Frequency selective surface inverse design method, device and equipment

Abstract

The invention provides a frequency selective surface inverse design method, device and equipment. The method comprises the steps of obtaining a target square wave; the target square wave is used for representing the expected transmission performance of the frequency selective surface; inputting the target square wave into a pre-trained multi-layer perceptron model to obtain a target design parameter corresponding to the frequency selection surface; the pre-trained multilayer perceptron model is obtained by training based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on a transmission line theory, and is calculated by adopting an equivalent inductive reactance and capacitive reactance empirical formula. According to the method, the equivalent circuit model based on the transmission line theory is used for replacing full-wave simulation, the complex electromagnetic field solving process is avoided, the sample data preparation time is remarkably shortened, due to the fact that the pre-trained multi-layer perceptron model is introduced, the target square waves are directly mapped to the corresponding design parameters, rapid and efficient reverse design is achieved, and the method is suitable for large-scale popularization and application. And the efficiency and the precision of the frequency selective surface design are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic inversion, and particularly to a method, device and equipment for inverse design of frequency selective surfaces. Background Art

[0002] In a wireless communication system, in order to avoid mutual interference between electromagnetic waves of different frequency bands, an electromagnetic structure with selective wave transmission characteristics is usually required. A frequency selective surface (FSS) is a periodic metal array attached to a dielectric substrate. By designing its topological structure, optimizing geometric parameters, and selecting appropriate dielectric materials, selective transmission of electromagnetic waves can be achieved, making it a typical spatial filter. With the rapid development of wireless communication technology, the performance requirements for FSS in engineering applications are continuously increasing, and efficient design also needs to be completed in a shorter time. However, traditional FSS design methods rely on full-wave simulation, which has a long calculation time and low iterative optimization efficiency, and it is difficult to meet the growing design requirements. Therefore, how to significantly improve the FSS design efficiency while ensuring high performance has become a technical problem to be solved urgently.

[0003] Currently, the prior art has proposed an FSS design method based on full-wave simulation. This method meets the target requirements by calculating the electromagnetic characteristics of each structure under different parameter configurations, and has high design accuracy. Full-wave simulation can accurately simulate the interaction between electromagnetic waves and the FSS structure, providing a reliable theoretical basis for design. In addition, in recent years, deep learning methods have been introduced into the FSS design field. By training a neural network to learn the relationship between structural parameters and electromagnetic characteristics, rapid inverse design from target performance to design parameters is realized, significantly improving the design efficiency.

[0004] However, the existing technical methods still have significant problems. First, the process of generating training data by full-wave simulation takes a long time and it is difficult to quickly meet the large-scale data requirements. Second, the unprocessed simulation data may lead to instability in the model training process, affecting the prediction accuracy and generalization ability of the neural network. In addition, the high dependence of deep learning methods on data quality and quantity limits their wide application in FSS inverse design. Therefore, the prior art still faces great challenges in achieving efficient and high-precision FSS design. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a method, device and equipment for inverse design of frequency selective surfaces.

[0006] The technical problems to be solved by the present invention are realized through the following technical solutions:

[0007] In a first aspect, the present invention provides a method for inverse design of a frequency selective surface, including:

[0008] Obtain a target square wave; the target square wave is used to characterize the desired transmission performance of the frequency selective surface;

[0009] Input the target square wave into a pre-trained multi-layer perceptron model to obtain the target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on the transmission line theory and calculated using the empirical formulas for equivalent inductive reactance and capacitive reactance.

[0010] Optionally, when the frequency selective surface is a double-loop metal patch structure, the target design parameters include: the period of the unit of the double-loop metal patch structure, the distance between the two outer loops of the double-loop metal patch structure, the loop width of the outer loop of the double-loop metal patch structure, the distance between the outer loop and the inner loop of the double-loop metal patch structure, and the loop width of the inner loop of the double-loop metal patch structure.

[0011] Optionally, the training process of the pre-trained multi-layer perceptron model includes:

[0012] Select an FSS training unit structure;

[0013] Establish an equivalent circuit model of the FSS training unit structure according to the transmission line theory;

[0014] Substitute the empirical formulas for equivalent inductive reactance and capacitive reactance into the equivalent circuit model to obtain the transmission performance calculation formula corresponding to the FSS training unit structure;

[0015] Use the transmission performance calculation formula and the preset target design parameter range to obtain a preset sample data set;

[0016] Train an initial multi-layer perceptron model using the preset sample data set to obtain the pre-trained multi-layer perceptron model.

