A method and system for optimizing high-entropy alloy coatings based on machine learning

Abstract

The present invention provides a method and system for optimizing high-entropy alloy coatings based on machine learning, which relates to the technical field of data processing. The method includes: obtaining a variety of material descriptors; screening the material descriptors; using the coating process, composition, and the selected material descriptors as inputs, and the coating performance as the output, to establish a mapping relationship between the inputs and outputs based on different machine learning models; generating training data through a generative adversarial network; training each machine learning model separately with the training data, and selecting the machine learning model with the best performance as the final prediction model; using the prediction model as a surrogate model for multi-objective optimization, and determining the optimal combination parameters of the high-entropy alloy coating through multi-objective optimization. The present invention solves the problem of data imbalance by generating training data through a generative adversarial network, providing convenience for the design and research and development of high-entropy alloy coatings.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method and system for optimizing high-entropy alloy coatings based on machine learning. Background Art

[0002] High-entropy alloys have received extensive attention from researchers due to their high hardness, high fatigue resistance, high temperature oxidation resistance, etc. At present, they have been successfully applied in engineering fields such as aerospace, mechanical manufacturing, and biomedicine. However, due to their production cost being higher than that of most traditional alloys, to a certain extent, it restricts their engineering applications. Therefore, preparing high-performance high-entropy alloy coatings on low-cost metals takes into account people's dual requirements for material cost and performance. However, materials with high hardness are often prone to fracture. Therefore, designing high-performance high-entropy alloy coatings with both high hardness and high toughness to improve the comprehensive performance of materials is of great significance for material research and development and applications.

[0003] As a data-driven tool, machine learning has been applied to various fields of high-entropy alloy material design and development, including studying the formation of high-entropy alloy phases, thermodynamic and mechanical properties, predicting the energy of different atomic configurations, etc. However, its application in high-entropy alloy coatings is not extensive because there are complex interactions between the "process - composition - structure performance" of high-entropy alloy coatings, and there is currently no complete and standardized high-entropy alloy coating dataset that can be directly used for data-driven methods, which causes great difficulties for the design and development of high-entropy alloy coatings. Summary of the Invention

[0004] In order to solve the technical problem in the prior art that there is a lack of a complete and standardized high-entropy alloy coating dataset that can be directly used for data-driven methods, which causes great difficulties for the design and development of high-entropy alloy coatings, the present invention provides a method and system for optimizing high-entropy alloy coatings based on machine learning.

[0005] The technical solutions provided by the embodiments of the present invention are as follows:

[0006] First Aspect

[0007] A method for optimizing high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention includes:

[0008] S1: Obtain a variety of material descriptors;

[0009] S2: Screen the material descriptors;

[0010] S3: Use the process, composition of the coating, and the selected material descriptors as inputs, and the coating performance as the output, and establish a mapping relationship between the input and the output based on different machine learning models;

[0011] S4: Generate training data through a generative adversarial network;

[0012] S5: Train each of the machine learning models using the training data, and select the machine learning model with the best performance as the final prediction model;

[0013] S6: Use the prediction model as a surrogate model for multi-objective optimization, and determine the optimal combination parameters of the high-entropy alloy coating through multi-objective optimization.

[0014] Second aspect

[0015] An optimization system for high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention includes:

[0016] A processor;

[0017] A memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, the method for optimizing high-entropy alloy coatings based on machine learning as described in the first aspect is implemented.

[0018] Third aspect

[0019] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored. When the program is executed by a processor, the method for optimizing high-entropy alloy coatings based on machine learning as described in the first aspect is implemented.

[0020] The beneficial effects brought by the technical solutions provided by the embodiments of the present invention at least include:

[0021] (1) In the present invention, in the case of lacking a complete and standardized high-entropy alloy coating data set, training data is generated through a generative adversarial network, and then each machine learning model is trained using the training data. The machine learning model with the best performance is selected as the final prediction model to solve the data imbalance problem and facilitate the design and research and development of high-entropy alloy coatings.

