An antenna array optical fiber grating layout method based on genetic algorithm and neural network

Abstract

The application provides an antenna array surface fiber grating layout method based on a genetic algorithm and a neural network, and relates to the technical field of deformation sensing of an antenna array surface structure and optimization of fiber grating layout. The method analyzes strain sensitive positions of the array surface structure through the genetic algorithm and the neural network, arranges fiber grating sensors at optimal layout positions of the array surface structure, and realizes high-precision deformation sensing of the array surface structure under complex service loads. The method solves the problems of node layout redundancy, high data processing cost and insufficient displacement prediction accuracy in traditional antenna monitoring, and provides technical support for efficient sensing of antenna structure deformation and stable guarantee of electromagnetic performance.

Description

Technical Field

[0001] This invention relates to the field of deformation sensing and fiber optic grating layout optimization technology for antenna array structures, and particularly to an antenna array fiber optic grating layout method based on genetic algorithms and neural networks. Background Technology

[0002] In actual service environments, wind loads, overloads, random vibrations, and other loads can cause structural deformation of the antenna array, leading to displacement of array elements and directly affecting the electromagnetic performance of the antenna. The fiber optic grating monitoring system for the array structure is the core means of sensing array deformation. Its monitoring accuracy is highly related to the layout design of the grating. A reasonable layout can maximize the coverage of key deformation areas of the array with a limited number of sensors, providing reliable data support for subsequent structural reconstruction and electrical performance compensation.

[0003] Existing methods for laying out fiber optic gratings in antenna arrays have two limitations:

[0004] On the one hand, traditional empirical layout methods (such as uniform distribution or arrangement according to structural nodes) do not take into account the deformation characteristics of the array under actual service loads, which can easily lead to blind spots in key areas of monitoring, or increase costs and data redundancy due to sensor redundancy. On the other hand, particle swarm optimization algorithms rely on a large amount of deformation simulation / measured data of the array structure to build a detection area model. However, the high-precision data sample size of the antenna array is small and the acquisition cost is high. In small sample scenarios, the reliability of the model will drop significantly, and the layout optimization results will deviate greatly from the actual working conditions.

[0005] On the other hand, some scenarios use neural networks to optimize fiber optic grating layouts. However, a single neural network lacks global search capabilities and is easily limited to the layout range covered by training data, making it unable to escape local optima. Furthermore, when faced with complex load combinations not seen during service, the model has weak generalization ability, and the optimized layout is difficult to adapt to actual deformation monitoring needs.

[0006] Currently, there are studies on the optimization of fiber optic sensor network layout based on genetic algorithms. However, this method lacks the ability to learn autonomously about nonlinear mapping relationships in complex scenarios. When faced with the deformation monitoring requirements of small sample size and high coupling of antenna arrays, it is difficult to accurately fit the correlation between "layout and monitoring accuracy". Summary of the Invention

[0007] Purpose of the invention: The purpose of this invention is to provide a fiber optic grating layout method for antenna arrays based on genetic algorithms and neural networks. By integrating finite element simulation data with intelligent algorithms (including genetic algorithms and neural networks), it achieves accurate selection and optimal layout design of strain-sensitive nodes in the antenna array, solving problems such as redundant node layout, high data processing costs, and insufficient displacement prediction accuracy in traditional antenna monitoring. This provides technical support for efficient sensing of antenna structural deformation and stable electromagnetic performance.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for laying out fiber optic gratings in an antenna array based on genetic algorithms and neural networks includes the following steps:

[0010] A finite element model of the antenna array structure was constructed, and different constraints and loads were set to simulate and obtain low-fidelity strain-displacement data of the antenna array structure under different service conditions.

[0011] Based on preset screening criteria, strain-sensitive nodes are selected to obtain a set of strain-sensitive nodes;

[0012] The strain-sensitive node set is encoded, and the population is initialized based on random sampling rules to generate a preset number of initial individuals, with each individual corresponding to a set of strain-sensitive nodes;

[0013] Define a fitness function that balances deformation representation capability with node quantity economy;

[0014] Genetic operations are performed on the initialized population, and the process is iterated until the population converges. The strain-sensitive node test group with fitness that meets the preset requirements is output.

