A collaborative optimization method for thick and thin models of traveling wave tubes and zigzag waveguides based on weight matrix parameterization
By using weight matrix parameterization and adaptive optimization strategies, the parameters of the coarse model of the traveling wave tube tortuous waveguide are directly reconstructed, solving the problems of modeling complexity and multi-phase point matching in traditional methods, and realizing efficient and stable traveling wave tube slow wave structure design.
Patent Information
- Application Number
- CN202611287036.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-24
- Publication Date
- 2026-11-17
AI Technical Summary
Traditional spatial mapping methods for designing traveling wave tubes, tortuous waveguides, and slow wave structures suffer from problems such as large modeling workload, lack of physical interpretability of mapping functions, limitations of single-point optimization, and difficulties in integration with commercial software, making it difficult to achieve efficient and stable multi-phase point collaborative matching.
By employing a weight matrix parameterization method, the internal parameters of the coarse model are reconstructed by constructing a 5×5 weight matrix. Combined with a three-stage adaptive optimization strategy and the CST automation framework, the dispersion characteristics of the slow-wave structure are directly calibrated, achieving collaborative optimization without the need for an independent coarse model.
It significantly reduces modeling complexity, enables rapid and accurate design of slow-wave structures for traveling wave tubes in the V-band and above, with a coarse model frequency response error of less than 1.0 and a design efficiency improvement of approximately 15 times.
Smart Images

Figure CN123116584A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of microwave vacuum electronic device design optimization technology. More specifically, it relates to a collaborative optimization method for coarse and fine models of traveling wave tube tortuous waveguides based on weight matrix parameterization, which is applicable to the rapid and accurate design of slow wave structures of traveling wave tubes in the V-band and above. Background Technology
[0002] Traveling wave tubes (TWTs) are core microwave power amplification devices in fields such as satellite communications, radar, and electronic warfare. Their slow-wave structure is a key component that determines the device's operating frequency, bandwidth, and output power. The zigzag waveguide slow-wave structure, due to its all-metal structure, strong heat dissipation, large power capacity, and suitability for millimeter-wave frequencies, has become the preferred solution for TWTs in the V-band and above.
[0003] In the design of tortuous waveguide slow-wave structures, it is necessary to accurately obtain their dispersion characteristics (frequency-phase relationship), which typically relies on three-dimensional full-wave electromagnetic simulation software (such as CST Studio Suite). However, high-precision simulation (fine model) requires fine mesh generation, and a single simulation can take tens of minutes to several hours, which would result in huge computational costs if used directly for optimization design. Therefore, a co-optimization strategy of coarse and fine models is often adopted in engineering: low-precision fast simulation (coarse model) is used to undertake the main optimization iterations, while only a small number of high-precision simulations (fine model) are used for calibration.
[0004] The traditional space mapping (SM) technique was proposed by Bandler in 1994. Its core idea is to establish a fine model parameter space. To coarse model parameter space mapping function , making This method has the following problems:
[0005] A coarse model needs to be built independently: Traditional spatial mapping requires that the coarse model be an independent model that physically corresponds to the fine model but simplifies the calculation (such as equivalent circuit or sparse mesh model), which results in a large amount of modeling work.
[0006] The physical meaning of the mapping function is unclear: mapping function Typically, these are black-box mathematical mappings, whose parameters lack physical interpretability, making it difficult for engineers to intervene based on experience.
[0007] Limitations of single-point optimization: Traditional spatial mapping is mostly optimized for specific frequency points or narrowband responses, and it is difficult to directly handle multi-phase point cooperative matching of wide dispersion curves of slow-wave structures;
[0008] Integration with commercial software is difficult: existing spatial mapping methods are mostly theoretical frameworks, lacking engineering and automated implementation for commercial simulation software such as CST, making it difficult to achieve unattended optimization. Summary of the Invention
[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a collaborative optimization method for coarse and fine models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization. This method does not require the construction of an independent coarse model, but directly reconstructs the internal parameters of the existing simulation model through the weight matrix. Combined with a three-stage adaptive optimization strategy and the CST automation framework, it achieves efficient, stable, and unattended calibration of the dispersion characteristics of slow-wave structures.
