A method for structural deformation reconstruction of phased array antenna based on multi-fidelity data fusion

Abstract

The present invention provides a method for reconstructing the structural deformation of a phased array antenna based on multi-fidelity data fusion, and relates to the technical field of structural deformation reconstruction. Different from the existing array structure prediction methods: on the one hand, the present invention improves a large amount of low-cost, low-fidelity data through limited high-cost, high-fidelity calibration data, wherein the low-fidelity data comes from the structural deformation simulation of Abaqus, and the high-fidelity data comes from the actual calibration test. On the other hand, the present invention uses images to represent the strain monitoring values ​​of fiber gratings arranged in a spatial position relationship on the structure, and then uses a convolutional neural network model to establish a mapping relationship between strain and deformation in data of different fidelity. Utilizing the technical solution proposed by the present invention, accurate prediction of the deformation of the array structure under complex service loads can be achieved, solving the problems of complex modeling and poor applicability of traditional methods, and providing a perception basis for active control of deformation of the array structure.

Description

Technical Field

[0001] The present invention relates to the technical field of structural deformation reconstruction, and in particular to a phased array antenna structural deformation reconstruction method based on multi-fidelity data fusion. Background Art

[0002] Phased array antennas are the core components of radars. Their performance directly determines the strength of radar detection capabilities and are widely used in detection, tracking, communications, and other fields. Phased array antennas produce beam deflection by changing the phase difference of their antenna radiation units to achieve signal positioning and focusing. The position accuracy of their array elements directly affects the electromagnetic performance of the antenna.

[0003] The new generation of phased array antenna array structure adopts active control technology, which realizes the control of array structure accuracy through online perception and compensation of array structure deformation. In this process, it is necessary to build the relationship between structural strain and deformation to realize the reconstruction of structural deformation. The mapping accuracy between strain and deformation directly determines the accuracy of structural deformation reconstruction.

[0004] The traditional array structure deformation reconstruction method based on theoretical models has two problems:

[0005] On the one hand, traditional modal methods and inverse finite element methods both require the establishment of accurate theoretical models. However, the antenna array structure is actually composed of many parts with different material properties and complex structures assembled through a variety of connection processes. It is difficult and time-consuming to establish an accurate model of the structure.

[0006] On the other hand, the improved traditional method relies on calibration data (based on experimentally measured stress-deformation data) to improve model accuracy. However, high-precision calibration data has problems such as small sample size and difficulty in obtaining, resulting in limited improvement in model accuracy. Summary of the invention

[0007] Purpose of the invention: To propose a phased array antenna structure deformation reconstruction method based on multi-fidelity data fusion, which can achieve accurate prediction of array surface structure deformation under complex service loads by fusing low-fidelity and low-cost simulation data with high-fidelity and high-cost calibration data, solve the problems of complex modeling and poor applicability of traditional methods, and provide a perception basis for active control of deformation of array surface structure.

[0008] In a first aspect of the present invention, a method for structural deformation reconstruction of a phased array antenna based on multi-fidelity data fusion is proposed, comprising the following steps:

[0009] A geometric model is established according to the phased array antenna surface structure, and a finite element model is constructed based on the geometric model;

[0010] Set the constraints and loads of the finite element model, simulate different service conditions of the phased array antenna array structure, and obtain several low-fidelity strain-deformation data sets ;

[0011] Fiber Bragg grating strain sensors are arranged on the array structure of the phased array antenna, and a binocular vision sensor is used to establish the coordinate system of the array structure. Different thermal loads and concentrated force loads are applied to the array structure of the phased array antenna to obtain a high-fidelity strain-deformation data set. ;

[0012] The dataset and dataset The strain data in the XY direction are decomposed into strain data sets , , , ;

[0013] The strain data set , , , According to the spatial position, it is reorganized into a low-fidelity image through a predetermined mapping relationship. , and high-fidelity images , ;

[0014] The low-fidelity strain-deformation dataset and high-fidelity strain-deformation datasets The deformation data in the workpiece coordinate system is decomposed into displacement component data sets in the three directions of XYZ , , , , , ;

[0015] Fitting low-fidelity images based on CNN network , With displacement component dataset , , The mapping relationship between them is used to train a low-fidelity regression prediction model ;

[0016] High-fidelity images , Input to a low-fidelity regression prediction model The low-fidelity prediction results of the phased array antenna structure are obtained ;

[0017] Construct a BP neural network and fit the strain data through the BP neural network , With displacement component dataset , , The mapping relationship between them is used to obtain a high-fidelity regression prediction model ;

[0018] Fusion of low-fidelity regression prediction models , high-fidelity regression prediction model , and obtain a final prediction model, which is used to reconstruct the deformation of the phased array antenna structure.

