Four-channel global electromagnetic metasurface and design method and related device thereof

Abstract

The invention belongs to the field of electromagnetic materials, and discloses a four-channel global electromagnetic metasurface and a design method and related device thereof, and the method comprises the steps: obtaining a phase profile map of the four-channel global electromagnetic metasurface in a target frequency band; obtaining an electromagnetic frequency response curve of each unit of the four-channel global electromagnetic metasurface in the target frequency band according to the phase profile map of the four-channel global electromagnetic metasurface in the target frequency band, and converting the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band; and generating a unit image of each unit through a pre-trained unit image generator according to the two-dimensional image of each unit in the target frequency band, and obtaining a unit structure of each unit according to the unit image of each unit. A one-dimensional electromagnetic frequency response curve is converted into a two-dimensional image to solve the problem that input and output dimensions of a unit image generator are not matched, enough effective features can be extracted to accurately describe and predict, and the design efficiency and stability are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of electromagnetic materials and relates to a four-channel full-domain electromagnetic supersurface and a design method and related devices thereof. Background Art

[0002] An electromagnetic metasurface is an artificially designed subwavelength planar unit that is periodically or non-periodically arranged in a two-dimensional plane through a pre-set electromagnetic frequency response curve. Electromagnetic metasurfaces are widely used in electronics, communications, optics and other fields because they can flexibly adjust the amplitude, phase, polarization and frequency of electromagnetic waves. In the traditional electromagnetic metasurface design paradigm, researchers mainly rely on numerical simulation software for design, which is usually based on the finite element method, time-domain finite difference method or other numerical calculation methods. The design process generally starts with the preliminary determination of the geometric structure and material properties of the metasurface unit, and the electromagnetic frequency response curve of the structure is simulated by simulation software to obtain electromagnetic parameters such as reflection coefficient, transmission coefficient and absorptivity, and then the structural parameters are repeatedly adjusted according to the simulation results until the design requirements are met. However, when faced with complex targets, this trial-and-error method has problems such as low efficiency, limited search solution space and long design cycle.

[0003] In recent years, artificial intelligence technologies represented by deep learning have provided new methods for the design of electromagnetic metasurfaces. Deep learning has powerful nonlinear mapping and high-dimensional data fitting capabilities. Using this feature, researchers can establish a direct mapping relationship between the structural parameters of the electromagnetic metasurface and its electromagnetic frequency response curve, thereby quickly predicting the electromagnetic frequency response curve under different structural parameters, greatly shortening the design time. In addition, deep learning can also reversely derive the corresponding structural parameters from the target electromagnetic performance, simplifying the design process, and combining optimization algorithms to efficiently search for the optimal design solution in the high-dimensional parameter space, improving the design efficiency and effect. The application of deep learning in the design of electromagnetic metasurfaces provides new ideas and tools for solving many problems in traditional methods, and is gradually changing the research and application methods in this field.

[0004] However, existing deep learning-based electromagnetic metasurface design methods usually lack pre-screening of training sets, resulting in a large number of low-quality data sets entering the training process. These low-quality data not only increase the consumption of computing resources, but also may make it difficult for the model to converge, reducing the design efficiency and accuracy. Specifically, the unscreened training data set contains many unit structures that are not suitable for the requirements of metasurface design. These structures perform poorly in terms of electromagnetic frequency response curves, such as failing to cover the required 2π full phase range or 0-1 full amplitude range, resulting in the model being unable to accurately learn the key electromagnetic frequency response curve characteristics, which in turn affects the final design effect. At the same time, the problem of mismatch between input and output dimensions is faced in the inverse electromagnetic metasurface design process. In existing methods, the electromagnetic frequency response curve is usually one-dimensional, while the electromagnetic metasurface unit structure is two-dimensional. This dimensional difference makes model training difficult. Because the feature representation capability of one-dimensional signals is limited, the model may not be able to extract enough effective features to accurately describe and predict the electromagnetic frequency response curve characteristics of the electromagnetic metasurface unit, resulting in poor training results and inefficient and unstable design processes. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art that the model may not be able to extract enough effective features to accurately describe and predict the electromagnetic frequency response curve characteristics of the electromagnetic metasurface unit, resulting in poor training effect and the design process becoming inefficient and unstable, and to provide a four-channel full-domain electromagnetic metasurface and its design method and related devices.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a four-channel full-domain electromagnetic metasurface design method, comprising: obtaining a phase profile of the four-channel full-domain electromagnetic metasurface in a target frequency band; obtaining an electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band according to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and converting the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band; generating a unit image of each unit through a pre-trained unit image generator according to the two-dimensional image of each unit in the target frequency band, and obtaining a unit structure of each unit according to the unit image of each unit; wherein , the pre-trained unit image generator is obtained in the following manner: obtaining unit images of a number of unit structure samples, and obtaining electromagnetic frequency response curves of a number of unit structure samples within a target frequency band, and converting the electromagnetic frequency response curves of the number of unit structure samples within the target frequency band into two-dimensional images of the number of unit structure samples within the target frequency band; based on the unit images of the number of unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as the pre-trained unit image generator.

