A radar absorbing structure optimization design method based on deep learning

Through deep learning-based methods, using electromagnetic performance calculation simulation models and databases, combined with neural networks and genetic algorithms, the perforated radar absorbing structure with high absorption bandwidth is optimized, which solves the design efficiency of radar absorbing structure in the existing technology The problem of low efficiency achieves efficient absorption effect.

CN114722666BActive Publication Date: 2024-11-01DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210315465.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-29
Publication Date
2024-11-01
Estimated Expiration
2042-03-29

AI Technical Summary

Technical Problem

There are a lot of repeated simulation calculations in the existing radar wave absorbing structure design methods, which leads to inefficiency and cannot effectively optimize the electromagnetic wave absorption effect of structural radar wave absorbing stealth materials.

Method used

Using a deep learning-based method, by establishing an electromagnetic performance calculation simulation model and database, combining a fully connected neural network and genetic algorithm, a perforated radar absorbing structure with high absorption frequency bandwidth and high absorption efficiency is optimized.

Benefits of technology

It realizes rapid and efficient design optimization of radar wave absorbing structure, improves absorption frequency bandwidth and absorption efficiency, simplifies the design process, and reduces calculation costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114722666B_ABST
    Figure CN114722666B_ABST
Patent Text Reader

Abstract

The present invention provides a radar absorbing structure optimization design method based on deep learning, which belongs to the field of electromagnetics. First, by using the interference destructive theory and the transmission line theory, through the electromagnetic field full-wave analysis software based on the finite element method, an electromagnetic performance calculation simulation model of the radar absorbing structure is established and an electromagnetic performance parameter database is further established; then, a fully connected neural network is established and fully trained to achieve rapid calculation of the electromagnetic performance of radar absorbing structures with different structural parameters; finally, an optimization design algorithm based on a genetic algorithm is established to obtain a perforated radar absorbing structure with a wide absorption bandwidth and high absorption efficiency. The radar absorbing structure optimization design method provided by the present invention can quickly and efficiently optimize the structural parameters and materials of the radar absorbing structure in response to the performance requirements of the radar absorbing structure, and optimize the design of a perforated radar absorbing structure with a wide absorption bandwidth and high absorption efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of electromagnetism and relates to a radar absorbing structure optimization design method based on deep learning. Background Art

[0002] Excellent radar stealth performance is an important feature of advanced weapons and equipment to improve penetration and survivability. Structural absorbing stealth materials have both absorbing performance and load-bearing capacity. They are composed of weapon skin and radar absorbing structure. They are designable, have wide absorption bandwidth, and high absorption efficiency. They are an important development direction of stealth material research. Therefore, it is crucial to develop a fast and efficient radar absorbing structure design method.

[0003] Radar absorbing structures are composed of periodic multi-dimensional gradient micro-unit array structural units, and their structural units have high design freedom and optimization complexity. Traditional research mainly explores the electromagnetic wave absorption characteristics of radar absorbing structures through experimental tests, and uses approximate engineering analysis methods after repeated experiments to complete the design of structural units. The experimental cost is high and the efficiency is low. At present, the electromagnetic field full-wave analysis method based on the finite element method and the time-domain finite difference method can improve the design efficiency through computer simulation, but it still requires a lot of repeated calculations. In recent years, deep learning algorithms have been widely used in science and technology and life, which can greatly improve the efficiency of radar absorbing structure research and design, reduce tedious experimental exploration work, and shorten the design cycle.

[0004] In 2015, Wang Qun et al. disclosed in the invention patent CN201510185532.1 a multi-layer frequency selective surface composite absorbing structure and material optimization design method based on differential evolution algorithm. This method uses electromagnetic simulation software to directly feed back the calculation results to the main program, which facilitates iterative optimization. By quickly searching for the optimal model's structural and material characteristic parameter combination, the design cycle is shortened, and directional guidance is provided for the production of absorbing materials. However, this absorbing structure optimization design method still cannot avoid tedious computer simulation work. In 2019, Song Lin et al. disclosed in the invention patent CN201910501055.3 a design method for metamaterial absorbing structure based on neural network. The metamaterial structure is modeled using full-wave simulation software, and an optimization algorithm is constructed based on the neural network to optimize the absorbing structure parameters that meet the required goals. However, the design method of the absorbing structure is not universal. The overall electromagnetic wave absorption effect of the structure in the band range is not considered during optimization, and it is not suitable for structural radar absorbing stealth materials. Summary of the invention

