6061 aluminum alloy rheological behavior constitutive relation prediction method based on physical information machine learning

Abstract

The invention provides a 6061 aluminum alloy rheological behavior constitutive relation prediction method based on physical information machine learning, and the method comprises the steps: carrying out the interception of a yield point of experimental stress-strain curve data, retaining a plastic segment, and combining with a corresponding experimental condition, thereby obtaining a data set for the training of a physical information machine learning model; performing ANN model training on the data set to obtain a data driving loss item between an ANN model prediction stress value and an experimental stress value; secondly, a physical constitutive model is embedded into a loss function of an ANN model, a physical law serves as a constraint condition to be introduced into neural network learning, and an additional physical driving loss item is formed by calculating the difference between neural network prediction and physical results described by the constitutive model; and establishing a loss function by combining the physical driving loss item and the data driving loss item according to different weights, training a physical information machine learning model according to the loss function, and outputting a final prediction stress-strain curve.

Description

Technical Field

[0001] The present invention relates to the technical field of constitutive behavior prediction of metal materials, and in particular to a constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning. Background Art

[0002] Hot deformation is a common processing technology to improve the ductility and machinability of aluminum alloys. However, the rheological behavior of aluminum alloys is highly sensitive to factors such as temperature, strain rate and strain, which makes it very difficult to study the rheological behavior of aluminum alloys. In order to accurately understand the rheological behavior of aluminum alloys, the development of effective constitutive models is crucial to predict the material behavior in complex environments. A large number of constitutive models have been developed in recent years, but since phenomenological constitutive models are usually regression analyses of experimental data, they lack the necessary physical basis. Although the physics-based constitutive models are more accurate than phenomenological constitutive models, these models require a large amount of experimental data and are difficult to obtain. Summary of the invention

[0003] In view of the defects and shortcomings of the prior art, the purpose of the present invention is to provide a constitutive relationship prediction method for the rheological behavior of 6061 aluminum alloy based on physical information machine learning, so as to solve the problem of low accuracy of traditional metal plastic hardening models when facing different temperatures and strain rates.

[0004] The present invention specifically adopts the following technical solutions:

[0005] A constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning:

[0006] The experimental stress-strain curves of 6061 aluminum alloy under different temperature and strain rate conditions are obtained, and recorded together with the experimental conditions, and divided into training set and test set;

[0007] The yield point of the experimental stress-strain curve data in the training set is intercepted and the plastic segment is retained and combined with the corresponding experimental conditions, so as to obtain a data set for physical information machine learning model training;

[0008] Performing ANN model training on the data set to obtain a data-driven loss term between the stress value predicted by the ANN model and the experimental stress value;

[0009] Then, by embedding the physical constitutive model into the loss function of the ANN model, the physical laws are introduced as constraints into the neural network learning, and an additional physical driven loss term is formed by calculating the difference between the neural network prediction and the physical results described by the constitutive model.

[0010] A loss function is established by combining the physical-driven loss term and the data-driven loss term according to different weights, and a physical-information machine learning model is trained according to the loss function to output the final predicted stress-strain curve.

[0011] Furthermore, the experimental temperatures for obtaining the experimental stress-strain curve are 20°C, 100°C, and 160°C, and the strain rate is 0.001s -1 , 0.1s -1 , 10s -1 .

[0012] Furthermore, the experimental stress-strain curve data and the corresponding experimental conditions are divided into a training set and a test set by the following method:

[0013] Randomly select some experimental stress-strain curve data and their corresponding experimental conditions from all data sets as the test set, denoted as {(x j ,y j )}, j = 1, 2, ..., M, x is the experimental condition, y is the experimental stress, M is the amount of test set data; all other remaining experimental stress-strain curve data and their corresponding experimental conditions are used as training sets, denoted as {(x i ,y i )}, i = 1, 2, ..., Z, x is the experimental condition, y is the experimental stress, and Z is the amount of training set data.

