ACA and improved AWE-based electrically large-size target broadband RCS acquisition method
An improved asymptotic waveform estimation method using an adaptive cross-approximation method and a pseudo-spectral derivative method solves the problems of low efficiency in acquiring the wideband radar cross section of electrically large targets and high memory requirements, achieving more efficient computation and lower memory consumption.
Patent Information
- Application Number
- CN202411157636.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-22
AI Technical Summary
Existing technologies are inefficient and require a lot of computer memory when acquiring broadband radar cross sections of electrically large targets, especially when iteratively calculating matrix-vector products and filling high-order impedance matrices, which involve excessive computational complexity and memory requirements.
An adaptive cross-approximation method (ACA) is used to compress the far-field impedance matrix, and the pseudospectral derivative method (PSDM) is used to improve the asymptotic waveform estimation method (AWE). The high-order impedance matrix is represented by the decomposition of the near-field and far-field impedance matrices and the pseudospectral differential matrix, which reduces the computational complexity and memory requirements.
It effectively improves the efficiency of radar cross section acquisition, reduces computer memory requirements, and enhances computing speed and efficiency.
Smart Images

Figure CN119026363A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of electromagnetic simulation, and relates to a method for acquiring a broadband RCS of an electrically large-sized target, and in particular to a method for acquiring a broadband RCS of an electrically large-sized target based on ACA and improved AWE. Background Art
[0002] Radar cross section (RCS) is a physical quantity that quantitatively characterizes the scattering intensity of incident waves by targets. With the development of science and technology, radar detection systems and stealth technology have been continuously improving, which has put forward higher requirements for the rapid and accurate acquisition of RCS. As one of the most commonly used numerical calculation methods for electromagnetic fields, the method of moments (MoM) can accurately analyze the RCS of targets. However, the size of electrically large targets is usually much larger than the wavelength of electromagnetic waves. When processing their electromagnetic characteristics, more physical effects and computational complexity need to be considered. MoM has large memory requirements and long calculation time, which makes it difficult to solve electromagnetic problems of electrically large targets. For broadband problems, MoM needs to repeatedly fill the impedance matrix at each frequency point, and because the method of moments generates a dense matrix, a lot of computing time will be wasted when solving it. Therefore, it is of great significance to study the broadband RCS acquisition method of electrically large targets.
[0003] Xidian University proposed a method for obtaining broadband RCS of periodic structures based on CBFM and AWE in its patent application "Method for obtaining broadband RCS of periodic structures based on CBFM and AWE" (application date: May 15, 2023, application number: 202310540021.1, application publication number: CN 116484642A), and the implementation steps are: initialization parameters; calculation of the excitation vector of each sub-region; calculation of the induced current of each sub-region at each frequency corresponding to the wave number; and acquisition of broadband RCS of periodic structures. The invention divides the entire target into many sub-regions, constructs a basis function on each sub-region using the characteristic basis function method CBFM, and reflects the interaction between sub-regions through secondary characteristic basis functions, thereby achieving the reduction of the impedance matrix, avoiding the defect of the impedance matrix being too large due to too many unknown quantities, thereby reducing the complexity of solving the impedance matrix, and using the progressive waveform estimation AWE to calculate the current of the remaining frequencies according to the induced current of the center frequency, avoiding the repeated filling of the impedance matrix at each frequency, and effectively improving the acquisition efficiency. However, the invention requires iterative calculation of a large number of matrix-vector products, which affects the further improvement of the acquisition efficiency. In addition, when the invention uses the progressive waveform to estimate AWE, it is also necessary to fill in a large number of high-order impedance matrices, which increases the requirement for computer memory. 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 broadband RCS acquisition method for electrically large-size targets based on ACA and improved AWE, which is used to solve the technical problems of low acquisition efficiency and high computer memory requirements in the prior art.
