Medium-entropy steel performance prediction system and method and computer readable storage medium

Abstract

The invention relates to the technical field of metal materials, and provides a medium-entropy steel performance prediction system and method and a computer readable storage medium, and the method comprises the following steps: obtaining a characteristic data set of medium-entropy steel performance; constructing a KAN network prediction model, optimizing the KAN network prediction model through ESOA, and obtaining an optimal hyper-parameter combination of the KAN network prediction model; training the KAN network prediction model according to the feature data set and the optimal hyper-parameter combination; and predicting the performance of the medium-entropy steel according to the trained KAN network prediction model. According to the method, the strong learning ability and calculation ability of the KAN network prediction model are utilized, so that the relationship between the medium-entropy steel performance and the components can be quickly and efficiently obtained, and the prediction of the medium-entropy steel performance is realized; the problem that in the prior art, time consumption is large when the relation between the performance and the components of the medium-entropy alloy is researched through an experimental means is solved.

Description

Technical Field

[0001] The present invention relates to the field of metal materials, and in particular to a medium-entropy steel performance prediction system, method and computer-readable storage medium. Background Art

[0002] Medium-Entropy Alloys (MEAs) are a new type of multi-principal alloys, usually composed of three to five main elements, with high mixing entropy and excellent physical, chemical and mechanical properties. Compared with traditional alloy design, the design concept of medium-entropy alloys is more innovative, and its high mixing entropy helps to form a simple and stable phase structure, such as face-centered cubic (FCC) or body-centered cubic (BCC) structure. This unique structure makes medium-entropy alloys perform well in high-temperature oxidation resistance, high strength, high toughness and corrosion resistance.

[0003] FeCrNi(AlTi)x alloy is a typical medium entropy alloy (medium entropy steel). Its microstructure and mechanical properties can be significantly changed by adjusting the content of Al and Ti. Studies have shown that with the increase of Al and Ti content, the phase structure of FeCrNi(AlTi)x alloy changes from single-phase FCC structure to dual-phase FCC and BCC structure, and the volume fraction of BCC phase increases significantly. This change in phase structure directly affects the hardness and compressive yield strength of the alloy. The study also found that the hardness of FeCrNi(AlTi)x alloy increases with the increase of Al and Ti content, indicating that the doping of Al and Ti elements plays a key role in improving the hardness of the alloy.

[0004] In the prior art, the relationship between the performance and composition of entropy alloys is mainly studied by experimental means. Each experiment requires multiple steps such as batching, smelting, heat treatment and testing, which has a long cycle and is time-consuming. Summary of the invention

[0005] In order to solve the problem of time-consuming experimental research on the relationship between entropy alloy performance and composition, the present application provides a medium-entropy steel performance prediction system that can efficiently obtain the relationship between medium-entropy steel performance and composition.

[0006] The medium entropy steel performance prediction system comprises:

[0007] A data processing module, used to obtain characteristic data sets of medium entropy steel performance;

[0008] A construction module is used to construct a KAN network prediction model, and optimize the KAN network prediction model through ESOA to obtain an optimal hyperparameter combination of the KAN network prediction model;

[0009] A training module, used for training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination;

[0010] The prediction module is used to predict the performance of medium entropy steel according to the trained KAN network prediction model.

[0011] The medium-entropy steel performance prediction system constructs a KAN network prediction model, optimizes the structure and parameters of the KAN network prediction model through ESOA, obtains the optimal hyperparameter combination of the KAN network prediction model, and then trains the KAN network prediction model according to the feature data set and the optimal hyperparameter combination. Finally, the medium-entropy steel performance is predicted based on the trained KAN network prediction model. The powerful learning and computing capabilities of the KAN network prediction model can be used to quickly and efficiently obtain the relationship between the performance and composition of medium-entropy steel, thereby realizing the prediction of the performance of medium-entropy steel, and solving the problem of time-consuming experimental research on the relationship between the performance and composition of entropy alloys in the prior art.

[0012] Preferably, the characteristic data set includes experimental data of a medium-entropy alloy composed of Fe, Cr, Ni matrix and Al and Ti elements with different molar fractions, and the experimental data includes the molar fractions of Al and Ti and the Vickers hardness values ​​corresponding to the molar fractions of Al and Ti.