[0005] The main technical problem solved by the present invention is to overcome the shortcomings of the existing methods. Aiming at the problem of a large amount of repeated simulation calculation work in the design of radar absorbing structures in structural absorbing stealth materials, a radar absorbing structure optimization design method based on deep learning is provided to achieve rapid and efficient design optimization of its structural parameters for the performance requirements of wider effective absorbing bandwidth and better absorption effect of the radar absorbing structure, and based on this method, a perforated radar absorbing structure with wide absorption bandwidth and high absorption efficiency is optimized and designed.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for optimizing the design of radar absorbing structures based on deep learning. First, using the interference destructive theory and transmission line theory, the electromagnetic performance calculation simulation model of the radar absorbing structure is established through the electromagnetic field full-wave analysis software based on the finite element method, and the electromagnetic performance parameter database is further established; then, a fully connected neural network is established and fully trained to achieve rapid calculation of the electromagnetic performance of radar absorbing structures with different structural parameters; finally, an optimization design algorithm based on a genetic algorithm is established to obtain a perforated radar absorbing structure with a wide absorption bandwidth and high absorption efficiency. The specific steps of the method are as follows:

[0008] The first step is to establish an electromagnetic performance calculation simulation model and an electromagnetic performance parameter database

[0009] Structural radar absorbing stealth materials are composed of a periodically arranged array of sub-wavelength structural units. The electromagnetic properties of the material are adjusted by designing the shape and size of the structural units. The current structure is a perforated radar absorbing structure, and its structural unit is obtained by adding a perforation with a circle diameter of d and a geometric center as the center to a square unit with a side length of h. The structural thickness is t, and the ratio of the perforation aperture d to the unit side length h is λ. The existing material is a silicon carbide composite material doped with iron powder.

[0010] Based on the theory of interference destructiveness and transmission line theory, the electromagnetic performance calculation simulation model of the radar absorbing structure is established using the electromagnetic field full-wave analysis software based on the finite element method. The electromagnetic performance calculation simulation model includes three parts: air domain, absorbing structure and bottom metal plate, where the metal plate represents the aircraft skin. The top boundary of the air domain is set as the scattering boundary, the bottom of the air domain is set as the only excitation source port, and the middle of the air domain is a perfect matching layer, ignoring its internal scattering effect. Part of the electromagnetic wave is reflected on the surface of the radar absorbing structure, and part of the electromagnetic wave enters the radar absorbing structure and is reflected at the bottom metal plate and is all collected by the excitation source port. The reflection loss RL parameter of the radar absorbing structure is calculated through software simulation, which represents the absorption effect of the radar absorbing structure on electromagnetic waves.

[0011] Step 1-1: First, in the electromagnetic field full-wave analysis software, a unit cell configuration of the absorbing structure of the periodic micro-unit array is established according to the geometric parameters of the perforated radar absorbing structure.

[0012] Step 1-2: Then, the basic electromagnetic parameters of the iron-doped silicon carbide composite material are imported and set as the material of the region where the unit cell configuration of the absorbing structure is located. The relevant electromagnetic parameters of air are imported and the air domain on the upper side of the absorbing structure is set as the air material.

[0013] Step 1-3: Then, set up the physical field of the simulation in the software, set the unit cell boundary as a periodic condition, select the incident plane at the bottom of the air domain as a periodic port, set it to vertical incidence of electromagnetic waves, set the air domain to avoid electromagnetic waves at the port at the top of the air domain as a scattering boundary, and set the air domain part above the incident port as a perfectly matched layer to avoid the scattering effect inside the air domain.

[0014] Step 1-4: Select frequency domain study, set the frequency domain range, and start simulation to obtain the reflection loss RL spectrum.

[0015] Step 1-5: The structural unit of the existing perforated radar absorbing structure is obtained by adding a perforation with an aperture d with its geometric center as the center of the circle on a square unit with a side length of h. The structural thickness is t, and the ratio of the perforation aperture d to the unit side length h is λ. Repeatedly bring different combinations of the side length h, aperture ratio λ, and structural thickness t within the value range into steps 1-1 to 1-4 to obtain multiple sets of structural parameters corresponding to the reflection loss RL spectrum of the radar absorbing structure, import the data into the database, and complete the construction of the electromagnetic performance parameter database.