[0005] To achieve the above object, the technical solution adopted by the present invention includes the following steps:
[0006] (1) Initialization parameters:
[0007] Initialize the electric field of the broadband uniform plane wave incident on the electrically large target surface as E in The frequency band f includes the frequency f1 as the starting frequency and f B B frequencies are the cutoff frequencies, each frequency f b The corresponding wavelength and wave number are λ b , k b , the center frequency is f0; the induced current vector I(k b ) is Q; the RWG basis function of each of the W triangular facet pairs formed by dividing the electrically large target surface is f w (r), where B ≥ 4, and when B is an even number, f0 is frequency, f0 is the first frequency when B is an odd number frequency, r is the field point position vector;
[0008] (2) Sampling the incident wave on the electrically large target surface:
[0009] Based on the GLC distribution, the incident wave on the surface of the electrically large target is sampled N times with k0 as the center, and N GLC sampling points are obtained, where N>(Q+1) and the nth GLC sampling point is σ n , σ n ∈(k0-Δk,k0+Δk), Δk is the wave number variation radius centered on k0;
[0010] (3) Construct the impedance matrix at each GLC sampling point:
[0011] Based on the octree algorithm, the RWG basis functions of W triangle face pairs are divided into S groups Ω={Ω1,Ω2,…,Ω s ,…,Ω S}, and based on the moment method MoM n Two Ω at a spatial distance less than λ1 μ and Ω ν Fill in the near-field impedance matrix formed between and obtain σ nAt the same time, the adaptive cross approximation method ACA is used to calculate σ n Two Ω at a spatial distance greater than λ1 χ and Ω ξ The far-field impedance matrix formed between the two is compressed and decomposed to obtain σ n Y far-field blocks at , and then through σ n Each far-field block at and near field blocks Calculate σ n The impedance matrix Z(σ n ), where χ≠ξ, 1≤μ,ν, χ,ξ≤S and is an integer, when μ≠ν is the mutual impedance near field block, and when μ = ν, it is the self-impedance near field block;
[0012] (4) Obtaining broadband RCS of electrically large targets:
[0013] The asymptotic waveform estimation method AWE is based on the pseudo-spectral derivative method PSDM, and through Z(σ n )Calculate the induced current vector I(k b )'s w-th element I w (k b ), then through I w (k b ), f w (r) and E in Calculate the radar cross section RCS (k b ), we can get the broadband RCS of electrically large target = {RCS(k1),RCS(k2),...,RCS(k b ),...,RCS(k B )}.
[0014] Compared with the prior art, the present invention has the following advantages:
[0015] (1) When the present invention adopts an improved progressive waveform estimation method to calculate the induced current vector of the electrically large-sized target surface at each wave number, the high-order impedance matrix of each order is represented by a pseudo-spectral differential matrix and an impedance matrix, thereby avoiding the defect of the prior art of directly filling in the high-order impedance matrix with high complexity, and effectively reducing the demand for computer memory space.
[0016] (2) Based on the adaptive cross approximation method, the present invention compresses and decomposes the far-field impedance matrix into the product of two low-rank matrices, and uses the low-rank matrix to accelerate the calculation of a large number of matrix-vector products in the asymptotic waveform estimation method, thereby avoiding the defect of the prior art that the calculation process is complicated due to the direct calculation of a large number of matrix-vector products in the asymptotic waveform estimation method, and effectively improving the efficiency of obtaining the radar scattering cross section. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 It is a flow chart for implementing the present invention. DETAILED DESCRIPTION
[0018] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0019] Reference Figure 1 , the present invention comprises the following steps:
[0020] Step 1) Initialize parameters:
[0021] Initialize the electric field of the broadband uniform plane wave incident on the electrically large target surface as E in The frequency band f includes the frequency f1 as the starting frequency and f B B frequencies are the cutoff frequencies, each frequency f b The corresponding wavelength and wave number are λ b , k b , the center frequency is f0; the induced current vector I(k b ) is Q; the RWG basis function of each of the W triangular facet pairs formed by dividing the electrically large target surface is f w (r), where B ≥ 4, and when B is an even number, f0 is frequency, f0 is the first frequency when B is an odd number frequency, r is the field point position vector; in this embodiment, f1=1GHz, f b =1.4GHz, B=41, f0 is the 21st frequency, f0=1.2GHz, W=18175.