[0008] Optionally, the unit structure of the unit structure sample is a mirror-symmetric structure or a rotationally symmetric structure.

[0009] Optionally, obtaining unit images of several unit structure samples includes: dividing each unit structure sample into several grids, setting each grid in a manner of 1 for metal and 2 for medium to form a Boolean matrix, obtaining a two-dimensional matrix of the unit structure sample and converting it into a binary image, to obtain a unit image of each unit structure sample; obtaining electromagnetic frequency response curves of several unit structure samples within a target frequency band includes: performing electromagnetic simulation on several unit structure samples within the target frequency band through simulation software, to obtain electromagnetic frequency response curves of several unit structure samples within the target frequency band.

[0010] Optionally, after obtaining the electromagnetic frequency response curves of several unit structure samples in the target frequency band, the unit structure samples whose phase change in the target frequency band is less than a phase change threshold are eliminated according to the electromagnetic frequency response curve of each unit structure sample in the target frequency band.

[0011] Optionally, the converting of the electromagnetic frequency response curve of each unit in the target frequency band into the two-dimensional image of each unit in the target frequency band includes: converting the electromagnetic frequency response curve of each unit in the target frequency band into the two-dimensional image of each unit in the target frequency band by using Gram angle difference field technology; the converting of the electromagnetic frequency response curves of several unit structure samples in the target frequency band into the two-dimensional image of several unit structure samples in the target frequency band includes: converting the electromagnetic frequency response curves of several unit structure samples in the target frequency band into the two-dimensional image of several unit structure samples in the target frequency band by using Gram angle difference field technology.

[0012] Optionally, the method of obtaining a phase profile of a four-channel full-domain electromagnetic metasurface within a target frequency band includes: obtaining an x-line polarized holographic image, orbital angular momentum vortex wave and a target frequency band, as well as a y-line polarized holographic image, orbital angular momentum vortex wave and a target frequency band of the four-channel full-domain electromagnetic metasurface; based on the x-line polarized and y-line polarized holographic images, orbital angular momentum vortex wave and the target frequency band, according to the diffraction propagation formula of orbital angular momentum vortex wave and electromagnetic wave, obtaining an ideal phase distribution diagram of x-line polarization and an ideal phase distribution diagram of y-line polarization; arranging the ideal phase distribution diagram of x-line polarization and the ideal phase distribution diagram of y-line polarization in an interlaced manner in the horizontal and vertical directions according to a preset step distance, to obtain a phase profile of the four-channel full-domain electromagnetic metasurface within the target frequency band.

[0013] In a second aspect, the present invention provides a four-channel full-domain electromagnetic metasurface design system, comprising: a data acquisition module, used to acquire a phase profile of the four-channel full-domain electromagnetic metasurface in a target frequency band; a two-dimensional conversion module, used to obtain an electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band according to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and convert the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band; a structure determination module, used to generate a unit image of each unit through a pre-trained unit image generator according to the two-dimensional image of each unit in the target frequency band, and convert the unit image of each unit into a two-dimensional image of each unit in the target frequency band; The unit structure of each unit is obtained by imaging; wherein the pre-trained unit image generator is obtained by the following method: obtaining unit images of a number of unit structure samples, and obtaining electromagnetic frequency response curves of a number of unit structure samples within a target frequency band, and converting the electromagnetic frequency response curves of the number of unit structure samples within the target frequency band into two-dimensional images of the number of unit structure samples within the target frequency band; based on the unit images of the number of unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as the pre-trained unit image generator.

[0014] According to a third aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned four-channel full-domain electromagnetic metasurface design method when executing the computer program.

[0015] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned four-channel full-domain electromagnetic metasurface design method are implemented.

[0016] According to a fifth aspect of the present invention, a four-channel full-area electromagnetic metasurface is provided, wherein the unit structure of each unit of the four-channel full-area electromagnetic metasurface is designed using the above-mentioned four-channel full-area electromagnetic metasurface design method.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] The four-channel full-domain electromagnetic metasurface design method of the present invention obtains the electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band based on the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and then converts the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band. Finally, according to the two-dimensional image of each unit in the target frequency band, a unit image of each unit is generated by a pre-trained unit image generator, and then a specific unit structure is obtained based on the unit image to realize the design of the four-channel full-domain electromagnetic metasurface. Specifically, by converting the one-dimensional electromagnetic frequency response curve into a two-dimensional image to solve the input and output dimension mismatch problem of the unit image generator, the unit image generator can process these data in an image recognition manner, especially in the convolution processing, it can effectively extract spatial features, patterns and local relationships, thereby improving the intuitiveness and effectiveness of the reverse design process, so as to overcome the problem that the training of the unit image generator becomes difficult due to the difference in dimensions, so that the unit image generator can extract enough effective features to accurately describe and predict the electromagnetic response characteristics of the electromagnetic metasurface unit, effectively improve the training effect, and thus make the design process of the four-channel full-domain electromagnetic metasurface efficient and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a flow chart of a four-channel full-domain electromagnetic metasurface design method according to an embodiment of the present invention.

[0020] Figure 2 Schematic diagram of two-dimensional matrix conversion of a mirror-symmetric structure according to an embodiment of the present invention.