[0016] The second step is to build a deep learning analysis model and conduct sufficient training

[0017] A fully connected neural network is established. The fully connected neural network is a typical structure of deep learning. By extracting the intrinsic laws and representation levels of learning sample data, it can realize the rapid calculation of the electromagnetic properties of radar absorbing structures with different structural parameters.

[0018] Step 2-1: Extract all structural parameters and their corresponding reflection loss RL from the database established in the first step, use the K-fold cross-validation method to efficiently extract data features, and divide the data into two sub-datasets D0 and D1. D0 is the test set, which is composed of M groups of data randomly selected from 8-12% of all data sets, where M is a positive integer that can make the number of remaining data sets divisible by K. Each group of data can be recorded as {[h,λ,t]; [RL]}, including the structural parameters of the radar absorbing structure and the corresponding reflection loss RL, which are used as the input data of the neural network and the predicted target control value respectively. The test set D0 is used to finally test the training results of the neural network. D1 consists of the remaining data, including the training set and the validation set, which are used to train the neural network and detect the training effect of the neural network in the process. The data set D1 is randomly divided into K equal parts, and the i1th part is selected as the validation set and recorded as The value of i1 is a positive integer from 0 to K, and the rest is the training set. K is set to 10 based on existing research conclusions.

[0019] Step 2-2: Establish the main structure of a fully connected neural network, including an input layer, several hidden layers and an output layer, each layer contains several nodes. Data is input into the neural network from the nodes in the input layer and output to the nodes in the next layer in turn. The number of nodes in the input and output layers, the number of hidden layers and nodes, and the expected mean square error threshold are determined based on the structural parameters and the number of reflection loss RL data in the data set.

[0020] Step 2-3: Use the training set to train the neural network and calculate the trained neural network on the validation set. The mean square error on

[0021]

[0022] Among them, m is the number of samples in each equal part, n is the frequency points of reflection loss RL of each sample, is the neural network prediction value of the k1th frequency point of the reflection loss of the radar absorbing structure, and the value of k1 is a positive integer from 1 to n. is the predicted target comparison value of the k1th frequency point of the reflection loss of the same structure, For the validation set The mean square error between the j1th structure RL prediction value and the predicted target control value, where j1 is a positive integer from 1 to m.

[0023] After repeating the training K times, calculate the mean square error E1 of this round of training.

[0024]

[0025] Step 2-4: Use the trained neural network to predict the data in the test set D0. If the mean square error E0 of the prediction result is less than the expected mean square error threshold, the training is completed. According to the previous test results, the expected mean square error threshold is 5×10 -5 .

[0026]

[0027] Where M is the number of samples in the test set, n is the frequency points of the reflection loss RL of each sample, is the neural network prediction value of the k2th frequency point of the reflection loss of the radar absorbing structure, and the value of k2 is a positive integer from 1 to n. is the predicted target comparison value of the k2th frequency point of the reflection loss of the same structure, is the mean square error between the j2th structure RL prediction value and the predicted target control value on the test set D0, and the value of j2 is a positive integer from 1 to M.

[0028] The third step is to establish an optimization design algorithm to optimize the parameters of the radar absorbing structure.

[0029] An optimization design algorithm based on genetic algorithm is established to optimize the structural parameters of the unit cell with the goal of wider effective bandwidth and better absorbing performance. Better individuals can be generated through crossover and mutation operations. The fitness function is used to judge the quality of individuals in the population, and bad individuals are eliminated through selection operations. As the number of repeated iterations of crossover, mutation, and selection operations increases, the population becomes better and eventually approaches the optimal solution.

[0030] Step 3-1: Define the fitness function and effectively establish a mathematical correlation model between the radar absorbing structure reflection loss, effective absorbing bandwidth and the quality of absorbing structure parameters. The specific expression is as follows:

[0031]

[0032] in, is the RL value calculated by the neural network for individual i2, and the value of i2 is a positive integer from 1 to N. is the effective absorbing bandwidth of the radar absorbing structure, is the average reflectivity over the effective absorbing bandwidth.

[0033] The optimization of structural parameters is achieved by finding the maximum value of the fitness function within the value range, that is, the maximum effective absorption bandwidth and the maximum absolute value of the average reflection loss within the effective absorption bandwidth range. Set the minimum value Min and maximum value Max of each structural parameter range, set the population size N, crossover probability P_cross, mutation probability P_variation, and maximum number of iterations I.

