Construction method and device of prediction model for fatigue performance of high-entropy alloy, equipment, medium and product

Abstract

The invention discloses a construction method and device for a prediction model of fatigue performance of a high-entropy alloy, equipment, a medium and a product. The construction method comprises the following steps: acquiring a feature data sample set and an initial prediction model of the high-entropy alloy; according to an attention module and a feature data sample set in the initial prediction model, determining global correlation features among the components in the high-entropy alloy; according to the global correlation features and a hybrid module in the initial prediction model, determining fusion time sequence features, the hybrid module comprising a long and short term memory sub-module and a gating cycle unit sub-module; according to a full connection layer and fusion features in the initial prediction model, determining a predicted value and a loss function of fatigue performance; and carrying out iterative training on the initial prediction model based on the loss function until a training ending condition is met, and obtaining a final prediction model. Important features are screened and strengthened through the attention module, potential time sequence dependence in heat treatment parameters and microscopic structure features is processed through the mixing module, and the accuracy of the prediction model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material prediction, and in particular to a method, device, equipment, medium and product for constructing a prediction model for fatigue performance of a high-entropy alloy. Background Art

[0002] High-entropy alloy (HEA) is a new type of metallic material composed of multiple elements in near-equiatomic proportions. This design concept differs from traditional alloys, and its notable features include high strength, high toughness, excellent corrosion resistance, and high-temperature stability, making it an important application prospect in aerospace, energy, nuclear industry, and other fields. However, HEA needs to withstand complex cyclic loads during service, making fatigue failure one of its key performance indicators. Therefore, accurately predicting the fatigue performance of HEA is of great significance for material design, process optimization, and practical application.

[0003] At present, the research on fatigue properties of high entropy alloys mainly focuses on the following aspects: First, the method based on experimental testing, which usually includes the following steps: preparing standardized specimens; applying specific loads on fatigue testing equipment; recording the fatigue life of the material (such as the number of cycles). Second, the numerical simulation method: finite element analysis and other methods simulate the material properties under fatigue loads. This method combines theoretical models and experimental data and is suitable for performance prediction under certain conditions. Third, machine learning methods: in recent years, algorithms such as decision trees, support vector machines, and random forests have gradually been applied to performance prediction. By training existing experimental data, a prediction model for fatigue performance is established.

[0004] However, the first approach requires significant time and resources, especially for high-entropy alloys with multiple compositional combinations. Experimental conditions are limited, making it difficult to cover all possible operating environments. Experimental unpredictability makes it difficult to quickly evaluate performance under untested conditions. For the second approach, traditional machine learning models are inadequate for characterizing nonlinear and multivariate coupling characteristics. Furthermore, they struggle to extract effective features from high-dimensional data, and traditional models have limited ability to model time series dependencies, as fatigue performance changes dynamically with test conditions. Summary of the Invention

[0005] The present invention provides a method, device, equipment, medium and product for constructing a prediction model for the fatigue performance of high-entropy alloys, so as to increase the prediction model's mining of potential temporal dependencies in high-entropy alloy data features and improve the accuracy of the prediction model.

[0006] According to a first aspect of the present invention, a method for constructing a prediction model for fatigue performance of a high entropy alloy is provided, the method comprising:

[0007] Obtaining a characteristic data sample set and an initial prediction model for high entropy alloys;

[0008] Determining global correlation features between components in the high-entropy alloy based on the attention module in the initial prediction model and the feature data sample set;

[0009] Determining a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, wherein the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule;

[0010] Determining a predicted value and a loss function of fatigue performance of the high-entropy alloy based on the fully connected layer and the fusion features in the initial prediction model;

[0011] The initial prediction model is iteratively trained based on the loss function until a training end condition is met to obtain a final prediction model.

[0012] According to a second aspect of the present invention, there is provided a device for constructing a prediction model for fatigue performance of a high entropy alloy, comprising:

[0013] A data acquisition module is used to obtain a characteristic data sample set and an initial prediction model for high entropy alloys;

[0014] A first determination module is configured to determine a global correlation feature between components in the high entropy alloy based on the attention module in the initial prediction model and the feature data sample set;

[0015] A second determination module is configured to determine a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, wherein the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule;

[0016] A third determination module is used to determine the predicted value and loss function of the fatigue performance of the high entropy alloy based on the fully connected layer and the fusion feature in the initial prediction model;

[0017] The model determination module is used to iteratively train the initial prediction model based on the loss function until the training end condition is met to obtain a final prediction model.

