An Electromagnetic Emission Testing Method for Smart Home Appliances

Abstract

This invention discloses an electromagnetic emission testing method for smart home appliances. By controlling multiple modules to operate simultaneously according to preset logic to simulate a dynamic interaction process, the method includes: real-time acquisition of electromagnetic signals and frequency domain conversion to obtain the spectrum distribution; identification of the target frequency band and its peak intensity caused by energy convergence due to nonlinear superposition; obtaining the superimposed waveform sequence and classifying the nonlinear superposition mode; identifying abnormal interference modes by combining the peak intensity and mode classification results; correlating the abnormal frequency band with the module operation logs, analyzing the interaction timing and logic to determine the cause; applying filtering or control strategies based on the cause, verifying whether the peak value decreases, and generating a report. This invention achieves accurate diagnosis and root cause location of concealed transient interference by identifying the energy convergence frequency band, classifying nonlinear superposition modes, tracing the cause of interference, directional control, and closed-loop verification. This improves testing depth, optimizes rectification efficiency, and enhances the predictability of product electromagnetic compatibility design.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic emission testing technology, and in particular to an electromagnetic emission testing method for smart home appliances. Background Technology

[0002] With the increasing prevalence of smart home appliances in modern households and their ever-increasing functional integration, electromagnetic compatibility (EMC) has become a core factor affecting reliable operation, user experience, and even market access. The industry's R&D and testing focus is shifting from static compliance checks on individual devices to dynamic EMC evaluation of complex systems under real-world operating conditions. This is because smart refrigerators, air conditioners, washing machines, and other products typically integrate multiple functional units such as compressors, motors, display units, and wireless communication modules. These units do not operate in isolation when performing complex tasks but are in a state of continuous interaction and collaboration.

[0003] The most prominent and unresolved problem currently facing the industry in testing electromagnetic emissions from smart home appliances is that existing mainstream testing methods and evaluation systems generally cannot accurately capture and evaluate the nonlinear superposition effects caused by signal interaction during the dynamic collaborative operation of multi-functional modules, and the resulting energy convergence phenomena in specific frequency bands. Current technologies typically employ methods such as testing the device in a laboratory environment under several standardized steady-state modes (e.g., single-module full-load operation) or performing a wideband scan of the overall radiation to check for exceeding limits. While these methods can detect some obvious and persistent interference, their fundamental flaw lies in treating the device as a static or linear system, ignoring the complex mutual modulation, harmonic generation, and energy accumulation processes of different signals in the time and frequency domains when multiple modules operate concurrently. Therefore, existing solutions often miss transient but potentially high-intensity interference peaks that only occasionally appear under specific module combinations and operating sequences. They also fail to explain the specific sources and interaction mechanisms of these interferences, leading to vague problem identification. Remedial measures often remain at the "trial and error" level, such as blindly strengthening shielding or adjusting filter parameters, which are ineffective and costly. Summary of the Invention

[0004] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of the prior art in effectively capturing and evaluating specific frequency band energy convergence interference and its root cause caused by nonlinear signal interaction in the dynamic collaborative working scenario of multiple modules of smart home appliances. The present invention provides an electromagnetic emission testing method for smart home appliances, which can accurately diagnose and locate the root cause of hidden transient interference by constructing dynamic interaction scenarios, identifying energy convergence frequency bands, classifying nonlinear superposition modes, tracing the cause of interference and performing closed-loop verification, thereby improving the depth of testing, optimizing rectification efficiency and enhancing the predictability of product electromagnetic compatibility design.

[0005] To address the aforementioned technical problems, this invention provides an electromagnetic emission testing method for smart home appliances. This method controls multiple functional modules of the smart home appliance under test to operate simultaneously according to a preset interaction logic, simulating the dynamic interaction process under actual working conditions. The method includes the following steps:

[0006] During the dynamic interaction process, electromagnetic signal data generated by smart home appliances are collected in real time, and the electromagnetic signal data is frequency domain converted to obtain the corresponding spectrum distribution.

[0007] Analyze the spectral distribution to identify one or more target frequency bands where energy converges due to the nonlinear superposition of signals from multiple modules, and record the peak intensity of each target frequency band;

[0008] The electromagnetic signal in the target frequency band is sampled in the time domain to obtain the corresponding superimposed waveform sequence. Based on the characteristics of the superimposed waveform sequence, the nonlinear superposition mode is classified.

[0009] By combining the peak intensity of the target frequency band and the classification results of nonlinear superposition modes, nonlinear superposition modes that meet the preset abnormal characteristics are identified and judged as abnormal interference modes with potential risks.

[0010] After determining that the interference characteristics are abnormal, the operation logs of the functional modules that generate frequency band signals during the dynamic interaction process are associated with the target frequency band. By analyzing the signal interaction timing and logic between modules, the cause of module interaction leading to energy convergence is determined.

[0011] Based on the causes of module interaction, corresponding filtering or control strategies are applied to electromagnetic signal data or the operating mode of smart home appliances. Then, it is verified whether the peak intensity of the target frequency band is reduced, and a test report is generated based on the verification results.

[0012] In one embodiment of the present invention, identifying one or more target frequency bands where energy convergence occurs due to the nonlinear superposition of signals from multiple modules includes:

[0013] Traverse each frequency point of the spectrum distribution and generate a continuous spectrum energy profile based on its signal energy value;

[0014] By analyzing the changing trends of the spectral energy profile, the starting frequency point where the energy change rate changes from flat to increasing, and the ending frequency point where the rate of change changes from increasing to flat or decreasing, can be identified.

[0015] The continuous frequency bands between each set of starting and ending frequency points are designated as potential energy convergence areas and used as candidate frequency bands.

[0016] The synchronous operation sequence of each functional module of the smart home appliance is retrieved during the dynamic interaction process. The time period when the energy of the candidate frequency band increases significantly is analyzed to see if there is a synchronous or related relationship with the time period when two or more functional modules enter a specific working state. The candidate frequency bands with synchronous or related relationships are finally determined as the target frequency bands generated by the nonlinear superposition of signals from multiple modules.

[0017] In one embodiment of the present invention, trend enhancement of the spectral energy profile includes:

[0018] By analyzing the energy fluctuation amplitude and frequency of the spectral energy profile in different frequency ranges, we can distinguish between the slow-changing component reflecting the macroscopic trend and the fast-changing component reflecting noise interference.

[0019] Based on the distribution characteristics of slowly varying components in the spectrum, a smoothing window that matches the width of the local energy trend is dynamically generated. A wider smoothing window is used in the wide frequency band where the energy trend is gentle, and a narrower smoothing window is used in the narrow frequency band where the energy change is potentially drastic.

[0020] By using a smooth window to convolve the spectral energy profile, the rapidly varying components are suppressed, while the macroscopic transition features of energy convergence and decay characterized by the slowly varying components are preserved and enhanced, resulting in an optimized and smooth spectral energy profile.

[0021] In one embodiment of the present invention, time-domain sampling is performed on the electromagnetic signal of the target frequency band to obtain the corresponding superimposed waveform sequence, including:

[0022] For a target frequency band, the original time-domain signal segment corresponding to the frequency band is extracted from the real-time acquired electromagnetic signal data;

[0023] Based on the switching time of the operating state of the associated functional module that generates the target frequency band signal, multiple synchronous trigger signals are generated;

[0024] Using each synchronous trigger signal as a reference point, a signal window of fixed time length is extracted from the original time domain signal segment to obtain multiple time domain waveform units that are aligned with each other.

[0025] All time-domain waveform units are arranged according to the order of events corresponding to their synchronization trigger signals to form a superimposed waveform sequence that characterizes how the target frequency band signal repeatedly appears with associated events during dynamic interaction.