[0034] Step 3-2: Randomly generate an initial population of N individuals. Each individual consists of x floating-point numbers, representing a set of structural parameters. A structural parameter is also called a gene locus.

[0035] Step 3-3: Perform crossover, mutation, and selection operations to randomly group the population. Each group contains two parent individuals u and v. After performing the crossover operation according to formulas (5) and (6), the offspring individuals u′ and v′ are obtained:

[0036]

[0037]

[0038] in, is the value of the i3th gene locus of individual u, the value of i3 is a positive integer from 1 to x, and rand(0,1) represents a random number between 0 and 1. express and The union of .

[0039] Traverse all individuals w in the population and perform mutation operations according to formula (7) to obtain the offspring individual w′:

[0040]

[0041] in, are the minimum and maximum values ​​of the value range of the i4th gene locus respectively. The value of i4 is a positive integer from 1 to x.

[0042] Perform a selection operation, calculate the fitness of the offspring generated in the crossover and mutation operations according to the fitness function, mix the parent and offspring to form a new population, sort the individuals from large to small according to their fitness, eliminate individuals with serial numbers greater than N, and keep the total number of individuals in the population stable.

[0043] Step 3-4: Repeat step 3-3 until the number of iterations reaches the maximum number of iterations I. The individual with the largest fitness in the current population is the optimal solution. The gene points of the optimal solution are extracted to obtain the structural parameters of the optimal result.

[0044] The beneficial effects of the present invention are:

[0045] (1) The present invention provides a database of electromagnetic performance parameters of perforated structures containing a series of different structural parameters through a calculation simulation model of the electromagnetic performance of radar absorbing structures based on finite element electromagnetic field full-wave analysis software.

[0046] (2) The present invention provides a radar absorbing structure optimization design method based on deep learning, which can quickly and efficiently optimize the structural parameters and materials of the radar absorbing structure according to the performance requirements of the structure.

[0047] (3) According to this method, the present invention optimizes and designs a perforated radar absorbing structure with a wide absorption bandwidth and high absorption efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Flowchart for optimizing the design of radar absorbing structures.

[0049] Figure 2 The electromagnetic wave absorption performance curve of the structure optimized based on the deep learning analysis method in the range of 8-12.6GHz.

[0050] Figure 3 Schematic diagram of the perforated radar absorbing structure optimized and designed based on the deep learning analysis method. The structural unit is obtained by adding a perforation with an aperture d of 7.6 mm with its geometric center as the center of a square unit with a side length h of 9.74 mm. The ratio λ of the perforation aperture d to the unit side length h is 0.78, and the structural thickness t is 2.5 mm. DETAILED DESCRIPTION

[0051] The embodiments of the present invention are described in detail in combination with the accompanying drawings and technical solutions, and the optimization design method of radar absorbing structure based on deep learning is explained.

[0052] The first step is to establish an electromagnetic performance calculation simulation model and an electromagnetic performance parameter database

[0053] Establish an electromagnetic performance calculation simulation model. First, establish the unit cell configuration of the periodic micro-unit array according to the geometric parameters of the perforated radar absorbing structure in the electromagnetic wave software; then import the basic electromagnetic parameters of the silicon carbide composite material doped with iron powder and set it as the material in the area where the unit cell configuration of the radar absorbing structure is located. Import the relevant electromagnetic parameters of the air and set the air domain on the upper side of the radar absorbing structure as the air material; then set the simulation physical field in the software, set the unit cell boundary as the periodic condition, select the incident plane at the bottom of the air domain as the periodic port, set it as the vertical incident of the electromagnetic wave, set the air domain to avoid electromagnetic waves at the port at the top of the air domain as the scattering boundary, and set the air domain part above the incident port as the perfect matching layer to avoid the scattering effect inside the air domain; finally, select the frequency domain study, set the frequency domain range to 8-12.6GHz, the frequency point calculation step to 0.01GHz, and simulate to obtain the reflection loss RL spectrum.

