A high-gain metasurface beamwave antenna array and its reverse design method

Abstract

This invention discloses a high-gain metasurface beamwave antenna array and its reverse design method, comprising a metasurface layer, a first dielectric layer, a slotted ground layer, a second dielectric layer, and a feed layer. The metasurface layer includes several antenna elements, each composed of several patch elements. The central patch element located in the central region of the antenna element and the peripheral patch elements located in the outer region of the antenna element are both based on rectangular patches. Chamfered regions are set on both sides of the rectangular patch to obtain an asymmetrical central patch element. A diamond-shaped slot is set in the center of the rectangular patch to obtain the peripheral patch element. In this invention, the chamfered region design of the central patch element breaks the structural symmetry of the traditional rectangular metasurface, significantly enhancing the surface current in the vertical direction. The diamond-shaped slot of the peripheral patch element further strengthens the current distribution in the vertical direction, resulting in a significant optimization of the radiation characteristics of the metasurface.

Description

Technical Field

[0001] This invention relates to the field of antenna design, and in particular to a high-gain metasurface beamwave antenna array and its reverse design method. Background Technology

[0002] As a core component of radio systems, the antenna's transmission efficiency and performance directly determine the reliability of the communication system. With advancements in radar imaging and modern wireless communication technologies, complex electromagnetic environments place higher demands on antenna design, particularly in terms of high gain and beamforming performance. Beamforming antennas, by controlling radiation directivity to improve signal transmission efficiency and directionality, are widely used in radar, satellite communication, and wireless sensor networks.

[0003] A metasurface (MTS) is a two-dimensional periodic or quasi-periodic planar array composed of subwavelength-scale artificial microstructure units. It allows for precise control of the phase, amplitude, and polarization of spatial electromagnetic waves by flexibly adjusting the geometry, size, and spatial arrangement of these microstructure units. Compared to traditional structures, metasurfaces not only possess powerful electromagnetic wavefront manipulation capabilities but also offer significant advantages such as low profile, ease of planar integration, and low manufacturing costs, providing a highly promising solution for achieving miniaturized, high-gain, and efficient directional beam radiation in antennas.

[0004] Existing metasurface antenna designs have several shortcomings. For example, invention patent CN118137116A discloses a substrate integrated cavity metasurface antenna. This antenna combines a substrate integrated cavity with a metasurface structure to achieve multimode resonance and uses a SIW (Solidated In-line Wave) as the feed structure, thereby improving the antenna's bandwidth and gain. However, the structure of an upper substrate integrated cavity and a lower SIW increases the antenna's complexity, and the symmetrical rectangular metasurface structure easily excites multiple modes, leading to a coupled radiation dead zone in the radiation pattern. In other designs, although the radiation dead zone is reduced by using certain asymmetric metasurface structures, the excitation of multiple characteristic modes of the metasurface results in poor beamforming, making it difficult to meet the stringent requirements of high-precision radar imaging for high resolution and strong anti-interference capabilities.

[0005] Furthermore, the forward design of traditional metasurface antennas relies heavily on human experience, and full-wave simulation parameter scanning is time-consuming and prone to getting trapped in local optima. Although some deep learning models are gradually being applied to the intelligent design of antennas, existing models often suffer from problems such as insufficient multi-scale feature extraction, weak global perception, slow convergence, and low inversion accuracy when dealing with the nonlinear mapping between target S-parameters and complex structures. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of the prior art by providing a high-gain metasurface beamwave antenna array and its reverse design method.

[0007] The objective of this invention is achieved through the following technical solution: a high-gain metasurface beamwave antenna array, comprising, from top to bottom, a metasurface layer, a first dielectric layer, a slotted ground layer, a second dielectric layer, and a feed layer. The metasurface layer includes several antenna elements arranged in a rectangular array, each antenna element consisting of several patch elements arranged in a rectangular array. The center patch element located in the central region of the antenna element and the peripheral patch elements located in the outer region of the antenna element are both based on rectangular patches. Chamfered areas are set on both sides of the rectangular patches to obtain an asymmetrical structure for the center patch element. A diamond-shaped slot is set in the center of the rectangular patch to obtain the peripheral patch element. The long diagonal of the diamond-shaped slot points to the center of the antenna element.

