Multi-fidelity composite material strength checking method, equipment, medium and product

Abstract

The invention discloses a multi-fidelity composite material strength checking method and device, a medium and a product, and relates to the technical field of composites.The method comprises the steps that a finite element model is used for constructing a three-dimensional unit cell model of an aviation composite material, and then a unit failure judgment data set is constructed; using the unit failure judgment data set to train the first full-connection neural network to obtain a low-fidelity neural network, performing a structural strength test on the aviation composite material to obtain test data, and using the test data and the low-fidelity neural network to train the second full-connection neural network to obtain a high-fidelity neural network; according to the generalized internal force state of the to-be-predicted working condition, the low-fidelity neural network and the high-fidelity neural network are utilized to complete strength checking of the aviation composite material under the to-be-predicted working condition, the multi-fidelity strength model of the aviation composite material is constructed based on the test data and the simulation data, and the strength checking efficiency and precision of the aviation composite material are improved.

Description

Technical Field

[0001] The present application relates to the technical field of composite materials, and in particular to a multi-fidelity composite material strength verification method, equipment, medium and product. Background Art

[0002] Composite structures are widely used in advanced aircraft structures. Such structures usually have complex internal topological configurations or include materials with complex internal topological configurations. Therefore, it is often impossible to perform full-scale detailed model simulation in structural analysis, but shell elements are used for modeling. However, due to the lack of strength criteria for composite shell models, the Cauchy stress in each layer can only be calculated based on the cross-sectional force of the shell, and then the stress-based strength model is applied for strength verification. There is a scale difference between the structural response based on shell elements and the material-level failure criteria used for strength assessment. It is of great significance to establish a reliable strength model that can be directly used for shell element structural failure prediction. Due to the different internal structures of composite materials, their strength characteristics are difficult to describe based on a unified mathematical expression. In the past few decades, data-driven methods have gradually developed as another method for building strength models for complex material systems, and have been applied to develop strength models for various composite materials and structures. With sufficient training data, data-driven strength models can describe complex strength behaviors and provide very accurate strength predictions for composite structures.

[0003] However, how to collect a sufficient amount of data has always been a key issue in developing data-driven models. Experimental test data is reliable, but the experiment is costly and time-consuming. On the other hand, compared with experiments, a large amount of data can be obtained through numerical simulation, and the time and cost are greatly reduced, but for complex structures or load conditions, the accuracy of numerical simulation results is difficult to guarantee. It is of great significance to take advantage of multi-source data. Summary of the invention

[0004] The purpose of this application is to provide a multi-fidelity composite material strength verification method, equipment, medium and product, to construct a multi-fidelity strength model of aviation composite materials based on test data and simulation data, and to improve the efficiency and accuracy of aviation composite material strength verification.

[0005] To achieve the above objectives, the present application provides the following solutions: In a first aspect, the present application provides a multi-fidelity composite material strength verification method, including: using a finite element model to construct a three-dimensional unit cell model of an aviation composite material.

[0006] Based on the three-dimensional unit cell model, multiple sampling simulations are performed to construct a unit failure judgment data set. The unit failure judgment data set includes the strength margin of the aviation composite material under different simulation conditions. The generalized internal force state of the aviation composite material under different simulation conditions is different.

[0007] Construct a first fully connected neural network and a second fully connected neural network.

[0008] The first fully connected neural network is trained using the unit failure judgment data set to obtain a low-fidelity neural network.

[0009] A structural strength test is performed on the aviation composite material to obtain test data. The test data includes the strength margin of the aviation composite material under different test conditions. The generalized internal force state of the aviation composite material under different test conditions is different.

[0010] The second fully connected neural network is trained using the test data and the low-fidelity neural network to obtain a high-fidelity neural network.

[0011] Obtain the generalized internal force state of the working condition to be predicted.

[0012] According to the generalized internal force state of the working condition to be predicted, the low-fidelity neural network and the high-fidelity neural network are used to complete the strength verification of the aviation composite material under the working condition to be predicted.

[0013] Optionally, the first fully connected neural network is trained using the unit failure judgment data set to obtain a low-fidelity neural network, including: taking the generalized internal force state of aviation composite materials under different simulation conditions as input, and taking the strength margin of aviation composite materials under corresponding simulation conditions as output, to train the first fully connected neural network to obtain a low-fidelity neural network.

