Near-far field joint prediction and structure parameter bidirectional inversion method of patch antenna array and its application

Abstract

The application relates to the technical field of computers, in particular to a near-far-field joint prediction and structure parameter bidirectional inversion method of a patch antenna array and application thereof. When receiving an antenna array input, the input type is identified as a structure parameter, a near-field distribution or a target far-field performance index; when the input is a structure parameter, a forward prediction model based on a graph attention network is used to first predict the target near-field distribution of the antenna array on an observation plane, then a near-field-to-far-field transformation layer embedding physical priori is used to calculate a far-field radiation field, and further, a far-field performance index is extracted; when the input is a near-field distribution and / or a target far-field performance index, an inverse inversion model is used to inversely obtain an array structure parameter satisfying the target far-field performance under the constraints of equivalent sources and a near-field. The application aims to solve the problem of how to shorten the design and optimization cycle of an antenna array.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays and its application. Background Technology

[0002] With the rapid development of technologies such as 5G / 6G communication, satellite communication, and phased array radar, the design complexity of antenna arrays is constantly increasing. Patch antenna arrays have become the preferred solution for these applications due to their compact structure, ease of integration, and low cost. However, traditional antenna array design heavily relies on electromagnetic simulation software (such as ANSYS HFSS, CST, etc.) for full-wave numerical calculations. For a 5×5 patch array, a single simulation takes 30 minutes to several hours, parameter optimization requires hundreds of simulation iterations, and the total design cycle can last for weeks to months.

[0003] Currently, research has attempted to accelerate antenna design using machine learning methods, primarily including fully connected neural networks and convolutional neural networks. However, unlike other conventional machine learning scenarios, patch antenna arrays exhibit strong electromagnetic coupling effects between adjacent elements. Current models cannot learn the topological relationships and electromagnetic coupling effects between array elements, making it difficult to model complex spatial correlations. Furthermore, most design efforts focus on unidirectional prediction of "structural parameters" to "far-field radiation patterns," lacking physical consistency constraints combined with near-field testing, and also lacking efficient inversion capabilities from "far-field radiation patterns" to "structural parameters."

[0004] Therefore, there is an urgent need for a new technical solution that can achieve rapid forward prediction of structural parameters to near-field and far-field performance, and reverse inversion of far-field performance to near-field / equivalent source and structural parameters, while taking into account electromagnetic coupling between elements and physical transformation from near field to far field, thereby shortening the design and optimization cycle of antenna arrays. Summary of the Invention

[0005] The main objective of this application is to provide a method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays, aiming to solve the problem of how to shorten the design and optimization cycle of antenna arrays.

[0006] To achieve the above objectives, this application provides a method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of a patch antenna array, the method comprising:

[0007] S10, upon receiving the antenna array parameters of the patch antenna array to be tested, determine the parameter type corresponding to the antenna array parameters;

[0008] S20, when the parameter type is structural parameter, a forward prediction model is used to predict the near-field distribution of the patch antenna array under test, and the near-field distribution is converted into a far-field radiation field and the target far-field performance index of the patch antenna array under test in the far-field radiation field is determined.

[0009] S30, when the parameter type is near-field distribution and / or far-field performance index, the target structural parameters of the patch antenna array under test are inverted using the inverse inversion model under near-field distribution and equivalent source constraints.

[0010] Optionally, in step S20, the forward prediction model employs a graph attention network, and the forward prediction model includes the following steps in predicting the near-field distribution of the patch antenna array under test:

[0011] S21, Obtain the graph structure G=(V, E) of the patch antenna array under test, where V is the node set, each node corresponds to a patch unit; E is the edge set representing the electromagnetic coupling relationship between units;

[0012] S22, Based on the multi-head attention mechanism of the graph attention network, capture the electromagnetic coupling relationship between nodes in the node set to obtain the node characteristics of each node;

[0013] S23, aggregate the node features to map them to the near-field grid of the observation plane, and predict and output the complex values ​​of the electric / magnetic fields of each near-field grid through a fully connected layer to obtain the near-field distribution.

[0014] Optionally, the step of determining the far-field radiation field in S20 specifically includes:

[0015] S24, determine the equivalent current value of the near-field distribution. and equivalent magnetic current value :

[0016] ;

[0017] ;

[0018] In the formula, , Let these represent the complex vectors of the magnetic field intensity and the electric field intensity, respectively, in the near-field distribution. The unit normal vector representing the observation plane. This represents the position vector of the field point on the observation plane used to characterize the near-field distribution;

[0019] S25, based on the equivalent current value and equivalent magnetic current value Determine the far-field radiation field :

[0020]

[0021] In the formula, For wave number, For air wave impedance, To be observed as a unit vector, Let r be the magnetic permeability of the medium, r be the distance from the observation point to the origin, and j be the imaginary part. ω is the angular frequency of the electromagnetic wave.