[0008] The slotted grounding layer is a grounding metal layer with cross-shaped slots etched on it. The cross-shaped slots on the slotted grounding layer correspond one-to-one with the patch groups on the metasurface layer. The feed layer is equipped with a microstrip feed network with a fixed resistance value. After the radio frequency signal is input, it is fed into the antenna element with equal amplitude and in phase through the microstrip feed network.

[0009] Preferably, the chamfered areas on both sides of the central patch unit are right-angled trapezoids to give the central patch unit a Z-shaped structure.

[0010] Preferably, the side length of the rectangular patch and the spacing between adjacent patch units satisfy the following formula: 3 W1+2 W2=2.0λ; In the formula, λ is the wavelength of the antenna array at the corresponding operating frequency, W1 is the side length of the rectangular patch, and W2 is the spacing between adjacent patch units.

[0011] Preferably, the microstrip feed network includes multiple feed ends, the width of which is greater than the width of its front section; the feed ends correspond to the center of the cross-shaped slot on the slotted grounding layer.

[0012] Preferably, the microstrip power supply network includes multiple power dividers, each power divider having two power supply branches. The subsequent power divider is located at the end of the power supply branch of the preceding power divider, and the end of the power supply branch of the last power divider is the power supply end.

[0013] Preferably, the cross-shaped slot consists of two perpendicularly intersecting rectangular slots; the center of the cross-shaped slot is aligned with the center of the antenna element.

[0014] Preferably, the center-to-center spacing of the adjacent antenna elements is greater than half the wavelength of the antenna array at the corresponding operating frequency.

[0015] Preferably, the rectangular patch has a side length of 8.9 mm, a spacing of 0.7 mm between adjacent patch units, a height of 6 mm for the chamfered area, and upper and lower bottom edges of the chamfered area of ​​1.5 mm and 2.6 mm, respectively; the long diagonal of the diamond-shaped slit is 5.0 mm; the width of the rectangular slit of the cross-shaped slit is 0.8 mm, and the length of the rectangular slit of the cross-shaped slit is 34 mm.

[0016] Preferably, the first dielectric layer is a Rogers RT3003 dielectric substrate with a thickness of 2.28 mm, and the second dielectric layer is a Rogers RT3003 dielectric substrate with a thickness of 0.76 mm; the metasurface layer, the slotted ground layer, and the power supply layer are all made of copper.

[0017] A reverse design method for a high-gain metasurface beamwave antenna array is disclosed. This method utilizes a 1D multi-scale CNN network model to achieve the reverse mapping from the target scattering parameter sequence to the key structural parameters of the antenna array. The 1D multi-scale CNN network model comprises an initial feature extraction module, a multi-scale parallel convolution module, a three-channel attention mechanism module, and a parameter prediction output layer, connected sequentially. The specific method is as follows: S1: Input the target scattering parameter sequence into the initial feature extraction module of the pre-trained 1D multi-scale CNN network model to perform preliminary feature extraction. The extracted preliminary features are then subjected to batch normalization and max pooling to achieve feature dimensionality reduction and obtain the basic electromagnetic features. S2: Input the basic electromagnetic features into the multi-scale parallel convolution module of the 1D multi-scale CNN network model. Through the parallel convolution branches of RefConv convolution kernels of different sizes, local features, mesoscale features and global context features are extracted respectively. After feature fusion, multi-scale electromagnetic features are obtained. S3: Input multi-scale electromagnetic features into the three-channel attention mechanism module of the 1D multi-scale CNN network model, and process them sequentially through the spatial attention layer, channel attention layer and non-local attention layer to enhance the perception of global and key features, thereby obtaining optimized features; S4: Input the optimized features into the parameter prediction output layer of the 1D multi-scale CNN network model, and output the key structural parameters after gradual dimensionality reduction through four fully connected networks; the key structural parameters include the side length of the rectangular patch, the spacing between patch units, and the length and width of the rectangular gap in the cross-shaped gap.

[0018] The beneficial effects of this invention are: 1. In this invention, the chamfered region design of the central patch unit breaks the structural symmetry of the traditional rectangular metasurface, significantly enhancing the surface current in the vertical direction while suppressing the surface current in the horizontal direction, thereby fundamentally strengthening the excitation efficiency of the vertical electric field. Furthermore, the rhomboid slots of the outer patch units, with their long diagonals pointing towards the center of the antenna unit, further enhance the current distribution in the vertical direction, significantly optimizing the radiation characteristics of the metasurface and improving the radiation efficiency and directivity of the antenna array.

