CN115588470A - Crystal plasticity constitutive parameter calibration method based on two-channel convolutional neural network - Google Patents

Abstract

The invention provides a crystal plasticity constitutive parameter calibration method based on a two-channel convolutional neural network, which comprises the following steps: firstly, establishing a magnesium alloy model by using Neper software, then establishing a magnesium alloy model with a texture by using MTEX, and then establishing a crystal plasticity simulation file by using Python; inputting the crystal plasticity simulation file into DAMASK simulation software for simulation to obtain an IPF (intrinsic stress) -true strain coordinate data and an IPF diagram; then cutting the IPF graph to obtain the IPF _S Drawing (1) toDrawing a true stress-true strain curve graph by using true stress-true strain coordinate data; then carrying out normalization processing on constitutive parameters; IPF _S The method comprises the steps of taking a graph and a true stress-true strain curve graph as input, taking normalized constitutive parameters as output, and establishing a universal two-channel convolution neural network model for calibrating crystal plastic constitutive parameters through training, wherein the degree of fitting between the parameters calibrated by the model and the parameters measured by experiments is more than 99%; the method provided by the invention has high efficiency and good accuracy, and is superior to the traditional algorithm.

Description

Calibration Method of Crystal Plastic Constitutive Parameters Based on Dual-Channel Convolutional Neural Network

technical field

The invention belongs to the field of performance detection of metal materials, and in particular relates to a method for calibrating constitutive parameters of crystal plasticity based on a double-channel convolutional neural network.

Background technique

Crystal plasticity is an important calculation tool for the deformation of crystalline materials. It can simulate the metal deformation process at the mesoscopic scale, including grain orientation, grain twist, texture change, etc. It is more and more widely used in the field of material research. However, since crystal plasticity involves many constitutive parameters of materials, its parameter identification and calibration are difficult and challenging. The traditional identification methods of crystal plastic constitutive parameters mainly use genetic algorithm, particle swarm algorithm, simulated annealing and other optimization methods. The optimization work takes a long time, low efficiency, and high trial and error costs. Operators need to have a deep understanding of the optimization algorithm and certain Experience. Therefore, it is urgent to establish a calibration method for crystal plastic constitutive parameters, which can significantly reduce the workload of optimization algorithms and improve optimization efficiency.

Contents of the invention

Based on this, the present invention provides a method for calibrating constitutive parameters of crystal plasticity based on dual-channel convolutional neural network, which includes the following steps:

(1) Use Neper software to establish a magnesium alloy model: input m group parameters x and y to Neper software, wherein x and y represent the random distribution of grains and the roundness of grains respectively, and the value range of m is 100-100000 and is an integer, the value range of x is 0-1, and the value range of y is 0.3-1.0, and the magnesium alloy model of group m is obtained, which is recorded as group m (x, y), and the model of each group (x, y) is as follows Shown in matrix 1;

matrix 1

i is one of 1, 2 or 3;

(2) Use the toolbox MTEX of Matlab to establish the magnesium alloy model with added texture: set up m group d, e, f, g, h, k parameters, d, e, f, g, h, k represent steps (1 ) of the six different texture strengths of magnesium alloys, the value ranges of d, e, f, g, h, k are all 0-1, and d+e+f+g+h+k=1; Input the m group d, e, f, g, h, k parameters and the m group (x, y) obtained in step (1) into MTEX, and use the Bunge formula of MTEX to obtain the m group added texture magnesium alloy Model (d, e, f, g, h, k) (x, y), recorded as m groups (d, e, f, g, h, k) (x, y), each group (d, e, The data model of f, g, h, k)(x, y) is shown in matrix 2;