[0014] Optionally, the first fully connected neural network or the second fully connected neural network is: . .

[0015] ; in, represents the first Tier neurons; Represents the generalized internal force state of aviation composite materials under simulation conditions; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron for Tier The weights of the neurons; It means that after nonlinear activation Tier neurons; It means that after nonlinear activation Tier neurons; represents a nonlinear activation function; represents the output of the first fully connected neural network; represents the first Tier neurons; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron weights; represents the first A neuron.

[0016] Optionally, the second fully connected neural network is trained using the test data and the low-fidelity neural network to obtain a high-fidelity neural network, including: expanding the test data to obtain an extended strength margin of the aviation composite material under different extended working conditions. The generalized internal force state of the aviation composite material under different extended working conditions is different.

[0017] The generalized internal force state of aviation composite materials under different extended working conditions is input into the low-fidelity neural network to obtain the low-fidelity strength margin.

[0018] The generalized internal force state and low-fidelity strength margin of aerospace composite materials under the same extended working condition are taken as input data pairs.

[0019] The second fully connected neural network is trained with the input data pair as input and the extended strength margin of the aviation composite material under the corresponding extended working condition as output to obtain a high-fidelity neural network.

[0020] Optionally, the test data is expanded to obtain the expansion strength margin of the aviation composite material under different expansion conditions, including: obtaining a scaling interval.

[0021] Extracts a preset number of amplitudes from the zoom interval.

[0022] Determine any amplitude as the current amplitude.

[0023] Determine any test condition as the current test condition.

[0024] Multiply the generalized internal force state corresponding to the current test condition by the current amplitude to obtain the current extended condition.

[0025] The extended strength margin of the aerospace composite material under the current test condition is multiplied by the current amplitude to obtain the extended strength margin of the aerospace composite material under the current extended condition.

[0026] Update the current test condition and return to the step "multiply the generalized internal force state corresponding to the current test condition by the current amplitude to obtain the current extended condition" until all test conditions are traversed.

[0027] Update the current amplitude and return to the step "determine any test condition as the current test condition" until all amplitudes are traversed to obtain the extended strength margin of the aviation composite material under different extended conditions.

[0028] Optionally, according to the generalized internal force state of the working condition to be predicted, the low-fidelity neural network and the high-fidelity neural network are used to complete the strength verification of the aviation composite material under the working condition to be predicted, including: inputting the generalized internal force state of the working condition to be predicted into the low-fidelity neural network to obtain a low-fidelity strength margin prediction value.

[0029] The low-fidelity strength margin prediction value and the generalized internal force state of the working condition to be predicted are input into the high-fidelity neural network to obtain the high-fidelity strength margin prediction value.

[0030] Based on high-fidelity strength margin prediction values, the strength verification of aviation composite materials under predicted working conditions is completed.

[0031] Optionally, completing strength verification of the aviation composite material under the working condition to be predicted based on the high-fidelity strength margin prediction value includes: when the high-fidelity strength margin prediction value is greater than or equal to a strength margin threshold, determining that the aviation composite material is in a failure state under the working condition to be predicted.

[0032] When the high-fidelity strength margin prediction value is less than the strength margin threshold, it is determined that the aviation composite material is in a non-failure state under the working condition to be predicted.

[0033] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned multi-fidelity composite material strength verification method.

[0034] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned multi-fidelity composite material strength verification method when executed by a processor.

[0035] In a fourth aspect, the present application provides a computer program product, including a computer program, which implements the above-mentioned multi-fidelity composite material strength verification method when executed by a processor.

[0036] According to the specific embodiments provided in the present application, the present application discloses the following technical effects: the present application provides a multi-fidelity composite material strength verification method, equipment, medium and product, which quickly obtains a large amount of low-fidelity strength evaluation data through numerical simulation, and then uses an artificial neural network to establish a low-fidelity data-driven composite material shell strength model. On this basis, a "two-step" neural network (i.e., a low-fidelity neural network and a high-fidelity neural network) is applied to fuse the low-fidelity data with the high-fidelity data obtained from the experiment, while retaining the failure behavior in the full load space of the structure learned from the simulation data, ensuring the numerical consistency with the experimental test results, and achieving a balance between efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 This is a flow chart of a multi-fidelity composite material strength verification method in one embodiment of the present application.

