Design method of high-entropy alloy components with multi-performance collaborative optimization based on machine learning

Abstract

The present invention discloses a method for designing high entropy alloy components with multi-performance collaborative optimization based on machine learning. In this method, a high entropy alloy component-performance initial data set is created, data cleaning is performed, the cleaned data is standardized, a high entropy alloy performance training set is obtained, a base model for predicting each performance of the high entropy alloy is determined by the high entropy alloy performance training set, an expected improvement (EI) value of each performance of the high entropy alloy in the component search space is determined, an alloy performance EI value Pareto front is determined in the component search space, a cluster analysis is performed to obtain the cluster center of the EI value Pareto front, a high entropy alloy and its related component information corresponding to the EI value are determined, the high entropy alloy is manufactured, and a characteristic evaluation test is performed for each target performance of the manufactured high entropy alloy, and if the test data does not meet the requirements, the high entropy alloy component-performance initial data set is updated according to the test data. The present invention can realize an efficient component design of the multi-performance collaborative optimization of high entropy alloy.

Description

[Technical field]

[0001] The present invention belongs to the field of metallic material design technology, and specifically relates to a design method for multi-performance collaborative optimization high-entropy alloy components based on machine learning. [Background technology]

[0002] With the rapid development of economy and society, the requirements for the performance of metallic materials are constantly improving. Traditional metal alloy systems based on single main element design are constantly improving, performance development tends to be saturated, and there are bottlenecks in the development, design and application of high-performance metallic materials. Taiwanese scholar Ye Junwei proposed the concept of multi-element alloy composition design in 2004. High entropy alloys defined in this way provide a rich search space for the design of metallic materials, greatly improving the freedom of alloy composition selection and performance adjustment. Research in recent years has shown that high entropy alloys are superior to traditional alloy materials in many aspects such as high temperature strength, low temperature toughness, thermal stability and corrosion resistance, and have a wide range of application prospects.

[0003] However, the composition system of high-entropy alloys is complex and relies on traditional trial-and-error material design methods, making it difficult to meet the rapid and accurate composition design requirements. Meanwhile, the research and development of high-entropy alloys for engineering applications usually requires synergistic optimization of multiple properties. Alloy design faces the problem that all performances cannot be fully optimized, resulting in low efficiency, long cycles, and high costs in the research and development of new high-performance alloy materials. Therefore, how to accelerate the search for target materials in a wide composition space and develop efficient design methods for high-performance alloy materials has become a technical challenge that requires urgent breakthroughs in the field of metal alloys and even the entire field of material design.

[0004] In recent years, the rise of material genetic engineering has created a new paradigm in materials research and development, and data technologies such as machine learning are playing an increasingly important role in material design. Machine learning methods can establish implicit relationships between material composition, process, etc. and target performance based on existing material data information, and can rapidly and accurately predict material properties by digging up hidden material laws and knowledge based on existing material data information, accelerating the design of material composition. At present, it has been reported that machine learning methods can be used to guide the composition and performance design of high-entropy alloys.

[0005] For example, Chinese Patent CN113870957A discloses a method and device for designing eutectic high entropy alloy components based on machine learning, which predicts the important elements and the elements closely related to them that have a great influence on the formation of eutectic by training a machine learning model, and adjusts the element contents based on the prediction of the eutectic high entropy alloy to obtain the components of the eutectic high entropy alloy.

[0006] Chinese patent CN114678086A discloses a machine learning method for low activation high entropy alloy design, which predicts the composition of Fe-Cr-VW-Mn based high entropy alloys with low activation properties and BCC phase structure by training classification and regression models.

[0007] In order to solve the time-consuming, labor-intensive and inaccurate problems existing in the conventional methods for predicting and searching for high-entropy alloys with excellent hardness properties, Chinese patent CN112216356A discloses a machine learning based method for predicting the hardness of high-entropy alloys, and Chinese patent CN115061435A discloses a machine learning method for quickly predicting the hardness of high-entropy alloys and its manufacturing process.

[0008] Chinese patent CN114464274A discloses a method for predicting the hardness of high entropy alloys by improving genetic algorithm feature screening based on machine learning. By training a machine learning model, improving the genetic algorithm, and screening feature combinations that are highly related to the hardness of high entropy alloys, the accuracy of hardness prediction is improved, and the design of high entropy alloy components according to hardness requirements is guided.

