Optimization method of GTN micro-damage model parameters based on GA-BP neural network algorithm

Abstract

The present invention belongs to the field of material analysis technology and discloses a method for optimizing parameters of a GTN mesoscopic damage model based on a GA-BP neural network algorithm. The method comprises the following steps: Step 10, experimentally obtaining a measured stress-strain curve of a material to be tested; Step 20, numerically simulating uniaxial tension of the GTN mesoscopic damage model using finite element analysis to obtain a simulated stress-strain curve, and correcting the damage parameters using a finite element inverse calibration method; Step 30, varying the maximum stress and maximum strain, and repeating uniaxial tension tests using finite element analysis to obtain multiple sets of sample data; Step 40, inputting a training set of data into the GA-BP neural network model to obtain an optimized BP neural network model; and Step 50, inputting a test set of data into the optimized BP neural network model, determining the optimal test set of data, and using the corresponding output as the final GTN mesoscopic damage model parameters. The present invention reduces simulation and correction time and improves the accuracy and universality of damage parameter determination.

Description

Technical Field

[0001] The present invention belongs to the technical field of material analysis, and in particular relates to a method for optimizing parameters of a Gurson-Tvergaard-Needleman (GTN) microscopic damage model based on a GA-BP neural network algorithm. Background Art

[0002] Traditional plasticity theory assumes that the yield surface of a material has nothing to do with the macroscopic hydrostatic stress. The yield surface of the GTN micro-damage model takes into account the influence of hydrostatic stress and links the yield of the material with damage. This allows the yield surface of the material to gradually decrease with increasing pore volume fraction, describing the process of continuous material deterioration as damage develops during deformation. The yield condition also considers the interaction between pores, making the prediction results more accurate.

[0003] GTN plasticity relationship:

[0004]

[0005] Where: is the equivalent stress, is the stress deviator tensor, is the hydrostatic stress, σ y is the flow stress of the matrix material, q1, q2, q3 are used as correction parameters of the GTN micro-damage model, f * It is used to explain the fracture of the final material caused by the aggregation of voids; its expression is:

[0006]

[0007] Where: f is the current pore volume fraction, f c is the critical volume fraction at which pores begin to aggregate, f f is the void volume fraction when the material finally fails;

[0008] For ductile materials, the anisotropic properties caused by damage are not obvious, so the damage of the GTN model is considered to be isotropic, and the total damage is expressed as the change of the pore volume fraction including the growth of the initial pore (df growth ) and new hole nucleation (df nucleation ), the evolution law of the pore volume fraction is as follows:

[0009] df=df growth +df nucleation

[0010] df growth =(1-f)dε ii p

[0011] df nucleation =Adε m p

[0012] The nucleation strain follows a normal distribution with respect to its mean value, so the strain-dominated nucleation evolution law is:

[0013]

[0014] Where: dε ii p is the plastic strain tensor, f N is the value of the nucleation part in the pore volume fraction, A is the pore nucleation coefficient, ε N is the average equivalent plastic strain at the time of hole initiation, s N is the standard deviation of the normal distribution.

[0015] Determining the parameters of the GTN damage model is a complex problem. A comprehensive range of methods has been developed within the research community, primarily including finite element back-calibration and micromorphology analysis. However, each method has its own shortcomings. The finite element back-calibration method preliminarily estimates damage parameters based on previous experience and then refines them through numerical simulation. However, this method is inherently time-consuming and requires extensive simulation and correction to obtain damage parameters within acceptable tolerances. Micromorphology analysis uses scanning electron microscopy to analyze the specimen surface at different stages during loading, calculating the void volume fraction at each stage. However, surface analysis cannot fully reflect internal phenomena, resulting in inaccurate results. Therefore, a method for obtaining damage parameters that is both convenient and accurate remains to be explored. The relationship between the experimental stress-strain curve and the parameters of the GTN microscopic damage model is an extremely complex and highly nonlinear relationship. Appropriate GTN damage parameters are crucial for accurately predicting the location and load at which fracture will occur. Researchers have gradually introduced optimization methods based on these two methods to optimize these parameters, but this research is still in its infancy, and practical application of these optimization methods remains to be explored.

