Equivalent electromagnetic parameter acquisition method and device and electronic equipment

Abstract

The invention provides an equivalent electromagnetic parameter acquisition method and device and electronic equipment, and the method comprises the steps: obtaining a radiation characteristic curve of a to-be-measured target before equivalence; inputting the radiation characteristic curve of the to-be-measured target into a pre-constructed equivalent sparse parameter matrix, performing information fusion, obtaining a characteristic vector matrix, and extracting correlation characteristics of the to-be-measured target; calling a pre-trained information fusion network model, taking the relevance characteristics of the to-be-detected target as input, and obtaining equivalent electromagnetic parameters of the to-be-detected target; and inputting the equivalent electromagnetism of the to-be-measured target into a uniform dielectric plate model for full-wave simulation reconstruction of radiation characteristics, and determining final equivalent electromagnetic parameters of the to-be-measured target according to a reconstruction result and an information strategy. According to the invention, the accuracy of the final equivalent electromagnetic parameter can be improved, the method is not limited by the target structure of the to-be-measured target, and the problem of poor feasibility in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of electromagnetic technology, and in particular to a method, device and electronic equipment for acquiring equivalent electromagnetic parameters. Background Art

[0002] Electromagnetic scattering modeling is a key technical means to evaluate and improve the stealth performance of weapon equipment targets. However, in the microwave frequency band where radars work, common targets such as aircraft, missiles and their composite materials have the problem of coexistence of electrically large and electrically small sizes. Direct modeling will encounter limitations and challenges in computing resources. As a part of targets such as aircraft and missiles, the structure of composite materials is closely connected with the metal surface of the target. This overall structure can be regarded as a composite material containing a metal substrate. Therefore, efficient and effective characterization of the electromagnetic properties of composite materials containing metal substrates has become the key to solving the problem. However, the components of each phase material in the composite material usually have different proportions and shapes, and the overall electromagnetic properties of the composite material cannot be fully and accurately reflected based solely on the electromagnetic parameters of each component. By equating the composite material containing a metal substrate to a uniform dielectric plate and obtaining equivalent electromagnetic parameters, the final electromagnetic absorption characteristics of the composite material can be characterized. Therefore, obtaining the equivalent electromagnetic parameters of composite materials containing metal substrates is of great significance to meet the needs of practical applications.

[0003] The current methods for obtaining equivalent electromagnetic parameters can be divided into two categories: one is to start from the structure and electromagnetic parameters of each phase of the composite target, use electromagnetic theory to analyze the mutual influence of each component of the target's internal structure, and obtain the equivalent electromagnetic parameters of the structure. The corresponding electromagnetic theories include strong perturbation theory and homogenization theory, but this method is not suitable for use when the parameters of each phase of the composite material are unknown, and some composite targets are not suitable for the current theory. The other is to obtain equivalent electromagnetic parameters by inverting the scattering (S) coefficient. Since the scattering coefficient is easier to obtain and its application range is not limited by the structure of the composite target, this method is widely used. The transmission reflection method (Nicolson-Ross-Weir, NRW) based on the S coefficient obtains the corresponding electromagnetic parameters by reverse analysis of physical equations, which has the characteristics of simple operation and high accuracy. However, NRW needs to obtain equivalent electromagnetic parameters based on two parameters: reflection coefficient and transmission coefficient. In scenarios where composite targets need to add metal substrates, such as electronic equipment shielding and radar cross section (RCS) reduction, there is no transmission coefficient parameter, which will cause NRW to fail. Therefore, it is particularly important to develop a simple and automatic retrieval method for obtaining equivalent electromagnetic parameters of composite materials containing metal substrates.

[0004] There is currently no effective solution to the problem of poor feasibility in existing related technologies. Summary of the invention

[0005] The present invention provides an equivalent electromagnetic parameter acquisition method, device and electronic equipment, which are used to solve the defect of poor feasibility in the prior art.

[0006] In a first aspect, the present invention provides a method for obtaining equivalent electromagnetic parameters, comprising: Obtain the radiation characteristic curve of the target to be measured before equivalence; Inputting the radiation characteristic curve of the target to be measured into a pre-constructed equivalent sparse parameter matrix, performing information fusion, obtaining a feature vector matrix and extracting the correlation characteristics of the target to be measured; Retrieving a pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining equivalent electromagnetic parameters of the target to be measured; The radiation characteristics of the equivalent electromagnetic input uniform dielectric plate model of the target to be measured are reconstructed by full-wave simulation, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction result and information strategy.

[0007] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, the parameter matrix is ​​constructed, including: The equivalent sparse parameter matrix is ​​constructed, including: Obtaining the upper and lower value ranges of electromagnetic parameters in each phase material of the target to be measured before equivalence; the electromagnetic parameters include dielectric parameters and magnetic permeability; Based on the upper and lower value ranges of the electromagnetic parameters, sparse sampling is performed, and permutations and combinations are performed to construct a radiation characteristic curve and obtain a radiation characteristic parameter matrix.

[0008] The electromagnetic parameter sparse matrix is ​​substituted into an equivalent rear uniform dielectric plate model containing a metal substrate, and a parameterized full-wave simulation is performed on the equivalent rear uniform dielectric plate model to obtain the equivalent rear sparse parameter matrix.

