Radio frequency component optimization method, electronic device, program product and base station antenna

Abstract

The invention provides an optimization method of a radio frequency component, which comprises the following steps of: performing parallel calculation on target response by utilizing a plurality of lower confidence bound functions to obtain a plurality of candidate parameters; performing electromagnetic simulation on the plurality of candidate parameters to obtain simulation response data corresponding to the candidate parameters; on the basis of the simulation response data corresponding to the candidate parameters, the optimization result of the radio frequency component is determined from the candidate parameters, and the optimization method can improve the optimization efficiency of the radio frequency component, effectively utilize computing resources and reduce waste of the computing resources. The invention also provides electronic equipment, a program product and a base station antenna.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a method for optimizing a radio frequency component, an electronic device, a program product, and a base station antenna. Background Art

[0002] In recent years, wireless communication technology has developed rapidly, and people's requirements for data transmission rate and signal quality have continued to increase, which has led to more stringent performance requirements for antennas and antenna arrays. In order to improve the performance of antennas, topology optimization has been introduced into antenna design to break through the traditional antenna structure and obtain a topology that meets different requirements. The functions used in the optimization of traditional topological structures are sequential, which means that the optimization process can only select candidate parameters one by one, resulting in low optimization efficiency of base station antennas. In addition, the computing power of computers cannot be effectively utilized, resulting in a waste of computing resources. Summary of the invention

[0003] The present disclosure provides a method for optimizing a radio frequency component, an electronic device, a computer-readable medium, and a product.

[0004] In a first aspect, an embodiment of the present disclosure provides a method for optimizing a radio frequency component, including:

[0005] Multiple lower confidence bound functions are used to parallelly calculate the target response to obtain multiple candidate parameters;

[0006] Performing electromagnetic simulation on the plurality of candidate parameters respectively to obtain simulation response data corresponding to the candidate parameters;

[0007] Based on the simulation response data corresponding to the candidate parameters, an optimization result of the radio frequency component is determined from the multiple candidate parameters.

[0008] In a second aspect, an embodiment of the present disclosure provides an electronic device, comprising a memory and a processor; the memory stores a computer program that can be executed by the processor, and when the computer program is executed by the processor, it implements the optimization method of the radio frequency component described in any one of the embodiments of the present disclosure.

[0009] In a third aspect, an embodiment of the present disclosure provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the optimization method of the radio frequency component described in any one of the embodiments of the present disclosure.

[0010] In a fourth aspect, an embodiment of the present disclosure provides a base station antenna, comprising:

[0011] At least one antenna element;

[0012] A feeding network, the feeding network being used to electrically connect the at least one antenna element;

[0013] The topological structure of the at least one antenna element and the feeding network is determined by the optimization method of the radio frequency component provided by the embodiment of the present disclosure.

[0014] The optimization method of the RF component in the embodiment of the present disclosure utilizes multiple lower confidence bound functions to perform parallel calculations on the target response, which can not only obtain multiple candidate parameters at the same time, but also perform electromagnetic simulations in parallel, thereby greatly improving the optimization efficiency of the RF component, and effectively utilizing computing resources and reducing the waste of computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In the accompanying drawings of the embodiments of the present disclosure:

[0016] Figure 1 A flowchart of a method for optimizing a radio frequency component provided in an embodiment of the present disclosure;

[0017] Figure 2 A flowchart of another method for optimizing a radio frequency component provided by an embodiment of the present disclosure;

[0018] Figure 3 is a top view of a MIMO antenna structure;

[0019] Figure 4 A bottom view of a MIMO antenna structure;

[0020] Figure 5 A flowchart of another method for optimizing a radio frequency component provided in an embodiment of the present disclosure;

[0021] Figure 6 Isolation curves for iterative optimization of decoupling for ultra-wideband MIMO antennas;

[0022] Figure 7 This is a comparison chart of return loss and isolation of ultra-wideband MIMO antenna;

[0023] Figure 8 This is a comparison chart of the algorithm performance of ultra-wideband MIMO antenna compared with traditional methods;

[0024] Fig. 9 A schematic diagram of the structure of a dual-polarized slot-coupled millimeter-wave base station antenna array provided in an embodiment of the present disclosure;

[0025] Fig.10 Isolation curves for decoupling optimization iterations for dual-polarized slot-coupled antennas;

[0026] Fig.11 The comparison diagram of return loss and isolation of dual-polarization slot-coupled antenna;

[0027] Fig.12A block diagram of an electronic device provided in an embodiment of the present disclosure;

[0028] Fig.13 A schematic diagram of the structure of a base station antenna provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, the embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0030] The present disclosure will be described more fully below with reference to the accompanying drawings, but the embodiments shown may be embodied in different forms, and the present disclosure should not be construed as being limited to the embodiments set forth below. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete, and will enable those skilled in the art to fully understand the scope of the present disclosure.

