CN115719027A - Method for realizing antenna design and related equipment - Google Patents
Method for realizing antenna design and related equipment Download PDFInfo
- Publication number
- CN115719027A CN115719027A CN202110972826.4A CN202110972826A CN115719027A CN 115719027 A CN115719027 A CN 115719027A CN 202110972826 A CN202110972826 A CN 202110972826A CN 115719027 A CN115719027 A CN 115719027A
- Authority
- CN
- China
- Prior art keywords
- model
- antenna
- deep
- kriging
- parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Variable-Direction Aerials And Aerial Arrays (AREA)
Abstract
The disclosure provides an antenna design realization method, an antenna design realization device, electronic equipment and a storage medium. The method for realizing the antenna design comprises the following steps: acquiring training data, wherein the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof; processing the design parameters of the antenna through a deep kriging model to obtain a prediction result; determining a fitness function of a genetic algorithm according to the simulation result and the prediction result of the antenna; and training the deep kriging model according to the fitness function to obtain a target deep kriging model. The method can reduce the dimension of input data and reduce the electromagnetic simulation times, thereby reducing the design time and improving the efficiency of antenna design.
Description
Technical Field
The present disclosure relates to the field of communications technologies, and in particular, to an antenna design implementation method and apparatus, an electronic device, and a storage medium.
Background
As SA/NSA (independent networking/non-independent networking) 5G (5 th Generation Mobile Communication Technology) systems are developed and deployed worldwide, the demand for designing 5G millimeter wave antennas is increasing, and the antennas are used as transmitting and receiving devices of Communication systems, and the performance thereof is especially important for the whole Communication systems.
Microstrip Patch Antennas (MPA) still have strong attraction for millimeter wave applications due to their advantages of small size, easy manufacturing, light weight, planar structure, etc. Substrate Integrated Waveguide (SIW) is a planar form of composite Rectangular Waveguide (RWG), which can be applied to many microwave, millimeter wave antennas and Radio Frequency (RF) circuits, and has the advantages of low loss, low cost, high power handling capability, and coplanar integration.
In the related art, the antenna design method can be used for completing the antenna design with a simpler structure by using electromagnetic simulation and manual parameter adjustment, but is not suitable for the antenna with a complex structure; by the method of designing the antenna through the proxy model, the calculation difficulty is increased and the calculation cost is overhigh due to overlarge input data dimension.
It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present disclosure and therefore may include information that does not constitute prior art known to a person of ordinary skill in the art.
Disclosure of Invention
The present disclosure provides an antenna design implementation method, an antenna design implementation apparatus, an electronic device, and a storage medium, where the method can reduce dimensionality of input data and reduce electromagnetic simulation times, thereby reducing design time and improving antenna design efficiency.
Additional features and advantages of the disclosure will be set forth in the detailed description which follows, or in part will be obvious from the description, or may be learned by practice of the disclosure.
The embodiment of the present disclosure provides an implementation method of an antenna design, including: acquiring training data, wherein the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof; processing the design parameters of the antenna through a deep kriging model to obtain a prediction result; determining a fitness function of a genetic algorithm according to the simulation result and the prediction result of the antenna; and training the deep kriging model according to the fitness function to obtain a target deep kriging model.
In some exemplary embodiments of the present disclosure, the deep kriging model includes an input layer of the convolutional neural network model, a convolutional layer of the convolutional neural network model, a pooling layer of the convolutional neural network model, a kriging proxy model, and an output layer of the convolutional neural network model.
In some exemplary embodiments of the present disclosure, the method further includes: constructing an antenna initial model according to the structural parameters of the antenna; randomly generating design parameters of a plurality of groups of antennas; and inputting the design parameters of the plurality of groups of randomly generated antennas into electromagnetic simulation software for simulation to obtain the simulation result of the plurality of groups of antennas.
In some exemplary embodiments of the present disclosure, processing the design parameters of the antenna through a deep kriging model to obtain a prediction result includes: initializing the deep kriging model, taking the network parameters of the convolutional neural network and the hyper-parameters of the kriging proxy model as the particle number of the genetic algorithm, and arranging the particle number in sequence; initializing the genetic algorithm, and inputting the design parameters of the antenna into the deep kriging model to obtain the prediction result.
In some exemplary embodiments of the present disclosure, determining a fitness function of a genetic algorithm according to the simulation result and the prediction result of the antenna includes: and taking the mean square error of the prediction result and the simulation result as a fitness function of the genetic algorithm.
In some exemplary embodiments of the present disclosure, training the deep kriging model according to the fitness function to obtain a target deep kriging model includes: when the fitness function meets a preset condition, completing the training of the deep kriging model to obtain target model parameters of the deep kriging model; and carrying out particle decoding on the target model parameters according to the sequence to obtain the target depth Krigin model.