[0009] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, training the information fusion network includes: Obtaining correlation characteristics of the isoradiative characteristic parameter matrix; The upper and lower correlation features between the electromagnetic parameter sparse matrix and the radiation characteristic parameter matrix are established to train the information fusion network; the data set input into the information fusion network for training is the processed radiation characteristic correlation features.

[0010] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, obtaining the correlation characteristics of the equivalent sparse parameter matrix includes: Decentralizing the parameter matrix and determining a covariance matrix; Based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, a correlation feature matrix is ​​determined; the correlation feature matrix is ​​composed of correlation features of the equivalent sparse parameter matrix.

[0011] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, the parameter matrix is ​​decentralized and a covariance matrix is ​​determined, including: Determining a mean value of each parameter sample in the parameter matrix, and generating an intermediate matrix based on a difference between the parameter matrix and the mean value; The covariance matrix is ​​determined based on the intermediate matrix.

[0012] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, the information fusion network model includes a pre-configured bidirectional gated recurrent unit and a convolutional layer; Before training the information fusion network, the following steps are included: The forward and backward sequence information between the correlation features of the parameter matrix is ​​learned by the bidirectional gated recurrent unit to obtain augmented key features; The augmented key features are processed longitudinally through the convolution layer.

[0013] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, training the information fusion network includes: The information fusion network is trained by taking the correlation features of the equivalent sparse parameter matrix as input and the electromagnetic parameters of the equivalent uniform dielectric plate model as labels.

[0014] According to an equivalent electromagnetic parameter acquisition method provided by the present invention, the equivalent electromagnetic input uniform dielectric plate model of the target to be measured is subjected to full-wave simulation to reconstruct the radiation characteristics, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction result and the information strategy, including: Re-inputting the equivalent electromagnetic parameters of the target to be measured into the uniform dielectric plate model to obtain an equivalent frequency-reflectivity curve; Obtain the frequency-reflectivity curve of the target to be measured before equivalence, and compare the resonance point position with the frequency-reflectivity curve after equivalence. Based on the comparison result, determine whether the equivalent electromagnetic parameters need to be adjusted based on the resonance point deviation, and obtain the final equivalent electromagnetic parameters of the target to be measured.

[0015] In a second aspect, the present invention further provides an equivalent electromagnetic parameter acquisition device, comprising: An acquisition module, used for acquiring a radiation characteristic curve of a target to be measured before equivalent; An extraction module, used to input the radiation characteristic curve of the target to be measured into a pre-built parameter matrix, perform information fusion, and extract the correlation characteristics of the target to be measured; A generation module is used to call the pre-trained information fusion network model, take the correlation characteristics of the target to be measured as input, and obtain the initial equivalent electromagnetic parameters of the target to be measured; The reconstruction module is used to reconstruct the initial equivalent electromagnetic parameters of the target to be measured, and determine the final equivalent electromagnetic parameters of the target to be measured according to the reconstruction result and the information strategy.

[0016] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for acquiring equivalent electromagnetic parameters as described in the first aspect above is implemented.

[0017] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the equivalent electromagnetic parameter acquisition method as described in the first aspect above.

[0018] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the equivalent electromagnetic parameter acquisition method as described in the first aspect above.

[0019] Compared with the prior art, the present invention has the following beneficial effects: The equivalent electromagnetic parameter acquisition method provided by the present invention determines the parameter range of the final equivalent electromagnetic parameters to be measured based on the electromagnetic parameter interval of each phase material of the target to be measured, and performs sparse sampling based on this range to obtain the equivalent electromagnetic parameter solution set of the target structure of the same type to be measured. The sparse sampling data is regarded as the equivalent electromagnetic parameter, and is substituted into the equivalent uniform dielectric plate model to obtain the radiation characteristic curve matrix. The radiation characteristic curve matrix is ​​subjected to global characteristic analysis and processing to obtain the correlation characteristics. The information fusion network model is used to process and obtain the equivalent electromagnetic parameters. Finally, the equivalent electromagnetic parameters are reconstructed and verified, and the position of the resonance point is compared to determine whether the equivalent electromagnetic parameters need to be optimized and adjusted. And correction is performed through the error threshold and information strategy of the information fusion network model. The above process is not limited by the target structure of the target to be measured, and eliminates the influence of the large deviation value of individual samples that may appear in the network model.

[0020] The correlation features in the above method not only have a qualitative relationship with their own equivalent electromagnetic parameters, but also fully consider the correlation changes between samples corresponding to different equivalent electromagnetic parameters. The processed correlation features are all represented based on the characteristic vector coordinate system. Each dimension of the correlation feature is the original different frequency points S11 Therefore, through the information fusion network model, after training, it can effectively learn the radiation characteristic curve of each dimension of electromagnetic parameters, that is, S 11 The influence of parameters is calculated to obtain super-approximate equivalent electromagnetic parameters.