[0010] To achieve the above-mentioned objectives, this invention provides a collaborative optimization method for coarse and thin models of traveling wave tubes and tortuous waveguides based on weighted matrix parameterization, characterized by the following steps:
[0011] (1) Constructing a coarse model with parameterized weight matrix
[0012] Establish a 5×5 initial weight matrix :
[0013]
[0014] The basic structural parameters of the meandering waveguide slow wave structure are: the reference length of the wide side of the waveguide. Waveguide narrow side reference length Waveguide bending diameter reference value Straight waveguide length reference value Electron injection channel radius reference value As input, the actual structural parameters of the coarse model are generated through linear combination:
[0015] , , , , ;
[0016] (2) Establish the optimization objective function
[0017] 2.1) Multi-phase point frequency response simulation
[0018] exist Target phase points The fine and coarse models were used to perform frequency simulations, obtaining two sets of frequency response sequences. and ,in, For the fine model at the phase point Frequency at that location Weight matrix Actual structural parameters of the coarse model generated by linear combination The frequency at the same phase point To optimize the number of iterations;
[0019] 2.2) Calculate the objective function value
[0020] The objective function value is the mean square error of the frequency response at multiple phase points of the coarse and fine models. :
[0021] ;
[0022] (3) Solving the optimal weight matrix using two-stage adaptive optimization
[0023] Initialize optimization iteration count =0;
[0024] Phase 1: Conservative Trust Region Sampling
[0025] 3.1) In the current weight matrix Trust region radius Inner, diagonal elements adopt a conservative step size Off-diagonal elements Use a larger step size Random sampling points are generated by grouping by row;
[0026] 3.2) Evaluate the objective function value of all sampling points. The optimal step is to update the weight matrix as follows: If the improvement is successful, the trust region radius will be expanded. Otherwise, shrink ;
[0027] 3.3) Determine the value of the objective function Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. Or trust region radius If the value is below 0.001, switch to Phase Two; otherwise... (Return to step 3.1).
[0028] Phase Two: Fine Coordinate Descent
[0029] 3.4) Selecting the weight matrix The diagonal elements are used as optimization variables, employing a fixed step size set:
[0030]
[0031] Perform a fine search and select the target function value Update the selected element with the minimum value, and traverse all weight matrices. The elements are used to complete one round of optimization, resulting in a weight matrix.
[0032] 3.5) Determine the value of the objective function. Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. If so, then The optimal weight matrix Otherwise, otherwise (Return to step 3.4).
[0033] (4) Coarse model verification and application
[0034] The optimal weight matrix It is applied to the coarse model to verify its frequency response consistency with the fine model across the entire phase range. Once the accuracy requirements are met, it is used for the design of tortuous waveguide slow wave structures. If the accuracy requirements are not met, the basic structural parameters are reset and optimized again until the accuracy requirements are met.
[0035] The objective of this invention is achieved as follows.
[0036] This invention proposes a collaborative optimization method for coarse and fine models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization. First, a coarse model parameterized by the weight matrix is constructed to generate the actual structural parameters of the coarse model. The system then uses both the fine and coarse models to establish the optimization objective function, and performs a two-stage adaptive optimization to solve for the optimal weight matrix. Then, a coarse model is verified and applied. Once the accuracy requirements are met, it is used for the design of tortuous waveguide slow wave structures.
[0037] This invention innovatively employs a weight matrix to linearly reconstruct the internal structural parameters of the coarse model. A 5×5 weight matrix maps the basic structural parameters to the actual geometric parameters of the coarse model. With the goal of minimizing the frequency response differences at multiple phase points, a two-stage adaptive optimization strategy (conservative trust region sampling → fine coordinate descent) is used to solve for the optimal weight matrix, achieving a high-precision approximation of the fine model from the coarse model. Compared to traditional spatial mapping methods, this invention eliminates the need to establish an independent coarse model, directly calibrating the existing coarse model through parameter reconstruction, significantly reducing modeling complexity. In the design of V-band tortuous waveguide slow-wave structures, the mean square error between the frequency response of the coarse model and the fine model can be reduced to below 1.0 after only 28 iterations. Attached Figure Description
[0038] Figure 1 This is a flowchart of the collaborative optimization method for coarse and thin models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization, as described in this invention.
