A shielding effectiveness monitoring method and system

Abstract

The application discloses a shielding effectiveness monitoring method and system, the method comprises: space electromagnetic environment modeling based on multi-sensor fusion and ray tracing; jointly analyzing the mixed signal collected by the sensor network in real time, extracting the fingerprint features representing different signal sources; based on the extracted fingerprint features and type identification results, dynamically defining the effective action boundary of the current shielding field, distinguishing the effective signal in the shielding area from the cross-border interference; at the same time, using a deep residual network to identify weak signals under strong interference and estimate power compensation; finally outputting the multi-dimensional evaluation results of shielding effectiveness monitoring. By using signal fingerprint analysis and intelligent separation, the application enhances the anti-confusion and anti-interference ability, and reduces the misjudgment rate. Even under strong background interference, the application can still effectively identify weak shielding test signals and accurately diagnose the root cause of the monitoring failure phenomenon, thereby improving the accuracy of the shielding effectiveness monitoring result.

Description

Technical Field

[0001] This invention relates to the field of shielding effectiveness monitoring technology, and specifically to a shielding effectiveness monitoring method and system. Background Technology

[0002] Shielding effectiveness monitoring instruments are used to monitor the real-time shielding effectiveness of shielding structures. Based on the shielding effectiveness requirements of various shielding structures and the main frequency bands and signal transmission characteristics of information leakage, this equipment is specifically designed for monitoring the shielding effectiveness of various shielded rooms and shielding structures. In modern network environments, shielding effectiveness monitoring has become particularly important, using professional technical means to evaluate the actual effectiveness of filtering or shielding measures implemented in the network. With the continuous improvement of various network monitoring and filtering methods, shielding effectiveness monitoring can not only help optimize network configuration but also ensure the security and smooth flow of information.

[0003] Existing shielding effectiveness monitoring systems often encounter challenges in complex environments. Obstacles such as walls and metal structures can block, reflect, and attenuate interference signals, creating signal "dead zones" or "blind spots." This results in incomplete or distorted monitoring results because these areas cannot be fully and uniformly captured. Furthermore, in some specialized scenarios, interference signals can extend beyond the shielded areas. If the monitoring instrument only monitors the signal strength within the shielded area, it cannot accurately reflect the monitoring results and range, leading to improper interference range control and false alarms. When the shielded area is too close to a base station, the base station signal strength is extremely strong, requiring higher power output for shielding effectiveness monitoring. However, if the shielding effectiveness monitoring instrument's power is insufficient or its frequency band setting is incompatible (e.g., unable to cover new 5G frequency bands), the instrument may misjudge the shielding as ineffective, when in reality the shielding device's performance is inadequate, resulting in a false alarm. Summary of the Invention

[0004] The purpose of this invention is to provide a shielding effectiveness monitoring method and system to solve the problems mentioned in the background art.

[0005] The specific technical solution provided by this invention is as follows: A shielding effectiveness monitoring method, comprising the following operational steps:

[0006] Step S1: Modeling the spatial electromagnetic environment based on multi-sensor fusion and ray tracing.

[0007] Preferably, the specific implementation includes:

[0008] Step S11: Obtain the three-dimensional structural data of the monitoring space through laser scanning or by importing a BIM model;

[0009] Step S12: Deploy sensor nodes for precise positioning and coordinate calibration;

[0010] Step S13: The sensor nodes synchronously acquire the intensity and phase information of the broadband background signal;

[0011] Step S14: Using an improved ray tracing algorithm, a correction term based on geometric diffraction theory is introduced to calculate the diffraction field at the edge of the obstacle and generate a predicted electromagnetic environment base map.

[0012] Step S2: Perform joint analysis on the mixed signals acquired in real time by the sensor network to extract fingerprint features that characterize different signal sources.

[0013] Preferably, the specific implementation includes:

[0014] Step S21: Preprocess the signal data collected by the sensor and perform synchronous short-time Fourier transform on the time domain signal of each sensor node to obtain the time spectrum.

[0015] Step S22: Apply improved blind source separation preprocessing to the time-spectrum graph to initially separate signal components from different sources;

[0016] Step S23: Based on the initially separated signal components and their corresponding time spectra, calculate a set of multidimensional features in parallel to form a feature vector.

[0017] Step S3: Based on the extracted fingerprint features and type recognition results, dynamically define the effective boundary of the current shielding field and distinguish between effective signals and out-of-bounds interference within the shielding area.

[0018] Preferably, the specific implementation includes:

[0019] Step S31: Construct a spatial-feature matrix, calculate feature similarity, and perform spatial field clustering analysis based on feature similarity;

[0020] Step S32: Define a boundary decision function based on signal intensity gradient and feature similarity;

[0021] Step S33: Extract the boundary surface, mark the cross-boundary interference, and determine the effective shielding effectiveness monitoring area at the current moment.

[0022] Step S4: Simultaneously, a deep residual network is used to identify weak signals under strong interference and to estimate power compensation.

[0023] Preferably, the specific implementation includes:

[0024] Step S41: Construct a training sample library by inputting the mixed signals collected in the strong interference region into the improved deep residual network for network pre-training;

[0025] Step S42: Detect weak signals in areas of strong interference and estimate the test signal power in the corresponding areas;

[0026] Step S43: Calculate the theoretically required critical power of the shield to completely suppress the base station signal under the current environment;

[0027] Step S44: Compare the calculated minimum power with the nominal maximum output power of the shield, and determine the cause of the output failure.