[0022] Optionally, the step of determining the target far-field performance index of the patch antenna array under test in the far-field radiation field in step S20 specifically includes:

[0023] S26. Determine the maximum gain after normalizing the far-field radiation field. :

[0024]

[0025] In the formula, This represents the normalized far-field radiation field. Represents the pitch angle in spherical coordinates. Represents the azimuth angle in spherical coordinates;

[0026] S27, Determine based on maximum gain Below, two angles on the preset principal tangent plane , :

[0027]

[0028] In the formula, This indicates the selected principal tangent azimuth angle; the preset principal tangent plane includes the E plane or the H plane;

[0029] S28, according to angle , Calculate the 3dB beamwidth :

[0030]

[0031] S29, maximum gain and 3dB beamwidth This is used as a target far-field performance indicator and output.

[0032] Optionally, in step S30, the step of determining the target structural parameters includes:

[0033] S31, when the input is a far-field performance index, the target far-field performance index is encoded into a conditional vector, and a transformed near-field distribution matching the target far-field is constructed through the inverse constraint of the near-field to far-field transformation layer or pre-training mapping.

[0034] S32, downsample the input near-field distribution and / or the transformed near-field distribution to form a field value grid corresponding one-to-one with the array nodes;

[0035] S33 uses the downsampled field values, predefined array patterns, and coupling features as node features to construct a graph structure input;

[0036] S34, capture the local electromagnetic coupling features of the graph structure input through the graph attention layer, and capture the global features through the Transformer layer;

[0037] S35, based on the local electromagnetic coupling characteristics and the global characteristics, predict the target structural parameters corresponding to each node.

[0038] Optionally, the method further includes:

[0039] S40 simulates the electrical coupling effect between array elements by using mutual impedance or scattering parameters.

[0040] Optionally, the step of simulating the electrical coupling effect between array elements using mutual impedance or scattering parameters specifically includes:

[0041] S41, the mutual impedance between array element m and array element n is defined as:

[0042]

[0043] In the formula, Represents array cell Open circuit voltage, Represents array cell The excitation current, Indicates the unit Other unit port currents;

[0044] S42, mutual impedance The construction is based on an integral approximation of the current distribution and the induced electric field:

[0045]

[0046] In the formula, , These represent surface mount units. and unit The effective radiation surface area. Representation unit The intensity of the induced electric field generated in space Representation unit Surface current density distribution on;

[0047] S43, scattering parameters obtained based on simulation or measurement Define normalized coupling coefficients :

[0048]

[0049] In the formula, Represents array cell With array unit The mutual scattering parameters between them and Representing array units respectively and array unit The self-scattering parameters;

[0050] S44, the center distance between array cell m and array cell n mutual impedance and coupling coefficient Through fusion function The edge weights of the graph are obtained after preprocessing and weighted fusion. :

[0051]

[0052] In the formula, the fusion function This is a linear weighted mapping operation;

[0053] S45, edge weights The input is fed into a graph attention network to simulate the electrical coupling effect between array cells.

[0054] Furthermore, to achieve the above objectives, this application also provides a device for joint near-field and far-field prediction and bidirectional inversion of structural parameters of a patch antenna array, comprising:

[0055] The parameter type identification module is used to determine the parameter type corresponding to the antenna array parameters when the antenna array parameters of the patch antenna array under test are received.

[0056] The forward near-field and far-field prediction module is used to predict the near-field distribution of the patch antenna array under test using a forward prediction model when the parameter type is structural parameter, and convert the near-field distribution into a far-field radiation field and determine the target far-field performance index of the patch antenna array under test in the far-field radiation field.

[0057] The inverse parameter inversion module is used to invert the target structural parameters of the patch antenna array under test under near-field distribution and equivalent source constraints using an inverse inversion model when the parameter type is near-field distribution and / or far-field performance index.

[0058] In addition, to achieve the above objectives, this application also provides a computer system comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the near-field and far-field joint prediction and bidirectional inversion method for structural parameters of patch antenna arrays as described in any of the preceding claims.

[0059] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the near-field and far-field joint prediction and bidirectional inversion method for structural parameters of patch antenna arrays as described in any of the preceding claims.

