Method for efficiently predicting rigidity of three-dimensional woven composite material based on neural network

Through a neural network-based method combined with a mesoscopic mechanic model, the in-plane equivalent stiffness of a three-dimensional woven composite material is predicted in stages, solving the problems of low accuracy and long calculation time in the existing technology, and achieving high efficiency and precision material stiffness prediction.

CN118942580AInactive Publication Date: 2024-11-12HARBIN INST OF TECH
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410916204.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-09
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art is difficult to efficiently predict the stiffness of three-dimensional woven composite materials, and the traditional method has low accuracy and long calculation time. Especially because the mesoporical structure of textile composite materials is complex, it is difficult to determine the mapping relationship between mesoporical parameters and mechanical properties.

Method used

Using a neural network-based method, the in-plane equivalent stiffness of three-dimensional woven composite materials is predicted in stages. The neural network model combined with the mesoscopic mechanical analytical model is used to screen out the key parameters that affect the stiffness of the sewing warp yarn sub-cells, and use it The Chamis model and the mixed law model were corrected to predict the meridian, zonal, and shear elastic modulus, and combined with the Voigt and Reuss models, the overall stiffness of the material was further predicted.

Benefits of technology

High-precision three-dimensional woven composite stiffness prediction is achieved, the calculation efficiency is greatly improved, and the error is controlled within 8%, which significantly saves computing resources and time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942580A_ABST
    Figure CN118942580A_ABST
Patent Text Reader

Abstract

The invention discloses a three-dimensional woven composite material rigidity efficient forecasting method based on a neural network, and belongs to the field of composite material static mechanical property evaluation. The method comprises the following steps: a first stage: predicting an in-plane equivalent elastic modulus of a sub unit cell containing a single seam warp yarn based on a neural network model; in the second stage, after the in-plane elasticity modulus of the seam warp yarn sub unit cells is obtained, weft-direction fiber bundles are inserted in the transverse direction of the unit cells to obtain the sub unit cells, and the in-plane elasticity modulus of the sub unit cells is predicted and comprises the radial elasticity modulus, the weft-direction elasticity modulus and the in-plane shear elasticity modulus; and a third stage: predicting the unit cell radial elastic modulus and in-plane shear elastic modulus of the three-dimensional woven composite material based on the Voigt model, and predicting the unit cell latitudinal elastic modulus of the three-dimensional woven composite material based on the Russ model. According to the three-dimensional woven composite material rigidity efficient prediction method based on the neural network, the neural network model and the mesomechanics analysis model are combined, and in-plane equivalent rigidity prediction of unit cells of the three-dimensional woven composite material can be achieved in stages.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of static mechanical performance evaluation of composite materials, and specifically relates to a method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network. Background Art

[0002] Three-dimensional woven composites are gradually gaining favor in aerospace and other industrial fields due to their higher damage tolerance and more balanced overall mechanical properties. Accurate analysis and evaluation of the mechanical properties of three-dimensional woven composites is the basis for structural optimization design. Three-dimensional woven composite structures often contain many types of mesoscopic unit cells. In order to accurately analyze the mechanical behavior of the structure, the equivalent performance of multiple unit cells is required as input parameters. Therefore, it is very important to efficiently analyze the mechanical properties of the unit cells.

[0003] At present, the stiffness prediction methods for textile composites are mainly based on analytical methods and finite element methods. The analytical method is simple in form, but its accuracy is easily affected by the material structure; the finite element method is mature in theory and has high calculation accuracy, but it often takes a long time from modeling to numerical calculation. The key to predicting the stiffness of composite materials is to establish a quantitative relationship between the mechanical properties of the material and the material structural parameters. However, due to the numerous microstructures and material parameters of textile composites, it is difficult for traditional mechanical models to effectively determine the mapping relationship between many microscopic parameters and mechanical properties. The machine learning methods that have emerged in recent years are very good at constructing a mapping relationship between multi-parameter features and material properties. At the same time, there are very few studies on combining neural network models with analytical methods to predict the equivalent stiffness performance of three-dimensional woven composites. Therefore, it is of great significance to propose an efficient prediction method based on a neural network model that does not rely on finite element calculations but can approach the accuracy of finite element calculations for the performance prediction and optimization design of three-dimensional woven composites. Summary of the invention

[0004] The purpose of the present invention is to solve the above problems existing in the prior art and to propose a method for efficiently predicting the stiffness of a three-dimensional woven composite material based on a neural network.

[0005] The present invention provides a high-efficiency prediction method for the stiffness of a three-dimensional woven composite material based on a neural network. In a six-step three-dimensional woven composite material (a three-dimensional woven composite material widely used now, the naming of the six-step method is related to the path period of the warp yarn, and the stiffness prediction method of the present invention is proposed for the six-step three-dimensional woven composite material), there are two types of yarns, namely, fluctuating warp yarns and straight weft fiber bundles. The overall structure interweaves the weft fiber bundles of the upper and lower layers through the warp yarns. Considering the complexity of the six-step fabric structure, the prediction of the stiffness of the complete unit cell is achieved in stages by decomposing it.

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

[0007] A method for efficiently predicting the stiffness of a three-dimensional woven composite material based on a neural network includes the following three stages:

[0008] The first stage: predicting the equivalent elastic modulus of the sub-unit cell containing a single warp yarn based on the neural network model;

[0009] The second stage: After obtaining the in-plane elastic modulus of the warp yarn sub-cell based on the neural network model, insert the weft fiber bundle in the transverse direction of the unit cell to obtain the sub-cell, and predict the in-plane elastic modulus of the sub-cell, including the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus;

[0010] The third stage: Based on the Voigt model, the warp elastic modulus and in-plane shear elastic modulus of the unit cell of the three-dimensional woven composite material are predicted, and based on the Reuss model, the weft elastic modulus of the unit cell of the three-dimensional woven composite material is predicted.