[0013] Preferably, the KAN network prediction model includes:

[0014] An input layer, comprising two nodes, wherein the two nodes correspond to the molar fractions of Al and Ti respectively;

[0015] Several hidden layers, each containing several neurons;

[0016] The output layer contains a single node for predicting the Vickers hardness value.

[0017] Preferably, the medium entropy steel performance prediction system further includes:

[0018] The model evaluation module is used to evaluate the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel.

[0019] Preferably, the data processing module is also used to clean, screen and standardize the feature data set.

[0020] The present application also provides a medium entropy steel performance prediction method, which is applied to the medium entropy steel performance prediction system described in the first aspect, and comprises the following steps:

[0021] S1, obtain the characteristic data set of medium entropy steel performance;

[0022] S2, constructing a KAN network prediction model, and optimizing the KAN network prediction model through ESOA to obtain an optimal hyperparameter combination of the KAN network prediction model;

[0023] S3, training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination;

[0024] S4, predicting the performance of medium entropy steel according to the trained KAN network prediction model.

[0025] Preferably, before training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination, the feature data set is cleaned, screened and standardized.

[0026] Preferably, in S3, the specific method for training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination includes:

[0027] Adopt Adam optimizer and mean square error loss function;

[0028] Implement an early stopping mechanism to monitor the change in loss value on the validation set. If multiple iterations fail to improve the performance on the validation set, terminate the training early.

[0029] After each round of training, the mean square error on the training set and the test set is recorded.

[0030] Preferably, the medium entropy steel performance prediction method further includes:

[0031] S5, obtaining the mean square error and the determination coefficient of the KAN network prediction model, and evaluating the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel according to the mean square error and the determination coefficient;

[0032] S7, obtaining a learning rate curve of the KAN network prediction model, and evaluating the fitting degree of the KAN network prediction model according to the learning rate curve.

[0033] The present application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the medium-entropy steel performance prediction method described in the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0035] Figure 1It is a schematic diagram of the overall process of a method for predicting properties of medium entropy steel in one embodiment of the present invention.

[0036] Figure 2 It is a schematic diagram of the structure of a KAN network prediction model in one embodiment of the present invention.

[0037] Figure 3 It is a flowchart of the egret group optimization algorithm in one embodiment of the present invention.

[0038] Figure 4 It is a schematic diagram of the training process of the KAN network prediction model in one embodiment of the present invention.

[0039] Figure 5 It is a schematic diagram of a mean square error variation curve in one embodiment of the present invention.

[0040] Figure 6 It is a schematic diagram of a determination coefficient variation curve in one embodiment of the present invention.

[0041] Figure 7 Schematic diagram of a learning rate curve in one embodiment of the present invention.

[0042] Figure 8 It is a schematic diagram of a scatter plot in one embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the protection scope of the present invention.

[0044] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0046] In the traditional method of studying the relationship between the performance and composition of entropy alloys, each experiment requires multiple steps such as batching, smelting, heat treatment and testing, which takes a long time and is time-consuming. Therefore, it is necessary to develop an efficient prediction system and method for obtaining the relationship between the performance and composition of medium-entropy steel.

[0047] In order to solve the problem of time-consuming traditional experimental methods for studying the relationship between the performance and composition of entropy alloys, an embodiment of the present application provides a medium-entropy steel performance prediction system that can efficiently obtain the relationship between the performance and composition of medium-entropy steel.

[0048] The medium entropy steel performance prediction system includes a data processing module, a construction module, a training module and a prediction module.

[0049] The data processing module is used to obtain a feature data set of medium entropy steel performance. The medium entropy steel performance includes but is not limited to medium entropy steel hardness, tensile strength, elongation, impact toughness, yield strength and electrochemical corrosion. The feature data set includes input features and output targets corresponding to the input features, the input features are the molar fractions of Al and Ti, and the output targets are the medium entropy steel performance.

[0050] Preferably, the characteristic data set includes experimental data of a medium-entropy alloy composed of Fe, Cr, Ni matrix and Al and Ti elements with different molar fractions, and the experimental data includes the molar fractions of Al and Ti and the medium-entropy steel properties corresponding to the molar fractions of Al and Ti.