[0019] In a further embodiment of the first aspect, the low-fidelity strain-deformation dataset ;

[0020] The i-th group of data Contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the low-fidelity strain vector, is the low-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction.

[0021] In a further embodiment of the first aspect, the high-fidelity strain-deformation dataset , the jth group of data Contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the high-fidelity strain vector, is the high-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction.

[0022] In a further embodiment of the first aspect, the strain dataset , , , Reassemble into images , , , , the mapping formula is as follows:

[0023]

[0024] Where A is the strain data set in the XY direction, is the minimum value of the elements in matrix A, is the maximum value of the elements in matrix A, is the image after mapping.

[0025] In a further embodiment of the first aspect, based on the CNN network, fitting the low-fidelity image , With displacement component dataset , , The mapping relationship between them is used to train a low-fidelity regression prediction model , and , and fix the parameters of each convolutional neural network , , ;

[0026] High-fidelity images , Enter the low-fidelity regression prediction model , and The low-fidelity prediction vector of the array structure deformation is obtained , , .

[0027] In a further embodiment of the first aspect, before constructing the BP neural network, the method further includes: constructing a mapping relationship F(*) between low-fidelity deformation data and high-fidelity deformation data:

[0028]

[0029] In the formula, represents low-fidelity deformation data, represents high-fidelity deformation data, Represents high-fidelity strain data.

[0030] In a further embodiment of the first aspect, a BP neural network is constructed according to the mapping relationship between the low-fidelity deformation data and the high-fidelity deformation data, and the parameters of each BP neural network are fixed. , , ;

[0031] The BP neural network expression is as follows:

[0032] ;

[0033] ;

[0034] ;

[0035] In the formula, , , It is the low-fidelity prediction vector of the front structure deformation in three directions; , , The feature vector extracted by the last convolutional layer.

[0036] In a further embodiment of the first aspect, the low-fidelity regression prediction model is integrated , high-fidelity regression prediction model , the fusion formula is as follows:

[0037] ;

[0038] ;

[0039] .

[0040] Where x is the model input, and its actual meaning is the strain value measured by the fiber Bragg grating sensor arranged on the phased array antenna structure.

[0041] According to a second aspect of the present invention, an electronic device is proposed, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store a plurality of executable instructions, and the executable instructions enable the processor to execute the phased array antenna structure deformation reconstruction method based on multi-fidelity data fusion as described in the first aspect.

[0042] According to a third aspect of the present invention, a computer-readable storage medium is provided, in which a plurality of executable instructions are stored. When the executable instructions are executed on an electronic device, the electronic device executes the method for structural deformation and reconstruction of a phased array antenna based on multi-fidelity data fusion as described in the first aspect.

[0043] Compared with the prior art, the present invention has at least the following beneficial effects:

[0044] (1) The present invention realizes the end-to-end prediction of the array structure deformation, avoids the problems of complex modeling and difficult calibration data acquisition in traditional methods, and simplifies the operation of strain-deformation reconstruction of the array structure;

[0045] (2) The present invention establishes prior knowledge of structural strain-deformation through a large amount of low-cost low-fidelity data (obtained through simulation experiments), and corrects the prior knowledge through limited high-cost high-fidelity calibration data. Compared with the existing machine learning methods, the accuracy and generalization ability of the reconstruction of the array structure deformation are effectively improved, and it can be applied to the prediction of structural deformation under complex loads;

[0046] (3) The present invention realizes the establishment of the structural stress-deformation mapping relationship based on the image, and then uses the convolutional neural network model to establish the mapping relationship between strain and deformation of data with different fidelity. Compared with the existing deep learning method, it can effectively reduce the number of model parameters, realize multi-level feature extraction of strain-deformation data, improve the training speed of the model, shorten the training time of the model, and enhance the robustness of the model;

[0047] (4) The present invention concatenates the feature vector extracted by the last convolutional layer in the low-fidelity regression prediction model with the high-fidelity data and uses them as the input of the high-fidelity prediction model, thereby providing a large number of low-fidelity features for the establishment of the high-fidelity prediction model and effectively suppressing the overfitting phenomenon of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of the overall solution of the phased array antenna structure deformation and reconstruction method in the embodiment.