[0014] Preferably, the ratio of the data volume of the test set to the data volume of the training set is 3:7.

[0015] Furthermore, the following method is used to intercept the yield point of the experimental stress-strain curve data in the training set and retain the plastic segment and combine it with the corresponding experimental conditions:

[0016] For any experimental stress-strain curve, the yield point is intercepted according to the following formula:

[0017]

[0018] Among them, σ 0 (T) is the yield stress, T is the current experimental temperature, T m is the melting point temperature of the material under consideration, K 0 =0.2295,σ 0 (T 0 ) is the initial yield stress corresponding to the benchmark test temperature;

[0019] Then the plastic deformation part of the retained experimental stress-strain curve and the corresponding experimental conditions are normalized:

[0020]

[0021] Where X i,norm represents the normalized value of the i-th variable in the dataset, X min and X max are the minimum and maximum values ​​of the original data, respectively, i Represents the raw value of the corresponding feature.

[0022] Furthermore, the ANN model includes: neurons as basic units, input layers, output layers and hidden layers; each input variable is connected to each neuron, and the number of input layers and output layers is one and only one, wherein the number of neurons in the input layer and the output layer is determined by the input variable x i With the output variable y i The input variables are strain ε, strain rate and temperature T, the output variable is stress σ; there is a hidden layer between the input layer and the output layer, with a total of 4 hidden layers, and the number of neurons in each layer is 64, 64, 64, and 32 respectively; the two layers of the ANN model are connected by a single neuron, and the model completes the mapping from the input vector to the corresponding output vector through basic calculations:

[0023]

[0024] Where f is the Relu activation function, M and N are the input vector dimensions of the input layer and output layer respectively, ω ij and ω ki are the weights from the input layer to the hidden layer and from the hidden layer to the output layer, respectively, i and b k are the biases of the hidden layer and the output layer respectively;

[0025] It is a data-driven loss term, which is used to measure the error of the ANN model's predicted data to the original data:

[0026]

[0027] where y i,pre For the ANN model to predict stress, y i is the experimental stress.

[0028] The training of ANN models needs to be carried out step by step. First, define the neural network architecture (for example, the number of layers, the number of neurons in each layer, etc.), initialize the network weights (usually using small random values ​​or specific initialization methods) and biases; next, the input data starts from the input layer, passes through the hidden layers layer by layer, and finally reaches the output layer. The output of each layer is nonlinearly transformed by the activation function (ReLU), and the loss function (mean square error) is used to calculate the difference between the predicted output of the model and the true label. According to the loss function, the gradient relative to each parameter (weight and bias) is calculated, and the gradient is passed back from the output layer to the input layer through the chain rule, and the gradient of each parameter is calculated layer by layer. Use gradient descent or other optimization algorithms (Adam) to update the weights and biases according to the calculated gradients to reduce losses; finally, through multiple iterations, continuously optimize the parameters of the model, and each iteration will update the weights through forward propagation and back propagation. After each iteration, use the validation set to evaluate the performance of the model to prevent overfitting.

[0029] Furthermore, the embedding of the physical constitutive model into the loss function of the ANN model is specifically:

[0030] The Voce model is used to represent the rheological behavior of 6061 aluminum alloy within a given temperature and strain rate range; the Voce strain hardening formula is as follows:

[0031]

[0032] Where K i and m i is the material parameter, σ 0 (T) is the initial yield stress, T is the current experimental temperature, T m is the melting point temperature of the material under consideration;

[0033] By adding terms representing physical laws to the loss function, the neural network can learn the Voce formula to describe the sensitivity of strain hardening to temperature and strain rate while learning the data. After embedding the constitutive model into the loss function of the ANN model, the expression is as follows:

[0034]

[0035] is a physics-driven loss term that quantifies the deviation between the output and the physical prior knowledge; w data and w Voce Weight used to balance the interaction between the two loss terms.