[0018] According to a third aspect of the present invention, there is provided an electronic device, comprising:

[0019] at least one processor; and

[0020] a memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for constructing a prediction model for fatigue performance of high entropy alloys described in any embodiment of the present invention.

[0022] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for constructing a prediction model for fatigue performance of a high-entropy alloy as described in any embodiment of the present invention when executed.

[0023] According to the fifth aspect of the present invention, an embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the method for constructing a prediction model for the fatigue performance of a high-entropy alloy according to any embodiment of the present invention.

[0024] The technical solution of the embodiment of the present invention obtains a feature data sample set and an initial prediction model of a high-entropy alloy; determines the global correlation characteristics between the components in the high-entropy alloy based on the attention module and the feature data sample set in the initial prediction model; determines the fused temporal characteristics based on the global correlation characteristics and the hybrid module in the initial prediction model, the hybrid module including the long-short-term memory submodule and the gated recurrent unit submodule; determines the predicted value and loss function of fatigue performance based on the fully connected layer and fused characteristics in the initial prediction model; iteratively trains the initial prediction model based on the loss function until the training end condition is met, thereby obtaining the final prediction model. The attention module selects and strengthens important features, and the hybrid module processes the potential temporal dependencies in heat treatment parameters and microstructural characteristics, thereby improving the accuracy of the prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 1 is a schematic structural diagram of a method for constructing a prediction model for fatigue performance of a high entropy alloy provided in accordance with the first embodiment of the present invention;

[0026] Figure 2 This is an example diagram of a loss curve in a method for constructing a prediction model for fatigue performance of a high entropy alloy provided in Example 1 of the present invention;

[0027] Figure 3 This is a comparison chart of predicted values in a method for constructing a prediction model for fatigue performance of a high entropy alloy provided in Example 1 of the present invention;

[0028] Figure 4 2 is a schematic structural diagram of a device for constructing a prediction model for fatigue performance of a high entropy alloy according to a second embodiment of the present invention;

[0029] Figure 5 It is a schematic structural diagram of an electronic device implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0031] Example 1

[0032] Figure 1 A flowchart of a method for constructing a prediction model of high entropy alloy fatigue performance is provided for the first embodiment of the present invention. This embodiment is applicable to the construction of a prediction model of high entropy alloy fatigue performance. The method can be executed by a device for constructing a prediction model of high entropy alloy fatigue performance. The device for constructing a prediction model of high entropy alloy fatigue performance can be implemented in the form of hardware and / or software. The device for constructing a prediction model of high entropy alloy fatigue performance can be configured in an electronic device. Figure 1 As shown, the method includes:

[0033] S110. Obtain a characteristic data sample set and an initial prediction model for a high entropy alloy.

[0034] In this embodiment, the feature data sample set can be understood as a set containing multiple feature data training samples for model training, for example, a public dataset Kaggle can be used. The initial prediction model can be understood as an untrained prediction model.

[0035] Among them, each feature data training sample in the feature data sample set includes the chemical composition, heat treatment parameters, microstructure characteristics and fatigue properties of the high entropy alloy.

[0036] For example, the characteristic data sample may include chemical composition (C, Si, Mn, Ni, Cr, Cu, Mo), heat treatment parameters (NT (normalizing temperature), THT (normalizing temperature), Ct (carburization time), Dt (diffusion time), Tt (tempering time)), microstructural characteristics (RedRatio (reduction ratio), dA (area proportion of inclusions deformed by plastic work), dB (area proportion of inclusions occurring in discontinuous array), dC (area proportion of isolated inclusions)), and fatigue properties (rotating bending fatigue strength (10^7 cycles)).

[0037] Specifically, the processor can obtain a characteristic data sample set and an initial prediction model of the high entropy alloy.

[0038] S120. Determine the global correlation characteristics between the components in the high entropy alloy based on the attention module and the feature data sample set in the initial prediction model.

[0039] In this embodiment, the attention module can be understood as a module that uses multiple attention heads to focus on different parts of the input, capture more diverse information, and improve the model's representation capabilities. For example, it can be the multi-head attention mechanism in the Transformer model. The global correlation feature can be understood as a global correlation feature associated with the composition of the high-entropy alloy.

[0040] Specifically, the processor can preprocess the feature data sample set, divide it into a training set and a test set, and input the processed feature data samples into the initial prediction model. The attention module in the initial prediction model is combined with the constraint characteristics between the components in the high-entropy alloy to determine the global correlation characteristics between the components in the high-entropy alloy.