In some exemplary embodiments of the present disclosure, the method further includes: acquiring actual design parameters of a plurality of groups of antennas; and simulating the actual design parameters of the multiple groups of antennas according to the target depth Kriging model to obtain the target design parameters of the antennas.
In some exemplary embodiments of the present disclosure, the antenna is a substrate integrated waveguide aperture coupled microstrip patch antenna.
The embodiment of the present disclosure provides an apparatus for implementing antenna design, including: the training data acquisition module is used for acquiring training data, and the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof; the prediction result obtaining module is used for processing the design parameters of the antenna through a deep kriging model to obtain a prediction result; a fitness function determining module for determining a fitness function of a genetic algorithm according to the simulation result of the antenna and the prediction result; and the model training module is used for training the deep kriging model according to the fitness function so as to obtain a target deep kriging model.
An embodiment of the present disclosure provides an electronic device, including: at least one processor; and the storage terminal device is used for storing at least one program, and when the at least one program is executed by at least one processor, the at least one processor is enabled to realize the implementation method of any one of the antenna designs.
The embodiment of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, wherein the computer program is executed by a processor to implement any one of the above-mentioned methods for implementing an antenna design.
According to the method for realizing the antenna design, the design parameters of the antenna are processed through the deep kriging model, so that the dimensionality of input data can be reduced, the electromagnetic simulation times can be reduced, the design time can be shortened, and the efficiency of the antenna design can be improved; the parameters of the deep kriging model are optimized by using the genetic algorithm, so that the network training time can be greatly shortened, and the local optimization problem of the back propagation optimization algorithm in the related technology is solved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure. It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those skilled in the art without the exercise of inventive faculty.
Fig. 1 is a flow diagram illustrating a method of implementing an antenna design in accordance with an exemplary embodiment.
Fig. 2 is a schematic diagram of a SIW aperture-coupled microstrip patch antenna configuration according to an example.
FIG. 3 is a schematic diagram illustrating the structure of a deep Kriging model in accordance with an exemplary embodiment.
Fig. 4 is a schematic diagram illustrating the structure of another deep kriging model in accordance with an exemplary embodiment.
Fig. 5 is a graph illustrating antenna predictions for a deep kriging model in comparison to simulation results output by simulation software according to an example.
Fig. 6 is a block diagram illustrating an apparatus for implementing one antenna design in accordance with an example embodiment.
Fig. 7 is a schematic diagram of a structure of an electronic device according to an exemplary embodiment.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor terminal devices and/or microcontroller terminal devices.
Hereinafter, the steps of the method for implementing the antenna design in the exemplary embodiment of the present disclosure will be described in more detail with reference to the drawings and the exemplary embodiment.
Fig. 1 is a flow diagram illustrating a method of implementing an antenna design in accordance with an exemplary embodiment.
As shown in fig. 1, an implementation method of an antenna design provided in the embodiments of the present disclosure may include the following steps.
In step S102, training data is obtained, where the training data includes design parameters of multiple groups of antennas and simulation results thereof.
In an exemplary embodiment, the antenna is a Substrate Integrated Waveguide (SIW) aperture coupled microstrip patch antenna.
It should be noted that the implementation method of the antenna design provided in the embodiment of the present disclosure may be applied to various types of antennas, and the SIM aperture coupled microstrip patch antenna is taken as an example for description below, but the present disclosure is not limited thereto.
In an exemplary embodiment, the above training data may be obtained by: constructing an antenna initial model according to the structural parameters of the antenna; randomly generating design parameters of a plurality of groups of antennas; and inputting the design parameters of the plurality of groups of randomly generated antennas into electromagnetic simulation software for simulation to obtain the simulation results of the plurality of groups of antennas.
For example, an antenna initial model, which may be an antenna structure model, may be constructed according to a SIW aperture-coupled microstrip patch antenna that needs to be designed.
For example, the design parameters of multiple sets of antennas may be randomly generated, with different parameters resulting in different performance and frequency response points for SIW aperture-coupled microstrip patch antennas.
Fig. 2 is a schematic diagram of a SIW aperture-coupled microstrip patch antenna configuration according to an example.
Referring to fig. 2, for example, an antenna with a dielectric constant of 2.33 and a loss tangent angle of 0.002 may be selected in which the upper and lower dielectric layers are made of low-loss material RT Duroid 5870. The microstrip patch antenna is coupled on the surface of the upper dielectric plate, and the lower dielectric plate is correspondingly provided with a slot. The thickness of the upper dielectric layer and the lower dielectric layer can be 1.575mm, and the length and width parameters can be 16mm and 10mm respectively.
For example, size parameters of 7 sets of SIW aperture-coupled microstrip patch antennas can be randomly generated as variables, and the size parameters are randomly combined to be used as input data of the deep kriging model, and specific parameters are shown in table 1.
