Testing system for testing transformer electromagnetic shielding coupler based on intelligent sensing

Abstract

The invention discloses a system for testing an electromagnetic shielding coupler of a testing transformer based on intelligent sensing, and relates to the technical field of electrical equipment testing, and the system comprises a flexible distributed intelligent sensing array which is composed of a plurality of high-density miniature electromagnetic sensors and a temperature-vibration composite sensing unit, a non-uniform topological structure is embedded into a seam and an insulating interface area on the surface of the electromagnetic shielding coupler. According to the test system of the test transformer electromagnetic shielding coupler based on intelligent sensing, through dynamic electromagnetic field reconstruction of the flexible distributed intelligent sensing array and multi-band interference coupling modeling, the limitation of a traditional static test is broken through; and real-time capture and multi-dimensional decoupling of shielding effectiveness attenuation characteristics in a complex alternating electromagnetic environment are realized. Compared with a traditional method, the method has the advantages that the shielding effectiveness dynamic quantization error is reduced, and the engineering guidance value of test data is improved.

Description

Technical Field

[0001] The invention relates to the technical field of electrical equipment testing, in particular to a testing system for a test transformer electromagnetic shielding coupler based on intelligent sensing. Background Art

[0002] In the field of performance testing of electromagnetic shielding couplers for test transformers, traditional testing methods have long relied on basic parameter measurements in a single frequency or static electromagnetic environment, making it difficult to truly reflect the impact of multi-band composite electromagnetic interference on shielding couplers in complex operating conditions. With the development of intelligent power systems, the electromagnetic environment faced by test transformers is becoming increasingly complex, with frequent overlapping interference signals such as high-frequency harmonics, transient pulses, and power-frequency magnetic fields. However, due to the limitations of sensing technology and analysis methods, existing test systems still use a static test mode with fixed-frequency excitation and discrete sampling, resulting in an inability to accurately capture the dynamic attenuation characteristics of shielding effectiveness. For example, during actual operation, the leakage flux distribution of electromagnetic shielding couplers will undergo nonlinear changes due to factors such as equipment temperature rise and mechanical vibration. However, traditional testing methods can only obtain local static data at specific moments. They cannot track the phase differences of electromagnetic fields at different spatial locations in real time, nor can they quantitatively evaluate the time-varying characteristics of shielding effectiveness under the superposition of multi-band interference. Furthermore, existing techniques for analyzing electromagnetic signals are often limited to independent dimensions in the time or frequency domains, lacking research on the correlation between interference waveform distortion and spectral energy distribution. This results in an inability to accurately identify critical frequency band thresholds that lead to shielding failure and the cumulative effects of pulse interference. These deficiencies lead to significant deviations between test results and actual operating conditions, severely restricting the reliability design and optimization of electromagnetic shielding couplers. Summary of the Invention

[0003] (1) Technical problems solved

[0004] In response to the shortcomings of the existing technology, the present invention provides a test system for the electromagnetic shielding coupler of a test transformer based on intelligent sensing, which solves the problem of how to achieve multi-dimensional real-time quantitative evaluation of the dynamic coupling efficiency of the shielding coupler in a complex alternating electromagnetic environment.

[0005] (2) Technical solution

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A test system for testing electromagnetic shielding couplers of test transformers based on intelligent sensing, comprising:

[0007] A flexible distributed intelligent sensor array, comprised of multiple high-density micro-electromagnetic sensors and temperature-vibration composite sensing units, is embedded in a non-uniform topology within the joints and insulating interfaces of electromagnetic shielding couplers. These sensor units dynamically adjust the sampling frequency from 1kHz to 10MHz using an adaptive sampling chip, and achieve microsecond-level synchronized data acquisition based on a spatiotemporal synchronization trigger mechanism. Deployed within the joints and insulating interfaces of electromagnetic shielding couplers, the flexible distributed intelligent sensor array uses a combination of high-density micro-electromagnetic sensors and temperature-vibration composite sensing units to sense electromagnetic field distribution and environmental parameter changes in real time. The adaptive sampling mechanism embedded in the sensor nodes dynamically adjusts the acquisition frequency based on the electromagnetic field gradient. When transient pulse interference is detected, it automatically switches to high-frequency sampling mode to capture rapidly changing electromagnetic signals. For stable power-frequency magnetic fields, low-frequency sampling is used to reduce data redundancy. Each node receives a synchronized trigger signal via a fiber optic network, ensuring microsecond-level time alignment of multi-source data. The edge computing unit compresses and removes noise from the raw data to generate a three-dimensional electromagnetic field dynamic map containing spatial coordinates, time-domain waveforms, and frequency-domain energy.

[0008] a dynamic coupling model construction module, connected to the output end of the flexible distributed intelligent sensor array, including an asymmetric frequency domain convolution kernel and a dynamic attenuation factor matrix, for simulating the penetration path of multi-band interference signals in the shielding structure and correcting the material nonlinear parameters in combination with real-time temperature and vibration data;

[0009] A frequency-time fusion analysis engine, connected to the output end of the dynamic coupling model building module, includes an improved wavelet packet transform module and an adaptive convolutional neural network, and is used to establish a mapping relationship between time domain waveform distortion and frequency domain energy distribution, and identify critical frequency band thresholds;

[0010] The dual-modal verification module consists of a physical verification unit and a digital verification unit. The physical verification unit replicates a complex interference scenario in a standard electromagnetic interference chamber, while the digital verification unit performs reverse simulation based on a digital twin that couples electromagnetic, thermal, and mechanical multi-physics fields. Both units use a dual screening mechanism to verify the error threshold of the evaluation results, triggering adaptive resampling and model iteration. The dual-modal verification module replicates multi-band complex interference scenarios in the physical verification unit. Programmable interference sources dynamically match the measured electromagnetic environment. When leakage flux density exceeds the specified limit, the module automatically enhances the high-frequency component and superimposes transient pulses. During the real-time comparison between the measured data and the model predictions, deviations exceeding the specified limit trigger iterative adjustment of the interference parameters until they converge within the threshold range. The digital verification unit performs reverse simulation using the multi-physics coupled digital twin, mapping real-time temperature and vibration data into material parameter boundary conditions. Exceeding the specified temperature rise limit activates the magnetic permeability attenuation model, and anomalies in the vibration spectrum apply equivalent alternating stress. Phase gradient analysis is used to verify the topological consistency of the simulation results with theoretical calculations. When the angular deviation exceeds the specified limit, the conductivity parameters are reversely optimized and the dynamic attenuation factor matrix is simultaneously updated.

[0011] Preferably, each sensor node of the flexible distributed intelligent sensor array has a built-in edge computing unit, which uses pulse code multiplexing transmission technology to compress and frequency-domain sparsify the raw data to generate a three-dimensional electromagnetic field dynamic distribution data set containing time domain waveforms, frequency domain energy, and spatial coordinates. Each sensor node of the flexible distributed intelligent sensor array has a built-in edge computing unit, which performs localized pre-processing on the raw electromagnetic field signal during the data acquisition phase, including: when transient pulse interference is detected, the edge unit preferentially extracts the amplitude slope characteristics of the rising edge of the waveform, and distributes signals of different frequency bands to independent transmission channels through pulse code multiplexing technology, where high-frequency harmonics use differential coding mode to suppress common-mode interference, and the power frequency magnetic field signal uses Manchester coding to ensure clock synchronization.

[0012] Preferably, the dynamic attenuation factor matrix is based on real-time temperature and vibration data, and dynamically corrects the nonlinear changes of material magnetic permeability and electrical conductivity through a parallel tensor decomposition algorithm, and outputs the shielding effectiveness attenuation curves of each frequency band under separate and synergistic effects.