[0039] Figure 2 This is a flow chart of a multi-fidelity composite material strength verification method in another embodiment of the present application.

[0040] Figure 3 This is a schematic diagram of the multi-fidelity neural network model architecture in one embodiment of the present application.

[0041] Figure 4 Schematic diagram of strength test specimens and test settings under different working conditions in one embodiment of the present application.

[0042] Figure 5 This is a diagram of the final failure results of laminate specimens under different working conditions in one embodiment of the present application.

[0043] Figure 6 It is a graph of load-displacement data of each group of tests in one embodiment of the present application; Figure 6 Part (a) is a load-displacement curve of a 0° tensile test in an embodiment of the present application; Figure 6 Part (b) is a load-displacement curve diagram of a 90° tensile test in an embodiment of the present application; Figure 6 Part (c) is a load-displacement curve diagram of a 0° compression test in an embodiment of the present application; Figure 6 Part (d) is a load-displacement curve diagram of a 90° compression test in an embodiment of the present application; Figure 6 Part (e) is a load-displacement curve of a 0° bending test in an embodiment of the present application; Figure 6 Part (f) is a load-displacement curve of a 90° bending test in an embodiment of the present application; Figure 6 Part (g) is a load-displacement curve of an in-plane shear test in one embodiment of the present application; Figure 6 Part (h) is a load-displacement curve of an out-of-plane shear test in an embodiment of the present application.

[0044] Figure 7 This is a diagram of low-fidelity model strength prediction results in one embodiment of the present application. Figure 7 Part (a) is an embodiment of the present application. Structural failure prediction result diagram in load space; Figure 7 Part (b) is an embodiment of the present application. Structural failure prediction result diagram in load space; Figure 7 Part (c) is an embodiment of the present application. Structural failure prediction result diagram in load space; Figure 7 Part (d) is an embodiment of the present application. Graph of structural failure prediction results in the load space.

[0045] Figure 8 This is a comparison chart of the prediction results of the low-fidelity model and the multi-fidelity model in one embodiment of the present application. Figure 8 Part (a) is an embodiment of the present application. Structural strength envelope diagram within the load space; Figure 8 Part (b) is an embodiment of the present application. Structural strength envelope diagram within the load space; Figure 8 Part (c) is an embodiment of the present application. Structural strength envelope diagram within the load space; Figure 8 Part (d) is an embodiment of the present application. Structural strength envelope diagram within the load space.

[0046] Fig. 9 It is a structural schematic diagram of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0049] In an exemplary embodiment, Figure 1 As shown, a multi-fidelity composite material strength verification method is provided, including: steps 101 to 108.

[0050] Step 101: Use the finite element model to construct a three-dimensional unit cell model of the aviation composite material.

[0051] Step 102: Perform multiple sampling simulations based on the three-dimensional unit cell model to construct a unit failure judgment data set. The unit failure judgment data set includes the strength margin of the aviation composite material under different simulation conditions. The generalized internal force state of the aviation composite material under different simulation conditions is different.

[0052] Step 103: Construct a first fully connected neural network and a second fully connected neural network.

[0053] Step 104: The first fully connected neural network is trained using the unit failure judgment data set to obtain a low-fidelity neural network. The generalized internal force state of the aviation composite material under different simulation conditions is used as input, and the strength margin of the aviation composite material under the corresponding simulation conditions is used as output to train the first fully connected neural network to obtain a low-fidelity neural network.

[0054] Step 105: Perform a structural strength test on the aviation composite material to obtain test data. The test data includes the strength margin of the aviation composite material under different test conditions. The generalized internal force state of the aviation composite material under different test conditions is different.

[0055] Step 106: Use the test data and the low-fidelity neural network to train the second fully connected neural network to obtain a high-fidelity neural network.

[0056] Step 107: Obtain the generalized internal force state of the working condition to be predicted.