Matrix 2

(3) Use Python to obtain m groups of crystal plasticity simulation files: select 16 parameters of crystal plasticity constitutive: basal plane slip critical shear stress T1, cylinder slip critical shear stress T2, conical surface <a> slip critical Slitting stress T3, critical slitting stress T4 of conical surface <c+a> slip, basal surface slip saturated slip slitting stress T5, cylindrical slip saturated slip slitting stress T6, conical surface <a> slip Saturated slip shear stress T7, cone surface <c+a> slip saturated slip shear stress T8, tensile twin critical shear stress T9, compression twin critical shear stress T10, initial twinning and twinning Hardening modulus T11, initial hardening modulus T12 of slip and slip, initial hardening modulus T13 of slip and twinning, sliding strain rate sensitive parameter T14, twinning strain rate sensitive parameter T15, slip hardening parameter T16; T1 The range of T2 is 5-80, the range of T2 is 10-150, the range of T3 is 20-200, the range of T4 is 50-500, the range of T5 is 10-160, the range of T6 is 20-300, the range of T7 40-400, the range of T8 is 100-1000, the range of T9 is 5-250, the range of T10 is 10-400, the range of T11 is 10-500, the range of T12 is 50-3000, and the range of T13 is 50 -1200, the range of T14 is 1-15, the range of T15 is 5-80, and the range of T16 is 0.5-10; within the above range design m groups T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16, each group T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, After T16 is added to a set of magnesium alloy models (d, e, f, g, h, i) (x, y) in step (2) through Python, a set of crystal plasticity simulation files is obtained, and finally m sets of crystal plasticity are obtained Simulation files, the form of each group of crystal plasticity simulation files is shown in matrix 3;

Matrix 3

(4) Obtain the IPF diagram and the simulated true stress-true strain coordinate data through simulation: DAMASK is used as the crystal plasticity finite element simulation software, and the crystal plasticity simulation file obtained in step (3) is input for the crystal plasticity finite element simulation, and the step ( 3) Input the m group of crystal plasticity simulation files into DAMASK software for m group simulation. DAMASK simulation follows the crystal plasticity constitutive equations. After the simulation is completed, the m group simulation results are obtained. Use Python to post-process the m group simulation results to obtain The IPF diagram of group m magnesium alloy and the simulated true stress-true strain coordinate data of group m;

The described crystal plasticity constitutive equations:

为晶体塑性剪切速率，值为1，

is the crystal plastic shear rate, the value is 1,

为初始剪切速率，值为0.001，

is the initial shear rate, the value is 0.001,

n is the initial hardening modulus of tension and compression twins, n is one of T9 and T10,

为临界分切应力，

代表4个不同滑移系的临界分切应力，

为T1，T2，T3，T4中的一种，

is the critical shear stress,

represents the critical shear stress of four different slip systems,

One of T1, T2, T3, T4,

h ₀ is the initial hardening modulus value between three different slip systems, between slip systems and twin systems, h ₀ is one of T11, T12, T13,

为4个不同滑移系的饱和滑移分切应力，

为T5，T6，T7，T8中的一种，

is the saturated slip shear stress of four different slip systems,

One of T5, T6, T7, T8,

q ^αβ is the coefficient between different slip systems α and β, q ^αβ is one of T14, T15, T16,

H ^αβ is the total hardening modulus;

(5) Use Python to obtain the IPFs diagram and the true stress-true strain curve diagram, and normalize the crystal plastic constitutive parameters: the m group of 16 crystal plastic constitutive parameters T1, T2, T3 in step (3) , T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 are normalized respectively to obtain m group normalized t1, t2, t3, t4, t5, t6 , t7, t8, t9, t10, t11, t12, t13, t14, t15, t16, use Python to draw the m sets of simulated true stress-true strain coordinate data in step (4) as m sets of true stress-true strain Curve diagram, the m group IPF figure in the step (4) is cut into the m group standard IPF figure of standard 50x50 pixel, is recorded as the m group IPFs figure; finally obtains the m group crystal plasticity data set, and each described data set is : IPFs diagram, true stress-true strain curve diagram and normalized t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16;

The normalization formula described is:

W _norm is one of t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16 parameters,

W is one of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 parameters,

W _min is the smallest number in the m group of parameters,

W _max is the largest number in the m group of parameters;

