Conversion fault prediction method and system for power supply cabinet

Abstract

The invention relates to the technical field of fault prediction, in particular to a conversion fault prediction method and system for a power cabinet, and the method comprises the following steps: collecting the inductance terminal voltage and inductance current of an inductor in the power cabinet, and the input side common-mode current of an EMI filter in the power cabinet; according to the EMI filter health state on-line monitoring method and device, the inherent frequency response characteristic of the passive filter network and the strong correlation between element parameters are utilized, and on-line monitoring of the EMI filter health state is achieved under the condition that power is not needed to be cut off or an excitation source is not needed to be additionally arranged. When the frequency drift exceeds the threshold value, the aging fault is judged, the problem that the filtering failure caused by capacitance attenuation or inductance magnetic conductivity reduction in long-term operation of a passive device is difficult to perceive is solved, the electromagnetic compatibility and the power conversion reliability of a power supply system in the whole life cycle are guaranteed, and the fault prediction precision is improved.

Description

Technical Field

[0001] This invention relates to the field of fault prediction technology, and in particular to a method and system for predicting switching faults in power supply cabinets. Background Technology

[0002] Fault prediction technology utilizes sensor technology to acquire key physical quantities (such as voltage, current, frequency, temperature, and vibration signals) during equipment operation in real time. Combined with signal processing, statistical analysis, or machine learning algorithms, it performs feature extraction and pattern recognition on the collected time-series data.

[0003] Current technologies for condition monitoring and health management of power electronic equipment typically focus on collecting statistical characteristics of macroscopic port parameters such as voltage, current, and temperature, or rely on general signal processing and machine learning models for black-box analysis of time-series data. This monitoring approach often ignores the nonlinear mechanisms underlying the dynamic changes in internal physical parameters of devices under operating conditions. In actual operation, due to the voltage regulation capabilities of the closed-loop control system of the switching power supply, slight decreases in inductance or saturation point drift of magnetic components are often compensated by the controller's duty cycle adjustment. This results in the output voltage and current indicators remaining normal in the early stages of a fault, thus masking the accumulated physical damage within the device. This compensation mechanism makes traditional port signal-based monitoring methods significantly lagging, making it difficult to issue early warnings before irreversible damage occurs. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a method and system for predicting switching faults in power cabinets.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for predicting switching faults in power supply cabinets, comprising the following steps: Collect the inductor terminal voltage and inductor current of the inductors in the power cabinet, as well as the input common-mode current of the EMI filter in the power cabinet; The real-time differential inductance value is calculated based on the current change rate of the inductor terminal voltage and the inductor current. The inductor current is divided into several current intervals, and the average differential inductance value in each current interval is statistically analyzed to construct a real-time inductor-current characteristic curve. Calculate the curve difference between the real-time inductor-current characteristic curve and the preset factory fingerprint curve, and determine the type of physical damage to the inductor based on the curve difference. Spectral analysis is performed on the input-side common-mode current to extract the resonant center frequency in the spectrum and calculate the frequency drift of the resonant center frequency relative to the preset reference frequency. If the frequency drift exceeds a preset drift threshold, the EMI filter is determined to have an aging fault.

[0006] Preferably, the steps for calculating the real-time differential inductance value specifically include: The rising slope of the inductor current during the conduction of the switching transistor is obtained, and the rising slope is used as the rate of change of the current. The real-time differential inductance value is calculated using the differential inductance calculation formula; The formula for calculating the differential inductance is: ,in, This represents the real-time differential inductance value. This represents the voltage at the inductor terminals. This indicates the upward slope.

[0007] Preferably, the steps for constructing the real-time inductor-current characteristic curve specifically include: Set the current segmentation step size, and divide the numerical range of the inductor current into several current intervals according to the current segmentation step size; The average differential inductance value is obtained by summing all the real-time differential inductance values ​​falling within each current range and calculating the average value. Using the aforementioned current ranges as the horizontal axis and the corresponding average differential inductance value as the vertical axis, the real-time inductance-current characteristic curve is generated by fitting.

[0008] Preferably, the step of determining the type of physical damage to the inductor specifically includes: Determine the low current range and the high current range from the aforementioned current ranges; If the average differential inductance value of the real-time inductance-current characteristic curve in the low current range is lower than the inductance value of the preset factory fingerprint curve in the same range, it is determined that the inductor has a crack fault with increased core air gap. If the inductance drop inflection point of the real-time inductance-current characteristic curve in the high current range is earlier than the inflection point of the preset factory fingerprint curve, it is determined that the inductor has a core saturation characteristic degradation fault.

[0009] Preferably, the steps for calculating the curve difference include: The curve difference degree is calculated using the weighted Euclidean distance algorithm; The formula for the weighted Euclidean distance algorithm is as follows: ,in, This indicates the degree of difference between the curves. This indicates the total number of the current ranges. This represents the k-th current interval being calculated. This represents the weighting coefficient for the k-th current interval. This represents the average differential inductance value of the real-time inductance-current characteristic curve in the k-th current interval. This represents the reference inductance value of the preset factory fingerprint curve in the k-th current range.