[0019] 2. The slotted grounding layer serves as both an upper and lower electromagnetic isolation layer and an energy coupling window. On the one hand, the slotted grounding layer can isolate the electromagnetic interference between the feed layer and the metasurface layer. On the other hand, the slotted grounding layer excites an alternating electric field in the vertical direction through the cross-shaped slot, so that the characteristic mode current on the metasurface is strictly controlled in the vertical direction.

[0020] 3. The microstrip feed network with fixed resistance value divides the input RF signal into four paths evenly, ensuring that the signals fed into each antenna element are of equal amplitude and in phase, so that the radiated electromagnetic waves of each element can be coherently superimposed in space, achieving a high gain effect. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structure of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0022] Figure 2 This is a schematic diagram of the metasurface layer structure of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0023] Figure 3 This is a cross-sectional view of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0024] Figure 4 This is a schematic diagram of the feed layer structure of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0025] Figure 5 This is a graph showing the modal significance of the metasurface antenna element of the high-gain metasurface beamwave antenna array designed for radar imaging in the present invention under metasurface modes 1–6.

[0026] Figure 6 This is a surface current distribution diagram of the metasurface antenna element of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention under metasurface modes 1–4.

[0027] Figure 7 This is the radiation pattern of the metasurface antenna element of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention in metasurface modes 1–4.

[0028] Figure 8 This is an electric field distribution diagram of the feeding structure of the metasurface antenna element of the high-gain metasurface beamwave antenna array for radar imaging and its reverse design method designed in this invention.

[0029] Figure 9 This invention describes the surface current distribution of the metasurface antenna element at 5.5 GHz and 5.9 GHz in the high-gain metasurface beamwave antenna array for radar imaging and its reverse design method.

[0030] Figure 10 This is the radiation pattern of the metasurface antenna element of the high-gain metasurface beamwave antenna array designed for radar imaging in this invention at 5.5 GHz and 5.9 GHz.

[0031] Figure 11 This is a framework diagram of a 1D multi-scale CNN network model involved in the reverse design method of this invention.

[0032] Figure 12 This is an ablation experiment result diagram based on the 1D multi-scale CNN network model of this invention.

[0033] Figure 13 The target S in the reverse design method of this invention 11 Curve and S predicted based on network model 11 Curve comparison chart.

[0034] Figure 14 This is a simulated return loss and gain diagram of the metasurface antenna element of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0035] Figure 15 This is a physical image of the metasurface structure of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0036] Figure 16 This is a schematic diagram of the feeding structure of the high-gain metasurface beamwave antenna array for radar imaging according to the present invention.

[0037] Figure 17 This is a comparison diagram of return loss and gain of the high-gain metasurface beamwave antenna array for radar imaging designed in this invention.

[0038] Figure 18 This invention presents the main polarization and cross-polarization radiation patterns of the high-gain metasurface beamwave antenna array designed for radar imaging at 5.5 GHz and 5.9 GHz in the xz and yz planes.

[0039] In the figure: 1. Metasurface layer, 1-1. Antenna element, 111. Patch element, 2. First dielectric layer, 3. Slotted ground layer, 31. Cross-shaped slot, 4. Second dielectric layer, 5. Feed layer, 51. Microstrip feed network, 52. First-stage power divider, 53. Second-stage power divider. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0041] Those skilled in the art should understand that, in the disclosure of this invention, the terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the above terms should not be construed as limiting this invention.

[0042] It is understood that the term "a" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "a" should not be understood as a limitation on the number.

[0043] like Figures 1 to 18 As shown, a high-gain metasurface beamwave antenna array comprises, from top to bottom, a metasurface layer 1, a first dielectric layer 2, a slotted ground layer 3, a second dielectric layer 4, and a feed layer 5. The metasurface layer 1 includes several antenna elements 1-1 arranged in a rectangular array, each antenna element 1-1 consisting of several patch elements 111 arranged in a rectangular array. The central patch element 111 located in the central region of the antenna element 1-1 and the peripheral patch elements 111 located in the peripheral region of the antenna element 1-1 are both based on rectangular patches as their basic structure, with features on both sides of the rectangular patches. A chamfered region is set to obtain a center patch unit 111 with an asymmetrical structure; a diamond-shaped slot is set in the center of the rectangular patch to obtain an outer patch unit 111; the long diagonal of the diamond-shaped slot points to the center of the antenna unit 1-1; the slotted ground layer 3 is a ground metal layer with cross-shaped slots 31 etched on it, and the cross-shaped slots 31 on the slotted ground layer 3 correspond one-to-one with the patch group on the metasurface layer 1; the feed layer 5 is provided with a microstrip feed network 51 with a fixed resistance value; after the radio frequency signal is input, it is fed into the antenna unit 1-1 with equal amplitude and in phase through the microstrip feed network 51.