[0021] Figure 3 Schematic diagram of two-dimensional matrix transformation of a rotationally symmetric structure according to an embodiment of the present invention.

[0022] Figure 4 1 is an electromagnetic response distribution diagram of the reflection amplitude of the unit structure sample of an embodiment of the present invention at 8 GHz and different metal filling rates.

[0023] Figure 5 1 is an electromagnetic response distribution diagram of the transmission amplitude of the unit structure sample of an embodiment of the present invention at 8 GHz and different metal filling rates.

[0024] Figure 6 1 is an electromagnetic response distribution diagram of the reflection phase of the unit structure sample of an embodiment of the present invention at 8 GHz and different metal filling rates.

[0025] Figure 7 1 is an electromagnetic response distribution diagram of the transmission phase of the unit structure sample of an embodiment of the present invention at 8 GHz and different metal filling rates.

[0026] Figure 8 It is a schematic diagram of a two-dimensional image of an electromagnetic frequency response curve corresponding to a unit structure sample according to an embodiment of the present invention after being converted by a Gram angle difference field.

[0027] Fig. 9 Schematic diagram of the generative adversarial network model architecture of an embodiment of the present invention.

[0028] Fig.10 This is a loss curve graph of the generator and the discriminator as the number of iterations changes during the training process of the generative adversarial network model in an embodiment of the present invention.

[0029] Fig.11 This is a schematic diagram of the functional implementation of a four-channel global electromagnetic metasurface according to an embodiment of the present invention.

[0030] Fig.12 Figure 1 is a phase profile diagram of a four-channel full-area electromagnetic metasurface according to an embodiment of the present invention, wherein (a), (b), (c), (d), (e) and (f) are respectively the upper phase distribution diagram of the array at 8 GHz, the lower phase distribution diagram at 8 GHz, the upper phase distribution diagram of the array at 12 GHz, the lower phase distribution diagram at 12 GHz, the upper phase distribution diagram of the array with an interlaced arrangement and the lower phase distribution diagram with an interlaced arrangement of the four-channel full-area electromagnetic metasurface.

[0031] Fig.13 This is the normalized radiation pattern of the first-order OAM wave of the four-channel full-area electromagnetic metasurface of an embodiment of the present invention when operating at 8 GHz.

[0032] Fig.14 This is the normalized radiation pattern of the second-order OAM wave of the four-channel full-area electromagnetic metasurface of an embodiment of the present invention when operating at 12 GHz.

[0033] Fig.15 This is a simulation diagram of a four-channel full-domain electromagnetic metasurface at 8 GHz according to an embodiment of the present invention. Figure (a) and Figure (b) are the first-order OAM near-field electric field amplitude distribution and phase distribution diagrams, respectively.

[0034] Fig.16 This is a simulation diagram of a four-channel full-domain electromagnetic metasurface at 12 GHz according to an embodiment of the present invention. Figure (a) and Figure (b) are the second-order OAM near-field electric field amplitude distribution and phase distribution diagrams, respectively.

[0035] Fig.17 Schematic diagram of the simulated holographic patterns "AI" and "XD" letter patterns at different diffraction distances when the four-channel full-area electromagnetic metasurface of an embodiment of the present invention operates at 8 GHz and 12 GHz respectively.

[0036] Fig.18 This is a structural block diagram of a four-channel full-domain electromagnetic metasurface design system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0038] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0039] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0040] See also Figure 1In one embodiment of the present invention, a four-channel full-area electromagnetic metasurface design method is provided, which can design a four-channel full-area electromagnetic metasurface that reflects or transmits orbital angular momentum vortex waves and specific holographic patterns under the excitation of waves of different frequencies and different linear polarizations.

[0041] Specifically, the four-channel full-domain electromagnetic metasurface design method of the present invention comprises the following steps:

[0042] S1: Obtain the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band.

[0043] S2: According to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, the electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band is obtained, and the electromagnetic frequency response curve of each unit in the target frequency band is converted into a two-dimensional image of each unit in the target frequency band.

[0044] S3: According to the two-dimensional image of each unit in the target frequency band, a unit image of each unit is generated by a pre-trained unit image generator, and a unit structure of each unit is obtained according to the unit image of each unit.

[0045] The pre-trained unit image generator is obtained in the following manner: obtaining unit images of a number of unit structure samples, and obtaining electromagnetic frequency response curves of a number of unit structure samples within a target frequency band, and converting the electromagnetic frequency response curves of the number of unit structure samples within the target frequency band into two-dimensional images of the number of unit structure samples within the target frequency band; based on the unit images of the number of unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as the pre-trained unit image generator.

[0046] The four-channel full-domain electromagnetic metasurface design method of the present invention obtains the electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band based on the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and then converts the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band. Finally, according to the two-dimensional image of each unit in the target frequency band, a unit image of each unit is generated by a pre-trained unit image generator, and then a specific unit structure is obtained based on the unit image to realize the design of the four-channel full-domain electromagnetic metasurface. Specifically, by converting the one-dimensional electromagnetic frequency response curve into a two-dimensional image to solve the input and output dimension mismatch problem of the unit image generator, the unit image generator can process these data in an image recognition manner, especially in the convolution processing, it can effectively extract spatial features, patterns and local relationships, thereby improving the intuitiveness and effectiveness of the reverse design process, so as to overcome the problem that the training of the unit image generator becomes difficult due to the difference in dimensions, so that the unit image generator can extract enough effective features to accurately describe and predict the electromagnetic response characteristics of the electromagnetic metasurface unit, effectively improve the training effect, and thus make the design process of the four-channel full-domain electromagnetic metasurface efficient and stable.

