Alloy material heat treatment process design method based on machine learning model
By constructing an alloy dataset and using machine learning models to screen and optimize process parameters, the problem of low efficiency in the design of traditional alloy material heat treatment processes was solved, and efficient and accurate multi-objective performance optimization was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH BEIJING
- Filing Date
- 2026-02-14
- Publication Date
- 2026-06-12
AI Technical Summary
Traditional alloy material heat treatment process design relies on engineering experience and a large number of experiments, which is inefficient and costly. It is difficult to achieve efficient and accurate process optimization in complex alloy systems with multiple compositions, especially in high-dimensional parameter spaces with coupled multi-objective performance, where model predictions and actual results lack consistency.
An alloy dataset was constructed based on a machine learning model. Multiple machine learning algorithms were used to select the optimal prediction model, a process parameter space was defined, performance was predicted, and the preferred process was selected. The optimized process was then verified by experiments.
It enables rapid and automated design of heat treatment processes for alloy materials, improves the ability to coordinate and optimize multi-objective performance in a multi-dimensional parameter space, and enhances the generalization ability and prediction accuracy of the model.
Smart Images
Figure CN122201495A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of heat treatment technology, and in particular to a method for designing heat treatment processes for alloy materials based on machine learning models. Background Technology
[0002] Heat treatment can effectively control the grain size, type and distribution of precipitated strengthening phases, and stress state of alloy materials, thereby achieving a synergistic improvement in properties such as strength, plasticity, and toughness. Traditional heat treatment process design mainly relies on engineering experience and extensive experimentation, resulting in low efficiency, high trial-and-error costs, and poor optimization capabilities, making it difficult to meet the current demands for diversified, personalized, and rapid development of alloy materials. Especially when dealing with complex, multi-component alloy systems, the relationship between process, microstructure, and properties is highly nonlinear, with significant coupling between parameters, making it difficult for existing methods to achieve efficient and precise process optimization.
[0003] CN119862683A discloses a heat treatment process design method based on the principle of large models, aiming to replace the traditional experience-based parameter trial and error with data-driven approaches, and to quickly and automatically deduce suitable heat treatment processes for given target material properties. CN117521518A proposes an iterative machine learning optimization framework for magnesium alloy heat treatment: initial data is obtained through orthogonal design, an SVM regression model with confidence intervals is constructed, the expected value improvement (EGO) strategy of Kriging is used to select the optimal next experimental point and incorporate new data, iterating until the performance meets the target, thereby reducing experiments and accelerating convergence under uncertainty guidance. Although there have been several studies on heat treatment process optimization based on different models, existing technologies still struggle to achieve efficient model transfer and adaptation between different alloy systems, lack consistency between model predictions and actual heat treatment effects, and in high-dimensional parameter spaces involving multiple objectives such as strength, toughness, and plasticity coupling, it is difficult to balance optimization efficiency and generalization ability when resources are limited.
[0004] Therefore, there is a need to improve the design methods for heat treatment processes of alloy materials in the existing technology. Summary of the Invention
[0005] In view of this, the purpose of this invention is to propose a heat treatment process design method for alloy materials based on machine learning models. This method is based on a constructed alloy heat treatment dataset and combines multiple machine learning models to quickly infer and optimize the heat treatment process parameters of alloy materials.
[0006] To achieve the above objectives, one aspect of the present invention provides a method for designing heat treatment processes for alloy materials based on machine learning models, comprising the following steps:
[0007] S1 obtains alloy composition, process parameters, and performance parameters to establish an alloy dataset;
[0008] S2 preprocesses the alloy dataset and uses various machine learning algorithms to train different target mechanical properties in order to select the corresponding optimal machine learning prediction model.
[0009] S3 defines a process space based on the target alloy composition, which includes multiple combinations of candidate process parameters;
[0010] S4 uses machine learning models to predict the performance of candidate process parameters in the process space and selects the preferred process based on preset performance targets.