[0015] Construct the actual antenna array structure, deploy fiber optic strain sensors at the locations corresponding to the strain-sensitive node set, establish the coordinate system of the array structure, load different types of loads onto the actual antenna array structure, and obtain high-fidelity strain-displacement data.

[0016] The low-fidelity strain-displacement data and the high-fidelity strain-displacement data are subjected to strain data decomposition and recombination and displacement data decomposition processing, respectively, to obtain strain image data and displacement component datasets.

[0017] A deformation prediction model is constructed by fitting the mapping relationship between strain image data and displacement component dataset corresponding to low-fidelity strain-displacement data using a neural network.

[0018] Based on the strain-sensitive node test group, the strain image data under its high-fidelity strain-displacement data is input into the deformation prediction model to obtain the predicted displacement data of each strain-sensitive node test group.

[0019] The deviation between the predicted displacement data and the measured displacement data is calculated. The strain-sensitive node corresponding to the strain-sensitive node test group with the smallest deviation value is selected as the optimal strain-sensitive node. Its spatial coordinate information is output as the optimized layout position of the antenna array fiber optic grating sensor.

[0020] As a preferred option, strain-sensitive nodes are selected based on preset screening criteria, which include strain fluctuation amplitude criteria and displacement correlation criteria. The criteria also satisfy that the sensitive nodes cover the key area of ​​the array surface and the spatial distance between any two sensitive nodes is not less than a preset threshold, thus obtaining a set of strain-sensitive nodes.

[0021] As a preferred embodiment, the strain fluctuation amplitude criterion is the extreme difference of strain at the array nodes under various service conditions. :

[0022]

[0023] in Let be the strain set of the i-th node under different working conditions; the screening threshold is .

[0024] As a preferred embodiment, the displacement correlation criterion uses the Pearson correlation coefficient to quantify the linear correlation between nodal strain and overall surface deformation:

[0025]

[0026] in The resultant value of the triaxial displacement of the array's centroid; the screening threshold is... .

[0027] As a preferred embodiment, the sensitive nodes cover the key areas of the array surface and the spatial distance between any two sensitive nodes is not less than a preset threshold, wherein the key areas of the array surface include the center area, the edge area, and the stress concentration area; the preset threshold is 5cm.

[0028] As a preferred embodiment, the set of strain-sensitive nodes is encoded using binary encoding. Each sensitive node corresponds to one gene bit. A value of "1" in the gene bit indicates that the node is selected, and a value of "0" indicates that the node is not selected. The encoding length is consistent with the total number of sensitive nodes.

[0029] As a preferred option, a fitness function that balances deformation representation capability and node quantity economy is defined, as follows:

[0030]

[0031] Where F(X) is the individual fitness value; R(X) is the complex correlation coefficient between the nodal strain and the centroid displacement of the front surface corresponding to test group X; Let X be the number of nodes in test group X, and K be the total number of nodes in the strain-sensitive node set. , These are the weighting coefficients.

[0032] As a preferred approach, genetic operations are performed on the initialized population, including selection, crossover, and mutation operations; the probability of each individual being selected as a parent is determined based on the proportion of the individual's fitness value F(X) to the total fitness value of the population.

[0033] Individuals with higher fitness values ​​F(X) are more likely to be selected for subsequent crossover and mutation operations, thereby preserving high-quality genes in the population.

[0034] Individuals with fitness values ​​F(X) higher than the preset value are retained, and the strain-sensitive node test group is output.

[0035] As a preferred option, the selection operation adopts the roulette wheel selection method, which determines the probability of each individual being selected as the parent based on the proportion of the individual's fitness value to the total fitness value of the population.

[0036] The crossover operation adopts a single-point crossover strategy, with an initial crossover probability of 0.7, which can be adjusted to between 0.6 and 0.8 depending on the convergence of the algorithm.

[0037] The mutation operation sets the initial mutation probability to 0.05, which can be adjusted to between 0.03 and 0.08 depending on the population optimization situation.

[0038] As a preferred approach, the deviation between the predicted displacement data and the measured displacement data is calculated. This deviation is quantified using the root mean square error (RMSE), and the calculation formula is as follows:

[0039]

[0040] in, , , To predict the x, y, and z components of the displacement data; , , represents the x, y, and z components of the measured displacement data; N represents the number of data samples.

