A Fast Optimization Method for SIW Filters Based on GA Improved J-Update Matrix
Through the GA-improved J update matrix SIW filter rapid optimization method, combined with genetic algorithm and BFGS quasi-Newtonian algorithm, the problems of initial point selection and local optimal solution in filter optimization are solved, and efficient filter design is realized, which is significant Reduced simulation times and optimization time.
Patent Information
- Application Number
- CN202411310474.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-20
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-09-20
AI Technical Summary
Existing filter optimization algorithms are difficult to determine the initial point and search space, and are easily trapped in the local optimal solution, and lack effective solutions.
The SIW filter rapid optimization method of GA improved J update matrix is adopted, combined with genetic algorithm and BFGS quasi-Newtonian algorithm, and through full-wave electromagnetic simulation and model update, the coupling matrix and feature parameters are gradually adjusted to ensure the initial point reliability of the optimization process and stability.
It significantly reduces the number of full-wave electromagnetic simulations, shortens the optimization time, improves design efficiency, and ensures the robustness and accuracy of the optimization process.
Smart Images

Figure CN118821632B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of filter optimization, and in particular to a SIW filter fast optimization method of a GA improved J update matrix. Background Art
[0002] In recent years, with the rapid development of mobile communication systems, filters, as one of the key components of millimeter wave communication systems, face design challenges such as miniaturization, high integration, and high performance. Their design is becoming increasingly complex, and the number of parameters that can be optimized, the difficulty of optimization, and the cost of optimization are increasing. Applying optimization algorithms to filter design to reduce computational costs and improve optimization efficiency is one of the current research hotspots. The most commonly used optimization algorithms include gradient optimization algorithms such as quasi-Newton method, simplex method, and sequential quadratic programming, as well as random optimization algorithms such as genetic algorithms and differential evolution.
[0003] However, due to the complexity of the filter itself, when the above algorithm is used to optimize the filter, it is difficult to determine a set of sufficiently good initial points or search spaces, and it is easy to fall into a local optimal solution.
[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention
[0005] 1. Technical issues to be resolved
[0006] In view of the deficiencies of the prior art, the present invention provides a SIW filter fast optimization method of a GA improved J update matrix to overcome the above technical problems existing in the prior art.
[0007] To this end, the specific technical solution adopted by the present invention is as follows:
[0008] A GA improved J update matrix SIW filter fast optimization method, the GA improved J update matrix SIW filter fast optimization method comprises the following steps:
[0009] S1. Initialize optimization parameters and load the target coupling matrix and filter initial electromagnetic model size parameters;
[0010] S2. Perform full-wave electromagnetic simulation on the electromagnetic model of the filter to obtain the filter response result, and determine whether the filter response result meets the preset target requirements;
[0011] S3. If the preset target requirements are not met, the coupling matrix is adjusted by genetic algorithm to obtain the coupling matrix deviation, and the filter characteristic parameter deviation is calculated according to the coupling matrix deviation. If the target requirements are met, the optimization is terminated.
[0012] S4, using the general model function to calculate the Jacobian matrix, and convert the characteristic parameter deviation into the electromagnetic model size parameter deviation, and update the filter electromagnetic model;
[0013] S5. Perform full-wave electromagnetic simulation on the updated filter electromagnetic model to obtain an updated filter response result, and determine whether the filter response result meets the preset target requirements;
[0014] S6. If the preset target requirements are not met, the quasi-Newton algorithm is used to approximate the response results of the filter electromagnetic model, extract the coupling matrix deviation, calculate the filter characteristic parameter deviation, and return to step S4. If the preset target requirements are not met, the optimization is terminated.
[0015] Preferably, the method of adjusting the coupling matrix by a genetic algorithm to obtain a coupling matrix deviation, and calculating the filter characteristic parameter deviation according to the coupling matrix deviation comprises the following steps:
[0016] S31, initializing the coupling matrix deviation, and randomly generating an initial population of coupling matrix deviations;
[0017] S32, calculating the fitness function value of each individual in the population;
[0018] S33, according to the fitness function value of each individual in the population, performing selection, crossover and mutation operations on each individual to generate a new population;
[0019] S34, calculating the fitness function value of each individual in the new population, and selecting the individual with the highest fitness function value as the new generation population;
[0020] S35, repeating steps S33-S34 until the preset end condition is reached, selecting the individual with the highest fitness function value as the output value, and obtaining the final coupling matrix deviation;
[0021] S26. Calculate the filter characteristic parameter deviation according to the characteristic parameter deviation calculation formula.