[0021] This method is simpler and more universal, improves the accuracy of corresponding equivalent electromagnetic parameter retrieval, has stronger generalization, and solves the problem of poor feasibility in existing related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 is a flow chart of the method for obtaining equivalent electromagnetic parameters provided by the present invention; Figure 2 is a schematic diagram of the structure of a multi-output regression information fusion network in the present invention; Figure 3 1. It is a schematic diagram of the honeycomb model before and after the equivalent comparison in the present invention; Figure 4 This is a schematic diagram of the information proportion of the first 25 dimensions; Figure 5 This is the MAE error trend diagram of the real part of the equivalent electromagnetic dielectric constant of three different models; Figure 6 This is the MAE error trend diagram of the imaginary part of the equivalent electromagnetic dielectric constant of three different models; Figure 7 This is the MSE error trend diagram of the real part of the equivalent dielectric constant of three different models; Figure 8 This is the MSE error trend diagram of the imaginary part of the equivalent dielectric constant of three different models; Fig. 9 is the error distribution diagram of the honeycomb equivalent dielectric constant; Fig.10 is the error distribution diagram of honeycomb equivalent permeability; Fig.11 This is a parameter comparison diagram of a honeycomb with an absorption thickness of 0.3375 mm in the experiment of the present invention; Fig.12 This is a parameter comparison diagram of a honeycomb with an absorption thickness of 0.36 mm in the experiment of the present invention; Fig.13 It is a structural block diagram of the equivalent electromagnetic parameter acquisition device provided by the present invention; Fig.14 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The present invention provides a method for obtaining equivalent electromagnetic parameters. Figure 1 is a flow chart of the method for obtaining equivalent electromagnetic parameters provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step S101, obtaining a radiation characteristic curve of the target to be measured before equivalence; Step S102, inputting the radiation characteristic curve of the target to be measured into the pre-constructed equivalent sparse parameter matrix, performing information fusion, obtaining the feature vector matrix and extracting the correlation characteristics of the target to be measured; Step S103, calling the pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining the equivalent electromagnetic parameters of the target to be measured; Step S104, the equivalent electromagnetic input of the target to be measured is input into a uniform dielectric plate model for full-wave simulation to reconstruct radiation characteristics, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction result and the information strategy.

[0026] In this method, first, the radiation characteristic curve of the target to be measured before equivalent is obtained. The target to be measured is a honeycomb structure composite material. The radiation characteristic curve includes S 11 Parameters. Conventional inversion methods require the use of S 11 Parameters (reflectivity) and S 21 Parameter (transmittance), this application only needs to use S 11 Then, the radiation characteristic curve of the target to be measured is input into the equivalent sparse parameter matrix to extract the correlation features. Correspondingly, the equivalent sparse parameter matrix is ​​S 11 parameter matrix, and S 11 The parameter matrix is ​​constructed based on the sampling within the upper and lower limits of the electromagnetic parameters of each phase material of the target to be measured. The electromagnetic parameters are regarded as possible solutions of equivalent electromagnetic parameters after the target to be measured is equivalent, and the electromagnetic parameter sparse matrix is ​​regarded as a collection of equivalent electromagnetic parameter solutions of the same target structure. Extract different equivalent electromagnetic parameters corresponding to S 11The correlation characteristics of the parameter matrix. Specifically, different electromagnetic parameters to radiation characteristic parameters were originally independent, but in the present invention, the processed correlation characteristics to radiation characteristic parameters not only have a corresponding relationship, but also fuse the differentiated representations of different electromagnetic parameters to radiation characteristic parameters. Then the correlation characteristics of the target to be measured are input into the information fusion network model to obtain the equivalent electromagnetic parameters of the target to be measured. In this method, the information fusion network model adopts a multi-output regression information fusion network model (Multi-output Regression Information Fusion Network, MRIFN). After training, the information fusion network model can effectively learn the radiation characteristic curves of each dimension of the electromagnetic parameters, that is, S 11 The influence of the parameters, so the accuracy of the equivalent electromagnetic parameters obtained is higher. Finally, in order to eliminate the possible large deviation values, the equivalent electromagnetic parameters of the target to be measured are reconstructed to determine whether the position of the resonance point is consistent. If not, the initial equivalent electromagnetic parameters are optimized and adjusted, wherein the optimization adjustment amplitude range of the initial electromagnetic parameters is the maximum error threshold of the sparse sampling data corresponding to the initial electromagnetic parameters, and the adjustment direction is based on the information strategy of the electromagnetic parameters and the reflectivity curve, that is, based on the position relationship of the resonant frequency point and the position relationship of the resonant amplitude point, the super-approximate equivalent electromagnetic parameters of the target to be measured are obtained. By reconstructing and verifying the initial equivalent electromagnetic parameters, the influence of large deviation values ​​is eliminated, and the generalization of the method for retrieving the final equivalent electromagnetic parameters is further improved. Moreover, the above process is not limited by the target structure of the target to be measured, which solves the problem of poor feasibility in the existing related technologies.

[0027] In some of the embodiments, constructing an equivalent post-sparse parameter matrix includes: obtaining an upper and lower value range of electromagnetic parameters in each phase material of the target to be measured before equivalent; the electromagnetic parameters include dielectric parameters and magnetic permeability; performing sparse sampling based on the upper and lower value ranges of the electromagnetic parameters, and performing permutations and combinations to construct an electromagnetic parameter sparse matrix; substituting the electromagnetic parameter sparse matrix into an equivalent post-uniform dielectric plate model containing a metal substrate, performing parameterized full-wave simulation on the equivalent post-uniform dielectric plate model, and obtaining a radiation characteristic curve, i.e., S 11 parameter curve, and construct S 11 Parametric curve matrix.