As a result, IPFs diagrams, true stress-true strain curve diagrams and large data samples of corresponding normalized constitutive parameters were established;

(6) training dual-channel convolutional neural network model: with the m group IPF _S figure and true stress-true strain curve figure described in step (5) as the input of training dual-channel convolutional neural network model, step (5) The t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16 of the m group are used as the output of the neural network to make a dual-channel convolutional neural network model , be randomly divided into p group training set and q group verification set with 5:1-30:1 ratio; Described dual-channel convolutional neural network model comprises: 5 layers of alxnet model convolution layers, 3 fully connected layers; Described The convolutional layer is: input p groups of IPF _S diagrams and true stress-true strain curves in the training set into the convolutional layer, after the convolution operation of the convolutional layer, the image is abstracted into a feature map, p groups of IPF After the _S map is processed by 5 layers of convolutional layers, the characteristic map 1 of group p is obtained, and the true stress-true strain curve of group p is processed by 5 layers of convolutional layers to obtain the characteristic map 2 of group p. Feature map 2 merges input fully connected layer processing; described fully connected layer processing is: obtain p group of output values by calculating the forward propagation formula containing weights and offsets, the weights and offsets are random, The range of weight is 1-2, the range of bias is 10-20, and the output value of each group is: t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16', then calculate p groups t1', t2', t3', t4', t5', t6', t7', t8 ', t9', t10', t11', t12', t13', t14', t15', t16' with p groups t1, t2, t3, t4, t5, t6, t7, t8, t9, t10 in the training set , the overall error between t11, t12, t13, t14, t15, and t16; then recalculate the weight and bias according to the overall error, and then recalculate using the forward propagation formula, and reciprocate until the overall error is 0.2%- Finish training after 1%, obtain the dual-channel convolutional neural network model that training is completed; Wherein the initial learning rate of described dual-channel convolutional neural network model is set to 0.00001-0.01, and the number of learning is 1000-3000 times; The dual-channel convolutional neural network model that training is completed The overall error of the channel convolutional neural network model for the p-group training set is 0.2%-1%, and the overall error of the trained dual-channel convolutional neural network model for the q-group verification set is 0.4%-1.1%;

The overall error formula is:

p is the number of training set groups,

σ is the overall error,

ν is a positive integer, the value is 1-16,

t' _v for t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16 one of ',

t _v is one of t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16,

The overall error, weight, and offset calculation formula:

σ＝ ^WT +b

σ is the overall error,

W is the weight,

T is the transpose symbol,

b is the bias.

Further, the magnesium alloy is one of AZ31, AZ61, AZ91, ZK60 or AM60;

Beneficial effects of the present invention:

Machine learning is applied in the macroscopic fields such as machine vision and speech recognition, but it is rarely used in the microscopic field. The present invention adopts the convolutional neural network in machine learning as a tool for solving nonlinear problems, and utilizes its higher efficiency and Intelligent advantages, borrowing the current advanced algorithm of machine learning, based on big data to establish a convolutional neural network model for the calibration of crystal plastic constitutive parameters, effectively changing the current problems of low-efficiency acquisition and reliance on optimization algorithms for crystal plastic constitutive parameter calibration.

The present invention uses the IPFs diagram and the true stress-true strain curve diagram as input and the normalized constitutive parameter as the output training by constructing the IPFs diagram and the true stress-true strain curve diagram and the large data samples corresponding to the normalized constitutive parameters Dual-channel convolutional neural network, the error of the trained neural network model for the training set is 0.2%-1%, and the error for the verification set is 0.4%-1.1%; the magnesium alloy anisotropy obtained by simulating the neural network of the present invention Coefficient R', section reduction rate and magnesium alloy anisotropy coefficient R, section reduction rate obtained by experiment are comparatively analyzed, and the relative parameter obtained by the present invention and the parameter fitting degree of experiment acquisition are 99.1-99.7%; Compared with traditional Genetic algorithm, particle swarm algorithm, simulated annealing and other methods, the advantage of the present invention is that it can calibrate 16 crystal plastic constitutive parameters at one time, the calibration speed is fast, the calibration accuracy is high, and the neural network model after training is highly universal , which can be applied to the calibration of crystal plastic constitutive parameters of the whole magnesium alloy system.