[0044] In this invention, the chamfered region design of the central patch unit 111 breaks the structural symmetry of the traditional rectangular metasurface, significantly enhancing the surface current in the vertical direction while suppressing the surface current in the horizontal direction, thereby fundamentally strengthening the excitation efficiency of the vertical electric field. Furthermore, the rhomboid slots of the outer patch units 111, with their long diagonals pointing towards the center of the antenna unit 1-1, further enhance the current distribution in the vertical direction, significantly optimizing the radiation characteristics of the metasurface and improving the radiation efficiency and directivity of the antenna array.

[0045] The slotted grounding layer 3 serves as both upper and lower electromagnetic isolation and an energy coupling window. On the one hand, the slotted grounding layer 3 can isolate the electromagnetic interference between the feed layer 5 and the metasurface layer 1; on the other hand, the slotted grounding layer 3 excites an alternating electric field in the vertical direction through the cross-shaped slot 31, so that the characteristic mode current on the metasurface is strictly controlled in the vertical direction.

[0046] The microstrip feed network 51 with a fixed resistance value divides the input radio frequency signal into four paths, ensuring that the signals fed into each antenna element 1-1 are of equal amplitude and in phase, so that the radiated electromagnetic waves of each element can be coherently superimposed in space, achieving a high gain effect.

[0047] To elicit excellent modal saliency in the target frequency band, the side length of the rectangular patch and the spacing between adjacent patch units 111 satisfy the following formula: 3 W1+2 W2=2.0λ; In the formula, λ is the wavelength of the antenna array at the corresponding operating frequency, W1 is the side length of the rectangular patch, and W2 is the spacing between adjacent patch units.

[0048] The chamfered areas on both sides of the central patch unit 111 are right-angled trapezoids to give the central patch unit 111 a Z-shaped structure. Specifically, the right-angled trapezoidal chamfered area on one side is located at the upper right corner of the rectangular patch, and the right-angled trapezoidal chamfered area on the other side is located at the lower left corner of the rectangular patch. This structure can precisely disrupt the current symmetry in the horizontal direction while preserving the continuity of the current in the vertical direction to the greatest extent.

[0049] like Figure 2 As shown, in this embodiment, antenna element 1-1 consists of 16 patch elements 111 arranged in a 4×4 pattern. The four central patch elements 111 located in the central region have chamfered areas on both sides, forming an asymmetrical structure. The twelve peripheral patch elements 111 located in the outer region have diamond-shaped gaps at their centers. The long diagonal of the diamond-shaped gaps in the twelve peripheral patch elements 111 all point towards the center of antenna element 1-1.

[0050] In this embodiment, the rectangular patch has a side length of 8.9 mm, the spacing between adjacent patch units 111 is 0.7 mm, the height of the chamfered area is 6 mm, and the upper and lower bottom edges of the chamfered area are 1.5 mm and 2.6 mm, respectively. The long diagonal of the rhomboid slit is 5.0 mm.

[0051] The chamfered areas on both sides of the rectangular patch and the diamond-shaped gap in the center can be obtained through etching.

[0052] In this embodiment, the first dielectric layer 2 is a Rogers RT3003 dielectric substrate with a thickness of 2.28 mm. The first dielectric layer 2 has low loss characteristics and is used to support the metasurface layer 1.

[0053] In this embodiment, the metasurface layer 1 is made of copper.

[0054] Figure 5 The modal saliency of characteristic modes 1–6 of this metasurface structure in the target frequency band is presented. These modes all exhibit high saliency and can effectively contribute radiated energy within the antenna's operating frequency band. See [link to relevant documentation] Figure 6 , Figure 6 Showing Figure 5 The surface current distribution of the first four characteristic modes shows that the surface current in the horizontal direction of the patch unit 111 is effectively suppressed.

