Phase-free near-field reconstruction far-field prediction method, apparatus, equipment and storage medium

Abstract

This application provides a method, apparatus, device, and storage medium for phase-free near-field reconstruction and far-field prediction. The method includes: acquiring near-field magnetic field amplitude data and geometric boundaries; performing multiple independent reconstructions to obtain multiple sets of equivalent dipole models; calculating the far-field maximum value of each model at the target distance to form a dataset containing multiple samples; and finally calculating the statistical mean and standard deviation of the dataset. The mean is used as the far-field prediction result, and the standard deviation plus or minus a preset multiple of the mean is output as the confidence interval. This method, for the first time, transforms the inherent non-uniqueness of solutions in phase-free near-field reconstruction into quantifiable statistical reliability. The output statistical mean eliminates the random bias of single reconstructions, exhibiting higher stability and repeatability. Simultaneously, the confidence interval provides an objective quantitative basis for determining engineering design margins and risk decisions.

Description

Technical Field

[0001] This application relates to the field of electromagnetic compatibility prediction, and in particular to a phase-free near-field reconstruction far-field prediction method, apparatus, device, and storage medium. Background Technology

[0002] With the rapid development of electronic information technology, the integration level and switching frequency of printed circuit boards (PCBs) and integrated circuits (ICs) continue to rise. The resulting electromagnetic interference (EMI) and electromagnetic compatibility (EMC) issues have become core challenges restricting the reliability of complex electronic systems. In fields such as automotive electronics, consumer electronics, communication equipment, and industrial control systems, the analysis and prediction of the radiation characteristics of devices under test (DUTs) are crucial for ensuring product compliance. For example, in automotive electronics R&D, electromagnetic radiation from vehicle controllers, radar modules, or communication modules may interfere with the normal operation of other electronic components; in the consumer electronics field, EMI issues in smartphones and wearable devices may affect signal transmission quality or pose health risks to users. Traditional EMC testing requires a 3-meter or 10-meter far-field environment, relying on large semi-anechoic chambers (SACs) to simulate a free-space environment. However, such testing places extremely high demands on site, equipment, and maintenance costs, especially for small and medium-sized enterprises or in the early stages of R&D, where the economic burden of frequent testing is unbearable.

[0003] Existing technologies output a set of equivalent dipole models through a single equivalent source reconstruction and use these models to calculate far-field predictions.

[0004] However, when the reconstruction algorithm converges to different local optima due to a lack of phase information, the far-field prediction results fluctuate significantly, and the stability and repeatability of the prediction results cannot be guaranteed. Summary of the Invention

[0005] This application provides a phase-free near-field reconstruction far-field prediction method, apparatus, device, and storage medium to solve the technical problem that the lack of statistical reliability of far-field prediction results due to the non-uniqueness of solutions in existing phase-free near-field reconstruction.

[0006] In a first aspect, embodiments of this application provide a phase-free near-field reconstruction far-field prediction method, including:

[0007] The near-field magnetic field strength amplitude data of the device under test is obtained as reference field data and the geometric boundary of the spatial cluster. The geometric boundary is generated by a preset method and has the same number of preset equivalent dipoles.

[0008] Multiple independent reconstructions are performed based on the reference field data and the geometric boundary to obtain multiple sets of equivalent dipole models, wherein each reconstruction uses different random initial conditions.

[0009] Based on the analytical formula of electromagnetic field, the maximum value of the far-field electric field intensity of each equivalent dipole model at the target distance is calculated, forming a dataset containing multiple far-field maximum value samples;

[0010] Perform statistical analysis on the dataset to calculate the statistical mean and standard deviation of the far-field maximum value samples in the dataset;

[0011] When the far-field maximum value sample follows a normal distribution, the statistical average is used as the final prediction result of the far-field radiation maximum value of the device under test, and the confidence interval of the prediction value is determined by combining the standard deviation.

[0012] In one possible implementation, the geometric boundaries are generated by a near-field radiation source spatial pre-calibration method based on K-means clustering. Each geometric boundary corresponds to a spatial cluster of concentrated radiation energy, and the geometric boundaries corresponding to different equivalent dipoles are independent and do not overlap.

[0013] In one possible implementation, the multiple independent reconstructions based on the reference field data and the geometric boundary include:

[0014] For each reconstruction, the population of the dynamic differential evolution algorithm is initialized using the geometric boundary as a hard constraint.

[0015] Based on the initialized population, mutation, crossover, and selection operations are performed, and boundary validity is determined and corrected. Iterative optimization is carried out until the preset termination condition is reached, and an equivalent dipole model that satisfies all geometric boundary constraints is output.

[0016] In one possible implementation, the method further includes:

[0017] A preset multiple is determined based on the confidence level requirement and a preset correspondence, wherein the correspondence is a mapping relationship between the confidence level and the preset multiple, and the preset multiple is used to determine the standard deviation.

[0018] In one possible implementation, the random initial condition includes any of the following:

[0019] Different initial populations can be randomly generated at the start of reconstruction, different random seeds can be used in the mutation operation of the differential evolution algorithm, or random perturbations can be introduced in the fitness function evaluation.

[0020] In one possible implementation, the preset termination condition includes at least one of the following: reaching a preset maximum number of iterations, the improvement of the population fitness function value being lower than a preset threshold for several consecutive generations, and the fitness function value of the best individual reaching a preset target value.

[0021] In one possible implementation, before performing statistical analysis on the dataset to calculate the statistical mean and standard deviation, the method further includes:

[0022] Identify and remove outliers in the dataset that significantly deviate from the overall sample population;

[0023] Accordingly, the statistical analysis of the dataset includes:

[0024] Perform statistical analysis on the processed dataset.

[0025] In one possible implementation, the method further includes:

[0026] When the far-field maximum value sample does not conform to the normal distribution characteristics, the far-field maximum value sample is subjected to data transformation processing to make the transformed data approach the normal distribution. Then, the statistical mean and standard deviation of the transformed data are calculated, and the data is inversely transformed back to the original scale as the prediction result.

[0027] Secondly, embodiments of this application provide a phase-free near-field reconstruction far-field prediction device, comprising:

[0028] The acquisition module is used to acquire the near-field magnetic field strength amplitude data of the device under test as reference field data and the geometric boundary of the spatial cluster. The geometric boundary is generated by a preset method and has the same number of preset equivalent dipoles.

[0029] The reconstruction module is used to perform multiple independent reconstructions based on the reference field data and the geometric boundary to obtain multiple sets of equivalent dipole models, wherein each reconstruction uses different random initial conditions;

[0030] The calculation module is used to calculate the maximum far-field electric field intensity of each equivalent dipole model at the target distance based on the electromagnetic field analytical formula, forming a dataset containing multiple far-field maximum value samples.

[0031] The statistical analysis module is used to perform statistical analysis on the dataset and calculate the statistical mean and standard deviation of the far-field maximum value samples in the dataset.

[0032] The output module is used to take the statistical average value as the final prediction result of the far-field radiation maximum value of the device under test when the far-field maximum value sample has a normal distribution characteristic, and to output the confidence interval of the prediction value by combining the standard deviation.

[0033] In one possible implementation, the geometric boundaries are generated by a near-field radiation source spatial pre-calibration method based on K-means clustering. Each geometric boundary corresponds to a spatial cluster of concentrated radiation energy, and the geometric boundaries corresponding to different equivalent dipoles are independent and do not overlap.

[0034] In one possible implementation, the reconstruction module includes:

[0035] For each reconstruction, the population of the dynamic differential evolution algorithm is initialized using the geometric boundary as a hard constraint.

[0036] Based on the initialized population, mutation, crossover, and selection operations are performed, and boundary validity is determined and corrected. Iterative optimization is carried out until the preset termination condition is reached, and an equivalent dipole model that satisfies all geometric boundary constraints is output.

[0037] In one possible implementation, the device further includes:

[0038] The determination module is used to determine a preset multiple based on the confidence level requirement and a preset correspondence, wherein the correspondence is a mapping relationship between the confidence level and the preset multiple, and the preset multiple is used to determine the standard deviation.

[0039] In one possible implementation, the random initial condition includes any of the following:

[0040] Different initial populations can be randomly generated at the start of reconstruction, different random seeds can be used in the mutation operation of the differential evolution algorithm, or random perturbations can be introduced in the fitness function evaluation.

[0041] In one possible implementation, the preset termination condition includes at least one of the following: reaching a preset maximum number of iterations, the improvement of the population fitness function value being lower than a preset threshold for several consecutive generations, and the fitness function value of the best individual reaching a preset target value.

[0042] In one possible implementation, the device further includes:

[0043] The preprocessing module is used to identify and remove outliers in the dataset that are significantly different from the overall sample population.

[0044] Accordingly, the statistical analysis module includes:

[0045] Perform statistical analysis on the processed dataset.

[0046] In one possible implementation, the device further includes:

[0047] The transformation processing module is used to perform data transformation processing on the far-field maximum value sample when the far-field maximum value sample does not conform to the normal distribution characteristics, so that the transformed data approaches the normal distribution, and then calculates the statistical mean and standard deviation of the transformed data, and inversely transforms it back to the original scale as the prediction result.

[0048] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0049] The memory stores computer-executed instructions;

[0050] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0051] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0052] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0053] The phase-free near-field reconstruction and far-field prediction method, apparatus, device, and storage medium provided in this application acquire near-field magnetic field strength amplitude data of the device under test as reference field data and the geometric boundary of the spatial cluster. Based on the reference field data and geometric boundary, multiple independent reconstructions are performed to obtain multiple sets of equivalent dipole models. Based on the analytical formula of electromagnetic field, the maximum far-field electric field strength of each set of equivalent dipole models is calculated at the target distance, forming a dataset containing multiple far-field maximum value samples. Statistical analysis is performed on the dataset to calculate the statistical mean and standard deviation of the far-field maximum value samples. When the far-field maximum value samples have a normal distribution characteristic, the statistical mean is used as the final prediction result of the far-field radiation maximum value of the device under test, and the confidence interval of the prediction value is determined by combining the standard deviation. The above method, through multiple independent reconstructions and statistical analysis, transforms the inherent non-uniqueness of the solution in phase-free near-field reconstruction from an "uncertainty defect" into quantifiable statistical reliability for the first time. Compared with existing technologies that only output far-field predictions from a single reconstruction, the statistical average value output by this scheme has higher stability and repeatability, and is not affected by the randomness of a single reconstruction. At the same time, the confidence interval provided provides an objective quantitative basis for the formulation of engineering design margins and risk decisions, filling the gap in existing technologies that cannot provide far-field predictions with statistical confidence. Attached Figure Description

[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0055] Figure 1 A flowchart illustrating the phase-free near-field reconstruction far-field prediction method provided in this application. Figure 1 ;

[0056] Figure 2 A flowchart illustrating the phase-free near-field reconstruction far-field prediction method provided in this application. Figure 2 ;

[0057] Figure 3 A schematic diagram of the phase-free near-field reconstruction far-field prediction device provided in this application;

[0058] Figure 4 A schematic diagram of the structure of the electronic device provided in this application.

[0059] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0060] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0061] With the rapid development of electronic information technology, the integration level and switching frequency of printed circuit boards (PCBs) and integrated circuits (ICs) continue to rise. The resulting electromagnetic interference (EMI) and electromagnetic compatibility (EMC) issues have become core challenges restricting the reliability of complex electronic systems. In fields such as automotive electronics, consumer electronics, communication equipment, and industrial control systems, the analysis and prediction of the radiation characteristics of devices under test (DUTs) are crucial for ensuring product compliance. For example, in automotive electronics R&D, electromagnetic radiation from vehicle controllers, radar modules, or communication modules may interfere with the normal operation of other electronic components; in the consumer electronics field, EMI issues in smartphones and wearable devices may affect signal transmission quality or pose health risks to users. Traditional EMC testing requires a 3-meter or 10-meter far-field environment, relying on large semi-anechoic chambers (SACs) to simulate a free-space environment. However, such testing places extremely high demands on site, equipment, and maintenance costs, especially for small and medium-sized enterprises or in the early stages of R&D, where the economic burden of frequent testing is unbearable. Existing technologies output a set of equivalent dipole models through single equivalent source reconstruction and use these to calculate far-field prediction values. However, when the reconstruction algorithm converges to different local optima due to a lack of phase information, the far-field prediction results fluctuate significantly, and the stability and repeatability of the prediction results cannot be guaranteed.

[0062] To address the aforementioned problems, this application provides a phase-free near-field reconstruction far-field prediction method, apparatus, device, and storage medium. This transforms the inherent non-uniqueness of solutions in phase-free near-field reconstruction from an uncertainty defect into quantifiable statistical reliability, thereby freeing the far-field prediction results from dependence on the randomness of a single reconstruction and making the prediction results more stable and repeatable. Specifically, existing technologies output a set of equivalent dipole models through a single equivalent source reconstruction and use these to calculate far-field prediction values. However, when the reconstruction algorithm converges to different local optima due to a lack of phase information, the far-field prediction results exhibit significant fluctuations, failing to guarantee the stability and repeatability of the prediction results. Considering the above problems, the inventors investigated whether a large-sample statistical mechanism could be used to calculate the far-field maximum value of multiple sets of equivalent dipole models obtained from multiple independent reconstructions and form a dataset. Then, the statistical mean and standard deviation could be calculated as the prediction result and confidence interval. For the first time, the inherent non-uniqueness of the solution in phase-free near-field reconstruction was transformed from an uncertainty defect into quantifiable statistical reliability. This freed the far-field prediction result from dependence on the randomness of a single reconstruction, and the prediction result has high stability and repeatability. At the same time, the provided confidence interval provides an objective quantitative basis for the formulation of engineering design margins and risk decisions, filling the gap in existing technologies that cannot provide far-field prediction values ​​with statistical confidence.

[0063] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0064] Figure 1 A flowchart illustrating the phase-free near-field reconstruction far-field prediction method provided in this application. Figure 1 ,like Figure 1 As shown, the method includes:

[0065] S101: Acquire the near-field magnetic field strength amplitude data of the device under test as reference field data and the geometric boundary of the space cluster.

[0066] Existing techniques only acquire near-field magnetic field amplitude data as a reference field and then perform a single reconstruction. Due to the lack of geometric boundary constraints, the reconstruction algorithm blindly searches within the entire plane, resulting in slow convergence and a tendency to get trapped in physically unreasonable local optima. More importantly, existing techniques rely solely on the results of a single reconstruction for far-field prediction, and when this reconstruction converges to different local optima, the prediction results exhibit significant fluctuations.

[0067] This step obtains two types of basic data required for subsequent large-sample statistical prediction: first, near-field measured reference data for equivalent source reconstruction; and second, spatial geometric boundaries used to constrain the search range for the equivalent dipole positions. The geometric boundaries are generated by a near-field radiation source spatial pre-calibration method based on K-means clustering. Each geometric boundary corresponds to a spatial cluster of concentrated radiation energy, and the geometric boundaries corresponding to different equivalent dipoles are independent and do not overlap.

[0068] Specifically, this step acquires two types of data simultaneously. The first type is the amplitude data of the phaseless magnetic field intensity in the scanning plane above the device under test, collected by the near-field scanning platform, which serves as a reference benchmark for subsequent equivalent source reconstruction. The second type is the geometric boundaries of spatial clusters with the same number of preset equivalent dipoles. Each geometric boundary corresponds to a spatial cluster with concentrated radiation energy, used for constraint optimization in reconstruction.

[0069] For example, for reference field data, first, the device under test (DUT) is fixed on the near-field scanning platform, and the scanning plane is set 6 mm above the DUT. The scanning area is set to cover the entire size of the DUT; for example, if the DUT is 100 mm × 80 mm, the scanning area is set to 100 mm × 80 mm. The sampling step size is set to 2 mm, so the number of sampling points in the x-direction is 100 / 2 + 1 = 51, the number of sampling points in the y-direction is 80 / 2 + 1 = 41, and the total number of sampling points is 51 × 41 = 2091.

[0070] Then, a three-dimensional precision mechanical motion control arm carries the magnetic field probe, moving point by point along a preset grid path. At each sampling point, the probe stops and collects the magnetic field signal, and a spectrum analyzer records the magnetic field amplitude at the target frequency (e.g., 1 GHz) at that point. After the data acquisition is complete, the coordinates of all sampling points are recorded. , ) and its corresponding magnetic field amplitude | , | Stored as a reference field data matrix.

[0071] For the geometric boundaries, the strong radiation region is divided into N spatial clusters (N equals the preset number of equivalent dipoles) using K-means clustering, and the geometric boundaries, including the minimum value on the x-axis, are output for each spatial cluster. maximum value on the x-axis minimum value of y-axis y-axis maximum value For example, when N=4, the following geometric boundaries may be obtained:

[0072] Cluster 1 boundary: x∈[0, 25], y∈[0, 40]

[0073] Cluster 2 boundary: x∈[25, 50], y∈[30, 80]

[0074] Cluster 3 boundary: x∈[50, 75], y∈[0, 45]

[0075] Cluster 4 boundary: x∈[60, 100], y∈[45, 80]

[0076] These geometric boundaries are stored in array form, where the i-th geometric boundary corresponds to the legal position range of the i-th equivalent dipole.

[0077] S102: Multiple independent reconstructions are performed based on reference field data and geometric boundaries to obtain multiple sets of equivalent dipole models.

[0078] Current techniques perform only one reconstruction, outputting a set of optimal equivalent dipole models. However, due to the lack of phase information in near-field data, the reconstruction problem itself suffers from severe non-uniqueness of solutions, and the result of a single reconstruction depends on random initial conditions, exhibiting randomness. When the process is repeated with a different random seed, different equivalent dipole distributions may be obtained, leading to significant differences in far-field prediction results. Current techniques cannot handle this uncertainty.

[0079] This step involves multiple independent reconstructions, sampling multiple sets of different equivalent dipole models to characterize the non-uniqueness of the reconstructed solutions, thus providing a data foundation for large-sample statistics. Each reconstruction uses different random initial conditions; optionally, the number of independent reconstructions is an integer greater than 30. The random initial conditions include any of the following:

[0080] Different initial populations can be randomly generated at the start of reconstruction, different random seeds can be used in the mutation operation of the differential evolution algorithm, or random perturbations can be introduced in the fitness function evaluation.

[0081] Specifically, a dynamic differential evolution equivalent source reconstruction method based on spatial boundary constraints is used to independently perform multiple reconstructions. Each reconstruction employs different random initial conditions, including but not limited to randomly generating different initial populations, using different random seeds in mutation operations, or introducing random perturbations in fitness evaluation. Each reconstruction outputs a set of equivalent dipole models that satisfy geometric boundary constraints. The number of reconstructions is set to an integer greater than 30, with 200 recommended to ensure that the sample size meets the statistical requirements of the law of large numbers. Because the dynamic differential evolution equivalent source reconstruction method based on spatial boundary constraints has an efficient constraint optimization mechanism and a relatively fast single reconstruction speed, multiple independent reconstructions are feasible in engineering.

[0082] For example, let's set the population size P=100, the equivalent number of dipoles N=4, and the maximum number of iterations. =200.

[0083] Initialization Phase: Generate 100 initial individuals. For the first equivalent dipole in each individual, its position coordinates are randomly generated within the boundary of cluster 1: = + rand() × ( - ), Similarly, the 2nd, 3rd, and 4th equivalent dipoles are randomly generated within the boundaries of clusters 2, 3, and 4, respectively. The dipole moment parameters are randomly generated across the entire range; for example, the dipole moment amplitude is set to a range of 0 to 1, and the phase range is set to 0 to 360 degrees.

[0084] Iterative optimization phase: For each target individual in the current population, three distinct individuals a, b, and c are randomly selected from the population. Differential mutation is performed: Candidate position = position of a + F × (position of b - position of c), where F is 0.8. After generating candidate individuals, it is checked whether the position of each equivalent dipole is within its corresponding geometric boundary. If the x-coordinate of the i-th dipole exceeds [x_min_i, ... If x < 0, then the boundary absorption strategy is used for correction: Then set as If x > Then set as The y-coordinate is corrected similarly.

[0085] Then, a crossover operation is performed between the target individual and the mutated individual, with a crossover probability CR of 0.9. For each parameter dimension, a random number rand is generated. If rand ≤ CR, the parameter is inherited from the mutated individual; otherwise, it is inherited from the target individual. The same boundary correction operation is performed again after the crossover.

[0086] Next, the fitness value of the experimental individual is calculated: based on the equivalent dipole model of the experimental individual, the near-field magnetic field amplitude at all sampling points on the scanning plane is calculated, and the root mean square error is calculated with the measured reference field data. If the root mean square error of the experimental individual is less than that of the target individual, the experimental individual is used to replace the target individual.

[0087] Repeat the above process for 200 generations, and output the individual with the best fitness value in this reconstruction, which is a set of equivalent dipole models that satisfy all geometric boundary constraints.

[0088] The specific process of multiple independent refactorings:

[0089] The number of reconstructions is set to 200. Before each reconstruction, the seed of the random number generator is reset to a different value. For example, seed 1 is used for the first reconstruction, seed 2 for the second reconstruction, and so on. Because the random seeds are different, the initial population and the random selection of mutation operations are different for each reconstruction. Therefore, 200 reconstructions will result in 200 sets of equivalent dipole models with comparable near-field fitting accuracy but different parameter distributions.

[0090] S103: Based on the analytical formula of electromagnetic field, calculate the maximum value of the far-field electric field intensity of each equivalent dipole model at the target distance, forming a dataset containing multiple far-field maximum value samples.

[0091] Current technology calculates the far-field maximum value for only a single equivalent dipole model, outputting a single predicted value. Due to the lack of comparison and statistical processing across multiple models, the reliability of this predicted value cannot be quantified, and engineers cannot determine its fluctuation range and credibility.

[0092] This step transforms multiple sets of equivalent dipole models into multiple far-field maximum samples, forming a dataset for statistical analysis.

[0093] Specifically, for the multiple equivalent dipole models obtained in S102, the maximum far-field electric field strength at the target distance for each model is calculated based on the analytical formula for electromagnetic fields. The target distance is determined according to the electromagnetic compatibility testing standards of the device under test, typically 1 meter, 3 meters, or 10 meters. The maximum far-field electric field strength is the maximum value across all spatial directions, i.e., the maximum value of the electric field strength is taken across all spatial angular directions. After the calculation is completed, multiple samples of the maximum far-field values ​​are organized into a dataset for subsequent statistical analysis.

[0094] For example, for a set of equivalent dipole models obtained in S102, assuming it contains N equivalent magnetic dipoles, the position of the i-th dipole is ( , , ), the magnetic dipole moment is ( , , The target distance is set to 3 meters.

[0095] In the far-field region, the relationship between the electric field strength and the magnetic dipole moment is given by the following formula:

[0096] For a magnetic dipole located at the origin, the far-field electric field components in spherical coordinates are:

[0097]

[0098]

[0099] in Here, is the free-space wave impedance, k is the wave number, and r is the target distance.

[0100] For a magnetic dipole located at an arbitrary position, the phase delay factor e^(jk·r') needs to be considered, where r' is the vector from the dipole position to the origin.

[0101] In actual calculations, in three-dimensional space... and Perform discrete sampling. From 0 to 180 degrees, with a step size of 5 degrees, a total of 37 sampling points; From 0 to 360 degrees, with a step size of 5 degrees, there are a total of 73 sampling points. The total number of directions is 37 × 73 = 2701.

[0102] For each direction ( , The total electric field strength is obtained by superimposing the electric field vectors generated by N dipoles in this direction. ( , Then iterate through all 2701 directions to find the maximum value of the electric field strength. .

[0103] The specific process of dataset formation:

[0104] The above calculations were performed on 200 equivalent dipole models respectively, resulting in 200 far-field maximum values: , , ..., Store these values ​​as a one-dimensional array, for example, [32.5, 31.8, 33.2, ..., 30.9], in dB. V / m. This array is the dataset used for statistical analysis.

[0105] S104: Perform statistical analysis on the dataset and calculate the statistical mean and standard deviation of the far-field maximum sample in the dataset.

[0106] Current technologies lack a step for statistical analysis of multiple far-field predictions. When anomalies occur in a single reconstruction result (e.g., the algorithm converges to a poor local optimum), current technologies cannot identify and handle such anomalies, potentially resulting in predictions that deviate significantly from the true values.

[0107] This step involves statistical analysis of the far-field maximum value sample dataset, calculating the statistical mean and standard deviation to provide a quantitative basis for the final prediction results and confidence intervals.

[0108] Specifically, the dataset is first subjected to statistical analysis to calculate the statistical mean and standard deviation of the far-field maximum values. The statistical mean reflects the central tendency of the sample, while the standard deviation reflects the dispersion of the sample.

[0109] Optionally, before step S104, it is also necessary to identify and remove outliers in the dataset that are significantly deviated from the overall sample population.

[0110] Before statistical analysis, outlier handling is performed: outliers that significantly deviate from the overall population are identified and removed. Outlier identification can employ box plots, Z-scores, or interquartile range (IMR)-based criteria. For example, box plots identify outliers as points below the first quartile minus 1.5 IMR or above the third quartile plus 1.5 IMR. Removing outliers eliminates interference from extreme values ​​caused by single reconstruction failures (such as algorithm non-convergence or getting trapped in a poor local optimum), improving the robustness of the statistical results.

[0111] For example, using the box plot method, the quartiles of the dataset are first calculated. The 200 far-field maximum values ​​are sorted in ascending order. The 50th value is the first quartile Q1, the 100th value is the median Q2, and the 150th value is the third quartile Q3. The interquartile range (IQR) is then calculated as IQR = Q3 - Q1.

[0112] Define outlier criteria: lower boundary = Q1 - 1.5 × IQR, upper boundary = Q3 + 1.5 × IQR. Iterate through 200 samples, identifying and removing samples with values ​​less than the lower boundary or greater than the upper boundary as outliers. For example, if Q1 = 31.2, Q3 = 33.5, and IQR = 2.3, then the lower boundary = 31.2 - 3.45 = 27.75, and the upper boundary = 33.5 + 3.45 = 36.95. Suppose a sample has a value of 38.2, which is greater than 36.95; in this case, it is identified as an outlier and removed.

[0113] After removing outliers, the number of remaining valid samples is M', for example, M'=198.

[0114] The specific calculation process for statistical mean and standard deviation:

[0115] Let the effective samples be x_1, x_2, ..., x_M'. The formula for calculating the statistical mean is: μ = (x_1 + x_2 + ... + x_M') / M'.

[0116] The formula for calculating the standard deviation is: .

[0117] For example, calculation =32.5 dB V / m, =1.2 dB.

[0118] S105: When the far-field maximum value sample has a normal distribution characteristic, the statistical mean is used as the final prediction result of the far-field radiation maximum value of the device under test, and the confidence interval of the prediction value is determined by combining the standard deviation.

[0119] Current technologies output a single far-field prediction value without a confidence interval, making it impossible to quantify the uncertainty of the prediction result. When faced with this prediction value, engineers cannot judge its accuracy and range of fluctuation, making it difficult to make reliable engineering decisions based on it.

[0120] This step outputs far-field prediction results with statistical confidence and their confidence intervals based on the statistical mean and standard deviation.

[0121] Specifically, when the far-field maximum value samples conform to a normal distribution, the statistical mean is used as the final predicted result of the far-field radiation maximum value of the device under test, and the standard deviation of the statistical mean plus or minus a preset multiple is used as the confidence interval of the predicted value. There is a preset correspondence between the preset multiple and the confidence level requirement. For example, when the confidence level requirement is approximately 68.3%, the preset multiple is 1; when the confidence level requirement is approximately 95.4%, the preset multiple is 2; and when the confidence level requirement is approximately 99.7%, the preset multiple is 3. This correspondence is based on the properties of the normal distribution: approximately 68.3% of the samples fall within the range of the mean plus or minus 1 standard deviation, approximately 95.4% fall within the range of plus or minus 2 standard deviations, and approximately 99.7% fall within the range of plus or minus 3 standard deviations.

[0122] Among them, the preset multiple is determined according to the confidence level requirement and the preset correspondence, which is the mapping relationship between the confidence level and the preset multiple.

[0123] Optionally, the method further includes:

[0124] When the far-field maximum sample does not conform to the characteristics of a normal distribution, the far-field maximum sample is processed by data transformation to make the transformed data approach a normal distribution. Then, the statistical mean and standard deviation of the transformed data are calculated, and the data is inversely transformed back to the original scale as the prediction result.

[0125] In practical applications, due to the radiation characteristics of the device under test or the characteristics of the reconstruction algorithm, multiple far-field maximum samples may not perfectly follow a normal distribution, but may exhibit a skewed or heavy-tailed distribution. In such cases, directly using the confidence interval calculation formula based on the normal distribution assumption may not be accurate enough.

[0126] Therefore, when the far-field maximum sample does not satisfy the normal distribution assumption, data transformation can be used to make it approximate a normal distribution, thus ensuring the applicability of the statistical prediction method.

[0127] Specifically, a normality test is performed on the original far-field maximum sample, such as calculating the skewness and kurtosis coefficients, or performing the Shapiro-Wilk test. If it is determined that the data does not conform to a normal distribution, an appropriate data transformation method is selected. Logarithmic transformation is suitable for right-skewed data, the Box-Cox transformation is a parametric transformation method that optimizes the parameter λ to maximize the normality of the transformed data, and the Johnson transformation is suitable for a wider range of distribution types. After transformation, the statistical mean and standard deviation are calculated for the transformed data, and then the statistics are mapped back to the original scale through an inverse transformation to obtain the final prediction result.

[0128] The phase-free near-field reconstruction and far-field prediction method provided in this application acquires the near-field magnetic field strength amplitude data of the device under test (DUT) as reference field data and the geometric boundary of the spatial cluster. Based on the reference field data and geometric boundary, multiple independent reconstructions are performed to obtain multiple sets of equivalent dipole models. The maximum far-field electric field strength of each equivalent dipole model is calculated at the target distance based on the analytical formula of the electromagnetic field, forming a dataset containing multiple far-field maximum value samples. Statistical analysis is performed on the dataset to calculate the statistical mean and standard deviation of the far-field maximum value samples. When the far-field maximum value samples exhibit a normal distribution characteristic, the statistical mean is used as the final prediction result of the far-field radiation maximum value of the DUT, and the confidence interval of the predicted value is determined by combining the standard deviation. This method, through multiple independent reconstructions and statistical analysis, transforms the inherent non-uniqueness of solutions in phase-free near-field reconstruction from an "uncertainty defect" into a quantifiable "statistical reliability" for the first time. Compared with existing technologies that only output far-field predictions from a single reconstruction, the statistical average value output by this scheme has higher stability and repeatability, and is not affected by the randomness of a single reconstruction. At the same time, the confidence interval provided provides an objective quantitative basis for the formulation of engineering design margins and risk decisions, filling the gap in existing technologies that cannot provide far-field predictions with statistical confidence.

[0129] Figure 2 A flowchart illustrating the phase-free near-field reconstruction far-field prediction method provided in this application. Figure 2 ,like Figure 2 As shown, based on the above embodiments, S102 specifically includes:

[0130] S201: For each reconstruction, initialize the population of the dynamic differential evolution algorithm with geometric boundaries as hard constraints.

[0131] In existing technologies, the position coordinates of equivalent dipoles are randomly initialized across the entire scanning plane. This results in a large number of initial dipoles being distributed in non-radiative or weakly radiative regions. For example, the actual radiation source of the device under test may be concentrated in the central region, but the initialized dipoles may appear at the edge of the scan. These dipoles located in non-radiative regions require multiple iterations in subsequent optimizations to gradually move into high-radiative regions, or they may remain in illegal regions for extended periods due to a lack of effective gradient information, severely slowing down the convergence speed. More seriously, multiple dipoles may be initialized in the same region, causing mutual interference during the optimization process and ultimately converging to a physically unreasonable overlapping distribution.

[0132] In the initial stage of the reconstruction algorithm, this step uses geometric boundaries to constrain and initialize the position parameters of the equivalent dipoles, ensuring that each equivalent dipole is located in its corresponding concentrated radiation energy region from the beginning, thus laying a physically reasonable foundation for subsequent iterative optimization.

[0133] Specifically, the strong radiation region is divided into N spatial clusters (N equals the preset number of equivalent dipoles) using K-means clustering. Each spatial cluster outputs its geometric boundary, including the minimum value in the x-axis direction. and maximum value Minimum value in the y-axis direction and maximum value These geometric boundaries are stored as an array `boundaries[N]`, where `boundaries[i]` corresponds to the legal position range of the i-th equivalent dipole.

[0134] Next, set the population size P, typically between 50 and 200. Let's take P=100 as an example. For each individual in the population, N equivalent dipole position parameters and dipole moment parameters need to be generated.

[0135] The position coordinates of the i-th equivalent dipole are randomly generated using a uniform distribution within the corresponding geometric boundary. The specific calculation formula is as follows:

[0136] = + rand() × ( - )

[0137] = + rand() × ( - )

[0138] The `rand()` function generates a random number that is uniformly distributed between 0 and 1. For example, if the geometric boundary of the third spatial cluster is x∈[50, 75] and y∈[0, 45], then the generated x-coordinates will be between 50 and 75, and the y-coordinates will be between 0 and 45.

[0139] For the dipole moment parameter, since the dipole moment reflects radiation intensity and is not directly related to spatial location, it is randomly generated across the entire range. The dipole moment is a complex number, including a real part and an imaginary part. The amplitude range of the dipole moment is set to 0 to M_max (M_max is preset according to the radiation level of the device under test, such as 1), and the phase range is 0 to 360 degrees. Real part = amplitude × cos(phase), imaginary part = amplitude × sin(phase). The real and imaginary parts can also be directly generated uniformly and randomly within the range [-M_max, M_max].

[0140] Repeat the above generation process P times to obtain an initial population containing P individuals. The N equivalent dipoles of each individual satisfy the positional constraint, that is, the i-th dipole must be located within the geometric boundary of the i-th spatial cluster.

[0141] S202: Based on the initialized population, perform mutation, crossover, and selection operations, and execute boundary validity judgment and correction. Iterate and optimize until the preset termination condition is reached, and output an equivalent dipole model that satisfies all geometric boundary constraints.

[0142] Existing techniques do not perform constraint checks on position parameters during mutation and crossover operations. In mutation, candidate position = position of a + F × (position of b - position of c). Since positions a, b, and c may be located in different regions, the difference vector may push candidate positions outside the geometric boundaries or even outside the scan plane. Although these illegal positions can theoretically compute the near field, they are physically highly unreasonable—dipoles located outside the device under test cannot accurately represent the internal radiation source. Existing techniques allow these illegal individuals to participate in fitness evaluation and selection, causing the algorithm to perform a large number of invalid searches in the illegal space, wasting computational resources, and easily converging to physically uninterpretable local optima.

[0143] In this step, during the iterative optimization process, the population evolution is driven by the mutation, crossover, and selection operations of the differential evolution algorithm. At the same time, the boundary legality judgment and correction mechanism ensures that all individuals always satisfy the geometric boundary constraints, and finally outputs the optimal equivalent dipole model that satisfies the constraints.

[0144] Specifically, for each target individual in the current population, three distinct individuals, denoted as a, b, and c, are randomly selected from the population. A differential mutation operation is then performed to generate candidate individuals. For the position parameter, the mutation formula is:

[0145] = + F × ( - )

[0146] = + F × ( - )

[0147] Where F is the scaling factor, which is usually between 0.5 and 1; in this scheme, F = 0.8 is used. The same variation formula is used for the dipole moment parameter.

[0148] After generating candidate individuals, boundary validity is immediately determined. For the i-th equivalent dipole, its candidate position is checked ( , Does it satisfy:

[0149] ≤ ≤

[0150] ≤ ≤

[0151] If all equivalent dipoles meet the conditions, the candidate individual is valid and directly used as a variant individual. If any equivalent dipole exceeds the boundary, boundary correction is performed. This embodiment provides three correction strategies:

[0152] Boundary absorption strategy: If < Then set = ;like > Then set = The same applies to the y-coordinate. This strategy is the simplest to implement and the fastest to calculate.

[0153] Boundary reflection strategy: If < Calculate the excess amount d = - ,set up = + d; if > Calculate the excess amount d = - ,set up = - d. If the result still exceeds the limit after correction, reflection continues until it falls within the boundary. This strategy maintains the continuity of the position distribution.

[0154] Regeneration strategy: Valid positions are randomly generated within the corresponding geometric boundaries to replace invalid positions. This strategy ensures diversity.

[0155] After the correction is completed, a mutant individual that satisfies the geometric boundary constraints is obtained.

[0156] Crossover operations and boundary correction:

[0157] The target individual and the mutant individuals are subjected to a binomial crossover operation to generate experimental individuals. For each parameter dimension of each equivalent dipole, a random number rand between 0 and 1 is generated. The crossover probability CR is set, usually between 0.8 and 0.9; this scheme uses CR=0.9. At the same time, a dimension j_rand is randomly selected to ensure that at least one dimension is inherited from the mutant individual.

[0158] The crossover rule is as follows: if rand ≤ CR or the current dimension is equal to j_rand, then the experimental parameters are inherited from the mutated individual; otherwise, they are inherited from the target individual.

[0159] After crossover, an initial experimental individual is generated. The same boundary validity determination and correction process as the mutation operation is performed on this experimental individual to obtain an experimental individual that satisfies the geometric boundary constraints.

[0160] Fitness evaluation and selection:

[0161] Calculate the fitness values ​​of the target individual and the experimental individual. If the RMSE of the experimental individual is less than that of the target individual (i.e., the experimental individual is better), then replace the target individual with the experimental individual in the next generation of the population; otherwise, retain the target individual.

[0162] Iterative optimization and termination:

[0163] Repeat the above mutation, crossover, and selection operations until the preset termination conditions are met. The termination conditions include: reaching the preset maximum number of iterations (e.g., 200 generations), the improvement in the fitness value of the population being lower than the preset threshold (e.g., 1e-6) for 10 consecutive generations, or the fitness value of the best individual reaching the preset target value.

[0164] After the iteration is completed, the individual with the best fitness value is selected from the current population and used as the output of this reconstruction—an equivalent dipole model that satisfies all geometric boundary constraints.

[0165] The phase-free near-field reconstruction and far-field prediction method provided in this application initializes the population of a dynamic differential evolution algorithm for each reconstruction, using geometric boundaries as hard constraints. Based on the initialized population, mutation, crossover, and selection operations are performed, and boundary validity is determined and corrected. Iterative optimization continues until a preset termination condition is met, outputting an equivalent dipole model that satisfies all geometric boundary constraints. This method compresses the equivalent dipole position search space from a global search across the entire plane to a local search within multiple independent subspaces. The search space dimension decreases exponentially, significantly improving the convergence speed, while ensuring that the final output equivalent dipole model has a clear physical location meaning, with each dipole located within a pre-calibrated high-radiation region.

[0166] Figure 3 This is a schematic diagram of the phase-free near-field reconstruction far-field prediction device provided in this application, as shown below. Figure 3 As shown, the phase-free near-field reconstruction far-field prediction device 300 provided in this embodiment specifically includes:

[0167] The acquisition module 301 is used to acquire the near-field magnetic field strength amplitude data of the device under test as reference field data and the geometric boundary of the space cluster. The geometric boundary is generated by a preset method and has the same number of preset equivalent dipoles.

[0168] The reconstruction module 302 is used to perform multiple independent reconstructions based on the reference field data and geometric boundaries to obtain M sets of equivalent dipole models, wherein each reconstruction uses different random initial conditions;

[0169] The calculation module 303 is used to calculate the maximum value of the far-field electric field intensity of each equivalent dipole model at the target distance based on the electromagnetic field analytical formula, forming a dataset containing multiple far-field maximum value samples.

[0170] The statistical analysis module 304 is used to perform statistical analysis on the dataset and calculate the statistical mean and standard deviation of the far-field maximum value samples in the dataset.

[0171] The output module 305 is used to take the statistical average as the final prediction result of the far-field radiation maximum value of the device under test when the far-field maximum value sample has a normal distribution characteristic, and to determine the confidence interval of the prediction value by combining the standard deviation.

[0172] In one possible implementation, the geometric boundaries are generated by a near-field radiation source spatial pre-calibration method based on K-means clustering. Each geometric boundary corresponds to a spatial cluster of radiation energy concentration, and the geometric boundaries corresponding to different equivalent dipoles are independent and do not overlap.

[0173] In one possible implementation, the reconstruction module 302 includes:

[0174] For each reconstruction, the population of the dynamic differential evolution algorithm is initialized using geometric boundaries as hard constraints.

[0175] Based on the initialized population, mutation, crossover, and selection operations are performed, and boundary validity is determined and corrected. Iterative optimization is carried out until the preset termination condition is reached, and an equivalent dipole model that satisfies all geometric boundary constraints is output.

[0176] In one possible implementation, the phase-free near-field reconstruction far-field prediction device 300 further includes:

[0177] The determination module 306 is used to determine the preset multiple based on the confidence level requirement and the preset correspondence, which is a mapping relationship between the confidence level and the preset multiple.

[0178] In one possible implementation, the random initial conditions include any of the following:

[0179] Different initial populations can be randomly generated at the start of reconstruction, different random seeds can be used in the mutation operation of the differential evolution algorithm, or random perturbations can be introduced in the fitness function evaluation.

[0180] In one possible implementation, the preset termination condition includes at least one of the following: reaching a preset maximum number of iterations, the improvement of the population fitness function value being lower than a preset threshold for several consecutive generations, or the fitness function value of the best individual reaching a preset target value.

[0181] In one possible implementation, the phase-free near-field reconstruction far-field prediction device 300 further includes:

[0182] The preprocessing module 307 is used to identify and remove outliers in the dataset that are significantly different from the overall sample population.

[0183] Correspondingly, the statistical analysis module 304 includes:

[0184] Perform statistical analysis on the processed dataset.

[0185] In one possible implementation, the phase-free near-field reconstruction far-field prediction device 300 further includes:

[0186] The transformation processing module 308 is used to perform data transformation processing on the far-field maximum value sample when the far-field maximum value sample does not conform to the normal distribution characteristics, so that the transformed data approaches the normal distribution, and then calculates the statistical mean and standard deviation of the transformed data, and inversely transforms it back to the original scale as the prediction result.

[0187] The phaseless near-field reconstruction far-field prediction device provided in this embodiment can execute the phaseless near-field reconstruction far-field prediction method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0188] Figure 4 A schematic diagram of the structure of the electronic device provided in this application. Figure 4 As shown, the electronic device 400 provided in this embodiment includes at least one processor 401 and a memory 402. Optionally, the electronic device 400 further includes a communication component 403. The processor 401, memory 402, and communication component 403 are connected via a bus 404.

[0189] In a specific implementation, at least one processor 401 executes computer execution instructions stored in memory 402, causing at least one processor 401 to perform the above-described method.

[0190] The specific implementation process of processor 401 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0191] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0192] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0193] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0194] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0195] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0196] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0197] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0198] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0199] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0200] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0201] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0202] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0203] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A phaseless near-field to far-field reconstruction method, characterized in that, include: The near-field magnetic field strength amplitude data of the device under test is obtained as reference field data and the geometric boundary of the spatial cluster. The geometric boundary is generated by a preset method and has the same number of preset equivalent dipoles. Multiple independent reconstructions are performed based on the reference field data and the geometric boundary to obtain multiple sets of equivalent dipole models, wherein each reconstruction uses different random initial conditions. Based on the analytical formula of electromagnetic field, the maximum value of the far-field electric field intensity of each equivalent dipole model at the target distance is calculated, forming a dataset containing multiple far-field maximum value samples; Perform statistical analysis on the dataset to calculate the statistical mean and standard deviation of the far-field maximum value samples in the dataset; When the far-field maximum value sample follows a normal distribution, the statistical average is used as the final prediction result of the far-field radiation maximum value of the device under test, and the confidence interval of the prediction value is determined by combining the standard deviation.

2. The method of claim 1, wherein, The geometric boundaries are generated by a near-field radiation source spatial pre-calibration method based on K-means clustering. Each geometric boundary corresponds to a spatial cluster of radiation energy concentration, and the geometric boundaries corresponding to different equivalent dipoles are independent and do not overlap.

3. The method of claim 1, wherein, The process of performing multiple independent reconstructions based on the reference field data and the geometric boundary includes: For each reconstruction, the population of the dynamic differential evolution algorithm is initialized using the geometric boundary as a hard constraint. Based on the initialized population, mutation, crossover, and selection operations are performed, and boundary validity is determined and corrected. Iterative optimization is carried out until the preset termination condition is reached, and an equivalent dipole model that satisfies all geometric boundary constraints is output.

4. The method according to claim 1, characterized in that, The method further includes: A preset multiple is determined based on the confidence level requirement and a preset correspondence, wherein the correspondence is a mapping relationship between the confidence level and the preset multiple, and the preset multiple is used to determine the standard deviation.

5. The method according to claim 1, characterized in that, The random initial conditions include any one of the following: Different initial populations can be randomly generated at the start of reconstruction, different random seeds can be used in the mutation operation of the differential evolution algorithm, or random perturbations can be introduced in the fitness function evaluation.

6. The method according to claim 1, characterized in that, Before performing statistical analysis on the dataset and calculating the statistical mean and standard deviation, the method further includes: Identify and remove outliers in the dataset that significantly deviate from the overall sample population; Accordingly, the statistical analysis of the dataset includes: Perform statistical analysis on the processed dataset.

7. The method according to claim 1, characterized in that, The method further includes: When the far-field maximum value sample does not conform to the normal distribution characteristics, the far-field maximum value sample is subjected to data transformation processing to make the transformed data approach the normal distribution. Then, the statistical mean and standard deviation of the transformed data are calculated, and the data is inversely transformed back to the original scale as the prediction result.

8. A phase-free near-field reconstruction far-field prediction device, characterized in that, include: The acquisition module is used to acquire the near-field magnetic field strength amplitude data of the device under test as reference field data and the geometric boundary of the spatial cluster. The geometric boundary is generated by a preset method and has the same number of preset equivalent dipoles. The reconstruction module is used to perform multiple independent reconstructions based on the reference field data and the geometric boundary to obtain multiple sets of equivalent dipole models, wherein each reconstruction uses different random initial conditions; The calculation module is used to calculate the maximum far-field electric field intensity of each equivalent dipole model at the target distance based on the electromagnetic field analytical formula, forming a dataset containing multiple far-field maximum value samples. The statistical analysis module is used to perform statistical analysis on the dataset and calculate the statistical mean and standard deviation of the far-field maximum value samples in the dataset. The output module is used to take the statistical average value as the final prediction result of the far-field radiation maximum value of the device under test when the far-field maximum value sample has a normal distribution characteristic, and to output the confidence interval of the prediction value by combining the standard deviation.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processing unit, are used to implement the method as described in any one of claims 1 to 7.

Images