[0011] Furthermore, in the first stage, the equivalent elastic modulus of the sub-unit cell plane containing a single warp yarn is predicted, including the following steps:

[0012] Step 1: By equating the path of the sewing warp yarn to a combination of a power elliptic function and a straight line, the sewing warp yarn corrugation ratio w, corrugation area size b, straight segment ratio d, wave segment power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell are screened out based on the comparison and demonstration of numerical results. f There are 7 mesoscopic design parameters that affect the equivalent stiffness of the warp yarn sub-cell;

[0013] First, the seam warp yarn path is geometrically simplified. The path is uniquely determined by four geometric parameters: the seam warp yarn corrugation ratio w, the corrugation area size b, the straight section ratio d, and the wave section power index P. Due to the unit cell length L, unit cell width, unit cell height H, and fiber bundle volume fraction V in the unit cell, ... f , fiber bundle cross-sectional height and cross-sectional width will affect the equivalent stiffness of the warp yarn sub-unit cell. Therefore, the PCA method was used to analyze the influence of the above 10 microscopic parameters on the equivalent stiffness, thereby screening out the warp yarn corrugation ratio w, corrugation area size b, straight section ratio d, wave section power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell. f These 7 mesoscopic design parameters that have a significant impact on stiffness are used as the 7 input dimensions of the neural network;

[0014] Step 2: A modified Chamis model based on neural network, namely NN-Chamis model, is proposed for the longitudinal elastic modulus, transverse elastic modulus and in-plane shear elastic modulus of the sub-cell of the seam warp yarn. This model introduces a fluctuation correction factor before the Chamis formula, takes the 7 mesoscopic design parameters screened out in step 1 as the input of the neural network model, and uses the fluctuation correction factor converted from the finite element results as the output of the neural network. With the help of several sample data obtained by mesoscopic high-fidelity finite element modeling, an agent model is established, and a quantitative relationship between the mesoscopic design parameters and the fluctuation correction factor is established to realize the prediction of the in-plane effective stiffness of the seam warp yarn sub-cell.

[0015] Furthermore, in step 1, the expression of the power elliptic function is:

[0016]

[0017] Where: x t and t They respectively represent the abscissa and ordinate values ​​of the power ellipse function in the Cartesian coordinate system, a and A respectively represent the semi-major axis and semi-minor axis of the power ellipse, P represents the power exponent of the power ellipse, t represents the independent variable of the power ellipse function, and the abscissa and ordinate values ​​are functions of the independent variable t; p represents the power exponent of the power ellipse.

[0018] Furthermore, in step 2, a modified Chamis model based on a neural network is proposed, as shown in formula (4):

[0019]

[0020] Where: f 1~3 Both represent the fluctuation correction factors for the Chamis model results, which are obtained by training the neural network model; E x 、E Y With G XY They represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of the warp yarn sub-cell respectively; V f is the volume fraction of fiber bundles in the warp yarn sub-cell; E X,Chamis 、E Y,Chamis and G XY,Chamis They respectively represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of a single cell of the warp yarn calculated using the Chamis formula without considering the fiber bundle fluctuation.

[0021] Furthermore, in the second stage, the in-plane elastic modulus of the sub-unit cell is predicted to be:

[0022] The modified mixed-law model is used to predict the warp and weft elastic moduli of the sub-unit cell, and the in-plane shear elastic modulus of the sub-unit cell is predicted based on the classical laminate theory.

[0023] Furthermore, the modified mixed law model is used to predict the warp elastic modulus and the weft elastic modulus of the sub-unit cell; the prediction process is as follows:

[0024] First, establish the series model in the mixed law algorithm, as shown in formula (5):

[0025]

[0026] Where: represents the longitudinal elastic modulus of the nth sub-unit cell, represents the transverse elastic modulus of the weft yarn, Indicates the longitudinal elastic modulus of the sewing warp yarn, E M Represents the elastic modulus of the resin matrix, C weft represents the volume fraction of the weft yarn in the sub-unit cell, C Binder represents the volume fraction of the sewing warp yarn in the sub-unit cell, C M represents the volume fraction of the resin matrix in the subunit cell; due to C weft +C warp +C M =1, formula (5) is rewritten as formula (6):

[0027]

[0028] When the warp yarn fluctuates, equation (6) needs to be improved. According to the idea of ​​the series model in classical micromechanics, the terms on the right side of equation (6) are regarded as the stiffness contributions of the warp yarn, weft yarn and resin matrix to the warp elastic modulus of the sub-unit cell. Then, when the warp yarn fluctuates, the stiffness contribution term of the warp yarn is Need to be updated; at the same time, the longitudinal elastic modulus of the warp yarn unit cell It includes the stiffness contribution of the resin matrix and the stiffness contribution of the warp yarn; the stiffness contribution of the resin matrix is ​​E M (1-C Binder ), thus the contribution of the warp yarn to the elastic modulus of the sub-unit cell in the warp direction is obtained According to the stiffness contribution term of the warp yarn without fluctuation phenomenon in the updated formula (6), the calculation formula of the warp elastic modulus of the sub-unit cell is obtained as follows:

[0029]

[0030] Similarly, taking the latitudinal elastic modulus of the sub-unit cell as the research object, we obtain formula (8):

[0031]

[0032] Where: represents the latitudinal elastic modulus of the nth sub-unit cell, It represents the elastic modulus in the weft direction of the warp yarn unit cell. It represents the longitudinal elastic modulus of the weft yarn.

[0033] Furthermore, the prediction process of the in-plane shear elastic modulus of the sub-unit cell based on the classical laminate theory is as follows:

[0034] The prediction of the in-plane shear elastic modulus of the sub-unit cell is based on the classical laminate theory. First, the sub-unit cell is divided into two layers, the warp yarns and weft yarns are separated into these two layers, and a laminate structure with two single layers is formed. The material properties of the upper and lower layers are equivalent to each other according to the classical laminate theory.

[0035] Under the action of in-plane load, the constitutive relation of the laminate is shown in formula (9):

[0036]

[0037] Where: N X 、N Y and N XY They represent the tension, pressure and shear force per unit width on the cross section of the laminate, and They represent the mid-surface strain of the laminate; k X , k Y and k XY are the mid-surface bending deflection and mid-surface twisting rate of the laminate, A ij and B ij denote the tensile and coupling stiffness matrices of the laminate, respectively, and A 11 , A 12 and A 22 Both represent the stiffness coefficient between tension and mid-surface tensile strain; A 66 A represents the stiffness coefficient between shear force and mid-surface shear strain; 16 , A 26 Both represent the coupling stiffness coefficient between shear and tension; B 11 , B 12 and B 22 Both represent the coupling stiffness coefficient between tension and bending; B 66 Represents the coupling stiffness coefficient between shear and torsion; B 16 , B 26 Represents the coupling stiffness coefficient between tension and torsion or shear and bending;

[0038] The third equation of equation (9) determines the calculation of the in-plane shear elastic modulus of the laminate, and it is expanded to obtain equation (10):

[0039]

