Method for acquiring parameters of stainless steel material av constitutive model based on neural network

Abstract

The application relates to a neural network-based stainless steel material AV constitutive model parameter acquisition method, which comprises the following steps: step 1, obtaining a uniaxial tensile stress-strain curve of a stainless steel material; step 2, establishing a sample database for training a neural network; step 3, establishing a neural network model for identifying AV constitutive model parameters; step 4, training the neural network by using sample data; and step 5, identifying AV constitutive model parameters of a test curve by using the trained neural network. The neural network is used to establish a network model for identifying AV constitutive model parameters, and compared with a traditional fitting and optimization method, the neural network has great improvement in efficiency and accuracy; the method is used for calculating stress-strain curves by using random number to form AV constitutive model parameter combinations and using AV constitutive models to obtain parameter sample data, and effectively solves the problem of difficulty in obtaining sample data required by neural network training.

Description

Technical Field

[0001] This application relates to the fields of machine learning and materials science, specifically to a method for obtaining parameters of the AV constitutive model of stainless steel based on neural networks. Background Technology

[0002] In practical engineering, due to different working environments, the stress and strain of components change accordingly when subjected to different loads, and different load parameters have different effects on local plastic deformation. The AV constitutive model can effectively describe the stress-strain curves of stainless steel materials under different loads, but there are many undetermined parameters in the model, and the coupling between these parameters is strong. How to effectively determine these parameters has always been a difficult problem.

[0003] Traditional parameter determination methods typically employ fitting or optimization techniques, using the material's mechanical response to inversely calculate constitutive model parameters, such as multi-objective genetic algorithms or trial-and-error methods. However, these methods are cumbersome to solve and the results are not very practical.

[0004] With the increasing complexity of science and engineering technologies, there is a growing need for algorithms capable of solving large-scale and complex problems. Many intelligent inversion analysis methods have been researched for different constitutive model problems. The AV constitutive model exhibits high nonlinearity, making it difficult to establish the mapping relationship between known and unknown quantities using traditional inversion analysis methods. However, artificial neural networks, by learning from large amounts of sample data, can find the characteristic relationships between input and output quantities, thereby establishing very complex nonlinear mapping relationships. Therefore, neural network algorithms can be introduced into the determination of AV constitutive model parameters, and a neural network-based method for obtaining AV constitutive model parameters for stainless steel materials can be developed. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention utilizes the generation of random numbers to construct the parameter combination of the AV constitutive model under uniaxial tensile conditions for stainless steel. The stress-strain curve is then calculated using the AV constitutive model via forward modeling to obtain parameter sample data, effectively solving the problem of obtaining the sample data required for neural network training. Furthermore, a network model for identifying the AV constitutive model parameters is established using a neural network, resulting in significant improvements in both efficiency and accuracy compared to traditional fitting and optimization methods.

[0006] To achieve the above objectives, the solution adopted by the present invention is as follows:

[0007] Step 1: Obtain the uniaxial tensile stress-strain curve of stainless steel material;

[0008] At room temperature, a uniaxial tensile test was conducted on stainless steel material at a preset loading rate to obtain the uniaxial tensile stress-strain curve of the stainless steel material.

[0009] Step 2: Establish a sample database for training the neural network;

[0010] The value ranges of the first, second, third, and fourth parameters in the AV constitutive model of stainless steel under uniaxial tensile conditions are determined as follows: the value range of the first parameter in the AV constitutive model of stainless steel under uniaxial tensile conditions is (20000, 50000), the value range of the second parameter is (200, 500), the value range of the third parameter is (0, 100), and the value range of the fourth parameter is (0.4, 0.8). Within these value ranges, the AV constitutive model parameters are randomly and uniformly generated to establish N combinations of AV constitutive model parameters. The AV constitutive model is shown below:

[0011]

[0012] In the formula: α represents the back stress tensor; ε p λ represents the plastic strain tensor; b represents the cumulative plastic strain; C, γ1, γ2, and δ represent the first, second, third, and fourth parameters of the stainless steel material in the AV constitutive model, respectively.

[0013] Stress-strain sample curves are obtained by forward modeling using the AV constitutive model and the N AV constitutive model parameter combinations; n sample feature points that can represent the shape of the stress-strain sample curves are selected on the stress-strain sample curves according to the set strain interval, and a sample database for training the neural network is established.