[0017] Optionally, using the transmission performance calculation formula and the preset target design parameter range to obtain a preset sample data set includes:

[0018] Traverse the target design parameter range to generate multiple groups of target training design parameters;

[0019] Substitute the multiple groups of target training design parameters into the transmission performance calculation formula in turn to generate corresponding transmission performance training curves;

[0020] Convert the transmission performance training curves into corresponding training square waves;

[0021] Form a set of training data with the training square waves corresponding to each group of target training design parameters;

[0022] Form an initial sample data set from all the training data;

[0023] Perform outlier screening on the initial sample data set to obtain a preset sample data set.

[0024] Optionally, the transmission performance calculation formula is expressed as:

[0025]

[0026] Among them, S21 represents the transmission performance value corresponding to the FSS training unit structure, and Y represents the equivalent final total impedance of the FSS training unit structure;

[0027]

[0028] B 1 = 0.75×B C1 ×(d 1 / p);

[0029]

[0030] B C1 = F(p, g 1 , λ);

[0031] B C2 = F(p, g 2 , λ);

[0032] d 1 = p - g 1 ;

[0033] d 2 = d 1 - w 1 *2 - g 2 *2;

[0034]

[0035] X 2 = F(p, 2w 2 , λ)×(d 2 / p);

[0036] X L1 = F(p, w 1 , λ);

[0037] X L2 = F(p, w 2 , λ);

[0038] Among them, j represents the imaginary unit, and B 1 represents B C1 and B C2The first capacitive coupling result of B 2 Indicates B C1 And B C2 The second capacitive coupling result of B C1 Indicates the first normalized capacitance, B C2 Indicates the second normalized capacitance, F(p,g 1 ,λ) represents substituting (p,g 1 ,λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, and the corresponding first normalized capacitance of the FSS training unit structure, F(p,g 2 ,λ) represents substituting (p,g 2 ,λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, and the corresponding second normalized capacitance of the FSS training unit structure, p represents the period of the unit of the FSS training unit structure, g 1 Represents the distance between two outer square rings of the FSS training unit structure, g 2 Represents the distance between the outer square ring and the inner square ring of the FSS training unit structure, w 1 Represents the ring width of the outer square ring of the FSS training unit structure, w 2 Represents the ring width of the inner square ring of the FSS training unit structure, d 1 Represents the side length of the outer square ring of the FSS training unit structure, d 2 Represents the side length of the inner square ring of the FSS training unit structure, Z 0 Is the air impedance, X 1 Indicates X L1 And X L2 The first inductive coupling result of X 2 Indicates X L1 And X L2 The second inductive coupling result of X L1 Indicates the first normalized inductance corresponding to the FSS training unit structure, X L2 Indicates the second normalized inductance corresponding to the FSS training unit structure, F(p,w 1 ,λ) represents substituting (p,w 1 ,λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, and the corresponding first normalized inductance of the FSS training unit structure, F(p,w 2 ,λ) represents substituting (p,w 2 ,λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, and the corresponding second normalized inductance of the FSS training unit structure, λ represents the wavelength of the current electromagnetic wave, and the wavelength of the current electromagnetic wave is obtained by selecting one by one from within the electromagnetic wave threshold range.

[0039] Optionally, use a preset sample data set to train the initial multi-layer perceptron model to obtain a pre-trained multi-layer perceptron model, including:

[0040] Input a preset sample data set into the initial multi-layer perceptron model;

[0041] Continuously train the initial multi-layer perceptron model in the direction of the decrease of the loss function value of the initial multi-layer perceptron model;

[0042] Use the initial multi-layer perceptron model that meets the preset stop condition as the pre-trained multi-layer perceptron model;

[0043] Among them, the mean square error loss function is used as the loss function.

[0044] Optionally, the preset stop condition includes:

[0045] The number of iterations meets the preset iteration threshold or the value of the loss function is continuously less than the loss threshold.