[0036] Embedding the constitutive equation into the loss function actually introduces the physical laws in reality as constraints into the learning of the neural network, and forms an additional loss term by calculating the difference between the neural network prediction and the physical results described by the constitutive model. This means that the neural network must not only make accurate predictions at the data level, but also make predictions that conform to the laws of physics at the physical level. During the training of the neural network, the model minimizes the loss function through gradient descent or other optimization algorithms, which includes consideration of physical constraints. This approach allows the neural network to consider not only how to better fit the data when updating parameters, but also how to better satisfy the laws of physics.

[0037] Furthermore, the training process of the physical information machine learning model is:

[0038] The input variables strain ε, strain rate and temperature T are provided to the physical information machine learning model for training, and the data-driven loss is obtained respectively. and physical drive losses

[0039] The physical information machine learning model assigns weights to the two loss terms according to the Adam optimizer to obtain the loss function

[0040] The physical information machine learning model continuously performs hyperparameter optimization according to the loss function to obtain the optimal training result.

[0041] And, a constitutive relationship prediction system for rheological behavior of 6061 aluminum alloy based on physical information machine learning, comprising:

[0042] The data generation module is used to obtain the experimental stress-strain curves of 6061 aluminum alloy under different temperature and strain rate conditions, and record them together with the experimental conditions, and divide them into training sets and test sets;

[0043] A data processing module, used to intercept the yield point of the experimental stress-strain curve data in the training set and retain the plastic segment and combine it with the corresponding experimental conditions, so as to obtain a data set for physical information machine learning model training;

[0044] A data-driven loss term acquisition module is used to perform ANN model training on the data set to obtain a data-driven loss term between the stress value predicted by the ANN model and the experimental stress value;

[0045] A physical driven loss term acquisition module is used to embed the physical constitutive model into the loss function of the ANN model, so as to introduce physical laws as constraints into the neural network learning, and to form an additional physical driven loss term by calculating the difference between the neural network prediction and the physical results described by the constitutive model;

[0046] The prediction module is used to establish a loss function by combining the physical-driven loss term and the data-driven loss term according to different weights, and train the physical information machine learning model according to the loss function to output the final predicted stress-strain curve.

[0047] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the constitutive relationship prediction method for the rheological behavior of 6061 aluminum alloy based on physical information machine learning are implemented as described above.

[0048] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting the constitutive relationship of the rheological behavior of 6061 aluminum alloy based on physical information machine learning as described above.

[0049] Compared with the prior art, the present invention and its preferred solution adopt machine learning methods to provide a new, more accurate, universal and flexible 6061 aluminum alloy rheological behavior constitutive relationship prediction solution, which can not only improve the prediction accuracy of metal material properties, but also make the prediction results of neural networks more consistent with physical laws. It can not only reduce the cost of research and engineering design, accelerate the progress of material research, but also provide new possibilities for various application fields.

[0050] The constitutive modeling based on machine learning in this invention is not based solely on empirical formulas or theoretical models, but uses experimental or simulation data to derive models of material or system behavior to address challenges encountered in this field. The prediction results of the physical information machine learning model are more consistent with the laws of physics, providing a new solution for accurately describing the rheological behavior of aluminum alloys during thermal deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0052] Figure 1 It is a schematic diagram of the structure of the constitutive relationship prediction model of the rheological behavior of 6061 aluminum alloy built based on the PIML model in an embodiment of the present invention.

[0053] Figure 2It is a comparison chart of the stress-strain curve predicted for 6061 aluminum alloy at temperatures of 20°C, 100°C, 160°C and strain rates of 0.001s-1, 0.1s-1, and 10s-1 in an embodiment of the present invention and the stress-strain curve obtained by testing under the same conditions. DETAILED DESCRIPTION

[0054] In order to make the features and advantages of the present invention more clearly understood, the following embodiments are specifically described in detail with reference to the accompanying drawings:

[0055] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0056] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0057] The specific steps of the overall construction process of the scheme for predicting the constitutive relationship of the rheological behavior of 6061 aluminum alloy based on physical information machine learning in the embodiment of the present invention are as follows:

