Ultra-high strength and toughness nickel-based corrosion-resistant alloy component design method based on machine learning

Abstract

The invention discloses an ultrahigh-toughness nickel-based corrosion-resistant alloy component design method based on machine learning, and belongs to the field of alloy component design. The method comprises the following steps: defining a target of alloy component optimization; an alloy strength prediction model is constructed through a machine learning algorithm and trained, and a trained strength prediction model is obtained through testing; thermodynamic calculation is conducted on multiple groups of alloy components in the target of alloy component optimization, and the precipitated phase volume fraction VF and the average radius R are obtained; inputting the calculated precipitated phase volume fraction VF and average radius R into the trained strength prediction model to obtain the strength value of each combination of alloy components, and performing fitness screening and sorting; the alloy components with high fitness are reserved, the other alloy components are subjected to crossover and mutation operation through a genetic algorithm and then subjected to thermodynamic calculation again with the alloy components with high fitness, and the corresponding precipitated phase volume fraction VF and the average radius R are obtained; and the operation is repeated until the preset maximum number of iterations is reached or the convergence condition is met, and the alloy component with the optimal fitness is selected as the ultrahigh-toughness nickel-based corrosion-resistant alloy component. According to the method, thermodynamic calculation and machine learning are combined to reversely design the nickel-based corrosion-resistant alloy, and efficient optimization of components of the ultrahigh-toughness corrosion-resistant alloy is achieved through a genetic preferential algorithm.

Description

Technical Field

[0001] The present invention relates to alloy composition design, and particularly to a method for designing the composition of a super-high strength and toughness nickel-based corrosion-resistant alloy based on machine learning. Background Art

[0002] With the extension of oil and gas exploration from land to sea, and the development of oil and gas from medium and deep layers to deep and ultra-deep layers, from medium and shallow waters to deep and ultra-deep waters, and from conventional oil and gas to unconventional oil and gas, the demand for super-high strength and toughness high-performance corrosion-resistant alloys in oil and gas drilling using measurement while drilling, directional drilling, and subsea wellhead equipment is increasing. The nickel-based corrosion-resistant alloy has a high degree of alloying and many elements, and the interaction between elements limits the design of alloy composition to a certain extent. The development of super-high strength nickel-based corrosion-resistant alloys is challenging. The traditional alloy design idea mainly uses the forward trial-and-error method to design new alloy compositions within a certain range, and then finally finds new alloys with ideal target properties through process optimization. The development cycle is long, the cost is high, and there is great uncertainty. With the continuous development of machine learning methods in the field of materials, establishing a machine learning model through existing data and combining with an optimization algorithm to optimize the ingredient formula has been successfully applied to the optimization and design of material composition. However, the machine learning prediction model directly constructed based on existing composition and performance data often has non-explainability and requires a large amount of reliable data accumulation to ensure the accuracy of the prediction model.

[0003] Therefore, there is an urgent need to propose a new alloy composition design method to solve the above problems. Summary of the Invention

[0004] For this reason, the purpose of the present invention is to provide a method for designing the composition of a super-high strength and toughness nickel-based corrosion-resistant alloy based on machine learning, which uses a reverse design method of nickel-based corrosion-resistant alloy by combining thermodynamic calculation and machine learning, and uses a genetic optimization algorithm to realize the efficient optimization of the composition of super-high strength and toughness corrosion-resistant alloy.

[0005] According to the technical solution of the present invention, there is provided a method for designing the composition of a super-high strength and toughness nickel-based corrosion-resistant alloy based on machine learning, wherein the method includes the following steps:

[0006] S1: Define the objectives of alloy composition optimization, including multiple alloy compositions;

[0007] S2: According to the alloy strength, the volume fraction VF of the precipitated phase, and the average radius R of the known alloy samples, construct an alloy strength prediction model through a machine learning algorithm, train and test it, and obtain a trained strength prediction model;

[0008] S3: Perform thermodynamic calculations on the multiple alloy compositions in the objectives of the alloy composition optimization to obtain the volume fraction VF of the precipitated phase and the average radius R;

[0009] S4: Input the calculated precipitation phase volume fraction VF and the average radius R into the trained strength prediction model to obtain the strength values of each alloy composition, and perform fitness screening and sorting.

