Method and device for identifying impact damage of carbon fiber composite laminates

Abstract

The present application belongs to the technical field of laminated plate damage identification, and provides a laminated plate impact response database construction method, a damage identification method and device. The laminated plate impact response database construction method comprises obtaining actual impact response signals and finite element simulation impact response signals of a laminated plate under multiple impact energies; a set of empirical mode - Euclidean distance similarity discrimination combined algorithm is used to extract feature vectors of the actual impact response signals and the finite element simulation impact response signals; based on the feature vectors of the actual response impact signals and the finite element simulation impact response signals and a limit learning machine regression prediction method, a nonlinear relationship between the feature vectors and damage types and positions is established; based on the nonlinear relationship between the feature vectors and damage types and positions, a damage identification model of the laminated plate structure is determined, and a laminated plate impact response database is constructed.

Description

Technical Field

[0001] This invention belongs to the field of laminate damage identification technology, and particularly relates to a method for constructing a laminate impact response database, a damage identification method, and a device. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Carbon fiber composites are crucial materials for spacecraft structures. During service, spacecraft are inevitably subjected to external loads, material aging, fatigue failure, and other factors, leading to visible damage such as cracks and deformation, as well as numerous internal damages that are not immediately apparent. This poses a significant threat to the overall operational reliability of the space system. Therefore, real-time monitoring of critical spacecraft components is essential to obtain accurate response data and promptly determine the type and location of structural damage.

[0004] Currently, traditional industrial methods often rely on expert experience to judge and estimate the structural condition of composite materials. This method is both time-consuming and labor-intensive. Therefore, embedding sensors and other sensitive components into composite materials to collect response information in real time and process the data to achieve structural damage identification has become a research hotspot. Currently, commonly used methods include frequency methods, mode shape methods, data-driven methods, and intelligent algorithms. Among these methods, the frequency method determines the damage location by comparing the natural frequencies of the structure before and after the damage. Studies have shown that frequency has low sensitivity to structural damage and poor discrimination effect. The mode shape method makes a judgment based on the difference in the mode shape of the structure before and after the damage. This method requires a high degree of integrity of the mode shape, and its accuracy depends on the accuracy of the collected data. It is more suitable for structures under laboratory conditions. The data-driven method refers to wavelet transform, Fourier transform, etc., which converts the signal to the frequency domain for analysis. In the process of conversion, it may cause the loss of detailed information of the signal. To address the above problems, intelligent algorithms have emerged. This algorithm can perform corresponding calculations without the need for a precise model of the system. It has a fast calculation speed and does not produce cumulative errors. It constructs a nonlinear mapping model between the signal feature vector and the damage location or type through self-learning. It uses a large amount of experimental data as known conditions, fully explores the detailed features of the signal, and constructs a damage identification model. Based on the principle of intelligent self-learning, the damage discrimination accuracy is high. However, the inventors found that the existing intelligent algorithms require a lot of prior knowledge, have a long experimental cycle, and the damaged structure cannot be reused. Summary of the Invention

[0005] To address the technical problems mentioned above, this invention provides a method for constructing a laminate impact response database, a damage identification method, and a device. Based on the idea of ​​extreme learning machine regression prediction, it establishes a mapping relationship between finite element simulation data and real data, using simulation data to infinitely approximate real data, thereby establishing a massive database of structural impact response. Based on the constructed massive database, an ELM classifier is used to construct a damage identification model, enabling the identification of damage location and type in carbon fiber composite laminate structures.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for constructing a laminate impact response database.

[0008] In one or more embodiments, a method for constructing a laminate impact response database includes:

[0009] Obtain the actual impact response signals and finite element simulation impact response signals of laminates under various impact energies;

[0010] The feature vectors of the actual impact response signal and the finite element simulation impact response signal are extracted using a combined algorithm of ensemble empirical mode and Euclidean distance similarity discrimination.

[0011] Based on the eigenvectors of actual impact response signals and finite element simulation impact response signals, and using the limit learning machine regression prediction method, a nonlinear relationship between eigenvectors and damage type and location is established.

[0012] Based on the nonlinear relationship between feature vectors and damage type and location, a damage identification model for laminated plate structures is determined, and a laminated plate impact response database is constructed.

