Neural network based antenna beam calibration method

Abstract

The present application belongs to the technical field of beam pointing accuracy correction of phased array antenna, and particularly relates to an antenna beam calibration method based on neural network. The specific process is as follows: a plurality of phased array antennas of the same specification and process are selected, the beam pointing is measured on a turntable according to the elevation angle interval kΔθ, and the data is preprocessed; the elevation angle is taken as the input of the neural network, and the beam pointing error is taken as the output of the neural network, and the neural network is trained; for each phased array antenna of the same specification and process, the beam pointing is measured on the turntable according to the elevation angle interval mΔθ, and the data is preprocessed; the preprocessed results corresponding to each phased array antenna are used to fine-tune the trained neural network, and the neural network corresponding to each phased array antenna is obtained; the beam pointing of the elevation angle interval Δθ is predicted by using the fine-tuned neural network, a lookup table is constructed and written into the corresponding phased array antenna, and the antenna beam calibration is completed by using the lookup table.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of beam pointing accuracy correction of phased array antennas, and particularly relates to an antenna beam calibration method based on a neural network. BACKGROUND

[0002] In recent years, phased array antennas have increasingly prominent advantages and are widely used in military and civilian fields. Beam pointing accuracy is one of the important technical indicators of phased array antennas. In order to improve the pointing accuracy, beam pointing correction is required. Traditional correction methods include turntable measurement and interpolation. The turntable measurement method measures all beam pointing directions in turn to establish a lookup table, and this method has high accuracy but high cost and long time consumption. The interpolation method uses partial beam pointing directions for interpolation to establish a lookup table, which can reduce the number of measurements, but the degree is limited, the efficiency is low, and the accuracy is not high.

[0003] Neural networks have unique network structures and information transmission mechanisms. The core feature of neural networks is to process data through multiple connected neuron layers, and the output of each layer is used as the input of the next layer. This structure enables neural networks to learn and capture complex patterns and relationships in input data. By adjusting the connection weights between neurons, neural networks can model and predict input data, which is very suitable for processing various types of data. For example, when processing beam pointing errors, neural networks can learn and identify error patterns in input data to more accurately predict and correct errors. Currently, no researchers have applied neural networks to beam pointing correction. SUMMARY

[0004] Therefore, the purpose of the present application is to provide an antenna beam calibration method based on a neural network to overcome the shortcomings of the prior art, ensure the accuracy of beam pointing correction, and significantly improve the calibration speed and system real-time performance.

[0005] The technical solutions of the present application are as follows:

[0006] In the first aspect, an antenna beam calibration method based on a neural network is used to calibrate the beam pointing of multiple phased arrays of the same specification and process. The specific process is as follows:

[0007] Select multiple phased array antennas of the same specification and process to measure the beam pointing on the turntable at an elevation angle interval of kΔθ, and pre-process the data. The elevation angle is used as the input of the neural network, and the beam pointing error is used as the output of the neural network for neural network training.

[0008] For each phased array antenna of the same specification and process, the beam pointing is measured on the turntable according to the elevation angle interval mΔθ, and the data is preprocessed, wherein m>k; the preprocessed result corresponding to each phased array antenna is used to fine-tune the trained neural network, so as to obtain the neural network corresponding to each phased array antenna;

[0009] The fine-tuned neural network is used to predict the beam pointing with the elevation angle interval Δθ, a lookup table is constructed and written into the corresponding phased array antenna, and the lookup table is used to complete the antenna beam calibration.

[0010] Further, the phased array antenna according to the application measures the beam pointing on the turntable according to the elevation angle interval kΔθ, and the specific process is as follows:

[0011] The measurement range of the elevation angle is set to -θ M ~ θ M , and the measurement range of the azimuth angle is set to -φ M ~ φ M .

[0012] The turntable completes a left-to-right azimuth plane scan, moves down kΔθ, and then completes a right-to-left azimuth plane scan, moves down kΔθ again, and so on, until the scan in the measurement range is completed, and the beam pointing is measured in the above scan process.

[0013] The measured beam pointing is subtracted from the actual beam pointing in the above scan process to obtain the beam pointing error data at all angles.

[0014] Further, the data preprocessing according to the application is as follows: the beam pointing error is equally spaced in the azimuth plane, and is averaged; the error data is separated according to the left and right scans of the azimuth plane, and is used as sample data for training of the neural network, wherein the neural network includes two, and the left / right scan data is input into the neural network for training.

