Filter automatic optimization design and debugging method fusing coupling matrix and agent model

Abstract

The application discloses a filter automatic optimization design and debugging method fusing a coupling matrix and an agent model, and comprises the steps that design indexes of the filter are set, a loop processing process is executed, when a loop end condition is met, the execution of the whole loop processing process is ended, and design parameters obtained in the last loop processing process are returned, and the like, in the loop processing process, a loss function vector set is determined, non-dominated sorting is performed, a non-dominated vector set is clustered, a clustering center obtained is taken as a search center for searching, simulation is performed on the obtained design parameters, and the design parameters are mapped into coupling matrix coefficients through an agent model. The application directly optimizes based on the design indexes, can realize simultaneous search on multiple sets of design parameters corresponding to the coupling matrixes meeting the design indexes, and therefore can expand the optimization target to all the coupling matrixes meeting the design indexes, and has high automation degree. The application is widely applied to the technical field of filters.

Description

Technical Field

[0001] This invention relates to the field of filter technology, and in particular to an automatic optimization design and debugging method for filters that integrates coupling matrix and surrogate model. Background Technology

[0002] In wireless communication systems, filters are essential frequency-selective devices, and their performance directly affects the overall system quality. With the rapid development of wireless communication systems, higher standards have been placed on filters, necessitating more complex structures and higher-order filters to achieve the required functions. This also makes the design, optimization, and debugging processes of filters more complex. Manual design and optimization processes are inefficient and typically cannot achieve device optimization in a short time.

[0003] Optimization algorithms can effectively accelerate the design and debugging process. Traditional filter design optimization techniques mainly target a few key design parameters, relying on prior knowledge of devices and designs, and requiring parameter sweep analysis. Furthermore, traditional filter design optimization techniques typically require specifying a fixed coupling matrix as the optimization target. Since the physical topology of the filter is closely related to the coupling matrix, traditional techniques essentially predetermine the physical topology of the filter in each optimization process. Due to the non-uniqueness of the coupling matrix selection, existing techniques can only achieve optimization when a suitable coupling matrix or physically implementable topology is chosen; that is, the designer's human judgment directly affects the optimization result. When optimization fails, the coupling matrix or topology must be reselected, resulting in low utilization of simulated data and low optimization efficiency, thus limiting the application and development of this type of design and optimization method. Summary of the Invention

[0004] To address at least one technical problem, such as the low optimization efficiency of current filter design techniques, the present invention aims to provide an automatic optimization design and debugging method for filters that integrates coupling matrices and surrogate models.

[0005] The automatic optimization design and debugging method for filters that fuse coupling matrices and surrogate models in this embodiment includes the following steps:

[0006] Set the design specifications for the filter;

[0007] The process involves executing a loop. When the loop is the first iteration, design parameters are initialized and generated. A surrogate model is constructed based on these parameters, mapping the space containing the design parameters to the space containing the coupling matrix coefficients. If the loop is not the first iteration, the loss function vector set for the current iteration is determined based on the coupling matrix coefficients from the previous iteration and the design indices. The loss function vector set is then sorted non-dominated to obtain the design parameter set corresponding to the non-dominated loss function vector set for the current iteration (i.e., the non-dominated vector set). This non-dominated vector set is then clustered to obtain at least one cluster center for the current iteration. Using this cluster center as a search center, an intelligent optimization algorithm is used to search for the search center on the surrogate model to obtain the design parameters for the current iteration. The design parameters for the current iteration are then simulated to obtain the filter response data for the current iteration. Finally, the design parameters for the current iteration are mapped to the coupling matrix coefficients for the current iteration through the surrogate model.

[0008] When the loop termination condition is met, the execution of all loop processing procedures ends, and the design parameters obtained in the last loop processing procedure are returned.

[0009] Furthermore, the initialization of design parameters includes:

[0010] Generate initial value centers;

[0011] The initialization generation design parameters are obtained by using a value-selection algorithm to select values ​​around the initial value center.

[0012] Furthermore, the value selection algorithm is a star-shaped distribution value selection algorithm, a Latin hypercube value selection algorithm, or an equal probability value selection algorithm.