[0057] Step 108: Based on the generalized internal force state of the working condition to be predicted, the low-fidelity neural network and the high-fidelity neural network are used to complete the strength verification of the aviation composite material under the working condition to be predicted.

[0058] Among them, the first fully connected neural network or the second fully connected neural network is as follows.

[0059] .

[0060] .

[0061] .

[0062] in, represents the first Tier neurons; Represents the generalized internal force state of aviation composite materials under simulation conditions; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron for Tier The weights of the neurons; It means that after nonlinear activation Tier neurons; It means that after nonlinear activation Tier neurons; represents a nonlinear activation function; represents the output of the first fully connected neural network; represents the first Tier neurons; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron weights; represents the first A neuron.

[0063] Step 106 includes: step 106-1 to step 106-4.

[0064] Step 106-1: Expand the test data to obtain the expansion strength margin of the aviation composite material under different expansion conditions. The generalized internal force state of the aviation composite material under different expansion conditions is different.

[0065] Step 106-2: Input the generalized internal force state of the aviation composite material under different extended working conditions into the low-fidelity neural network to obtain the low-fidelity strength margin.

[0066] Step 106-3: The generalized internal force state and low-fidelity strength margin of the aviation composite material under the same extended working condition are used as input data pairs.

[0067] Step 106-4: Taking the input data pair as input and the extended strength margin of the aviation composite material under the corresponding extended working condition as output, the second fully connected neural network is trained to obtain a high-fidelity neural network.

[0068] Step 106-1 includes: step 106-1-1 to step 106-1-8.

[0069] Step 106-1-1: Obtain the zoom range.

[0070] Step 106-1-2: extract a preset number of amplitudes from the zoom interval.

[0071] Step 106-1-3: Determine any amplitude as the current amplitude.

[0072] Step 106-1-4: Determine any test condition as the current test condition.

[0073] Step 106-1-5: Multiply the generalized internal force state corresponding to the current test condition by the current amplitude to obtain the current extended condition.

[0074] Step 106-1-6: Multiply the strength margin of the aviation composite material under the current test condition by the current amplitude to obtain the extended strength margin of the aviation composite material under the current extended condition.

[0075] Step 106-1-7: Update the current test condition and return to step 106-1-6 until all test conditions are traversed.

[0076] Step 106-1-8: Update the current amplitude and return to step 106-1-4 until all amplitudes are traversed to obtain the expansion strength margin of the aviation composite material under different expansion conditions.

[0077] Step 108 includes: step 108-1 to step 108-3.

[0078] Step 108-1: Input the generalized internal force state of the working condition to be predicted into the low-fidelity neural network to obtain a low-fidelity strength margin prediction value.

[0079] Step 108-2: Input the low-fidelity strength margin prediction value and the generalized internal force state of the working condition to be predicted into the high-fidelity neural network to obtain a high-fidelity strength margin prediction value.

[0080] Step 108-3: Complete the strength check of the aviation composite material under the working condition to be predicted based on the high-fidelity strength margin prediction value. When the high-fidelity strength margin prediction value is greater than or equal to the strength margin threshold, it is determined that the aviation composite material is in a failed state under the working condition to be predicted. When the high-fidelity strength margin prediction value is less than the strength margin threshold, it is determined that the aviation composite material is not in a failed state under the working condition to be predicted.

[0081] Based on neural networks, the fusion of test and simulation data is realized, and a multi-fidelity data-driven strength model for aviation composite materials is constructed. The generalization of limited test data in the full load space is realized, and the failure behavior of the structure in the full load space obtained by learning from simulation data is retained, while ensuring the numerical consistency with the experimental test results. It effectively solves the contradiction between the insufficient accuracy of low-cost simulation data and the difficulty in obtaining high-fidelity simulation data in the construction of traditional data-driven composite material strength models, so as to achieve a balance between efficiency and accuracy in the construction of strength models. In actual engineering applications, this application effectively improves the analysis efficiency while ensuring the analysis accuracy, and can help realize high-reliability real-time rapid failure prediction and strength verification of aviation composite plate and shell structures under complex external load conditions. In addition, this application has good versatility. In addition to the laminated plate structure given in the case, it is also applicable to various forms of aviation plate and shell structures such as metamaterials, sandwich, and reinforcement, and has great application potential.