[0026] In one embodiment of the present invention, the classification result of the nonlinear superposition mode includes at least:

[0027] Steady-state resonance mode: The waveform units in the superimposed waveform sequence have highly similar envelope shapes and stable repetition periods, indicating the existence of continuous, periodic signal superposition in phase;

[0028] Transient impact mode: One or more waveform units with sharp pulse-like envelopes appear intermittently in the superimposed waveform sequence. The amplitude of these waveform units is significantly higher than that of other units in the sequence, and their occurrence time is highly correlated with the power-on / off or load change events of a specific module, indicating the existence of transient energy injection and superposition caused by drastic state changes.

[0029] Random modulation mode: The envelope shape, amplitude and period of each waveform unit in the superimposed waveform sequence all show irregular fluctuations, indicating that there are multiple independent source signals that are asynchronously superimposed in an unpredictable manner.

[0030] In one embodiment of the present invention, identifying a nonlinear superposition pattern that conforms to preset abnormal characteristics includes:

[0031] Establish an anomaly feature knowledge base, where each knowledge unit defines a preset anomaly feature, including: nonlinear superposition mode category, target frequency band peak intensity condition, and indicated risk type;

[0032] Generate a feature description of the current mode, including: the category of the nonlinear superposition mode and the real-time peak intensity value of the target frequency band;

[0033] Matching feature descriptions with anomaly feature knowledge bases includes: filtering knowledge units with consistent categories based on the categories in the feature descriptions; and verifying whether the real-time peak intensity values ​​in the feature descriptions meet the intensity conditions defined by the filtered knowledge units.

[0034] If the verification is successful, the nonlinear superposition mode is determined to be an abnormal interference mode that meets the preset abnormal characteristics, and a judgment report containing risk type and feature description is generated.

[0035] In one embodiment of the present invention, the causes of module interactions leading to energy convergence are determined by analyzing the signal interaction timing and logic between modules, including:

[0036] Key time feature points of waveform envelope are extracted from the superimposed waveform sequence of abnormal interference mode, and precise timestamps of state switching events are extracted from the operation log of functional module. The key time feature points and precise timestamps are aligned on a unified time axis.

[0037] For each key time feature point, find the corresponding state transition event and mark the functional module that executes the state transition event as a suspicious module associated with the current feature point;

[0038] During the statistical dynamic interaction process, the co-occurrence relationship and co-occurrence frequency among different functional modules marked as suspicious modules were analyzed.

[0039] Based on co-occurrence relationships and co-occurrence frequencies, one or more functional modules that repeatedly co-occur in time and whose co-occurrence times correspond to key time feature points are identified as the module interaction causes leading to energy convergence.

[0040] In one embodiment of the present invention, the corresponding filtering strategy includes:

[0041] Frequency domain notch filtering: Applying a band-stop filter with a specific center frequency and suppression depth to the signal path for the target frequency band;

[0042] Time-domain gated filtering: Selectively attenuating or shielding electromagnetic signal data within the critical time window corresponding to the abnormal interference mode.

[0043] In one embodiment of the present invention, the corresponding control strategy includes:

[0044] Module work timing staggered: Adjust the start time, stop time or work cycle of the functional modules involved in the module interaction to stagger the high load operation periods of each module in time;

[0045] Module power ramp-up rate control: Reduce the power change rate of at least one functional module during state switching in the module interaction cause.

[0046] The technical solution of the present invention has the following advantages over the prior art:

[0047] The electromagnetic emission testing method for smart home appliances described in this invention achieves accurate and in-depth diagnosis of the electromagnetic interference characteristics of smart home appliances under real-world operating conditions, rather than merely determining compliance. By identifying energy convergence frequency bands and classifying superposition modes, it can reveal hidden transient interference risks that traditional methods cannot detect. By establishing a traceability path from interference phenomena to the specific module interaction causes, it greatly improves the efficiency and accuracy of problem localization, directly addressing the core issues and avoiding blind rectification. Ultimately, the entire closed-loop process transforms electromagnetic compatibility testing from a passive, post-event detection into a proactive, predictable, and optimizable design phase, thereby significantly shortening product development cycles, reducing post-modification costs and market risks caused by electromagnetic compatibility issues, and fundamentally improving the intrinsic reliability and quality of smart home appliance products. Attached Figure Description

[0048] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:

[0049] Figure 1 This is a flowchart of the electromagnetic emission testing method for smart home appliances according to the present invention;

[0050] Figure 2 This is a flowchart of the steps in this invention to identify the target frequency band where energy convergence occurs due to the nonlinear superposition of signals from multiple modules;

[0051] Figure 3 This is a flowchart illustrating the steps of the present invention to obtain a corresponding superimposed waveform sequence by time-domain sampling of electromagnetic signals in the target frequency band.

[0052] Figure 4 This is a flowchart of the steps in the present invention to identify a nonlinear superposition pattern that conforms to preset abnormal characteristics;

[0053] Figure 5 This invention provides a flowchart of steps to determine the causes of module interactions that lead to energy convergence by analyzing the signal interaction timing and logic between modules. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0055] Reference Figure 1 As shown, this invention discloses an electromagnetic emission testing method for smart home appliances, which constructs a closed-loop dynamic testing and diagnostic method from a principle level, including the following steps:

[0056] First, this method controls multiple functional modules of the smart home appliance under test to operate simultaneously according to a preset interaction logic, simulating the dynamic interaction process under actual working conditions. This control process can be implemented through a pre-programmed test script, which defines the timing relationships of each functional module's start-up, stop, and mode switching. For example, the compressor, fan, and display screen of a smart refrigerator can be set to work alternately or concurrently within a specific time period to simulate daily user operations such as opening and closing doors and adjusting temperatures. As another implementation method, a sequence of commands can be sent to the smart home appliance through an external control interface to drive its internal modules to operate collaboratively according to a predetermined process. This fundamentally ensures that the test signal source can simulate the necessary conditions for generating complex interference in the real world.

[0057] During dynamic interaction, electromagnetic signal data generated by smart home appliances is acquired in real time, and frequency domain transformation is performed on the electromagnetic signal data to obtain the corresponding spectral distribution. Electromagnetic signal data can be acquired using a broadband antenna or near-field probe and digitized by a spectrum analyzer or oscilloscope. The acquired time-domain signal data is then sent to a processor and converted into frequency-domain data using algorithms such as Fast Fourier Transform (FFT), thereby obtaining the signal energy distribution at different frequency points, i.e., the spectral distribution. Through real-time acquisition and frequency domain transformation, continuous spectral snapshots are obtained during the dynamic process, allowing not only observation of the spectrum but also targeted identification of target frequency bands where energy convergence occurs due to nonlinear superposition. This differs from simply searching for out-of-range frequencies; rather, it actively detects the phenomenon of abnormal concentration of signal energy in the frequency domain, which is a typical manifestation of increased interference caused by dynamic interaction.

[0058] Subsequently, the spectral distribution is analyzed to identify one or more target frequency bands where energy convergence occurs due to the nonlinear superposition of signals from multiple modules, and the peak intensity of each target frequency band is recorded. This identification process can be based on a preset energy threshold. For example, a global energy threshold can be set, and when the energy of a frequency band in the spectrum is consistently higher than this threshold, it is marked as a potential energy convergence region. Alternatively, the spectrum can be manually observed, and the presence of abnormal energy spikes can be determined based on experience. The highest energy value of the identified target frequency band will be recorded as the peak intensity of that band.

[0059] Furthermore, the electromagnetic signals in the target frequency band are sampled in the time domain to obtain the corresponding superimposed waveform sequence. Based on the characteristics of the superimposed waveform sequence, the nonlinear superposition modes are classified. For the identified target frequency band, signal segments corresponding to that frequency band can be extracted from the original acquired time-domain signal. These signal segments can be simply arranged in chronological order to form a continuous waveform sequence. Subsequently, by observing the overall shape, periodicity, amplitude changes, and other macroscopic characteristics of the waveform sequence, it is classified into different nonlinear superposition modes.