[0039] Figure 2 This is a schematic diagram illustrating the principle of weight matrix parameterization in this invention;
[0040] Figure 3 This is a three-dimensional vacuum structure diagram of the tortuous waveguide slow wave structure of the present invention;
[0041] Figure 4 These are frequency response comparison curves of the coarse and fine models at six phase points in this embodiment of the invention;
[0042] Figure 5 This is the convergence curve of the objective function during the optimization process of this invention. Detailed Implementation
[0043] The specific embodiments of the present invention will now be described with reference to the accompanying drawings to enable those skilled in the art to better understand the invention. It should be particularly noted that in the following description, detailed descriptions of known functions and designs that might obscure the main content of the invention will be omitted here.
[0044] This invention, based on a weighted matrix parameterized method for collaborative optimization of coarse and fine models of traveling wave tubes and zigzag waveguides, innovatively employs a weighted matrix to linearly reconstruct the internal structural parameters of the coarse model. It achieves coarse model response calibration through a linear combination of shared physical parameter spaces, differing from the traditional spatial mapping paradigm of "two independent models + external mapping function." Specifically, in this embodiment, as... Figure 1 As shown, the collaborative optimization method for coarse and thin models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization of the present invention includes the following steps:
[0045] Step S1: Construct a coarse-grained model with parameterized weight matrix
[0046] This invention employs a 5×5 weight matrix to linearly reconstruct the five structural parameters of the coarse model, a core innovation that distinguishes it from traditional spatial mapping. Specifically:
[0047] Establish a 5×5 initial weight matrix :
[0048]
[0049] The basic structural parameters of the meandering waveguide slow wave structure are: the reference length of the wide side of the waveguide. Waveguide narrow side reference length Waveguide bending diameter reference value Straight waveguide length reference value Electron injection channel radius reference value As input, the actual structural parameters of the coarse model are generated through linear combination:
[0050] , , , , .
[0051] Let the basic structure parameter vector be The weight matrix is Then, the actual structural parameters of the coarse model generated by linear combination are:
[0052]
[0053] Right now:
[0054]
[0055] The physical significance of this parameterization method lies in allowing each structural parameter to borrow the physical influence of other parameters to compensate for electromagnetic response deviations caused by mesh sparsity. For example, in the coarse model... The equivalent value can be fine-tuned Contribution (through This is used to compensate for the frequency response without changing the grid density.
[0056] To ensure physical plausibility, boundary constraints are imposed on the elements of the weight matrix:
[0057] diagonal elements Ensure that the principal contribution of each structural parameter is close to 1, and that no physical distortion occurs; off-diagonal elements... Limit the cross-coupling strength, maintain the interpretability of structural parameters, and provide sufficient degrees of freedom for optimization.
[0058] In this embodiment, the initial weight matrix The elements are set as perturbations of the identity matrix:
[0059] ,
[0060] in, This is a small perturbation, typically ±0.005, used to break the strict symmetry of the identity matrix and help the optimization algorithm escape saddle points.
[0061] Step S2: Establish the optimization objective function
[0062] Step S2.1: Multi-phase point frequency response simulation
[0063] exist Target phase points The fine and coarse models were used to perform frequency simulations, obtaining two sets of frequency response sequences. and ,in, For the fine model at the phase point Frequency at that location Weight matrix Actual structural parameters of the coarse model generated by linear combination The frequency at the same phase point To optimize the number of iterations.
[0064] In this embodiment, the fine model uses a high-precision CST eigenmode solver and the coarse model uses a low-precision CST eigenmode solver. The fine model uses a fine-mesh CST eigenmode solver with a mesh density of λ / 20 or higher, while the coarse model uses a sparse-mesh CST eigenmode solver with a mesh density of λ / 10 or lower; the simulation time ratio between the two is greater than 5:1.
[0065] Step S2.2: Calculate the objective function value
[0066] The objective function value is the mean square error of the frequency response at multiple phase points of the coarse and fine models. :
[0067] .
[0068] Traditional spatial mapping often optimizes specific frequency points or S-parameters. This invention, however, targets the dispersion characteristics of slow-wave structures and employs a multi-phase-point simultaneous optimization strategy.
[0069] In this embodiment, phase point This strategy ensures that the dispersion curves of the coarse model remain globally consistent with those of the fine model across the entire operating frequency band, rather than only matching at individual frequency points.
[0070] Step S3: Solve for the optimal weight matrix using two-stage adaptive optimization.
[0071] To address the highly nonlinear and multi-extremum characteristics of the weight matrix optimization problem, this invention designs an adaptive three-stage optimization strategy, with automatic switching between stages, balancing global exploration and local fine-grained search. Specifically:
[0072] Initialize optimization iteration count =0;
[0073] Phase 1: Conservative Trust Region Sampling
[0074] Conservative trust region sampling is responsible for global exploration. It balances the stability of master parameters and cross-coupling exploration through a differentiated step size strategy, specifically including the following steps:
[0075] Step S3.1: Trust Region Sampling
[0076] In the current weight matrix Trust region radius Inner, diagonal elements adopt a conservative step size Off-diagonal elements Use a larger step size Random sampling points are generated by grouping by row;
[0077] Step S3.2: Evaluation and Trust Region Radius Adjustment
[0078] Evaluate the objective function value at all sampling points. The optimal step is to update the weight matrix as follows: If the improvement is successful, the trust region radius will be expanded. Otherwise, shrink ;
[0079] Step S3.3: Determine whether the objective function value has not improved or whether the trust region radius is less than or below 0.001.
[0080] Determine the value of the objective function Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. Or trust region radius If the value is below 0.001, switch to Phase Two; otherwise... Return to step S3.1.
[0081] Phase Two: Fine Coordinate Descent
[0082] Focusing on key dimensions for a refined local search involves the following steps:
[0083] Step S3.4: Fine-tune the search and update the weight matrix
[0084] Selecting the weight matrix The diagonal elements are used as optimization variables, employing a fixed step size set:
[0085]
[0086] Perform a fine search and select the target function value Update the selected element with the minimum value, and traverse all weight matrices. The elements are used to complete one round of optimization, resulting in a weight matrix. .
[0087] Step S3.5: Determine if the objective function value has not improved.
[0088] Determine the value of the objective function Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. If so, then The optimal weight matrix Otherwise, otherwise Return to step S3.4.
[0089] Step S4: Coarse Model Validation and Application
[0090] The optimal weight matrix It is applied to the coarse model to verify its frequency response consistency with the fine model across the entire phase range. Once the accuracy requirements are met, it is used for the design of tortuous waveguide slow wave structures. If the accuracy requirements are not met, the basic structural parameters are reset and optimized again until the accuracy requirements are met.
[0091] Example
[0092] In this example, coarse and fine models of a V-band planar tortuous waveguide slow-wave structure are jointly optimized, with a target operating frequency band of 56–76 GHz. The basic structural parameters are as follows: , , , , .
[0093] Step 1: Configure coarse and fine models
[0094] (1) Detailed model: The CST intrinsic mode solver was used, the mesh density was set to λ / 25, and the simulation time was about 120 seconds per phase point;
[0095] (2) Coarse model: The CST intrinsic mode solver was used, the mesh density was set to λ / 8, and the simulation time was about 24 seconds per phase point;
[0096] (3) The time ratio of coarse to fine models is about 5:1, which meets the computational efficiency requirements of the spatial mapping method.
[0097] Step 2: Initialize the weight matrix
[0098] Set the initial weight matrix:
[0099]
[0100] Apply tiny perturbations (±0.005) to the diagonal elements to break the strict symmetry of the identity matrix, which helps optimize saddle point escape.