[0022] Preferably, the characteristic parameter deviation calculation formula is:
[0023]
[0024] In the formula, Represents the filter characteristic parameter deviation; Represents the characteristic parameter vector, including positive coupling coefficient, negative coupling coefficient, external quality factor, resonant frequency, etc.; represents the target coupling matrix; represents the coupling matrix deviation.
[0025] Preferably, the method of calculating the Jacobian matrix using a general model function and converting the characteristic parameter deviation into an electromagnetic model size parameter deviation, and updating the filter electromagnetic model comprises the following steps:
[0026] S41, constructing a general model function of a SIW filter;
[0027] S42, deriving each function in the SIW filter general model function, and constructing a Jacobian matrix according to the derivative results;
[0028] S43, converting the characteristic parameter deviation into the electromagnetic model size parameter deviation according to the Jacobian matrix;
[0029] S44. Based on the electromagnetic model size parameter deviation, the size parameters of the filter electromagnetic model are updated through a model parameter update formula.
[0030] Preferably, the SIW filter universal model function includes a positive coupling coefficient function, a negative coupling coefficient function, an external quality factor function and a resonant frequency function.
[0031] Preferably, the model parameter updating formula is:
[0032] ;
[0033] In the formula, represents the updated filter electromagnetic model size parameters; Represents the size parameters of the initial electromagnetic model of the filter; Represents the correction factor.
[0034] Preferably, the method of using a quasi-Newton algorithm to approximate the response result of the filter electromagnetic model, extracting the coupling matrix deviation, and calculating the filter characteristic parameter deviation comprises the following steps:
[0035] S61, initializing quasi-Newton algorithm parameters;
[0036] S62, based on the updated electromagnetic model of the filter, obtaining the response result of the electromagnetic model of the filter through electromagnetic simulation, and calculating the gradient value of the current coupling matrix deviation;
[0037] S63, calculating the search direction according to the initial value of the positive definite symmetric matrix and the gradient value of the current coupling matrix deviation, and using the Armijo step search method to solve the step length that satisfies the preset inequality;
[0038] S64, updating the electromagnetic model parameters of the filter according to the obtained step size and search direction, and performing electromagnetic simulation again based on the updated electromagnetic model parameters of the filter to obtain a new filter response result;
[0039] S65, determining a deviation of a new coupling matrix according to a new filter response result, and calculating a filter characteristic parameter deviation again by a characteristic parameter deviation calculation formula according to the deviation of the new coupling matrix;
[0040] S66. If the obtained filter characteristic parameter deviation does not meet the preset target requirement, the positive definite symmetric matrix is updated using the matrix update formula and the process returns to step S62.
[0041] Preferably, the quasi-Newton algorithm parameters include a step size search parameter, a termination error, and an initial value of a positive definite symmetric matrix.
[0042] Preferably, the calculation formula for calculating the search direction is:
[0043] ;
[0044] In the formula, Indicates the search direction of the kth iteration; represents a positive definite symmetric matrix; Represents the gradient value of the k-th coupling matrix deviation.
[0045] Preferably, the matrix update formula is:
[0046] ;
[0047] In the formula, represents the positive definite symmetric matrix at the k+1th iteration; represents the positive definite symmetric matrix at the kth iteration; Indicates the step size at the kth iteration; Represents the change in the gradient value in the kth iteration; T represents the transpose operation.
[0048] (III) Beneficial effects
[0049] Compared with the prior art, the present invention provides a fast optimization method for SIW filter using GA improved J update matrix, which has the following beneficial effects:
[0050] 1. The present invention extracts characteristic parameter deviations through GA algorithm and BFGS quasi-Newton algorithm, which ensures the reliability of initial point selection of optimization algorithm and the stability of subsequent optimization. This optimization method greatly reduces the number of full-wave electromagnetic simulations in the SIW filter design optimization process, shortens the time spent on simulation optimization, and improves design efficiency.