[0060] Based on this, by combining the peak intensity of the target frequency band with the classification results of nonlinear superposition modes, nonlinear superposition modes that meet the preset abnormal characteristics are identified and judged as anomalous interference modes with potential risks. By combining peak intensity and mode classification results to identify anomalous interference modes, its judgment logic goes beyond a single threshold comparison and incorporates intelligent recognition of interference behavior characteristics. It can discover interference whose intensity may not be absolutely excessive but whose mode is abnormally dangerous (such as having high transient energy).

[0061] After determining that the interference characteristics are abnormal, the operational logs of the target frequency band and the functional modules that generate signals in that frequency band during dynamic interaction are correlated. By analyzing the timing and logic of signal interactions between modules, the causes of module interactions leading to energy convergence can be identified. The operational logs of functional modules typically record events such as the startup, shutdown, and state transitions of each module, along with their occurrence times. By comparing the time points when the target frequency band exhibits anomalies with the event timestamps in the logs of each module, it is possible to preliminarily infer which module interactions might be related to energy convergence. For example, if it is found that the startup time of a certain module is always synchronized with the energy increase in the target frequency band, then that module is considered one of the suspected causes. This is equivalent to establishing a traceable link for electromagnetic interference problems, tracing back from the phenomenon (anomaly in a specific frequency band) to the root cause (which specific modules generated harmful interactions at what timing).

[0062] Finally, based on the causes of module interaction, corresponding filtering or control strategies are applied to the electromagnetic signal data or the operating mode of smart home appliances. The peak intensity of the target frequency band is then verified to see if it decreases, and a test report is generated based on the verification results. Filtering strategies can involve connecting a simple band-stop filter in series in the signal path to attenuate the signal in the target frequency band. Control strategies can involve manually adjusting the operating timing of the modules involved, for example, staggering the startup times of two modules that might cause overlap. After applying the strategies, electromagnetic signal acquisition and frequency domain analysis are performed again to compare whether the peak intensity of the target frequency band has decreased. The verification results, including the original peak value, the peak value after applying the strategies, and the change in peak value, are recorded and a test report is generated. This not only provides a rectification verification method, but more importantly, it forms a complete closed loop of test-analysis-location-measures-verification, enabling feedback debugging in electromagnetic compatibility design.

[0063] This method simulates the dynamic interaction process of smart home appliances, capturing in real time the nonlinear superposition effect generated when multifunctional modules work together, and accurately identifying the target frequency band of energy convergence. By analyzing and classifying the superimposed waveform sequences, the generation mechanism of transient high-intensity interference peaks can be effectively revealed. Furthermore, by tracing the causes of module interactions through associated module operation logs, the source of electromagnetic interference can be precisely located, enabling the application of targeted filtering or control strategies to effectively reduce interference. This avoids the blindness and high cost of traditional "trial and error" rectification, improving the accuracy and efficiency of electromagnetic compatibility testing for smart home appliances.

[0064] In practice, a preliminary analysis of the spectrum distribution alone may not be sufficient to accurately distinguish between strong signals and background noise generated by a single module, and the nonlinear superposition of energy resulting from the complex interaction of multiple functional modules. This ambiguity can lead to misjudgments, making subsequent interference mode classification and cause tracing less accurate, thus affecting the reliability of test results and the effectiveness of subsequent optimization strategies.

[0065] In this regard, refer to Figure 2 As shown, this application further proposes a specific method for identifying one or more target frequency bands where energy convergence occurs due to the nonlinear superposition of multi-module signals, including:

[0066] When generating a continuous spectral energy profile, the electromagnetic signal data after frequency domain transformation is processed. For example, the energy values ​​(such as power spectral density or amplitude squared) at each frequency point are obtained through Fast Fourier Transform (FFT). These energy values ​​are then arranged and connected in frequency order to form a continuous curve or spectrum, which visually shows the distribution of signal energy across the entire frequency range.

[0067] Based on this, trend analysis of the spectral energy profile is performed to identify the boundaries of regions where energy increases or decreases significantly. This can be achieved by calculating the first or second derivative of the spectral energy profile to detect significant changes in the slope. For example, when the derivative value changes from near zero (gradual) to a significantly positive value (increase), a starting frequency can be determined; when the derivative value changes from a significantly positive value (increase) to near zero or a negative value (gradual or decrease), an ending frequency can be determined.

[0068] In practice, the trend analysis can be achieved through the following steps:

[0069] Differential sequence calculation: For the energy values ​​of each frequency point arranged in frequency order in the spectral energy profile, calculate its first-order forward differential sequence. This differential sequence characterizes the rate of change sequence of energy with frequency variation.

[0070] Quantitative Judgment of Significance of Change: To objectively identify the transition from "gradual" to "increased" or "increased" to "gradual / decreased" from continuous difference values, quantitative criteria need to be established. One possible implementation is to set a positive threshold, which can be determined based on the overall fluctuation level of the background noise or spectral energy profile of the test system. For example, it can be obtained by multiplying the statistical value (such as the mean, median, or a certain percentile) of the absolute value sequence of the difference sequence by an empirical coefficient (such as 0.5 to 2); traverse the difference sequence to identify frequency points that meet the following conditions:

[0071] Starting frequency: If the energy change rate value corresponding to a certain frequency exceeds the judgment threshold, and the absolute value of the change rate corresponding to the previous frequency is lower than the threshold, then it is determined that the energy change trend at that frequency changes from flat to rising, and the frequency is recorded as a starting frequency.

[0072] End frequency point: If the absolute value of the energy change rate corresponding to a certain frequency point is lower than the judgment threshold (or is a negative value exceeding the threshold), and the change rate value corresponding to the previous frequency point exceeds the threshold, then it is determined that the energy change trend at that frequency point changes from increasing to flat or decreasing, and the frequency point is recorded as an end frequency point.

[0073] Specifically, the empirical coefficient is set to a range of 0.5 to 2.0, which is a reasonable range derived from statistical analysis and engineering practice of a large amount of electromagnetic signal data from typical smart home appliances.

[0074] The lower limit of 0.5 is set based on the following principle: when the coefficient is below 0.5, the calculated threshold is too low and overly sensitive, easily misjudging a large amount of background noise or slight normal fluctuations as significant changes. This leads to too many invalid candidate frequency bands in subsequent steps, greatly increasing the computational burden and causing false alarms. This lower limit ensures that the threshold can effectively filter out normal background fluctuations.

[0075] The upper limit of 2.0 is set based on the following principle: when the coefficient is higher than 2.0, the calculated threshold is too high and too lenient, which may lead to the underestimation of real, moderately intense energy convergence phenomena. This upper limit ensures that the system has the necessary detection sensitivity for abnormal energy convergences that meet the definition.

[0076] Practice shows that for the vast majority of smart home appliances, the effective discrimination coefficient, after optimization through the above calibration process, falls within the range of 0.5 to 2.0. This range provides technicians with clear parameters for tuning, avoiding blind, large-scale searches, and is sufficient to cover the differences between different product models, ensuring the universality and feasibility of the method.

[0077] Subsequently, the continuous frequency bands between each set of start and end frequencies are designated as potential energy convergence regions and used as candidate frequency bands. This step visualizes the results of the aforementioned trend analysis as a series of discrete frequency intervals. For example, if the analysis results show a significant increase in energy between 100MHz and 120MHz, then [100MHz, 120MHz] is marked as a candidate frequency band.