[0013] As one implementation method, the process of acquiring the actual impact response signal is as follows:

[0014] The actual impact response signal of the laminate under various impact energy tests is obtained based on the impact damage monitoring system; wherein, the impact damage monitoring system is constructed by embedding Y-shaped FBG sensors into the carbon fiber composite plate.

[0015] As one implementation method, the process of obtaining the finite element simulation impact response signal is as follows:

[0016] The laminate was subjected to simulation analysis under the same conditions as the impact energy test to obtain the finite element simulation impact response signal.

[0017] In one implementation, in the impact damage monitoring system, Y-shaped FBG sensors are arranged at the four vertices and the center of the laminate.

[0018] As one implementation method, the eigenvectors of the actual impact response signal and the finite element simulation impact response signal are structural strain-time curves.

[0019] As one implementation method, the process of establishing the nonlinear relationship between feature vectors and damage type and location is as follows:

[0020] The eigenvectors of the finite element simulation impact response signal are used as the input matrix, and the corresponding eigenvectors of the actual impact response signal are used as the output matrix. The data feature learning model is trained to narrow the gap between the finite element simulation impact response signal and the actual impact response signal.

[0021] Based on the trained data feature learning model and the feature vector of the finite element simulation impact response signal, a set of data consisting of the feature vector, damage type, and location is output.

[0022] A second aspect of the present invention provides a method for identifying laminate loss.

[0023] In one or more embodiments, a laminate loss identification method includes:

[0024] Obtain the impact response signal of the laminate and extract the corresponding feature vector;

[0025] Based on the damage identification model of laminate structure in the laminate impact response database, the damage type and location are obtained; wherein, the laminate impact response database is constructed using the laminate impact response database construction method described above.

[0026] A third aspect of the present invention provides an apparatus for constructing a laminate impact response database.

[0027] In one or more embodiments, a laminate impact response database construction apparatus includes:

[0028] The impact response signal acquisition module is used to acquire the actual impact response signal and finite element simulation impact response signal of the laminate under various impact energies.

[0029] The feature vector extraction module is used to extract feature vectors from the actual impact response signal and the finite element simulation impact response signal using a combined algorithm of ensemble empirical mode-Euclidean distance similarity discrimination.

[0030] The nonlinear relationship construction module is used to establish a nonlinear relationship between the feature vector and the damage type and location based on the feature vector of the actual response impact signal and the finite element simulation impact response signal and the limit learning machine regression prediction method.

[0031] The impact response database construction module is used to determine the damage identification model of the laminate structure based on the nonlinear relationship between feature vectors and damage type and location, and to construct the impact response database of the laminate.

[0032] A fourth aspect of the present invention provides a laminate loss identification device.

[0033] In one or more embodiments, a laminate loss identification device includes:

[0034] The actual impact feature extraction module is used to acquire the actual impact response signal of the laminate and extract the corresponding feature vector.

[0035] The laminated plate loss identification module is used to obtain the damage type and location based on the damage identification model of the laminated plate structure in the laminated plate impact response database; wherein, the laminated plate impact response database is constructed using the laminated plate impact response database construction method described above.

[0036] A fifth aspect of the present invention provides an electronic device.

[0037] In one or more embodiments, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the laminate impact response database construction method described above.

[0038] In other embodiments, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps in the laminate loss identification method described above.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] (1) This invention addresses the problems of long experimental cycles, non-reusability of damaged structures, and high human and material costs associated with current intelligent learning algorithms. It proposes a method for constructing a massive database based on the ELM regression algorithm. By establishing a nonlinear mapping relationship between simulated and real data under various impact energies and locations, and employing the EEMD-Euclidean joint algorithm to fully exploit signal details, the gap between simulated and real data is narrowed, establishing a damage data feature learning model. Further impact simulations under different conditions are performed on the simulation model and input into the constructed model, thereby building a large-sample structural damage database that approximates real-world conditions.

[0041] (2) To maximize the strain sensitivity of the embedded FBG sensor and minimize the impact of the embedded sensing element on the laminate structure, this invention utilizes the characteristic that FBG sensors are more sensitive to stress waves in the vertical direction. Y-shaped pairs of FBG sensors are embedded at the four vertices and center of the laminate to ensure accurate reception of impact signals from all directions. This FBG sensor layout inherits the advantages of high-density, large-area distributed sensor networks for high-sensitivity and high-completeness signal reception while minimizing the number of sensors and effectively reducing the impact on structural performance.