TABLE 1
Where d denotes the diameter of the aperture of the antenna, s denotes the distance between the apertures, ys denotes the distance between the first row of apertures and the coupling slot, lslot denotes the length of the coupling slot, wslot denotes the width of the coupling slot, lmpa denotes the length of the patch, and Wmpa denotes the width of the patch.
For example, 200 sets of different antenna parameters are randomly combined to call electromagnetic simulation software HFSS (High Frequency Structure Simulator) for simulation, the obtained simulation result is used as training output, and 7 antenna size parameters of each set are used as training input, so that the deep kriging model is trained.
In step S104, the design parameters of the antenna are processed by the deep kriging model, and a prediction result is obtained.
In an exemplary embodiment, the deep kriging model includes an input layer of the convolutional neural network model, a convolutional layer of the convolutional neural network model, a pooling layer of the convolutional neural network model, a kriging proxy model, and an output layer of the convolutional neural network model.
The deep Kriging Model in the embodiment of the present disclosure may be a combination of a Convolutional Neural network Model (CNN) and a Kriging proxy Model (KM), and the deep Kriging Model may be used instead of a full connection layer of the Convolutional Neural network Model.
FIG. 3 is a schematic diagram illustrating the structure of a deep Kriging model in accordance with an exemplary embodiment.
FIG. 4 is a diagram illustrating the structure of another deep Kriging model in accordance with an exemplary embodiment.
Referring to fig. 3 and 4, in an embodiment of the present disclosure, the deep kriging model may include an input layer, a convolutional layer, a pooling layer, a kriging proxy model, and an output layer, where there may be one or more of the convolutional layer and the pooling layer, which is not limited by the present disclosure.
The following description will be given by taking an example in which the deep kriging model includes two convolution layers and two pooling layers. The input layer, the convolutional layer 1, the pooling layer 1, the convolutional layer 2, the pooling layer 2, the kriging proxy model and the output layer of the deep kriging model are sequentially connected, the input layer is used for inputting data, the convolutional layer is used for keeping characteristic quantity of the input data, the pooling layer is used for reducing data dimension, and the kriging proxy model is used for outputting a predicted object.
The model structure of the deep kriging model in the embodiment of the disclosure is evolved from LeNet-5 (a common conventional structure of a convolutional neural network) in general, but is more flexible than LeNet-5, the overall structure thereof is improved, two convolutional layers and pooling layers are provided, and multiple convolutions or the order of convolution pooling can be changed according to the size of data.
In an exemplary embodiment, the processing the design parameters of the antenna through the deep kriging model to obtain the prediction result may include: initializing a deep kriging model, taking network parameters of a convolutional neural network and hyper-parameters of a kriging proxy model as the number of particles of a genetic algorithm, and arranging the particles in sequence; initializing a genetic algorithm, and inputting design parameters of the antenna into the deep kriging model to obtain a prediction result.
The network parameters of the convolutional neural network may include a weight threshold and an offset. After the input data of the model is subjected to convolution operation, the number of neurons in the characteristic surface of the convolution layer or the size of the characteristic surface is equal to the following formula:
wherein Outsize represents the number of output characteristic surface neurons; inSize represents the number of input characteristic surface neurons; CSize is the size of the convolution kernel and CInterval represents the step size of the sliding translation of the convolution kernel.
In general, the division term operation result of the formula is guaranteed to be an integer, so that additional subsequent processing on the network can be omitted, and calculation of the subsequent network is facilitated. Thus, convolutional layers may have trainable parameters as follows:
CPN＝(InSize×CSize+1)×OutSize
wherein, CPN is the number of training parameters, and InSize is the number of input characteristic facial neurons; CSize represents the size of the convolution kernel; 1 represents the number of thresholds, typically only one shared threshold is set per layer; outsize represents the number of output characteristic surface neurons.
The hyper-parameters of the kriging proxy model may be trainable parameters of the kriging model itself.
For example, consider a sample set X = [ X ] with m design points (1) ,x (2) ,...,x (m) ] T And the dimension of each design point x is n. The response corresponding to the sample set is Y = [ Y (1) ,y (2) ,...,y (m) ] T The dimension of the response value is q. The design point and response of the sample are normalized to satisfy the following equation:
μ[X :,j ]＝1 Var[X :,j ,X :,j ]＝1 j＝1,2,...,n
μ[Y :,j ]＝1 Var[Y :,j ,Y :,j ]＝1 j＝1,2,...,q
wherein X :,j Represents the jth column vector, μ, of matrix X.]And Var [.]Mean and covariance are indicated.
