Antenna design method and device based on current distribution and multi-port network theory

Abstract

This application relates to the field of antenna structure design technology, providing an antenna design method and apparatus based on current distribution and multi-port network theory. By establishing a mapping relationship between current distribution and antenna topology, the effective and redundant regions on the antenna radiator are intelligently identified according to current strength, and then optimized to maintain electromagnetic performance while achieving lightweight antenna topology. Principal component analysis is used to extract the first left singular vector as the principal feature vector, constructing a principal current feature distribution map. The map is then adaptively binarized using K-means clustering to construct a preliminary antenna topology. By converting the multi-port network and genetic algorithm into a performance feedforward compensator, impedance and radiation characteristics are optimized using the same model, achieving a dynamic balance between lightweight design and high performance within the same process.

Description

Technical Field

[0001] This application relates to the field of antenna structure design technology, and in particular to an antenna design method and apparatus based on current distribution and multi-port network theory. Background Technology

[0002] With the rapid development of 5G / 6G communication technology, the integration requirements of base station antennas are constantly increasing. Existing base station antenna designs generally face the problem of balancing lightweight structure with high performance indicators. Furthermore, traditional design methods rely on the experience of designers and have low simulation efficiency, which seriously restricts the integration and cost reduction of base station antennas. It is urgent to break through the relevant technical bottlenecks.

[0003] There is an inherent contradiction between "lightweight" and "high performance" in existing base station antennas. To ensure key performance characteristics such as antenna bandwidth, gain, and cross-polarization isolation, traditional designs often employ complex metal geometries and large metal surface areas, resulting in bulky and costly antenna structures that are not conducive to integrated deployment. On the other hand, simply reducing the amount of metal material empirically to achieve lightweighting can easily disrupt the antenna's inherent current path and electromagnetic coupling mechanism, directly causing impedance mismatch, gain reduction, and pattern distortion, making it difficult to maintain or even improve electromagnetic performance while significantly reducing weight.

[0004] Existing design methods fail to effectively utilize the physical mapping mechanism of "current-radiation," and structural tailoring lacks a scientific basis. The essence of antenna radiation characteristics stems from the current distribution on the conductor surface. High current density regions are the core of carrying electromagnetic energy, while low current density regions are mostly structural redundancies. However, existing optimization methods ignore this core mechanism, only performing mathematical optimization at the geometric level. They cannot quantify the actual contribution of metallic materials to performance, and it is difficult to accurately identify key current paths and redundant regions, leading to easy failure of lightweight topologies or poor weight reduction effects.

[0005] After lightweight antenna trimming, changes in metal boundary conditions lead to input impedance drift and radiation mode perturbations, requiring secondary optimization to compensate for performance losses. Currently, the industry-standard full-wave electromagnetic simulation iterative verification method is time-consuming with each calculation. When dealing with discrete mesh topologies with massive degrees of freedom, the serial trial-and-error approach becomes a computational bottleneck, unable to quickly traverse the solution space to find the optimal compensation scheme, and can only accept suboptimal solutions. Furthermore, existing current distribution extraction methods and multi-port network optimization methods have limitations. The former extracts features lacking representativeness, and the latter cannot guide changes in metal structure topology, resulting in fragmented and open-loop design processes, making it difficult to achieve a balance between lightweight design and high performance. Therefore, an efficient, integrated, lightweight antenna design method is urgently needed. Summary of the Invention

[0006] Therefore, it is necessary to provide an antenna design method and device based on current distribution and multi-port network theory to address the above-mentioned technical problems.

[0007] An antenna design method based on current distribution and multi-port network theory includes the following steps:

[0008] Obtain the surface current distribution of the antenna structure to be optimized under multiple operating states, and construct the current intensity matrix;

[0009] The current intensity matrix is ​​subjected to singular value decomposition, and the first left singular vector is extracted as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct the principal current feature distribution map.

[0010] Using metal and air as cluster types respectively, the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm, and the segmentation results are mapped to a preliminary antenna topology.

[0011] The initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model.

[0012] Using the on / off state of the parasitic loading end of the multi-port network model as the optimization variable, and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives, the multi-port network model is iteratively optimized using a genetic algorithm to obtain the port on / off states that meet the target performance, and then mapped to the final lightweight antenna structure.

[0013] In one embodiment, the surface current distribution of the antenna structure to be optimized under multiple operating states is obtained, and a current intensity matrix is ​​constructed, including:

[0014] The surface of the antenna structure to be optimized is discretized into an M×N pixel grid. Full-wave electromagnetic simulation is then performed within the operating frequency band. Internal selection Each sampling frequency point, within a complete time period Internal selection discrete time Extract each grid cell Surface current amplitude at all sampling frequencies and at all discrete times The construction dimension is The time-frequency joint current characteristic matrix:

[0015]

[0016] in, This is the current intensity matrix; for frequency The M×N matrix consists of the surface current values ​​of all pixels at time t. Column vector.