[0042] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0043] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0044] Figure 1 This is a flowchart of database construction and damage identification according to an embodiment of the present invention;

[0045] Figure 2 This is an embodiment of the carbon fiber laminate impact damage monitoring system of the present invention;

[0046] Figure 3(a) is a set of impact test points of carbon fiber composite laminates taken before ELM regression training in an embodiment of the present invention;

[0047] Figure 3(b) is a set of impact test points for carbon fiber composite laminates after ELM regression training in an embodiment of the present invention.

[0048] Figure 4 This is a displacement distribution cloud map from a finite element simulation impact experiment according to an embodiment of the present invention;

[0049] Figure 5 This describes the working principle of the Extreme Learning Machine in this embodiment of the invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0051] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0052] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0053] Example 1

[0054] Reference Figure 1 This embodiment provides a method for constructing a laminate impact response database, which includes:

[0055] Step 1: Obtain the actual impact response signal and finite element simulation impact response signal of the laminate under various impact energies.

[0056] In the specific implementation process, the acquisition process of the actual impact response signal is as follows:

[0057] The actual impact response signal of the laminate under various impact energy tests is obtained based on the impact damage monitoring system; wherein, the impact damage monitoring system is constructed by embedding Y-shaped FBG sensors into the carbon fiber composite plate.

[0058] In the impact damage monitoring system, Y-shaped FBG sensors are arranged at the four vertices and the center of the laminate.

[0059] Specifically, the Y-shaped sensor pairs FBGs1 to FBGs5 are arranged at the four vertices and the center of the laminate, respectively. See [link to specific location information] for details. Figure 2 This forms a 360*360mm monitoring area, where each sensor pair consists of 3 FBG sensors. The included angle between each pair of FBG (sensor grid length: 5mm) sensors is 120°.

[0060] In this embodiment, the process of preparing the embedded carbon fiber composite plate is as follows:

[0061] Step a: When selecting the layup layers for the carbon fiber composite laminate, the principles of balance and symmetry and layup orientation are followed, with a total of 16 layers of prepreg laid symmetrically [±45° / 0° / 0° / [90°0°]2]. 2s ;

[0062] Step b: The FBG sensor is implanted between the 3rd and 4th layers of prepreg. During FBG sensor implantation, the optical fiber is relatively fragile, and the flow of resin during curing can cause the fiber to shift. Therefore, necessary protection is required. First, the optical fiber is pre-fixed by placing the FBG sensor in the desired position and gently pressing the fiber. To keep the fiber straight, weights are added to the fiber's tail end, and adhesive is applied to both ends. After the adhesive solidifies, the weights are removed. Second, a Teflon sleeve is fitted to the fiber's lead-out end, with 15mm embedded inside the structure and the remainder outside. This sleeve has excellent high-temperature and high-pressure resistance, effectively preventing fiber breakage. Finally, a secondary curing process is used to pre-cure both the upper (layers 1-3) and lower (layers 4-16) sections where the FBG sensor is laid, curing the epoxy resin portion. After the fiber grating is pasted, the pre-formed composite laminate is then fully cured a second time. The pre-curing process ensures that the substrate has sufficient rigidity during fiber Bragg grating sensor installation, making it easier to maintain the straightness of the optical fibers. In addition, the reduced fluidity of the pre-cured substrate resin during the secondary curing process also reduces its impact on the position of the optical fibers laid on the prefabricated plate.

[0063] It should be noted that other existing methods can also be used in the preparation process of embedded carbon fiber composite plates. Those skilled in the art can choose the specific preparation method of embedded carbon fiber composite plates according to the actual situation, which will not be described in detail here.

[0064] In this embodiment, the process of setting up the impact test system is as follows:

[0065] Impact damage monitoring system for carbon fiber composite laminates, such as Figure 2 As shown, it consists of a spring impact hammer (capable of applying load energy: 0.5J, 5J, 10J, 15J), an embedded carbon fiber composite laminate (400*400*3.2mm), a high-precision fiber optic demodulator (sampling frequency: 1000Hz), and a control computer.