[0051] The data processing module is also used to clean, screen and standardize the feature data set. The feature database can come from an online public database. After obtaining the feature data set, it is cleaned and screened to remove noise data and outliers to ensure the quality of the data. The role of standardizing the feature data set is to eliminate the impact of data magnitude differences.

[0052] The construction module is used to construct a KAN network prediction model and obtain the optimal hyperparameter combination of the KAN network prediction model through ESOA (Egret Swarm Optimization Algorithm).

[0053] The KAN (Kolmogorov-Arnold) network is based on the Kolmogorov-Arnold representation theorem and can effectively represent any multivariate continuous function. Unlike the traditional multilayer perceptron (MLP), KAN places a learnable activation function on the edge instead of a fixed activation function on the node.

[0054] Kolmogorov-Arnold representation theorem: Any continuous multivariate function can be represented as a finite combination of one-dimensional functions and addition operations Among them, φ q,p and Φ q is the one-dimensional function to be learned.

[0055] In the implementation, the KAN network includes an input layer, several hidden layers and an output layer. The weight parameters of each layer are represented by a B-spline curve and learned. Specifically, the input layer includes two nodes, the two nodes correspond to the molar fractions of Al and Ti respectively; the several hidden layers, each of which includes several neurons; the output layer includes a single node, which is used to predict the performance of medium entropy steel.

[0056] ESOA optimizes problem solving by simulating the foraging behavior of egrets and has good global search capabilities. The fitness value of each egret is determined by the objective function. First, the parameters of the ESOA algorithm are initialized, including the population size, the maximum number of iterations, and the parameter boundary. Then, the fitness of each egret is evaluated, and the optimal egret of the current population is determined based on the fitness value. On this basis, the position and speed of the egret population are updated.

[0057] After each iteration, the fitness of the new position is evaluated and the optimal position is updated. If the predetermined number of iterations is reached or the convergence condition is met, the iteration is terminated and the position of the optimal egret is output as the final result. In this way, the ESOA algorithm can effectively find the optimal hyperparameter combination of the KAN network prediction model.

[0058] The training module is used to train the KAN network prediction model according to the feature data set and the optimal hyperparameter combination.

[0059] The prediction module is used to predict the performance of medium entropy steel based on the trained KAN network prediction model.

[0060] The medium-entropy steel performance prediction system constructs a KAN network prediction model, optimizes the structure and parameters of the KAN network prediction model through ESOA, obtains the optimal hyperparameter combination of the KAN network prediction model, and then trains the KAN network prediction model according to the feature data set and the optimal hyperparameter combination. Finally, the medium-entropy steel performance is predicted based on the trained KAN network prediction model. The powerful learning and computing capabilities of the KAN network prediction model can be used to quickly and efficiently obtain the relationship between the performance and composition of medium-entropy steel, thereby realizing the prediction of the performance of medium-entropy steel, and solving the problem of time-consuming experimental research on the relationship between the performance and composition of entropy alloys in the prior art.

[0061] In addition, existing technologies mainly rely on experimental methods to study the relationship between the performance and composition of entropy alloys, and have the following shortcomings: (1) The experimental cost is high. The experiment requires a lot of time and resources, including high-purity raw materials, precision instruments and professional technicians; (2) The data is limited. Due to the limitations of experimental conditions, the amount of data obtained is limited, and it is difficult to fully reflect the changing laws of alloy properties.

[0062] The medium-entropy steel performance prediction system described in this embodiment trains the KAN network prediction model according to the characteristic data set and the optimal hyperparameter combination, and finally predicts the medium-entropy steel performance according to the trained KAN network prediction model. By utilizing the powerful learning and computing capabilities of the KAN network prediction model, not only can the relationship between the performance and composition of medium-entropy steel be obtained quickly and efficiently, and the performance of medium-entropy steel can be predicted, but a large amount of data can also be quickly obtained, thereby reducing experimental costs and saving resources including high-purity raw materials, precision instruments and professional technicians.

[0063] The characteristic data set includes experimental data of a medium entropy alloy composed of Fe, Cr, Ni matrix and Al and Ti elements with different molar fractions, and the experimental data includes the molar fractions of Al and Ti and Vickers hardness values ​​corresponding to the molar fractions of Al and Ti.