[0047] In a possible implementation, obtaining unit images of several unit structure samples includes: dividing each unit structure sample into several grids, setting each grid in a manner such that metal is 1 and medium is 2 and forming a Boolean matrix, obtaining a two-dimensional matrix of the unit structure sample and converting it into a binary image, and obtaining a unit image of each unit structure sample.

[0048] Specifically, when obtaining unit images of a plurality of unit structure samples, the plurality of unit structure samples may be preliminarily generated by using MATLAB software.

[0049] This implementation uses the grid type as the basic topological structure, that is, the unit structure is divided into several grids, and the attributes of each grid are randomly assigned to metal or dielectric, and are represented by 1 and 0 respectively. In this way, the unit structure can be encoded into a two-dimensional matrix, which can then be converted into a binary image, i.e., a unit image, and saved in a data set.

[0050] See also Figure 2 and 3 In order to fully explore the design space of the structure, reduce the data dimension and model complexity, and fully explore the design parameter space, the unit structure of the unit structure sample can be in the form of a mirror symmetric structure and a rotational symmetric structure. The mirror symmetric structure has symmetry about the central axis, while the rotational symmetric structure remains unchanged after rotating a certain angle. These two structural forms can not only significantly reduce the number of parameters, but also improve symmetry and stability, making the design more flexible and efficient.

[0051] In a possible implementation, a four-channel full-domain electromagnetic metasurface design under two linear polarization excitations of x and y is taken as an example, and the frequency range is 8 to 12 GHz. The electromagnetic metasurface as a whole is composed of three metal layers and two F4B dielectric plates. The thickness of the dielectric plates is 1 mm, the relative dielectric constant is 2.65, and the loss tangent is 0.001. The unit period p = 10 mm remains unchanged. In order to ensure that the two linear polarizations can work in two different areas of transmission and reflection to the greatest extent and are not coupled with each other, the middle metal layer is a polarization grid, and the upper and lower layers are divided into 16×16 square grids with a side length of 0.5 mm; each grid attribute is randomly assigned by the program as metal (1) or dielectric (0), thereby encoding the structure of the entire electromagnetic metasurface into two Boolean matrices.

[0052] Optionally, obtaining electromagnetic frequency response curves of the plurality of unit structure samples within the target frequency band includes: performing electromagnetic simulation on the plurality of unit structure samples within the target frequency band by simulation software to obtain electromagnetic frequency response curves of the plurality of unit structure samples within the target frequency band.

[0053] Specifically, the generated unit structure can be subjected to electromagnetic simulation using commercial simulation software CST to obtain its electromagnetic frequency response curve. The electromagnetic frequency response includes reflection and transmission coefficients within a specific frequency range, and the corresponding amplitude and phase characteristics are obtained.

[0054] Optionally, after obtaining the electromagnetic frequency response curves of several unit structure samples in the target frequency band, the unit structure samples whose phase change in the target frequency band is less than a phase change threshold are eliminated according to the electromagnetic frequency response curve of each unit structure sample in the target frequency band.

[0055] Specifically, by analyzing the proportion of different metal areas in the unit structure, some meaningless unit structure samples are excluded. For example, unit structure samples with small or no obvious phase change in the target frequency band, these phase responses have no practical significance for the target function, so they are not included in the training sample set. In this way, according to the CST simulation results, it is possible to collect some high-quality sample sets with practical significance, lay the foundation for subsequent network training, and ensure the stability of the design.

[0056] This embodiment uses four violin plots to analyze the electromagnetic response of 30,000 samples collected at 8 GHz. Figure 4 and 5 , respectively represent the reflection phase ∠R under x polarization xx and amplitude |R xx |; Figure 6 and Figure 7 They represent the transmission phase ∠T under y polarization respectively. yy and amplitude |T yy|. Each figure contains different metal fill ratios (MFR) of the unit under mirror symmetry and rotational symmetry. Figure 4 and Figure 5 It can be seen that the reflection and transmission amplitudes are basically maintained above 0.8. Except when the MFR is around 50%, the transmission amplitude |T yy | has decreased, and some samples have values ​​below 0.5. On the other hand, Figure 6 and Figure 7 The displayed phase distribution also covers a 360° range. For example, for a mirror-symmetrical unit, when the MFR is equal to 20%, the reflection phase ∠R xx The phase values ​​are concentrated between 60° and 120°, and when the MFR increases to 70%, the phase values ​​are concentrated between -180° and -60°. Combining the different MFR distributions, it can be judged that the entire sample set basically covers the 2π range while ensuring high reflection / transmission amplitude.

