Deep learning-based method and system for predicting plastic properties of anisotropic material by using indentation response data

Abstract

The present invention provides a deep learning-based method and system for predicting the firing properties of an anisotropic material by using indentation response data, the method and the system ena

Description

Method and system for predicting plastic properties of anisotropic materials based on deep learning using indentation response data

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

[0002] When a metallic material is subjected to plastic processing such as rolling, drawing, or extrusion, or is formed into a fiber-reinforced material, or a film layer is deposited or coated, a texture is formed and grown, which may result in plastic anisotropy. Since the plastic anisotropy may change the formability required for forming processes such as bending, tension, and deep drawing, it is very important to accurately measure or predict the plastic anisotropy.

[0003] Traditionally, to measure the plastic anisotropy of a material, uniaxial tensile or uniaxial compression tests were performed repeatedly at varying angles. However, these tests are expensive and time-consuming, and because they inherently destruct the specimen, they are not applicable to limited sample volumes or small quantities. Therefore, a method for analyzing plastic anisotropy more easily, rapidly, and non-destructively is needed.

[0004] The technical task to be achieved by the technical idea of ​​the present invention is to provide a deep learning-based method and system for predicting the plastic properties of anisotropic materials using indentation response data, which can easily, quickly, and non-destructively obtain the plastic properties of anisotropic materials.

[0005] However, these tasks are exemplary and the technical idea of ​​the present invention is not limited thereto.

[0006] According to one aspect of the present invention, a method and system for predicting plastic properties of anisotropic materials based on deep learning using indentation response data capable of easily, quickly, and non-destructively obtaining plastic properties of anisotropic materials are provided.

[0007] According to one embodiment of the present invention, a method for predicting 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 each composed of indentation response data for learning and plastic property data for learning of anisotropic materials; causing a computer system to perform deep learning using the indentation response data for learning as an input value and the plastic property data for learning as an output value; providing actual indentation response data of a target 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 target anisotropic material to be predicted.

[0008] According to one embodiment of the present invention, a method for predicting plastic properties of an anisotropic material based on deep learning using the indentation response data may include the steps of: providing a computer system that performs deep learning using the indentation response data for learning and the plastic property data for learning of anisotropic materials as input values ​​and the plastic property data for learning as output values ​​in a plurality of data sets; providing actual indentation response data of a target 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 target anisotropic material to be predicted.

[0009] According to one embodiment of the present invention, a system for predicting plastic properties of an anisotropic material based on deep learning using the indentation response data comprises a finite element simulation performing module and a deep learning performing module, and comprises the steps of: a) preparing a plurality of data sets consisting of indentation response data for learning and plastic property data for learning of anisotropic materials for learning; b) causing a computer system to perform deep learning using the indentation response data for learning as an input value and the plastic property data for learning as an output value; c) providing actual indentation response data of an anisotropic material to be predicted; And d) a step of inputting the actual indentation response data into the deep-learned computer system to predict the plastic properties of the predicted target anisotropic material, wherein a deep-learning-based method for predicting the plastic properties of anisotropic materials using indentation response data is performed, and a system for predicting the plastic properties of anisotropic materials using indentation response data is performed, wherein step a) is performed by the finite element simulation performing module, and steps b) and d) can be performed by the deep-learning performing module.

[0010] According to the technical idea of ​​the present invention, a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data can easily, quickly, and non-destructively obtain plastic properties of anisotropic materials by using a non-destructive and highly efficient indentation test instead of a tensile test that entails destruction of the material, and by using an artificial neural network system that can correlate the indentation test results with the plastic properties.

[0011] The effects of the present invention described above are illustrative, and the scope of the present invention is not limited by these effects.

[0012] FIG. 1 is a flowchart illustrating a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0013] FIG. 2 is a schematic diagram illustrating an example of a deep learning-based plasticity characteristic prediction system for anisotropic materials using indentation response data according to an embodiment of the present invention.

[0014] FIG. 3 is a flowchart illustrating a step of preparing multiple data sets in a method for predicting plastic properties of an anisotropic material based on deep learning using the indentation response data of FIG. 1 according to one embodiment of the present invention.

[0015] FIGS. 4 and 5 are graphs showing displacement due to indentation calculated using finite element simulation in a deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0016] FIGS. 6 to 8 are graphs showing linearity between displacements due to indentation obtained using finite element simulation in a deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0017] FIGS. 9 to 11 are graphs comparing the results of finite element simulations and actual experiments in a deep learning-based method for predicting the plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0018] FIG. 12 is a schematic diagram of an artificial neural network performing deep learning in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0019] FIG. 13 is a graph showing the influence of data linearity on the performance of an artificial neural network performing deep learning in a 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.

[0020] Figures 14 to 16 are graphs comparing the results of predicting plastic properties using a deep-learning artificial neural network and experimental results in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0021] Fig. 17 is a graph showing plastic properties predicted using a deep-learning artificial neural network in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0022] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. Embodiments of the present invention are provided to more completely explain the technical idea of ​​the present invention to those skilled in the art. The following embodiments may be modified in various different 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 more faithfully and completely convey the technical idea of ​​the present invention to those skilled in the art. Like reference numerals throughout this specification denote like elements. Furthermore, various elements and areas in the drawings are schematically drawn. Therefore, the technical idea of ​​the present invention is not limited by the relative sizes or intervals drawn in the attached drawings.

[0023] Indentation technology measures hardness by measuring the size and depth of indentations formed by applying pressure to a specimen with an indenter. It has the advantage of being able to precisely measure the plastic properties of small specimens. 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. The hardness, elastic modulus, and other hardening properties of the target material can be measured through the load-depth curve and analysis method. In addition, a protocol has been proposed to analyze the global uniaxial tensile behavior using the finite element method based on the local load-depth curve acquired from high-resolution nano-indentation test data. The plastic properties of the target material can be inversely estimated from the load-depth curve acquired from the instrumented indentation test using a spherical or sharp indenter, based on finite element simulation and optimization algorithms. Additionally, to address the non-uniqueness that arises when numerically determining mechanical properties from load-depth curves, additional indentation information, such as dual indentation, or vertical residual indentation marks around the indentation, such as pile-up or sink-in, are sometimes used. Furthermore, instead of load-depth curves, pile-up / sink-in, in-plane displacement, and residual indentation mark profiles are sometimes used to obtain plastic properties.