The basis function vectors of the design points in the sample set form a design matrix F with m rows and p columns:
F＝[f(x (1) ),f(x (2) ),...,f(x (m) )] T
further, R represents a matrix of m rows and m columns of m correlation functions between two of the m design points:
R ij ＝R(θ,x (i) ,x (j) )i,j∈{1,2,...,m}
for a design point x waiting for prediction, r (x) represents a vector of m rows and 1 columns formed by correlation function values between the point and the current m design points:
r(x)＝[R(θ,x,x (1) ),R(θ,x,x (2) ),...,R(θ,x,x (m) )] T
the correlation function R (θ, x, w) is a gaussian-type correlation function, and is expressed as follows:
let d j ＝w j -x j Considering only a single dimension, where θ is a hyper-parameter of the correlation function, the optimum of which can pass through the poleThe large likelihood method obtains:
for convenience of description, it can be assumed that each design point x (i) Corresponding response y (i) Q =1 since the calculations performed for each dimension of the response to build the model are identical, i.e. β = β is implied in the formula listed subsequently :,1 And Y = Y :,1 。
Consider first the existence of a linear predictor:
where c = c (x) is an m-dimensional vector and is a function of x. The predictor-to-sample error can be expressed as:
wherein Z = [ Z ] (1) ,z (2) ,...,z (m) ] T Showing the observed error at the design point.
In order for the predictor to need to satisfy the unbiased estimation condition, it needs to satisfy:
F T c-f(x)＝0
under this condition, the mean square error of the predictor is:
c (x) is solved using the lagrange method to minimize the mean square error phi (x), the lagrange equation being as follows:
L(c,λ)＝σ 2 (1+c T Rc-2c T r)-λ T (F T c-f)
the derivative of c is obtained
L c '(c,λ)＝2σ 2 (Rc-r)-Fλ
The equation set can be obtained by combining the following equations:
F T c＝f
wherein
Solving the equation yields:
due to the need to solve the inverse R of the correlation matrix R -1 Are all symmetrical, one can get:
in the case of the regression problem,
the generalized least squares solution under the correlation function R is:
β*＝(F T R -1 F) -1 F T R -1 Y
in combination with the above, a final expression of the kriging model can be obtained:
in the embodiment of the present disclosure, a deep kriging model may be initialized first, and the network parameters of the convolutional neural network and the hyper-parameters of the kriging proxy model are used as the number of particles of the genetic algorithm and are arranged in sequence; initializing a genetic algorithm, and inputting design parameters of the antenna into the deep kriging model to obtain a prediction result.
In step S106, an fitness function of the genetic algorithm is determined based on the simulation result and the prediction result of the antenna.
In the basic genetic algorithm, only 3 basic genetic operators, namely a selection operator, a crossover operator and a mutation operator, are generally used, and the genetic evolution operation process is simple and easy to understand, and is the prototype and the basis of other genetic algorithms. The components of the basic genetic algorithm may include:
(1) Chromosome coding method
The basic genetic algorithm uses a string of binary symbols of fixed length to represent individuals in a population whose alleles are composed of a binary set of symbols {0,1 }. The gene values for individual individuals in the initial population may be generated using uniformly distributed random numbers.
(2) Evaluation of individual fitness
The basic genetic algorithm determines how many chances each individual in the current population will be inherited into the next generation population with a probability proportional to the fitness of the individual. To calculate this probability correctly, it is required here that the fitness of all individuals must be positive or zero. Thus, according to different kinds of problems, the rule for converting the objective function value into the individual fitness needs to be determined in advance, and particularly, the processing method when the objective function value is negative needs to be determined in advance.
(3) Genetic operator
The basic genetic algorithm uses the following 3 genetic operators:
(1) the selection operation uses a proportional selection operator.
(2) The crossover operation uses a single-point crossover operator.
(3) The mutation operation uses a base mutation operator or a uniform mutation operator.
(4) The operating parameters of the basic genetic algorithm.
The basic genetic algorithm has the following 4 operating parameters to be set in advance:
(1) m: the population size, i.e., the number of individuals contained in the population, is generally taken to be 20 to 100.
(2) T: the final evolution algebra of genetic operation is generally 100-5000.
③p c : the crossover probability is generally 0.4 to 0.99.
④p m : the variation probability is generally 0.0001 to 0.1.
The solving results and solving efficiency of the 4 operating parameters have certain influence, but at present, no theoretical basis for reasonably selecting the 4 operating parameters exists. In practical application of genetic algorithm, the reasonable value size or value range of the parameters can be determined after a plurality of trial calculations.
Genetic algorithms provide a general framework for solving complex system optimization problems independent of the field and type of problem. For a practical application problem that needs to be optimized, a genetic algorithm for solving the problem can be constructed by the following steps:
firstly, the method comprises the following steps: decision variables and their various constraints are determined.