[0078] To ultimately determine the target frequency band, this application further retrieves the synchronous runtime sequence of each functional module of the smart home appliance during dynamic interaction. This runtime sequence records the precise timestamps of startup, shutdown, mode switching, or critical operations of each functional module (such as the Wi-Fi module, Bluetooth module, display module, and processor module) during the testing process. Then, the specific periods of significant energy increase within each candidate frequency band are analyzed and compared with the synchronous runtime sequence of the functional modules. If the energy increase period of a candidate frequency band has a temporal synchronization or close correlation with the periods when two or more functional modules simultaneously enter a specific working state, then that candidate frequency band is ultimately determined as the target frequency band generated by the nonlinear superposition of signals from multiple modules.

[0079] In practice, the original spectral energy profile often contains a lot of random noise and subtle fluctuations. These rapidly changing components may mask the true energy convergence trend, leading to deviations in identifying the start and end frequencies where the rate of energy change changes from flat to rising or from rising to flat / falling, thus affecting the accuracy of target frequency band identification.

[0080] Furthermore, a trend enhancement approach for the spectral energy profile is proposed, including: analyzing the energy fluctuation amplitude and frequency of the spectral energy profile within different frequency ranges, and distinguishing between slowly varying components reflecting macroscopic trends and rapidly varying components reflecting noise interference. This step aims to initially decompose the original spectral energy profile, classifying its signal components into two categories: one is the "slowly varying component" reflecting the overall energy change trend, typically corresponding to actual signal superposition or energy convergence; the other is the "rapidly varying component," mainly caused by random noise, transient interference, or minor non-critical fluctuations. This distinction provides a basis for subsequent smoothing processing, ensuring that important macroscopic trend information is not lost while removing noise. In terms of implementation, various signal processing techniques can be employed. For example, Fourier transform or wavelet transform can be used to decompose the spectral energy profile into different frequency scales, where low-frequency components correspond to slowly varying components and high-frequency components correspond to rapidly varying components. Alternatively, preliminary smoothing methods such as moving average and Gaussian filtering can be used, identifying rapidly varying components by comparing the differences between the original signal and the smoothed signal. In addition, statistical analysis, such as calculating local variance or standard deviation, can be used to quantify the amplitude and frequency of energy fluctuations in different frequency ranges, thereby setting thresholds for differentiation.

[0081] Subsequently, based on the distribution characteristics of the slowly varying components in the spectrum, a smoothing window is dynamically generated to match the width of the local energy trend. A wider smoothing window is used in wide frequency bands with gentle energy trends, while a narrower smoothing window is used in narrow frequency bands with potentially drastic energy changes. The core of this step lies in introducing the concept of a "dynamic smoothing window" to adapt to the different characteristics exhibited by the spectral energy profile in different frequency bands. Traditional fixed-window smoothing methods may over-smooth in regions with drastic energy changes, leading to loss of detail; while in regions with gentle energy changes, they may under-smooth, failing to effectively suppress noise. Dynamically generated smoothing windows can intelligently adjust the smoothing intensity according to the local characteristics of the slowly varying components, thereby preserving key trend information while suppressing noise to the greatest extent. Implementing dynamic smoothing windows can be based on real-time analysis of the local gradient, second derivative, or local variance of the slowly varying components. For example, a threshold can be set: when the local gradient of the slowly varying components is less than a certain threshold (indicating a gentle energy trend), a larger smoothing window (e.g., 50 frequency points) is used; when the local gradient is greater than a certain threshold (indicating drastic energy changes), a smaller smoothing window (e.g., 5 frequency points) is used. The smooth window can be a rectangular window, a Gaussian window, a Hamming window, etc., and its width is dynamically adjusted according to the smoothness or intensity of the local energy trend.

[0082] Specifically, when calculating the local gradient, for the current spectral energy profile sequence P[i] (where i is the frequency index), a local neighborhood is considered with it as the center.

[0083] The gradient calculation method is as follows: the approximate gradient of the energy profile in the neighborhood is calculated using the first-order central difference method. For frequency point i, its local gradient G[i] can be calculated as follows:

[0084] G[i]=(P[i+L]-P[iL]) / (2×L×Δf);

[0085] Where L is the half-window length used to calculate the gradient, and Δf is the frequency interval. This formula characterizes the average rate of change of energy within a frequency band centered at i and with a width of 2L.

[0086] Specifically, based on the calculation of the local gradient described above, the dynamic mapping rule for the smoothing window width is as follows:

[0087] Gradient normalization: To avoid the absolute gradient value being affected by the signal amplitude, the calculated local gradient sequence G[i] is first normalized, for example, by mapping it to the [0,1] interval:

[0088] Gnorm[i]=(|G[i]|-Gmin) / (Gmax-Gmin)

[0089] Gmin and Gmax can be the minimum and maximum gradient values ​​predicted for the current spectral profile, or they can be set as fixed empirical values.

[0090] Width mapping function: Sets the minimum width Wmin (e.g., 5 frequency points) and maximum width Wmax (e.g., 50 frequency points) of the smoothing window. The width W[i] of the smoothing window at frequency point i is determined by the normalized gradient Gnorm[i] through the following mapping function:

[0091] W[i]=Wmax-(Wmax-Wmin)×Gnorm[i]

[0092] Mapping function description: This is a simple linear inverse mapping. When Gnorm[i] = 0 (gradient is minimal, trend is most gradual), W[i] = Wmax, using the widest smoothing window to maximize noise suppression. When Gnorm[i] = 1 (gradient is maximum, trend is most drastic), W[i] = Wmin, using the narrowest smoothing window to preserve details of rapidly changing conditions to the maximum extent. When Gnorm[i] is an intermediate value, the window width transitions linearly.

[0093] Building upon this, a smoothing window is used to convolve the spectral energy profile, suppressing rapidly changing components while preserving and enhancing the macroscopic transition features of energy convergence and attenuation characterized by slowly changing components. This generates an optimized, smoothed spectral energy profile. Convolution is a commonly used smoothing technique in signal processing. By convolving the spectral energy profile with the dynamically generated smoothing window, the suppression of rapidly changing components can be effectively achieved. This process is equivalent to weighted averaging of the energy values ​​of each frequency point and its neighborhood, thereby eliminating random noise and high-frequency fluctuations. Simultaneously, since the smoothing window is dynamically adjusted based on the slowly changing components, it better preserves and enhances the macroscopic trends reflecting actual energy convergence and attenuation, making the optimized spectral energy profile clearer and more accurate, facilitating subsequent identification of energy convergence regions. The specific implementation of the convolution operation can be accomplished using software algorithms on a digital signal processor (DSP) or a general-purpose processor. For each frequency point, the energy values ​​around it are multiplied and summed with the weights of the smoothing window to obtain a new smoothed value for that frequency point. For example, if a Gaussian smoothing window is used, the closer the frequency point is to the center frequency, the greater its weight. In this way, the randomness of the rapidly varying components (high-frequency noise) is averaged out, while the slowly varying components (low-frequency trends) are preserved and enhanced. The final optimized and smoothed spectral energy profile will have clearer energy peaks and valleys, as well as smoother transition regions.

[0094] In some embodiments described above, a method is proposed to perform frequency domain conversion on electromagnetic signal data to identify the target frequency band, and to perform time domain sampling on the electromagnetic signals of the target frequency band to obtain a superimposed waveform sequence. However, in practice, if only simple time domain sampling is performed on the electromagnetic signals of the target frequency band, the obtained superimposed waveform sequence may not accurately reflect the true mode of nonlinear superposition of multi-module signals due to a lack of consideration for the signal generation background and module interaction timing, thereby affecting the subsequent effective classification and identification of abnormal interference modes.