[0028] Step S5: Output the multi-dimensional evaluation results of shielding effectiveness monitoring.

[0029] Preferably, the reliability of shielding effectiveness monitoring is evaluated from five aspects: model matching degree, signal classification confidence degree, boundary clarity, strong interference suppression confidence degree, and data integrity.

[0030] On the other hand, the technical solution of the present invention also includes a shielding effectiveness monitoring system, which performs a shielding effectiveness monitoring method and includes the following functional units:

[0031] 3D Electromagnetic Environment Module: Used to construct digital models of the electromagnetic environment for analysis;

[0032] Multi-source signal processing module: used to deconstruct and identify mixed signals acquired in real time;

[0033] Adaptive monitoring module: used to define the effective range of the current shielding measures;

[0034] Root cause diagnosis and attribution module: used for strong interference suppression and fault root cause diagnosis;

[0035] Multi-dimensional information output module: used to generate final quantifiable and reliable shielding effectiveness monitoring conclusions.

[0036] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0037] (1) This invention realizes continuous spatial coverage monitoring of complex spatial electromagnetic fields through three-dimensional digital twins and dynamic boundary delineation. At the same time, it uses feature similarity clustering and fusion decision function to generate a three-dimensional boundary surface that changes smoothly with the electromagnetic environment. This clearly excludes external "boundary crossing" interference from the internal performance evaluation and locks in internal abnormal signals, thereby solving the problem of monitoring range confusion and false alarm caused by interference crossing the boundary and ensuring the spatial accuracy of the evaluation object.

[0038] (2) This invention utilizes signal fingerprint analysis and intelligent separation to deepen the monitoring basis from the single signal strength that is susceptible to interference to the intrinsic multidimensional features of the signal, thereby enhancing the anti-aliasing and anti-interference capabilities and reducing the false judgment rate.

[0039] (3) Even under strong background interference (such as nearby high-power base stations), the present invention can still effectively identify weak shielding test signals and accurately diagnose whether the root cause of the “monitoring failure” phenomenon is “insufficient shielding performance” or “shortcomings in shielding equipment capabilities”. Attached Figure Description

[0040] Figure 1 This is a flowchart of the operation steps of a shielding effectiveness monitoring method provided in an embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram of the operating architecture of a shielding effectiveness monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0043] Terminology Definitions: In this invention, "insufficient shielding performance" refers to the failure of physical shielding structures such as shielding rooms and shielding covers to meet design standards due to material aging, structural gaps, or design defects. "Shortcomings in shielding device capabilities" refers to insufficient hardware / functional indicators of the shielding device, such as output power, frequency band coverage, and signal suppression capabilities, to meet the electromagnetic signal suppression requirements of the current monitoring scenario. "Insufficient shielding capability" is a general term encompassing both, referring to a state where any component of the shielding body or shielding device fails to achieve the expected shielding effect. "Shielding failure" refers to a critical state where insufficient shielding performance or shortcomings in shielding device capabilities reach a critical point, resulting in the complete inability to achieve the preset electromagnetic signal shielding effect.

[0044] Example 1:

[0045] like Figure 1 As shown in the figure, the shielding effectiveness monitoring method described in this embodiment includes the following operating steps:

[0046] Step S1: Modeling the spatial electromagnetic environment based on multi-sensor fusion and ray tracing.

[0047] In this embodiment, to address the signal dead zones caused by obstacles such as walls, the present invention first models the non-uniform electromagnetic environment of the monitoring area. Unlike conventional single-point signal strength measurement, the present invention deploys multiple heterogeneous sensor nodes (including broadband antennas, magnetic field probes, and electric field probes) both inside and outside the shielded area. It also utilizes a pre-acquired three-dimensional model of the building structure, combined with an improved ray tracing algorithm, to simulate the direct, reflected, and diffracted paths of electromagnetic waves in complex spaces.

[0048] For example, the specific implementation process includes:

[0049] Step S11: Obtain the three-dimensional structural data of the monitoring space through laser scanning or by importing a BIM model;

[0050] For example, this invention employs two acquisition methods for data acquisition, and performs preprocessing and model reconstruction on the acquired data. The two acquisition methods include:

[0051] Laser scanning: A 3D laser scanner is used to perform a panoramic scan of the shield and its internal space. The scanner emits laser pulses from multiple stations, receives the reflected signals, and records the 3D coordinates of a massive number of points. and reflection intensity This generates "point cloud" data, and the scanning process covers all critical structures, such as walls, doors and windows, ventilation waveguides, cable entry and exit points, and the surface of large equipment, avoiding voids caused by obstruction.

[0052] BIM Model Import: When a pre-built Building Information Model (BIM) exists, the complete model containing geometry and material properties is directly imported. The core is to extract the key surfaces and boundaries for electromagnetic calculations from the model and ensure that its geometric accuracy meets the requirements of ray tracing calculations. In this invention, the model error is set to be less than one-tenth of the wavelength of the main frequency band.

[0053] For example, data processing and model reconstruction include:

[0054] Point cloud registration and denoising: The point cloud data from multiple stations are registered with coordinates using a common target and merged into a unified point cloud in a unified coordinate system. Then, algorithms such as statistical filtering are applied to remove discrete noise points.