[0009] However, the machine learning methods reported so far are usually designed for a specific high-entropy alloy system in a limited alloy space, making it difficult to quickly search for the desired alloy in a high-dimensional complex composition space. Meanwhile, the reported methods are mainly aimed at material composition design for the optimization of single-objective performance, and cannot solve the problem of high-entropy alloy composition design under the requirement of joint optimization design of multi-objective performance.

[0010] In view of the above, in order to address the problem of material design that requires the collaborative optimization of a huge space of high-entropy alloy components and multiple performances, it is necessary to design a new design method for high-entropy alloy components that is data-driven and addresses multiple performance requirements using machine learning technology based on the concept of materials genetic engineering. Summary of the Invention

[0011] In view of the above, the main objective of the present invention is to provide a method for designing multi-performance collaborative optimization high-entropy alloy components based on machine learning.

[0012] In order to achieve the above object, the technical means of the present invention are as follows.

[0013] An embodiment of the present invention provides a machine learning based multi-performance collaborative optimization method for designing high entropy alloy components, the method comprising: creating an initial high entropy alloy composition-performance data set based on historical data information of the high entropy alloy; Performing data cleaning for each performance parameter of the high entropy alloy, and standardizing the cleaned data to obtain a high entropy alloy performance training set; Using the high entropy alloy performance training set, a number of machine learning models are trained for each target performance of the high entropy alloy, and then screening and determining a base model for predicting each performance of the high entropy alloy; Determining the expected improvement (EI) value of each performance of the high entropy alloy obtained by searching in the component space using the base model; determining a search space of the target component of the high entropy alloy in the high entropy alloy performance training set to be optimized, and performing genetic optimization search using the EI value of each performance of the high entropy alloy as the target, and determining the Pareto front of the alloy performance EI value in the search component space; Performing cluster analysis on the Pareto front of the EI value and screening to obtain high entropy alloys corresponding to the EI value and their related component information; A step of producing a high entropy alloy corresponding to the EI value related composition information by a vacuum arc melting furnace melting method; Conducting a characteristic evaluation test for each target performance of the manufactured high entropy alloy, and updating the initial data set of high entropy alloy composition-performance based on the test data if the test data does not meet the requirements; Includes.

[0014] In the above embodiment, in the step of creating an initial data set of high entropy alloy components-performance based on the historical data information of the high entropy alloy, specifically, a literature survey is conducted on the refractory high entropy alloy to create an initial data set of high entropy alloy components-performance, which includes 10 common elements, namely Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W, and includes alloy components, high temperature yield strength, and room temperature compressive plasticity performance.

[0015] In the above embodiment, data cleaning is performed on each performance parameter of the high entropy alloy, specifically, outliers and missing values ​​of the reported alloy performance are deleted, and the relative error of each reported value is determined for performance data of the same component reported by different documents, and the duplicate values ​​whose relative error is smaller than the error threshold are averaged, and the duplicate values ​​whose relative error is larger than the error threshold are deleted to obtain the final alloy performance value. The alloy performance includes high temperature yield strength and room temperature plasticity. The cleaned alloy performance value is normalized to determine a high temperature yield strength training data set including a plurality of samples and a room temperature plasticity training data set including a plurality of samples.

[0016] In the above embodiment, in the step of respectively training a plurality of machine learning models for each target performance of the high entropy alloy using the high entropy alloy performance training set, and sequentially screening and determining a base model for predicting each performance of the high entropy alloy, specifically, the high temperature yield strength training data set and the room temperature plasticity training data set are evenly divided into N groups, and the first group of data is used to train and fit the machine learning model, and the remaining N-1 groups of training data are predicted to obtain the performance prediction value thereof, and a first root mean square error between the actual value and the predicted value is determined, and the same prediction operation is sequentially performed on the second group to the Nth group of data to obtain the predicted second root mean square error, and the cross-validation error of each machine learning model is determined according to the first root mean square error and the second root mean square error, and the machine learning model with the smallest cross-validation error is the base model for predicting the high temperature strength and room temperature plasticity performance of the alloy, respectively.