[0024] However, while predicting plastic properties based on indentation data has been extensively studied for isotropic materials, research on anisotropic materials, which are closer to real materials, remains insufficient. Conventional methods assume transverse isotropy to simplify unknown material parameters and obtain plastic properties of anisotropic materials from indentation data. However, this method has limitations in predicting overall mechanical anisotropy. Furthermore, while conventional methods consider the load-depth curve and residual vertical displacement, they do not consider the residual in-plane displacement field. Therefore, we propose the use of artificial neural networks (NNs), which demonstrate excellent performance as universal approximators to extract general anisotropic plasticity from various indentation results. These neural networks can model complex relationships between input and output values ​​with high accuracy, based on statistical deep learning algorithms without using equations. To date, research using artificial neural networks on anisotropic materials remains insufficient.

[0025] The present inventors have established a general framework for analyzing the plastic properties of bulk materials. Using this framework, anisotropic properties can be comprehensively analyzed using finite element-deep learning modeling from spherical indentation responses, which consist of load-depth curves, pile-up / sink-in, and in-plane displacement fields. In the finite element-deep learning modeling, the anisotropic plastic properties obtained through deep learning using an artificial neural network with tuned hyperparameters were compared with actual experimental results, demonstrating the robustness and effectiveness of the finite element-deep learning modeling.

[0026] Below, the plastic properties of anisotropic materials will be described in detail.

[0027] The elastic behavior of continuous materials such as metals is described by the mathematical equation 1 known as Hooke's law.

[0028] [Mathematical Formula 1]

[0029] σ = Eε

[0030] In the above mathematical expression 1, σ represents stress, E represents elastic modulus, and ε represents strain. For simplicity and to focus on analyzing the plastic properties of the material, the influence on the indentation curve can be ignored and the Poisson's ratio can be set to 0.3.

[0031] To describe the strain hardening behavior from the onset of plastic yielding, a nonlinear isotropic hardening model can be adopted, and the Swift equation of Equation 2, which is a power law, can be introduced in the user-defined subroutine UHARD.

[0032] [Equation 2]

[0033]

[0034] In the above mathematical formula 2, is the Swift effective stress, is the equivalent plastic strain, k is the strength coefficient, ε0 is the strain parameter, and n is the strain hardening exponent. Hereinafter, k, ε0, and n will be referred to as Swift hardening parameters.

[0035] 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 for the Poly6 model is as shown in Mathematical Equation 3.

[0036] [Equation 3]

[0037]

[0038] In the above mathematical expression 3, σ xx , σ yy and σ zz is the normal stress related to the vertical direction, and σ xy , σ yz and σ zx is the shear stress. a1 to a 16 are independent poly6 parameters and can be defined based on uniaxial tensile test data and biaxial tensile test data.

[0039] The most common data set considered in this method is as shown in Equation 4.

[0040] [Equation 4]

[0041] Biaxial curve data set: {σ0, r0, σ b , r b , σ 90 , r 90}

[0042] Directional data set: {σ 15 , r 15 , σ 30 , r 30 , σ 45 , r 45 , σ 60 , r 60 , σ 75 , r 75}

[0043] In the above mathematical expression 4, σ b is the balanced biaxial yield stress, and r b is r b = dε yy / dε xx is the r-value defined as .

[0044] Mathematical expression 5 shows the uniaxial stress state function according to the angle.

[0045] [Equation 5]

[0046] σ(θ) = σ θ (cos 2 θ, sin 2 θ, sinθcosθ)

[0047] In the above mathematical expression 5, θ refers to an angle in a clockwise or counterclockwise direction from a reference direction, for example, a rolling direction (RD).

[0048] The r-value of a specimen is the ratio of the transverse strain (the strain component perpendicular to the direction of the indentation load) to the thickness strain. Equation 6 shows the r-value state function as a function of angle.

[0049] [Equation 6]

[0050]

[0051] Mathematical expression 7 shows the r-value considering the normality rule, the rigid-plastic approximation, and the volume-invariant plastic strain.

[0052] [Equation 7]

[0053]

[0054] Hereinafter, a method and system for predicting the plastic properties of anisotropic materials based on deep learning using indentation response data according to the technical idea of ​​the present invention will be described.

[0055] According to the technical idea of ​​the present invention, a method and system for predicting plastic properties of anisotropic materials based on deep learning using indentation response data, which can easily, quickly, and non-destructively obtain plastic properties of anisotropic materials, are provided.

[0056] The method for predicting the plastic properties of an anisotropic material based on deep learning using the above indentation response data may include 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 a target 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 target anisotropic material to be predicted.

[0057] The step of preparing the plurality of data sets may include: providing tensile properties of the learning anisotropic material; acquiring poly6 anisotropy parameters, elastic moduli, and isotropic hardening parameters from the tensile properties of the learning anisotropic material; performing a finite element simulation using the poly6 anisotropy parameters, elastic moduli, and isotropic hardening parameters; and acquiring the learning indentation response data of the learning anisotropic material as a result of performing the finite element simulation.

[0058] The above tensile properties may include tensile stress and r-value of the anisotropic material for learning.

[0059] The above poly6 anisotropy parameter can be obtained from the following equation.

[0060]

[0061] (Here, σ xx , σ yy and σ zz is the normal stress related to the vertical direction, and σ xy , σ yz and σ zx is the shear stress)

[0062] The above elastic modulus may include Young's modulus (E) and Poisson's ratio (ν) of the above learning anisotropic material.

[0063] The above isotropic hardening parameters may include a strength coefficient (k), a strain parameter (ε0), and a strain hardening exponent (n) obtained from the following equation.

[0064]

[0065] (Here, is the Swift effective stress, is the equivalent plastic strain)

[0066] After performing the step of acquiring the above poly6 anisotropy parameter, a step of evaluating whether the poly6 anisotropy parameter of the learning anisotropic material satisfies the convexity for the yield criterion of the learning anisotropic material may be further included.