[0010] Preferably, the step of extracting the resonant center frequency from the spectrum specifically includes: The input-side common-mode current is processed using a power spectral density analysis algorithm to obtain the common-mode current spectrum; Identify the local maxima of the common-mode current spectrum within a preset frequency band to obtain a set of resonance peaks; The position of the maximum energy peak in the set of resonance peaks is traced using the spectral centroid algorithm to determine the resonance center frequency.

[0011] Preferably, the resonant center frequency is associated with the parasitic parameters of the EMI filter, and the association relationship satisfies: ,in, Indicates the resonant center frequency, Represents pi (π). This represents the inductance value of the common-mode choke in the EMI filter. This indicates the capacitance value of the Y capacitor in the EMI filter.

[0012] Preferably, the step of determining that the EMI filter has an aging fault specifically includes: Monitor the changing trend of the frequency drift; If the frequency drift shows a monotonically increasing trend and the frequency drift is greater than the preset drift threshold, it is determined that the capacitance of the Y capacitor in the EMI filter has decayed.

[0013] The present invention also provides a system comprising: The data acquisition module is used to acquire the inductor terminal voltage and inductor current of the inductors in the power cabinet, as well as the input common-mode current of the EMI filter in the power cabinet. The feature construction module is used to calculate the real-time differential inductance value based on the current change rate of the inductor terminal voltage and the inductor current, divide the inductor current into several current intervals, count the average differential inductance value in each current interval, and construct a real-time inductor-current feature curve. The damage determination module is used to calculate the curve difference between the real-time inductor-current characteristic curve and the preset factory fingerprint curve, and to determine the physical damage type of the inductor based on the curve difference. The spectrum analysis module is used to perform spectrum analysis on the input-side common-mode current, extract the resonant center frequency in the spectrum, and calculate the frequency drift of the resonant center frequency relative to the preset reference frequency. An aging fault determination module is used to determine that the EMI filter has an aging fault if the frequency drift exceeds a preset drift threshold.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the voltage and current at the inductor terminals in the power supply cabinet are collected, and the real-time differential inductance value is calculated based on the rate of change of current. By dividing the inductor current into several intervals and statistically analyzing the average differential inductance value of each interval, a real-time inductance-current characteristic curve is constructed, which can map the nonlinear physical characteristics of magnetic components under different magnetic flux densities. By comparing the difference between the real-time characteristic curve and the factory fingerprint curve, the microscopic physical state inside the magnetic core can be deeply analyzed, distinguishing and identifying faults such as increased air gap cracks caused by mechanical stress or saturation characteristic degradation caused by material aging. Thus, early signs of damage to magnetic components can be captured before macroscopic electrical parameters become abnormal. Spectral analysis of the input common-mode current is performed, and the resonant center frequency is extracted. Its drift relative to the reference frequency is calculated. Utilizing the inherent frequency response characteristics of the passive filter network and the strong correlation between component parameters, online monitoring of the health status of the EMI filter is achieved without power interruption or external excitation source. When the frequency drift exceeds the threshold, an aging fault is identified. This solves the problem that filtering failures caused by capacitor capacity decay or inductor permeability reduction in passive components during long-term operation are difficult to detect. This ensures the electromagnetic compatibility performance and power conversion reliability of the power system throughout its entire life cycle and improves the accuracy of fault prediction. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the steps of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] Please see Figure 1 This invention provides a technical solution, a method for predicting switching faults in power cabinets, comprising the following steps: Collect the inductor terminal voltage and inductor current of the inductors in the power cabinet, as well as the input common-mode current of the EMI filter in the power cabinet; The real-time differential inductance value is calculated based on the rate of change of inductor terminal voltage and inductor current. The inductor current is divided into several current intervals, and the average differential inductance value in each current interval is statistically analyzed to construct a real-time inductor-current characteristic curve. Calculate the curve difference between the real-time inductor-current characteristic curve and the preset factory fingerprint curve, and determine the type of physical damage to the inductor based on the curve difference. Perform spectrum analysis on the input-side common-mode current, extract the resonant center frequency in the spectrum, and calculate the frequency drift of the resonant center frequency relative to the preset reference frequency. If the frequency drift exceeds the preset drift threshold, the EMI filter is determined to have an aging fault.

[0018] In this embodiment, the step of calculating the real-time differential inductance value specifically includes: obtaining the rising slope of the inductance current during the conduction period of the switching transistor, using the rising slope as the rate of change of current, and calculating the real-time differential inductance value using the differential inductance calculation formula. The differential inductance calculation formula is as follows: ,in, This represents the real-time differential inductance value. Indicates the voltage across the inductor terminals. Indicates the upward slope.