[0058] S01: respectively obtaining experimental stress-strain curves of 6061 aluminum alloy under different temperature and strain rate conditions, and recording them together with the experimental conditions; respectively preprocessing the obtained experimental stress-strain curves, and dividing the processed experimental stress-strain curve data and the corresponding experimental conditions into a training set and a test set according to the required ratio;

[0059] S02: for the experimental stress-strain curve data in the training set of the experimental plastic stress-strain curve obtained in step S01, intercept the yield point and retain the plastic segment, combine all processed experimental stress-strain curve data with the corresponding experimental conditions, so as to obtain a data set for physical information machine learning (PIML) model training;

[0060] S03: Building a prediction model based on the ANN model, part 1, performing ANN model training on the yield stress-strain curve data set at different temperatures and strain rates obtained in step S02, and obtaining a data-driven loss term between the stress value predicted by the ANN model and the experimental stress value from the training;

[0061] S04: Based on embedding the physical constitutive model into the loss function of the ANN model, the prediction model is built. Part 2, which introduces the physical laws in reality as constraints into the neural network learning, and forms an additional physical drive loss term by calculating the difference between the neural network prediction and the physical results described by the constitutive model. This means that the neural network must not only make accurate predictions at the data level, but also make predictions that conform to the laws of physics at the physical level;

[0062] S05: The physical driven loss term obtained in step S04 and the data driven loss term obtained in step S03 are combined according to different weights to establish a loss function, and the PIML model is continuously trained according to the loss function to output the final predicted stress-strain curve.

[0063] In step S01, processing the stress-strain curve obtained from the experiment includes the following sub-steps:

[0064] S01-1: at different temperatures 20℃, 100℃, 160℃ and different strain rates 0.001s -1 , 0.1s -1 , 10s -1 The 6061 aluminum alloy tensile test was carried out under the following conditions, and 7 groups of original experimental stress-strain curves were obtained. The specific tensile test conditions are shown in Table 1;

[0065] Table 1

[0066] Temperature(℃) <![CDATA[Strain rate (s -1 )]]> 1 20℃ <![CDATA[0.001s -1 ]]> 2 160℃ <![CDATA[0.001s -1 ]]> 3 100℃ <![CDATA[0.1s -1 ]]> 4 160℃ <![CDATA[0.1s -1 ]]> 5 20℃ <![CDATA[10s -1 ]]> 6 100℃ <![CDATA[10s -1 ]]> 7 160℃ <![CDATA[10s -1 ]]>

[0067] S01-2: Preprocess the stress-strain curves obtained in step S01-1 respectively. The specific operations of the preprocessing are as follows:

[0068] (S01-21): For any experimental stress-strain curve, according to:

[0069]

[0070] The formula intercepts the yield point. Among them, σ 0 (T) is the yield stress, T is the current experimental temperature, Tn ＝ 617℃ is the melting point of the material under consideration, K 0 =0.2295,σ 0 (T 0 ) = 148.5 MPa is the initial yield stress at the experimental temperature of 20°C. The plastic deformation part of the experimental stress-strain curve is retained;

[0071] (S01-22): Normalize the plastic deformation part of the retained experimental stress-strain curve and the corresponding experimental conditions:

[0072]

[0073] Where X i,norm represents the normalized value of the i-th variable in the dataset, X min and X max are the minimum and maximum values ​​of the original data, respectively, i Indicates the original value of the corresponding feature. Table 2 and Table 3 below are the original data and part of the normalized data of the experimental data respectively;

[0074] Table 2 Original data sample (part)

[0075] strain <![CDATA[Strain rate (s -1 )]]> Temperature(℃) 0.03554 0.001 20 0.12175 0.001 20 0.2029 0.001 20 0.03564 0.001 160 0.08971 0.001 160 0.14727 0.001 160 0.06005 0.1 100 0.11101 0.1 100 0.13193 0.1 100 0.03286 0.1 160 0.03692 0.1 160 0.04728 0.1 160 0.05936 10 20 0.12453 10 20 0.14606 10 20 0.02123 10 100 0.0887 10 100 0.11608 10 100 0.01327 10 160 0.04925 10 160 0.07613 10 160

