A transmission structure design method based on the inverse problem
By designing transmission structures using electromagnetic backscattering methods and multi-frequency point algorithms, the problems of low computational efficiency and dataset dependence in existing technologies are solved, realizing an efficient and flexible inverse design process that is applicable to various application scenarios and frequency bands.
Patent Information
- Application Number
- CN202410294967.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-14
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2044-03-14
AI Technical Summary
Existing techniques are computationally inefficient in inverse scattering problems, rely on large datasets to train deep learning models and have poor generalization capabilities, making it impossible to achieve real-time computation and flexible design.
The electromagnetic backscattering method and multi-frequency backscattering algorithm are used to design the transmission structure through initialization, forward process, inverse process and evaluation process. The dielectric constant distribution is updated by electromagnetic backscattering integral equation and the design target is optimized by multi-frequency iterative algorithm.
It achieves an efficient transmission structure design that does not rely on prior knowledge, reduces computational costs, and improves design efficiency and flexibility, making it suitable for different application scenarios and frequency bands.
Smart Images

Figure CN117951421B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of antenna communication technology, and specifically relates to a transmission structure design method based on inverse problems. Background Technology
[0002] Numerical calculations of the scattering and inverse scattering characteristics of a medium have wide applications in many fields, such as Earth remote sensing, biomedical imaging, and nondestructive testing. These problems generally involve reconstructing the properties of the scatterer, such as its dielectric constant distribution function, from knowledge of its scattering data, which is typically expressed in the form of transmission parameters of the scatterer, such as reflection or transmission coefficients. Therefore, the goal of solving the inverse problem is generally to design a scatterer that achieves a specified frequency response. Many studies have shown that inverse scattering methods have been applied to the design of rippled waveguides, such as rippled filters and the design of selective excitation pulses for nuclear magnetic resonance (NMR) imaging.
[0003] The forward design method has obvious drawbacks, specifically the following two points: First, the forward design method relies on existing knowledge bases and requires a high level of engineer experience and prior knowledge; second, the forward design method has high design time costs, especially the time cost of full-wave simulation in the process of adjusting and optimizing design parameters.
[0004] Numerous researchers have proposed algorithms to solve inverse scattering problems. The Born iterative algorithm utilizes the Born approximation to approximate the electric field and iteratively calculates the dielectric constant distribution of the medium. Optimization algorithms such as the contrast source method and subspace optimization have also been applied to inverse scattering calculations. Furthermore, in recent years, the field of nanophotonics has introduced the concept of inverse design, distinct from forward design methods, by incorporating deep learning. The existing technology "Nanophotonic particle simulation and inverse design using artificial neural networks" utilizes backpropagation to apply trained neural networks to inverse design problems in nanophotonics. The existing technology "Generative model for the inverse design of metasurfaces" applies an unsupervised learning system to the inverse design of nanophotonic structures. This system, without a clear guiding objective when processing data, can obtain new results from entirely new data.
[0005] Traditional algorithms for solving electromagnetic scattering and inverse scattering suffer from low computational efficiency due to the involvement of large matrix inverse operations, making real-time computation impossible and hindering the use of complete frequency band electric field information. Deep learning-based inverse design methods also have significant drawbacks, specifically: first, the results are highly sensitive to dataset quality, requiring the generation of high-quality and large datasets to train a robust neural network, leading to high computational costs; second, these methods require specific datasets and neural network training for particular inverse design problems, limiting their generalization capabilities. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a transmission structure design method based on the inverse problem. For the first time, electromagnetic backscattering method and multi-frequency backscattering method are introduced to realize the transmission structure design based on the inverse problem.
[0007] The technical problem addressed by this invention is solved as follows:
[0008] A transmission structure design method based on the inverse problem includes the following steps:
[0009] S1, Initialization
[0010] The position S′ of the inverse design scatterer is manually set, and the scatterer position S′ is located in the internal region of the transmission structure waveguide; the performance indicators of the transmission structure waveguide are set as the reflection parameter S. 11 (f) and transmission parameter S 21 (f), where f is the frequency; defines the desired incident electric field at the incident port of the waveguide transmission structure.