[0022] (2) In the present invention, through machine learning and multi-objective optimization methods, the process, composition, and material descriptors of the coating can be comprehensively considered to find the optimal combination parameters of the high-entropy alloy coating, thereby significantly improving the overall performance of the coating. Description of the drawings

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 Schematic flowchart of a method for optimizing high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention;

[0025] Figure 2 Schematic structural diagram of a system for optimizing high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention. Detailed implementation manners

[0026] The technical solutions in the present invention will be described below with reference to the accompanying drawings.

[0027] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used for exemplification, illustration or explanation. Any embodiment or design solution described as "exemplary" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "exemplary" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0028] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0029] In the embodiments of the present invention, sometimes subscripts such as W 1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meanings they express are the same.

[0030] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0031] Referring to the accompanying drawings of the specification Figure 1 , a schematic flowchart of a method for optimizing high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention is shown.

[0032] Referring to the accompanying drawings of the specification Figure 2 , a schematic structural diagram of a system for optimizing high-entropy alloy coatings based on machine learning provided by an embodiment of the present invention is shown.

[0033] An embodiment of the present invention provides a method for optimizing a high-entropy alloy coating based on machine learning. This method can be implemented by a device for optimizing a high-entropy alloy coating based on machine learning, and this device for optimizing a high-entropy alloy coating based on machine learning can be a terminal or a server. The processing flow of the method for optimizing a high-entropy alloy coating based on machine learning can include the following steps:

[0034] S1: Obtain a variety of material descriptors.

[0035] It should be noted that by combining the atomic properties of elements, including atomic radius, first ionization energy, density, etc., we can obtain a set of material descriptors containing more information. These material descriptors may affect the hardness and modulus of the coating. Existing research has selected descriptors that have a significant impact on the target performance through feature selection methods, thereby improving the prediction ability of the model and revealing the internal relationship between the descriptors and the performance.

[0036] Optionally, calculate a variety of material descriptors based on the Xenonpy library.

[0037] In a possible implementation manner, the material descriptors specifically include:

[0038]

[0039] Among them, f ave,i represents the average descriptor of the i-th material, w k,i represents the weight of the k-th element of the i-th material, f k,i represents the physicochemical property value of the k-th element of the i-th material, m represents the total number of elements, f ave,i represents the variance descriptor of the i-th material.

[0040] Optionally, the weight is given by the atomic percentage of the material elements.

[0041] S2: Screen the material descriptors.

[0042] In a possible implementation manner, S2 specifically includes sub-steps S201 to S203:

[0043] S201: Calculate the Spearman correlation coefficient between each pair of material descriptors and the Spearman correlation coefficient between each material descriptor and the performance of the high-entropy alloy coating.

[0044] Optionally, the calculation method of the Spearman correlation coefficient is specifically:

[0045]

[0046] Among them, ρ represents the Spearman correlation coefficient, r(x i ) represents the rank of the i-th value in the material descriptor x, r(yi ) represents the rank of the i-th value in the material descriptor y, represents the average of the ranks of each value in the material descriptor x, represents the average of the ranks of each value in the material descriptor y, and n represents the total number of numerical values in the material descriptor.

[0047] It should be noted that by calculating the Spearman correlation coefficient between material descriptors, highly correlated descriptors can be identified and removed, thereby reducing redundancy. This helps to reduce the dimensionality of the feature space and the complexity of the model.

[0048] S202: When the Spearman correlation coefficient between material descriptors is greater than the correlation threshold, compare the magnitudes of the Spearman correlation coefficients between the two material descriptors and the properties of the high-entropy alloy coating, retain the larger one, and delete the smaller one.

[0049] S203: Input the retained material descriptors into the random forest model embedded with the recursive feature elimination algorithm, and screen the material descriptors through the random forest model embedded with the recursive feature elimination algorithm.

