Method for monitoring internal mechanical state of carbon fiber wing in real time

Abstract

The invention discloses a method for monitoring the internal mechanical state of a carbon fiber wing in real time, and the method specifically comprises the steps: S1, collecting and obtaining the wing strain data of a carbon fiber wing skin in real time through an embedded sensor array; s2, performing one-way order reduction processing on the wing strain data to obtain carbon fiber wing damage characteristics; s3, obtaining wing stress data of each carbon fiber layer node of the carbon fiber wing according to the damage characteristics of the carbon fiber wing based on the constructed radial basis function proxy model; s4, training the constructed high-fidelity proxy model according to the wing stress data to obtain a state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing; and S5, constructing a digital twinning system according to the state evolution prediction model, and realizing visual real-time monitoring of the internal mechanical state of the carbon fiber wing according to the digital twinning system. The problems that when an existing structure health monitoring system faces a composite material complex structure, sensors are difficult to arrange, and damage recognition precision is insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft structural health monitoring, and particularly to a method for real-time monitoring of the internal mechanical state of a carbon fiber wing. Background Art

[0002] Due to its excellent properties such as high specific strength and high specific stiffness, carbon fiber composite materials are widely used in the wing structures of modern aircraft. However, during long-term service, especially under complex aerodynamic loads, internal defects such as delamination damage and matrix cracks are likely to occur in carbon fiber composite materials. These defects are characterized by strong concealment and fast propagation speed, and composite materials are extremely sensitive to micro-damage. Even a tiny defect may lead to a significant decline in structural performance.

[0003] During flight, when the aircraft encounters air turbulence, the wing bears severe alternating loads, which exacerbates the initiation and propagation of internal damage in the composite material. Due to the complex laminated structure of the carbon fiber composite wing, the internal stress distribution shows non-linear characteristics, there are obvious anisotropies between layers of composite materials, and the damage evolution process involves the coupling effect of multiple failure modes. In addition, a large deformation amplitude will also occur during the wing's swinging process, which requires a large amount of computational resources and data support when calculating mechanical properties. Traditional non-destructive testing techniques (such as ultrasonic testing, X-ray testing, etc.) are mainly used for ground maintenance and cannot achieve real-time monitoring during flight. Existing structural health monitoring systems have problems such as difficult sensor layout and insufficient damage identification accuracy when facing the complex structure of composite materials. Summary of the Invention

[0004] The present invention provides a method for real-time monitoring of the internal mechanical state of a carbon fiber wing to overcome the above technical problems.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] A method for real-time monitoring of the internal mechanical state of a carbon fiber wing specifically includes the following steps:

[0007] S1: Through the set embedded sensor array, the wing strain data of the carbon fiber wing skin is collected and obtained in real time;

[0008] S2: Perform one-way order reduction processing on the wing strain data to obtain the damage characteristics of the carbon fiber wing;

[0009] The damage characteristics of the carbon fiber wing include strain distribution characteristics, dynamic response characteristics, local characteristics, and strain pattern recognition characteristics;

[0010] And the strain distribution characteristics at least include the maximum strain value, strain gradient, and strain distribution pattern; the dynamic response characteristics at least include the frequency characteristics of strain changing with time; the local characteristics at least include local strain concentration and local strain outliers; the strain pattern recognition characteristics at least include principal component analysis characteristics and clustering analysis characteristics;

[0011] S3: Based on the constructed radial basis function surrogate model, obtain the wing stress data of each carbon fiber layer node of the carbon fiber wing according to the damage characteristics of the carbon fiber wing;

[0012] S4: Train the constructed high-fidelity surrogate model according to the wing stress data to obtain a state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing;

[0013] S5: Construct a digital twin system according to the state evolution prediction model, and realize the visual real-time monitoring of the mechanical state of the carbon fiber wing skin according to the digital twin system.

[0014] Further, the S1 specifically includes the following steps:

[0015] S11: Based on the finite element analysis method, construct a finite element simulation model of the carbon fiber wing according to the solid model of the carbon fiber wing;

[0016] S12: Use high-order elements to mesh the finite element simulation model of the carbon fiber wing to obtain the carbon fiber wing area mesh;

[0017] And the high-order elements include hexahedral elements or shell elements;

[0018] S13: By fixing the wing end of the finite element simulation model of the carbon fiber wing, simulate the connection constraint between the wing and the fuselage, and apply a simulated constant displacement load to the top of the wing of the finite element simulation model to simulate and obtain the simulated wing deformation data of the carbon fiber wing in the swinging state;

[0019] S14: According to the simulated wing deformation data, confirm the key mesh areas in the carbon fiber wing area mesh: the key mesh areas include high-stress mesh areas, easily damaged mesh areas, and aerodynamic load areas;

[0020] And the high-stress mesh areas at least include the wing root, wing beam connection, flap and aileron connection;

[0021] The easily damaged mesh areas at least include the wing leading edge, wing trailing edge, joints, and around the connection holes;

[0022] The aerodynamic load areas at least include the area near the wing tip on the upper surface of the wing;

[0023] S15: Conduct local meshing on the key mesh areas to obtain high-order meshes in the key areas;

[0024] An embedded sensor array is set according to the high-order grid of the key area for real-time collecting and obtaining the wing strain data inside the carbon fiber wing.