[0028] On the basis of this embodiment, training the information fusion network includes: obtaining the correlation characteristics of the equivalent sparse parameter matrix; establishing the upper and lower correlation characteristics between the electromagnetic parameter sparse matrix and the equivalent sparse parameter matrix, and training the information fusion network; the data set input into the information fusion network for training is the processed radiation characteristic correlation characteristics. Among them, the electromagnetic parameters include two-output parameters and four-output parameters, and the specific types are determined according to the objects of different targets to be measured. For example, the equivalent electromagnetic parameters corresponding to non-magnetic microwave materials are two-output parameters, the real part and the imaginary part of the equivalent dielectric constant. For two-output and four-output targets to be measured, the number of layers and neurons of the information fusion network has been determined. The overshoot for training and adjusting the information fusion network model is only the batch size and the training batch, which reduces the degree of freedom of the model parameters. Specifically, obtaining the correlation characteristics of the equivalent sparse parameter matrix includes: decentralizing the parameter matrix and determining the covariance matrix; determining the correlation characteristic matrix based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; the correlation characteristic matrix is ​​composed of the correlation characteristics of the equivalent sparse parameter matrix. The correlation characteristics not only have a qualitative relationship with the equivalent electromagnetic parameters themselves, but also fully consider the correlation changes between samples corresponding to different equivalent electromagnetic parameters.

[0029] More specifically, the parameter matrix is ​​decentralized and the covariance matrix is ​​determined, including: determining the mean of each parameter sample in the parameter matrix, and generating an intermediate matrix based on the difference between the parameter matrix and the mean; and determining the covariance matrix based on the intermediate matrix.

[0030] For example, first, the upper and lower value ranges of the real and imaginary parts of the dielectric constants and the real and imaginary parts of the permeability of each phase of the target to be measured are obtained. Then, sparse sampling is performed within the respective ranges of the four dimensions, and permutations and combinations are performed, and the equivalent possible solutions of the equivalent electromagnetic parameters are substituted into the uniform dielectric plate model containing a metal substrate. The uniform dielectric plate model is parametrically simulated by the VBA (Visual Basic for Applications) script integrated with Matlab and CST to establish the radiation characteristic parameter matrix, namely S 11 Parameter matrix. Among them, a single parameter sample, for S 11 The -f curve is sampled at 0.1 GHz in the microwave frequency band of the radar operation, from 2 to 18 GHz, to obtain 161 S 11 Parameters. Construct the real and imaginary parts of the dielectric constant, the real and imaginary parts of the magnetic permeability, 161 S 11 Parameters of composite materials with metal substrates.

[0031] Due to the combination of electromagnetic parameters of a single target (real and imaginary parts of dielectric constant, real and imaginary parts of magnetic permeability) and S 11The -f curve is obtained through full-wave simulation and has a physical function relationship. However, by controlling variables, individual electromagnetic parameters of different targets (such as the real part of the dielectric constant) and S 11 The change of the resonance point in the -f curve has an obvious linear relationship. The corresponding information strategy is as follows: For the case of large loss, increase the imaginary part of the dielectric constant and the imaginary part of the magnetic permeability, and the corresponding S 11 The resonant frequency of the parameter will shift to low frequency, and the resonant amplitude will fluctuate up and down; for the case of smaller loss, increase the imaginary part of the dielectric constant and the imaginary part of the magnetic permeability, and the corresponding S 11 The resonant frequency of the parameter will not shift, but the resonant amplitude will have a small dB range deviation.

[0032] In order to fully integrate the correlation change characteristic information between different samples in the entire electromagnetic parameter range, the S 11 The parameter matrix is ​​global and features are extracted as follows: Group m into 161 S 11 The parameter features are converted into a new set of variables, and the original data set X is represented as follows:

[0033] in, m Represents the original data set X The total number of samples of the original data set X The parameter matrix is ​​decentralized to obtain the intermediate matrix X * , that is, subtract the mean of each parameter sample in the parameter matrix , the calculation formula of the parameter sample mean is as follows:

[0034] in, i and j is the position number of the parameter sample, m is the total number of samples. Then, calculate the intermediate matrix X * The covariance matrix of is as follows:

[0035] in, C represents the covariance matrix, C The elements on the upper diagonal are the S at each frequency point. 11 The variance of the parameter and itself. Then, calculate the eigenvalues ​​of the covariance matrix and the corresponding eigenvectors ,in, i =1,2,3..., k , the calculation formula is as follows:

[0036] in, v k Indicates k feature vectors, is a scalar. Reduce the dimension of the intermediate matrix to obtain the matrix Y , the specific formula is as follows:

[0037] in, W is the eigenvector matrix, , the variance information proportion of each correlation feature can be obtained by the following formula:

[0038] in, r i Indicates the proportion of variance information of correlation features. r i Calculate the cumulative information ratio, calculate the cumulative information ratio of the first few feature dimensions to reach 99%, and determine the corresponding feature dimension, that is, m × k The number of samples is the combination of the magnitude values ​​of the electromagnetic parameters in each dimension. Therefore, in addition to the qualitative relationship between the electromagnetic parameters of each sample, the correlation characteristics of each sample also fully consider the correlation changes between samples with different parameters.

[0039] In some embodiments, the information fusion network model includes a pre-configured bidirectional gated recurrent unit and a convolutional layer. Before training the information fusion network, it includes: learning the forward and backward sequence information between the correlation features of the parameter matrix through the bidirectional gated recurrent unit to obtain the augmented key features; and processing the augmented key features longitudinally through the convolutional layer.

[0040] Since S of samples with different parameters 11 The correlation characteristics of the parameter matrix are based on the eigenvector matrix The key features of each dimension are obtained from the original different frequency points S 11 The linear superposition of parameters, therefore, the multi-output regression information fusion network is used to deeply learn the effects of various electromagnetic parameter dimensions on S 11 The influence of parameter correlation characteristics. First, a bidirectional gated recurrent unit (BiGRU) is used to learn the corresponding S 11The forward and backward sequence information between parameter correlation features. The two-layer BiGRU expands the original key features several times. This approach can be regarded as a data augmentation form of key features in order to extract deeper feature relationships. The convolution layer processes the augmented key features processed horizontally by BiGRU vertically, further fuses the frequency information characteristics extracted by BiGRU, and gradually extracts and abstracts them. When the output is two target single values, the specific parameters of the corresponding multi-output regression information fusion network are as follows: Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of the multi-output regression information fusion network in the present invention. When the output is four target single values, the corresponding neurons can be doubled. Through the model structure of the above information fusion network model, the output of two target single values ​​and four target single values ​​can also be cited to prove that the information fusion network model has good generalization. The comparison of the model before and after equivalence is as follows Figure 3 As shown, Figure 3 : is a schematic diagram comparing the pre-equivalent model and the post-equivalent model in the present invention. In the figure, h Indicates the model height.

[0041] In some embodiments, the information fusion network is trained, including: taking the correlation characteristics of the equivalent sparse parameter matrix as input, taking the electromagnetic parameters of the equivalent uniform dielectric plate model as labels, and training the information fusion network. In this embodiment, the data for training the information fusion network model is discrete sparse sampling data within the upper and lower limits of the electromagnetic parameters, so after the model is trained, the maximum error threshold of the sparse sampling data can be used as the amplitude range value of the equivalent electromagnetic parameter optimization adjustment within the upper and lower limits of the entire electromagnetic parameter.

[0042] Exemplarily, the real part of the dielectric constant, the imaginary part of the dielectric constant, the real part of the magnetic permeability, and the imaginary part of the magnetic permeability are constructed, S 11 -f curve correlation feature data set. 11 -f curve correlation features as input, electromagnetic parameters as labels, training (multi-output regression information fusion network) MRIFN model. In addition, for this data set, the data set is divided into training set and validation set, and the training set is used to train the MRIFN model. After the MRIFN model is trained, the model accuracy is verified by the validation set, and the corresponding errors of different equivalent electromagnetic parameters are visualized, and the maximum error threshold of sparse data is used as the range value of the adjustment amplitude to be optimized for dense data.

[0043] In some embodiments, step S104, inputting the equivalent electromagnetic of the target to be measured into the uniform dielectric plate model for full-wave simulation to reconstruct the radiation characteristics, and determining the final equivalent electromagnetic parameters of the target to be measured according to the reconstruction result and the information strategy, includes: inputting the equivalent electromagnetic parameters of the target to be measured into the uniform dielectric plate model to obtain the equivalent S11 -f curve; obtain the S of the target to be tested before it is equivalent 11 -f curve, and the equivalent S 11 -f curve, and based on the comparison result, the equivalent electromagnetic parameters are optimized and adjusted to obtain the final equivalent electromagnetic parameters of the target to be measured. In this embodiment, the method of obtaining the frequency-reflectivity curve before equivalent includes but is not limited to the coaxial line method, microstrip line method, waveguide and free space method. The basis for optimization and adjustment is to calculate the equivalent electromagnetic parameter single variable S 11 -f curve linear influence, information strategy obtained, to judge and optimize the adjustment direction of equivalent electromagnetic parameters.

[0044] In the above method, the information strategy is as follows: 1. For the case of large loss, increase the imaginary part of the dielectric constant and the imaginary part of the magnetic permeability, and the corresponding S 11 The resonant frequency of the parameter will shift to low frequency, and the resonant amplitude will fluctuate up and down; 2. For the case of smaller loss, increase the imaginary part of the dielectric constant and the imaginary part of the magnetic permeability, and the corresponding S 11 The resonant frequency of the parameter will not shift, but the resonant amplitude will have a small dB range deviation. Determine the fine-tuning direction of the retrieved equivalent electromagnetic parameters, such as upward adjustment or downward adjustment.

[0045] Secondly, according to the maximum mean absolute error of the dielectric real part of the training model, the error range of different equivalent electromagnetic parameters is locked to determine the adjustment range. Figure 8 and Fig. 9 The Euclidean distance trained in the training is used to determine the adjustment range of the super-approximate equivalent electromagnetic parameters. It is worth noting that whether to optimize the adjustment is determined based on the reconstruction result, and which equivalent electromagnetic parameter dimension to optimize is a strategic choice. This method is suitable for eliminating the error problem of individual samples and has strong generalization for large-scale parameterized test targets.

[0046] In order to verify the effectiveness of the above method, the present invention constructs two data sets for verification. Data set 1 and data set 2 are respectively denoted as No.1 and No.2. The difference is that the target object to be measured in data set 1 is a non-magnetic microwave absorbing material, and the constructed data set is S 11 (reflectivity) is the input, and the real and imaginary parts of the effective electromagnetic parameters are the output (two-dimensional, two output values). The target object of data set 2 is the honeycomb structure composite material widely used in military fighters. The constructed data set is S 11 (reflectivity) is the input, and the real and imaginary parts of the effective electromagnetic parameters and the real and imaginary parts of the magnetic permeability are the outputs (four-dimensional, four output values).