[0050] Among them, the random forest model embedded with the recursive feature elimination algorithm (Embedded Recursive Feature Elimination with Random Forests, RFE-RF) is an ensemble learning method for feature selection. This method combines recursive feature elimination (RFE) and random forest (Random Forest, RF), and can effectively improve the performance and robustness of the model during the feature selection process.

[0051] It should be noted that the random forest model embedded with the recursive feature elimination algorithm is already a mature existing technology, and the present invention will not elaborate on it.

[0052] In the present invention, by combining the Spearman correlation coefficient and the recursive feature elimination algorithm, a multi-level feature screening method is adopted, which helps to select the most useful features more comprehensively and meticulously. Among the descriptors with higher correlations, retain the features with higher correlations with the properties of the high-entropy alloy coating to ensure that the features used by the model are more meaningful for performance prediction. By screening and retaining the features with high correlations, the prediction accuracy and robustness of the model can be improved.

[0053] S3: Use the process, composition of the coating, and the selected material descriptors as inputs, and the coating properties as outputs, and establish a mapping relationship between the inputs and outputs based on different machine learning models.

[0054] Optionally, the machine learning model specifically includes an AdaBoost model, an RF model, a GBDT model, an ETR model, a BG model, and an XGBoost model. The AdaBoost model, RF model, GBDT model, ETR model, BG model, and XGBoost model are already mature existing technologies, and will not be elaborated in this invention.

[0055] Furthermore, based on the six models, selections are made respectively on the hardness dataset and the modulus dataset, and the selected models are used as performance prediction models. Both model selection and training are implemented based on Python.

[0056] S4: Generate training data through a generative adversarial network.

[0057] It should be noted that the experimental data of high-entropy alloy coatings may be scarce and expensive. Through GAN, a large amount of realistic data can be generated, enriching the dataset and improving the training effect.

[0058] It should be noted that to adapt to the characteristics of the current dataset, this paper proposes a two-dimensional deheterogeneity conditional generative adversarial network (2D-dhgCGAN) based on CGAN (conditional Generative Adversarial Network, cGAN) to solve the problem of small sample size in the current dataset.

[0059] Among them, cGAN is a data augmentation technique that has been widely used in the field of computer vision. cGAN consists of two multi-layer perceptrons, namely a generator G and a discriminator D, and they have a zero-sum game relationship, that is, the sum of the cost functions is zero.

[0060] Let the real data satisfy the probability distribution function P data(x) , and the generated data z satisfy the probability distribution function P z (z). Random samples are drawn from P data(x) and P z (z) to obtain x ∈ X and z ∈ Z. D is optimized to correctly identify the real data x and the generated data G(z), and is assigned 1 and 0 respectively. G is optimized to confuse D, that is, to approximate the real data distribution to the greatest extent. The trained G can be used to generate virtual samples to solve the problem of insufficient samples as a generative model. An additional space Y is introduced, and this space is used to provide external information of the training data. G is described as follows:

[0061] G: Z × Y → X

[0062] Similar to G, D can be described as:

[0063] D: X×Y→[0, 1].

[0064] In a possible implementation, the cost functions of the generator and discriminator in the generative adversarial network are specifically:

[0065]

[0066] where min represents minimization, G represents the generator, max represents maximization, D represents the discriminator, V represents the cost function, E represents the mathematical expectation operation, x represents the real data, P data(x) represents the real data distribution, log represents the logarithmic function, z represents the noise data, P z (z) represents the noise data distribution.

[0067] In the present invention, the data set contains two different types of feature classes, namely process features and composition features. Although both the process and composition are from the manual design of researchers, there is no inherent correlation between them. However, the mechanisms by which the process and composition affect performance are different. During the preparation process, the process directly affects the performance, while different components interact with each other due to their unique physical and chemical properties and indirectly affect the process. In order to eliminate the heterogeneity existing between the feature classes to generate virtual data that better conforms to the current data distribution, the cGAN algorithm is improved in this paper.