[0022] The method for dividing the electrically large target surface is as follows: dividing the electrically large target surface into H triangular patches, and forming a triangular patch pair with each triangular patch and a triangular patch with a common edge therewith, to obtain W triangular patch pairs, where H≥4 and is an even number; in this embodiment, H=12650.
[0023] The side length of each triangular patch obtained by subdivision shall not exceed And every two triangular faces either do not intersect or have only one common edge.
[0024] RWG basis function f w The expression of (r) is:
[0025]
[0026] Among them, l w Represents two triangular patches and The length of the common edge of the w-th triangle patch pair, and Respectively and The area of express The vector from the vertex to the field point, express The vector from the field point to the vertex.
[0027] Step 2) Sampling the incident wave on the electrically large target surface:
[0028] Based on the GLC distribution, the incident wave on the surface of the electrically large target is sampled N times with k0 as the center, and N GLC sampling points are obtained, where N>(Q+1) and the nth GLC sampling point is σ n , σ n ∈(k0-Δk,k0+Δk), Δk is the wave number variation radius centered on k0.
[0029] GLC sampling point σ n The calculation formula is:
[0030]
[0031] Here, π represents the ratio of a circle to a circle.
[0032] Step 3) Construct the impedance matrix at each GLC sampling point:
[0033] Before constructing the impedance matrix, the RWG basis functions need to be grouped.
[0034] Based on the octree algorithm, the RWG basis functions of W triangle face pairs are divided into S groups Ω={Ω1,Ω2,…,Ω s ,...,Ω S}, the implementation method is:
[0035] Based on the octree algorithm, the (x min ,y min ,z min ) as the starting point, the cube with side length θ is evenly divided into eight equal parts, and each cube after division is evenly divided into eight equal parts again, and so on, until the side length of each cube after division is less than Then, S cubes containing RWG basis functions are selected from all cubes obtained by the last segmentation, and the RWG basis functions in each cube are grouped into a group to obtain the RWG basis function group Ω = {Ω1, Ω2, …, Ω s ,…,ΩS}. The calculation formula of θ is:
[0036] θ=max{|x max -x min |,|y max -y min |,|z max -z min |}
[0037] where x max , x min ,y max ,y min 、z max , z min They represent the maximum and minimum values of the x-, y-, and z-axis components of the coordinates of the H triangle patches’ vertices, respectively. max{·} represents the maximum value operation, and |·| represents the modulus operation.
[0038] After the RWG basis functions are grouped, the impedance matrix formed between the two groups of RWG basis functions needs to be divided into a near-field impedance matrix and a far-field impedance matrix according to the spatial distance between the RWG basis functions. Then, at each GLC sampling point σ n Based on the method of moments MoM, the near-field impedance matrix is filled to obtain σ n Based on the adaptive cross approximation method ACA, the far-field impedance matrix is compressed and decomposed to obtain σ n The far-field block at σ n All near-field blocks and far-field blocks at the location are calculated to obtain σ n The impedance matrix at .
[0039] Based on the MoM method n Two Ω at a spatial distance less than λ1 μ and Ω ν Fill in the near-field impedance matrix formed between and obtain σ n There are X near-field blocks at , and the xth near-field block is Among them, 1≤μ,ν≤S and are integers, when μ≠ν It is a mutual impedance near-field block, and when μ=ν it is a self-impedance near-field block.
[0040] Based on the adaptive cross approximation method ACA n Two Ω at a spatial distance greater than λ1 χ and Ω ξ The far-field impedance matrix formed between the two is compressed and decomposed to obtain σ n There are Y far-field blocks at , and the yth far-field block is The specific steps are:
[0041] First, set up two low-rank matrices calculate The corresponding approximate far-field block The calculation formula is:
[0042]
[0043] Then, calculate and The error matrix R y (σ n ), and ensure that the following conditions are met:
[0044]
[0045] Finally, use replace
[0046]
[0047] Among them, 1≤χ,ξ≤S and is an integer, χ≠ξ; express The corresponding dimension is u y ×v y The approximate far-field block is and Respectively The corresponding two low-rank matrices, ι y express The effective rank, R y (σ n )express and The error matrix of , ||·|| represents the matrix Frobenus norm, and ε represents the given error iteration threshold.