[0017] In one embodiment, singular value decomposition is performed on the current intensity matrix to extract the first left singular vector as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct a principal current feature distribution map. The method further includes:

[0018] The numerical values ​​of the main current characteristic distribution map are linearly mapped to the standard grayscale image range [0, 255] and then normalized.

[0019]

[0020] in, Main current characteristic distribution map coordinates The value at that location; for The corresponding grayscale value; Main current characteristic distribution diagram;

[0021] Obtain the grayscale image of the current characteristics corresponding to the main current characteristic distribution map. .

[0022] In one embodiment, metal and air are used as cluster types, and the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm. The segmentation result is then mapped to a preliminary antenna topology, including:

[0023] Initialize the center of the corresponding clusters of metal and air , ;

[0024] Will Each pixel is assigned to the category represented by the cluster center closest to the pixel's grayscale value;

[0025] Based on all the pixels currently assigned to each cluster, recalculate the new centroid of the corresponding cluster:

[0026]

[0027] in, For the first t The new centroid of the cluster is obtained by +1 calculation; Indicates the first Assigned to in round clustering The number of samples in a cluster; For the first t The next assignment is to the first k A collection of pixels in a cluster; For pixels; For the first k The centroid of a cluster, ;

[0028] Pixels are redistributed based on the new centroid of the cluster to minimize the sum of squared errors within the cluster:

[0029]

[0030] in, The set of pixels assigned to a cluster;

[0031] Map the pixel classification results to a preliminary antenna topology:

[0032]

[0033] in, The results are for pixel classification.

[0034] In one embodiment, the initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model, including:

[0035] The initial antenna topology is discretized into a rectangular grid array. Lumped ports are set at the boundaries of the grid cells to construct a multi-port network model. The lumped ports include one main feed port and... One parasitic loading port;

[0036] Pre-simulation was performed on the constructed multi-port network model to extract the open-circuit impedance matrix of the multi-port network model in the target frequency band. and far-field radiation characteristic data of each port, the far-field radiation characteristic data including the first n Far-field electric field distribution at each port under independent excitation , include Quantity and Quantity .

[0037] In one embodiment, the parasitic loading ports of the multi-port network model are used as individual chromosomes in a genetic algorithm. With the goal of compensating for and optimizing antenna impedance matching and radiation performance, the genetic algorithm optimizes the load impedance of the parasitic loading ports, including:

[0038] Vectors constructed from parasitic loading ports of a multi-port network model As individual chromosomes in a genetic algorithm, For the first i The connection status of a parasitic loading port. When, the corresponding applied impedance , When, the corresponding applied impedance ;

[0039] Based on vectors Construct a diagonal load matrix , The diagonal elements are The mapped load impedance values ​​for each port;

[0040] Obtain the voltage-current relationship of the multi-port network model, and based on the diagonal load matrix. Calculate the current distribution across the entire port;

[0041] Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and all frequency points is calculated in parallel.

[0042] Based on full-port current distribution and far-field radiation characteristic data, obtain radiation pattern data;

[0043] A cost function is constructed based on pattern data and return loss:

[0044]

[0045]

[0046] in, The cost function; for S Parameter penalty; and These are the weighting coefficients; For far-field penalty; For frequency point Return loss; The target threshold;

[0047] Minimize the cost function and output the optimized connection method for the parasitic loading port.

[0048] In one embodiment, the voltage-current relationship of the multi-port network model is obtained, and based on the diagonal load matrix... Calculate the total port current distribution, including:

[0049] Obtain the voltage-current relationship in the multiport network model:

[0050]

[0051] in, Port voltage; Port current; for Extended to A matrix of dimensionality;

[0052] Separate the main power supply port from the parasitic load port:

[0053]

[0054] in, The mutual impedance of all ports except port 1 to port 1; This is the mutual impedance matrix between all ports except port 1; The current at the main feed port; The induced current vector of the parasitic loading port;

[0055] Using the GPU's batch linear solver, the solution is obtained. Output the full-port current distribution vector .

[0056] In one embodiment, based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and all frequencies is calculated in parallel, including:

[0057] Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the input impedance of the antenna is calculated:

[0058]

[0059]

[0060] in, This is the input impedance of the antenna; The self-impedance of port 1; This is the mutual impedance of port 1 to all other ports; This is the total voltage at port 1;

[0061] Calculate the return loss in the current state:

[0062]

[0063] in, For return loss; This is the reference impedance.