[0095] In this regard, refer to Figure 3 As shown, this application further proposes to perform time-domain sampling of electromagnetic signals in the target frequency band to obtain the corresponding superimposed waveform sequence, specifically including:

[0096] For a target frequency band, the original time-domain signal segment corresponding to the band is extracted from the real-time acquired electromagnetic signal data. This step aims to focus the analysis on the specific frequency band where energy convergence issues have been identified. By applying a digital bandpass filter (e.g., a finite impulse response (FIR) filter or an infinite impulse response (IIR) filter), the frequency components corresponding to the target frequency band in the real-time acquired broadband electromagnetic signal data are separated, thereby obtaining the original signal representation of that frequency band on the time axis. This original time-domain signal segment contains detailed information on the time-varying changes of all electromagnetic energy within the target frequency band, laying the foundation for subsequent refined analysis.

[0097] Based on the switching times of the associated functional modules that generate the target frequency band signal, multiple synchronization trigger signals are generated. This step is crucial for establishing the correlation between electromagnetic signals and the internal operating state of the smart home appliance. Associated functional modules refer to those modules whose changes in operating state are identified as having a causal relationship with the energy convergence of the target frequency band. Operating state switching times can include events such as module startup, shutdown, mode switching, data transmission start / end, and load changes. These times can be precisely acquired through the smart home appliance's internal log system, dedicated hardware monitoring interfaces (such as GPIO signals), or software event listening mechanisms, and used as time reference points, i.e., synchronization trigger signals, for synchronous sampling of electromagnetic signals.

[0098] It should be further explained that, in this embodiment, the associated functional modules are not completely unknown at the beginning of the analysis, but rather refer to one or more candidate functional modules that are predetermined based on prior information of the target frequency band or system design knowledge of smart home appliances.

[0099] The candidate set can be determined using the spectral feature mapping method: based on electromagnetic compatibility (EMC) design principles, a mapping relationship library is established between the typical operating frequency bands or interference characteristic frequency bands of specific functional modules and the target frequency bands. For example, the main harmonic frequency bands of switching power supply modules, the operating frequency and harmonics of clock circuits, and the carrier frequency bands of wireless communication modules (such as Wi-Fi and Bluetooth) are all known design parameters. When the target frequency band is within the known characteristic frequency band range of a module, that module is included in the candidate set.

[0100] Based on the candidate functional module set determined in the above manner, the state transition times recorded in their operation logs are extracted and used as the basis for generating the synchronization trigger signal. The subsequent "interaction cause tracing" step, based on this candidate set, uses precise timing and logical correlation analysis to ultimately determine which module(s) is the true cause of the energy convergence in the target frequency band.

[0101] Using each synchronization trigger signal as a reference point, a signal window of fixed time length is extracted from the original time-domain signal segment, resulting in multiple aligned time-domain waveform units. This step standardizes the electromagnetic signal segments related to specific internal events through a unified time window. For example, a fixed time window can be set, containing several microseconds before and after the trigger point. When a synchronization trigger signal occurs, the system extracts a preset length of signal data from the previously extracted original time-domain signal segment, centered on or starting from the trigger signal, forming a time-domain waveform unit. Since all waveform units are extracted using their corresponding synchronization trigger signals as reference points, they are aligned in time, allowing for direct comparison and analysis of the electromagnetic responses at different events.

[0102] All time-domain waveform units are arranged according to the order of events corresponding to their synchronization trigger signals, forming a superimposed waveform sequence that characterizes how the target frequency band signal repeatedly occurs with associated events during dynamic interaction. This step logically arranges all independently extracted, aligned time-domain waveform units according to the actual occurrence time order of their respective associated internal events. This sequence is not simply a mathematical superposition of all waveforms, but rather forms a collection of time series that visually demonstrates the repeatability, consistency, or variability of the electromagnetic signal in the target frequency band during each state switch of the associated functional module. In this way, the dynamic behavior pattern of the target frequency band signal under different event triggers can be clearly observed.

[0103] In practical applications, based on the characteristics of superimposed waveform sequences, nonlinear superposition modes can be classified into at least the following categories: steady-state resonance mode, transient impulse mode, and random modulation mode.

[0104] The steady-state resonance mode is characterized by highly similar envelope shapes and stable repetition periods among the waveform units in the superimposed waveform sequence, indicating the existence of continuous, periodic signal superposition. For example, when multiple functional modules within a smart home appliance operate synchronously at a fixed frequency or phase relationship, the electromagnetic signals they generate may undergo continuous in-phase superposition within a specific frequency band, resulting in a stable energy convergence. In practice, this type of mode can be identified by performing spectral analysis or autocorrelation analysis on the superimposed waveform sequence to detect the presence of significant and stable periodic components and by observing the consistency of the waveform envelope.

[0105] Specifically, spectral analysis or autocorrelation analysis can be used to identify steady-state resonance modes through the following steps:

[0106] For periodic detection, the autocorrelation analysis method involves calculating the autocorrelation function of the superimposed waveform sequence and searching for significant autocorrelation peaks outside of zero time delay. If a series of equally spaced significant peaks exist, the interval represents the repetition period of the signal. The significance can be quantified by determining whether the peak amplitude exceeds a preset threshold (e.g., twice the average absolute value of all autocorrelation coefficients). The spectral analysis method involves performing a Fourier transform on the superimposed waveform sequence to obtain its spectrum and searching for significant spectral peaks at the fundamental frequency and its harmonics. The significance can also be quantified by determining whether the peak amplitude exceeds a preset threshold (e.g., three times the average level of the spectral background noise).

[0107] Periodic stability determination: By performing the above periodic detection on different segments of the sequence, the standard deviation of each periodic value is calculated. If the standard deviation is less than a predetermined proportion of the average period (e.g., 5%), it is determined to have a stable repetition period.

[0108] For envelope shape consistency analysis: Envelope extraction: For each waveform unit in the superimposed waveform sequence, its envelope signal is extracted by Hilbert transform; Consistency quantification: The correlation coefficients between each pair of the envelope signals of all waveform units are calculated to form a correlation coefficient matrix, and the average value of the off-diagonal elements in the matrix is ​​calculated as the overall envelope similarity; Consistency determination: If the overall envelope similarity is greater than a preset first similarity threshold (e.g., 0.85), the waveform units are determined to have highly similar envelope shapes.

[0109] A nonlinear superposition mode is determined to be a steady-state resonance mode if and only if both conditions are met simultaneously: a stable repetition period and a highly similar envelope shape.

[0110] Transient impact patterns manifest as one or more waveform units with sharp, pulse-like envelopes appearing intermittently in a superimposed waveform sequence. These waveform units have significantly higher amplitudes than other units in the sequence, and their occurrence is highly correlated with power-on / off or load surge events of specific modules, indicating the presence of transient energy injection and superposition caused by abrupt state changes. Such patterns are typically closely related to transient state changes of functional modules within smart home appliances, such as power switching, relay activation or deactivation, motor start-up or shutdown, and rapid switching of high-current loads. For identification, peak detection algorithms can be used to capture abnormally high-amplitude pulses, and precise time correlation analysis can be performed by combining this with the timestamps of state transition events recorded in the functional module's operation log.

[0111] The characteristic of random modulation modes is that the envelope shape, amplitude, and period of each waveform unit in the superimposed waveform sequence exhibit irregular fluctuations, indicating the asynchronous superposition of multiple independent source signals in an unpredictable manner. This may stem from the asynchronous, non-periodic, or random superposition of electromagnetic signals from multiple functional modules, such as due to the randomness of module operating timing, the randomness of internal noise sources, or the simultaneous presence of multiple independent and unrelated signal sources in the same frequency band. To identify such modes, statistical analysis can be performed on the envelope characteristics of the waveform units, such as calculating the mean, variance, kurtosis, and skewness, and assessing their periodicity. If the statistics exhibit high randomness and irregularity, it can be determined to be a random modulation mode.

