Phased-array antenna fault detection method based on sparse signal reconstruction algorithm

Abstract

The invention provides a phased-array antenna fault detection method based on a sparse signal reconstruction algorithm. The phased-array antenna fault detection method comprises the following implementation steps: initializing parameters; constructing an amplitude square data vector; constructing a sparse signal reconstruction model based on a sparse signal reconstruction algorithm; solving the sparse signal reconstruction model; and obtaining a phased-array antenna fault detection result. Based on the sparse signal reconstruction algorithm, the sparse signal reconstruction model is constructed by introducing the sparsity priori knowledge of the sparse signal matrix into the signal reconstruction model rewritten by the fault factor, the sparse signal reconstruction model is solved by using the convex optimization algorithm, and the fault state of the antenna unit is judged by using the solving result. The sparse priori knowledge is introduced into the objective function, additional guidance information is added in the solving process, and more measurement data is not needed, so that compared with the prior art, the detection efficiency is improved on the premise of ensuring the detection precision.

Description

Technical Field

[0001] The present invention belongs to the technical field of antennas and relates to a phased array antenna fault detection method, and in particular to a phased array antenna fault detection method based on a sparse signal reconstruction algorithm. Background Art

[0002] Phased array antenna is an antenna that changes the direction or shape of the beam by controlling the feed excitation of the antenna unit. It is widely used in radar, navigation, communication, microwave wireless energy transmission, etc. Since each antenna unit in the phased array antenna has a very obvious influence on the beamforming result, when a unit fails, it may cause adverse effects such as antenna beam deviation or beam shape change. After locating the faulty unit, the feed excitation of the antenna unit can be changed to eliminate the influence of the faulty unit. Therefore, it is very important to detect the fault status of each antenna unit.

[0003] For example, Xiong Can disclosed a phased array antenna fault detection method based on a signal reconstruction algorithm in "Excitation Retrieval for Phased Arrays With Magnitude-Only Fields Measured at a Fixed Location" published in the journal IEEE Antennas and Wireless Propagation Letters in February 2021. This method constructs a signal reconstruction model using the signal reconstruction algorithm and solves it to obtain the antenna unit excitation, and judges the fault state of the antenna unit by the solved antenna unit excitation. This method has high detection accuracy, but in the process of obtaining the antenna unit excitation, only the accuracy of the signal reconstruction algorithm is considered, and the sparsity of the faulty unit of the phased array antenna is ignored. A large amount of measurement data is required, which affects the improvement of detection efficiency. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and propose a phased array antenna fault detection method based on a sparse signal reconstruction algorithm, aiming to improve the detection efficiency while ensuring the detection accuracy.

[0005] To achieve the above object, the technical solution adopted by the present invention includes the following steps:

[0006] (1) Initialization parameters:

[0007] The initialization phased array antenna includes N antenna units arranged periodically, each antenna unit includes a phase shifter, the measurement point P with the microwave probe is located in the far field area with a distance D from the plane where the phased array antenna aperture is located, and the projection of P on the phased array antenna aperture is located at the center of the phased array antenna. The initial excitation applied to the nth antenna unit is w n The initial excitation vector and diagonal excitation matrix of the phased array antenna are w and W respectively, and the fault factor vector and matrix of the phased array antenna are s and S respectively, where N ≥ 10, and the initial value of the fault factor of the nth antenna unit is s n , 0≤s n ≤1;

[0008] (2) Construct the magnitude square data vector:

[0009] Each antenna element has a n Under the excitation of , the electric field amplitude at the measuring point P is measured M times by the microwave probe, and the electric field amplitude data b obtained by each measurement is m Calculate the magnitude squared data y m , and then the M amplitude square data are combined into an amplitude square data vector y = [y1, y2, ..., y m ,y M ] T , M ≥ 10, [·] T Represents a transpose operation;

[0010] (3) Construct a sparse signal reconstruction model based on the sparse signal reconstruction algorithm:

[0011] Based on the sparse signal reconstruction algorithm, the initial value of the sparse signal matrix ΔS is calculated by the initial value of the fault factor matrix S, and the sparse signal reconstruction model is constructed using ΔS and the amplitude square data vector y.