[0011] In some implementations, the method further includes:
[0012] S5 performs experimental verification of the preferred process and feeds the verification results back to the dataset. S2 to S5 are repeated until an optimized process that meets the preset performance target is obtained.
[0013] In some implementations, in S1,
[0014] The alloy composition includes the base elements of the alloy and one or more alloying elements;
[0015] The process parameters include at least the temperature and time of solution treatment and the temperature and time of aging treatment;
[0016] Performance parameters include tensile strength, yield strength, and elongation at break.
[0017] In some implementations, preprocessing the alloy dataset in S2 includes:
[0018] Data with missing values exceeding a preset threshold in the alloy dataset is removed, data records with missing values within the preset threshold are filled in, and the feature units in the dataset are standardized.
[0019] In some implementations, in S2, multiple machine learning algorithms include at least two of the following: random forest regression, gradient boosting decision tree, support vector machine, and K-nearest neighbor algorithm; for different target mechanical properties, the model accuracy is evaluated by cross-validation and by combining the coefficient of determination and root mean square error index, and the corresponding optimal machine learning prediction model is independently selected.
[0020] In some implementations, the coefficient of determination The calculation method is as follows:
[0021]
[0022] The root mean square error is calculated as follows:
[0023]
[0024] In the formula: Represents the actual value. This represents the average value. This represents an estimated value.
[0025] In some implementations, in S3, defining a process space that includes multiple combinations of candidate process parameters includes:
[0026] Based on the basic process parameters, the search range and step size are set for the temperature and time parameters respectively, thereby generating multiple sets of candidate process parameter combinations.
[0027] In some implementations, the basic process parameters are obtained in the following ways:
[0028] Based on the target alloy composition, samples with similar compositions and meeting mechanical property standards are retrieved from the alloy dataset, and the process parameters of these samples are used as the basic process parameters.
[0029] Alternatively, process parameters recommended for the target alloy composition may be used from heat treatment process manuals, industry-standard materials, or material databases.
[0030] In some implementations, in S4, performance prediction of candidate process parameters within the process space includes:
[0031] The composition of the target alloy and each set of candidate process parameters in the process space are combined and input into the optimal machine learning prediction model for the corresponding mechanical properties selected in S2 for performance prediction.
[0032] In some implementations, in S5, experimental verification of the preferred process includes: performing actual heat treatment on the target alloy using the selected preferred process, and testing tensile strength, yield strength, and elongation at break;
[0033] If the test results meet the preset performance target, the process terminates and outputs that the process is an optimized process.
[0034] If the conditions are not met, the alloy composition, process parameters, and performance parameters obtained in this experiment will be added to the alloy dataset of S1 as new data, and S2-S5 will be repeated.
[0035] The present invention has at least the following beneficial technical effects:
[0036] The method of this invention constructs a "composition-process-performance" dataset and trains a machine learning model, enabling it to automatically learn complex nonlinear relationships from massive historical data and quickly predict the performance of a given alloy composition under different heat treatment processes. By defining a "process parameter space" and using the model to perform full-space performance prediction, this method can systematically traverse and evaluate a large number of candidate process combinations, thereby discovering potential optimal process parameters that are difficult to reach through experience or local experiments alone, significantly improving the possibility of obtaining a globally better solution. Attached Figure Description
[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.
[0038] Figure 1 A schematic diagram illustrating an embodiment of the alloy material heat treatment process design method based on machine learning model provided by the present invention;
[0039] Figure 2 A schematic diagram showing the mechanical properties corresponding to all process points in the process space;
[0040] Figure 3 This is a diagram showing the comparison of mechanical performance test results before and after optimization. Detailed Implementation
[0041] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to specific examples and the accompanying drawings.
[0042] It should be noted that all uses of "first" and "second" in the embodiments of the present invention are for the purpose of distinguishing two entities or parameters with the same name but different names. It is clear that "first" and "second" are only for the convenience of expression and should not be construed as limiting the embodiments of the present invention. Subsequent embodiments will not explain this in detail.