[0010] S5: Retain the alloy compositions with higher fitness. For the remaining alloy compositions, after performing crossover and mutation operations through the genetic algorithm, perform thermodynamic calculations again with the alloy compositions with higher fitness to obtain the corresponding precipitation phase volume fraction VF and average radius R.

[0011] S6: Repeat S4 - S5 until the preset maximum number of iterations is reached or the convergence condition is satisfied, and select the alloy composition with the optimal fitness as the ultra-high strength and toughness nickel-based corrosion-resistant alloy composition.

[0012] Further, the S1 specifically includes: Initialize a binary population, where each individual represents a set of alloy compositions.

[0013] Further, each binary individual is converted into the corresponding alloy composition through decoding.

[0014] Further, the size of the population is set according to the problem complexity and computing power.

[0015] Further, the S2 specifically includes:

[0016] S2.1: Collect the composition, heat treatment process, and their corresponding yield strength data of known nickel-based corrosion-resistant alloys to form an initial data set.

[0017] S2.2: Perform standardization processing on the initial data set to obtain a standard data set, and calculate the precipitation phase volume fraction VF and average radius R of each alloy through thermal / kinetic calculation software. Combine the precipitation phase volume fraction VF and average radius R with the standard data set to form a final data set.

[0018] S2.3: Divide the final data set into a training set and a test set according to the ratio n:1 through the multiple holdout method, where n is a positive integer.

[0019] S2.4: Use a machine learning algorithm to construct an Adaptive Boosting Regression (ABR) model. With the precipitation phase volume fraction VF and average radius R as input variables and the yield strength as the output variable, train and test the ABR model to obtain the trained ABR model as the strength prediction model.

[0020] Further, in the S2.1, the data collection methods in the initial data set include experimental data, literature data, and database queries.

[0021] Further, in S2.1, if there are missing parts in the initial dataset, data integrity repair is performed by filling in the missing values.

[0022] Further, the methods for filling in the missing values include mean filling, the nearest neighbor method, and other suitable interpolation methods.

[0023] Further, in S2.1, each set of data in the initial dataset includes the composition of a nickel-based corrosion-resistant alloy (including element types and their contents), heat treatment processes (including parameters such as heat treatment temperature and time), and the corresponding yield strength.

[0024] Further, in S2.2, the formula for standardization is: z = (x - μ) / σ,

[0025] where z is the standardized data, x is the original data in the initial dataset, μ is the mean of each dimension variable in the original data, and σ is the standard deviation of each dimension variable in the original data.

[0026] Further, in S2.2, the thermo / kinetic calculation software is Thermo-Calc.

[0027] Further, in S2.3, n = 4.

[0028] Further, in S2.4, during the training and testing of the ABR model, the hyperparameters of the ABR model are adjusted using the grid search or random search method to optimize the hyperparameters of the model.

[0029] Further, in S3, the thermo / kinetic calculation software Thermo-Calc is used to obtain the precipitation phase volume fraction (VF), average radius (R), and the content of Cr in the matrix phase of each alloy, and thermodynamic calculations are performed for multiple combinations of alloy compositions in the target of optimizing the alloy composition.

[0030] Further, in S4, the fitness screening further includes:

[0031] Evaluating the corrosion resistance based on the Cr content in the matrix phase and the pitting corrosion resistance equivalent value PREN, where the mass fraction of Cr exceeds 18%, PREN ≥ 32, and the formula for calculating the pitting corrosion resistance equivalent is PREN = %Cr + 3.3 * %Mo + 16 * %N.

[0032] Further, in S5, sorting according to the fitness from high to low, the top 5% are used as alloy compositions with higher fitness.

