A neural network optimization method, system, device and medium for microwave filter
Optimizing the geometric variables of microwave filters through neural network optimization methods solves the problem of low efficiency in existing design methods and realizes efficient and accurate microwave filter design.
Patent Information
- Application Number
- CN202411096193.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2025-02-07
- Estimated Expiration
- 2044-08-12
AI Technical Summary
The existing microwave filter design method is inefficient and difficult to quickly converge and optimize parameters, resulting in slow and inefficient design process.
Using neural network optimization method, the sample data is generated using the simulation tool HFSS by defining the geometric variables of the microwave filter, and the training samples and test samples are divided using the orthogonal allocation sampling method. Then, the neural network model is constructed using the quasi-Newtonian algorithm for optimization, and iteratively optimized until the optimization termination condition is met.
It improves the optimization and design efficiency of the microwave filter model, has high model accuracy, reliable design principles, simple structure, and has a wide range of application prospects.
Smart Images
Figure CN118627572B_ABST
Abstract
Description
Background Art
[0002] At present, wireless communication is not applicable in environments such as underground pipeline corridors due to the existence of electromagnetic shielding. Optical fiber communication has strong anti-interference and can effectively solve the problem of electromagnetic shielding. Therefore, it is very suitable for use in electromagnetic shielding environments such as underground pipeline corridors. As an intelligent network system for optical fiber transmission, the intelligent optical fiber system first has the characteristics of traditional optical fiber. It uses optical fiber as the transmission medium and transmits data in the form of optical signals. Secondly, it can also be used as a sensor to detect various physical quantities. Thirdly, it can also analyze and process the data of the optical fiber sensor, monitor the operating status of the entire intelligent optical fiber system, and realize intelligent feedback and control. Microwave filters are important components of intelligent optical fiber systems. They are devices that filter microwave or radio frequency signals in the optical domain. They greatly reduce the complexity of baseband signal processing, reduce the cost of optical fiber communication, and improve the quality of optical fiber communication. Traditional microwave filter design optimization usually uses EM simulation software combined with global optimization algorithms. Repeated simulations are required to adjust the set parameter values. This process is slow, difficult to converge, and inefficient.
[0003] This is a shortcoming of the prior art. Therefore, in view of the above-mentioned defects in the prior art, it is very necessary to provide a neural network optimization method, system, equipment and medium for microwave filters. Summary of the invention
[0004] The above-mentioned intelligent optical fiber system in the prior art plays an important role in the electromagnetic shielding environment. As an important component of the intelligent optical fiber, the microwave filter can improve the performance of the intelligent optical fiber communication and reduce the cost. However, the existing microwave filter design method has the defect of low efficiency. The present invention provides a neural network optimization method, system, device and medium of the microwave filter to solve the above technical problems.
[0005] In a first aspect, the present invention provides a neural network optimization method for a microwave filter, comprising the following steps:
[0006] S1. Define the geometric variables of the microwave filter, design the microwave filter according to the defined geometric variables, input the designed geometric variable values of the microwave filter into the simulation tool HFSS, obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to the set ratio;
[0007] S2. Use the model tool NeuroModeler to build a neural network model and initialize the parameters for training and optimizing the neural network model;
[0008] S3. Use training samples and test samples to train and test the neural network model to obtain optimized microwave filter geometric variable values and model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination conditions are met based on the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable values when they are met.
[0009] Furthermore, the specific steps of step S1 are as follows:
[0010] S11. Define geometric variables of the microwave filter, wherein the geometric variables include the internal dimensions of the cavity of the microwave filter and the height of the resonant column;
[0011] S12. Collect the internal dimensions of the cavity of the designed microwave filter and the corresponding resonant column height to generate a microwave filter design value;
[0012] S13. Input the microwave filter design value into the simulation tool HFSS for parallel expansion to obtain a set number of sample data;
[0013] S14. Use orthogonal distribution sampling method to divide the data samples into training samples and test samples according to the set ratio.
[0014] Furthermore, the specific steps of step S2 are as follows:
[0015] S21. Use the model tool NeuroModeler to build a neural network model, define the input of the neural network model as the geometric variables of the microwave filter, define the output of the neural network model as the filtering performance parameters, and use the quasi-Newton algorithm to optimize the neural network model;
[0016] S23. Initialize the starting point and number of iterations of the neural network model training;
[0017] S24. Initialize the trust region radius for neural network model optimization.