[0064] As a preferred technical solution, the KAN network prediction model includes an input layer, several hidden layers and an output layer. The input layer includes two nodes, and the two nodes correspond to the molar fractions of Al and Ti respectively; the several hidden layers, each of which includes several neurons; and the output layer includes a single node, which is used to predict the performance of medium entropy steel.

[0065] As a preferred technical solution, the medium entropy steel performance prediction system also includes a model evaluation module.

[0066] The model evaluation module is used to evaluate the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel. Specifically, in order to quantify the prediction accuracy of the KAN network prediction model, the mean square error (MSE) and the coefficient of determination (R 2 ) as the evaluation index.

[0067] In addition, the model evaluation module also evaluates the fitting degree of the KAN network prediction model by acquiring the learning rate curve of the KAN network prediction model.

[0068] As a preferred technical solution, in the medium entropy steel performance prediction system, the FeCrNi(AlTi)x two-phase medium entropy alloy is taken as the research object, and the feature data set includes input features and output targets corresponding to the input features. The input features are the molar fractions of Al and Ti, and the output target is the Vickers hardness value of the medium entropy steel.

[0069] The data format of the feature data set is: {(x Al ,x Ti ,HV)}. Among them, x Al and x Ti They represent the mole fractions of Al and Ti respectively, and HV represents the corresponding Vickers hardness value.

[0070] The formula for standardizing the feature data set is: Among them, x is the original data, μ is the mean, σ is the standard deviation, and x′ is the standardized data. Specifically, the method for standardizing the feature data set is: standardizing the mole fraction and Vickers hardness of Al and Ti respectively.

[0071] In this implementation, if Figure 2 As shown, the KAN network prediction model includes an input layer, several hidden layers and an output layer. The weight parameters of each layer are represented by a B-spline curve and learned. Specifically, the input layer includes two nodes, corresponding to the molar fractions of Al and Ti respectively; each of the hidden layers includes several neurons, and the activation function of each neuron is represented by a learnable PReLU function; the output layer includes a single node for predicting the Vickers hardness value.

[0072] The formula of the PReLU activation function is: f(x) = max(0,x) + amin(0,x), where a is the parameter to be learned.

[0073] The training module is used to train the KAN network prediction model according to the feature data set and the optimal hyperparameter combination. The prediction module is used to predict the hardness of medium entropy steel according to the trained KAN network prediction model.

[0074] This embodiment optimizes the KAN network prediction model through ESOA, which can improve the accuracy of medium entropy steel hardness prediction, reduce test costs, and save time.

[0075] An embodiment of the present application also provides a medium entropy steel performance prediction method, which is applied to the medium entropy steel performance prediction system, such as Figure 1 As shown, it includes the following steps:

[0076] S1, obtaining a characteristic data set of medium entropy steel properties. The medium entropy steel properties include medium entropy steel hardness, tensile strength, elongation, impact toughness, yield strength, and electrochemical corrosion.

[0077] Preferably, before training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination, the feature data set is cleaned, screened and standardized.

[0078] Cleaning and screening the feature data set can remove noise data and outliers to ensure data quality. Standardizing the feature data set eliminates the impact of data magnitude differences and can improve the training effect of the KAN network prediction model.

[0079] S2, constructing a KAN network prediction model, and optimizing the KAN network prediction model through ESOA to obtain the optimal hyperparameter combination of the KAN network prediction model.

[0080] The KAN network prediction model includes an input layer, several hidden layers and an output layer. The weight parameters of each layer are represented by a B-spline curve and learned. Specifically, the input layer includes two nodes, and the two nodes correspond to the molar fractions of Al and Ti respectively; the several hidden layers, each of which includes several neurons; and the output layer includes a single node, which is used to predict the performance of medium entropy steel.

[0081] As a neural network with enhanced knowledge, KAN network can improve the prediction ability and stability of the model by introducing prior knowledge. Combined with ESOA, the structure and parameters of KAN network prediction model can be effectively optimized to improve the accuracy of medium entropy steel performance prediction.