[0033] Further, in S6, the genetic parameters are as follows: the number of iterations is 100 generations; the mutation rate is set to 0.1; the crossover rate is 0.6.

[0034] Further, in S6, when the change in fitness of the convergence condition is less than a certain threshold, the algorithm terminates.

[0035] The beneficial effects of adopting the above technical solution are as follows: A method for designing the composition of a super high-strength and high-toughness nickel-based corrosion-resistant alloy based on machine learning is provided. This method is different from the traditional machine learning prediction model directly constructed based on existing composition and performance data. The volume fraction VF and average radius R of the precipitated phases of the alloy sample are calculated through a thermo / kinetic software and used as inputs. A prediction model of the alloy strength with respect to the volume fraction VF and average radius R of the precipitated phases is constructed through a machine learning algorithm, establishing the correlation relationship between the composition, precipitated phases, and strength. Using the constructed strength prediction model as the fitness function, the genetic algorithm is applied to rapidly optimize the strength within the preset composition range. This method solves to a certain extent the problems that the machine learning prediction model directly constructed based on existing composition and performance data usually has non-explainability and requires a large amount of reliable data accumulation to ensure the accuracy of the prediction model.

[0036] Among them, in the present invention, a machine learning algorithm is used to construct an Adaptive Boosting Regression (ABR) model, and the prediction accuracy of the regression model is improved by weighted combination of multiple weak learners. Specifically, the calculated volume fraction of the precipitated phase and the radius of the precipitated phase are used as input features to predict the strength of the alloy. The relationship between the alloy composition and mechanical properties is usually complex and highly non-linear, and traditional methods are often difficult to accurately model. However, ABR can effectively improve the prediction accuracy by adaptively adjusting the weights of each weak learner and can better handle this complex relationship. The prediction error of each weak learner when processing samples will be weighted, so that the model gradually corrects the error and optimizes the prediction effect.

[0037] Among them, the present invention uses a machine learning algorithm to construct an Adaptive Boosting Regression (ABR) model, which improves the prediction accuracy of the regression model by weighted combination of multiple weak learners. Specifically, the calculated volume fraction of the precipitated phase and the radius of the precipitated phase are used as input features to predict the strength of the alloy. The relationship between the alloy composition and mechanical properties is usually complex and highly non-linear, and traditional methods are often difficult to accurately model. However, ABR can effectively improve the prediction accuracy by adaptively adjusting the weights of each weak learner and can better handle this complex relationship. The prediction error of each weak learner when processing samples will be weighted, so that the model gradually corrects the error and optimizes the prediction effect.

[0038] In addition, through the standardization of z = (x - μ) / σ, the data is transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1, eliminating the dimensional differences of different features, making the contributions of each feature to the model equal under the same scale, thus avoiding the improper influence of certain features on the model training process, accelerating the convergence speed of the model, and improving the stability and accuracy of the model. Description of the Drawings

[0039] Figure 1 It is a flowchart of the method provided by the embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the strength prediction result provided by the embodiment of the present invention. Detailed Embodiments

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] The present invention provides a method for designing the composition of a super high-strength and tough nickel-based corrosion-resistant alloy based on machine learning, which adopts a method of reverse designing nickel-based corrosion-resistant alloys by combining thermodynamic calculations and machine learning. A strength prediction model of the volume fraction VF of the precipitated phase and the volume fraction of the average radius R is established through machine learning and used as a strength fitness screening function. Combining with the genetic optimization algorithm, the efficient optimization of the composition of the super high-strength and tough corrosion-resistant alloy is realized.

[0043] Specifically, the present invention first provides a method for designing the composition of a super high-strength and tough nickel-based corrosion-resistant alloy based on machine learning:

[0044] The volume fraction VF of the precipitated phase and the average radius R of the known alloy samples are calculated through thermal / kinetic software and used as inputs. A prediction model of the alloy strength and the volume fraction VF of the precipitated phase and the average radius R is constructed through a machine learning algorithm; the strength prediction model established by selecting the algorithm with the highest accuracy is used as the strength fitness screening function for subsequent genetic algorithm composition optimization. Through genetic algorithm optimization iteration until the preset maximum iteration number is reached or the convergence condition is satisfied, the preferred alloy composition is obtained.