[0040] Where: A 66The stiffness coefficient between the shear force and the mid-plane shear strain, the in-plane shear elastic modulus G of the laminate XY The calculation formula is:

[0041] G XY =A 66 / h0 (11)

[0042] Where: h0 represents the thickness of the laminate, for the sub-unit cell, A 66 The calculation formula is shown in formula (12):

[0043]

[0044] Where: h1 and h2 represent the thickness of the single layer of weft yarn and the single layer of warp yarn respectively, z k and z k-1 They represent the z-direction coordinates of the upper and lower surfaces of the kth single-layer plate in the laminate in the local coordinate system, k represents the number of the single-layer plate in the laminate, and m represents the total number of single-layer plates in the laminate. It represents the stiffness coefficient of the single-layer plate in the general coordinate system. Its relationship with the stiffness coefficient in the main direction of the material is shown in formula (13):

[0045]

[0046] Where: Q ij represents the stiffness coefficient of the single-layer plate in the material coordinate system, θ represents the angle between the main direction of the material and the X-axis of the global coordinate system; since the warp yarn extends along the X-axis of the global coordinate system, and the weft yarn extends along the Y-axis of the global coordinate system; the single-layer deflection angle of the weft yarn θ1 = 0°, and the single-layer deflection angle of the warp yarn θ2 = 90°, substituting into formula (13) to obtain formula (14):

[0047]

[0048] Where: (G 12 )1 and (G 12 )2 represent the in-plane shear elastic modulus of the weft yarn single layer and the sewing warp yarn single layer respectively; Substitute equation (14) into equation (12), simplify it and then substitute it into equation (11), and the prediction formula (15) of the in-plane shear elastic modulus of the sub-unit cell is obtained:

[0049]

[0050] Where: represents the in-plane shear elastic modulus of the sub-unit cell, C1 and C2 represent the volume fractions of the weft yarn single layer and the sewn warp yarn single layer in the equivalent laminate, respectively.

[0051] Compared with the prior art, the invention has the following beneficial effects: the neural network-based three-dimensional woven composite material stiffness efficient prediction method of the invention, combined with the neural network model and the micromechanical analytical model, can realize the in-plane equivalent stiffness prediction of the three-dimensional woven composite material unit cell in stages. The prediction accuracy is high, and the absolute value of the error with the finite element calculation result is controlled within 8%; the calculation efficiency is high, which greatly saves calculation resources and time. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is the technical roadmap for the neural network-based efficient stiffness prediction method of three-dimensional woven composite materials of the present invention.

[0053] Figure 1 In, x i represents seven micro-design parameters, yi represents the fluctuation correction factor for the warp elastic modulus, weft elastic modulus and in-plane shear modulus, Weft and Warp represent the weft and warp directions, respectively.

[0054] Figure 2 Schematic diagram of the simplified and reconstructed warp yarn geometric path.

[0055] Figure 3 Flowchart of the high-fidelity modeling method for mesoscopic unit cells of three-dimensional woven composites.

[0056] Figure 4 Flowchart for predicting the equivalent elastic modulus of warp yarn sub-cell in the first stage.

[0057] Figure 5 Diagram of the mesomechanical model for predicting the in-plane stiffness of the complete three-dimensional woven composite material for the third stage.

[0058] Figure 5 In the figure: ε1 represents the strain in one direction of the unit cell, which corresponds to the uniform strain boundary condition of the unit cell in the Voigt model; σ2 represents the stress in two directions of the unit cell, which corresponds to the uniform stress boundary condition of the unit cell in the Reuss model.

[0059] Figure 6 The following are the validation diagrams of the stiffness prediction examples of four three-dimensional woven composite structures: Figure 6 (a) is the stiffness prediction example verification diagram of the three-dimensional woven composite structure A. Figure 6 (b) is the stiffness prediction example verification diagram of the three-dimensional woven composite structure B. Figure 6 (c) is the stiffness prediction example verification diagram of the three-dimensional woven composite structure C. Figure 6 (d) is the stiffness prediction example verification diagram of the three-dimensional woven composite structure D.

[0060] The technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. DETAILED DESCRIPTION

[0061] Specific implementation method 1: Figure 1-Figure 5 As shown, this embodiment describes a method for efficiently predicting the stiffness of a three-dimensional woven composite material based on a neural network, and the method includes the following three stages:

[0062] The first stage: predicting the equivalent elastic modulus of the sub-unit cell containing a single warp yarn based on the neural network model;

[0063] The second stage: After obtaining the in-plane elastic modulus of the warp yarn sub-cell based on the neural network model, insert the weft fiber bundle in the transverse direction of the unit cell to obtain the sub-cell, and predict the in-plane elastic modulus of the sub-cell, including the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus;

[0064] The third stage: Based on the Voigt model, the warp elastic modulus and in-plane shear elastic modulus of the unit cell of the three-dimensional woven composite material are predicted, and based on the Reuss model, the weft elastic modulus of the unit cell of the three-dimensional woven composite material is predicted.

[0065] Furthermore, in the first stage, the prediction of the equivalent elastic modulus of the sub-unit cell plane containing a single warp yarn comprises the following steps:

[0066] Step 1: By equating the path of the sewing warp yarn to a combination of a power elliptic function and a straight line, the sewing warp yarn corrugation ratio w, corrugation area size b, straight segment ratio d, wave segment power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell are screened out based on the comparison and demonstration of numerical results. f There are 7 mesoscopic design parameters that affect the equivalent stiffness of the warp yarn sub-cell;

[0067] First, the warp yarn path is geometrically simplified (e.g. Figure 2 As shown in Figure 2, the path is uniquely determined by the four geometric parameters: the corrugation ratio w of the sewing warp yarn, the size of the corrugated area b, the proportion of the straight section d, and the power index P of the fluctuating section; due to the unit cell length L, unit cell width, unit cell height H, and the volume fraction of the fiber bundle in the unit cell V f, fiber bundle cross-sectional height and cross-sectional width will affect the equivalent stiffness of the warp yarn sub-unit cell. Therefore, the PCA method was used to analyze the influence of the above 10 microscopic parameters on the equivalent stiffness, thereby screening out the warp yarn corrugation ratio w, corrugation area size b, straight section ratio d, wave section power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell. f These 7 mesoscopic design parameters that have a significant impact on stiffness are used as the 7 input dimensions of the neural network;

[0068] Step 2: A modified Chamis model based on neural network, namely NN-Chamis model, is proposed for the longitudinal elastic modulus, transverse elastic modulus and in-plane shear elastic modulus of the sub-cell of the seam warp yarn. The model introduces a fluctuation correction factor before the Chamis formula, takes the 7 mesoscopic design parameters screened in step 1 as the input of the neural network model (NN model), and the fluctuation correction factor converted from the finite element results as the output of the neural network. With the help of several sample data obtained by mesoscopic high-fidelity finite element modeling (which is the existing technology), an agent model is established, and a quantitative relationship between the mesoscopic design parameters and the fluctuation correction factor is established to realize the prediction of the in-plane effective stiffness of the seam warp yarn sub-cell.