[0015] Further, the application also includes a verification link for the precision of the neural network, and the verification indexes include two, which are azimuth 3σ and elevation 3σ; wherein the calculation method of the azimuth and elevation angle 3σ is as follows: the three times standard deviation 3σ of the difference between the predicted beam pointing angle and the actual measured beam pointing angle is calculated.

[0016] Further, the fine-tuning of the trained neural network by using the preprocessed result corresponding to each phased array antenna is as follows:

[0017] The hidden layer parameters of the pre-trained network are frozen, and two additional hidden layers are added, then the measured beam pointing is input into the pre-trained neural network, and a small learning rate is used for training, so as to realize the fine-tuning of the network weight parameters.

[0018] Further, the application predicts all scanning data, including: predicting left scanning data and right scanning data respectively by using the trained neural network according to the pitch angle interval Δθ; and integrating the predicted left scanning data and right scanning data to obtain all beam pointing directions.

[0019] Further, when the phased array is directed to a certain beam pointing angle, the phased array no longer issues the theoretical beam pointing direction to the phased array according to the lookup table, but the corresponding predicted beam pointing direction, and the predicted beam pointing direction replaces the theoretical beam pointing direction.

[0020] In a second aspect, a neural network-based antenna beam calibration method is used to calibrate the beam pointing direction of a phased array, and the specific process is as follows:

[0021] The beam pointing direction of the phased array antenna on the turntable is measured at a pitch angle interval of kΔθ, and the data is preprocessed;

[0022] The pitch angle is used as the input of the neural network, and the beam pointing error is used as the output of the neural network, and the neural network is trained;

[0023] The trained neural network is used to predict the beam pointing direction at a pitch angle interval Δθ, a lookup table is constructed and written into the phased array antenna, and the lookup table is used to complete the antenna beam calibration.

[0024] Advantages:

[0025] The phased array antenna beam pointing fast calibration method provided by the application can predict the error at all angles under the condition that the measurement is not more than 1 / k angle. By introducing transfer learning, the error at all angles can be further predicted under the condition that the measurement is not more than 1 / m (m>k) angle. The standard deviation 3σ of the absolute value of the deviation between the predicted error and the true error is less than σ std (σ std , which is the standard of the specified prediction result, and the prediction accuracy is high. The predicted beam pointing direction is introduced into the lookup table and written into the phased array antenna, so that the pointing accuracy of the phased array antenna can be corrected. The method provided by the application effectively improves the beam pointing accuracy and real-time performance of the phased array system, and has low cost and low system complexity. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0027] Figure 1A diagram illustrating the beam pointing angle error at similar elevation angles;

[0028] Figure 2 A diagram illustrating the difference probability density distribution of the beam pointing angle error between similar elevation angles;

[0029] Figure 3 A diagram illustrating the beam pointing angle error at non-similar elevation angles;

[0030] Figure 4 A diagram illustrating the flow of the beam pointing fast calibration method based on the neural network provided by the present application;

[0031] Figure 5 A diagram illustrating the measurement trajectory required by the existing calibration method;

[0032] Figure 6 A diagram illustrating the measurement trajectory provided by the method embodiment of the present application;

[0033] Figure 7 A diagram illustrating the interpretation of the result verification index in the method embodiment of the present application. DETAILED DESCRIPTION

[0034] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0035] It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict; and based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present disclosure.

[0036] It should be noted that the various aspects of the embodiments described below are within the scope of the appended claims. It should be apparent that the aspects described herein can be embodied in a wide variety of forms and that any specific structure and / or function described herein is merely illustrative. Based on the teachings herein one skilled in the art should appreciate that an aspect described herein can be implemented independently of any other aspects and that two or more aspects can be combined in any suitable manner. For example, an apparatus can be implemented and / or a method can be practiced using any number of the aspects described herein. In addition, such an apparatus can be implemented and / or such a method can be practiced using other structure and / or functionality in addition to or other than one or more of the aspects described herein.

[0037] During the research process, the applicant discovered that existing methods for antenna calibration using neural networks typically collect data according to accuracy requirements. For example, when the accuracy requirement for antenna beam measurement is Δθ, data is collected at intervals of Δθ. The neural network is then trained based on the collected data, which theoretically can meet the corresponding accuracy requirements. However, the above process overlooks an important phenomenon: the overall distribution characteristics of beam pointing error are similar and gradually change between similar elevation angles.