[0076] Table 3 Normalized data samples (part)

[0077]

[0078]

[0079] S01-3: Divide the stress-strain curve data processed in step S01-2 and its corresponding experimental conditions into a training set and a test set according to the following method:

[0080] Randomly select some experimental stress-strain curve data and their corresponding experimental conditions from all data sets as the test set, denoted as {(x j ,y j )}, j = 1, 2, ..., M, x is the experimental condition, y is the experimental stress, M is the amount of test set data; all other remaining experimental stress-strain curve data and their corresponding experimental conditions are used as training sets, denoted as {(x i ,y i )}, i = 1, 2, ..., Z, x is the experimental condition, y is the experimental stress, and Z is the amount of data in the training set. The ratio of the amount of data in the test set to the amount of data in the training set is 3:7;

[0081] In step S02, the structural schematic diagram of the constitutive relationship prediction model of the rheological behavior of 6061 aluminum alloy built by the PIML model is as follows: Figure 1 As shown;

[0082] In step S03, the method of establishing and training the ANN model is as follows:

[0083] ANN is composed of neurons as basic units, input layer, output layer and hidden layer. Each input variable is connected to each neuron, and there is only one input layer and one output layer, where the number of neurons in the input layer and the output layer is determined by the input variable x. i With the output variable y iIn this work, the input variables are strain ε, strain rate and temperature T, and the output variable is stress σ. The layer between the two is called the hidden layer, and 4 hidden layers are set, with 64, 64, 64, and 32 neurons in each layer. The two layers of the ANN model are connected by a single neuron, and the model completes the mapping from the input vector to the corresponding output vector through basic calculations:

[0084]

[0085] Where f is the Relu activation function, M and N are the input vector dimensions of the input layer and output layer respectively, ω ij and ω ki are the weights from the input layer to the hidden layer and from the hidden layer to the output layer, respectively, i and b k are the biases of the hidden layer and the output layer respectively;

[0086] The loss function is a function that measures the difference between the model's predicted value and the true value and is used to guide the training process of the neural network. The data-driven loss term is used to measure the error of the ANN model's predicted data to the original data. The mean square error (MSE) loss is used, and its calculation formula is:

[0087]

[0088] where y i,pre For the ANN model to predict stress, y i is the experimental stress;

[0089] The model was trained for 200 epochs, and the learning rate was set to 10 during all training processes. -3 , using the Adam optimizer with default hyperparameters, the model training uses a batch size of 100, and the mean absolute percentage error (MAPE) and correlation coefficient (R) are used to evaluate the model performance:

[0090]

[0091] The smaller the value of mean absolute percentage error MAPE, the better the prediction effect of the model; the value of the correlation coefficient R is a value between -1 and 1, which indicates the goodness of fit. The closer its value is to 1, the better the model fitting effect;

[0092] In step S04, the specific method of embedding the physical constitutive model into the loss function of the ANN model in step S03 is as follows:

[0093] The Voce model is selected to represent the rheological behavior of 6061 aluminum alloy in the above temperature and strain rate range. This model can be used to express the sensitivity of the rheological behavior of 6061 aluminum alloy to temperature and strain rate when large plastic strain is applied under equibiaxial load. The Voce strain hardening formula is as follows:

[0094]

[0095] Where K i and m i is the material parameter, σ 0 (T) is the initial yield stress, T is the current experimental temperature, T m is the melting point temperature of the material under consideration;

[0096] By adding terms representing physical laws to the loss function, the neural network can learn the Voce formula to describe the sensitivity of strain hardening to temperature and strain rate while learning the data. The constitutive model is embedded in the loss function of the ANN model. The expression is as follows:

[0097]