[0082] Preferably, in step S2, Figure 3 As shown, the specific method of obtaining the optimal hyperparameter combination of the KAN network prediction model through ESOA includes:

[0083] S21, initializing ESOA algorithm parameters, wherein the parameters include population size, maximum number of iterations, and parameter boundaries.

[0084] S22, define the objective function as the mean square error (MSE) of the KAN network. ESOA optimizes the problem by simulating the foraging behavior of egrets and has good global search capabilities. The fitness value of each egret is determined by the objective function.

[0085] S23, iterative optimization parameters.

[0086] S24, evaluate the fitness value of each individual. Determine the optimal egret of the current population based on the fitness, and on this basis, update the position and speed of the egret population. The update formula is:

[0087] v i (t+1)=ωv i (t)+c1r1(p i -x i (t))+c2r2(g i -x i (t))

[0088] x i (t+1)=x i (t)+v i (t+1)

[0089] Among them, v i (t) and x i (t) represents the speed and position of the egret at the time; ω is the inertia weight; c1 and c2 are acceleration coefficients; r1 and r2 are random numbers between [0,1]; p i is the optimal individual position of the i-th egret; g i is the global optimal position.

[0090] S25, update the global optimal position and the individual optimal position, and determine whether the stop condition is met. If not, jump back to step S23, otherwise, enter step S26.

[0091] S26, stop iteration and output the best hyperparameter combination.

[0092] After each iteration, the fitness of the new position is evaluated and the optimal position is updated. If the predetermined number of iterations is reached or the convergence condition is met, the iteration is terminated and the position of the optimal egret is output as the final result. In this way, the ESOA algorithm can effectively find the optimal hyperparameter combination of the KAN network prediction model.

[0093] The optimal hyperparameter combination includes learning rate, batch size, number of training rounds, and the number of neurons in each hidden layer.

[0094] The optimal hyperparameter combination of the KAN network prediction model is obtained through ESOA, which significantly improves the prediction accuracy and computational efficiency.

[0095] S3: training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination.

[0096] Preferably, in S3, if Figure 4As shown, the specific method for training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination includes:

[0097] S31, using the feature data set and the optimal hyperparameter combination to train the KAN network prediction model, initialize the parameters, use the Adam optimizer and the mean square error loss function, and calculate the mean square error loss through the mean square error loss function.

[0098] The parameter update formula of the Adam optimizer is: Among them, η is the learning rate and are the estimates of the first-order moment and the second-order moment of the gradient, and ∈ is the smoothing term.

[0099] S32, implements the early stopping mechanism to monitor the change of loss value on the validation set. If multiple iterations fail to improve the performance on the validation set, the training is terminated early.

[0100] Specifically, during the training process, the changes in the loss value on the validation set are monitored. If multiple iterations fail to improve the performance on the validation set, the training is terminated early to prevent overfitting and enhance the robustness and generalization ability of the KAN network prediction model.

[0101] S33, after each round of training, the mean square error on the training set and the test set is recorded for subsequent evaluation of model performance.

[0102] S4, predicting the performance of medium entropy steel according to the trained KAN network prediction model.

[0103] The medium-entropy steel performance prediction method trains the KAN network prediction model through the characteristic data set and the optimal hyperparameter combination, and predicts the performance of medium-entropy steel according to the trained KAN network prediction model. The powerful learning ability and computing power of the KAN network prediction model are utilized, and a large amount of resources and time do not need to be spent on experiments. The acquisition of data is not limited by experimental conditions. It can not only quickly and efficiently obtain the relationship between the performance and composition of medium-entropy steel, and realize the prediction of medium-entropy steel performance, but also quickly obtain a large amount of data, reduce experimental costs, and save resources including high-purity raw materials, precision instruments, and professional technicians.

[0104] In one embodiment, the medium entropy steel performance prediction method further includes:

[0105] S5, obtaining the mean square error (MSE) and the determination coefficient of the KAN network prediction model, and evaluating the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel based on the mean square error and the determination coefficient.

[0106] Mean Square Error:

[0107] Among them, y i is the actual performance value, is the prediction performance value, and N is the number of samples. A lower MSE value indicates that the error between the prediction performance value and the actual value is smaller.