[0038] Similarity refers to the fact that the overall distribution characteristics of beam pointing error are similar between similar elevation angles, such as... Figure 1 As shown, Figure 1 Taking azimuth error as an example, it shows that the beam pointing error at similar elevation angles is within the entire azimuth plane (-θ). M ~θ M The fluctuation characteristics are similar to those of other words. Figure 2 The figure shows the probability distribution density of the differences between beam pointing errors at various elevation angles. In the figure, 98% or more of the differences between beam pointing errors at various elevation angles are within ±Δθ, indicating that the magnitudes of the beam pointing errors at various elevation angles are relatively similar, demonstrating the similarity of the distribution characteristics of the beam pointing errors.

[0039] Gradual change refers to the fact that the overall distribution characteristics of beam pointing error gradually change with the elevation angle, specifically as follows: Figure 6 . Figure 6 The beam error was plotted with intervals of kΔθ (k=10, pitch angle 1, pitch angle 11, pitch angle 21, pitch angle 31), and compared. Figure 1 and Figure 3 It can be seen that the overall distribution of the pitch angles 1-7 is similar, while the overall distribution of the pitch angles 1, 11, 21, and 31 changes gradually with the pitch angle, indicating that the distribution characteristics of the beam pointing error have a gradual change.

[0040] The following explains the principle behind the distribution characteristics of neural network learning errors: The core of a neural network is the state h of its hidden layer neurons. n It is updated for each sample n and depends on the current input sample x. n :

[0041] h n =f(W x ·x n +b)

[0042] Among them, h n It represents the hidden neuron state of sample n; W x It is the input weight matrix; x n b is the input of sample n; b is the bias vector; f is the activation function (such as tanh or ReLU).

[0043] In the embodiment, for the similar pitch angles, the corresponding input sequences x n will have small changes. When updating the hidden neuron state h n at each sample, the neural network mainly depends on the current input x n . Due to the small changes in the input, the changes in the hidden neuron state are also small, and the characteristics of the hidden neuron state show similarity. When the characteristics between the input samples gradually change, the hidden neuron state of the neural network is also gradually adjusted, thereby capturing the gradual changes in the input data.

[0044] In the output stage, the neural network predicts the output y n according to the current hidden neuron state h n , that is, the beam pointing angle:

[0045] y n =g(W y ·h n +c)

[0046] where y n is the output of sample n; W y is the weight matrix of the output layer; c is the bias vector of the output layer; and g is the activation function of the output layer.

[0047] By training the neural network, that is, training and optimizing the weight matrices W x and W y , the relationship between the input sequence x n and the output beam pointing angle y n is learned, and the neural network can accurately predict the beam pointing angle error according to the new input pitch angle.

[0048] Based on the analysis of similarity and gradual change, in the neural network training, the input data does not have to be the data collected at the interval of Δθ, but the interval of K times Δθ is set based on the required accuracy of Δθ, that is, the pitch angle is collected at the interval of kΔθ, thereby overcoming the problem of too long measurement time caused by small measurement interval.

[0049] An embodiment of the application is a neural network-based antenna beam calibration method, which is used for calibrating the beam pointing of a phased array, and the specific process is as follows:

[0050] S11: Select the phased array antenna to measure the beam pointing at the interval of kΔθ on the turntable according to the pitch angle, and pre-process the data;

[0051] S12: Take the pitch angle as the input of the neural network and the beam pointing error as the output of the neural network, and train the neural network.

[0052] S13: predicting the beam pointing of the pitch angle interval Δθ by using the trained neural network, constructing a lookup table and writing it into the phased array antenna, and completing the antenna beam calibration by using the lookup table.

[0053] On this basis, the application further uses the concept of transfer learning to use sample data of multiple phased arrays of the same specification and process, pre-train the neural network by using the collected results, and then collect small batch sample data for each phased array to fine-tune the pre-trained neural network, so that the fine-tuned neural network meets the demand for rapid calibration of the corresponding phased array beam pointing.

[0054] As shown in FIG. Figure 4 The embodiment of the application provides a neural network-based antenna beam calibration method, and the specific process is as follows:

[0055] S21: selecting multiple phased array antennas of the same specification and process to measure the beam pointing at the interval of kΔθ on the turntable, and pre-processing the data; taking the pitch angle as the input of the neural network and the beam pointing error as the output of the neural network, and training the neural network;

[0056] S22: for each phased array antenna of the same specification and process, measuring the beam pointing at the interval of mΔθ on the turntable, and pre-processing the data, wherein m>k; fine-tuning the trained neural network by using the pre-processing result corresponding to each phased array antenna to obtain the neural network corresponding to each phased array antenna;

[0057] S23: predicting the beam pointing of the pitch angle interval Δθ by using the fine-tuned neural network, constructing a lookup table and writing it into the corresponding phased array antenna, and completing the antenna beam calibration by using the lookup table.