[0098] is a physics-driven loss term that quantifies the deviation between the output and the physical prior knowledge. data and w Voce are weights used to balance the interaction between the two loss terms. These weights can be defined by the user or adjusted automatically, and play an important role in improving the trainability of the model;

[0099] In step S05, the physical driven loss term obtained in step S04 and the data driven loss term obtained in step S03 are combined according to different weights to establish a loss function, and the PIML model is continuously trained according to the loss function to output the final predicted stress-strain curve;

[0100] The model's prediction results are as follows Figure 2 As shown. Figure 2 From a to c, it can be seen that the numerical prediction of the model is in good agreement with the experimentally obtained stress-strain curve. The physical information machine learning model can predict the rheological behavior of 6061 aluminum alloy with sufficient accuracy and reliability. Table 4 shows the mean absolute percentage error (MAPE) and correlation coefficient (R) of the model for each set of data. As can be seen from the table, the prediction accuracy of the physical information machine learning model is very high, and the average MAPE and R values ​​are 0.807% and 0.9972, respectively, which meet the expected goals;

[0101] Table 4 Prediction results of the physical information machine learning model under different experimental parameters

[0102]

[0103]

[0104] It has been verified by the examples that the constitutive relationship prediction method of the rheological behavior of 6061 aluminum alloy based on physical information machine learning of the present invention can effectively predict the plastic part stress-strain curve (i.e., the constitutive relationship of the rheological behavior) of 6061 aluminum alloy under different temperature and strain rate conditions after being trained with the training set data.

[0105] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0106] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps for the functions specified in one or more boxes.

[0109] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of constitutive relationship prediction methods for the rheological behavior of 6061 aluminum alloy based on physical information machine learning under the inspiration of the present invention. All equal changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning, characterized by: The experimental stress-strain curves of 6061 aluminum alloy under different temperature and strain rate conditions are obtained, and recorded together with the experimental conditions, and divided into training set and test set; The yield point of the experimental stress-strain curve data in the training set is intercepted and the plastic segment is retained and combined with the corresponding experimental conditions, so as to obtain a data set for physical information machine learning model training; Performing ANN model training on the data set to obtain a data-driven loss term between the stress value predicted by the ANN model and the experimental stress value; Then, by embedding the physical constitutive model into the loss function of the ANN model, the physical laws are introduced as constraints into the neural network learning, and an additional physical driven loss term is formed by calculating the difference between the neural network prediction and the physical results described by the constitutive model. A loss function is established by combining the physical-driven loss term and the data-driven loss term according to different weights, and a physical-information machine learning model is trained according to the loss function to output the final predicted stress-strain curve.

2. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 1 is characterized in that: The experimental temperatures for obtaining the experimental stress-strain curves are 20°C, 100°C, and 160°C, and the strain rate is 0.001s -1 , 0.1s -1 , 10s -1 .

3. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 1 is characterized in that: The experimental stress-strain curve data and the corresponding experimental conditions are divided into a training set and a test set using the following method: Randomly select some experimental stress-strain curve data and their corresponding experimental conditions from all data sets as the test set, denoted as {(x j ,y j )}, j = 1, 2, ..., M, x is the experimental condition, y is the experimental stress, M is the amount of test set data; all other remaining experimental stress-strain curve data and their corresponding experimental conditions are used as training sets, denoted as {(x i ,y i )}, i = 1, 2, ..., Z, x is the experimental condition, y is the experimental stress, and Z is the amount of training set data.

4. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 1, characterized in that: The following method is used to intercept the yield point of the experimental stress-strain curve data in the training set and retain the plastic segment and combine it with the corresponding experimental conditions: For any experimental stress-strain curve, the yield point is intercepted according to the following formula: Among them, σ0(T) is the yield stress, T is the current experimental temperature, and T m is the melting point temperature of the material under consideration, K0 = 0.2295, σ0(T0) is the initial yield stress corresponding to the benchmark experimental temperature; Then the plastic deformation part of the retained experimental stress-strain curve and the corresponding experimental conditions are normalized: Where X i,norm represents the normalized value of the i-th variable in the dataset, X min and X max are the minimum and maximum values ​​of the original data, respectively, i Represents the raw value of the corresponding feature.

5. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 1, characterized in that: The ANN model includes: neurons as basic units, input layers, output layers and hidden layers; each input variable is connected to each neuron, and the number of input layers and output layers is one and only one, wherein the number of neurons in the input layer and the output layer is determined by the input variable x. i With the output variable y i The input variables are strain ε, strain rate and temperature T, the output variable is stress σ; there is a hidden layer between the input layer and the output layer, with a total of 4 hidden layers, and the number of neurons in each layer is 64, 64, 64, and 32 respectively; the two layers of the ANN model are connected by a single neuron, and the model completes the mapping from the input vector to the corresponding output vector through basic calculations: Where f is the Relu activation function, M and N are the input vector dimensions of the input layer and output layer respectively, ω ij and ω ki are the weights from the input layer to the hidden layer and from the hidden layer to the output layer, respectively, i and b k are the biases of the hidden layer and the output layer respectively; It is a data-driven loss term, which is used to measure the error of the ANN model's predicted data to the original data: where y i,pre For the ANN model to predict stress, y i is the experimental stress.

6. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 5 is characterized in that: The embedding of the physical constitutive model into the loss function of the ANN model is specifically: The Voce model is used to represent the rheological behavior of 6061 aluminum alloy within a given temperature and strain rate range; the Voce strain hardening formula is as follows: Where K i and m i is the material parameter, σ0(T) is the initial yield stress, T is the current experimental temperature, T m is the melting point temperature of the material under consideration; By adding terms representing physical laws to the loss function, the neural network can learn the Voce formula to describe the sensitivity of strain hardening to temperature and strain rate while learning the data. After embedding the constitutive model into the loss function of the ANN model, the expression is as follows: is a physics-driven loss term that quantifies the deviation between the output and the physical prior knowledge; w data and w Voce Weight used to balance the interaction between the two loss terms.

7. The constitutive relationship prediction method for rheological behavior of 6061 aluminum alloy based on physical information machine learning according to claim 6, characterized in that: The training process of the physical information machine learning model is: The input variables strain ε, strain rate and temperature T are provided to the physical information machine learning model for training, and the data-driven loss is obtained respectively. and physical drive losses The physical information machine learning model assigns weights to the two loss terms according to the Adam optimizer to obtain the loss function The physical information machine learning model continuously performs hyperparameter optimization according to the loss function to obtain the optimal training result.

8. A constitutive relationship prediction system for rheological behavior of 6061 aluminum alloy based on physical information machine learning, characterized in that: include: The data generation module is used to obtain the experimental stress-strain curves of 6061 aluminum alloy under different temperature and strain rate conditions, and record them together with the experimental conditions, and divide them into training sets and test sets; A data processing module, used to intercept the yield point of the experimental stress-strain curve data in the training set and retain the plastic segment and combine it with the corresponding experimental conditions, so as to obtain a data set for physical information machine learning model training; A data-driven loss term acquisition module is used to perform ANN model training on the data set to obtain a data-driven loss term between the stress value predicted by the ANN model and the experimental stress value; A physical driven loss term acquisition module is used to embed the physical constitutive model into the loss function of the ANN model, so as to introduce physical laws as constraints into the neural network learning, and to form an additional physical driven loss term by calculating the difference between the neural network prediction and the physical results described by the constitutive model; The prediction module is used to establish a loss function by combining the physical-driven loss term and the data-driven loss term according to different weights, and train the physical information machine learning model according to the loss function to output the final predicted stress-strain curve.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for predicting the constitutive relationship of the rheological behavior of 6061 aluminum alloy based on physical information machine learning are implemented as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting the constitutive relationship of the rheological behavior of 6061 aluminum alloy based on physical information machine learning as described in any one of claims 1 to 7 are implemented.

Images