[0025] Further, the S2 specifically includes the following steps:

[0026] S21: Obtain the wing strain data of the high-order grid of the key area;

[0027] And based on the RBF interpolation method, map the wing strain data of the high-order grid of the key area to the preset low-order grid of the key area;

[0028] S22: Fit the wing strain data in the low-order grid of the key area through the RBF function, and construct an RBF feature function to obtain the eigenvalue of the wing strain data;

[0029] And the expression of the RBF feature function is

[0030]

[0031] In the formula: f(x, y, z) represents the eigenvalue of the wing strain data; (x, y, z) represents the node coordinates of the low-order grid of the key area; (x i , y i , z i ) represents the node coordinates of the high-order grid of the key area; ω i represents the weight coefficient determined by fitting the high-order grid data; represents the radial basis function; N represents the number of nodes of the high-order grid of the key area;

[0032] S23: Perform one-way reduction processing on the eigenvalue of the wing strain data based on the one-way reduction model to obtain the damage characteristics of the carbon fiber wing;

[0033] And the expression of the one-way reduction model is

[0034]

[0035] In the formula: represents the i-th eigenvalue sample after one-way reduction, that is, the discretized result obtained by integrating or sampling the output f(x, y, z) of the RBF feature function at the i-th layer of key nodes; X i represents the eigenvalue of the wing strain data; β ij represents the interlayer relationship between the wing strain data in the i-th low-order grid of the key area and the wing strain data in the j-th high-order grid of the key area; X j represents the wing strain data of the full-order nodes without processing.

[0036] Further, S3 specifically includes the following steps:

[0037] S31: Fit and simulate through the K-nearest neighbor algorithm to obtain the complex non-linear relationship between the damage characteristics of each carbon fiber wing, and interpolate the low-order grid of the key area based on the simulation results of the K-nearest neighbor algorithm fitting to expand the resolution of the carbon fiber wing damage characteristics and obtain the optimized carbon fiber wing damage characteristics;

[0038] S32: Based on the constructed radial basis function surrogate model, obtain the wing stress data of each carbon fiber layer node of the carbon fiber wing according to the optimized carbon fiber wing damage characteristics;

[0039] And the expression of the radial basis function surrogate model is

[0040]

[0041] In the formula: F(x) represents the predicted wing stress data of each carbon fiber layer node of the carbon fiber wing; n represents the number of radial basis functions; represents the radial basis function; w i represents the weight coefficient of the i-th RBF model; x represents the coordinate of an unknown point, that is, the node position to be predicted; x i represents the training sample point in the RBF model, that is, the node position of the known node coordinates and stress values; ||x - x i || represents the fixed point, that is, the Euclidean distance between the node position to be predicted and the source point, that is, the node position of the known node coordinates and stress values.

[0042] Further, S4 specifically includes the following steps:

[0043] S41: Construct a high-fidelity surrogate model based on the radial basis function surrogate model;

[0044] And the expression of the high-fidelity surrogate model is

[0045]

[0046] In the formula: represents the high-fidelity function prediction of the node position x to be predicted; ρ represents the trainable low-fidelity scaling factor; F L (x) represents the low-fidelity function prediction obtained according to the radial basis function surrogate model F(x); d(x) represents the difference function prediction;

[0047] S42: Take the material properties and applied loads of the carbon fiber wing as feature data, and the wing stress data as label data to obtain a sample data set;

[0048] And the material properties of the carbon fiber wing include at least the elastic modulus and Poisson's ratio;

[0049] Randomly divide the sample data set into a training set and a test set;

[0050] S43: Train the constructed high-fidelity surrogate model according to the training set to obtain the trained high-fidelity surrogate model:

[0051] S44: Based on the mean square error function as the model loss function, and evaluate the trained high-fidelity surrogate model according to the test set to determine whether the output of the trained high-fidelity surrogate model converges;

[0052] If so, the trained high-fidelity surrogate model at this time is the state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing;

[0053] Otherwise, adaptively adjust the model parameter weights of the trained high-fidelity surrogate model based on the backpropagation method, and repeat step S43; and the model parameters at least include the weight coefficients and basis function parameters of the basis function surrogate model, and the low-fidelity scaling factor.

[0054] Furthermore, the method for realizing the visual real-time monitoring of the internal mechanical state of the carbon fiber wing according to the digital twin system in S5 includes:

[0055] S51: Use the Unity software as the system building platform of the digital twin system, and the digital twin system built through the system building platform includes a carbon fiber wing physical space module, a communication unit, a digital space module, and a visual monitoring server;

[0056] And the carbon fiber wing physical space module includes a carbon fiber wing entity model and an embedded sensor array, and the embedded sensors include distributed optical fiber sensors and piezoelectric sensors;

[0057] The communication unit includes a serial communication module, a Socket communication program module, and a Redis server;

[0058] The carbon fiber wing physical space module is communicatively connected to the Redis server through the serial communication module;

[0059] The carbon fiber wing physical space module is communicatively connected to the digital space module through the Socket communication program module;

[0060] The digital space module realizes the prediction of the internal mechanical state of the carbon fiber wing according to the state evolution prediction model by calling the data stored or cached in the Redis server;

[0061] S52: Transmit the output data of the digital space module to the visual monitoring server;

[0062] And the visual monitoring server is provided with a data rendering module and a visualization module;

[0063] The predicted internal mechanical state data of the carbon fiber wing is rendered onto a preset solid model of the carbon fiber wing through a data rendering module, and the rendered solid model of the carbon fiber wing is visually displayed through a visualization module, thereby realizing the real-time visual monitoring of the internal mechanical state of the carbon fiber wing.

[0064] Beneficial effects: The present invention provides a method for real-time monitoring of the internal mechanical state of a carbon fiber wing. Through an embedded sensor array and one-way reduction processing, efficient acquisition and dimensionality reduction processing of the internal strain data of the carbon fiber wing are realized, significantly improving the efficiency and accuracy of data processing; based on the constructed radial basis function surrogate model and high-fidelity surrogate model, the internal mechanical state of the carbon fiber wing can be predicted according to the stress data of the carbon fiber wing skin, providing a reliable basis for damage identification and evaluation of the carbon fiber wing. In addition, through the constructed digital twin system, real-time visualization and dynamic monitoring of the wing mechanical state are realized, providing a strong technical guarantee for flight safety. Description of the Drawings

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0066] Figure 1 It is a flowchart of the method for real-time monitoring of the internal mechanical state of the carbon fiber wing of the present invention;

[0067] Figure 2 It is a schematic diagram of the embedding installation position of the sensor in this embodiment;

[0068] Figure 3 It is a stress distribution diagram of a layer of carbon fiber obtained by the finite element analysis method in this embodiment;

[0069] Figure 4 It is a stress nephogram predicted by the high-fidelity surrogate model in this embodiment;

[0070] Figure 5 It is a schematic diagram of the digital twin system interface in this embodiment. Detailed Embodiments

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0072] This embodiment provides a method for real-time monitoring of the internal mechanical state of a carbon fiber wing, as Figure 1 shown, which specifically includes the following steps:

[0073] S1: Through the provided embedded sensor array, wing strain data of the carbon fiber wing skin is collected and obtained in real time, which specifically includes the following steps:

[0074] S11: Based on the finite element analysis method, a finite element simulation model of the carbon fiber wing is constructed according to the solid model of the carbon fiber wing;

[0075] Specifically, in this embodiment, the material properties of the simulation model are set according to the carbon fiber material used in the existing solid model of the carbon fiber wing, and the anisotropic characteristics and boundary conditions of the material are fully considered. The ply angles of the finite element simulation model are set according to the ply angles of the wing solid (such as 0°, ±45°, 90°) to ensure that the material properties are consistent with the actual materials of the carbon fiber wing solid model; in addition, carbon fiber wings usually adopt a laminated plate structure, which is stacked by multiple layers of unidirectional carbon fiber prepregs. The fiber direction, thickness, and material properties of each layer may be different. The design of the laminated plate directly affects the mechanical properties of the wing. The number of laying layers, laying sequence, and laying angles need to be set according to the actual situation of the wing; the division of carbon fiber layer nodes is based on the structure of the laminated plate and the requirements of finite element modeling. Each layer of single layer is discretized into finite element meshes to define nodes or elements. The coordinates of each carbon fiber layer node include plane coordinates and thickness direction coordinates. Each carbon fiber layer node contains mechanical properties (such as displacement, stress, strain) and material properties (such as elastic modulus, fiber direction);

[0076] S12: The finite element simulation model of the carbon fiber wing is meshed using high-order elements to obtain the wing area mesh;

[0077] And the high-order elements include hexahedral elements or shell elements;

[0078] In the mesh division stage of this embodiment, to ensure the accuracy and reliability of the simulation results, high-order elements (such as hexahedral elements or shell elements) are used to discretize the wing model;

[0079] S13: Fix the wing tip of the finite element simulation model of the carbon fiber wing, simulate the connection constraint between the wing and the fuselage, and apply a simulated constant displacement load to the top of the wing of the finite element simulation model to simulate and obtain the simulated wing deformation data of the carbon fiber wing in the swinging state;

[0080] S14: According to the simulated wing deformation data, confirm the key grid areas in the carbon fiber wing area grid: The key grid areas include high-stress grid areas, easily damaged grid areas, and aerodynamic load areas;

[0081] And the high-stress grid area includes at least the wing root, the wing beam connection, the flap and aileron connection;

[0082] The easily damaged grid area includes at least the leading edge of the wing, the trailing edge of the wing, the joint, and the area around the connection hole;

[0083] The aerodynamic load area includes at least the area near the wing tip on the upper surface of the wing;

[0084] This embodiment also includes obtaining the corresponding wing stress nephogram according to the simulated wing deformation data. As Figure 3 shown, it can intuitively display the high-stress area and easily damaged area of the wing, which is convenient for extracting the stress distribution data of the key area and used to optimize the layout scheme of the sensors;

[0085] S15: Perform local mesh division on the key grid areas to obtain high-order grids in the key areas;

[0086] And set an embedded sensor array according to the high-order grids in the key areas to be used for real-time collection and obtaining the wing strain data inside the carbon fiber wing.

[0087] Specifically, this embodiment can measure the phase change of the optical signal in the optical fiber of the distributed optical fiber sensor in the arranged embedded sensor array, monitor the strain distribution inside the wing in real time, identify local stress concentration areas, and realize continuous monitoring of the overall structural health state of the wing. According to the structural characteristics and force characteristics of the wing, and the specific embedding layout rules of the distributed optical fiber sensor:

[0088] (1) Arranged in the wing-fuselage connection area, that is, the wing root:

[0089] Lay 3 - 5 optical fibers along the wing beam direction to form a high-density monitoring network. The spacing between each optical fiber is 10 - 20 cm, the layout spacing of the optical fiber sensor is 5 - 10 cm, and the length of each optical fiber should cover the entire area of the wing root. Through embedded installation, ensure that the optical fiber is closely attached to the wing surface. The purpose is to monitor the high-stress distribution at the root and identify possible fatigue damage;

[0090] (2) Arranged on the upper and lower surfaces of the main wing beam:

[0091] One or two optical fibers are respectively arranged on the upper surface and the lower surface of the main wing beam to form a symmetric monitoring network, which is arranged along the wingspan direction. The arrangement spacing of the optical fiber sensors is 10 cm. The spacing can be reduced to 5 cm at the connection between the wing beam and the wing skin and in the wing beam joint area. One optical fiber is arranged on the lateral stiffener of the main wing beam or the auxiliary wing beam along the direction of the stiffener to capture the bending and torsional strains of the wing beam and assist in identifying the torsional strain;

[0092] (3) Arranged on the leading edge (windward side) and the trailing edge (the area near the flap and aileron) of the wing:

[0093] Two or three optical fibers are respectively arranged along the arc paths of the leading edge and the trailing edge. The arrangement spacing of the optical fiber sensors is 10 - 15 cm. The spacing can be reduced to 5 cm at the connection between the leading edge and the wing skin and at the connections between the trailing edge and the flap and aileron. The purpose is to monitor the local stress concentration caused by the aerodynamic load;

[0094] (4) Arranged in the skin areas on the upper surface and the lower surface of the wing:

[0095] A grid layout is adopted to form a crisscross optical fiber network. The arrangement spacing of the optical fiber sensors is 20 - 30 cm. Three to five optical fibers are arranged on each of the upper and lower surfaces. The purpose is to achieve full - coverage monitoring of the wing skin and capture the overall strain distribution;

[0096] In this embodiment, piezoelectric sensors in the arranged embedded sensor array are used to capture high - frequency vibration and acoustic emission signals, realize active monitoring of local damage of the wing and dynamic response analysis. According to the dynamic characteristics and damage - sensitive areas, the specific embedding rules of the piezoelectric sensors are as follows:

[0097] (1) Arranged in the connection areas between the flap and aileron and the wing body:

[0098] Four to six piezoelectric sensors are arranged in a circular array around the connection to form a closed monitoring loop. The arc spacing between the sensors is 5 - 10 cm. Two to three additional piezoelectric sensors are arranged radially around the connection bolts or rivets, and the spacing is reduced to 2 - 5 cm to cover the stress conduction path. The purpose is to capture the high - frequency acoustic emission signals generated by the loosening of the connectors and the generation of micro - cracks.

[0099] (2) Arranged in the wing tip part of the wing:

[0100] Three to four piezoelectric sensors are arranged along the arc structure of the wing tip with a spacing of 10 - 15 cm to form a vibration monitoring network. One triaxial piezoelectric accelerometer is arranged at the connection between the wing tip and the wing skin, and the spacing is reduced to 5 cm. Monitor the wing tip flutter (1 - 500 Hz) and the high - frequency vibration (>1 kHz) caused by the aerodynamic load, and assist in identifying the aeroelastic instability state.

[0101] (3) Arranged in the connection area between the main spar joint and the fuselage:

[0102] Four piezoelectric sensors are arranged around the spar-fuselage docking flange, symmetrically distributed in a cross shape, with a spacing of 8 - 12 cm. One micro piezoelectric sheet (diameter < 5 mm) is arranged around each connection bolt hole, with a spacing of 2 - 3 cm. The purpose is to monitor the fretting wear and impact events of the joint.

[0103] This embodiment is provided with a differential amplification circuit module connected to the embedded sensor array to amplify the signals collected by the sensors through the differential amplification circuit, and use a preset analog-to-digital conversion module to convert the analog signals into digital signals, and then transmit them to the computer in real time through the Redis server, and synchronize the data of the distributed fiber optic sensors and piezoelectric sensors through timestamps to ensure the consistency of multi-source data;

[0104] S2: Perform one-way order reduction processing on the wing strain data to obtain the damage characteristics of the carbon fiber wing;

[0105] The damage characteristics of the carbon fiber wing include strain distribution characteristics, dynamic response characteristics, local characteristics, and strain pattern recognition characteristics;

[0106] And the strain distribution characteristics at least include the maximum strain value, strain gradient, and strain distribution pattern; the dynamic response characteristics at least include the frequency characteristics of the strain changing with time; the local characteristics at least include the local strain concentration and local strain outliers; the strain pattern recognition characteristics at least include the principal component analysis characteristics and clustering analysis characteristics;

[0107] In this embodiment, the characteristic function is extracted from the full-order model through a one-way order reduction model, simplifying the model and extracting the characteristic information related to damage. The full-order model is a high-precision and high-resolution mechanical modeling and simulation of the carbon fiber wing structure based on finite element analysis, including the distribution and change of physical quantities such as stress, strain, and displacement, and can describe the mechanical behavior of the wing in detail;

[0108] Specifically, it includes the following steps:

[0109] S21: Obtain the wing strain data of the high-order grid in the key area;

[0110] And based on the RBF interpolation method, map the wing strain data of the high-order grid in the key area to the preset low-order grid in the key area;

[0111] S22: Fit the wing strain data in the low-order grid of the key area through the RBF function, construct the RBF characteristic function to obtain the characteristic values of the wing strain data;

[0112] In this embodiment, the feature function is a mathematical function extracted from the full-order model, which is used to describe the key features related to damage. First, the RBF interpolation method is used to map the strain data of the high-order grid to the low-order grid, and then the RBF feature function is obtained by fitting the strain data of the low-order grid nodes with the RBF function;

[0113] And the expression of the RBF feature function is

[0114]

[0115] In the formula: f(x, y, z) represents the eigenvalue of the wing strain data, that is, the continuous space function generated according to the RBF interpolation, which is used to fit the characteristic distribution of the high-order strain data at the low-order grid nodes (x, y, z), and the output value is the strain eigenvalue at this position; (x, y, z) represents the node coordinates of the low-order grid in the key area; (x i , y i , z i ) represents the node coordinates of the high-order grid in the key area; ω i represents the weight coefficient determined by fitting the high-order grid data; represents the radial basis function; N represents the number of nodes in the high-order grid in the key area;

[0116] S23: Perform one-way reduction processing on the eigenvalues of the wing strain data based on the one-way reduced-order model to obtain the damage characteristics of the carbon fiber wing; in this embodiment, the data in the input layer of the single-item reduced-order model only affects the output layer unidirectionally, and there is no coupling relationship where the output affects the input conversely, which can ensure the independence of the materials of each carbon fiber layer of the carbon fiber wing;

[0117] And the expression of the one-way reduced-order model is

[0118]

[0119] In the formula: represents the i-th eigenvalue sample after one-way reduction; X i represents the eigenvalue of the wing strain data; β ij represents the interlayer relationship between the wing strain data in the i-th low-order grid in the key area and the wing strain data in the j-th high-order grid in the key area; X j represents the wing strain data of the full-order nodes obtained from simulation and experiment without processing; is the discrete feature vector output by the one-way reduced-order model, and each element corresponds to a damage feature after dimensionality reduction; Each eigenvalue of is the discretization result obtained by integrating or sampling f(x, y, z) in a specific area (such as the key nodes in the i-th layer);

[0120] S3: Based on the constructed radial basis function surrogate model, obtain the wing stress data of each carbon fiber layer node of the carbon fiber wing according to the damage characteristics of the carbon fiber wing;

[0121] Specifically, it includes the following steps:

[0122] S31: Fit and simulate through the K-nearest neighbor algorithm to obtain the complex non-linear relationship between the damage characteristics of each carbon fiber wing, and interpolate the low-order grids in the key areas based on the simulation results of the K-nearest neighbor algorithm fitting to improve the resolution of the carbon fiber wing damage characteristics and the prediction accuracy of the model, and obtain the optimized carbon fiber wing damage characteristics;

[0123] Before fitting and simulating through the K-nearest neighbor algorithm to obtain the complex non-linear relationship between the damage characteristics of each carbon fiber wing in this embodiment, it also includes data processing of the damage characteristics of the carbon fiber wing based on the finite element analysis method, which specifically includes:

[0124] S001: Establish a three-dimensional geometric model of the carbon fiber wing based on the finite element analysis method, import it into the finite element analysis software ANSYS, and obtain the non-repeated nodes and node indexes of each carbon fiber layer through the finite element. The coordinates (x, y, z) of each carbon fiber layer node are uniquely present in the three-dimensional geometric model. If multiple elements share the same node coordinates, only one node is retained; in the finite element analysis software ANSYS, through the node merging function, nodes with the same coordinates can be merged into one non-repeated node; the node index is the unique identifier of each carbon fiber layer node in the finite element model, which is used to quickly locate and access node data during calculations; the node index starts from 1 and increases sequentially. The node index of each carbon fiber layer node has nothing to do with its position in the three-dimensional geometric model, and only represents its order in the nodes;

[0125] S002: Extract the grid data in the finite element simulation, that is, the damage characteristics of the carbon fiber wing, magnify and remove the repeated nodes, match the node indexes, and perform data calibration and grid repair, and finally save the optimized carbon fiber layer node and index data; specifically, numerically magnify the node coordinates through a Python script; use the drop_duplicates method in the Pandas library to remove the repeated nodes; use the merge method in the Pandas library to match the node indexes to calibrate the carbon fiber wing damage characteristic data; call the PyVista tool and the PyMeshFix tool for grid repair, including removing invalid elements, filling holes, and smoothing the grid, etc.;

[0126] S003: Traverse and read the characteristic data such as the node coordinates, stress distribution, and strain distribution of each carbon fiber layer node after data processing, that is, the carbon fiber wing damage characteristic data, to provide basic data support for subsequent analysis;

[0127] S32: Based on the constructed radial basis function surrogate model, obtain the wing stress data of each carbon fiber layer node of the carbon fiber wing according to the optimized damage characteristics of the carbon fiber wing;

[0128] And the expression of the radial basis function surrogate model is

[0129]

[0130] In the formula: F(x) represents the predicted wing stress data of each carbon fiber layer node of the carbon fiber wing; n represents the number of radial basis functions; represents the radial basis function; w i represents the weight coefficient of the i-th RBF model; x represents the coordinate of an unknown point, that is, the node position to be predicted; x i represents the training sample point in the RBF model, that is, the node position of the known node coordinates and stress values; ||x - x i || represents the fixed point, that is, the Euclidean distance between the node position to be predicted and the general source, that is, the node position of the known node coordinates and stress values;

[0131] S4: Train the constructed high-fidelity surrogate model according to the wing stress data to obtain a state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing;

[0132] In this embodiment, the high-fidelity surrogate model can accurately predict the internal mechanical state of the carbon fiber wing by fusing the high-fidelity data measured by experiments and the low-fidelity data obtained by simulation calculations; in addition, based on the measured historical data, the model can dynamically update the training parameters, further improving the accuracy and robustness of the prediction. The high-fidelity surrogate model can accurately reflect the non-linear mechanical behavior of carbon fiber composites through the joint training of finite element simulation and experimental data;

[0133] Specifically, it includes the following steps:

[0134] S41: Construct a high-fidelity surrogate model based on the radial basis function surrogate model;

[0135] And the expression of the high-fidelity surrogate model is

[0136]

[0137] In the formula: represents the high-fidelity function prediction of the node position x to be predicted; ρ represents the trainable low-fidelity scaling factor; F L(x) represents the low-fidelity function prediction obtained according to the radial basis function surrogate model F(x); d(x) represents the difference function prediction; where F(x) represents the direct output of the RBF surrogate model, i.e., the radial basis function surrogate model, which predicts the value of the unknown point x based on the known node data of simulation and experiment, and is used to provide the prediction of the global initial stress field; F L (x) is the prediction function of the low-fidelity data source, usually from fast but low-precision simulations (such as coarse-grid finite element analysis) or simplified physical models, and its role is to quickly generate baseline predictions; in the low-fidelity prediction stage, the RBF form of F(x) is used to quickly generate the global stress distribution F L (x); in this way, the stress data of the carbon fiber wing can be quickly obtained, and the progressive optimization from low-cost approximation to high-fidelity prediction can be realized;

[0138] S42: Take the material properties and applied loads of the carbon fiber wing as feature data, and the wing stress data as label data to obtain a sample data set;

[0139] This embodiment also includes preprocessing the sample data set: including normalization, denoising, and data augmentation, to improve the generalization ability and robustness of the model, and ensure that the model can adapt to different data distributions and noise interferences;

[0140] And the material properties of the carbon fiber wing at least include elastic modulus and Poisson's ratio;

[0141] Randomly divide the sample data set into a training set and a test set;

[0142] S43: Train the constructed high-fidelity surrogate model according to the training set to obtain the trained high-fidelity surrogate model:

[0143] S44: Based on the mean square error function as the model loss function, and evaluate the trained high-fidelity surrogate model according to the test set to judge whether the output of the trained high-fidelity surrogate model converges;

[0144] If so, the trained high-fidelity surrogate model at this time is the state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing;

[0145] Otherwise, adaptively adjust the model parameter weights of the trained high-fidelity surrogate model based on the backpropagation method, and repeat step S43; and the model parameters at least include the weight coefficients and basis function parameters of the basis function surrogate model, the number of neighbors / distance metric in the KNN algorithm, and the low-fidelity scaling factor for adjusting the weights between the low-fidelity model and the high-fidelity model;

[0146] Based on the measured historical data, this embodiment dynamically updates the training parameters of the high-fidelity proxy model to ensure that the model can adapt to the mechanical behaviors under different flight states and environmental conditions, and further improves the prediction accuracy and adaptability of the model by continuously optimizing the model parameters;

[0147] S5: Construct a digital twin system based on the state evolution prediction model, and realize the visual real-time monitoring of the mechanical state of the carbon fiber wing skin according to the digital twin system;

[0148] Specifically, as Figures 4 to 5 shown, the method for realizing the visual real-time monitoring of the internal mechanical state of the carbon fiber wing according to the digital twin system includes:

[0149] S51: Use the Unity software as the system construction platform of the digital twin system, and the digital twin system constructed through the system construction platform includes a carbon fiber wing physical space module, a communication unit, a digital space module, and a visual monitoring server;

[0150] And the carbon fiber wing physical space module includes a carbon fiber wing entity model and an embedded sensor array, and the embedded sensors include distributed optical fiber sensors and piezoelectric sensors; as Figure 2 shown, A represents a distributed optical fiber sensor; B represents a piezoelectric sensor;

[0151] The communication unit includes a serial port communication module, a Socket communication program module, and a Redis server;

[0152] The carbon fiber wing physical space module is communicatively connected to the Redis server through the serial port communication module;

[0153] The carbon fiber wing physical space module is communicatively connected to the digital space module through the Socket communication program module; specifically, the Socket communication program module is used to realize the real-time data interaction between the carbon fiber wing physical space module and the digital space module, and uses the TCP protocol for streaming communication to ensure the stability and reliability of data transmission;

[0154] The digital space module predicts the internal mechanical state of the carbon fiber wing according to the state evolution prediction model by calling the data stored or cached in the Redis server; specifically, the digital space model processes the received data through the KNN algorithm, the unidirectional reduction model, the radial basis function proxy model, and the high-fidelity proxy model, and predicts the stress distribution, damage location, and evolution trend inside the wing;

[0155] S52: Transmit the output data of the digital space module to the visual monitoring server;

[0156] Moreover, the visualization monitoring server is provided with a data rendering module and a visualization module;

[0157] Through the data rendering module, the predicted internal mechanical state data of the carbon fiber wing is rendered onto a preset solid model of the carbon fiber wing, and through the visualization module, the rendered solid model of the carbon fiber wing is visually displayed, thereby realizing the visual real-time monitoring of the internal mechanical state of the carbon fiber wing.

[0158] The working principle of the digital twin system in this embodiment: The data collected by the sensor is transmitted to the computer in real time through the serial port and stored in the Redis server; The digital twin system uses the TCP protocol for streaming communication, and sends the data in the Redis server to the high-fidelity proxy model in the background application through Socket two-way communication, and returns the trained prediction data to the digital twin system. Finally, the real-time processing and visual display of the data are realized through the UI design. The background application uses python to package the background running programs such as models and algorithms to realize the encapsulation and independent operation of the background programs.

[0159] In summary, the beneficial effects of the method described in this embodiment: Through the embedded sensor array and one-way reduction processing, this embodiment realizes the efficient acquisition and dimensionality reduction processing of the internal strain data of the carbon fiber wing, significantly improving the efficiency and accuracy of data processing; Based on the constructed radial basis function proxy model and high-fidelity proxy model, it is possible to predict the internal mechanical state of the carbon fiber wing according to the stress data of the carbon fiber wing skin, providing a reliable basis for the damage identification and evaluation of the carbon fiber wing. In addition, through the constructed digital twin system, the real-time visualization and dynamic monitoring of the wing mechanical state are realized, providing a strong technical guarantee for flight safety.

[0160] Finally, it should be noted that: The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: They can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for real-time monitoring of the internal mechanical state of a carbon fiber wing, characterized in that, Specifically, it includes the following steps: S1: Through the set embedded sensor array, wing strain data of the carbon fiber wing skin is collected and obtained in real time; S2: Perform one-way order reduction processing on the wing strain data to obtain the damage characteristics of the carbon fiber wing; The damage characteristics of the carbon fiber wing include strain distribution characteristics, dynamic response characteristics, local characteristics, and strain pattern recognition characteristics; And the strain distribution characteristics at least include the maximum strain value, strain gradient, and strain distribution pattern; the dynamic response characteristics at least include the frequency characteristics of the strain changing with time; the local characteristics at least include local strain concentration and local strain outliers; the strain pattern recognition characteristics at least include principal component analysis characteristics and clustering analysis characteristics; S3: Based on the constructed radial basis function surrogate model, wing stress data of each carbon fiber layer node of the carbon fiber wing is obtained according to the damage characteristics of the carbon fiber wing; S4: Train the constructed high-fidelity surrogate model according to the wing stress data to obtain a state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing; S5: Construct a digital twin system according to the state evolution prediction model, and realize visual real-time monitoring of the internal mechanical state of the carbon fiber wing according to the digital twin system.

2. The method for real-time monitoring of the internal mechanical state of a carbon fiber wing according to claim 1, characterized in that The specific steps of S1 are as follows: S11: Based on the finite element analysis method, a finite element simulation model of the carbon fiber wing is constructed according to the solid model of the carbon fiber wing; S12: Use high-order elements to mesh the finite element simulation model of the carbon fiber wing to obtain the wing area mesh; And the high-order elements include hexahedral elements or shell elements; S13: By fixing the wing end of the finite element simulation model of the carbon fiber wing, simulating the connection constraint between the wing and the fuselage, and applying a simulated constant displacement load to the wing top of the finite element simulation model, the simulated wing deformation data of the carbon fiber wing in the swinging state is simulated and obtained; S14: According to the simulated wing deformation data, confirm the key grid areas in the wing area mesh of the carbon fiber wing: the key grid areas include high-stress grid areas, vulnerable grid areas, and aerodynamic load areas; And the high-stress grid areas at least include the wing root, wing beam connection, flap, and aileron connection; The vulnerable grid areas at least include the wing leading edge, wing trailing edge, joint, and around the connection hole; The aerodynamic load areas at least include the area near the wing tip on the upper surface of the wing; S15: Perform local meshing on the key grid areas to obtain high-order grids in the key areas; And an embedded sensor array is set according to the high-order grids in the key areas to be used for collecting and obtaining the wing strain data of the carbon fiber wing skin in real time.

3. A method for real-time monitoring of the internal mechanical state of a carbon fiber wing according to claim 2, characterized in that, The specific steps of S2 are as follows: S21: Obtain the wing strain data of the high-order grids in the key areas; And based on the RBF interpolation method, map the wing strain data of the high-order grids in the key areas to the preset low-order grids in the key areas; S22: Fit the wing strain data in the low-order grids in the key areas through the RBF function, and construct an RBF feature function to obtain the eigenvalue of the wing strain data; And the expression of the RBF feature function is where: f(x, y, z) represents the eigenvalue of the wing strain data; (x, y, z) represents the node coordinates of the low-order grid in the key area; (x i , y i , z i ) represents the node coordinates of the high-order grid in the key area; ω i represents the weight coefficient determined by fitting the high-order grid data; represents the radial basis function; N represents the number of nodes in the high-order grid of the key area; S23: Perform one-way reduced-order processing on the eigenvalue of the wing strain data based on the one-way reduced-order model to obtain the damage characteristics of the carbon fiber wing; And the expression of the one-way reduced-order model is In the formula: represents the i-th eigenvalue sample after one-way order reduction, that is, the discretized result obtained by integrating or sampling the output f(x, y, z) of the RBF feature function at the i-th layer of key nodes; X i represents the eigenvalue of the wing strain data; β ij represents the interlayer relationship between the wing strain data in the low-order grid of the i-th key area and the wing strain data in the high-order grid of the j-th key area; X j represents the wing strain data of the full-order nodes without processing.

4. A method for real-time monitoring of the internal mechanical state of a carbon fiber wing according to claim 3, characterized in that, The specific steps of S3 are as follows: S31: Fit and simulate through the K-nearest neighbor algorithm to obtain the complex non-linear relationship between the damage characteristics of each carbon fiber wing, and interpolate the low-order grid in the key area based on the fitting and simulation results of the K-nearest neighbor algorithm to expand the resolution of the damage characteristics of the carbon fiber wing and obtain the optimized damage characteristics of the carbon fiber wing; S32: Based on the constructed radial basis function surrogate model, obtain the wing stress data of each carbon fiber layer node of the carbon fiber wing according to the optimized damage characteristics of the carbon fiber wing; And the expression of the radial basis function surrogate model is Where: F(x) represents the wing stress data of each carbon fiber layer node of the predicted carbon fiber wing; n represents the number of radial basis functions; represents the radial basis function; w i represents the weight coefficient of the i-th RBF model; x represents the coordinate of an unknown point, that is, the node position to be predicted; x i represents the training sample point in the RBF model, that is, the node position of the known node coordinates and stress values; ||x - x i || represents the fixed point, that is, the Euclidean distance between the node position to be predicted and the generalization source, that is, the node position of the known node coordinates and stress values.

5. A method for real-time monitoring of the internal mechanical state of a carbon fiber wing according to claim 4, characterized in that, The specific steps of S4 are as follows: S41: Construct a high-fidelity surrogate model based on the radial basis function surrogate model; And the expression of the high-fidelity surrogate model is In the formula: represents the high-fidelity function prediction of the node position x to be predicted; ρ represents the trainable low-fidelity scaling factor; F L (x) represents the low-fidelity function prediction obtained according to the radial basis function surrogate model F(x); d(x) represents the difference function prediction; S42: Use the material properties and applied loads of the carbon fiber wing as feature data, and the wing stress data as label data to obtain a sample data set; And the material properties of the carbon fiber wing at least include elastic modulus and Poisson's ratio; Randomly divide the sample data set into a training set and a test set; S43: Train the constructed high-fidelity surrogate model according to the training set to obtain the trained high-fidelity surrogate model: S44: Based on the mean square error function as the model loss function, and evaluate the trained high-fidelity surrogate model according to the test set to determine whether the output of the trained high-fidelity surrogate model converges; If so, the trained high-fidelity surrogate model at this time is the state evolution prediction model for predicting the internal mechanical state of the carbon fiber wing; Otherwise, adaptively adjust the model parameter weights of the trained high-fidelity surrogate model based on the backpropagation method, and repeat step S43; and the model parameters at least include the weight coefficients and basis function parameters of the basis function surrogate model, and the low-fidelity scale factor.

6. A method for real-time monitoring of the internal mechanical state of a carbon fiber wing according to claim 5, characterized in that, The method for realizing visual real-time monitoring of the internal mechanical state of the carbon fiber wing according to the digital twin system in S5 includes: S51: Use the Unity software as the system construction platform of the digital twin system, and the digital twin system constructed through the system construction platform includes a carbon fiber wing physical space module, a communication unit, a digital space module, and a visual monitoring server; And the carbon fiber wing physical space module includes a carbon fiber wing entity model and an embedded sensor array, and the embedded sensors include distributed optical fiber sensors and piezoelectric sensors; The communication unit includes a serial communication module, a Socket communication program module, and a Redis server; The carbon fiber wing physical space module is communicatively connected to the Redis server through the serial communication module; Communicatively connect the carbon fiber wing physical space module and the digital space module through the Socket communication program module; The digital space module realizes the prediction of the internal mechanical state of the carbon fiber wing according to the state evolution prediction model by calling the data stored or cached by the Redis server; S52: Transmit the output data of the digital space module to the visual monitoring server; Moreover, the visualization monitoring server is equipped with a data rendering module and a visualization module; The predicted internal mechanical state data of the carbon fiber wing is rendered onto a preset carbon fiber wing solid model through the data rendering module, and the rendered carbon fiber wing solid model is visually displayed through the visualization module, thereby realizing the visual real-time monitoring of the internal mechanical state of the carbon fiber wing.

Images