[0013] Furthermore, the automatic optimization design and debugging method for filters that integrates the coupling matrix and the surrogate model also includes:

[0014] The coupling matrix coefficients for this round are solved using the filter response data of this round, and the phase loading effect is eliminated from the filter response data of this round.

[0015] Furthermore, the process of eliminating phase loading effects on the design parameters for this round includes:

[0016] The filter response data for this round is processed using a genetic algorithm to obtain the phase quantity;

[0017] Eliminate the phase quantity present in the filter response data of this round.

[0018] Furthermore, the surrogate model is a Kriging model, a radial basis function model, a support vector machine model, or an artificial neural network model.

[0019] Furthermore, the design specifications include the filter's center frequency, bandwidth, passband return loss, passband insertion loss, and out-of-passband suppression characteristics.

[0020] Furthermore, the types of design parameters include device structure parameters, device material parameters, and device fabrication subsequent tuning parameters.

[0021] Furthermore, the intelligent optimization algorithm is a genetic algorithm, a differential evolution algorithm, a genetic algorithm based on non-dominated sorting, or a multi-objective evolutionary algorithm based on decomposition.

[0022] Furthermore, the loop termination condition is specifically as follows:

[0023] The filter response data obtained in the final round of processing meets the design specifications.

[0024] or

[0025] The cumulative number of rounds in the cyclic processing reaches the round number threshold.

[0026] The beneficial effects of this invention are as follows: The automatic optimization design and debugging method for filters that integrates coupling matrices and surrogate models in the embodiments achieves direct optimization based on design metrics by setting design indices, executing multiple loop processes, and updating design parameters, surrogate models, and coupling matrix coefficients in each loop process. This differs from the traditional filter optimization technique that specifies a fixed coupling matrix as the optimization target. The automatic optimization design and debugging method for filters that integrates coupling matrices and surrogate models in this embodiment can simultaneously search for design parameters corresponding to multiple sets of coupling matrices that meet the design indices. Therefore, the optimization target can be expanded to all coupling matrices that meet the design indices. Since each step can be implemented by computer algorithms, the degree of automation is high, which can be used to accelerate the design and debugging process of filters. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the automatic optimization design and debugging method for filters that integrates coupling matrices and surrogate models in this embodiment.

[0028] Figure 2 This is a schematic diagram of the automatic optimization design and debugging method for filters that integrates coupling matrix and surrogate model in the embodiment.

[0029] Figure 3 This is a schematic diagram of the structure and parameters of a microstrip line filter used in the automatic optimization design and debugging method for filters in the application embodiment.

[0030] Figure 4 , Figure 5 and Figure 6 For the example Figure 3The diagram illustrates the effect achieved by the automatic optimization design and debugging method for microstrip line filters in the application embodiment shown. Detailed Implementation

[0031] In this embodiment, refer to Figure 1 The automatic optimization design and debugging method for filters that integrates coupling matrices and surrogate models includes the following steps:

[0032] S1. Set the design specifications for the filter;

[0033] Execute the loop process; specifically, the number of iterations of the loop process is represented by a positive integer variable i.

[0034] For the i-th round of loop processing, if i=1, that is, the loop processing process being executed is the first round of loop processing, then in the i-th round of loop processing, design parameters are initialized and generated, and a proxy model is constructed based on the design parameters of this round; wherein, the proxy model is used to map the space where the design parameters are located to the space where the coupling matrix coefficients are located;

[0035] For the i-th round of loop processing, if i>1, meaning the loop processing being executed is not the first round of loop processing, then in this round of loop processing, the following steps S2-S6 are executed:

[0036] S2. Determine the set of loss function vectors for this round based on the coupling matrix coefficients and design metrics from the previous round;

[0037] S3. Perform non-dominated sorting on the loss function vector set to obtain the design parameter set corresponding to the non-dominated loss function vector set in this round, i.e., the non-dominated vector set;

[0038] S4. Cluster the set of non-dominated vectors to obtain at least one cluster center in this round;

[0039] S5. Using the cluster center as the search center, use an intelligent optimization algorithm to search for the search center on the surrogate model to obtain the design parameters for this round;

[0040] S6. Simulate the design parameters for this round to obtain the filter response data for this round, and map the design parameters for this round to the coupling matrix coefficients for this round through a surrogate model;

[0041] When the loop termination condition is met, execute step S7: end the execution of all loop processing procedures and return the design parameters obtained in the last loop processing procedure.