[0055] The radiation patterns of the first four characteristic modes of the metasurface of this invention are as follows: Figure 7 As shown, modes 2 and 3 have significant radiation blind zones in the main radiation direction and should not be excited in practical applications. In contrast, modes 1 and 4 exhibit better beamforming effects, concentrating radiated energy in a specific direction, thereby effectively improving the directionality and gain of the signal. This makes modes 1 and 4 very suitable for radiation applications in the target frequency band.

[0056] The present invention, through the combined application of the central patch unit 111 (with a chamfered region and asymmetric structure) and the peripheral patch unit 111 (with a diamond-shaped gap), enables the metasurface to achieve high gain and optimized radiation directionality while minimizing the excitation of invalid modes.

[0057] The slotted ground layer 3, the second dielectric layer 4, and the feed layer 5 together constitute the microstrip slot feed structure. The microstrip slot feed structure is used to excite the target characteristic modes of the metasurface to form the operating bandwidth of the antenna.

[0058] like Figure 4 and Figure 16 As shown, the microstrip feed network 51 disposed on the feed layer 5 includes multiple feed ends, the width of which is greater than the width of its front section; the feed ends correspond to the center of the cross-shaped slot 31 on the slotted grounding layer 3 respectively.

[0059] Furthermore, the microstrip power supply network 51 includes a multi-stage power divider, each stage of which includes two power supply branches. The subsequent power divider is located at the end of the power supply branch of the preceding power divider, and the end of the power supply branch of the last stage power divider is the power supply end.

[0060] In this embodiment, the antenna array includes 2×2 arranged antenna elements 1-1 (a total of 4 antenna elements 1-1). Under this array, the microstrip feed network 51 includes a first-stage power divider 52 and two second-stage power dividers 53. The first-stage power divider 52 branches out two feed branches, and the two second-stage power dividers 53 are respectively connected to the ends of the two feed branches branched out by the first-stage power divider 52. The two second-stage power dividers 53 each branch out two feed branches, forming a total of four feed ends. The four feed ends correspond to the four antenna elements 1-1 respectively. The input radio frequency signal is fed into the four antenna elements 1-1 with equal amplitude and in phase through the four feed ends, thereby realizing the efficient in-phase superposition of electromagnetic waves in space.

[0061] The microstrip feed network 51 can be manufactured using laser engraving technology, which maintains a compact physical structure while ensuring processing accuracy.

[0062] In this embodiment, the characteristic impedance of the microstrip feed network 51 is 50Ω. The input terminal of the microstrip feed network 51 is connected to an SMA microwave high-frequency connector. After the radio frequency signal is input through the SMA microwave high-frequency connector, it is fed into the antenna element 1-1 through the microstrip feed network 51 with equal amplitude and in phase.

[0063] Specifically, the cross-shaped slot 31 consists of two perpendicularly intersecting rectangular slots; the center of the cross-shaped slot 31 is aligned with the center of the antenna element 1-1. In this embodiment, the material of the slotted grounding layer 3 is copper, the width of the rectangular slot of the cross-shaped slot 31 is 0.8mm, and the length of the rectangular slot of the cross-shaped slot 31 is 34mm.

[0064] In this embodiment, the second dielectric layer 4 is a Rogers RT3003 dielectric substrate with a thickness of 0.76 mm, which is used to support the bottom microstrip power supply network 51 and ensure low-loss transmission of energy of the microstrip power supply network 51.

[0065] In this embodiment, the feed layer 5 is made of copper.

[0066] The design of the microstrip slot feed structure directly determines the excitation efficiency of metasurface characteristic modes and the final radiation performance of the antenna. To achieve precise control and efficient excitation of the target mode, this invention employs a microstrip slot feed structure with etched cross-shaped slots 31.

[0067] like Figure 8As shown in the electric field distribution, the cross-shaped slot designed in this invention possesses extremely strong electric field concentration capabilities, and the excited electric field modes exhibit alternating electric fields in the vertical direction. Coupled by this alternating electric field, the characteristic mode current excited on the metasurface is strictly controlled in the vertical direction. This design effectively filters out interference modes and accurately excites the desired target characteristic modes within the target frequency band, thereby endowing the antenna array with excellent gain and directivity. The above excitation mechanism also yields the surface current distribution (…). Figure 6 This is a valid confirmation.