[0043] This invention primarily addresses the following technical problems in the design of existing heat treatment processes for alloy materials: Traditional heat treatment process design mainly relies on engineering experience and numerous trial-and-error experiments, resulting in low R&D efficiency, long cycles, and high costs; Furthermore, when facing complex alloy systems with multiple compositions, existing methods struggle to accurately construct highly nonlinear mapping relationships between composition, process, and performance, and cannot achieve efficient synergistic optimization of multiple target properties such as strength and plasticity in a multidimensional parameter space. The generalization ability and prediction accuracy of the models are insufficient to meet practical engineering needs.
[0044] Based on the above-mentioned technical problems, the first aspect of the present invention proposes an embodiment of a heat treatment process design method for alloy materials based on a machine learning model. Figure 1 The diagram illustrates an embodiment of the alloy material heat treatment process design method based on a machine learning model provided by the present invention. Figure 1 As shown, the alloy material heat treatment process design method based on machine learning model according to an embodiment of the present invention includes the following steps:
[0045] S1 obtains alloy composition, process parameters, and performance parameters to establish an alloy dataset;
[0046] S2 preprocesses the alloy dataset and uses various machine learning algorithms to train different target mechanical properties in order to select the corresponding optimal machine learning prediction model.
[0047] S3 defines a process space based on the target alloy composition, which includes multiple combinations of candidate process parameters;
[0048] S4 uses machine learning models to predict the performance of candidate process parameters in the process space and selects the preferred process based on preset performance targets.
[0049] Furthermore, the method also includes:
[0050] S5 performs experimental verification of the preferred process and feeds the verification results back to the dataset. S2 to S5 are repeated until an optimized process that meets the preset performance target is obtained.
[0051] Furthermore, in S1, by collecting and organizing data on aluminum alloy material composition, heat treatment process, and mechanical properties from literature, an alloy "composition-process-performance" dataset is established.
[0052] The alloy composition includes the base element of the alloy and one or more alloying elements; wherein the base element is such as Al, Cu, Fe, Ti, Ni, etc. and the alloying element is such as Zn, Mg, Si, Mn, etc. In some embodiments, the composition of aluminum alloy includes Al, Zn, Mg, Cu, etc., and copper alloy includes Cu, Zn, Sn, Ni, Al, etc.
[0053] The alloy composition includes the matrix elements of the alloy and one or more alloying elements; the process parameters include at least the temperature and time of solution treatment and the temperature and time of aging treatment; the performance parameters include tensile strength, yield strength and elongation at break.
[0054] Furthermore, in S2, the preprocessing of the alloy dataset includes:
[0055] Data with missing values exceeding a preset threshold in the alloy dataset is removed, data records with missing values within the preset threshold are filled in, and the feature units in the dataset are standardized.
[0056] Furthermore, in S2, various machine learning algorithms are employed, including: Random Forests Regression (RF), Extreme Gradient Boosting (XGBoost), Gradient Boosting Decision Tree (GBDT), Decision Tree Regression (DTR), Polynomial Support Vector Regression (polySVR), K-Nearest Neighbors (KNN), NaiveBayes (Bayes), and Radial Basis Function Support Vector Regression (rbfSVR). For different target mechanical properties, cross-validation is used, combined with the coefficient of determination and root mean square error (RMSE) to evaluate model accuracy, independently selecting the optimal machine learning prediction model. In some implementations, the coefficient of determination... The calculation method is as follows:
[0057]
[0058] The root mean square error is calculated as follows:
[0059]
[0060] In the formula: Represents the actual value. This represents the average value. This represents an estimated value.
[0061] Furthermore, in S3, the process space defined to include multiple combinations of candidate process parameters includes:
[0062] Based on the basic process parameters, the search range and step size are set for the temperature and time parameters respectively, thereby generating multiple sets of candidate process parameter combinations.