[0011] S2, Forward Process
[0012] The scatterer is equivalent to a secondary source, and is divided into several grids, with each grid serving as a scatterer source point r′; based on the current background dielectric constant distribution function... Update the total field within the scatterer position S′ using the electromagnetic inverse scattering integral equation. x′ and y′ represent the x and y coordinates of the scattering source point r′, respectively; specify any point at the emission port as the port sampling point, and update the Green's function from the scattering source point r′ to the port sampling point.
[0013] S3, Reverse Process
[0014] Desired electric field The total field within the scatterer position S′ Green's function from the scattering source point r′ to the port sampling point Substituting into the electromagnetic inverse scattering integral equation and solving it using a multi-frequency method, we obtain the dielectric constant distribution function εr(x′, y′) within the scatterer position S′, and update the background dielectric constant distribution function.
[0015] S4, Assessment
[0016] Based on the transmission structure waveguide, the emitter position S′, and the corresponding dielectric constant distribution function εr(x′, y′), the reflection parameters S corresponding to the current structure are obtained. 11 ′(f) and transmission parameter S 21 ′(f), determine whether the current performance index has converged; if so, determine the current dielectric constant distribution function ε. r (x′, y′) represents the desired solution, and the current dielectric constant distribution function ε r The structure corresponding to (x′, y′) is taken as the structure at the scatterer position S′; otherwise, let the dielectric constant distribution function ε r (x′, y′) is the background dielectric constant distribution function. Execute S2-S4 until convergence.
[0017] Furthermore, the formula for updating the Green's function in step S2 is as follows:
[0018]
[0019] in, For the Laplace operator, k0 is the free space wavenumber, δ(r) far -r′) represents the distance from the scatterer source point r′ to the port sampling point r. far The pulse excitation function.
[0020] Furthermore, in steps S2 and S3, the integral equation for electromagnetic inverse scattering is:
[0021]
[0022]
[0023] Where r is any point inside the waveguide; As the variable to be solved, E represents the distribution function of the difference dielectric constant; s (r) represents the scattered field;
[0024] Compared with the current background dielectric constant distribution function The relationship is represented as:
[0025]
[0026] Furthermore, in step S3, the process of solving the electromagnetic inverse scattering integral equation using a multi-frequency point method is as follows:
[0027] The discretized integral equation for electromagnetic inverse scattering is expressed as:
[0028]
[0029] Among them, [E s ] represents the discretized scattering field, [ε r [] denotes the discretized dielectric constant distribution function, and [G] denotes the discretized Green's function. Represents the discretized total field;
[0030] The multi-frequency electromagnetic inverse scattering integral equation can be expressed in matrix operation form as follows:
[0031]
[0032] Where n is the number of frequency points divided into the waveguide operating frequency band, f1, f2, ..., f n This represents the frequency values corresponding to different frequency points. G1, G2, ..., G represent the scattered field values corresponding to each frequency point. n This represents the magnitude of the Green's function at each frequency point during the current iteration. This represents the total field value at the output port corresponding to each frequency point;
[0033] Solving the above matrix operation formula yields the discretized dielectric constant distribution function within the scatterer position S′.
[0034] Furthermore, in step S4, the specific process for determining whether the current performance metric has converged is as follows:
[0035] The convergence value between the current performance metric and the set expected performance metric is:
[0036]
[0037] Where || represents the modulus value; if the iterative convergence value ITV ≤ the set threshold, then convergence is determined.
[0038] The beneficial effects of this invention are:
[0039] This invention introduces electromagnetic backscattering and multi-frequency backscattering algorithms for the first time in transmission structure design based on inverse problems. This allows for the rational design of different transmission structures according to various application scenarios and operating frequency bands. Compared to existing forward design methods, the inverse design method of this invention uses the design objective as a known condition in the inverse design of the transmission structure, without relying on existing prior knowledge, resulting in higher efficiency and overcoming the shortcomings of forward design methods. Compared to traditional backscattering algorithms, the multi-frequency algorithm utilizes complete prior information and can iteratively obtain a dielectric constant distribution function that better matches the design objective.
[0040] Compared to existing deep learning-based inverse design methods, the method described in this invention does not require the time overhead of acquiring large training datasets and training neural networks, thus offering higher efficiency. Furthermore, the method can update the desired scattering field according to different design objectives to achieve inverse design processes for different goals, without needing to generate specific datasets or train corresponding neural networks for each specific inverse design problem, making the method more flexible. Attached Figure Description
[0041] Figure 1 This is a flowchart illustrating the reverse design method described in this invention;
[0042] Figure 2 The figure shows the expected performance index curves of the two-dimensional parallel plate waveguide with baffle in the embodiment.