[0055] Surface Reconstruction and Simplification: Point cloud data is triangulated (e.g., using the Poisson reconstruction algorithm) to generate a continuous triangular mesh model describing the spatial boundary. The mesh is adaptively simplified while reducing the number of faces, thus balancing computational accuracy and speed, while preserving key geometric features (such as edges and corners). In the triangular mesh model, a triangular facet refers to the smallest geometric unit obtained after surface reconstruction of the 3D point cloud data of the monitoring space through triangulation; it is the basic unit constituting the 3D geometric model of the monitoring space. A material region refers to a spatial area formed by a set of multiple triangular faces with the same electromagnetic material properties; it is a secondary partitioning unit of triangular faces based on material properties. One material region corresponds to at least one triangular facet, and the same triangular facet belongs to only one material region.

[0056] Electromagnetic property assignment: Assigning electromagnetic property parameters to each triangular facet or material region in a continuous triangular mesh model, mainly including the relative permittivity. and conductivity These parameters can be obtained by consulting material databases or by sampling and measuring actual shielding materials (such as steel plates, conductive coatings, and concrete) using a handheld dielectric constant measuring instrument.

[0057] Step S12: Deploy sensor nodes for precise positioning and coordinate calibration;

[0058] For example, the sensor nodes in this invention employ a heterogeneous design, including a broadband omnidirectional antenna (for capturing the main signal), a high-sensitivity magnetic field probe (for detecting near-field and gap leaks), and a directional electric field probe. The number and location of the nodes are pre-simulated and optimized based on the size and complexity of the space, ensuring deployment in key areas such as corners where signal attenuation may be severe, door and window seams, and ventilation openings, and ensuring that some nodes are located outside the shield (as a reference). The node deployment principle is to form multi-angle, multi-layered coverage of the monitoring space, avoiding all nodes being in the same potential signal shadow area.

[0059] Precise positioning and coordinate calibration are performed: a global coordinate system consistent with the coordinate system of the 3D model is established in space, and high-precision measuring equipment (such as a total station or an indoor positioning system based on ultra-wideband (UWB)) is used to determine the 3D coordinates of the antenna phase center of each sensor node in the global coordinate system. The positioning error is set to be much smaller than the minimum wavelength of the monitoring frequency band (for example, for the 6GHz band, the wavelength is 5cm, and the positioning error is set to be controlled at the millimeter level). At the same time, the type, number and precise coordinates of each node are recorded to form a "sensor network coordinate mapping table".

[0060] Step S13: The node synchronously acquires the intensity and phase information of the broadband background signal (including base station signal);

[0061] Preferably, in this invention, all sensor nodes are equipped with a high-precision clock synchronization module (using GPS timing or a precision temperature-controlled crystal oscillator in conjunction with a synchronization trigger signal line) to ensure that all nodes simultaneously sample the electromagnetic environment with timestamps aligned at the millisecond or even microsecond level. With the shielding function disabled, all nodes start simultaneously and perform spectrum scanning according to a preset scanning frequency band (covering from critical low frequencies to high frequencies of 5G / 6G), collecting power spectral density, instantaneous amplitude, phase, and I / Q data of signals at various frequency points over a period of time (e.g., several minutes). The center frequency, bandwidth, and typical power level of known strong signal sources in the environment (such as 5G base station signals from specific operators, Wi-Fi signals) are specifically recorded. Each node performs preliminary preprocessing on the collected raw data (such as noise floor subtraction and gain calibration), and then aggregates data packets with precise timestamps and node location information to the central processing server via a wired (preferred) or wireless data transmission network.

[0062] Step S14: Run the improved ray tracing algorithm, introducing a correction term based on geometric diffraction theory (GTD) to more accurately calculate the diffraction field at the obstacle edge and generate a predicted electromagnetic environment base map.

[0063] For example, the core of the improved ray tracing algorithm used in this invention is the superposition of field strengths along multiple main propagation paths from the virtual signal source to each sensor. Specifically:

[0064] Path tracing: An improved ray tracing algorithm is used to track the propagation of each ray in the model. When a ray intersects a triangular facet, the direction of the reflected ray is calculated according to Snell's law. When a ray approaches a sharp edge or curved surface, an improved diffraction field calculation module is triggered.

[0065] Diffraction field calculation: A modified GTD model with an adaptive factor is adopted. For encountered diffraction edges, not only are the traditional GTD diffraction coefficients calculated, but also the local curvature radius of the edge is considered. and angle of incidence By querying a preset correction factor table or calculating a function in real time. The diffraction coefficient is dynamically corrected to more accurately simulate the diffraction effect of complex edges (such as curved door frames and ventilation grilles).

[0066] Multipath field strength vector superposition: Obtain the node position of each deployed sensor node (or virtual grid points set for generating a continuous base map), and use an improved ray tracing algorithm to collect all direct, reflected, and diffracted ray paths reaching the sensor nodes, based on the length of each path. The number of reflections and diffractions and their coefficients Calculate the complex field strength generated at the sensor node for each path according to the following formula. :

[0067]

[0068] in, For the source of strength, For wave number, Let m be the reflection coefficient of the m-th reflection. The standard diffraction coefficient for the nth diffraction is... As the corresponding adaptive correction factor, the complex field intensities of all N paths leading to the point are vector-superimposed to obtain the total predicted complex field intensity of the sensor node. Then, the predicted power density or equivalent received power of the sensor node is calculated.