[0049] Figure 2 It is a structural deformation reconstruction framework diagram of multi-fidelity data fusion in the embodiment.

[0050] Figure 3 This is the Abaqus finite element simulation model of the array structure established in the embodiment.

[0051] Figure 4 To obtain the displacement data of the array structure measurement points during the simulation process.

[0052] Figure 5 Schematic diagram of the reconstruction error of the actual deformation of the array structure using a fully connected neural network and simulation data. DETAILED DESCRIPTION

[0053] In the following description, a large number of specific details are provided to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that the present invention can be implemented without one or more of these details. In other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described.

[0054] High-precision online perception of phased array antenna deformation is the basis for realizing array surface structural deformation compensation and signal electrical compensation, and its perception accuracy directly determines the compensation effect. However, existing phased array antenna structural deformation perception methods generally have shortcomings such as bulky devices, long measurement cycles, susceptibility to interference from environmental factors, and low reconstruction accuracy. It is difficult to achieve online perception of structural deformation in complex environments. For example, the displacement sensor measurement method requires the addition of additional measurement equipment and structures, and cannot be highly integrated with the array surface structure; the accuracy of the optical measurement method is seriously affected by environmental interference, and it is impossible to achieve deformation measurement in complex service environments. Fiber Bragg Grating (FBG) sensors have the characteristics of high sensitivity, wide measurement range, high temperature resistance and high stability, and can more conveniently and accurately obtain strain data at each point on the structure.

[0055] However, the fiber Bragg grating online detection is the strain data of the structure, and it is necessary to further build the relationship between the structural strain and deformation to realize the reconstruction of the structural deformation. The mapping accuracy between the strain and deformation directly determines the accuracy of the structural deformation reconstruction.

[0056] Compared with the limitations of traditional deformation reconstruction methods, the structural reconstruction method based on neural network algorithms does not require an accurate model of the structure. The reconstruction process is simple and has a wide range of applications. It is an effective way to achieve fiber Bragg grating monitoring and reconstruction of phased array antenna surface structures. However, the structural deformation reconstruction method based on neural network algorithms is essentially a regression problem. The reconstruction accuracy depends on the number and quality of samples. Although it does not require the establishment of an accurate structural model, it also faces the problem of a small number of high-precision calibration data samples and time-consuming and labor-intensive problems. The classic structural deformation reconstruction method based on neural networks either obtains a large number of low-fidelity samples (sample data with low reconstruction accuracy) through a large number of virtual simulations, or directly trains based on calibrated high-fidelity sample data to establish the stress-deformation mapping relationship of the structure. [5] Both methods have limitations. The former is based on the neural network model trained with low-fidelity samples, and the trained model will have a large error with the actual situation due to the low accuracy of the samples. Although the latter is based on high-fidelity samples for training, the number of samples is small. Under the complex load in the service process, the generalization ability of the model is weak, and the overfitting problem occurs.

[0057] In order to solve the problems of high dependence on the number of real samples, weak generalization ability of the reconstruction model and low reconstruction accuracy, the present invention proposes a high-precision reconstruction method of structural deformation fiber Bragg grating based on multi-fidelity data fusion. Figure 1. Different from the existing array structure prediction methods: On the one hand, the present invention proposes a high-precision reconstruction method of fiber Bragg grating of phased array antenna structure deformation based on multi-fidelity data fusion, which improves a large amount of low-cost low-fidelity data through limited high-cost high-fidelity calibration data, wherein the low-fidelity data comes from Abaqus structural deformation simulation, and the high-fidelity data comes from actual calibration tests. On the other hand, the multi-fidelity data fusion method proposed in the present invention uses images to represent the strain monitoring values ​​of fiber Bragg gratings arranged in a spatial position relationship on the structure, and then uses a convolutional neural network model to establish a mapping relationship between strain and deformation in data of different fidelity. Compared with the existing methods, the number of model parameters can be reduced, multi-level feature extraction of strain-deformation data can be achieved, and the robustness of the model can be enhanced.