[0054] The side length h of the perforated radar absorbing structure is taken in the range of 6-16 mm with a step size of 1 mm, the ratio λ of the perforated aperture d to the unit side length h is taken in the range of 0.1-0.8 with a step size of 0.05, and the structure thickness t is taken in the range of 1.5-2.5 mm with a step size of 0.1. Different combinations of the above structural parameters are repeatedly brought into the electromagnetic performance calculation simulation model, and the reflection loss RL spectra of the radar absorbing structure corresponding to 1815 structural parameter combinations are obtained. The data are imported into the database to complete the construction of the electromagnetic performance parameter database.

[0055] The second step is to build a deep learning analysis model and conduct sufficient training

[0056] All structural parameters and their corresponding reflection loss RL are extracted from the database established in the first step. The K-fold cross-validation method is used to efficiently extract data features. K is set to 10, and the data is divided into two sub-datasets, D0 and D1. D0 is the test set consisting of 185 randomly selected groups of data. Each group of data can be recorded as {[h, λ, t]; [RL]}, including the structural parameters of the radar absorbing structure and the corresponding reflection loss RL, which are used as the input data of the neural network and the predicted target control value respectively. The test set D0 is used to finally test the training results of the neural network. D1 consists of the remaining data, including the training set and the validation set, which are used to train the neural network and detect the training effect of the neural network in the process. The data set D1 is randomly divided into 10 equal parts, and the i1th part is selected as the validation set and recorded as D 1i , the value of i1 is a positive integer from 0 to K, and the rest is the training set.

[0057] The main structure of a fully connected neural network was established, which included an input layer with 3 nodes, four hidden layers with 50, 100, 200, and 400 nodes respectively, and an output layer with 461 nodes. The expected mean square error threshold was set to 5×10 -5 .

[0058] Use the training set to train the neural network, and use formula (1) to calculate the trained neural network's performance on the validation set. The mean square error on After repeating the training 10 times, use formula (2) to calculate the mean square error E1 of this round of training, use the trained neural network to predict the data in the test set D0, and use formula (3) to calculate the mean square error E0 of the prediction result. If it is less than the expected mean square error threshold, the training is completed.

[0059] The third step is to establish an optimization design algorithm to optimize the parameters of the radar absorbing structure.

[0060] Set the minimum value Min of the structural parameter [h, λ, t] to [6, 0.1, 1.5] and the maximum value Max to [16, 0.8, 2.5], and use the fitness function defined by formula (4). Set the population size N to 1000, the crossover probability P_cross to 0.6, the mutation probability P_variation to 0.1, and the maximum number of iterations I to 10000.

[0061] The optimization process of the genetic algorithm begins. An initial population of 1000 individuals is randomly generated. Each individual consists of 3 floating-point numbers, representing a set of structural parameters. A structural parameter is also called a gene locus. Then, crossover, mutation, and selection operations are performed. The population is randomly divided into groups. Each group contains two parent individuals u and v. After crossover operations are performed according to formulas (5) and (6), the offspring individuals u′ and v′ are obtained. All individuals w in the population are traversed, and mutation operations are performed according to formula (7) to obtain the offspring individual w′. Then, the selection operation is performed. The fitness of the offspring generated in the crossover and mutation operations is calculated according to the fitness function. The parents and offspring are mixed to form a new population. The individuals are sorted from large to small according to fitness. Individuals with sequence numbers greater than 1000 are deleted to keep the total number of individuals in the population stable. Finally, the optimization process of the genetic algorithm is repeated until the maximum number of iterations reaches 10000. The individual with the highest fitness in the current population is the optimal solution. The gene position of the optimal solution is extracted to obtain the structural parameters of the optimization result. The side length h of the structural unit is 9.74 mm, the center aperture d is 7.6 mm, that is, the ratio λ of the perforated aperture d to the unit side length h is 0.78, and the structural thickness t is 2.5 mm. The reflection loss of the structure in the 8-12.6 GHz band is as low as -18.6 dB, and the effective absorbing bandwidth reaches 3.64 GHz. A perforated radar absorbing structure with wide absorption bandwidth and high absorption efficiency in the 8-12.6 GHz band is obtained.

[0062] The present invention realizes rapid and efficient design optimization of radar absorbing structures, and based on the method optimizes and designs a perforated radar absorbing structure with a wide absorption bandwidth and high absorption efficiency, which is simple to operate and reliable.