Description of drawings

FIG. 1 is a diagram of a dual-channel convolutional neural network model used in step S6 in Example 1.

Specific implementation method

The present invention will be further described below in combination with specific embodiments and accompanying drawings.

Example 1

Selecting the AZ31 magnesium alloy, the method of calibrating the crystal plastic constitutive parameters of the AZ31 magnesium alloy by establishing a dual-channel convolutional neural network includes the following steps:

S1. Use Neper software to build an AZ31 magnesium alloy model, and input 10,000 sets of parameters x and y to the Neper software, where x and y represent the random distribution of grains and the roundness of grains, and the value range of x is 0.1-0.8, The value range of y is 0.3-0.9, and 10,000 sets of magnesium alloy models are obtained, which are recorded as 10,000 sets (x, y), and the model of each set (x, y) is shown in matrix 1;

matrix 1

i is one of 1, 2 or 3;

S2. Use the toolbox MTEX of Matlab to establish the AZ31 magnesium alloy model with added texture: set up 10000 groups of d, e, f, g, h, k parameters, d, e, f, g, h, k respectively represent the steps in S1 The six different texture strengths of the magnesium alloy, the value ranges of d, e, f, g, h, k are all 0-1, and d+e+f+g+h+k=1; 10000 Group d, e, f, g, h, k parameters and 10000 groups (x, y) obtained in step S1 are input into MTEX, and 10000 groups of magnesium alloy models with added texture (d, e, f, g, h, k) (x, y), recorded as 10000 groups (d, e, f, g, h, k) (x, y), each group (d, e, f, g, h, k)(x, y) data model is shown in matrix 2;

Matrix 2

S3. Use Python to obtain 10,000 sets of crystal plastic simulation files: select 16 parameters of the crystal plastic constitutive: basal plane slip critical shear stress T1, cylindrical slip critical shear stress T2, conical surface <a> slip critical split Shear stress T3, cone surface <c+a> slip critical shear stress T4, basal surface slip saturated slip shear stress T5, cylindrical slip saturated slip shear stress T6, cone surface <a> slip Saturated slip shear stress T7, cone surface <c+a> slip saturated slip shear stress T8, tensile twin critical shear stress T9, compression twin critical shear stress T10, initial hardening of twins and twins Modulus T11, initial hardening modulus T12 of slip and slip, initial hardening modulus T13 of slip and twinning, sliding strain rate sensitive parameter T14, twinning strain rate sensitive parameter T15, slip hardening parameter T16; The range is 5-40, the range of T2 is 10-100, the range of T3 is 20-100, the range of T4 is 50-250, the range of T5 is 10-100, the range of T6 is 20-150, and the range of T7 is 40-200, the range of T8 is 100-500, the range of T9 is 5-150, the range of T10 is 10-200, the range of T11 is 10-250, the range of T12 is 50-1500, and the range of T13 is 50- 600, the range of T14 is 1-10, the range of T15 is 5-40, and the range of T16 is 0.5-8; design 10000 groups of T1, T2, T3, T4, T5, T6, T7, T8, T9 within the above range , T10, T11, T12, T13, T14, T15, T16, each group T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 After adding to a set of magnesium alloy models (d, e, f, g, h, i) (x, y) in step S2 through Python, a set of crystal plasticity simulation files is obtained, and finally 10,000 sets of crystal plasticity simulation files are obtained, The format of each group of crystal plasticity simulation files is shown in matrix 3;