[0013] Preferably, the improved wavelet packet transform module of the frequency-time fusion analysis engine decomposes the time domain waveform into multiple frequency band components and calculates the energy entropy value of each component. The adaptive convolutional neural network extracts the correlation between the time domain distortion characteristics and the frequency band energy distribution, and generates a frequency-time correlation matrix to quantify the coupling contribution of different interference modes. The frequency-time fusion analysis engine uses the improved wavelet packet transform to perform multi-layer decomposition of the time domain waveform, selects the optimal wavelet basis function according to different interference characteristics, uses high-order vanishing moment wavelets to capture steep fronts for transient pulses, and uses symmetric wavelets to separate harmonic components for power frequency distortion. The energy entropy value calculation of each frequency band after decomposition introduces a temperature vibration compensation factor, and adjusts the energy weight of the corresponding frequency band when the local temperature rises or the vibration exceeds the limit. The adaptive convolutional neural network extracts the key modes of time domain distortion through a pre-trained feature library, focuses on zero-crossing rate mutation detection for high-frequency harmonics, identifies the waveform inflection point characteristics for transient pulses, and outputs a frequency-time correlation matrix to quantify the contribution of the interference mode. The Markov chain-based prediction model analyzes historical attenuation trends, triggers parameter calibration when the instantaneous attenuation rate is abnormal, and verifies the prediction reliability in combination with the digital twin simulation results.

[0014] Preferably, the dual screening mechanism requires that the error between the measured data and the model prediction value of the shielding effectiveness evaluation results of any frequency band in the physical verification unit shall not exceed 5%, and the topological consistency error between the reverse simulation and the theoretical calculation in the digital verification unit shall not exceed 3%. If any error exceeds the limit, the adaptive resampling of the sensor node and the iterative update of the weights of the asymmetric frequency domain convolution kernel will be triggered.

[0015] Preferably, the parallelized tensor decomposition algorithm performs spatiotemporal separation on the superimposed disturbance sources of high-frequency harmonics, transient pulses and power-frequency magnetic fields, and predicts the maximum attenuation rate and critical failure time parameters of shielding effectiveness based on a Markov chain.

[0016] Preferably, the output end of the frequency-time correlation matrix is connected to a visualization evaluation module to generate a multi-dimensional test report including a heat map of electromagnetic field distribution, frequency band energy proportion and dynamic attenuation trend.

[0017] Preferably, the multi-physics field coupling solver of the digital verification unit updates the material thermal expansion coefficient boundary conditions of the digital twin according to the real-time temperature data, and updates the mechanical stress distribution parameters according to the vibration data, thereby realizing dynamic calibration of the electromagnetic field reverse simulation.

[0018] Preferably, the trigger signal of the adaptive sampling chip is synchronously transmitted to all sensor nodes via optical fiber, ensuring that the clock deviation of multi-node data collection is less than 0.1 μs.

[0019] Preferably, the dynamic coupling model construction module, the frequency-time fusion analysis engine and the dual-modal verification module are connected through a closed-loop feedback link to form an iterative optimization architecture for dynamic quantitative evaluation of shielding effectiveness. After receiving the pre-processed data, the dynamic coupling model construction module uses an asymmetric frequency domain convolution kernel to simulate the penetration path of interference signals in different frequency bands. High-frequency harmonics use narrowband filtering to track the skin effect. Transient pulses capture multiple reflection trajectories through a time-delay compensation algorithm. The power-frequency magnetic field calculates the leakage flux distribution based on the magnetoresistance model. The model introduces a dynamic attenuation factor matrix and establishes a nonlinear correction rule for material parameters based on real-time temperature and vibration data. When the temperature rise exceeds the threshold, the magnetic permeability weight is automatically reduced, and the conductivity correction amount is increased when the vibration exceeds the limit. The parallelized tensor decomposition algorithm performs time-space separation on the multi-band coupling effect, gives priority to the time-domain reflection energy of the transient pulse, and dynamically adjusts the weight coefficient of the high-frequency harmonics based on the working conditions, and outputs the shielding effectiveness attenuation curve under independent and synergistic effects of each frequency band.

[0020] The closed-loop feedback architecture dynamically optimizes model parameters based on dual verification results. Physical verification deviations prioritize adjustment of the magnetic permeability correction factor, while digital verification anomalies reset the sensor synchronization mechanism. The system performs global parameter updates at fixed intervals, shortening the feedback cycle and compressing the confidence interval under high-temperature and high-vibration conditions. The visualization module integrates and presents the electromagnetic field distribution, frequency band energy share, and attenuation trends. The heat map rendering accurately matches the sensor spatial topology, and critical frequency bands are marked with pulsed haloes and associated with the digital twin coordinates. Under extreme conditions, an anti-interference mode is activated to reduce computational complexity and apply signal smoothing to ensure accurate identification of key parameters. In the event of a data link interruption, virtual feedback is generated based on historical data to maintain system operation, and a fast resynchronization mechanism is used to restore and calibrate the system. Through a closed-loop process of perception, modeling, verification, and optimization, the entire system achieves multi-dimensional dynamic evaluation of electromagnetic shielding effectiveness and reliability design optimization.

[0021] (3) Beneficial effects

[0022] The present invention provides a test system for testing transformer electromagnetic shielding couplers based on intelligent sensing. It has the following beneficial effects:

[0023] (1) This intelligent sensor-based test system for electromagnetic shielding couplers of test transformers breaks through the limitations of traditional static testing by reconstructing dynamic electromagnetic fields and modeling multi-band interference coupling using a flexible distributed intelligent sensor array. It achieves real-time capture and multi-dimensional decoupling of shielding effectiveness attenuation characteristics in complex alternating electromagnetic environments. A frequency-time fusion analysis engine is used to establish a correlation mapping between time-domain distortion and frequency-domain energy, accurately identifying critical failure frequency band thresholds. A dual-modal verification mechanism is combined to ensure the self-consistency of the evaluation results. Compared to traditional methods, this system reduces the dynamic quantification error of shielding effectiveness and enhances the engineering guidance value of the test data.

[0024] (2) This intelligent sensor-based test system for electromagnetic shielding couplers of test transformers provides a dynamic optimization basis for the reliability design of electromagnetic shielding couplers through a closed-loop feedback architecture and digital twin reverse simulation technology, solving the pain points of operating condition simulation distortion and unclear failure mechanisms in traditional testing. Through the physical-digital dual-modal verification and visual evaluation module, it can quickly locate weak shielding areas and predict life degradation trends, thereby improving the efficiency of shielding structure optimization. At the same time, it supports adaptive switching of anti-interference modes under extreme working conditions, providing an innovative testing method for the safe operation and maintenance of smart grid equipment in complex electromagnetic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the overall framework of the present invention;

[0026] Figure 2 This is a control logic timing diagram of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] See also Figure 1 and Figure 2 The present invention provides a technical solution: a test system for electromagnetic shielding couplers of test transformers based on intelligent sensing, comprising:

[0029] The flexible distributed intelligent sensor array is composed of multiple high-density micro-electromagnetic sensors and temperature-vibration composite sensing units, which are embedded in the seams and insulating interface areas on the surface of the electromagnetic shielding coupler with a non-uniform topology structure. The sensor unit dynamically adjusts the sampling frequency from 1kHz to 10MHz through an adaptive sampling chip, and realizes microsecond-level synchronous data acquisition based on a time-space synchronization trigger mechanism; the flexible distributed intelligent sensor array uses a non-uniform topology layout strategy to embed high-density micro-electromagnetic sensors and temperature-vibration composite sensing units into areas with high incidence of electromagnetic leakage on the surface of the electromagnetic shielding coupler. Areas with high incidence of electromagnetic leakage include seams and insulating interfaces. Among them, the sensor density in the seam area is 2-3 times that of the insulating interface area to match the spatial distribution characteristics of the leakage magnetic flux.

[0030] Each sensor unit has a built-in adaptive sampling chip that monitors local electromagnetic field gradient changes in real time. When transient pulse interference is detected, the sampling frequency automatically increases to 10MHz to capture nanosecond-level waveform distortion. If stable interference dominated by power-frequency magnetic fields is detected, it switches to a low-frequency sampling mode of 1kHz to reduce data redundancy. All sensor nodes transmit synchronized trigger signals via optical fiber. When triggered by the rising edge of an external composite interference signal, the clock deviation of the multi-node data collection is ensured to be less than 0.1μs. Pulse code multiplexing technology is used to classify the raw data stream according to its frequency band characteristics and transmit it to the edge computing node.