[0017] In the above embodiment, in the step of obtaining the expected improvement (EI) value of each target performance of the high-entropy alloy obtained by searching in the component space using the base model, specifically, the alloy performance training set is resampled by a bootstrap method to obtain M resampling data sets, the number of samples contained in each resampling data set is the same as the number of training sets, and a base model is trained using each resampling data set to obtain M performance prediction base models, and the M performance prediction base models are used to perform performance prediction on the high-entropy alloys obtained by searching in the component space using the M performance prediction base models to obtain M predicted values, and the average value μ and variance σ of the performance predicted values ​​are determined by the formula:

number

[0018] In the above embodiment, in the step of determining the search space of high entropy alloy target components in the high entropy alloy performance training set to be optimized, specifically, a search space of high entropy target components consisting of elements of Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W is determined according to the types of alloy constituent elements in the alloy performance training set, and the search alloy is limited to contain 3-6 types of elements, and the molar content of each element in the alloy is not more than 35% and not less than 5%, and the variable step size of the content is 1%.

[0019] In the above embodiment, in the step of performing a genetic optimization search with the EI value of each performance of the high entropy alloy as the target, and determining the Pareto front of the EI value in the search component space, specifically, in the search space of the target component of the high entropy alloy, each time a genetic search is performed, a Pareto front of one alloy EI value is obtained. The genetic search is performed N times to obtain N Pareto fronts. Non-dominated sorting is performed on the N Pareto front alloys, and the Pareto front of the EI value of the alloy performance to be genetically searched is finally determined.

[0020] In the above embodiment, in the step of performing cluster analysis according to the Pareto front of the alloy performance EI value and screening to obtain high-entropy alloys corresponding to the performance EI value and their related component information, specifically, the Pareto front of the alloy performance EI value is cluster analyzed by the K-means clustering method, and the optimal number of cluster centers is determined by the elbow method. This number is used to perform clustering and screening on the Pareto front of the alloy performance EI value to obtain alloys corresponding to the EI value and their related component information.

[0021] In the above embodiment, the method further includes the step of finalizing the design if the test data meets the requirements, resulting in a multi-performance co-optimized high entropy alloy.

[0022] In the above embodiment, the method further comprises the step of designing each high entropy alloy in the high entropy alloy composition-performance initial dataset until the multiple performances of the designed alloys simultaneously meet the optimized and improved requirements or reach the budget upper limit.

[0023] Compared with the prior art, the composition design method of the present invention can realize a fast search for a target alloy in a high-dimensional composition space, and can simultaneously improve the multi-performance of a high-entropy alloy. [Brief description of the drawings]

[0024] The drawings described herein are intended to provide a better understanding of the present invention and are a part of the present invention. The schematic examples of the present invention and the description thereof are intended to illustrate the present invention and are not intended to limit the present invention.

[0025] [Figure 1] FIG. 1 is a logical framework diagram of the present invention. [Figure 2a] FIG. 2 is a model performance evaluation chart for predicting high temperature strength performance of the alloy of the present invention. [Figure 2b] FIG. 2 is a model performance evaluation diagram for predicting room temperature plasticity performance of the alloy of the present invention. [Diagram 3] FIG. 1 is a comparison diagram between the multi-performance collaborative optimization of the alloy designed in the present invention and the performance of existing alloys. [Figure 4] 1 is a room temperature mechanical performance test curve of the alloy designed in the present invention. [Diagram 5] 1 is a high temperature mechanical performance test curve of the alloy designed in the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0026] In order to make the objectives, technical means and advantages of the present invention clearer, the present invention will be described in more detail with reference to the drawings and examples below. The following specific examples are only for illustrating the present invention, and do not limit the present invention.

[0027] As used herein, the terms "comprise," "contain," and other variations thereof are intended to imply a non-exclusive inclusion, i.e., a process, article, or device comprising a set of elements is intended to include not only those elements, but also other elements not expressly listed or inherent in the process, article, or device. In the absence of further limitations, an element qualified with the term "comprising a..." does not exclude other elements that are included in the process, article, or device that includes the element.

[0028] An embodiment of the present invention provides a machine learning-based multi-performance collaborative optimization method for designing high entropy alloy components. As shown in Figure 1, the method includes the following steps:

[0029] S1: Create an initial data set of high entropy alloy composition-performance data based on historical data information of high entropy alloys.

[0030] Specifically, a literature survey was conducted on refractory high-entropy alloys, and an initial data set of high-entropy alloy composition-performance data was created that included 10 common elements, namely Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W, including alloy composition, high-temperature yield strength, and room-temperature compressive plasticity performance.

[0031] S2: Perform data cleaning for each performance parameter of the high entropy alloy, and standardize the cleaned data to obtain a high entropy alloy performance training set.

[0032] Specifically, outliers and missing values ​​of reported alloy performance are removed, a relative error of each reported value is determined for performance data of the same component reported by different documents, duplicate values ​​whose relative error is smaller than an error threshold are averaged, and duplicate values ​​whose relative error is larger than an error threshold are removed to obtain a final alloy performance value. The alloy performance values ​​include high temperature yield strength and room temperature plasticity. The cleaned alloy performance values ​​are normalized to determine a high temperature yield strength training dataset including a plurality of samples and a room temperature plasticity training dataset including a plurality of samples.

[0033] In some embodiments, the error threshold is 10%.

[0034] In some embodiments, the high temperature yield strength training data set includes 140 samples and the room temperature plasticity training data set includes 50 samples.

[0035] S3: Using the high-entropy alloy performance training set, train multiple machine learning models for each target performance of the high-entropy alloy respectively, and screen and determine the base model for predicting each performance of the high-entropy alloy in turn.

[0036] Specifically, the model input is the content of each constituent element of the high entropy alloy, and the model output is each target performance of the high entropy alloy.

[0037] A number of machine learning models are trained for the high-temperature yield strength training data set and the room-temperature plasticity training data set, respectively, the contents of each constituent element of the high-entropy alloy are used as inputs of the machine learning models, the high-temperature strength and room-temperature plasticity performance are used as outputs of the machine learning models, and the mean square error index parameters of each machine learning model are obtained.

[0038] The data for the high temperature yield strength training data set and the room temperature plasticity training data set are evenly divided into N groups. The first group of data is used to train and fit a machine learning model, and the remaining N-1 groups of training data are predicted to obtain their performance prediction values, and a first root mean square error between the actual value and the predicted value is determined, and the same prediction operation is then performed on the second group to the Nth group of data in sequence to obtain a second predicted root mean square error.

[0039] In some embodiments, N is 10, that is, the data for the high temperature yield strength training data set and the room temperature plasticity training data set are evenly divided into 10 groups. The first group of data is used to train and fit the machine learning model, and the remaining 9 groups of training data are predicted to obtain their performance prediction values, and the root mean square error between the actual value and the predicted value is determined. The same prediction operation is then performed on the second group to the tenth group of data in sequence to obtain the predicted root mean square error, and finally the cross-validation error of the machine learning model is obtained. The model with the smallest error is the prediction base model for the alloy high temperature strength and room temperature plasticity performance. Figures 2a and 2b show performance evaluation diagrams of the alloy high temperature strength and room temperature plasticity performance prediction models, respectively, where the high temperature strength prediction base model is the SVR model and the room temperature plasticity prediction base model is the GR model.

[0040] The machine learning model employs at least one of the following five models: support vector machine (SVR), K nearest neighbor (KNN), Gaussian regression (GR), neural network (NN), and kernel ridge regression (KRR).

[0041] S4: Determine the expected improvement (EI) value of each performance of the high entropy alloy obtained by searching within the component space based on the base model.

[0042] Specifically, the alloy performance training set is resampled using a bootstrap method to obtain M groups of resampled data sets. The number of samples included in each group of resampled data sets is the same as the number of training sets. A base model is trained using each group of resampled data sets to obtain M performance prediction base models. Performance prediction is performed on the high-entropy alloys obtained by searching in the component space using the M performance prediction base models, M predicted values ​​are obtained, and the mean μ and variance σ of the performance prediction values ​​are determined, and the formula:

number

[0043] In some embodiments, the training set is resampled by the bootstrap method to obtain 1000 groups of resampled data sets, the number of samples in each group of resampled data sets is the same as the number of training sets, and a base model is trained using each group of resampled data sets to obtain 1000 performance prediction base models, and the 1000 base models are used to perform performance prediction for alloys in the component search space to obtain 1000 predicted values. If it is assumed that the prediction follows a normal distribution, the mean value μ and variance σ of the predicted values ​​of each performance of the alloy are calculated. Formula:

number

[0044] S5: Determine the search space of the target components of the high entropy alloy to be optimized, and set the EI value of each performance of the high entropy alloy as the target. Perform genetic optimization search to determine the Pareto front of the alloy performance EI value in the search component space.