[0031] The accompanying drawings of the embodiments of the present disclosure are used to provide a further understanding of the embodiments of the present disclosure, and constitute a part of the specification, and are used together with the detailed embodiments to explain the present disclosure, and do not constitute a limitation of the present disclosure. By describing the detailed embodiments with reference to the accompanying drawings, the above and other features and advantages will become more apparent to those skilled in the art.

[0032] The present disclosure may be described with reference to plan views and / or cross-sectional views by means of ideal schematic views of the present disclosure. Therefore, the exemplary illustrations may be modified according to manufacturing techniques and / or tolerances.

[0033] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0034] The terms used in the present disclosure are only used to describe specific embodiments and are not intended to limit the present disclosure. The term "and / or" as used in the present disclosure includes any and all combinations of one or more related enumerated items. The singular forms "one" and "the" as used in the present disclosure are also intended to include plural forms, unless the context clearly indicates otherwise. The terms "including", "made of..." as used in the present disclosure specify the presence of the features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof.

[0035] Unless otherwise defined, all terms (including technical and scientific terms) used in this disclosure have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly defined in this disclosure.

[0036] The present disclosure is not limited to the embodiments shown in the drawings, but includes modifications of the configurations formed based on the manufacturing process. Therefore, the regions illustrated in the drawings have schematic properties, and the shapes of the regions shown in the drawings illustrate the specific shapes of the regions of the elements, but are not intended to be limiting.

[0037] The functions used in the traditional optimization methods for RF components are serial functions, which have low optimization efficiency. Even with the current more advanced multi-core computers, the improvement in optimization efficiency is relatively limited, which in turn results in a waste of computing resources.

[0038] In some related technologies, although a batch Bayesian optimization algorithm is used, the algorithm still selects candidate parameters in a batch one by one and still cannot improve the optimization efficiency.

[0039] Based on the above reasons, an embodiment of the present disclosure provides a method for optimizing a radio frequency component, which is used to improve the optimization efficiency of the radio frequency component and effectively utilize computing resources.

[0040] In the embodiment of the present disclosure, the radio frequency component includes one or more of a base station antenna and a filter.

[0041] In a first aspect, an embodiment of the present disclosure provides a method for optimizing a radio frequency component.

[0042] Figure 1 The present invention provides a flowchart of a method for optimizing a radio frequency component according to an embodiment of the present invention. Figure 1 As shown, the embodiment of the present disclosure provides a method for optimizing a radio frequency component, including:

[0043] Step S101, using multiple lower confidence bound functions to perform parallel calculations on the target response to obtain multiple candidate parameters.

[0044] The target response is the response data obtained by predicting the sampling points of the initial topological structure. The candidate parameters include candidate topological parameters and candidate structural parameters.

[0045] The initial topological structure can be generated by using a normalized Gaussian network according to pre-set topological parameters and structural parameters. After the initial topological structure is obtained, the initial topological structure is sampled to obtain sampling points.

[0046] In the disclosed embodiment, the lower confidence bound function (genetic algorithm) is applied to the reinforcement learning function to balance development, exploration and exploitation, and tends to explore actions that currently perform poorly but have greater potential for improvement. The genetic algorithm is used to search the RF components in the design space to find the optimal optimization parameters.

[0047] Step S102: performing electromagnetic simulation on a plurality of candidate parameters respectively to obtain simulation response data corresponding to the candidate parameters.

[0048] The specific method of electromagnetic simulation is not limited in the disclosed embodiments. The candidate parameters are simulated by electromagnetic simulation to obtain simulation response data, which can be used to characterize the performance of the radio frequency component.

[0049] In the embodiment of the present disclosure, since multiple lower confidence bound functions can be used for parallel calculation, multiple candidate parameters can be obtained in parallel compared to the existing sequential functions. Therefore, during electromagnetic simulation, multiple candidate parameters can also be simulated in parallel, thereby improving the efficiency of electromagnetic simulation.

[0050] It should be noted that electromagnetic simulation is the simulation of the electrical and magnetic properties of RF components. The simulation response data is the manifestation of the electrical and magnetic properties. The quality of the topological structure of the RF component can be judged based on the simulation response data.

[0051] Step S103: determining an optimization result of the radio frequency component from a plurality of candidate parameters based on the simulation response data corresponding to the candidate parameters.

[0052] In the embodiment of the present disclosure, the simulation response data is used to verify the performance of the RF component. According to the simulation response data corresponding to the multiple candidate parameters, the optimization performance of the RF component can be determined, and the optimal optimization result can be determined according to the optimization performance.

[0053] The optimization method of the RF component in the embodiment of the present disclosure utilizes multiple lower confidence bound functions to perform parallel calculations on the target response, which can not only obtain multiple candidate parameters at the same time, but also perform electromagnetic simulations in parallel, thereby greatly improving the efficiency of RF component optimization, and effectively utilizing computing resources and reducing the waste of computing resources.