[0058] The embodiment of the application is aimed at phased array radar antennas of the same specification and process, and only needs to train a basic neural network once, and then only needs to fine-tune a small amount of data for each phased array individual, so that the purpose of calibrating multiple phased array antennas of the same specification and batch can be quickly achieved. Meanwhile, kΔθ is used as the interval in the training in the embodiment, so that the number of samples required for measurement can be reduced to 1 / k, and the measurement time is also reduced to 1 / k.

[0059] In the determination of the appropriate k value, the sample data at the pitch angle interval Δθ can be measured in advance, and the pitch angle interval kΔθ of the sample data is sampled on this basis, the sampled data is input into the network training and the data of the pitch angle interval Δθ is predicted, and then the prediction result is compared with the result measured in advance, and the accuracy reaches the required accuracy to determine the appropriate k value.

[0060] In another embodiment of the present application, the phased array antenna measures the beam pointing at an elevation angle interval of kΔθ on the turntable, and the specific process is as follows:

[0061] The measurement range of the elevation angle is set to -θ M ~ θ M , and the measurement range of the azimuth angle is set to -φ M ~ φ M .

[0062] The turntable completes a left-to-right azimuth plane scan, moves down by kΔθ, then completes a right-to-left azimuth plane scan, moves down by kΔθ again, and so on until the scan in the measurement range is completed. The beam pointing (including the elevation angle and the azimuth angle) is measured in the above scanning process. The azimuth plane angle interval in this process is determined by the turntable rotation speed and the PRF.

[0063] The measured beam pointing is subtracted from the actual beam pointing in the above scanning process to obtain the beam pointing error data at all angles.

[0064] This embodiment is executed according to the above process to measure all beam pointing errors. As shown in FIG. 8, for ease of display, the elevation angle is only taken as -θ1~θ1, θ1<θ M , and the traditional beam acquisition is as shown in FIG. 9. Figure 6 Figure 5

[0065] In another embodiment of the present application, the data is preprocessed as follows: the beam pointing error is equally spaced in the azimuth plane and is averaged; the error data is separated by left and right scanning in the azimuth plane as sample data for training of a neural network, and the neural network includes two, and the left / right scanning data is input into the neural network for training. In this embodiment, the measured elevation angle is taken as a sample feature, and the left and right scanning data are taken as sample labels and input into the neural network for training.

[0066] In another embodiment of the present application, the neural network type is selected as MLP, LSTM or GRU; the neural network structure adopted in this embodiment includes three hidden neuron layers with 512 units, and each hidden neuron layer is followed by a ReLU activation layer to enhance the nonlinear feature extraction capability. The input is a single variable sequence of the elevation angle, which is processed through these hidden layers and connected to a fully connected layer to output the regression result. The neural network uses the He initialization method, adopts the Adam optimization algorithm, sets an adaptive learning rate adjustment strategy, and is trained on a GPU, which is suitable for processing and predicting sequence data.

[0067] ​​The embodiment of the application further includes a verification link for the precision of the neural network, and the verification indexes include two indexes, i.e., azimuth 3σ and elevation 3σ. The calculation method of the azimuth-elevation angle 3σ is that the three times standard deviation 3σ of the difference between the predicted beam pointing angle and the actual measured beam pointing angle is calculated. In this embodiment, the beam pointing error output by the neural network is used to verify the accuracy of the neural network trained in step S2, and the verification indexes include two indexes, i.e., azimuth 3σ and elevation 3σ. The calculation method of the azimuth-elevation angle 3σ is that the three times standard deviation (3σ) of the difference between the predicted beam pointing angle and the actual measured beam pointing angle is calculated.