[0063] Furthermore, the methods for obtaining basic process parameters include:
[0064] Based on the target alloy composition, samples with similar compositions and meeting mechanical property standards are retrieved from the alloy dataset, and the process parameters of these samples are used as the basic process parameters.
[0065] Alternatively, process parameters recommended for the target alloy composition may be used from heat treatment process manuals, industry-standard materials, or material databases.
[0066] Furthermore, in S4, based on the initially determined heat treatment process, temperature and time are used as key process parameters, and an appropriate step size is used to establish the heat treatment process space. For example, based on the initially determined solution temperature or aging temperature, a temperature range (e.g., ±50℃) and a time range (e.g., ±10 h) are set, and a temperature step size (e.g., 5℃) and a time step size (e.g., 0.5 h) are set, thereby constructing a high-dimensional process space containing a large number of potential combinations of process parameters.
[0067] In S4, performance prediction of candidate process parameters within the process space includes:
[0068] The composition of the target alloy and each set of candidate process parameters in the process space are combined and input into the optimal machine learning prediction model for the corresponding mechanical properties selected in S2 for performance prediction.
[0069] Specifically, all heat treatment process parameters within the process space, combined with aluminum alloy material composition data, are input into a mechanical property machine learning model for prediction. The predicted results for tensile strength, yield strength, and elongation at break corresponding to all process points are then summarized and organized. Based on the mechanical property data predicted by the model, a mechanical property space is established using yield strength, tensile strength, and elongation at break as coordinate axes. Target mechanical property screening criteria are set (e.g., tensile strength greater than a set threshold and elongation at break not lower than a set threshold), and data points in the mechanical property space are screened to optimize the heat treatment process parameters that meet the target performance requirements. The optimized heat treatment process is used to treat the alloy material to be optimized, and standard mechanical property tests, including tensile strength, yield strength, and elongation at break, are performed on the treated material to obtain actual performance data.
[0070] Furthermore, in S5, the experimental verification of the preferred process includes: performing actual heat treatment on the target alloy using the selected preferred process, and testing the tensile strength, yield strength and elongation at break;
[0071] If the test results meet the preset performance target, the process terminates and outputs that the process is an optimized process.
[0072] If the conditions are not met, the alloy composition, process parameters, and performance parameters obtained in this experiment will be added to the alloy dataset of S1 as new data, and S2-S5 will be repeated.
[0073] The specific embodiments of the present invention are further described below with reference to specific examples.
[0074] Taking a 7xxx series aluminum alloy as an example, this paper describes the rapid design, optimization, and verification of a heat treatment process, illustrating the specific application process of the rapid design and optimization method for alloy heat treatment processes based on machine learning models described in this invention.
[0075] Establishment of a dataset for the "composition-processing-properties" of S1 alloy
[0076] A dedicated dataset for 7xxx series aluminum alloys was constructed by collecting data from publicly published academic papers and historical laboratory experimental records. This dataset includes characteristic variables such as the mass percentages of the main alloy elements Zn, Mg, and Cu, as well as heat treatment process characteristics, namely solution treatment temperature, solution treatment time, aging treatment temperature, and aging treatment time. The target variables are the mechanical properties of the alloys, including tensile strength, yield strength, and elongation at break. After constructing the dataset, the collected raw data underwent standardization and cleaning, removing samples with missing key information to ensure data integrity and validity.
[0077] S2 preprocesses the alloy dataset and uses various machine learning algorithms to train different target mechanical properties in order to select the corresponding optimal machine learning prediction model.
[0078] Using the preprocessed dataset, machine learning prediction models were trained for three target performance parameters: tensile strength, yield strength, and elongation at break. The algorithms used in training included Random Forest Regression (RF), Extreme Gradient Boosting Tree (XGBoost), Gradient Boosting Decision Tree (GBDT), Decision Tree Regression (DTR), Polynomial Kernel Support Vector Machine (polySVR), K-Nearest Neighbors (KNN), Bayesian Ridge Regression (Bayes), and Gaussian Kernel Support Vector Machine (rbfSVR). It is important to note that the machine learning algorithms are not limited to the types mentioned above. During this process, ten-fold cross-validation was used to divide the training and test sets, and the coefficient of determination (R²) and root mean square error (RMSE) were used as evaluation metrics to assess the prediction accuracy of each model.