[0068] Compared with taking the entire feature as the input, we propose 2D-dhg CGAN based on the cGAN framework. Before inputting into cGAN, an independent neural network is first applied to different types of feature subsets, and the transformed features are concatenated as the input of cGAN. The features transformed by the neural network can, to a certain extent, eliminate the heterogeneity between the feature classes and better represent the original data.

[0069] In a possible implementation, S4 specifically includes:

[0070] S401: Select the first noise Z 1 and the second noise Z 2 .

[0071] S402: Input the first noise and the second noise into a transformer constructed based on a neural network to obtain the transformed first feature Z' 1 and the second feature Z' 2 .

[0072] It should be noted that by using an independent neural network to transform different types of feature subsets, the heterogeneity between features can be effectively eliminated, making the spliced features more uniform and consistent, thereby improving the quality of the generated data. The neural network can extract and transform important information in the feature subsets, optimize the feature representation, enabling the generator to better capture and simulate the distribution of real data when generating data, and improving the diversity of the data.

[0073] S403: Perform splicing processing on the first feature Z′ 1 and the second feature Z′ 2 to obtain a spliced feature.

[0074] S404: Use the spliced feature and the attribute p as the input z of the generator in the generative adversarial network to obtain the virtual feature G(z).

[0075] S405: Splice the attribute p with the real feature to construct real data, and splice the attribute p with the virtual feature G(z) to construct fake data.

[0076] S406: Use the real data and the fake data as the input of the discriminator in the generative adversarial network.

[0077] S407: Through adversarial games, train and optimize the generator and the discriminator in the generative adversarial network.

[0078] S408: Use the trained generator to generate training data.

[0079] In the present invention, by generating training data through the 2D-dhg CGAN framework based on cGAN, the data quality and diversity can be significantly improved, the problems of data imbalance and scarcity can be solved, the generalization ability and performance of the model can be enhanced, and the processing and analysis of complex data can be supported. This not only improves the efficiency and effectiveness of model training, but also provides strong data support and innovative ideas for the optimization of high-entropy alloy coatings.

[0080] Specifically, both the generator G and the discriminator D use four fully connected layers as hidden layers, where the transformers C 1 and C 2 respectively simulate the internal conversions of process features and composition features, and the output of the generator reflects the data distribution conditional on the attribute p. To improve the model stability and accelerate the model convergence, batch normalization is used between the hidden layers. In addition, to prevent overfitting and improve the generalization ability of the model, Dropout with Rate = 0.2 is added to individual layers. During the training process, the Adam optimizer is used, and the learning rates of G and D are set to 0.0005 and 0.0001 respectively.

[0081] S5: Train each machine learning model using the training data, and select the machine learning model with the best performance as the final prediction model.

[0082] Optionally, use the coefficient of determination (R2) as the evaluation metric for the machine learning model:

[0083]

[0084] where y i , and represent the test value, average value, and predicted value of the variable, respectively.

[0085] S6: Use the prediction model as the surrogate model for multi-objective optimization, and determine the optimal combination parameters of the high-entropy alloy coating through multi-objective optimization.

[0086] In a possible implementation, S6 specifically includes sub-steps S601 to S603:

[0087] S601: Construct the parameter variables of the high-entropy alloy coating and the search space of the parameter variables:

[0088] X = [x 1 , x 2 , x 3 , x 4 , x 5 , x 6 , x 7 , x 8 , x 9 , x 10

[0089] = [pressure, bias, flow, c Zr , c V , c Nb , c Ti , c Cr , c W , c Ta

[0090] where X represents the parameter variables, and x 1 , x 2 , x 3 , x 4 , x 5 , x 6 , x 7 , x 8 , x 9 , x 10 represent 10 parameters respectively, pressure represents air pressure, bias represents bias voltage, flow represents flow rate, and cZr represents the zirconium element content, c V represents the vanadium element content, c Nb represents the niobium element content, c Ti represents the titanium element content, c Cr represents the chromium element content, c W represents the tungsten element content, c Ta represents the tantalum element content.

[0091] It should be noted that before optimization, we need to transform the goal of maximizing hardness into a minimization problem. By minimizing 100 - f 1 (X) to achieve the maximization of f 1 (X).

[0092] S602: Construct a multi - objective optimization function based on hardness and modulus:

[0093] minF(X) = (100 - f 1 (X), f 2 (X))

[0094] where min represents minimization, F represents the multi - objective optimization function, X represents the variable, f 1 represents the hardness optimization function, f 2 represents the modulus optimization function.

[0095] Furthermore, the constraint conditions are specifically:

[0096] Subject to ∑x i = 1, i = 4, 8,..., 10

[0097] x 1 ∈[0, 1]

[0098] x 2 ∈[0, 1]

[0099] x 3 ∈[0, 1]

[0100] x 4 ∈[0.14, 0.27]

[0101] x 5 ∈[0.04, 0.31]

[0102] x 6 ∈[0.05, 0.38]

[0103] x 7 ∈[0.01, 0.28]

[0104] x 8 ∈[0.03, 0.37]

[0105] x 9 ∈ [0.06, 0.69]

[0106] x 10 ∈ [0.06, 0.41]

[0107] S603: Aiming to minimize the multi-objective optimization function, determine the optimal combination parameters of the high-entropy alloy coating through the genetic algorithm.

[0108] Among them, the genetic algorithm (Genetic Algorithm, GA) is a search and optimization algorithm based on natural selection and genetic mechanisms. It mimics the biological evolution process and performs well in solving complex optimization problems through operations such as selection, crossover, and mutation.

[0109] Optionally, the genetic algorithm is specifically the non-dominated sorting genetic algorithm. The genetic algorithm is already a mature existing technology, and the present invention will not elaborate further.

[0110] Among them, the non-dominated sorting genetic algorithm (NSGA-II, Non-dominated Sorting Genetic Algorithm II) is a genetic algorithm specifically for multi-objective optimization problems. It deals with the optimization of multiple objectives through non-dominated sorting and crowding distance comparison. The non-dominated sorting genetic algorithm is already a mature existing technology, and the present invention will not elaborate further.

[0111] In the present invention, through multi-objective optimization, especially by using the genetic algorithm, the combination parameters of the high-entropy alloy coating can be optimized efficiently and comprehensively, which not only improves the overall performance of the coating but also enhances the efficiency, applicability, and automation of the optimization process, providing a powerful tool and method for the design and development of high-performance materials.

[0112] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0113] (1) In the present invention, in the case of lacking a complete and standardized high-entropy alloy coating data set, training data is generated through the generative adversarial network, and then each machine learning model is trained separately with the training data, and the machine learning model with the best performance is selected as the final prediction model to solve the data imbalance problem, providing convenience for the design and research and development of high-entropy alloy coatings.

[0114] (2) In the present invention, through machine learning and multi-objective optimization methods, the process, composition, and material descriptors of the coating can be comprehensively considered to find the optimal combination parameters of the high-entropy alloy coating, thereby significantly improving the overall performance of the coating.

[0115] Refer to the attached instructions Figure 2, which shows a schematic structural diagram of a high-entropy alloy coating optimization system based on machine learning provided by the present invention.

[0116] The present invention also provides a high-entropy alloy coating optimization system 20 based on machine learning, which is applied to the above-mentioned high-entropy alloy coating optimization method based on machine learning, and includes:

[0117] A processor 201.

[0118] A memory 202, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor 201, the high-entropy alloy coating optimization method based on machine learning as in the method embodiment is implemented.

[0119] The high-entropy alloy coating optimization system 20 provided by the present invention can execute the above-mentioned high-entropy alloy coating optimization method based on machine learning and achieve the same or similar technical effects. To avoid repetition, the present invention will not be elaborated herein.

[0120] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0121] (1) In the present invention, in the case of lacking a complete and standardized high-entropy alloy coating data set, training data is generated through a generative adversarial network, and then each machine learning model is trained respectively with the training data. The machine learning model with the best performance is selected as the final prediction model to solve the data imbalance problem, which provides convenience for the design and research and development of high-entropy alloy coatings.

[0122] (2) In the present invention, through machine learning and multi-objective optimization methods, the process, composition, and material descriptors of the coating can be comprehensively considered to find the optimal combination parameters of the high-entropy alloy coating, thereby significantly improving the overall performance of the coating.

[0123] It should be understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), and this processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0124] It should also be understood that the memory in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0125] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0126] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, and there can be three relationships. For example, A and / or B can be: A exists alone, A and B exist simultaneously, or B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context.

[0127] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or a similar expression means any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can be: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0128] It should be understood that in various embodiments of the present invention, the sequence numbers of the above processes do not indicate the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0129] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0130] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described devices, apparatuses, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0131] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in an electrical, mechanical, or other form.

[0132] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, the functional units in each embodiment of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0134] When the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0135] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and is characterized in that when the program is executed by a processor, it implements the method for optimizing high-entropy alloy coatings based on machine learning as described in the method embodiment.

[0136] The computer-readable storage medium provided by the present invention can implement the steps and effects of the method for optimizing high-entropy alloy coatings based on machine learning in the above method embodiment. To avoid repetition, the present invention will not elaborate further.

[0137] The beneficial effects brought by the technical solution provided by the embodiment of the present invention at least include:

[0138] (1) In the present invention, in the case of lacking a complete and standardized high-entropy alloy coating data set, training data is generated through a generative adversarial network, and then each machine learning model is trained separately with the training data. The machine learning model with the best performance is selected as the final prediction model to solve the data imbalance problem, providing convenience for the design and research and development of high-entropy alloy coatings.

[0139] (2) In the present invention, through machine learning and multi-objective optimization methods, the process, composition, and material descriptors of the coating can be comprehensively considered to find the optimal combination parameters of the high-entropy alloy coating, thereby significantly improving the overall performance of the coating.

[0140] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0141] The following points need to be explained:

[0142] (1) The accompanying drawings of the embodiments of the present invention only relate to the structures involved in the embodiments of the present invention, and other structures can refer to the general design.

[0143] (2) For clarity, in the accompanying drawings used to describe the embodiments of the present invention, the thickness of layers or regions is enlarged or reduced, that is, these drawings are not drawn to actual scale. It can be understood that when an element such as a layer, film, region, or substrate is referred to as being "on" or "under" another element, the element can be "directly" on or under the other element or there can be intermediate elements.

[0144] (3) Without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0145] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A high entropy alloy coating optimization method based on machine learning, characterized in that: include: S1: Get multiple material descriptors; S2: screening the material descriptors; S3: Taking the coating process, composition and selected material descriptors as input and the coating performance as output, the mapping relationship between input and output is established based on different machine learning models; S4: Generate training data through generative adversarial network; S5: training each of the machine learning models using the training data, and selecting the machine learning model with the best performance as the final prediction model; S6: Using the prediction model as a proxy model for multi-objective optimization, the optimal combination parameters of the high entropy alloy coating are determined by constructing a multi-objective optimization function based on hardness and modulus; wherein the constructed multi-objective optimization function is: minF(X)=(100-f1(X),f2(X)) Among them, min means minimization, F means multi-objective optimization function, X means variable, f1 means hardness optimization function, and f2 means modulus optimization function.

2. The high entropy alloy coating optimization method based on machine learning according to claim 1 is characterized in that: The material descriptor specifically includes: Among them, f ave,i represents the average descriptor of the ith material, w k,i represents the weight of the kth element of the ith material, f k,i represents the physical and chemical property value of the kth element of the ith material, m represents the total number of elements, and f var,i represents the variance descriptor of the ith material.

3. The high entropy alloy coating optimization method based on machine learning according to claim 1 is characterized in that: The S2 specifically includes: S201: Calculating the Spearman correlation coefficients between the material descriptors and the Spearman correlation coefficients between the material descriptors and the performance of the high entropy alloy coating; S202: when the Spearman correlation coefficient between the material descriptors is greater than the correlation threshold, comparing the Spearman correlation coefficients between the two material descriptors and the performance of the high entropy alloy coating, retaining the larger one and deleting the smaller one; S203: inputting the retained material descriptors into the random forest model embedded with the recursive feature elimination algorithm, and screening the material descriptors through the random forest model embedded with the recursive feature elimination algorithm.

4. The high entropy alloy coating optimization method based on machine learning according to claim 3 is characterized in that: The Spearman correlation coefficient is calculated as follows: Where ρ represents the Spearman correlation coefficient, r(x i ) represents the rank of the i-th value in the material descriptor x, r(y i ) represents the rank of the i-th value in the material descriptor y, represents the average of the ranks of the values ​​in the material descriptor x, represents the average of the ranks of the values ​​in the material descriptor y, and n represents the total number of values ​​in the material descriptor.

5. The high entropy alloy coating optimization method based on machine learning according to claim 1, characterized in that: The machine learning models specifically include AdaBoost model, RF model, GBDT model, ETR model, BG model and XGBoost model.

6. The high entropy alloy coating optimization method based on machine learning according to claim 1, characterized in that: The S4 specifically includes: S401: Selecting a first noise Z1 and a second noise Z2 from random noise; S402: Inputting the first noise and the second noise into a converter based on a neural network to obtain a first feature Z1′ and a second feature Z2′ after conversion; S403: performing splicing processing on the first feature Z1′ and the second feature Z2′ to obtain a splicing feature; S404: Using the concatenated feature and the attribute p as input z of a generator in a generative adversarial network to obtain a virtual feature G(z); S405: Concatenate the attribute p with the real feature to construct real data, and concatenate the attribute p with the virtual feature G(z) to construct false data; S406: Using the real data and the false data as inputs of a discriminator in a generative adversarial network; S407: Training and optimizing the generator and discriminator in the generative adversarial network through adversarial games; S408: Generate the training data through the trained generator.

7. The high entropy alloy coating optimization method based on machine learning according to claim 1 is characterized in that: The cost functions of the generator and discriminator in the generative adversarial network are specifically: Among them, min means minimization, G means generator, max means maximization, D means discriminator, V means cost function, E means mathematical expectation operation, x means real data, P data(x) represents the real data distribution, log represents the logarithmic function, z represents the noise data, P z (z) represents the noise data distribution.

8. The high entropy alloy coating optimization method based on machine learning according to claim 1, characterized in that: The S6 specifically includes: S601: Construct parameter variables of the high entropy alloy coating and a search space of parameter variables: X=[x1,x2,x3,x4,x5,x6,x7,x8,x9,x 10 ] =[pressure,bias,flow,c Zr ,c V ,c Nb ,c Ti ,c Cr ,c W ,c Ta ] Among them, X represents the parameter variable, x1,x2,x3,x4,x5,x6,x7,x8,x9,x 10 Represents 10 parameters respectively, pressure represents air pressure, bias represents bias pressure, flow represents flow rate, c Zr Indicates the zirconium content, c V Indicates the vanadium content, c Nb Indicates the niobium content, c Ti Indicates the titanium content, c Cr Indicates the chromium content, c W Indicates the tungsten content, c Ta Indicates the tantalum content; S602: construct a multi-objective optimization function based on hardness and modulus; S603: With the goal of minimizing the multi-objective optimization function, determine the optimal combination parameters of the high entropy alloy coating through a genetic algorithm.

9. The high entropy alloy coating optimization method based on machine learning according to claim 8, characterized in that: The genetic algorithm is specifically a non-dominated sorting genetic algorithm.

10. A high entropy alloy coating optimization system based on machine learning, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the high-entropy alloy coating optimization method based on machine learning as described in any one of claims 1 to 9 is implemented.

Images