CN-122021259-A - Deep learning-based antenna array construction method, device and storage medium

CN122021259ACN 122021259 ACN122021259 ACN 122021259ACN-122021259-A

Abstract

The invention provides an antenna array construction method, device and storage medium based on deep learning, belonging to the technical field of antenna construction, wherein the method comprises the steps of importing key parameters of antenna array elements to be constructed, key parameters of the antenna array to be constructed, construction data of a plurality of historical antenna array elements and construction data of a plurality of historical antenna arrays; and respectively carrying out parameter analysis on the key parameters of the antenna array element to be constructed and the key parameters of the antenna array to be constructed to obtain the target gradient slot line coordinates and the antenna array factors. The invention solves the problem of long construction period in the construction of the antenna array, greatly improves the construction efficiency, can reduce the dependence on the experience of constructors, realizes the automation and global optimization construction of the construction process of the antenna array, improves the stability and reliability of the construction result, and breaks through the dependence of the correction precision of the traditional method on the correction frequency step and the space-domain step interval.

Inventors

TANG LONG
WEI JIDA
YE ZUCHANG
BIN XUE
LI HUI

Assignees

桂林长海发展有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (10)

1. The method for constructing the antenna array based on the deep learning is characterized by comprising the following steps of: Importing key parameters of an antenna array element to be constructed, key parameters of the antenna array to be constructed, construction data of a plurality of historical antenna array elements and construction data of a plurality of historical antenna arrays; Respectively carrying out parameter analysis on the key parameters of the antenna array elements to be constructed and the key parameters of the antenna array to be constructed to obtain target gradient slot line coordinates and antenna array factors; building a training model, and carrying out predictive analysis on the key parameters of the antenna array elements to be built, the key parameters of the antenna array to be built, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element building data and all the historical antenna array building data to obtain target antenna array data; And carrying out consistency analysis on the target antenna array data to obtain an antenna array construction result.
2. The method for constructing an antenna array based on deep learning according to claim 1, wherein the process of performing parameter analysis on the key parameters of the antenna array element to be constructed and the key parameters of the antenna array to be constructed to obtain the target gradient slot line coordinates and the antenna array factors includes: calculating target gradual change slot line coordinates for the key parameters of the antenna array elements to be constructed to obtain the target gradual change slot line coordinates; and calculating the antenna array factors for the key parameters of the antenna array to be constructed to obtain the antenna array factors.
3. The method for constructing an antenna array based on deep learning according to claim 2, wherein the key parameters of the antenna array element to be constructed include a gradient slot line start coordinate, a gradient slot line end coordinate and a gradient slot line curvature, and the process of calculating the target gradient slot line coordinate for the key parameters of the antenna array element to be constructed to obtain the target gradient slot line coordinate includes: calculating the initial coordinate of the gradient groove line, the final coordinate of the gradient groove line and the curvature of the gradient groove line by a first formula to obtain a target gradient groove line coordinate, wherein the first formula is as follows: , Wherein, the For the coordinates of the target taper line, For the starting coordinates of the graded slot line, For the graded slot line termination coordinates, Is the curvature of the gradual change groove line.
4. The method for constructing an antenna array based on deep learning according to claim 2, wherein the key parameters of the antenna array to be constructed include an antenna array beam scanning angle, an operating wavelength, a plurality of excitation signal amplitudes and a plurality of antenna array element positions, and the process of calculating the antenna array factors for the key parameters of the antenna array to be constructed includes: Calculating the antenna array beam scanning angle, the working wavelength, the excitation signal amplitudes and the antenna array element positions by a second formula to obtain an antenna array factor, wherein the second formula is as follows: , Wherein, the , Wherein, the As a factor of the antenna array, For the number of antenna elements, Is the first The excitation signal amplitude corresponding to the antenna array elements, Is the number of waves to be used, Is the first The antenna element positions corresponding to the antenna elements, For the antenna array beam scan angle, Is the operating wavelength.
5. The method for constructing an antenna array based on deep learning according to claim 1, wherein the training model includes a first convolutional neural network and an original antenna array prediction network, and the process of obtaining target antenna array data by performing predictive analysis on the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data includes: Training the first convolutional neural network according to all the historical antenna array element construction data to obtain a target antenna array element construction model; Predicting key parameters of the antenna elements to be constructed and the coordinates of the target gradient slot line through the target antenna element construction model to obtain target antenna element prediction data; Training the original antenna array prediction network according to the target antenna array element prediction data and all the historical antenna array construction data to obtain a target antenna array prediction network; And predicting the key parameters of the antenna array to be constructed and the antenna array factors through the target antenna array prediction network to obtain target antenna array data.
6. The method for constructing an antenna array based on deep learning according to claim 5, wherein the process of performing model analysis on the first convolutional neural network according to all the historical antenna array element construction data to obtain a target antenna array element construction model comprises: s311, collecting all the historical antenna array element construction data to obtain a historical antenna array element construction data set; S312, dividing the historical antenna array element construction data set into an antenna array element construction training set and an antenna array element construction test set; s313, predicting the antenna array element construction training set through the first convolutional neural network to obtain a plurality of first antenna array element prediction data, and integrating all the first antenna array element prediction data to obtain a first antenna array element prediction data set; S314, importing a first antenna element real data set corresponding to the antenna element construction training set and a second antenna element real data set corresponding to the antenna element construction test set; S315, calculating the prediction errors of the first antenna array element prediction data set and the first antenna array element real data set to obtain a first prediction error; S316, judging whether the first prediction error is smaller than or equal to a first preset prediction error threshold, if not, reconstructing a training model, and carrying out prediction analysis on the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data; S317, calculating a second antenna array element prediction data set and a second antenna array element real data set