[0112] Specifically, the randomness quantification criterion utilizes entropy analysis: the approximate entropy or sample entropy of the envelope sequence is calculated. The higher the entropy value, the stronger the complexity and unpredictability of the sequence, and the higher the randomness. When the entropy value exceeds a preset entropy threshold, it indicates that the sequence has a high degree of randomness and is determined to be a random modulation mode. The entropy threshold can be obtained through experimental calibration, for example, by analyzing the entropy distribution of known random modulation signal samples and deterministic signal samples, and selecting a critical value that can effectively distinguish between the two types of signals.

[0113] Specifically, the entropy threshold is achieved through the following training and calibration process:

[0114] Constructing a calibration sample set includes: Positive sample set acquisition: In a controlled test environment or on a smart home appliance product known to have random interference, the dynamic interaction process is run, and the steps of the method of this invention are executed until multiple instances initially identified as "random modulation modes" are identified. The superimposed waveform sequences and their envelope sequences corresponding to these instances are recorded, constituting a positive sample set for random modes. Negative sample set acquisition: Similarly, on products known to have steady-state resonance or transient impact modes, or by artificially injecting periodic or transient deterministic signals in a laboratory, the envelope sequences of multiple instances identified as "steady-state resonance modes" or "transient impact modes" are obtained, constituting a negative sample set for deterministic modes. The sample set is ensured to be of sufficient size and representativeness.

[0115] Feature extraction and distribution analysis: For each sample in the positive and negative sample sets, calculate its preset randomness quantification features, and draw the feature value distribution histogram or probability density curve of the positive and negative sample sets respectively.

[0116] Based on the feature value distribution, the mean midpoint method is adopted: calculate the mean of the feature values ​​of the positive sample set and the mean of the feature values ​​of the negative sample set, and take the midpoint as the initial threshold candidate value.

[0117] The calculated threshold is applied to an independent validation sample set to evaluate its classification accuracy, false positive rate, and false negative rate. If the performance meets the preset engineering requirements, the threshold is fixed as the preset entropy threshold for use in subsequent tests of the same or similar products.

[0118] In electromagnetic emission testing of smart home appliances, when identifying potential abnormal interference modes by combining the peak intensity of the target frequency band and the classification results of nonlinear superposition modes, if there is a lack of a systematic and standardized definition and matching mechanism for abnormal features, the identification process of abnormal modes may rely on human experience, which is inefficient and inconsistent, and makes it difficult to accurately and objectively determine which nonlinear superposition modes truly pose a potential risk.

[0119] In this regard, refer to Figure 4 As shown, this application further proposes a method for identifying nonlinear superposition patterns that conform to preset anomalous characteristics, which includes:

[0120] An anomaly feature knowledge base is established, where each knowledge unit defines a preset anomaly feature, including: a nonlinear superposition mode category, a target frequency band peak intensity condition, and an indicated risk type. The anomaly feature knowledge base is a structured dataset used to store and manage predefined electromagnetic interference anomaly modes and their associated risk information. Each knowledge unit is an entry in this knowledge base, detailing a specific anomaly feature. The "nonlinear superposition mode category" refers to the mode classified based on the characteristics of the superimposed waveform sequence after time-domain sampling of the electromagnetic signal in the above method, such as a steady-state resonance mode, a transient impulse mode, or a random modulation mode. The "target frequency band peak intensity condition" refers to the threshold or range condition that the peak intensity of the electromagnetic signal in the corresponding target frequency band must meet for a specific nonlinear superposition mode category. For example, it can be set as "peak intensity greater than X dBm" or "peak intensity between Y and Z dBm". The "indicated risk type" refers to the nature of the potential risk represented when the anomaly feature defined by the knowledge unit is met, such as "communication interruption risk," "equipment malfunction risk," "user experience degradation risk," or "security hazard risk."

[0121] Specifically, when the system identifies a transient impulse pattern, if its energy is concentrated in a sensitive communication frequency band and its intensity is too high, it may indicate a risk of communication interruption. Based on this, a rule can be constructed: if the center frequency of the target frequency band is within the Wi-Fi operating frequency band (2.4-2.4835GHz) and the peak intensity exceeds -50dBm, then a risk of Wi-Fi communication interruption is identified. This rule stems from the suppressive effect of transient pulses on the receiving sensitivity of wireless modules.

[0122] When the system identifies a steady-state resonance mode, it is necessary to pay attention to whether its frequency coincides with the inherent frequency or harmonics of the critical circuits inside the device. For example, if the resonance frequency is an integer multiple of the display screen's horizontal scanning frequency (e.g., 100MHz) ((100×N)MHz±1%), and the peak intensity exceeds -40dBm, it may cause display screen flicker risk by coupling interference with the display drive timing.

[0123] For random modulation modes, the danger lies not only in the absolute intensity but also in the drastic fluctuations in amplitude. If the peak intensity fluctuation of the target frequency band in this mode exceeds 20dB, it indicates that the electromagnetic environment is extremely unstable and may overwhelm the dynamic margin of adaptive devices such as Bluetooth, thus indicating a risk of Bluetooth audio interruption.

[0124] Furthermore, the knowledge base needs to cover composite pattern rules to cope with complex interference scenarios. For example, if transient impacts and steady-state resonances are detected to alternate in time, and the peak intensity corresponding to either pattern exceeds -45dBm, it indicates that the system has a continuous instability problem in power supply or clock, and it can be determined that there is a higher risk of microcontroller reset or crash.

[0125] This knowledge base can be built and updated through human experience, historical test data analysis, or industry standards to ensure its coverage and accuracy. The above examples are just a few instances illustrating the logic of knowledge base construction. In practical applications, the strength threshold, frequency range, risk type, etc., can be adjusted and expanded in a targeted manner according to the specific product's module composition, circuit layout, and reliability requirements to form a proprietary knowledge base that matches the product's electromagnetic compatibility characteristics.

[0126] Subsequently, a feature description of the current pattern is generated. This description is a quantitative representation of the nonlinear superposition pattern of the electromagnetic signal currently being analyzed, including the category of the nonlinear superposition pattern and the real-time peak intensity value of the target frequency band, which is matched with the anomaly feature knowledge base.

[0127] Next, the feature descriptions are matched against an anomaly feature knowledge base. The matching process aims to compare the currently detected electromagnetic interference pattern with predefined anomaly features. This includes filtering knowledge units with matching categories based on the categories in the feature descriptions; that is, identifying knowledge units from the anomaly feature knowledge base that share the same nonlinear superposition mode category as the current pattern (e.g., "steady-state resonance mode"). Then, the real-time peak intensity value in the feature descriptions is verified to meet the intensity conditions defined by the filtered knowledge units; that is, for each filtered knowledge unit with matching categories, the real-time peak intensity value of the current pattern is checked to ensure it meets the preset peak intensity conditions for that knowledge unit.

[0128] If the verification passes, meaning that both the nonlinear superposition category and the real-time peak intensity value of the current pattern perfectly match a knowledge unit in the anomaly feature knowledge base, then the nonlinear superposition pattern is determined to be an anomalous interference pattern that conforms to the preset anomalous features. This means that the currently detected electromagnetic signal superposition pattern does indeed pose a potential risk. Finally, a determination report containing the risk type and feature description will be generated. This report details the category of the determined anomalous interference pattern, the real-time peak intensity value, and the risk type indicated by the knowledge base, providing a basis for subsequent fault investigation and strategy adjustment.

[0129] Furthermore, in the process of identifying nonlinear superposition patterns that conform to preset abnormal characteristics and associating the operation logs of functional modules that generate signals in the target frequency band with those of the frequency band, it is challenging to accurately locate the specific module interaction behavior that leads to energy convergence from the complex module operation logs. This is especially true in dynamic scenarios where multiple functional modules are running simultaneously and interacting frequently. Simply associating logs may not reveal deep causal relationships and make it difficult to accurately trace the source of interference.