[0058] To explain the specific embodiments of the present invention in detail, Figure 2 The present invention is further illustrated by a structural deformation reconstruction framework diagram of multi-fidelity data fusion.

[0059] 1) Determine the geometric model parameters of the array antenna based on the structural parameters and material properties of the phased array antenna surface structure;

[0060] 2) Establish a finite element model of the array structure in Abaquas;

[0061] 3) According to the wind load, deadweight and temperature loads that the phased array antenna is subjected to during service, the constraints, loads and workpiece coordinate system of the finite element model are established. The strain at the specified position of the antenna structure and the position change in the x, y and z directions under various working conditions are simulated through Python scripts, and the data of the above strain values ​​and position changes are stored;

[0062] 4) Perform stress-deformation finite element simulation under group M working conditions in 3) to obtain low-fidelity strain-deformation data of the phased array antenna surface structure under group M working conditions , , the i-th group of data (where i is the index of the low-fidelity data, ) contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the low-fidelity strain vector, is the low-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction;

[0063] 5) According to the strain measurement point position of the finite element model in abaqus, the fiber Bragg grating strain sensor is arranged on the actual array structure of the phased array antenna, and the fiber Bragg grating measurement points are arranged on the physical structure of the phased array antenna according to the specified spatial position relationship;

[0064] 6) By applying different thermal loads and concentrated force loads to the physical structure of the phased array antenna array, the array structure is deformed, and a binocular vision sensor is used to establish the same workpiece coordinate system as the finite element simulation model;

[0065] 7) Load different thermal loads and concentrated force loads on the actual structure of the phased array antenna array to obtain N sets of high-fidelity data sets of array surface structure strain-deformation , , the jth group of data (where j is the index of the experimental data, ) contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the high-fidelity strain vector, is the high-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction;

[0066] 8) The i-th group of data The strain vector in Center front The elements are reorganized into a low-fidelity strain matrix representing the measurement in the x direction according to the spatial position distribution of the array structure coordinate system established by binocular vision in 3). ( ), the specific expression is:

[0067]

[0068] in , , low-fidelity strain matrix Each element in satisfies the following conditions: , , , ;

[0069] 9) Step 8) Indicates obtaining the x-direction coordinate of the corresponding element in the array structure coordinate system established by binocular vision. Indicates obtaining the y-direction coordinate of the corresponding array structure coordinate system;

[0070] 10) The i-th group of data The strain vector in The index is The elements of the elements are reorganized into a low-fidelity strain matrix representing the measurement of the y direction according to the spatial position distribution of the array structure coordinate system established by binocular vision in 4). ( ), the specific expression is:

[0071]

[0072] in , , low-fidelity strain matrix Each element in satisfies the following conditions: , , , ;

[0073] 11) The jth group of data The strain vector in Center front The elements are reorganized into a high-fidelity strain matrix representing the measurement in the x direction according to the spatial position distribution of the array structure coordinate system established by binocular vision in 3). ( ), the specific expression is:

[0074]

[0075] in , , high-fidelity strain matrix Each element in satisfies the following conditions: , , , ;

[0076] 12) The jth group of data The strain vector in The index is The elements of the elements are reorganized into a high-fidelity strain matrix representing the measurement of the y direction according to the spatial position distribution of the array structure coordinate system established by binocular vision in 3). ( ), the specific expression is:

[0077]

[0078] in , , high-fidelity strain matrix Each element in satisfies the following conditions: , , , ;

[0079] 13) Convert the low-fidelity strain matrix from 4) to 6) , With high fidelity strain matrix , Composition of the dataset , , , ,in , , , ;

[0080] 14) Combine the low-fidelity deformation data from 2) and 3) With high-fidelity deformation data Composition of the dataset , , , , , ,in , , , , , ;

[0081] 15) High- and low-fidelity strain datasets , Mapped to an image through the following formula: , where IMG is the grayscale value of each pixel in the converted grayscale image, A is the strain or deformation matrix, is the minimum value of the elements in matrix A, is the maximum value of the elements in the A matrix, and the transformed strain image set is , ;