[0069] Base map generation and verification: The above calculations are repeated on a 3D grid (voxels) of the monitoring space to generate a "3D distribution map of predicted signal intensity" at a specific frequency point, i.e., a "predicted electromagnetic environment base map." The predicted values ​​of the base map at the sensor node locations are compared with the reference signal intensity actually collected by the nodes. By minimizing the mean square error and other algorithms, the virtual source parameters or a small number of unknown environmental material parameters are optimized in reverse to achieve the best fit between the prediction model and the actual measurement. The calibrated model will have significantly enhanced prediction reliability, especially in the "blind spot" areas where no sensors are deployed.

[0070] Step S2: Perform joint analysis on the mixed signals acquired in real time by the sensor network to extract fingerprint features that can uniquely characterize different signal sources.

[0071] In this embodiment, to address the misjudgment caused by frequency band mismatch and signal interference, the present invention performs in-depth analysis on the monitored real-time signals, not only looking at the signal strength, but also extracting their unique time-frequency domain "fingerprint" features to distinguish signals from shielding effectiveness test signals, external base stations, other interference sources, and signals leaked from the shield itself.

[0072] For example, the implementation of the joint parsing and feature extraction process includes:

[0073] Step S21: Preprocess the signal data collected by the sensor, perform synchronous short-time Fourier transform (STFT) on the time domain signal of each sensor node, and obtain the time spectrum.

[0074] For example,

[0075] Data Access and Alignment: The central processing server receives synchronized data streams from all sensor nodes. First, based on the recorded high-precision timestamps and node IDs, all data is strictly aligned on the timeline to ensure that signals from different spatial points at the same time can be jointly analyzed.

[0076] Background noise estimation and filtering: For the data of each node, select a time period or frequency band where there is known no strong signal, and estimate its background noise power spectral density. An adaptive Wiener filter based on this noise estimation is applied to perform preliminary noise reduction on the full-band data, thereby improving the signal-to-noise ratio.

[0077] High-resolution time-frequency transformation: time-domain signal for each channel (sensor node) Time-frequency spectrograms were generated using synchronous short-time Fourier transform (STFT). To achieve high resolution, an adjustable window function is used: a longer window is used during signal stability to obtain high frequency resolution; and a shorter window is automatically switched at signal abrupt changes to obtain high time resolution. The transformation formula is:

[0078]

[0079] in For time-dependent Adaptive window functions for local signal characteristics, The window length parameter is determined by the instantaneous frequency change rate of the signal.

[0080] Background removal: Using the generated predicted electromagnetic environment base map, the time-frequency profiles of the predicted static background signals (mainly stable external base station signals) at each sensor node location are removed from the measured data. The mid-vector subtraction yields the residual time-spectrum diagram, which mainly contains the shielding test signal, the shielding leakage signal, and random interference. This key preprocessing greatly reduces the masking effect of strong background interference on subsequent analysis.

[0081] Step S22: Apply improved blind source separation (BSS) preprocessing to the time-spectrum graph to initially separate signal components from different sources;

[0082] For example, the implementation of blind source separation preprocessing based on space-time-frequency constraints includes:

[0083] Constructing the space-time-frequency tensor: Residual time-frequency spectrum of all sensor nodes within the same time period. Combined into a three-dimensional tensor ,in For time points, For frequency points, Given the number of sensors, this tensor simultaneously contains information about the signal in the time, frequency, and spatial domains.

[0084] Improved Blind Source Separation: An improved blind source separation algorithm based on spatiotemporal sparsity constraints is adopted, defining the existence of... An unknown source signal will affect the three-dimensional tensor. Decomposed into Time-frequency distribution of individual source signals and its corresponding spatial mixing vector An approximation of a linear combination. Further, spatial constraints and time-frequency masking constraints are introduced into the cost function. In the spatial constraints, the precise coordinates of the sensor nodes are used to constrain the mixing vector. This aligns with a planar wavefront model originating from a local direction in space, used to distinguish signals from different directions. In time-frequency mask constraints, energy accumulation regions are automatically identified in the time-frequency domain using clustering algorithms, and a binary mask is generated, forcing each source signal to be non-zero only within a specific time-frequency region, thus promoting separation.

[0085] Output preliminary separated components: Output the preliminary separated signal components and its corresponding time spectrum The estimated direction of arrival (DOA) of each component is used as the direct object for subsequent feature extraction.

[0086] Step S23: Using the initially separated signal components and their corresponding time spectra, calculate a set of multidimensional features in parallel to form a feature vector.

[0087] For example, parallel computation of a set of multidimensional features specifically includes:

[0088] Conventional feature extraction:

[0089] Instantaneous frequency trajectory: calculated by time spectrum The instantaneous frequency is obtained at the center of mass at each moment. And calculate its variance over a period of time. The smaller the variance, the more stable the signal carrier frequency;

[0090] Spectral kurtosis: Calculation of signal components The spectral kurtosis value, quantization signal pulse characteristics or non-Gaussianity;

[0091] Specific frequency band energy ratio: Calculate the proportion of the energy of this component in the preset shielded test signal frequency band to its total energy;

[0092] Modulation identification parameters: By analyzing the cyclic spectrum of the signal, its modulation type (e.g., ...) is estimated. ) and possible orders.

[0093] Weighted singular spectral entropy: for signal components Time-frequency spectrum matrix Perform Singular Value Decomposition (SVD): Obtain the singular value diagonal matrix, for the ... For each singular value, an approximate component of the original matrix is ​​reconstructed using its corresponding left and right singular vectors. Then, the energy and weighted probability represented by this reconstructed component are calculated, considering the magnitude of the singular value (representing the importance of the mode) and its actual energy contribution. Finally, the weighted singular spectral entropy is obtained based on the weighted probability; this value effectively characterizes the structural complexity of the signal in the time-frequency plane.