[0042] In this embodiment, each step in the automatic optimization design and debugging method of the filter that integrates the coupling matrix and the surrogate model, including step S1 and steps S2-S6 in each round of loop processing, can be executed by a computer.

[0043] In this embodiment, the principle of steps S1-S7 is as follows: Figure 2 As shown.

[0044] In step S1, the design specifications for the filter can specifically include parameters such as the filter center frequency f0, bandwidth BW, in-passband return loss RL, in-passband insertion loss IL, and out-of-passband rejection characteristics. These design specifications can be in vector form, consisting of multiple values, such as all parameters including the filter center frequency f0, bandwidth BW, in-passband return loss RL, in-passband insertion loss IL, and out-of-passband rejection characteristics, or only some of these parameters.

[0045] In step S1, the objective for each round of the cyclic processing can be defined by the following formula:

[0046]

[0047] In the above formula, U(x) represents the loss function, which can be further expanded into U k The sum of all sub-functions such as (x,f). For example, taking the design specifications set in step S1 as the S-parameter curve in the filter response, the center frequency and bandwidth specifications can be converted into frequency-limited f. min and f max In the above formula, f represents the frequency, x represents the design parameters, and the return loss can be converted to... Insertion loss and out-of-band rejection can be converted into The summation term with index m in the above formula is about S 11 The design specifications, with n as the subscript, are summation terms about S. 21 The design specifications, specifically the S-parameters, are typically in the form of magnitudes. When the design specifications are based on equality constraints, such as... The function F can be either L1 norm or L2 norm; when the design index is an upper limit constraint, such as... Function F can be a maximum value function; when the design index is a lower limit constraint, such as The function F can be taken as the minimum value function.

[0048] After completing step S1 and setting the filter design parameters, the counter i is set to 1, and the first round of loop processing begins. During the first round of loop processing, the computer initializes and generates design parameters, and constructs a surrogate model based on these design parameters.

[0049] In the first round of processing, during the initialization of design parameters, the computer can first generate an initial value center, and then use a star-shaped distribution algorithm, a Latin hypercube algorithm, or an equal probability algorithm to perform star-shaped distribution, Latin hypercube, or equal probability value-taking around the initial value center, thereby obtaining the initialized design parameters. Specifically, the design parameters can be parameters of device structure, device material, and subsequent tuning parameters after device fabrication. For example, design parameters can be the positional parameters of components such as tuning screws or flybars in a filter.

[0050] In the first round of processing, and in subsequent rounds, the proxy model used has the following form:

[0051] M p =p(X)

[0052] In the formula, the row vectors of matrix X correspond to a set of design parameters x, and matrix M... p The row vectors of matrix X are the coefficients of the coupling matrix corresponding to the row vectors of matrix X, and p is the mapping function of the surrogate model. Therefore, the surrogate model can map the space of design parameters to the space of coupling matrix coefficients. Specifically, the surrogate model can be a kriging model, a radial basis function model, a support vector machine model, or an artificial neural network model, and p is the corresponding function or algorithm. For example, when the surrogate model is a kriging model, p is the kriging algorithm.

[0053] After the first round of loop processing is completed, the next round of loop processing, namely the second round, can be executed. During the second round of loop processing, the computer executes steps S2-S6. Since the principles of the second and subsequent rounds of loop processing are similar, describing the second round is equivalent to describing the subsequent rounds. In this embodiment, unless otherwise specified, the description refers to the second round of loop processing.

[0054] In step S2, the set of loss function vectors for this round is determined based on the coupling matrix coefficients of the previous round (that is, the coupling matrix coefficients obtained in the (i-1)th round of iterative processing) and the design index set in step S1.

[0055] Specifically, based on the objectives defined in step S1, the mean value of the loss function corresponding to each objective of the coupling matrix coefficients in the previous round can be calculated at the frequency points. Where k = 1, 2, ..., m+n, and m and n are S0 and S1, respectively. 11 and S 21 The number of targets. The row vectors in the coupling matrix coefficients correspond to each... The loss function vectors are formed, and the loss function vectors corresponding to each row vector form a set of loss function vectors.

[0056] In step S3, a non-dominated sort is performed. The principle of the non-dominated sort is as follows: Taking a minimization problem as an example, for two vectors x... 1 and x 2 If and only if x 2 All elements are greater than or equal to x 1 The corresponding element, and x 2 There exists at least one element that is strictly greater than x. 1 When the corresponding element is x, then x 1 <x 2 That is, x 1 Dominate x 2 The relationship holds. If in the vector set X, there is no vector that dominates x. 1 If x is a vector, then x is called x 1 It is a non-dominated vector.

[0057] Based on the above definition, specifically in step S3, the loss function vectors obtained in step S2 are subjected to non-dominated sorting to obtain the design parameter set corresponding to the set of non-dominated loss function vectors, i.e., the non-dominated vector set X. np Its corresponding sorting optimization level is the highest.

[0058] In step S4, the set of non-dominated vectors X obtained in step S3 is... np Clustering can be performed, specifically using the K-means algorithm to cluster the set of non-dominated vectors X. np The clusters are divided into K classes, resulting in K cluster centers.

[0059] In step S5, the K cluster centers obtained in step S4 are used as search centers. An intelligent optimization algorithm is then used to search for these search centers on the surrogate model to obtain the design parameters for this round. Specifically, the intelligent optimization algorithm used can be a genetic algorithm (GA), a differential evolution algorithm (DE), a non-dominated sorting-based genetic algorithm (NSGA-II), or a decomposition-based multi-objective evolutionary algorithm (MOEA / D), etc.

[0060] The principle of steps S2-S5 is as follows: In step S2, for the filter response curves corresponding to different design parameters x, the loss function is calculated according to the objective defined in step S1, and the non-dominated sorting in step S3 is performed. Finally, each search center has multiple non-dominated solutions x. In step S4, multiple search centers x are given. In step S5, the intelligent optimization algorithm searches within its respective search range under the given multiple search centers x, and selects the design parameter with the smallest sum of loss functions, thereby completing the search for design parameters.

[0061] In step S6, the simulation software is started to simulate the design parameters selected in step S5 for this round, and the filter response data for this round is obtained. The filter response data can specifically be in the form of S-parameters. Multiple parallel simulations can be used to speed up the process during step S6.

[0062] In step S6, the design parameters for this round are mapped to the coupling matrix coefficients for this round using a surrogate model. According to the surrogate model, different design parameters x can be mapped to corresponding coupling matrix coefficients, forming the coupling matrix M. Since the filter response data in step S6 is obtained by simulating the design parameters, there is a correspondence between the filter response data (specifically, the S-parameters) and the design parameters. In step S6, the corresponding filter response curve is predicted using the following conversion relationship between the coupling matrix and the S-parameters:

[0063] A=-jR+ω′U+M

[0064]

[0065] S 11 =1+2j[A -1 ] 1,1

[0066] S 21 =-2j[A -1 ] N+2,1

[0067] In the formula, N is the order of the filter, and R is an (N+2)×(N+2) matrix whose elements, except for R... 1,1 =R N+2,N+2 All values ​​except 1 are 0, ω′ is the normalized frequency, and U is an identity matrix of type (N+2)×(N+2), except for U 1,1 =U N+2,N+2 =0, M is a coupling matrix of (N+2)×(N+2).

[0068] In this embodiment, the coupling matrix coefficients for the current round are calculated using the filter response data, and the phase loading effect is eliminated from the filter response data. Specifically, a genetic algorithm (GA) is used to process the filter response data for the current round to obtain the phase quantity. Taking the S-parameters as an example, the phase quantity present in the filter response data that needs to be eliminated can be expressed as: ψ and θ can be solved using a genetic algorithm. The solution operation for the above phase quantity only needs to be performed once. In subsequent iterations, when extracting the coefficients of the coupling matrix, the phase quantity Δφ can be reused, thus saving time.

[0069] In step S6, the coupling matrix coefficients for this round can be obtained using methods such as Cauchy extraction and vector fitting. Taking vector fitting as an example, it is necessary to first extract the coefficients of the coupling matrix S... ij Parameters converted to Y ij Parameters. Randomly generate a set of vectors a = [a1 a2 … a N ] T Substitute into formula (1) multiple times to update vector a:

[0070]

[0071] In the formula, the k-th row of A1 and A2 are represented as follows:

[0072] (A1) k =[(A2) k 1]∈C 1×(N+1)

[0073]

[0074] ω′ k The normalized frequency of the k-th sampling point; the meanings of the other parameters are as follows:

[0075]

[0076]

[0077]

[0078]

[0079] In the above iteration process, the intermediate variables are determined. Then, vector a can be updated using the vector formed by the eigenvalues ​​of matrix H in formula (2). In the formula, b is an N×1 all-one vector.

[0080]

[0081] The iteration process stops when it converges, i.e., when vector a remains essentially unchanged. The variables a and r obtained from the above process... ij The topology of the extracted coupling matrix corresponds to the form of a transverse resonator circuit, namely:

[0082]

[0083] In the formula, the coefficients of each coupling matrix are calculated as follows:

[0084] M k,k =ja k k = 1, ..., N

[0085]

[0086]

[0087] M S,L =-jY 21 (s)| s＝∞ (If and only if N = number of transmission zeros N) TZ )

[0088] Traditional filter optimization schemes typically require the coupling matrix to undergo multiple similarity transformations to eliminate unimportant stray coupling terms, yielding the coupling matrix corresponding to the physical topology. Based on the linear nature of similarity transformations, the automatic filter optimization design and debugging method in this embodiment, which integrates the coupling matrix and the surrogate model, only requires the coupling matrix corresponding to the lateral resonator circuit model. This matrix already contains all the information of the coupling matrix corresponding to the physical topology. Therefore, this invention can be used for filter topology optimization, achieving optimization for different topologies, or for optimizing the performance of filter devices under unknown physical topologies.

[0089] After executing steps S2-S6 in the i-th round of loop processing, check whether the loop termination condition is met. In this embodiment, the loop termination condition can be either "the filter response data obtained in the last round of loop processing meets the design specifications" or "the cumulative number of rounds in the loop processing reaches the round threshold". That is, as long as it is detected that "the filter response data obtained in the last round of loop processing meets the design specifications" or "the cumulative number of rounds in the loop processing reaches the round threshold", the next round of loop processing will not be executed. However, if it is detected that "the filter response data obtained in the last round of loop processing does not meet the design specifications" and "the cumulative number of rounds in the loop processing has not reached the round threshold", then the counter is incremented by i = i + 1, and the next round of loop processing is executed.

[0090] Specifically, when verifying the loop termination condition that "the filter response data obtained in the last loop processing process meets the design specifications," the following method can be used: Obtain the filter response data obtained in the last loop processing process, and observe the S in the filter response data. 11 (Its meaning is the input reflection coefficient, which is also the input return loss), S 11 In the design specifications For comparison, when the S in the filter response data 11 Less than If the filter response data obtained in the last round of processing meets the design specifications, it is determined that the filter response data obtained in the last round of processing does not meet the design specifications.

[0091] Specifically, when verifying the loop termination condition "the cumulative number of loop processing cycles reaches the number of cycles threshold", the following method can be used: read the value of i in the counter, compare i with the preset number of cycles threshold i0, and when i reaches the number of cycles threshold i0, the next round (the i+1th round) of loop processing will not be executed.

[0092] In actual implementation, you can refer to Figure 1 First, verify that the filter response data obtained in the last round of loop processing meets the design specifications in the loop termination condition. If it does, proceed to step S7. If it does not, then verify that the cumulative number of loop processing rounds reaches the round number threshold in the loop termination condition. If it does, proceed to step S7. If it does not, proceed to the next round (the i+1th round) of loop processing.

[0093] In step S7, the design parameters obtained from step S5 of the last iteration are returned. Specifically, the returned design parameters can be stored in memory, printed out, or displayed on a monitor. Technicians can design filters based on the design parameters returned in step S7. Since the design parameters returned in step S7 are simulation-optimized, technicians do not need to perform further optimization on these parameters before designing the filter, thus saving them manual optimization work and improving filter design and debugging efficiency.

[0094] The automatic filter optimization design and debugging method integrating coupling matrix and surrogate model in this embodiment achieves direct optimization based on design indicators by setting design parameters, executing multiple iterative processes, and updating design parameters, surrogate model, and coupling matrix coefficients in each iteration. This differs from traditional filter optimization techniques that specify a fixed coupling matrix as the optimization target. The automatic filter optimization design and debugging method integrating coupling matrix and surrogate model in this embodiment can simultaneously search for design parameters corresponding to multiple sets of coupling matrices that meet the design indicators, thus expanding the optimization target to all coupling matrices that meet the design indicators. Each step in the automatic filter optimization design and debugging method integrating coupling matrix and surrogate model in this embodiment can be implemented by computer algorithms, resulting in a high degree of automation and accelerating the filter design and debugging process.

[0095] In this embodiment, it is specifically applied to the optimization design of microstrip line filters. The structure and parameter definitions of the microstrip line filter are as follows: Figure 3 As shown. The dielectric substrate parameters of the filter are: relative permittivity ε r=10.2, thickness h = 1.27mm. The fixed parameters of the microstrip line are: W0 = 1.18mm, L0 = 3mm, W1 = W2 = 0.4mm, and the optimized parameters are x = [L1, L2, S1, S2]. In this embodiment, the design index can be defined as:

[0096] f0 = 2GHz, BW = 400MHz

[0097] F1:|S 11 |≤-10dB,1.8GHz≤f≤2.2GHz

[0098] F2:|S 21 |≥-2dB, 1.8GHz≤f≤2.2GHz

[0099] F3:|S 21 |≤-20dB,1GHz≤f≤1.5GHz

[0100] F4:|S 21 |≤-20dB,2.5GHz≤f≤3GHz

[0101] With initial value x 0 Simulation was performed using [14.7, 15.3, 0.62, 0.5], and the results are as follows: Figure 4 As shown, the filter's passband characteristics are poor, the three modes deviate significantly from their design specifications, and the center frequency has shifted relative to the design parameters.

[0102] Based on the initial simulation results, the phase quantity extracted using the genetic algorithm is:

[0103] ΔΦ=[ψ 11 ,ψ 21 ,ψ 22 ,θ 11 ,θ 21 ,θ 22 ]

[0104] =[1.186,-1.448,1.48,0.2574,-0.3124,-0.1287]rad

[0105] After eliminating the phase quantity from the S-parameters and extracting the coupling matrix, the calculated S-parameters are as follows: Figure 4 As shown, the curves of the two graphs are highly consistent, proving the accuracy of the extraction results.

[0106] The initialization uses a star-shaped distribution method, where each variable is changed sequentially by ±0.2 mm. Including the initial value, a total of 4*2+1=9 simulations are performed. The search range of the subsequent intelligent optimization algorithm is limited to ±0.2 mm for each variable at each search center. After 5 iterations of the aforementioned process, two non-dominated vectors that meet the design specifications can be found. The simulation results are as follows Figure 5 As shown, I and II represent respectively It can be seen that the simulation curves corresponding to the two vectors both meet the design specifications, verifying the effectiveness of the proposed solution.

[0107] Furthermore, the convergence curve of the objective-loss function during the optimization process is as follows: Figure 6 As shown, the loss functions for the same generation are plotted using the same line type and sign. This is because the mean of each loss function at the frequency point... To minimize the quantity, the vertical axis uses a logarithmic coordinate system, and the transformation relationship is as follows: To avoid infinite terms, the minimum value of the loss function is set to 1e-5. It can be seen that after iteration, the values ​​of each loss function decrease rapidly, and eventually the loss function values ​​of all objectives are reduced to 1e-5 or below, completing the optimization and further verifying the efficiency and practicality of the proposed solution.

[0108] It should be noted that, unless otherwise specified, when a feature is referred to as "fixed" or "connected" to another feature, it can be directly fixed or connected to the other feature, or indirectly fixed or connected to the other feature. Furthermore, the descriptions of "upper," "lower," "left," and "right" used in this disclosure are only relative to the relative positional relationships of the various components of this disclosure in the accompanying drawings. The singular forms "a," "described," and "the" used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. Moreover, unless otherwise defined, all technical and scientific terms used in this embodiment have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this embodiment specification is only for describing particular embodiments and is not intended to limit the invention. The term "and / or" as used in this embodiment includes any combination of one or more of the associated listed items.

