Method and device for optimizing low alloy steel performance based on differential evolution and machine learning algorithm

Abstract

The present invention provides a method and device for optimizing the performance of low alloy steel using differential evolution and machine learning algorithms, and relates to the technical field of machine learning. The method comprises: obtaining low alloy steel performance data, and preprocessing and feature conversion of the low alloy steel performance data, and after obtaining the first data, outputting a target subset through a two-stage feature selection method using a differential evolution algorithm and model importance sorting; then performing hyperparameter search using a differential evolution algorithm, constructing a machine learning model, and then performing a multi-objective optimization algorithm to obtain a target solution set. Performing hyperparameter search using a differential evolution algorithm improves the efficiency of rapidly screening potential candidate materials, and at the same time, reduces the time cost of screening potential candidate materials; then performing optimization processing using a multi-objective optimization algorithm, quickly and effectively balancing the mutually exclusive properties of carbon steel tensile strength and ductility, helps to find materials that have both high strength and sufficient ductility to meet actual engineering needs.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and in particular to a method and device for optimizing the performance of low alloy steel using differential evolution and machine learning algorithms. Background Art

[0002] Steel has always been an important material in engineering structures, and its mechanical properties are crucial to ensuring the reliability and safety of the structure. Among these steels, carbon steel is a common and important material. Due to its relatively low price, low alloy element content, and excellent mechanical properties, it is widely used in various fields such as construction, bridges, automobile manufacturing, and marine construction.

[0003] With the promotion and popularization of artificial intelligence, computational materials science has developed rapidly, promoting the shift of material research and development from the traditional trial and error and historical experience model to the fourth paradigm of data-driven development. Machine learning has been widely used in various fields of materials science, including predicting material properties, selecting materials, and optimizing design, thanks to its ability to autonomously learn and efficiently solve complex tasks. Since the search space for the chemical composition of carbon steel is huge and its mechanical properties are often contradictory, a method combining data-driven machine learning and multi-objective optimization has become popular in the field of materials for designing new materials and analyzing and optimizing material properties.

[0004] However, traditional technologies cannot quickly and effectively optimize the mutually exclusive properties of carbon steel tensile strength and elongation during optimization; they are also faced with low efficiency and high cost in quickly screening potential candidate materials from large amounts of data. Summary of the invention

[0005] In order to solve the problem that traditional technology optimization cannot quickly and effectively optimize the mutually exclusive properties of carbon steel tensile strength and elongation, and faces the technical problems of low efficiency and high cost in quickly screening potential candidate materials in large amounts of data, the present invention provides a low alloy steel performance optimization method and device based on differential evolution and machine learning algorithms.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] A first aspect of the present invention provides a method for optimizing low alloy steel performance using differential evolution and machine learning algorithms, the method comprising:

[0008] S1, obtaining low alloy steel performance data, and preprocessing and feature conversion of the low alloy steel performance data to obtain first data;

[0009] S2, outputting a target subset according to the first data through a two-stage feature selection method of differential evolution algorithm and model importance ranking;

[0010] S3, performing hyperparameter optimization on the target subset through a differential evolution algorithm to construct a machine learning model;

[0011] S4, performing a multi-objective optimization algorithm on the machine learning model to obtain a target solution set.

[0012] The second aspect of the present invention provides a low alloy steel performance optimization device based on differential evolution and machine learning algorithm, comprising:

[0013] An acquisition module, used for acquiring low alloy steel performance data, and performing preprocessing and feature conversion on the low alloy steel performance data to obtain first data;

[0014] An output module, used to output a target subset according to the first data by a two-stage feature selection method of differential evolution algorithm and model importance ranking;

[0015] A construction module is used to perform hyperparameter optimization on the target subset through a differential evolution algorithm to construct a machine learning model;

[0016] The acquisition module is used to perform a multi-objective optimization algorithm on the machine learning model to obtain a target solution set.