[0031] The edge computing node performs frequency-domain sparsification on the raw data, removing the ambient noise floor signal, and reconstructs a dynamic electromagnetic field distribution map containing time-domain waveform amplitude, frequency-domain energy fraction, and three-dimensional spatial coordinates. When the temperature-vibration composite sensing unit detects a local temperature rise exceeding 50°C or a vibration acceleration exceeding 2g, the adaptive sampling chip activates overload protection mode, prioritizing electromagnetic field phase difference data in the joint area and triggering a coordinated compensation mechanism among adjacent nodes to prevent sensor data distortion caused by mechanical deformation.

[0032] The dynamic coupling model construction module is connected to the output end of the flexible distributed intelligent sensor array, including an asymmetric frequency domain convolution kernel and a dynamic attenuation factor matrix, which is used to simulate the penetration path of multi-band interference signals in the shielding structure and correct the nonlinear parameters of the material in combination with real-time temperature and vibration data; after the dynamic coupling model construction module receives the three-dimensional electromagnetic field dynamic distribution data output by the flexible distributed intelligent sensor array, it first uses the asymmetric frequency domain convolution kernel to analyze the superimposed penetration path of high-frequency harmonics, transient pulses and power frequency magnetic fields: for high-frequency harmonic interference, the convolution kernel uses narrowband filtering characteristics to extract its skin effect attenuation characteristics at the edge of the shielding structure; for transient pulse interference, the time delay compensation algorithm is used to track its multiple reflection paths at the joints; the power frequency magnetic field calculates its leakage flux distribution penetrating the insulation interface based on the spatial magnetoresistance model.

[0033] The dynamic attenuation factor matrix establishes nonlinear correction rules for material parameters based on real-time data from the temperature-vibration composite sensing unit. Specifically, when the local temperature rise exceeds 40°C, the magnetic permeability weight coefficient is reduced based on the Arrhenius accelerated aging model. If the vibration acceleration reaches 1.5g, the conductivity correction is dynamically updated based on the stress-hysteresis loop characteristics. The model prioritizes the coupled effects of multi-band interference: when transient pulses and high-frequency harmonics coexist, the local saturation effect of the shielding structure caused by the pulse leading edge is prioritized, followed by the cumulative effect of the skin depth of the high-frequency harmonics. When temperature rise and vibration are combined, a thermal-mechanical coupling iterative algorithm is used to first update the material parameter boundary conditions based on the temperature field, and then correct the equivalent magnetic resistance of the penetration path based on the vibration stress distribution.

[0034] In cases where sensor data conflicts with model predictions, such as when the measured leakage flux exceeds the theoretical value by 20%, the tensor decomposition algorithm's spatiotemporal recalibration mode is triggered, marking the anomalous data as a potential shielding defect area and initiating targeted retesting of high-density sensor nodes to eliminate the risk of misjudgment. The model automatically compares historical attenuation curves with real-time data trends every 10 seconds. If the prediction error exceeds 8% for three consecutive times, it is determined to be material parameter drift or structural deformation, triggering a global reweighting of the dynamic attenuation factor matrix and reshaping of the convolution kernel.

[0035] The frequency-time fusion analysis engine is connected to the output end of the dynamic coupling model construction module, including an improved wavelet packet transform module and an adaptive convolutional neural network, which are used to establish a mapping relationship between time domain waveform distortion and frequency domain energy distribution, and identify critical frequency band thresholds; after the frequency-time fusion analysis engine receives the multi-band penetration path data output by the dynamic coupling model, the improved wavelet packet transform module first adaptively divides the time domain waveform into frequency bands according to an eight-layer decomposition depth: for transient pulse signals with a duration of less than 10μs, a high-order vanishing moment wavelet basis function is used to capture its steep rising edge characteristics; for the power frequency magnetic field fundamental wave distortion signal, a symmetric wavelet basis is used to separate the odd and even harmonic components.

[0036] The energy entropy calculation for each frequency band incorporates a temperature-vibration compensation factor. Specifically, when the temperature sensor detects a local temperature rise exceeding 45°C, the energy weight coefficient for the corresponding frequency band is automatically increased by 20% to enhance thermal failure characteristics. If the vibration acceleration exceeds 1.2g, a pulse interference suppression algorithm is applied to the energy entropy of the frequency band in the insulation interface region. An adaptive convolutional neural network uses a pre-trained library of interference patterns to extract key distortion features from the time domain waveform. For periodic oscillation distortion caused by high-frequency harmonics, the network prioritizes identifying zero-crossing rate mutation points. For asymmetric clipping distortion caused by transient pulses, the network focuses on detecting extreme points in the waveform's second-order derivative.

[0037] When the network output layer constructs the frequency-time correlation matrix, a dynamic threshold decision mechanism is adopted. That is, when the energy proportion of a certain frequency band exceeds 15% of the total interference energy and the corresponding time domain distortion exceeds the preset threshold, the frequency band is marked as a potential critical failure band; if the energy entropy values of three adjacent frequency bands increase by more than 5% for three consecutive iterations, a cross-band coupling warning is triggered.

[0038] To address multi-feature conflicts under complex operating conditions, such as when the high-frequency energy ratio meets the target while the time-domain distortion remains within the limit, the network backpropagation process redistributes attention weights, strengthening the contribution of temperature-vibration data to the decision-making factors. When there is a spatial deviation between the model's predicted critical frequency band and the measured shielding failure location, the wavelet packet decomposition layer is automatically switched to the sixth layer for localized frequency band refinement analysis. A dynamic convolution kernel size adjustment mechanism is used to match the characteristic scales of shielding defects of varying sizes.

[0039] The analysis engine performs a full-band health assessment every 5 seconds. If the same band is marked as critical in all three assessments, it is determined to be a permanent shield failure and a band lock instruction is generated. The shielded coupler will automatically avoid this band in subsequent tests to protect the safety of the device.

[0040] The dual-modal verification module includes a physical verification unit and a digital verification unit. The physical verification unit reproduces the complex interference scenario in a standard electromagnetic interference chamber, and the digital verification unit performs reverse simulation based on the digital twin coupled with electromagnetic-thermal-mechanical multi-physical fields. Both use a dual screening mechanism to check the error threshold of the evaluation results, triggering adaptive resampling and model iteration.

[0041] In the physical verification unit, the dual-modal verification module uses a programmable interference source in a standard electromagnetic interference chamber to generate a multi-band composite interference signal that matches the real-time monitoring data of the flexible distributed intelligent sensor array. When the leakage flux density in the seam area of the tested shielded coupler exceeds 200μT, the module automatically boosts the high-frequency harmonic component to 1.5 times its original amplitude and superimposes a transient pulse sequence with a pulse width of 50ns to accurately replicate the electromagnetic stress concentration scenario in actual working conditions. The physical verification unit compares the measured shielding effectiveness attenuation rate with the predicted value of the dynamic coupling model in real time. If the deviation exceeds 5% for three consecutive measurements within the 2MHz-10MHz frequency band, the interference parameter adaptive adjustment mechanism is triggered. Specifically, the power frequency magnetic field intensity is reduced in 10% steps, while the transient pulse repetition rate is increased to 100 times per second until the deviation between the measured data and the predicted value converges to within the threshold.

[0042] The digital verification unit performs reverse simulation using a digital twin coupled with electromagnetic, thermal, and mechanical multi-physics fields. After importing the three-dimensional electromagnetic field distribution data preprocessed by the edge computing node into the twin model, it first updates the material thermal expansion coefficient boundary conditions based on the temperature sensor data. When the local temperature rise reaches 60°C, the temperature decay curve of the material's magnetic permeability is automatically activated. Simultaneously, the mechanical stress distribution map is reconstructed based on the vibration acceleration data. If a resonance peak between 100Hz and 500Hz is detected in the vibration spectrum of the insulating interface region, an equivalent alternating stress load is applied in the simulation. The topological consistency check between the simulation results and the theoretical calculations uses a gradient difference analysis method. Specifically, for the phase difference of the leakage flux in the joint region, if the angle deviation between the simulated value and the theoretical value for five consecutive sampling points exceeds 3 degrees, the material interface contact resistance is determined to be abnormal, triggering the reverse optimization calculation of the conductivity parameters in the dynamic attenuation factor matrix and simultaneously adjusting the delay compensation coefficient of the asymmetric frequency-domain convolution kernel.