[0057] In a possible implementation manner, converting the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band includes: converting the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band by using a Gram angle difference field technology; converting the electromagnetic frequency response curves of a number of unit structure samples in the target frequency band into a two-dimensional image of a number of unit structure samples in the target frequency band includes: converting the electromagnetic frequency response curves of a number of unit structure samples in the target frequency band into a two-dimensional image of a number of unit structure samples in the target frequency band by using a Gram angle difference field technology.

[0058] The Gram Angle Difference Field (GADF) technique is used to convert the one-dimensional frequency response curve into a two-dimensional image, and then a labeled sample pair is formed with the corresponding unit image. This sample pair contains both the spectral characteristics of the electromagnetic frequency response curve and the corresponding physical structure information in the form of an image, thereby ensuring the close association between the electromagnetic frequency response curve and the structural design. In this way, the subsequent network model can learn the mapping relationship between these data in the form of image recognition, especially when performing convolution operations, it can effectively extract the spatial features and patterns of the data, thereby improving the accuracy and efficiency of the design. Specifically, after completing the sample screening, the one-dimensional electromagnetic frequency response curve is converted into a two-dimensional image through GADF. The steps are as follows:

[0059] For a one-dimensional sequence array x containing n values, n}, and normalize it according to the following formula:

[0060]

[0061] The normalized sequence values ​​are then mapped to the polar coordinate system using the following formula, where their values ​​and indices are encoded as cosine angles and radius r:

[0062]

[0063] Where N is a constant used as a regularization factor for the polar coordinate system, which is set to 1. Based on the above formula, the value of the Gram angle difference field GADF can be obtained by the following formula:

[0064]

[0065] Where I is the unit row vector.

[0066] See also Figure 8 , showing the effect of the electromagnetic frequency response curve of one of the unit structure samples after GADF conversion, it can be seen that the converted image shows rich feature information. These feature information clearly reflects the amplitude and phase distribution in the original electromagnetic response through color changes, thus providing important input data for subsequent network training.

[0067] After completing the collection and processing of the sample set, the model training phase begins. In the model training phase, a generative adversarial network (GAN) is used for reverse design. GAN consists of two parts: a generator and a discriminator. The generator first receives a noise vector and a two-dimensional image converted by GADF as input. Then, after multiple layers of convolution operations, a fake unit image is generated. The discriminator receives the fake unit image and compares it with the real unit image in the sample set. The network guides the parameter update of the two modules by calculating the loss function of the generator and the discriminator until the fake image generated by the generator becomes more and more realistic, so that the discriminator cannot distinguish between the real and the fake, and finally the network converges.

[0068] See also Fig. 9, showing the overall structure of GAN. First, a noise vector of size 128×128×1 and a two-dimensional image converted by GADF are taken as the input of the generator. The noise vector is used to introduce randomness, and the two-dimensional image converted by GADF provides the initial feature information. The input data is first fused in the generator through an addition operation. Then, the fused data passes through the first convolution layer, with a convolution kernel size k of 4, a number n of 64, and a stride s of 2, denoted as k4n64s2. After the convolution operation, the data passes through a batch normalization layer (BN) to standardize the feature map, and then nonlinearity is introduced through the ReLU activation function. Subsequently, the data enters the downsampling process, passing through multiple layers of convolution and batch normalization layers. The specific parameters of each layer are: the second layer has a convolution kernel size of 4, a number of 128, and a stride of 2 (k4n128s2); the third layer has a convolution kernel size of 4, a number of 256, and a stride of 2 (k4n256s2); the fourth layer has a convolution kernel size of 4, a number of 512, and a stride of 2 (k4n512s2); the fifth layer has a convolution kernel size of 4, a number of 1024, and a stride of 2 (k4n1024s2). Each layer is processed by the ReLU activation function to gradually reduce the size of the feature map and extract higher-level features. The network then enters the upsampling stage after a symmetric transformation operation, and directly transfers the input features to the output stage through jump connections to enhance feature transfer. Finally, through the deconvolution operation, the feature map dimension is enlarged to the target unit pattern (16×16×2) to generate the final output image.

[0069] The fake unit images, real unit images and two-dimensional images generated by the generator are input into the discriminator together. The discriminator first performs a series of convolution and downsampling operations on the input image to extract image features. The specific convolution parameters are: the first layer convolution kernel size is 4, the number is 64, and the stride is 2 (k4n64s2); the second layer convolution kernel size is 4, the number is 128, and the stride is 2 (k4n128s2); the third layer convolution kernel size is 4, the number is 256, and the stride is 2 (k4n256s2); the fourth layer convolution kernel size is 4, the number is 512, and the stride is 2 (k4n512s2). Each layer of convolution is processed by batch normalization and ReLU activation function. After all convolution operations are completed, the discriminator flattens the feature map into a one-dimensional vector, and outputs the probability of the image being true or false through the Sigmoid activation function. By calculating the generator loss (L G ) and the discriminator loss (L D ), the network updates the weights through back propagation to achieve high-quality image generation and discrimination.

[0070] In this embodiment, the PyTorch deep learning framework is used. The learning rate is 1e-4, and the optimizer uses the default Adam optimizer of PyTorch. The network performs 1500 rounds of iterations, and the batch size of each round is 64. The hardware used by the network is the Nvidia RTX 4080 graphics card. Fig.10 The figure shows the loss curves of the generator and the discriminator. It can be seen that in the early stage of training, the generator loss L G and the discriminator loss L D They are around 200 and -80 respectively. With the iteration of training, the generator gradually shows the ability to generate high-quality samples. Its loss and the discriminator loss finally stabilize at around 80 and -30 respectively. The network converges and the training ends.

[0071] After the training is completed, the unit image generator can be used to reversely design a four-channel full-domain electromagnetic metasurface. The functions implemented by the four-channel full-domain electromagnetic metasurface in this embodiment are as follows: Fig.11 As shown, under the excitation of x-line polarized plane waves, the four-channel full-domain electromagnetic metasurface projects a holographic image "AI" working at 8 GHz and a second-order OAM vortex wave working at 12 GHz in the backward space; under the excitation of y-line polarized plane waves, the four-channel full-domain electromagnetic metasurface projects a first-order OAM vortex wave working at 8 GHz and a holographic image "XD" working at 12 GHz in the forward space.

[0072] In a possible embodiment, the method of obtaining a phase profile of a four-channel full-domain electromagnetic metasurface within a target frequency band includes: obtaining an x-line polarized holographic image, orbital angular momentum vortex wave, and a target frequency band, as well as a y-line polarized holographic image, orbital angular momentum vortex wave, and a target frequency band of the four-channel full-domain electromagnetic metasurface; based on the x-line polarized and y-line polarized holographic images, orbital angular momentum vortex wave, and target frequency band, according to the diffraction propagation formula of orbital angular momentum vortex wave and electromagnetic wave, obtaining an ideal phase distribution diagram of x-line polarization and an ideal phase distribution diagram of y-line polarization; arranging the ideal phase distribution diagram of x-line polarization and the ideal phase distribution diagram of y-line polarization in an interlaced manner in the horizontal and vertical directions according to a preset step size, to obtain a phase profile of the four-channel full-domain electromagnetic metasurface within the target frequency band.

[0073] Specifically, the design process of a four-channel full-domain electromagnetic metasurface integrating OAM and holographic imaging is as follows: Fig.12 First, the total number of units of the four-channel full-area electromagnetic metasurface array is determined to be 40×40, covering a total area of ​​400×400mm 2 Then, the ideal phase distribution diagram of the first-order OAM wave and hologram “AI” operating at 8 GHz is obtained through theoretical calculation, as shown in Fig.12 As shown in Figure (a) and Figure (b).

[0074] For vortex waves, the spiral phase distribution can be obtained by the following formula:

[0075]

[0076] Among them, x mn and y mn are the unit coordinate positions corresponding to the mth row and nth column, and l is the order of the OAM wave.

[0077] For holograms, the calculation process is as follows:

[0078] First, a random phase φ(x,y) distribution is initialized to form a wavefront function f(x,y)=exp(jφ(x,y)), and then the wavefront function is converted to the frequency domain using the fast Fourier transform FFT to obtain g(x,y):

[0079]

[0080] in, and represent fast Fourier transform and inverse fast Fourier transform respectively; λ is the wavelength; f x and f y is the spatial frequency component; z is the imaging distance.

[0081] In this embodiment, the imaging distance of the two holographic images is set to 180 mm. In the frequency domain, the amplitude |G(x,y)| of the target holographic image is combined with the phase g(u,v) to form an intermediate function:

[0082]

[0083] Perform an inverse Fourier transform on the intermediate function and convert it back to the spatial domain to obtain the updated wavefront function:

[0084]

[0085] The initial phase φ(x, y) is replaced by the phase of the updated wavefront function f′(x, y), and the above process is repeated until f′(x, y) meets the requirements of the desired hologram.

[0086] The same method can be used to obtain the ideal phase distribution diagram of the second-order OAM wave and hologram "XD" operating at 12 GHz, such as Fig.12 As shown in Figure (c) and Figure (d).

[0087] Finally, the two frequency types of units are arranged alternately, and the corresponding upper and lower layers of phase are also interlaced and fused as shown in the figure. Fig.12As shown in Figures (e) and (f), the units at the two frequencies form a chessboard-like alternating layout on the same array plane to obtain the final phase profile and achieve the best distribution effect.

[0088] Then, according to the above phase profile, the trained unit image generator is used to generate the corresponding unit structure for each unit position in the array in turn. Specifically, the electromagnetic response at each position calculated theoretically, that is, the electromagnetic frequency response curve, is converted by GADF to obtain a two-dimensional image as the input of the unit image generator. After the data passes through multiple operation modules such as convolution, batch normalization, activation function, jump connection and transposition, a unit image representing the unit structure is output. After the unit image output at all positions is completed, the entire structure is full-wave simulated in CST to verify its electromagnetic performance under dual-frequency dual-polarization excitation.

[0089] In yet another embodiment of the present invention, a four-channel full-area electromagnetic metasurface is provided, and the unit structure of each unit of the four-channel full-area electromagnetic metasurface is designed using the above-mentioned four-channel full-area electromagnetic metasurface design method.