[0054] In some embodiments, the lower confidence bound function includes a constraint term, and the constraint term is used to constrain the optimization target of the radio frequency component.

[0055] In some embodiments, the constraints are determined by: obtaining the number of constraints, and the mean and variance of each constraint; determining the cumulative probability density function value of the standard normal distribution corresponding to the constraint based on the mean and variance of the constraint; and performing a product operation on the cumulative probability density function values ​​of the standard normal distribution of all constraints to obtain the constraints.

[0056] Adding the constraint term to the lower confidence bound function can avoid local optimization of the RF component and improve the convergence speed, thereby speeding up the optimization of the RF component.

[0057] For example, the constraint It can be expressed using formula (1).

[0058] (1)

[0059] in, represents the constraint item, is the number of constraints, is the cumulative probability density function of the standard normal distribution, and They represent the output of the Gaussian Processes Regression (GPR) proxy model. The mean and variance of the constraints.

[0060] In some embodiments, the lower confidence bound function includes an exploration term parameter, which is used to expand the candidate parameters.

[0061] In the embodiment of the present disclosure, multiple lower confidence bound functions to be constrained are used as fitness functions, and exploration and development item parameters are set in the lower confidence bound functions, so that both development and exploration can be taken into account while achieving parallelism.

[0062] For example, multiple lower confidence bound functions in the embodiments of the present disclosure can be expressed by formula (2).

[0063] (2)

[0064] in, represents the fitness function value, and is the mean and variance of the sampling points output by the Gaussian process regression proxy model, To explore and develop parameters for the control algorithm, To handle constraints.

[0065] In some embodiments, the development item parameters will be explored Set to one vector, then for each exploration and development item parameter Each of these will generate a candidate parameter with different exploration and development properties, and then The candidate parameters are simulated in parallel to obtain the corresponding response data.

[0066] In some embodiments, the target response is to use a Gaussian process regression proxy model to predict the sampling points based on the prior sampling points and their response data to obtain the response data. The Gaussian process regression proxy model can obtain the performance of the RF component based on the RF component parameters and structural parameters. For example, the Gaussian process regression proxy model inputs the topological parameters and structural parameters, and outputs the mean and standard deviation of the performance of the RF component.

[0067] In the embodiment of the present disclosure, the Gaussian process regression proxy model is trained by machine learning, and the present disclosure does not limit the training method of machine learning.

[0068] In some embodiments, the sample set for training the Gaussian process regression proxy model includes initial samples and intermediate samples; wherein the initial samples are samples obtained from the initial topological structure using Latin hypercube sampling; and the intermediate samples are samples obtained based on candidate parameters and their simulated response data.

[0069] In some embodiments, the initial topological structure is obtained by: dividing the area to be optimized to obtain multiple optimization units, where each optimization unit corresponds to a normalized Gaussian function; using topological parameters to perform weighted summation on the normalized Gaussian function to obtain a normalized Gaussian function value; using the normalized Gaussian function value to characterize the topological structure of the RF component; based on the topological structure of the RF component and the sampled structural parameters, the initial topological structure is obtained.

[0070] For example, the initial topology can be obtained by following the steps below:

[0071] Step A1: Divide the optimization area into cells and place a normalized Gaussian function on each cell.

[0072] The normalized Gaussian function can be expressed using formula (3).

[0073] (3)

[0074] in, and is a two-dimensional Gaussian function expression, which can be expressed by formula (4); is the number of optimized cells for partitioning.

[0075] (4)

[0076] in, represents the location of the sampling point, Indicates the location where each Gaussian function is placed, that is, the mean of the Gaussian function, is the variance.

[0077] Step A2: Use the topological parameters sampled by the Latin hypercube to perform weighted summation on the normalized Gaussian function.

[0078] The weighted summation can be expressed by formula (5).

[0079] (5)

[0080] in, is the weight, represents the topological parameters of the sampling points, Represents the normalized Gaussian function value obtained by normalizing the Gaussian function.

[0081] Step A3: The part of the Gaussian network that is greater than 0 is set as the first topological element, and the part that is less than 0 is set as the second topological element, so as to characterize the topological structure of the radio frequency component.

[0082] The present disclosure does not limit the first topological element and the second topological element. For example, the first topological element is metal and the second topological element is air.

[0083] The first topological element is used to replace the part of the Gaussian network that is greater than 0, and the second topological element is used to replace the part of the Gaussian network that is less than 0, so as to establish the topological structure of the radio frequency component.

[0084] Step A4: The topological structure and the sampled structural parameters are input into the updated antenna model to generate a radio frequency component.

[0085] The disclosed embodiments do not limit the antenna model. Any antenna model that can generate a radio frequency component based on the topological structure and structural parameters can be used in the embodiments.

[0086] After the RF component is generated, electromagnetic simulation may be performed to obtain response data of the RF component and record the response data.