[0046] In a second aspect, the present invention provides a frequency selective surface inverse design device, which includes an acquisition unit and a parameter design unit;

[0047] The acquisition unit is used to: acquire a target square wave; the target square wave is used to characterize the expected transmission performance of the frequency selective surface;

[0048] The parameter design unit is used to: input the target square wave into the pre-trained multi-layer perceptron model to obtain the target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on the transmission line theory and calculated using the equivalent inductive reactance and capacitive reactance empirical formulas.

[0049] In a third aspect, the present invention provides a frequency selective surface inverse design device, including: a processor, a storage medium and a bus. The storage medium stores machine-readable instructions executable by the processor. When the frequency selective surface inverse design device runs, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to perform the steps of any one of the frequency selective surface inverse design methods in the first aspect as described above.

[0050] The present invention provides a method, apparatus and device for inverse design of frequency selective surfaces. Among them, a method for inverse design of frequency selective surfaces includes: obtaining a target square wave; the target square wave is used to characterize the expected transmission performance of the frequency selective surface; inputting the target square wave into a pre-trained multi-layer perceptron model to obtain the target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on the transmission line theory and calculated using the empirical formulas of equivalent inductive reactance and capacitive reactance. In the present invention, the equivalent circuit model based on the transmission line theory is used instead of full-wave simulation, and the equivalent circuit model quickly calculates the response characteristics of the electromagnetic structure through the empirical formulas of inductive reactance and capacitive reactance, avoiding the complex electromagnetic field solving process, significantly shortening the sample data preparation time. Secondly, the equivalent circuit model based on the transmission line theory is constrained by physical laws (empirical formulas of equivalent inductive reactance and capacitive reactance), and the generated data is more in line with physical laws, reducing noise and outliers in the data, thereby improving the stability and prediction accuracy of the multi-layer perceptron model training; finally, due to the introduction of the pre-trained multi-layer perceptron model, the target square wave is directly mapped to the corresponding design parameters, realizing fast and efficient inverse design. The multi-layer perceptron model can output high-precision design parameters in a short time by learning the mapping relationship in the preset sample data set, significantly improving the efficiency and accuracy of the frequency selective surface design.

[0051] The following will further describe the present invention in detail with reference to the drawings and embodiments. Description of the Drawings

[0052] Figure 1 It is a schematic flowchart of the method for inverse design of frequency selective surfaces provided by an embodiment of the present invention;

[0053] Figure 2 Exemplarily shows a schematic structural diagram of the equivalent circuit model after equivalent transformation of the double-ring metal patch structure based on the transmission line theory;

[0054] Figure 3 Exemplarily shows a schematic diagram of the result comparison between the real target design parameters and the predicted target design parameters based on the present invention;

[0055] Figure 4 It is a schematic structural diagram of an apparatus for inverse design of frequency selective surfaces provided by an embodiment of the present invention;

[0056] Figure 5 It is a schematic structural diagram of a device for inverse design of frequency selective surfaces provided by an embodiment of the present invention. Detailed Embodiments

[0057] The following further describes the present invention in detail with specific embodiments, but the embodiments of the present invention are not limited thereto.

[0058] To improve the efficiency and accuracy of frequency selective surface (FSS) design, an embodiment of the present invention provides a method for inverse design of frequency selective surfaces. Figure 1 It is a schematic flow chart of the method for inverse design of frequency selective surfaces provided by the embodiment of the present invention. As Figure 1 shown, it includes:

[0059] S101. Obtain a target square wave.

[0060] It should be noted that the target square wave is used to characterize the expected transmission performance of the frequency selective surface, and can be specifically set according to the functional requirements of the FSS. Exemplarily, according to the band-pass filter, band-stop filter, high-pass filter, and low-pass filter of the FSS, the basic shape and characteristics of the target square wave can be determined.

[0061] S102. Input the target square wave into a pre-trained multi-layer perceptron model to obtain the target design parameters corresponding to the frequency selective surface.

[0062] The pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on the transmission line theory, and is calculated using the empirical formulas of equivalent inductive reactance and capacitive reactance.