[0014] Step 3: Establish a neural network model for recognizing the parameters of the AV constitutive model;

[0015] The number of nodes in the input layer of the neural network corresponds to the number of sample feature points n of the stress-strain sample curve, and the number of nodes in the output layer is determined by the four AV constitutive model parameters. It contains three hidden layers, and the number of nodes in the hidden layers is set according to the geometric pyramid rule, with the number of nodes in each layer decreasing continuously from the input layer to the output layer. The activation function used is the Sigmoid function.

[0016] Step 4: Train the neural network using sample data;

[0017] The stress values ​​corresponding to the sample feature points of the stress-strain sample curves are normalized and used as the input values ​​of the neural network. The parameters of the constitutive model corresponding to the stress-strain sample curves are normalized and used as the output values ​​to train the constructed neural network. The normalization principle is that the normalized data should be kept in the interval [0, 1]. The normalization method for the feature values ​​of the stress-strain sample curves is as follows:

[0018]

[0019] Where: σ nor The normalized eigenvalues ​​are represented by σ; the eigenvalues ​​of the sample curve are represented by σ; and the stress characteristic scaling factor is represented by h1.

[0020] The normalization method for the parameters of the AV constitutive model is as follows:

[0021]

[0022] In the formula: p represents the normalized AV constitutive model parameters. i p represents the parameters of the AV constitutive model before normalization. imax p represents the maximum value within the range of each parameter. imin This indicates the minimum value within the range of each parameter;

[0023] Step 5: Use the trained neural network to identify the AV constitutive model parameters of the experimental curve;

[0024] Following the sample feature point selection method in step 2, feature points are selected on the experimental curve. The stress values ​​corresponding to the feature points are normalized and used as input values ​​for the neural network trained in step 4. The neural network will output the normalized AV constitutive model parameters corresponding to the experimental curve. The AV constitutive model parameters are then denormalized to obtain the AV constitutive model parameters corresponding to the experimental curve, specifically:

[0025]

[0026] The predicted stress-strain curves were calculated using the AV constitutive model and compared with the experimental curves to verify the accuracy of the prediction results, thereby obtaining the parameters required for the AV constitutive model of stainless steel.

[0027] Preferably, the random and uniform generation of AV constitutive model parameters in step 2 is specifically as follows:

[0028] Uniformly distributed random numbers are generated within the interval of the AV constitutive model parameters. The probability density function of the random numbers within the interval of the AV constitutive model parameters is shown below:

[0029]

[0030] In the formula: f represents the probability density function; X represents the abscissa of the probability density function; x represents the independent variable of the probability density function; a represents the lower limit of the parameter interval; b represents the upper limit of the parameter interval.

[0031] Preferably, the forward modeling calculation in step 2 specifically involves: using the generated AV constitutive model parameter combination to calculate the stress-strain curve through the constitutive model, with an initial strain value of 0 and a strain increment set to 1e. -5 The maximum strain is 0.3. At each strain point, the generated AV constitutive model parameters C, γ1, γ2, and δ are combined and substituted into the AV constitutive model to calculate the stress value corresponding to each strain point, thus obtaining the stress-strain sample curve.

[0032] Preferably, the geometric pyramid rule in step 3 is as follows: from the input layer to the output layer of the neural network structure model, the number of neural network nodes in each layer continuously decreases, and the number of neural network nodes in each layer is reduced by 2 compared to the previous layer, forming a geometric pyramid structure.

[0033] Preferably, the preset loading rate in step 1 is within the range of 0.1% / s to 0.3% / s.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] (1) The range of AV constitutive model parameters determined in this invention can cover the vast majority of stainless steel materials and can effectively identify the AV constitutive model parameters of most stainless steel materials under uniaxial tensile conditions.

[0036] (2) This invention utilizes the generation of uniformly distributed random numbers to form the parameter combination of the AV constitutive model, and uses the AV constitutive model to calculate the stress-strain curve through forward modeling to obtain sample data. This effectively replaces the method of obtaining sample data through experiments and finite element simulation, and generates any number of samples required in a short time, greatly improving efficiency. It also effectively solves the problem of difficulty in obtaining sample data required for neural network training.