[0087] In some embodiments, the optimized parameters of the RF component include topological parameters of the RF component and / or structural parameters of the RF component, that is, the topological parameters and / or structural parameters of the RF component are optimized.

[0088] In some embodiments, the optimized parameters also include a normalized Gaussian function variance. In the process of optimizing the radio frequency component, the normalized Gaussian function variance can be optimized.

[0089] The normalized Gaussian function variance includes the variance in the x-direction and the y-direction and , the variance in the x-direction and the y-direction can be obtained by Latin hypercube sampling, and the variance can be used as an optimized parameter to be optimized in subsequent steps.

[0090] In some embodiments, after determining the optimization result of the RF component from multiple candidate parameters based on the simulation response data corresponding to the candidate parameters, the following includes: when the optimization result cannot meet the optimization goal, using the candidate parameters and their simulation response data as training samples to continue training the Gaussian process regression proxy model.

[0091] In the disclosed embodiment, whether the optimized RF component meets the optimization target can be determined by simulation response data, or by the number of times the Gaussian process regression proxy model is trained.

[0092] When the optimization goal is not achieved, the candidate parameters that have been obtained can be used as intermediate samples and added to the sample set, so that the intermediate samples can be used to train the Gaussian process regression proxy model in subsequent training.

[0093] In some embodiments, after the input vector x is input into the Gaussian process regression proxy model, the response data f(x) output by the Gaussian process regression proxy model obeys a normal distribution, that is, ,in, is the mean function, is the covariance function; using the properties of Gaussian process regression, through the prior sampling points and their response data Predict the next sampling point Response data .

[0094] In some embodiments, the target response is calculated in parallel using multiple lower confidence bound functions, and before obtaining multiple candidate parameters, it also includes: obtaining the optimization target, optimization constraints, the position and size of the topological structure to be optimized, the topological parameters to be optimized and their ranges, and the structural parameters and their ranges.

[0095] In some embodiments, the optimization target and the optimization constraint both include one or more of isolation, return loss, and antenna gain, wherein the isolation, return loss, and antenna gain can be arbitrarily set by the user.

[0096] Figure 2 A flowchart of another method for optimizing a radio frequency component provided by an embodiment of the present disclosure. Figure 2 , an optimization method for a radio frequency component provided by an embodiment of the present disclosure includes:

[0097] Step S201: Generate an initial topology structure using a normalized Gaussian network according to preset topology parameters and structural parameters of a base station antenna.

[0098] Step S202: training the base station antenna GPR proxy model.

[0099] First, Latin hypercube sampling is used to sample from the initial topological structure to obtain initial samples, electromagnetic simulation is performed on the initial samples within a preset working frequency band to obtain simulation response data, and then the topological parameters, structural parameters and simulation response data are organized into a sample set. It should be noted that Latin hypercube sampling can obtain multiple initial samples.

[0100] Then, the GPR proxy model is trained using the samples in the sample set to obtain a trained base station antenna GPR proxy model.

[0101] It should be noted that when training the base station antenna GPR proxy model, the normalized Gaussian function variance is also taken as an optimization object.

[0102] Step S203, using multiple lower confidence bound functions with constraints to perform parallel calculations on the target response output by the base station antenna GPR proxy model to obtain multiple candidate parameters, wherein the target response includes isolation and echo loss at a certain frequency point.

[0103] Step S204: performing electromagnetic simulation on the multiple candidate parameters to obtain simulation response data.

[0104] Step S205, judging whether the optimization target is achieved according to the simulation response data, if so, ending the training and obtaining the trained GPR proxy model; if not, executing step S206.

[0105] Step S206, update the sample set using the candidate parameters and their response data, and then return to step S202 to continue training the base station antenna GPR proxy model.

[0106] The topological parameters and standing waves of the base station antenna generated by the base station antenna GPR proxy model can achieve the optimal effect.

[0107] In the disclosed embodiment, the base station antenna GPR proxy model may be trained through multiple rounds of iterations, and the specific steps are as follows:

[0108] Using multiple lower confidence bound functions with constraints, the target response of the RF component output by the proxy model of the i-th round is calculated to obtain multiple candidate solutions; electromagnetic simulation is performed on the multiple candidate solutions respectively to obtain corresponding RF component response data; based on the RF component response data, it is determined whether the RF component meets the optimization goal; if not, the RF component response data is added to the training set, the proxy model is continued to be trained, the i+1-th round proxy model is obtained, and the i+1-th round proxy model is used to continue to optimize the RF component; if satisfied, the candidate parameters corresponding to the RF component response data are used as the optimization result of the RF component.

[0109] In order to better understand the optimization method of the radio frequency component provided by the present disclosure, the ultra-wideband MIMO antenna is used as an example for explanation below.