[0016] BP neural networks, based on biomimetic principles, possess powerful self-learning and prediction capabilities. They use error inverse adjustment to build a model describing the complex nonlinear relationship between decision variables X and objective functions Y. However, in practical applications, BP neural networks have slow convergence speeds and are prone to local minima. Summary of the Invention

[0017] The present invention overcomes the deficiencies of the prior art and aims to solve the following technical problem: providing an optimization method for the parameters of a GTN mesoscopic damage model based on a GA-BP neural network to accurately determine the parameters of the GTN mesoscopic damage model.

[0018] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for optimizing the parameters of the GTN micro-damage model based on the GA-BP neural network algorithm, comprising the following steps:

[0019] Step 10: Experimentally obtain the measured stress-strain curve of the material to be tested, as well as the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f ;

[0020] Step 20: According to the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f Finite element analysis is used to perform uniaxial tensile numerical simulation on the GTN mesoscopic damage model to obtain a simulated stress-strain curve. The finite element reverse calibration method is used to correct the damage parameters until the deviation between the simulated stress-strain curve and the experimental stress-strain curve is less than a threshold value, thereby obtaining the correction parameters and void nucleation parameters of the modified GTN mesoscopic damage model; the correction parameters include the first correction parameter q1, the second correction parameter q2 and the third correction parameter q3, and the void nucleation parameter includes the average equivalent plastic strain ε at the time of void initiation N , the standard deviation s of the normal distribution N and the volume fraction of nucleation pores f N ;

[0021] Step 30: Based on the correction parameters and pore nucleation parameters obtained in step 20, the initial pore volume fraction f0 and the critical pore volume fraction f are changed by the control variable method. c , fracture void volume fraction f f , nucleation hole volume fraction f N , multiple groups of uniaxial tensile tests were simulated using finite element analysis to obtain the corresponding maximum stress and maximum strain, and multiple groups of sample data were obtained, including the maximum stress and maximum strain, as well as the corresponding initial void volume fraction f0, critical void volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N ;

[0022] Step 40: Use the maximum stress and maximum strain as network inputs, and the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f NAs the network output, a GA-BP neural network model is built; the sample data is divided into a training set and a test set, and the nonlinear mapping relationship between the GTN damage parameters and stress and strain is established using the training set data. Then, the corresponding output of the training set data is obtained using the optimized GA-BP neural network model;

[0023] Step 50: The output of the GA-BP neural network model is compared with the initial pore volume fraction f0 and the critical pore volume fraction f corresponding to the test set data. c , fracture void volume fraction f f , nucleation hole volume fraction f N The optimal test set data is determined according to its error, and the network output corresponding to the optimal test set data is used as the final GTN micro-damage model parameter.

[0024] The step 10 comprises the following steps:

[0025] Step 101: Cutting the metal material to be tested into a uniaxial tensile specimen;

[0026] Step 102: performing a uniaxial tensile test on the uniaxial tensile specimen along the rolling direction to obtain elastic-plastic parameters and a measured stress-strain curve of the metal material to be tested;

[0027] Step 103: Wire cutting, manually grinding to metallographic standards, and polishing are performed on the necking area of ​​the tensile specimen before deformation and the tensile specimen, and the microscopic morphology of micropore aggregation in the fracture area of ​​the tensile specimen is analyzed by scanning electron microscopy to obtain microscopic morphology photographs;

[0028] Step 104: Identify the microscopic morphology photos, calculate the pore volume fractions at each stage, and obtain the initial pore volume fraction f0, the critical pore volume fraction f c , volume fraction at final fracture f f .

[0029] It is characterized in that the step 20 specifically includes:

[0030] Step 201: Obtain GTN damage model correction parameters q1, q2, q3, and hole nucleation parameter ε N 、S N 、f N The initial pore volume fraction f0 and the critical pore volume fraction f are obtained by combining the experimental results. c , fracture void volume fraction f f The value of , the uniaxial tensile numerical simulation of the GTN mesoscopic damage model was carried out using finite element analysis to obtain the simulated stress-strain curve;

[0031] Step 202: Based on the theory of microscopic damage mechanics and void evolution law, combined with the measured plastic deformation stress-strain relationship, the finite element reverse calibration method is used to repeatedly correct the damage parameters until the deviation between the simulated stress-strain curve and the experimental stress-strain curve is less than a threshold, thereby obtaining the correction parameters and void nucleation parameters of the corrected GTN microscopic damage model.