[0060] This application has at least the following beneficial effects:

[0061] 1. By establishing a fast prediction model from structural parameters to near-field distribution, the predicted near-field distribution is converted into a far-field pattern, and far-field performance indicators are calculated, thus realizing a physically consistent, end-to-end trainable forward link from structural parameters to near-field to far-field performance.

[0062] 2. For far-field performance indicators, an equivalent near-field / equivalent source constraint is constructed based on the near-field to far-field transformation layer. The target far-field information is mapped to a set of target near-field distributions or near-field features. Then, through a hybrid attention model, the array structure parameters that satisfy the target far-field performance are obtained by inversion, realizing the "performance-driven" reverse design.

[0063] 3. Compared with the traditional iterative design process of "structural parameters to simulation and then parameter tuning", it can significantly shorten the antenna array design cycle and provide a new technical approach for the near-field and far-field collaborative design and optimization of patch antenna arrays. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating the near-field and far-field joint prediction and bidirectional inversion method of structural parameters for patch antenna arrays involved in the embodiments of this application;

[0065] Figure 2 This is a schematic diagram of the overall architecture of the near-field and far-field joint prediction and bidirectional inversion method of structural parameters of patch antenna array involved in the embodiments of this application;

[0066] Figure 3 This is a visual comparison of the minimum error, median error, and random representative samples involved in the embodiments of this application.

[0067] Figure 4 This is a scatter plot showing the corresponding comparison results of the embodiments of this application;

[0068] Figure 5 This is another scatter plot comparison result related to an embodiment of this application;

[0069] Figure 6 This is a schematic diagram of the architecture of the near-field and far-field joint prediction and two-way inversion device for structural parameters of the patch antenna array in this application.

[0070] Figure 7 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0071] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0072] To better understand the above technical solutions, exemplary embodiments of this disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art.

[0073] First Embodiment

[0074] Reference Figure 1 and Figure 2 This embodiment provides a method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays. The method includes the following steps:

[0075] S10, upon receiving the antenna array parameters of the patch antenna array to be tested, determine the parameter type corresponding to the antenna array parameters;

[0076] In this embodiment, the antenna array parameters are the relevant parameters of the patch antenna array, which are divided into structural parameters, near-field distribution and far-field performance indicators according to the parameter type.

[0077] In some alternative implementations, the structural parameters may include, but are not limited to: patch width. Patch length Array spacing Power supply position offset Array mode pattern, operating frequency Near-field distribution refers to the electric field intensity distribution on the observation plane above the antenna array, usually expressed as the field strength value on the spatial grid; far-field performance indicators include maximum gain and 3dB beamwidth.

[0078] After the parameter type is determined, when the input is determined to be a structural parameter, the forward prediction model is called to execute the forward inference process of "structural parameter to near field and then to far field performance"; when the input is determined to be a near field distribution or target far field performance index, the inverse model is called to execute the reverse design process of "target far field / near field to structural parameter".

[0079] S20, when the parameter type is structural parameter, a forward prediction model is used to predict the near-field distribution of the patch antenna array under test, and the near-field distribution is converted into a far-field radiation field and the target far-field performance index of the patch antenna array under test in the far-field radiation field is determined.

[0080] In the forward prediction branch, this embodiment models the array structure using a graph attention network, predicts the near-field distribution of the observation plane, and obtains the far-field radiation pattern and various far-field performance indicators based on the near-field to far-field transformation layer. In the inverse inversion branch, this embodiment uses a hybrid attention mechanism (GAT+Transformer) to solve for the matching structural parameters under the constraints of the near-field distribution and / or the target far-field performance, thereby achieving target performance-driven structural design.

[0081] It should be noted that, by using a graph neural network to model the antenna array, the topological relationships and electromagnetic coupling effects between array elements are explicitly represented, and a near-field to far-field physical transformation layer is integrated within the network, enabling rapid prediction from structural parameters to near-field and far-field performance.

[0082] Further and optionally, the forward prediction model employs a graph attention network, and the forward prediction model, in predicting the near-field distribution of the patch antenna array under test, specifically includes the following steps:

[0083] S21, Obtain the graph structure G=(V, E) of the patch antenna array under test, where V is the node set, each node corresponds to a patch unit; E is the edge set representing the electromagnetic coupling relationship between units;

[0084] S22, Based on the multi-head attention mechanism of the graph attention network, capture the electromagnetic coupling relationship between nodes in the node set to obtain the node characteristics of each node;

[0085] S23, aggregate the node features to map them to the near-field grid of the observation plane, and predict and output the complex values ​​of the electric / magnetic fields of each near-field grid through a fully connected layer to obtain the near-field distribution.