[0094] The feature vector is composed of all features, including instantaneous frequency trajectory, spectral kurtosis, specific band energy ratio, modulation identification parameters, and weighted singular spectral entropy, after normalization. .

[0095] Preliminary source type classification: The feature vector of each component... Input a pre-trained classifier (Support Vector Machine, SVM), and combine its DOA information to preliminarily determine the possible types of the output: "shielding effectiveness test signal", "known external interference (such as specific Wi-Fi)", "shielding harmonic leakage", "unknown transient interference", etc.

[0096] Step S3: Based on the extracted signal features and type identification results, dynamically define the effective boundary of the current shielding field and distinguish between effective signals and out-of-bounds interference within the shielding area.

[0097] In this embodiment, the "effective shielding monitoring area" and the "potential interference area" are dynamically defined based on the real-time electromagnetic environment, and fixed monitoring boundaries are no longer preset. The specific implementation process includes:

[0098] Step S31: Construct the spatial-feature matrix, calculate the feature similarity matrix, and perform spatial field clustering analysis based on feature similarity;

[0099] For example, the specific implementation process includes:

[0100] Constructing a spatial-feature matrix: For each location point in the monitoring space where sensors are deployed... The feature vectors of all signal components initially classified as "shielding effectiveness test signals" are extracted, and the mean vector of these feature vectors is calculated. The mean vector is used to represent the location point. The comprehensive characteristics of the test signal.

[0101] Calculate the feature similarity matrix: calculate all sensor point pairs The feature similarity between them, i.e., using Mahalanobis distance Similarity with cosine For sensor nodes and Overall similarity between The calculation is performed while considering both the distribution of eigenvalues ​​and the direction of the vector:

[0102]

[0103]

[0104]

[0105] in, For nodes The corresponding mean vector at that point The covariance matrix of the global eigenvectors. To adjust parameters, the overall similarity is... yes The scalar values ​​between the two points indicate that the closer the value is to 1, the more similar the two points are in terms of features, and the more likely they are to belong to the same region (such as an effective shielded internal region).

[0106] Spatial clustering: Using spectral clustering algorithms, the sensor point set is divided into several clusters. Ideally, points in areas effectively covered by strong shielding and with uniform internal test signal characteristics will cluster into a large "core cluster". Points located at the edge, whose signal characteristics are distorted due to external interference, or whose characteristics are abnormal due to severe obstruction by obstacles may form small edge clusters or outliers.

[0107] Step S32: Define a boundary decision function based on signal intensity gradient and feature similarity.

[0108] For example, at the edge of the shielded area, the test signal strength gradient is large, and its characteristics are highly similar to those inside, while cross-boundary interference exhibits a gentle gradient and abrupt changes in characteristics. The specific implementation process includes:

[0109] Spatial interpolation generates feature fields: utilizing the feature mean vector of the "core cluster" sensor points. Its precise coordinates are obtained, and Kriging spatial interpolation is used to interpolate on a three-dimensional grid across the entire monitoring space to generate a continuous test signal characteristic distribution field. .

[0110] Calculate the gradient field and characteristic divergence field: Calculate the spatial gradient field of the test signal power obtained by integrating the time-spectral energy of the signal components. This gradient value is expected to be large near the effective shielding boundary; at each point in space... Calculate its interpolation eigenvector. Average eigenvector of the "core cluster" Improved Bartholomew's Distance This is used to quantify the decay of feature similarity.

[0111]

[0112] in This is for traversing the feature dimensions.

[0113] The dynamic boundary decision function is formed by fusion: the decision function value at any point in the space is defined as:

[0114]

[0115] in, , These are adaptive weights. The first term is a normalized power gradient term, used to capture the physical boundaries of signal strength. The second term is an exponential decay term based on feature similarity, used to capture the logical boundaries of signal features. This is the scale parameter. It is negatively correlated with the total multipath richness of the current environment, because when multipath is severe, relying solely on intensity gradients to determine boundaries is unreliable. It is positively correlated with the average classification confidence of the test signal components output by the classifier. The higher the confidence, the more dependent it is on the feature consistency judgment boundary.

[0116] Step S33: Extract the boundary surface, mark the cross-boundary interference, and determine the effective shielding effectiveness monitoring area at the current moment.

[0117] For example, the specific implementation process includes:

[0118] Threshold segmentation and isosurface extraction: Set a dynamic threshold The value is automatically determined based on the statistical distribution of the decision function values ​​of points within the "core cluster" (e.g., the mean minus a certain number of standard deviations). This is done in 3D mesh data. In this context, the Moving Cubes algorithm is used to extract the function value equal to... The isosurface is the dynamic boundary surface that is initially determined.

[0119] Surface smoothing and optimization: The extracted original isosurface is smoothed by filtering (such as Laplacian smoothing) and constrained by a 3D geometric model to ensure that the boundary surface does not pass through insurmountable obstacles such as solid walls.

[0120] Cross-boundary interference determination: For identified interference components that are not "test signals" (such as "known external interference"), check the origin direction of their DOA estimation or the sensor location where their signal strength peak occurs. If the origin direction clearly originates from outside the aforementioned dynamic boundary surface, or if its main energy appears on a sensor node outside the boundary surface, it is determined to be "external cross-boundary interference," recorded, but not included in the internal shielding effectiveness evaluation. If the interference source is determined to be located within the boundary surface, it is marked as "internal potential leakage or interference," which requires special attention in subsequent steps.