[0018] Furthermore, the geometric variables of the microwave filter used as inputs to the neural network model are the internal dimensions of the cavity and the corresponding resonant column height;
[0019] The filter performance parameter that is the output of the neural network model is the input return loss parameter.
[0020] Furthermore, in step S23, a set of geometric variables of the microwave filter is set from the training samples as the inner dimensions of the cavity and the corresponding resonant column height as the starting point of the neural network model.
[0021] Furthermore, the specific steps of step S3 are as follows:
[0022] S31. Use the training samples to train the neural network model, and start the iteration to output the optimized microwave filter geometry variable values and the corresponding model filtering performance values according to the initial starting point during the training process;
[0023] S32. The optimized microwave filter geometric variable value is input into the simulation tool HFSS for simulation to obtain the filter performance simulation value;
[0024] S33. Record a set of microwave filter geometric variable values, model filtering performance values and filtering performance simulation values, and calculate the error between the model filtering performance values and the filtering performance simulation values as the model error;
[0025] S34. Determine whether the number of iterations is completed;
[0026] If yes, go to step S36;
[0027] If not, proceed to step S35;
[0028] S35. Perform the next iteration, select the geometric variable values of the microwave filter corresponding to the next set of training samples according to the trust region radius, input them into the neural network model, and output the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values, and return to step S32;
[0029] S36. Determine whether the optimization termination condition is met;
[0030] If yes, go to step S38;
[0031] If not, proceed to step S37;
[0032] S37. Update the microwave filter design value, update the number of iterations and the trust region radius, and return to step S13;
[0033] S38. Testing the neural network model using the test sample, inputting a set of geometric variable values of the microwave filter in the test sample into the neural network model, and obtaining the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values;
[0034] S39. Calculate the model error in accordance with step S32-step S33, and determine whether the model error is less than a set threshold;
[0035] If so, the test is determined to be passed, and a set of microwave filter cavity internal dimensions and resonant column heights with the best microwave filter performance values recorded are output as the optimal microwave filter model;
[0036] If not, the test is deemed to have failed.
[0037] Furthermore, the optimization termination condition in step S36 is that the model error of this iteration is greater than the model error of the previous iteration.
[0038] In a second aspect, the present invention provides a neural network optimization system for a microwave filter, comprising:
[0039] The sample generation module is used to define the geometric variables of the microwave filter, and design the microwave filter according to the defined geometric variables, and input the designed geometric variable values of the microwave filter into the simulation tool HFSS to obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to the set ratio;
[0040] The model building module is used to build a neural network model using the model tool NeuroModeler and initialize the parameters for training and optimizing the neural network model;
[0041] The model optimization module is used to train and test the neural network model using training samples and test samples to obtain optimized microwave filter geometric variable values and model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination condition is met according to the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable value when it is met.
[0042] In a third aspect, the present invention provides a device, comprising a processor and a memory;
[0043] The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the device executes the method described in the first aspect above.
[0044] In a fourth aspect, the present invention provides a storage medium,
[0045] The storage medium stores instructions, which, when executed on a computer, enable the computer to execute the method described in the first aspect.
[0046] The beneficial effects of the present invention are:
[0047] The neural network optimization method, system, device and medium of the microwave filter provided by the present invention improve the optimization and design efficiency of the microwave filter model through the neural network model, and the model has high accuracy.
[0048] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect.
[0049] It can be seen that compared with the prior art, the present invention has outstanding substantive features and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] 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 or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0051] Figure 1 It is a schematic flow chart of the neural network optimization method for the microwave filter of the present invention.
[0052] Figure 2 It is a schematic diagram of the neural network optimization system of the microwave filter of the present invention.
[0053] Figure 3 It is a schematic diagram of the cavity structure of the microwave filter of the present invention. DETAILED DESCRIPTION
[0054] The neural network optimization method of microwave filters can be applied to one or more devices, which are devices that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and whose hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.
[0055] The device can be any electronic product that can interact with the user, such as a personal computer, tablet computer, smart phone, personal digital assistant (PDA), interactive network television (IPTV), etc. The network where the device is located includes but is not limited to the Internet, wide area network, metropolitan area network, local area network, virtual private network (VPN), etc.
[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0057] See also Figure 1 The figure shows a neural network optimization method for a microwave filter in a specific implementation, comprising the following steps:
[0058] S1. Define the geometric variables of the microwave filter, design the microwave filter according to the defined geometric variables, input the designed geometric variable values of the microwave filter into the simulation tool HFSS, obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to the set ratio;
[0059] S2. Use the model tool NeuroModeler to build a neural network model and initialize the parameters for training and optimizing the neural network model;
[0060] S3. Use training samples and test samples to train and test the neural network model to obtain optimized microwave filter geometric variable values and model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination conditions are met based on the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable values when they are met.