[0064] In one embodiment, radiation pattern data is obtained based on full-port current distribution and far-field radiation characteristic data, including:

[0065] Based on the full-port current distribution and far-field radiation characteristics data, the total electric field amplitude is calculated:

[0066]

[0067] in, This represents the total electric field amplitude; For full-port current distribution The n One element;

[0068] Converting the total electric field amplitude to decibels yields the radiation pattern data:

[0069]

[0070]

[0071] in, For directional pattern data; For the total electric field amplitude Polarization components; For the total electric field amplitude Polarization components; This is a stability parameter.

[0072] An antenna design device based on current distribution and multiport network theory includes:

[0073] The current matrix construction module is used to obtain the surface current distribution of the antenna structure to be optimized under multiple operating states and construct the current intensity matrix.

[0074] The feature distribution map construction module is used to perform singular value decomposition on the current intensity matrix, extract the first left singular vector as the principal feature vector, and rearrange the elements of the principal feature vector according to the mapping relationship between the principal feature vector and the current matrix to construct the principal current feature distribution map.

[0075] The preliminary structure construction module is used to adaptively binarize the main current feature distribution map using the K-means clustering algorithm, with metal and air as cluster types respectively, and map the segmentation results to the preliminary antenna topology.

[0076] The network model construction module is used to perform pixel-griding of the preliminary antenna topology, establish internal ports between the grids, and construct a multi-port network model.

[0077] The lightweight structure optimization module is used to iteratively optimize the multi-port network model using the on / off state of the parasitic loading end of the multi-port network model as the optimization variable, and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives. It obtains the port on / off state that meets the target performance and maps it to the final lightweight antenna structure.

[0078] The aforementioned antenna design method and device based on current distribution and multi-port network theory establishes a mapping relationship between current distribution and antenna topology. It intelligently identifies effective and redundant regions on the antenna radiator based on current strength, and then optimizes the antenna topology to maintain its electromagnetic performance while achieving lightweight design. Principal component analysis is used to extract the first left singular vector as the principal feature vector, constructing a principal current feature distribution map. This map is then adaptively binarized using K-means clustering to construct a preliminary antenna topology. By converting the multi-port network and genetic algorithm into a performance feedforward compensator, and utilizing the same model for impedance and radiation characteristic optimization, a dynamic balance between lightweight design and high performance is achieved within the same process.

[0079] This invention constructs an automated sequence of principal component feature extraction, K-means adaptive segmentation, structure mapping, multi-port network modeling, and genetic algorithm performance compensation. The intermediate processes do not require manual intervention in decision-making, which greatly shortens the design cycle of lightweight antennas. Attached Figure Description

[0080] Figure 1 This is a flowchart illustrating an antenna design method based on current distribution and multi-port network theory in one embodiment.

[0081] Figure 2 This is a comparison chart of S-parameters in one embodiment;

[0082] Figure 3 Here are some comparison images of the far-field plots in one embodiment, where (a) is a comparison image of the far-field plots at 2.0 GHz, (b) is a comparison image of the far-field plots at 2.2 GHz, (c) is a comparison image of the far-field plots at 2.4 GHz, (d) is a comparison image of the far-field plots at 2.6 GHz, (e) is a comparison image of the far-field plots at 2.8 GHz, and (f) is a comparison image of the far-field plots at 3.0 GHz.

[0083] Figure 4 This is a block diagram of an antenna design device based on current distribution and multi-port network theory in one embodiment. Detailed Implementation

[0084] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0085] In one embodiment, such as Figure 2 As shown, an antenna design method based on current distribution and multi-port network theory is provided, including the following steps:

[0086] Obtain the surface current distribution of the antenna structure to be optimized under multiple operating states, and construct the current intensity matrix;

[0087] The current intensity matrix is ​​subjected to singular value decomposition, and the first left singular vector is extracted as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct the principal current feature distribution map.

[0088] Using metal and air as cluster types respectively, the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm, and the segmentation results are mapped to a preliminary antenna topology.

[0089] The initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model.

[0090] Using the on / off state of the parasitic loading end of the multi-port network model as the optimization variable, and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives, the multi-port network model is iteratively optimized using a genetic algorithm to obtain the port on / off states that meet the target performance, and then mapped to the final lightweight antenna structure.

[0091] In the antenna design based on current distribution and multi-port network theory, a mapping relationship between current distribution and antenna topology is established. Effective and redundant regions on the antenna radiator are intelligently identified based on current strength, and then optimized to maintain electromagnetic performance while achieving a lightweight antenna topology. Principal component analysis is used to extract the first left singular vector as the principal feature vector, constructing a principal current feature distribution map. This map is then adaptively binarized using K-means clustering to construct a preliminary antenna topology. By converting the multi-port network and genetic algorithm into a performance feedforward compensator, impedance and radiation characteristics are optimized using the same model, achieving a dynamic balance between lightweight design and high performance within the same process.