[0041] S130 , determining a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, where the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule.

[0042] In this embodiment, a hybrid module can be understood as a module that is mixed with submodules. For example, it can include a long-short-term memory submodule for extracting long-term dependencies in time series, and a gated recurrent unit submodule for reducing the number of parameters and improving computational efficiency. Fusion of temporal features can be understood as fusing features that incorporate potential temporal dependencies between heat treatment parameters and microstructural features.

[0043] Specifically, the processor can input the obtained global correlation features into the hybrid module in the initial prediction model, process the global correlation features respectively through the long short-term memory sub-module and the gated recurrent unit sub-module in the hybrid module, obtain the results output by the two sub-modules and fuse them to obtain the final fused time series features.

[0044] S140. Determine the predicted value and loss function of the fatigue performance of the high entropy alloy based on the fully connected layer and fusion features in the initial prediction model.

[0045] In this embodiment, a fully connected layer can be understood as a layer that connects all neurons in the previous layer of the neural network to every neuron in the current layer to obtain the final output. The predicted value can be understood as a prediction of the fatigue performance of the high-entropy alloy. The loss function can be understood as a function that represents the difference between the predicted value and the true value.

[0046] Specifically, the fusion features are reduced in dimension through the fully connected layer in the initial prediction model to generate a predicted fatigue performance value, i.e., a predicted value. At the same time, the loss function can be determined by the predicted value.

[0047] S150. Iteratively train the initial prediction model based on the loss function until the training end condition is met to obtain the final prediction model.

[0048] In this embodiment, the training end condition can be understood as the deviation of the predicted value reaching the expected value, or the number of iterations has reached the maximum. The final prediction model can be understood as a model that meets the requirements after training.

[0049] Specifically, when the loss function does not meet the training end condition, the processor can adjust the parameters in the initial prediction model and perform iterative training until the loss function reaches the training end condition or the number of iterations reaches the training end condition, thereby obtaining the final prediction model.

[0050] The technical solution of the embodiment of the present invention obtains a feature data sample set and an initial prediction model of a high-entropy alloy; determines the global correlation characteristics between the components in the high-entropy alloy based on the attention module and the feature data sample set in the initial prediction model; determines the fused temporal characteristics based on the global correlation characteristics and the hybrid module in the initial prediction model, the hybrid module including the long-short-term memory submodule and the gated recurrent unit submodule; determines the predicted value and loss function of fatigue performance based on the fully connected layer and fused characteristics in the initial prediction model; iteratively trains the initial prediction model based on the loss function until the training end condition is met, thereby obtaining the final prediction model. The attention module selects and strengthens important features, and the hybrid module processes the potential temporal dependencies in heat treatment parameters and microstructural characteristics, thereby improving the accuracy of the prediction model.

[0051] Furthermore, based on the above embodiment, the steps of determining the global correlation features between the components in the high entropy alloy according to the attention module and the feature data sample set in the initial prediction model can be refined as follows:

[0052] Each feature data training sample in the feature data sample set is preprocessed to obtain a preprocessed training sample; the preprocessed training sample is linearly transformed based on the attention module in the initial prediction model to obtain an output feature space; the attention output is determined according to the output feature space and the chemical composition interaction weight matrix related to the high entropy alloy; and the global correlation feature is determined based on the attention output.

[0053] In this embodiment, the preprocessed training samples can be understood as the processed training samples. The output feature space can be understood as the vector space obtained after linear transformation. The attention output can be understood as the attention score obtained after normalization.

[0054] Specifically, the processor can perform data cleaning and preprocessing on each feature data training sample in the feature data sample set, remove noise data and outliers to ensure the quality of the data, and obtain preprocessed training samples. Based on the attention module in the initial prediction model (such as the multi-head attention mechanism in Transformer), the preprocessed training samples are linearly transformed to obtain three output feature spaces; the initial prediction model can calculate the attention score and scale it according to the output feature space through dot product calculation, and introduce a physical constraint attention mechanism, that is, the chemical composition interaction weight matrix related to high entropy alloys, to determine the attention output. The attention output is then normalized by Softmax, and the attention output is converted into weights to determine the global correlation features.

[0055] Exemplary methods for data cleaning and preprocessing may include: Outlier processing: Use statistical methods (such as box plots or 3σ criteria) to identify outliers and remove or interpolate them. Unit unification: Ensure that the units of heat treatment parameters are consistent (such as degrees Celsius and seconds). Missing value filling: For samples with missing data such as chemical composition and heat treatment parameters, mean filling and regression prediction filling can be used. Feature normalization: Normalize or standardize all input features to ensure that features of different dimensions have a consistent numerical range. The formula is as follows:

[0056]

[0057] Among them, X norm is the original data, μ is the mean, and σ is the standard deviation.