Secondly, the method comprises the following steps: and (3) establishing an optimization model, namely determining the type of the objective function and the mathematical description form or the quantization method thereof.
Thirdly, the method comprises the following steps: a chromosome coding method representing a feasible solution is determined.
Fourthly: and determining a decoding method, namely determining the corresponding relation or a conversion method from the individual genotype X to the individual phenotype X.
Fifth, the method comprises the following steps: and determining a quantitative evaluation method of the individual fitness.
Sixth: designing genetic operators, namely determining specific operation methods of genetic operators such as a selection operator, a crossover operator, a mutation operator and the like.
Seventh: determining relevant operating parameters of genetic algorithms, i.e. determining M, T, p of genetic algorithms c 、 p m And the like.
It can be seen from the above construction steps that the design of the encoding method and genetic operator of feasible solution is two main problems to be considered when constructing the genetic algorithm, and is also two key steps when designing the genetic algorithm. Different encoding methods and different manipulated genetic operators may be used for different optimization problems.
In an exemplary embodiment, the mean square error of the prediction result and the simulation result may be used as a fitness function of the genetic algorithm.
In the disclosed embodiment, the fitness function of the Genetic Algorithm (GA) may be the mean square error between the given output prediction and the simulation result of the untrained deep kriging model on the input data.
In step S108, the deep kriging model is trained according to the fitness function to obtain a target deep kriging model.
In an exemplary embodiment, when the fitness function meets a preset condition, the training of the deep kriging model can be completed, and target model parameters of the deep kriging model are obtained; and carrying out particle decoding on the target model parameters according to the sequence to obtain a target depth Krigin model.
The preset condition may be that the function value of the fitness function reaches the maximum value, or the iteration number reaches the preset number.
For example, when the fitness function value is minimal, deep kriging model training is complete.
In the embodiment of the present disclosure, the specific steps of updating the GA algorithm may include:
(1) Initialization: a set of initial solutions is randomly generated as an initial population.
(2) Individual evaluation: the fitness of each individual in the population, i.e., the value of the objective function corresponding to the string of each chromosome, is calculated.
(3) Selecting and operating: the selection operator is applied to the population.
(4) And (3) cross operation: the crossover operator is applied to the population.
(5) Performing mutation operation: and (4) acting mutation operators on the population. The population is selected, crossed and mutated to obtain the next generation new population.
(6) And (4) judging termination conditions: if the evolution algebra (iteration times) is smaller than a set value, turning to (2); and on the contrary, outputting the individual with the maximum fitness obtained in the evolution process as the optimal solution, and terminating the calculation.
In the embodiment of the disclosure, the population 1 may be determined according to the actual problem parameter set, the fitness value of the population 1 is calculated, whether the stop condition is met is judged, and if the stop condition is met, the process may be ended; if the stopping condition is not met, performing genetic operation, wherein the genetic operation can use a basic operator or other high-level operators, obtaining a population 2 after performing the genetic operation, and continuously calculating the fitness value by using the population 2 instead of the population 1.
In the embodiment of the present disclosure, the 200 sets of antenna data may be used for training the model, a mean square error of the deep kriging model predicting 200 sets of data output and a simulation result each time is used as a fitness function of the genetic algorithm GA in the training process, and if the prediction precision of the model does not meet the requirement, the genetic algorithm continues iterative training until the model meets the precision requirement.
The genetic algorithm in the disclosed embodiment has the following advantages: the breadth of the representation of feasible solutions: the processing objects of the genetic algorithm are not parameters per se, but are gene individuals encoded by parameter sets. This encoding operation allows genetic algorithms to operate directly on structural objects. This feature makes genetic algorithms have a wide range of applications. Group search characteristics: the genetic algorithm differs from the single point search of the traditional search method in that it processes multiple individuals in a population simultaneously, i.e., evaluates multiple solutions in the search space simultaneously. The characteristic enables the genetic algorithm to have better global search performance and also enables the genetic algorithm to be easy to parallelize. Genetic algorithms do not easily trap partial optima during the search, and find globally optimal solutions with great probability even in cases where the fitness function defined is discontinuous, irregular, or noisy. The genetic algorithm adopts a natural evolution mechanism to express a complex phenomenon, and can quickly and reliably solve the problem of difficult solution. Genetic algorithms have inherent parallelism and scalability capabilities and are easily mixed with other technologies.
Referring to fig. 4, in the embodiment of the present disclosure, the input data size of each group of antennas may be 1 × 7, and the output size may be 1 × 31, where the output value is a frequency response value of a frequency point of a frequency range.
For example, a deep kriging model constructed by an embodiment of the present disclosure may be constructed with two convolutional layers, two pooling layers, and a kriging proxy model, where the number of convolutional layer channels is 3. And the pooling layer is subjected to average sampling and pooling, the sampling scale is 1 multiplied by 2, and the data enters the kriging proxy model after being output by the second pooling layer and is used as input data of the kriging proxy model.