[0101] Step 3: Multi-phase point reference simulation
[0102] At 6 phase points ( Perform detailed model simulation to obtain the reference frequency response:
[0103]
[0104] Initial simulation frequency response of the coarse model:
[0105]
[0106] Initial objective function value:
[0107]
[0108] Step 4: Solve for the optimal weight matrix using two-stage adaptive optimization.
[0109] (1) Phase 1: The initial radius of the trust region is Δ0 = 0.03, and sampling is performed in rows. Due to the initial exploration after relaxing the boundary, some parameters went out of bounds. After the trust region showed no improvement for 5 consecutive times, the process was switched to Phase 2.
[0110] (2) Phase Two (Iterations 6-27): Focus on fine-grained search of diagonal elements and adjacent off-diagonal elements. The 6th iteration involved adjustments... From 1.00 to 0.98, the objective function decreased from 273.52 to 213.63; subsequent dimension-by-dimensional fine-tuning caused the objective function to continue to decrease, reaching 1.10 after 22 rounds of coordinate reduction. Iteration 28: Adjustment From 1.00 to 0.988, the objective function decreased to 0.98. After five consecutive iterations without improvement, the optimal weight matrix was obtained. :
[0111]
[0112] Step 5: Optimization Result Verification
[0113] Frequency response comparison of the coarse model at 6 phase points with that of the fine model:
[0114]
[0115] Table 1
[0116] The total number of simulations was 134 (6 fixed simulations for the fine model + 128 simulations for the coarse model), taking about 1.5 hours. Compared with direct fine model optimization (which is expected to require hundreds of simulations and tens of hours), the efficiency is improved by about 15 times.
[0117] Figure 2 This demonstrates the core principle of weight matrix parameterization in this invention. Basic structural parameters ( , , , , By linearly combining 5×5 weight matrices W, the actual structural parameters of the coarse model are generated. Diagonal elements Controlling the contribution of main parameters, off-diagonal elements By introducing cross-coupling compensation, the coarse model can approximate the electromagnetic response of the fine model through equivalent adjustment of geometric parameters without changing the mesh density.
[0118] Figure 3 A three-dimensional vacuum structure diagram of the tortuous waveguide slow wave structure according to an embodiment of the present invention is shown, with the length of the wide side of the waveguide marked. Waveguide narrow side length Waveguide bending diameter Straight waveguide length and electron beam channel radius Five key structural parameters.
[0119] Figure 4 The example demonstrates a comparison of the frequency responses of the coarse and fine models before and after optimization. Before optimization (red dashed line), the coarse model deviates significantly from the fine model (green solid line) at six phase points, and the initial objective function... =273.52; After optimization (marked by blue triangles), the frequency response of the coarse model is in high agreement with that of the fine model, and the objective function is reduced to 273.52. =0.98, and especially at the 450° phase point, the error is only 0.06 GHz, which verifies the effectiveness of the weight matrix parameterization method.
[0120] Figure 5 The convergence curve of the objective function during the optimization process is shown. The curve exhibits distinct phases: Phase 1 (iterations 1-5) shows no improvement in trust region exploration; Phase 2 (iterations 6-27) enters the fine-grained coordinate descent phase, with the objective function rapidly decreasing from 273.52 to 1.10; finally, in the 28th iteration, it converges to 0.98 through w44 fine-tuning, satisfying the condition. <1.0.
[0121] This example demonstrates that the proposed collaborative optimization method for coarse and fine models of traveling wave tube tortuous waveguides based on weight matrix parameterization eliminates the need to construct an independent coarse model. Instead, it directly reconstructs the internal parameters of existing simulation models through the weight matrix. Combined with a two-stage adaptive optimization strategy and the CST automation framework, it can obtain a fast coarse model that approximates the fine model with high accuracy in a short time, significantly improving the design efficiency of traveling wave tube slow wave structures.