[0058] For example, the global correlation between high entropy alloy components can be extracted from the preprocessed training samples through the Transformer multi-head attention mechanism of the initial prediction model. First, the input feature data is mapped into three output feature spaces Query (Q), Key (K) and Value (V) through linear transformation:

[0059] Q=XW Q

[0060] K=XW K

[0061] V=XW V

[0062] Where X is the input feature matrix, W Q , W K , W V It is the learned weight matrix that maps the output features to the output feature space of Query, Key and Value respectively.

[0063] Then, the similarity between the query and the key is calculated, and the attention score is obtained and scaled by dot product calculation. In addition, a physical constraint attention mechanism is introduced, and the known chemical composition interaction (such as the effect of Ni and Cr interaction on fatigue performance) is introduced, that is, the chemical composition interaction weight matrix related to high entropy alloys is introduced. An additional weighting matrix is added when calculating the attention distribution:

[0064]

[0065] in, It is a scaled form of the dot product to ensure that the calculation result is not too large, d k is the dimension of query and key, and M is the chemical composition interaction weight matrix. After introducing physical constraint features, the model can more effectively model complex material systems and reduce overfitting of meaningless features.

[0066] Softmax is used to normalize the calculation results, converting attention scores into weights to control the degree of attention paid to the relationships between different components. Furthermore, through the sparse attention mechanism, only the attention interactions between highly correlated features are retained, reducing computational complexity and improving model efficiency. A multi-head attention mechanism is then used to parallelize the attention outputs of multiple heads and concatenate them:

[0067] MultiHead(Q,K,V)=Concat(head1,head2,…,head h )W O

[0068] Among them, h is the number of heads, and the calculation results of each head will be combined with the weight matrix W O Finally, the attention output is normalized to maintain a stable training process.

[0069] Furthermore, based on the above embodiment, the step of determining the fused time series features based on the global correlation features and the hybrid module in the initial prediction model can be refined as follows:

[0070] Determine whether the dimension of the global correlation feature meets the module requirements; if so, process the global correlation feature based on the hybrid module in the initial prediction model to determine the fused time series feature; if not, perform a linear transformation on the global correlation feature to obtain the input feature that meets the module requirements, input the input feature into the hybrid module to obtain the fused time series feature.

[0071] In this embodiment, the module requirement can be understood as whether the dimension of the input feature is consistent with the dimension of the feature processed by the hybrid module. The input feature can be understood as a feature whose dimension meets the requirement.

[0072] Specifically, the initial prediction model can determine whether the dimension of the global correlation feature meets the module requirements. If so, the global correlation feature is processed based on the hybrid module in the initial prediction model to determine the fused time series feature; if not, the global correlation feature is linearly transformed, and the dimension of the global correlation feature is adjusted to the dimension that meets the module requirements to obtain the input feature, and the input feature is input into the hybrid module for processing to obtain the fused time series feature.

[0073] Exemplarily, the global correlation feature is a set of global feature vectors, usually of shape (N, T, D), where N is the batch size, T is the time step, and D is the feature dimension. The global correlation feature is then used as the input of the hybrid module. Here, the global correlation feature can be directly used as the time series data of the hybrid module when the dimension matches the hybrid module dimension. If the dimension does not match, the global feature vector needs to be adjusted to the dimension required by the hybrid module through the adaptation layer. The above D can be adjusted using the following linear transformation formula:

[0074] x lstm =W t ·x Transformer +b t

[0075] Where W t is the weight of the linear transformation, b t For paranoid items.

[0076] The step of processing the global correlation features based on the hybrid module in the initial prediction model to determine the fused time series features may include:

[0077] The long-term dependency features of the global correlation features are extracted through the long-short-term memory submodule to obtain the first hidden state; the short-term dynamic features of the global correlation features are extracted through the gated recurrent unit submodule to obtain the second hidden state; the first hidden state and the second hidden state are fused according to the preset fusion algorithm to obtain the fused temporal features.

[0078] In this embodiment, the first hidden state can be understood as the result obtained after processing by the long short-term memory submodule. The second hidden state can be understood as the result obtained after processing by the gated recurrent unit submodule. The preset fusion algorithm can be understood as being used for fusing two features, for example, including splicing fusion and weighted fusion.