[0032] In step 30, 10 to 20 groups of sample data are obtained.

[0033] In step 40, 70-90% of the data are selected as training set data.

[0034] In step 50, the test set data with the lowest error percentage is selected as the optimal test set data.

[0035] In step 30, multiple groups of uniaxial tensile tests are performed using the control variable method. When multiple groups of sample data are obtained, the change rate of the maximum stress is less than 35%, and the change rate of the maximum strain is less than 5%.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The present invention provides a method for optimizing the parameters of a GTN micro-damage model based on a GA-BP neural network. First, the micro-morphology analysis method and the finite element reverse calibration method are used to preliminarily formulate the GTN model damage parameters based on the theoretical knowledge of micro-damage mechanics and pore evolution laws, combined with the measured plastic deformation stress-strain relationship. Then, a BP neural network model optimized by a genetic algorithm is constructed, and a training set and test set are formed based on a large amount of data to obtain a nonlinear input-output relationship between the damage parameters and the maximum stress and strain. Finally, the output with the lowest error percentage is selected from the test set to determine the optimal GTN micro-damage model parameters. Compared with existing parameter determination methods, the present invention integrates the disciplines of biomimetics, computer science, and materials science, reduces the time for simulation correction, improves the accuracy and universality of damage parameter determination, and promotes comprehensive development.

[0038] 2. The method for determining the parameters of the GTN micro-damage model based on GA-BP neural network optimization can establish a nonlinear mapping relationship between GTN damage parameters and stress and strain by analyzing a large amount of data, thereby inferring the output results based on the input-output relationship, optimizing the GTN model damage parameters and improving their accuracy, which has important practical significance for accurately predicting material fracture problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A diagram showing the structure of a specimen for a uniaxial tensile test according to an embodiment of the present invention;

[0040] Figure 2Measured stress-strain curve of uniaxial tensile test of an example of the present invention;

[0041] Figure 3 The micromorphology analysis diagrams at different stages of the uniaxial tensile test of the present invention are shown; (a) is the initial hole, (b) is the critical hole, and (c) is the fracture hole;

[0042] Figure 4 A comparison diagram of the stress-strain curve measured by the uniaxial tensile test of an example of the present invention and the numerically simulated stress-strain curve based on the preliminarily proposed GTN damage parameters;

[0043] Figure 5 The algorithm flow chart of the BP neural network optimized by the genetic algorithm of the embodiment of the present invention;

[0044] Figure 6 This is a comparison chart of the stress-strain curve measured in the uniaxial tensile test of an example of the present invention and the numerical simulation stress-strain curve of the GTN damage parameters optimized based on the genetic algorithm and BP neural network. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] The embodiment of the present invention provides a method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm, comprising the following steps:

[0047] Step 10: Experimentally obtain the measured stress-strain curve of the material to be tested, as well as the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f .

[0048] The step 10 comprises the following steps:

[0049] Step 101: Cutting the metal material to be tested into a uniaxial tensile specimen;

[0050] Step 102: performing a uniaxial tensile test on the uniaxial tensile specimen along the rolling direction to obtain elastic-plastic parameters and a measured stress-strain curve of the metal material to be tested;

[0051] Step 103: Wire cutting, manually grinding to metallographic standards, and polishing are performed on the necking area of ​​the tensile specimen before deformation and the tensile specimen, and the microscopic morphology of micropore aggregation in the fracture area of ​​the tensile specimen is analyzed by scanning electron microscopy to obtain microscopic morphology photographs;

[0052] Step 104: Identify the microscopic morphology photos, calculate the pore volume fractions at each stage, and obtain the initial pore volume fraction f0, the critical pore volume fraction f c , volume fraction at final fracture f f .