[0066] The impact test in this embodiment is a drop hammer impact test, and the experimental procedure is as follows:

[0067] First, an orthogonal network was generated for the carbon fiber composite laminate as shown in Figure 3(a). The spring hammer was manually released, and the center (4*4) of the grid was impacted sequentially. The strain information (0.5s) caused by the load impact was collected by the FBG sensor, demodulated by the fiber optic grating demodulator, and then transmitted to the control computer for storage and processing. Each experiment included wavelength information from 10 FBG sensors at 16 impact locations.

[0068] Where 0.5 J represents the damage-free energy, and 5 J, 10 J, and 15 J represent the lossy energy. The specific experimental setup is as follows:

[0069] (1) 0.5J lossless energy, repeat the experiment 5 times on the same carbon fiber laminate according to the above experimental steps;

[0070] (2) 5J lossy energy: Take 16 carbon fiber composite laminates T1 to T16, first impact the C1 center of T1, the C2 center of T2, ... the C16 center of T16, and collect the structural damage data at a single location. Then impact the C16, C15, C14, ... C1 centers of T1 to T16 in sequence as secondary damage structural response sensing data. Repeat the experiment 3 times. The 10J and 15J lossy energy impact experiment settings are the same as above.

[0071] In the specific implementation process, the acquisition process of the finite element simulation impact response signal is as follows:

[0072] The laminate was subjected to simulation analysis under the same conditions as the impact energy test to obtain the finite element simulation impact response signal.

[0073] Specifically, the process of the drop hammer impact finite element simulation experiment in this embodiment is as follows:

[0074] The composite laminate structure was modeled at a 1:1 scale using ABAQUS software. The carbon fiber composite laminate measures 400*400*3.2mm, with a single-layer thickness of 0.2mm and a density of ρ = 1520kg / m³. 3 The elastic modulus is E11 = 121 GPa; E12 = 8.6 GPa; the shear modulus is G12 = 5.9 GPa; G23 = 4.3 GPa; Poisson's ratio v = 0.13; a solid sphere made of aluminum alloy is used to simulate the spring impact hammer, with a Poisson's ratio v = 0.33 and a density ρ = 2700 kg / m³. 3 The elastic modulus E = 71 GPa and the shear modulus G = 27 GPa.

[0075] According to the work-energy theorem, the work done by an object with an initial velocity of 0 is:

[0076]

[0077] Given a solid sphere with a mass m of 2 kg (radius 13.3 cm), the velocities at W = 0.5 J, 5 J, 10 J, and 15 J are 0.707 m / s, 2.236 m / s, 3.162 m / s, and 3.873 m / s, respectively. The impact time is set to 0.5 s, and the data is collected in 500 frames. The structural displacement contour plots under different impact energies are shown below, with the impact location always set to the center. Figure 4 As shown.

[0078] The simulation experiment is set up as follows:

[0079] (1) 0.5J non-destructive energy impact, impact simulation experiment was carried out on the center points C1~C16 of the orthogonal grid in Figure 3(a);

[0080] (2) Lossy energy impacts of 5J, 10J, and 15J: The simulation analysis under lossy energy is divided into three steps: the initial analysis step is the original state of the structure when it is not subjected to load impact; analysis step 1 is a single-point load on Ci (i = 1, 2, ... 16) to simulate single-point damage; analysis step 2 takes the final state of the structure in analysis step 1 as the initial state and performs C... j (j=16,8,…1) were subjected to impact to simulate the structural response data under secondary damage.

[0081] The strain of the fourth layer of prepreg in the laminate at the actual grating area (5mm) covered by the sensor was collected. The average strain at all grating areas was taken as the simulated strain value, and a strain-time curve was plotted. The strain ε-wavelength λ was then used as the model. B τ B A linear formula is used to obtain a wavelength-time curve.

[0082]

[0083] In the formula, P e λ represents the strain optical sensitivity coefficient. B Δλ represents the center wavelength of the Bragg wavelength. B This represents the change in wavelength.

[0084] Step 2: Use the ensemble empirical mode-Euclidean distance similarity discrimination joint algorithm to extract the feature vectors of the actual impact response signal and the finite element simulation impact response signal.