[0068] It should be noted that in this embodiment, the target of calibration is that the error of the theoretical beam pointing of the phased array after calibration and the measured beam pointing satisfies 3σ < σ std , and the calculated index is the 3σ of the error between the predicted beam pointing of the phased array and the measured beam pointing. The specific explanation is as follows:

[0069] After the lookup table is constructed and written into the phased array, when the phased array points to a certain beam pointing angle, the phased array no longer issues the theoretical beam pointing to the phased array according to the lookup table, but the corresponding predicted beam pointing. This process can be regarded as using the predicted beam pointing to replace the theoretical beam pointing, as shown in FIG. 8. Figure 7 Therefore, only the 3σ of the error between the predicted angle before calibration and the measured angle needs to be guaranteed. std .

[0070] In another embodiment of the application, the trained neural network is fine-tuned by using the pre-processing result corresponding to each phased array antenna, specifically, all the hidden layer parameters of the pre-trained network are frozen, and two additional hidden layers are added, then the measured beam pointing is input into the pre-trained neural network, and the network is trained by using a small learning rate, so as to realize fine-tuning of the network weight parameters.

[0071] In another embodiment of the application, under the premise of distinguishing the left-scan and right-scan data, the elevation angle (elevation angle interval Δθ) of the unmeasured beam pointing is sequentially input into the trained neural network model as a sample feature, and the sample label output by the network is the predicted beam pointing error under the corresponding elevation angle.

[0072] In another embodiment of the application, the trained neural network is used to predict the left-scan data and the right-scan data according to the elevation angle interval Δθ; the predicted left-scan data and the right-scan data are integrated to obtain all the beam pointings, a lookup table is constructed, the lookup table is written into the phased array antenna, and the calibration is completed.

[0073] The measured part and the predicted part are used to construct the lookup table with the azimuth-elevation angle interval Δθ.

[0074] Based on the above method specific implementation process S11-S13, 10 different phased array of measured data were simulated (in simulation experiment, Δθ = 0.2°, k = 10, σ std

[0075] Table 1 MLP network prediction beam pointing error results

[0076] Phased array number Azimuth 3σ (°) Elevation 3σ (°) 1 0.1690 0.1747 2 0.1554 0.1929 3 0.1739 0.1687 4 0.1788 0.2232 5 0.1681 0.1646 6 0.1915 0.1649 7 0.1901 0.2035 8 0.1550 0.1839 9 0.1626 0.1760 10 0.1787 0.1956

[0077] Table 2 LSTM network prediction beam pointing error results

[0078] Phased array number Azimuth 3σ (°) Elevation 3σ (°) 1 0.1601 0.1565 2 0.1489 0.1791 3 0.1709 0.1623 4 0.1598 0.1897 5 0.1538 0.1597 6 0.1781 0.1491 7 0.1845 0.1751 8 0.1542 0.1768 9 0.1466 0.1664 10 0.1639 0.1646

[0079] Table 3 GRU network prediction beam pointing error results

[0080]

[0081]

[0082] In the verification results, the simulation verification of 10 groups of data of the three networks all reached the standard, and the effectiveness of the method can be seen.

[0083] Based on the above method specific implementation process S21-S23, the beam pointing data of 10 phased arrays of the same specification and process were used to pre-train the MLP network model, and the beam pointing data of two phased arrays measured at m = 2k and 3k were input into the network for fine-tuning and prediction, and the simulation verification results are shown in Table 4 and Table 5.

[0084] Table 4 Phased array 1 prediction beam pointing error results

[0085] Sample ratio Azimuth 3σ (°) Elevation 3σ (°) 1 / 30 0.2031 0.2479 1 / 20 0.1806 0.2080

[0086] Table 5 Phased array 2 prediction beam pointing error results

[0087] Sample ratio Azimuth 3σ (°) Elevation 3σ (°) 1 / 30 0.2158 0.2491 1 / 20 0.1665 0.1892

[0088] In the verification results, the simulation verification of 2 phased arrays all reached the standard, and the effectiveness of the method can be seen.

[0089] Further, in specific practice, the beam pointing accuracy of the phased array antenna can be corrected by searching for the corresponding predicted beam pointing error of the designed beam pointing.

[0090] ​The beam pointing fast calibration method provided by the embodiment can predict the beam pointing at all angles under the condition of measuring the beam pointing at only a few angles. Compared with the prior art, the method has fewer measurement times, greatly improved efficiency, high precision, low cost and is easy to implement in engineering.

[0091] It should be noted that in the description of the present application, the terms "first", "second", etc. are only for illustrative purposes, and do not indicate or imply relative importance. In addition, in the description of the present application, unless otherwise explicitly stated, the term "a plurality of" refers to at least two.