[0045] Specifically, according to the types of alloy constituent elements in the alloy performance training set, a high-dimensional component search space consisting of elements of Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W is determined, and the search alloy is limited to contain 3-6 elements, and the molar content of each element in the alloy is 35% or less and 5% or more, and the variable step size of the content is 1%.

[0046] In the search space of the target components of the high entropy alloy, a Pareto front of the EI value of one alloy is obtained each time a genetic search is performed. The genetic search is performed N times to obtain N Pareto fronts. Non-dominated sorting is performed on all alloys related to the N Pareto fronts, and the Pareto front of the EI value of the alloy performance to be genetically searched is finally determined.

[0047] In some embodiments, the target component search is performed by NSGA-II non-dominated sorting genetic algorithm, the initial population number is selected as 500, the genetic evolution algebra is selected as 20, the population crossover rate is selected as 0.8, and the population mutation rate is selected as 0.02. The genetic optimization search is performed 100 times by randomizing the initial population, and the Pareto front of the EI values ​​of the high temperature yield strength and room temperature plasticity performance of 100 alloys is obtained, and non-dominated sorting is performed for all alloys related to the 100 Pareto fronts, and finally the Pareto front of the EI values ​​of the high entropy alloy performance in the component space is determined.

[0048] S6: Perform cluster analysis according to the Pareto front of the alloy performance EI value, and screen to obtain high entropy alloys corresponding to the EI value and their related component information.

[0049] Specifically, the Pareto front of the alloy EI values ​​is cluster-analyzed using the K-means clustering method, and the optimal number of cluster centers is determined using the elbow method. This number is then used to perform clustering and screening on the Pareto front of the alloy performance EI values, to obtain the alloys corresponding to the EI values ​​and their related component information.

[0050] Before cluster analysis, the target performance EI values ​​on the Pareto front are logarithmically transformed and then normalized, and then the change in the within-group sum of squares (i.e., the sum of squares of the distances from each sample point to its cluster center) with the change in the number of cluster centers is calculated. If the within-group sum of squares drops very slowly with the change in the number of cluster centers, an elbow in the number of clusters appears. The number of cluster centers K corresponding to the elbow is the number of cluster centers actually adopted. Clustering and screening are performed on the Pareto front of the performance EI values ​​with the K value to obtain the alloys and their related component information corresponding to the EI values.

[0051] S7: Produce a high entropy alloy corresponding to the EI value-related component information by a vacuum arc melting furnace melting method.

[0052] Specifically, the steps for synthesizing and manufacturing the high entropy alloy are as follows: Step 1: Use metals with a purity of more than 99.9% as raw materials for alloy production, polish the surface of the metal raw material to remove the surface oxide film, ultrasonically clean it, and then dry it in a drying box. Step 2: According to the alloy components to be screened, calculate the mass of the required metal raw material based on the molar content of its constituent elements, and then weigh the metal raw material ultrasonically cleaned in step 1 and store it until use. Step 3: The alloy is smelted using a non-consumable vacuum arc furnace. The metal raw materials are placed in the melting tank in the furnace in order from low to high melting point, and the sample chamber is evacuated and protected by introducing high-purity argon gas. Step 4: Turn on the welding machine power of the non-consumable vacuum arc furnace, generate an arc, and then melt the titanium sponge placed in the combustion chamber in advance to deoxidize the furnace. Step 5: The alloy is repeatedly smelted 5-8 times, and after each smelting, the alloy is cooled and turned over to obtain a high-entropy alloy ingot.

[0053] S8: Conduct characteristic evaluation tests for each target performance of the manufactured high entropy alloy.

[0054] S9: Evaluate whether the performance of the designed material meets the requirements. If it does, terminate the design and obtain a high-entropy alloy with multi-performance co-optimization. If it does not meet the requirements, execute S10.

[0055] S10: Feed back new alloy composition and performance information into the high entropy alloy composition-performance initial data set. Repeat S4-S9 to execute the closed-loop path of multi-performance collaborative optimization high-entropy alloy component design until the designed alloy meets the requirements of multi-performance synchronous optimization and improvement or reaches the budget upper limit.