[0092] In one embodiment, the surface current distribution of the antenna structure to be optimized under multiple operating states is obtained, and a current intensity matrix is ​​constructed, including:

[0093] The surface of the antenna structure to be optimized is discretized into an M×N pixel grid. Full-wave electromagnetic simulation is then performed within the operating frequency band. Internal selection Each sampling frequency point, within a complete time period Internal selection discrete time Extract each grid cell Surface current amplitude at all sampling frequencies and at all discrete times The construction dimension is The time-frequency joint current characteristic matrix:

[0094]

[0095] in, This is the current intensity matrix; for frequency The M×N matrix consists of the surface current values ​​of all pixels at time t. Column vector.

[0096] In this embodiment, by constructing a current matrix, the current information that was originally scattered in the time domain is integrated into a unified mathematical framework, laying the foundation for subsequent optimization.

[0097] In one embodiment, singular value decomposition is performed on the current intensity matrix, and the first left singular vector is extracted as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct a principal current feature distribution map. Specifically, for the current intensity matrix... Perform singular value decomposition:

[0098]

[0099] in, Let be an M×M orthogonal matrix containing left singular vectors, representing Spatial modes; Given an M×N singular value diagonal matrix arranged in descending order, with K singular values ​​on its main diagonal ( ); It is an N×N orthogonal matrix containing right singular vectors, whose row vectors represent the time evolution coefficients corresponding to the spatial modes.

[0100] Current matrix It can be regarded as all spatial modes A linear combination, where the weight of each mode is determined by its corresponding singular value. Determined by the time coefficient. The first left singular vector. Corresponding to the largest singular value This represents the spatial distribution pattern with the largest variance among all time-sampled current distributions, i.e., the most significant and stable common feature of current flow. Therefore, the first left singular vector is used as the principal feature vector of the antenna lightweight topology.

[0101] Based on the mapping relationship between the principal eigenvector and the current matrix, the first left singular vector is... The elements, according to The mapping relationship with the original M×N grid is rearranged to construct the main current characteristic distribution map. .

[0102] It should be noted that the current characteristic distribution map quantifies the "contribution" of each location on the antenna surface to maintaining the dominant radiation mode, and its value directly indicates the importance of the metal material at that location.

[0103] In one embodiment, the singular value decomposition of the current intensity matrix is ​​performed, the first left singular vector is extracted as the principal eigenvector, and the elements of the principal eigenvector are rearranged according to the mapping relationship between the principal eigenvector and the current matrix to construct a principal current feature distribution map. The method further includes:

[0104] The numerical values ​​of the main current characteristic distribution map are linearly mapped to the standard grayscale image range [0, 255] and then normalized.

[0105]

[0106] in, Main current characteristic distribution map coordinates The value at that location; for The corresponding grayscale value; Main current characteristic distribution diagram;

[0107] Obtain the grayscale image of the current characteristics corresponding to the main current characteristic distribution map. .

[0108] Understandably, the principal eigenvectors obtained through singular value decomposition and the constructed principal current characteristic distribution map have numerical ranges of arbitrary scale. It should be noted that the principal eigenvectors have sign uncertainty (±...). (Both are solutions). Before or after normalization, a judgment needs to be made: ensure that in the final normalized graph, high current density regions (the metal regions we want to retain) correspond to higher grayscale values ​​(close to 255), while low current density regions correspond to lower grayscale values ​​(close to 0). If the direction is reversed, then... Perform grayscale inversion operation .

[0109] In this embodiment, the values ​​of the main current feature distribution map are linearly mapped to the standard grayscale image range [0, 255] by using the max-min normalization method, thereby completing the normalization and obtaining a standardized grayscale map of current features where high grayscale values ​​represent "high structural importance", which facilitates subsequent image processing.

[0110] In one embodiment, metal and air are used as cluster types, and the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm. The segmentation result is then mapped to a preliminary antenna topology, including:

[0111] Initialize the center of the corresponding clusters of metal and air , ;

[0112] Will Each pixel is assigned to the category represented by the cluster center closest to the pixel's grayscale value;

[0113] Based on all the pixels currently assigned to each cluster, recalculate the new centroid of the corresponding cluster:

[0114]

[0115] in, For the first t The new centroid of the cluster is obtained by +1 calculation; Indicates the first Assigned to in round clustering The number of samples in a cluster; For the first t The next assignment is to the first k A collection of pixels in a cluster; For pixels; For the first k The centroid of a cluster, ;

[0116] Pixels are redistributed based on the new centroid of the cluster to minimize the sum of squared errors within the cluster:

[0117]

[0118] in, The set of pixels assigned to a cluster;

[0119] Map the pixel classification results to a preliminary antenna topology:

[0120]

[0121] in, The results are for pixel classification.