[0090] The present invention is further described below in conjunction with simulation results:

[0091] like Fig.13 and Fig.14 As shown in the figure, the normalized radiation patterns of the 8 GHz and 12 GHz OAM vortex waves are expressed in polar coordinates. The amplitude zero depth at 8 GHz is -14 dB, corresponding to an angle of 2°; the amplitude zero depth at 12 GHz is -22 dB, corresponding to an angle of -5°.

[0092] The electric field amplitude and phase distribution of the first-order OAM vortex wave in near-field simulation are shown in Fig.15 Figure (a) and Fig.15 In Figure (b), it can be seen that the energy zero point is in the middle of the amplitude distribution corresponding to 8 GHz, and the phase is in the form of a first-order spiral; the electric field amplitude and phase distribution of the second-order OAM vortex wave simulated in the near field are shown in Fig.16 Figure (a) and Fig.16 In Figure (b), the amplitude corresponding to 12 GHz also shows a hollow characteristic, and the phase is in the form of a second-order spiral. Therefore, the designed four-channel global electromagnetic metasurface can generate OAM vortex electromagnetic waves with distinct characteristics and different modes at different frequencies.

[0093] like Fig.17As shown in the figure, holographic images at different diffraction distances were extracted through near-field simulation. The backward area (z<0) is the word "AI" working at 8GHz, and the forward area (z>0) is the word "XD" working at 12GHz. It can be seen that at the theoretical value z=±180mm, the two holographic images can basically be clearly distinguished. At the positions of z=±160mm and z=±200mm, which deviate from the theoretical value, the imaging quality is slightly reduced, but the overall outline of the letters can also be clearly distinguished. Therefore, the designed four-channel full-domain electromagnetic metasurface can generate different holographic images at different frequency points.

[0094] The four-channel full-domain electromagnetic metasurface design method of the present invention has the following characteristics:

[0095] 1) Data pre-screening improves model accuracy: By manually screening the metal ratio of the unit, we ensure that the generated training data set covers the full phase range of 2π. This pre-screening process effectively eliminates low-quality data, reduces the waste of computing resources, enables the deep learning model to converge faster, and accurately learns key electromagnetic response characteristics during the training process, improving design efficiency and accuracy.

[0096] 2) Solve the dimensionality mismatch problem: The Gram angle difference field is introduced into the generative adversarial network, and the dimensionality mismatch problem between the electromagnetic response and the unit structure is solved by converting the one-dimensional electromagnetic response curve into a two-dimensional image. The model can process these data in the form of image recognition, especially in the convolution processing, it can effectively extract spatial features, patterns and local relationships, which improves the intuitiveness and effectiveness of the reverse design process.

[0097] 3) Save computing resources: Through data pre-screening and image processing, the consumption of computing resources during the training process is effectively reduced. High-quality data sets and efficient model training processes shorten the design cycle, reduce costs, and significantly improve resource utilization.

[0098] 4) Enhanced model generalization ability: During the model training process, the pre-screened high-quality data set and effective feature extraction methods enable the model to have stronger generalization ability. The model can adapt to more diverse design requirements and application scenarios, improving the flexibility and adaptability in practical applications.

[0099] The following are device embodiments of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0100] See also Fig.18In another embodiment of the present invention, a four-channel full-domain electromagnetic metasurface design system is provided, which can be used to implement the above-mentioned four-channel full-domain electromagnetic metasurface design method. Specifically, the four-channel full-domain electromagnetic metasurface design system includes a data acquisition module, a two-dimensional conversion module and a structure determination module.

[0101] Among them, the data acquisition module is used to obtain the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band; the two-dimensional conversion module is used to obtain the electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band according to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and convert the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band; the structure determination module is used to generate a unit image of each unit through a pre-trained unit image generator according to the two-dimensional image of each unit in the target frequency band, and obtain the unit structure of each unit according to the unit image of each unit; Among them, The pre-trained unit image generator is obtained in the following manner: obtaining unit images of a number of unit structure samples, and obtaining electromagnetic frequency response curves of a number of unit structure samples within a target frequency band, and converting the electromagnetic frequency response curves of the number of unit structure samples within the target frequency band into two-dimensional images of the number of unit structure samples within the target frequency band; based on the unit images of the number of unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as the pre-trained unit image generator.

[0102] All relevant contents of each step involved in the aforementioned embodiment of the four-channel full-domain electromagnetic metasurface design method can be referred to the functional description of the functional modules corresponding to the four-channel full-domain electromagnetic metasurface design system in the embodiment of the present invention, and will not be repeated here.

[0103] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0104] In another embodiment of the present invention, a computer device is provided, the computer device including a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the four-channel global electromagnetic metasurface design method.

[0105] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both a built-in storage medium in a computer device and an extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space that stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the four-channel global electromagnetic metasurface design method in the above embodiment.