[0012] (4) Solve the sparse signal reconstruction model:

[0013] The sparse signal reconstruction model is solved using a convex optimization algorithm, and the sparse signal matrix ΔS is updated using the solution results;

[0014] (5) Obtaining phased array antenna fault detection results:

[0015] The average value of the sparse signal matrix ΔS is calculated row by row, and the fault factor vector s is updated using the calculation result. Then, each fault factor s is updated by the threshold η. n Discretization is performed and the result after discretization is used to calculate the fault factor s n Update and finally determine the fault factor s n =1 is true, if so, then s nThe corresponding antenna unit is in a fault state, otherwise it is in a normal state.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] The present invention is based on a sparse signal reconstruction algorithm. A sparse signal reconstruction model is constructed by introducing sparsity prior knowledge of a sparse signal matrix into a signal reconstruction model rewritten by a fault factor, and a convex optimization algorithm is used to solve it. The fault state of the antenna unit is judged by using the solution result. Since the sparsity prior knowledge is introduced into the objective function, additional guiding information is added in the solution process, and no more measurement data is required. Compared with the prior art, the detection efficiency is improved while ensuring the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart for implementing the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0020] Reference Figure 1 , the present invention comprises the following steps:

[0021] Step 1), initialization parameters:

[0022] The initialization phased array antenna includes N antenna units arranged periodically, each antenna unit includes a phase shifter, the measurement point P with the microwave probe is located in the far field area with a distance D from the plane where the phased array antenna aperture is located, and the projection of P on the phased array antenna aperture is located at the center of the phased array antenna. The initial excitation applied to the nth antenna unit is w n The initial excitation vector and diagonal excitation matrix of the phased array antenna are w and W respectively, and the fault factor vector and matrix of the phased array antenna are s and S respectively, where N ≥ 10, and the initial value of the fault factor of the nth antenna unit is s n , 0≤s n ≤1;

[0023] Each phase shifter is used to change the excitation phase of the antenna unit to which it belongs. The microwave probe is used to measure the electric field amplitude data at the measurement point P. The initial excitation w applied to the nth antenna unit n It contains the unit excitation amplitude and unit excitation phase. The diagonal excitation matrix W is a diagonal matrix with a dimension of N×N, which is composed of the initial excitation of each antenna unit. The distance D between the measurement point P and the plane where the phased array antenna aperture is located and the calculation formula of the fault factor matrix S are:

[0024] D>2L 2 / λ

[0025] S=s×s T

[0026] Where L and λ represent the aperture size and wavelength of the radiated electromagnetic wave of the phased array antenna, respectively. T Represents a transpose operation;

[0027] In this embodiment, the total number of phased array antenna units N=25, the antenna unit type is a dipole antenna, the wavelength λ of the electromagnetic wave radiated by the antenna unit is 5.17 cm, the unit spacing is 0.5λ, the aperture size L of the antenna is 3.7λ, and the distance D between the measurement point P and the surface where the phased array antenna aperture is located is 3L 2 / λ, the initial excitation w applied to each antenna element n All are unit excitation, and the failure factor of each antenna unit is s n The initial value of is 1;

[0028] For each antenna unit, the fault state of the antenna unit can be determined by the fault factor. When the fault factor is equal to 1, the antenna unit corresponding to the fault factor is in a normal state. When the fault factor is equal to 0, the antenna unit corresponding to the fault factor is in a fault state. The fault factor s n The initial value of does not represent the fault status of each antenna unit, but is only the initial value used for subsequent iterative solution using the convex optimization algorithm.

[0029] Step 2) Construct the magnitude square data vector:

[0030] Each antenna element has a n Under the excitation of , the electric field amplitude at the measurement point P is measured M times by the microwave probe. In each measurement process, an additional excitation phase is applied to the phase shifter of each antenna unit. And the electric field amplitude data b obtained by each measurement m Calculate the magnitude squared data y m , and then the M amplitude square data are combined into an amplitude square data vector y = [y1, y2, ..., y m ,...y M ] T , the amplitude square data calculation formula is:

[0031]

[0032] In this embodiment, M=100, the additional excitation phase The value of is obtained through Gaussian random distribution.