[0122] In this embodiment, the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm to construct a preliminary antenna topology.

[0123] In one embodiment, the initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model, including:

[0124] The initial antenna topology is discretized into a rectangular grid array. Lumped ports are set at the boundaries of the grid cells to construct a multi-port network model. The lumped ports include one main feed port and... One parasitic loading port (total) P (ports)

[0125] Pre-simulation was performed on the constructed multi-port network model to extract the open-circuit impedance matrix of the multi-port network model in the target frequency band. and far-field radiation characteristic data of each port, the far-field radiation characteristic data including the first n Far-field electric field distribution at each port under independent excitation , include Quantity and Quantity .

[0126] It should be noted that in a multi-port network model, the lumped port is... The lumped port includes one main power supply port ( )and One parasitic loading port ( The boundaries of mesh cells are the interfaces between meshes, such as adjacent areas on the top, bottom, left, and right. Pre-simulation of the constructed multi-port network model is performed using CST Microwave Studio. Open-circuit impedance matrix. The mutual impedance coupling relationship between all ports was recorded; the far-field radiation characteristic data contained the spatial coupling information of the structure, as well as the radiation characteristic information. Once extracted, the time-consuming full-wave simulation was not required again in the subsequent genetic algorithm iterations, only matrix algebra operations were required.

[0127] In one embodiment, the parasitic loading ports of the multi-port network model are used as individual chromosomes in a genetic algorithm. With the goal of compensating for and optimizing antenna impedance matching and radiation performance, the genetic algorithm optimizes the load impedance of the parasitic loading ports, including:

[0128] Vectors constructed from parasitic loading ports of a multi-port network model As individual chromosomes in a genetic algorithm, For the first i The connection status of a parasitic loading port. When, the corresponding applied impedance , When, the corresponding applied impedance ;

[0129] Based on vectors Construct a diagonal load matrix , The diagonal elements are The mapped load impedance values ​​for each port;

[0130] Obtain the voltage-current relationship of the multi-port network model, and based on the diagonal load matrix. Calculate the current distribution across the entire port;

[0131] Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and all frequency points is calculated in parallel.

[0132] Based on full-port current distribution and far-field radiation characteristic data, obtain radiation pattern data;

[0133] A cost function is constructed based on pattern data and return loss:

[0134]

[0135]

[0136] in, The cost function; Penalize the S-parameters; These are the weighting coefficients; For far-field penalty; For frequency point Return loss; The target threshold;

[0137] Minimize the cost function and output the optimized connection method for the parasitic loading port.

[0138] It should be noted that, When this occurs, it indicates that a metal connection is maintained at that location; in the circuit model, this corresponds to the applied impedance. ; When this occurs, it indicates that the metal at that location has been etched or cut off; in the circuit model, this corresponds to the applied impedance. .

[0139] Regarding the cost function, it should be noted that... S The parameter penalty formula represents the penalty at a certain frequency point. Cost is only incurred when the threshold is exceeded (performance fails to meet standards); if If the value is below the threshold (performance meets the standard), the cost is 0. By introducing the ReLU mechanism, the optimization process focuses on correcting "bad points" rather than ineffectively suppressing the values ​​of frequencies that have already met the standard, thus improving convergence efficiency. For the far-field penalty, specifically, the weighted Euclidean distance between the synthesized radiation pattern and the ideal target mask is calculated to constrain the main lobe gain and side lobe level.

[0140] In this embodiment, in the cost function, SThe parameter penalty employs a one-sided penalty mechanism based on the ReLU activation function (i.e., introducing a penalty mechanism specifically for the ReLU activation function into the genetic algorithm). S The "reach the target and stop" (ReLU mechanism) strategy for parameters, and the beam shape constraint of the far-field radiation pattern, guide the genetic algorithm to automatically search for the best structure that meets the requirements of lightweighting while ensuring bandwidth and directional radiation performance from a huge number of structural combinations.

[0141] In one embodiment, the voltage-current relationship of the multi-port network model is obtained, and based on the diagonal load matrix... Calculate the total port current distribution, including:

[0142] Obtain the voltage-current relationship in the multiport network model:

[0143]

[0144] in, Port voltage; Port current; for Extended to A matrix of dimension (with non-zero values ​​only on the diagonal corresponding to the parasitic port).