[0079] Specifically, the global correlation features / the above-mentioned input features are simultaneously input into the long short-term memory sub-module and the gated recurrent unit sub-module, and the long-term dependency features of the global correlation features are extracted through the long short-term memory sub-module to obtain a first hidden state; the short-term dynamic features of the global correlation features are extracted through the gated recurrent unit sub-module to obtain a second hidden state; the first hidden state and the second hidden state are fused according to a preset fusion algorithm to obtain a fused temporal feature.

[0080] For example, at each time step passed to the hybrid module network, the hybrid module captures the dynamic changes of fatigue performance from the time series data and calculates the forget gate (f t ), output gate (i t ), candidate memory units Update memory unit (C t), output gate (o t ), update gate (z t ) and reset gate (r t ).

[0081] The specific calculation formula is as follows:

[0082] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0083] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0084]

[0085] o t =σ(W o ·[h h-1 ,x t ]+b o )

[0086] h t =o t tanh(C t )

[0087] z t =σ(W z ·[h t-1 ,x t ]+b z )

[0088] r t =σ(W r ·[h t-1 ,x t ]+b r )

[0089] Where W f 、W i 、W c 、W o 、W z 、W r and b f 、b i 、b C 、b o 、b z 、b r are the learned weights and biases that control the flow of information. t-1is the hidden state of the previous time step. σ is the Sigmoid activation function, which is used to map the result to [0,1].

[0090] Then the output hidden state of the hybrid module is fused by weighting or splicing to form a unified time series feature h t fused , which is used to predict the fatigue performance of the subsequent fully connected layer. The formula is as follows:

[0091] Splicing and fusion:

[0092] h t fused =Concat(h t LSTM ,h t GRU )

[0093] Weighted fusion:

[0094] h t fused =α·h t LSTM +(1-α)·h t GRU

[0095] where h t LSTM represents the first hidden state of the long short-term memory submodule, h t GRU Represents the second hidden state of the gated recurrent unit submodule, Concat represents the concatenation of two hidden state vectors to form a higher-dimensional fusion feature. α represents a learnable weighting parameter with a value range of [0,1], which represents the weight ratio of the long short-term memory submodule and the gated recurrent unit submodule. Finally, the fused feature h is quantized by dimensionality reduction method. t fused Optimization is performed to preserve key time step characteristics and provide high-quality input for fatigue performance prediction.

[0096] Furthermore, based on the above embodiment, the steps of determining the predicted value and loss function of the fatigue performance of the high-entropy alloy according to the fully connected layer and fusion features in the initial prediction model can be refined as follows:

[0097] Based on the attention mechanism of the initial prediction model, the time step weight of each time step under the fusion feature is determined; the weighted summary feature is determined based on the weight of each time step and the fusion feature; the weighted summary feature is processed according to the fully connected layer in the initial prediction model to determine the predicted value of the fatigue performance of the high entropy alloy; and the loss function is determined based on the predicted value and the true value in the feature data sample.

[0098] In this embodiment, a time step can be understood as each time point in the sequence data of the fused feature. A time step weight can be understood as the weight of each time point. A weighted summary feature can be understood as the weighted feature. A true value can be understood as the true fatigue performance.

[0099] Specifically, the attention mechanism of the initial prediction model is used to determine the time step weight of each time step under the fusion feature; the weighted summary feature is determined based on the weight of each time step and the fusion feature; the weighted summary feature is processed according to the fully connected layer in the initial prediction model to determine the predicted value of the fatigue performance of the high entropy alloy; and the loss function is determined based on the predicted value and the true value in the feature data sample.

[0100] For example, first, the output fusion feature h of the hybrid module network t (usually a high-dimensional vector) is passed to the fully connected layer. Then, an attention mechanism is added to the output layer of the hybrid module to calculate the weight of each time step and emphasize the influence of key time steps:

[0101] α t =Softmax(W a h t +b a )

[0102]

[0103] Among them, α t is the weight of the t-th time step, h w is the weighted summary feature.

[0104] Then, the fully connected layer performs a linear transformation, which is to transform h t Perform dimensionality reduction:

[0105] y=W fc ·h t +b fc

[0106] Among them, W fc is the weight matrix of the fully connected layer, b fc is the bias term, and y is the final predicted value. The function of the fully connected layer is to transform the high-dimensional time series feature vector h t Mapped to a scalar output y, representing the fatigue performance value. Finally, since this is a regression task, an activation function is usually not used, and the output value is the predicted value of fatigue performance predicted by the model.