Matrix 3

S4. Obtain the IPF diagram and simulated true stress-true strain coordinate data through simulation: use DAMASK as the crystal plasticity finite element simulation software, input the crystal plasticity simulation file obtained in step S3 to carry out the crystal plasticity finite element simulation, and convert the 10000 groups in step S3 The crystal plasticity simulation file is input into DAMASK software for 10,000 sets of simulations. The DAMASK simulation follows the crystal plasticity constitutive equations. After the simulation is completed, 10,000 sets of simulation results are obtained. Python is used to post-process the 10,000 sets of simulation results to obtain 10,000 sets of magnesium alloys. IPF diagram and 10,000 sets of simulated true stress-true strain coordinate data;

The described crystal plasticity constitutive equations:

为晶体塑性剪切速率，值为1，

is the crystal plastic shear rate, the value is 1,

为初始剪切速率，值为0.001，

is the initial shear rate, the value is 0.001,

n is the initial hardening modulus of tension and compression twins, n is one of T9 and T10,

为临界分切应力，

代表4个不同滑移系的临界分切应力，

为T1，T2，T3，T4中的一种，

is the critical shear stress,

represents the critical shear stress of four different slip systems,

One of T1, T2, T3, T4,

h ₀ is the initial hardening modulus value between three different slip systems, between slip systems and twin systems, h ₀ is one of T11, T12, T13,

为4个不同滑移系的饱和滑移分切应力，

为T5，T6，T7，T8中的一种，

is the saturated slip shear stress of four different slip systems,

One of T5, T6, T7, T8,

q ^αβ is the coefficient between different slip systems α and β, q ^αβ is one of T14, T15, T16,

H ^αβ is the total hardening modulus;

S5. Use Python to obtain the IPFs diagram and the true stress-true strain curve diagram, and normalize the crystal plastic constitutive parameters: the 10000 groups of 16 crystal plastic constitutive parameters T1, T2, T3, T4 in step S3, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 were normalized to obtain 10000 groups of normalized t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16, use Python to draw the 10,000 sets of simulated true stress-true strain coordinate data in step S4 as 10,000 sets of true stress-true strain curves, and step The 10,000 sets of IPF images in S4 are cut into 10,000 sets of standard IPF images of standard 50x50 pixels, which are recorded as 10,000 sets of IPFs images; finally, 10,000 sets of crystal plasticity data sets are obtained, and each set of data sets is: IPFs images, true stress- True strain curve and normalized t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16;

The normalization formula described is:

W _norm is one of t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16 parameters,

W is one of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 parameters,

W _min is the minimum number among 10000 groups of parameters,

W _max is the largest number among 10000 groups of parameters;

As a result, IPFs diagrams, true stress-true strain curve diagrams and large data samples of corresponding normalized constitutive parameters were established;

S6, training dual-channel convolutional neural network model: 10000 groups of IPF _S diagrams and true stress-true strain curves described in step S5 are used as the input of training dual-channel convolutional neural network model, and the 10000 groups described in step S5 are t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16 are used as the output of the neural network, making a two-channel convolutional neural network model, with a ratio of 9:1 The ratio is randomly divided into 9000 groups of training sets and 1000 groups of verification sets; the two-channel convolutional neural network model includes: 5 layers of alxnet model convolution layers, 3 fully connected layers; the convolution layer is: the training set The 9,000 sets of IPF _S maps and true stress-true strain curves are input into the convolutional layer. After the convolution operation of the convolutional layer, the image is abstracted into a feature map. The 9,000 sets of IPF _S maps are processed by 5 layers of convolutional layers. Finally, 9000 sets of feature maps 1 are obtained, and 9000 sets of true stress-true strain curves are processed by 5 layers of convolutional layers to obtain 9000 sets of feature maps 2, and 9000 sets of feature maps 1 and 9000 sets of feature maps 2 are combined and input into the fully connected layer for processing ; The process of the fully connected layer is: 9000 sets of output values are obtained by calculating the forward propagation formula containing weights and offsets, the weights and offsets are random, and the range of weights is 1-2 , the range of bias is 10-20, where the output value of each group is: t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16', then calculate 9000 sets of t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11 ', t12', t13', t14', t15', t16' with 9000 groups t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, The overall error between t15 and t16; then recalculate the weight and bias according to the overall error, and then recalculate using the forward propagation formula, repeating the cycle until the overall error is 0.2%-1% and the training ends, and the training is completed The dual-channel convolutional neural network model Z of the dual-channel convolutional neural network model Z; wherein the initial learning rate of the dual-channel convolutional neural network model Z is set to 0.00001, and the number of studies is 3000 times; the training completed dual-channel convolutional neural network model Z is for 9000 groups of training The overall error of the set is 0.2%-0.8%, and the overall error of the trained dual-channel convolutional neural network model Z for 1000 sets of verification sets is 0.4%-0.9%;