[0145] Separate the main power supply port from the parasitic load port:

[0146]

[0147] in, This represents the mutual impedance of all ports except port 1 (the feed port) to port 1. This is the mutual impedance matrix between all ports except port 1; The current at the main feed port; The induced current vector of the parasitic loading port;

[0148] Using the GPU's batch linear solver, the solution is obtained. Output the full-port current distribution vector .

[0149] Specifically, the voltage-current relationship matrix of the multi-port network model is divided into blocks, and the main feed port is divided into blocks. (Referred to as the excitation port) and the parasitic loading port Separation:

[0150]

[0151] Because there is no external excitation source at the parasitic loading port (i.e. And let the excitation current be normalized to Therefore, it is converted to:

[0152]

[0153] The solution can be obtained directly using the GPU's batch linear solver. .

[0154] In one embodiment, based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and all frequencies is calculated in parallel, including:

[0155] Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the input impedance of the antenna is calculated:

[0156]

[0157]

[0158] in, This is the input impedance of the antenna; The self-impedance of port 1; This is the mutual impedance of port 1 to all other ports; This is the total voltage at port 1;

[0159] Calculate the return loss in the current state:

[0160]

[0161] in, For return loss; The reference impedance is typically 1. .

[0162] It should be noted that the downgrading of a multi-port network to a single-port network is done using the Schur complement principle.

[0163] By using the formula for calculating return loss, the return loss of all individuals and all frequency points can be calculated in parallel on the GPU to obtain the corresponding return loss. The curves do not require repeated field simulations.

[0164] In one embodiment, radiation pattern data is obtained based on full-port current distribution and far-field radiation characteristic data, including:

[0165] Based on the full-port current distribution and far-field radiation characteristics data, the total electric field amplitude is calculated:

[0166]

[0167] in, This represents the total electric field amplitude; For full-port current distribution Then One element;

[0168] Converting the total electric field amplitude to decibels yields the radiation pattern data:

[0169]

[0170]

[0171] in, For directional pattern data; For the total electric field amplitude Polarization components; For the total electric field amplitude Polarization components; This is a stability parameter.

[0172] Specifically and Calculate using the following formula:

[0173] .

[0174] In this embodiment, by calculating the radiation pattern data, the performance of each body in a specific cross-section (such as...) is evaluated in real time. The shape of the radiation beam.

[0175] To verify the effectiveness of the proposed method, the generated lightweight antenna structure was imported into the full-wave electromagnetic simulation software CST Microwave Studio for verification and compared with the "original reference antenna (unlightweight)" and the "antenna with only preliminary current trimming but no optimization (intermediate state)" (the intermediate state being the preliminary antenna topology). The simulation frequency was set to the typical frequency band of mobile communication base stations (1.5 GHz - 3.6 GHz). The results are as follows: Figure 2 , Figure 3 As shown. (Refer to...) Figure 2 ,from Figure 2 As can be seen, the intermediate state, compared to the original reference antenna, has... The curves did not deteriorate significantly, indicating that the current-intensity-based pruning logic could preserve the main radiation paths. However, due to the alteration of the local reactance distribution caused by the removal of the metal structure, the resonance depth of the intermediate-state antenna at some frequencies became slightly shallower, and while frequency offset existed, it remained within a controllable range. After optimization using a genetic algorithm based on a multi-port network, a lightweight antenna structure was generated. The curve was further corrected and improved. Through fine-tuning of the on / off states of the mesh ports, the algorithm successfully compensated for the slight impedance mismatch caused by the clipping. Although limited by the reduction in physical size (smaller electrical size), its low-frequency band... The depth loss is slightly higher than the original all-metal antenna (i.e., the return loss reading is slightly higher, but still remains below the -10dB engineering standard), which is a reasonable trade-off under the laws of physics. It can also be seen that, due to the reduced metal volume of the antenna after weight reduction, some parasitic effects in the high-frequency band are suppressed. The resulting lightweight antenna structure exhibits superior matching characteristics in the high-frequency region compared to the original antenna, thus expanding the effective operating bandwidth of the antenna overall.

[0176] from Figure 3 As can be seen, the optimized lightweight antenna maintains stable directional radiation characteristics throughout the entire operating frequency band. Although the low-frequency gain fluctuates slightly, the overall beamwidth and aspect ratio are basically the same as the original antenna. Comparative data shows that the peak gain of the final lightweight antenna has a very small decrease (<2 dB) compared to the original all-metal antenna. This is due to the substantial reduction in the antenna's physical aperture and the metal radiator, which is a normal phenomenon under the laws of electromagnetic physics. However, compared to the intermediate state antenna, the optimized gain still recovers. This indicates that the present invention has successfully found the optimal balance between "significant weight reduction" and "maintaining radiation performance." Furthermore, calculations show that the final designed antenna's metal volume is reduced by approximately 36% compared to the original reference antenna. This not only significantly reduces the antenna's weight and manufacturing cost but also effectively reduces wind resistance load.