[0107] The technical solution of the embodiment of the present invention obtains a set of characteristic data samples of high-entropy alloys and an initial prediction model. The characteristic data samples include the comprehensive effects of chemical composition, heat treatment parameters, and microstructural characteristics on fatigue performance. The attention module performs a linear transformation on the processed training samples to obtain an output feature space. Then, combined with the chemical composition interaction weight matrix, the global correlation features are determined. A physical constraint attention mechanism is introduced to reduce overfitting of unobjectionable features and retain only the attention interactions between highly correlated features, thereby reducing computational complexity and improving model efficiency. By processing the dimensions of the global correlation features to meet the module requirements and inputting them into the hybrid module, the long-term short-term memory submodule processes the long-term dependencies of the time series, and the gated recurrent unit submodule processes short-term dynamic changes. This not only retains the efficient modeling capability of the long-term short-term memory submodule, but also reduces the number of parameters and improves computational efficiency. The hidden states processed by the two submodules are then fused to obtain a fused time series feature, providing high-quality input for fatigue performance prediction. Finally, the final prediction result and loss function are obtained through a fully connected layer. The initial prediction model is iteratively trained based on the loss function to obtain the final prediction model. This method effectively improves the accuracy of fatigue performance prediction for high-entropy alloys. Compared with existing technologies, this method reduces prediction costs and computational complexity, while enhancing the model's prediction accuracy and generalization capabilities, achieving rapid prediction of fatigue performance for high-entropy alloys with high accuracy, low cost, and wide applicability.

[0108] For example, taking the final prediction model as an example, Figure 2 This is an example diagram of a loss curve in a method for constructing a prediction model for fatigue performance of a high entropy alloy according to the first embodiment of the present invention. The loss curve is shown in FIG. Figure 2 As shown in the figure, the blue curve represents the loss value of the model on the training dataset, which reflects the degree of fit of the model on the training data. The loss function is used to calculate the error between the predicted value and the true value. The downward curve indicates that the model is gradually learning the characteristics of the training data. The orange curve represents the loss value of the model on the validation dataset, which is used to evaluate the performance of the model on unseen data. The validation set is an independent part divided from the original data and does not participate in the learning of the model during the training process. If the validation loss decreases with the training rounds, it means that the generalization performance of the model on the validation set has improved. The horizontal axis represents the training rounds, and each epoch represents a complete traversal of the entire training set by the model. The vertical axis represents the loss value, which is the error between the model prediction and the target value. The smaller the value, the closer the model prediction is to the true value.

[0109] Figure 3 This is a comparison diagram of predicted values in a method for constructing a prediction model for fatigue performance of a high entropy alloy provided in Example 1 of the present invention. The comparison diagram of predicted values and true values is shown in FIG. Figure 3As shown in the figure, the comparison between the predicted value of the optimization model and the actual value is shown. The horizontal axis of each blue point is the actual fatigue performance value (rotational bending fatigue strength of the material, with a value of 10 7 The vertical axis represents the predicted fatigue performance value. The middle red dashed line represents the ideal state (indicating that the predicted value is equal to the true value), and the blue dots indicate the accuracy of the fatigue performance prediction. The closer the dots are to the middle red dashed line, the more accurate the model prediction. In summary, judging by the loss function evaluation indicators and scatter plots, the composite algorithm performs satisfactorily in predicting the fatigue performance of high-entropy alloys, demonstrating high prediction speed and stability.

[0110] Example 2

[0111] Figure 4 This is a schematic diagram of a device for constructing a prediction model for fatigue performance of a high entropy alloy provided in the second embodiment of the present invention. Figure 4 As shown, the device includes: a data acquisition module 41, a first determination module 42, a second determination module 43, a third determination module 44 and a model determination module 45.

[0112] The data acquisition module 41 is used to obtain a characteristic data sample set and an initial prediction model of a high entropy alloy;

[0113] A first determination module 42 is configured to determine a global correlation feature between components in the high entropy alloy based on the attention module in the initial prediction model and the feature data sample set;

[0114] A second determining module 43 is configured to determine a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, wherein the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule;

[0115] A third determination module 44 is configured to determine a predicted value and a loss function of fatigue performance of the high entropy alloy based on the fully connected layer and the fusion features in the initial prediction model;

[0116] The model determination module 45 is used to iteratively train the initial prediction model based on the loss function until the training end condition is met to obtain a final prediction model.