[0130] In this regard, refer to Figure 5 As shown, this application further proposes a method to determine the causes of module interactions leading to energy convergence by analyzing the signal interaction timing and logic between modules. The specific steps include:

[0131] First, key time feature points of the waveform envelope are extracted from the superimposed waveform sequence of abnormal interference modes, and precise timestamps of state switching events are extracted from the operation logs of functional modules. These key time feature points and precise timestamps are aligned on a unified time axis to establish a precise temporal correspondence between electromagnetic interference phenomena (waveform envelope feature points) and the behaviors of internal functional modules of smart home appliances (state switching events). By aligning them on a unified time axis, it is possible to visually observe and analyze which module behaviors are likely closely related to the occurrence of electromagnetic energy convergence. Key time feature points of the waveform envelope can be identified by performing envelope detection on the superimposed waveform sequence, for example, using Hilbert transform, peak detection, or moving average methods, and then identifying significant change points in the envelope, such as rising edges, falling edges, peak points, and valley points, as key time feature points. These points represent moments when the electromagnetic signal energy or mode undergoes significant changes. The operation logs of functional modules typically record events such as module startup, shutdown, mode switching, and load changes, along with precise timestamps, which can be directly parsed from the logs. Unified timeline alignment ensures that all timestamps and feature points are referenced to the same system clock or calibrated through a time synchronization mechanism for accurate comparison and correlation.

[0132] Furthermore, for each key time feature point, the corresponding state transition event is searched, and the functional module that executed the state transition event is marked as a suspicious module associated with the current feature point. The purpose is to initially screen out functional modules that highly coincide with the time of electromagnetic interference occurrence. In this way, the analysis scope can be narrowed down from all functional modules to a few "suspicious" modules, improving the efficiency and accuracy of subsequent analysis. For each extracted key time feature point, state transition events occurring in the functional module's operation log can be searched within a preset time window (e.g., a few milliseconds to tens of milliseconds before and after the feature point). If a module undergoes a state transition within this time window, such as from standby to operation, or from low power to high power, then the module is temporarily marked as a "suspicious module" associated with that key time feature point. The size of the time window can be dynamically adjusted according to the response speed and signal propagation delay of the smart home appliance.

[0133] Specifically, the time window can be a fixed duration interval centered on a key time feature point, such as −Tms, +Tms, where T can be preset based on the typical circuit response and signal propagation delay of smart home appliances (e.g., T=20ms). As a more preferred implementation, the time window can be dynamically adjusted: initially, a wider window (e.g., ±50ms) is used to ensure the capture of associated events; subsequently, based on the statistical event density of modules marked near multiple feature points, if the density is too high, the window is narrowed (e.g., reduced to ±10ms) to reduce accidental associations; otherwise, the window is maintained or appropriately expanded. Within the preset or dynamic associated time window, state switching events occurring in the function module's operation log are searched, and the function module executing the event is marked as a suspicious module associated with the current key time feature point.

[0134] Subsequently, the co-occurrence relationships and frequencies among different functional modules marked as suspicious during the dynamic interaction process are statistically analyzed to discover whether there are patterns of collaborative work or simultaneous state switching among different suspicious modules. The nonlinear superposition of electromagnetic energy is often the result of the combined action of multiple modules; therefore, identifying the co-occurrence relationships between modules is crucial for revealing deep interaction mechanisms. A co-occurrence matrix or list can be established to record all combinations of modules marked as suspicious near each key time feature point. For example, if both module A and module B are marked as suspicious at a key time point, the "AB" combination is recorded once. Then, the total number of times each module combination occurs throughout the entire dynamic interaction process, i.e., the co-occurrence frequency, is counted. This can be accomplished by traversing all key time feature points and their corresponding sets of suspicious modules.

[0135] Finally, based on co-occurrence relationships and frequencies, one or more groups of functional modules that repeatedly co-occur in time, with their co-occurrence times corresponding to key time feature points, are identified as the module interaction causes leading to energy convergence. This is a crucial step in ultimately determining the root cause of energy convergence. By analyzing co-occurrence relationships and frequencies, accidental and unrelated module behaviors can be eliminated, focusing on module combinations that continuously and stably occur synchronously with electromagnetic interference phenomena, thereby accurately locating the module interaction causes leading to nonlinear superposition and energy convergence. A co-occurrence frequency threshold can be set; only when the co-occurrence frequency of a module combination exceeds this threshold is it considered statistically significant. Further analysis of these high-frequency co-occurring module combinations, combined with their specific behaviors at key time feature points, such as whether they start simultaneously or transmit data simultaneously, identifies the specific interaction patterns most likely to lead to energy convergence. For example, if modules A and B always start simultaneously, and their start times are always synchronized with the peak of a certain electromagnetic interference, then "the coordinated start-up of modules A and B" can be identified as a module interaction cause.

[0136] Specifically, the co-occurrence frequency threshold can be set in one of the following ways:

[0137] (a) Absolute frequency method: Set a minimum integer threshold Nmin (e.g., Nmin=3). When the co-occurrence frequency of the module combination is ≥ Nmin, it is considered significant.

[0138] (b) Relative Proportion Method: Set a proportion threshold P (e.g., P=0.1, i.e. 10%). When the co-occurrence frequency of the module combination is greater than or equal to (total number of key time feature points × P), it is considered significant.

[0139] After identifying the module interactions that cause energy accumulation, the key challenge in ensuring the electromagnetic compatibility of smart home appliances is how to specifically and efficiently eliminate or significantly reduce these abnormal electromagnetic emissions. Without effective intervention methods, even if the problem is accurately identified, it is difficult to achieve practical improvements.

[0140] To address this, this application further proposes corresponding filtering and control strategies: the filtering strategies include frequency-domain notch filtering and time-domain gated filtering. Frequency-domain notch filtering applies a band-stop filter with a specific center frequency and suppression depth to the signal path for the target frequency band. Frequency-domain notch filtering is a filtering technique that suppresses signals within a specific frequency range. Its implementation typically involves introducing a band-stop filter into the signal processing link. This filter is designed to have maximum attenuation at the center frequency of the target frequency band, along with a certain suppression depth and bandwidth. For example, a digital notch filter can be implemented using a digital signal processor (DSP), precisely controlling the center frequency and suppression depth by setting its coefficients; alternatively, analog circuit design can be used, employing LC resonant circuits or active filters to achieve attenuation at specific frequencies. This filtering strategy can accurately suppress energy in the identified target frequency band without affecting normal signals in other frequency bands.

[0141] Time-domain gated filtering selectively attenuates or shields electromagnetic signal data within a critical time window corresponding to an abnormal interference mode. It's a filtering technique that selectively processes signals within a specific time window. For example, when transient interference is detected when a module switches states at a specific moment, the amplitude of the acquired electromagnetic signal data can be attenuated using digital signal processing algorithms within the time window of the transient interference, or the signal transmission can be temporarily blocked or attenuated at the physical level using hardware gating circuits. This method is particularly suitable for handling transient, non-periodic interference, avoiding over-processing of continuous signals and thus reducing the impact on normal system functions.

[0142] The control strategies include staggered module operating timing and controlled module power ramp-up rates. Staggered module operating timing refers to adjusting the start-up time, stop time, or duty cycle of functional modules involved in the module interactions causing energy accumulation, so that the high-load operating periods of each module are staggered in time. Specifically, after identifying the module interactions causing energy accumulation, the system analyzes the key operating timing of these modules in normal operating mode, such as data transmission, motor startup, and high-power or high-frequency signal transmission periods like sensor sampling. Subsequently, through software scheduling, firmware updates, or hardware logic adjustments, the key operating timing of these modules is reallocated or shifted to ensure that their high-load or high-transmission activities do not occur within the same time window. For example, if two modules will cause superimposed interference when started simultaneously, the start-up delay of one module can be adjusted so that it starts working only after the other module has started and stabilized.