[0063] An embodiment of the present invention provides a method for inverse design of frequency selective surfaces, which uses an equivalent circuit model based on the transmission line theory instead of full-wave simulation. The equivalent circuit model quickly calculates the response characteristics of the electromagnetic structure through the empirical formulas of inductive reactance and capacitive reactance, avoiding the complex electromagnetic field solving process, significantly shortening the sample data preparation time. Secondly, the equivalent circuit model based on the transmission line theory is constrained by physical laws (empirical formulas of equivalent inductive reactance and capacitive reactance), and the generated data is more in line with physical laws, reducing noise and outliers in the data, thereby improving the stability and prediction accuracy of the multi-layer perceptron model training. Finally, due to the introduction of the pre-trained multi-layer perceptron model, the target square wave is directly mapped to the corresponding design parameters, realizing fast and efficient inverse design. The multi-layer perceptron model can output high-precision design parameters in a short time by learning the mapping relationship in the preset sample data set, significantly improving the efficiency and accuracy of frequency selective surface design.

[0064] Optionally, when the frequency selective surface is a double-square-loop metal patch structure, the target design parameters include: the period of the unit of the double-square-loop metal patch structure, the distance between the two outer square loops of the double-square-loop metal patch structure, the loop width of the outer square loop of the double-square-loop metal patch structure, the distance between the outer square loop and the inner square loop of the double-square-loop metal patch structure, and the loop width of the inner square loop of the double-square-loop metal patch structure.

[0065] Furthermore, the frequency selective surface may further include: a single-sided ring metal patch structure, a Jerusalem cross patch structure, a single-ring metal patch structure, and a double-ring metal patch structure. For example, when the frequency selective surface is a single-sided ring metal patch structure, the corresponding target design parameters include: the period of the unit of the single-sided ring metal patch structure, the ring width of the single-sided ring metal patch structure, and the spacing between the single-sided ring metal patch structures. The selection of the frequency selective surface and the selection of its corresponding target design parameters can be flexibly set according to specific demand scenarios, and this embodiment does not limit this.

[0066] Optionally, the training process of the pre-trained multi-layer perceptron model includes:

[0067] Select the FSS training unit structure;

[0068] Establish an equivalent circuit model of the FSS training unit structure according to the transmission line theory;

[0069] Substitute the equivalent inductive reactance and capacitive reactance empirical formulas into the equivalent circuit model to obtain the transmission performance calculation formula corresponding to the FSS training unit structure;

[0070] Use the transmission performance calculation formula and the preset target design parameter interval to obtain a preset sample data set;

[0071] Train the initial multi-layer perceptron model using the preset sample data set to obtain the pre-trained multi-layer perceptron model.

[0072] The preset target design parameter interval can be set according to experience. In this embodiment, the target design parameter interval can be taken as p = [6mm, 14mm], g 1 = [0.1mm, 2mm], w 1 = [0.1mm, 2mm], g 2 = [0.1mm, 2mm], w 2 = [0.1mm, 2mm].

[0073] Optionally, using the transmission performance calculation formula and the preset target design parameter interval to obtain a preset sample data set includes:

[0074] Traverse the target design parameter interval to generate multiple groups of target training design parameters;

[0075] Substitute the multiple groups of target training design parameters into the transmission performance calculation formula in turn to generate corresponding transmission performance training curves;

[0076] Convert the transmission performance training curve into a corresponding training square wave;

[0077] Form a set of training data from the training square waves corresponding to each group of target training design parameters;

[0078] Form an initial sample data set from all training data;

[0079] Perform outlier screening on the initial sample data set to obtain a preset sample data set.

[0080] The conversion method of converting the transmission performance training curve into a training square wave is as follows: 200 points can be set in the range of 2 - 25Ghz, and the y - coordinate less than - 10dB is set to 1, and the rest are set to 0.

[0081] The training square wave expression: 200 points in the range of 2 - 25Ghz, for example, the range of 4 - 6Ghz is set to 1, the range of 9 - 11Ghz is set to 1, and the rest are set to 0.

[0082] Performing outlier screening on the initial sample data set includes removing the sample rows corresponding to non - double - resonance frequency points in the transmission performance training curve (S21) of the initial sample data set and removing the sample rows containing outliers (Nan), and finally obtaining a preset sample data set.