[0117] The technical solution of the embodiment of the present invention obtains a feature data sample set and an initial prediction model of a high-entropy alloy; determines the global correlation characteristics between the components in the high-entropy alloy based on the attention module and the feature data sample set in the initial prediction model; determines the fused temporal characteristics based on the global correlation characteristics and the hybrid module in the initial prediction model, the hybrid module including the long-short-term memory submodule and the gated recurrent unit submodule; determines the predicted value and loss function of fatigue performance based on the fully connected layer and fused characteristics in the initial prediction model; iteratively trains the initial prediction model based on the loss function until the training end condition is met, thereby obtaining the final prediction model. The attention module selects and strengthens important features, and the hybrid module processes the potential temporal dependencies in heat treatment parameters and microstructural characteristics, thereby improving the accuracy of the prediction model.

[0118] Wherein, each feature data training sample in the feature data sample set includes the chemical composition, heat treatment parameters, microstructure characteristics and fatigue performance of the high entropy alloy.

[0119] Furthermore, the first determining module 42 is specifically configured to:

[0120] Preprocessing each feature data training sample in the feature data sample set to obtain a preprocessed training sample;

[0121] Performing a linear transformation on the preprocessed training samples based on the attention module in the initial prediction model to obtain an output feature space;

[0122] determining an attention output based on the output feature space and a chemical composition interaction weight matrix associated with the high entropy alloy;

[0123] Based on the attention output, a global correlation feature is determined.

[0124] Furthermore, the second determining module 43 includes:

[0125] A dimension judgment unit, configured to judge whether the dimension of the global correlation feature meets the module requirements;

[0126] a first determining unit, configured to, if yes, process the global correlation feature based on the hybrid module in the initial prediction model to determine a fused temporal feature;

[0127] The second determining unit is configured to, if not, perform a linear transformation on the global correlation feature to obtain an input feature that meets the module requirements, and input the input feature into the hybrid module to obtain the fused time series feature.

[0128] The first determining unit is specifically configured to:

[0129] Extracting the long-term dependency features of the global correlation features through the long short-term memory submodule to obtain a first hidden state;

[0130] Extracting the short-term dynamic features of the global correlation features through the gated recurrent unit submodule to obtain a second hidden state;

[0131] The first hidden state and the second hidden state are fused according to a preset fusion algorithm to obtain a fused time series feature.

[0132] Furthermore, the third determining module 44 is specifically configured to:

[0133] Determining the time step weight of each time step under the fusion feature based on the attention mechanism of the initial prediction model;

[0134] Determining a weighted summary feature for each of the time step weights and the fusion feature;

[0135] Processing the weighted summary features according to the fully connected layer in the initial prediction model to determine a predicted value of the fatigue performance of the high entropy alloy;

[0136] A loss function is determined based on the predicted value and the true value in the feature data sample.

[0137] The video tracking device provided in the embodiment of the present invention can execute the method for constructing a prediction model for fatigue performance of high-entropy alloys provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0138] Example 3

[0139] Figure 5 A schematic diagram of the structure of an electronic device 50 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0140] like Figure 5As shown, the electronic device 50 includes at least one processor 51 and a memory, such as a read-only memory (ROM) 52, a random access memory (RAM) 53, etc., which is communicatively connected to the at least one processor 51. The memory stores a computer program that can be executed by the at least one processor. The processor 51 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 52 or the computer program loaded from the storage unit 58 into the random access memory (RAM) 53. Various programs and data required for the operation of the electronic device 50 can also be stored in the RAM 53. The processor 51, ROM 52, and RAM 53 are connected to each other via a bus 54. An input / output (I / O) interface 55 is also connected to the bus 54.

[0141] Multiple components in the electronic device 50 are connected to the I / O interface 55, including an input unit 56, such as a keyboard, a mouse, etc.; an output unit 57, such as various types of displays, speakers, etc.; a storage unit 58, such as a magnetic disk, an optical disk, etc.; and a communication unit 59, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 59 allows the electronic device 50 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0142] The processor 51 can be a variety of general and / or specialized processing components with processing and computing capabilities. Some examples of the processor 51 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 51 executes the various methods and processes described above, such as the method for constructing a predictive model for fatigue properties of high-entropy alloys.

[0143] In some embodiments, the method for constructing a prediction model of fatigue performance of a high entropy alloy can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 58. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 50 via the ROM 52 and / or the communication unit 59. When the computer program is loaded into the RAM 53 and executed by the processor 51, one or more steps of the method for constructing a prediction model of fatigue performance of a high entropy alloy described above can be performed. Alternatively, in other embodiments, the processor 51 can be configured to execute the method for constructing a prediction model of fatigue performance of a high entropy alloy by any other appropriate means (e.g., by means of firmware).