prediction error to obtain a second prediction error; S318, judging whether the second prediction error is smaller than or equal to a second preset prediction error threshold, if not, reconstructing a training model, and reconstructing the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data; S319, respectively predicting the optimized antenna array element construction data through the first convolutional neural network to obtain antenna array element performance prediction values corresponding to the optimized antenna array element construction data; S3110, respectively calculating the fitness value of a preset antenna element performance target value and each antenna element performance predicted value to obtain an antenna element fitness value corresponding to each optimized antenna element construction data; S3111, judging whether all the antenna array element fitness values are larger than or equal to a first preset fitness threshold value, if not, carrying out parameter updating on the first convolutional neural network according to all the antenna array element fitness values, and carrying out prediction analysis on the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data again after the parameter updating; S3112, judging whether the antenna array element prediction simulation result is a first preset simulation result, if not, inputting all the antenna array element performance prediction values into the historical antenna array element construction data set, returning to S312, and if so, taking the first convolutional neural network as a target antenna array element construction model.
7. The deep learning-based antenna array construction method according to claim 5, wherein the original antenna array prediction network includes a deep neural network and a second convolutional neural network, the training the original antenna array prediction network according to the target antenna array element prediction data and all the historical antenna array construction data, and the obtaining the target antenna array prediction network includes: S331, collecting target antenna array element prediction data and all historical antenna array construction data to obtain a historical antenna array construction data set; s332, dividing the historical antenna array construction data set into a first historical antenna array construction data subset and a second historical antenna array construction data subset according to a preset space parameter division rule; S333, extracting features of the first historical antenna array construction data subset through the deep neural network to obtain a plurality of first antenna array construction features; S334, extracting features of the second historical antenna array construction data subset through the second convolutional neural network to obtain a plurality of second antenna array construction features; S335, respectively performing feature stitching on each first antenna array construction feature and each second antenna array construction feature to obtain a plurality of first antenna array performance prediction values, and collecting all the first antenna array performance prediction values to obtain an antenna array performance prediction set; S336, importing an antenna array real data set corresponding to the historical antenna array construction data set, and calculating a prediction error of the antenna array performance prediction set and the antenna array real data set to obtain a third prediction error; S337, judging whether the third prediction error is smaller than or equal to a third preset prediction error threshold, if not, reconstructing a training model, and reconstructing the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data; s338, dividing the original optimization variable group into a first divided optimization variable group and a second divided optimization variable group according to the preset space parameter dividing rule; s339, extracting features of the first divided optimization variable group through the deep neural network to obtain a plurality of first optimization variable features; s3310, extracting features of the second divided optimization variable group through the second convolutional neural network to obtain a plurality of second optimization variable features; s3311, respectively performing feature stitching on each first optimized variable feature and each second optimized variable feature to obtain a plurality of second antenna array performance predicted values; S3312, calculating the fitness value of the preset antenna array performance target value and each second antenna array performance predicted value respectively to obtain an antenna array fitness value corresponding to each second antenna array performance predicted value; S3313, judging whether all the antenna array fitness values are larger than or equal to a second preset fitness threshold value, if not, carrying out parameter updating on the original antenna array prediction network through a preset parameter updating rule, and carrying out prediction analysis on the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data after parameter updating; s3314, judging whether the antenna array prediction simulation result is a second preset simulation result, if not, inputting all second antenna array performance prediction values into the historical antenna array construction data set, returning to S332, and if so, taking the original antenna array prediction network as a target antenna array prediction network.
8. The method for constructing an antenna array based on deep learning according to claim 1, wherein the process of performing a consistency analysis on the target antenna array data to obtain an antenna array construction result comprises: Performing outlier processing on the target antenna array data to obtain processed antenna array data; Filtering the processed antenna array data to obtain filtered antenna array data; Normalizing the filtered antenna array data to obtain normalized antenna array data; extracting features of the normalized antenna array data through a third convolutional neural network to obtain original antenna array features; Extracting time sequence features of the original antenna array features through a cyclic neural network to obtain the antenna array time sequence features; weighting the time sequence characteristics of the antenna array through an attention mechanism layer to obtain an original antenna array predicted value; performing inverse normalization processing on the original antenna array predicted value to obtain an inverse normalized antenna array predicted value; and carrying out smoothing treatment on the inverse normalized antenna array predicted value to obtain a smoothed antenna array predicted value, and taking the smoothed antenna array predicted value as an antenna array construction result.
9. Antenna array construction device based on degree of depth study, characterized by comprising: The system comprises an importing module, a judging module and a judging module, wherein the importing module is used for importing key parameters of an antenna array element to be built, key parameters of an antenna array to be built, a plurality of historical antenna array element building data and a plurality of historical antenna array building data; The parameter analysis module is used for respectively carrying out parameter analysis on the key parameters of the antenna array element to be constructed and the key parameters of the antenna array to be constructed to obtain target gradient slot line coordinates and antenna array factors; The prediction analysis module is used for constructing a training model, and predicting and analyzing the key parameters of the antenna array elements to be constructed, the key parameters of the antenna array to be constructed, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element construction data and all the historical antenna array construction data to obtain target antenna array data; and the construction result obtaining module is used for carrying out consistency analysis on the target antenna array data to obtain an antenna array construction result.
10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the deep learning-based antenna array construction method according to any one of claims 1 to 8.