[0061] In this embodiment, the specific steps of step S1 are as follows:
[0062] S11. Define geometric variables of the microwave filter, wherein the geometric variables include the internal dimensions of the cavity of the microwave filter and the height of the resonant column;
[0063] The cavity structure of microwave filter is as follows Figure 3 As shown, the height h1, height h2, and height h3 are the heights of the resonant columns in the coupling window, and the heights hc1 and hc2 are the heights of the square cutouts placed in the center of the resonant cavity;
[0064] S12. Collect the internal dimensions of the cavity of the designed microwave filter and the corresponding resonant column height to generate a microwave filter design value;
[0065] S13. Input the microwave filter design value into the simulation tool HFSS for parallel expansion to obtain a set number of sample data;
[0066] S14. Use orthogonal distribution sampling method to divide the data samples into training samples and test samples according to the set ratio;
[0067] The specific steps of step S2 are as follows:
[0068] S21. Use the model tool NeuroModeler to build a neural network model, define the input of the neural network model as the geometric variables of the microwave filter, define the output of the neural network model as the filtering performance parameters, and use the quasi-Newton algorithm to optimize the neural network model;
[0069] The specific implementation process of step S21 is as follows:
[0070] S211. Select the starting values of the neural network model parameters , and initialize a positive definite matrix ;
[0071] Corresponding to a set of initial microwave filter geometric variable values, including the internal dimensions of the cavity and the corresponding resonant column height, As an approximation of the Hesse matrix of the starting value, the identity matrix is taken at the beginning;
[0072] The internal dimensions of the cavity correspond to the height of the square cross section at the center of the resonant cavity of the microwave filter, and the height of the resonant column corresponds to the height of the tuning column in the coupling window of the microwave filter;
[0073] S212. Define the objective function of the neural network model as the error function between the filtering performance parameter and the expected performance parameter;
[0074] The filter performance parameter can be input return loss, and the error function can be mean square error MSE;
[0075] S213. Construct a quadratic model of the objective function of the neural network model:
[0076]
[0077] k is the number of iterations, is the current value, is an approximation of the Hesse matrix of the current value, The objective function is The value at The objective function is The gradient at, p is the search direction;
[0078] S214. Solve the extreme value of the quadratic model of the objective function of the neural network model and find the optimal solution :
[0079] ;
[0080] S215. With the optimal solution For the search direction, choose a step size , and obtain the new iteration value that satisfies the Wolfe condition:
[0081] Among them, the Wolfe condition ensures that the search direction is the descending direction, and the step length Not too small;
[0082] S216. Using iterative values and the current value The gradient information updates the Hesse matrix :
[0083] Depend on get:
[0084] S217. Detect iteration value Gradient Is it greater than the gradient threshold? ;
[0085] If yes, go to step S22;
[0086] If not, return to step S213;
[0087] S22. Initialize the starting point and the number of iterations of the neural network model training; set a set of geometric variables of the microwave filter as the internal dimensions of the cavity and the corresponding resonant column height from the training samples as the starting point of the neural network model;
[0088] S23. Initialize the trust region radius for optimizing the neural network model;
[0089] It should be noted that the geometric variables of the microwave filter used as the input of the neural network model are the internal dimensions of the cavity and the corresponding resonant column height;
[0090] The filter performance parameter as the output of the neural network model is the input return loss parameter;
[0091] The specific steps of step S3 are as follows:
[0092] S31. Use the training samples to train the neural network model, and start the iteration to output the optimized microwave filter geometry variable values and the corresponding model filtering performance values according to the initial starting point during the training process;
[0093] S32. The optimized microwave filter geometric variable value is input into the simulation tool HFSS for simulation to obtain the filter performance simulation value;
[0094] S33. Record a set of microwave filter geometric variable values, model filtering performance values and filtering performance simulation values, and calculate the error between the model filtering performance values and the filtering performance simulation values as the model error;
[0095] S34. Determine whether the number of iterations is completed;
[0096] If yes, go to step S36;
[0097] If not, proceed to step S35;
[0098] S35. Perform the next iteration, select the geometric variable values of the microwave filter corresponding to the next set of training samples according to the trust region radius, input them into the neural network model, and output the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values, and return to step S32;
[0099] S36. Determine whether the optimization termination condition is met; the optimization termination condition is that the model error of this iteration is greater than the model error of the previous iteration;
[0100] If yes, go to step S38;
[0101] If not, proceed to step S37;
[0102] S37. Update the microwave filter design value, update the number of iterations and the trust region radius, and return to step S13;
[0103] S38. Testing the neural network model using the test sample, inputting a set of geometric variable values of the microwave filter in the test sample into the neural network model, and obtaining the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values;
[0104] S39. Calculate the model error in accordance with step S32-step S33, and determine whether the model error is less than a set threshold;
[0105] If so, the test is determined to be passed, and a set of microwave filter cavity internal dimensions and resonant column heights with the best microwave filter performance values recorded are output as the optimal microwave filter model;
[0106] If not, the test is deemed to have failed.