[0048] By σ n Each near-field block at and far-field blocks Calculate σ n The impedance matrix Z(σ n ):
[0049]
[0050] Step 4) Obtain the broadband RCS of the electrically large target:
[0051] The key to obtaining broadband RCS of electrically large-sized targets is to obtain the induced current vector I(k b ).
[0052] The asymptotic waveform estimation method AWE is based on the pseudo-spectral derivative method PSDM, and through Z(σ n)Calculate the induced current vector I(k b )'s w-th element I w (k b ), the specific steps are:
[0053] 4a) Based on MoM, and through f w (r) and E in Calculate the excitation vector Γ(k0) at k0 and its κ-order derivative Γ (κ) The w-th element Γ of (k0) w (k0),Γ w (κ) (k0):
[0054] Γ w (k0)=∫f w (r)·E in dr
[0055] Γ w (κ) (k0) = ∫E in(κ) ·f w (r)dr
[0056] Among them, E in(κ) For E in The κ-order derivative of , κ ≥ 1.
[0057] 4b) Calculate the element D in the αth row and βth column of the N×N dimensional pseudospectral differential matrix D based on the pseudospectral derivative method PSDM αβ :
[0058]
[0059] Among them, σ α , σ β is the αth and βth GLC sampling points σ α , σ β The corresponding sampling points after coordinate transformation are: is a coefficient related to the values of α and β, 1≤α,β≤N.
[0060] 4c) Calculate the 0th and qth order Taylor expansion coefficient vectors m0 and m1 through the results of 4b) and 4c) q :
[0061] m0=Z -1 (k0)·Γ(k0)
[0062]
[0063]
[0064] Where Z(k0) is the impedance matrix at the center wave number k0, Z -1 (k0) is the inverse matrix of Z(k0), Z (i) (k0) is the i-th order high-order impedance matrix at the center wave number k0, D i is the i-th power of matrix D, is the matrix D i No. Elements of the row, i ≥ 1, 1 ≤ q ≤ Q.
[0065] In step 3), σ n The impedance matrix Z(σ n ) is represented by two low-rank matrices based on ACA: The product of , using low-rank matrices can speed up the solution of m in 4c). q The matrix-vector product Z(σ n )m q-i The calculation speed can improve the efficiency of radar cross section RCS acquisition.
[0066] In 4c) solve for m q In the process, the pseudo-spectral differential matrix D and σ are constructed by AWE based on the pseudo-spectral derivative method PSDM. n The impedance matrix Z(σ n ) represents the complex i-order high-order impedance matrix Z (i) (k0), avoiding direct filling of Z (i) (k0), thus reducing the need for computer memory space.
[0067] Obtain m0 and m through the above steps q After that, I can be obtained through Taylor expansion formula w (k b ), but since the Taylor expansion is limited by the bandwidth, the calculation result may have a large deviation outside the bandwidth. In order to obtain a larger bandwidth, the Padé approximation is introduced in 4d) to transform the Taylor series into a rational function.
[0068] 4d) Through m0, m q The w-th element m w,0 、m w,q Calculate each k b The induced current vector I(k b )'s w-th element I w (k b ):
[0069]
[0070]
[0071] L+O=Q
[0072] Among them, L, a w,l Polynomial The number and corresponding to I w (k b )'s first-order coefficient, O, d w,o Polynomial The number and corresponding to I w (k b ), L and O are positive integers. If Q is an even number, L=O. If Q is an odd number, the absolute value of the difference between L and O |LO|=1.
[0073] Then through I w (k b ), f w (r) and E in Calculate each wave number k b Radar cross section RCS(k b ), we get the wideband RCS of electrically large target = {RCS(k1),RCS(k2),...,RCS(k b ),...RCS(k B )},RCS(k b ) is calculated as:
[0074]
[0075] Where R represents the distance between the electrically large target surface and the radar, represents R approaches infinity, j represents an imaginary unit, η represents the wave impedance in air, represents the unit vector of the observation point position vector r', Λ represents the outer surface of the electrically large target, ∫∫ Λ represents a surface integral operation on Λ, e represents a natural constant, and |·| represents a modulo operation.