[0045] The method for designing the composition of a super high-strength and high-toughness nickel-based corrosion-resistant alloy provided by the present invention includes the following steps:

[0046] S1: Initialize the population

[0047] Define the objectives of alloy composition optimization, including the types of elements and their composition ranges. Initialize a binary population, where each individual represents a set of alloy compositions. Each binary individual is converted into the corresponding alloy composition through decoding. The size of the population should be set according to the problem complexity and computing power.

[0048] S2: Establish a fitness screening function

[0049] S2.1: Data collection and processing

[0050] Collect the composition, heat treatment process, and their corresponding yield strength data of nickel-based corrosion-resistant alloys. The data are obtained through various methods, including experimental data, literature data, and database queries. For the missing parts in the dataset, data integrity is repaired by filling in the missing values using methods such as mean filling, nearest neighbor method, or other suitable interpolation methods to form an initial dataset. Each set of data includes the composition of a nickel-based corrosion-resistant alloy (including the types of elements and their contents), the heat treatment process (including parameters such as heat treatment temperature and time), and the corresponding yield strength.

[0051] S2.2: Standardization processing

[0052] Perform standardization processing on each set of original data to ensure that the dimensions of different feature data are consistent and avoid affecting the model training effect due to differences in feature dimensions. The standardization formula is: z = (x - μ) / σ,

[0053] where z is the standardized data, x is the original data, μ is the mean of each dimension variable in the original data, and σ is the standard deviation of each dimension variable in the original data. Then, using the thermodynamics / kinetics calculation software Thermo-Calc, based on the composition and heat treatment process of the nickel-based corrosion-resistant alloy, calculate the precipitation phase volume fraction (VF) and average radius (R) of each alloy, and combine VF and R with the standard dataset to form the final standard dataset.

[0054] S2.3: Dataset division

[0055] Divide the standard dataset into a training set and a test set according to the ratio n:1, ensuring that the training set can fully represent the variation range of alloy composition and performance. Preferably, n = 4. The training set is used for model training, and the test set is used for model performance evaluation. The training set and test set are divided by the multiple holdout method. Each time, data randomly selected from the standard dataset at a ratio of n / (n + 1) is used as the training set, and the remaining data is used as the test set.

[0056] S2.4: Build a machine learning model

[0057] Based on the training set, a machine learning algorithm is used to construct an ABR (Adaptive Boosting Regression) model. Specifically, the ABR model improves the performance of the weak regressor through weighted averaging. The volume fraction of the precipitated phase (VF) and the average radius (R) are combined as input variables, and the yield strength is used as the output variable to train the ABR model to obtain the prediction ability of alloy properties. Hyperparameter optimization is performed on the ABR model, and grid search or random search methods are used to adjust the hyperparameters of the model to obtain the ABR model with the best performance on the test set.

[0058] S3: Fitness function ranking

[0059] Thermodynamic calculations are performed on each individual (alloy composition) in the population. The volume fraction of the precipitated phase (VF), the average radius (R), and the mass fraction of Cr in the matrix phase of each alloy are obtained using the thermo / kinetic calculation software Thermo-Calc. The calculation results of the volume fraction of the precipitated phase (VF) and the average radius (R) are used as inputs and fed into the trained ABR model for strength prediction. The model predicts the strength value of each combination of alloys based on VF and R.

[0060] S4: Selection operation

[0061] The strength values predicted by the machine learning model are used to perform fitness screening on all individuals in the population, and individuals with higher fitness are selected and retained. According to the fitness ranking, the top 5% of the excellent individuals are retained for the next generation, and the other individuals are eliminated.

