Method and System for Predicting Plastic Characteristics of Anisotropic Substances Based on Deep Learning Using Press-In Response Data

Abstract

The present invention provides a method and system for predicting plastic properties of an anisotropic material based on deep learning using indentation response data capable of easily and quickly obtaining plastic properties of an anisotropic material in a non-destructive manner. The method for predicting plastic properties of an anisotropic material based on deep learning using indentation response data according to an embodiment of the present invention includes the steps of preparing a plurality of data sets consisting of learning indentation response data and learning plastic property data of a learning anisotropic material; causing a computer system to perform deep learning using the learning indentation response data as an input value and the learning plastic property data as an output value; providing actual indentation response data of an anisotropic material to be predicted; and inputting the actual indentation response data into the deep-learned computer system to predict the plastic properties of the anisotropic material to be predicted.

Description

Technical Field

[0001] The technical idea of the present invention relates to a method for predicting the plastic properties of anisotropic materials, and more particularly, to a method and system for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data.

Background Art

[0002] When a metallic material is plastically processed such as rolling, drawing, or extrusion, or formed with a fiber reinforcement, or a film layer is vapor-deposited or coated, a texture is formed and grows, and thereby plastic anisotropy can appear. Since the plastic anisotropy can change the formability required in forming processes such as bending, tension, and deep drawing, it is very important to precisely measure or predict the plastic anisotropy.

[0003] Conventionally, in order to measure the plastic anisotropy of a material, a uniaxial tensile test or a uniaxial compression test has been performed several times while changing the angle. However, such tests are costly and time-consuming, and are essentially performed while destroying the test piece, so there is a limit in that the volume of the test piece is limited or it is difficult to apply in small amounts. Therefore, there is a need for a method for more easily and quickly analyzing plastic anisotropy in a non-destructive manner.

Summary of the Invention

Problems to be Solved by the Invention

[0004] The technical problem to be achieved by the technical idea of the present invention is to provide a method and system for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data that can easily and quickly obtain plastic properties of anisotropic materials in a non-destructive manner.

[0005] However, such problems are exemplary and the technical idea of the present invention is not limited thereto.

Means for Solving the Problems

[0006] According to one aspect of the present invention, there is provided a method and system for predicting the plastic properties of an anisotropic material based on deep learning using indentation response data capable of easily and quickly obtaining plastic properties of an anisotropic material in a non-destructive manner.

[0007] According to one embodiment of the present invention, the method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data may include: preparing a plurality of data sets including learning indentation response data and learning plastic property data of a learning anisotropic material; causing a computer system to perform deep learning using the learning indentation response data as an input value and the learning plastic property data as an output value; providing actual indentation response data of an anisotropic material to be predicted; and inputting the actual indentation response data into the computer system that has been deep-learned to predict the plastic properties of the anisotropic material to be predicted.

[0008] According to one embodiment of the present invention, the method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data may include: providing a computer system that has performed deep learning using the learning indentation response data as an input value and the learning plastic property data as an output value in a plurality of data sets including learning indentation response data and learning plastic property data of a learning anisotropic material; providing actual indentation response data of an anisotropic material to be predicted; and inputting the actual indentation response data into the computer system that has been deep-learned to predict the plastic properties of the anisotropic material to be predicted.

[0009] According to an embodiment of the present invention, a prediction system for plastic properties of anisotropic materials based on deep learning using the indentation response data includes a finite element simulation execution module and a deep learning execution module, and comprises: a) preparing a plurality of data sets consisting of learning indentation response data and learning plastic property data of anisotropic materials for learning; b) causing a computer system to perform deep learning with the learning indentation response data as input values and the learning plastic property data as output values; c) providing actual indentation response data of an anisotropic material to be predicted; and d) inputting the actual indentation response data into the computer system that has been deep learned to predict the plastic properties of the anisotropic material to be predicted. A prediction system for plastic properties of anisotropic materials based on deep learning using indentation response data, which performs a prediction method for plastic properties of anisotropic materials based on deep learning using indentation response data, wherein the step a) is performed by the finite element simulation execution module, and the steps b) and d) can be performed by the deep learning execution module.

Advantages of the Invention

[0010] In the case of the technical idea of the present invention, a prediction method for plastic properties of anisotropic materials based on deep learning using indentation response data uses a non-destructive and highly efficient indentation test instead of a tensile test involving material fracture, and by using an artificial neural network system that can associate the results of the indentation test with plastic properties, the plastic properties of anisotropic materials can be easily and quickly obtained in a non-destructive manner.

[0011] The advantages of the present invention described above are exemplarily described, and the scope of the present invention is not limited by such advantages.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Best Mode for Carrying Out the Invention

[0013] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The embodiments of the present invention are provided to more fully explain the technical idea of the present invention to those having ordinary knowledge in the relevant technical field. The following embodiments can be modified into various other forms, and the scope of the technical idea of the present invention is not limited to the following embodiments. Rather, these embodiments are provided to further enrich and complete the present disclosure and to fully convey the technical idea of the present invention to those skilled in the art. Throughout this specification, the same reference numerals mean the same elements. Further, various elements and regions in the drawings are schematically shown. Therefore, the technical idea of the present invention is not limited by the relative sizes and intervals shown in the accompanying drawings.

[0014] The indentation technique measures hardness using the size and depth of an indentation formed by pressing an indenter onto a test piece, and has the advantage of being able to precisely measure the plastic properties of a small amount of test piece. The instrumented indentation test (IIT) is a test method that continuously records the load applied by the indenter and the indentation depth of the indenter, and through the load-depth curve and analysis method, the hardness, elastic modulus, and other hardening properties of the target material can be measured. In addition to this, a protocol for analyzing the wide-area uniaxial tensile behavior by the finite element method based on the local load-depth curve obtained from high-resolution nano-indentation test data has been proposed. From the load-depth curve obtained in an instrumented indentation test using a spherical or sharp indenter, the plastic properties of the target material can be inversely estimated based on finite element simulation and optimization algorithms. Also, to solve the non-uniqueness caused when numerically determining mechanical properties from the load-depth curve, additional indentation information such as dual indentation may be used, or vertical residual indentations around the indentation such as pile-up or sink-in may be considered. Also, instead of the load-depth curve, pile-up / sink-in, in-plane displacement, and the profile of the residual indentation may be used to obtain plastic properties.