Claims
1. A method for acquiring wideband RCS of electrically large-size targets based on ACA and improved AWE, characterized in that: By introducing the pseudo-spectral derivative method PSDM into the progressive waveform estimation method AWE, the improvement of AWE is achieved, which specifically includes the following steps: (1) Initialization parameters: Initialize the electric field of the broadband uniform plane wave incident on the electrically large target surface as E in The frequency band f includes the frequency f1 as the starting frequency and f B B frequencies are the cutoff frequencies, each frequency f b The corresponding wavelength and wave number are λ b , k b , the center frequency is f0; the induced current vector I(k b ) has a truncation order of Q; The RWG basis function of each of the W triangular facet pairs formed by dividing the electrically large target surface is f w (r), where B ≥ 4, and when B is an even number, f0 is frequency, f0 is the first frequency when B is an odd number frequency, r is the field point position vector; (2) Sampling the incident wave on the electrically large target surface: Based on the GLC distribution, the incident wave on the surface of the electrically large target is sampled N times with k0 as the center, and N GLC sampling points are obtained, where N>(Q+1) and the nth GLC sampling point is σ n , σ n ∈(k0-Δk,k0+Δk), Δk is the wave number variation radius centered on k0; (3) Construct the impedance matrix at each GLC sampling point: Based on the octree algorithm, the RWG basis functions of W triangle face pairs are divided into S groups Ω={Ω1,Ω2,…,Ω s ,…,Ω S }, and based on the moment method MoM n Two Ω at a spatial distance less than λ1 μ and Ω ν Fill in the near-field impedance matrix formed between n At the same time, the adaptive cross approximation method ACA is used to calculate σ n Two Ω at a spatial distance greater than λ1 χ and Ω ξ The far-field impedance matrix formed between the two is compressed and decomposed to obtain σ n Y far-field blocks at , and then through σ n Each far-field block at and near field blocks Calculate σ n The impedance matrix Z(σ n ), where χ≠ξ, 1≤μ,ν, χ,ξ≤S and is an integer, when μ≠ν is the mutual impedance near field block, and when μ = ν, it is the self-impedance near field block; (4) Obtaining broadband RCS of electrically large targets: The asymptotic waveform estimation method AWE is based on the pseudo-spectral derivative method PSDM, and through Z(σ n )Calculate the induced current vector I(k b )'s w-th element I w (k b ), then through I w (k b ), f w (r) and E in Calculate each wave number k b Radar cross section RCS(k b ), we can get the broadband RCS of electrically large target = {RCS(k1),RCS(k2),...,RCS(k b ),...RCS(k B )}.
2. The method according to claim 1, characterized in that: The electrically large target surface is segmented as described in step (1), and the implementation method is as follows: the electrically large target surface is segmented into H triangular facets, and each triangular facet and the triangular facets with which it has a common edge form a triangular facet pair to obtain W triangular facet pairs, where H ≥ 4 and is an even number.
3. The method according to claim 1, characterized in that The RWG basis function f described in step (1) w (r), its expression is: Among them l w Represents two triangular patches and The length of the common edge of the w-th triangle patch pair, and Respectively and The area of express The vector from the vertex to the field point, express The vector from the field point to the vertex.
4. The method according to claim 1, characterized in that: The σ mentioned in step (2) n , the calculation formula is: Here, π represents the ratio of a circle to a circle.