[0106] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0107] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A four-channel full-domain electromagnetic metasurface design method, characterized in that: include: Obtain the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band; According to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, the electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band is obtained, and the electromagnetic frequency response curve of each unit in the target frequency band is converted into a two-dimensional image of each unit in the target frequency band; According to the two-dimensional image of each unit in the target frequency band, a unit image of each unit is generated by a pre-trained unit image generator, and a unit structure of each unit is obtained according to the unit image of each unit; Among them, the pre-trained unit image generator is obtained by the following method: The unit images of several unit structure samples are obtained, as well as the electromagnetic frequency response curves of several unit structure samples within a target frequency band, and the electromagnetic frequency response curves of several unit structure samples within the target frequency band are converted into two-dimensional images of several unit structure samples within the target frequency band; based on the unit images of several unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as a pre-trained unit image generator.

2. The four-channel full-domain electromagnetic metasurface design method according to claim 1 is characterized in that: The unit structure of the unit structure sample is a mirror symmetric structure or a rotational symmetric structure.

3. The four-channel full-domain electromagnetic metasurface design method according to claim 1 is characterized in that: Acquiring unit images of a plurality of unit structure samples includes: dividing each unit structure sample into a plurality of grids, setting each grid in a manner of metal as 1 and medium as 2 to form a Boolean matrix, obtaining a two-dimensional matrix of the unit structure sample and converting it into a binary image, thereby obtaining a unit image of each unit structure sample; The obtaining of electromagnetic frequency response curves of the plurality of unit structure samples within the target frequency band includes: performing electromagnetic simulation on the plurality of unit structure samples within the target frequency band by means of simulation software to obtain electromagnetic frequency response curves of the plurality of unit structure samples within the target frequency band.

4. The four-channel full-domain electromagnetic metasurface design method according to claim 1 is characterized in that: After obtaining the electromagnetic frequency response curves of several unit structure samples in the target frequency band, the unit structure samples whose phase change in the target frequency band is less than the phase change threshold are eliminated according to the electromagnetic frequency response curves of each unit structure sample in the target frequency band.

5. The four-channel full-domain electromagnetic metasurface design method according to claim 1 is characterized in that: The converting of the electromagnetic frequency response curve of each unit in the target frequency band into the two-dimensional image of each unit in the target frequency band includes: converting the electromagnetic frequency response curve of each unit in the target frequency band into the two-dimensional image of each unit in the target frequency band by using the Gram angle difference field technology; the converting of the electromagnetic frequency response curves of a plurality of unit structure samples in the target frequency band into the two-dimensional image of a plurality of unit structure samples in the target frequency band includes: converting the electromagnetic frequency response curves of a plurality of unit structure samples in the target frequency band into the two-dimensional image of a plurality of unit structure samples in the target frequency band by using the Gram angle difference field technology.

6. The four-channel full-domain electromagnetic metasurface design method according to claim 1 is characterized in that: The step of obtaining a phase profile of a four-channel global electromagnetic metasurface within a target frequency band comprises: Obtaining the x-polarized holographic image, orbital angular momentum vortex wave and target frequency band of the four-channel full-domain electromagnetic metasurface, as well as the y-polarized holographic image, orbital angular momentum vortex wave and target frequency band; Based on the holographic images of x-line polarization and y-line polarization, orbital angular momentum vortex waves and target frequency bands, and according to the diffraction propagation formula of orbital angular momentum vortex waves and electromagnetic waves, the ideal phase distribution diagram of x-line polarization and the ideal phase distribution diagram of y-line polarization are obtained; The ideal phase distribution diagram of x-line polarization and the ideal phase distribution diagram of y-line polarization are alternately arranged in the horizontal and vertical directions according to a preset step distance to obtain the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band.

7. A four-channel full-domain electromagnetic metasurface design system, characterized in that: include: A data acquisition module, used to obtain a phase profile of a four-channel full-domain electromagnetic metasurface within a target frequency band; A two-dimensional conversion module is used to obtain an electromagnetic frequency response curve of each unit of the four-channel full-domain electromagnetic metasurface in the target frequency band according to the phase profile of the four-channel full-domain electromagnetic metasurface in the target frequency band, and convert the electromagnetic frequency response curve of each unit in the target frequency band into a two-dimensional image of each unit in the target frequency band; The structure determination module is used to generate a unit image of each unit through a pre-trained unit image generator according to the two-dimensional image of each unit in the target frequency band, and obtain the unit structure of each unit according to the unit image of each unit; wherein the pre-trained unit image generator is obtained by the following method: The unit images of several unit structure samples are obtained, as well as the electromagnetic frequency response curves of several unit structure samples within a target frequency band, and the electromagnetic frequency response curves of several unit structure samples within the target frequency band are converted into two-dimensional images of several unit structure samples within the target frequency band; based on the unit images of several unit structure samples and the two-dimensional images within the target frequency band, a preset generative adversarial network model is trained with the two-dimensional images within the target frequency band as input and the unit images as output, and the generator of the generative adversarial network model after the training is used as a pre-trained unit image generator.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the four-channel full-domain electromagnetic metasurface design method as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the four-channel full-domain electromagnetic metasurface design method as described in any one of claims 1 to 6 are implemented.

10. A four-channel full-range electromagnetic metasurface, characterized in that: The unit structure of each unit of the four-channel full-area electromagnetic metasurface is designed using the four-channel full-area electromagnetic metasurface design method described in any one of claims 1 to 6.

Images