[0177] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0178] In one embodiment, such as Figure 4 As shown, an antenna design device based on current distribution and multi-port network theory is provided, comprising:

[0179] The current matrix construction module 901 is used to obtain the surface current distribution of the antenna structure to be optimized under multiple operating states and construct the current intensity matrix.

[0180] The feature distribution map construction module 902 is used to perform singular value decomposition on the current intensity matrix, extract the first left singular vector as the principal feature vector, and rearrange the elements of the principal feature vector according to the mapping relationship between the principal feature vector and the current matrix to construct the principal current feature distribution map.

[0181] The preliminary structure construction module 903 is used to adaptively binarize the main current feature distribution map using the K-means clustering algorithm with metal and air as cluster types respectively, and map the segmentation result to the preliminary antenna topology.

[0182] The network model construction module 904 is used to perform pixel meshing on the preliminary antenna topology, establish internal ports between the meshes, and construct a multi-port network model.

[0183] The lightweight structure optimization module 905 is used to iteratively optimize the multi-port network model using a genetic algorithm, with the on / off state of the parasitic loading end of the multi-port network model as the optimization variable and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives, to obtain the port on / off states that meet the target performance, and map them to the final lightweight antenna structure.

[0184] Specific limitations regarding the antenna design device based on current distribution and multiport network theory can be found in the limitations of the antenna design method based on current distribution and multiport network theory mentioned above, and will not be repeated here. Each module in the aforementioned antenna design device based on current distribution and multiport network theory can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0185] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0186] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An antenna design method based on current distribution and multi-port network theory, characterized in that, Includes the following steps: Obtain the surface current distribution of the antenna structure to be optimized under multiple operating states, and construct the current intensity matrix; The current intensity matrix is ​​subjected to singular value decomposition, and the first left singular vector is extracted as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct the principal current feature distribution map. Using metal and air as cluster types respectively, the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm, and the segmentation results are mapped to a preliminary antenna topology. The initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model. Using the on / off state of the parasitic loading end of the multi-port network model as the optimization variable, and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives, the multi-port network model is iteratively optimized using a genetic algorithm to obtain the port on / off states that meet the target performance, and then mapped to the final lightweight antenna structure.

2. The antenna design method based on current distribution and multi-port network theory according to claim 1, characterized in that, Obtain the surface current distribution of the antenna structure to be optimized under multiple operating states, and construct the current intensity matrix, including: The surface of the antenna structure to be optimized is discretized into an M×N pixel grid. Full-wave electromagnetic simulation is then performed within the operating frequency band. Internal selection Each sampling frequency point, within a complete time period Internal selection discrete time Extract each grid cell Surface current amplitude at all sampling frequencies and at all discrete times The construction dimension is The time-frequency joint current characteristic matrix: in, This is the current intensity matrix; for frequency An M×N column vector consisting of the surface current values ​​of all pixels at time 1. .

3. The antenna design method based on current distribution and multi-port network theory according to claim 1, characterized in that, The current intensity matrix is ​​subjected to singular value decomposition, and the first left singular vector is extracted as the principal eigenvector. Based on the mapping relationship between the principal eigenvector and the current matrix, the elements of the principal eigenvector are rearranged to construct the principal current feature distribution map. The method also includes: The numerical values ​​of the main current characteristic distribution map are linearly mapped to the standard grayscale image range [0, 255] and then normalized. in, Main current characteristic distribution map coordinates The value at that location; for The corresponding grayscale value; Main current characteristic distribution diagram; Obtain the grayscale image of the current characteristics corresponding to the main current characteristic distribution map. .

4. The antenna design method based on current distribution and multi-port network theory according to claim 3, characterized in that, Using metal and air as cluster types respectively, the main current feature distribution map is adaptively binarized and segmented using the K-means clustering algorithm. The segmentation results are then mapped to a preliminary antenna topology, including: Initialize the center of the corresponding clusters of metal and air , ; Will Each pixel is assigned to the category represented by the cluster center closest to the pixel's grayscale value; Based on all the pixels currently assigned to each cluster, recalculate the new centroid of the corresponding cluster: in, For the first t The new centroid of the cluster is obtained by +1 calculation; Indicates the first Assigned to in round clustering The number of samples in a cluster; For the first t The next assignment is to the first k A collection of pixels in a cluster; For pixels; For the first k The centroid of a cluster, ; Pixels are redistributed based on the new centroid of the cluster to minimize the sum of squared errors within the cluster: in, The set of pixels assigned to a cluster; Map the pixel classification results to a preliminary antenna topology: in, The results are for pixel classification.