[0144] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system comprising at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0145] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0146] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0147] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0148] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0149] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0150] In one embodiment, the embodiment of the present invention also includes a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the method for constructing a prediction model for fatigue performance of high entropy alloys of any embodiment of the present invention.

[0151] The computer program product may be implemented by writing computer program code for performing the operations of the present invention in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0152] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0153] The above embodiments are only preferred embodiments of the present invention and cannot be used to limit the scope of protection of the present invention. Any non-substantial changes and replacements made by technicians in this field on the basis of the present invention fall within the scope of protection required by the present invention.

Claims

1. A method for constructing a prediction model for fatigue performance of high entropy alloys, characterized in that: include: Obtaining a characteristic data sample set and an initial prediction model for high entropy alloys; Determining global correlation features between components in the high-entropy alloy based on the attention module in the initial prediction model and the feature data sample set; Determining a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, wherein the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule; Determining a predicted value and a loss function of fatigue performance of the high-entropy alloy based on the fully connected layer and the fusion features in the initial prediction model; The initial prediction model is iteratively trained based on the loss function until a training end condition is met to obtain a final prediction model.

2. The method according to claim 1, characterized in that Each feature data training sample in the feature data sample set includes the chemical composition, heat treatment parameters, microstructure characteristics and fatigue properties of the high entropy alloy.

3. The method according to claim 1, characterized in that Determining the global correlation features between the components in the high-entropy alloy based on the attention module in the initial prediction model and the feature data sample set includes: Preprocessing each feature data training sample in the feature data sample set to obtain a preprocessed training sample; Performing a linear transformation on the preprocessed training samples based on the attention module in the initial prediction model to obtain an output feature space; determining an attention output based on the output feature space and a chemical composition interaction weight matrix associated with the high entropy alloy; Based on the attention output, a global correlation feature is determined.

4. The method according to claim 1, wherein The determining of the fused time series features according to the global correlation features and the hybrid module in the initial prediction model includes: Determining whether the dimension of the global correlation feature meets the module requirements; If so, processing the global correlation features based on the hybrid module in the initial prediction model to determine a fused temporal feature; If not, a linear transformation is performed on the global correlation feature to obtain an input feature that meets the requirements of the module, and the input feature is input into the hybrid module to obtain the fused temporal feature.

5. The method according to claim 4, characterized in that The processing of the global correlation features based on the hybrid module in the initial prediction model to determine the fused time series features includes: Extracting the long-term dependency features of the global correlation features through the long short-term memory submodule to obtain a first hidden state; Extracting the short-term dynamic features of the global correlation features through the gated recurrent unit submodule to obtain a second hidden state; The first hidden state and the second hidden state are fused according to a preset fusion algorithm to obtain a fused time series feature.

6. The method according to claim 1, characterized in that Determining the predicted value and loss function of the fatigue performance of the high-entropy alloy based on the fully connected layer and the fusion features in the initial prediction model includes: Determining the time step weight of each time step under the fusion feature based on the attention mechanism of the initial prediction model; Determining a weighted summary feature for each of the time step weights and the fusion feature; Processing the weighted summary features according to the fully connected layer in the initial prediction model to determine a predicted value of the fatigue performance of the high entropy alloy; A loss function is determined based on the predicted value and the true value in the feature data sample.

7. A device for constructing a prediction model for fatigue performance of high entropy alloys, characterized in that: include: A data acquisition module is used to obtain a characteristic data sample set and an initial prediction model for high entropy alloys; A first determination module is configured to determine a global correlation feature between components in the high entropy alloy based on the attention module in the initial prediction model and the feature data sample set; A second determination module is configured to determine a fused temporal feature based on the global correlation feature and a hybrid module in the initial prediction model, wherein the hybrid module includes a long short-term memory submodule and a gated recurrent unit submodule; A third determination module is used to determine the predicted value and loss function of the fatigue performance of the high entropy alloy based on the fully connected layer and the fusion feature in the initial prediction model; The model determination module is used to iteratively train the initial prediction model based on the loss function until the training end condition is met to obtain a final prediction model.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for constructing a prediction model for fatigue performance of a high-entropy alloy according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for constructing a prediction model for fatigue performance of a high-entropy alloy according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the method for constructing a prediction model for fatigue performance of a high-entropy alloy according to any one of claims 1 to 6.

Images