[0083] This embodiment is described with the FSS training unit structure being a double - loop metal patch structure. Specifically, when the FSS training unit structure is a double - loop metal patch structure, the corresponding transmission performance calculation formula is expressed as:

[0084]

[0085] Among them, S21 represents the transmission performance value corresponding to the FSS training unit structure, and Y represents the final total impedance equivalent to the FSS training unit structure;

[0086]

[0087] B 1 = 0.75×B C1 ×(d 1 / p);

[0088]

[0089] B C1 = F(p,g 1 ,λ);

[0090] B C2 = F(p,g 2 ,λ);

[0091] d 1 = p - g 1 ;

[0092] d 2 = d 1 - w 1 *2 - g 2*2;

[0093]

[0094] X 2 = F(p, 2w 2 , λ) × (d 2 / p);

[0095] X L1 = F(p, w 1 , λ);

[0096] X L2 = F(p, w 2 , λ);

[0097] where j represents the imaginary unit, B 1 represents the first capacitive coupling result of B C1 and B C2 , B 2 represents the second capacitive coupling result of B C1 and B C2 , B C1 represents the first normalized capacitance, B C2 represents the second normalized capacitance, F(p, g 1 , λ) represents the first normalized capacitance corresponding to the FSS training unit structure when (p, g 1 , λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, F(p, g 2 , λ) represents the second normalized capacitance corresponding to the FSS training unit structure when (p, g 2 , λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, p represents the period of the unit of the FSS training unit structure, g 1 represents the distance between the two outer square rings of the FSS training unit structure, g 2 represents the distance between the outer square ring and the inner square ring of the FSS training unit structure, w 1 represents the ring width of the outer square ring of the FSS training unit structure, w 2 represents the ring width of the inner square ring of the FSS training unit structure, d 1 represents the side length of the outer square ring of the FSS training unit structure, d 2 represents the side length of the inner square ring of the FSS training unit structure, Z 0 is the air impedance, X 1 represents X L1 and X L2 's first inductive coupling result, X 2 represents X L1 and X L2 's second inductive coupling result, X L1 represents the first normalized inductance corresponding to the FSS training unit structureL2 represents the second normalized inductance corresponding to the FSS training unit structure, F(p, w 1 , λ) represents substituting (p, w 1 , λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, the first normalized inductance corresponding to the FSS training unit structure, F(p, w 2 , λ) represents substituting (p, w 2 , λ) into the empirical formula of equivalent inductive reactance and capacitive reactance, the second normalized inductance corresponding to the FSS training unit structure, λ represents the wavelength of the current electromagnetic wave, and the wavelength of the current electromagnetic wave is obtained by selecting one by one from within the electromagnetic wave threshold range.

[0098] Figure 2 Exemplarily shows a schematic diagram of the equivalent circuit model structure after equivalent substitution of the double - loop metal patch structure based on the transmission line theory. As Figure 2 shown, Figure 2 Figure (a) in Figure 2 is the double - loop metal patch structure,

[0099] Optionally, training the initial multi - layer perceptron model with a preset sample data set to obtain a pre - trained multi - layer perceptron model, including:

[0100] Inputting the preset sample data set into the initial multi - layer perceptron model;

[0101] Continuously training the initial multi - layer perceptron model in the direction of the decreasing loss function value of the initial multi - layer perceptron model;

[0102] Taking the initial multi - layer perceptron model that meets the preset stop condition as the pre - trained multi - layer perceptron model;

[0103] Among them, the loss function uses the mean square error loss function.

[0104] Optionally, the preset stop condition includes:

[0105] The number of iterations meets the preset iteration threshold or the value of the loss function is continuously less than the loss threshold.

[0106] In summary, the frequency - selective surface inverse design method proposed by the present invention has the following technical advantages:

[0107] 1. Using an equivalent circuit model to replace full - wave simulation, significantly reducing the time cost of data generation.

[0108] 2. Generating high - quality data with the constraint of physical laws, improving the stability of the multi - layer perceptron model training.

[0109] 3. Combining deep learning methods and physical prior knowledge, alleviating the dependence on large - scale real data.

[0110] 4. Implement fast and high-precision inverse design of FSS using a pre-trained multi-layer perceptron model.

[0111] To verify the effectiveness of the frequency selective surface inverse design method proposed by the present invention, simulation experiments were also conducted in the embodiments of the present invention. Figure 3 Exemplarily, a schematic diagram of the result comparison between the real target design parameters and the predicted target design parameters based on the present invention is shown, as Figure 3 shown, Figure 3 Figures (a)-(e) of show the differences between the real values and the predicted values of the spacing between the two outer rings of the FSS unit structure corresponding to the target FSS unit structure, the spacing between the outer ring and the inner ring of the FSS unit structure, the ring width of the outer ring of the FSS unit structure, the ring width of the inner ring of the FSS unit structure, and the period of the unit of the FSS unit structure, respectively. And according to Figure 3 it can be seen that the predicted values obtained by the frequency selective surface inverse design method proposed by the present invention basically coincide with the real values, verifying the effectiveness of the method of the present invention.