5. The antenna design method based on current distribution and multi-port network theory according to claim 4, characterized in that, The initial antenna topology is pixelated, and internal ports are established between the grids to construct a multi-port network model, including: The initial antenna topology is discretized into a rectangular grid array. Lumped ports are set at the boundaries of the grid cells to construct a multi-port network model. The lumped ports include one main feed port and... One parasitic loading port; Pre-simulation was performed on the constructed multi-port network model to extract the open-circuit impedance matrix of the multi-port network model in the target frequency band. and far-field radiation characteristic data of each port, the far-field radiation characteristic data including the first n Far-field electric field distribution at each port under independent excitation , include Quantity and Quantity .

6. The antenna design method based on current distribution and multi-port network theory according to claim 5, characterized in that, Using parasitic load ports in a multi-port network model as individual chromosomes in a genetic algorithm, and aiming to compensate for and optimize antenna impedance matching and radiation performance, the genetic algorithm optimizes the load impedance of the parasitic load ports, including: Vectors constructed from parasitic loading ports of a multi-port network model As individual chromosomes in a genetic algorithm, For the first i The connection status of a parasitic loading port. When, the corresponding applied impedance , When, the corresponding applied impedance ; Based on vectors Construct a diagonal load matrix , The diagonal elements are The mapped load impedance values ​​for each port; Obtain the voltage-current relationship of the multi-port network model, and based on the diagonal load matrix. Calculate the current distribution across all ports; Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and all frequency points is calculated in parallel. Based on full-port current distribution and far-field radiation characteristic data, obtain radiation pattern data; A cost function is constructed based on pattern data and return loss: in, The cost function; for S Parameter penalty; and These are the weighting coefficients; For far-field penalty; For frequency point Return loss; The target threshold; Minimize the cost function and output the optimized connection method for the parasitic loading port.

7. The antenna design method based on current distribution and multi-port network theory according to claim 6, characterized in that, Obtain the voltage-current relationship of the multi-port network model, and based on the diagonal load matrix. Calculate the total port current distribution, including: Obtain the voltage-current relationship in the multiport network model: in, Port voltage; Port current; for Expand to A matrix of dimensionality; Separate the main power supply port from the parasitic load port: in, The mutual impedance of all ports except port 1 to port 1; This is the mutual impedance matrix between all ports except port 1; The current at the main feed port; The induced current vector of the parasitic loading port; Using the GPU's batch linear solver, the solution is obtained. Output the full-port current distribution vector .

8. The antenna design method based on current distribution and multi-port network theory according to claim 7, characterized in that, Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the return loss of all individuals and at all frequencies is calculated in parallel, including: Based on the full-port current distribution, the multi-port network is reduced to a single-port network, and the input impedance of the antenna is calculated: in, This is the input impedance of the antenna; The self-impedance of port 1; This is the mutual impedance of port 1 to all other ports; This is the total voltage at port 1; Calculate the return loss in the current state: in, For return loss; This is the reference impedance.

9. The antenna design method based on current distribution and multi-port network theory according to claim 8, characterized in that, Based on full-port current distribution and far-field radiation characteristic data, radiation pattern data is obtained, including: Based on the full-port current distribution and far-field radiation characteristics data, the total electric field amplitude is calculated: in, This represents the total electric field amplitude; For full-port current distribution The n One element; Converting the total electric field amplitude to decibels yields the radiation pattern data: in, For directional pattern data; For the total electric field amplitude Polarization components; For the total electric field amplitude Polarization components; This is a stability parameter.

10. An antenna design device based on current distribution and multi-port network theory, characterized in that, include: The current matrix construction module is used to obtain the surface current distribution of the antenna structure to be optimized under multiple operating states and construct the current intensity matrix. The feature distribution map construction module is used to perform singular value decomposition on the current intensity matrix, extract the first left singular vector as the principal feature vector, and rearrange the elements of the principal feature vector according to the mapping relationship between the principal feature vector and the current matrix to construct the principal current feature distribution map. The preliminary structure construction module is used to adaptively binarize the main current feature distribution map using the K-means clustering algorithm, with metal and air as cluster types respectively, and map the segmentation results to the preliminary antenna topology. The network model construction module is used to perform pixel-griding of the preliminary antenna topology, establish internal ports between the grids, and construct a multi-port network model. The lightweight structure optimization module is used to iteratively optimize the multi-port network model using the on / off state of the parasitic loading end of the multi-port network model as the optimization variable, and the antenna impedance matching performance and far-field radiation performance as the joint optimization objectives. It obtains the port on / off state that meets the target performance and maps it to the final lightweight antenna structure.

Images