5. The method according to claim 1, characterized in that The octree algorithm described in step (3) divides the RWG basis functions of the W triangle face pairs into S groups, and the implementation steps are as follows: Based on the octree algorithm, the (x min ,y min ,z min ) as the starting point, the cube with side length θ is evenly divided into eight equal parts, and each cube after division is evenly divided into eight equal parts again, and so on, until the side length of each cube after division is less than Then, S cubes containing RWG basis functions are selected from all cubes obtained by the last segmentation, and the RWG basis functions in each cube are grouped into a group to obtain the RWG basis function group Ω = {Ω1, Ω2, …, Ω s ,…,Ω S }, where the calculation formula of θ is: θ=max{|x max -x min |,|y max -y min |,|z max -z min |} where x max , x min ,y max ,y min 、z max , z min They represent the maximum and minimum values of the x-, y-, and z-axis components of the coordinates of the H triangle patches’ vertices, respectively. max{·} represents the maximum value operation, and |·| represents the modulus operation.
6. The method according to claim 1, characterized in that The σ mentioned in step (3) n Each far-field block at The calculation formula is: in, express The corresponding dimension is u y ×v y The approximate far-field block is and Respectively The corresponding two low-rank matrices, ι y express The effective rank, R y (σ n )express and The error matrix of , ||·|| represents the matrix Frobenus norm, and ε represents the given error iteration threshold.
7. The method according to claim 1, characterized in that The σ mentioned in step (3) n The impedance matrix Z(σ n ), the calculation formula is:
8. The method according to claim 1, characterized in that The calculation of the induced current vector I(k b )'s w-th element I w (k b ), the implementation steps are: (4a) Based on MoM, and through f w (r) and E in Calculate the excitation vector Γ(k0) at k0 and its κ-order derivative Γ (κ) The w-th element Γ of (k0) w (k0),Γ w (κ) (k0): C w (k0)=∫f w (r)·E in doctor C w (κ) (k0)=∫E in(κ) ·f w (r)dr Among them, E in(κ) For E in The κ-order derivative of , κ ≥ 1; (4b) Based on the pseudo-spectral derivative method PSDM, the element D in the αth row and βth column of the N×N dimensional pseudo-spectral differential matrix D is calculated. αβ : Among them, σ α , σ β is the αth and βth GLC sampling points σ α , σ β The corresponding sampling points after coordinate transformation are: is the coefficient related to the values of α and β, 1≤α,β≤N; (4c) Calculate the 0th and qth order Taylor expansion coefficient vectors m0 and m1 through the results of steps (4a) and (4b). q : m0=Z -1 (k0)·Γ(k0) Where Z(k0) is the impedance matrix at the center wave number k0, Z -1 (k0) is the inverse matrix of Z(k0), Z (i) (k0) is the i-th order high-order impedance matrix at the center wave number k0, D i is the i-th power of matrix D, is the matrix D i No. The elements of the row, i ≥ 1, 1 ≤ q ≤ Q; (4d) Through m0, m q The w-th element m w,0 、m w,q Calculate each k b The induced current vector I(k b )'s w-th element I w (k b ): L+O=Q Among them, L, a w,l Polynomial The number and corresponding to I w (k b )'s first-order coefficient, O, d w,o Polynomial The number and corresponding to I w (k b ), L and O are positive integers. If Q is an even number, L=O. If Q is an odd number, the absolute value of the difference between L and O |LO|=1.
9. The method according to claim 1, characterized in that: Each wave number k described in step (4) b Radar cross section RCS(k b ), the calculation formula is: Where R represents the distance between the electrically large target surface and the radar, represents that R approaches infinity, j represents an imaginary unit, and η represents the wave impedance in air. represents the unit vector of the observation point position vector r', Λ represents the outer surface of the electrically large target, ∫∫ Λ represents a surface integral operation on Λ, e represents a natural constant, and |·| represents a modulo operation.
Citation Information
Patent Citations
CBFM and AWE-based periodic structure broadband RCS acquisition method
CN116484642A
Method for estimating electromagnetic scattering characteristics of inhomogeneous medium target based on PMCHWT integral equation
CN115859613A
Novel electromagnetic property calculation method for electrically large metal target with uncertain appearance based on Maehly approximation
CN117436273A
Method and apparatus for analyzing numerical value of electromagnetic wave using moment method
KR101501371B1
Method and system for adaptive modeling and simulation of lossy transmission lines
US8386216B1