[0082] In an exemplary embodiment, a multi-fidelity composite material strength verification method is provided, including: composite material structure finite element model simulation, low-fidelity model construction, structural strength test under typical working conditions, multi-fidelity model construction and structural strength verification. A composite material strength model that can integrate different fidelity data obtained from experiments and simulations is firstly based on a large number of low-cost low-fidelity data sets obtained from simulations to establish a strength verification proxy model suitable for real-time, fast and efficient simulation, which can perform structural failure judgment based on the cross-sectional internal forces of shell units. On this basis, through a small number of structural tests, high-fidelity test results of structural strength under typical working conditions are obtained, and through a multi-fidelity model construction method, the test data is used to correct the low-fidelity model. The generalization of limited test data in the full load space is achieved, and the efficiency of data-driven composite shell unit strength model construction is effectively improved while ensuring accuracy, which can help achieve high-reliability real-time fast failure prediction and strength verification under complex external load conditions for aviation composite plate and shell structures. In addition, this application has good versatility. In addition to the laminated plate structure given in the case, it is also applicable to various forms of aviation plate and shell structures such as metamaterials, sandwich, and reinforcement. It has great application potential. The implementation flow chart is shown in Figure 2 and Figure 3 .

[0083] Step 1. Simulation of finite element model of composite structure.

[0084] Step 1.1 Construct a three-dimensional unit cell model of the composite material structure. This is achieved by constructing a finite element model of the structural unit cell and applying equation constraints to impose periodic boundary conditions on the unit cell.

[0085] Step 1.2 Perform a large number of sampling simulations to build a unit failure judgment data set.

[0086] Unit forces are applied to each reference node in turn, and the stress and interface force responses of the unit cell under 8 unit generalized stress conditions including tension, bending, torsion, and in-plane shear are simulated and calculated. On this basis, random sampling is performed in the 8-dimensional generalized internal force space. Based on the linear superposition of the field, the structural response under different generalized internal force states in each sample is calculated. The strength judgment criterion based on Cauchy stress at the material scale is applied to judge the failure state of each layer of the structure. The interlayer interface strength judgment criterion is applied to judge the delamination failure state and obtain a strength simulation data set.

[0087] Step 2. Low-fidelity model (i.e. low-fidelity neural network) construction.

[0088] Step 2.1 Build a fully connected neural network.

[0089] Build includes layer neural network, The width of the layer is The neural network input is the generalized force state: ,in, is the pulling force in one direction, is the pulling force in two directions, is the in-plane shear force, is the bending moment in direction 1, is the bending moment in two directions, is the torque, is the shear force in direction 1, is the shear force in two directions. Output is the strength margin , which is obtained based on the strength criterion of the structural details, and a value greater than or equal to 1 indicates structural failure.

[0090] .

[0091] .

[0092] .

[0093] in, The value range is ; The value range is .

[0094] Step 2.2: Train the model parameters based on the simulation data set.

[0095] Low-fidelity neural networks It is based on low-fidelity data obtained from simulation For training. is the training data, is the data size, Indicates data points. Set the loss function and continuously reduce the loss function , train the parameters of each hidden layer of the neural network, that is: .

[0096] in, are vectors consisting of all weight and bias parameters respectively.

[0097] Step 3. Structural strength test under typical working conditions.

[0098] Carry out specimen-level tests to test the structural strength of the structure under the action of a single generalized force and obtain high-fidelity data.

[0099] It is usually a standard test under a single load, such as unidirectional tension, compression, bending or shear in all directions. And this method aims to construct the initial failure strength criterion, that is, the focus of the test needs to be on obtaining the generalized load state when the structure begins to fail, rather than the ultimate load at the final failure.

[0100] Step 4. Multi-fidelity model (i.e. high-fidelity neural network) construction.

[0101] Step 4.1: Expand the high-fidelity data.

[0102] It is assumed here that the damage mode does not change within a small range of the initial damage load, that is, the structural strength margin It changes approximately linearly with the load. Therefore, by setting a certain scaling interval , evenly take the preset number in the interval Amplitude , expand each set of test data: , To test the data size, Indicates data points.