And after the model reaches the maximum iteration times or meets the fitness function, the optimal size combination of the SIW aperture coupling microstrip patch antenna can be output.
The optimization indexes of the SIW aperture coupling microstrip patch antenna can be as follows: return loss S at 30GHz of working frequency 11 Less than-10 dB, with a bandwidth range of around 10 GHz. The set of line size parameters for the optimization is shown in table 2.
TABLE 2
After the genetic algorithm iteration is completed, the deep kriging model training is completed, and the deep kriging model can be compared with data obtained by HFSS simulation to verify the effectiveness and accuracy of the deep kriging model.
Fig. 5 is a graph illustrating antenna predictions for a deep kriging model in comparison to simulation results output by simulation software, according to an example.
As can be seen from fig. 5, the antenna prediction result output by the deep kriging model provided by the present disclosure is substantially consistent with the simulation result output by the simulation software, and the antenna prediction result output by the deep kriging model provided by the present disclosure satisfies n257 (26.5 GHz-29.5 GHz) and n258 (24.25 GHz-27.5 GHz) frequency bands covering 5G wireless communication millimeter wave FR2, that is, satisfies the existing commercial range of 5G millimeter wave frequency bands.
In an exemplary embodiment, after the model training is completed, the method may further include: acquiring actual design parameters of a plurality of groups of antennas; and simulating actual design parameters of the multiple groups of antennas according to the target depth Krigin model to obtain target design parameters of the antennas.
In the embodiment of the disclosure, the trained deep kriging model can be used to replace electromagnetic simulation software in the related art, and the response of the antenna design parameters in the design process of the SIW aperture coupling microstrip patch antenna is simulated, so that the design of the antenna is completed.
According to the method for realizing the antenna design, the design parameters of the antenna are processed through the deep kriging model, so that the dimensionality of input data can be reduced, the electromagnetic simulation times can be reduced, the design time can be shortened, and the efficiency of the antenna design can be improved; the parameters of the deep kriging model are optimized by using the genetic algorithm, so that the network training time can be greatly shortened, and the local optimization problem of the back propagation optimization algorithm in the related technology is solved.
In addition, the deep kriging model can be obtained by combining a convolutional neural network model and a kriging proxy model, the dimensionality of input data can be reduced on the premise of ensuring that the data value is not lost by using the convolutional neural network, and the fitting output in the complex time-consuming design field can be subjected to antenna optimization design by using the kriging proxy model, so that the calculation time is reduced, and the efficiency is improved.
The following is an embodiment of a terminal device of the present disclosure, which can be used to execute an embodiment of the method of the present disclosure. For details that are not disclosed in the embodiments of the terminal device of the present disclosure, refer to the embodiments of the method of the present disclosure.
Fig. 6 is a block diagram illustrating an apparatus for implementing one antenna design in accordance with an example embodiment.
As shown in fig. 6, an apparatus 600 for implementing antenna design may include: a training data acquisition module 602, a prediction result acquisition module 604, a fitness function determination module 606, and a model training module 608.
The training data obtaining module 602 is configured to obtain training data, where the training data includes design parameters of multiple groups of antennas and simulation results thereof; the prediction result obtaining module 604 is configured to process the design parameters of the antenna through a deep kriging model to obtain a prediction result; a fitness function determining module 606 is configured to determine a fitness function of a genetic algorithm according to the simulation result of the antenna and the prediction result; the model training module 608 is configured to train the deep kriging model according to the fitness function to obtain a target deep kriging model.
In an exemplary embodiment, the deep kriging model includes an input layer of a convolutional neural network model, a convolutional layer of a convolutional neural network model, a pooling layer of a convolutional neural network model, a kriging proxy model, and an output layer of a convolutional neural network model.
In an exemplary embodiment, the apparatus 600 for implementing antenna design further includes: the initial model building module is used for building an antenna initial model according to the structural parameters of the antenna; the design parameter generation module is used for randomly generating design parameters of a plurality of groups of antennas; and the simulation result obtaining module is used for inputting the design parameters of the plurality of groups of randomly generated antennas into electromagnetic simulation software for simulation so as to obtain the simulation results of the plurality of groups of antennas.
In an exemplary embodiment, the prediction result obtaining module 604 includes: the model initialization unit is used for initializing the deep kriging model, taking the network parameters of the convolutional neural network and the hyper-parameters of the kriging proxy model as the particle number of the genetic algorithm, and arranging the particle number in sequence; and the genetic algorithm initialization unit is used for initializing the genetic algorithm, inputting the design parameters of the antenna into the deep kriging model and obtaining the prediction result.