[0041] As a preferred approach, the finite element model of the antenna array structure is constructed using Abaqus software; different constraints and loads are set for the finite element model using Python scripts; a coordinate system for the array structure is established using a binocular vision sensor; and different types of loads are applied, including thermal loads and concentrated force loads.

[0042] Compared with the prior art, the present invention has at least the following beneficial effects:

[0043] (1) This invention uses a genetic algorithm to narrow down the optimization range based on "pre-screening of strain-sensitive nodes" and combines a dual-objective fitness function to balance deformation representation capability and node economy. This avoids the blindness of the traditional genetic algorithm's full node traversal and can quickly obtain a fiber grating layout scheme that meets the needs of array deformation monitoring. The optimization process is more in line with actual engineering scenarios. It achieves high-precision reconstruction of array structure deformation and simplifies the array deformation reconstruction steps. Compared with traditional methods (curvature method, inverse finite element method, etc.), this invention does not require the establishment of a complex model and can solve the problems of difficult calibration data acquisition and limited sample data in traditional methods.

[0044] (2) This invention introduces a neural network to establish a “strain-deformation” mapping relationship, replacing the reconstruction logic that relies on complex structural models in traditional methods. The direct correlation and restoration of array deformation can be achieved through strain data, solving the problems of difficult model calibration and strong sample dependence in traditional methods, improving the accuracy and flexibility of deformation characterization, and can be used for array fiber grating layout under complex loads.

[0045] (3) The present invention outputs multiple sets of differentiated fiber optic grating layout test groups based on genetic algorithm, which can adapt to the requirements of the number of nodes and distribution location under different monitoring scenarios. At the same time, the introduction of neural network enables the layout scheme to adapt to deformation monitoring of various service conditions without additional adjustment, improves the training speed of the model, shortens the training time of the model, enhances the robustness of the model, and enhances the engineering practical value of the method. Attached Figure Description

[0046] Figure 1 This is a general technical roadmap of the antenna array fiber optic grating layout method in the embodiment.

[0047] Figure 2 The flowchart shows the point selection criteria and genetic algorithm in this embodiment.

[0048] Figure 3 This is a schematic diagram of the neural network-driven antenna strain-displacement correlation analysis in the embodiment.

[0049] Figure 4 The diagram shows the Abaqus finite element simulation model of the array structure in the embodiment.

[0050] Figure 5 The diagram shows the simulated strain data of the array structure in the embodiment.

[0051] Figure 6 The diagram shows the simulated displacement data of the array structure in the embodiment.

[0052] Figure 7 This is a diagram of the experimental fixture for the array structure model in the embodiment.

[0053] Figure 8The diagram shows the actual deformation of the array structure in the embodiment.

[0054] Figure 9 This is a root mean square error plot of the genetic algorithm test group of this invention. Detailed Implementation

[0055] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0056] To address the issues of fiber Bragg grating (FBG) layout relying on manual rules, poor adaptability with small samples, and insufficient monitoring accuracy in key areas, this invention proposes an optimized layout method for antenna array FBGs based on genetic algorithms and neural networks. This combined approach leverages the global search capability of genetic algorithms to quickly identify potentially efficient layout schemes within the constraints of the number of sensors, while the neural network fits the mapping relationship between array deformation and monitoring data, reducing reliance on high-precision sample data. The synergy of these two methods can balance monitoring coverage and sensor utilization efficiency with limited samples, representing an effective way to improve the monitoring performance of antenna array FBGs.

[0057] Unlike existing array structure layout methods, this invention first constructs an array finite element model, obtains a large amount of low-cost, low-fidelity strain-displacement data through Abaqus simulation of multiple service conditions, and then selects strain-sensitive nodes according to strain fluctuation amplitude, displacement correlation criteria, and spatial constraints. A genetic algorithm is used to iteratively encode the sensitive nodes, generating multiple candidate layouts. Secondly, based on the candidate layouts output by the genetic algorithm, this invention constructs a physical array structure and lays fiber optic gratings at the candidate nodes. A small amount of high-cost, high-fidelity strain-displacement data is collected through actual calibration tests. Simultaneously, the strain data is converted into images with spatial location features, and a neural network is used to construct a mapping model of "strain image - displacement component". Thirdly, this invention inputs the high-fidelity strain images corresponding to the candidate layouts into this mapping model. By evaluating the error between predicted and measured displacements, the candidate node combination with the smallest root mean square error is selected, and its spatial coordinates represent the optimal layout position of the fiber optic gratings. This approach reduces dependence on a large number of real samples through multi-fidelity data fusion while ensuring the deformation monitoring accuracy and engineering practicality of the layout.