[0056] Example 1 An embodiment of the present invention provides a machine learning-based multi-performance collaborative optimization method for designing high entropy alloy components, as shown in Figure 1, including the following steps:

[0057] S1: A literature survey was conducted on refractory high-entropy alloys, and a data set of refractory high-entropy alloys containing 10 common elements, namely Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W, was created, including alloy components, 1000°C high-temperature yield strength, and room temperature compressive plasticity performance.

[0058] S2: Outliers and missing values ​​of reported alloy performance were removed, the relative error of each reported value was calculated against performance data reported by different literatures for the same components, duplicate values ​​with an error of less than 10% were averaged to obtain the final alloy performance value, and duplicate values ​​with an error of more than 10% were removed. The cleaned alloy performance values ​​were normalized to determine a training dataset for high temperature yield strength containing 140 alloy samples and a training dataset for room temperature alloy plasticity containing 50 samples.

[0059] S3: Five machine learning models, including support vector machine (SVR), K-nearest neighbor (KNN), Gaussian regression (GR), neural network (NN) and kernel ridge regression (KRR), were trained on the high-temperature strength performance training set and the room-temperature plasticity performance training set. The content of each constituent element of the high-entropy alloy was used as the model input, and the high-temperature strength and room-temperature plasticity were used as the model output. A grid search method was used to optimize each model parameter using the mean squared error index parameter, and the determined optimal parameter model was used to train the training set.

[0060] In addition, the data for each of the two training sets was divided equally into 10 groups, and the following model evaluation operations were performed independently. The first group of data was used to fit the machine learning model, and the remaining 9 groups of training data were predicted to obtain the performance prediction values, and the root mean square error between the actual value and the predicted value was calculated. The same prediction operation was then performed on the second to tenth groups of data in sequence to obtain the predicted root mean square error, and finally the cross-validation error of the machine learning model was obtained. The model with the smallest error was the prediction base model for alloy high temperature strength and room temperature plasticity performance. Figures 2a and 2b show the performance evaluation diagrams of the alloy high temperature strength and room temperature plasticity performance prediction models, respectively, in which the high temperature strength prediction base model was selected as the SVR model and the room temperature plasticity prediction base model was selected as the GR model.

[0061] S4: For the target performance of the alloy's high temperature strength and room temperature plasticity, the expected improvement (EI) value of each target performance was calculated using the following operations. The training set was resampled using the bootstrap method to obtain 1000 groups of resampled data sets, and the number of samples contained in each group of resampled data sets was the same as the number of training sets. A base model was trained using each group of resampled data sets to obtain 1000 performance prediction base models, and performance predictions were made for alloys in the component search space using these 1000 base models to obtain 1000 predicted values. If it is assumed that this prediction follows a normal distribution, the mean μ and variance σ of the predicted values ​​of each target performance were calculated. Formula:

number

[0062] S5: According to the types of alloy constituent elements in the training set, a high-dimensional component search space consisting of 10 elements, namely Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W, was determined, and the search alloy was restricted to contain 3-6 elements, the molar content of each element in the alloy was 35% or less and 5% or more, and the variable step size of the content was 1%.

[0063] According to the calculation method of the alloy performance EI value determined in S4, the target component search was performed by the NSGA-II non-dominated sorting genetic algorithm, the initial population number was selected as 500, the genetic evolution algebra was selected as 20, the population crossover rate was selected as 0.8, and the population mutation rate was selected as 0.02, and the genetic optimization search was performed 100 times by randomizing the initial population to obtain the Pareto front of the EI values ​​of the high temperature strength and room temperature plasticity performance of 100 alloys, and non-dominated sorting was performed on all alloys related to the 100 Pareto fronts to determine the Pareto front of the EI value of high entropy alloy performance in the component space.

[0064] S6: The Pareto front of the EI values ​​of the high entropy alloy performance determined in S5 was analyzed by the K-means clustering method, and the number of cluster centers was determined by the "elbow method". Specifically, the target performance EI values ​​on the Pareto front surface were logarithmically transformed, then normalized, and then the change in the sum of squares within the group (i.e., the sum of squares of the distances from each sample point to its cluster center) with the change in the number of cluster centers was calculated. If the drop in the sum of squares within the group was very slow with the change in the number of cluster centers, an elbow in the number of clusters appeared. The number of cluster centers K corresponding to the elbow was the number of cluster centers actually adopted. Clustering and screening were performed on the Pareto front of the performance EI values ​​with the K value, and the alloys corresponding to the EI values ​​and their related component information were obtained.