[0085] In practical implementation, the raw data collected by the FBG sensor inevitably contains noise, such as baseline drift and power frequency, which can significantly interfere with damage identification. Therefore, eliminating noise interference in the response signal is essential. This invention employs Ensemble Empirical Mode Decomposition (EEMD), which requires no further prior knowledge, to decompose the raw signal into IMFs, and selects the IMF components with higher eigenvalues ​​as the effective signals based on the Euclidean distance similarity discrimination principle.

[0086] The decomposition principle of EEMD is as follows:

[0087] Adding normally distributed white noise n(t) to the original sensing signal x(t) yields a new signal f(t):

[0088] f(t)=x(t)+n(t) (3)

[0089] EMD decomposition is performed on the signal with added white noise to obtain the IMF component matrix IMF1 = [imf 11 imf 12….Imf 1m ] and the residual component r(t).

[0090]

[0091] Repeat the EMD decomposition process n times (n = 50~100) following the steps described above, adding a new white noise sequence before each decomposition to obtain n IMF component matrices [IMF1 IMF2…IMF]. n The average of the IMF values ​​obtained each time is used as the final result:

[0092]

[0093] IMF components are evaluated based on Euclidean distance: imf = [x1 x2 ... x ... ... q ] and the original signal W = [w1 w2….w q The higher the similarity coefficient, the more effective the IMF component.

[0094]

[0095] After normalizing the similarity coefficients, sort them from largest to smallest and sum them sequentially:

[0096]

[0097] ρ h =ρ1+ρ2+…+ρ h h = 2, 3...m (8)

[0098] The calculation stops when the sum of the similarity coefficients exceeds 0.9. The IMF values ​​corresponding to the similarity coefficients involved in the summation are taken as effective components and summed to obtain the final effective signal W. r =imf1+imf2+..+imf q .

[0099] Analysis shows that the impact response signal of the structure contains rich characteristic signals, so the processed wavelength-time curve of the FBG sensor can be used as a feature vector.

[0100] Step 3: Based on the eigenvectors of the actual impact response signal and the finite element simulation impact response signal, and using the limit learning machine regression prediction method, establish the nonlinear relationship between the eigenvectors and the damage type and location.

[0101] The eigenvectors of the actual impact response signal and the finite element simulation impact response signal are structural strain-time curves.

[0102] Step 4: Based on the nonlinear relationship between feature vectors and damage type and location, determine the damage identification model of the laminate structure and construct the laminate impact response database.

[0103] Specifically, the process of establishing the nonlinear relationship between feature vectors and damage type and location is as follows:

[0104] The eigenvectors of the finite element simulation impact response signal are used as the input matrix, and the corresponding eigenvectors of the actual impact response signal are used as the output matrix. The data feature learning model is trained to narrow the gap between the finite element simulation impact response signal and the actual impact response signal.

[0105] Based on the trained data feature learning model and the feature vector of the finite element simulation impact response signal, a set of data consisting of the feature vector, damage type, and location is output.

[0106] Among them, such as Figure 5 As shown, the Extreme Learning Machine (ELM) is a single-hidden-layer neural network. It does not require adjusting weight parameters during operation, has a fast learning speed, and effectively avoids local minima and overfitting. Therefore, it is widely used in regression fitting and classification problems. Its working principle is as follows:

[0107] Suppose there are N sample data (P, Q), where P = [x1 x2 ... x... N ] T The input matrix represents the strain information features, Q = [y1y2...y]. N ] T The expected output matrix represents the impact location and damage type. The network structure is as follows: Figure 5 As shown, n, m, and L represent the number of nodes in the input layer, the number of nodes in the output layer, and the number of neurons in the hidden layer, respectively.

[0108] Combination Figure 4 The relationship between the hidden layer and the output layer can be obtained as follows:

[0109]

[0110] In the formula: β represents the output weight matrix.