[0107] It should be understood that the order of execution of the steps in the above embodiment does not mean that the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.
[0108] The following is an embodiment of a neural network optimization system for a microwave filter provided by an embodiment of the present disclosure. The system and the neural network optimization method for a microwave filter in the above-mentioned embodiments belong to the same inventive concept. For details not described in detail in the embodiment of the neural network optimization system for a microwave filter, reference can be made to the embodiment of the neural network optimization method for a microwave filter in the above-mentioned embodiments.
[0109] like Figure 2 As shown, the system includes:
[0110] The sample generation module is used to define the geometric variables of the microwave filter, and design the microwave filter according to the defined geometric variables, and input the designed geometric variable values of the microwave filter into the simulation tool HFSS to obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to the set ratio;
[0111] The model building module is used to build a neural network model using the model tool NeuroModeler and initialize the parameters for training and optimizing the neural network model;
[0112] The model optimization module is used to train and test the neural network model using training samples and test samples to obtain optimized microwave filter geometric variable values and model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination condition is met according to the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable value when it is met.
[0113] The neural network optimization method for microwave filters provided by the present invention is a combination of the modules and algorithm steps of each example described in the embodiments disclosed herein, and can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0114] Those skilled in the art will appreciate that various aspects of the neural network optimization method for microwave filters provided by the present invention can be implemented as a system, method or program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, which can be collectively referred to as a "circuit", "module" or "system" here.
[0115] In the non-transitory computer-readable storage medium of the neural network optimization method for microwave filters, it can be, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
[0116] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.
Claims
1. A neural network optimization method for microwave filters, characterized in that: The steps include: S1. Define the geometric variables of the microwave filter, design the microwave filter according to the defined geometric variables, input the designed geometric variable values of the microwave filter into the simulation tool HFSS, obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to a set ratio; Step S1 includes the following steps: S11. Define geometric variables of the microwave filter, wherein the geometric variables include the internal dimensions of the cavity of the microwave filter and the height of the resonant column; S13. Input the microwave filter design value into the simulation tool HFSS for parallel expansion to obtain a set number of sample data; S14. Use orthogonal distribution sampling method to divide the data samples into training samples and test samples according to the set ratio; S2. Use the model tool NeuroModeler to build a neural network model and initialize the parameters for training and optimizing the neural network model; step S2 includes the following steps: S21. Use the model tool NeuroModeler to build a neural network model, define the input of the neural network model as the geometric variables of the microwave filter, define the output of the neural network model as the filtering performance parameters, and use the quasi-Newton algorithm to optimize the neural network model; the specific implementation process of step S21 is as follows: S211. Select the starting values of the neural network model parameters , and initialize a positive definite matrix ; Corresponding to a set of initial microwave filter geometric variable values, including the internal dimensions of the cavity and the corresponding resonant column height, As an approximation of the Hesse matrix of the starting value, the identity matrix is taken at the beginning; The internal dimensions of the cavity correspond to the height of the square cross section at the center of the resonant cavity of the microwave filter, and the height of the resonant column corresponds to the height of the tuning column in the coupling window of the microwave filter; S212. Define the objective function of the neural network model as the error function between the filtering performance parameter and the expected performance parameter; The filter performance parameter can be input return loss, and the error function can be mean square error MSE; S3. Use the training samples and the test samples to train and test the neural network model to obtain the optimized microwave filter geometric variable values and the model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination condition is met according to the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable value when it is met; the specific steps of step S3 are as follows: S31. Use the training samples to train the neural network model, and start the iteration to output the optimized microwave filter geometry variable values and the corresponding model filtering performance values according to the initial starting point during the training process; S32. The optimized microwave filter geometric variable value is input into the simulation tool HFSS for simulation to obtain the filter performance simulation value; S33. Record a set of microwave filter geometric variable values, model filtering performance values and filtering performance simulation values, and calculate the error between the model filtering performance values and the filtering performance simulation values as the model error; S34. Determine whether the number of iterations is completed; If yes, go to step S36; If not, proceed to step S35; S35. Perform the next iteration, select the geometric variable values of the microwave filter corresponding to the next set of training samples according to the trust region radius, input them into the neural network model, and output the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values, and return to step S32; S36. Determine whether the optimization termination condition is met; If yes, go to step S38; If not, proceed to step S37; S37. Update the microwave filter design value, update the number of iterations and the trust region radius, and return to step S13; S38. Testing the neural network model using the test sample, inputting a set of geometric variable values of the microwave filter in the test sample into the neural network model, and obtaining the optimized geometric variable values of the microwave filter and the corresponding model filtering performance values; The specific optimization during the training process is as follows: S213. Construct a quadratic model of the objective function of the neural network model: k is the number of iterations, is the current value, is an approximation of the Hesse matrix of the current value, The objective function is The value at The objective function is The gradient at, p is the search direction; S214. Solve the extreme value of the quadratic model of the objective function of the neural network model and find the optimal solution : ； S215. With the optimal solution For the search direction, choose a step size , and obtain the new iteration value that satisfies the Wolfe condition: Among them, the Wolfe condition ensures that the search direction is the descending direction, and the step length Not too small; S216. Using iterative values and the current value The gradient information updates the Hesse matrix : Depend on get: S217. Detect iteration value Gradient Is it greater than the gradient threshold? If yes, initialize the starting point and number of iterations of neural network model training; If not, return to step S213.
2. The neural network optimization method for a microwave filter according to claim 1, wherein: Step S1 also includes the following steps: S12. Collect the internal dimensions of the cavity of the designed microwave filter and the corresponding resonant column height to generate the microwave filter design value.
3. The neural network optimization method for a microwave filter as claimed in claim 2, characterized in that: Step S2 also includes the following steps: S22. Initialize the starting point and number of iterations of the neural network model training; S23. Initialize the trust region radius for neural network model optimization.
4. The neural network optimization method for a microwave filter as claimed in claim 3, characterized in that: The geometric variables of the microwave filter used as inputs to the neural network model are the internal dimensions of the cavity and the corresponding resonant column height; The filter performance parameter that is the output of the neural network model is the input return loss parameter.
5. The neural network optimization method for microwave filter according to claim 3, characterized in that: In step S23, a set of geometric variables of the microwave filter is set from the training samples as the internal dimensions of the cavity and the corresponding resonant column height as the starting point of the neural network model.
6. The neural network optimization method for microwave filter according to claim 5, characterized in that: Step S3 also includes the following steps: S39. Calculate the model error in accordance with step S32-step S33, and determine whether the model error is less than a set threshold; If so, the test is determined to be passed, and a set of microwave filter cavity internal dimensions and resonant column heights with the best microwave filter performance values recorded are output as the optimal microwave filter model; If not, the test is deemed to have failed.
7. The neural network optimization method for a microwave filter according to claim 6, characterized in that: The optimization termination condition in step S36 is that the model error of this iteration is greater than the model error of the previous iteration.
8. A neural network optimization system for a microwave filter that implements the neural network optimization method for a microwave filter according to any one of claims 1 to 7, characterized in that: include: The sample generation module is used to define the geometric variables of the microwave filter, and design the microwave filter according to the defined geometric variables, and input the designed geometric variable values of the microwave filter into the simulation tool HFSS to obtain sample data, and use the orthogonal allocation sampling method to divide the sample data into training samples and test samples according to the set ratio; The model building module is used to build a neural network model using the model tool NeuroModeler and initialize the parameters for training and optimizing the neural network model; The model optimization module is used to train and test the neural network model using training samples and test samples to obtain optimized microwave filter geometric variable values and model filtering performance values, input the optimized microwave filter geometric variable values into the simulation tool HFSS to calculate the filtering performance simulation value, and judge whether the optimization termination condition is met according to the optimized model filtering performance value and the filtering performance simulation value, and output the optimal microwave filter geometric variable value when it is met.
9. A device, characterized in that: including a processor and a memory; The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the device executes the method described in any one of claims 1 to 7.
10. A storage medium, characterized in that: The storage medium stores instructions, which, when executed on a computer, enable the computer to execute the method described in any one of claims 1 to 7.