[0043] During the dual-screening mechanism, the data synchronization period between the physical verification unit and the digital verification unit is locked to 2 seconds. When the measured attenuation rate error of the physical verification exceeds 5% and the phase consistency error of the digital verification exceeds 3%, the sensor node's adaptive resampling mode is prioritized, increasing the sampling frequency in the joint area to 20MHz for 10 interference cycles. If only a single mode is out of tolerance, an optimization path is selected based on the characteristics of the out-of-tolerance frequency band: high-frequency band out-of-tolerance triggers convolution kernel weight iteration, while low-frequency band out-of-tolerance triggers material parameter boundary reset. For frequency bands that fail to meet the standard after three consecutive dual-screening tests, such as 8MHz-10MHz, the defect localization protocol is activated: the spatial hotspot coordinates corresponding to the frequency band are marked in the digital twin, and the physical verification unit is instructed to continuously test the area with double the interference intensity. If the shielding effectiveness decay rate accelerates by more than 50% / minute, it is determined to be a structural shielding failure, a priority repair instruction is generated, and the test channel for that frequency band is frozen. Under high temperature and high vibration combined working conditions, the double screening cycle is shortened to 0.5 seconds, and the multi-physics field solver of the digital verification unit uses a reduced-order model to accelerate calculations, ensuring that the dynamic tracking error between the simulation and measured data is maintained at less than 2% when the temperature change rate exceeds 5°C / s.

[0044] Each sensor node of the flexible distributed intelligent sensor array has a built-in edge computing unit, which uses pulse coded multiplexing transmission technology to compress and perform frequency domain sparsification processing on the raw data, generating a three-dimensional electromagnetic field dynamic distribution data set containing time domain waveforms, frequency domain energy and spatial coordinates.

[0045] It is important to further clarify that, in its implementation, each sensor node in the flexible distributed intelligent sensor array incorporates a built-in edge computing unit, which performs localized preprocessing of the raw electromagnetic field signals during the data acquisition phase. When transient pulse interference is detected, the edge unit prioritizes extracting the amplitude slope characteristics of the rising edge of the waveform and, through pulse code multiplexing, allocates signals of different frequency bands to independent transmission channels. High-frequency harmonics (i.e., harmonics greater than 5 MHz) are differentially encoded to suppress common-mode interference, while Manchester encoding is used for power-frequency magnetic field signals to ensure clock synchronization. During frequency-domain sparsification, the edge computing unit filters the raw data based on a dynamic noise floor threshold. Specifically, if the energy value of a frequency point is less than 1.2 times the mean ambient noise value for three consecutive samples, it is identified as an invalid interference component and removed. For high-density sensor nodes in seam areas, an additional spatial correlation filtering algorithm is applied. When the phase difference of the magnetic field detected by three adjacent nodes at the same time is less than 2 degrees, only the data from the central node is retained to reduce redundancy. The processed data set is reorganized according to the spatial coordinate grid. The joint area is mapped using a three-dimensional polar coordinate system with an accuracy of 0.5mm, and the insulating interface area is converted to a two-dimensional cylindrical coordinate system. Combined with the timestamp of the temperature-vibration composite sensing data, a dynamic distribution map is generated that binds the time domain waveform amplitude, frequency domain energy ratio and spatial position.

[0046] When the data transmission channel load exceeds 80%, the edge computing unit initiates an adaptive compression strategy: for high-frequency data greater than 2MHz, run-length encoding is used to compress the repeated waveform fragments of transient pulses; for low-frequency data less than 1kHz, linear interpolation is used to reduce the sampling point density to ensure that the original waveform fidelity of the key frequency band is not less than 95%.

[0047] When strong electromagnetic interference causes communication interruption of some nodes, the data reconstruction compensation mechanism of the neighboring nodes is triggered: based on the historical data trend of the interrupted node 10 seconds before, combined with the current temperature-vibration parameters, the electromagnetic field distribution during the missing period is predicted, and the predicted data is marked in red for subsequent verification modules to trace the error.

[0048] Based on real-time temperature and vibration data, the dynamic attenuation factor matrix dynamically corrects the nonlinear changes in the material's magnetic permeability and electrical conductivity through a parallelized tensor decomposition algorithm, outputting shielding effectiveness attenuation curves for each frequency band, both individually and in combination. It should be further explained that, in its implementation, the dynamic attenuation factor matrix receives real-time temperature and vibration data from the flexible distributed intelligent sensor array and first establishes nonlinear correction rules for material parameters: when the temperature in the joint area exceeds 45°C, the magnetic permeability correction factor is reduced by 0.8% per degree Celsius based on the pre-stored thermal aging curve for the metal material; if the peak vibration acceleration in the insulating interface area reaches 2g, the electrical conductivity correction is increased to 1.3 times the baseline value based on the stress-dielectric constant relationship table for epoxy resin composite materials.

[0049] The parallelized tensor decomposition algorithm adopts a priority strategy when performing spatiotemporal separation of the superimposed interference of high-frequency harmonics, transient pulses and power-frequency magnetic fields. That is, for transient pulses with a duration of less than 50 μs, the time-domain reflection path of the pulse penetrating the seam area is preferentially extracted; for high-frequency harmonics with a frequency higher than 5 MHz, the eddy current loss component on the surface of the shielding layer is separated based on the skin depth model.

[0050] The algorithm introduces a dynamic weight allocation mechanism during the time-space separation process: when the temperature sensor detects a local temperature rise rate exceeding 3°C / s, the weight coefficient of the high-frequency harmonic component in the corresponding area is automatically increased by 20% to enhance the impact of the thermal accumulation effect on shielding attenuation; if a resonance peak appears in the vibration spectrum within the range of 200Hz-800Hz, a 1.5-fold safety margin compensation is applied to the magnetic resistance calculation of the power frequency magnetic field.

[0051] For abnormal data under the action of multi-band coupling, such as a sudden increase of 15% in the measured attenuation rate in a certain frequency band compared with the model prediction value, the abnormal tracing mode of tensor decomposition is triggered, that is, the electromagnetic field distribution data 10 seconds before the current moment is subjected to time domain deconvolution operation to locate the maximum energy mutation point, and combined with the temperature-vibration data to determine whether it is caused by material degradation or mechanical deformation.

[0052] When outputting the shielding effectiveness attenuation curve, confidence interval thresholds are set for different frequency bands, including: for high frequency bands greater than 2MHz, the instantaneous fluctuation amplitude is allowed to not exceed 8%, and for low frequency bands less than 1kHz, the deviation of three consecutive samples is required to be less than 3%.

[0053] If the data of a certain frequency band exceeds the confidence interval five times in a row, the output of this frequency band will be frozen and a directional retest protocol will be initiated: 1.2 times the nominal interference intensity will be applied to the frequency band through the physical verification unit, and the reverse simulation data of the digital twin will be combined to verify whether it is a permanent shielding defect or a transient interference misjudgment.

[0054] Under high-temperature and high-vibration combined working conditions, the algorithm automatically switches to anti-interference mode, that is, shortening the update period of material parameters from 10 seconds to 2 seconds, and introducing a vibration displacement compensation factor into the reflection path calculation of transient pulses to ensure that the time-space separation accuracy can still maintain an error of less than 5% when the vibration acceleration reaches 3g.

[0055] The improved wavelet packet transform module of the frequency-time fusion analysis engine decomposes the time-domain waveform into multiple frequency band components and calculates the energy entropy of each component. An adaptive convolutional neural network extracts the correlation between time-domain distortion characteristics and frequency band energy distribution, generating a frequency-time correlation matrix to quantify the coupling contribution of different interference modes. It should be further explained that, in its specific implementation, the improved wavelet packet transform module of the frequency-time fusion analysis engine dynamically selects wavelet basis functions based on the transient characteristics of the interference signal when decomposing the time-domain waveform. For fast-rising pulses with a duration of less than 20μs, an eight-layer decomposition using the Daubechies wavelet basis with high-order vanishing moments is used to capture narrowband high-frequency components. For wide pulses caused by distortion of the power-frequency magnetic field fundamental wave, a six-layer decomposition using the Symlets wavelet basis is used to separate odd and even harmonic components. The energy entropy calculation at each decomposition level incorporates an operating condition compensation factor. Specifically, when the temperature sensor detects a temperature rise exceeding 55°C at the insulation interface, the energy entropy weight of the corresponding frequency band is increased by 30% to amplify the impact of thermally induced dielectric loss. If the vibration accelerometer measures an impact signal exceeding 2.5g in the joint area, a 20% attenuation compensation is applied to the energy entropy of high-frequency components above 1MHz to suppress spurious high-frequency noise caused by mechanical vibration. An adaptive convolutional neural network extracts key distortion patterns using a pretrained interference feature library. For periodic oscillation distortion caused by high-frequency harmonics, the network's first-layer convolution kernel focuses on detecting local extreme points where the waveform's zero-crossing rate suddenly changes. For asymmetric clipping distortion caused by transient pulses, a second-order derivative gradient threshold is used to identify waveform inflection points.

[0056] When constructing the frequency-time correlation matrix at the network output layer, a dynamic decision threshold mechanism is adopted. That is, when the energy proportion of a certain frequency band exceeds 18% of the total interference energy and the time domain distortion index increases by more than 4% for three consecutive iterations, it is judged as a critical failure frequency band; if the energy entropy values of two adjacent frequency bands show an alternating upward trend within five sampling cycles, the cross-band coupling warning is activated, and the potential thermal-mechanical coupling failure area is marked in the digital twin.

[0057] For scenarios where the time domain distortion characteristics conflict with the frequency band energy distribution, such as when the high-frequency energy ratio meets the standard but the time domain waveform integrity does not exceed the limit, the network attention weight redistribution mechanism is triggered, including: using the temperature rise rate of the temperature sensor and the resonance peak ratio of the vibration spectrum as auxiliary decision-making factors, increasing their weight coefficients to 40% of the main feature parameters, and recalculating the coupling contribution ranking of the frequency-time correlation matrix.

[0058] When the spatial deviation between the critical frequency band predicted by the model and the measured shielding failure position of the physical verification unit exceeds 10 mm, the wavelet packet decomposition layer is automatically switched to the seventh layer for local frequency band refinement, and the convolution kernel size is adjusted to a 3×3 grid to match the geometric characteristics of the seam defect.

[0059] The analysis engine performs a full-band health assessment every 8 seconds. If the same frequency band is marked as critical in two consecutive assessments and the digital twin's reverse simulation verification error is less than 2%, a frequency band lock instruction is generated. The shielded coupler automatically limits the interference intensity of this frequency band to 50% of the nominal value in subsequent tests to prevent accelerated degradation. Under the combined operating conditions of high temperatures exceeding 60°C and high vibration exceeding 3g, the engine activates anti-disturbance mode. This fixes the number of wavelet packet decomposition layers to five to reduce the computational load and applies time-domain waveform smoothing filtering to the input layer of the convolutional neural network to ensure that the error rate of critical frequency band identification remains below 5% despite vibration-induced signal jitter.

[0060] The dual-screening mechanism requires that the shielding effectiveness evaluation results for any frequency band must have a maximum error of 5% between the measured data and the model predictions in the physical verification unit, and a maximum error of 3% between the topological consistency of the reverse simulation and theoretical calculations in the digital verification unit. If any of these errors exceed the limit, adaptive resampling of the sensor nodes and iterative updates of the weights of the asymmetric frequency domain convolution kernel are triggered. It should be further explained that during the specific implementation, when the dual-modal verification module performs dual screening, the physical verification unit dynamically matches the composite interference scenario reproduced in the standard electromagnetic interference chamber with the real-time monitoring data of the flexible distributed intelligent sensor array. When the peak leakage flux density in the seam area is detected to exceed 250μT, the physical verification unit automatically enhances the high-frequency harmonic component to 1.8 times the nominal value and injects a transient pulse train with a pulse width of 30ns to simulate the electromagnetic stress impact under extreme working conditions. In the real-time comparison between the measured physical data and the predicted values of the dynamic coupling model, if the attenuation rate deviation exceeds 5% three times in a row within the 5MHz-20MHz frequency band, the adaptive adjustment of the interference parameters will be triggered, that is, the power frequency magnetic field intensity will be reduced in steps of 15%, and the transient pulse repetition rate will be increased to 150 times per second until the deviation converges to within the threshold.

[0061] When the digital verification unit performs reverse simulation using a digital twin coupled with electromagnetic, thermal, and mechanical multi-physics fields, it maps the real-time temperature data of the joint area into a material thermal expansion coefficient gradient field. This includes activating a nonlinear attenuation model of magnetic permeability with temperature when the local temperature rise reaches 70°C, and applying an equivalent alternating stress load in areas where the vibration acceleration exceeds 3g. The topological consistency between the simulation results and theoretical calculations is verified using phase gradient analysis. Specifically, if the angular deviation between the simulated and theoretical values of the leakage flux phase difference in the joint area exceeds 2 degrees at five consecutive sampling points, the interface contact resistance is determined to be abnormal, triggering inverse optimization of the dynamic attenuation factor matrix and simultaneously adjusting the delay compensation coefficient of the asymmetric frequency-domain convolution kernel to 1.2 times the current dominant frequency of the vibration spectrum.

[0062] The dual-screening mechanism performs error threshold verification with a 2-second synchronization cycle. When the measured attenuation error (physical verification) exceeds 5% and the phase consistency error (digital verification) exceeds 3%, directional resampling of the sensor nodes in the joint area is prioritized, increasing the sampling frequency to 25MHz for 15 interference cycles. If only a single mode exceeds the tolerance, an optimization path is selected based on the frequency band characteristics. For high-frequency bands exceeding 10MHz, the weight coefficients of the asymmetric frequency-domain convolution kernel are iteratively updated using the gradient descent method. For low-frequency bands below 1kHz, the material parameter boundary conditions are reset and the tensor decomposition operation is restarted. For frequency bands that fail to meet the standards three times in a row, such as 12MHz-15MHz, the digital twin marks the spatial hotspot coordinates corresponding to that frequency band and instructs the physical verification unit to apply three times the interference intensity to the area for stress testing. This includes: if the shielding effectiveness decay rate accelerates by more than 60% within one minute, it is determined to be a structural failure, a maintenance instruction is generated, and the test channel for that frequency band is shut down. Under the combined conditions of high temperatures exceeding 80°C and high vibrations exceeding 4g, the screening cycle is compressed to 0.3 seconds. The digital verification unit's multiphysics solver switches to a reduced-order model. When the temperature change rate exceeds 8°C / s, low-frequency components below 200Hz in the vibration spectrum are automatically ignored to ensure the simulation dynamic error is less than 1.5%. If a frequency band passes the physical verification but fails the digital verification, a parameter sensitivity analysis of the digital twin is initiated. This involves sequentially perturbing the material's magnetic permeability, electrical conductivity, and dielectric constant parameters to find the parameter combination that best matches the measured data. This combination then updates the correction rule base for the dynamic attenuation factor matrix.

[0063] The parallelized tensor decomposition algorithm performs spatiotemporal separation of the superimposed disturbance sources of high-frequency harmonics, transient pulses, and power-frequency magnetic fields, and predicts the maximum attenuation rate and critical failure time parameters of the shielding effectiveness based on a Markov chain. It should be further explained that, in its specific implementation, the parallelized tensor decomposition algorithm first classifies the interference signals based on their spatiotemporal characteristics when processing the superimposed disturbances of high-frequency harmonics, transient pulses, and power-frequency magnetic fields. For transient pulses with a duration of less than 30 μs, time-domain reflection path tracing technology is used to separate the multiple reflection energy components in the seam area by matching the time delay difference between the incident and reflected waveforms. For high-frequency harmonics with a frequency greater than 8 MHz, the eddy current loss distribution on the shielding layer surface is calculated based on the skin depth model, and the additional attenuation error caused by the decrease in material conductivity due to temperature rise is eliminated.