[0069] Furthermore, in step 1, the expression of the power elliptic function is:

[0070]

[0071] Where: x t and t They respectively represent the abscissa and ordinate values ​​of the power ellipse function in the Cartesian coordinate system, a and A respectively represent the semi-major axis and semi-minor axis of the power ellipse, P represents the power exponent of the power ellipse, t represents the independent variable of the power ellipse function, and the abscissa and ordinate values ​​are functions of the independent variable t; p represents the power exponent of the power ellipse.

[0072] Furthermore, in step 2, a modified Chamis model based on a neural network is proposed, as shown in formula (4):

[0073]

[0074] Where: f 1~3 Both represent the fluctuation correction factors for the Chamis model results, which are obtained by training the neural network model; E x 、E Y With G XY They represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of the warp yarn sub-cell respectively; V f is the volume fraction of fiber bundles in the warp yarn sub-cell; E X,Chamis 、E Y,Chamis and GXY,Chamis They respectively represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of a single cell of the warp yarn calculated using the Chamis formula without considering the fiber bundle fluctuation.

[0075] Furthermore, in the second stage, the in-plane elastic modulus of the sub-unit cell is predicted to be:

[0076] The modified mixed-law model is used to predict the warp and weft elastic moduli of the sub-unit cell, and the in-plane shear elastic modulus of the sub-unit cell is predicted based on the classical laminate theory.

[0077] Furthermore, the modified mixed law model is used to predict the warp elastic modulus and the weft elastic modulus of the sub-unit cell; the prediction process is as follows:

[0078] First, establish the series model in the mixed law algorithm, as shown in formula (5):

[0079]

[0080] Where: represents the longitudinal elastic modulus of the nth sub-unit cell, represents the transverse elastic modulus of the weft yarn, Indicates the longitudinal elastic modulus of the sewing warp yarn, E M Represents the elastic modulus of the resin matrix, C weft represents the volume fraction of the weft yarn in the sub-unit cell, C Binder represents the volume fraction of the sewing warp yarn in the sub-unit cell, C M represents the volume fraction of the resin matrix in the subunit cell; due to C weft +C warp +C M =1, formula (5) is rewritten as formula (6):

[0081]

[0082] When the warp yarn fluctuates, equation (6) needs to be improved. According to the idea of ​​the series model in classical micromechanics, the terms on the right side of equation (6) are regarded as the stiffness contributions of the warp yarn, weft yarn and resin matrix to the warp elastic modulus of the sub-unit cell. Then, when the warp yarn fluctuates, the stiffness contribution term of the warp yarn is Need to be updated; at the same time, the longitudinal elastic modulus of the warp yarn unit cell It includes the stiffness contribution of the resin matrix and the stiffness contribution of the warp yarn; the stiffness contribution of the resin matrix is ​​E M (1-C Binder ), thus the contribution of the warp yarn to the elastic modulus of the sub-unit cell in the warp direction is obtained According to the stiffness contribution term of the warp yarn without fluctuation phenomenon in the updated formula (6), the calculation formula of the warp elastic modulus of the sub-unit cell is obtained as follows:

[0083]

[0084] Similarly, taking the latitudinal elastic modulus of the sub-unit cell as the research object, we obtain formula (8):

[0085]

[0086] Where: represents the latitudinal elastic modulus of the nth sub-unit cell, It represents the elastic modulus in the weft direction of the warp yarn unit cell. It represents the longitudinal elastic modulus of the weft yarn.

[0087] Furthermore, the prediction process of the in-plane shear elastic modulus of the sub-unit cell based on the classical laminate theory is as follows:

[0088] The prediction of the in-plane shear elastic modulus of the sub-unit cell is based on the classical laminate theory. First, the sub-unit cell is divided into two layers, the warp yarns and weft yarns are separated into these two layers, and a laminate structure with two single layers is formed. The material properties of the upper and lower layers are equivalent to each other according to the classical laminate theory.

[0089] Under the action of in-plane load, the constitutive relation of the laminate is shown in formula (9):

[0090]

[0091] Where: N X 、N Y and N XY They represent the tension, pressure and shear force per unit width on the cross section of the laminate, and They represent the mid-surface strain of the laminate; k X , k Y and k XY are the mid-surface bending deflection and mid-surface twisting rate of the laminate, A ij and B ij denote the tensile and coupling stiffness matrices of the laminate, respectively, and A 11 , A 12 and A 22 Both represent the stiffness coefficient between tension and mid-surface tensile strain; A 66 A represents the stiffness coefficient between shear force and mid-surface shear strain; 16 , A 26 Both represent the coupling stiffness coefficient between shear and tension; B 11 , B 12 and B 22Both represent the coupling stiffness coefficient between tension and bending; B 66 Represents the coupling stiffness coefficient between shear and torsion; B 16 , B 26 Represents the coupling stiffness coefficient between tension and torsion or shear and bending;

[0092] The third equation of equation (9) determines the calculation of the in-plane shear elastic modulus of the laminate, and it is expanded to obtain equation (10):

[0093]

[0094] Where: A 66 The stiffness coefficient between the shear force and the mid-plane shear strain, the in-plane shear elastic modulus G of the laminate XY The calculation formula is:

[0095] G XY =A 66 / h0 (11)

[0096] Where: h0 represents the thickness of the laminate, for the sub-unit cell, A 66 The calculation formula is shown in formula (12):

[0097]

[0098] Where: h1 and h2 represent the thickness of the single layer of weft yarn and the single layer of warp yarn respectively, z k and z k-1 They represent the z-direction coordinates of the upper and lower surfaces of the kth single-layer plate in the laminate in the local coordinate system (the order of the single-layer plates is related to the definition of the coordinate system, which is defined from the origin of the coordinate system to the z-direction), k represents the number of the single-layer plate in the laminate, and m represents the total number of single-layer plates in the laminate. It represents the stiffness coefficient of the single-layer plate in the general coordinate system. Its relationship with the stiffness coefficient in the main direction of the material is shown in formula (13):