[0063] The above-described embodiments merely express the implementation methods of the present invention, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A radar absorbing structure optimization design method based on deep learning, It is characterized in that Firstly, by using the interference destructive theory and transmission line theory, an electromagnetic performance calculation simulation model of the radar absorbing structure is established through the electromagnetic field full-wave analysis software based on the finite element method, and further an electromagnetic performance parameter database is established; then, a fully connected neural network is established and fully trained to realize the rapid calculation of the electromagnetic performance of radar absorbing structures with different structural parameters; finally, an optimization design algorithm based on the genetic algorithm is established to obtain a perforated radar absorbing structure with wide absorption bandwidth and high absorption efficiency; the following steps are included: The first step is to establish an electromagnetic performance calculation simulation model and an electromagnetic performance parameter database Based on the theory of destructive interference and transmission line theory, the electromagnetic performance calculation simulation model of the radar absorbing structure is established using the electromagnetic field full-wave analysis software based on the finite element method. The electromagnetic performance calculation simulation model includes three parts: air domain, absorbing structure and bottom metal plate, where the metal plate represents the aircraft skin; the top boundary of the air domain is set as the scattering boundary, the bottom of the air domain is set as the only excitation source port, and the middle of the air domain is a perfect matching layer, ignoring its internal scattering effect; part of the electromagnetic wave is reflected on the surface of the radar absorbing structure, and part of the electromagnetic wave enters the radar absorbing structure and is reflected at the bottom metal plate and is all collected by the excitation source port. The reflection loss RL parameter of the radar absorbing structure is calculated through software simulation, which represents the absorption effect of the radar absorbing structure on electromagnetic waves; Step 1-1: First, in the electromagnetic field full-wave analysis software, a unit cell configuration of the absorbing structure of the periodic micro-unit array is established according to the geometric parameters of the perforated radar absorbing structure; Step 1-2: Then, the basic electromagnetic parameters of the iron-doped silicon carbide composite material are imported and set as the material of the region where the unit cell configuration of the absorbing structure is located; the relevant electromagnetic parameters of the air are imported and the air domain on the upper side of the absorbing structure is set as the air material; Step 1-3: Next, set the simulation physical field in the software, set the unit cell boundary as a periodic condition, select the incident plane at the bottom of the air domain as a periodic port, set it to vertical incidence of electromagnetic waves, set the top of the air domain as a scattering boundary, and set the air domain part above the incident port as a perfect matching layer to avoid the scattering effect inside the air domain; Step 1-4: Select frequency domain study, set the frequency domain range, and start simulation to obtain the reflection loss RL spectrum; Step 1-5: The structural unit of the existing perforated radar absorbing structure is obtained by adding a perforation with an aperture d with its geometric center as the center of the circle on a square unit with a side length of h, the structural thickness is t, and the ratio of the perforation aperture d to the unit side length h is λ; different combinations of the side length h, the aperture ratio λ, and the structural thickness t within the value range are repeatedly brought into steps 1-1 to 1-4 to obtain multiple groups of structural parameters corresponding to the reflection loss RL spectrum lines of the radar absorbing structure, and the data are imported into the database to complete the construction of the electromagnetic performance parameter database; The second step is to build a deep learning analysis model and conduct sufficient training Establish a fully connected neural network, which is a typical structure of deep learning. By extracting the inherent laws and representation levels of learning sample data, it can realize the rapid calculation of the electromagnetic performance of radar absorbing structures with different structural parameters. Step 2-1: Extract all structural parameters and their corresponding reflection loss RL from the database established in the first step, use K-fold cross-validation method to efficiently extract data features, and divide the data into D 0 , D 1 Two sub-datasets, where D 0 The test set includes M groups of data. Each group of data includes the structural parameters of the radar absorbing structure and the corresponding reflection loss RL, which are used as the input data of the neural network and the predicted target control value respectively. The test set D 0 The training results for the final test of the neural network; D 1 The rest of the data consists of a training set and a validation set, which are used to train the neural network and detect the effect of the neural network training in the process respectively; Step 2-2: Establish the main structure of a fully connected neural network, including an input layer, several hidden layers and an output layer, each layer contains several nodes; the data is input into the neural network from the nodes in the input layer, and is output to the nodes in the next layer in turn. The number of nodes in the input layer and output layer, the number of hidden layers and nodes, and the expected mean square error