[0121] Output: The final output is a dynamic, smooth 3D boundary surface model, and a spatial relationship report that marks the positional relationship between various interference signals and the boundary surface. The internal space of the surface is defined as the effective shielding effectiveness monitoring area at the current moment.

[0122] Step S4: Use a deep residual network to identify weak signals under strong interference and estimate power compensation.

[0123] In this embodiment, the present invention addresses the difficulty in detecting weak shielding test signals under background interference such as strong base station signals, and the problem of distinguishing between "shielding failure caused by insufficient shielding performance or limitations in shielding equipment capabilities" and simply "insufficient shielding performance." The invention utilizes sensor data within a defined effective monitoring area, focusing on analyzing node data located in areas with strong interference (such as the side near the base station). The specific implementation process includes:

[0124] Step S41: Construct a training sample library by inputting the mixed signals (strong base station signals + possible weak test signals) collected in the strong interference area into a pre-trained deep residual network (ResNet) for network pre-training.

[0125] For example, by utilizing historical data accumulation, under the conditions of known shield performance (nominal maximum power) and known external base station signal strength, a large number of time-domain signals received by sensor nodes in different scenarios are collected. For each sample, according to its corresponding real situation (judged manually or with the assistance of high-precision instruments), two labels are marked: a binary label indicating whether a shield test signal exists (1 for existence, 0 for non-existence), and a continuous value label indicating, if it exists, its pure test signal power value (which can be obtained by measuring it alone at close range to the shield or by estimating it through known attenuation relationships).

[0126] Furthermore, each sample undergoes preprocessing to obtain a residual signal, which is then converted into a time-spectrum graph and used as input to the neural network. Simultaneously, the known base station signal power at the corresponding sensor node is used as an auxiliary input feature. An improved deep residual network (ResNet) model with dual-branch output is constructed. The inputs are the time-spectrum graph (two-dimensional matrix) and the base station power (scalar, normalized and used as an additional channel or fully connected layer input). The backbone network is a ResNet-18 or similar structure used to extract deep features. After the penultimate layer, it splits into two branches: Branch 1 (existence judgment): outputs a 2D vector through a fully connected layer and softmax, representing the probability of the test signal "not existing" and "existing". Branch 2 (power regression): outputs a scalar through a fully connected layer, representing the estimated test signal power. The model is then trained using a combined loss function on a large number of historical samples until convergence, resulting in a pre-trained model.

[0127] Step S42: Detect weak signals in areas of strong interference and estimate the test signal power in the corresponding areas.

[0128] For example, the specific implementation includes: based on the output dynamic boundary surface and the predicted base map, identifying the region located outside the shield, close to the direction of a known strong base station signal, and within the dynamic boundary (i.e., the region where strong interference may intrude), and selecting all sensor nodes within this region. For each selected node, the time-domain signal within the current monitoring period is acquired, and after the same preprocessing as in the training phase (background removal, time-frequency transformation), a time-spectrum map is generated. Simultaneously, the currently measured base station signal power of the node is acquired. The time-spectrum map and the base station signal power are input into a pre-trained improved ResNet model. The network outputs two results: the probability of the test signal presence and the estimated pure test signal power. Finally, inference is performed on multiple nodes within the strong interference region, and the probability of the test signal presence of each node is statistically analyzed. If the pure test signal power of more than half of the nodes is greater than the color interaction threshold, it is determined that a weak test signal exists in the region. Then, the median of the pure test signal powers of these determined nodes is taken as the final estimate of the test signal power in the region.

[0129] Step S43: Calculate the theoretically required critical power of the shield to completely suppress the base station signal under the current environment.

[0130] For example, the specific implementation includes: subtracting the estimated received power from the known transmitted power of the shielded test signal (set and recorded at the start of the test) to obtain the path loss for the area. To completely suppress the base station signal in this area, the interference signal power generated by the shield in this area must at least reach a level comparable to the base station signal. Considering a certain safety margin, the critical power that the shield needs to achieve in this area is the median of the measured base station power in this area plus the safety margin. Since the base station signal and the test signal experience different spatial paths, the critical power here is the interference field strength that the shield needs to create in this area, rather than directly compensating for the power of the test signal. This invention considers that the shield signal also experiences path loss from the transmission point to this area. Assuming that the path loss from the shield to this area is similar to that of the test signal (because the frequency and location are similar), the minimum output power required by the shield is the critical power plus the path loss for this area.

[0131] Step S44: Compare the calculated minimum power with the nominal maximum output power of the shield and determine the cause of failure.

[0132] For example, the specific implementation includes: the invention compares the calculated minimum power with the nominal maximum output power of the shield. If the minimum power is less than or equal to the nominal maximum output power of the shield, and the test signal power in that area is lower than a set expected value (e.g., more than 20 dB lower than in an unobstructed area), it is determined to be "shielding failure caused by insufficient shielding performance or a shortcoming in the shielding device's capabilities," meaning the shielding device is functioning normally but there is leakage in the shielding body. If the minimum power is greater than the nominal maximum output power of the shield, it is determined to be "a shortcoming in the shielding device's capabilities," meaning the existing shield cannot generate sufficient field strength in that area to suppress strong external interference, rather than the shielding body itself failing. The final output includes the detection results (presence or absence, estimated power) of the weak test signal in the strong interference area, the theoretical critical suppression power, and the final conclusion regarding the cause of failure.

[0133] Step S5: Output quantifiable and reliable multi-dimensional evaluation results of shielding effectiveness monitoring.