[0043] Figure 3 This is a structural diagram of the scatterer position S′ corresponding to the backscattering of the two-dimensional parallel plate waveguide with baffle in the embodiment;
[0044] Figure 4 The transmission parameter curves of the two-dimensional parallel plate waveguide with baffle in the embodiment are obtained based on the transmission structure, the scatterer position S′ and the corresponding dielectric constant distribution function.
[0045] Figure 5 This is a schematic diagram illustrating the expected performance parameters of a three-dimensional rectangular waveguide in an embodiment.
[0046] Figure 6 This is a schematic diagram showing the dielectric constant distribution at one-eighth of the waveguide length along the waveguide transmission direction in a waveguide structure obtained by inverse design of the desired performance index of a three-dimensional rectangular waveguide, as an example.
[0047] Figure 7 This is a schematic diagram showing the dielectric constant distribution at a position one-quarter of the waveguide length along the waveguide transmission direction in the waveguide structure obtained by inverse design of the desired performance index of a three-dimensional rectangular waveguide, as an example.
[0048] Figure 8This is a schematic diagram showing the dielectric constant distribution at three-eighths of the waveguide length along the waveguide transmission direction in a waveguide structure obtained by inverse design of the desired performance index of a three-dimensional rectangular waveguide, as an example.
[0049] Figure 9 This is a schematic diagram showing the dielectric constant distribution at the center of the waveguide in the waveguide transmission direction, obtained by inverse design of the waveguide structure based on the desired performance index of a three-dimensional rectangular waveguide in an example embodiment.
[0050] Figure 10 The diagram shows the performance index of the waveguide structure obtained by inverse design based on the desired performance index of a three-dimensional rectangular waveguide, as an example. Detailed Implementation
[0051] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.
[0052] This embodiment provides a transmission structure design method based on the inverse problem, the flowchart of which is shown below. Figure 1 As shown, the process is as follows: Initialization: Set the position of the scatterer and determine the corresponding desired electric field according to the set performance indicators; Forward process: Update the total field within the scatterer and the Green's function from the source point to the sampling point based on the background dielectric constant distribution function; Inverse process: Substitute the desired scattered field, the total field within the scatterer, and the Green's function from the source point to the sampling point into the integral equation of the electromagnetic inverse scattering method, and thereby introduce a multi-frequency iterative algorithm to obtain the dielectric constant distribution function within the scatterer; Evaluation: If the current performance indicators match the set performance indicators, it is determined to be the desired solution; otherwise, the process is repeated until the set performance indicators are met.
[0053] The method described in this embodiment specifically includes the following steps:
[0054] S1, Initialization
[0055] The position S′ of the inverse design scatterer is manually set, and the scatterer position S′ is located in the internal region of the transmission structure waveguide; the performance indicators of the transmission structure waveguide are set as the reflection parameter S. 11 (f) and transmission parameter S 21 (f), where f is the frequency; defines the desired incident electric field at the incident port of the waveguide transmission structure.
[0056] In this embodiment, the process of setting the performance indicators and desired electric field in step S1 is as follows:
[0057] To verify the effectiveness of this invention, a parallel plate waveguide with a baffle was first designed according to engineering requirements. The baffle was located at the center of the parallel plate waveguide. The frequency band was selected from DC to 1GHz, the distance between the upper and lower plates was d, the baffle had no thickness, and its length was d / 3. Electromagnetic simulation was performed on the parallel plate waveguide with the baffle to obtain the reflection parameter S. 11 (f) Transmission parameter S 21 (f) Incident electric field at the incident port Total Field at Incident Port Scattering field and the total field of the exit port The performance indicators and desired electric field are set in this embodiment. Figure 2 The graph shows the expected performance index of a two-dimensional parallel plate waveguide with baffles.
[0058] S2, Forward Process
[0059] In this context, the scatterer can be equivalently represented as a secondary source. The scatterer is divided into several grids, with each grid serving as a scatterer source point r′; based on the current background dielectric constant distribution function... (Initially set to 1), update the total field within the scatterer position S′ using the electromagnetic inverse scattering integral equation. x′ and y′ represent the x and y coordinates of the scattering source point r′, respectively; specify any point at the emission port as the port sampling point, and update the Green's function from the scattering source point r′ to the port sampling point.
[0060] The formula for updating the Green's function in step S2 is:
[0061]
[0062] in, For the Laplace operator, k0 is the free space wavenumber, δ(r) far -r′) represents the distance from the scatterer source point r′ to the port sampling point r. far The pulse excitation function.
[0063] S3, Reverse Process
[0064] Desired electric field The total field within the scatterer position S′ Green's function from the scattering source point r′ to the port sampling point Substituting into the electromagnetic inverse scattering integral equation, we obtain the dielectric constant distribution function ε within the scatterer position S′. r (x′,y′).
[0065] In steps S2 and S3, the integral equation for electromagnetic inverse scattering is:
[0066]
[0067]
[0068] Where r is any point inside the waveguide, As the variable to be solved, E represents the distribution function of the difference dielectric constant; s (r) represents the scattered field;
[0069] Compared with the current background dielectric constant distribution function The relationship is represented as:
[0070]
[0071] After the above calculations, the dielectric constant distribution function based on prior information at a single frequency point can be obtained. However, since the calculation only involves prior information at one frequency point, the scattering parameters corresponding to the calculated dielectric constant distribution function have a large error compared to the scattering parameters across the entire frequency band. To better obtain the dielectric constant distribution function, a calculation method involving multiple frequency points is introduced.
[0072] The process of solving the electromagnetic inverse scattering integral equation using a multi-frequency method is as follows:
[0073] The discretized integral equation for electromagnetic inverse scattering is expressed as:
[0074]
[0075] Among them, [E s ] represents the discretized scattering field, [ε r [] denotes the discretized dielectric constant distribution function, and [G] denotes the discretized Green's function. Represents the discretized total field;
[0076] In the above formula, the electric field and Green's function at different frequencies are connected by the same dielectric constant matrix. Therefore, the electric field data of all ports at the multiple frequency points are stored in a matrix. This matrix, which contains all prior information, also corresponds to the same dielectric constant distribution matrix.
[0077] The multi-frequency electromagnetic inverse scattering integral equation can be expressed in matrix operation form as follows:
[0078]
[0079] Where n is the number of frequency points divided into the waveguide operating frequency band, f1, f2, ..., f n This represents the frequency values corresponding to different frequency points. G1, G2, ..., G represent the scattered field values corresponding to each frequency point.n This represents the magnitude of the Green's function at each frequency point during the current iteration. This represents the total field value at the output port corresponding to each frequency point.
[0080] S4, Assessment
[0081] Based on the transmission structure waveguide, the emitter position S′, and the corresponding dielectric constant distribution function εr(x′, y′), the reflection parameters S corresponding to the current structure are obtained. 11 ′(f) and transmission parameter S 21 ′(f), determine whether the current performance index has converged; if so, determine the current dielectric constant distribution function ε. r (x′, y′) represents the desired solution, and the current dielectric constant distribution function ε r The structure corresponding to (x′, y′) is taken as the structure at the scatterer position S′; otherwise, let the dielectric constant distribution function ε r (x′, y′) is the background dielectric constant distribution function. Execute S2-S4 until convergence.
[0082] In step S4, the specific process for determining whether the current performance metric has converged is as follows:
[0083] The convergence value between the current performance metric and the set expected performance metric is:
[0084]
[0085] Where || represents the modulus value; if the iterative convergence value ITV ≤ the set threshold, then convergence is determined.
[0086] Figure 3 This is a structural diagram of the scatterer position S′ corresponding to the backscattering of the two-dimensional parallel plate waveguide with baffle in this embodiment.
[0087] Figure 4 The transmission parameter curves for a two-dimensional parallel plate waveguide with a baffle are obtained based on the transmission structure, the scatterer position S′, and the corresponding dielectric constant distribution function. It can be seen that the curves fit the desired transmission parameters.