[0111] First, randomly generate the input weights ω = [ω1 ω2 ... ω L ] T And bias B = [b1 b2...b L ] T Then, in order to map the output layer data from its original space to the feature space of ELM, this paper chooses the Sigmoid function as the activation function g(x). This function performs well when the feature difference of the input matrix is ​​small, thus obtaining the output matrix H of the hidden layer. g,ω,b :

[0112]

[0113]

[0114] The relationship between the hidden layer and the output layer is as follows:

[0115]

[0116] Finally, to obtain the optimal solution for β on the training sample set, it is necessary to ensure the minimum limit of the training error. This is achieved by solving for β using the minimum approximate squared difference:

[0117] min||Hβ-Q|| 2 , β∈R (13)

[0118] Through relevant mathematical derivations, we can obtain:

[0119]

[0120]

[0121] In the formula, H + H represents the generalized inverse matrix of matrix H. T Let H be the transpose matrix, (H T H) -1 This represents the inverse matrix of HTH.

[0122] In summary, the simulation feature data is used as the input matrix, and the corresponding experimental feature data is used as the output matrix. The dataset includes 368 samples. During the experiment, training data and test data are randomly selected at a ratio of approximately 3:1, and these two are independent and have no overlap. After training, a data feature learning model is obtained.

[0123] Collect more simulation data under impact conditions. The specific impact points are selected as shown in Figure 3(b). Perform numerical simulations according to the experimental settings in step 3.

[0124] (1) A 0.5J lossless energy impact was applied to the center points W1 to W2 of the orthogonal grid in Figure 3(b). 169 Conduct impact simulation experiments;

[0125] (2) 5J, 10J, and 15J lossy energy impacts, first affecting W i (i = 1, 2… 169) Single-point loading is applied to simulate single-point damage; then, the final state of the structure in analysis step 1 is used as the initial state to apply W. j (j=169,99,…1) were subjected to impact to simulate the structural response data under secondary damage.

[0126] The strain-time of the structure after load impact is integrated and processed as input to the trained data feature learning model. The output of ELM is summarized into a database. Each set of data in the database consists of signal feature vectors and damage location and type (whether there is damage).

[0127] Example 2

[0128] Reference Figure 1 This embodiment provides a laminate loss identification method, including:

[0129] Obtain the impact response signal of the laminate and extract the corresponding feature vector;

[0130] Based on the damage identification model of laminate structure in the laminate impact response database, the damage type and location are obtained; wherein, the laminate impact response database is constructed using the laminate impact response database construction method described above.

[0131] In this embodiment, the feature vector extraction process can be performed using the ensemble empirical mode-Euclidean distance similarity discrimination joint algorithm as described above, and the specific process will not be repeated here.

[0132] Example 3

[0133] This embodiment provides a laminate impact response database construction device, which includes:

[0134] The impact response signal acquisition module is used to acquire the actual impact response signal and finite element simulation impact response signal of the laminate under various impact energies.

[0135] The feature vector extraction module is used to extract feature vectors from the actual impact response signal and the finite element simulation impact response signal using a combined algorithm of ensemble empirical mode-Euclidean distance similarity discrimination.

[0136] The nonlinear relationship construction module is used to establish a nonlinear relationship between the feature vector and the damage type and location based on the feature vector of the actual response impact signal and the finite element simulation impact response signal and the limit learning machine regression prediction method.

[0137] The impact response database construction module is used to determine the damage identification model of the laminate structure based on the nonlinear relationship between feature vectors and damage type and location, and to construct the impact response database of the laminate.

[0138] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0139] Example 4

[0140] This embodiment provides a laminate loss identification device, which includes:

[0141] The actual impact feature extraction module is used to acquire the actual impact response signal of the laminate and extract the corresponding feature vector.

[0142] The laminated plate loss identification module is used to obtain the damage type and location based on the damage identification model of the laminated plate structure in the laminated plate impact response database; wherein, the laminated plate impact response database is constructed using the laminated plate impact response database construction method described above.

[0143] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment two, and their specific implementation process is the same, so it will not be repeated here.

[0144] Example 5

[0145] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the laminate impact response database construction method described above.

[0146] It should be noted that the steps of the laminate impact response database construction method in this embodiment are the same as those in Embodiment 1, and the specific implementation process is the same, so it will not be repeated here.

[0147] Example 6

[0148] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the laminate loss identification method described above.

[0149] It should be noted that the steps of the laminate loss identification method in this embodiment are the same as those in Embodiment 2, and the specific implementation process is the same, so it will not be repeated here.