[0062] S5: Crossover and mutation

[0063] Crossover and mutation operations are performed on individuals with lower fitness to generate new offspring individuals. The crossover operation can select two individuals for information exchange to generate new offspring individuals; the mutation operation randomly modifies some genes to increase the diversity of the population and avoid the algorithm falling into a local optimal solution. The crossover and mutation operations should be carried out according to the preset probability to ensure that the search space is fully explored.

[0064] S6: Iteration and termination

[0065] The above process is iterated repeatedly until the preset maximum number of iterations is reached or the convergence condition is satisfied. The convergence condition can be set to terminate the algorithm when the change in the fitness of the population is less than a certain threshold. Finally, the individual with the optimal fitness is selected as the optimization design result of the alloy composition.

[0066] The corrosion resistance is evaluated based on the Cr content in the matrix phase and the pitting resistance equivalent number PREN value, where the mass fraction of Cr exceeds 18% and PREN ≥ 32. The formula for calculating the pitting resistance equivalent number is PREN = %Cr + 3.3 * %Mo + 16 * %N.

[0067] Example

[0068] As Figure 1 shown, the method of this example is as follows:

[0069] A method for designing the composition of a super high-strength and tough nickel-based corrosion-resistant alloy based on machine learning is as follows:

[0070] S1: Data collection and processing

[0071] S1.1: First, collect 42 pieces of data on the composition, heat treatment process, and corresponding yield strength of nickel-based corrosion-resistant alloys. The data sources include experimental data, relevant literature, and public databases. To ensure data integrity, for missing parts, interpolation processing is performed using the mean filling method and the nearest neighbor method to repair missing values, forming an initial data set. Each data record includes the composition (element types and their contents) of the nickel-based alloy, heat treatment temperature, and yield strength value.

[0072] Table 1 Data distribution in the standard data set

[0073]

[0074]

[0075] S1.2: Standardize all the original data using the standardization formula z = (x - μ) / σ. Based on the heat treatment process, use the thermodynamics / kinetics calculation software Thermo-Calc to calculate the precipitation phase volume fraction (VF) and average radius (R) of each nickel-based corrosion-resistant alloy composition in the data set, and incorporate the obtained results into the standard data set to form the final data set.

[0076] S2: Data set division

[0077] S2.1: To train and evaluate the machine learning model, first divide the standard data set into a training set and a test set in a ratio of 4:1. The training set is used to train the machine learning model, and the test set is used for model performance evaluation.

[0078] S3: Construct a machine learning model

[0079] S3.1: Based on the training set, use a machine learning algorithm to construct an ABR (Adaptive Boosting Regression) model. This model combines the precipitation phase volume fraction (VF) and average radius (R) as input variables, and the yield strength as the output variable, and improves the performance of the weak regressor through weighted averaging. During the training process, the ABR model learns the relationship between the alloy composition and the yield strength so as to be able to predict the yield strength of the alloy.

[0080] S3.2: After model training, the hyperparameters of the ABR model are optimized using the grid search method to obtain the model with the best performance on the test set. During the optimization process, the prediction accuracy of the model is improved by adjusting hyperparameters such as the learning rate and the number of weak regressors of the model. The strength prediction results of the model are as Figure 2 shown.

[0081] S4: Genetic algorithm optimization design

[0082] S4.1: In the alloy composition optimization stage, first set the types and composition ranges of each element in the alloy, and initialize the binary population. Each individual represents an alloy composition. The alloy composition optimization design range is shown in Table 2. By decoding, each binary individual is converted into the corresponding alloy composition. For each composition, the precipitation phase volume fraction (VF) and average radius (R) are calculated using the thermal / kinetic calculation software Thermo-Calc.

[0083] Table 2 Alloy composition optimization design range

[0084] Name Minimum Quantity Maximum Quantity Chromium (wt.%) 18 23 Molybdenum (wt.%) 3 10 Titanium (wt.%) 0 3 Aluminum (wt.%) 0 1 Niobium (wt.%) 0 7.5 Iron (wt.%) 10 20 Copper (wt.%) 0 3

[0085] S4.2: Use the trained ABR model to predict the yield strength according to the VF and R values of each alloy composition. The strength values of all compositions are used as the input of the fitness function into the genetic algorithm for fitness ranking. The alloy compositions in the top 5% are retained for the next generation, and other compositions are eliminated. The corrosion resistance is evaluated by the Cr content in the matrix phase and the pitting resistance equivalent number (PREN) value. Among them, the Cr mass fraction exceeds 18%, and PREN ≥ 32. The pitting resistance equivalent number calculation formula is PREN = %Cr + 3.3 * %Mo + 16 * %N.

[0086] S4.3: During the evolution process of the genetic algorithm, new offspring alloy compositions are generated through selection, crossover, and mutation operations. This process is repeated until the preset maximum number of iterations is reached or the final fitness change is less than 0.1%, at which point the genetic algorithm terminates. Finally, the optimal alloy composition design is obtained. The genetic parameters set in this embodiment are as follows:

[0087] The number of iterations is 100 generations; the mutation rate is set to 0.1; the crossover rate is 0.6.

[0088] S5: Optimization results and experimental verification

[0089] Through the implementation of this embodiment, some nickel-based corrosion-resistant alloy compositions with excellent yield strength are successfully designed. Select the best-performing data group in the last generation as the verification alloy Alloy1, and output its alloy composition as shown in Table 3.

[0090] Table 3 Optimized alloy composition Alloy1 (mass percentage)

[0091] Cr Mo Fe Al Ti Nb Cu Alloy 1 20.94 3.44 12.22 0.55 0.93 6.24 0.07

[0092] Melt Alloy1. For better comparability, three systems are adopted for the solution temperature, and the double-stage aging heat treatment system of typical nickel-based corrosion-resistant alloys in the literature is used for the aging system. The experimental verification results are shown in Table 4. The predicted yield strength of Alloy1 is 1355 MPa. The actually measured yield strength of the alloy is 1370 MPa, exceeding the yield strength of the alloys in the dataset. This indicates that the composition design method of this nickel-based corrosion-resistant alloy can accurately and effectively design materials.

[0093] Table 4 Strength and elongation of Alloy1

[0094]

[0095] In summary, the technical solution of the present invention provides a composition design method for ultra-high strength and toughness nickel-based corrosion-resistant alloys based on machine learning. Calculate the volume fraction VF and average radius R of the precipitated phases of known alloy samples through thermal / kinetic software, and use these as inputs. Construct a prediction model of alloy strength with the volume fraction VF of the precipitated phases and the average radius R through a machine learning algorithm; select the strength prediction model established by the algorithm with the highest accuracy as the subsequent genetic algorithm composition optimization strength fitness screening function, and use the Cr content in the matrix phase and the pitting corrosion resistance equivalent PREN value as the corrosion resistance evaluation function. Optimize and iterate through the genetic algorithm until the preset maximum number of iterations is reached or the convergence condition is met to obtain the optimized alloy composition. The present invention establishes an association model among composition, precipitated phases and strength through machine learning methods, and then uses the strength prediction model as the fitness function to apply the genetic optimization algorithm to efficiently optimize the composition within the preset composition range.

[0096] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, the element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0097] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0098] Through the description of the above embodiments, those skilled in the art can clearly understand that the above implementation methods can be realized by means of software plus a necessary general hardware platform. Of course, it can also be realized by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0099] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. All of these are within the protection scope of the present invention.

Claims

1. A method for designing the composition of ultra-high strength and toughness nickel-based corrosion-resistant alloy based on machine learning, characterized in that: The method comprises the following steps: S1: Define the goal of alloy composition optimization, including multiple sets of alloy compositions; S2: Based on the alloy strength, precipitate phase volume fraction VF and average radius R of the known alloy samples, an alloy strength prediction model is constructed through a machine learning algorithm and trained and tested to obtain a trained strength prediction model; S3: performing thermodynamic calculations on multiple sets of alloy components in the target of the alloy component optimization to obtain the volume fraction VF and the average radius R of the precipitated phase; S4: input the calculated precipitate phase volume fraction VF and average radius R into the trained strength prediction model to obtain the strength value of each group of alloy components, and perform fitness screening and sorting; S5: retaining the alloy components with higher fitness, and for the remaining alloy components, performing crossover and mutation operations through a genetic algorithm, and then performing thermodynamic calculations again with the alloy components with higher fitness to obtain the corresponding precipitate phase volume fraction VF and average radius R; S6: Repeat S4-S5 until the preset maximum number of iterations is reached or the convergence condition is met, and select the alloy composition with the best fitness as the ultra-high strength and toughness nickel-based corrosion-resistant alloy composition.

2. The method according to claim 1, characterized in that The S1 specifically includes: initializing a binary population, each individual represents a set of alloy components, and each binary individual is converted into a corresponding alloy component by decoding; The size of the population is set according to the complexity of the problem and computing power.

3. The method according to claim 1, characterized in that The S2 specifically includes: S2.1: Collect the composition, heat treatment process and corresponding yield strength data of known nickel-based corrosion-resistant alloys to form an initial data set; S2.2: performing standardization processing on the initial data set to obtain a standard data set, and calculating the volume fraction VF and average radius R of the precipitated phase of each alloy by thermal / kinetic calculation software, and combining the volume fraction VF and average radius R of the precipitated phase with the standard data set to form a final data set; S2.3: Divide the final data set into a training set and a test set in a ratio of n:1 by multiple holdout method, where n is a positive integer; S2.4: Use a machine learning algorithm to build an adaptive enhanced regression model, take the precipitated phase volume fraction VF and the average radius R as input variables, and take the yield strength as the output variable, train and test the ABR model, and obtain the trained ABR model as a strength prediction model.

4. The method according to claim 3, characterized in that In S2.1, the data in the initial data set are collected in a manner including experimental data, literature data, and database query; If there are missing parts in the initial data set, data integrity is repaired by filling the missing values; wherein the method of filling the missing values ​​includes mean filling, nearest neighbor method and other suitable interpolation methods; Each set of data in the initial data set includes a composition, a heat treatment process and a corresponding yield strength of a nickel-based corrosion-resistant alloy.

5. The method according to claim 3, characterized in that: In S2.2, the formula for standardization is: z = (x-μ) / σ, where z is the standardized data, x is the original data in the initial data set, μ is the mean of each dimensional variable in the original data, and σ is the standard deviation of each dimensional variable in the original data; Wherein, in said S2.2, said thermal / kinetic calculation software is Thermo-Calc.

6. The method according to claim 3, characterized in that: In S2.4, during the training and testing of the ABR model, a grid search or random search method is used to adjust the hyperparameters of the ABR model to optimize the model hyperparameters.

7. The method according to claim 1, characterized in that In S3, the volume fraction of the precipitated phase, the average radius and the content of Cr in the matrix phase of each alloy are obtained using the thermal / dynamic calculation software Thermo-Calc, so as to perform thermodynamic calculations on multiple sets of alloy components in the goal of optimizing the alloy components.

8. The method according to claim 1, characterized in that In S4, the fitness screening further includes: The corrosion resistance is evaluated based on the Cr content in the matrix phase and the pitting resistance equivalent PREN value, where the mass fraction of Cr exceeds 18%, PREN ≥ 32, and the pitting resistance equivalent calculation formula is PREN = %Cr + 3.3*%Mo + 16*%N.

9. The method according to claim 1, characterized in that: In S5, the fitness is sorted from high to low, and the top 5% is regarded as the alloy components with higher fitness.

10. The method according to claim 1, characterized in that In S6, the genetic parameters are: the number of iterations is 100 generations; the mutation rate is set to 0.1; and the crossover rate is 0.6.

Images