[0067] The step of performing the above finite element simulation can be performed for a spherical indenter.

[0068] The output values ​​obtained by performing the above finite element simulation may include a load-depth curve, in-plane displacement field information, and vertical displacement field information from the results of the learning anisotropic material.

[0069] The above learning indentation response data may include indentation load data, radial displacement data, and vertical displacement data for indentation formed by indenting the learning anisotropic material.

[0070] The above indentation load data may include a load value for the depth of the indentation.

[0071] The above radial displacement data may include radial displacement values ​​for an angle from the reference direction of the indentation.

[0072] The above radial displacement data may include radial displacement values ​​at a distance from the center of the indentation that is a multiple of the radius (R) of the indenter.

[0073] The above vertical displacement data may include vertical displacement with respect to an angle from the reference direction of the indentation.

[0074] The above vertical displacement data may include a vertical displacement value at a distance of the radius (R) of the indenter from the center of the indentation.

[0075] The above actual indentation response data may include indentation load data, radial displacement data, and vertical displacement data for indentation formed by indenting the predicted target anisotropic material.

[0076] The plastic properties of the predicted anisotropic material may include the poly6 anisotropy parameter, elastic modulus, and isotropic hardening parameter of the predicted anisotropic material.

[0077] After performing the step of predicting the plastic properties of the predicted anisotropic material, the step of evaluating whether the poly6 anisotropy parameter of the predicted anisotropic material satisfies the convexity for the yield criterion of the predicted anisotropic material may be further included.

[0078] After performing the step of predicting the plastic properties of the predicted anisotropic material, the step of comparing the predicted plastic properties of the predicted anisotropic material with the actual plastic properties of the predicted anisotropic material may be further included.

[0079] At least one of the above learning anisotropic material and the predicted anisotropic material may include a metal alloy.

[0080] The method for predicting plastic properties of an anisotropic material based on deep learning using the above indentation response data may include the steps of: providing a computer system that performs deep learning using the learning indentation response data and the learning plastic property data of a learning anisotropic material as input values ​​and the learning plastic property data as output values ​​in a plurality of data sets; providing actual indentation response data of a target 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 target anisotropic material. In this case, the method can also be applied when there is a time interval between the time at which the computer system performs deep learning and the time at which the plastic property is predicted.

[0081] The system for predicting the plastic properties of an anisotropic material based on deep learning using the above indentation response data comprises a finite element simulation performing module and a deep learning performing module, and comprises the steps of: a) preparing a plurality of data sets consisting of learning indentation response data and learning plastic property data of a learning anisotropic material; b) causing a computer system to perform deep learning using the learning indentation response data as input values ​​and the learning plastic property data as output values; c) providing actual indentation response data of a target anisotropic material to be predicted; and d) inputting the actual indentation response data into the deep-learned computer system to predict the plastic properties of the target anisotropic material, wherein step a) may be performed by the finite element simulation performing module, and steps b) and d) may be performed by the deep learning performing module.

[0082] FIG. 1 is a flowchart illustrating a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0083] Referring to FIG. 1, the deep learning-based method for predicting plastic properties of anisotropic materials (S100) includes the steps of: preparing a plurality of data sets consisting of learning indentation response data and learning plasticity property data of a learning anisotropic material (S110); performing deep learning by causing a computer system to use the learning indentation response data as input values ​​and the learning plasticity property data as output values ​​(S120); providing actual indentation response data of a target anisotropic material to be predicted (S130); and inputting the actual indentation response data into the deep learning-enabled computer system to predict the plasticity properties of the target anisotropic material to be predicted (S140).

[0084] At least one of the above learning anisotropic material and the predicted anisotropic material may include a material having plastic anisotropy, for example, may include a metal alloy, for example, may include an ultra-high-strength steel or an aluminum alloy, for example, may include at least one of DP780, TBF1050, AA7075, and AA 2090. However, this is exemplary, and the technical idea of ​​the present invention is not limited thereto.

[0085] In this specification, when indentation response data and plasticity characteristic data are included in a dataset for deep learning, the term "learning" is added to the name. When measured through actual experiments, the term "actual" is added to the name. Of course, actual indentation response data and actual plasticity characteristic data can be used in the deep learning of an artificial neural network to improve learning accuracy.

[0086] FIG. 2 is a schematic diagram illustrating an example of a deep learning-based plasticity characteristic prediction system for anisotropic materials using indentation response data according to an embodiment of the present invention.

[0087] Referring to FIG. 2, the deep learning-based anisotropic material plasticity property prediction system may include a finite element simulation execution module (FE) and a deep learning execution module (NN).

[0088] The above finite element simulation performing module (FE) can perform finite element simulation to derive indentation response elements of anisotropic materials for learning. The finite element simulation performing module (FE) can verify whether the indentation response simulation results are consistent with the experimental indentation results, thereby verifying the validity and reliability of data collection for deep learning of an artificial neural network. The finite element simulation performing module (FE) that performs the above finite element simulation can be configured as a computer system.

[0089] The above finite element simulation execution module (FE) can perform the step (S110) of preparing multiple data sets of FIG. 1.

[0090] The number of training indentation response data and training plasticity characteristic data included in the above data set may be, for example, 100 or more sets to ensure the accuracy of the output value, and may be, for example, 1,000 or more sets to further increase the accuracy, and may be, for example, 50,000 or more sets. However, the technical idea of ​​the present invention is not limited thereto. The greater the number of data sets, the higher the accuracy of the output value, but it may be appropriately selected taking time and cost into consideration. The above data set may be provided through actual experimental results, or may be provided by repeatedly performing a finite element simulation while changing the input value.

[0091] FIG. 3 is a flowchart illustrating a step of preparing multiple data sets in a method for predicting plastic properties of an anisotropic material based on deep learning using the indentation response data of FIG. 1 according to one embodiment of the present invention.

[0092] Referring to FIG. 3, the step of preparing the plurality of data sets (S110) may include the step of providing the tensile properties of the learning anisotropic material (S111); the step of acquiring poly6 anisotropy parameters, elastic moduli, and isotropic hardening parameters from the tensile properties of the learning anisotropic material (S112); the step of performing a finite element simulation using the poly6 anisotropy parameters, elastic moduli, and isotropic hardening parameters (S113); and the step of acquiring the learning indentation response data of the learning anisotropic material as a result of performing the finite element simulation (S114).