[0068] Figure 9 The surface current distribution of the antenna element on the metasurface is demonstrated. Under excitation by the cross-shaped slot, the surface current of the metasurface exhibits a highly consistent vertical distribution at both 5.5 GHz and 5.9 GHz frequencies. This phenomenon fully confirms that the core radiation mode (Mode 1) of the metasurface is precisely excited. This ideal vertical current distribution is highly consistent with the expected metasurface structure design, fundamentally optimizing the antenna's gain, directivity, and beamforming effect, ensuring efficient energy transfer and high-quality radiation within the target frequency band.

[0069] Figure 10 The radiation pattern of the antenna element is further illustrated. It can be clearly observed from the figure that the antenna element exhibits extremely excellent radiation directivity at the two key frequency points of 5.5 GHz and 5.9 GHz. The radiated energy is highly concentrated in the main radiation direction, while sidelobe and other directional radiation interference are significantly suppressed. Figure 10 The effectiveness of the metasurface design and feeding structure matching was effectively verified, demonstrating that the antenna successfully achieved ideal beam focusing and obtained high gain. This highly optimized directional radiation mode fully meets the application requirements of high-performance communication and radar imaging systems for high gain, high directivity, and strong anti-interference capabilities.

[0070] The overall arrangement and physical structure of the antenna element array in this embodiment are as follows: Figure 15 As shown in the figure. In this embodiment, the four antenna elements are arranged in a 2×2 matrix. To effectively reduce the electromagnetic coupling effect between adjacent elements and ensure a good array radiation pattern, the center-to-center distance between adjacent antenna elements is greater than half the wavelength of the antenna array at the corresponding operating frequency. In this embodiment, the center-to-center distance between adjacent antenna elements is set to 46mm.

[0071] Simulated and measured return loss S of metasurface antenna array 11 And the gain curve is as follows Figure 17As shown in the figure. Experimental results demonstrate that this antenna array exhibits excellent impedance matching characteristics over a wide bandwidth of 5.2–6.4 GHz. Within this operating frequency band, the measured gain of the antenna array reaches 13.4–16.2 ​​dBi, with a peak gain as high as 16.2 dBi. The measured data are in high agreement with the simulation results, and the antenna exhibits significant beam focusing effect and excellent directional radiation capability within the target frequency band. The main polarization and cross-polarization radiation patterns of the antenna array in the xz and yz planes at 5.5 GHz and 5.9 GHz are shown in the figure. Figure 18 As shown, the array exhibits highly concentrated main lobe energy at key frequencies, while side lobes and cross-polarization levels are significantly suppressed, demonstrating excellent beamforming and polarization isolation. These results fully demonstrate that the antenna array is compact, stable, and perfectly meets the stringent application requirements of modern high-performance systems such as radar imaging, satellite communication, and high-speed wireless networks for long-distance transmission, high-quality signal coverage, and strong anti-interference capabilities.

[0072] A reverse design method for a high-gain metasurface beamwave antenna array is disclosed. This method utilizes a 1D multi-scale CNN network model to achieve the reverse mapping from the target scattering parameter sequence to the key structural parameters of the antenna array. The 1D multi-scale CNN network model comprises an initial feature extraction module, a multi-scale parallel convolution module, a three-channel attention mechanism module, and a parameter prediction output layer, connected sequentially. The specific method is as follows: S1: Input the target scattering parameter sequence into the initial feature extraction module of the pre-trained 1D multi-scale CNN network model to perform preliminary feature extraction. The extracted preliminary features are then subjected to batch normalization and max pooling to achieve feature dimensionality reduction, resulting in basic electromagnetic features.

[0073] S2: Input the basic electromagnetic features into the multi-scale parallel convolution module of the 1D multi-scale CNN network model. Through the parallel convolution branches of RefConv convolution kernels of different sizes, local features, mesoscale features and global context features are extracted respectively. After feature fusion, multi-scale electromagnetic features are obtained. S3: Input multi-scale electromagnetic features into the three-channel attention mechanism module of the 1D multi-scale CNN network model, and process them sequentially through the spatial attention layer, channel attention layer and non-local attention layer to enhance the perception of global and key features, thereby obtaining optimized features; S4: Input the optimized features into the parameter prediction output layer of the 1D multi-scale CNN network model, and output the key structural parameters after gradual dimensionality reduction through four fully connected networks; the key structural parameters include the side length of the rectangular patch, the spacing between patch units, and the length and width of the rectangular gap in the cross-shaped gap.