[0051] 2. The present invention uses a genetic algorithm with stronger global optimization search capability for the first extraction of feature parameter deviations to ensure the reliability of initial point extraction, and uses the BFGS quasi-Newton algorithm with a faster convergence speed in the subsequent optimization iteration process to ensure the accuracy of feature parameter deviation extraction and the robustness of the overall optimization process. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0053] Figure 1 is a flow chart of a SIW filter fast optimization method according to an embodiment of the present invention using a GA improved J update matrix;
[0054] Figure 2 is a schematic diagram of a SIW coupled resonator in a SIW filter fast optimization method according to a GA improved J update matrix according to an embodiment of the present invention;
[0055] Figure 3 1 is a schematic diagram of positive coupling coefficients under different resonant cavity sizes in a SIW filter fast optimization method according to a GA improved J update matrix according to an embodiment of the present invention;
[0056] Figure 4 2. It is a schematic diagram of a negative coupling structure of a SIW symmetrical rectangular slot in a SIW filter fast optimization method according to a GA improved J update matrix according to an embodiment of the present invention;
[0057] Figure 5 is a schematic diagram of negative coupling coefficients under different resonant cavity sizes in a SIW filter fast optimization method based on a GA improved J update matrix according to an embodiment of the present invention;
[0058] Figure 6 1 is a schematic diagram of a single-port model of a SIW resonant cavity in a SIW filter fast optimization method according to a GA improved J update matrix according to an embodiment of the present invention;
[0059] Figure 7 1 is a schematic diagram of external quality factors under different external coupling window sizes in a SIW filter fast optimization method using a GA improved J update matrix according to an embodiment of the present invention;
[0060] Figure 8 1. It is a schematic diagram of the structure of a 4th-order cross-coupled SIW filter in a SIW filter fast optimization method based on a GA improved J-update matrix according to an embodiment of the present invention;
[0061] Fig. 9 is a schematic diagram of initial response results in a SIW filter fast optimization method using a GA improved J update matrix according to an embodiment of the present invention;
[0062] Fig.10 1 is a schematic diagram of the first iteration response result in the SIW filter fast optimization method according to the GA improved J update matrix of the embodiment of the present invention;
[0063] Fig.11 is a schematic diagram of the fourth response result in the SIW filter fast optimization method according to the GA improved J update matrix of an embodiment of the present invention;
[0064] Fig.12 It is a schematic diagram of the optimal solution response result in the SIW filter fast optimization method of the GA improved J update matrix according to an embodiment of the present invention. DETAILED DESCRIPTION
[0065] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0066] According to an embodiment of the present invention, a SIW filter fast optimization method with a GA improved J update matrix is provided.
[0067] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to the GA improved J update matrix SIW filter fast optimization method of the embodiment of the present invention, the GA improved J update matrix SIW filter fast optimization method comprises the following steps:
[0068] S1. Initialize optimization parameters and load the target coupling matrix and filter initial electromagnetic model size parameters;
[0069] It should be noted that the optimization parameters include: the number of electromagnetic simulation iterations n=0, the filter electromagnetic model size parameters and correction factors .
[0070] S2. Perform full-wave electromagnetic simulation on the electromagnetic model of the filter to obtain the filter response result, and determine whether the filter response result meets the preset target requirements;
[0071] S3. If the preset target requirements are not met, the coupling matrix is adjusted by genetic algorithm to obtain the coupling matrix deviation, and the filter characteristic parameter deviation is calculated according to the coupling matrix deviation. If the target requirements are met, the optimization is terminated.
[0072] It should be noted that the genetic algorithm is a random search algorithm that simulates genetic selection and natural elimination in the biological world. It searches for the optimal solution by performing three basic operations: selection, crossover, and mutation on chromosomes in each iteration. All feasible solutions (called chromosomes, individuals) generated in each iteration constitute a population. A feasible solution generally consists of multiple elements, and each of these elements is called a gene on the chromosome. Starting from the initial population, the fitness function value corresponding to each chromosome is calculated, and relatively good chromosomes are selected based on the fitness probability for crossover and mutation to produce the next generation of populations. In this way, the evolution of life is imitated and iterated optimization is continuously carried out until the target requirements are met.
[0073] As a preferred implementation, the method of adjusting the coupling matrix using a genetic algorithm to obtain a coupling matrix deviation, and calculating the filter characteristic parameter deviation according to the coupling matrix deviation includes the following steps:
[0074] S31, initializing the coupling matrix deviation, and randomly generating an initial population of coupling matrix deviations;
[0075] S32, calculating the fitness function value of each individual in the population;
[0076] S33, according to the fitness function value of each individual in the population, performing selection, crossover and mutation operations on each individual to generate a new population;
[0077] S34, calculating the fitness function value of each individual in the new population, and selecting the individual with the highest fitness function value as the new generation population;
[0078] S35, repeating steps S33-S34 until the preset end condition is reached, selecting the individual with the highest fitness function value as the output value, and obtaining the final coupling matrix deviation;
[0079] S26. Calculate the filter characteristic parameter deviation according to the characteristic parameter deviation calculation formula.
[0080] Specifically, the characteristic parameter deviation calculation formula is:
[0081] ;
[0082] In the formula, Represents the filter characteristic parameter deviation; Represents the characteristic parameter vector, including positive coupling coefficient, negative coupling coefficient, external quality factor, resonant frequency, etc.; represents the target coupling matrix; represents the coupling matrix deviation.
[0083] S4, using the general model function to calculate the Jacobian matrix, and convert the characteristic parameter deviation into the electromagnetic model size parameter deviation, and update the filter electromagnetic model;
[0084] As a preferred implementation, the Jacobian matrix is calculated using a general model function, and the characteristic parameter deviation is converted into an electromagnetic model size parameter deviation, and updating the filter electromagnetic model includes the following steps:
[0085] S41, constructing a general model function of a SIW filter;
[0086] The general model function of the SIW filter includes a positive coupling coefficient function, a negative coupling coefficient function, an external quality factor function and a resonant frequency function.
[0087] It should be noted that for the positive coupling coefficient function, the structure used to extract the positive coupling coefficient model is as follows Figure 2 As shown, the positive coupling coefficient is mainly related to the size of the coupling window, and the calculation formula is as follows:
[0088] ;
[0089] In the formula, Indicates the magnitude of the positive coupling coefficient; represents the main mode resonant frequency; represents the first higher-order mode resonant frequency;
[0090] Here, electromagnetic simulation software HFSS is used to solve the eigenmode of the SIW coupled resonator and obtain the main mode resonant frequency and the first higher-order mode resonant frequency Based on the consideration of establishing a universal coupling coefficient model applicable to a wide frequency range, the eigenmodes of different coupling window sizes are solved under different resonant cavity sizes, and the corresponding coupling coefficients are calculated. Figure 3 As shown in the figure, the positive coupling coefficient extraction results are given when the resonant cavity sizes are 1.7mm, 1.8mm, 1.9mm and 2mm respectively. Based on curve fitting, the following general model function for the positive coupling coefficient function can be obtained:
[0091] ;
[0092] In the formula, represents the positive coupling coefficient function; Indicates the magnitude of the positive coupling coefficient;
[0093] By taking the derivative of the positive coupling coefficient function, the relationship between the coupling window size deviation and the positive coupling coefficient deviation is obtained as follows:
[0094] ;
[0095] In the formula, represents the relationship between the coupling window size deviation and the positive coupling coefficient deviation; represents the equivalent length of the waveguide cavity; Indicates physical length; Indicates the hole diameter; Indicates and +1 coupling coefficient between cavities; Indicates a small change in the coupling coefficient.
[0096] For the negative coupling coefficient function, compared with Figure 2 The coupling method shown is as follows Figure 4 The symmetrical rectangular slot structure shown can introduce cross negative coupling in SIW filter design by changing the rectangular slot width W. s And the rectangular slot length L nc The negative coupling strength can be effectively controlled, thereby controlling the transmission zero point position. In order to simplify the optimization complexity, the present invention fixes the rectangular slot width to W. s =0.1mm. Similar to the process of extracting the positive coupling coefficient, under different resonant cavity sizes, the eigenmode is used to solve the change of the main mode and the first high-order mode resonant frequency of the coupled resonator with the length of the rectangular slot, and then the corresponding coupling coefficient is calculated. Figure 5 As shown in the figure, the negative coupling coefficient extraction results are given when the resonant cavity sizes are 1.8mm, 1.9mm and 2mm respectively. Through fitting, a general model function for the negative coupling coefficient function can be established as follows:
[0097] ;
[0098] In the formula, represents the negative coupling coefficient function; Indicates the magnitude of the positive coupling coefficient;
[0099] By taking the derivative of the negative coupling coefficient function, the relationship between the rectangular slot length deviation and the negative coupling coefficient deviation can be obtained as follows:
[0100] ;
[0101] In the formula, It represents the relationship between the deviation of rectangular slot length and the deviation of negative coupling coefficient; Indicates the magnitude of the positive coupling coefficient; represents the equivalent length of the waveguide cavity; Indicates a small change in the coupling coefficient.
[0102] For the external quality factor function, the extraction of the external quality factor of the filter requires the establishment of its single-port model, such as Figure 6 As shown, a HFSS full-wave simulation is performed, the window width is changed and the corresponding external quality factor value is calculated according to the following formula:
[0103] ;
[0104] In the formula, represents the external quality factor value; represents the center frequency of the resonator, which can be expressed by Figure 6 The group delay peak value of the reflection coefficient corresponding to the structure is obtained; Indicates the bandwidth range corresponding to a phase shift of ±90° about the center frequency phase;
[0105] exist Figure 7 The external quality factor extraction results under different window widths are given in . Based on curve fitting, the model function of the external quality factor function can be obtained as:
[0106] ;
[0107] In the formula, represents the external quality factor function; represents the external quality factor value;
[0108] By taking the derivative of the external quality factor function, the relationship between the window width deviation and the external quality factor deviation can be obtained as:
[0109] ;
[0110] In the formula, represents the relationship between the window width deviation and the external quality factor deviation; Indicates physical length; represents the external quality factor value; Represents the external quality factor deviation.
[0111] For the resonant frequency function, the resonant frequency function of a general SIW rectangular cavity can be approximately calculated by the following formula:
[0112] ;
[0113] ;
[0114] In the formula, represents the resonant frequency function of the SIW rectangular cavity; represents the speed of light in a vacuum; Indicates the relative magnetic permeability of the dielectric substrate; Represents the relative dielectric constant of the dielectric substrate; and They represent the length and width of the equivalent dielectric rectangular waveguide filled with the SIW rectangular cavity; and Represents an integer; Represents the SIW equivalent waveguide length and physical length , aperture And the hole spacing It matters.
[0115] In actual SIW filter design, in order to reduce the number of design variables, usually only one side is selected to control the resonant frequency. The resonant frequency of the above SIW rectangular cavity is derived, and the relationship function between the resonant frequency deviation and the resonant cavity size deviation is obtained as follows:
[0116] ;
[0117] In the formula, It represents the relationship between the deviation of the resonance frequency and the deviation of the resonance cavity size; represents the size of the resonant cavity in the x direction; represents the size of the resonant cavity in the y direction; represents the resonant frequency; Indicates SIW band cavity; Indicates the resonant frequency deviation; Represents the deviation of the resonant cavity size in the y direction.
[0118] S42, deriving each function in the SIW filter general model function, and constructing a Jacobian matrix according to the derivative results;
[0119] S43, converting the characteristic parameter deviation into the electromagnetic model size parameter deviation according to the Jacobian matrix;
[0120] S44. Based on the electromagnetic model size parameter deviation, the size parameters of the filter electromagnetic model are updated through a model parameter update formula.
[0121] Specifically, the model parameter update formula is:
[0122] ;
[0123] In the formula, represents the updated filter electromagnetic model size parameters; Represents the size parameters of the initial electromagnetic model of the filter; Represents the correction factor.
[0124] S5. Perform full-wave electromagnetic simulation on the updated filter electromagnetic model to obtain an updated filter response result, and determine whether the filter response result meets the preset target requirements;
[0125] S6. If the preset target requirements are not met, the quasi-Newton algorithm is used to approximate the response results of the filter electromagnetic model, extract the coupling matrix deviation, calculate the filter characteristic parameter deviation, and return to step S4. If the preset target requirements are not met, the optimization is terminated.
[0126] Specifically, the quasi-Newton algorithm is one of the most effective methods for solving nonlinear optimization problems. It overcomes the defects of the Newton method, such as the need to solve the inverse matrix of the complex Hessian matrix each time, and approximates the Hessian matrix (or the inverse of the Hessian matrix) by constructing a positive definite symmetric matrix, thereby simplifying the complexity of the operation. Different construction methods produce different quasi-Newton methods. The present invention adopts the famous BFGS quasi-Newton algorithm. The specific principle implementation process of the BFGS quasi-Newton algorithm is as follows:
[0127] Step 1: Given parameters , , initialization point , termination error , initialize the positive definite symmetric matrix , let the number of iterations ,in express dimensional identity matrix, represents the initialization point of the parameter x to be optimized, express dimensional real vector space;
[0128] Step 2: Calculate the numerical gradient ,like , stop, output As an approximate minimum point;
[0129] in, It means that the function f(x) is The gradient value at ;
[0130] Step 3: Take , and at the same time ;
[0131] Step 4: Use the Armijo step search method to find the smallest non-negative integer m that satisfies the following inequality;
[0132] ,make , ;
[0133] In the formula, Represents the objective function after the search; and Both represent the step length along the search direction; Represents the output value of the k+1th iteration; Indicates the search direction of the kth iteration;
[0134] Step 5: Calculate by the matrix update formula ,make , and jump to step 2.
[0135] As a preferred implementation, the use of a quasi-Newton algorithm to approximate the response results of the filter electromagnetic model, extract the coupling matrix deviation, and calculate the filter characteristic parameter deviation includes the following steps:
[0136] S61, initializing quasi-Newton algorithm parameters; the quasi-Newton algorithm parameters include step length search parameters and , termination error and initial values for positive definite symmetric matrices;
[0137] Based on the updated electromagnetic model of the filter, a response result of the electromagnetic model of the filter is obtained through electromagnetic simulation, and a gradient value of the current coupling matrix deviation is calculated;
[0138] In addition, it should be noted that if the gradient value of the current coupling matrix deviation is , stop updating and directly output the current coupling matrix deviation.
[0139] S63, calculating the search direction according to the initial value of the positive definite symmetric matrix and the gradient value of the current coupling matrix deviation, and using the Armijo step search method to solve the step length that satisfies the preset inequality;
[0140] Specifically, the calculation formula for calculating the search direction is:
[0141] ;
[0142] In the formula, Indicates the search direction of the kth iteration; represents a positive definite symmetric matrix; Represents the gradient value of the k-th coupling matrix deviation.
[0143] S64, updating the electromagnetic model parameters of the filter according to the obtained step size and search direction, and performing electromagnetic simulation again based on the updated electromagnetic model parameters of the filter to obtain a new filter response result;
[0144] S65, determining a deviation of a new coupling matrix according to a new filter response result, and calculating a filter characteristic parameter deviation again by a characteristic parameter deviation calculation formula according to the deviation of the new coupling matrix;
[0145] S66. If the obtained filter characteristic parameter deviation does not meet the preset target requirement, the positive definite symmetric matrix is updated using the matrix update formula and the process returns to step S62.
[0146] Specifically, the matrix update formula is:
[0147] ;
[0148] In the formula, represents the positive definite symmetric matrix at the k+1th iteration; represents the positive definite symmetric matrix at the kth iteration; Indicates the step size at the kth iteration; Represents the change in the gradient value in the kth iteration; T represents the transpose operation.
[0149] in, ; ;
[0150] In the formula, Indicates the step size at the kth iteration; represents the output value of the k+1th iteration, i.e., the coupling matrix deviation; represents the output value of the kth iteration, i.e., the coupling matrix deviation; Represents the change in the gradient value in the kth iteration;
[0151] It means that the function f(x) is The gradient at .
[0152] In order to better understand the above technical solutions of the present invention, the above technical solutions of the present invention are further explained from the perspective of architecture and principle as follows:
[0153] Assume that the characteristic parameter vector of the SIW filter is ,in, represents the resonant frequency of the ith cavity; represents the coupling coefficient between the i-th cavity and the j-th cavity; It represents the no-load quality factor; represents the external quality factor; the design parameter vector of the electromagnetic model is ,in, represents the physical length of the waveguide cavity; represents the coupling width of the waveguide cavity; represents the size of the cross-coupling structure; represents the external coupling width of the port; the corresponding full-wave electromagnetic simulation result, i.e., the filter response, is recorded as , the filter response with the normalized coupling matrix M as the basic parameter is recorded as . Figure 1 It is a flowchart of the improved method for fast optimization of updating filter based on general model Jacobian matrix of the present invention, and the process is summarized as follows:
[0154] Step 1: Initialize the number of electromagnetic simulation iterations n=0, and the correction factor , load the target coupling matrix and filter initial electromagnetic model size parameters ;
[0155] Step 2: Perform full-wave electromagnetic simulation on the filter electromagnetic model to obtain the filter response , if the response meets the target requirements, stop and exit the optimization;
[0156] Step 3: Use the GA algorithm to make the filter response corresponding to the coupling matrix Approximate electromagnetic simulation response results , and obtain the coupling matrix deviation , calculate the filter characteristic parameter deviation ;
[0157] Step 4: Calculate the Jacobian matrix based on the established general model function , and the characteristic parameter deviation Converted into electromagnetic model size parameter deviation , update the filter electromagnetic model ;
[0158] Step 5. Settings , according to the electromagnetic model size parameters Perform full-wave electromagnetic simulation on the filter to obtain the electromagnetic response results ;
[0159] Step 6: Use the BFGS algorithm to approximate the electromagnetic model response results and extract the coupling matrix deviation , calculate the filter characteristic parameter deviation ; and return to step 4.
[0160] For example, a 4th-order cross-negative coupled SIW Chebyshev filter is used to verify the feasibility of the optimization method proposed in the present invention. The specific design index requirements are: the center frequency is 80GHz, the bandwidth is 2GHz, the return loss is ≤-20dB, and the transmission zero position is 77GHz and 83GHz. The corresponding target coupling matrix is:
[0161] ;
[0162] Among them, the model structure of the 4th-order cross-negatively coupled SIW filter used for HFSS full-wave electromagnetic simulation is as follows: Figure 8 As shown, a Rogers 5880 plate with a thickness of 0.508 mm and a dielectric constant of 2.2 is used. is the width of one side of the resonant cavity, fixed at 2.52mm, The width of the symmetrical rectangular slot is fixed at 0.1 mm. Used to control the resonant frequency of the resonant cavity, the resonant cavity coupling window Used to adjust the positive coupling coefficient, rectangular slot length Used to control the cross negative coupling coefficient, external coupling window width Used to adjust the external quality factor. Since the 4th order filter is a symmetrical structure (the 4th order filter has four resonant cavities, the structure is centrally symmetrical, and the four resonant cavities are sequentially connected with a total of three coupling windows, namely , , , with central symmetry; negative coupling between resonant cavity 1 and resonant cavity 4), in order to reduce the complexity of the design, the size parameter vector of the filter electromagnetic model is: ;
[0163] In the formula, represents the size parameter vector of the filter electromagnetic model; and Represent the lengths of the first two resonant cavities (the lengths of the last two resonant cavities are L 2 and L 1 ); , Both represent coupling windows; Represents the length of the rectangular slot gap, which is used to change the magnitude of the negative coupling; represents the external coupling window width;
[0164] The corresponding characteristic parameter vector is: ;
[0165] In the formula, represents the feature parameter vector; represents the resonant frequency of resonant cavity 1; represents the resonant frequency of resonant cavity 2; represents the coupling coefficient between resonant cavities 1 and 2; represents the coupling coefficient between resonant cavities 2 and 3; represents the coupling coefficient between resonant cavities 1 and 4; represents the external quality factor value;
[0166] Among them, the resonant frequencies of the four resonant cavities are ; Their size and cavity dimensions related.
[0167] The optimization method of the present invention is used to iteratively optimize the 4th order cross negative coupling filter, such as Figure 9-12 As shown in the figure, the response results of several iterations and the response results of the final solution are given. From the iterative results, it can be seen that with the progress of iterative optimization, the electromagnetic simulation response results of the filter gradually improve, and the target response results are achieved after 7 iterations. The final response results show that the operating frequency band of the filter is 78.95-81.02GHz, the overall return loss in the band is less than 20dB, and the two transmission zeros are located at 77.05GHz and 82.96GHz, which meets the design target requirements. The results verify the feasibility of the proposed optimization method for fast optimization of SIW cross-negative coupling filters.
[0168] In summary, with the aid of the above technical scheme of the present invention, the present invention extracts characteristic parameter deviations through the GA algorithm and the BFGS quasi-Newton algorithm, thereby ensuring the reliability of the initial point selection of the optimization algorithm and the stability of the subsequent optimization. The optimization method greatly reduces the number of full-wave electromagnetic simulations in the SIW filter design optimization process, shortens the time spent on simulation optimization, and improves the design efficiency. The present invention uses a genetic algorithm with stronger global optimization search capability for the first extraction of characteristic parameter deviations to ensure the reliability of initial point extraction, and uses the BFGS quasi-Newton algorithm with a faster convergence speed in the subsequent optimization iteration process to ensure the accuracy of characteristic parameter deviation extraction and the robustness of the overall optimization process.
[0169] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.
[0170] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A fast optimization method for SIW filter with GA improved J update matrix, characterized in that: The SIW filter fast optimization method of the GA improved J update matrix includes the following steps: S1. Initialize optimization parameters and load the target coupling matrix and filter initial electromagnetic model size parameters; S2. Perform full-wave electromagnetic simulation on the electromagnetic model of the filter to obtain the filter response result, and determine whether the filter response result meets the preset target requirements; S3. If the preset target requirements are not met, the coupling matrix is adjusted by genetic algorithm to obtain the coupling matrix deviation, and the filter characteristic parameter deviation is calculated according to the coupling matrix deviation. If the target requirements are met, the optimization is terminated. S4, using the general model function to calculate the Jacobian matrix, and convert the characteristic parameter deviation into the electromagnetic model size parameter deviation, and update the filter electromagnetic model; S5. Perform full-wave electromagnetic simulation on the updated filter electromagnetic model to obtain an updated filter response result, and determine whether the filter response result meets the preset target requirements; S6. If the preset target requirements are not met, the quasi-Newton algorithm is used to approximate the response results of the filter electromagnetic model, extract the coupling matrix deviation, calculate the filter characteristic parameter deviation, and return to step S4. If the preset target requirements are not met, the optimization is terminated.
2. The SIW filter fast optimization method of a GA improved J update matrix according to claim 1 is characterized in that: The method of adjusting the coupling matrix by a genetic algorithm to obtain a coupling matrix deviation and calculating the filter characteristic parameter deviation according to the coupling matrix deviation comprises the following steps: S31, initializing the coupling matrix deviation, and randomly generating an initial population of coupling matrix deviations; S32, calculating the fitness function value of each individual in the population; S33, according to the fitness function value of each individual in the population, performing selection, crossover and mutation operations on each individual to generate a new population; S34, calculating the fitness function value of each individual in the new population, and selecting the individual with the highest fitness function value as the new generation population; S35, repeating steps S33-S34 until the preset end condition is reached, selecting the individual with the highest fitness function value as the output value, and obtaining the final coupling matrix deviation; S26. Calculate the filter characteristic parameter deviation according to the characteristic parameter deviation calculation formula.
3. The SIW filter fast optimization method of a GA improved J update matrix according to claim 2 is characterized in that: The characteristic parameter deviation calculation formula is: ; In the formula, Represents the filter characteristic parameter deviation; represents the feature parameter vector; represents the target coupling matrix; represents the coupling matrix deviation.
4. The SIW filter fast optimization method of a GA improved J update matrix according to claim 1 is characterized in that: The method of calculating the Jacobian matrix by using the general model function and converting the characteristic parameter deviation into the electromagnetic model size parameter deviation and updating the filter electromagnetic model comprises the following steps: S41, constructing a general model function of a SIW filter; S42, deriving each function in the SIW filter general model function, and constructing a Jacobian matrix according to the derivative results; S43, converting the characteristic parameter deviation into the electromagnetic model size parameter deviation according to the Jacobian matrix; S44. Based on the electromagnetic model size parameter deviation, the size parameters of the filter electromagnetic model are updated through a model parameter update formula.
5. The SIW filter fast optimization method of a GA improved J update matrix according to claim 4 is characterized in that: The SIW filter universal model function includes a positive coupling coefficient function, a negative coupling coefficient function, an external quality factor function and a resonant frequency function.
6. The SIW filter fast optimization method of a GA improved J update matrix according to claim 4 is characterized in that: The model parameter update formula is: ; In the formula, represents the updated filter electromagnetic model size parameters; Represents the size parameters of the initial electromagnetic model of the filter; Represents the correction factor.
7. The SIW filter fast optimization method of a GA improved J update matrix according to claim 1 is characterized in that: The method of using the quasi-Newton algorithm to approximate the response result of the filter electromagnetic model, extracting the coupling matrix deviation, and calculating the filter characteristic parameter deviation includes the following steps: S61, initializing quasi-Newton algorithm parameters; S62, based on the updated electromagnetic model of the filter, obtaining the response result of the electromagnetic model of the filter through electromagnetic simulation, and calculating the gradient value of the current coupling matrix deviation; S63, calculating the search direction according to the initial value of the positive definite symmetric matrix and the gradient value of the current coupling matrix deviation, and using the Armijo step search method to solve the step length that satisfies the preset inequality; S64, updating the electromagnetic model parameters of the filter according to the obtained step size and search direction, and performing electromagnetic simulation again based on the updated electromagnetic model parameters of the filter to obtain a new filter response result; S65, determining a deviation of a new coupling matrix according to a new filter response result, and calculating a filter characteristic parameter deviation again by a characteristic parameter deviation calculation formula according to the deviation of the new coupling matrix; S66. If the obtained filter characteristic parameter deviation does not meet the preset target requirement, the positive definite symmetric matrix is updated using the matrix update formula and the process returns to step S62.
8. The SIW filter fast optimization method of a GA improved J update matrix according to claim 7 is characterized in that: The quasi-Newton algorithm parameters include a step size search parameter, a termination error, and an initial value of a positive definite symmetric matrix.
9. The SIW filter fast optimization method of a GA improved J update matrix according to claim 7, characterized in that: The calculation formula for calculating the search direction is: ; In the formula, Indicates the search direction of the kth iteration; represents a positive definite symmetric matrix; Represents the gradient value of the k-th coupling matrix deviation.
10. The SIW filter fast optimization method of a GA improved J update matrix according to claim 7, characterized in that: The matrix update formula is: ; In the formula, represents the positive definite symmetric matrix at the k+1th iteration; represents the positive definite symmetric matrix at the kth iteration; Indicates the step size at the kth iteration; Represents the change in the gradient value in the kth iteration; T represents the transpose operation.
Citation Information
Patent Citations
PMSM parameter recognition method based on covariance coupling UFK (Unscented Kalman Filter) algorithm
CN106487297A
Filter coupling matrix decoupling transformation method based on genetic algorithm
CN110852009A
Filter reverse design and optimization method based on deep learning
CN113128119A
Method for automatically setting PID (Proportion Integration Differentiation) parameters in AGV (Automatic Guided Vehicle) speed regulation based on genetic algorithm convolutional neural network
CN116165880A
Kalman filter parameter automatic tuning method and system and readable storage medium
CN117459024A