[0092] Any flowchart or other manner described process or method described in the present application can be understood as including one or more code modules, segments or portions of executable instructions for performing specific logical functions or processes. The scope of the preferred embodiments of the present application also includes other implementations that can not be in the order shown, as understood by those skilled in the art of the technology involved.

[0093] It can be understood that when describing the terms such as "one embodiment", "some embodiments", "example", "specific example" or "some examples" in the present application, it refers to at least one embodiment or example having specific features, structures, materials or characteristics. The use of these terms does not necessarily involve the same embodiment or example, and the specific features, structures, materials or characteristics in the description can be appropriately combined in one or more embodiments or examples.

[0094] The above detailed description further describes the purpose, technical method and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A neural network based antenna beam calibration method for calibrating the beam pointing of a plurality of phased arrays of the same specification and technology, characterized in that, The specific process is as follows: A plurality of phased array antennas of the same specification and process are selected to measure the beam pointing at an elevation angle interval of kΔθ on the turntable, and the data is preprocessed; the elevation angle is taken as the input of the neural network, and the beam pointing error is taken as the output of the neural network, and the neural network is trained; For each phased array antenna of the same specification and process, the beam pointing is measured at an elevation angle interval of mΔθ on the turntable, and the data is preprocessed, wherein m>k; the preprocessed results corresponding to each phased array antenna are used to fine-tune the trained neural network, and a neural network corresponding to each phased array antenna is obtained; The fine-tuned neural network is used to predict the beam pointing at an elevation angle interval of Δθ, a lookup table is constructed and written into the corresponding phased array antenna, and the lookup table is used to complete the antenna beam calibration.

2. The neural network based antenna beam calibration method of claim 1, wherein, The specific process is as follows: The measurement range of the pitch angle is set to -θ M ~ θ M , and the measurement range of the azimuth angle is set to -φ M ~ φ M ; The turntable completes a scan in the azimuth plane from left to right, moves downward by kΔθ, then completes a scan in the azimuth plane from right to left, moves downward by kΔθ again, and so on, until the scan in the measurement range is completed, and the beam pointing is measured in the above scan process; The measured beam pointing is subtracted from the actual beam pointing in the above scan process to obtain the beam pointing error data at all angles. 3.The neural network based antenna beam calibration method of claim 1, wherein, The preprocessed data is: the beam pointing error is equally spaced in the azimuth plane, and is averaged; the error data is separated according to the left scan and the right scan in the azimuth plane, and is used as sample data for training of the neural network, wherein the neural network includes two, and the left / right scan data is input into the neural network for training.

4. The neural network based antenna beam calibration method of claim 1, wherein, The verification link of the precision of the neural network is also included, and the verification indexes include two, which are azimuth 3σ and elevation 3σ; wherein the calculation method of the azimuth 3σ is that the three times standard deviation 3σ of the difference between the predicted beam pointing angle and the actual measured beam pointing angle is calculated.

5. The neural network based antenna beam calibration method of claim 1, wherein, The fine-tuning of the trained neural network by using the preprocessed results corresponding to each phased array antenna is as follows: All hidden layer parameters of the pre-trained network are frozen, and two additional hidden layers are added, then the measured beam pointing is input into the pre-trained neural network, and a small learning rate is used for training, so as to realize fine-tuning of the network weight parameters.

6. The neural network based antenna beam calibration method of claim 1, wherein, The predicted all-scan data includes: the left-scan data and the right-scan data are respectively predicted by using the trained neural network at an elevation angle interval of Δθ; and the predicted left-scan data and right-scan data are integrated to obtain all beam pointings. 7.The neural network based antenna beam calibration method of claim 1, wherein, After the lookup table is constructed and written into the phased array, when the phased array points to a certain beam pointing angle, the phased array no longer issues the theoretical beam pointing according to the lookup table, but the corresponding predicted beam pointing, and the predicted beam pointing replaces the theoretical beam pointing.

8. A neural network-based antenna beam calibration method for calibrating the beam pointing of a phased array, characterized in that, The specific process is as follows: A phased array antenna is selected to measure the beam pointing at an elevation angle interval of kΔθ on the turntable, and the data is preprocessed; The elevation angle is taken as the input of the neural network, and the beam pointing error is taken as the output of the neural network, and the neural network is trained; The trained neural network is used to predict the beam pointing of the pitch angle interval Δθ, a lookup table is constructed and written into the phased array antenna, and the antenna beam calibration is completed by using the lookup table.

Images