[0064] The algorithm introduces a dynamic priority strategy during the time-space separation process. That is, when the temperature sensor detects that the temperature rise rate in the insulation interface area exceeds 5°C / s, the weight coefficient of the high-frequency harmonics is automatically increased by 25% to enhance the nonlinear influence of the heat accumulation effect on the shielding effectiveness. If a resonance peak appears in the vibration spectrum within the range of 300Hz-1kHz, a 2-fold safety margin compensation is applied to the magnetic resistance calculation of the power frequency magnetic field to avoid the magnetic circuit saturation effect caused by mechanical vibration.

[0065] The Markov chain prediction model constructs a state transition matrix based on historical attenuation curves. When the instantaneous attenuation rate of shielding effectiveness in a frequency band exceeds 1.5 times its historical maximum, a three-level warning mechanism is triggered. Specifically, if the predicted critical failure time deviates from the digital twin simulation result by more than 10%, a dynamic calibration procedure for model parameters is initiated, adjusting the initial weights of state transition probabilities through a backpropagation algorithm. For prediction conflicts caused by multi-band coupling, such as when a high-band prediction fails while a low-band remains within the safety threshold, the algorithm initiates a collaborative analysis mode. This involves inputting the correlation coefficient matrix of shielding effectiveness attenuation rates for each frequency band into the Markov chain, calculating the cross-band coupling failure probability, and marking the geometric coordinates of high-risk areas in the digital twin.

[0066] When the predicted results deviate from the measured data of the physical verification unit by more than 8% for three consecutive times, the spatiotemporal recalibration protocol is triggered, including: freezing the current prediction model, reloading the electromagnetic field distribution data of the previous 30 seconds for tensor decomposition calculations, and if the matching degree between the attenuation trend of the decomposed independent frequency band and the measured data is improved to more than 95%, it is determined to be a model offset caused by transient interference; otherwise, it is marked as material parameter drift and a parameter reset instruction is generated.

[0067] Under high temperatures greater than 70°C or high vibration conditions greater than 3.5g, the algorithm switches to anti-disturbance mode, namely: shortening the state transition period of the Markov chain from 5 seconds to 1 second, and introducing a vibration displacement compensation factor into the reflection path calculation of the transient pulse, compensating the path length by 0.2mm for every 1g of vibration acceleration, to ensure that the prediction error of the maximum attenuation rate under extreme conditions does not exceed 6%.

[0068] Regarding the confidence interval setting of the shielding effectiveness curve, the high-frequency band greater than 5MHz allows the instantaneous fluctuation amplitude to not exceed 10%, and the low-frequency band less than 2kHz requires the deviation of five consecutive sampling times to be less than 2%; if the data of a certain frequency band exceeds the confidence interval twice in a row, the directional verification process will be initiated: 1.5 times the nominal interference intensity will be applied to the frequency band through the physical verification unit, and the parameter sensitivity analysis of the digital twin will be triggered simultaneously to verify whether it is a permanent defect or a transient interference anomaly.

[0069] The output of the frequency-time correlation matrix is connected to a visualization evaluation module, generating a multi-dimensional test report containing an electromagnetic field distribution heat map, frequency band energy percentages, and dynamic attenuation trends. It should be further explained that during implementation, after the output data of the frequency-time correlation matrix is connected to the visualization evaluation module, the rendering accuracy of the electromagnetic field distribution heat map is first aligned with the sensor spatial topology. This includes: heat map mapping with a 0.2mm grid resolution for the joint area, and dynamic adjustment of the color gradient threshold based on the leakage flux density. Specifically, when the local density exceeds 150μT, the color gradient switches to a red warning color scale with a flashing effect superimposed. For the insulating interface area, a blue-green gradient color scale based on 50μT is used.

[0070] The frequency band energy share analysis module calculates the percentage of total interference energy in each frequency band in real time based on the output of the frequency-time correlation matrix. When the energy share of the 8MHz-12MHz band exceeds 18% three times in a row, the corresponding spatial region is automatically marked with a pulse halo in the 3D model and cross-validated with the hot zone coordinates of the digital twin. The dynamic attenuation trend curve is displayed synchronously using a dual-axis coordinate system: the primary axis represents the shielding effectiveness attenuation rate, and the secondary axis represents the temperature-vibration composite impact factor index. When the slope of the attenuation rate curve exceeds twice the historical average and the impact factor index exceeds 80%, the red dashed line warning indicator on the trend prediction line is triggered.

[0071] Under high temperature conditions greater than 60°C, the visualization module activates the thermal failure enhancement mode: the color gradient threshold of the thermal map in the joint area is lowered by 20%, and the electromagnetic field phase difference cloud map within a radius of 5mm around the temperature sensor is forced to be displayed; if the vibration acceleration exceeds 2g at the same time, the vibration spectrum characteristic peak mark is inserted in the dynamic trend chart, and the frequency band range that may cause mechanical resonance is marked with yellow shading.

[0072] When there is a spatial deviation between the reverse simulation results of the digital verification unit and the physical measured data, the visualization interface automatically splits the screen to display the thermal map of the electromagnetic field distribution differences between the two, and calculates the maximum deviation coordinate point in the joint area. That is, if the deviation value exceeds 10%, a rotating alarm icon is generated at the coordinate point, and the historical deviation change curve of the previous 30 seconds is displayed in conjunction.

[0073] For critical failure frequency bands, the module generates multi-dimensional test reports using a graded warning mechanism. This includes: if a critical condition is detected only once, the report will indicate a recommended recheck in orange. If a critical condition is detected three times consecutively in the same frequency band with a digital verification error less than 3%, the report will indicate a priority repair item in red. An additional snapshot function is included to save a snapshot of the electromagnetic field distribution at the moment of failure. When extreme electromagnetic interference causes data interruption, the visualization module activates a historical data infill algorithm. Based on the attenuation curve trend of the previous five minutes, a gray dashed line is used to predict the shielding effectiveness change during the missing period. Once data is restored, a color gradient is used to smoothly transition between the predicted and measured values.

[0074] The digital verification unit's multi-physics coupling solver updates the digital twin's material thermal expansion coefficient boundary conditions based on real-time temperature data and updates the mechanical stress distribution parameters based on vibration data, enabling dynamic calibration of electromagnetic field inverse simulation. It should be further explained that, in specific implementations, the digital verification unit's multi-physics coupling solver dynamically updates the digital twin's material parameter boundary conditions after receiving real-time temperature and vibration data. This includes: when the temperature sensor in the joint region detects a temperature rise exceeding 50°C, the solver corrects the expansion of the mesh elements in the insulating interface region to 1.2 times the theoretical value based on the nonlinear curve of the material thermal expansion coefficient versus temperature, and simultaneously reduces the magnetic permeability weight coefficient in this region to 85% of the baseline value. If the vibration accelerometer measures an impact signal exceeding 2.2g in the insulating interface region, an equivalent alternating stress load is applied to the digital twin, with an amplitude of 1.5 times the amplitude of the measured main frequency of the vibration spectrum and aligned with the vibration sensor's coordinate axes.

[0075] During the reverse simulation process, the solver uses a phase gradient matching algorithm to verify the electromagnetic field distribution. That is, if the angle deviation between the simulated value of the leakage flux phase difference in the joint area and the theoretical calculated value at three consecutive sampling points exceeds 1.5 degrees, it is judged as a contact resistance anomaly, triggering the reverse optimization of the material conductivity parameters: the conductivity correction coefficient is iteratively adjusted in steps of 0.1% until the phase deviation converges to within the threshold. At the same time, the optimized parameters are synchronized to the real-time update queue of the dynamic attenuation factor matrix.

[0076] Under high-temperature conditions greater than 65°C, the solver activates the thermally induced material degradation compensation model, including: when the temperature rise rate exceeds 4°C / s, the update period of the thermal expansion coefficient is shortened from 10 seconds to 2 seconds, and an accelerated aging factor is applied to the temperature decay curve of the magnetic permeability. The accelerated aging factor increases the aging rate by 15% for every 10°C increase; if a resonance peak of vibration acceleration is detected in the 500Hz-1kHz frequency band, a displacement compensation algorithm is introduced for the mechanical stress distribution in the simulation, that is, every 1g of vibration acceleration corresponds to a compensation of 0.25mm of grid node displacement to avoid distortion of the electromagnetic field path calculation caused by vibration deformation.