[0134] In this embodiment, M representative evaluation points are uniformly selected within the "effective monitoring area" inside the defined dynamic boundary surface. These points can be actual sensor node locations or virtual grid points. For each evaluation point, the traditional shielding effectiveness, spatial uniformity index, and characteristic stability index are calculated to obtain the power compensation discrimination result. The comprehensive shielding effectiveness is also calculated, and the overall reliability of the monitoring is evaluated and output from five aspects: model matching degree, signal classification confidence, boundary clarity, strong interference suppression confidence, and data integrity.

[0135] Example 2

[0136] like Figure 2 As shown in the figure, the shielding effectiveness monitoring system described in this embodiment includes the following functional modules:

[0137] The 3D electromagnetic environment module, serving as the system's foundational sensing layer, is responsible for constructing a digital model of the electromagnetic environment that closely mirrors the physical world. It accurately senses and digitally reproduces the complete propagation pattern of electromagnetic waves interacting with obstacles (walls, equipment) within complex shielded spaces. By integrating 3D geometric scanning, multi-type sensor networks, and high-precision ray tracing algorithms, this module not only provides a spatial "geometric base map" but also generates a "predicted electromagnetic field strength distribution base map" that includes direct, reflected, and diffracted effects. This base map accurately predicts the residual field strength in signal "dead zones," providing a calibrated and reliable spatiotemporal reference for all subsequent analyses, fundamentally solving the monitoring blind spots and distortion problems caused by complex spatial layouts.

[0138] Multi-source signal processing module: Used to "deconstruct" and "identify" mixed signals acquired in real time. By highly separating components from different sources in the mixed signal (such as shielded test signals, external base station signals, shield leakage harmonics, and random noise), it extracts highly discriminative multi-dimensional "identity fingerprints." Through the introduction of spatiotemporal-frequency joint blind source separation and innovative weighted singular spectral entropy features, this module surpasses the limitations of traditional intensity-based judgments, accurately identifying target test signals and preliminarily classifying various types of interference. This provides a clean signal source and rich feature data for accurately evaluating shielding effectiveness and distinguishing interference types, effectively avoiding misjudgments caused by signal mixing and frequency band mismatch.

[0139] Adaptive Monitoring Module: Used to define the effective range of current shielding measures. By dynamically and intelligently delineating the physical boundaries of the "effective shielding monitoring zone" based on real-time signal characteristics and intensity changes, it automatically identifies "boundary-crossing" interference signals. This module utilizes feature similarity clustering and a fusion decision function, rather than pre-setting fixed boundaries, to generate a three-dimensional boundary surface that smoothly changes with the electromagnetic environment. It explicitly excludes external "boundary-crossing" interference from the internal effectiveness assessment, while simultaneously locking onto internal anomalous signals, resolving the issues of monitoring range confusion and false alarms caused by interference crossing boundaries, and ensuring the spatial accuracy of the assessment object.

[0140] Root Cause Diagnosis and Attribution Module: Used for strong interference suppression and fault root cause diagnosis. Even under strong background interference (such as from nearby high-power base stations), this module can effectively identify weak shielding test signals and accurately diagnose whether the root cause of "monitoring failure" is "insufficient shielding performance" or "weakness in the shielding equipment." It directly estimates the true power of the test signal through a deep learning model and inversely calculates the theoretically required suppression power, improving the accuracy of monitoring conclusions.

[0141] Multi-dimensional information output module: As the system's comprehensive evaluation and report generation layer, this module generates final quantifiable and reliable monitoring conclusions. By integrating information from all preceding modules, it calculates a comprehensive shielding effectiveness index from multiple dimensions, including field strength, spatial uniformity, and signal characteristic consistency, and simultaneously generates a "credibility level" characterizing the reliability of the results. This module no longer outputs a single, isolated effectiveness value, but instead provides a comprehensive report including "Comprehensive Effectiveness Value," "Spatial Uniformity Index," "Characteristic Stability Index," and "High / Medium / Low Credibility" labels, providing refined decision-making basis for the optimization and rectification of the shielding structure.

[0142] It should be noted that, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0143] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring shielding effectiveness, characterized in that: The following steps are included: Step S1: Modeling the spatial electromagnetic environment based on multi-sensor fusion and ray tracing; The modeling includes: Step S11: Obtain the three-dimensional structural data of the monitoring space through laser scanning or importing a BIM model, and preprocess the acquired data and reconstruct the model; Step S12: Deploy sensor nodes for precise positioning and coordinate calibration; Step S13: The sensor nodes synchronously acquire the intensity and phase information of the broadband background signal; Step S14: Using an improved ray tracing algorithm, a correction term based on geometric diffraction theory is introduced to calculate the diffraction field at the edge of the obstacle and generate a predicted electromagnetic environment base map. Step S2: Perform joint analysis on the mixed signals acquired in real time by the sensor network to extract fingerprint features that characterize different signal sources; Step S3: Based on the extracted fingerprint features and type recognition results, dynamically define the effective boundary of the current shielding field and distinguish between effective signals and out-of-bounds interference within the shielding area; Step S4: Simultaneously, a deep residual network is used to identify weak signals under strong interference and to estimate power compensation. Step S5: Output the multi-dimensional evaluation results of shielding effectiveness monitoring.