[0015] However, the prediction of plastic properties based on indentation data has been well studied for isotropic materials, but the research on anisotropic materials closer to actual materials is still insufficient. As a conventional method, in order to simplify unknown material parameters, there is a method of obtaining the plastic properties of anisotropic materials from indentation data assuming transverse isotropy, but this has the limitation that it is difficult to predict general mechanical anisotropy. In addition, conventionally, the load-depth curve and the residual vertical displacement have been considered, but there is a limitation that the residual in-plane displacement field has not been considered. Therefore, in order to extract general anisotropic plasticity from various indentation results, it can be proposed to use an artificial neural network (neural network, NN) that exhibits excellent performance as a universal approximator. The artificial neural network can model the complex relationship between input values and output values with very high accuracy based on a statistical deep learning algorithm without using equations. So far, the research using artificial neural networks for anisotropic materials is in an insufficient state.

[0016] The inventors of the present invention have established a general framework for analyzing the plastic properties of bulk materials. By using such a framework, the anisotropic properties can be generally analyzed from the spherical indentation response composed of the load-depth curve, pile-up / sink-in, and in-plane displacement field using finite element-deep learning modeling. In the finite element-deep learning modeling, the anisotropic plastic properties obtained by deep learning using an artificial neural network with adjusted hyperparameters were compared with the actual experimental results. As a result, it can be seen that the finite element-deep learning modeling is robust and effective.

[0017] The following will explain in detail in relation to the plastic properties of anisotropic materials.

[0018] The elastic behavior of continuous materials such as metals is as shown in Mathematical Formula 1 known as Hooke's law. [Equation 1] σ = Eε

[0019] In Equation 1 above, σ is stress, E is the elastic modulus, and ε is strain. For the sake of simplification and focusing on the analysis of the plastic characteristics of the material, the influence on the indentation curve can be ignored, and the Poisson's ratio can be set to 0.3.

[0020] To explain the strain hardening behavior from the onset of plastic yielding, a non-linear isotropic hardening model can be adopted, and the Swift equation of Equation 2, which is a power law, can be introduced using the user-defined subroutine UHARD. [Equation 2] [Number]

[0021] [Number]

[0022] In the present invention, a sixth order polynomial yield criterion can be applied as a constitutive equation. Hereinafter, the sixth order polynomial yield criterion will be referred to as Poly6. The three-dimensional shape representation of the yield criterion of the Poly6 model is as shown in Equation 3. [Equation 3] [Number]

[0023] In Equation 3 above, σ xx , σ yy and σ zz are the normal stresses in the vertical direction, and σ xy , σ yz and σ zx are the shear stresses. a1 to a 16is an independent poly-6 anisotropy parameter and can be defined based on uniaxial tensile test data and biaxial tensile test data.

[0024] The most common data set considered in this method is as shown in Equation 4. [Equation 4]

[0025] Biaxial curve data set: {σ0, r0, σ b , r b , σ 90 , r 90}

[0026] Directional data set: {σ 15 , r 15 , σ 30 , r 30 , σ 45 , r 45 , σ 60 , r 60 , σ 75 , r 75}

[0027] In Equation 4 above, σ b is the balanced biaxial yield stress, and r b is the r value defined as r b = dε yy / dε xx .

[0028] Equation 5 shows the uniaxial stress state function depending on the angle. [Equation 5]

[0029] σ(θ) = σ θ (cos 2 θ, sin 2 θ, sinθcosθ)

[0030] In Equation 5 above, θ refers to the angle from a reference direction, for example, the rolling direction (RD), clockwise or counterclockwise.

[0031] The r-value of the test piece is the ratio of the transverse strain (the strain component perpendicular to the indentation load direction) to the thickness strain. Equation 6 shows the state function of the r-value with respect to the angle. [Equation 6] [Number]

[0032] Equation 7 shows the r-value considering the normality rule, the rigid-plastic approximation, and the non-change of the volume of the plastic strain. [Equation 7] [Number]

[0033] Hereinafter, a method and a system for predicting the plastic properties of an anisotropic material based on deep learning using indentation response data according to the technical idea of the present invention will be described.

[0034] According to the technical idea of the present invention, there are provided a method and a system for predicting the plastic properties of an anisotropic material based on deep learning using indentation response data with which the plastic properties can be easily and quickly obtained in a non-destructive manner for the anisotropic material.

[0035] The method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data may include the steps of preparing a plurality of data sets including learning indentation response data and learning plastic property data of a learning anisotropic material, causing a computer system to perform deep learning with the learning indentation response data as an input value and the learning plastic property data as an output value, providing actual indentation response data of an anisotropic material to be predicted, and inputting the actual indentation response data into the computer system that has been deep-learned to predict the plastic properties of the anisotropic material to be predicted.

[0036] The step of preparing the plurality of data sets can include a step of providing tensile characteristics of the anisotropic material for learning, a step of obtaining poly6 anisotropic parameters, an elastic modulus, and an isotropic hardening parameter from the tensile characteristics of the anisotropic material for learning, a step of performing a finite element simulation using the poly6 anisotropic parameters, the elastic modulus, and the isotropic hardening parameter, and a step of obtaining the learning push-in response data of the anisotropic material for learning as a result of performing the finite element simulation.

[0037] The tensile characteristics can include a tensile stress and an r-value of the anisotropic material for learning.

[0038] The poly6 anisotropic parameter can be obtained from the following formula.

Equation

[0039] (Here, σ xx , σ yy and σ zz are vertical stresses in the vertical direction, and σ xy , σ yz and σ zx are shear stresses.)

[0040] The elastic modulus can include a Young's modulus (E) and a Poisson's ratio (ν) of the anisotropic material for learning.

[0041] The isotropic hardening parameter can include a strength coefficient (k), a strain parameter (ε0), and a strain hardening index (n) obtained from the following formula.

Equation

[0042] After performing the step of obtaining the poly 6 anisotropy parameter, it is further possible to include a step of evaluating whether the poly 6 anisotropy parameter of the anisotropic substance for learning satisfies the convexity with respect to the yield criterion of the anisotropic substance for learning.

[0043] The step of performing the finite element simulation can be performed on a spherical indenter.

[0044] The output value obtained by performing the finite element simulation can include a load-depth curve, in-plane displacement field information, and vertical displacement field information from the results of the anisotropic substance for learning.

[0045] The learning indentation response data can include indentation load data, radial displacement data, and vertical displacement data for an indentation formed by pushing in the anisotropic substance for learning.

[0046] The indentation load data can include load values with respect to the depth of the indentation.