[0058] To illustrate the specific embodiments of the present invention in detail, according to Figure 1 The overall technology roadmap shown is as follows: Figure 2 The point selection criteria and genetic algorithm flowchart shown below further illustrate the present invention. The specific process is as follows:

[0059] 1) Based on the theoretical geometric model of the antenna array structure, construct the Abaqus finite element model of the array structure;

[0060] 2) By setting different constraints and loads on the finite element model using Python scripts, different service conditions of the antenna array structure are simulated, and a large amount of low-fidelity strain-displacement data of the antenna array structure under different operating conditions is obtained. ;

[0061] 3) The selection of strain-sensitive locations needs to meet the strain fluctuation amplitude criterion and the displacement correlation criterion. The strain fluctuation amplitude criterion is the difference in strain extreme values ​​between the frontal nodes under various service conditions. ,in The set of strains of the i-th node under different working conditions; Filtering threshold: The minimum strain resolution of the antenna needs to be covered to ensure that the node has sufficient response sensitivity to load changes.

[0062] 4) Displacement Correlation Criterion: The Pearson correlation coefficient is used to quantify the linear relationship between nodal strain and overall surface deformation. ,in The composite value of the three-dimensional displacement of the centroid of the array surface; Filtering threshold: (Strong correlation level) to ensure that the nodal strain can effectively reflect the overall deformation trend of the array surface.

[0063] 5) Sensitive nodes must cover the center area, edge area, and stress concentration area of ​​the array surface (such as the connection between the array element and the substrate). The spatial distance between any two sensitive nodes should not be less than 5cm to avoid information redundancy and the risk of local failure. Combine the strain fluctuation amplitude criterion and the displacement correlation criterion to select strain sensitive nodes with strong characterization of array surface deformation. The location of the sensitive node is denoted as S.

[0064] 6) Using the selected strain-sensitive nodes as the optimization targets, the set of strain-sensitive nodes is encoded using binary encoding: each sensitive node corresponds to one gene bit, where a value of "1" indicates that the node is selected, and a value of "0" indicates that the node is not selected. Therefore, the encoding length of each individual is consistent with the total number of sensitive nodes. The population is initialized based on random sampling rules, generating 50 initial individuals (each individual corresponds to a set of strain-sensitive nodes) to ensure that the initial population covers different node distribution patterns and avoids subsequent iterations from getting trapped in local optima.

[0065] 7) To simultaneously consider the deformation representation capability of the test group and the economy of node quantity, the fitness function is defined as:

[0066]

[0067] Where: R(X) is the complex correlation coefficient between the nodal strain and the displacement of the centroid of the front surface corresponding to test group X (the value range is [0,1], and the closer the coefficient is to 1, the stronger the deformation characterization ability of the test group).

[0068] Let X be the number of nodes in test group X, and K be the total number of nodes in the strain-sensitive node set.

[0069] Initial setting of weight coefficients , The weighting coefficients need to be dynamically adjusted according to the actual optimization objectives, prioritizing the deformation representation accuracy of the test group, while controlling the number of nodes to reduce monitoring costs.

[0070] 8) Genetic algorithms consist of three parts: selection, crossover, and mutation. Selection: A roulette wheel selection method is used, based on the individual's fitness value. The proportion of the total fitness value of the population determines the probability of each individual being selected as a parent; fitness value The taller the individual, the greater the probability of being selected to participate in subsequent crossover and mutation operations, thereby preserving high-quality genes in the population;

[0071] 9) The crossover operation adopts a single-point crossover strategy, with the initial crossover probability set to 0.7 (if the algorithm converges slowly, it can be appropriately increased to 0.75-0.8 to enhance the gene recombination frequency; if the results fluctuate greatly, it can be reduced to 0.6-0.65 to reduce the probability of high-quality genes being disrupted); a crossover position is randomly selected, and the gene segments of the two parent individuals after the crossover position are exchanged to generate two offspring individuals, enriching the node combination pattern of the population through gene recombination;

[0072] 10) The mutation operation sets the initial mutation probability to 0.05 (if the algorithm gets stuck in a local optimum, it can be increased to 0.07-0.08 to enhance population diversity; if the result is unstable, it can be reduced to 0.03-0.04 to reduce meaningless mutations). For newly generated offspring individuals, perform bit flip mutation, randomly select several gene bits and flip their values ​​from "1" to "0" or from "0" to "1" to avoid the population getting stuck in a local optimum due to gene convergence and to ensure global search capability. Perform the above genetic operation to achieve population evolution, iterate until the population converges, and finally output multiple sets of nodes with high fitness ranking to form multiple sets of strain-sensitive node test groups. Each test group contains a different number and different positions of sensitive nodes.

[0073] 11) Construct the actual antenna array structure, lay fiber optic strain sensors at the corresponding positions of sensitive node S, and establish the coordinate system of the array structure using a binocular vision sensor. Apply different thermal loads and concentrated force loads to the actual antenna array structure, and obtain a small set of high-fidelity strain-displacement data samples of the array structure. ;

[0074] 12) , The strain data is decomposed into datasets based on the orthogonality of spatial location. , , , ,in , ;

[0075] 13) Reassemble the decomposed strain dataset into an image based on its spatial location through a certain mapping relationship. , , , ;

[0076] 14) , The displacement data is decomposed into displacement component datasets in the x, y, and z directions according to the workpiece coordinate system. , , , , , ;

[0077] 15) Based on neural network fitting , and , , The mapping relationship of the output features is used to obtain model M;

[0078] 16) High-fidelity data images corresponding to multiple test groups determined by the genetic algorithm. , Input model M to obtain the output prediction , , ;

[0079] 17) Calculate the predicted displacement for each test group. , , Compared with the measured displacement , , The root mean square error between the predicted and measured displacements is used to quantify the deviation between the predicted and measured displacements. The formula is as follows:

[0080]

[0081] 18) Compare the root mean square error results of all test groups, select the test group with the smallest error, and the strain sensitive node corresponding to this group is the "optimal strain sensitive node"; output the spatial coordinate information of this group of nodes, and use it as the optimized layout position of the antenna array fiber optic grating sensor to complete the layout design.

[0082] Figure 3 This illustration demonstrates the neural network-driven antenna strain-displacement correlation analysis of this invention. In this embodiment, an existing open-source BP neural network project is used, without any special limitations.

[0083] Figure 4 The Abaqus finite element simulation model of the array structure was established; Figure 5 For the simulation strain data of the array structure, Figure 5 (a) shows the displacement components in the X direction of low-fidelity strain-displacement data. Figure 5 (b) shows the displacement components in the Y direction of the low-fidelity strain-displacement data; Figure 6 For the simulation displacement data of the array structure, Figure 6 (a) shows the displacement component dataset. , Figure 6 (b) shows the displacement component dataset. , Figure 6 (c) shows the displacement component dataset. ;like Figure 7 As shown, this invention uses an antenna array structure model test fixture. Figure 8 This is a diagram showing the actual deformation of the array structure. Figure 9 The genetic algorithm was used to test the root mean square error (RMSE) map of the test group. The maximum RMSE in the test group was 2.6934 mm, and the minimum RMSE was 0.2707 mm. The test group with the minimum RMSE was selected as the optimal position for fiber optic grating layout.

[0084] In summary, the present invention is applied to the deformation sensing and fiber optic grating layout optimization of antenna array structures. By analyzing the strain-sensitive locations of the array structure through genetic algorithms and neural networks, fiber optic grating sensors are arranged at the optimal layout locations in the array structure to achieve high-precision deformation sensing of the array structure under complex service loads.

[0085] The logical ideas behind the methods disclosed in the above embodiments can be implemented, in whole or in part, through software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs.

[0086] When computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).

