonic pulping treatment on the molten magnesium alloy and regulate the ultrasonic process parameters, including vibration frequency, power, time, and horn specifications;

[0032] Step S3: Process the semi-solid slurry into parts through the rheological extrusion forming process and regulate the forming parameters, including nominal force, ejection force, and slider speed;

[0033] Step S4: Measure the mechanical properties of the parts and establish a database containing alloy composition - process parameters (ultrasonic pulping process & rheological forming process) - mechanical properties;

[0034] Step S5: Preprocess the database, including feature screening and data normalization;

[0035] Step S6: Build a neural network model. The input layer includes composition and process parameters, and the output is the predicted value of mechanical properties. Improve the model accuracy through hyperparameter optimization;

[0036] Step S7: Based on the optimized model, use the genetic algorithm to optimize multiple mechanical properties simultaneously and obtain the best combination of composition and process parameters;

[0037] Step S8: Expand the database through a generative adversarial network, generate new data that conforms to the original distribution, and update the prediction model;

[0038] Step S9: Output the corresponding composition and process parameter ranges according to the target mechanical property requirements.

[0039] Preferably, in step S1, according to the strengthening effect of rare earth elements on magnesium alloys, a variety of rare earth magnesium alloys with different element contents are designed according to experience for melting.

[0040] Preferably, in step S1, the designed rare earth magnesium alloy is melted using a pit resistance furnace and protected using a mixed gas of SF6 + N2.

[0041] Preferably, in step S5, the tested mechanical properties include tensile strength, yield strength, and elongation.

[0042] Preferably, in step S6, the data preprocessing methods used include Pearson correlation coefficient, recursive feature elimination, etc.

[0043] Among them, the calculation method of the Pearson correlation coefficient is:

[0044]

[0045] The model used for recursive feature elimination is a multi-layer feedforward neural network model.

[0046] Preferably, in step S7, the established neural network models include multi-layer feedforward neural network models, cascade feedforward neural network models, function fitting neural network models, etc.

[0047] Preferably, in step S7, the hyperparameter optimization method is the optuna optimizer, and the optimized parameters include: database division random seed number, hidden layer depth, learning rate, activation function, etc.

[0048] Preferably, in step S8, the mechanical property optimization targets include tensile strength and elongation.

[0049] Furthermore, in step S8, the optimization process is as follows: Determine the initial population size according to the requirements, and generate an initial population based on the initial database. Use the number of times of tensile strength, yield strength, and elongation as the fitness of the population, and set their importance to 1, and optimize in the direction of maximizing the population fitness. Select the initial population according to a specific selection strategy, and perform crossover and mutation operations on the selected individuals to generate a new population. Reorder the new population, select the optimal individuals to form a new population for the next generation of evolution. Repeat the steps of selection, crossover, mutation, and new population until the preset genetic algebra is reached. The final solution can be used as the optimal solution.

[0050] Preferably, in step S9, the deep learning method is a generative adversarial network method.

[0051] Preferably, in step S10, the main basis of the data analysis method is the tensile strength, yield strength, and elongation range.

[0052] Preferably, in step S2, the vibration frequency of the ultrasonic pulping process is 20 - 50 kHz, the power is 100 - 1500 W, the vibration time is 10 - 300 seconds, and the diameter of the horn is 5 - 50 mm.

[0053] Preferably, in step S3, the nominal force of the rheological extrusion forming process is 50 - 500 MPa, the ejection force is 10 - 200 kN, and the slider speed is 0.1 - 5 mm / s.

[0054] The above method will be described below with reference to a specific embodiment.

[0055] Embodiment 1

[0056] 1. Design a magnesium alloy containing Zn (3 - 6 wt.%), Y (1 - 3 wt.%), and Zr (0.3 - 0.6 wt.%). After melting, perform ultrasonic pulping (frequency 40 kHz, power 1200 W, time 60 s);

[0057] 2. Perform rheological extrusion forming (nominal force 200 MPa, slider speed 2 mm / s);

[0058] 3. Test the tensile strength (210 MPa) and elongation (7.5%) and enter them into the database;

[0059] 4. Build a neural network (2 hidden layers, 29 neurons, learning rate 0.0002), with the prediction error RMSE ≤ 5%;

[0060] 5. Obtain the optimal parameters after genetic algorithm optimization: Zn 4.2 wt.%, vibration time 80 s, extrusion pressure 180 MPa, and the tensile strength is increased to 230 MPa.

[0061] Table 1 and Table 2 show the partial composition and process database of rare earth magnesium alloys established according to the steps S1 - S5 of the method of the embodiment of the present invention.

[0062] Table 1 Composition data of partial ultrasonic pulping semi - solid rheological forming rare earth magnesium alloys

[0063]

[0064]

[0065] Table 2 Process and property data of partial ultrasonic pulping semi - solid rheological forming rare earth magnesium alloys

[0066]

[0067]

[0068] The hyperparameters used in the established neural network are as follows: the number of hidden layers is 2, the number of neurons is 29, the activation function is tanh, the α value is 0.0029, the learning rate is 0.000163. The comparison between the predicted values and the actual values of the neural network model for tensile strength, yield strength and elongation is as Figure 4 、 Figure 5 and Figure 6 shown. The abscissa in the figure is the actual value of the mechanical property, and the ordinate is the predicted value of the neural network model. And Figure 4 、 Figure 5 and Figure 6 , for the prediction results of tensile strength, the RMSE is 6.68, the MSE is 5.60, and the R2 is 0.89; for the prediction results of yield strength, the RMSE is 5.25, the MSE is 4.92, and the R2 is 0.83; for the prediction results of elongation, the RMSE is 0.89, the MSE is 0.72, and the R2 is 0.84. Table 3 shows the results of the composition and process parameters of the rare earth magnesium alloy predicted based on the established neural network model.