Description

Deep learning-based antenna array construction method, device and storage medium Technical Field The invention mainly relates to the technical field of antenna construction, in particular to an antenna array construction method and device based on deep learning and a storage medium. Background The Vivaldi antenna array has wide band characteristics, high gain, good directivity and other advantages, and is widely applied to the fields of radar detection, wireless communication, electronic countermeasure and the like. The traditional Vivaldi antenna array construction mainly depends on experience and theoretical deduction of constructors, in the construction process, multiple parameter analyses such as antenna array element technical indexes, array element spacing, array element position distribution, array element number and the like are needed to be considered, and then repeated simulation iterative optimization is carried out through electromagnetic simulation software so as to obtain a construction scheme meeting performance requirements. The traditional antenna array construction method mainly has the following defects that (1) the antenna array construction period is long, the antenna array construction is difficult to meet the market demand of rapid change, which is often required to be completed in weeks or months, (2) the antenna array construction is more involved in parameters, the construction result is greatly influenced by engineers experience, the performance of the array constructed by different engineers is possibly greatly different, the overall optimal construction is difficult to realize, the side lobe level of the array is higher, the directivity is not ideal, and the like, and (3) the antenna array beam pattern distortion, main lobe pointing offset and side lobe level elevation are caused due to the influence of hardware difference, processing assembly errors, mutual coupling among array elements and the like of the antenna array, the antenna array beam pattern uniformity is difficult to be caused by actual measurement data of the antenna array, the antenna array amplitude uniformity data acquisition is carried out by the traditional method based on the combination of an external radiation source and an internal correction source, the antenna array amplitude uniformity data acquisition is carried out according to a certain frequency step and a certain step, the antenna array amplitude uniformity data acquisition is complex, the operation is easy to realize, the work load is huge, and the antenna array amplitude uniformity is difficult to be fully adapted to the environment variation due to the fact that the antenna array uniformity is not uniform at all frequency points and space intervals. Disclosure of Invention The invention aims to solve the technical problem of providing an antenna array construction method, an antenna array construction device and a storage medium based on deep learning aiming at the defects of the prior art. The technical scheme for solving the technical problems is as follows, the method for constructing the antenna array based on deep learning comprises the following steps: Importing key parameters of an antenna array element to be constructed, key parameters of the antenna array to be constructed, construction data of a plurality of historical antenna array elements and construction data of a plurality of historical antenna arrays; Respectively carrying out parameter analysis on the key parameters of the antenna array elements to be constructed and the key parameters of the antenna array to be constructed to obtain target gradient slot line coordinates and antenna array factors; building a training model, and carrying out predictive analysis on the key parameters of the antenna array elements to be built, the key parameters of the antenna array to be built, the target gradient slot line coordinates and the antenna array factors through the training model, all the historical antenna array element building data and all the historical antenna array building data to obtain target antenna array data; And carrying out consistency analysis on the target antenna array data to obtain an antenna array construction result. The invention solves the technical problems as follows, the antenna array constructing device based on deep learning comprises: The system comprises an importing module, a judging module and a judging module, wherein the importing module is used for importing key parameters of an antenna array element to be built, key parameters of an antenna array to be built, a plurality of historical antenna array element building data and a plurality of historical antenna array building data; The parameter analysis module is used for respectively carrying out parameter analysis on the key parameters of the antenna array element to be constructed and the key parameters of the antenna array to be constructed to obtain target gradient slot line coordinates and antenna