[0047] The radial displacement data can include radial displacement values with respect to the angle from the reference direction of the indentation.

[0048] The radial displacement data can include radial displacement values at a separation distance that is a multiple larger than the radius (R) of the indenter from the center of the indentation.

[0049] The vertical displacement data can include vertical displacements with respect to the angle from the reference direction of the indentation.

[0050] The vertical displacement data can include vertical displacement values at a separation distance equal to the radius (R) of the indenter from the center of the indentation.

[0051] The actual indentation response data can include indentation load data, radiation displacement data, and vertical displacement data for an indentation formed by pressing in the anisotropic material to be predicted.

[0052] The plastic properties of the anisotropic material to be predicted can include the poly6 anisotropic parameters, elastic modulus, and isotropic hardening parameters of the anisotropic material to be predicted.

[0053] After performing the step of predicting the plastic properties of the anisotropic material to be predicted, the method can further include the step of evaluating whether the poly6 anisotropic parameters of the anisotropic material to be predicted satisfy convexity with respect to the yield criterion of the anisotropic material to be predicted.

[0054] After performing the step of predicting the plastic properties of the anisotropic material to be predicted, the method can further include the step of comparing the predicted plastic properties of the anisotropic material to be predicted with the actual plastic properties of the anisotropic material to be predicted.

[0055] At least one of the anisotropic material for learning and the anisotropic material to be predicted can include a metal alloy.

[0056] The method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data includes providing a computer system that performs deep learning with the learning indentation response data of the anisotropic material for learning and the learning plastic property data as input values and the learning plastic property data as output values in a plurality of data sets, providing the actual indentation response data of the anisotropic material to be predicted, and inputting the actual indentation response data into the computer system that has been deep learned to predict the plastic properties of the anisotropic material to be predicted. In this case, it is also applicable when there is a time interval between the time when the computer system performs deep learning and the time when the plastic properties are predicted.

[0057] The prediction system for the plastic properties of anisotropic materials based on deep learning using the indentation response data is composed of a finite element simulation execution module and a deep learning execution module, and includes: a) preparing a plurality of data sets consisting of learning indentation response data and learning plastic property data of the anisotropic material for learning; b) causing a computer system to perform deep learning with the learning indentation response data as input values and the learning plastic property data as output values; c) providing actual indentation response data of the anisotropic material to be predicted; and d) inputting the actual indentation response data into the computer system that has been deep-learned to predict the plastic properties of the anisotropic material to be predicted. The prediction system for the plastic properties of anisotropic materials based on deep learning using the indentation response data, which performs the prediction method for the plastic properties of anisotropic materials based on deep learning using the indentation response data, wherein the step a) is performed by the finite element simulation execution module, and the steps b) and d) can be performed by the deep learning execution module.

[0058] Figure 1 is a flowchart showing a method for predicting the plastic properties of anisotropic materials based on deep learning using the indentation response data according to an embodiment of the present invention.

[0059] Referring to Figure 1, the method for predicting the plastic properties of anisotropic materials based on deep learning (S100) includes: preparing a plurality of data sets consisting of learning indentation response data and learning plastic property data of the anisotropic material for learning (S110); causing a computer system to perform deep learning with the learning indentation response data as input values and the learning plastic property data as output values (S120); providing actual indentation response data of the anisotropic material to be predicted (S130); and inputting the actual indentation response data into the computer system that has been deep-learned to predict the plastic properties of the anisotropic material to be predicted (S140).

[0060] At least one of the anisotropic substances for learning and the anisotropic substances to be predicted can include a substance having plastic anisotropy. For example, it can include a metal alloy. For example, it can include a high-strength steel or an aluminum alloy. For example, it can include at least one of DP780, TBF1050, AA7075, and AA2090. However, this is an example, and the technical idea of the present invention is not limited thereto.

[0061] In this specification, when the indentation response data and the plastic property data are included in the data set for deep learning, the term "for learning" is added to the name. When measured by actual experiments, the term "actual" is added to the name. Of course, for improving the accuracy of learning, the actual indentation response data and the actual plastic property data can be used for the deep learning of the artificial neural network.

[0062] FIG. 2 is a schematic diagram exemplarily showing a prediction system for the plastic properties of anisotropic substances based on deep learning using the indentation response data according to an embodiment of the present invention.

[0063] Referring to FIG. 2, the prediction system for the plastic properties of the anisotropic substances based on deep learning can include a finite element simulation execution module FE and a deep learning execution module NN.

[0064] The finite element simulation execution module FE can perform a finite element simulation for deriving the indentation response elements of the anisotropic substances for learning. In the finite element simulation execution module FE, it can be confirmed whether the result of the indentation response simulation matches the experimental indentation result, and the validity and reliability of data collection for the deep learning of the artificial neural network can be verified. The finite element simulation execution module FE for performing the finite element simulation can be configured by a computer system.

[0065] The finite element simulation execution module FE can perform the step (S110) of preparing a plurality of data sets in FIG. 1.

[0066] From the viewpoint of ensuring the accuracy of the output value, the learning indentation response data and the learning plastic property data included in the data set may be, for example, 100 sets or more, and in order to further improve the accuracy, for example, 1,000 sets or more, and for example, 50,000 sets or more. However, the technical idea of the present invention is not limited thereto. The larger the number of the data sets, the higher the accuracy of the output value, but it can be appropriately selected in consideration of time and cost. The data set may be provided through actual experimental results or may be provided by repeatedly performing finite element simulation while changing input values.

[0067] FIG. 3 is a flowchart showing the step of preparing a plurality of data sets in a method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data of FIG. 1 according to an embodiment of the present invention.

[0068] Referring to FIG. 3, the step (S110) of preparing the plurality of data sets may include a step (S111) of providing the tensile properties of the learning anisotropic material, a step (S112) of obtaining the poly6 anisotropic parameters, the elastic modulus, and the isotropic hardening parameters from the tensile properties of the learning anisotropic material, a step (S113) of performing a finite element simulation using the poly6 anisotropic parameters, the elastic modulus, and the isotropic hardening parameters, and a step (S114) of obtaining the learning indentation response data of the learning anisotropic material as a result of performing the finite element simulation.

[0069] In the step (S111) of providing the tensile properties of the learning anisotropic material, as shown in the formula 4, as an example of the tensile properties of the learning anisotropic material, the tensile stress (σ0~σ 90 ,σb ), and r values (r0 to r 90 , r b ) are provided. The tensile stress of the anisotropic material for learning and the r value can include test results obtained by directly performing a tensile experiment on the anisotropic material for learning.