[0088] The effectiveness of the invention was further verified using a three-dimensional rectangular waveguide. The desired performance indicators of the three-dimensional rectangular waveguide were set, such as... Figure 5 As shown. Figure 6 , Figure 7 , Figure 8 and Figure 9 These are schematic diagrams showing the dielectric constant distribution at one-eighth, one-quarter, three-eighths of the waveguide length and at the center of the waveguide in the waveguide structure obtained by inverse design based on the desired performance index of the three-dimensional rectangular waveguide, respectively. Figure 10The performance index of the waveguide structure is obtained by inverse design from the desired performance index of the three-dimensional rectangular waveguide, and it can be seen that it fits the desired performance index.
[0089] The method described in this invention solves the scattering problem of two-dimensional and three-dimensional transmission structures by introducing a multi-frequency point algorithm, and realizes the goal of designing microwave devices using waveguide structures through inverse scattering.
Claims
1. A transmission structure design method based on the inverse problem, characterized in that, Includes the following steps: S1, Initialization Manually set the position of the inverse design scatterer scatterer position Located within the internal region of the transmission structure waveguide; the performance parameters of the transmission structure waveguide are defined as reflection parameters. and transmission parameters f is the frequency; the desired electric field incident at the transmission structure wave input port is defined. ; S2, Forward Process The scatterer is equivalent to a secondary source, and the scatterer is divided into several grids, with each grid serving as a scatterer source point. Based on the current background dielectric constant distribution function The total field within the scatterer position S' is updated using the electromagnetic inverse scattering integral equation. , and These represent the scattering source points, respectively. The x and y coordinates are specified; any point at the emission port is designated as the port sampling point, and the scattering source point is updated. Green's function to port sampling point ; S3, Reverse Process Desired electric field scatterer position The main field inside and scattering source point Green's function to port sampling point Substituting the electromagnetic inverse scattering integral equation and solving it using a multi-frequency method, the position of the scattering body can be obtained. Dielectric constant distribution function within Update the background dielectric constant distribution function ; S4, Assessment Based on the transmission structure waveguide and scatterer position and the corresponding dielectric constant distribution function Obtain the reflection parameters corresponding to the current structure. and transmission parameters Determine whether the current performance index has converged; if so, determine the current dielectric constant distribution function. For the desired solution, the current dielectric constant distribution function The corresponding structure serves as the location of the scatterer. The structure; otherwise, let the dielectric constant distribution function Background dielectric constant distribution function Execute S2-S4 until convergence.
2. The transmission structure design method based on the inverse problem according to claim 1, characterized in that, The formula for updating the Green's function in step S2 is: in, For the Laplace operator, , For free space wavenumber, Source point of scattering volume to port sampling point The pulse excitation function.
3. The transmission structure design method based on the inverse problem according to claim 1, characterized in that, In steps S2 and S3, the integral equation for electromagnetic inverse scattering is: in, For any point within the waveguide; As the variable to be solved, it represents the distribution function of the difference dielectric constant; Represents the scattered field; Compared with the current background dielectric constant distribution function The relationship is represented as: 。 4. The transmission structure design method based on the inverse problem according to claim 3, characterized in that, In step S3, the process of solving the electromagnetic inverse scattering integral equation using the multi-frequency point method is as follows: The discretized integral equation for electromagnetic inverse scattering is expressed as: in, Represents the discretized scattering field. This represents the discretized dielectric constant distribution function. Represents the discretized Green's function. Represents the discretized total field; The multi-frequency electromagnetic inverse scattering integral equation can be expressed in matrix operation form as follows: Where n is the number of frequency points divided into the waveguide's operating frequency band. This represents the frequency values corresponding to different frequency points. This represents the scattered field value corresponding to each frequency point. This represents the magnitude of the Green's function at each frequency point during the current iteration. This represents the total field value at the output port corresponding to each frequency point; Solving the above matrix operation formula can yield the position of the scatterer. Discretized dielectric constant distribution function within.
5. The transmission structure design method based on the inverse problem according to claim 1, characterized in that, In step S4, the specific process for determining whether the current performance metric has converged is as follows: The convergence value between the current performance metric and the set expected performance metric is: in, This represents the modulo value; if the iterative convergence value ITV is less than or equal to the set threshold, then convergence is determined.
Citation Information
Patent Citations
Systems and methods for virtual index-matching of diffusive media
CN102781307A
Rapid predictive analysis of very large data sets using the distributed computational graph
US20170124464A1