[0053] Specifically, in step 10, AZ31 magnesium alloy is used as the research object, and the hot-rolled plate is used as the original billet. Its heat treatment state is the fully annealed state (350°C×12h). The grain size of the treated AZ31 magnesium alloy plate is uniform and is basically equiaxed grains.

[0054] According to the national standard GBT 4338-2006 High-temperature tensile test method for metallic materials, a plate-shaped tensile specimen with a thickness of 1 mm was cut from the magnesium alloy sheet along the rolling direction using an electric spark wire cutting machine. Figure 1 As shown, the electronic universal tensile testing machine was heated to 200 ° C, the sample was placed in it, kept warm for 5 minutes, and the strain rate was 10 -3 The loading direction is consistent with the rolling direction until the fracture damage, and the deformation is completed by air cooling and unloading to obtain the elastic-plastic parameters of the metal material and the measured stress-strain curve, such as Figure 2 shown.

[0055] Repeat the above experiment and pause when the stress reaches the maximum value. This stage is the necking process during plastic deformation. The samples before stretching and when necking occur are cut along the axial direction with an electric spark cutting machine, and manually polished to the metallographic standard with sandpaper of different mesh sizes. The samples are then polished with a polishing machine and cleaned with acetone solution to remove surface impurities and oil stains. Finally, the micromorphology of the above two samples and the fracture surface is analyzed, as shown in the figure. Figure 3 shown.

[0056] Image-Pro Plus software was used to identify the microscopic morphology photos, and the void volume fractions at each stage were counted. The initial void volume fraction f0 = 0.0003, the critical void volume fraction f c =0.0075, volume fraction at final fracture f f =0.06.

[0057] Step 20: Preliminary formulation of other parameters of the GTN micro-damage model: Based on the initial void volume fraction f0, the critical void volume fraction f c , fracture void volume fraction f fFinite element analysis is used to perform uniaxial tensile numerical simulation on the GTN mesoscopic damage model to obtain a simulated stress-strain curve. The finite element reverse calibration method is used to correct the damage parameters until the deviation between the simulated stress-strain curve and the experimental stress-strain curve is less than a threshold value, thereby obtaining the correction parameters and void nucleation parameters of the modified GTN mesoscopic damage model; the correction parameters include the first correction parameter q1, the second correction parameter q2 and the third correction parameter q3, and the void nucleation parameter includes the average equivalent plastic strain ε at the time of void initiation N , the standard deviation s of the normal distribution N and the volume fraction of nucleation pores f N .

[0058] In this embodiment, the elastic-plastic parameters and f0=0.0003, f c =0.0075, f f =0.06 and the GTN damage model correction parameters q1, q2, q3, and the hole nucleation parameter ε obtained from relevant literature N 、S N 、f N The nine damage parameters were input into the finite element analysis software together, and the uniaxial tension numerical simulation based on the GTN mesoscopic damage model was applied.

[0059] Based on the theoretical knowledge of microscopic damage mechanics and void evolution law, combined with the measured plastic deformation stress-strain relationship, the finite element reverse calibration method was used to repeatedly correct the damage parameters. Finally, the damage parameters of the GTN model were preliminarily formulated as follows: correction parameters q1 = 1.5, q2 = 1, q3 = 2.25, void nucleation parameter ε N =0.15, S N =0.05, f N =0.0045, the stress-strain curves obtained by numerical simulation and measured results are compared as follows Figure 4 Wherein, σ1 and ε1 are the maximum stress and strain values ​​obtained by numerical simulation; σ2 and ε2 are the maximum stress and strain values ​​of the measured results.

[0060] Step 30: Corrected parameters q1, q2, q3 and pore nucleation parameter ε obtained in step 20 N 、S N By controlling the variable method, the initial pore volume fraction f0 and the critical pore volume fraction f are changed respectively. c , fracture void volume fraction f f , nucleation hole volume fraction f N, using finite element analysis simulation to conduct multiple groups of uniaxial tensile tests, obtain the corresponding maximum stress and maximum strain, obtain multiple groups of sample data, and when obtaining sample data by the control variable method, the change rate of the obtained maximum stress should be less than 35%, and the change rate of the maximum strain should be less than 5%; the sample data includes the maximum stress and maximum strain, as well as the corresponding initial void volume fraction f0, critical void volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N .