The overall error formula is:

σ is the overall error,

ν is a positive integer, the value is 1-16,

t' _v for t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16 one of ',

t _v is one of t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16,

The overall error, weight, and offset calculation formula:

σ＝ ^WT +b

σ is the overall error,

W is the weight,

T is the transpose symbol,

b is the bias.

Example 2

The universality of the dual-channel convolutional neural network model Z obtained by training in Example 1 is verified by experiments;

S1. Select the magnesium alloy commonly used in the laboratory: AZ91, take the EDSD map, select the magnesium alloy IPF map, use Python to cut the IPF map into a standard 50×50 pixel IPFs map, and obtain the anisotropy coefficient R, Section reduction rate, and then perform uniaxial tensile experiments on the material to obtain the true stress-true strain coordinate data, and draw the true stress-true strain coordinate data as a true stress-true strain curve through Python;

S2. Input the IPFs diagram and the true stress-true strain curve diagram obtained in step S1 into the dual-channel convolutional neural network model Z trained in Example 1, and the neural network outputs a set of t1, t2, t3, t4 for calibration , t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16, the neural network output calibration value is converted into 16 crystal plastic constitutive parameters by denormalization method: T1, T2 , T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16;

S3. The 16 crystal plastic constitutive parameters T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15, T16 and In step S1, the crystal plasticity simulation file is established through the IPFs diagram obtained by the experiment, and the crystal plasticity simulation file is input into the DAMASK simulation software for crystal plasticity simulation, and the simulation result is obtained, and the simulation result is processed by Python to obtain the simulated AZ91 magnesium alloy isotropic Anisotropy coefficient R', section reduction rate;

S4. Perform an error analysis on the simulated AZ91 magnesium alloy anisotropy coefficient R' and area reduction rate obtained in step S3 and the AZ91 magnesium alloy anisotropy coefficient R and area reduction rate obtained in the step S1 experiment, and the results show: parameter fitting The degree is 99.5-99.7%, which shows that the present invention adopts the dual-channel convolutional neural network to calibrate the accuracy and universality of 16 crystal plastic constitutive parameters.

In summary: the present invention applies the convolutional neural network used in the macroscopic field to the field of microscopic crystal structure parameter calibration, and the obtained dual-channel convolutional neural network can accurately and synchronously calibrate 16 parameters of the crystal plastic constitutive at one time, and the calibration speed Both the accuracy and accuracy are better than the calibration methods disclosed in the prior art. In addition, the successfully trained neural network model of the present invention has universality and is applicable to the magnesium alloy system, and can accurately and synchronously calibrate the crystal plastic constitutive parameters of various magnesium alloys in the magnesium alloy system at one time, without the need to repeatedly establish a model database and repeatedly Training convolutional neural networks simplifies the computational process.

Claims (2)

1. A crystal plasticity constitutive parameter calibration method based on a two-channel convolution neural network is characterized by comprising the following steps: it comprises the following steps:

(1) The magnesium alloy model was established using the Neper software: inputting m groups of parameters x and y into Neper software, wherein x and y respectively represent the random distribution and the crystal grain roundness of crystal grains, the value range of m is 100-100000 and is an integer, the value range of x is 0-1, the value range of y is 0.3-1.0, m groups of magnesium alloy models are obtained and are marked as m groups (x, y), and the model of each group (x, y) is shown as a matrix 1;

matrix 1

i is one of 1, 2 or 3;