[0079] Based on the evaluation results, the optimal combined prediction model was finally determined: for yield strength, the gradient boosting decision tree (GBDT) model with the highest prediction accuracy was selected; for tensile strength, the extreme gradient boosting tree (XGBoost) model with the best prediction performance was selected; and for fracture elongation, the random forest regression (RF) model with the best fitting effect was selected.
[0080] Preliminary determination of S3 heat treatment process and establishment of process space
[0081] Combining the chemical composition information of the alloy material to be optimized with the pre-processed "composition-process-performance" dataset established in step 1, one or more candidate heat treatment processes that meet the target mechanical performance requirements are preliminarily determined through retrieval and screening. In this embodiment, for the target aluminum alloy material, solution treatment combined with two-stage aging treatment is selected as the initial process route. The solution treatment regime is fixed at 470℃ for 2 hours; the aging treatment adopts a two-stage aging regime, with the first-stage aging temperature set at 120℃ and the second-stage aging temperature set at 160℃. Specific process parameters are shown in Table 1.
[0082] Table 1 Initial Heat Treatment Process Parameters
[0083]
[0084] Based on the defined initial process, a high-dimensional process space was established with aging temperature and aging time as the core optimization variables. The solution temperature range was set at 460-480 ℃, with a step size of 5, meaning a data point was taken every 5 ℃ starting from 460 ℃. The product of the number of data points in all spaces is the total number of data entries in the process space. After being divided according to Table 2, the process space contains a total of 7680 process groups.
[0085] Table 2 Process Space Division
[0086]
[0087] S4 uses machine learning models to predict the performance of candidate process parameters within the process space and selects the optimal process based on preset performance targets.
[0088] All temperature and time parameters generated within the process space described in S3 are combined with the fixed composition data of the aluminum alloy and input into the three mechanical property prediction models established in S2 for prediction. For example... Figure 2 As shown in the figure, each scatter point represents the predicted mechanical properties under a set of virtual process parameters. A multi-objective screening strategy is set, that is, using the fracture elongation meeting a preset requirement as a constraint, the set of process parameters with the highest predicted tensile strength value is selected as the preferred process from the prediction results. According to this strategy, the model selects a specific set of solid solution + two-stage aging processes, and the prediction results show that this process can significantly improve the material strength while ensuring plasticity.
[0089] S5 Experimental Verification
[0090] The same batch of aluminum alloy materials was treated using the heat treatment process obtained from S4 optimization, and optimized group samples were prepared for room temperature tensile testing, marked as corresponding attachments. Figure 3 Sample #2 was selected from the samples listed in Table 1. Simultaneously, mechanical property tests were conducted on the six groups of samples after preliminary heat treatment. The group with the highest tensile strength and yield strength was selected as the control group and marked as the corresponding sample. Figure 3 Sample 1# in the sample. Figure 3 The results shown in the diagram and mechanical property tests indicate that the horizontal axis represents strain and the vertical axis represents stress (MPa). Compared to sample #1, which showed the best performance in the initial experimental scheme, sample #2, optimized by this invention, exhibited a 30.08 MPa increase in tensile strength and a 41.86 MPa increase in yield strength. Although the elongation at break decreased by 3.03%, it remained within the range required for use. These significant improvements in strength indicators demonstrate that by establishing a process space and combining it with a machine learning model for optimization, better process parameters than those obtained through traditional experimental screening can be identified. This confirms the reliability and effectiveness of the heat treatment process optimization method proposed in this invention.