[0047] In this experiment, we first use dataset No.1 as the model dataset and11 The parameters are processed to obtain their correlation characteristics. Figure 4 As shown, Figure 4 It is a schematic diagram of the information proportion of each of the first 25 dimensions. In the figure, the columnar part is the variance information proportion of the correlation characteristics, and the curve part represents the total information proportion. The sum of the information proportions of the first 15 correlation features is 99%, which can fully summarize the information of the original 161 features. Therefore, this experiment selects the first 15 correlation features as input features for reverse extraction of electromagnetic parameters.

[0048] In order to verify the effectiveness of the MRIFN model, this method uses the BiGRU (Bidirectional Gated Recurrent Unit) model and the CNN (Convolutional Neural Networks) model for comparative experiments. In order to compare the results in more detail, the data of 4000 test samples are fully displayed. Figure 5-8 As shown, Figure 5 This is the trend diagram of the real part error of the equivalent electromagnetic dielectric constant of three different models. Figure 6 This is the trend diagram of the imaginary part error of the equivalent electromagnetic dielectric constant of three different models. Figure 7 This is the trend diagram of the real part error of equivalent permeability of three different models. Figure 8 The following are the trends of the imaginary part errors of the equivalent permeability of three different models. Based on the above trends, the error distribution of the honeycomb equivalent electromagnetic parameters is as follows: Figure 9-10 As shown, Fig. 9 is the error distribution diagram of the honeycomb equivalent dielectric constant, Fig.10 It is the error distribution diagram of honeycomb equivalent magnetic permeability. Since different combinations of real and imaginary parts of dielectric constants constitute different composite materials, the same electromagnetic parameter value in the figure corresponds to multiple samples. It can be seen from the figure that the MRIFN model has a higher retrieval accuracy for a single composite material target. When the imaginary part of the dielectric constant is large, the error curves of the mean absolute error (MAE) of the real part of the dielectric and the mean square error (MSE) of the real part of the dielectric show an upward trend. This is because in the part with a large imaginary part of the dielectric constant, the S of individual samples is 11 The resonant amplitude of the curve is too low, which is inconsistent with the S 11 The curve trends are quite different. Nevertheless, the overall error remains within a small range, with MAE not exceeding 0.07 and MSE not exceeding 0.01. Therefore, the results of the MRIFN model are closer to the actual value and show better stability in the entire electromagnetic parameter range. The above MAE is used as the error threshold of the total equivalent electromagnetic parameter interval. Specifically, the calculation formula for the mean absolute error is as follows:

[0049] in, n represents the number of samples,i Indicates the sample number, Indicates the reference value, Represents the estimated value. The calculation formula of mean square error is as follows:

[0050] in, n represents the number of samples, i Indicates the sample number, Indicates the reference value, Indicates an estimated value.

[0051] In order to verify the superiority and generalization ability of this method, a comparative experiment was conducted with the BP neural network and RBP neural network prediction methods. At the same time, in order to study the situation where the scattering characteristics are affected by both the dielectric constant and the magnetic permeability, the data set No.2 was used for training verification. The experimental results are shown in Table 1: Table 1 Comparison of prediction errors output by different methods

[0052] In the table, ε' represents the real part of the equivalent electromagnetic dielectric constant, ε" represents the imaginary part of the equivalent electromagnetic dielectric constant, μ' represents the real part of the equivalent magnetic permeability, and μ" represents the imaginary part of the equivalent magnetic permeability. The evaluation indicators in Table 1 show that compared with other machine learning models, this method significantly improves the accuracy of the model. Since both the dielectric constant and the magnetic permeability are complex numbers, in order to better demonstrate the prediction effect, the Euclidean distance (ED) between the predicted value and the actual value of the real and imaginary parts of the dielectric constant is calculated, and the same is true for the magnetic permeability. Fig. 9 and Fig.10 The error distribution of samples corresponding to different electromagnetic parameter ranges is visualized, and the contour line of error ED=0.1 is set, which shows that the error ED of most samples is less than 0.1, and the error of samples in some parameter ranges is greater than 0.1 but less than 0.15.

[0053] To explore the equivalent pre-S 11 The influence of parameters on retrieval effect, this paper selects Figure 2 The three-layer honeycomb structure shown in the figure is used for verification. The aperture of the honeycomb structure is set to 0.5 mm, the thickness of the skeleton material is 0.05 mm, and the thickness of the absorbent material is 0.3375 mm and 0.36 mm. 11 The parameters are added to the equivalent sparse parameter data set to perform information fusion and extract correlation features. The extracted correlation features are normalized and input into the MRIFN model. The model output results are denormalized and the retrieved results are compared with the HS theory. The results are shown in Table 2: Table 2 Comparison of electromagnetic parameter retrieval results

[0054] In the table, Ours represents this method. The effective electromagnetic parameters retrieved by this method and the effective electromagnetic parameters obtained by HS theory are input into the uniform dielectric plate containing metal substrate for reconstruction and compared with the equivalent S 11 Parameters are compared. Fig.11 and Fig.12 As shown, Fig.11 This is a parameter comparison diagram of a honeycomb with an absorption thickness of 0.3375 mm in the experiment of the present invention. Fig.12 This is a parameter comparison diagram of a honeycomb with an absorption thickness of 0.36mm in the experiment of the present invention. The resonance point frequency obtained by this method is closer to the true value than the HS theory. There are slight differences between the resonance amplitude point and the data before equivalent, but this difference is within the acceptable 1dB error range in engineering practice. In addition, for the honeycomb of absorbent material with a thickness of 0.3375mm, the corresponding resonance point amplitude of this method is close to that of the HS theory, which is attributed to the fact that the loss tangent of the electromagnetic parameters under the two methods is similar, which makes the amplitude difference of the resonance point smaller.