[0077] When the shielding effectiveness attenuation rate deviation between the physical verification unit and the digital twin exceeds 8%, the parameter sensitivity analysis mode is activated: the magnetic permeability, electrical conductivity and dielectric constant parameters are perturbed in turn, with the perturbation amplitude within plus or minus 5%, and the parameter combination with the highest match to the measured data is screened out, and the correction rule base of the dynamic attenuation factor matrix is reconstructed accordingly.

[0078] For combined operating conditions of high temperatures exceeding 70°C and high vibrations exceeding 3g, the solver switches to reduced-order operation mode. This mode retains only the high-precision meshes in the joint area and the insulating interface area, while the remaining areas use an equivalent homogenized model. This ensures that the simulation refresh rate can be maintained at no less than 5Hz even when the temperature change rate exceeds 6°C / s.

[0079] If the topological consistency between the reverse simulation results and theoretical calculations fails three consecutive times, the digital twin reconstruction protocol is triggered. This involves reconstructing the initial boundary conditions based on the previous 30 seconds of sensor data and using a genetic algorithm to globally optimize the material parameter combination until the maximum phase deviation is reduced to within 2 degrees. If extreme electromagnetic interference causes some sensors to fail, the solver initiates a data substitution mechanism. This interpolation generates missing data based on the electromagnetic field gradient trends of adjacent nodes, and the replacement data area is marked with a purple dashed line for subsequent manual review and verification.

[0080] The trigger signal from the adaptive sampling chip is synchronously transmitted to all sensor nodes via optical fiber, ensuring that the clock deviation of data collected by multiple nodes is less than 0.1μs. It should be further explained that during the specific implementation, when the trigger signal from the adaptive sampling chip is distributed to all sensor nodes via the optical fiber synchronous transmission network, a hierarchical compensation mechanism is used to ensure clock deviation control: After the master control node generates a reference clock signal, it dynamically calculates the optical transmission delay compensation value based on the spatial distribution distance of the sensor nodes. Specifically, for high-density node clusters in the seam area, the compensation delay is increased by 0.05ps per millimeter of transmission distance; for sparse nodes in the insulating interface area, a fixed compensation delay of 1.2ns is used to reduce the computational load. The spacing between high-density node clusters is less than 5cm, while the spacing between sparse nodes is greater than 10cm.

[0081] The optical fiber transmission path adopts a ring topology. When a node's signal transmission abnormality is detected, such as optical power attenuation exceeding 3dB, it automatically switches to the backup path and marks the node as a low-priority data source until the fault is eliminated.

[0082] At high temperatures exceeding 55°C, the optical fiber's refractive index temperature drift compensation algorithm is triggered. This algorithm adjusts the optical pulse width by 0.003% per degree Celsius increase based on real-time temperature sensor data, ensuring that the clock signal's rising edge jitter under thermally induced deformation is less than 5ps. If vibration acceleration exceeds 1.8g, causing a sudden increase in fiber microbend losses, a redundant signal retransmission mechanism is activated. Specifically, the same trigger signal is transmitted in parallel through three adjacent optical fiber paths. The receiving end uses majority voting logic to filter valid signals and marks the abnormal path as offline for maintenance.

[0083] The clock synchronization accuracy check is performed every 10 seconds: the master control node sends a check pulse sequence, and each node returns the timestamp deviation data. If the node deviation in the seam area exceeds 0.1μs, the dynamic compensation coefficient update is triggered, that is, the pre-compensation delay of the subsequent trigger pulse is adjusted by 120% of the deviation value; if the node deviation in the insulation interface area exceeds 0.15μs for two consecutive times, it is judged that the optical fiber connector has poor contact, and a physical maintenance instruction is generated and the sampling frequency of the area is reduced to 80% of the nominal value.

[0084] In extreme electromagnetic interference environments, the fiber optic transmission mode switches to a noise-resistant coding protocol, including modulating the clock signal into differential phase-coded pulses, extracting the effective signal at the receiving end through a correlation detection algorithm, and suppressing the false trigger rate caused by common-mode interference to below 0.01%; among them, the field strength of extreme electromagnetic interference is greater than 300V / m.

[0085] When the system detects a multi-node synchronization failure, it initiates a global clock reconstruction process. This involves using the three nodes with the smallest deviations as the new reference source, recalculating the network-wide delay compensation table, and rebuilding the electromagnetic field acquisition timing model in the digital twin to ensure that the time alignment error of subsequent data does not exceed 0.08μs. Synchronization failure refers to a node deviation exceeding 30%.

[0086] The dynamic coupling model construction module, the frequency-time fusion analysis engine, and the dual-modal verification module are connected via a closed-loop feedback link, forming an iterative optimization architecture for dynamic quantitative shielding effectiveness evaluation. It should be further explained that during implementation, when these modules form an iterative optimization architecture through a closed-loop feedback link, the data synchronization mechanism is based on a timestamp alignment strategy. This includes matching packets every one second between the shielding effectiveness attenuation curve output by the dynamic coupling model and the critical frequency band determination results from the frequency-time fusion analysis engine. If there is a conflict between the model-predicted attenuation rate in the 6MHz-8MHz frequency band and the coupling contribution ranking of the frequency-time correlation matrix, dynamic model weight adjustment is triggered. Specifically, the temperature sensor's temperature rise rate and the amplitude of the main frequency of the vibration spectrum are added as additional factors, weighted at 20%, and injected into the penetration path calculation of the asymmetric frequency-domain convolution kernel. This also synchronizes the material parameter boundary conditions of the digital twin. Conflicts in the coupling contribution ranking include situations where the attenuation rate exceeds the limit but the contribution does not meet the threshold.

[0087] After the dual screening results of the physical verification unit and the digital verification unit are input into the closed-loop link, the system executes the priority decision logic, including: if the measured attenuation rate error of the physical verification exceeds 5% and the digital verification error is less than 2%, it is judged as a model parameter offset, and the magnetic permeability correction coefficient of the dynamic attenuation factor matrix is adjusted first; if the phase consistency error of the digital verification exceeds 3% and the physical verification error meets the standard, it is judged as a sensor clock synchronization anomaly, triggering the reset of the delay compensation coefficient of the optical fiber transmission network. For compound error scenarios, if both errors exceed the limit, the full-link traceability analysis is initiated, that is, the input and output data streams of each module in the previous 30 seconds are reversely traced to locate the operation node where the deviation first occurred, such as the space-time separation unit of tensor decomposition or the frequency band decomposition level of wavelet packet transform, and the operation process of the node is restarted in a targeted manner.

[0088] During the iterative optimization process, the system dynamically adjusts the feedback cycle based on operating conditions: under normal operating conditions, global parameter updates are performed every 5 seconds. When a temperature rise rate exceeding 4°C / s or a vibration acceleration exceeding 3g is detected, the feedback cycle is shortened to 1 second, and the confidence interval of the shielding effectiveness attenuation curve is forced to be compressed to 70% of the nominal value. If the fluctuation of key parameters still exceeds 10% after three consecutive iterations, the system switches to conservative optimization mode. This freezes the morphological changes of the asymmetric frequency-domain convolution kernel and gradually approaches the optimal solution only through linear correction of the dynamic attenuation factor matrix until the fluctuation converges to within 5%. Key parameters include the predicted critical failure time.

[0089] In the event of extreme electromagnetic interference, such as field strengths exceeding 500V / m, leading to data link interruption, the closed-loop feedback architecture activates disaster recovery mechanisms. This includes generating a virtual feedback signal based on the previous 10 seconds of historical data, maintaining the progressive update of model parameters, and marking the virtual data segment with a yellow warning icon. If the interruption lasts for more than 20 seconds, the system switches to local optimization mode, in which each module independently calculates based on the most recent valid data. After communication is restored, parameters are rapidly resynchronized through a data consistency check, which includes ensuring greater than 90% similarity in frequency band energy distribution. For optimization failures under high-temperature and high-vibration conditions, such as material parameter drift exceeding 0.5% / minute, an expert intervention interface is triggered. This interface outputs a snapshot of intermediate variables and correlation heat maps for all current modules, and recommends three alternative parameter combinations for manual decision-making, ensuring system robustness under unpredictable conditions.

[0090] A test method for an electromagnetic shielding coupler of a test transformer based on intelligent sensing includes the following steps:

[0091] Step S1: Through the non-uniform topology deployment of the flexible distributed intelligent sensor array, multi-band electromagnetic field data from the surface seams and insulating interface areas of the electromagnetic shielding coupler are collected in real time; the adaptive sampling chip dynamically adjusts the 1kHz-10MHz sampling frequency according to the electromagnetic field gradient, and realizes microsecond-level multi-node data synchronization based on the optical fiber synchronization trigger mechanism; the edge computing node performs pulse code multiplexing transmission and frequency domain sparse processing on the raw data to generate a three-dimensional dynamic electromagnetic field distribution map;

[0092] Step S2: The dynamic coupling model construction module receives preprocessed data and uses an asymmetric frequency domain convolution kernel to analyze the penetration paths of high-frequency harmonics, transient pulses, and power-frequency magnetic fields. Combining temperature-vibration composite sensing data, the module uses a dynamic attenuation factor matrix to correct the nonlinear parameters of the material's magnetic permeability and electrical conductivity in real time. A parallel tensor decomposition algorithm is used to separate the spatiotemporal coupling effects of multi-band interference and output shielding effectiveness attenuation curves for each frequency band.

[0093] Step S3: The frequency-time fusion analysis engine uses an improved wavelet packet transform to decompose the time domain waveform and calculate the energy entropy value of each frequency band. An adaptive convolutional neural network extracts the mapping relationship between the time domain distortion characteristics and the frequency band energy distribution to generate a frequency-time correlation matrix. Based on the Markov chain prediction model, the critical failure frequency band threshold and dynamic attenuation trend parameters are identified.

[0094] Step S4: The physical verification unit reproduces the composite interference scenario in a standard electromagnetic interference chamber and compares the measured shielding effectiveness attenuation rate with the model prediction. The digital verification unit performs reverse simulation using a digital twin coupled with electromagnetic, thermal, and mechanical multi-physics fields to verify the consistency of the electromagnetic field distribution topology. The dual screening mechanism performs threshold judgment on the error between the physical and digital modes. If the error exceeds the limit, adaptive resampling and model parameter iteration are triggered.

[0095] Step S5: Dynamically adjust the asymmetric frequency domain convolution kernel weights and dynamic attenuation factor matrix parameters based on the verification results; generate electromagnetic field distribution heat maps, frequency band energy proportions, and attenuation trend curves through a visual evaluation module; the system continuously iterates and optimizes until the test results meet the dual screening threshold requirements, and outputs a multi-dimensional evaluation report and shielding structure optimization recommendations.

[0096] Through dynamic electromagnetic field reconstruction and multi-band interference coupling modeling using a flexible distributed intelligent sensor array, the limitations of traditional static testing have been overcome, enabling real-time capture and multi-dimensional decoupling of shielding effectiveness attenuation characteristics in complex alternating electromagnetic environments. A frequency-time fusion analysis engine is used to establish a correlation mapping between time-domain distortion and frequency-domain energy, accurately identifying critical failure frequency band thresholds. Combined with a dual-modal verification mechanism, the test results are self-consistent, reducing the error in dynamic shielding effectiveness quantification compared to traditional methods and enhancing the engineering guidance value of the test data.

[0097] Closed-loop feedback architecture and digital twin reverse simulation technology provide a dynamic optimization basis for the reliability design of electromagnetic shielding couplers, resolving the pain points of distorted operating condition simulation and ambiguous failure mechanisms in traditional testing. Through a physical-digital dual-modal verification and visual evaluation module, weak shielding areas can be quickly located and life degradation trends predicted, improving the efficiency of shielding structure optimization. It also supports adaptive switching of anti-interference modes under extreme operating conditions, providing an innovative testing method for the safe operation and maintenance of smart grid equipment in complex electromagnetic environments.

[0098] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A test system for electromagnetic shielding coupler of test transformer based on intelligent sensing, characterized in that: include: A flexible distributed intelligent sensor array, consisting of multiple high-density micro-electromagnetic sensors and temperature-vibration composite sensing units, is embedded in the seams and insulating interface areas of the electromagnetic shielding coupler surface in a non-uniform topology. The sensor units dynamically adjust the sampling frequency from 1kHz to 10MHz through an adaptive sampling chip, and achieve microsecond-level synchronous data acquisition based on a spatiotemporal synchronization trigger mechanism. a dynamic coupling model construction module, connected to the output end of the flexible distributed intelligent sensor array, including an asymmetric frequency domain convolution kernel and a dynamic attenuation factor matrix, for simulating the penetration path of multi-band interference signals in the shielding structure and correcting the material nonlinear parameters in combination with real-time temperature and vibration data; A frequency-time fusion analysis engine, connected to the output end of the dynamic coupling model building module, includes an improved wavelet packet transform module and an adaptive convolutional neural network, and is used to establish a mapping relationship between time domain waveform distortion and frequency domain energy distribution, and identify critical frequency band thresholds; The dual-modal verification module includes a physical verification unit and a digital verification unit. The physical verification unit reproduces the complex interference scenario in a standard electromagnetic interference chamber, and the digital verification unit performs reverse simulation based on a digital twin coupled with electromagnetic, thermal, and mechanical multi-physical fields. Both use a dual screening mechanism to check the error threshold of the evaluation results, triggering adaptive resampling and model iteration.

2. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: Each sensor node of the flexible distributed intelligent sensor array has a built-in edge computing unit, which uses pulse code multiplexing transmission technology to compress and perform frequency domain sparsification processing on the original data to generate a three-dimensional electromagnetic field dynamic distribution data set containing time domain waveforms, frequency domain energy and spatial coordinates.

3. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: The dynamic attenuation factor matrix is based on real-time temperature and vibration data, and dynamically corrects the nonlinear changes of material magnetic permeability and electrical conductivity through a parallel tensor decomposition algorithm, outputting the shielding effectiveness attenuation curves under the individual and synergistic effects of each frequency band.

4. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: The improved wavelet packet transform module of the frequency-time fusion analysis engine decomposes the time domain waveform into multiple layers of frequency band components and calculates the energy entropy value of each component. The adaptive convolutional neural network extracts the correlation between the time domain distortion characteristics and the frequency band energy distribution, and generates a frequency-time correlation matrix to quantify the coupling contribution of different interference modes.

5. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: The dual screening mechanism requires that the error between the measured data and the model prediction value of the shielding effectiveness evaluation results of any frequency band in the physical verification unit shall not exceed 5%, and the topological consistency error between the reverse simulation and theoretical calculation in the digital verification unit shall not exceed 3%. If any error exceeds the limit, the adaptive resampling of the sensor node and the iterative update of the weights of the asymmetric frequency domain convolution kernel will be triggered.

6. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 3, characterized in that: The parallelized tensor decomposition algorithm performs spatiotemporal separation on the superimposed disturbance sources of high-frequency harmonics, transient pulses and power-frequency magnetic fields, and predicts the maximum attenuation rate and critical failure time parameters of shielding effectiveness based on a Markov chain.

7. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 4, characterized in that: The output end of the frequency-time correlation matrix is connected to a visualization evaluation module to generate a multi-dimensional test report including an electromagnetic field distribution heat map, frequency band energy proportion, and dynamic attenuation trend.

8. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 5, characterized in that: The multi-physics field coupling solver of the digital verification unit updates the material thermal expansion coefficient boundary conditions of the digital twin according to the real-time temperature data, and updates the mechanical stress distribution parameters according to the vibration data, thereby realizing dynamic calibration of the electromagnetic field reverse simulation.

9. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: The trigger signal of the adaptive sampling chip is synchronously transmitted to all sensor nodes through optical fibers, ensuring that the clock deviation of multi-node data collection is less than 0.1 μs.

10. The test system for testing transformer electromagnetic shielding couplers based on intelligent sensing according to claim 1, characterized in that: The dynamic coupling model construction module, the frequency-time fusion analysis engine and the dual-modal verification module are connected through a closed-loop feedback link to form an iterative optimization architecture for dynamic quantitative evaluation of shielding effectiveness.

Images