[0091] The mechanical property test results of the optimized sample (2#) are compared with those of the control group (1#) and the preset target performance requirements. In this embodiment, the strength index of the optimized sample (2#) is significantly better than that of the best sample (1#) selected by traditional experiments, and the plasticity remains within the qualified range, proving that the optimization is successful. This process is output as the final heat treatment scheme. If the measured results do not achieve the expected improvement effect, the new data generated in this experiment are fed back to the dataset, the model is retrained and the process space is adjusted, and S2-S5 are repeated until the optimal process that meets the requirements is obtained.
[0092] It should be particularly noted that each step in the various embodiments of the above-mentioned alloy material heat treatment process design method based on machine learning model can be interleaved, substituted, added, or deleted. Therefore, these reasonable permutations and combinations of the alloy material heat treatment process design method based on machine learning model should also fall within the protection scope of this invention, and the protection scope of this invention should not be limited to the embodiments.
[0093] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.
[0094] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.
[0095] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0096] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.
[0097] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.
Claims
1. A method for designing heat treatment processes for alloy materials based on machine learning models, characterized in that, Includes the following steps: S1 obtains alloy composition, process parameters, and performance parameters to establish an alloy dataset; S2 preprocesses the alloy dataset and trains it using multiple machine learning algorithms based on different target mechanical properties to select the corresponding optimal machine learning prediction model. S3 defines a process space based on the target alloy composition, which includes multiple combinations of candidate process parameters; S4 uses the machine learning model to predict the performance of candidate process parameters in the process space, and selects the preferred process based on the preset performance target.
2. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, Also includes: S5 performs experimental verification of the preferred process and feeds back the verification results to the dataset. S2 to S5 are repeated until an optimized process that meets the preset performance target is obtained.
3. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, In S1, The alloy composition includes the base elements of the alloy and one or more alloying elements; The process parameters include at least the temperature and time of solution treatment and the temperature and time of aging treatment; The performance parameters include tensile strength, yield strength, and elongation at break.
4. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, In S2, the preprocessing of the alloy dataset includes: Data in the alloy dataset with missing values exceeding a preset threshold are removed, data records with missing values within the preset threshold are completed, and the feature units in the dataset are standardized.
5. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, In S2, the various machine learning algorithms include at least two of the following: random forest regression, gradient boosting decision tree, support vector machine, and K-nearest neighbor algorithm; for different target mechanical properties, the model accuracy is evaluated by cross-validation and by combining the coefficient of determination and root mean square error index, and the corresponding optimal machine learning prediction model is independently selected.
6. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 5, characterized in that, Coefficient of determination The calculation method is as follows: The method for calculating the root mean square error is as follows: In the formula: Represents the true value. This represents the average value. This represents an estimated value.
7. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, In S3, the process space defined to include multiple combinations of candidate process parameters includes: Based on the basic process parameters, search ranges and step sizes are set for temperature and time parameters respectively, thereby generating the multiple sets of candidate process parameter combinations.
8. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 7, characterized in that, The methods for obtaining the basic process parameters include: Based on the target alloy composition, samples with similar composition and meeting mechanical properties are retrieved from the alloy dataset, and the process parameters of these samples are used as the basic process parameters. Alternatively, process parameters recommended for the target alloy composition may be used from heat treatment process manuals, industry-standard materials, or material databases.
9. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 1, characterized in that, In S4, performance prediction of candidate process parameters within the process space includes: The composition of the target alloy is combined with each set of candidate process parameters in the process space and input into the optimal machine learning prediction model for the corresponding mechanical properties selected in S2 for performance prediction.
10. The method for designing heat treatment processes for alloy materials based on machine learning models according to claim 2, characterized in that, In S5, the experimental verification of the preferred process includes: performing actual heat treatment on the target alloy using the selected preferred process, and testing the tensile strength, yield strength and elongation at break; If the test results meet the preset performance target, the process terminates and outputs the process as the optimized process. If the conditions are not met, the alloy composition, process parameters, and performance parameters obtained in this experiment will be added as new data to the alloy dataset in S1, and S2-S5 will be repeated.