[0055] Based on the above experiments, it can be seen that the real part of the equivalent electromagnetic parameters retrieved by this method is more accurate, and the characterization effect of the imaginary part is consistent with that calculated by HS theory. The overall effect is better than that of HS theory.

[0056] The present invention also provides an equivalent electromagnetic parameter acquisition device. The equivalent electromagnetic parameter acquisition device provided by the present invention is described below. The equivalent electromagnetic parameter acquisition device described below and the equivalent electromagnetic parameter acquisition method described above can refer to each other. Fig.13 is a structural block diagram of the equivalent electromagnetic parameter acquisition device provided by the present invention, such as Fig.13 As shown, the device comprises: An acquisition module 1301 is used to acquire a radiation characteristic curve of a target to be measured before equivalence; The extraction module 1302 is used to input the radiation characteristic curve of the target to be measured into the pre-constructed equivalent sparse parameter matrix, perform information fusion, obtain the feature vector matrix and extract the correlation characteristics of the target to be measured; The generation module 1303 is used to retrieve the pre-trained information fusion network model, take the correlation characteristics of the target to be measured as input, and obtain the equivalent electromagnetic parameters of the target to be measured; The reconstruction module 1304 is used to input the equivalent electromagnetic of the target to be measured into the uniform dielectric plate model for full-wave simulation to reconstruct the radiation characteristics, and determine the final equivalent electromagnetic parameters of the target to be measured according to the reconstruction result and the information strategy.

[0057] When the device is used, first, the acquisition module 1301 acquires the radiation characteristic curve of the target to be measured before equivalent. The target to be measured is a composite material. The radiation characteristic curve includes S 11 Then, the extraction module 1302 inputs the radiation characteristic curve of the target to be measured into the parameter matrix to extract the correlation features. Correspondingly, the parameter matrix is ​​S 11 Parameter matrix, and the parameter matrix is ​​constructed based on composite materials of the same type as the target to be measured. The generation module 1303 then inputs the correlation characteristics of the target to be measured into the information fusion network model to obtain the initial equivalent electromagnetic parameters of the target to be measured. In this method, the information fusion network model adopts a multi-output regression information fusion network model (Multi-output Regression Information Fusion Network, MRIFN). Finally, in order to improve the accuracy of the obtained results, the reconstruction module 1304 reconstructs the initial equivalent electromagnetic parameters of the target to be measured, verifies the error of the initial equivalent electromagnetic parameters, and fine-tunes the initial equivalent electromagnetic parameters to obtain the final equivalent electromagnetic parameters of the target to be measured. In the above process, for the target to be measured, a global feature analysis and processing is performed based on the parameter matrix constructed based on the electromagnetic parameters of the same material as it, and the correlation characteristics are obtained, and the information fusion network model is used for processing to obtain the initial equivalent electromagnetic parameters. After training, the information fusion network model can effectively learn the radiation characteristic curve of each dimension of the electromagnetic parameters, that is, S 11 The influence of the parameters is eliminated, so the authenticity of the obtained initial equivalent electromagnetic parameters is better. Finally, the initial equivalent electromagnetic parameters are reconstructed and verified to further improve the accuracy of the final equivalent electromagnetic parameters. Moreover, the above process is not limited by the target structure of the target to be measured, which solves the problem of poor feasibility in the existing related technologies.

[0058] Fig.14 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.14 As shown, the electronic device may include: a processor 1401, a communication interface 1402, a memory 1403 and a communication bus 1404, wherein the processor 1401, the communication interface 1402 and the memory 1403 communicate with each other through the communication bus 1404. The processor 1401 may call the logic instructions in the memory 1403 to execute the equivalent electromagnetic parameter acquisition method, which includes: Obtain the radiation characteristic curve of the target to be measured before equivalence; The radiation characteristic curve of the target to be measured is input into the pre-constructed equivalent sparse parameter matrix, information fusion is performed, the feature vector matrix is ​​obtained and the correlation characteristics of the target to be measured are extracted; Retrieving the pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining the equivalent electromagnetic parameters of the target to be measured; The radiation characteristics of the target to be measured are reconstructed by full-wave simulation of the equivalent electromagnetic input uniform dielectric plate model, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction results and information strategy.

[0059] In addition, the logic instructions in the above-mentioned memory 1403 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0060] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the equivalent electromagnetic parameter acquisition method provided by the above methods, the method includes: Obtain the radiation characteristic curve of the target to be measured before equivalence; The radiation characteristic curve of the target to be measured is input into the pre-constructed equivalent sparse parameter matrix, information fusion is performed, the feature vector matrix is ​​obtained and the correlation characteristics of the target to be measured are extracted; Retrieving the pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining the equivalent electromagnetic parameters of the target to be measured; The radiation characteristics of the target to be measured are reconstructed by full-wave simulation of the equivalent electromagnetic input uniform dielectric plate model, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction results and information strategy.

[0061] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the equivalent electromagnetic parameter acquisition method provided by the above methods, the method comprising: Obtain the radiation characteristic curve of the target to be measured before equivalence; The radiation characteristic curve of the target to be measured is input into the pre-constructed equivalent sparse parameter matrix, information fusion is performed, the feature vector matrix is ​​obtained and the correlation characteristics of the target to be measured are extracted; Retrieving the pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining the equivalent electromagnetic parameters of the target to be measured; The radiation characteristics of the target to be measured are reconstructed by full-wave simulation of the equivalent electromagnetic input uniform dielectric plate model, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction results and information strategy.