[0093] In the step (S111) of providing the tensile properties of the anisotropic material for learning, as shown in the mathematical expression 4, the tensile stress (σ0~ σ) of the anisotropic material for learning is provided as an example of the tensile properties of the anisotropic material for learning. 90 , σ b ) and r-value (r0~ r 90 , r b ) is provided. The tensile stress and the r-value of the above learning anisotropic material may include test results obtained by directly performing a tensile test on the above learning anisotropic material.

[0094] In the above acquisition step (S112), the learning plasticity characteristic data is acquired from the tensile characteristics, for example, the poly6 anisotropy parameter (a1~ a 16), elastic constant, and isotropic hardening parameter can be obtained. The poly6 anisotropic parameter can be obtained from the above mathematical expression 3. From the tensile properties, the elastic constant and isotropic hardening parameter of the learning anisotropic material 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), the strain parameter (ε0), and the strain hardening exponent (n) obtained from the above mathematical expression 2.

[0095] After performing the step (S112) of acquiring the poly6 parameter, a step of evaluating whether the poly6 parameter of the learning anisotropic material satisfies the convexity for the yield criterion of the learning anisotropic material may be further included. If the poly6 parameter does not satisfy the convexity, the step of providing tensile properties by feedback is performed again. If the poly6 parameter satisfies the convexity, the poly6 anisotropy parameter, the elastic modulus, and the isotropic hardening parameter are set as input values ​​for finite element simulation.

[0096] In the step (S113) of performing the finite element simulation, the finite element simulation is performed using the poly6 anisotropy parameter, the elastic modulus, and the isotropic hardening parameter as input values. The step (S113) of performing the finite element simulation may be performed for a spherical indentation, but this is exemplary and the technical idea of ​​the present invention is not limited thereto. The output value of the finite element simulation may include a load-depth curve of the learning anisotropic material, an in-plain displacement field of the surface, and a vertical displacement field.

[0097] In the step (S114) of acquiring the above learning pressure response data, a process of extracting feature data from the output value is performed for subsequent deep learning performance, and thus, the learning pressure response data can be acquired as the feature data.

[0098] The above learning pressure response data may include indentation load data, radial displacement data, and vertical displacement data of the learning anisotropic material.

[0099] Referring again to FIG. 2, the deep learning execution module (NN) can perform the step (S120) of performing deep learning of FIG. 1.

[0100] The above deep learning execution module (NN) can cause a computer system to perform deep learning using the training pressure response data as input and the training plasticity characteristic data as output. The deep learning can be performed at least once, and accordingly, the artificial neural network equipped in the computer system can perform deep learning.

[0101] The deep learning execution module (NN) that performs the above deep learning may be configured as a computer system. The finite element simulation execution module (FE) and the deep learning execution module (NN) may be configured as the same computer system or as separate computer systems. The computer system referred to herein includes an artificial intelligence program, an artificial neural network program, or any system equipped with such a program.

[0102] Referring back to FIGS. 1 and 2, a step (S130) of providing actual indentation response data of a target anisotropic material is performed. The actual indentation response data may include indentation load data, radial displacement data, and vertical displacement data of the target anisotropic material. The actual indentation response data may include test results obtained by directly indenting the target anisotropic material. For example, after obtaining a load-depth curve, in-plane displacement field information of a surface, and vertical displacement field information by directly indenting the target anisotropic material, the indentation load data, the radial displacement data, and the vertical displacement data may be obtained by extracting characteristic data therefrom. Alternatively, the actual indentation response data may be obtained by performing the finite element simulation.

[0103] Referring back to FIGS. 1 and 2, a step (S140) is performed to input the actual indentation response data into the deep learning computer system to predict the plastic properties of the target anisotropic material. The plastic properties may include the poly6 anisotropy parameter, elastic modulus, and isotropic hardening parameter of the target anisotropic material.

[0104] After performing the step (S140) of predicting the plastic properties of the anisotropic material to be predicted, the method may further include a step of evaluating whether the poly6 anisotropy parameter of the anisotropic material to be predicted predicted by the deep-learned computer system satisfies the convexity for the yield criterion of the anisotropic material to be predicted. If the poly6 parameter does not satisfy the convexity, it is fed back and deep learning is performed again. If the poly6 parameter satisfies the convexity, the poly6 anisotropy parameter, the elastic modulus, and the isotropic hardening parameter of the anisotropic material to be predicted are output as the plastic properties of the anisotropic material to be predicted.

[0105] The above poly6 anisotropy parameters are the parameters (a1~ a) shown in the above mathematical expression 3. 16 ) may include. The elastic modulus may include Young's modulus (E) and Poisson's ratio (ν). The isotropic hardening parameter may include the strength coefficient (k), the strain parameter (ε0), and the strain hardening exponent (n) shown in the above mathematical expression 2.

[0106] The step of evaluating the convexity of the predicted target anisotropic material can also be performed in the step of performing deep learning (S120).

[0107] Optionally, after performing the step of predicting the plastic properties of the target anisotropic material, the method may further include a step of comparing the predicted plastic properties of the target anisotropic material with actual plastic properties of the target anisotropic material. The actual plastic properties may be test results obtained by directly experimenting on the target anisotropic material. 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 may be terminated, and if it is outside the certain error rate range, the prediction step may be performed again.

[0108] The above-described deep learning is a term widely used in the field of artificial intelligence, and the present invention does not specifically limit its method. According to one embodiment of the present invention, by inputting the data set into an artificial neural network (or artificial intelligence) and performing deep learning (or machine learning), an artificial neural network can be constructed that predicts plasticity characteristic data as output when indentation response data is input. The principle of constructing such an artificial neural network is similar to a regression problem for a general linear function. That is, a function that indicates the relationship between the input and output values ​​of given data sets is created. The deep learning method of the artificial neural network is a widely known method such as general deep learning, and the present invention may adopt any method among the already known methods.

[0109] As an example of the above deep learning, a case will be described by way of example in which an artificial neural network is configured with 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.

[0110] An input value is entered into the first column, and then each node in the first column is connected to each node in the second column by a line, and the node values ​​in the second column are determined using the input value in the first column through each line. The first node at the top of the second column follows the same calculation as relational expression A.

[0111] [Relationship A]

[0112] value1 = 1 / (1 + exp(-(a1_1

[0113] In the above relational expression A, value1 represents the value of the second column, the first node, a1_1 to a1_19 each represent the weight of each input value for determining the value of the first node of the second column, and ε1 to ε19 each represent 19 input values.

[0114] Next, the second node of the second column also follows the same calculation, and specifically follows the calculation of the following relation B.

[0115] [Relationship B]

[0116] Value2 = 1 / (1 + exp(-(a2_1

[0117] In the above relational expression B, value2 represents the value of the second column and the second node, a2_1 to a2_19 each represent the weight of each input value for determining the value of the second node of the second column, and ε1 to ε19 each represent 19 input values.

[0118] This same process can be repeated from the third column to the fifth column, which is the last in the sequence. Then, by changing the values ​​of the unknown variables, denoted as a#_# (where # represents a number), a function called an artificial neural network can be used to find the relationship between the input and output of real data.

[0119] To this end, a random number is initially assigned to the a#_#, and the difference between the calculated output value and the output of the actual data is calculated, and the values ​​of the a#_# are modified using gradient descent to reduce this difference. This process of modifying the values ​​of unknown variables by utilizing the difference between the output values ​​is called back-propagation. This process of the artificial neural network finding the relationship between actual data while performing back-propagation corresponds to deep learning (learning) in artificial intelligence.

[0120] Unlike conventional methods that require separate user intervention, the present invention utilizes the deep learning process of artificial intelligence, enabling the user to simply input measured input values ​​into a deep-learning computer system, resulting in instantaneous predictions. Therefore, the system can be easily integrated with existing experimental equipment, resulting in significant time and cost savings.

[0121] According to the technical concept of the present invention, a deep artificial neural network (NN) model capable of identifying nonlinear correlations between discontinuous input data and output data can be applied. To improve the performance of the artificial neural network model, the basic hyperparameters, i.e., the number of hidden layers and the number of neurons per hidden layer, can be tuned using a Bayesian optimization algorithm to optimize the artificial neural network model structure. The number of neurons per hidden layer can be set to the same value, and by optimizing the hyperparameters, a 3-layer model with 45 neurons per hidden layer can be applied. The prepared data set can be randomly assigned to 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 trained artificial neural network can be evaluated using a test data set excluded during deep learning. New information, unused in deep learning, can be provided to a deep-learned artificial neural network, and the calculated and measured parameters can be compared. Since the initial weights, biases, and dataset partitioning can be configured differently for deep learning, sufficient generalization can be ensured by independently repeating deep learning multiple times, for example, five times, and calculating the mean squared error and output. The maximum number of epochs for deep learning can be set to, for example, 30 or more, for example, 200. This is sufficiently large to stabilize the artificial neural network system and enable early stopping methods to operate.

[0122] Experimental example

[0123] Below, experimental examples are described to aid understanding of the present invention. The following experimental examples are presented to aid understanding of the invention, and are not limited to the following experimental examples of the present invention.

[0124] Materials exhibiting plastic anisotropy were selected as learning anisotropic materials, and specifically, two ultra-high-strength steels (DP780 and TBF1050) and two aluminum alloys (AA7075 and AA 2090) were selected.

[0125] To verify the finite element simulation for the above indentation as a means of creating a database and to confirm the validity of the deep learning prediction results, a tensile test was performed on the above learning anisotropic material.

[0126] The above learning anisotropic material was subjected to a tensile test to obtain tensile properties. The tensile properties were obtained by measuring the tensile stress (σ0~ σ) of the learning anisotropic material. 90 , σ b ) and r-value (r0~ r 90 , r b ) were included. The tensile test of the above learning anisotropic materials was performed as follows. First, each of the learning anisotropic materials was manufactured into a tensile specimen with a thickness of 1 mm. The tensile specimens were subjected to a uniaxial tensile test according to the ASTM-E08 standard using a general-purpose tensile tester (Instron 5584, USA). The tensile specimens were subjected to a 10 -3 s -1 Quasi static tensile tests were performed in the rolling direction (RD), diagonal direction (DD), and transverse direction (TD) at a nominal strain rate.

[0127] Using the analysis solution for the above tensile test results, a total of 19 parameters were obtained by combining 16 anisotropic parameters of the poly6 model and 3 isotropic parameters of the Swift hardening model.

[0128] The indentation test of the above-mentioned study anisotropic materials was performed as follows. Each of the above-mentioned study anisotropic materials was prepared as an indentation specimen having a size of 60 x 50 x 1 mm and a rectangular shape. Prior to performing the indentation test, the surface of the indentation specimen was mechanically ground and polished using an alumina suspension containing alumina particles with a diameter of up to 1 μm. An instrumented indentation test was performed on the indentation specimens using an AIS3000 system (Frontics, Inc., Korea) with 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. Since a sufficient area under the indentation must be allowed to accurately describe the tensile behavior of the bulk material, the indentation depth was controlled to a maximum of 150 μm using displacement control. To ensure reliability, the indentation was performed five times and the average value was obtained.

[0129] In addition, tensile tests and indentation tests of the predicted anisotropic material were performed in the same manner.

[0130] In order to obtain the in-plane displacement field information and vertical displacement field information of the above press-fit specimen, images of the above press-fit specimen before and after the press-fit test were obtained.

[0131] Since the acquisition of in-plane displacement field information of the above-mentioned indentation specimen requires a non-contact, non-destructive, and simple optical setup, the Digital Image Correlation (DIC) technique used for strain measurement was used. The spot pattern was formed by spray-painting the surface of the indentation specimen before indentation. The in-plane displacement field of the above-mentioned indentation specimen was calculated by acquiring digital images before and after deformation, and tracking each point of the images using the MATLAB DIC application.

[0132] The above 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 multiple 2D images at regular intervals and accumulating them, the 3D topography of the indentation can be reconstructed. Image processing was performed using VK-analyzer software, and pile-up / sync-in information was obtained by observing the vertical cross-section of the 3D topography.

[0133] Finite element simulations of the above-mentioned anisotropic material for study were performed using commercial ABAQUS software. A 1 / 4 finite element indentation model consisting of three-dimensional eight-node continuous brick elements (C3D8R) with reduced integration points was used. In the finite element simulations, the indenter shape was a spherical indenter, identical to that used in the experiments. 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, which could eliminate geometrical effects compared to the indenter size and indentation depth. The mechanical boundary conditions of lateral axisymmetric and bottom-fixed were imposed on the specimen model. As in the experiments, a displacement condition of 0.3 mm / min in the z-axis direction was applied to the indenter until the indentation depth reached a maximum of 150 μm. During indentation, the interaction between the indenter and the specimen surface was controlled using the surface-to-surface contact pair algorithm, normal hard contact, and tangential friction. Since the friction coefficient has a significant effect on the load-depth curve for deep spherical indentations greater than 50 μm, an optimal friction coefficient of 0.12 was used. A preliminary investigation was conducted on contact area mesh improvement and convergence.

[0134] Table 1 is a table showing the mechanical parameters of anisotropic materials for learning in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0135] substancea1a2a3a4a5a6a7a8ke0na9a 10 a 11 a 12 a 13 a 14 a 15 a 16 TBF10501-2.926.42-7.695.99-2.350.7465.4417170.0060.16-13.6723.18-17.808.6724.99-24.3220.9421. 97DP7801-2.645.85-7.385.89-2.801.049.6813150.0050.14-19.730.42-18.409.6828.63-22.9027.2433.16 AA2090-T31-1.052.19-4.755.97-4.321.7615.366460.0250.227-3.63-10.885.1516.067.43-9.0485.58117. 42AA7075-T61-2.417.28-11.188.38-3.1818.978820.0450.154-20.9730.58-15.406.0417.17-13.2923.0127

[0136] Referring to Table 1, for all the above-mentioned learning anisotropic materials, since a1 = 1, there are 18 independent parameters. Since these parameters are predicted based on indentation response data, it is necessary to carefully consider the strain field formed by indentation. The parameter analysis was performed using a random generation method for each parameter, resulting in a total of 50,000 data sets.

[0137] Table 2 shows the parameters and their ranges input to the finite element simulation 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.

[0138] Model Swift hardening yield stress r-value parameter k (MPa) ε 0 n σ 0 σ 15 ~σ 90, σ b r0~r 90, r b RangeMinimum 3000.00010.0110.80.2Maximum 20000.050.51.22.0

[0139] In Table 2, the Young's modulus was set to 70 GPa for the aluminum alloy and 200 GPa for the iron alloy. The Poisson's ratio was set to 0.3 for the aluminum alloy and the iron alloy.

[0140] Next, a finite element simulation was performed using the poly6 anisotropy parameters, elastic modulus, and isotropic hardening parameters (Swift hardening parameters) obtained from the tensile properties of the above-mentioned learning anisotropic material.

[0141] The above finite element simulation can be performed repeatedly to produce output values. The output values ​​include a load-depth curve, an in-plain displacement field, and a vertical displacement field.

[0142] To perform subsequent deep learning, characteristic data was extracted from the output values, and training indentation response data was acquired as the characteristic data. The training indentation response data included indentation load data, radial displacement data, and vertical displacement data of the training anisotropic material.

[0143] FIGS. 4 and 5 are graphs showing displacement due to indentation calculated using finite element simulation in a deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0144] Referring to Figures 4 and 5, the radial displacement (u) obtained by performing a finite element simulation on DP780 steel r ) and vertical displacement (u z ) are shown respectively. The above radial displacement (u r ) and vertical displacement (u z ) were acquired at various angles from the rolling direction (RD) from the residual strain field around the indentation, for example, 0 degrees, 15 degrees, 30 degrees, 45 degrees, 60 degrees, 75 degrees, and 90 degrees, respectively. The acquisition locations of the radial displacement and the vertical displacement are each shown on the upper side. The acquisition locations were set by changing the angle at 15-degree intervals from the rolling direction (RD) at a multiple of the radius (R) of the indenter based on the center of the indentation. The multiple of the distance was 1.0R, 1.2R, 1.4R, and 1.6R.

[0145] The above radial displacement and the vertical displacement have the same value at the same separation distance regardless of the angle in the case of an isotropic material. However, it can be seen that different values ​​are shown in an anisotropic material with plastic anisotropy, such as the DP780 steel. That is, even though the values ​​were acquired at the same distance from the indentation center, a deviation with different values ​​appeared as the angle changed, and the minimum value was shown when the value was 45 degrees from the rolling direction at all distances. The deviation was larger in the radial displacement. The behavior of the radial displacement and the behavior of the vertical displacement appeared similar at 1.2R, 1.4R, and 1.6R, whereas they appeared different at a distance of 1.0R. This anisotropic behavior was also similarly observed in TBF1050 steel, AA7075 aluminum alloy, and AA2090 aluminum alloy.

[0146] The plastic region of the uniaxial tensile curve in the rolling direction (RD) can be used as a reference state for determining the hardening parameters for each material. Conventionally, 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, the anisotropic parameters can be defined from the results of the indentation test, as shown in FIGS. 4 and 5.

[0147] This similar behavior suggests that the data presented in Figures 4 and 5 are not valid input values ​​for deep learning. Therefore, to verify the validity of the data, we examine the linearity of the radial displacement and the vertical displacement.

[0148] FIGS. 6 to 8 are graphs showing linearity between displacements due to indentation obtained using finite element simulation in a deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0149] Referring to Fig. 6, the radial displacement (u) of 1.2R distance obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy r ) and the radial displacement at a distance of 1.4R. The correlation between the above radial displacements is R 2 = Shows strong linearity at the 95% level.

[0150] Referring to Fig. 7, the vertical displacement (u) of 1.2R distance obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy z ) and the vertical displacement of 1.4R distance. The correlation between the vertical displacements is R 2 = It shows strong linearity at the 94% level.

[0151] Referring to Fig. 8, the radial displacement (u) of 1.2R distance obtained for the DP780 steel, the TBF1050 steel, the AA7075 aluminum alloy, and the AA2090 aluminum alloy r ) and vertical displacement (u) at a distance of 1.2R z ) shows the correlation. The above DP780 steel is R 2 = It showed linearity of 92%, and the TBF1050 steel was R 2 = It showed linearity of 88%, and the AA7075 aluminum alloy was R 2 = showed linearity of 70%, and the AA2090 aluminum alloy was R 2 = It showed linearity at the level of 98%. When the results of the above four substances were integrated and calculated, the average R 2 = It showed linearity at the level of 87%. Compared to the results of Figs. 6 and 7, it shows somewhat lower linearity, but still shows strong linearity.

[0152] From the results of FIGS. 6 to 8, for example, the 1.2R radial displacement, the 1.4R radial displacement, the 1.6R radial displacement, the 1.2R vertical displacement, the 1.4R vertical displacement, and the 1.6R vertical displacement all have linearity, so one data set among them can be selected as an input value as a representative data set. However, the radial displacement measured at 1.0R may contain a large error due to the pile-up and sink-in phenomena formed around the indentation. Therefore, the 1.0R vertical displacement, which exhibits different behavior, can be selected as an additional input value. For example, the 1.4R radial displacement and the 1.0R vertical displacement can be selected as input values.

[0153] In the following, the finite element simulations for load versus indentation depth derived from the load-depth curve, radial displacement versus angle derived from the in-plane strain field, and vertical displacement versus angle derived from the vertical strain field are compared with experimental results to verify the results.

[0154] FIGS. 9 to 11 are graphs comparing the results of finite element simulations and actual experiments in a deep learning-based method for predicting the plastic properties of anisotropic materials using indentation response data according to an embodiment of the present invention.

[0155] Referring to Fig. 9, the loads for indentation depths obtained from the finite element simulation results and experimental results for the DP780 steel and the AA7075 aluminum alloy, respectively, are shown. The indentation depths were varied from 30 μm to 30 μm, 60 μm, 90 μm, 120 μm, and 150 μm in 30 μm increments, and the indentation loads at each indentation depth were measured. It was confirmed that the finite element simulation results and the experimental results were very consistent with each other regarding the load-depth behavior for the DP780 steel and the AA7075 aluminum alloy. Therefore, the validity of the indentation model can be guaranteed, and thus the load-depth data derived from the finite element simulation can be used as input values ​​in deep learning.

[0156] Referring to Fig. 10, for the DP780 steel and the AA7075 aluminum alloy, the radial displacement (u) at a distance of 1.4R from the reference direction, for example, the rolling direction (RD), obtained from the finite element simulation results and the experimental results, respectively, for the DP780 steel and the AA7075 aluminum alloy. r) 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 varied with angle, and had the lowest radial displacement at 45 degrees, which was the same in both the finite element simulation results and the experimental results. It was confirmed that the finite element simulation results and the experimental results were in good overall agreement with respect to the radial displacement behavior with respect to angle for the DP780 steel and the AA7075 aluminum alloy. Therefore, the validity of the indentation model can be guaranteed, and thus the radial displacement data with respect to angle obtained by the finite element simulation can be used as an input value in deep learning.

[0157] Referring to Fig. 11, for the DP780 steel and the AA7075 aluminum alloy, the vertical displacement (u) at a distance of 1.0R from the reference direction, for example, the rolling direction (RD), obtained from the finite element simulation results and the experimental results, respectively, for the DP780 steel and the AA7075 aluminum alloy z) is shown. The experimental results of 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 varied with angle and had the lowest radial displacement at 90 degrees, which was the same in both the finite element simulation results and the experimental results. It was confirmed that the finite element simulation results and the experimental results were in good overall agreement with respect to the vertical displacement behavior with respect to angle for the DP780 steel and the AA7075 aluminum alloy. Therefore, the validity of the indentation model can be guaranteed, and therefore, the vertical displacement data with respect to angle obtained by the finite element simulation can be used as an input value in deep learning.

[0158] FIG. 12 is a schematic diagram of an artificial neural network performing deep learning in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0159] Referring to Fig. 12, the input values ​​in the artificial neural network may be composed of indentation response data, and may be composed of, for example, a total of 19. The input values ​​may be a total of 5 loads (F) extracted at indentation depths of 30 μm intervals from the load-depth curve. i , i=5), a total of seven radial displacements (u) at 15-degree intervals from 0 to 90 degrees from the rolling direction at a distance of 1.4R from the indentation center. r, j , j=7), and a total of seven vertical displacements (u) at 15 degree intervals from 0 to 90 degrees from the rolling direction at a distance of 1.0R from the indentation center. z, j , j=7) can be composed.

[0160] In the above artificial neural network, the output value can be composed of plasticity characteristic data, and can be composed of, for example, a total of 19. The output value can be composed of three hardening parameters (k, ε0, n) and 16 poly6 anisotropy parameters (a1 to a 16 ) can be composed of.

[0161] In the above artificial neural network, the hidden layer can be configured as a 3-layer model with 45 neurons per layer derived as a result of hyperparameter optimization.

[0162] When performing a prediction step after the above artificial neural network has been deep-learned, the step can be performed by setting the indentation response data of the target anisotropic material to be predicted as an input value and setting the predicted plasticity characteristic data of the target anisotropic material to be predicted as an output value.

[0163] The above input values, output values, and the number and configuration of the hidden layers are exemplary, and the technical idea of ​​the present invention is not limited thereto.

[0164] The above artificial neural network can apply the feed-forward & back propagation method and can be configured with the LM algorithm (Levenberg-Marquardt algorithm).

[0165] FIG. 13 is a graph showing the influence of data linearity on the performance of an artificial neural network performing deep learning in a 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.

[0166] Referring to Figure 13, the mean square error (MSE) according to epoch for data sets C1 to C4 input to the artificial neural network is shown. The data set C1 has a 1.4R radial displacement (u r ) and vertical displacement (u z), and the data set C2 includes 1.4R radial displacement and 1.6R radial displacement, the data set C3 includes 1.4R radial displacement and 1.4R vertical displacement, and the data set C4 includes 1.4R radial displacement and 1.0R vertical displacement.

[0167] Depending on the above data sets, the tendency of decreasing mean square error in the deep learning process of the artificial neural network was different. In the case of the above data set C1, since it does not include the vertical displacement, the number of data for deep learning was relatively insufficient, resulting in a high mean square error. 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 do not have linearity with respect to data corresponding to other distances. Therefore, it can be seen that the above C4, which includes the 1.4R radial displacement and the 1.0R vertical displacement that exclude linearity, has a low mean square error.

[0168] Therefore, it is analyzed that the performance of the artificial neural network increases as the data linearity is absent or low. When configuring a data set that minimizes linearity, as in C4 above, the prediction of plastic anisotropy using the artificial neural network can be more accurate. Therefore, the selection of 1.4R radial displacement and 1.0R vertical displacement as input values ​​is analyzed as appropriate.

[0169] Figures 14 to 16 are graphs comparing the results of predicting plastic properties using a deep-learning artificial neural network and experimental results in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0170] In FIGS. 14 to 16, in the case of the DP780 steel and the AA7075 aluminum alloy, the indentation data measured in an actual indentation test was used as input values ​​for deep learning of an artificial neural network, and 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 an artificial neural network.

[0171] Referring to FIGS. 14 to 16, the prediction results of plastic properties predicted from the learned artificial neural network are shown. Specifically, FIG. 14 shows a yield locus derived from the correlation of σ2 / σ0 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. When the artificial neural network is deep-learned using the indentation data acquired through the finite element simulation, or when the artificial neural network is deep-learned using the indentation data acquired through the actual indentation experiment, it can be seen that the plastic properties such as the yield locus, the yield stress, and the r-value predicted by the deep-learned artificial neural network are very consistent with the actual experimental results.

[0172] Fig. 17 is a graph showing plastic properties predicted using a deep-learning artificial neural network in a method for predicting plastic properties of anisotropic materials based on deep learning using indentation response data according to an embodiment of the present invention.

[0173] Referring to Fig. 17, the true stress for equivalent plastic strain predicted using a deep-learning 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 relationship of Equation 2. The data points are based on actual experimental results, and the solid line is the predicted curve (NN prediction) predicted by the deep-learning artificial neural network. In all cases, it can be seen that the experimental results and the predicted curves are in good agreement.

[0174] Therefore, by deep learning an artificial neural network using radial and vertical displacements at various indentation distances along with load-distance curves as inputs, numerous parameters describing plastic anisotropy can be successfully determined. Therefore, the deep learning-based plastic properties prediction method for metallic materials according to the present invention can be used to reliably and efficiently implement metal forming simulations for various materials.

[0175] It will be apparent to a person skilled in the art 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 above-described embodiments and the attached drawings, and that various substitutions, modifications, and changes are possible within a scope that does not depart from the technical idea of ​​the present invention.

[0176] Using the deep learning-based method for predicting the plasticity properties of metal-based materials according to the present invention, metal forming simulations for various materials can be implemented reliably and efficiently.

Claims

1. A step of preparing multiple data sets consisting of learning indentation response data and learning plasticity characteristic data of a learning anisotropic material; A step of causing a computer system to perform deep learning using the above learning pressure response data as input values ​​and the above learning plasticity characteristic data as output values; A step of providing actual indentation response data of a predicted target anisotropic material; and A step of inputting the actual pressure response data into the deep learning computer system to predict the plastic properties of the target anisotropic material, A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

2. In paragraph 1, The step of preparing the above multiple data sets is: A step of providing tensile properties of the above learning anisotropic material; A step of obtaining poly6 anisotropy parameters, elastic modulus, and isotropic hardening parameters from the tensile properties of the above learning anisotropic material; A step of performing a finite element simulation using the above poly6 anisotropic parameter, the elastic modulus, and the isotropic hardening parameter; and As a result of performing the above finite element simulation, a step of acquiring the above learning indentation response data of the above learning anisotropic material is included. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

3. In paragraph 2, The above tensile properties include the tensile stress and r-value of the anisotropic material for learning. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

4. In paragraph 2, The above poly6 anisotropy parameter is obtained from the following equation: (Here, σ xx , σ yy and σ zz is the normal stress related to the vertical direction, and σ xy , σ yz and σ zx is the shear stress) A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

5. In paragraph 2, The above elastic modulus includes the Young's modulus (E) and Poisson's ratio (ν) of the learning anisotropic material. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

6. In paragraph 1, The above isotropic hardening parameters include the strength coefficient (k), the strain parameter (ε0), and the strain hardening exponent (n) obtained from the following equation. (Here, is the Swift effective stress, is the equivalent plastic strain) A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

7. In paragraph 2, After performing the step of obtaining the above poly6 anisotropy parameters. Further comprising a step of evaluating whether the poly6 anisotropic parameter of the learning anisotropic material satisfies the convexity for the yield criterion of the learning anisotropic material. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

8. In paragraph 2, The output value of the above finite element simulation includes a load-depth curve, in-plane displacement field information, and vertical displacement field information from the results of the learning anisotropic material. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

9. In paragraph 1, The above learning indentation response data includes indentation load data, radial displacement data, and vertical displacement data for indentation formed by indenting the learning anisotropic material. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

10. In paragraph 9, The above indentation load data includes a load value for the depth of the indentation, A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

11. In paragraph 9, The above radial displacement data includes radial displacement values ​​for an angle from the reference direction of the indentation. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

12. In paragraph 11, The above radial displacement data includes radial displacement values ​​at a distance multiple of the radius (R) of the indenter from the center of the indentation. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

13. In paragraph 9, The above vertical displacement data includes vertical displacement with respect to an angle from the reference direction of the indentation, A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

14. In paragraph 13, The above vertical displacement data includes the vertical displacement value at a distance of the radius (R) of the indenter from the center of the indentation. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

15. In paragraph 1, After performing the step of predicting the plastic properties of the above predicted target anisotropic material, Further comprising a step of comparing the predicted plastic properties of the predicted anisotropic material with the actual plastic properties of the predicted anisotropic material. A deep learning-based method for predicting plastic properties of anisotropic materials using indentation response data.

Images