[0074] The framework of the 1D multi-scale CNN network model involved in this method is as follows: Figure 11As shown, the model adopts an end-to-end architecture, which includes an initial feature extraction module, a multi-scale parallel convolution module, a three-channel attention mechanism module, and a parameter prediction output layer.

[0075] The initial feature extraction module uses 1D convolutional kernels to perform preliminary processing on the input S-parameter sequence (target scattering parameter sequence) of length 201. It achieves feature dimensionality reduction through batch normalization and max pooling, and extracts basic electromagnetic features. The multi-scale parallel convolution module (RefConv) innovatively introduces a multi-branch RefConv structure, which includes 3×1 (local features), 5×1 (mesoscale features), and 7×1 (global context) parallel convolutional branches. This module can mine resonant point information in the S-parameter curve from multiple dimensions, significantly improving feature extraction efficiency.

[0076] The three-channel attention mechanism module: Based on channel and spatial attention, a non-local attention layer is connected in series. The non-local mechanism captures long-range dependencies by calculating the correlation between any two positions within the sequence, enhancing the network's ability to perceive global response features and ensuring the stability of the training process.

[0077] Parameter prediction output layer: The key structural parameters of the antenna are finally output by gradually reducing the dimensionality through a four-layer fully connected network (FC).

[0078] To verify the contributions of each module in the 1D multi-scale CNN network model, this invention conducted ablation experiments on the model. The experiment used four key structural parameters of the antenna array—the side length of the rectangular patch, the spacing between patch elements, and the length and width of the rectangular slot in the cross-shaped slot—as the targets for inverse retrieval. The experimental results are as follows: Figure 12 As shown, Figure 12 Experimental data shows that after introducing the RefConv multi-branch structure, the convergence cycle of the model test set is significantly reduced from 500 to 250, significantly accelerating the network's efficiency in capturing multi-scale electromagnetic features. Further introduction of a Non-Local layer significantly reduces the mean squared error of the test set from 0.023 to 0.013. This experimental data fully demonstrates that the global attention mechanism can effectively correct biases in local feature extraction and suppress gradient oscillations during training, thereby greatly enhancing the model's generalization ability. The final complete model achieves stable convergence at 0.013 in only 200 training epochs, exhibiting excellent prediction accuracy and training efficiency.

[0079] Furthermore, to verify the inversion design capability of the 1D multi-scale CNN network model, this invention uses the trained network model to predict the four sets of target S-parameter (scattering parameter) curves in the test set. The model successfully and accurately outputs the corresponding four sets of antenna structure parameter combinations (corresponding to the side length of the rectangular patch, the spacing of the patch unit, and the length and width of the cross-shaped slot, respectively), with the specific predicted dimensions being: [8.792, 0.890, 33.956, 0.889], [8.698, 0.708, 34.673, 0.942], [8.903, 0.699, 33.073, 0.614], and [8.905, 0.778, 33.005, 0.916]. Figure 13 The initial target S is displayed intuitively. 11 The curve and the S generated after substituting the above predicted parameters into the electromagnetic simulation 11 The curves are compared. The results show that the simulated curves generated by the predicted parameters are highly consistent with the target curves in key indicators such as the core resonant frequency, effective bandwidth, and return loss depth. Faced with test samples exhibiting different frequency band shifts and bandwidth evolution characteristics, this method maintains extremely high prediction accuracy, fully demonstrating the feasibility and high reliability of this network model in intelligent antenna design tasks.

[0080] This invention is not limited to the preferred embodiments described above. Anyone can derive other products in various forms under the guidance of this invention. However, regardless of any changes in shape or structure, any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A high-gain metasurface beamlet antenna array, characterized in that, From top to bottom, the antenna consists of a metasurface layer, a first dielectric layer, a slotted ground layer, a second dielectric layer, and a feed layer. The metasurface layer comprises several antenna elements arranged in a rectangular array, each antenna element consisting of several patch elements arranged in a rectangular array. The center patch element located in the central region of the antenna element and the peripheral patch elements located in the outer region of the antenna element are both based on rectangular patches. Chamfered areas are set on both sides of the rectangular patch to obtain an asymmetrical center patch element. A diamond-shaped slot is set in the center of the rectangular patch to obtain the peripheral patch element. The long diagonal of the diamond-shaped slot points to the center of the antenna element. The slotted grounding layer is a grounding metal layer with cross-shaped slots etched on it. The cross-shaped slots on the slotted grounding layer correspond one-to-one with the patch groups on the metasurface layer. The feed layer is equipped with a microstrip feed network with a fixed resistance value. After the radio frequency signal is input, it is fed into the antenna element with equal amplitude and in phase through the microstrip feed network.