[0103] Step 4.2: Using the low-fidelity model, based on the generalized force input in each set of high-fidelity data, obtain low-fidelity strength prediction results.

[0104] The high-fidelity dataset is obtained in step 4.1 In this step, As input, the low-fidelity neural network model constructed in step 2 is used for prediction to obtain the low-fidelity failure prediction result of the structure under this load state. ,Right now: .

[0105] Step 4.3: Based on the high-fidelity data and the low-fidelity model prediction results, train a second neural network for model correction.

[0106] The input of training data consists of two parts: Part 1 Input comes from high frequency data, i.e., section forces from structural tests. Part II This is the predicted value of the low-fidelity model obtained in step 4.2 under the input of high-fidelity data. The output of the training data is the expanded data of each group of test results. , that is, the data set finally used for training the neural network model is: The model construction and training methods are the same as step 2.

[0107] Step 5. Structural strength verification.

[0108] Based on the given generalized internal force state, the failure of plate and shell elements is judged through a multi-fidelity surrogate model to perform structural strength verification.

[0109] Given the generalized internal force state As input, it first passes through the first neural network to make a low-fidelity prediction. Then, this prediction is submitted to the second neural network along with the initial input, which corrects the result, namely: .

[0110] Get the final high-fidelity prediction result.

[0111] Taking a 24-ply ZT7H5429 material laminate as an example, a multi-fidelity composite material strength verification method provided in this embodiment is specifically described.

[0112] Step 1. Simulation of finite element model of composite structure.

[0113] Construct a finite element model of the structural unit cell, and apply equation constraints to impose periodic boundary conditions on the unit cell. Apply unit force to each reference node in turn, and simulate and calculate the stress and interface force response of the unit cell under 8 unit generalized stress conditions including tension, bending, torsion, and in-plane shear. On this basis, random sampling is performed in the 8-dimensional generalized internal force space, and based on the linear superposition of the field, the structural response under different generalized internal force states in each sample is calculated. The strength criterion of the material is used to check the failure state of the structure. In this case, the Hashin criterion and the quadratic interface force criterion are used to judge the failure of the laminate.

[0114] .

[0115] in, Characterize fiber breakage under tension, Characterize fiber buckling and kinking under compression, Characterize matrix failure under tension and compression, Characterizes delamination failure between layers. The action represents the Cauchy stress. They represent fiber tension, fiber compression, matrix tension, matrix compression, in-plane shear, and in-plane tensile strength, respectively.

[0116] Based on this, the strength simulation data set can be calculated .

[0117] Step 2. Low-fidelity model construction.

[0118] Construct a fully connected neural network and train the model based on the simulation data set. In this case, the nonlinear activation function uses a linear rectifier function, that is, The loss function used for parameter training uses the mean square error loss criterion, namely: , This represents the neural network.

[0119] Step 3. Structural strength test under typical working conditions.

[0120] Eight sets of laminate strength tests were conducted to measure the tensile, compressive, bending and shear strength of the laminates. The test results are high-fidelity data. The types and standards of the eight sets of laminate strength tests are summarized in Table 1.

[0121] Table 1 Strength test and test standards

[0122] Each test set consists of three samples. Figure 4 The load is increased in each test until the specimen is destroyed. The failure mode diagram of each group of tests is shown in Figure 5 The load-displacement curves of the eight groups of tests obtained in the experiment are shown in Figure 6 As shown, for clarity, the test curves of different specimens in each figure are offset along the x-axis.

[0123] Since this is the initial failure strength of the present application, when the load-displacement curve begins to deviate from the initial slope (indicated by the dotted line), it is Figure 6 The test results and the corresponding shell generalized section forces are shown in Table 2.

[0124] Table 2 Strength test results

[0125] Based on the test results, 8 pairs of high-fidelity data were obtained. The input of each pair is the section force , the corresponding output is (indicates the onset of damage) and is calculated from the test load.

[0126] Step 4. Multi-fidelity model construction.

[0127] By constructing Figure 3 The two-step multi-fidelity model shown in the figure realizes the fusion of experimental data and simulation data. First, it expands on the basis of the existing 8 pairs of high-frequency data. This application is in [0.9,1.1] to The amplitude is uniformly taken within the interval , expand each set of test data: .