2. The shielding effectiveness monitoring method according to claim 1, characterized in that: Preprocessing and model reconstruction of the acquired data include: Point cloud registration and denoising: Point cloud data from multiple stations are registered using a common target and merged into a unified point cloud in a unified coordinate system, removing discrete noise points; Surface reconstruction and simplification: Triangulation is performed using point cloud data to generate a continuous triangular mesh model describing the spatial boundary, and the mesh is adaptively simplified. Electromagnetic property assignment: Based on the geometric division of the continuous triangular mesh model, each triangular facet in the model is first assigned basic electromagnetic property parameters. Then, according to the material properties of the triangular facets, the electromagnetic property parameters of the material region composed of the set of triangular facets of the same material are uniformly calibrated and assigned. The electromagnetic property parameters include relative permittivity and conductivity.

3. The shielding effectiveness monitoring method according to claim 2, characterized in that: The improved ray tracing algorithm used in step S14 includes: An improved ray tracing algorithm is used to track the propagation process of each ray in the model. When the ray intersects with the triangular facet, the direction of the reflected ray is calculated according to Snell's law. When the ray approaches a sharp edge or curved surface, an improved diffraction field calculation is triggered. The GTD correction model with an adaptive factor is adopted. For the diffraction edge encountered, not only is the traditional GTD diffraction coefficient calculated, but the diffraction coefficient is also dynamically corrected by querying a preset correction factor table or a real-time calculation function based on the local curvature radius and incident angle of the edge. The location of each deployed sensor node is obtained, and an improved ray tracing algorithm is used to collect all direct, reflected, and diffracted ray paths reaching the sensor node. Based on the length of each path, the number of reflections and diffractions experienced, and the coefficients, the complex field strength generated by each path at the sensor node is calculated. Repeat the above calculations on a three-dimensional grid in the monitoring space to generate a predicted electromagnetic environment base map of the entire space at a specific frequency.

4. The shielding effectiveness monitoring method according to claim 3, characterized in that: The implementation of step S2 includes: Step S21: Preprocess the signal data collected by the sensor and perform synchronous short-time Fourier transform on the time domain signal of each sensor node to obtain the time spectrum. Step S22: Apply improved blind source separation preprocessing to the time-spectrum graph to initially separate signal components from different sources; Step S23: Based on the initially separated signal components and their corresponding time spectra, calculate a set of multidimensional features in parallel to form a feature vector.

5. The shielding effectiveness monitoring method according to claim 4, characterized in that: The improved blind source separation preprocessing applied in step S22 includes: The residual time-frequency spectra of all sensor nodes within the same time period are combined into a three-dimensional tensor, which simultaneously contains the time domain, frequency domain, and spatial domain information of the signal. An improved blind source separation algorithm based on space-time-frequency sparsity constraints is adopted, defining the existence of... An unknown source signal is used to decompose a three-dimensional tensor into an approximation of the time-frequency distribution of multiple source signals and the linear combination of their corresponding spatial mixing vectors. Spatial constraints and time-frequency masking constraints are introduced into the cost function; The output includes the initially separated signal components and their corresponding time-frequency spectra, as well as the estimated direction of arrival for each component.

6. The shielding effectiveness monitoring method according to claim 5, characterized in that: The implementation of step S3 includes: Step S31: Construct a spatial-feature matrix, calculate feature similarity, and perform spatial field clustering analysis based on feature similarity; Step S32: Define a boundary decision function based on signal intensity gradient and feature similarity; Step S33: Extract the boundary surface, mark the cross-boundary interference, and determine the effective shielding effectiveness monitoring area at the current moment.

7. The shielding effectiveness monitoring method according to claim 6, characterized in that: Step S32 includes: By utilizing the characteristic mean vector and its precise coordinates of the "core cluster" sensor points, and employing the Kriging space interpolation method, interpolation is performed on a three-dimensional grid across the entire monitoring space to generate a continuous test signal characteristic distribution field. Calculate the spatial gradient field of the test signal power obtained by integrating the time-spectrum energy of the signal components; Calculate the improved Bach distance between the interpolated eigenvector at each point in space and the average eigenvector of the "core cluster" sensor points; The normalized power gradient term and the exponential decay term based on feature similarity are fused to form a dynamic boundary decision function.

8. The shielding effectiveness monitoring method according to claim 7, characterized in that: The implementation of step S4 includes: Step S41: Construct a training sample library and input the mixed signals collected from the strong interference region into the improved deep residual network for network pre-training; Step S42: Detect weak signals in areas of strong interference and estimate the test signal power in the corresponding areas; Step S43: Calculate the theoretically required critical power of the shield to completely suppress the base station signal under the current environment; Step S44: Compare the calculated minimum power with the nominal maximum output power of the shield, and determine the cause of the output failure.

9. The shielding effectiveness monitoring method according to claim 8, characterized in that: Step S5 includes evaluating the credibility of shielding effectiveness monitoring from five aspects: model matching degree, signal classification confidence degree, boundary clarity, strong interference suppression confidence degree, and data integrity.

10. A shielding effectiveness monitoring system, characterized in that, The shielding effectiveness monitoring method as described in any one of claims 1-9 includes the following functional units: 3D Electromagnetic Environment Module: Used to construct digital models of the electromagnetic environment for analysis; Multi-source signal processing module: used to deconstruct and identify mixed signals acquired in real time; Adaptive monitoring module: used to define the effective range of the current shielding measures; Root cause diagnosis and attribution module: used for strong interference suppression and fault root cause diagnosis; Multi-dimensional information output module: used to generate final quantifiable and reliable shielding effectiveness monitoring conclusions.

Images