[0033] Step 3) Construct a sparse signal reconstruction model based on the sparse signal reconstruction algorithm:

[0034] The initial value of the sparse signal matrix ΔS is calculated by the initial value of the fault factor matrix S, and the sparse signal reconstruction model is constructed using ΔS and the amplitude square data vector y. The steps are as follows:

[0035] Step 3a) constructs a quadratic nonlinear optimization model for the relationship between the magnitude squared data vector y and the fault factor vector s:

[0036] find s

[0037] sty=|AWs 2

[0038]

[0039] Among them, A is the perception matrix with dimension M×N, e is a natural constant, j is an imaginary unit, and AEP n represents the array directional pattern of the nth unit, where the array directional pattern AEP of the nth unit n The simulation is obtained by applying unit excitation to the nth unit and connecting the remaining units to matching loads;

[0040] find s and sty = | AWs 2 They represent the failure factor vector s as the optimization variable and y = |AWs 2 is a constraint condition.

[0041] Step 3b) Rewrite the quadratic nonlinear optimization model through the fault factor matrix S to obtain a linear optimization model:

[0042] find S

[0043]

[0044] Where Tr(·) represents the trace operation of the matrix, Indicates a m The conjugate transpose result of can reduce the difficulty of solving by rewriting the quadratic nonlinear optimization model into a linear optimization model.

[0045] Step 3c) Rewrite the linear optimization model by using the low-rank property of the fault factor matrix S to obtain a rank minimization optimization model:

[0046] find S

[0047] min rank(S)

[0048]

[0049] S≥0

[0050] Among them, rank(·) is the rank operation. Since the failure factor matrix S = s×sT , so its rank is 1, which has the low rank property. Min rank(S) means that the minimum value of the rank(S) of the fault factor matrix S is used as the objective function.

[0051] Step 3d) When the fault factor matrix S is a Hermitian matrix and the linear operator When the finite isometry property is satisfied, the rank minimization optimization model is equivalent to the trace minimization optimization model:

[0052] find S

[0053] min Tr(S)

[0054]

[0055] S≥0

[0056] Among them, the fault factor matrix S is equal to the product of the fault factor vector s and its transpose. Therefore, the fault factor matrix S must satisfy the definition of the Hermitian matrix. The additional excitation angles with Gaussian random distribution can ensure the linear operator Satisfying the finite isometry property, since the objective function min rank(S) of the rank minimization optimization model is a non-convex function, equating it to a convex function min Tr(S) can reduce the difficulty of solving the problem;

[0057] Step 3e) transforms the trace minimization optimization model into a signal reconstruction model represented by the fault factor:

[0058] find S

[0059] min Tr(S)

[0060]

[0061] S≥0

[0062] Among them, ||·||2 represents the operation of obtaining the l2 norm of the matrix, ε is the noise level parameter, ε>0;

[0063] Constraints in the trace minimization optimization model is non-convex, rewrite it as a convex constraint It can reduce the difficulty of solving the model;

[0064] Step 3f) Calculate the initial value of the sparse signal matrix ΔS by using the all-one vector I with dimension N×1 and the initial value of the fault factor matrix S:

[0065] ΔS=II T -S

[0066] Step 3g) Add the sparsity prior knowledge of the sparse signal matrix ΔS to the objective function to construct a sparse signal reconstruction model:

[0067] find ΔS

[0068] min Tr(S)+ρ||ΔS|| TV

[0069]

[0070] S=II T -ΔS

[0071] S≥0

[0072]

[0073] Where ρ is the regularization parameter, ρ>0, ||·|| TV represents the operation of obtaining the total variation norm of the matrix, vec(·) represents the vectorization operator, and ||·||1 represents the operation of obtaining the 1-norm. and Respectively represent the horizontal and vertical discrete gradient matrices with dimensions of N×N;

[0074] Since the number of faulty units in the phased array antenna is much smaller than the total number of units, the true solution of the sparse signal matrix ΔS is sparse. The sparse signal reconstruction algorithm is used to introduce the sparsity prior knowledge, that is, the total variation norm of the sparse signal matrix ΔS ||ΔS|| is added to the objective function. TV By introducing additional constraints to guide the direction of iteration when solving the model, the constraints are reduced. The demand for intensity, that is, the demand for the amount of measurement data, is reduced, thus achieving the purpose of improving detection efficiency.

[0075] Step 4) Solve the sparse signal reconstruction model:

[0076] The sparse signal reconstruction model is solved using a convex optimization algorithm, and the solution result is used to update the sparse signal matrix.

[0077] In this embodiment, the sparse signal reconstruction model is solved by using the default algorithm of the CVX2.9 toolbox, which is an open source convex optimization model solver and includes a variety of iteration-based convex optimization algorithms.