[0086] For example, a graph attention network (GAT) is used to update node features. For each layer of GAT, the attention weight of neighboring node j to node i is... Defined as:

[0087]

[0088] in,

[0089] in, For the first Layer nodes Features and For trainable parameters, Indicates splicing, As a non-linear activation function, multi-head attention can be achieved by applying several groups of... It is obtained by parallel computation and splicing.

[0090] The node is updated to:

[0091]

[0092] Through multi-layer GAT, the electromagnetic coupling modes between array elements can be adaptively learned.

[0093] The node features output from several layers of GAT are mapped onto the near-field grid of the observation plane through an aggregation operation, and the complex values ​​of the electric or magnetic field at each near-field grid point are predicted through a fully connected layer (MLP). The predicted near-field distribution can be expressed as:

[0094]

[0095] in, This represents the mapping function from node features to spatial near-field distribution. L represents the location point on the observation plane, and L is the GAT layer number.

[0096] In some alternative implementations, the aggregation operation may include taking the center subarray node, global pooling, or tiling it to a two-dimensional mesh.

[0097] Further, and optionally, the steps for determining the far-field radiation field specifically include:

[0098] S24, determine the equivalent current value of the near-field distribution. and equivalent magnetic current value :

[0099]

[0100]

[0101] In the formula, , Let these represent the complex vectors of the magnetic field intensity and the electric field intensity, respectively, in the near-field distribution. The unit normal vector representing the observation plane. This represents the position vector of the field point on the observation plane used to characterize the near-field distribution;

[0102] S25, based on the equivalent current value and equivalent magnetic current value Determine the far-field radiation field :

[0103]

[0104] In the formula, For wave number, For air wave impedance, To be observed as a unit vector, Let r be the magnetic permeability of the medium, r be the distance from the observation point to the origin, and j be the imaginary part. ω is the angular frequency of the electromagnetic wave.

[0105] For example, assume the observation plane is located at At that location, the normal vector is Equivalent current and equivalent magnetic current can be defined:

[0106]

[0107] The far-field electric field can be approximated as:

[0108]

[0109] When implemented on a discrete grid, the integral can be discretized into a summation, and further efficient computation can be achieved using FFT. This transformation is encapsulated as a differentiable layer in a neural network, allowing errors to be backpropagated to the near-field prediction module and the graph neural network parameters.

[0110] Further, and optionally, the step of determining the target far-field performance index of the patch antenna array under test in the far-field radiation field specifically includes:

[0111] S26. Determine the maximum gain after normalizing the far-field radiation field. :

[0112]

[0113] In the formula, This represents the normalized far-field radiation field. Represents the pitch angle in spherical coordinates. Represents the azimuth angle in spherical coordinates;

[0114] S27, Determine based on maximum gain Below, two angles on the preset principal tangent plane , :

[0115]

[0116] In the formula, This indicates the selected principal tangent azimuth angle; the preset principal tangent plane includes the E plane or the H plane;

[0117] S28, according to angle , Calculate the 3dB beamwidth :

[0118]

[0119] S29, maximum gain and 3dB beamwidth This is used as a target far-field performance indicator and output.

[0120] In some alternative implementations, the sidelobe level and zero-point direction can be calculated as far-field performance indicators for multi-objective optimization or as part of the training supervision signal.

[0121] S30, when the parameter type is near-field distribution and / or far-field performance index, the target structural parameters of the patch antenna array under test are inverted using the inverse inversion model under near-field distribution and equivalent source constraints.

[0122] In this embodiment, the reverse design process from "target far-field performance to near-field / equivalent source to structural parameters" specifically includes the following steps:

[0123] S31, when the input is a far-field performance index, the target far-field performance index is encoded into a conditional vector, and a transformed near-field distribution matching the target far-field is constructed through the inverse constraint of the near-field to far-field transformation layer or pre-training mapping.

[0124] S32, downsample the input near-field distribution and / or the transformed near-field distribution to form a field value grid corresponding one-to-one with the array nodes;

[0125] S33 uses the downsampled field values, predefined array patterns, and coupling features as node features to construct a graph structure input;

[0126] S34, capture the local electromagnetic coupling features of the graph structure input through the graph attention layer, and capture the global features through the Transformer layer;

[0127] S35, based on the local electromagnetic coupling characteristics and the global characteristics, predict the target structural parameters corresponding to each node.

[0128] It is worth noting that in steps S31 and S32, if the input parameter type is a near-field distribution, then it directly enters S32 for downsampling. If the parameter type is a far-field performance index, then it needs to be converted into a transformed near-field distribution that matches the target far field in step S31 before the downsampling step in S32 is executed.

[0129] In some alternative implementations, the inverse model employs a hybrid structure of several GAT layers at the beginning and several Transformer layers at the end. The GAT layers are used to capture electromagnetic coupling characteristics between cells within a local neighborhood; the Transformer layers model long-range dependencies across the entire array using a self-attention mechanism, adapting to different array topologies and sizes.

[0130] For nodes The inverse model outputs predicted structural parameters. And it is trimmed by predefined constraints (such as upper and lower limits of parameters, process constraints).

[0131] During the training phase, the actual structural parameters can be simulated using HFSS. The difference defines the inversion loss, for example, using SmoothL1 loss:

[0132]

[0133] in This refers to the structural parameters obtained through the reconstruction of the "structural parameters to near field to far field and then to inverse structural parameters" link.

[0134] In the technical solution provided in this embodiment, when receiving antenna array input, the input type is identified as structural parameters, near-field distribution, or target far-field performance index. When the input is structural parameters, a forward prediction model based on a graph attention network first predicts the target near-field distribution of the antenna array on the observation plane, then calculates the far-field radiation field through a near-field to far-field transformation layer with embedded physical priors, and further extracts the far-field performance index. When the input is near-field distribution and / or target far-field performance index, the array structural parameters that satisfy the target far-field performance are obtained by inversion under equivalent source and near-field constraints through an inverse inversion model.

[0135] Second Embodiment

[0136] Based on the first embodiment, considering the interference caused by strong electromagnetic coupling between adjacent units, this embodiment proposes to simulate the electromagnetic coupling effect between array units using mutual impedance or scattering parameters. Specifically, it includes the following steps:

[0137] S41, the mutual impedance between array element m and array element n is defined as:

[0138]

[0139] In the formula, Represents array cell Open circuit voltage, Represents array cell The excitation current, Indicates the unit Other unit port currents;

[0140] S42, mutual impedance The construction is based on an integral approximation of the current distribution and the induced electric field:

[0141]

[0142] In the formula, , These represent surface mount units. and unit The effective radiation surface area. Representation unit The intensity of the induced electric field generated in space Representation unit Surface current density distribution on;

[0143] S43, scattering parameters obtained based on simulation or measurement Define normalized coupling coefficients :

[0144]

[0145] In the formula, Represents array cell With array unit The mutual scattering parameters between them and Representing array units respectively and array unit The self-scattering parameters;

[0146] S44, the center distance between array cell m and array cell n mutual impedance and coupling coefficient Through fusion function The edge weights of the graph are obtained after preprocessing and weighted fusion. :

[0147]

[0148] In the formula, the fusion function This is a linear weighted mapping operation.

[0149] Specifically, firstly , , Dimensionless preprocessing was performed using min-max standardization to obtain normalized parameters. , , Then, edge weights are constructed by linearly superimposing preset weighting coefficients. The specific formula is as follows:

[0150]

[0151] In the formula, the minimum-maximum normalization mapping rule is: , represent , or The original parameters, , For the minimum and maximum values ​​of the corresponding parameter dataset, , , For pre-calibrated weighting coefficients and Wherein, the coupling coefficient corresponds to the weighting coefficient. Value greater than , To prioritize characterizing the electromagnetic coupling strength between units.

[0152] S45, edge weights The input is fed into a graph attention network to simulate the electrical coupling effect between array cells.

[0153] Third Embodiment

[0154] Based on the first embodiment, this embodiment provides a data normalization and training strategy on the basis of the aforementioned forward and inverse model structures, as follows:

[0155] First, the mean and standard deviation of the structural parameters, near-field data, and far-field data are calculated separately, and Z-score normalization is applied:

[0156]

[0157] Near-field data can be encoded using amplitude-phase or real-imaginary encoding, while far-field patterns can be represented using logarithmic amplitude (dB) to enhance the contrast between the main lobe and side lobes.

[0158] Next, define the joint loss function:

[0159]

[0160] in, The mean square error (MSE) of positive near-field prediction; This refers to the error in the far-field radiation pattern or far-field performance indicators. This refers to the error in the inverse structural parameter retrieval. This refers to cyclic consistency error; This is a weighting coefficient that can be adjusted based on the performance on the validation set.

[0161] Finally, the AdamW optimizer and cosine annealing learning rate scheduling strategy are used for training, and early stopping and gradient pruning are introduced to prevent overfitting and gradient explosion. The training data is generated through HFSS simulation and includes samples with different array sizes, operating frequencies, and target orientation patterns to enhance the model's generalization ability.