[0017] The third aspect of the present invention provides a low alloy steel performance optimization system based on differential evolution and machine learning algorithms, comprising:

[0018] processor;

[0019] A memory having computer-readable instructions stored therein, which, when executed by a processor, implements the low alloy steel performance optimization method based on the differential evolution and machine learning algorithm of the first aspect.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the low alloy steel performance optimization method using the differential evolution and machine learning algorithm of the first aspect.

[0021] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0022] In the present invention, hyperparameter search is performed through differential evolution algorithm, which improves the efficiency of rapid screening of potential candidate materials and reduces the time cost of screening potential candidate materials; optimization processing is performed through multi-objective optimization algorithm, which quickly and effectively balances the mutually exclusive properties of carbon steel tensile strength and ductility, and helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A schematic flow chart of a method for optimizing low alloy steel performance using a differential evolution and machine learning algorithm provided in an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of the structure of a low alloy steel performance optimization device using a differential evolution and machine learning algorithm provided in an embodiment of the present invention;

[0026] Figure 3 A schematic structural diagram of a low alloy steel performance optimization system using a differential evolution and machine learning algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0028] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0029] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0030] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0031] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0032] Reference Manual Attached Figure 1, showing a flow chart of a low alloy steel performance optimization method using a differential evolution and machine learning algorithm provided in an embodiment of the present invention.

[0033] The embodiment of the present invention provides a method for optimizing the performance of low alloy steel using differential evolution and machine learning algorithms. The method can be implemented by a low alloy steel performance optimization device using differential evolution and machine learning algorithms. The low alloy steel performance optimization device using differential evolution and machine learning algorithms can be a terminal or a server. The processing flow of the method for optimizing the performance of low alloy steel using differential evolution and machine learning algorithms can include the following steps:

[0034] S1, obtaining low alloy steel performance data, and preprocessing and feature conversion of the low alloy steel performance data to obtain first data.

[0035] In a possible implementation, S1 specifically includes sub-steps S101 and S102:

[0036] S101: Acquire multiple observation values ​​of low alloy carbon steel as an initial data set, wherein each observation value includes an input value and an output value, the input value includes 9 chemical elements and 3 process parameters, and the output value includes a tensile strength value and an elongation value.

[0037] S102: Convert the 9 chemical elements into 155 physical and chemical descriptors, and obtain first data by calculating the mean, variance and sum of the 155 physical and chemical descriptors. The first data includes 174 characteristic values ​​and 3 process parameters.

[0038] It should be noted that the first data consists of 177 features.

[0039] Specifically, the present invention searches for 367 observations of various low-alloy carbon steels as the initial data set. Each observation in the initial data set consists of 9 chemical elements and 3 process parameters as input values, and two target properties, tensile strength and elongation, as outputs. Any element can be represented by a series of atomic-scale physical and chemical properties. The present invention converts the 9 elements into 58 physicochemical descriptors, calculates their mean, variance and sum totaling 177 features, screens out the physical and chemical properties of key atoms, and synergistically optimizes the mutually exclusive properties of low-alloy steel, tensile strength and elongation.

[0040] In the present invention, by calculating the mean, variance and sum of these descriptors, 177 eigenvalues ​​are obtained, which can provide a richer feature representation. These statistical features help to reveal the potential patterns and relationships in the data; at the same time, by screening key physical and chemical properties, redundant or irrelevant features can be removed and the most informative features can be retained, which helps to improve the predictive performance and generalization ability of the model. The coordinated optimization of the two mutually exclusive properties of tensile strength and ductility of low alloy steel can find the best balance in multi-objective optimization and improve the comprehensive performance of the model under multi-objective conditions.

[0041] Furthermore, through preprocessing and feature conversion, an efficient data training set was constructed. Using the differential evolution algorithm and multi-objective optimization algorithm, the optimal model and parameter combination can be found more quickly, providing effective guidance for the design of low-alloy steel and obtaining materials with better performance.

[0042] In this paper, 9 chemical elements are converted into 58 physicochemical descriptors, and the mean, variance and sum of these descriptors are calculated, and a total of 177 features are obtained. These rich feature representations help to reveal the potential patterns and relationships in the data; calculating statistical features such as mean, variance and sum can capture the behavior of chemical elements under different combinations and conditions, providing a deeper understanding of the properties of low alloy steel.

[0043] S2: Output a target subset based on the first data through a two-stage feature selection method of differential evolution algorithm and model importance ranking.

[0044] In a possible implementation, S2 specifically includes sub-steps S201 to S203:

[0045] S201: Convert the first data into individuals, perform initialization population calculation, and sort the individual fitness.

[0046] Among them, the initialization population calculation adopts the formula:

[0047]

[0048] Where X i,G is a hyperparameter, NP is the population size, D is the dimension of the target vector, i.e. the parameter to be optimized, G is the current generation, and when G = 0, it indicates the initial population. When a hyperparameter is converted to a vector, the jth dimension of the vector. rand is a uniform random number in the interval [0,1]. and are the upper and lower bounds of the j-th dimension of the vector respectively.

[0049] It should be noted that in the differential evolution algorithm process, the feature combination is encoded as a binary string of 0 and 1, "1" means the corresponding feature is selected, and "0" means the corresponding feature is not selected. Each string represents an individual, and the collection of individuals forms a population. After the initial population is established, the individuals are sorted using the model prediction evaluation index as the fitness value.

[0050] Specifically, the preprocessed data training set is input into the differential evolution algorithm to search for the feature subset of the best prediction model. In the differential evolution algorithm process, the feature combination is encoded as a binary string of 0 and 1, 1 means selecting the corresponding feature and 0 means not selecting the corresponding feature. Each string represents an individual, and the collection of individuals forms a population. After the initial population is established, the individuals are sorted using the model prediction evaluation index as the fitness value. After the crossover mutation operation, the individuals with higher rankings based on the roulette wheel selection method are more likely to be selected into the next generation in the early selection process. Then, the corresponding best subset is obtained through the differential evolution algorithm. DE is run multiple times, and the features that frequently appear in the best prediction subset output by each iteration will be included in the final best subset. Then, the model importance ranking FIS is applied to the new data set composed of the features selected by the DE differential evolution algorithm. The features that contribute more to the model are screened through feature importance evaluation and the selected features are automatically modeled, so that a more concise model can be built. Finally, the parameters are adjusted during the training process. The differential evolution algorithm is also used here. Each individual represents a candidate hyperparameter setting. The model performance is evaluated on the training set using the ten-fold cross validation method. The optimal solution in the hyperparameter space is searched iteratively. The selected features are used as input, and the mechanical properties of tensile strength and elongation are used as output. DE tunes the model's hyperparameters and establishes the final prediction model for atomic physicochemical descriptors.

[0051] S202: Perform mutation operations and crossover strategies on individuals and select target individuals with high rankings.

[0052] Among them, the mutation operation adopts the formula:

[0053] V i，G =X best,G +F×(X r1，G -X r2,G ).

[0054] Among them, the mutation operation is to generate disturbances through individual differences to maintain population diversity, which is also the most important step in the DE algorithm. The basic idea of ​​the mutation operation is: first randomly select two vectors, then multiply their difference by a scaling factor as a perturbation variable, and finally add the basis vector to the perturbation variable to obtain the mutation vector; in order to improve the diversity of the population, the DE algorithm crosses the basis vector and the mutation vector in the crossover operation to obtain a recombinant test vector.

[0055] Among them, the crossover strategy selects binomial crossover, and the formula is:

[0056]

[0057] Among them, V i,G is the mutation vector. r1,G , X r2,G is a randomly selected individual from the population. rbest,G is the optimal solution. F is the variation factor, which is usually limited to the interval [0,2]. i,G is the basis vector. i,G The mutation vector is hybridized to obtain the recombination test vector. is the test vector U i,G The jth dimension of . CR is the crossover rate. rand∈[0,1] is the random number corresponding to the jth dimension component, and Jrand is a random integer between dimensions [0,D].

[0058] It should be noted that individuals (target individuals) with higher rankings based on the roulette wheel selection method after the crossover mutation operation are more likely to be selected into the next generation in the early stage of the selection process.

[0059] S203: Perform differential evolution algorithm and selection operation on the target individual.

[0060] Among them, the selection operation adopts the formula:

[0061]

[0062] Among them, X i,G is the target vector, X i,G+1 is the target vector of the G+1th generation, and f(·) is the fitness function.

[0063] It should be noted that the selection operation process determines the evolutionary direction of the entire population. DE uses a fitness evaluation selection mechanism to ensure that the population continues to converge to the global optimum. According to the fitness value, the optimal selection is made between the target vector and its corresponding test vector to generate the next generation of individuals.

[0064] S204: According to the built-in importance ranking of the model, the feature subset obtained after the differential evolution algorithm screens the features is subjected to a second stage screening to obtain a target subset.

[0065] Specifically, the corresponding best subset is obtained through differential evolution algorithm. DE is run multiple times, and the frequently repeated features (selection operations) in the best prediction subset output by each iteration will be included in the final best subset (target subset).

[0066] Furthermore, in the second stage, the model importance ranking FIS is applied to the new data set consisting of the features (target subset) selected by the DE differential evolution algorithm. The features that contribute most to the model are screened through feature importance evaluation and the selected features are automatically modeled, so that a more concise model (machine learning model) can be constructed.

[0067] In the present invention, the two-stage feature selection method of differential evolution algorithm and model importance ranking can effectively improve model performance, reduce overfitting risk, accelerate the training process, improve interpretability, has strong global search capability and robustness, and is suitable for various feature selection and model optimization tasks.

[0068] S3, the target subset is optimized through the differential evolution algorithm to find hyperparameters and build a machine learning model.

[0069] It should be noted that the machine learning model includes a regression prediction model for tensile strength values ​​and a regression prediction model for elongation values.

[0070] Optionally, the present invention divides the initial data set into a training set for model optimization and a test set for evaluating the prediction accuracy of the model. The training set is 80% of the total data set, i.e., 293 data, and the test set is 20% of the total data set. We selected five regression prediction models, namely Gaussian kernel support vector regression (SVR), gradient boosting regression (GBR), random forest (RF), adaptive boosting regression (AdaBoost) and multi-layer perceptron (MLP), and modeled 174 descriptors generated from 9 chemical elements and 58 atomic physicochemical properties, respectively, to determine the best prediction model for the corresponding mechanical properties.

[0071] Among them, the SHAP method provides an interpretable explanation of the prediction results of the machine learning model by calculating the contribution of each feature to the model prediction. Local interpretability and global interpretability are achieved, and the importance of features is analyzed and evaluated. In the present invention, we analyze the importance of chemical composition or physicochemical descriptors and process parameters to the prediction of target properties by comparing SHAP values, which is helpful for subsequent material design.

[0072] Specifically, the differential evolution algorithm is used to find hyperparameters, the root mean square error and determination coefficient are used to evaluate the prediction accuracy of the model, and the results of ten-fold cross validation on the training set are used as indicators for selecting model hyperparameters. After the training, the model with the best fitting accuracy is selected to predict the test set data, and Shapley is used to analyze and explain the characteristics of the training model.

[0073] In the present invention, hyperparameter optimization and model construction are performed through differential evolution algorithm, combined with SHAP analysis and feature importance evaluation, which can improve the model prediction accuracy, interpretability and generalization ability, reduce the risk of overfitting, simplify the model structure, accelerate the training process, and provide effective methods and tools for data-driven material design.

[0074] S4, performs a multi-objective optimization algorithm on the machine learning model to obtain the target solution set.

[0075] In a possible implementation, S4 specifically includes:

[0076] The machine learning model is used as the objective function, and multi-objective optimization is performed through the NSGA-II algorithm to obtain the target solution set, where the target solution set is a numerical value with a balance between the tensile strength value and the elongation value.

[0077] Specifically, the genetic evolutionary algorithm simulates the animal evolution process in nature. First, the population is initialized, and the objective function is converted into a fitness function to calculate and evaluate each individual in the population. Then, new individuals are generated through mutation, crossover, and selection to form a new population. In order to solve the multi-objective optimization problem in the present invention, we used NAGA-II based on genetic algorithm, efficient non-dominated sorting and crowding distance measurement as the main optimization algorithm. The parameter settings of NSGA-II, including population size, number of iterations, etc., were repeated many times to ensure the stability of the experiment. The top 200 largest candidate solutions were selected to draw the Pareto frontier. The algorithm finds a trade-off between multiple objectives. Excellent individuals that are not dominated by other individuals are called non-dominated solutions, that is, Pareto optimal solutions. The surface formed by the Pareto optimal solution set in the target space is the Pareto frontier.

[0078] In the present invention, the tensile strength value and the ductility value are used as the objective function, and a solution set with balanced performance between the two objectives is found through a multi-objective optimization algorithm such as NSGA-II. This helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs. At the same time, the parameters of NSGA-II are set, such as population size, number of iterations, etc., to ensure the stability and reliability of the algorithm.

[0079] Furthermore, by optimizing the comprehensive model using the multi-objective optimization algorithm NSGA-II, it is possible to find a solution set with balanced performance among multiple objectives, which helps guide material design and engineering applications and improve the comprehensive effect of material performance.

[0080] In the present invention, the machine learning model is comprehensively optimized through a multi-objective optimization algorithm to obtain a solution set of balanced performance, guide material design and engineering applications, improve the comprehensive effect of the material, save time and cost, and promote the progress of material research and development.

[0081] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0082] In the present invention, hyperparameter search is performed through differential evolution algorithm, which improves the efficiency of rapid screening of potential candidate materials and reduces the time cost of screening potential candidate materials; optimization processing is performed through multi-objective optimization algorithm, which quickly and effectively balances the mutually exclusive properties of carbon steel tensile strength and ductility, and helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs.

[0083] Reference Manual Attached Figure 2 , showing a structural schematic diagram of a low alloy steel performance optimization device using a differential evolution and machine learning algorithm provided in an embodiment of the present invention.

[0084] The embodiment of the present invention provides a low alloy steel performance optimization device 20 using differential evolution and machine learning algorithms, comprising:

[0085] The acquisition module 201 is used to acquire low alloy steel performance data, and preprocess and transform the low alloy steel performance data to obtain first data.

[0086] In a possible implementation, the acquisition module 201 is specifically used for:

[0087] A plurality of observation values ​​of low alloy carbon steel are obtained as an initial data set, wherein each observation value includes an input value and an output value, the input value includes 9 chemical elements and 3 process parameters, and the output value includes a tensile strength value and an elongation value.

[0088] The nine chemical elements are converted into 155 physicochemical descriptors, and the physicochemical descriptors are converted into first data by calculating the mean, variance and sum, and the first data includes 174 characteristic values ​​and three process parameters.

[0089] The output module 202 is used to output the target subset according to the first data through a two-stage feature selection method of differential evolution algorithm and model importance sorting.

[0090] In a possible implementation, the output module 202 is specifically configured to:

[0091] The first data is converted into individuals, and the population is initialized and calculated, and the fitness of the individuals is sorted.

[0092] Among them, the initialization population calculation adopts the formula:

[0093]

[0094] Where X i,Gis a hyperparameter, NP is the population size, D is the dimension of the target vector, i.e. the parameter to be optimized, G is the current generation, and when G = 0, it indicates the initial population. When a hyperparameter is converted to a vector, the jth dimension of the vector. rand is a uniform random number in the interval [0,1]. and are the upper and lower bounds of the j-th dimension of the vector respectively.

[0095] Perform mutation operations and crossover strategies on individuals and select the top-ranked target individuals.

[0096] Among them, the mutation operation adopts the formula:

[0097] V i,G =X best,G +F×(X r1,G -X r2,G ).

[0098] Among them, the crossover strategy selects binomial crossover, and the formula is:

[0099]

[0100] Among them, X i,G is the mutation vector. r1,G , X r2,G is a randomly selected individual from the population. rbest,G is the optimal solution. F is the variation factor, which is usually limited to the interval [0,2]. i,G is the basis vector. i,G The mutation vector is hybridized to obtain the recombination test vector. is the test vector U i,G The jth dimension of . CR is the crossover rate. rand∈[0,1] is the random number corresponding to the jth dimension component, and Jrand is a random integer between dimensions [0,D]

[0101] Perform differential evolution algorithm and selection operations on the target individuals.

[0102] Among them, the selection operation adopts the formula:

[0103]

[0104] Among them, X i,G is the target vector, X i,G+1 is the target vector of the G+1th generation, and f(·) is the fitness function.

[0105] According to the built-in importance ranking of the model, the feature subset obtained after the differential evolution algorithm screens the features is subjected to the second stage screening to obtain the target subset.

[0106] The construction module 203 is used to perform hyperparameter optimization on the target subset through the differential evolution algorithm to construct a machine learning model.

[0107] The acquisition module 204 is used to perform a multi-objective optimization algorithm on the machine learning model to obtain a target solution set.

[0108] In a possible implementation, the obtaining module 204 is specifically used for:

[0109] The machine learning model is used as the objective function, and multi-objective optimization is performed through the NSGA-II algorithm to obtain the target solution set, where the target solution set is a numerical value with a balance between the tensile strength value and the elongation value.

[0110] The low alloy steel performance optimization device 20 based on differential evolution and machine learning algorithm provided by the present invention can implement the above-mentioned low alloy steel performance optimization method based on differential evolution and machine learning algorithm, and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0111] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0112] In the present invention, hyperparameter search is performed through differential evolution algorithm, which improves the efficiency of rapid screening of potential candidate materials and reduces the time cost of screening potential candidate materials; optimization processing is performed through multi-objective optimization algorithm, which quickly and effectively balances the mutually exclusive properties of carbon steel tensile strength and ductility, and helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs.

[0113] Reference Manual Attached Figure 3 , showing a structural schematic diagram of a low alloy steel performance optimization system based on a differential evolution and machine learning algorithm provided by the present invention.

[0114] The present invention also provides a low alloy steel performance optimization system 30 using differential evolution and machine learning algorithms, which is applied to the low alloy steel performance optimization method using differential evolution and machine learning algorithms, including:

[0115] Processor 301.

[0116] The memory 302 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 301, a low alloy steel performance optimization method using differential evolution and machine learning algorithms as in the method embodiment is implemented.

[0117] The low alloy steel performance optimization system 30 based on differential evolution and machine learning algorithm provided by the present invention can execute the above-mentioned low alloy steel performance optimization method based on differential evolution and machine learning algorithm, and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0118] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0119] In the present invention, hyperparameter search is performed through differential evolution algorithm, which improves the efficiency of rapid screening of potential candidate materials and reduces the time cost of screening potential candidate materials; optimization processing is performed through multi-objective optimization algorithm, which quickly and effectively balances the mutually exclusive properties of carbon steel tensile strength and ductility, and helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs.

[0120] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0121] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0122] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by 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 process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). 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 data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0123] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0124] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0125] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0126] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0127] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0128] In the 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 only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as 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 mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0129] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0131] If the functions are implemented in the form of software functional 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 can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0132] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, a low alloy steel performance optimization method using differential evolution and machine learning algorithms as described in the method embodiment is implemented.

[0133] A computer-readable storage medium provided by the present invention can implement the steps and effects of the low alloy steel performance optimization method of differential evolution and machine learning algorithm in the above-mentioned method embodiment. To avoid repetition, the present invention will not go into details.

[0134] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0135] In the present invention, hyperparameter search is performed through differential evolution algorithm, which improves the efficiency of rapid screening of potential candidate materials and reduces the time cost of screening potential candidate materials; optimization processing is performed through multi-objective optimization algorithm, which quickly and effectively balances the mutually exclusive properties of carbon steel tensile strength and ductility, and helps to find materials that have not only high strength but also sufficient ductility to meet actual engineering needs.

[0136] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0137] There are a few points to note:

[0138] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0139] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of the layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is 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 may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0140] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0141] The above are only specific embodiments 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 based on the protection scope of the claims.

Claims

1. A low alloy steel performance optimization method based on differential evolution and machine learning algorithm, characterized in that: Methods include: S1, obtaining low alloy steel performance data, and preprocessing and feature conversion of the low alloy steel performance data to obtain first data; S2, outputting a target subset according to the first data through a two-stage feature selection method of differential evolution algorithm and model importance ranking; S3, performing hyperparameter optimization on the target subset through a differential evolution algorithm to construct a machine learning model; S4, performing a multi-objective optimization algorithm on the machine learning model to obtain a target solution set; Wherein, the S2 specifically includes: S201: converting the first data into individuals, performing initialization population calculation, and ranking the fitness of the individuals; S202: performing mutation operation and crossover strategy on the individuals, and selecting target individuals with high rankings; S203: performing a differential evolution algorithm and a selection operation on the target individual; S204: According to the built-in importance ranking of the model, the feature subset obtained after the differential evolution algorithm screens the features is subjected to a second stage screening to obtain a target subset.

2. The low alloy steel performance optimization method based on differential evolution and machine learning algorithm according to claim 1 is characterized in that: The S1 specifically includes: S101: Acquire a plurality of observation values ​​of low alloy carbon steel as an initial data set, wherein each of the observation values ​​includes an input value and an output value, the input value includes 9 chemical elements and 3 process parameters, and the output value includes a tensile strength value and an elongation value; S102: Convert the 9 chemical elements into 155 physicochemical descriptors, and obtain first data by calculating the mean, variance and sum of the 155 physicochemical descriptors, where the first data includes 174 characteristic values ​​and 3 process parameters.

3. The low alloy steel performance optimization method based on differential evolution and machine learning algorithm according to claim 1, characterized in that: The S2 further includes: The initial population calculation adopts the formula: Where X i,G is a hyperparameter, NP is the population size; D is the dimension of the target vector, that is, the parameter to be optimized; G is the current generation, when G = 0, it means the initial population; When a hyperparameter is converted to a vector, the jth dimension of the vector; rand is a uniform random number in the interval [0,1]; and are the upper and lower bounds of the j-th dimension of the vector respectively; The mutation operation adopts the formula: V i,G =X best,G +F×(X r1,G -X r2,G ); The crossover strategy selects binomial crossover, and the formula is: Among them, V i,G is the mutation vector; X r1,G , X r2,G is a randomly selected individual from the population; X rbest,G is the optimal solution; F is the variation factor, which is usually limited to the interval [0,2]; X i,G is the basis vector; U i,G Perform hybridization for the mutation vector to obtain a recombinant test vector; is the test vector U i,G CR is the crossover rate; rand∈[0,1] is the random number corresponding to the j-th dimension component, and Jrand is a random integer between dimensions [0,D]; The selection operation uses the formula: Among them, X i,G is the target vector, X i,G+1 is the target vector of the G+1th generation, and f(·) is the fitness function.

4. The low alloy steel performance optimization method based on differential evolution and machine learning algorithm according to claim 1, characterized in that: The S4 specifically includes: The machine learning model is used as the objective function, and multi-objective optimization is performed through the NSGA-II algorithm to obtain the target solution set, wherein the target solution set is a numerical value with a balance between the tensile strength value and the elongation value.

5. A low alloy steel performance optimization device using differential evolution and machine learning algorithm, wherein the low alloy steel performance optimization device using differential evolution and machine learning algorithm is used to implement a low alloy steel performance optimization method using differential evolution and machine learning algorithm as described in any one of claims 1 to 4, characterized in that: The device comprises: An acquisition module, used for acquiring low alloy steel performance data, and performing preprocessing and feature conversion on the low alloy steel performance data to obtain first data; An output module, used to output a target subset according to the first data by a two-stage feature selection method of differential evolution algorithm and model importance ranking; A construction module is used to perform hyperparameter optimization on the target subset through a differential evolution algorithm to construct a machine learning model; An acquisition module, used to perform a multi-objective optimization algorithm on the machine learning model to obtain a target solution set; Wherein, the output module is specifically used for: Convert the first data into individuals, perform initialization population calculation, and sort the individual fitness; Perform mutation operations and crossover strategies on the individuals to select target individuals with top rankings; Performing a differential evolution algorithm and a selection operation on the target individual; According to the built-in importance ranking of the model, the feature subset obtained after the differential evolution algorithm screens the features is subjected to the second stage screening to obtain the target subset.

6. The low alloy steel performance optimization device based on differential evolution and machine learning algorithm according to claim 5 is characterized in that: The acquisition module is specifically used for: Acquire a plurality of observation values ​​of low alloy carbon steel as an initial data set, wherein each of the observation values ​​comprises an input value and an output value, the input value comprises 9 chemical elements and 3 process parameters, and the output value comprises a tensile strength value and an elongation value; The nine chemical elements are converted into 155 physicochemical descriptors, and the physicochemical descriptors are subjected to calculation of means, variances and sums to obtain first data, wherein the first data includes 174 characteristic values ​​and three process parameters.

7. The low alloy steel performance optimization device based on differential evolution and machine learning algorithm according to claim 5, characterized in that: The output module is also used for: The initial population calculation adopts the formula: Where X i,G is a hyperparameter, NP is the population size; D is the dimension of the target vector, that is, the parameter to be optimized; G is the current generation, when G = 0, it means the initial population; When a hyperparameter is converted to a vector, the jth dimension of the vector; rand is a uniform random number in the interval [0,1]; and are the upper and lower bounds of the j-th dimension of the vector respectively; The mutation operation adopts the formula: V i,G =X best,G +F×(X r1,G -X r2,G ); The crossover strategy selects binomial crossover, and the formula is: Among them, V i,G is the mutation vector; X r1,G , X r2,G is a randomly selected individual from the population; X rbest,G is the optimal solution; F is the variation factor, which is usually limited to the interval [0,2]; X i,G is the basis vector; U i,G Perform hybridization for the mutation vector to obtain a recombinant test vector; is the test vector U i,G CR is the crossover rate; rand∈[0,1] is the random number corresponding to the j-th dimension component, and Jrand is a random integer between dimensions [0,D]; The selection operation uses the formula: Among them, X i,G is the target vector, X i,G+1 is the target vector of the G+1th generation, and f(·) is the fitness function.

8. The low alloy steel performance optimization device based on differential evolution and machine learning algorithm according to claim 7, characterized in that: The module is specifically used for: The machine learning model is used as the objective function, and multi-objective optimization is performed through the NSGA-II algorithm to obtain the target solution set, wherein the target solution set is a numerical value with a balance between the tensile strength value and the elongation value.

Images