(2) A magnesium alloy model with added texture was established using Matlab kit MTEX: setting m groups of d, e, f, g, h and k parameters, d, e, f, g, h and k respectively representing six different texture strengths of the magnesium alloy in the step (1), wherein the numeric area of d, e, f, g, h and k is 0-1, and d + e + f + g + h + k =1; inputting m groups of d, e, f, g, h and k parameters and m groups of (x, y) obtained in the step (1) into MTEX, and obtaining m groups of magnesium alloy models (d, e, f, g, h and k) (x, y) with texture added through the Bunge formula operation of MTEX, wherein the m groups of magnesium alloy models are marked as m groups of (d, e, f, g, h and k) (x, y), and data models of each group of (d, e, f, g, h and k) (x, y) are shown as a matrix 2;

matrix 2

(3) Using Python to obtain m sets of crystal plastic simulation files: selecting 16 parameters of a crystal plasticity constitutive structure: basal plane slippage critical slitting stress T1, cylindrical surface slippage critical slitting stress T2, conical surface < a > slippage critical slitting stress T3, conical surface < c + a > slippage critical slitting stress T4, basal plane slippage saturated slippage slitting stress T5, cylindrical surface slippage saturated slippage slitting stress T6, conical surface < a > slippage saturated slippage slitting stress T7, conical surface < c + a > slippage saturated slippage slitting stress T8, stretching twin crystal critical slitting stress T9, compressing twin crystal critical slitting stress T10, twin and twin initial hardening modulus T11, slippage and slippage initial hardening modulus T12, slippage and twin initial hardening modulus T13, sliding strain rate sensitive parameter T14, twin strain rate sensitive parameter T15 and sliding hardening parameter T16; t1 ranges from 5 to 80, T2 ranges from 10 to 150, T3 ranges from 20 to 200, T4 ranges from 50 to 500, T5 ranges from 10 to 160, T6 ranges from 20 to 300, T7 ranges from 40 to 400, T8 ranges from 100 to 1000, T9 ranges from 5 to 250, T10 ranges from 10 to 400, T11 ranges from 10 to 500, T12 ranges from 50 to 3000, T13 ranges from 50 to 1200, T14 ranges from 1 to 15, T15 ranges from 5 to 80, and T16 ranges from 0.5 to 10; designing m groups of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16 in the range, adding each group of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16 to the group of magnesium alloy models (d, e, f, g, h, i) (x and y) in the step (2) through Python to obtain a group of crystal plastic simulation files, and finally obtaining m groups of crystal plastic simulation files, wherein each group of crystal plastic simulation files is in a form shown in a matrix 3;

matrix 3

(4) IPF plots and simulated true stress-true strain coordinate data were obtained by simulation: adopting DAMASK as crystal plasticity finite element simulation software, inputting the crystal plasticity simulation file obtained in the step (3) to perform crystal plasticity finite element simulation, inputting m groups of crystal plasticity simulation files in the step (3) into the DAMASK software to perform m groups of simulation, following the crystal plasticity constitutive equation set in the DAMASK simulation, obtaining m groups of simulation results after the simulation is completed, and performing post-processing on the m groups of simulation results by using Python to obtain an IPF (in-plane stress) -true strain coordinate data of m groups of magnesium alloys and m groups of simulated true stress-true strain coordinate data;

the crystal plasticity constitutive equation set is as follows:

is the crystal plastic shear rate, with a value of 1,

an initial shear rate, value of 0.001,

n is the initial hardening modulus of the tensile and compressive twins, n is one of T9 and T10,

in order to obtain the critical value of the slitting stress,

representing the critical part stress of 4 different slip systems,

is one of T1, T2, T3 and T4, h ₀ The initial hardening modulus values between 3 different slip systems, between slip system and twin system, h ₀ Is one of T11, T12 and T13,

is the saturated sliding cutting stress of 4 different sliding systems,

is one of T5, T6, T7 and T8,

q ^αβ is a coefficient between the different slip systems alpha and beta, q ^αβ Is one of T14, T15 and T16,

H ^αβ total hardening modulus;

(5) IPFs plots and true stress-true strain plots were obtained using Python and the crystal plasticity constitutive parameters were normalized: respectively normalizing the m groups of 16 crystal plasticity constitutive parameters T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16 in the step (3) to obtain m groups of normalized T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16, drawing the m groups of simulated true stress-true strain coordinate data in the step (4) into m groups of true stress-true strain graphs by using Python, and cutting the m groups of IPF graphs in the step (4) into m groups of standard IPF graphs of standard 50x50 pixels, and marking the m groups of standard IPF graphs as m groups of IPFs graphs; finally obtaining m groups of crystal plasticity data sets, wherein each group of data sets comprises: IPFs diagrams, true stress-true strain graphs, and normalized t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16;

the normalization formula is as follows:

W _norm is one of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16 parameters, W is one of T1, T2, T3, T4, T5, T6, T7, T8, T9, T10, T11, T12, T13, T14, T15 and T16 parameters,

W _min is the smallest number of the m sets of parameters,

W _max in m sets of parametersMaximum number;

therefore, an IPFs graph, a true stress-true strain curve graph and a big data sample corresponding to the normalized constitutive parameters are established;

(6) Training a two-channel convolutional neural network model: subjecting the m sets of IPFs of step (5) to _S Taking the graph and a true stress-true strain graph as input of a training dual-channel convolutional neural network model, taking t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15 and t16 of the m groups in the step (5) as output of the neural network, manufacturing the dual-channel convolutional neural network model, and randomly dividing the model into p training sets and q verification sets according to a proportion of 5-30; the two-channel convolution neural network model comprises: 5 layers of alxnet model convolution layer, 3 full connection layers; the convolution layer is as follows: grouping p IPFs in a training set _S Inputting the graph and the real stress-real strain curve graph into the convolution layer, abstracting the image into a characteristic graph after convolution operation of the convolution layer, and p groups of IPFs _S P groups of characteristic diagrams 1 are obtained after 5 layers of convolution layer processing, p groups of real stress-real strain graphs obtain p groups of characteristic diagrams 2 after 5 layers of convolution layer processing, and the p groups of characteristic diagrams 1 and the p groups of characteristic diagrams 2 are merged and input into full connection layer processing; the full connection layer treatment comprises the following steps: p groups of output values are obtained through calculation of a forward propagation formula containing weights and biases, wherein the weights and the biases are random, the range of the weights is 1-2, the range of the biases is 10-20, and each group of output values is as follows: the overall error between the p groups t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16' is then calculated, and the p groups t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16' and the p groups t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16 in the training set; then, recalculating the weight and the bias according to the integral error, then, recalculating by adopting a forward propagation formula again, and performing reciprocating circulation until the integral error is 0.2-1%, and finishing training to obtain a trained dual-channel convolution neural network model; wherein an initial learning rate of the two-channel convolutional neural network model is set to0.00001-0.01, and the learning times are 1000-3000; the overall error of the trained dual-channel convolutional neural network model to the p groups of training sets is 0.2% -1%, and the overall error of the trained dual-channel convolutional neural network model to the q groups of verification sets is 0.4% -1.1%;

the overall error formula is as follows:

p is the number of training set groups,

the sigma is the overall error of the signal,

nu is a positive integer and takes the value of 1-16,

t′ _v is one of t1', t2', t3', t4', t5', t6', t7', t8', t9', t10', t11', t12', t13', t14', t15', t16',

t _v is one of t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15 and t16,

the integral error and weight value and bias calculation formula is as follows:

σ＝W ^T +b

the sigma is the overall error of the signal,

w is a weight value,

t is the transposed symbol and is the symbol,

b is an offset.

2. The method for calibrating the crystal plasticity constitutive parameters based on the dual-channel convolutional neural network as claimed in claim 1, wherein: the magnesium alloy is one of AZ31, AZ61, AZ91, ZK60 or AM 60.

Images