[0110] Figure 3 A top view of a MIMO antenna structure. Figure 4 A bottom view of a MIMO antenna structure. Figure 3 and Figure 4 As shown, the front of the MIMO antenna is a copper symmetrical antenna, and the top of the MIMO antenna is a radius The semicircular structure is 4.8 mm, and the lower part of the semicircular structure includes a first rectangle and a second rectangle, wherein the length of the first rectangle is He Kuan 4.1 millimeters (mm) and 14.0 mm respectively; the difference between the width of the second rectangle and the first rectangle is 3.5 mm, and the length of the second rectangle The gap between two adjacent MIMO antennas is 3.1 mm. The bottom layer of the MIMO antenna is a dielectric substrate with a size of 34×21 mm, thickness 1.6mm, made of FR4, relative dielectric constant The back of the antenna is the ground plane, and the size is 7.4×34 mm.

[0111] The specific structural parameters of the MIMO antenna are detailed in Table 1, and the optimization objectives and optimization constraints are detailed in Table 2.

[0112] Table 1 is the structural parameter table of MIMO antenna

[0113]

[0114] Table 2 shows the optimization objectives and optimization constraints of MIMO antennas.

[0115]

[0116] An optimization method for a radio frequency component provided by an embodiment of the present disclosure is a parallel joint optimization method with constraints, such as Figure 5 As shown, the optimization methods for RF components include:

[0117] Step S501, obtaining the number of normalized Gaussian network units according to preset topological parameters and structural parameters.

[0118] like Figure 3 As shown in the figure, taking the ultra-wideband MIMO antenna as an example, the topology optimization range is the dotted box between the two base station antennas. Assuming that the size of the dotted box is 4.5mm×8mm, the topology optimization range is divided into 6×10 optimization units. According to the symmetry of the base station antenna, the number of optimization units can be compressed to 3×5, that is, 15 dimensions; the structural parameter to be optimized is L, the optimization range is [3, 6], and the unit is mm; at the same time, the variance of the normalized Gaussian network function in the x and y directions is

[0119] The optimization parameters are added, and the variance range is set to 1 / 5 to 1 / 2 of the normalized Gaussian network function interval according to the number of divided optimization units.

[0120] The variance in the x and y directions are expressed by formula (6) and formula (7) respectively.

[0121] (6)

[0122] (7)

[0123] Therefore, the total number of optimization parameters is 18 dimensions.

[0124] Step S502, obtaining a topology parameter sample set and a structure parameter sample set through Latin hypercube sampling, synthesizing the two sample sets into a mixed sample set one-to-one according to the sorting order, and then adjusting the topology and structure parameters of the antenna model using the samples in the mixed sample set to obtain an initial topology structure.

[0125] Step S503, using the Latin hypercube sampling method to obtain 50 18-dimensional optimization parameters.

[0126] Among them, the first 15 dimensions are the weights of the normalized Gaussian network, the 16th and 17th dimensions are the variances of the normalized Gaussian network, and the 18th dimension is the structural parameter. According to the parameters generated by Latin hypercube sampling, the normalized Gaussian network is used to generate the initial samples, and the electromagnetic simulation of the initial samples is performed in the working frequency range of 3.6-9.8GHz, the simulation response data is recorded, and the simulation response data is organized into a data set for training the GPR proxy model.

[0127] Step S504, using the samples in the data set to train the GPR proxy model, using the 18-dimensional antenna topology parameters and structural parameters as the GPR proxy model input, and outputting the target response echo loss |S11| and isolation |S21|.

[0128] The target responses of 50 initial samples are obtained by electromagnetic simulation, and the target responses are processed according to the optimization objectives to obtain the training set.

[0129] Step S505, using a genetic algorithm to search within the topology optimization range, and using multiple lower confidence bound functions with constraints to perform parallel evaluations on the echo loss |S11| and isolation |S21| output by the GPR proxy model to obtain multiple candidate parameters.

[0130] In the embodiment of the present disclosure, the multiple lower confidence bound functions are fitness functions, and the multiple lower confidence bound functions are expressed in the form of formula (2).

[0131] in, are the mean and variance of |S21| output by the GPR proxy model, The exploration and development parameter can be a 1×5 vector, [0, 1, 2, 3, 4], and five different candidate parameters will be generated in each round of iteration. The exploration and development parameter is used to balance the exploration and development of the genetic algorithm.

[0132] In formula (2), there is also a constraint term , the constraint term can also be called the function term of the constraint. In the embodiment of the present disclosure, the constraint term can be expressed as formula (8).

[0133] (8)

[0134] in, is the cumulative probability density function of the standard normal distribution, and are the mean and variance of |S11| output by the GPR proxy model, respectively.

[0135] Step S506: Perform electromagnetic simulation on the multiple candidate parameters to obtain simulation response data.

[0136] Step S507, determining whether the maximum number of iterations has been reached within the optimization frequency band, if so, exiting the loop to obtain the best optimization result; if not, executing step S508.