[0162] Verification Implementation Examples

[0163] To verify the effectiveness of the proposed "Joint Near-Field and Far-Field Prediction and Two-Way Inversion Method for Patch Antenna Arrays," this embodiment constructs training and testing datasets based on HFSS parameter scanning simulation data. For each array design point, it includes structural parameters, near-field electric field data of the observation surface, and far-field performance indicators extracted from the far-field radiation pattern. In this embodiment, the near-field observation surface is set 11 mm above the array. The near-field data selects the central grid region of the observation surface above the array's central element as the target prediction region, demonstrating the method's ability to learn coupling effects within a limited observation area and its inversion robustness when constraint information is insufficient.

[0164] (I) Forward prediction verification: from structural parameters to near-field distribution and then to far-field performance indicators

[0165] 1) Structural parameter to near-field distribution prediction

[0166] A forward prediction model based on graph attention networks is employed to map the array structure parameters to the near-field electric field amplitude (|E|) distribution at the observation surface. To visually demonstrate the prediction results, the minimum error, median error, and randomly representative samples are selected for comparative visualization. Figure 3 As shown in the figure, the true value (|E|), the predicted value (|E|), and the absolute error (|E|) are given. It can be seen that the predicted near field is consistent with the actual near field in terms of main energy distribution pattern, main hot spot location and overall spatial distribution trend. The error is mainly concentrated in the local high gradient region, indicating that the model can effectively learn the coupling effect between array units and achieve rapid prediction of near field distribution.

[0167] 2) Near-field to far-field performance index prediction

[0168] Based on the near-field representation, further predictions of far-field performance metrics are made, including the maximum gain in the E-plane and H-plane. The quantitative evaluation results on the test set are shown in Table 1.

[0169] Table 1. Quantitative evaluation results on the test set

[0170]

[0171] The corresponding scatter comparison results are as follows Figure 4 As shown, the predicted and true values ​​are closely aligned with the (y=x) reference line, indicating that the near-field and far-field joint prediction link of this application can output far-field performance indicators with high accuracy, verifying the effectiveness of the forward inference of the method.

[0172] (II) Inverse verification: from far-field performance indicators / near-field constraints to structural parameters

[0173] This embodiment further verifies the capability of "far-field performance index-driven structural parameter inversion". According to the method of this application, when the input is a far-field performance index (and / or near-field constraint information), the inverse inversion model outputs structural parameters that satisfy the target far-field performance. Considering that only a small number of far-field scalar indices (such as...) are used... When the beamwidth is 3dB, some structural parameters may be weakly sensitive or have multiple solutions. In this embodiment, we prioritize the inversion of the dominant structural parameters (array spacing (D) and patch length (L)) that are more sensitive to far-field indicators and can be stably identified, and introduce near-field features as constraint inputs to enhance the identifiability.

[0174] The inversion results on the test set are shown in Table 2:

[0175] Table 2. Inversion results on the test set

[0176]

[0177] The corresponding scatter comparison results are as follows Figure 5 As shown, the predicted values ​​generally closely follow the (y=x) reference line, indicating that the inverse model can accurately recover structural parameters under given far-field target constraints, thus verifying the feasibility and effectiveness of this application in a "performance-driven" structural inversion scenario. Regarding the feeder position offset... The weakly identifiable parameters can be further enhanced in engineering applications by introducing front-to-back ratio, sidelobe level, or more complete pattern constraints to improve the determinism of the inversion and achieve accurate prediction of each structural parameter; this extension does not affect the overall framework and implementation path of the method in this application.

[0178] (III) Efficiency Verification: Design cycle significantly shortened

[0179] Traditional array design processes typically require full-wave simulation iterations in HFSS. According to statistics from this embodiment, during the generation of this dataset, the average time for HFSS full-wave simulation, converted to a single data point, is approximately 56 minutes. When using the model described in this application for inference, the average inference time per sample is approximately 0.0117 seconds, corresponding to a speedup of approximately 2.88 × 10⁻⁶. 5 Therefore, the method of this application can significantly reduce the evaluation cycle from "inputting structural parameters to obtaining far-field performance indicators" from minutes to seconds, effectively supporting the rapid design and optimization of patch antenna arrays, thereby achieving the expected technical effect of "significantly shortening the design cycle".

[0180] (iv) Conclusion

[0181] In summary, this embodiment verifies the effectiveness of the proposed method from two aspects: the forward link (structural parameters to near-field to far-field indicators) and the reverse link (far-field indicators / near-field constraints to structural parameters). While ensuring the accuracy of far-field performance indicator prediction and structural parameter inversion, it significantly improves the evaluation efficiency of a single sample, which can be used for rapid design and performance-driven optimization of patch antenna arrays.