[0061] According to the controlled variable method, 20 groups of uniaxial tensile tests were simulated by finite element method, and the sample data obtained are shown in Table 1 below.

[0062] Table 1 Sample data

[0063]

[0064]

[0065] Step 40: Use the maximum stress and maximum strain as network inputs, and the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N As the network output, a BP neural network model is built; the weight coefficients of the BP neural network model are optimized through a genetic algorithm to obtain a neural network optimization model based on GA-BP; the sample data are randomly divided into a training set and a test set, and the training set data is used for repeated training to establish a nonlinear mapping relationship between GTN damage parameters and stress and strain, and then the optimized GA-BP neural network model is used to obtain the corresponding output of the training set data.

[0066] In this embodiment, the maximum strain and maximum stress obtained from the uniaxial tensile test are used as the input of the GA-BP neural network, and the initial pore volume fraction f0, the critical pore volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N As the output of the GA-BP neural network.

[0067] In this embodiment, the training sample data of the 20 groups of uniaxial tensile simulations with different damage parameters are randomly selected, 16 groups are used as training sets, and 4 groups are used as test sets. All data are normalized to [-1.1] using the mapminmax function, and the number of hidden layers of the BP neural network model is set to 15, the number of training times is 1000, the training target is 1e-5, and the learning rate is 0.01. The genetic algorithm is introduced into the BP neural network model of this embodiment to initialize the parameters, set the evolutionary generation to 20, the population size to 30, the crossover probability to 0.7, and the mutation probability to 0.05, and the population is real-coded and the sum of the absolute values ​​of the errors between the predicted data and the expected data is used as the fitness value. The roulette wheel method is used to select the best from the population by selecting the genetic operator, the crossover genetic operator is used to obtain a progeny group with a higher average fitness, the mutation genetic operator is used to compensate for the lack of population diversity, the fitness operation is calculated, and finally the optimal weights and thresholds are obtained by running to the number of evolutions. The optimal weights and thresholds are used to construct a BP neural network to obtain a nonlinear mapping relationship with better fitting. The algorithm flow chart is as follows: Figure 5 shown.

[0068] Step 50: Determine the final GTN micro-damage model parameters: The output of the GA-BP neural network model is compared with the initial void volume fraction f0 and the critical void volume fraction f corresponding to the test set data. c , fracture void volume fraction f f , nucleation hole volume fraction f N The optimal test set data is determined according to its error, and the network output corresponding to the optimal test set data is used as the final GTN micro-damage model parameter.

[0069] In this embodiment, the test set data with the lowest error percentage is selected as the optimal test set data, and the network output corresponding to the optimal test set data is used to determine the optimal GTN micro-damage model parameters, that is, the model parameters are: f0 = 0.00029624, f c＝ 0.0075, f n =0.0059, f f =0.06, q1=1.5, q2=1, q3=2.25, ε N =0.15, S N =0.05, the optimized damage parameters are introduced into the finite element simulation of the uniaxial tensile test, and the stress-strain curves obtained by the simulation results are compared with those obtained by the measured results, as shown in Figure 2. Figure 6 As shown, σ1 and ε1 are the maximum stress and strain values ​​of the measured results; σ2 and ε2 are the maximum stress and strain values ​​after optimization based on the GA-BP neural network, which shows that the optimization method of the GTN micro-damage model parameters based on the GA-BP neural network according to the embodiment of the present invention can fit the measured results well.