[0112] The method provided by the embodiments of the present invention can be applied to electronic devices. Specifically, the electronic device can be: a desktop computer, a portable computer, a smart mobile terminal, a server, etc., which are not limited in the embodiments of the present invention.

[0113] Based on the same inventive concept, the embodiments of the present invention also provide a frequency selective surface inverse design device. Figure 4 As shown in is a schematic structural diagram of a frequency selective surface inverse design device provided by an embodiment of the present invention, as Figure 4 shown, including: an acquisition unit 401 and a parameter design unit 402;

[0114] The acquisition unit 401 is configured to: acquire a target square wave; the target square wave is used to characterize the desired transmission performance of the frequency selective surface;

[0115] The parameter design unit 402 is configured to: input the target square wave into a pre-trained multi-layer perceptron model to obtain the target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on the transmission line theory and calculated using the equivalent inductive reactance and capacitive reactance empirical formulas.

[0116] Figure 5A structural schematic diagram of a frequency selective surface inverse design device provided by an embodiment of the present invention includes: a processor 510, a storage medium 520, and a bus 530. The storage medium 520 stores machine-readable instructions executable by the processor 510. When the frequency selective surface inverse design device runs, the processor 510 communicates with the storage medium 520 through the bus 530, and the processor 510 executes the machine-readable instructions to perform the steps of the above method embodiment. The specific implementation manners and technical effects are similar and will not be elaborated here.

[0117] The storage medium may include a random access memory (RAM), or may also include a non-volatile memory (NVM), such as at least one disk memory. Optionally, the storage medium may also be at least one storage device located away from the aforementioned processor.

[0118] The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0119] It should be noted that terms such as "first" and "second" are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present invention described here can be implemented in an order other than those illustrated or described here. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present invention. On the contrary, they are only examples of devices and methods consistent with some aspects of the present invention.

[0120] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification.

[0121] Although the present invention has been described in connection with various embodiments herein, however, in the process of implementing the claimed invention, those skilled in the art can understand and implement other variations of the above-described disclosed embodiments by viewing the drawings and the disclosure. In the description of the present invention, the term "comprising" does not exclude other components or steps, the word "a" or "one" does not exclude a plurality of cases, and the meaning of "a plurality" is two or more unless otherwise specifically defined. In addition, certain measures are described in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0122] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. A frequency selective surface inverse design method, characterized in that: include: Get the target square wave; The target square wave is used to characterize the expected transmission performance of the frequency selective surface; Inputting the target square wave into a pre-trained multi-layer perceptron model to obtain target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; The preset sample data set is an equivalent circuit model established based on transmission line theory and is calculated using an empirical formula of equivalent inductive reactance and capacitive reactance.

2. The frequency selective surface inverse design method according to claim 1, characterized in that: When the frequency selective surface is a double-ring metal patch structure, the target design parameters include: the period of the unit of the double-ring metal patch structure, the spacing between the two outer square rings of the double-ring metal patch structure, the ring width of the outer square ring of the double-ring metal patch structure, the spacing between the outer square ring and the inner square ring of the double-ring metal patch structure, and the ring width of the inner square ring of the double-ring metal patch structure.

3. The frequency selective surface inverse design method according to claim 1, characterized in that: The training process of the pre-trained multi-layer perceptron model includes: Select the FSS training unit structure; Establishing an equivalent circuit model of the FSS training unit structure according to transmission line theory; Substituting the empirical formula of equivalent inductive reactance and capacitive reactance into the equivalent circuit model, obtaining a transmission performance calculation formula corresponding to the FSS training unit structure; Using the transmission performance calculation formula and the preset target design parameter range, a preset sample data set is obtained; The preset sample data set is used to train the initial multi-layer perceptron model to obtain the pre-trained multi-layer perceptron model.