[0150] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0151] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing a laminate impact response database, characterized in that, include: Obtain the actual impact response signals and finite element simulation impact response signals of laminates under various impact energies; The feature vectors of the actual impact response signal and the finite element simulation impact response signal are extracted using a combined algorithm of ensemble empirical mode and Euclidean distance similarity discrimination. Based on the eigenvectors of actual impact response signals and finite element simulation impact response signals, and using the limit learning machine regression prediction method, a nonlinear relationship between eigenvectors and damage type and location is established. Based on the nonlinear relationship between feature vectors and damage type and location, a damage identification model for laminated plate structures is determined, and a laminated plate impact response database is constructed. The process of establishing the nonlinear relationship between feature vectors and damage type and location is as follows: The eigenvectors of the finite element simulation impact response signal are used as the input matrix, and the corresponding eigenvectors of the actual impact response signal are used as the output matrix. The data feature learning model is trained to narrow the gap between the finite element simulation impact response signal and the actual impact response signal. Based on the trained data feature learning model and the feature vector of the finite element simulation impact response signal, a set of data consisting of the feature vector, damage type, and location is output.

2. The method for constructing a laminate impact response database as described in claim 1, characterized in that, The process of obtaining the actual impact response signal is as follows: The actual impact response signal of the laminate under various impact energy tests is obtained based on the impact damage monitoring system; wherein, the impact damage monitoring system is constructed by embedding Y-shaped FBG sensors into the carbon fiber composite plate.

3. The method for constructing a laminate impact response database as described in claim 1 or 2, characterized in that, The process of obtaining the impact response signal in the finite element simulation is as follows: The laminate was subjected to simulation analysis under the same conditions as the impact energy test to obtain the finite element simulation impact response signal.

4. The method for constructing a laminate impact response database as described in claim 2, characterized in that, In the impact damage monitoring system, Y-shaped FBG sensors are arranged at the four vertices and the center of the laminate.

5. The method for constructing a laminate impact response database as described in claim 1, characterized in that, The eigenvectors of the actual impact response signal and the finite element simulation impact response signal are structural strain-time curves.

6. A method for identifying laminate loss, characterized in that, include: Obtain the impact response signal of the laminate and extract the corresponding feature vector; Based on the damage identification model of the laminate structure in the laminate impact response database, the damage type and location are obtained; wherein, the laminate impact response database is constructed using the laminate impact response database construction method as described in any one of claims 1-5.

7. A device for constructing a laminate impact response database, characterized in that, include: The impact response signal acquisition module is used to acquire the actual impact response signal and finite element simulation impact response signal of the laminate under various impact energies. The feature vector extraction module is used to extract feature vectors from the actual impact response signal and the finite element simulation impact response signal using a combined algorithm of ensemble empirical mode-Euclidean distance similarity discrimination. The nonlinear relationship construction module is used to establish a nonlinear relationship between the feature vector and the damage type and location based on the feature vector of the actual response impact signal and the finite element simulation impact response signal and the limit learning machine regression prediction method. The impact response database construction module is used to determine the damage identification model of the laminate structure based on the nonlinear relationship between feature vectors and damage type and location, and to construct the impact response database of the laminate. The process of establishing the nonlinear relationship between feature vectors and damage type and location is as follows: The eigenvectors of the finite element simulation impact response signal are used as the input matrix, and the corresponding eigenvectors of the actual impact response signal are used as the output matrix. The data feature learning model is trained to narrow the gap between the finite element simulation impact response signal and the actual impact response signal. Based on the trained data feature learning model and the feature vector of the finite element simulation impact response signal, a set of data consisting of the feature vector, damage type, and location is output.

8. A laminate loss identification device, characterized in that, include: The actual impact feature extraction module is used to acquire the actual impact response signal of the laminate and extract the corresponding feature vector. A laminated plate loss identification module is used to obtain the damage type and location based on the damage identification model of the laminated plate structure in the laminated plate impact response database; wherein, the laminated plate impact response database is constructed using the laminated plate impact response database construction method as described in any one of claims 1-5.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the laminate impact response database construction method as described in any one of claims 1-5; or When the processor executes the program, it implements the steps in the laminate loss identification method as described in claim 6.

Images