[0143] Module power ramp-up rate control refers to reducing the rate of power change of at least one functional module during state transitions in module interaction. When a functional module switches from one operating state to another (e.g., from standby to operation, or from low speed to high speed), its power consumption or signal transmission strength often changes drastically. This rapid power change can generate broadband transient noise and nonlinearly superimpose with the signals of other modules. By controlling the power ramp-up rate, the power or signal strength change process of the module can be made smoother. This can be achieved by introducing a soft-start circuit in the module's power management unit, adjusting the PWM (Pulse Width Modulation) duty cycle curve of the digital controller, or implementing a power gradient algorithm at the firmware level. For example, when a motor module starts up, its current rises rapidly from zero to the rated value. By controlling the power ramp-up rate, the current can be gradually increased over a set time, thereby avoiding electromagnetic interference caused by instantaneous large current surges.

[0144] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for testing the electromagnetic emission of a smart home appliance, comprising controlling multiple functional modules of the smart home appliance under test to operate simultaneously according to a preset interaction logic, so as to simulate the dynamic interaction process under actual working conditions, characterized in that, Includes the following steps: During the dynamic interaction process, electromagnetic signal data generated by smart home appliances are collected in real time, and the electromagnetic signal data is frequency domain converted to obtain the corresponding spectrum distribution. The process involves analyzing the spectral distribution to identify one or more target frequency bands where energy converges due to the nonlinear superposition of signals from multiple modules, and recording the peak intensity of each target frequency band. This includes: traversing each frequency point in the spectral distribution and generating a continuous spectral energy profile based on its signal energy value; analyzing the trend of the spectral energy profile to identify the starting frequency point where the energy change rate changes from flat to increasing, and the ending frequency point where it changes from increasing to flat or decreasing; marking the continuous frequency bands between each set of starting and ending frequency points as potential energy convergence areas, serving as candidate frequency bands; retrieving the synchronous operation sequence of each functional module of the smart home appliance during dynamic interaction, and analyzing whether the periods of significant energy increase within the candidate frequency bands are synchronous or correlated with the periods when two or more functional modules enter a specific working state; and finally determining the candidate frequency bands with synchronous or correlated relationships as the target frequency bands generated by the nonlinear superposition of signals from multiple modules. Electromagnetic signals in the target frequency band are sampled in the time domain to obtain corresponding superimposed waveform sequences. Based on the characteristics of the superimposed waveform sequences, nonlinear superposition modes are classified. The classification results of nonlinear superposition modes include at least the following: Steady-state resonance mode: Each waveform unit in the superimposed waveform sequence has a highly similar envelope shape and a stable repetition period, indicating the existence of continuous, periodic signal in-phase superposition; Transient impulse mode: One or more waveform units with sharp pulse-like envelope shapes appear intermittently in the superimposed waveform sequence. The amplitude of these waveform units is significantly higher than that of other units in the sequence, and their occurrence time is highly correlated with the power-on / off or load change events of specific modules, indicating the existence of instantaneous energy injection and superposition caused by drastic state changes; Random modulation mode: The envelope shape, amplitude, and period of each waveform unit in the superimposed waveform sequence all exhibit irregular fluctuations, indicating the existence of multiple independent source signals asynchronously superimposed in an unpredictable manner. By combining the peak intensity of the target frequency band and the classification results of nonlinear superposition modes, nonlinear superposition modes that meet the preset abnormal characteristics are identified and judged as abnormal interference modes with potential risks. After determining that the interference characteristics are abnormal, the operation logs of the functional modules that generate frequency band signals during the dynamic interaction process are associated with the target frequency band. By analyzing the signal interaction timing and logic between modules, the cause of module interaction leading to energy convergence is determined. Based on the causes of module interaction, corresponding filtering or control strategies are applied to electromagnetic signal data or the operating mode of smart home appliances. Then, it is verified whether the peak intensity of the target frequency band is reduced, and a test report is generated based on the verification results.

2. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: Trend enhancement of the spectral energy profile includes: By analyzing the energy fluctuation amplitude and frequency of the spectral energy profile in different frequency ranges, we can distinguish between the slow-changing component reflecting the macroscopic trend and the fast-changing component reflecting noise interference. Based on the distribution characteristics of slowly varying components in the spectrum, a smoothing window that matches the width of the local energy trend is dynamically generated. A wider smoothing window is used in the wide frequency band where the energy trend is gentle, and a narrower smoothing window is used in the narrow frequency band where the energy change is potentially drastic. By using a smooth window to convolve the spectral energy profile, the rapidly varying components are suppressed, while the macroscopic transition features of energy convergence and decay characterized by the slowly varying components are preserved and enhanced, resulting in an optimized and smooth spectral energy profile.

3. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: Time-domain sampling of electromagnetic signals in the target frequency band yields the corresponding superimposed waveform sequence, including: For a target frequency band, the original time-domain signal segment corresponding to the frequency band is extracted from the real-time acquired electromagnetic signal data; Based on the switching time of the operating state of the associated functional module that generates the target frequency band signal, multiple synchronous trigger signals are generated; Using each synchronous trigger signal as a reference point, a signal window of fixed time length is extracted from the original time domain signal segment to obtain multiple time domain waveform units that are aligned with each other. All time-domain waveform units are arranged according to the order of events corresponding to their synchronization trigger signals to form a superimposed waveform sequence that characterizes how the target frequency band signal repeatedly appears with associated events during dynamic interaction.

4. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: Identify nonlinear superposition patterns that conform to preset abnormal characteristics, including: Establish an anomaly feature knowledge base, where each knowledge unit defines a preset anomaly feature, including: nonlinear superposition mode category, target frequency band peak intensity condition, and indicated risk type; Generate a feature description of the current mode, including: the category of the nonlinear superposition mode and the real-time peak intensity value of the target frequency band; Matching feature descriptions with anomaly feature knowledge bases includes: filtering knowledge units with consistent categories based on the categories in the feature descriptions; and verifying whether the real-time peak intensity values ​​in the feature descriptions meet the intensity conditions defined by the filtered knowledge units. If the verification is successful, the nonlinear superposition mode is determined to be an abnormal interference mode that meets the preset abnormal characteristics, and a judgment report containing risk type and feature description is generated.

5. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: By analyzing the timing and logic of signal interactions between modules, the causes of module interactions leading to energy accumulation were identified, including: Key time feature points of waveform envelope are extracted from the superimposed waveform sequence of abnormal interference mode, and precise timestamps of state switching events are extracted from the operation log of functional module. The key time feature points and precise timestamps are aligned on a unified time axis. For each key time feature point, find the corresponding state transition event and mark the functional module that executes the state transition event as a suspicious module associated with the current feature point; During the statistical dynamic interaction process, the co-occurrence relationship and co-occurrence frequency among different functional modules marked as suspicious modules were analyzed. Based on co-occurrence relationships and co-occurrence frequencies, one or more functional modules that repeatedly co-occur in time and whose co-occurrence times correspond to key time feature points are identified as the module interaction causes leading to energy convergence.

6. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: The corresponding filtering strategies include: Frequency domain notch filtering: Applying a band-stop filter with a specific center frequency and suppression depth to the signal path for the target frequency band; Time-domain gated filtering: Selectively attenuating or shielding electromagnetic signal data within the critical time window corresponding to the abnormal interference mode.

7. The electromagnetic emission testing method for smart home appliances according to claim 1, characterized in that: The corresponding regulatory strategies include: Module work timing staggered: Adjust the start time, stop time or work cycle of the functional modules involved in the module interaction to stagger the high load operation periods of each module in time; Module power ramp-up rate control: Reduce the power change rate of at least one functional module during state switching in the module interaction cause.

Images