[0099]

[0100] Where: Q ij represents the stiffness coefficient of the single-layer plate in the material coordinate system, θ represents the angle between the main direction of the material and the X-axis of the global coordinate system; since the warp yarn extends along the X-axis of the global coordinate system, and the weft yarn extends along the Y-axis of the global coordinate system; the single-layer deflection angle of the weft yarn θ1 = 0°, and the single-layer deflection angle of the warp yarn θ2 = 90°, substituting into formula (13) to obtain formula (14):

[0101]

[0102] Where: (G 12 )1 and (G 12)2 represent the in-plane shear elastic modulus of the weft yarn single layer and the sewing warp yarn single layer respectively; Substitute equation (14) into equation (12), simplify it and then substitute it into equation (11), and the prediction formula (15) of the in-plane shear elastic modulus of the sub-unit cell is obtained:

[0103]

[0104] Where: represents the in-plane shear elastic modulus of the sub-unit cell, C1 and C2 represent the volume fractions of the weft yarn single layer and the sewn warp yarn single layer in the equivalent laminate, respectively.

[0105] Example:

[0106] This embodiment describes a method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network. The overall technical route is as follows: Figure 1 As shown, the specific steps include:

[0107] The first stage: Combined with the neural network, the equivalent elastic modulus of the sub-unit cell containing only a single warp yarn is predicted; in the early stage of the prediction, the PCA method (principal component analysis) is used to screen out seven microscopic design parameters that have a significant impact on the stiffness for the undulating fiber bundle sub-unit cell, namely: corrugation ratio w, corrugation area size b, straight section ratio d, undulating section power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell. f , thus establishing a database consisting of 7 mesoscopic design parameters after PCA data dimensionality reduction to reduce the cost of collecting samples. Establishing the NN-Chamis model; this model introduces a fluctuation correction factor before the Chamis formula, takes the 7 selected mesoscopic design parameters as the input of the neural network model, and the fluctuation correction factor converted from the finite element results as the output of the neural network. With the help of the sample data obtained by mesoscopic high-fidelity finite element modeling, an agent model is established, and a quantitative relationship between the mesoscopic design parameters and the fluctuation correction factor is established to achieve the prediction of the effective stiffness of the warp yarn sub-cell in the plane.

[0108] By observing the warp yarn path, Figure 2 As shown, its path function can be equivalent to a combined line segment of a power elliptic function and a straight line. The expression of the power elliptic function is:

[0109]

[0110] Where: x t and t They respectively represent the abscissa and ordinate values ​​of the power ellipse function in the Cartesian coordinate system, a and A respectively represent the semi-major axis and semi-minor axis of the power ellipse, P represents the power exponent of the power ellipse, t represents the independent variable of the power ellipse function, and the abscissa and ordinate values ​​are functions of the independent variable t; p represents the power exponent of the power ellipse.

[0111] like Figure 2 As shown in the figure, the rising and falling sections of the path are fitted with two 1 / 4 power ellipse boundaries, and the falling and rising sections are also fitted with two 1 / 4 power ellipse boundaries. The rising and falling sections are controlled by parameters a, A and P. In addition, a fluctuating platform section is formed between the rising and falling sections, and its length is y. These three sections constitute the fluctuation area of ​​the sewing warp yarn, and the length is l. The line segment MJ is a fluctuation period of the sewing warp yarn, and this parameter model is used as a general model to characterize the sewing warp yarn path. In this parameter model, 5 geometric parameters characterizing the fiber bundle path are involved. They are dimensionless and simplified into 4 fiber bundle fluctuation parameters, namely, the sewing warp yarn corrugation ratio w, the corrugation area size b, the straight segment ratio d and the power exponent P of the power ellipse (a path parameter of the sewing warp yarn); its expression is:

[0112]

[0113] Where: L represents the unit cell length;

[0114] According to the mesomechanics theory, the volume fraction of fiber bundles in a single cell contributes significantly to the material stiffness. Therefore, it is analyzed as the main mesoscopic design parameter, and its definition is as follows:

[0115] V f =V 纤维束 / V 子单胞 (3)

[0116] Where: V f represents the volume fraction of fiber bundles in a single cell, V 纤维束 represents the volume of fiber bundles in the subunit cell, V 子单胞 Represents the sub-unit cell volume. Controlling other mesoscopic variables (for example, when studying the effect of unit cell length L on equivalent performance, keeping unit cell height H and width W constant), the influence of unit cell size specifications (L, H, W) on the equivalent elastic performance of the unit cell was studied. It was confirmed that the unit cell length L and unit cell height H are the mesoscopic design variables that significantly affect the equivalent elastic performance of the unit cell.

[0117] The Chamis formula has a good accuracy level in predicting the equivalent elastic properties of unidirectional composite materials, but it is not suitable for the case where the fiber has bending fluctuations. Therefore, the present invention proposes a modified Chamis model that takes into account the fluctuation effect of the fiber bundle, as shown in formula (4). By introducing a fluctuation correction factor before the Chamis formula, the equivalent elastic modulus of the warp yarn sub-cell in the plane is predicted;

[0118]

[0119] Where: f 1~3Both represent the fluctuation correction factors for the Chamis model results, which are obtained by training the neural network model; E x 、E Y With G XY They represent the radial elastic modulus, weft elastic modulus and in-plane shear elastic modulus of the warp yarn sub-cell respectively; V f E represents the volume fraction of fiber bundles in the warp yarn sub-unit cell; X,Chamis 、E Y,Chamis and G XY,Chamis They represent the radial elastic modulus, weft elastic modulus and in-plane shear elastic modulus of the warp yarn unit cell calculated by Chamis formula without considering the fiber bundle fluctuation;

[0120] A database for neural network training is built to establish a quantitative relationship between the mesoscopic design parameters of the warp yarn sub-cell and the fluctuation correction factor. Considering that the input dimension of the network is large, and the values ​​of the corrugation ratio w of the warp yarn and the corrugation area size b in the six-step fabric structure are relatively fixed and coupled to each other, an improved equidistant database is used to collect data (introducing physical constraints to the database construction of the network). Since the prediction object of the present invention is the equivalent stiffness performance of the six-step woven composite material, the constructed database considers the actual fluctuation form of the warp yarn in the six-step fabric structure, and the value range of each mesoscopic design variable will be defined according to the actual six-step fabric structure characteristics, rather than using the traditional method to set a larger value range, so that the amount of data can be simplified, and the network training accuracy will not be reduced to simplify the amount of data. The required sample distribution is shown in Table 1.