In an exemplary embodiment, fitness function determination module 606 includes: and the fitness function determining unit is used for taking the mean square error of the prediction result and the simulation result as the fitness function of the genetic algorithm.
In an exemplary embodiment, model training module 608 includes: the model training completion unit is used for completing the training of the deep kriging model when the fitness function meets a preset condition to obtain target model parameters of the deep kriging model; and the model obtaining unit is used for carrying out particle decoding on the target model parameters according to the sequence to obtain the target depth Krigin model.
In an exemplary embodiment, the apparatus 600 for implementing antenna design further includes: the design parameter acquisition module is used for acquiring actual design parameters of the multiple groups of antennas; and the simulation unit is used for simulating the actual design parameters of the multiple groups of antennas according to the target depth Kriging model to obtain the target design parameters of the antennas.
In an exemplary embodiment, the antenna is a substrate integrated waveguide aperture coupled microstrip patch antenna.
It is noted that the block diagrams shown in the above figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor terminal devices and/or microcontroller terminal devices.
Fig. 7 is a schematic diagram of a structure of an electronic device according to an exemplary embodiment. It should be noted that the electronic device shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 7, the electronic device 700 includes a Central Processing Unit (CPU) 701, which can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data necessary for the operation of the system 700 are also stored. The CPU 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
The following components are connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.
In particular, according to an embodiment of the present disclosure, the process described above with reference to the flow diagrams may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program containing program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section 709, and/or installed from the removable medium 711. The above-described functions defined in the system of the present disclosure are performed when the computer program is executed by a Central Processing Unit (CPU) 701.
It should be noted that the computer readable media shown in the present disclosure may be computer readable signal media or computer readable storage media or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, terminal device, or apparatus, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, terminal device, or apparatus. In contrast, in the present disclosure, a computer-readable signal medium may include a propagated data signal with computer-readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, terminal device, or apparatus. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present disclosure may be implemented by software or hardware. The described units may also be provided in a processor, which may be described as: a processor includes a transmitting unit, an obtaining unit, a determining unit, and a first processing unit. The names of these units do not constitute a limitation to the unit itself in some cases, and for example, the sending unit may also be described as "a unit that sends a picture acquisition request to a connected server".
As another aspect, the present disclosure also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be separate and not assembled into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise: acquiring training data, wherein the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof; processing the design parameters of the antenna through a deep kriging model to obtain a prediction result; determining a fitness function of a genetic algorithm according to the simulation result and the prediction result of the antenna; and training the deep kriging model according to the fitness function to obtain a target deep kriging model.
Exemplary embodiments of the present disclosure are specifically illustrated and described above. It is to be understood that the disclosure is not limited to the precise arrangements, instrumentalities, or instrumentalities described herein; on the contrary, the disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
Claims (11)
1. An implementation method of an antenna design, comprising:
acquiring training data, wherein the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof;
processing the design parameters of the antenna through a deep kriging model to obtain a prediction result;
determining a fitness function of a genetic algorithm according to the simulation result and the prediction result of the antenna;
and training the deep kriging model according to the fitness function to obtain a target deep kriging model.
2. The method of claim 1, wherein the deep kriging models comprise an input layer of a convolutional neural network model, a convolutional layer of a convolutional neural network model, a pooling layer of a convolutional neural network model, a kriging proxy model, and an output layer of a convolutional neural network model.
3. The method of claim 1 or 2, further comprising:
constructing an antenna initial model according to the structural parameters of the antenna;
randomly generating design parameters of a plurality of groups of antennas;
and inputting the design parameters of the plurality of groups of randomly generated antennas into electromagnetic simulation software for simulation to obtain the simulation results of the plurality of groups of antennas.
4. The method of claim 1 or 2, wherein the processing the design parameters of the antenna through the deep kriging model to obtain the prediction result comprises:
initializing the deep kriging model, taking the network parameters of the convolutional neural network and the hyper-parameters of the kriging proxy model as the particle number of the genetic algorithm, and arranging the particle number in sequence;
initializing the genetic algorithm, and inputting the design parameters of the antenna into the deep kriging model to obtain the prediction result.
5. The method of claim 4, wherein determining a fitness function for a genetic algorithm based on the simulation results and the prediction results for the antenna comprises:
and taking the mean square error of the prediction result and the simulation result as a fitness function of the genetic algorithm.
6. The method of claim 5, wherein training the deep kriging model according to the fitness function to obtain a target deep kriging model comprises:
when the fitness function meets a preset condition, completing the training of the deep kriging model to obtain target model parameters of the deep kriging model;
and carrying out particle decoding on the target model parameters according to the sequence to obtain the target depth Krigin model.