[0070] The present invention provides a method for optimizing the parameters of a GTN microscopic damage model based on a GA-BP neural network algorithm, which cross-compatibly integrates the disciplines of biomimetics, computer science, and materials science, reduces the time for simulation correction, improves the accuracy and universality of damage parameter determination, and enables the acquisition of a more accurate GTN model.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm, characterized in that: The following steps are involved: Step 10: Experimentally obtain the measured stress-strain curve of the material to be tested, as well as the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f ; Step 20: According to the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f Finite element analysis is used to perform uniaxial tensile numerical simulation on the GTN micro-damage model to obtain a simulated stress-strain curve. The finite element reverse calibration method is used to correct the damage parameters until the deviation between the simulated stress-strain curve and the experimental stress-strain curve is less than a threshold value, and the correction parameters and pore nucleation parameters of the corrected GTN micro-damage model are obtained; the correction parameters include the first correction parameter q1, the second correction parameter q2 and the third correction parameter q 3， The void nucleation parameters include the average equivalent plastic strain ε at the time of void initiation N , the standard deviation s of the normal distribution N and the volume fraction of nucleation pores f N ; Step 30: Based on the correction parameters and pore nucleation parameters obtained in step 20, the initial pore volume fraction f0 and the critical pore volume fraction f are changed by the control variable method. c , fracture void volume fraction f f , nucleation hole volume fraction f N , multiple groups of uniaxial tensile tests were simulated using finite element analysis to obtain the corresponding maximum stress and maximum strain, and multiple groups of sample data were obtained, including the maximum stress and maximum strain, as well as the corresponding initial void volume fraction f0, critical void volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N ; Step 40: Use the maximum stress and maximum strain as network inputs, and the initial pore volume fraction f0 and the critical pore volume fraction f c , fracture void volume fraction f f , nucleation hole volume fraction f N As the network output, a GA-BP neural network model is built; the sample data is divided into a training set and a test set, and the nonlinear mapping relationship between the GTN damage parameters and stress and strain is established using the training set data. Then, the corresponding output of the training set data is obtained using the optimized GA-BP neural network model; Step 50: The output of the GA-BP neural network model is compared with the initial pore volume fraction f0 and the critical pore volume fraction f corresponding to the test set data. c , fracture void volume fraction f f , nucleation hole volume fraction f N The optimal test set data is determined according to its error, and the network output corresponding to the optimal test set data is used as the final GTN micro-damage model parameter.

2. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: The step 10 comprises the following steps: Step 101: Cutting the metal material to be tested into a uniaxial tensile specimen; Step 102: performing a uniaxial tensile test on the uniaxial tensile specimen along the rolling direction to obtain elastic-plastic parameters and a measured stress-strain curve of the metal material to be tested; Step 103: Wire cutting, manually grinding to metallographic standards, and polishing are performed on the necking area of ​​the tensile specimen before deformation and the tensile specimen, and the microscopic morphology of micropore aggregation in the fracture area of ​​the tensile specimen is analyzed by scanning electron microscopy to obtain microscopic morphology photographs; Step 104: Identify the microscopic morphology photos, calculate the pore volume fractions at each stage, and obtain the initial pore volume fraction f0, the critical pore volume fraction f c , volume fraction at final fracture f f .

3. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: The step 20 specifically includes: Step 201: Obtain GTN damage model correction parameters q1, q2, q3, and hole nucleation parameter ε N 、S N 、f N The initial pore volume fraction f0 and the critical pore volume fraction f are obtained by combining the experiment. c , fracture void volume fraction f f The value of , the uniaxial tensile numerical simulation of the GTN mesoscopic damage model was carried out using finite element analysis to obtain the simulated stress-strain curve; Step 202: Based on the theory of microscopic damage mechanics and void evolution law, combined with the measured plastic deformation stress-strain relationship, the finite element reverse calibration method is used to repeatedly correct the damage parameters until the deviation between the simulated stress-strain curve and the experimental stress-strain curve is less than a threshold, thereby obtaining the correction parameters and void nucleation parameters of the corrected GTN microscopic damage model.

4. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: In step 30, 10 to 20 groups of sample data are obtained.

5. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: In step 40, 70-90% of the data are selected as training set data.

6. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: In step 50, the test set data with the lowest error percentage is selected as the optimal test set data.

7. The method for optimizing parameters of a GTN micro-damage model based on a GA-BP neural network algorithm according to claim 1, characterized in that: In step 30, multiple groups of uniaxial tensile tests are performed using the control variable method. When multiple groups of sample data are obtained, the change rate of the maximum stress is less than 35%, and the change rate of the maximum strain is less than 5%.

Images