[0069] Table 3 Results of the composition and process parameters of the rare earth magnesium alloy designed based on machine learning

[0070] Tensile strength Predicted value Error Yield strength Predicted value Error Elongation Predicted value Error 1 212 204 3.64％ 189 173.57 8.16％ 2.2 2.2 0.00％ 2 248 255.15 2.88％ 149 148.61 0.00％ 10.3 9.7 5.40％ 3 215 207.96 3.27％ 172 166.77 3.04％ 8.7 8.7 0.40％ 4 192 194.60 1.35％ 95 98.02 3.18％ 17 17.1 0.34％ 5 227 244.68 7.79％ 108 129.94 20.32％ 11 10.6 3.26％

[0071] It can be seen from the results that for the rare earth magnesium alloy designed by the neural network model constructed by the method of the embodiment of the present invention, the difference between the actual value and the predicted value of the tensile strength and elongation is relatively small.

[0072] In an alternative embodiment of the present invention, a random forest can be used to replace the neural network to establish a preliminary prediction model; the particle swarm optimization algorithm can be used to replace the genetic algorithm for multi-objective optimization; electromagnetic stirring can be used to assist ultrasonic pulping to further enhance the uniformity of the slurry. The process of implementing the method of the embodiment of the present invention using the above alternative solutions is similar to that of Embodiment 1 and will not be elaborated here.

[0073] Using the deep learning method, learn the database of the composition, process and mechanical properties of the rare earth magnesium alloy to generate multiple groups of new data that conform to the distribution of the original database, and use the optimized neural network model to predict the performance to obtain a new database. Perform data analysis on the new database to obtain the range of the composition and process parameters of the rare earth magnesium alloy under different mechanical properties, as shown in the following table.

[0074] Table 4 Range of the composition and process parameters of the rare earth magnesium alloy obtained by prediction under different properties ​

[0075]

[0076] In summary, the design model of the present invention guided by mechanical properties provides a rapid and effective method for realizing high-strength and high-toughness ultrasonic pulping semi-solid rheoforming rare earth magnesium alloys and their compositions and processes.

[0077] The present invention can be applied to the production of high-performance magnesium alloy components in fields such as aerospace and automotive, such as engine brackets, housings, etc. By guiding the process design with the model, the optimal balance between material properties and cost can be achieved.

[0078] In the description of the embodiments of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "center", "top", "bottom", "top part", "bottom part", "inner", "outer", "inner side", "outer side", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. Among them, the "inner side" refers to the internal or enclosed area or space. The "periphery" refers to the area around a specific component or specific area.

[0079] In the description of the embodiments of the present invention, the terms "first", "second", "third", "fourth" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first", "second", "third", "fourth" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise stated, the meaning of "a plurality" is two or more.

[0080] In the description of the embodiments of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", "joined", "assembled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0081] In the description of the embodiments of the present invention, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0082] In the description of the embodiments of the present invention, it should be understood that "-" and "~" represent the range between two numerical values, and this range includes the endpoints. For example, "A - B" represents a range greater than or equal to A and less than or equal to B. "A ~ B" represents a range greater than or equal to A and less than or equal to B.

[0083] In the description of the embodiments of the present invention, the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0084] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the mechanical properties and optimizing the process of rare earth magnesium alloy based on machine learning, characterized in that It includes the following steps: (a) Construct a database containing the composition of rare earth magnesium alloy, process parameters and corresponding mechanical properties; (b) Perform data preprocessing on the database, including feature screening and data normalization; (c) Based on the preprocessed data, establish a machine learning model with alloy composition and process parameters as inputs and mechanical properties as outputs; (d) Use a multi-objective optimization algorithm to optimize the machine learning model to generate a combination of alloy composition and process parameters that meet the target mechanical properties; (e) Based on the optimization results, output the range of preparation process parameters of the rare earth magnesium alloy.

2. The method according to claim 1, characterized in that, The machine learning model in step (c) includes at least one of neural network, random forest, and support vector machine.

3. The method according to claim 1, characterized in that, The multi-objective optimization algorithm in step (d) includes genetic algorithm, particle swarm algorithm or simulated annealing algorithm, and the optimization objectives include at least two of tensile strength, yield strength, and elongation.

4. The method according to claim 1, characterized in that It further includes expanding the database through a generative adversarial network, generating new data that conforms to the original distribution, and updating the machine learning model.

5. The method according to claim 1, wherein The process parameters include at least three of the vibration frequency, power, and time of ultrasonic pulping, and the nominal force, ejection force, and slider speed of the rheoforming process.

6. The method according to claim 1, wherein The rare earth elements include at least two of Zn, Y, Ce, La, Zr, and Ni.

7. The method according to claim 2, characterized in that, The hyperparameters of the machine learning model are automatically tuned by an optimizer, and the optimizer includes optuna optimizer, GridSearch or Bayesian optimization algorithm.

8. The method according to claim 1, wherein The data preprocessing method in step (b) includes Pearson correlation coefficient analysis, recursive feature elimination or principal component analysis.

9. The method according to claim 4, characterized in that The synthetic data generated by the generative adversarial network needs to pass the Kolmogrov-Smirnov test to ensure consistency with the original data distribution.

10. A rare earth magnesium alloy product, characterized in that, Designed and produced by the method described in any one of claims 1-9, its tensile strength ≥ 180 MPa and elongation ≥ 6%.

Images