7. The method of claim 1 or 6, further comprising:
acquiring actual design parameters of a plurality of groups of antennas;
and simulating the actual design parameters of the multiple groups of antennas according to the target depth Kriging model to obtain the target design parameters of the antennas.
8. The method according to any of claims 1-7, wherein the antenna is a substrate integrated waveguide aperture coupled microstrip patch antenna.
9. An apparatus for implementing antenna design, comprising:
the training data acquisition module is used for acquiring training data, wherein the training data comprises design parameters of a plurality of groups of antennas and simulation results thereof;
the prediction result obtaining module is used for processing the design parameters of the antenna through a deep kriging model to obtain a prediction result;
a fitness function determining module, configured to determine a fitness function of a genetic algorithm according to the simulation result of the antenna and the prediction result;
and the model training module is used for training the deep kriging model according to the fitness function so as to obtain a target deep kriging model.
10. An electronic device, comprising:
at least one processor;
storage means for storing at least one program which, when executed by the at least one processor, causes the at least one processor to carry out the method of any one of claims 1 to 8.
11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110972826.4A CN115719027A (en) | 2021-08-24 | 2021-08-24 | Method for realizing antenna design and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110972826.4A CN115719027A (en) | 2021-08-24 | 2021-08-24 | Method for realizing antenna design and related equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115719027A true CN115719027A (en) | 2023-02-28 |
Family
ID=85254682
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110972826.4A Pending CN115719027A (en) | 2021-08-24 | 2021-08-24 | Method for realizing antenna design and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115719027A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117574783A (en) * | 2024-01-16 | 2024-02-20 | 天津工业大学 | Antenna optimization method, device, equipment and medium based on depth Gaussian process |
-
2021
- 2021-08-24 CN CN202110972826.4A patent/CN115719027A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117574783A (en) * | 2024-01-16 | 2024-02-20 | 天津工业大学 | Antenna optimization method, device, equipment and medium based on depth Gaussian process |
| CN117574783B (en) * | 2024-01-16 | 2024-03-22 | 天津工业大学 | Antenna optimization method, device, equipment and medium based on depth Gaussian process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Koziel et al. | Efficient multi-objective simulation-driven antenna design using co-kriging | |
| Wu et al. | Multistage collaborative machine learning and its application to antenna modeling and optimization | |
| Naseri et al. | A generative machine learning-based approach for inverse design of multilayer metasurfaces | |
| El Misilmani et al. | A review on the design and optimization of antennas using machine learning algorithms and techniques | |
| Dong et al. | Fast multi-objective optimization of multi-parameter antenna structures based on improved BPNN surrogate model | |
| Massa et al. | On the design of complex EM devices and systems through the system-by-design paradigm: A framework for dealing with the computational complexity | |
| Zhang et al. | Surrogate-assisted quasi-Newton enhanced global optimization of antennas based on a heuristic hypersphere sampling | |
| Koziel et al. | Reduced‐cost electromagnetic‐driven optimisation of antenna structures by means oftrust‐region gradient‐search with sparse Jacobian updates | |
| Michiels et al. | Full-wave simulations of electromagnetic scattering problems with billions of unknowns | |
| CN109271695B (en) | Multi-target antenna design method based on neural network | |
| Koziel et al. | Rapid multi‐objective design optimisation of compact microwave couplers by means of physics‐based surrogates | |
| Ding et al. | High‐efficiency scheme and optimisation technique for design of fragment‐type isolation structure between multiple‐input and multiple‐output antennas | |
| Zhang et al. | Antenna optimization design based on deep Gaussian process model | |
| CN115719027A (en) | Method for realizing antenna design and related equipment | |
| Johanesson et al. | Feasible space boundary search for improved optimisation‐based miniaturisation of antenna structures | |
| de Sena et al. | Harmonic suppression using optimised hexagonal defected ground structure by genetic algorithm | |
| CN115203935A (en) | Frequency selection surface structure topology inverse prediction method and device based on deep learning | |
| CN108446503A (en) | Near field bare cloth aerial array optimization method based on compressed sensing Yu convex optimization | |
| Koziel et al. | Rapid optimisation of omnidirectional antennas using adaptively adjusted design specifications and kriging surrogates | |
| Ahmed et al. | Sub-6 GHz MIMO antenna design for 5G smartphones: A deep learning approach | |
| Chen | Application of multi‐objective fractional factorial design for ultra‐wideband antennas with uniform gain and high fidelity | |
| Koziel et al. | Low-cost quasi-global optimization of expensive electromagnetic simulation models by inverse surrogates and response features | |
| Ding et al. | Antenna Optimization Based on Auto‐Context Broad Learning System | |
| CN116341360A (en) | Antenna topology pixel optimization design method adopting machine learning auxiliary optimization | |
| Hu et al. | Vehicular task scheduling strategy with resource matching computing in cloud‐edge collaboration |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |