Method for identifying and positioning electromagnetic interference source based on multi-band EMI (Electro-Magnetic Interference)

Abstract

The invention provides an electromagnetic interference source identification and positioning method based on multi-band EMI. The method comprises the following steps: carrying out preprocessing and standardization processing on acquired electromagnetic characteristic data, real-time environment data of a sensor network and operation state data of electronic equipment; based on the predicted electromagnetic environment perception result, extracting electromagnetic environment feature data, performing noise reduction and preprocessing on the electromagnetic environment feature data of different dimensions, performing weight distribution and optimal combination, and performing iterative calculation to obtain an interference source position coordinate; the optimized sensor array layout and parameter setting are applied to the test process of the electronic equipment, performance parameters of the electronic equipment in different electromagnetic environments are obtained, the performance parameters comprise electromagnetic compatibility and anti-interference capacity, and a test result is compared with a preset standard value and calibrated.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for identifying and locating electromagnetic interference sources based on multi-band EMI. Background Art

[0002] In electronic equipment testing laboratories, in order to achieve comprehensive perception and analysis of the electromagnetic environment, it is necessary to integrate the electromagnetic characteristic data of the equipment and the real-time environmental data of the sensor network. However, in actual operation, due to the complex and changeable electromagnetic environment in the laboratory, the electromagnetic characteristics of various types of equipment vary greatly, and the layout of the sensor array is difficult to optimize, resulting in many technical difficulties in the data collection and analysis process. First, different types and models of equipment in the laboratory will generate various electromagnetic interferences during operation, and the frequency characteristics, time characteristics and spatial distribution characteristics of these interference sources are different. How to accurately locate and characterize the interference sources is a major problem. Secondly, in order to fully perceive the electromagnetic environment, a large number of sensors need to be deployed in the laboratory, but the layout of the sensor array will directly affect the accuracy and efficiency of data collection. How to optimize the layout of the sensor array to obtain high-quality data is also a thorny problem. Furthermore, in the data analysis process, how to effectively integrate the electromagnetic characteristic data of the equipment with the environmental data collected by the sensor network, and conduct a comprehensive analysis combined with the operating status of the equipment, so as to achieve accurate characterization of the electromagnetic environment and precise positioning of the interference source, is also a technical problem that needs to be solved urgently. Finally, due to the complexity and uncertainty of the electromagnetic environment, how to use the results obtained from perception and analysis to calibrate the test results of the equipment to improve the accuracy and reliability of the test is also an issue worthy of in-depth study. Summary of the invention

[0003] The present invention provides a method for identifying and locating electromagnetic interference sources based on multi-band EMI, which mainly includes:

[0004] Preprocess and standardize the acquired electromagnetic characteristic data, real-time environmental data of the sensor network, and operating status data of the electronic equipment;

[0005] By collecting the electromagnetic field strength data generated by various operating equipment in the laboratory, the frequency characteristics of different interference sources, the time-varying characteristics of electromagnetic signals, and the spatial distribution relationship information of electronic equipment, a probability distribution model of interference sources is constructed to distinguish and identify multi-source interference, and the characteristics and spatial positioning of each interference source in the laboratory electromagnetic environment are characterized;

[0006] The standardized electromagnetic characteristic data of electronic equipment, the real-time environmental data of the sensor network and the operating status data of the electronic equipment are integrated and input into the trained and optimized interference source probability distribution model to predict and output the electromagnetic environment perception results, and visualize and analyze the electromagnetic environment perception results;

[0007] Based on the predicted electromagnetic environment perception results, the electromagnetic environment feature data is extracted, the electromagnetic environment feature data of different dimensions are denoised and preprocessed, and weight allocation and optimization combination are performed, and the interference source location coordinates are obtained through iterative calculation;

[0008] According to the location coordinates of the interference source, the array of the sensor network is evaluated in different regions, and the areas where the detection coverage or signal quality is lower than the preset requirements are identified. The sensor position and orientation within the preset range of the identification area are adjusted. At the same time, the sensitivity parameters of the sensor are adjusted according to the electromagnetic environment perception results. During the adjustment process, the detection effect of each area is evaluated and analyzed, and the sensor array layout and parameters are optimized;

[0009] The optimized sensor array layout and parameter settings are applied to the testing process of electronic equipment to obtain the performance parameters of electronic equipment in different electromagnetic environments. The performance parameters include electromagnetic compatibility and anti-interference ability, and the test results are compared and calibrated with preset standard values.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses a method for identifying and locating electromagnetic interference sources based on multi-band EMI. The method obtains the electromagnetic characteristics, environment and operating status data of electronic equipment, and constructs a mixed Gaussian model of the electromagnetic environment to identify and locate multi-source interference. Based on the model prediction results, the present invention extracts electromagnetic environment characteristics, calculates the location of the interference source, and performs regional evaluation and optimization on the sensor network. By adjusting the sensor position, orientation and sensitivity parameters, the present invention improves the detection coverage and signal quality. Finally, the optimized sensor layout is applied to electronic equipment testing to evaluate its electromagnetic compatibility and anti-interference ability. The present invention realizes accurate perception of the electromagnetic environment, accurate positioning of the interference source and dynamic optimization of the sensor network, effectively improving the performance and reliability of electronic equipment in complex electromagnetic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention is a flowchart of a method for identifying and locating a multi-band EMI electromagnetic interference source. DETAILED DESCRIPTION

[0013] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] like Figure 1 In this embodiment, a method for identifying and locating a multi-band EMI electromagnetic interference source may specifically include:

[0015] S101, preprocessing and standardizing the acquired electromagnetic characteristic data, the real-time environment data of the sensor network, and the operation status data of the electronic equipment.

[0016] Electromagnetic characteristic data of electronic equipment is acquired from a plurality of electromagnetic sensor nodes, and the electromagnetic characteristic data records timing information at a preset sampling frequency; wavelet denoising is performed on the electromagnetic characteristic data to obtain a first type of processed data, and ambient temperature and humidity data and light intensity data are collected through a distributed sensor network and mean filtered to obtain a second type of processed data; the first type of processed data and the second type of processed data are time-series aligned through timestamps, and the correlation between the first type of processed data and the second type of processed data is calculated to obtain an abnormal data set; a fixed-length sliding time window is used to segment the abnormal data set, and the segmented data is clustered to obtain a device state feature vector, and an operation state correspondence database is established according to the device state feature vector.

[0017] Specifically, electromagnetic characteristic data generated by the electronic device in different working frequency bands are obtained from multiple electromagnetic sensor nodes, and the electromagnetic characteristic data are recorded in time series according to a preset sampling frequency. At the same time, environmental temperature and humidity data and light intensity data are collected through a distributed sensor network, and the collected data are stored according to a unified timestamp. Wavelet denoising is performed on the collected electromagnetic characteristic data to obtain the first type of processed data, and the electromagnetic interference level is determined by comparing the amplitude of the first type of processed data with a preset threshold value. At the same time, mean filtering is performed on the environmental temperature and humidity data and the light intensity data to obtain the second type of processed data. The equipment operation parameters are obtained from the electronic equipment operation controller and the data is standardized to obtain the third type of processed data. The first type of processed data, the second type of processed data and the third type of processed data are aligned in time series by the timestamp. A multivariate regression model is established based on the data after time series alignment to determine the correlation between the electromagnetic characteristic data and the environmental data, and the data points whose correlation exceeds the preset range are marked as abnormal to obtain an abnormal data set. The abnormal data set is segmented using a fixed-length sliding time window, and the segmented data is clustered by a density clustering method to obtain the device state feature vector, and a database of the corresponding relationship between the feature vector and the device operation state is established. According to the database of correspondence between the feature vector and the equipment operation status, the state of the real-time collected electromagnetic characteristic data, environmental data and operating parameter data is identified to obtain the equipment operation status judgment result. The electromagnetic characteristic data collection involves the layout of many sensor nodes. Multiple electromagnetic sensors are arranged around the electronic equipment at different azimuth angles. The sensor spacing is 20 cm and the sampling frequency is set at 100 Hz to achieve all-round monitoring of the electromagnetic field intensity. The electromagnetic data obtained by each sensor node contains amplitude and phase information. The amplitude range is between 0-100mV and the phase range is between 0-360 degrees. Millisecond timestamps are added when the data is recorded. At the same time, temperature sensors are arranged in the sensor network to collect ambient temperature data. The measurement range is -40 to 85 degrees Celsius and the accuracy is 0.1 degrees Celsius. The humidity sensor has a measurement range of 0-100% and an accuracy of 1%. The light sensor has a measurement range of 0-65000 lux. The wavelet denoising process for electromagnetic characteristic data uses the db4 wavelet basis function, sets a 4-layer decomposition, and suppresses high-frequency noise. The electromagnetic interference level is divided into segmented threshold judgment methods, and 0-20mV is defined as low interference, 20-50mV is defined as medium interference, and 50mV and above are defined as high interference. The mean filtering of environmental data uses a 5-second sliding window to reduce the impact of instantaneous fluctuations. The equipment operation parameters include key indicators such as voltage, current, and power factor. The voltage measurement range is 0-380V, the current measurement range is 0-100A, and the power factor range is 0-1. Data standardization uses the maximum and minimum value normalization method to map all parameters to the 0-1 interval. Data timing alignment adopts the principle of proximity, matching data from different sources according to timestamps, and the maximum allowable time difference is 50 milliseconds.In the multivariate regression model analysis, the electromagnetic characteristic data is used as the dependent variable, and the environmental data is used as the independent variable to establish a linear relationship model. The Pearson correlation coefficient is used to measure the correlation. When the absolute value of the correlation coefficient exceeds 0.8, it is determined to be strongly correlated, and when it is less than 0.3, it is determined to be weakly correlated. Abnormal data marking is performed on data points whose correlation coefficients deviate significantly from the historical mean. The sliding time window length is set to 60 seconds, and the window sliding step is 10 seconds. The DBSCAN algorithm is used for density clustering, and the neighborhood radius is set to 0.1 and the minimum number of samples is 5 points. The equipment state feature vector contains 10 dimensions, which correspond to the statistical characteristics of different operating parameters. The state recognition results are divided into multiple categories such as normal operation, load fluctuation, and efficiency reduction. Each state has a corresponding feature value range definition. In the operating state monitoring of an industrial motor, 500,000 sets of valid data were collected during the 3-month operation period, and 85 abnormal states were successfully identified, of which 78 were consistent with the actual fault records, verifying the reliability of the monitoring method. The average response time of the operating state determination result is 200 milliseconds, which meets the real-time monitoring requirements.

[0018] S102. By collecting the electromagnetic field strength data generated by various operating equipment in the laboratory, the frequency characteristics of different interference sources, the time-varying characteristics of electromagnetic signals, and the spatial distribution relationship information of electronic equipment, a probability distribution model for distinguishing and identifying multi-source interference is constructed, and the characteristics and spatial positioning of each interference source in the laboratory electromagnetic environment are characterized.

[0019] The original electromagnetic field data collected by the distributed electromagnetic sensor array is obtained, and Fourier transform is performed on the original electromagnetic field data to obtain spectrum feature data; characteristic frequency components of the interference source are extracted according to the spectrum feature data to obtain first-category feature data, and independent component analysis is performed on the first-category feature data to obtain independent signal source data; a mixed probability density function with multiple Gaussian components is constructed according to the independent signal source data, and the Gaussian component parameters are updated through expectation maximization iterative calculation to obtain an interference source probability distribution model; each Gaussian component in the interference source probability distribution model is spatially located using a triangulation positioning method, and signal strength data is obtained from the electromagnetic sensor array for spatial interpolation processing to obtain electromagnetic field spatial distribution data; the electromagnetic field gradient of each sampling point is calculated according to the electromagnetic field spatial distribution data, and time domain mutation characteristics and periodic characteristics are extracted in combination with the independent signal source data to obtain time-varying characteristic data of the interference source.

[0020] Specifically, the original electromagnetic field data is collected synchronously at multiple spatial locations in the laboratory through a distributed electromagnetic sensor array, and the electromagnetic field intensity and phase information of each sampling point are recorded according to the preset sampling sequence. The original electromagnetic field data is Fourier transformed to obtain the spectrum feature data. The adaptive frequency threshold is set according to the spectrum feature data, the characteristic frequency component of the interference source is extracted to obtain the first type of feature data, and the first type of feature data of each sampling point is subjected to independent component analysis to obtain multiple independent signal source data. Based on the independent signal source data, a mixed probability density function with multiple Gaussian components is constructed, and the Gaussian component parameters are updated through expectation maximization iterative calculation to obtain the optimized probability distribution model of the interference source. The triangulation positioning method is used to spatially locate the interference source corresponding to each Gaussian component, and the signal strength data is obtained from multiple electromagnetic sensor array nodes, and the spatial interpolation processing is performed to obtain the electromagnetic field spatial distribution data. The electromagnetic field gradient of each sampling point is calculated according to the electromagnetic field spatial distribution data, and the time domain mutation characteristics and periodic characteristics are extracted in combination with the independent signal source data to obtain the time-varying feature data of the interference source. The feature space probability distribution of the time-varying feature data of the interference source is established by the kernel density estimation method, and it is integrated with the optimized probability distribution model of the interference source to achieve the characterization and spatial positioning of various interference sources. The electromagnetic sensor array adopts a 4×4 matrix layout, with a spacing of 30 cm between adjacent sensors, covering an area of ​​100 square meters in the laboratory. The sampling frequency of each sensor is set to 1000 Hz, the measurement frequency range is 10 Hz to 100 kHz, and the sensitivity reaches 0.1 mV / m. The original electromagnetic field data contains two dimensions: amplitude and phase. The spectrum characteristics are obtained through 1024-point fast Fourier transform, and the frequency resolution is 0.98 Hz. In the spectrum feature data, the peak detection method is used to extract the characteristic frequency components, and the dynamic threshold is set to 3 times the signal mean to identify the significant frequency peak. Taking a laboratory as an example, the main interference frequencies detected include the characteristic frequencies of 50 Hz power frequency and its harmonics, 20 kHz switching power supply, and 100 MHz digital device clock. Independent component analysis is performed on these characteristic frequency data, and the FastICA algorithm is used to separate 5 independent signal sources. When the mixed Gaussian model is initialized, the number of Gaussian components is set to 8, and each component contains two parameters: mean vector and covariance matrix. The parameters were iteratively optimized using the expectation maximization algorithm, with the convergence threshold set to 0.001 and the maximum number of iterations set to 100. During the optimization process, when the weight of a Gaussian component was less than 0.05, the component was automatically removed, and 6 valid Gaussian components were finally obtained. In triangulation positioning, the three sensor nodes with the highest signal strength were selected to construct the positioning equation, and the distance to the interference source was calculated based on the inverse attenuation model of signal strength. In the laboratory environment, the positioning accuracy reached 0.5 meters, and on this basis, spatial interpolation was performed to obtain electromagnetic field distribution data within a range of 10×10×3 meters, with a spatial resolution of 0.1 meters. The electromagnetic field gradient calculation uses the central difference method to calculate the rate of change of field strength at adjacent sampling points.Time domain feature extraction focuses on the change of signal amplitude. When the signal mutation amplitude exceeds 2 times the mean, it is marked as a mutation point, and the time interval that recurs in continuous sampling is used as a periodic feature. Obvious mutation features were observed during the startup of a desktop computer, with a peak value of 5 times the background field strength and accompanied by a pulse signal with a period of 50ms. The kernel density estimation uses the Gaussian kernel function, and the kernel width parameter is determined to be 0.15 through cross-validation. The feature space probability distribution fusion uses a Bayesian framework to give equal weight to time-varying features and spatial distribution features. In the laboratory scenario, different types of interference sources such as printers, air conditioners, and computers were successfully distinguished, with an interference source type recognition accuracy of 90% and a spatial positioning error of less than 0.8 meters. In the actual deployment of a research and development laboratory, electromagnetic interference generated by 15 different types of electronic equipment was monitored for a long time, and a complete interference source feature library was established to achieve rapid identification and positioning of new interference sources. The amount of data collected during the system operation exceeded 100GB, and the feature library contained more than 1,000 sets of typical interference features.

[0021] S103, integrating the standardized electromagnetic characteristic data of the electronic equipment, the real-time environment data of the sensor network and the operating status data of the electronic equipment, inputting them into the trained and optimized interference source probability distribution model, predicting and outputting the electromagnetic environment perception results, and visually displaying and analyzing the electromagnetic environment perception results.

[0022] The electromagnetic characteristic data, environmental data and state data are standardized by maximum and minimum values ​​according to preset normalization parameters to obtain an electromagnetic data set, an environmental data set and a state data set; waveform amplitude and phase characteristics are extracted from the electromagnetic data set, temperature, humidity and light parameters are extracted from the environmental data set, voltage and current operating parameters are extracted from the state data set, and feature vectors are established through corresponding timestamps; a weighted summation method is used to perform data fusion on the feature vectors, and a fused feature vector is obtained according to the electromagnetic characteristic weights, environmental parameter weights and state parameter weights; a mixed Gaussian probability density function is calculated through the fused feature vector, and the Gaussian component parameters are updated online. If the difference between the newly added data and the historical distribution exceeds a preset threshold, the Gaussian component parameter update is triggered; a heat map is constructed based on the electromagnetic environment distribution data, abnormal field strength areas are marked, and contour visualization charts are generated.

[0023] Specifically, the electromagnetic characteristic data of the equipment is normalized by the maximum and minimum values ​​according to the preset normalization parameters to obtain the electromagnetic data set, the real-time environmental data of the sensor network is normalized by the maximum and minimum values ​​to obtain the environmental data set, and the operating status data of the electronic equipment is normalized by the maximum and minimum values ​​to obtain the state data set. The waveform amplitude and phase characteristics are extracted from the electromagnetic data set, the temperature, humidity and light parameters are extracted from the environmental data set, and the voltage and current operating parameters are extracted from the state data set. The feature vector is established according to the corresponding timestamp. The weighted summation method is used to fuse the feature vector data, and the electromagnetic feature weight is set to 0.5, the environmental parameter weight is set to 0.3, and the state parameter weight is set to 0.2 to generate a fused feature vector. The fused feature vector is input into the mixed Gaussian probability density function, the probability distribution of the electromagnetic field intensity at each sampling point is calculated, and the Gaussian component parameters are optimized by the expectation maximization method. The Gaussian component parameters are updated online based on the stochastic gradient descent method, and the parameter update is triggered when the difference between the new data and the historical distribution exceeds the preset threshold. A spatial grid is established in a three-dimensional rectangular coordinate system, and the expected value and variance of the electromagnetic field intensity are calculated for each grid node to generate electromagnetic environment distribution data. According to the electromagnetic environment distribution data, a heat map is constructed, the field strength abnormal area is marked, the field strength gradient vector is calculated, and the contour visualization chart is generated. The data is normalized by the maximum and minimum value standardization method. The original field strength range of the electromagnetic characteristic data is 0-120mV / m and is mapped to the 0-1 interval. The temperature in the environmental data is mapped from -10 to 40 degrees Celsius to the 0-1 interval, the humidity is mapped from 30% to 90% to the 0-1 interval, and the voltage in the equipment operation state is mapped from 180V to 250V to the 0-1 interval, and the current is mapped from 0 to 50A to the 0-1 interval. Features are extracted from the standardized data. The electromagnetic waveform features include the peak-to-peak amplitude, the root mean square value, and the peak factor. The phase features include the phase angle and the phase stability. The environmental parameters extract the temperature change rate, the humidity change rate, and the light intensity change rate. The equipment operation status parameters extract the voltage fluctuation rate, the current fluctuation rate, and the power factor. All features are aligned according to the millisecond timestamp to form a 32-dimensional feature vector. Data fusion adopts the weighted summation method, and the weight distribution is set based on the domain rules. The higher weight of electromagnetic characteristics reflects the importance of electromagnetic environment monitoring, the environmental parameters are second to reflect the external influence, and the equipment status parameters have the lowest weight as an auxiliary judgment basis. In a certain laboratory scenario, the fused feature vector shows that the electromagnetic field intensity is positively correlated with temperature and negatively correlated with humidity. When the mixed Gaussian model is initialized, 6 Gaussian components are set, each of which contains a 32-dimensional mean vector and a 32×32 covariance matrix. The parameters are iteratively optimized by the expectation maximization algorithm, and the convergence threshold is set to 0.001 and the maximum number of iterations is 200. In a certain experimental scenario, 4 significant Gaussian components are retained after the model converges, corresponding to different working states.In online learning, a mini-batch method is used. Each time 100 sets of new data are collected, parameter updates are triggered. The initial value of the learning rate is set to 0.01, and it gradually decreases as the amount of data increases. When the KL divergence between the new data and the historical distribution exceeds 0.5, increasing the learning rate accelerates model adaptation. In the experiment, it was observed that the model parameters were updated quickly in scenarios such as equipment startup and load changes. The spatial grid division adopts a 0.5-meter resolution, and 960 grid nodes are generated within a 10×8×3-meter spatial range. Each node calculates the expected value and 95% confidence interval of the electromagnetic field intensity to generate spatial distribution data. The field strength in a certain area of ​​the laboratory was abnormal, and the expected value reached 5 times the background value, and the variance increased significantly. The visualization uses a red, yellow and blue three-color heat map, blue indicates that the field strength is less than 1mV / m, yellow indicates 1-5mV / m, and red indicates greater than 5mV / m. The field strength gradient is marked with arrows, and the length of the arrow is proportional to the gradient size. The contour line interval is set to 1mV / m to clearly display the field strength distribution contour. In an industrial field application, the thermal map visually displays the electromagnetic field distribution within 3 meters around the equipment and identifies two abnormal field strength points.

[0024] S104. Based on the predicted electromagnetic environment perception results, extract electromagnetic environment feature data, perform noise reduction and preprocessing on electromagnetic environment feature data of different dimensions, and perform weight allocation and optimization combination to obtain the interference source location coordinates through iterative calculation.

[0025] The electromagnetic field intensity characteristic data, phase difference characteristic data, frequency distribution characteristic data and signal attenuation characteristic data are obtained from the electromagnetic environment perception results, and the filtered characteristic data are obtained through wavelet denoising and median filtering. The characteristic covariance matrix is ​​calculated according to the filtered characteristic data, and the characteristic vector corresponding to the cumulative contribution rate exceeding the contribution rate threshold is selected for the characteristic covariance matrix to obtain the reduced-dimensional characteristic data. The initial weights are allocated to the reduced-dimensional characteristic data using the spatial electromagnetic propagation attenuation law, and the characteristic weights are iteratively optimized using the gradient descent method to obtain the optimal characteristic weight coefficient. The characteristic data weighted by the optimal characteristic weight coefficient are calculated to obtain the rough coordinates of the interference source, and a search space is constructed according to the rough coordinates, and the position optimization equation is solved using the Newton iteration method to obtain the precise coordinates of the interference source.

[0026] Specifically, the four-dimensional feature data of electromagnetic field strength, phase difference, frequency distribution, and signal attenuation are extracted from the electromagnetic environment perception results. Wavelet denoising is performed on the extracted data to obtain the first type of feature data, and the second type of feature data is obtained by eliminating outliers through median filtering. The feature covariance matrix is ​​calculated based on the second type of feature data, and the corresponding feature vectors whose cumulative contribution rate exceeds the preset threshold are selected. The second type of feature data is transformed into the third type of feature data by dimensionality reduction. The initial feature weight allocation scheme is constructed based on the attenuation law of spatial electromagnetic propagation, and the initial weights are assigned to the four dimensions of field strength, phase, frequency, and attenuation in the third type of feature data. The weight optimization objective function is established using the minimum mean square error criterion, and the feature weights are iteratively optimized using the gradient descent method to obtain the optimal feature weight coefficient. The third type of feature data is weighted and combined according to the optimal feature weight coefficient to construct an objective function based on the electromagnetic field strength distribution, and the particle swarm optimization algorithm is used to calculate the rough coordinates of the interference source position. The search space is constructed with the rough coordinates of the interference source position as the center, and the least squares position optimization equation is established based on the received signal strength of multiple sensor nodes in three-dimensional space. The Newton iteration method is used to solve the position optimization equation. When the position coordinate update amount is less than the preset threshold, the iteration is stopped and the precise position coordinates of the interference source are output. The electromagnetic environment perception results contain multi-dimensional feature data, in which the electromagnetic field strength range is 0-100mV / m, the phase difference range is 0-360 degrees, the frequency distribution includes 50Hz power frequency and its harmonics, and the signal attenuation follows the inverse square attenuation law. For these raw data, the db4 wavelet basis function is used for 4-layer decomposition and denoising to effectively suppress high-frequency random noise. The median filter uses a 5-point sliding window to successfully eliminate 95% of outliers, and the signal-to-noise ratio is improved by 8dB after processing. The characteristic covariance matrix calculation reflects the correlation between the features of each dimension. The field strength and phase correlation coefficient is 0.72, and the field strength and frequency correlation coefficient is 0.45. The cumulative contribution rate threshold is set to 0.85. Finally, three principal component eigenvectors are selected. After data dimension reduction, the original information volume is retained to 87%. The law of electromagnetic propagation attenuation in space shows that the field strength attenuation rate is inversely proportional to the square of the distance, and the phase difference increases linearly with the propagation distance. Based on this, the initial weight of the field strength feature is set to 0.4, the phase feature weight is 0.3, the frequency feature weight is 0.2, and the attenuation feature weight is 0.1. The minimum mean square error criterion is used to optimize the weight, and the learning rate is set to 0.01. After 200 iterations, the weight coefficient converges. In the particle swarm optimization algorithm, the number of particles is set to 50, the search space is set to 10×10×3 meters, and the particle speed is limited to 0.1-1 meters per second. Taking a laboratory scenario as an example, when 5 sensor nodes are arranged, the particle swarm optimization obtains the rough coordinates of the interference source position, and the positioning error is within 1 meter. The least squares position optimization adopts the Gauss-Newton iteration method, and the particle swarm optimization result is used as the initial value to construct a local search space of 3 meters × 3 meters × 1 meter.An overdetermined set of equations is established based on the signal strength measured by the five sensor nodes. During the iterative optimization process, convergence is determined when the coordinate update amount is less than 1 cm. In the laboratory verification, 100 repeated positioning tests were conducted on the fixed-position interference source, with an average positioning error of 0.15 meters and a maximum error of no more than 0.3 meters. Taking an industrial field application as an example, eight sensor nodes are arranged in a 100-square-meter workshop to locate three mobile interference sources in real time. The sampling rate of the original electromagnetic field data is 1000Hz. Through noise reduction, dimension reduction and optimization of the data processing chain, the millisecond-level tracking of the interference source position is finally achieved. The positioning results show that the positioning accuracy of the static interference source is better than 0.2 meters, and the positioning accuracy of the moving interference source (speed less than 1 meter / second) is better than 0.5 meters. This method has good resolution for multi-source interference in complex electromagnetic environments. When different types of interference sources such as power frequency equipment, switching power supplies, and inverters coexist, the accuracy rate reaches 92%. During the operation of the system, the average delay of data processing is less than 50 milliseconds, which meets the requirements of real-time monitoring and positioning.

[0027] S105. According to the location coordinates of the interference source, the array of the sensor network is evaluated in different regions, and the areas where the detection coverage or signal quality is lower than the preset requirements are identified. The sensor position and orientation within the preset range of the identified area are adjusted. At the same time, the sensitivity parameters of the sensor are adjusted according to the electromagnetic environment perception results. During the adjustment process, the detection effect of each area is evaluated and analyzed, and the sensor array layout and parameters are optimized.

[0028] A grid evaluation area is constructed according to the location coordinates of the interference source, and the regional signal coverage is calculated using the sensor node detection radius. The area to be optimized is marked for the area where the signal coverage is lower than the coverage threshold; the signal reliability value is calculated based on the ratio of the collected signal of the sensor node to the background noise in the area to be optimized, and the sensor sensitivity adjustment parameters are generated according to the signal reliability value; the coverage optimization function and the signal strength optimization function are constructed using the sensitivity adjustment parameters, and the sensor node location coordinates and direction angles in the area to be optimized are optimized and calculated to obtain the sensor layout optimization parameters; a sensor node configuration database is established according to the sensor layout optimization parameters, and update instructions are sent to the sensor nodes through the database content to obtain the optimized sensor array configuration.

[0029] Specifically, a grid evaluation area is constructed based on the location coordinates of the interference source, and the regional signal coverage is calculated according to the detection radius of each sensor node. The area with signal coverage lower than the preset threshold is marked as the area to be optimized, and the signal strength calculator is used to measure the electromagnetic field intensity distribution data of the area to be optimized. The acquisition performance of the sensor nodes in the area to be optimized is measured, and the signal reliability value is obtained by calculating the ratio of the sensor received signal to the background noise. The sensor sensitivity adjustment parameter is generated according to the signal reliability value. The coverage optimization function and the signal strength optimization function are constructed by the optimization algorithm. The location coordinates and direction angles of the sensor nodes in the area to be optimized are jointly optimized and calculated to generate the sensor layout optimization parameters. Based on the layout optimization parameters, a sensor node configuration database is established to record the sensor location coordinates, direction angles, and sensitivity parameters, and a sensor parameter optimization record table is constructed. The data fusion method is used to perform online quality evaluation on the sensor acquisition data, calculate the regional signal coverage and field strength detection accuracy, and generate sensor performance evaluation data. The optimization results are judged according to the sensor performance evaluation data. When the regional signal coverage and field strength detection accuracy meet the preset threshold, the sensor configuration data is written into the parameter configuration database. The content of the parameter configuration database is read by the feedback controller, and update instructions are issued to each sensor node to complete the sensor array parameter configuration. The grid evaluation area is divided into 0.5-meter intervals, forming 400 evaluation grid points in a 100-square-meter laboratory space. The sensor node detection radius is set to 3 meters, and the signal coverage distribution map is obtained through spatial superposition calculation. When the coverage is less than 70%, it is marked as an area to be optimized. In a certain laboratory scene, three areas to be optimized are identified, with a total area of ​​about 15 square meters, mainly distributed in corners and equipment-intensive areas. The sensor node performance measurement adopts the signal-to-noise ratio evaluation method. Under the condition of background noise of 0.1mV / m, the effective signal strength range is measured to be 0.5-10mV / m. The signal reliability value is calculated in sections, and the signal-to-noise ratio greater than 20dB is recorded as 1.0, 10-20dB is recorded as 0.8, and 5-10dB is recorded as 0.5. For low-reliability areas, the sensor sensitivity parameter is increased by 20% on the original basis, and the signal sampling time is increased. In the joint optimization calculation, the weight of the coverage optimization function is set to 0.6, and the weight of the signal strength optimization function is set to 0.4. The position optimization range is limited to a radius of 1 meter from the original position, and the direction angle adjustment range is plus or minus 30 degrees. The optimization results show that by adjusting the position of two sensors and the direction angle of three sensors, the average coverage of the area to be optimized is increased to 85%. The parameter configuration database uses a key-value pair storage structure to record parameters such as sensor identification number, spatial coordinates, direction angle, gain coefficient, etc. The database supports incremental updates, and only the changed parameters are updated after each optimization. In an industrial field application, the database capacity reaches 100MB, including parameter optimization records within 1 year.The online quality assessment adopts the sliding window method, and the window length is set to 1 hour. The average coverage and detection accuracy are calculated. The coverage threshold is set to 80%, and the accuracy threshold is set to 90%. When the evaluation index exceeds the threshold for 4 hours, the parameter configuration update process is triggered. Experimental data show that the optimized sensor array maintains a stable working state for 95% of the time. The parameter configuration update adopts a distributed collaborative mechanism, first sending a pre-configuration instruction to the target sensor, and then executing the formal update after receiving confirmation. A double buffer mechanism is used during the update process to ensure data continuity. In a certain field application, the total parameter update time of 12 sensor nodes does not exceed 10 seconds, and data collection is uninterrupted during this period. The feedback control adopts a proportional-integral algorithm to adjust the sensor parameters according to the coverage deviation and cumulative error. The control cycle is set to 1 minute, the proportional coefficient is 0.8, and the integration time is 10 minutes. In actual operation, the system's response time to environmental changes is less than 5 minutes, and the parameter adjustment process is smooth and oscillating. Statistics show that the average annual number of failures of the optimized sensor network in industrial field applications is reduced by 60%, the maintenance cost is reduced by 50%, and the detection accuracy is always maintained above 95%.

[0030] The position and orientation information of the sensors within the preset range in the identification area is obtained, and several groups of candidate sensor adjustment schemes are generated. For the candidate adjustment schemes, the detection coverage and signal quality of the adjusted sensors are simulated and calculated, and compared with the preset optimization goals, and the target adjustment scheme is screened out. If the target adjustment scheme meets the preset optimization goals, the sensor position and orientation adjustment parameters in the scheme are sent to the corresponding sensors.

[0031] A spatial locator is used to read the three-dimensional coordinates and direction angle data of the sensor node, and the detection coverage distribution data is obtained according to the sensor detection radius; the sensor position adjustment range and the direction adjustment angle are set according to the detection coverage distribution data, and the sensor adjustment candidate scheme is obtained through the Monte Carlo sampling method; the signal coverage rate is calculated for the sensor adjustment candidate scheme, and the signal coverage rate score and the regional signal quality score of the detection area are calculated by the numerical integration method, and the optimization objective function is constructed according to the preset weight coefficient; the sensor adjustment candidate scheme is iteratively optimized, and if the optimization objective function value exceeds the function value threshold, the adjustment instruction of the sensor position coordinate and the direction angle is generated.

[0032] Specifically, a spatial locator is used to read the three-dimensional coordinates and direction angle data of sensor nodes within a preset range, and detection coverage distribution data is generated according to the sensor detection radius. The signal strength calculator is used to measure and record the signal quality distribution data of the detection area. Based on the detection coverage distribution data, the sensor position adjustment range and direction adjustment angle are set, and several groups of sensor adjustment candidate solutions are generated through the Monte Carlo sampling method. A signal coverage calculation model is established for each group of candidate solutions, and the signal coverage score of the detection area is calculated by the numerical integration method, and the regional signal quality score is calculated according to the signal strength threshold. A weight coefficient of 0.6 is assigned to the signal coverage score of the detection area, and a weight coefficient of 0.4 is assigned to the regional signal quality score, and the candidate solution optimization objective function is constructed. The particle swarm optimization algorithm is used to iteratively optimize the candidate solutions, and the position constraints and direction constraints are set. The iteration is stopped when the optimization objective function value exceeds the preset threshold. The optimized sensor position coordinates and direction angle data are written into the adjustment instruction, and the parameter adjustment instruction is sent to the corresponding sensor through the communication network. The online monitoring method is used to track the sensor adjustment execution process, record the sensor position and direction adjustment data, and generate the execution process record. The actual adjusted detection coverage and signal quality indicators are calculated based on the execution process records. When the calculation results meet the preset optimization goals, the sensor adjustment plan is executed. The spatial locator adopts the principle of ultrasonic ranging. Four positioning base stations are arranged in the laboratory with a measurement accuracy better than 5 mm. The initial layout of the sensor nodes adopts a 4×4 matrix layout, with a distance of 2 meters between adjacent sensors and a detection radius of 3 meters to form a regional coverage grid. Through signal strength calculation, it is found that the signal strength of 78% of the positions in the coverage area is greater than 1mV / m, and 22% of the positions are below the threshold. The sensor position adjustment range is limited to a 1.5-meter radius sphere from the origin, and the direction angle adjustment range is positive and negative 45 degrees in the horizontal direction and positive and negative 30 degrees in the vertical direction. 1000 Monte Carlo samplings are used to generate adjustment candidate solutions within the constraint range. The position coordinates of each sampling point are generated by three-dimensional Gaussian distribution, and the direction angle is generated by uniform distribution. After eliminating the solutions that do not meet the constraints, 856 valid candidate solutions are retained. The signal coverage calculation adopts a grid division method to divide the detection area into 0.5 m × 0.5 m grids, and calculate whether the received signal of each grid point exceeds the detection threshold. The signal quality score is calculated based on the signal-to-noise ratio, mapping the ratio of the measured signal strength to the background noise to the range of 0-1. In an industrial environment, the average background noise is 0.2mV / m, the effective signal threshold is set to 1mV / m, and a signal-to-noise ratio greater than 14dB is recorded as a full score. The optimization objective function comprehensively considers coverage and signal quality. The higher weight of coverage reflects the importance of spatial coverage, and the lower weight of signal quality is used as an auxiliary judgment basis. Particle swarm optimization sets the number of particles to 50 and the maximum number of iterations to 200. The calculation stops when the objective function value exceeds 0.85 or the improvement is less than 0.001 for 20 consecutive iterations.The parameter adjustment instruction adopts a standardized data format, including the sensor identification number, target position coordinates, target direction angle, and adjustment time window. The communication network adopts a star topology structure, and reliable instruction transmission is achieved through industrial Ethernet, with an average transmission delay of less than 10 milliseconds. Online monitoring uses real-time data stream processing to collect sensor position and direction data every 50 milliseconds and record the adjustment trajectory. The execution process record includes the adjustment start and end time, position change curve, direction change curve, and signal quality change curve. In a laboratory scenario, the sensor adjustment takes an average of 45 seconds, and the adjustment accuracy is better than 1 cm. Based on the execution effect evaluation data, after the adjustment, the signal strength of 92% of the positions in the detection area exceeds the threshold, and the average signal-to-noise ratio is improved by 6dB. By continuously monitoring the sensor network for 48 hours, the stability of the adjustment scheme is verified, and the fluctuation range of signal coverage and quality indicators is controlled within 5%. In practical applications, this method has good adaptability to environmental changes, and can quickly complete sensor optimization and adjustment in scenarios such as changes in equipment layout and the emergence of electromagnetic interference sources.

[0033] S106. Apply the optimized sensor array layout and parameter settings to the test process of the electronic device to obtain the performance parameters of the electronic device under different electromagnetic environments, the performance parameters including electromagnetic compatibility and anti-interference capability, and compare and calibrate the test results with preset standard values.

[0034] Electromagnetic field strength data is collected through a sensor array, and electronic equipment performance parameters are recorded based on the electromagnetic field strength data, wherein the performance parameters include output voltage, output current and output power; the electromagnetic field strength data is subjected to Fourier transform to obtain frequency domain feature data, wherein the frequency domain feature data characterizes the electromagnetic field distribution characteristics; electromagnetic compatibility indicators are extracted based on the equipment performance parameters to obtain an electromagnetic compatibility feature vector, wherein the electromagnetic compatibility indicator includes a spurious emission level, a conducted interference level and a radiation sensitivity; a mapping relationship between the electromagnetic compatibility feature vector and an anti-interference feature vector is established to obtain a prediction result of equipment performance change trend; performance deviation is calculated based on the prediction result and a preset standard value, and compensation adjustment is performed on the sensor array sensitivity parameter.

[0035] Specifically, an electromagnetic field test space is constructed according to the optimized sensor array layout, and environmental electromagnetic field data is collected from sensor nodes. The performance parameters such as output voltage, current, and power of electronic equipment under different electromagnetic intensities are recorded by a data collector, and the collected data are recorded and marked in time series. The performance parameters are collected by a variable sampling rate method, and the sampling frequency is increased in the interval of drastic signal changes. The collected data is subjected to wavelet denoising to obtain equipment operation data, and the electromagnetic field intensity data is subjected to Fourier transform to obtain frequency domain feature data. Electromagnetic compatibility indicators are extracted from the equipment operation data, including spurious emission level, conducted interference level, and radiation sensitivity, and an electromagnetic compatibility feature vector is generated. The electromagnetic interference intensity distribution is calculated based on the frequency domain feature data, and the signal detector is used to measure the degree of influence of external electromagnetic interference on equipment performance to generate an anti-interference feature vector. The support vector regression method is used to establish the mapping relationship between the electromagnetic compatibility feature vector and the anti-interference feature vector to predict the performance change trend of the equipment under different electromagnetic environments. The performance deviation is calculated based on the predicted results and the preset standard value, and a calibration compensation function is constructed to compensate and adjust the sensitivity parameters of the sensor array. The test results are corrected by the calibration compensation function, and the calibration is completed when the error between the corrected performance parameter and the standard value is less than the preset threshold. The electromagnetic field test space adopts an octahedral structure with a side length of 6 meters. The sensor nodes are arranged at eight vertices according to the principle of uniform spatial distribution to achieve all-round electromagnetic field monitoring. Performance parameter collection includes output voltage range 0-380V, current range 0-100A, power factor range 0-1, and sampling resolution of 16 bits. In the test of a certain industrial equipment, the voltage fluctuation amplitude was recorded to be 0.5V and the current fluctuation amplitude was 0.2A. In the variable sampling rate method, the basic sampling frequency is set to 1kHz, and it is automatically increased to 10kHz when the signal change rate exceeds 10%, so as to accurately capture the transient process. Wavelet denoising uses db4 wavelet basis function and 4-layer decomposition to successfully suppress 95% random noise. Fourier transform uses 1024-point FFT with a frequency resolution better than 1Hz, and the operating characteristic frequency of the equipment is identified in the range of 0-100kHz. In the extraction of electromagnetic compatibility indicators, the spurious emission level measurement range is 30MHz-1GHz, and the typical value is less than 30dBμV / m. The conducted interference level is measured in the 150kHz-30MHz frequency band, with a typical value of less than 60dBμV. The radiated sensitivity test field strength is set to 3V / m, with a frequency step of 20MHz, and the performance changes of the equipment are recorded. The feature vector contains 50 dimensions, covering the interference characteristics of key frequency bands. The electromagnetic interference intensity distribution calculation adopts the spatial interpolation method with a grid resolution of 0.2 meters to generate a three-dimensional field strength distribution map. The signal detector sensitivity is 0.1mV / m. Under the working frequency condition, the background interference is 3mV / m, the switching power supply interference is 8mV / m, and the inverter interference is 15mV / m. The anti-interference feature vector records the response characteristics of the equipment under the action of different interference sources. The support vector regression uses the radial basis kernel function, and the kernel parameters are optimized by cross-validation.The training data contains 1,000 sets of performance test records, covering typical working conditions and extreme working conditions. The prediction results show that the equipment performance indicators show a nonlinear relationship with the electromagnetic environment, and the performance degradation rate is accelerated in a strong interference environment. The calibration compensation function uses piecewise linear interpolation, and the compensation coefficient is set at the key point. The sensor sensitivity adjustment range is 80%-120% of the nominal value, and the step accuracy is 1%. In a certain test scenario, the original measurement error reached 15%, which was reduced to less than 3% after compensation. The calibration process adopts an iterative method, and the calibration is completed when the error of the measurement results for three consecutive times is less than 5%. Practical application verification shows that this test method has achieved good results on different types of equipment. The industrial inverter was tested for 72 hours continuously, and 12 electromagnetic interference anomalies were successfully identified, accurately predicting the performance change trend of the equipment. The repeatability error of the test results was controlled within 2%, meeting the metrological certification requirements.

[0036] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. In addition, the various different embodiments of the present invention can also be combined arbitrarily, as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A method for identifying and locating electromagnetic interference sources based on multi-band EMI, characterized in that: The method comprises: Preprocess and standardize the acquired electromagnetic characteristic data, real-time environmental data of the sensor network, and operating status data of the electronic equipment; By collecting the electromagnetic field strength data generated by various operating equipment in the laboratory, the frequency characteristics of different interference sources, the time-varying characteristics of electromagnetic signals, and the spatial distribution relationship information of electronic equipment, a probability distribution model of interference sources is constructed to distinguish and identify multi-source interference, and the characteristics and spatial positioning of each interference source in the laboratory electromagnetic environment are characterized; The standardized electromagnetic characteristic data of electronic equipment, the real-time environmental data of the sensor network and the operating status data of the electronic equipment are integrated and input into the trained and optimized interference source probability distribution model to predict and output the electromagnetic environment perception results, and visualize and analyze the electromagnetic environment perception results; Based on the predicted electromagnetic environment perception results, the electromagnetic environment feature data is extracted, the electromagnetic environment feature data of different dimensions are denoised and preprocessed, and weight allocation and optimization combination are performed, and the interference source location coordinates are obtained through iterative calculation; According to the location coordinates of the interference source, the array of the sensor network is evaluated in different regions, and the areas where the detection coverage or signal quality is lower than the preset requirements are identified. The sensor position and orientation within the preset range of the identification area are adjusted. At the same time, the sensitivity parameters of the sensor are adjusted according to the electromagnetic environment perception results. During the adjustment process, the detection effect of each area is evaluated and analyzed, and the sensor array layout and parameters are optimized; The optimized sensor array layout and parameter settings are applied to the testing process of electronic equipment to obtain the performance parameters of electronic equipment in different electromagnetic environments. The performance parameters include electromagnetic compatibility and anti-interference ability, and the test results are compared and calibrated with preset standard values.

2. The method according to claim 1, characterized in that The obtained electromagnetic characteristic data, the real-time environment data of the sensor network and the operation status data of the electronic device are preprocessed and standardized, including: Acquire electromagnetic characteristic data of electronic equipment from multiple electromagnetic sensor nodes, wherein the electromagnetic characteristic data records timing information at a preset sampling frequency; Performing wavelet denoising on the electromagnetic characteristic data to obtain first-category processed data, and performing mean filtering on environmental temperature and humidity data and light intensity data collected through a distributed sensor network to obtain second-category processed data; Performing time series alignment on the first type of processed data and the second type of processed data through timestamps, and calculating the correlation between the first type of processed data and the second type of processed data to obtain an abnormal data set; The abnormal data set is segmented using a fixed-length sliding time window, the segmented data is clustered to obtain a device state feature vector, and an operation state correspondence database is established according to the device state feature vector.

3. The method according to claim 1, characterized in that The method collects electromagnetic field strength data generated by various operating equipment in the laboratory, frequency characteristics of different interference sources, time-varying characteristics of electromagnetic signals, and spatial distribution relationship information of electronic equipment to build an interference source probability distribution model for distinguishing and identifying multi-source interference, and characterizes and spatially locates each interference source in the laboratory electromagnetic environment, including: Acquire original electromagnetic field data collected by a distributed electromagnetic sensor array, and perform Fourier transform on the original electromagnetic field data to obtain frequency spectrum feature data; Extracting interference source characteristic frequency components according to the frequency spectrum characteristic data to obtain first type characteristic data, and performing independent component analysis on the first type characteristic data to obtain independent signal source data; Constructing a mixed probability density function with multiple Gaussian components according to the independent signal source data, and updating Gaussian component parameters by expectation maximization iterative calculation to obtain an interference source probability distribution model; Using a triangulation positioning method to spatially locate each Gaussian component in the interference source probability distribution model, obtaining signal strength data from an electromagnetic sensor array and performing spatial interpolation processing to obtain electromagnetic field spatial distribution data; The electromagnetic field gradient of each sampling point is calculated according to the electromagnetic field spatial distribution data, and the time domain mutation characteristics and periodic characteristics are extracted in combination with the independent signal source data to obtain the time-varying characteristic data of the interference source.

4. The method according to claim 1, characterized in that: The standardized electromagnetic characteristic data of the electronic equipment, the real-time environment data of the sensor network and the operation status data of the electronic equipment are integrated and input into the trained and optimized interference source probability distribution model, the electromagnetic environment perception results are predicted and output, and the electromagnetic environment perception results are visualized and analyzed, including: Performing maximum and minimum value normalization processing on the electromagnetic characteristic data, environmental data and state data according to preset normalization parameters to obtain an electromagnetic data set, an environmental data set and a state data set; Extracting waveform amplitude and phase features from the electromagnetic data set, extracting temperature, humidity and light parameters from the environmental data set, extracting voltage and current operating parameters from the state data set, and establishing feature vectors through corresponding timestamps; The characteristic vectors are fused by using a weighted summation method, and a fused characteristic vector is obtained according to the electromagnetic characteristic weights, the environmental parameter weights and the state parameter weights; The mixed Gaussian probability density function is calculated by the fused feature vector, and the Gaussian component parameters are updated online. If the difference between the new data and the historical distribution exceeds a preset threshold, the Gaussian component parameter update is triggered; Construct a heat map based on the electromagnetic environment distribution data, mark areas with abnormal field strength, and generate contour visualization charts.

5. The method according to claim 1, characterized in that The electromagnetic environment perception result based on the prediction is used to extract the electromagnetic environment feature data, perform noise reduction and preprocessing on the electromagnetic environment feature data of different dimensions, and perform weight allocation and optimization combination, and obtain the interference source position coordinates through iterative calculation, including: Obtain electromagnetic field intensity characteristic data, phase difference characteristic data, frequency distribution characteristic data and signal attenuation characteristic data from the electromagnetic environment perception results, and obtain filtered characteristic data through wavelet denoising and median filtering; Calculate a feature covariance matrix according to the filtered feature data, and select a feature vector corresponding to a cumulative contribution rate exceeding a contribution rate threshold value for the feature covariance matrix to obtain reduced-dimensional feature data; For the dimension-reduced feature data, initial weights are assigned using the spatial electromagnetic propagation attenuation law, and the feature weights are iteratively optimized using the gradient descent method to obtain the optimal feature weight coefficient; The feature data weighted by the optimal feature weight coefficient is calculated to obtain the rough coordinates of the interference source, a search space is constructed according to the rough coordinates, and the position optimization equation is solved by using the Newton iteration method to obtain the precise coordinates of the interference source.

6. The method according to claim 1, characterized in that According to the interference source location coordinates, the array of the sensor network is evaluated in different regions, the detection coverage or signal quality of the area is identified, and the sensor position and orientation within the preset range of the identification area are adjusted. At the same time, the sensitivity parameters of the sensor are adjusted according to the electromagnetic environment perception results. During the adjustment process, the detection effect of each area is evaluated and analyzed, and the sensor array layout and parameters are optimized, including: Construct a gridded evaluation area according to the location coordinates of the interference source, calculate the regional signal coverage using the sensor node detection radius, and mark the area to be optimized for the area where the signal coverage is lower than the coverage threshold; The signal reliability value is calculated based on the ratio of the collected signal of the sensor node in the area to be optimized to the background noise, and the sensor sensitivity adjustment parameter is generated according to the signal reliability value; The sensitivity adjustment parameters are used to construct a coverage optimization function and a signal strength optimization function, and the sensor node position coordinates and direction angles in the optimized area are optimized and calculated to obtain the sensor layout optimization parameters; Establishing a sensor node configuration database according to the sensor layout optimization parameters, and sending update instructions to the sensor nodes through the database content to obtain the optimized sensor array configuration; The method also includes: obtaining the position and orientation information of the sensor within a preset range in the identification area, generating several groups of candidate sensor adjustment schemes, simulating and calculating the detection coverage and signal quality of the adjusted sensor for the candidate adjustment schemes, and comparing them with the preset optimization targets, screening out the target adjustment scheme, and if the target adjustment scheme meets the preset optimization targets, sending the sensor position and orientation adjustment parameters in the scheme to the corresponding sensors.

7. The method according to claim 6, characterized in that The obtaining of the position and orientation information of the sensor within the preset range of the identification area, generating a plurality of groups of candidate sensor adjustment schemes, simulating and calculating the detection coverage and signal quality of the adjusted sensor for the candidate adjustment schemes, and comparing them with the preset optimization target, screening out the target adjustment scheme, and if the target adjustment scheme meets the preset optimization target, sending the sensor position and orientation adjustment parameters in the scheme to the corresponding sensor, including: The spatial locator is used to read the three-dimensional coordinates and direction angle data of the sensor node, and the detection coverage distribution data is obtained according to the sensor detection radius; Setting a sensor position adjustment range and a direction adjustment angle according to the detection coverage distribution data, and obtaining a sensor adjustment candidate solution by a Monte Carlo sampling method; Calculate the signal coverage of the candidate sensor adjustment scheme, calculate the detection area signal coverage score and the area signal quality score using a numerical integration method, and construct an optimization objective function according to a preset weight coefficient; The candidate sensor adjustment solutions are iteratively optimized, and if the optimization objective function value exceeds a function value threshold, an adjustment instruction for the sensor position coordinates and direction angle is generated.

8. The method according to claim 1, characterized in that The optimized sensor array layout and parameter settings are applied to the test process of the electronic device to obtain the performance parameters of the electronic device under different electromagnetic environments, the performance parameters including electromagnetic compatibility and anti-interference ability, and the test results are compared and calibrated with the preset standard values, including: Collecting electromagnetic field strength data through a sensor array, and recording performance parameters of the electronic device according to the electromagnetic field strength data, wherein the performance parameters include output voltage, output current and output power; Performing Fourier transformation on the electromagnetic field intensity data to obtain frequency domain characteristic data, wherein the frequency domain characteristic data represents the electromagnetic field distribution characteristics; Extracting electromagnetic compatibility indicators according to the equipment performance parameters to obtain electromagnetic compatibility feature vectors, wherein the electromagnetic compatibility indicators include spurious emission levels, conducted interference levels, and radiation sensitivity; Establishing a mapping relationship between the electromagnetic compatibility feature vector and the anti-interference feature vector to obtain a prediction result of the equipment performance change trend; The performance deviation is calculated based on the predicted results and the preset standard value, and the sensitivity parameters of the sensor array are compensated and adjusted.

Images