[0108] Coefficient of determination: is the mean of the actual hardness values. 2 The closer the value is to 1, the better the prediction performance of the model.

[0109] S7, obtaining a learning rate curve of the KAN network prediction model, and evaluating the fitting degree of the KAN network prediction model according to the learning rate curve.

[0110] Preferably, the medium-entropy steel performance prediction method also includes using the trained KAN network prediction model to predict the hardness, tensile strength, elongation, impact toughness, yield strength, electrochemical corrosion and other properties of the medium-entropy steel and generate a scatter plot comparison of the true value and the predicted value, and evaluate the performance of the KAN network prediction model by drawing a learning curve and an error curve.

[0111] By evaluating the accuracy and fit of the KAN network prediction model, the prediction ability of the medium-entropy steel property prediction method can be more intuitively seen, and the performance of the KAN network prediction model can be evaluated.

[0112] In one of the embodiments, in the medium entropy steel performance prediction method, the FeCrNi(AlTi)x two-phase medium entropy alloy is taken as the research object, and the feature data set includes input features and output targets corresponding to the input features, the input features are the molar fractions of Al and Ti, and the output target is the Vickers hardness value of the medium entropy steel.

[0113] The data format of the feature data set is: {(x Al ,x Ti ,HV)}. Among them, x Al and x Ti They represent the mole fractions of Al and Ti respectively, and HV represents the corresponding Vickers hardness value.

[0114] The formula for standardizing the feature data set is: Among them, x is the original data, μ is the mean, σ is the standard deviation, and x′ is the standardized data. Specifically, the method for standardizing the feature data set is: standardizing the mole fraction and Vickers hardness of Al and Ti respectively.

[0115] In step S4, the hardness of medium entropy steel is predicted according to the trained KAN network prediction model.

[0116] In step S5, the mean square error (MSE) and the determination coefficient of the KAN network prediction model are obtained, and the accuracy of the KAN network prediction model in predicting the hardness of medium entropy steel is evaluated based on the mean square error and the determination coefficient.

[0117] The mean square error formula is:

[0118] Among them, y i is the actual performance value, is the prediction performance value, and N is the number of samples. A lower MSE value indicates that the error between the predicted hardness value and the actual hardness value is smaller.

[0119] MSE change curve is as follows Figure 5 As shown. Figure 5 It can be seen that as the number of training rounds increases, the mean square error decreases rapidly and stabilizes after about 25 rounds, indicating that the model quickly learns and effectively fits the training data in the initial stage.

[0120] Coefficient of determination: is the mean of the actual hardness values. 2 The closer the value is to 1, the better the prediction performance of the model.

[0121] R 2 The change curve is as Figure 6 shown. Figure 6 Shows R 2 How the score changes during the training process. 2 The score rises rapidly in the early stage and then stabilizes at a high value close to 1, indicating that the model has strong explanatory power and fitting effect.

[0122] In step S7, a learning rate curve of the KAN network prediction model is obtained, and the fitting degree of the KAN network prediction model is evaluated according to the learning rate curve.

[0123] like Figure 7 As shown in the figure, it can be seen from the learning curve that with the increase of the number of training iterations (Epoch), both the training error and the test error decrease significantly and stabilize after about 40 Epochs. Finally, both the training error and the test error converge to a lower level, indicating that the model performs well on both the training set and the test set, without overfitting or underfitting.

[0124] from Figure 7From the learning curve, we can see that the training error and the test error both drop rapidly in the early stage and remain low and stable in the later stage, indicating that the model has not been overfitted and has good generalization performance.

[0125] Preferably, the medium entropy steel performance prediction method also includes using a trained KAN network prediction model to predict the hardness of the medium entropy steel and generate a scatter plot comparison of the true value and the predicted value to evaluate the performance of the KAN network prediction model.

[0126] like Figure 8 As shown, a scatter plot of the actual hardness value and the predicted hardness value is shown, showing the relationship between the hardness value predicted by the KAN network prediction model and the actual hardness value. Figure 8 In the figure, the green dots represent the predicted hardness values, and the blue dots represent the actual hardness values. As can be seen from the figure, the predicted hardness values ​​and the actual hardness values ​​are roughly distributed on a diagonal line, and the predicted hardness values ​​and the true hardness values ​​show a high linear correlation, which verifies the high prediction accuracy of the model.

[0127] In summary, the medium-entropy steel performance prediction method based on the KAN network prediction model optimized by the egret swarm optimization algorithm performs well in the prediction of medium-entropy steel hardness. By accurately tuning the hyperparameters, the prediction performance of the model is significantly improved. This method not only has an efficient training speed, but also prevents overfitting during the training process, and ultimately achieves accurate prediction of medium-entropy steel performance. This method provides a new and efficient tool for the prediction and optimization of metal and alloy properties in the field of materials science.

[0128] In addition, the medium-entropy steel performance prediction method is suitable for the performance prediction of FeCrNi(AlTi)x alloy and other medium-entropy alloys, has wide applicability and powerful data processing capabilities, and has the characteristics of high precision, high computational efficiency, strong applicability, and significant visualization effect.

[0129] In one of the embodiments, the present application also provides a computer-readable storage medium storing a computer program, which implements the medium-entropy steel performance prediction method when the computer program is executed by a processor.

[0130] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A medium entropy steel performance prediction system, characterized in that: The medium entropy steel performance prediction system comprises: A data processing module, used to obtain characteristic data sets of medium entropy steel performance; A construction module is used to construct a KAN network prediction model, and optimize the KAN network prediction model through ESOA to obtain an optimal hyperparameter combination of the KAN network prediction model; A training module, used for training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination; The prediction module is used to predict the performance of medium entropy steel according to the trained KAN network prediction model.

2. A medium entropy steel performance prediction system as claimed in claim 1, characterized in that: The characteristic data set includes experimental data of a medium entropy alloy composed of Fe, Cr, Ni matrix and Al and Ti elements with different molar fractions, and the experimental data includes the molar fractions of Al and Ti and Vickers hardness values ​​corresponding to the molar fractions of Al and Ti.

3. A medium entropy steel performance prediction system as claimed in claim 2, characterized in that: The KAN network prediction model includes: An input layer, comprising two nodes, wherein the two nodes correspond to the molar fractions of Al and Ti respectively; Several hidden layers, each containing several neurons; The output layer contains a single node for predicting the Vickers hardness value.

4. A medium entropy steel performance prediction system as claimed in claim 1, characterized in that: The medium entropy steel performance prediction system also includes: The model evaluation module is used to evaluate the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel.

5. A medium entropy steel performance prediction system as claimed in claim 1, characterized in that: The data processing module is also used to clean, screen and standardize the feature data set.

6. A medium entropy steel performance prediction method, applied to the medium entropy steel performance prediction system according to any one of claims 1 to 5, characterized in that: The medium entropy steel performance prediction method comprises the following steps: S1, obtain the characteristic data set of medium entropy steel performance; S2, constructing a KAN network prediction model, and optimizing the KAN network prediction model through ESOA to obtain an optimal hyperparameter combination of the KAN network prediction model; S3, training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination; S4, predicting the performance of medium entropy steel according to the trained KAN network prediction model.

7. A method for predicting properties of medium entropy steel according to claim 6, characterized in that: Before training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination, the feature data set is cleaned, screened and standardized.

8. A method for predicting properties of medium entropy steel according to claim 6, characterized in that: In S3, the specific method for training the KAN network prediction model according to the feature data set and the optimal hyperparameter combination includes: Adopt Adam optimizer and mean square error loss function; Implement an early stopping mechanism to monitor the change in loss value on the validation set. If multiple iterations fail to improve the performance on the validation set, terminate the training early. After each round of training, the mean square error on the training set and the test set is recorded.

9. A method for predicting properties of medium entropy steel according to any one of claims 6 to 8, characterized in that: The medium entropy steel performance prediction method also includes: S5, obtaining the mean square error and the determination coefficient of the KAN network prediction model, and evaluating the accuracy of the KAN network prediction model in predicting the properties of medium entropy steel according to the mean square error and the determination coefficient; S7, obtaining a learning rate curve of the KAN network prediction model, and evaluating the fitting degree of the KAN network prediction model according to the learning rate curve.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the medium-entropy steel performance prediction method as described in any one of claims 6 to 9 is implemented.

Images