threshold are determined based on the structural parameters and the number of reflection loss RL data in the data set; Step 2-3: Use the training set to train the neural network and calculate the trained neural network on the validation set. The mean square error on Among them, m is the number of samples in each equal part, n is the frequency points of reflection loss RL of each sample, is the kth reflection loss of the radar absorbing structure 1 The neural network prediction value of k frequency points, 1 The value of is a positive integer from 1 to n. is the kth reflection loss of the same structure 1 The predicted target control value for each frequency point, For the validation set Previous 1 The mean square error between the predicted value of the structure RL and the predicted target control value, j 1 The value of is a positive integer from 1 to m; After repeating the training K times, calculate the mean square error E of this round of training 1 , Step 2-4: Use the trained neural network to train the test set D 0 If the mean square error E of the prediction result is 0 If it is less than the expected mean square error threshold, the training is completed; Where M is the number of samples in the test set, n is the frequency points of the reflection loss RL of each sample, is the kth reflection loss of the radar absorbing structure 2 The neural network prediction value of k frequency points, 2 The value of is a positive integer from 1 to n. is the kth reflection loss of the same structure 2 The predicted target control value for each frequency point, For the test set D 0 Previous 2 The mean square error between the predicted value of the structure RL and the predicted target control value, j 2 The value of is a positive integer from 1 to M; The third step is to establish an optimization design algorithm to optimize the parameters of the radar absorbing structure. An optimization design algorithm based on genetic algorithm is established to optimize the structural parameters of the unit cell with the goal of wider effective bandwidth and better wave absorption performance. Through crossover and mutation operations, better individuals can be generated. The fitness function is used to judge the quality of individuals in the population, and bad individuals are eliminated through selection operations. With the increase of repeated iterations of crossover, mutation, and selection operations, the population becomes better and finally approaches the optimal solution. Step 3-1: Define the fitness function and establish a mathematical correlation model between the radar absorbing structure reflection loss, effective absorbing bandwidth and the quality of absorbing structure parameters; the specific expression is as follows: in, For individual i 2 The RL value calculated by the neural network, i 2 The value of is a positive integer from 1 to N. is the effective absorbing bandwidth of the radar absorbing structure, is the average reflectivity over the effective absorbing bandwidth; The optimization of structural parameters is achieved by finding the maximum value of the fitness function within the value range, that is, the maximum effective absorbing bandwidth and the maximum absolute value of the average reflection loss within the effective absorbing bandwidth range; the minimum value Min and the maximum value Max of each structural parameter range are set, and the population size N, the crossover probability P_cross, the mutation probability P_variation, and the maximum number of iterations I are set; Step 3-2: Randomly generate an initial population of N individuals, each of which consists of x floating-point numbers, representing a set of structural parameters. A structural parameter is also called a gene locus. Step 3-3: Perform crossover, mutation, and selection operations to randomly group the population. Each group contains two parent individuals u and v. After performing the crossover operation according to formulas (5) and (6), the offspring individuals u′ and v′ are obtained: in, For individual u i 3 The value of the gene point, i 3 The value of is a positive integer from 1 to x, rand(0,1) represents a random number between 0 and 1, express and The union of; Traverse all individuals w in the population and perform mutation operations according to formula (7) to obtain the offspring individual w′: in, The i 4 The minimum and maximum values ​​of the value range of each gene point, i 4 The value of is a positive integer from 1 to x; Perform selection operations, calculate the fitness of the offspring generated in the crossover and mutation operations according to the fitness function, mix the parent and offspring to form a new population, sort the individuals from large to small according to fitness, eliminate individuals with sequence numbers greater than N, and keep the total number of individuals in the population stable; Step 3-4: Repeat step 3-3 until the number of iterations reaches the maximum number of iterations I. The individual with the largest fitness in the current population is the optimal solution. The gene points of the optimal solution are extracted to obtain the structural parameters of the optimal result.

2. According to the method for optimizing the design of radar absorbing structures based on deep learning according to claim 1, It is characterized in that In step 2-1, the test set D 0 It consists of M groups of data that are randomly selected, accounting for 8-12% of all data sets, where M is a positive integer that can make the number of remaining data sets divisible by K. Each group of data can be recorded as {[h,λ,t]; [RL]}; Data set D 1 Randomly divide into K equal parts, select the i-th 1 The set is used as the validation set and is recorded as i 1 The value of is a positive integer from 0 to K, and the rest is the training set.

3. According to the radar absorbing structure optimization design method based on deep learning according to claim 1, It is characterized in that In steps 2-4, the expected mean square error threshold is 5×10 -5 .