[0078] Step 5) Obtain phased array antenna fault detection results:

[0079] The average value of the updated sparse signal matrix ΔS is calculated row by row, and the fault factor vector s is updated using the calculation result. Then, each fault factor s is averaged by the threshold η. nDiscretization is performed, and the fault factor vector is updated using the discretization result. Finally, the fault factor s is determined. n =1 is true, if so, then s n The corresponding antenna unit is in a fault state, otherwise it is in a normal state, where the formula for discretization processing is:

[0080]

[0081] Among them, 0<η<1.

Claims

1. A phased array antenna fault detection method based on sparse signal reconstruction algorithm, characterized in that: The steps include: (1) Initialization parameters: The initialization phased array antenna includes N antenna units arranged periodically, each antenna unit includes a phase shifter, the measurement point P with the microwave probe is located in the far field area with a distance D from the plane where the phased array antenna aperture is located, and the projection of P on the phased array antenna aperture is located at the center of the phased array antenna. The initial excitation applied to the nth antenna unit is w n The initial excitation vector and diagonal excitation matrix of the phased array antenna are w and W respectively, and the fault factor vector and matrix of the phased array antenna are s and S respectively, where N ≥ 10, and the initial value of the fault factor of the nth antenna unit is s n , 0≤s n ≤1; (2) Construct the magnitude square data vector: Each antenna element has a n Under the excitation of , the electric field amplitude at the measuring point P is measured M times by the microwave probe, and the electric field amplitude data b obtained by each measurement is m Calculate the magnitude squared data y m , and then the M amplitude square data are combined into an amplitude square data vector y = [y1, y2, ..., y m ,y M ] T , M ≥ 10, [·] T Represents a transpose operation; (3) Construct a sparse signal reconstruction model based on the sparse signal reconstruction algorithm: Based on the sparse signal reconstruction algorithm, the initial value of the sparse signal matrix ΔS is calculated by the initial value of the fault factor matrix S, and the sparse signal reconstruction model is constructed using ΔS and the amplitude square data vector y. (4) Solve the sparse signal reconstruction model: The sparse signal reconstruction model is solved using a convex optimization algorithm, and the sparse signal matrix ΔS is updated using the solution results; (5) Obtaining phased array antenna fault detection results: The average value of the sparse signal matrix ΔS is calculated row by row, and the fault factor vector s is updated using the calculation result. Then, each fault factor s is updated by the threshold η. n Discretization is performed and the result after discretization is used to calculate the fault factor s n Update and finally determine the fault factor s n =1 is true, if so, then s n The corresponding antenna unit is in a fault state, otherwise it is in a normal state.

2. The method according to claim 1, characterized in that: The distance D from the plane where the phased array antenna aperture is located in step (1) is calculated as follows: D>2L 2 / λ Among them, L and λ represent the aperture size and wavelength of the radiated electromagnetic wave of the phased array antenna respectively.

3. The method according to claim 1, characterized in that: The diagonal excitation matrix W of the phased array antenna described in step (1) represents a diagonal matrix of dimension N×N with the initial excitation of each antenna unit as an element.

4. The method according to claim 1, characterized in that: The magnitude square data y described in step (2) m , the calculation formula is:

5. The method according to claim 1, characterized in that The steps of constructing the sparse signal reconstruction model described in step (3) are as follows: (3a) The initial value of the sparse signal matrix ΔS is calculated by the initial value of the all-one vector I with dimension N×1 and the fault factor matrix S: ΔS=II T -S Among them, I T represents the transposed result of I; (3b) The linear operator of the squared amplitude data y on the fault factor matrix S Constraints are imposed and the total variation norm ||ΔS|| of the trace Tr(S) of S and the sparse signal matrix ΔS is constructed. TV The minimum value of the weighted sum is the sparse signal reconstruction model of the objective function: findΔS min Tr(S)+ρ||ΔS|| TV S≥0 Among them, ||·|| TV represents the operation of obtaining the total variation norm, ρ is the regularization parameter, ε represents the noise level parameter, ε>0, ρ>0, and ||·||2 represents the operation of obtaining the 2-norm.

6. The method according to claim 1, characterized in that For each fault factor s described in step (5), n Discretization is performed, and the formula is: Among them, 0<η<1.

Images