[0137] Step S508, adding the response data corresponding to the candidate parameters obtained in this round to the sample set, and returning to step S504.

[0138] For example, the maximum number of optimization iterations is set to 50. It is determined whether the number of iterations reaches 50. If so, the loop is exited to obtain the best optimization result; if not, the response data corresponding to the candidate parameters obtained in this round are added to the sample set, and the process returns to step S504.

[0139] Figure 6 This is the isolation curve of the ultra-wideband MIMO antenna decoupling optimization iteration, where the horizontal axis represents the number of electromagnetic simulations and the vertical axis represents the isolation (in dB). A total of 250 electromagnetic simulations were performed in 50 iterations. From the change curve of the isolation |S21| of the MIMO antenna, it can be seen that during the optimization process, the optimization target value |S21| has been decreasing. At the 183rd electromagnetic simulation, when the return loss meets the optimization constraint ( ) under the condition that the isolation drops to the minimum value.

[0140] Figure 7This is a comparison chart of the return loss and isolation of the ultra-wideband MIMO antenna, where the gray dotted line represents the return loss before optimization, the black dotted line represents the isolation before optimization, the gray solid line represents the return loss after optimization, the black solid line represents the isolation after optimization (in dB), the horizontal axis represents the frequency (in GHz), and the vertical axis represents the S parameter (return loss and isolation). This method converges quickly, and because of the use of parallel optimization, the time required is only one-fifth of that of the traditional algorithm under the premise of the same number of simulations. At the same time, the algorithm has a stronger optimization capability, and the isolation of the optimized antenna is improved by more than 1dB. Figure 7 As shown in the figure, the S parameter curve of the 183rd generation base station antenna is compared with the return loss and isolation before optimization. The optimized antenna topology meets the requirements of return loss |S11| lower than -10dB and isolation |S21| lower than -18dB in the working frequency band of 3.6-9.8GHz. The isolation is improved by 8dB compared with that before optimization.

[0141] Figure 8 The figure is a comparison of the algorithm performance of ultra-wideband MIMO antennas compared with traditional methods. The gray solid line represents the optimization performance curve of multiple lower confidence bound functions with constraints, the black solid line represents the optimization performance curve of a single lower confidence bound function, the horizontal axis represents the number of iterations, and the vertical axis represents the isolation degree. Figure 8 It can be seen that under the same optimization goal, the optimization method of the masked autoencoder (MAE) proxy model requires more than 3,000 electromagnetic simulations, while the method provided in the present invention only requires 250 electromagnetic simulations to obtain a base station antenna structure with equivalent performance, and the optimization speed is increased by more than 90%.

[0142] Fig. 9 The structure diagram of a dual-polarized slot-coupled millimeter-wave base station antenna array provided by an embodiment of the present disclosure is shown in FIG. Fig. 9 As shown, the decoupling structure of the base station antenna is designed in the ground plane. The base station antenna is used to radiate ±45° linear polarization mode in the working bandwidth of 24-27.5 GHz.

[0143] The optimized frequency band, optimization target and optimization constraint of the dual-polarized slot-coupled millimeter-wave base station antenna array are shown in Table 3.

[0144] Table 3 shows the optimization objectives and optimization constraints of dual-polarization slot-coupled millimeter-wave base station antenna array

[0145]

[0146] The optimization method of the radio frequency component provided in the embodiment of the present disclosure is a parallel joint optimization method with constraints, which includes:

[0147] Step S1001, obtaining the number of normalized Gaussian network units according to preset topological parameters and structural parameters.

[0148] The scope of topology optimization is Fig. 9 Assuming that the size of the dotted box is 2 mm × 5 mm, the topology optimization range is divided into 4 × 10 optimization units. According to the symmetry of the base station antenna, the number of optimization units can be compressed to 2 × 5, that is, 10 dimensions.

[0149] Step S1002, obtain 50 12-dimensional initial optimization parameters through Latin hypercube sampling, generate initial samples using a normalized Gaussian network, perform electromagnetic simulation in the working frequency band of 24-27.5 GHz and record the simulation response data, and organize the simulation response data into a data set for training the GPR proxy model.

[0150] Step S1003: train the GPR proxy model using the data set.

[0151] Step S1004, using a genetic algorithm to search within the topology optimization range, and using multiple lower confidence bound functions with constraints to perform parallel evaluations on the echo loss |S11| and isolation |S21| output by the GPR proxy model to obtain multiple candidate parameters.

[0152] The specific method of step S1004 is the same as that of step S505 and will not be repeated here.

[0153] Step S1005, determine whether the maximum number of iterations has been reached within the optimization frequency band. If so, exit the loop to obtain the best optimization result; if not, add the response data corresponding to the candidate parameters obtained in this round to the sample set and return to step S1003.

[0154] Assume that the maximum number of iterations is set to 100. Whether the number of iterations reaches 100 in the optimization frequency band, if yes, then exit the loop to obtain the best optimization result; if no, then add the response data corresponding to the candidate parameters obtained in this round to the sample set, and return to step S1003.