[0065] S7: The target alloy screened in S6 was synthesized and manufactured by the following steps: Step 1: Metals with a purity of more than 99.9% were used as raw materials for alloy production, and the surfaces of the raw metal materials were polished to remove the oxide film on the surfaces, ultrasonically cleaned, and then dried in a drying box. Step 2: According to the alloy components to be screened, the mass of the required metal raw materials was calculated based on the molar contents of their constituent elements, respectively, and the metal raw materials ultrasonically cleaned in step 1 were weighed and stored until use. Step 3: The alloy was smelted using a non-consumable vacuum arc furnace. The metal raw materials were placed in the melting tank in the furnace in order from low to high melting point, and the sample chamber was evacuated and protected by introducing high-purity argon gas. Step 4: Turn on the welding machine power of the non-consumable vacuum arc furnace, generate an arc, and then melt the titanium sponge that was placed in the combustion chamber in advance to deoxidize the furnace. Step 5: The alloy was repeatedly refined six times, and after each smelting, the alloy was cooled and turned over to obtain a high-entropy alloy ingot.

[0066] S8: The high entropy alloy produced in S7 was machined to obtain a test sample. Performance tests for 1000°C high temperature strength and room temperature plasticity were carried out respectively to obtain the composition and target performance data information of the newly produced alloy. The results were then fed back to the initial data set to perform iterative collaborative optimization of the two target performances.

[0067] According to the above implementation steps, the information of two kinds of refractory high entropy alloys designed and manufactured through three repeated experiments is as follows: Design alloy A-Mo 0.14 Nb 0.28 Ta 0.2 Hf 0.15 Zr 0.23 Design alloy B-Mo 0.2 Nb 0.26 Ta 0.19 Hf 0.16 Zr 0.19

[0068] FIG. 3 shows the high temperature strength and room temperature plasticity of the refractory high entropy alloy designed by the method of the present invention, which shows a significantly high multi-performance collaborative optimization effect compared with the conventional alloy performance. FIG. 4 shows the room temperature mechanical performance curves of two designed alloys A and B. FIG. 5 shows the high temperature mechanical performance curves of two designed alloys A and B. As can be seen from FIG. 4 and FIG. 5, the room temperature plasticity of alloy A is 35.7%, and the 1000°C high temperature yield strength is 894MPa, while the room temperature plasticity of alloy B is 30.5%, and the 1000°C high temperature yield strength is 974MPa, which shows the high efficiency of the multi-performance collaborative optimization design for the high entropy alloy of the present invention.

[0069] The above descriptions are only preferred embodiments of the present invention, and do not limit the protection scope of the present invention.

Claims

1. A machine learning-based multi-performance collaborative optimization method for designing high entropy alloy components, comprising: creating an initial high entropy alloy composition-performance data set based on historical data information of the high entropy alloy; Performing data cleaning for each performance parameter of the high entropy alloy, and standardizing the cleaned data to obtain a high entropy alloy performance training set; Using the high entropy alloy performance training set, a number of machine learning models are trained for each target performance of the high entropy alloy, and then screening and determining a base model for predicting each performance of the high entropy alloy; Determining the expected improvement (EI) value of each performance of the high entropy alloy obtained by searching in the component space using the base model; Determine the search space of the target components of the high entropy alloy in the training set of the high entropy alloy performance to be optimized, and perform genetic optimization search using the EI value of each performance of the high entropy alloy as the target, and determine the Pareto front of the alloy performance EI value in the search component space; Performing cluster analysis on the Pareto front of the EI value and screening to obtain high entropy alloys corresponding to the EI value and their related component information; producing a high entropy alloy corresponding to the EI value-related composition information by a vacuum arc melting furnace method; Conducting a characteristic evaluation test for each target performance of the manufactured high entropy alloy, and updating the initial data set of high entropy alloy composition-performance based on the test data if the test data does not meet the requirements; A method comprising:

2. The method of claim 1, characterized in that in the step of creating a high entropy alloy composition-performance initial data set based on the historical data information of the high entropy alloy, a literature survey is conducted on the refractory high entropy alloy to create a high entropy alloy composition-performance initial data set including 10 common elements, namely Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W, and including alloy composition, high temperature yield strength, and room temperature compressive plasticity performance.