4. The frequency selective surface inverse design method according to claim 3, characterized in that: The transmission performance calculation formula and the preset target design parameter range are used to obtain a preset sample data set, including: Traversing the target design parameter interval to generate multiple groups of target training design parameters; Substituting the multiple groups of target training design parameters into the transmission performance calculation formula in sequence to generate corresponding transmission performance training curves; Converting the transmission performance training curve into a corresponding training square wave; The training square waves corresponding to each group of the target training design parameters constitute a group of training data; All the training data form an initial sample data set; The initial sample data set is processed for outlier screening to obtain the preset sample data set.

5. The frequency selective surface inverse design method according to claim 3, characterized in that: The transmission performance calculation formula is expressed as: Wherein, S21 represents the transmission performance value corresponding to the FSS training unit structure, and Y represents the final total impedance equivalent to the FSS training unit structure; B1=0.75×B C1 ×(d1 / p); B C1 =F(p,g1,λ); B C2 =F(p,g2,λ); d1 = p-g1; d2=d1-w1*2-g2*2; X2=F(p,2w2,λ)×(d2 / p); X L1 =F(p,w1,λ); X L2 =F(p,w2,λ); Where j represents the imaginary unit, B1 represents B C1 and B C2 The first capacitive coupling result, B2 represents B C1 and B C2 The second capacitive coupling result, B C1 represents the first normalized capacitance, B C2 represents the second normalized capacitance, F(p,g1,λ) represents the first normalized capacitance corresponding to the FSS training unit structure when (p,g1,λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, F(p,g2,λ) represents the second normalized capacitance corresponding to the FSS training unit structure when (p,g2,λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, p represents the period of the unit of the FSS training unit structure, g1 represents the spacing between the two outer square rings of the FSS training unit structure, g2 represents the spacing between the outer square ring and the inner square ring of the FSS training unit structure, w1 represents the ring width of the outer square ring of the FSS training unit structure, w2 represents the ring width of the inner square ring of the FSS training unit structure, d1 represents the side length of the outer square ring of the FSS training unit structure, d2 represents the side length of the inner square ring of the FSS training unit structure, Z0 represents the air impedance, X1 represents X L1 and X L2 The first inductive coupling result, X2 represents X L1 and X L2 The second inductive coupling result, X L1 represents the first normalized inductance corresponding to the FSS training unit structure, X L2 represents the second normalized inductance corresponding to the FSS training unit structure, F(p,w1,λ) represents the first normalized inductance corresponding to the FSS training unit structure when (p,w1,λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, F(p,w2,λ) represents the second normalized inductance corresponding to the FSS training unit structure when (p,w2,λ) is substituted into the empirical formula of equivalent inductive reactance and capacitive reactance, λ represents the wavelength of the current electromagnetic wave, and the wavelength of the current electromagnetic wave is selected one by one from the electromagnetic wave threshold range.

6. The frequency selective surface inverse design method according to claim 3, characterized in that: The method of using the preset sample data set to train the initial multi-layer perceptron model to obtain the pre-trained multi-layer perceptron model includes: Inputting the preset sample data set into the initial multi-layer perceptron model; Continuously training the initial multilayer perceptron model in a direction in which the loss function value of the initial multilayer perceptron model decreases; Using the initial multi-layer perceptron model that meets the preset stop condition as the pre-trained multi-layer perceptron model; The loss function adopts the mean square error loss function.

7. The frequency selective surface inverse design method according to claim 6, characterized in that: The preset stop conditions include: The number of iterations meets the preset iteration threshold or the value of the loss function is continuously less than the loss threshold.

8. A frequency selective surface inverse design device, characterized in that: The frequency selective surface inverse design device comprises: an acquisition unit and a parameter design unit; The acquisition unit is used to: acquire a target square wave; the target square wave is used to characterize the expected transmission performance of the frequency selective surface; The parameter design unit is used to: input the target square wave into a pre-trained multi-layer perceptron model to obtain target design parameters corresponding to the frequency selective surface; the pre-trained multi-layer perceptron model is trained based on a preset sample data set; the preset sample data set is an equivalent circuit model established based on transmission line theory, and is calculated using an empirical formula for equivalent inductive reactance and capacitive reactance.

9. A frequency selective surface inverse design device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the frequency selective surface inverse design device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the frequency selective surface inverse design method as described in any one of claims 1 to 7.

Images