[0122] Although the illustrative specific embodiments of the present invention have been described above to enable those skilled in the art to understand the invention, it should be understood that the 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 invention as defined and determined by the appended claims, and all inventions utilizing the concept of the present invention are protected.
Claims
1. A method for collaborative optimization of a roughness model of a TWT meander waveguide based on weight matrix parameterization, characterized in that, Includes the following steps: (1) Construct a coarse model parameterized by the weight matrix; Establish a 5×5 initial weight matrix : ; The basic structural parameters of the meandering waveguide slow wave structure are: the reference length of the wide side of the waveguide. Waveguide narrow side reference length Waveguide bending diameter reference value Straight waveguide length reference value Electron injection channel radius reference value As input, the actual structural parameters of the coarse model are generated through linear combination: , , , , ; (2) Establish the objective function for optimization; 2.1) Multi-phase point frequency response simulation; exist Target phase points The fine and coarse models were used to perform frequency simulations, obtaining two sets of frequency response sequences. and ,in, For the fine model at the phase point Frequency at that location Weight matrix Actual structural parameters of the coarse model generated by linear combination The frequency at the same phase point To optimize the number of iterations; 2.2) Calculate the objective function value ; The objective function value is the mean square error of the frequency response at multiple phase points of the coarse and fine models. : ; (3) Solving the optimal weight matrix using two-stage adaptive optimization ; Initialize optimization iteration count =0; Phase 1: Conservative Trust Region Sampling; 3.1) In the current weight matrix Trust region radius Inner, diagonal elements adopt a conservative step size Off-diagonal elements Use a larger step size Random sampling points are generated by grouping by row; 3.2) Evaluate the objective function value of all sampling points. The optimal step is to update the weight matrix as follows: If the improvement is successful, the trust region radius will be expanded. Otherwise, shrink ; 3.3) Determine the value of the objective function Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. Or trust region radius If the value is below 0.001, switch to Phase Two; otherwise... (Return to step 3.1). Phase Two: Fine Coordinate Descent 3.4) Selecting the weight matrix The diagonal elements are used as optimization variables, employing a fixed step size set: ; Perform a fine search and select the target function value Update the selected element with the minimum value, and traverse all weight matrices. The elements are used to complete one round of optimization, resulting in a weight matrix. ; 3.5) Determine the value of the objective function. Is it continuous? No improvement is achieved if the decrease is less than the set minimum threshold. If so, then The optimal weight matrix Otherwise, otherwise (Return to step 3.4). (4) Coarse model verification and application; The optimal weight matrix It is applied to the coarse model to verify its frequency response consistency with the fine model across the entire phase range. Once the accuracy requirements are met, it is used for the design of tortuous waveguide slow wave structures. If the accuracy requirements are not met, the basic structural parameters are reset and optimized again until the accuracy requirements are met.
2. The method for collaborative optimization of thick and thin models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization according to claim 1, characterized in that, For the weight matrix Apply boundary constraints to elements: diagonal elements Ensure that the principal contribution of each structural parameter is close to 1, and that no physical distortion occurs; off-diagonal elements... Limit the cross-coupling strength, maintain the interpretability of structural parameters, and provide sufficient degrees of freedom for optimization.
3. The method for collaborative optimization of thick and thin models of traveling wave tubes and tortuous waveguides based on weight matrix parameterization according to claim 1, characterized in that, The initial weight matrix The elements are set as perturbations of the identity matrix: , ; in, This is a small perturbation, typically ±0.005, used to break the strict symmetry of the identity matrix and help the optimization algorithm escape saddle points.
4. The method for coarse and fine model optimization of traveling wave tube tortuous waveguide based on weight matrix parameterization according to claim 1, wherein the fine model is a high-precision CST eigenmode solver and the coarse model is a low-precision CST eigenmode solver. The fine model uses a fine-grid CST eigenmode solver with a grid density of λ / 20 or higher, and the coarse model uses a sparse-grid CST eigenmode solver with a grid density of λ / 10 or lower; the simulation time ratio between the two is greater than 5:1.