[0128] Based on this High-fidelity data , train the second neural network As a revised model, it forms a multi-fidelity model together with the first neural network.

[0129] Step 5. Structural strength verification.

[0130] In the application of the model, given the generalized internal force state As input, the failure state of plate and shell units is judged through the multi-fidelity proxy model, and the structural strength is checked. At this point, the construction method of the multi-fidelity strength model of aviation composite materials that integrates test and simulation data has been applied.

[0131] The implementation of the large-scale sampling simulation in step 1.2 includes but is not limited to the method of applying linear superposition as described above. Any other method that can obtain the response results of the unit cell model under a certain generalized plate and shell load state through simulation can use this application.

[0132] The data obtained from the specimen-level test in step 3 include but are not limited to standard tests under a single load. Any strength test, as long as a specific generalized load state when the structure begins to fail is obtained, can be used as high-fidelity data for subsequent multi-fidelity model training.

[0133] This case study presents the strength prediction results of the low-fidelity and multi-fidelity data-driven strength models for ZT7H / 5429 laminates, and applies the proposed strength model to the strength assessment of a composite wing structure.

[0134] Based on low-fidelity RVE simulation data, a neural network with 4 layers and 64-132-64-8 units was constructed and trained. The model prediction results in some typical load section spaces are as follows: Figure 7 As shown. Figure 7 (a) and Figure 7 (b) It can be seen that under biaxial tension, compression or bending, the strength envelope, i.e. the boundary between the blue and orange regions, is a diagonally oriented diamond shape. This is because the deformation in one direction is compensated by the deformation in the other direction due to the Poisson effect. Figure 7 The strength envelope in (c) is roughly circular. This is because under the out-of-plane shear load, the failure of the laminate is mainly caused by delamination. Since the interlaminar shear strength is independent of direction, the out-of-plane shear strength of the laminate is the same in all directions. Figure 7 (d) shows the result of coupled axial force and bending loading. Under this loading, the laminate close to one laminate surface fails first due to the superposition of axial force and bending tensile load. In some loading cases, such as Figure 7 (b) and 7(d), the predictions agree well with the test data, but for some other cases, e.g. Figure 7 (a) and 7(c), the predictions are not accurate enough. But in all cases, the predicted intensity envelopes capture the pattern of the test data well.

[0135] On this basis, a modified neural network model with 3 layers and 64-32-8 neurons was trained based on the test data, achieving a loss of 0.00001 on the test data and a failure prediction accuracy of 99.8%. Figure 8 As shown in the figure, it can be seen that the overall pattern of the strength envelope remains almost unchanged from the low-fidelity prediction to the multi-fidelity prediction, but the multi-fidelity prediction is in better agreement with the test data. The results show that the multi-fidelity model proposed in this application can effectively measure the strength of the laminate using the shell section force, and can efficiently build the model using a small amount of test data and a large amount of simulation data.

[0136] In addition, the present application has good versatility. In addition to the laminate structure given in the case, it is also applicable to various forms of aviation plate and shell structures such as metamaterials, sandwich, and reinforcement, and has great application potential.

[0137] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a multi-fidelity composite material strength verification method is implemented.

[0138] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0139] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0140] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0141] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0142] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0143] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0144] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A multi-fidelity composite material strength verification method, characterized in that: include: Use finite element model to construct three-dimensional unit cell model of aviation composite materials; Perform multiple sampling simulations based on the three-dimensional unit cell model to construct a unit failure judgment data set; The unit failure judgment data set includes the strength margin of the aviation composite material under different simulation conditions; the generalized internal force state of the aviation composite material under different simulation conditions is different; Constructing a first fully connected neural network and a second fully connected neural network; Using the unit failure judgment data set to train the first fully connected neural network to obtain a low-fidelity neural network; Conducting a structural strength test on the aviation composite material to obtain test data; the test data includes the strength margin of the aviation composite material under different test conditions; the generalized internal force state of the aviation composite material under different test conditions is different; Using the test data and the low-fidelity neural network to train the second fully connected neural network to obtain a high-fidelity neural network; Obtain the generalized internal force state of the working condition to be predicted; According to the generalized internal force state of the working condition to be predicted, the low-fidelity neural network and the high-fidelity neural network are used to complete the strength verification of the aviation composite material under the working condition to be predicted.