[0155] Fig.10 This is the isolation curve of the decoupling optimization iteration of the dual-polarized slot-coupled antenna, where the horizontal axis represents the number of electromagnetic simulations and the vertical axis represents the isolation (in dB). A total of 500 electromagnetic simulations were performed in 100 iterations. From the change curve of the isolation |S21| of the dual-polarized slot-coupled antenna, it can be seen that the optimization target value |S21| has been decreasing during the optimization process. At the 473rd electromagnetic simulation, the isolation dropped to the lowest value under the condition that the return loss met the optimization constraint (|S11|<-12dB).

[0156] Fig.11This is a comparison chart of the return loss and isolation of the dual-polarized slot-coupled antenna, where the gray dotted line represents the return loss before optimization, the black dotted line represents the isolation before optimization, the gray solid line represents the return loss after optimization, the black solid line represents the isolation after optimization (in dB), the horizontal axis represents the frequency (in GHz), and the vertical axis represents the S parameter (return loss and isolation). Based on the S parameter curve of the 473rd generation base station antenna, the antenna topology finally optimized meets the requirement of return loss |S11| lower than -12dB in the working frequency band of 24-27.5GHz, and the isolation |S21| reaches -22.8dB, which is more than 7.5dB higher than before optimization.

[0157] The following uses a filter as an example to illustrate the filter optimization method provided by the embodiment of the present disclosure.

[0158] In an embodiment of the present disclosure, a filter optimization method includes:

[0159] Step A01, generating an initial topological structure of the filter using a normalized Gaussian network according to preset topological parameters and structural parameters of the filter.

[0160] Step A02, training the filter GPR proxy model.

[0161] The Latin hypercube sampling method is used to sample from the initial topological structure to obtain the initial sample, and the electromagnetic simulation is performed on the initial sample within the preset working frequency band to obtain the simulation response data, and then the topological parameters, structural parameters and simulation response data are sorted into a sample set. It should be noted that Latin hypercube sampling can obtain multiple initial samples. Then, the filter GPR proxy model is trained using the samples in the sample set to obtain a trained filter GPR proxy model.

[0162] Step A03, using multiple lower confidence bound functions with constraints to parallel calculate the target response output by the GPR proxy model to obtain multiple candidate parameters. The target response includes isolation and echo loss at a certain frequency point.

[0163] Step A04: Perform electromagnetic simulation on multiple candidate parameters to obtain simulation response data.

[0164] Step A05, judging whether the topological parameters, structural parameters and normalized Gaussian function variance reach the optimization target according to the simulation response data, if so, ending the training and obtaining the trained filter GPR proxy model; if not, executing step A06.

[0165] Step A06, update the sample set using the candidate parameters and their response data, and then return to step A02 to continue training the filter GPR proxy model.

[0166] After steps A01 to A06, a trained filter GPR proxy model is obtained. During use, the filter's function parameters only need to be input into the filter GPR proxy model to obtain a filter with an optimal topological structure.

[0167] In a second aspect, an embodiment of the present disclosure provides an electronic device.

[0168] Fig.12 The following is a block diagram of an electronic device provided by an embodiment of the present disclosure. Fig.12 As shown, the electronic device provided by the embodiment of the present disclosure includes a memory 1201 and a processor 1202; the memory 1201 stores a computer program that can be executed by the processor 1202, and when the computer program is executed by the processor, it implements any one of the optimization methods of the radio frequency components of the embodiment of the present disclosure.

[0169] In some embodiments, the electronic device also includes an I / O interface (read-write interface) 1203, which is connected between the processor 1201 and the memory 1202, and can realize information interaction between the memory 1202 and the processor 1201, including but not limited to a data bus (Bus), etc.

[0170] Among them, the processor is a device with data processing capabilities, including but not limited to the central processing unit (CPU); the memory is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); the I / O interface (read-write interface) is connected between the processor and the memory, and can realize information exchange between the memory and the processor, including but not limited to the data bus (Bus), etc.

[0171] In a third aspect, an embodiment of the present disclosure provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements any method for optimizing a radio frequency component of the embodiment of the present disclosure.

[0172] In a fourth aspect, an embodiment of the present disclosure provides a base station antenna.

[0173] Fig.13 The following is a schematic diagram of the structure of a base station antenna provided by an embodiment of the present disclosure. Fig.13 As shown, a base station antenna provided by an embodiment of the present disclosure includes:

[0174] At least one antenna element 1301;

[0175] A feeding network 1302 , wherein the feeding network 1302 is used to electrically connect the at least one antenna element 1301 .

[0176] The topological structure of the at least one antenna element 1301 and the feeding network 1302 is determined by the optimization method of the radio frequency component provided by the embodiment of the present disclosure.