[0087] 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, and 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 illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0088] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0089] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for laying out fiber optic gratings in an antenna array based on genetic algorithms and neural networks, characterized in that, Includes the following steps: A finite element model of the antenna array structure was constructed, and different constraints and loads were set to simulate and obtain low-fidelity strain-displacement data of the antenna array structure under different service conditions. Based on preset screening criteria, strain-sensitive nodes are selected to obtain a set of strain-sensitive nodes; The strain-sensitive node set is encoded, and the population is initialized based on random sampling rules to generate a preset number of initial individuals, with each individual corresponding to a set of strain-sensitive nodes; Define a fitness function that balances deformation representation capability with node quantity economy; Genetic operations are performed on the initialized population, and the process is iterated until the population converges. The strain-sensitive node test group with fitness that meets the preset requirements is output. Construct the actual antenna array structure, deploy fiber optic strain sensors at the locations corresponding to the strain-sensitive node set, establish the coordinate system of the array structure, load different types of loads onto the actual antenna array structure, and obtain high-fidelity strain-displacement data. The low-fidelity strain-displacement data and the high-fidelity strain-displacement data are subjected to strain data decomposition and recombination and displacement data decomposition processing, respectively, to obtain strain image data and displacement component datasets. A deformation prediction model is constructed by fitting the mapping relationship between strain image data and displacement component dataset under low-fidelity strain-displacement data using neural networks. Based on the strain-sensitive node test group, the strain image data under its high-fidelity strain-displacement data is input into the deformation prediction model to obtain the predicted displacement data of each strain-sensitive node test group. The deviation between the predicted displacement data and the measured displacement data is calculated. The strain-sensitive node corresponding to the strain-sensitive node test group with the smallest deviation value is selected as the optimal strain-sensitive node. Its spatial coordinate information is output as the optimized layout position of the antenna array fiber optic grating sensor.

2. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 1, characterized in that, Strain-sensitive nodes are selected based on preset screening criteria, which include strain fluctuation amplitude criteria and displacement correlation criteria. The criteria also satisfy that the sensitive nodes cover the key area of ​​the array surface and the spatial distance between any two sensitive nodes is not less than a preset threshold, thus obtaining a set of strain-sensitive nodes.

3. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 2, characterized in that, The strain fluctuation amplitude criterion is the extreme strain difference of the array nodes under various service conditions. : in Let be the strain set of the i-th node under different working conditions; the screening threshold is . .

4. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 2, characterized in that, The displacement correlation criterion uses the Pearson correlation coefficient. The linear relationship between quantified nodal strain and overall surface deformation: in The resultant value of the triaxial displacement of the array's centroid; the screening threshold is... .

5. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 2, characterized in that, The sensitive nodes cover the key areas of the array surface and the spatial distance between any two sensitive nodes is not less than a preset threshold. The key areas of the array surface include the center area, the edge area, and the stress concentration area.

6. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 1, characterized in that, The set of strain-sensitive nodes is encoded using binary encoding. Each sensitive node corresponds to a 1-bit gene. A value of "1" in the gene bit indicates that the node is selected, and a value of "0" indicates that the node is not selected. The encoding length is the same as the total number of sensitive nodes.

7. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 1, characterized in that, A fitness function that balances deformation representation capability and node quantity economy is defined as follows: Where F(X) is the individual fitness value; R(X) is the complex correlation coefficient between the nodal strain and the centroid displacement of the front surface corresponding to test group X; Let X be the number of nodes in test group X, and K be the total number of nodes in the strain-sensitive node set. , These are the weighting coefficients.

8. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 7, characterized in that, Genetic operations are performed on the initialized population, including selection, crossover, and mutation; the probability of each individual being selected as a parent is determined based on the proportion of the individual's fitness value F(X) to the total fitness value of the population. Individuals with fitness values ​​F(X) higher than the preset value are retained, and the strain-sensitive node test group is output.

9. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 1, characterized in that, The deviation between the predicted displacement data and the measured displacement data is calculated. This deviation is quantified using the root mean square error (RMSE), and the calculation formula is as follows: in, , , To predict the x, y, and z components of the displacement data; , , represents the x, y, and z components of the measured displacement data; N represents the number of data samples.

10. The antenna array fiber optic grating layout method based on genetic algorithm and neural network according to claim 1, characterized in that: A finite element model of the antenna array structure was constructed using Abaqus software; Set different constraints and loads for the finite element model using Python scripts; A coordinate system for the array structure is established using a binocular vision sensor, and different types of loads are applied, including thermal loads and concentrated force loads.

Images