[0037] (3) The present invention selects feature points on the stress-strain curve as input data of the neural network, which can be applied to various loading conditions such as uniaxial tension and cyclic deformation of stainless steel materials. It is only necessary to select feature points that can represent the curve characteristics on the stress-strain curve under different working conditions as input and adjustment parameter range of the neural network.

[0038] (4) This invention utilizes a neural network to establish a network model for identifying AV constitutive model parameters, which greatly improves efficiency and accuracy compared with traditional fitting and optimization methods. Attached Figure Description

[0039] Figure 1 This is a control block diagram of the method for obtaining parameters of the AV constitutive model of stainless steel material based on neural networks according to an embodiment of the present invention;

[0040] Figure 2 This is a uniaxial tensile stress-strain curve obtained from experiments in an embodiment of the present invention.

[0041] Figure 3 This is a schematic diagram illustrating the selection of characteristic points on the stress-strain curve in an embodiment of the present invention;

[0042] Figure 4 This is a comparison chart of the predicted curve and the experimental curve for uniaxial tensile conditions in an embodiment of the present invention. Detailed Implementation

[0043] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0044] This invention, in its embodiments, generates uniform random numbers to construct the parameter combination of the AV constitutive model under uniaxial tensile conditions for stainless steel. Using the AV constitutive model, it calculates the stress-strain curve through forward modeling to obtain parameter sample data, effectively solving the problem of obtaining the sample data required for neural network training. Utilizing a neural network to establish a network model for identifying the AV constitutive model parameters significantly improves both efficiency and accuracy. For example... Figure 1 The diagram shown is a control block diagram of the method for obtaining parameters of the AV constitutive model of stainless steel material based on neural networks according to an embodiment of the present invention.

[0045] This invention provides a method for obtaining AV constitutive model parameters of stainless steel material based on neural networks. To demonstrate the applicability of this invention, it is applied to an example, specifically including the following steps:

[0046] S1: Obtain the uniaxial tensile stress-strain curve of stainless steel material.

[0047] At room temperature, a uniaxial tensile test was conducted on stainless steel at a preset loading rate to obtain the stress-strain curve of the stainless steel under uniaxial tensile conditions; Figure 2 The figure shown is a uniaxial tensile stress-strain curve obtained from an experiment according to an embodiment of the present invention. The preset loading rate is within the range of 0.1% / s to 0.3% / s, preferably 0.2% / s.

[0048] S2: Establish a sample database for training the neural network.

[0049] First, the value ranges of the AV constitutive model parameters under uniaxial tension of stainless steel are determined. Specifically, the value range of the first parameter of the AV constitutive model under uniaxial tension of stainless steel is (20000, 50000), the value range of the second parameter is (200, 500), the value range of the third parameter is (0, 100), and the value range of the fourth parameter is (0.4, 0.8). Within these ranges, the AV constitutive model parameters are randomly generated, establishing N combinations of AV constitutive model parameters. There are four AV constitutive model parameters in total: the first parameter C, the second parameter γ1, the third parameter γ2, and the fourth parameter δ.

[0050] The random generation of AV constitutive model parameters specifically involves generating uniformly distributed random numbers within the interval of the AV constitutive model parameters. The probability density function of the random numbers within the interval of the AV constitutive model parameters is shown below:

[0051]

[0052] In the formula: f represents the probability density function; X represents the abscissa of the probability density function; x represents the independent variable of the probability density function; a represents the lower limit of the parameter interval; b represents the upper limit of the parameter interval.

[0053] Stress-strain sample curves are obtained by forward modeling using the AV constitutive model and the aforementioned N parameter combinations. Then, n feature points that can represent the curve shape are selected on the curve according to a certain strain interval to establish a sample database.

[0054] This example selects a strain value.

[0055] The feature points corresponding to 0.02, 0.04, 0.06, 0.08, 0.1, 0.12, 0.14, 0.16, 0.18, 0.2, 0.22, 0.24, 0.26, 0.28, and 0.3 are 15 in total.

[0056] The AV constitutive model is shown below:

[0057]

[0058] In the formula: α represents the back stress tensor; ε p λ represents the plastic strain tensor; b represents the cumulative plastic strain; C, γ1, γ2, and δ represent the first, second, third, and fourth parameters of the material's AV constitutive model.