[0121] Table 1 Distribution of database samples

[0122]

[0123] The finite element model of the warp yarn sub-cell under the specified micro-design parameters is established. Figure 3As shown, first, a fabric geometry model is established in Texgen, and then a periodic grid is divided in Hypermesh to obtain a high-fidelity unit cell model (a high-fidelity unit cell model is a professional term for finite element modeling in the field of composite materials, indicating that the unit cell is discretized using a finite element grid. The unit cell model in the present invention is discretized using tetrahedral first-order units, and the number of tetrahedral units used in each sample is different, basically between 200,000 and 300,000 units). Finally, in Abaqus, a Python script is secondary developed by the user (the key to defining the direction of the local coordinate system of the fiber bundle is to determine the 1 direction. First, the centroid coordinates of each fiber bundle unit need to be determined, and the tangent direction of the fiber bundle path corresponding to the centroid coordinates is determined according to the built-in library function of Texgen. The spatial vector in the tangent direction is used as the 1 direction of the fiber bundle. Since the fiber bundle is assumed to be a transversely isotropic material in the unit cell model, the spatial vector in the transverse 2 direction can be defined as 0 with the dot product of the vector in the 1 direction, and the 3-direction definition method is consistent with the 2-direction), the local coordinate direction is assigned to the fiber bundle unit, and the establishment of the finite element model of the warp yarn sub-unit cell is completed. The in-plane equivalent elastic modulus database of the unit cell samples was obtained through finite element analysis. A total of 267 groups of samples were formed, 240 of which (90% of the total number of samples were selected as training sets and 10% as test sets of the neural network) were selected as neural network training sets, and 27 groups were selected as neural network test sets to evaluate the training effect of the network; the prediction fluctuation correction factor f was established 1~3 After the neural network proxy model is constructed, the in-plane elastic modulus of the warp yarn sub-cell can be predicted according to formula (4). The technical route of this stage is as follows: Figure 4 shown.

[0124] The second stage: insert a fiber bundle in another direction (weft fiber bundle) in the transverse direction of the unit cell to predict the in-plane elastic modulus of the sub-unit cell. The prediction method of the radial elastic modulus of the sub-unit cell is studied. When the warp yarn does not fluctuate, the series model in the mixed law algorithm can be used for calculation, as shown in formula (5):

[0125]

[0126] Where: represents the longitudinal elastic modulus of the nth sub-unit cell, represents the transverse elastic modulus of the weft yarn, Indicates the longitudinal elastic modulus of the sewing warp yarn, E M Represents the elastic modulus of the resin matrix, C weft represents the volume fraction of the weft yarn in the sub-unit cell, C Binder represents the volume fraction of the sewing warp yarn in the sub-unit cell, C M represents the volume fraction of the resin matrix in the subunit cell. weft +C warp +C M=1, formula (5) can also be written as formula (6):

[0127]

[0128] When the warp yarn fluctuates, equation (6) needs to be improved. According to the idea of ​​the series model in micromechanics, the terms on the right side of equation (6) are regarded as the stiffness contributions of the warp yarn, weft yarn and resin matrix to the warp elastic modulus of the sub-unit cell. Then, when the warp yarn fluctuates, the stiffness contribution term originally representing the warp yarn is Need to be updated. At the same time, the longitudinal elastic modulus of the warp yarn unit cell It includes the stiffness contribution of the resin matrix and the stiffness contribution of the warp yarn. The stiffness contribution of the resin matrix is ​​E M (1-C Binder ), thus the contribution of the warp yarn to the radial elastic modulus of the sub-unit cell is obtained The stiffness contribution term of the warp yarn without fluctuation phenomenon in equation (6) is updated, and the calculation formula of the warp elastic modulus of the sub-unit cell is obtained as follows:

[0129]

[0130] Similarly, taking the latitudinal elastic modulus of the sub-unit cell as the research object, we obtain formula (8):

[0131]

[0132] Where: represents the latitudinal elastic modulus of the nth sub-unit cell, represents the transverse modulus of the warp yarn sub-unit cell, It represents the longitudinal elastic modulus of the weft yarn.

[0133] The prediction of the in-plane shear elastic modulus of the sub-unit cell is based on the classical laminate theory. First, the sub-unit cell is divided into two layers, the warp yarns and weft yarns are separated into these two layers respectively, forming a laminate structure with two single layers. According to the classical laminate theory, the material properties of the upper and lower layers are equivalent together.

[0134] Under the action of in-plane load, the constitutive relation of the laminate is shown in formula (9):

[0135]

[0136] Where: N X 、N Y and N XY They represent the tension, pressure and shear force per unit width on the cross section of the laminate, and They represent the mid-surface strain of the laminate; k X , k Y and kXY are the mid-surface bending deflection and mid-surface twisting rate of the laminate, A ij and B ij denote the tensile and coupling stiffness matrices of the laminate, respectively, and A 11 , A 12 and A 22 Both represent the stiffness coefficient between tension and mid-surface tensile strain; A 66 A represents the stiffness coefficient between shear force and mid-surface shear strain; 16 , A 26 Both represent the coupling stiffness coefficient between shear and tension; B 11 , B 12 and B 22 Both represent the coupling stiffness coefficient between tension and bending; B 66 Represents the coupling stiffness coefficient between shear and torsion; B 16 , B 26 Represents the coupling stiffness coefficient between tension and torsion or shear and bending;

[0137] The third equation of equation (9) determines the calculation of the in-plane shear elastic modulus of the laminate, and it is expanded to obtain equation (10):

[0138]

[0139] Where: A 66 The stiffness coefficient between the shear force and the mid-plane shear strain, the in-plane shear elastic modulus G of the laminate XY The calculation formula is:

[0140] G XY =A 66 / h0 (11)

[0141] Where: h0 represents the thickness of the laminate, for the sub-unit cell, A 66 The calculation formula is shown in formula (12):

[0142]

[0143] Where: h1 and h2 represent the thickness of the single layer of weft yarn and the single layer of warp yarn respectively; z k and z k-1 They respectively represent the z-direction coordinates of the upper and lower surfaces of the k-th single-layer plate in the laminate in the local coordinate system (the order of the single-layer plates is related to the definition of the coordinate system, which is defined from the origin of the coordinate system to the z-direction); represents the stiffness coefficient of the single-layer plate in the general coordinate system, and its relationship with the stiffness coefficient in the main direction of the material is shown in formula (13): k represents the number of the single-layer plate in the laminate; n is the total number of single-layer plates in the laminate;