[0082] 16) Collect low-fidelity strain images , With low-fidelity deformation dataset , , Divide into training set and test set;

[0083] 17) Construct multiple dual-channel CNN network models to fit low-fidelity strain data , and , , The mapping relationship is obtained to obtain a low-fidelity regression prediction model , and , and fix the parameters of each convolutional neural network , and ;

[0084] 18) Specifically, the dual-channel CNN network model in 17) includes convolution layer 1, pooling layer 1, convolution layer 2, pooling layer 2, convolution layer 3, pooling layer 3, convolution layer 4, pooling layer 4, full connector layer 1, full connector layer 2, and full connection layer 3, and the parameters of each layer are reasonably set;

[0085] 19) The mean square error loss function is used to measure the accuracy of deformation prediction, and the Adam optimizer is used to calculate the gradient of the loss function to the network parameters through the back propagation algorithm;

[0086] 20) Obtain a low-fidelity regression prediction model through training in 18) and 19) , and ;

[0087] 21) High-fidelity strain dataset images , Enter the low-fidelity regression prediction model , and The low-fidelity prediction vector of the array structure deformation is obtained , , ;

[0088] 22) Constructing low-fidelity deformation data With high-fidelity deformation data The mapping relationship ;

[0089] 23) According to the mapping relationship in 22), by constructing a BP neural network , , , achieving high-fidelity strain data , low-fidelity prediction vector of array structure deformation , , High-fidelity deformation data of the array structure , , Fitting of mapping relationships;

[0090] 24) Specifically, in order to improve the accuracy and generalization ability of the regression prediction model, avoid , , In order to solve the overfitting phenomenon of the high-fidelity regression prediction model, this embodiment additionally proposes a data fusion method for fusing the feature vector of the low-fidelity convolutional neural network to transform the low-fidelity regression prediction model into , and The feature vector extracted by the last convolutional layer , , and 21) , , , Perform linear splicing as input to the above regression prediction model;

[0091] 25) High-fidelity deformation dataset , , As , , Output of regression prediction model;

[0092] 26) Initialize the BP neural network parameters, input the parameters in 16) into the neural network regression prediction model, output the predicted value of the array structure deformation after forward propagation, set the error threshold, and if the error between the predicted value of the structure deformation and the actual structure deformation value does not meet the requirements, then back propagate the error in the neural network;

[0093] 27) The Adam optimization method is used to adjust the learning rate to optimize the model and update the weights , , , and its update weights are as follows: , where and are the estimates of the first and second order moments of the corrected gradient, is the learning rate for model training, It is a parameter to prevent division by 0;

[0094] 28) Repeat steps 26) and 27) until the model prediction error is less than the error threshold, and a high-fidelity prediction model is obtained. , , ;

[0095] 29) Low-fidelity regression prediction model in fusion 20) , , With the high-fidelity prediction model in 25), the fusion formula is , , ;

[0096] 30) In the actual reconstruction process, after the strain data is obtained, it is reorganized into an image according to the operation in 4), and the accurate reconstruction of the array structure deformation driven by the fiber Bragg grating strain data is realized according to formula 21).