[0062] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0063] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for obtaining equivalent electromagnetic parameters, characterized in that: include: Obtain the radiation characteristic curve of the target to be measured before equivalence; Inputting the radiation characteristic curve of the target to be measured into a pre-constructed equivalent sparse parameter matrix, performing information fusion, obtaining a feature vector matrix and extracting the correlation characteristics of the target to be measured; Retrieving a pre-trained information fusion network model, taking the correlation characteristics of the target to be measured as input, and obtaining equivalent electromagnetic parameters of the target to be measured; The radiation characteristics of the equivalent electromagnetic input uniform dielectric plate model of the target to be measured are reconstructed by full-wave simulation, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction result and information strategy.

2. The method for obtaining equivalent electromagnetic parameters according to claim 1, characterized in that: Constructing the equivalent sparse parameter matrix includes: Obtaining the upper and lower value ranges of electromagnetic parameters in each phase material of the target to be measured before equivalence; the electromagnetic parameters include dielectric parameters and magnetic permeability; Based on the upper and lower value ranges of the electromagnetic parameters, sparse sampling is performed, and permutations and combinations are performed to construct an electromagnetic parameter sparse matrix; The electromagnetic parameter sparse matrix is ​​substituted into an equivalent rear uniform dielectric plate model containing a metal substrate, a parameterized full-wave simulation is performed on the equivalent rear uniform dielectric plate model, a radiation characteristic curve is constructed, and a radiation characteristic parameter matrix is ​​obtained.

3. The method for obtaining equivalent electromagnetic parameters according to claim 2, characterized in that: Training the information fusion network includes: Obtaining correlation characteristics of the radiation characteristic parameter matrix; The upper and lower correlation features between the electromagnetic parameter sparse matrix and the radiation characteristic parameter matrix are established to train the information fusion network; the data set input into the information fusion network for training is the processed radiation characteristic correlation features.

4. The method for obtaining equivalent electromagnetic parameters according to claim 3, characterized in that: Obtaining the correlation characteristics of the radiation characteristic parameter matrix includes: Decentralizing the parameter matrix and determining a covariance matrix; Based on the eigenvalues ​​and corresponding eigenvectors of the covariance matrix, a correlation characteristic matrix is ​​determined; the correlation characteristic matrix is ​​composed of correlation characteristics of the radiation characteristic parameter matrix.

5. The method for obtaining equivalent electromagnetic parameters according to claim 4, characterized in that: The parameter matrix is ​​decentralized and a covariance matrix is ​​determined, including: Determining a mean value of each parameter sample in the parameter matrix, and generating an intermediate matrix based on a difference between the parameter matrix and the mean value; The covariance matrix is ​​determined based on the intermediate matrix.

6. The method for obtaining equivalent electromagnetic parameters according to claim 3, characterized in that: The information fusion network model includes a pre-configured bidirectional gated recurrent unit and a convolutional layer; Before training the information fusion network, the following steps are included: The forward and backward sequence information between the correlation features of the parameter matrix is ​​learned by the bidirectional gated recurrent unit to obtain augmented key features; The augmented key features are processed longitudinally through the convolution layer.

7. The method for acquiring equivalent electromagnetic parameters according to claim 3, characterized in that: Training the information fusion network includes: The information fusion network is trained by taking the correlation features of the radiation characteristic parameter matrix as input and taking the electromagnetic parameters of the equivalent uniform dielectric plate model as labels.

8. The method for acquiring equivalent electromagnetic parameters according to claim 2, characterized in that: The equivalent electromagnetic input uniform dielectric plate model of the target to be measured is subjected to full-wave simulation to reconstruct radiation characteristics, and the final equivalent electromagnetic parameters of the target to be measured are determined according to the reconstruction result and the information strategy, including: Re-inputting the equivalent electromagnetic parameters of the target to be measured into the uniform dielectric plate model to obtain an equivalent frequency-reflectivity curve; The frequency-reflectivity curve of the target to be measured before equivalence is obtained, and the resonance point position is compared with the frequency-reflectivity curve after equivalence, and the equivalent electromagnetic parameters are optimized and adjusted based on the comparison result to obtain the final equivalent electromagnetic parameters of the target to be measured.

9. An equivalent electromagnetic parameter acquisition device, characterized in that: include: An acquisition module, used for acquiring a radiation characteristic curve of a target to be measured before equivalent; An extraction module, used for inputting the radiation characteristic curve of the target to be measured into a pre-constructed equivalent sparse parameter matrix, performing information fusion, obtaining a feature vector matrix and extracting the correlation characteristics of the target to be measured; A generation module is used to call the pre-trained information fusion network model, take the correlation characteristics of the target to be measured as input, and obtain the equivalent electromagnetic parameters of the target to be measured; The reconstruction module is used to perform full-wave simulation on the equivalent electromagnetic input of the target to be measured into a uniform dielectric plate model to reconstruct the radiation characteristics, and determine the final equivalent electromagnetic parameters of the target to be measured according to the reconstruction result and information strategy.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the equivalent electromagnetic parameter acquisition method as described in any one of claims 1 to 8 is implemented.

Images