[0070] In the step of obtaining (S112), the plastic property data for learning is obtained from the tensile property. For example, the poly-6 anisotropic parameters (a1 to a 16 ), elastic constant, and isotropic hardening parameter can be obtained. The poly-6 anisotropic parameter can be obtained from Equation 3. From the tensile property, the elastic constant and isotropic hardening parameter of the anisotropic material for learning are obtained. The elastic constant can include Young's modulus (E) and Poisson's ratio (ν). The isotropic hardening parameter can include the strength coefficient (k), strain parameter (ε0), and strain hardening index (n) obtained from Equation 2.

[0071] After performing the step of obtaining the poly-6 anisotropic parameter (S112), it can further include a step of evaluating whether the poly-6 anisotropic parameter of the anisotropic material for learning satisfies the convexity with respect to the yield criterion of the anisotropic material for learning. If the poly-6 anisotropic parameter does not satisfy the convexity, feedback is given and the step of providing the tensile property is performed again. If the poly-6 anisotropic parameter satisfies the convexity, the poly-6 anisotropic parameter, the elastic constant, and the isotropic hardening parameter are set as input values for finite element simulation.

[0072] In the step (S113) of performing the finite element simulation, the finite element simulation is performed using the polycrystalline anisotropic parameters, the elastic coefficients, and the isotropic hardening parameters as input values. The step (S113) of performing the finite element simulation can be performed on a spherical indentation, however, this is an example and the technical idea of the present invention is not limited thereto. The output values obtained by performing the finite element simulation can include the load-depth curve, the in-plane displacement field information, and the vertical displacement field information of the anisotropic material for learning.

[0073] In the step (S114) of obtaining the learning push-in response data, a process of extracting characteristic data (feature) from the output values is performed for subsequent execution of deep learning, whereby the learning push-in response data can be obtained as the characteristic data.

[0074] The learning push-in response data can include the indentation load data, the radial displacement data, and the vertical displacement data of the anisotropic material for learning.

[0075] Referring to FIG. 2 again, the deep learning execution module NN can perform the step (S120) of causing the deep learning of FIG. 1 to be performed.

[0076] In the deep learning execution module NN, deep learning can be caused to be performed on a computer system using the learning push-in response data as an input value and the learning plastic property data as an output value. The deep learning can be performed at least once or more, whereby the artificial neural network provided in the computer system can be deep-learned.

[0077] The deep learning execution module NN that performs the deep learning can be configured by a computer system. The finite element simulation execution module FE and the deep learning execution module NN may be configured by the same computer system or by separate computer systems. The computer system referred to in this specification includes an artificial intelligence program, an artificial neural network program, or any system equipped with such a program.

[0078] Referring again to FIGS. 1 and 2, a step (S130) of providing actual indentation response data of the anisotropic substance to be predicted is performed. The actual indentation response data may include indentation load data, radial displacement data, and vertical displacement data of the anisotropic substance to be predicted. The actual indentation response data may include test results obtained by directly indenting the anisotropic substance to be predicted. For example, after directly indenting the anisotropic substance to be predicted to obtain a load-depth curve, in-plane displacement field information of the surface, and vertical displacement field information, and then extracting characteristic data therefrom, the indentation load data, the radial displacement data, and the vertical displacement data can be obtained. Alternatively, the actual indentation response data can be obtained by performing the finite element simulation.

[0079] Referring again to FIGS. 1 and 2, a step (S140) of inputting the actual indentation response data into the computer system that has been deep learned to predict the plastic properties of the anisotropic substance to be predicted is performed. The plastic properties may include poly 6 anisotropic parameters, elastic coefficients, and isotropic hardening parameters of the anisotropic substance to be predicted.

[0080] After performing the step (S140) of predicting the plastic properties of the anisotropic material to be predicted, it is further possible to include a step of evaluating whether the poly6 anisotropic parameters of the anisotropic material to be predicted predicted by the deep-learned computer system satisfy the convexity with respect to the yield criterion of the anisotropic material to be predicted. If the poly6 anisotropic parameters do not satisfy the convexity, feedback is given and deep learning is performed again. If the poly6 anisotropic parameters satisfy the convexity, the poly6 anisotropic parameters, the elastic modulus, and the isotropic hardening parameters of the anisotropic material to be predicted are output as the plastic properties of the anisotropic material to be predicted.

[0081] The poly6 anisotropic parameters can include the parameters (a1 to a 16 ) shown in Equation 3. The elastic modulus can include Young's modulus (E) and Poisson's ratio (ν). The isotropic hardening parameters can include the strength coefficient (k), the strain parameter (ε0), and the strain hardening index (n) shown in Equation 2.

[0082] The step of evaluating the convexity of the anisotropic material to be predicted can also be performed in the step (S120) of causing the deep learning to be performed.

[0083] Optionally, after performing the step of predicting the plastic properties of the anisotropic material to be predicted, it is further possible to include a step of comparing the predicted plastic properties of the anisotropic material to be predicted with the actual plastic properties of the anisotropic material to be predicted. The actual plastic properties can be test results obtained by directly experimenting on the anisotropic material to be predicted. As a result of the comparison, if the difference between the predicted plastic properties and the actual plastic properties is within a certain error rate range, the deep learning can be terminated, and if it is outside the certain error rate range, the prediction step can be performed again.

[0084] The above-mentioned deep learning is a term widely used in the field of artificial intelligence, and the present invention does not particularly limit its method. According to an embodiment of the present invention, by inputting the dataset into an artificial neural network (or artificial intelligence) to perform deep learning (or machine learning), when the push-in response data is input as an input value, an artificial neural network can be constructed to predict the plastic property data as an output value. The principle of constructing such an artificial neural network is similar to the regression problem for a general linear function. That is, it is to create a function that teaches the relationship between the input value and the output value of a given dataset. The deep learning method of the artificial neural network is a method widely known such as general deep learning, and in the present invention, any method among the known methods can be adopted.

[0085] As an example of the above-mentioned deep learning, an example will be described in which the artificial neural network is composed of 19 nodes in the first column included in the input layer, 45 nodes in the second column included in the hidden layer, 45 nodes in the third column, 45 nodes in the fourth column, and 19 nodes in the fifth column included in the output layer.

[0086] Input values are input into the first column, and then each node in the first column is connected by a line to each node in the second column. Through each line, the node values in the second column are determined using the input values in the first column. The topmost first node in the second column will follow calculations such as relational expression A. [Relational Expression A] value1=1 / (1+exp(-(a1_1Xε1+a1_2Xε2+...+a1_19Xε19))

[0087] In the above relational expression A, value1 represents the value of the first node in the second column, a1_1 to a1_19 respectively represent the weights of each input value for determining the value of the first node in the second column, and ε1 to ε19 respectively represent 19 input values.

[0088] Next, the second node in the second column will also follow the same calculation, specifically, it will follow a calculation such as the following relational expression B. [Relational Expression B] Value2 = 1 / (1 + exp(-(a2_1 × ε1 + a2_2 × ε2 +... + a2_19 × ε19)))

[0089] In the relational expression B, value2 represents the value of the second node in the second column, a2_1 to a2_19 respectively represent the weights of each input value for determining the value of the second node in the second column, and ε1 to ε19 respectively represent 19 input values.

[0090] Such the same process can be repeatedly executed from the third column to the fifth column which is the last order. Next, while changing the values of the unknowns represented as a#_# (where # is number), a function called an artificial neural network is used to find the relationship between the input and output of the actual data.

[0091] For this purpose, first, arbitrary numerical values are assigned to the a#_#, the difference between the calculated output value and the output of the actual data is obtained, and in order to reduce this difference, the value of the a#_# is corrected using the gradient descent method. The process of correcting the value of the unknown using the difference between the output values in this way is called back-propagation. The process of the artificial neural network finding the relationship between the actual data while performing such back-propagation corresponds to the artificial intelligence performing deep learning.

[0092] Through the deep learning process of artificial intelligence as described above, according to the present invention, different from the conventional method which required separate operations by the actual user, without any separate operations by the actual user, just by inputting the input values measured in the deep learning computer system, a predicted value can be immediately obtained. Therefore, it can be easily used in combination with existing experimental equipment, enabling an epoch-making reduction in time and cost.

[0093] According to the technical idea of the present invention, a deep artificial neural network (NN) model capable of identifying a non-linear correlation relationship between discontinuous input data and output data can be applied. In order to improve the performance of the artificial neural network model, basic hyperparameters, that is, the number of hidden layers and the number of neurons per hidden layer, can be tuned with a Bayesian optimization algorithm to optimize the structure of the artificial neural network model. The number of neurons per hidden layer can be set to be the same as each other, and by optimizing the hyperparameters, a three-layer model with 45 neurons per layer of the hidden layer can be applied. The prepared dataset can be randomly assigned to the deep learning, validation, and test sets. During the deep learning process, the mean squared error (MSE) of the validation set can be monitored to determine the completion of deep learning while avoiding overfitting. The performance of the learned artificial neural network can be evaluated using the test dataset excluded during deep learning. New information not used in deep learning can be provided to the deep-learned artificial neural network to compare the calculated and measured intermediate variables. Since the initial weights, biases, and dataset partitioning for deep learning can be configured differently, deep learning can be repeated independently multiple times, for example, 5 times, to calculate the mean squared error and output to ensure sufficient generalization. The maximum number of epochs for deep learning can be set to, for example, 30 or more, for example, 200, which is large enough to stabilize the artificial neural network system and allows the early stopping method to operate.

[0094] Experimental Example Hereinafter, experimental examples for helping the understanding of the present invention will be described. The following experimental examples are presented to help the understanding of the invention and are not limited to the following experimental examples of the present invention.

[0095] As the anisotropic material for learning, a material exhibiting plastic anisotropy was selected. Specifically, two types of ultra-high strength steels (DP780 and TBF1050) and two types of aluminum alloys (AA7075 and AA2090) were selected.

[0096] As the database generation means, the finite element simulation for the pressing was verified, and a tensile test was conducted on the anisotropic material for learning to confirm the validity of the deep learning prediction result.

[0097] The anisotropic material for learning was subjected to a tensile test to obtain tensile properties. The tensile properties include the tensile stress (σ0~σ 90 ,σ b ) and r-values (r0~r 90 ,r b ). The tensile test of the anisotropic material for learning was conducted as follows. First, each of the anisotropic materials for learning was manufactured into a tensile test specimen having a thickness of 1 mm. The tensile test specimen was subjected to a uniaxial tensile test according to the ASTM-E08 standard using a general-purpose tensile testing machine (Instron5584, USA). The tensile test specimen was subjected to a quasi-static tensile test in the rolling direction (RD), diagonal direction (DD), and transverse direction (TD) at a nominal strain rate of 10 -3 s -1 .

[0098] The results of the tensile test were obtained using an analysis solution, combining 16 anisotropic parameters of the poly6 model and 3 isotropic parameters of the Swift hardening model to obtain a total of 19 types of parameters.

[0099] The indentation test of the anisotropic material for learning was performed as follows. Each of the anisotropic materials for learning was fabricated into an indentation test piece having a size of 60×50×1 mm and a rectangular shape. Before performing the indentation test, the surface of the indentation test piece was mechanically ground and polished using an alumina suspension containing alumina with a maximum diameter of 1 μm. An instrumented indentation test on the indentation test piece was conducted using an AIS3000 system (Frontics, Inc., Korea) having a load resolution of 0.05 N and a displacement resolution of 0.1 μm. In the indentation test, a tungsten carbide spherical indenter with a radius of 250 μm was used. In order to accurately explain the tensile behavior of the bulk material, since a sufficient area below the indentation must be allowed, the depth of the indentation was controlled to a maximum of 150 μm using displacement control. The indentation was repeated 5 times to obtain an average value in order to ensure reliability.

[0100] Also, a tensile test and an indentation test of the anisotropic material to be predicted were conducted in the same manner.

[0101] In order to obtain the in-plane displacement field information and the vertical displacement field information of the indentation test piece, images before and after the indentation test of the indentation test piece were obtained.

[0102] The acquisition of the in-plane displacement field information of the indentation test piece used the DIC (Digital Image Correlation) technique used for strain measurement because a non-contact, non-destructive, and simple optical setup is required. The speckle pattern was formed by spray painting the surface of the indentation test piece before indentation. The in-plane displacement field of the indentation test piece was obtained by acquiring digital images before and after deformation, respectively, and tracking and calculating each point of the images using a MATLAB (registered trademark) DIC application program.

[0103] The vertical displacement field was measured using a 3D laser scanning confocal microscope (LSCM) system (VK-X21 series, Keyence, Japan) equipped with a 10x objective lens. By acquiring a plurality of two-dimensional images at regular intervals and accumulating them, the three-dimensional topography at the indentation can be reconstructed. Image processing was performed using VK-analyzer software to observe the vertical cross-sectional area of the three-dimensional topography and obtain pile-up / sink-in information.

[0104] The finite element simulation for the anisotropic material for learning was performed using commercial ABAQUS software. A quarter finite element indentation model composed of three-dimensional 8-node continuous brick elements (C3D8R) with reduced integration points was used. In the finite element simulation, the shape of the indenter was selected as a spherical indenter identical to that used in the experiment. The Young's modulus of the spherical indenter was 700 GPa, and the Poisson's ratio was 0.3. The specimen model for the finite element simulation had a radius and thickness of 3 mm each, and such a size can exclude geometric effects compared to the size of the indenter and the depth of indentation. Mechanical boundary conditions of lateral axis symmetry and bottom fixation were imposed on the specimen model. Similar to the experiment, a displacement condition of 0.3 mm / min in the z-axis direction was applied to the indenter until the depth of indentation reached a maximum of 150 μm. During the indentation, the interaction between the indenter and the surface of the specimen was controlled using a surface-to-surface contact pair algorithm, normal hard contact, and tangential friction. The friction coefficient was set to 0.12 as the optimal friction coefficient because it has a significant impact on the load-depth curve in deep spherical indentation of 50 μm or more. Preliminary investigations were conducted on the improvement of the mesh in the contact area and convergence.

[0105] Table 1 is a table showing the mechanical parameters of the anisotropic materials for learning in a method for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention. [Table 1]

[0106] Referring to Table 1, for all the anisotropic materials for learning, since a1 = 1, the number of independent parameters is 18. Since such parameters are to be predicted based on the indentation response data, it is necessary to deeply consider the strain field formed by indentation. The consideration of the parameters was carried out using a random generation method for each parameter, and a total of 50,000 data sets were generated.

[0107] Table 2 is a table showing the parameters input to the finite element simulation and their ranges in a method for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention. [Table 2]

[0108] In Table 2, the Young's modulus was set to 70 GPa for aluminum alloys and 200 GPa for iron alloys. The Poisson's ratio was set to 0.3 for both aluminum alloys and iron alloys.

[0109] Next, a finite element simulation was performed using the six poly anisotropic parameters, elastic modulus, and isotropic hardening parameter (Swift hardening parameter) obtained from the tensile properties of the anisotropic materials for learning.

[0110] The finite element simulation can be repeatedly performed to output an output value. The output value includes a load-depth curve, an in-plane displacement field, and a vertical displacement field.

[0111] For subsequent execution of deep learning, through a process of extracting characteristic data from the output value, learning push-in response data was obtained as the characteristic data. The learning push-in response data includes the indentation load data, the radial displacement data, and the vertical displacement data of the learning anisotropic material.

[0112] FIG. 4 and FIG. 5 are graphs showing displacements due to pushing calculated using finite element simulation in a method for predicting plastic properties of an anisotropic material based on deep learning using push-in response data according to an embodiment of the present invention.

[0113] Referring to FIGS. 4 and 5, the radial displacement (u r ) and the vertical displacement (u z ) obtained by performing finite element simulation on DP780 steel are shown respectively. The radial displacement (u r ) and the vertical displacement (u z ) were obtained from the residual strain field around the indentation at various angles from the rolling direction (RD), for example, at 0°, 15°, 30°, 45°, 60°, 75°, and 90°. The acquisition positions of the radial displacement and the vertical displacement are shown on the upper side respectively. The acquisition positions were set by changing the angle at 15° intervals from the rolling direction (RD) at a separation distance that is a multiple of the radius (R) of the indenter with respect to the center of the indentation. The multiple separation distances were 1.0R, 1.2R, 1.4R, and 1.6R.

[0114] In the case of isotropic materials, the radial displacement and the vertical displacement have the same value at the same separation distance regardless of the angle. However, in anisotropic materials with plastic anisotropy such as the DP780 steel, it can be seen that they show different values. That is, despite being obtained at the same distance from the center of the indentation, deviations with different numerical values appear due to the change in angle, and the minimum value was shown when the angle was 45° from the rolling direction at all distances. The deviation was more significantly shown in the radial displacement. The behavior of the radial displacement and the behavior of the vertical displacement are similar at 1.2R, 1.4R, and 1.6R, but different at a distance of 1.0R. Such anisotropic behavior was similarly shown in TBF1050 steel, AA7075 aluminum alloy, and AA2090 aluminum alloy.

[0115] In the plastic region of the uniaxial tensile curve in the rolling direction (RD), it can be used as a reference state for determining the hardening medium variable for each material. Conventionally, in order to define the anisotropic parameters of anisotropic materials, the yield stress and r-value obtained by performing uniaxial tensile tests at various angles were converted. However, according to the technical idea of the present invention, as shown in FIGS. 4 and 5, the anisotropic parameters can be defined from the results of the indentation test.

[0116] Such similar behavior means that not all of the data shown in FIGS. 4 and 5 are valid as input values to be input into deep learning. Therefore, in order to verify the valid data, the linearity with respect to the radial displacement and the vertical displacement will be examined.

[0117] FIGS. 6 to 8 are graphs showing the linearity between the displacements due to indentation obtained using finite element simulation in a method for predicting the plastic properties of anisotropic materials based on deep learning using the indentation response data according to an embodiment of the present invention.

[0118] Referring to FIG. 6, the correlation between the radiation displacement (u r ) at a distance of 1.2R and the radiation displacement at a distance of 1.4R obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy is shown. The correlation between the radiation displacements shows a strong linearity at the R 2 = 95% level.

[0119] Referring to FIG. 7, the correlation between the vertical displacement (u z ) at a distance of 1.2R and the vertical displacement at a distance of 1.4R obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy is shown. The correlation between the vertical displacements shows a strong linearity at the R 2 = 94% level.

[0120] Referring to FIG. 8, the correlation between the radiation displacement (u r ) at a distance of 1.2R and the vertical displacement (u z ) at a distance of 1.2R obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy is shown. The DP780 steel shows a linearity at the R 2 = 92% level, the TBF1050 steel shows a linearity at the R 2 = 88% level, the AA7075 aluminum alloy shows a linearity at the R 2 = 70% level, and the AA2090 aluminum alloy shows a linearity at the R 2 = 98% level. Calculating by integrating the results of the four substances, the average R 2 = 87% level of linearity is shown. Compared with the results of FIGS. 6 and 7, it shows a slightly lower linearity but still shows a strong linearity.

[0121] From the results of FIGS. 6 to 8, for example, the radial displacements of 1.2R, 1.4R, 1.6R, the vertical displacements of 1.2R, 1.4R, and 1.6R all have linearity. Therefore, one of these data sets can be selected as a representative data set and used as an input value. However, due to the pile-up and sink-in phenomena formed around the indentation, the radial displacement measured at 1.0R can include large errors. Therefore, the vertical displacement of 1.0R showing different behavior can be selected as an additional input value. Exemplarily, the radial displacement of 1.4R and the vertical displacement of 1.0R can be selected as input values.

[0122] Hereinafter, a finite element simulation of the load with respect to the indentation depth calculated from the load-depth curve, the radial displacement by the angle calculated from the in-plane strain field, and the vertical displacement by the angle calculated from the vertical strain field will be verified by comparing with the experimental results.

[0123] FIGS. 9 to 11 are graphs comparing the results of finite element simulation and actual experimental results in a method for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0124] Referring to FIG. 9, the load with respect to the depth of the indentation obtained from the results of the finite element simulation and the experimental results for the DP780 steel and the AA7075 aluminum alloy is shown. The depth of the indentation was changed in 30-μm units as 30 μm, 60 μm, 90 μm, 120 μm, and 150 μm, and the indentation load at each indentation depth was measured. It was confirmed that the results of the finite element simulation and the experimental results very closely match the load-depth behavior for the DP780 steel and the AA7075 aluminum alloy. Therefore, since the effectiveness of the indentation model can be guaranteed, the load-depth data derived from the finite element simulation can be used as an input value in deep learning.

[0125] Referring to FIG. 10, for the DP780 steel and the AA7075 aluminum alloy, the radial displacement (u r ) at a distance of 1.4R from the reference direction, for example, from the rolling direction (RD), with respect to the angle is shown. The experimental results of the radial displacement were measured using the DIC (Digital Image Correlation) method. The dotted line was obtained using the cubic spline interpolation method. Due to the anisotropy of the DP780 steel and the AA7075 aluminum alloy, the radial displacement changes with the angle and has the lowest radial displacement at 45°, which was shown identically in both the finite element simulation results and the experimental results. It was confirmed that the finite element simulation results and the experimental results generally agree well with respect to the radial displacement behavior with respect to the angle for the DP780 steel and the AA7075 aluminum alloy. Therefore, since the validity of the indentation model can be guaranteed, the radial displacement data with respect to the angle obtained by the finite element simulation can be used as the input value in deep learning.

[0126] Referring to FIG. 11, for the DP780 steel and the AA7075 aluminum alloy, the vertical displacement (u z) is shown. The experimental results of the vertical displacement were measured using a confocal laser scanning microscope (CLSM). The dotted line was obtained using the cubic spline interpolation method. Due to the anisotropy of the DP780 steel and the AA7075 aluminum alloy, the vertical displacement changes with the angle and has the lowest vertical displacement at 90°, which was shown identically in both the results of the finite element simulation and the experimental results. It was confirmed that the results of the finite element simulation and the experimental results are in good agreement overall with respect to the vertical displacement behavior with respect to the angle for the DP780 steel and the AA7075 aluminum alloy. Therefore, since the effectiveness of the indentation model can be guaranteed, the vertical displacement data with respect to the angle obtained by the finite element simulation can be used as the input value in deep learning.

[0127] FIG. 12 is a schematic diagram of an artificial neural network for performing deep learning in a method for predicting the plastic properties of an anisotropic material using indentation response data according to an embodiment of the present invention.

[0128] Referring to FIG. 12, in the artificial neural network, the input value (Input) may be composed of indentation response data and may be composed of, for example, a total of 19. The input value is a total of 5 loads (F i ,i = 5) extracted from the depth of indentation at 30-μm intervals from the load-depth curve, a total of 7 radial displacements (u r , j ,j = 7) at a distance of 1.4R from the center of the indentation in the rolling direction at intervals of 15° from 0° to 90°, and a total of 7 vertical displacements (u z , j ,j = 7) at a distance of 1.0R from the center of the indentation in the rolling direction at intervals of 15° from 0° to 90°.

[0129] In the artificial neural network, the output value (Output) can be composed of plastic property data, for example, it may be composed of a total of 19 items. The output value may be composed of 3 hardening parameters (k, ε0, n) and 16 poly-6 anisotropy parameters (a1~a 16 )

[0130] In the artificial neural network, the hidden layer (Hidden layer) can be composed of a three-layer model with 45 neurons per layer derived as a result of hyperparameter optimization.

[0131] After the artificial neural network is deep-learned, when performing the prediction step, the indentation response data of the anisotropic substance to be predicted can be set as the input value, and the predicted plastic property data of the anisotropic substance to be predicted can be set as the output value and then performed.

[0132] The input value, the output value, and the number and configuration of the hidden layers are examples, and the technical idea of the present invention is not limited thereto.

[0133] The artificial neural network can apply the feed-forward & back propagation method and can be configured with the LM algorithm (Levenberg-Marquardt algorithm).

[0134] FIG. 13 is a graph showing the influence of the linearity of data on the performance of an artificial neural network for deep learning in a method for predicting the plastic properties of an anisotropic substance using indentation response data according to an embodiment of the present invention.

[0135] Referring to FIG. 13, the mean squared error (MSE) according to the epoch for the data sets C1~C4 input as input values to the artificial neural network is shown. The data set C1 has a radial displacement of 1.4R (u r) and vertical displacement (u z ), data set C2 includes a radial displacement of 1.4R and a radial displacement of 1.6R, data set C3 includes a radial displacement of 1.4R and a vertical displacement of 1.4R, and data set C4 includes a radial displacement of 1.4R and a vertical displacement of 1.0R.

[0136] The data sets showed different trends of decreasing mean square error in the deep learning process of the artificial neural network. In the case of the data set C1, since the vertical displacement was not included, the number of data for deep learning was relatively insufficient, and the mean square error was high. In addition, as described above, the data corresponding to 1.2R, 1.4R, and 1.6R have linearity, and the data corresponding to 1.0R does not have linearity with respect to the data corresponding to other distances, so it can be seen that the data C4, which includes the radial displacement of 1.4R and the vertical displacement of 1.0R, which excludes linearity, has a low mean square error.

[0137] Therefore, it is analyzed that the less linear the data is, the higher the performance of the artificial neural network is. When a data set is constructed that eliminates linearity as much as possible, as in C4, the prediction of plastic anisotropy using the artificial neural network can be more accurate. Therefore, it is analyzed that the selection of a radial displacement of 1.4R and a vertical displacement of 1.0R as input values is appropriate.

[0138] 14 to 16 are graphs comparing the prediction results of plastic properties using a deep learning artificial neural network with experimental results in a deep learning-based method for predicting the plastic properties of anisotropic materials using indentation response data according to one embodiment of the present invention.

[0139] In FIGS. 14 to 16, in the case of the DP780 steel and the AA7075 aluminum alloy, the indentation data measured in the actual indentation test was used as input values for deep learning of the artificial neural network. In the case of the TBF1050 steel and the AA2090 aluminum alloy, the indentation data derived using finite element simulation was used as input values for deep learning of the artificial neural network.

[0140] Referring to FIGS. 14 to 16, the prediction results of the plastic properties predicted from the learned artificial neural network are shown. Specifically, FIG. 14 shows the yield locus derived from the correlation of σ2 / σ0 with respect to σ1 / σ0. FIG. 15 shows the yield stress with respect to the angle from the reference direction. FIG. 16 shows the r-value with respect to the angle from the reference direction. It can be seen that when the artificial neural network is deep-learned using the indentation data obtained by finite element simulation or the indentation data obtained by the actual indentation experiment, the plastic properties such as the yield locus, the yield stress, and the r-value predicted by the deep-learned artificial neural network are in very good agreement with the actual experimental results.

[0141] FIG. 17 is a graph showing the plastic properties predicted using a deep-learned artificial neural network in a method for predicting the plastic properties of an anisotropic material based on deep learning using indentation response data according to an embodiment of the present invention.

[0142] Referring to Fig. 17, the true stress with respect to the equivalent plastic strain predicted using a deep - learned artificial neural network for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy is shown. This corresponds to the Swift hardening relational expression of Equation 2 above. The data points are based on actual experimental results, and the solid line is the NN prediction curve predicted by the deep - learned artificial neural network. It can be seen that in all cases, the experimental results and the prediction curve are in good agreement.

[0143] Therefore, together with the load - distance curve, when the artificial neural network is deep - learned with the radiation displacement and the vertical displacement at various indentation distances as input values, many parameters explaining plastic anisotropy can be successfully determined. Therefore, using the method for predicting the plastic properties of metal - based materials based on deep learning according to the present invention, metal forming simulations for various materials can be realized reliably and efficiently.

[0144] It will be apparent to those of ordinary skill in the technical field to which the technical idea of the present invention pertains that the technical idea of the present invention described above is not limited to the foregoing embodiments and the accompanying drawings, and various substitutions, modifications, and changes are possible without departing from the technical idea of the present invention.

[0145] Using the method for predicting the plastic properties of metal - based materials based on deep learning according to the present invention, metal forming simulations for various materials can be realized reliably and efficiently.

Claims

1. Preparing a plurality of data sets including learning push-in response data and learning plastic property data of an anisotropic material for learning; Causing a computer system to perform deep learning using the learning push-in response data as an input value and the learning plastic property data as an output value; Providing actual push-in response data of an anisotropic material to be predicted; Predicting the plastic properties of the anisotropic material to be predicted by inputting the actual push-in response data into the computer system that has been deep-learned. A method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data, comprising the steps of:

2. The step of preparing the plurality of data sets includes: Providing the tensile properties of the anisotropic material for learning; Obtaining poly6 anisotropic parameters, elastic modulus, and isotropic hardening parameters from the tensile properties of the anisotropic material for learning; Performing a finite element simulation using the poly6 anisotropic parameters, the elastic modulus, and the isotropic hardening parameters; Obtaining the learning push-in response data of the anisotropic material for learning as a result of performing the finite element simulation. The method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data according to claim 1, comprising the steps of:

3. The tensile properties include the tensile stress and r value of the anisotropic material for learning. The method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data according to claim 2,

4. The poly6 anisotropic parameter is obtained from the following formula. The method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data according to claim 2, 【Number 1】 (Here, σ ｘｘ , σ ｙｙ and σ ｚｚ are vertical stresses in the vertical direction, and σ ｘｙ , σ ｙｚ and σ ｚｘ are shear stresses.)

5. The elastic modulus includes the Young's modulus (E) and Poisson's ratio (ν) of the anisotropic material for learning. The method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data according to claim 2,

6. The isotropic hardening parameter is a strength coefficient (k), a strain parameter (ε ０ ), and a strain hardening index (n) obtained from the following formula, and a method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data according to claim 2. 【Number 2】

7. After performing the step of obtaining the poly6 anisotropic parameter, Further comprising the step of evaluating whether the poly6 anisotropic parameter of the anisotropic material for learning satisfies the convexity with respect to the yield criterion of the anisotropic material for learning. The method for predicting the plastic properties of an anisotropic material based on deep learning using push-in response data according to claim 2,

8. The output value obtained by performing the finite element simulation is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 2, which includes a load-depth curve, in-plane displacement field information, and vertical displacement field information from the results of the anisotropic material for learning.

9. The indentation response data for learning includes indentation load data, radial displacement data, and vertical displacement data for an indentation formed by indenting the anisotropic material for learning, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 1.

10. The indentation load data includes load values with respect to the depth of the indentation, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 9.

11. The radial displacement data includes radial displacement values with respect to the angle from the reference direction of the indentation, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 9.

12. The radial displacement data includes radial displacement values at a separation distance that is a multiple larger than the radius (R) of the indenter from the center of the indentation, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 11.

13. The vertical displacement data includes vertical displacements with respect to the angle from the reference direction of the indentation, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 9.

14. The vertical displacement data includes vertical displacement values at a separation distance equal to the radius (R) of the indenter from the center of the indentation, and is a deep learning-based method for predicting the plastic properties of an anisotropic material using the indentation response data according to claim 13.

15. After performing the step of predicting the plastic properties of the anisotropic material to be predicted, The method for predicting the plastic properties of an anisotropic material based on deep learning using the indentation response data according to claim 1 further includes a step of comparing the predicted plastic properties of the anisotropic material to be predicted with the actual plastic properties of the anisotropic material to be predicted.

Images