[0177] Among them, the topological structure includes topological parameters and structural parameters. In the process of optimizing the topological structure of the base station antenna, the variance of the normalized Gaussian function can also be optimized. In order to save space, the specific steps of the optimization method of the RF component will not be described in detail here.

[0178] A base station antenna provided by an embodiment of the present disclosure has topological parameters and structural parameters of an antenna element and a feeding network that are calculated in parallel on a target response using multiple lower confidence bound functions. This not only enables multiple candidate parameters to be obtained simultaneously, but also allows electromagnetic simulations to be performed in parallel, thereby greatly improving the optimization efficiency of the base station antenna, and effectively utilizing computing resources and reducing the waste of computing resources.

[0179] Those skilled in the art will appreciate that all or some of the steps, systems, and functional modules / units in the apparatus disclosed above may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0180] In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed by several physical components in cooperation.

[0181] Some or all physical components may be implemented as software executed by a processor, such as a central processing unit (CPU), a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is well known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH) or other disk storage; read-only compact disk (CD-ROM), digital versatile disk (DVD) or other optical disk storage; magnetic cassettes, tapes, disk storage or other magnetic storage; any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0182] The present disclosure has disclosed example embodiments, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for limiting purposes. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly stated, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, those skilled in the art will appreciate that various changes in form and detail may be made without departing from the scope of the present disclosure as set forth in the appended claims.

Claims

1. A method for optimizing a radio frequency component, characterized in that: include: Multiple lower confidence bound functions are used to parallelly calculate the target response to obtain multiple candidate parameters; Performing electromagnetic simulation on the plurality of candidate parameters respectively to obtain simulation response data corresponding to the candidate parameters; Based on the simulation response data corresponding to the candidate parameters, an optimization result of the radio frequency component is determined from the multiple candidate parameters.

2. The method according to claim 1, characterized in that The lower confidence bound function includes a constraint term, and the constraint term is used to constrain the optimization target of the radio frequency component.

3. The method according to claim 2, characterized in that The constraints are determined in the following way: Obtaining the number of constraint items, and the mean and variance of each constraint item; Determine the cumulative probability density function value of the standard normal distribution corresponding to the constraint item based on the mean and variance of the constraint item; The constraint items are obtained by performing a product operation on the cumulative probability density function values ​​of the standard normal distribution of all the constraint items.

4. The method according to claim 1, characterized in that The lower confidence bound function includes exploration and development item parameters, and the exploration and development item parameters are used to expand the candidate parameters.

5. The method according to claim 1, characterized in that The target response is the response data predicted for the sampling point based on the prior sampling point and its response data by using the Gaussian process regression proxy model.

6. The method according to claim 5, characterized in that The sample set for training the Gaussian process regression proxy model includes initial samples and intermediate samples; wherein the initial samples are samples obtained from the initial topological structure using Latin hypercube sampling; and the intermediate samples are samples obtained based on the candidate parameters and their simulation response data.

7. The method according to claim 6, characterized in that The initial topology is obtained by: Segmenting the area to be optimized to obtain a plurality of optimization units, wherein each of the optimization units corresponds to a normalized Gaussian function; Performing weighted summation on the normalized Gaussian function using topological parameters to obtain a normalized Gaussian function value; Characterizing the topological structure of the radio frequency component by using the normalized Gaussian function value; The initial topology structure is obtained based on the topology structure of the radio frequency component and the sampled structural parameters.

8. The method according to claim 5, characterized in that After determining the optimization result of the radio frequency component from the plurality of candidate parameters based on the simulation response data corresponding to the candidate parameters, the method further comprises: When the optimization result cannot meet the optimization target, the candidate parameters and the simulation response data thereof are used as training samples to continue training the Gaussian process regression proxy model.

9. The method according to claim 1, characterized in that: The optimized parameters of the RF component include one or more of the topological parameters, structural parameters, and normalized Gaussian function variance of the RF component.

10. The method according to claim 5, characterized in that Before the method uses multiple lower confidence bound functions to perform parallel calculations on the target response to obtain multiple candidate parameters, the method further includes: Obtain the optimization goal, optimization constraints, the location and size of the topological structure to be optimized, the topological parameters to be optimized and their ranges, and the structural parameters and their ranges.

11. The method according to claim 10, characterized in that The optimization target and the optimization constraint both include one or more of isolation, return loss and antenna gain.

12. An electronic device, characterized in that: It comprises a memory and a processor; the memory stores a computer program that can be executed by the processor, and when the computer program is executed by the processor, it implements the optimization method of the radio frequency component described in any one of claims 1 to 11.

13. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the optimization method of the radio frequency component according to any one of claims 1 to 11.

14. A base station antenna, characterized in that: include: At least one antenna element; A feeding network, the feeding network being used to electrically connect the at least one antenna element; The topological structure of the at least one antenna element and the feeding network is determined by the optimization method of the radio frequency component according to any one of claims 1 to 11.

Images