[0109] It should be understood that although the terms first, second, third, etc., may be used to describe various elements in this disclosure, these elements should not be limited to these terms. These terms are only used to distinguish elements of the same type from each other. For example, a first element may also be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may also be referred to as a first element. The use of any and all instances or exemplary language (“e.g.,” “such as,” etc.) provided in this embodiment is intended only to better illustrate embodiments of the invention and, unless otherwise required, does not impose a limitation on the scope of the invention.

[0110] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0111] Furthermore, the procedures described in this embodiment can be performed in any suitable order unless otherwise indicated by this embodiment or clearly contradicted by the context. The procedures (or variations and / or combinations thereof) described in this embodiment can be executed under the control of one or more computer systems configured with executable instructions, and can be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program includes a plurality of instructions executable by one or more processors.

[0112] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The invention described in this embodiment includes these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques described in the invention, the invention also includes the computer itself.

[0113] A computer program can be applied to input data to perform the functions described in this embodiment, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices, such as a display. In a preferred embodiment of the invention, the transformed data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on the display.

[0114] The above description is merely a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention, as long as they achieve the technical effects of the present invention by the same means, should be included within the scope of protection of the present invention. Within the scope of protection of the present invention, the technical solutions and / or implementation methods can have various modifications and variations.

Claims

1. A method for automatic optimization design and debugging of filters that integrates coupling matrix and surrogate model, characterized in that, The automatic optimization design and debugging method for filters that integrate the coupling matrix and the surrogate model includes: Set the design specifications for the filter; The process involves executing a loop processing procedure. When the loop processing procedure is the first round, design parameters are initialized and generated. A surrogate model is constructed based on the design parameters of this round. The surrogate model maps the space containing the design parameters to the space containing the coupling matrix coefficients. When the loop processing procedure is not the first round, the loss function vector set for this round is determined based on the coupling matrix coefficients of the previous round and the design index. The loss function vector set is then sorted non-dominated to obtain the non-dominated vector set for this round. The non-dominated vector set is the design parameter set corresponding to the non-dominated loss function vector set. The non-dominated vector set is clustered to obtain at least one cluster center for this round. Using the cluster center as the search center, an intelligent optimization algorithm is used to search for the search center on the surrogate model to obtain the design parameters for this round. The design parameters for this round are simulated to obtain the filter response data for this round. The design parameters for this round are then mapped to the coupling matrix coefficients for this round through the surrogate model. When the loop termination condition is met, the execution of all loop processing procedures ends, and the design parameters obtained in the last loop processing procedure are returned.

2. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model as described in claim 1, characterized in that, The initialization of design parameters includes: Generate initial value centers; The initialization generation design parameters are obtained by using a value-selection algorithm to select values ​​around the initial value center.

3. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 2, characterized in that, The value selection algorithm is either a star distribution value selection algorithm, a Latin hypercube value selection algorithm, or an equal probability value selection algorithm.

4. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model as described in claim 1, characterized in that, The automatic optimization design and debugging method for filters that integrates the coupling matrix and the surrogate model also includes: The coupling matrix coefficients for this round are solved using the filter response data of this round, and the phase loading effect is eliminated from the filter response data of this round.

5. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 4, characterized in that, The process of eliminating phase loading effects on the filter response data of this round includes: The filter response data for this round is processed using a genetic algorithm to obtain the phase quantity; Eliminate the phase quantity present in the filter response data of this round.

6. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 1, characterized in that, The surrogate model is a Kriging model, a support vector machine model, or an artificial neural network model.

7. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 1, characterized in that, The design specifications include the filter's center frequency, bandwidth, passband return loss, passband insertion loss, and out-of-passband suppression characteristics.

8. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 1, characterized in that, The types of design parameters include device structure parameters, device material parameters, and device fabrication post-processing tuning parameters.

9. The automatic optimization design and debugging method for filters based on the fusion of coupling matrix and surrogate model according to claim 1, characterized in that, The intelligent optimization algorithm is a genetic algorithm, a differential evolution algorithm, or a decomposition-based multi-objective evolution algorithm.

10. The automatic optimization design and debugging method for filters based on the fusion coupling matrix and surrogate model according to any one of claims 1-9, characterized in that, The loop termination condition is specifically as follows: The filter response data obtained in the final round of processing meets the design specifications. or The cumulative number of rounds in the cyclic processing reaches the round number threshold.

Images