[0182] In addition, refer to Figure 6 As one implementation scheme, this embodiment also provides a device for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays, including:

[0183] The parameter type identification module 100 is used to determine the parameter type corresponding to the antenna array parameters when it receives the antenna array parameters of the patch antenna array under test.

[0184] The forward near-field and far-field prediction module 200 is used to predict the near-field distribution of the patch antenna array under test using a forward prediction model when the parameter type is structural parameter, and convert the near-field distribution into a far-field radiation field and determine the target far-field performance index of the patch antenna array under test in the far-field radiation field.

[0185] The inverse parameter inversion module 300 is used to invert the target structural parameters of the patch antenna array under test under near-field distribution and equivalent source constraints using an inverse inversion model when the parameter type is near-field distribution and / or far-field performance index.

[0186] As one implementation scheme, Figure 7 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0187] like Figure 7As shown, the computer system may include: a processor 1001, such as a CPU; a memory 1005; a user interface 1003; a network interface 1004; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0188] Those skilled in the art will understand that Figure 7 The computer system architecture shown does not constitute a limitation on the computer system and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0189] like Figure 7 As shown, the memory 1005, as a storage medium, may include an operating system, a network communication module, a user interface module, and computer programs. The operating system is a program that manages and controls the hardware and software resources of the computer system, as well as the operation of the computer programs and other software or programs.

[0190] exist Figure 7 In the computer system shown, the user interface 1003 is mainly used to connect to the terminal and communicate data with the terminal; the network interface 1004 is mainly used to communicate data with the backend server; and the processor 1001 can be used to call the computer program stored in the memory 1005.

[0191] In this embodiment, the computer system includes: a memory 1005, a processor 1001, and a computer program stored in the memory and executable on the processor, wherein:

[0192] When processor 1001 calls a computer program stored in memory 1005, it performs the following operations:

[0193] S10, upon receiving the antenna array parameters of the patch antenna array to be tested, determine the parameter type corresponding to the antenna array parameters;

[0194] S20, when the parameter type is structural parameter, a forward prediction model is used to predict the near-field distribution of the patch antenna array under test, and the near-field distribution is converted into a far-field radiation field and the target far-field performance index of the patch antenna array under test in the far-field radiation field is determined.

[0195] S30, when the parameter type is near-field distribution and / or far-field performance index, the target structural parameters of the patch antenna array under test are inverted using the inverse inversion model under near-field distribution and equivalent source constraints.

[0196] Furthermore, those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in a computer system to implement the process steps of the embodiments of the above methods.

[0197] Therefore, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the various steps of the near-field and far-field joint prediction and bidirectional inversion method for structural parameters of patch antenna arrays as described in the above embodiments.

[0198] The computer-readable storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0199] It should be noted that, since the storage medium provided in the embodiments of this application is the storage medium used to implement the methods of the embodiments of this application, those skilled in the art can understand the specific structure and variations of the storage medium based on the methods described in the embodiments of this application, and therefore will not be repeated here. All storage media used in the methods of the embodiments of this application fall within the scope of protection of this application.

[0200] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0201] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0202] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0203] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0204] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0205] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of a patch antenna array, characterized in that, The method includes the following steps: S10, upon receiving the antenna array parameters of the patch antenna array to be tested, determine the parameter type corresponding to the antenna array parameters; S20, when the parameter type is structural parameter, a forward prediction model is used to predict the near-field distribution of the patch antenna array under test, and the near-field distribution is converted into a far-field radiation field and the target far-field performance index of the patch antenna array under test in the far-field radiation field is determined. S30, when the parameter type is near-field distribution and / or far-field performance index, the target structural parameters of the patch antenna array under test are inverted using the inverse inversion model under near-field distribution and equivalent source constraints.

2. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 1, characterized in that, In step S20, the forward prediction model employs a graph attention network. The forward prediction model, in predicting the near-field distribution of the patch antenna array under test, includes the following steps: S21, Obtain the graph structure G=(V, E) of the patch antenna array under test, where V is the node set, each node corresponds to a patch unit; E is the edge set representing the electromagnetic coupling relationship between units; S22, Based on the multi-head attention mechanism of the graph attention network, capture the electromagnetic coupling relationship between nodes in the node set to obtain the node characteristics of each node; S23, aggregate the node features to map them to the near-field grid of the observation plane, and predict and output the complex values ​​of the electric / magnetic fields of each near-field grid through a fully connected layer to obtain the near-field distribution.

3. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 1 or 2, characterized in that, In step S20, the step of determining the far-field radiation field specifically includes: S24, determine the equivalent current value of the near-field distribution. and equivalent magnetic current value : ； ； In the formula, , Let these represent the complex vectors of the magnetic field intensity and the electric field intensity, respectively, in the near-field distribution. The unit normal vector representing the observation plane. This represents the position vector of the field point on the observation plane used to characterize the near-field distribution; S25, based on the equivalent current value and equivalent magnetic current value Determine the far-field radiation field : ； In the formula, For wave number, For air wave impedance, To be observed as a unit vector, Let r be the magnetic permeability of the medium, r be the distance from the observation point to the origin, and j be the imaginary part. ω is the angular frequency of the electromagnetic wave.

4. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 1 or 2, characterized in that, The target far-field performance indicators include maximum gain and 3dB beamwidth. In step S20, the step of determining the target far-field performance indicators of the patch antenna array under test in the far-field radiation field specifically includes: S26. Determine the maximum gain after normalizing the far-field radiation field. : ； In the formula, This represents the normalized far-field radiation field. Represents the pitch angle in spherical coordinates. Represents the azimuth angle in spherical coordinates; S27, Determine based on maximum gain Below, two angles on the preset principal tangent plane , : ； In the formula, This indicates the selected principal tangent azimuth angle; the preset principal tangent plane includes the E plane or the H plane; S28, according to angle , Calculate the 3dB beamwidth : ； S29, maximum gain and 3dB beamwidth This is used as a target far-field performance indicator and output.

5. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 1, characterized in that, In step S30, the step of determining the target structural parameters includes: S31, when the input is a far-field performance index, the target far-field performance index is encoded into a conditional vector, and a transformed near-field distribution matching the target far-field is constructed through the inverse constraint of the near-field to far-field transformation layer or pre-training mapping. S32, downsample the input near-field distribution and / or the transformed near-field distribution to form a field value grid corresponding one-to-one with the array nodes; S33 uses the downsampled field values, predefined array patterns, and coupling features as node features to construct a graph structure input; S34, capture the local electromagnetic coupling features of the graph structure input through the graph attention layer, and capture the global features through the Transformer layer; S35, based on the local electromagnetic coupling characteristics and the global characteristics, predict the target structural parameters corresponding to each node.

6. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 1, 2, or 5, characterized in that, The method further includes: S40 simulates the electrical coupling effect between array elements by using mutual impedance or scattering parameters.

7. The method for joint near-field and far-field prediction and bidirectional inversion of structural parameters of patch antenna arrays as described in claim 6, characterized in that, The steps for simulating the electrical coupling effect between array elements using mutual impedance or scattering parameters specifically include: S41, the mutual impedance between array element m and array element n is defined as: ； In the formula, Represents array cell Open circuit voltage, Represents array cell The excitation current, Indicates the unit Other unit port currents; S42, mutual impedance The construction is based on an integral approximation of the current distribution and the induced electric field: ； In the formula, , These represent surface mount units. and unit The effective radiation surface area. Representation unit The intensity of the induced electric field generated in space Representation unit Surface current density distribution on; S43, scattering parameters obtained based on simulation or measurement Define normalized coupling coefficients : ； In the formula, Represents array cell With array unit The mutual scattering parameters between them and Representing array units respectively and array unit The self-scattering parameters; S44, the center distance between array cell m and array cell n mutual impedance and coupling coefficient Through fusion function The edge weights of the graph are obtained after preprocessing and weighted fusion. : ； In the formula, the fusion function This is a linear weighted mapping operation; S45, edge weights The input is fed into a graph attention network to simulate the electrical coupling effect between array cells.

8. A device for joint near-field and far-field prediction and bidirectional inversion of structural parameters of a patch antenna array, characterized in that, include: The parameter type identification module is used to determine the parameter type corresponding to the antenna array parameters when the antenna array parameters of the patch antenna array under test are received. The forward near-field and far-field prediction module is used to predict the near-field distribution of the patch antenna array under test using a forward prediction model when the parameter type is structural parameter, and convert the near-field distribution into a far-field radiation field and determine the target far-field performance index of the patch antenna array under test in the far-field radiation field. The inverse parameter inversion module is used to invert the target structural parameters of the patch antenna array under test under near-field distribution and equivalent source constraints using an inverse inversion model when the parameter type is near-field distribution and / or far-field performance index.

9. A computer system, characterized in that, The computer system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the near-field and far-field joint prediction and bidirectional inversion method for the structure parameters of the patch antenna array as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the near-field and far-field joint prediction and bidirectional inversion method for structural parameters of the patch antenna array as described in any one of claims 1 to 7.

Images