[0144]

[0145] Where: Q ij represents the stiffness coefficient of the single-layer plate in the material coordinate system, and θ represents the angle between the main direction of the material and the X-axis of the global coordinate system. Since the warp yarns extend along the X-axis of the global coordinate system, and the weft yarns extend along the Y-axis of the global coordinate system. The deflection angle of the weft yarn single layer θ1 = 0°, and the deflection angle of the warp yarn single layer θ2 = 90°, substituting into formula (13) to obtain formula (14):

[0146]

[0147] Where: (G 12 )1 and (G 12 )2 represent the in-plane shear elastic modulus of the weft yarn single layer and the sewing warp yarn single layer respectively. Substituting formula (14) into formula (12), simplifying it and then substituting it into formula (11), we get the prediction formula (15) of the in-plane shear elastic modulus of the sub-unit cell:

[0148]

[0149] Where: C1 and C2 represent the volume fractions of the weft yarn single layer and the sewn warp yarn single layer in the equivalent laminate, respectively.

[0150] The third stage: According to the equal strain model (Voigt model), Figure 5 As shown in Figure 1, the radial elastic modulus and in-plane shear elastic modulus of the unit cell of the three-dimensional woven composite material are calculated as follows:

[0151]

[0152]

[0153] Where: and They represent the radial elastic modulus and in-plane shear elastic modulus of the unit cell of the three-dimensional woven composite material respectively; C Cn Represents the volume fraction of the nth sub-unit cell in the complete unit cell.

[0154] According to the isostress model (Reuss model), Figure 6 As shown, the latitudinal elastic modulus of the unit cell is:

[0155]

[0156] Where: represents the latitudinal elastic modulus of the unit cell.

[0157] In order to verify the accuracy of the proposed method for efficiently predicting the stiffness of three-dimensional woven composite materials based on neural networks, four different three-dimensional woven composite material structures were selected, such as Figure 6As shown in Table 2, the prediction results of this method are compared with the finite element numerical calculation results. X , latitudinal elastic modulus E Y And the in-plane shear elastic modulus G XY The absolute value of the error between the efficient prediction solution of the present invention and the finite element result is within 8%, which shows that the method has good accuracy. X Prediction, E X The radial elastic modulus of the three-dimensional woven composite material and the prediction time required by the finite element method and the efficient prediction method are shown in Table 3. It can be seen that the method proposed in the present invention greatly saves computing resources while ensuring the calculation accuracy.

[0158] Table 2 Validation of three-dimensional woven composite material stiffness prediction examples

[0159]

[0160] Table 3 Comparison of prediction time of two prediction methods

[0161]

[0162] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other forms of assembly without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations within the meaning and range of equivalents of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0163] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network, characterized in that: The method comprises the following three stages: The first stage: predicting the equivalent elastic modulus of the sub-unit cell containing a single warp yarn based on the neural network model; The second stage: After obtaining the in-plane elastic modulus of the warp yarn sub-cell based on the neural network model, insert the weft fiber bundle in the transverse direction of the unit cell to obtain the sub-cell, and predict the in-plane elastic modulus of the sub-cell, including the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus; The third stage: Based on the Voigt model, the warp elastic modulus and in-plane shear elastic modulus of the unit cell of the three-dimensional woven composite material are predicted, and based on the Reuss model, the weft elastic modulus of the unit cell of the three-dimensional woven composite material is predicted.

2. The method for efficiently predicting the stiffness of a three-dimensional woven composite material based on a neural network according to claim 1, characterized in that: In the first stage, the in-plane equivalent elastic modulus of the sub-unit cell containing a single sewing warp yarn is predicted, which includes the following steps: Step 1: By equating the path of the sewing warp yarn to a combination of a power elliptic function and a straight line, the sewing warp yarn corrugation ratio w, corrugation area size b, straight segment ratio d, wave segment power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell are screened out based on the comparison and demonstration of numerical results. f There are 7 mesoscopic design parameters that affect the equivalent stiffness of the warp yarn sub-cell; First, the seam warp yarn path is geometrically simplified. The path is uniquely determined by four geometric parameters: the seam warp yarn corrugation ratio w, the corrugation area size b, the straight section ratio d, and the wave section power index P. Due to the unit cell length L, unit cell width, unit cell height H, and fiber bundle volume fraction V in the unit cell, ... f , fiber bundle cross-sectional height and cross-sectional width will affect the equivalent stiffness of the warp yarn sub-unit cell. Therefore, the PCA method was used to analyze the influence of the above 10 microscopic parameters on the equivalent stiffness, thereby screening out the warp yarn corrugation ratio w, corrugation area size b, straight section ratio d, wave section power index P, unit cell length L, unit cell height H and fiber bundle volume fraction V in the unit cell. f These 7 mesoscopic design parameters that have a significant impact on stiffness are used as the 7 input dimensions of the neural network; Step 2: A modified Chamis model based on neural network, namely NN-Chamis model, is proposed for the longitudinal elastic modulus, transverse elastic modulus and in-plane shear elastic modulus of the sub-cell of the seam warp yarn. This model introduces a fluctuation correction factor before the Chamis formula, takes the 7 mesoscopic design parameters screened out in step 1 as the input of the neural network model, and uses the fluctuation correction factor converted from the finite element results as the output of the neural network. With the help of several sample data obtained by mesoscopic high-fidelity finite element modeling, an agent model is established, and a quantitative relationship between the mesoscopic design parameters and the fluctuation correction factor is established to realize the prediction of the in-plane effective stiffness of the seam warp yarn sub-cell.

3. The method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network according to claim 2, characterized in that: In step 1, the expression of the power elliptic function is: Where: x t and t They respectively represent the abscissa and ordinate values ​​of the power ellipse function in the Cartesian coordinate system, a and A respectively represent the semi-major axis and semi-minor axis of the power ellipse, P represents the power exponent of the power ellipse, t represents the independent variable of the power ellipse function, and the abscissa and ordinate values ​​are functions of the independent variable t; p represents the power exponent of the power ellipse.

4. The method for efficiently predicting the stiffness of a three-dimensional woven composite material based on a neural network according to claim 1, characterized in that: In step 2, a modified Chamis model based on a neural network is proposed, as shown in formula (4): Where: f 1~3 Both represent the fluctuation correction factors for the Chamis model results, which are obtained by training the neural network model; E x 、E Y With G XY They represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of the warp yarn sub-cell respectively; V f is the volume fraction of fiber bundles in the warp yarn sub-cell; E X,Chamis 、E Y,Chamis and G XY,Chamis They respectively represent the warp elastic modulus, weft elastic modulus and in-plane shear elastic modulus of a single cell of the warp yarn calculated using the Chamis formula without considering the fiber bundle fluctuation.

5. The method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network according to claim 1, characterized in that: In the second stage, the in-plane elastic modulus of the sub-unit cell is predicted to be: The modified mixed-law model is used to predict the warp and weft elastic moduli of the sub-unit cell, and the in-plane shear elastic modulus of the sub-unit cell is predicted based on the classical laminate theory.

6. The method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network according to claim 5, characterized in that: The modified mixed law model is used to predict the warp elastic modulus and the weft elastic modulus of the sub-unit cell; the prediction process is as follows: First, establish the series model in the mixed law algorithm, as shown in formula (5): Where: represents the longitudinal elastic modulus of the nth sub-unit cell, represents the transverse elastic modulus of the weft yarn, Indicates the longitudinal elastic modulus of the sewing warp yarn, E M Represents the elastic modulus of the resin matrix, C weft represents the volume fraction of the weft yarn in the sub-unit cell, C Binder represents the volume fraction of the sewing warp yarn in the sub-unit cell, C M represents the volume fraction of the resin matrix in the subunit cell; due to C weft +C warp +C M =1, formula (5) is rewritten as formula (6): When the warp yarn fluctuates, equation (6) needs to be improved. According to the idea of ​​the series model in classical micromechanics, the terms on the right side of equation (6) are regarded as the stiffness contributions of the warp yarn, weft yarn and resin matrix to the warp elastic modulus of the sub-unit cell. Then, when the warp yarn fluctuates, the stiffness contribution term of the warp yarn is Need to be updated; at the same time, the longitudinal elastic modulus of the warp yarn unit cell It includes the stiffness contribution of the resin matrix and the stiffness contribution of the warp yarn; the stiffness contribution of the resin matrix is ​​E M (1-C Binder ), thus the contribution of the warp yarn to the elastic modulus of the sub-unit cell in the warp direction is obtained According to the stiffness contribution term of the warp yarn without fluctuation phenomenon in the updated formula (6), the calculation formula of the warp elastic modulus of the sub-unit cell is obtained as follows: Similarly, taking the latitudinal elastic modulus of the sub-unit cell as the research object, we obtain formula (8): Where: represents the latitudinal elastic modulus of the nth sub-unit cell, It represents the elastic modulus in the weft direction of the warp yarn unit cell. It represents the longitudinal elastic modulus of the weft yarn.

7. The method for efficiently predicting the stiffness of three-dimensional woven composite materials based on a neural network according to claim 5, characterized in that: The prediction process of predicting the in-plane shear elastic modulus of the sub-unit cell based on the classical laminate theory is as follows: The prediction of the in-plane shear elastic modulus of the sub-unit cell is based on the classical laminate theory. First, the sub-unit cell is divided into two layers, the warp yarns and weft yarns are separated into these two layers, and a laminate structure with two single layers is formed. The material properties of the upper and lower layers are equivalent to each other according to the classical laminate theory. Under the action of in-plane load, the constitutive relation of the laminate is shown in formula (9): Where: N X 、N Y and N XY They represent the tension, pressure and shear force per unit width on the cross section of the laminate, and They represent the mid-surface strain of the laminate; k X , k Y and k XY are the mid-surface bending deflection and mid-surface twisting rate of the laminate, A ij and B ij denote the tensile and coupling stiffness matrices of the laminate, respectively, and A 11 , A 12 and A 22 Both represent the stiffness coefficient between tension and mid-surface tensile strain; A 66 A represents the stiffness coefficient between shear force and mid-surface shear strain; 16 , A 26 Both represent the coupling stiffness coefficient between shear and tension; B 11 , B 12 and B 22 Both represent the coupling stiffness coefficient between tension and bending; B 66 Represents the coupling stiffness coefficient between shear and torsion; B 16 , B 26 Represents the coupling stiffness coefficient between tension and torsion or shear and bending; The third equation of equation (9) determines the calculation of the in-plane shear elastic modulus of the laminate, and it is expanded to obtain equation (10): Where: A 66 The stiffness coefficient between the shear force and the mid-plane shear strain, the in-plane shear elastic modulus G of the laminate XY The calculation formula is: G XY =A 66 / h0 (11) Where: h0 represents the thickness of the laminate, for the sub-unit cell, A 66 The calculation formula is shown in formula (12): Where: h1 and h2 represent the thickness of the single layer of weft yarn and the single layer of warp yarn respectively, z k and z k-1 They represent the z-direction coordinates of the upper and lower surfaces of the kth single-layer plate in the laminate in the local coordinate system, k represents the number of the single-layer plate in the laminate, and m represents the total number of single-layer plates in the laminate. It represents the stiffness coefficient of the single-layer plate in the general coordinate system. Its relationship with the stiffness coefficient in the main direction of the material is shown in formula (13): Where: Q ij represents the stiffness coefficient of the single-layer plate in the material coordinate system, θ represents the angle between the main direction of the material and the X-axis of the global coordinate system; since the warp yarn extends along the X-axis of the global coordinate system, and the weft yarn extends along the Y-axis of the global coordinate system; the single-layer deflection angle of the weft yarn θ1 = 0°, and the single-layer deflection angle of the warp yarn θ2 = 90°, substituting into formula (13) to obtain formula (14): Where: (G 12 )1 and (G 12 )2 represent the in-plane shear elastic modulus of the weft yarn single layer and the sewing warp yarn single layer respectively; Substitute equation (14) into equation (12), simplify it and then substitute it into equation (11), and the prediction formula (15) of the in-plane shear elastic modulus of the sub-unit cell is obtained: Where: represents the in-plane shear elastic modulus of the sub-unit cell, C1 and C2 represent the volume fractions of the weft yarn single layer and the sewn warp yarn single layer in the equivalent laminate, respectively.

Citation Information

Cited By

  • Three-dimensional woven composite material reverse design method based on generative adversarial neural network

    CN120783913A

  • Three-dimensional woven composite reverse design method based on generative adversarial neural network

    CN120783913B