2. The high-gain metasurface beamlet array of claim 1, wherein, The chamfered areas on both sides of the central patch unit are right-angled trapezoids to give the central patch unit a Z-shaped structure.

3. The high-gain metasurface beamlet array of claim 1, wherein, The side length of the rectangular patch and the spacing between adjacent patch units satisfy the following formula: 3 W1+2 W2=2.0λ; In the formula, λ is the wavelength of the antenna array at the corresponding operating frequency, W1 is the side length of the rectangular patch, and W2 is the spacing between adjacent patch units.

4. The high-gain metasurface beamlet array of claim 1, wherein, The microstrip feed network includes multiple feed ends, each with a width greater than its front section; the feed ends correspond to the center of the cross-shaped slot on the slotted grounding layer.

5. The high-gain metasurface beamlet array of claim 4, wherein, The microstrip power supply network includes multiple power dividers, each of which includes two power divider branches. The power divider of the next stage is located at the end of the power divider branch of the previous stage, and the power divider branch of the last stage is the power supply end.

6. The high-gain metasurface beamlet array of claim 1, wherein, The cross-shaped slot consists of two perpendicularly intersecting rectangular slots; the center of the cross-shaped slot is aligned with the center of the antenna element.

7. A high-gain metasurface beamwave antenna array according to claim 1, characterized in that, The center-to-center distance between adjacent antenna elements is greater than half the wavelength of the antenna array at the corresponding operating frequency.

8. A high-gain metasurface beamwave antenna array according to claim 2, characterized in that, The rectangular patch has a side length of 8.9 mm, a spacing of 0.7 mm between adjacent patch units, a height of 6 mm for the chamfered area, and upper and lower bottom edges of the chamfered area of ​​1.5 mm and 2.6 mm, respectively; the long diagonal of the diamond-shaped slit is 5.0 mm; the width of the rectangular slit of the cross-shaped slit is 0.8 mm, and the length of the rectangular slit of the cross-shaped slit is 34 mm.

9. A high-gain metasurface beamwave antenna array according to claim 1, characterized in that, The first dielectric layer is a Rogers RT3003 dielectric substrate with a thickness of 2.28 mm, and the second dielectric layer is a Rogers RT3003 dielectric substrate with a thickness of 0.76 mm; the metasurface layer, the slotted ground layer, and the power supply layer are all made of copper.

10. A reverse design method for a high-gain metasurface beamwave antenna array based on any one of claims 1-9, characterized in that, A 1D multi-scale CNN network model is used to achieve the inverse mapping from the target scattering parameter sequence to the key structural parameters of the antenna array. The 1D multi-scale CNN network model includes an initial feature extraction module, a multi-scale parallel convolution module, a three-channel attention mechanism module, and a parameter prediction output layer connected in sequence. The specific method is as follows: S1: Input the target scattering parameter sequence into the initial feature extraction module of the pre-trained 1D multi-scale CNN network model to perform preliminary feature extraction. The extracted preliminary features are then subjected to batch normalization and max pooling to achieve feature dimensionality reduction and obtain the basic electromagnetic features. S2: Input the basic electromagnetic features into the multi-scale parallel convolution module of the 1D multi-scale CNN network model. Through the parallel convolution branches of RefConv convolution kernels of different sizes, local features, mesoscale features and global context features are extracted respectively. After feature fusion, multi-scale electromagnetic features are obtained. S3: Input multi-scale electromagnetic features into the three-channel attention mechanism module of the 1D multi-scale CNN network model, and process them sequentially through the spatial attention layer, channel attention layer and non-local attention layer to enhance the perception of global and key features, thereby obtaining optimized features; S4: Input the optimized features into the parameter prediction output layer of the 1D multi-scale CNN network model, and output the key structural parameters after gradual dimensionality reduction through four fully connected networks; the key structural parameters include the side length of the rectangular patch, the spacing between patch units, and the length and width of the rectangular gap in the cross-shaped gap.

Images