2. The multi-fidelity composite material strength verification method according to claim 1, characterized in that: The first fully connected neural network is trained using the unit failure judgment data set to obtain a low-fidelity neural network, including: Taking the generalized internal force state of aviation composite materials under different simulation conditions as input and the strength margin of aviation composite materials under corresponding simulation conditions as output, the first fully connected neural network is trained to obtain a low-fidelity neural network.

3. The multi-fidelity composite material strength verification method according to claim 1, characterized in that: The first fully connected neural network or the second fully connected neural network is: ； ； ； in, represents the first Tier neurons; Represents the generalized internal force state of aviation composite materials under simulation conditions; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron for Tier The weights of the neurons; It means that after nonlinear activation Tier neurons; It means that after nonlinear activation Tier neurons; represents a nonlinear activation function; represents the output of the first fully connected neural network; represents the first Tier neurons; Indicates Tier The bias parameters of neurons; Indicates The width of the layer; Indicates Tier The neuron weights; represents the first A neuron.

4. The multi-fidelity composite material strength verification method according to claim 1, characterized in that: The second fully connected neural network is trained using the test data and the low-fidelity neural network to obtain a high-fidelity neural network, including: By expanding the test data, the extended strength margin of aviation composite materials under different extended working conditions is obtained; the generalized internal force state of aviation composite materials under different extended working conditions is different; The generalized internal force state of the aviation composite material under different extended working conditions is input into the low-fidelity neural network to obtain the low-fidelity strength margin; The generalized internal force state and low-fidelity strength margin of aerospace composite materials under the same extended working condition are taken as input data pairs; The second fully connected neural network is trained with the input data pair as input and the extended strength margin of the aviation composite material under the corresponding extended working condition as output to obtain a high-fidelity neural network.

5. The multi-fidelity composite material strength verification method according to claim 4, characterized in that: The test data is expanded to obtain the expansion strength margin of aviation composite materials under different expansion conditions, including: Get the zoom range; Extract a preset number of amplitudes from the zoom interval; Determine any amplitude value as the current amplitude value; Determine any test condition as the current test condition; Multiply the generalized internal force state corresponding to the current test condition by the current amplitude to obtain the current extended condition; The strength margin of the aerospace composite material under the current test condition is multiplied by the current amplitude to obtain the extended strength margin of the aerospace composite material under the current extended condition; Update the current test condition and return to step "multiply the generalized internal force state corresponding to the current test condition by the current amplitude to obtain the current extended condition" until all test conditions are traversed; Update the current amplitude and return to step "determine any test condition as the current test condition" until all amplitudes are traversed to obtain the extended strength margin of the aviation composite material under different extended conditions.

6. The multi-fidelity composite material strength verification method according to claim 1, characterized in that: According to the generalized internal force state of the working condition to be predicted, the low-fidelity neural network and the high-fidelity neural network are used to complete the strength verification of the aviation composite material under the working condition to be predicted, including: Inputting the generalized internal force state of the working condition to be predicted into the low-fidelity neural network to obtain a low-fidelity strength margin prediction value; Inputting the low-fidelity strength margin prediction value and the generalized internal force state of the working condition to be predicted into the high-fidelity neural network to obtain the high-fidelity strength margin prediction value; Based on high-fidelity strength margin prediction values, the strength verification of aviation composite materials under predicted working conditions is completed.

7. The multi-fidelity composite material strength verification method according to claim 6, characterized in that: Based on high-fidelity strength margin prediction values, the strength verification of aviation composite materials under the predicted working conditions is completed, including: When the high-fidelity strength margin prediction value is greater than or equal to the strength margin threshold, it is determined that the aviation composite material is in a failure state under the working condition to be predicted; When the high-fidelity strength margin prediction value is less than the strength margin threshold, it is determined that the aviation composite material is in a non-failure state under the working condition to be predicted.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the multi-fidelity composite material strength verification method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multi-fidelity composite material strength verification method described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the multi-fidelity composite material strength verification method described in any one of claims 1 to 7 is implemented.

Images