3. The method according to claim 1 or 2, characterized in that: data cleaning is performed on each performance parameter of the high entropy alloy, specifically, outliers and missing values ​​of reported alloy performance are deleted, and relative errors of each reported value are determined for performance data reported by different documents for the same component, and duplicate values ​​whose relative errors are smaller than an error threshold are averaged, and duplicate values ​​whose relative errors are larger than an error threshold are deleted to obtain final alloy performance values, the alloy performances including high temperature yield strength and room temperature plasticity are normalized, and a high temperature yield strength training dataset including a plurality of samples and a room temperature plasticity training dataset including a plurality of samples are determined.

4. 4. The method according to claim 3, wherein in the step of respectively training a plurality of machine learning models for each target performance of the high entropy alloy using the high entropy alloy performance training set, and sequentially screening and determining a base model for predicting each performance of the high entropy alloy, the high temperature yield strength training data set and the room temperature plasticity training data set are evenly divided into N groups, a first group of data is used to train and fit a machine learning model, and the remaining N-1 groups of training data are predicted to obtain the performance prediction value thereof, a first root mean square error between the actual value and the predicted value is determined, and the same prediction operation is sequentially performed on the second group to the Nth group of data to obtain a second predicted root mean square error, and a cross-validation error of each machine learning model is determined based on the first root mean square error and the second root mean square error, and the machine learning model with the smallest cross-validation error is the base model for predicting the high temperature strength and room temperature plasticity performance of the alloy, respectively.

5. In the step of obtaining an expected improvement (EI) value of each target performance of the high-entropy alloy obtained by searching in the component space using the base model, specifically, the alloy performance training set is resampled by a bootstrap method to obtain M resampling data sets, the number of samples contained in each resampling data set is the same as the number of training sets, and a base model is trained using each resampling data set to obtain M performance prediction base models, and the performance prediction is performed on the high-entropy alloy obtained by searching in the component space using the M performance prediction base models to obtain M predicted values, and the average value μ and variance σ of the performance predicted values ​​are determined by the formula: [0010] Determine the expected improvement (EI) value of the alloy performance by: ＊ 5. The method of claim 4, wherein: denotes an optimum value of alloy performance in the training set; φ denotes a probability density function; and Φ denotes a probability distribution function.

6. The method according to claim 5, characterized in that, in the step of determining the search space of high entropy alloy target components in the high entropy alloy performance training set to be optimized, specifically, the search space of high entropy target components is determined to be composed of elements of Al, Ti, V, Zr, Cr, Nb, Mo, Hf, Ta, and W according to the types of alloy constituent elements in the alloy performance training set, and the search alloy is limited to contain 3-6 types of elements, the molar content of each element in the alloy is not more than 35% and not less than 5%, and the variable step size of the content is 1%.

7. The method according to claim 6, characterized in that in the step of performing a genetic optimization search with the EI value of each performance of a high entropy alloy as the target, and determining the Pareto front of the EI value in the search component space, specifically, in the search space of the target component of the high entropy alloy, a Pareto front of one alloy EI value is obtained each time a genetic search is performed, the genetic search is performed N times, N Pareto fronts are obtained, non-dominated sorting is performed on the N Pareto front alloys, and finally the Pareto front of the EI value of the alloy performance to be genetically searched is determined.

8. The method according to claim 7, characterized in that, in the step of performing cluster analysis according to the Pareto front of the alloy performance EI value and screening to obtain high entropy alloys corresponding to the performance EI value and their related component information, the Pareto front of the alloy performance EI value is cluster analyzed by K-means clustering method, an optimal number of cluster centers is determined by elbow method, and the optimal number is used to perform clustering and screening on the Pareto front of the alloy performance EI value to obtain alloys corresponding to the EI value and their related component information.

9. 10. The method of claim 8, further comprising the step of: terminating the design if the test data meets the requirements, resulting in a multi-performance co-optimized high entropy alloy.

10. 10. The method of claim 8, further comprising the step of designing each high entropy alloy in the high entropy alloy composition-performance initial data set until the multiple performances of the designed alloys simultaneously meet the optimized and improved requirements or reach the budget upper limit.

Images