[0097] The present invention uses a phased array antenna array structure array model to carry out experiments. Figure 3 To establish the Abaqus finite element simulation model of the array structure, Figure 4 In order to obtain the displacement data of the array structure measurement points during the simulation process, the red point set is the spatial position of the array structure measurement points before and after deformation, and the green line segment set surface measurement points are displaced in space due to deformation. Figure 5 The figure is a comparison diagram of the deformation prediction error of the phased array antenna structure based on multi-fidelity data fusion and the deformation prediction error of the phased array antenna based on the fully connected neural network, wherein the maximum prediction error of the fully connected neural network prediction method based on simulation data is 0.28mm, and the average prediction error is 0.19mm. The maximum prediction error proposed by the present invention is 0.12mm, and the average prediction error is 0.05mm. The reconstruction accuracy of the present invention is higher. The phased array antenna structure deformation reconstruction method disclosed in the above embodiment can be embedded in the field system by writing it as an executable code. The executable code runs in the field system in the form of software, and the software can be written into a computer-readable storage medium. More specific examples of the computer-readable storage medium in this embodiment may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0098] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, which carry readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0099] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A method for structural deformation reconstruction of phased array antenna based on multi-fidelity data fusion, characterized in that: The steps include: Establish the finite element model of the phased array antenna according to its array structure; Set constraints and loads for the finite element model to obtain a low-fidelity strain-deformation dataset ; Apply different thermal loads and concentrated force loads to the phased array antenna array structure, establish its coordinate system, and obtain a high-fidelity strain-deformation data set ; The dataset and dataset The strain data in is decomposed into strain data sets in XY directions; The strain data set in the XY direction is reorganized into a low-fidelity image and a high-fidelity image according to the spatial position mapping; The dataset and dataset The deformation data in is decomposed into a low-fidelity displacement component data set and a high-fidelity displacement component data set in the three directions of XYZ; Fitting low-fidelity images based on CNN network , With low-fidelity displacement component dataset , , The mapping relationship between them is used to train a low-fidelity regression prediction model , and , and fix the parameters of each convolutional neural network , , ; High-fidelity images , Enter the low-fidelity regression prediction model , and The low-fidelity prediction vector of the array structure deformation is obtained , , ; Construct the mapping relationship F(*) between low-fidelity deformation data and high-fidelity deformation data: ； In the formula, represents low-fidelity deformation data, represents high-fidelity deformation data, Represents high-fidelity strain data; Construct a BP neural network based on the mapping relationship between low-fidelity deformation data and high-fidelity deformation data, and fix the parameters of each BP neural network , , ; The BP neural network expression is as follows: ； ； ； In the formula, , , It is the low-fidelity prediction vector of the front structure deformation in three directions; , , The feature vector extracted by the last convolutional layer; Fit the mapping relationship between strain data and high-fidelity displacement component data sets to obtain a high-fidelity regression prediction model ; Fusion of low-fidelity regression prediction models , high-fidelity regression prediction model , the final prediction model is obtained, and the deformation results of the phased array antenna structure are reconstructed using the final prediction model.

2. The method for phased array antenna structure deformation reconstruction based on multi-fidelity data fusion according to claim 1, characterized in that: The low-fidelity strain-deformation dataset ; The i-th group of data Contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the low-fidelity strain vector, is the low-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction.

3. The method for phased array antenna structure deformation reconstruction based on multi-fidelity data fusion according to claim 1, characterized in that: The high-fidelity strain-deformation dataset The data is collected by the fiber Bragg grating strain sensor arranged on the array structure of the phased array antenna, and the coordinate system of the array structure is established by using the binocular vision sensor; The high-fidelity strain-deformation dataset ; The jth group of data Contains the strain values ​​of P feature points and the deformation values ​​of Q feature points, , is the high-fidelity strain vector, is the high-fidelity deformation vector, and Center front The element is the strain data measured in the x direction, and the index is The element is the strain data measured in the y direction.

4. The method for phased array antenna structure deformation reconstruction based on multi-fidelity data fusion according to claim 1, characterized in that: The strain data set in the XY direction , , , Reassemble into low-fidelity images , and high-fidelity images , , the mapping formula is as follows: ； Where A is the strain data set in the XY direction, is the minimum value of the elements in matrix A, is the maximum value of the elements in matrix A, is the image after mapping.

5. The method for phased array antenna structure deformation reconstruction based on multi-fidelity data fusion according to claim 1, characterized in that: The fusion low-fidelity regression prediction model , high-fidelity regression prediction model , the fusion formula is as follows: ； ； ； Where x is the strain value measured by the fiber grating sensor arranged on the phased array antenna structure; , , They are the displacement values ​​of the measured points of the antenna array structure in the x, y, and z directions in space respectively.

6. An electronic device, characterized in that: include: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store a plurality of executable instructions, and the executable instructions enable the processor to execute the phased array antenna structure deformation reconstruction method based on multi-fidelity data fusion as described in any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: The storage medium stores a plurality of executable instructions, and when the executable instructions are executed on the electronic device, the electronic device executes the phased array antenna structure deformation reconstruction method based on multi-fidelity data fusion as described in any one of claims 1 to 5.

Images