[0059] The range of values ​​for the AV constitutive model parameters under uniaxial tensile conditions in this example is shown in Table 1.

[0060] Table 1. Range of AV constitutive model parameters under uniaxial tensile conditions.

[0061]

[0062] Within the value range shown in the table above, parameters are randomly generated to establish 50,000 parameter combinations. Stress-strain sample curves are obtained using the AV constitutive model. Feature points corresponding to the strain values ​​are selected on the curves, as illustrated in the diagram below. Figure 3 The diagram shown is a schematic diagram of the selection of feature points of the stress-strain curve in an embodiment of the present invention, and a sample database containing 50,000 samples is established.

[0063] Stress-strain sample curves are obtained by forward modeling using the AV constitutive model and N combinations of AV constitutive model parameters. Specifically, the forward modeling calculation involves using the generated AV constitutive model parameter combinations to calculate the stress-strain curves, with an initial strain value of 0 and a strain increment set to 1e. -5 The maximum strain is 0.3. At each strain point, the generated AV constitutive model parameters C, γ1, γ2, and δ are combined and substituted into the AV constitutive model to calculate the stress value corresponding to each strain point, thus obtaining the stress-strain sample curve.

[0064] S3: Establish a neural network model for recognizing the parameters of the AV constitutive model.

[0065] The number of nodes in the input layer of the neural network corresponds to the number of sample feature points n of the stress-strain sample curve, which is 15 in this example. The number of nodes in the output layer is the four undetermined AV constitutive model parameters. There are a total of 3 hidden layers. The number of nodes in the hidden layers is set according to the geometric pyramid rule. From the input layer to the output layer, the number of nodes in each layer decreases continuously. The geometric pyramid rule is as follows: from the input layer to the output layer of the neural network structure model, the number of neural network nodes in each layer decreases continuously. The number of neural network nodes in each layer is 2 fewer than the previous layer, forming a geometric pyramid structure.

[0066] The activation function uses the Sigmoid function.

[0067] S4: Train the neural network using sample data.

[0068] The stress values ​​corresponding to the sample feature points of the stress-strain sample curves were normalized and used as the input values ​​of the neural network. The constitutive model parameters corresponding to the stress-strain sample curves were normalized and used as the output values. The constructed neural network was then trained. The normalization principle was that the normalized data remained within the interval [0, 1]. The number of training iterations was 3000, the learning rate was 0.1, and the training precision was 1×10⁻⁶. -7The normalization method for the eigenvalues ​​of the stress-strain sample curves is shown below:

[0069]

[0070] Where: σ nor σ represents the normalized eigenvalue; h1 represents the eigenvalue of the sample curve; h1 represents the stress characteristic scaling factor, which is taken as h1 = 400 in this example.

[0071] The normalization method for the parameters of the AV constitutive model is as follows:

[0072]

[0073] In the formula: p represents the normalized AV constitutive model parameters. i p represents the parameters of the AV constitutive model before normalization. imax p represents the maximum value within the range of each parameter. imin This indicates the minimum value within the range of each parameter.

[0074] S5: Use the trained neural network to identify the AV constitutive model parameters of the experimental curve.

[0075] Following the sample feature point selection method in S2, feature points are selected on the experimental curve. The stress values ​​corresponding to the feature points are normalized and used as input values ​​for the neural network trained in S4. The neural network will output the normalized AV constitutive model parameters corresponding to the experimental curve. The AV constitutive model parameters are then denormalized to obtain the AV constitutive model parameters corresponding to the experimental curve, as shown below:

[0076]

[0077] The predicted stress-strain curves were calculated using the forward modeling method of the AV constitutive model and compared with the experimental curves to verify the accuracy of the prediction results; finally, the parameters of the AV constitutive model were identified.

[0078] The root mean square error (RMSE) was used to calculate the deviation between the predicted and experimental curves. To more objectively observe the fit between the predicted and experimental curves, a dimensionless evaluation index, the coefficient of determination (R²), was introduced. 2 The formulas for the two indicators are shown below:

[0079]

[0080] In the formula: RMSE represents the root mean square error; This indicates the stress value at a characteristic point of the predicted curve; The value represents the stress at a characteristic point of the test curve; n represents the number of characteristic points of the stress-strain curve selected.

[0081]

[0082] In the formula: R 2 Indicates the coefficient of determination; This represents the average stress at the characteristic point.

[0083] The prediction results are shown in Table 2 and Figure 4 The figure shown is a comparison between the predicted curve and the experimental curve for uniaxial tensile conditions in an embodiment of the present invention. Figure 4 It is known that the predicted curve and the experimental curve highly coincide. The root mean square error (RMSE) reflects the deviation between the predicted and experimental values. It is sensitive to extremely large or small errors in the simulation process; the smaller the value, the smaller the deviation between the simulated and experimental values, and the better the simulation results. The coefficient of determination (R²) 2 The coefficient of determination is a statistical indicator used to reflect the reliability of a regression model in explaining changes in the dependent variable. A larger coefficient of determination indicates a stronger correlation between the two variables. The coefficient of determination ranges from 0 to 1. 2 When the value is ≥0.5, it indicates that the two have a strong correlation in a mathematical sense. As shown in Table 2, the root mean square error of the prediction results in this embodiment is only 1.792, and the coefficient of determination is as high as 0.9985, indicating that the prediction curve and the experimental curve have a very high degree of overlap and a very strong correlation.

[0084] The results above show that the AV constitutive model parameter identification method for stainless steel under uniaxial tensile conditions based on neural networks proposed in this invention can obtain the AV constitutive model parameters of the test steel under uniaxial tensile conditions very accurately.

[0085] Table 2. Prediction parameters and curve fitting error statistics for uniaxial tensile conditions.

[0086]

[0087] In summary, the prediction results of the neural network-based AV constitutive model parameter acquisition method for stainless steel materials in this embodiment demonstrate its good performance.

[0088] (1) In this embodiment of the invention, the AV constitutive model is used to describe the stress-strain curve of stainless steel under uniaxial tension. The range of values ​​of the AV constitutive model parameters determined in step S2 is relatively wide, and the AV constitutive model parameters under uniaxial tension of most stainless steel materials can be identified more accurately.

[0089] (2) In this embodiment, a neural network is used to identify constitutive model parameters. Within the parameter value range, a uniformly distributed random number is first generated to form an AV constitutive model parameter combination. The stress-strain sample curve is obtained by forward modeling using the AV constitutive model. The 50,000 sets of sample data used in this embodiment can be generated within one minute, which can greatly save time.

[0090] (3) The embodiments of the present invention cleverly select feature points that can represent the shape of the stress-strain curve to form a feature vector as input data for the neural network; and the method can be applied to stress-strain curves under various working conditions such as uniaxial tension and cyclic deformation that can be described by the AV constitutive model. It is only necessary to select feature points that can represent the characteristics of the stress-strain curves under different working conditions and adjust the range of parameter values. In this way, the AV constitutive model parameters of the stress-strain curves of stainless steel under different loading conditions can be effectively identified.

[0091] (4) Compared with traditional fitting and optimization methods, such as multi-objective genetic algorithms and trial-and-error methods, the method described in this embodiment of the invention has high accuracy in obtaining the AV constitutive model parameters of stainless steel under uniaxial tensile conditions; the root mean square error (RMSE) between the predicted results and the experimental results in the embodiment is as low as 1.792, and the coefficient of determination (R²) is also low. 2 With a value as high as 0.9985, it can effectively and accurately identify the AV constitutive model parameters of stainless steel materials under uniaxial tensile conditions.

[0092] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for obtaining parameters of an AV constitutive model of stainless steel material based on a neural network, characterized in that, It includes the following steps: Step 1: Obtain the uniaxial tensile stress-strain curve of stainless steel material; At room temperature, a uniaxial tensile test was conducted on stainless steel material at a preset loading rate to obtain the uniaxial tensile stress-strain curve of the stainless steel material. Step 2: Establish a sample database for training the neural network; The value ranges of the first, second, third, and fourth parameters in the AV constitutive model under uniaxial tension of stainless steel are determined as follows: the value range of the first parameter in the AV constitutive model under uniaxial tension of stainless steel is (20000, 50000), the value range of the second parameter is (200, 500), the value range of the third parameter is (0, 100), and the value range of the fourth parameter is (0.4, 0.8). Within these value ranges, the AV constitutive model parameters are randomly and uniformly generated to establish N combinations of AV constitutive model parameters. The AV constitutive model is shown below: ； In the formula: Represents the back stress tensor; Represents the plastic strain tensor; Indicates cumulative plastic strain; Indicates the additional follower variable; , , , These represent the first, second, third, and fourth parameters of the stainless steel material in the AV constitutive model, respectively. Stress-strain sample curves were obtained by using the AV constitutive model and the forward modeling method combining the parameters of the N AV constitutive models; On the stress-strain sample curve, n sample feature points that can represent the shape of the stress-strain sample curve are selected according to the set strain interval, and a sample database for training the neural network is established. Step 3: Establish a neural network model for recognizing the parameters of the AV constitutive model; The number of nodes in the input layer of the neural network corresponds to the number of sample feature points n of the stress-strain sample curve, and the number of nodes in the output layer is determined by the four AV constitutive model parameters. It contains three hidden layers, and the number of nodes in the hidden layers is set according to the geometric pyramid rule, with the number of nodes in each layer decreasing continuously from the input layer to the output layer. The activation function used is the Sigmoid function. Step 4: Train the neural network using sample data; The stress values ​​corresponding to the sample feature points of the stress-strain sample curves are normalized and used as the input values ​​of the neural network. The constitutive model parameters corresponding to the stress-strain sample curves are normalized and used as the output values ​​to train the constructed neural network. The normalization principle is that the normalized data should be kept between [0,1]. The stress-strain sample curve feature value normalization method is as follows: ； In the formula: Represents the normalized eigenvalues; Represents the characteristic values ​​of the sample curve; Indicates the stress characteristic scaling factor; The normalization method for the parameters of the AV constitutive model is as follows: ； In the formula: This represents the normalized AV constitutive model parameters; This represents the parameters of the AV constitutive model before normalization; This indicates the maximum value within the range of each parameter; This indicates the minimum value within the range of each parameter; Step 5: Use the trained neural network to identify the AV constitutive model parameters of the experimental curve; Following the sample feature point selection method in step 2, feature points are selected on the experimental curve. The stress values ​​corresponding to the feature points are normalized and used as input values ​​for the neural network trained in step 4. The neural network will output the normalized AV constitutive model parameters corresponding to the experimental curve. The AV constitutive model parameters are then denormalized to obtain the AV constitutive model parameters corresponding to the experimental curve, specifically: ； The predicted stress-strain curves were calculated using the AV constitutive model and compared with the experimental curves to verify the accuracy of the prediction results, thereby obtaining the parameters required for the AV constitutive model of stainless steel.

2. The method for obtaining AV constitutive model parameters of stainless steel material based on neural networks according to claim 1, characterized in that, The specific steps for randomly and uniformly generating the AV constitutive model parameters in step 2 are as follows: Uniformly distributed random numbers are generated within the interval of the AV constitutive model parameters. The probability density function of the random numbers within the interval of the AV constitutive model parameters is shown below: ； In the formula: Represents the probability density function; The x-axis represents the probability density function; The independent variable represents the probability density function; Indicates the lower limit of the parameter range; This indicates the upper limit of the parameter range.

3. The method for obtaining AV constitutive model parameters of stainless steel material based on neural networks according to claim 1, characterized in that, The forward modeling calculation in step 2 specifically involves: using the generated AV constitutive model parameter combination to calculate the stress-strain curve through the constitutive model, with an initial strain value of 0 and a strain increment set to 1e. -5 The maximum strain is 0.3, and the generated AV constitutive model parameters are used at each strain point. , , , The stress values ​​at each strain point are calculated by substituting the combined values ​​into the AV constitutive model, thus obtaining the stress-strain sample curves.

4. The method for obtaining AV constitutive model parameters of stainless steel material based on neural networks according to claim 1, characterized in that, The geometric pyramid rule in step 3 is as follows: from the input layer to the output layer of the neural network structure model, the number of neural network nodes in each layer decreases continuously, and the number of neural network nodes in each layer is reduced by 2 compared to the previous layer, forming a geometric pyramid structure.

5. The method for obtaining AV constitutive model parameters of stainless steel material based on neural networks according to claim 1, characterized in that, In step 1, the preset loading rate is within the range of 0.1% / s to 0.3% / s.

Images