[0019] Specifically, a high-frequency sampling probe is used to continuously collect the inductor current during the conduction cycle of the switching transistor. The sampling frequency is set to be more than 100 times the switching frequency to ensure that minute details of current changes are captured. Linear regression fitting is performed on the collected discrete current data points, and a subset of data in the linear rising segment of the current is selected. The slope of this data subset on the time axis is calculated using the least squares method. The slope value of the calculated fitted line is determined as the current change rate of the inductor current. At the same time, the voltage drop data across the inductor is collected synchronously. Considering that the voltage during conduction is mainly determined by the difference between the input voltage and the output voltage, the average of multiple collected voltage values ​​is taken to eliminate measurement jitter. The obtained average voltage value and current change rate are substituted into the pre-constructed division operation logic. The current inductor physical quantity is deduced from the physical definition, and the specific numerical calculation process is executed. The real-time differential inductance value is calculated using the differential inductance calculation formula, which is: ,in, This represents the real-time differential inductance value. Indicates the voltage across the inductor terminals. Indicates the upward slope.

[0020] In this embodiment, the steps for constructing the real-time inductor-current characteristic curve specifically include: setting a current segmentation step size, dividing the numerical range of the inductor current into several current intervals according to the current segmentation step size, accumulating all real-time differential inductance values ​​falling within each current interval, calculating the average value, obtaining the average differential inductance value, using several current intervals as the horizontal axis, using the corresponding average differential inductance value as the vertical axis, and fitting to generate the real-time inductor-current characteristic curve.

[0021] Specifically, the current segmentation step size is set according to the rated output current parameter of the power cabinet. For example, when the rated current is 100A, the current segmentation step size is set to 1A. Based on this step size, the continuous numerical range from 0A to the maximum overcurrent protection value is discretized into more than 100 independent intervals. The corresponding number of accumulators and counters are initialized. All calculated real-time differential inductance values ​​are traversed. According to the instantaneous current magnitude corresponding to each inductance value, it is mapped to the corresponding current interval. The differential inductance values ​​falling into the same interval are accumulated and the number of samples falling into the interval is recorded. At the end of each statistical cycle, the average value under the interval is calculated by dividing the accumulated value by the number of samples. The average value is used as the representative inductance value of the current interval to eliminate noise interference from a single calculation. The center current values ​​of all intervals are formed into an abscissa sequence, and the corresponding calculated average inductance values ​​are formed into a ordinate sequence. Cubic spline interpolation is used to smoothly connect these discrete coordinate points, fill in the intervals with missing samples, and fit to generate a real-time inductance-current characteristic curve.

[0022] In this embodiment, the step of determining the type of physical damage to the inductor specifically includes: determining a low current range and a high current range from several current ranges; if the average differential inductance value of the real-time inductance-current characteristic curve in the low current range is lower than the inductance value of the preset factory fingerprint curve in the same range, it is determined that the inductor has a crack fault with increased air gap in the magnetic core; if the inductance drop inflection point of the real-time inductance-current characteristic curve in the high current range is earlier than the inflection point of the preset factory fingerprint curve, it is determined that the inductor has a fault with degraded magnetic core saturation characteristics.

[0023] Specifically, based on the BH curve characteristics of the inductor core, several current ranges are divided into different monitoring areas. The range where the current value is within 0% to 30% of the rated current is defined as the low current range, and the range where the current value is within 70% to 100% of the rated current is defined as the high current range. The standard characteristic data of the inductor under factory conditions, pre-saved in memory, is retrieved as a preset factory fingerprint curve. The real-time inductor-current characteristic curve is compared point-by-point with the preset factory fingerprint curve for values ​​within the low current range. If the real-time curve is found to be in the low current range... The overall amplitude of the current section decreased by more than 5% compared to the fingerprint curve. Based on the inverse relationship between the air gap in the magnetic circuit and the inductance, it was determined that a physical separation occurred in the magnetic circuit that led to an increase in magnetic reluctance. In other words, it was determined that the inductor had a crack fault with an increased air gap in the magnetic core. At the same time, second derivative analysis was performed on the high current range to find the inflection point where the inductance value dropped sharply. If the current value corresponding to the inflection point of the real-time curve was less than 90% of the current value corresponding to the inflection point of the preset factory fingerprint curve, it was determined that the magnetic core entered a saturation state under a smaller current excitation. Based on this, it was determined that the inductor had a fault of magnetic core saturation characteristic degradation.

[0024] In this embodiment, the step of calculating the curve difference specifically includes: calculating the curve difference using a weighted Euclidean distance algorithm, the formula of which is: ,in, Indicates the degree of difference in the curves. Indicates the total number of current ranges. This represents the k-th current interval being calculated. This represents the weighting coefficient for the k-th current interval. This represents the average differential inductance value of the real-time inductor-current characteristic curve in the k-th current interval. This represents the reference inductance value of the preset factory fingerprint curve in the k-th current range.

[0025] Specifically, to quantitatively assess the current health status, a weighted Euclidean distance algorithm is used to calculate the similarity distance between two curves. Considering that changes in inductance in the saturation critical region are more sensitive to fault prediction, a weighting coefficient array is manually set. Larger weight values ​​(e.g., 1.5) are assigned to current intervals near the saturation point, while a standard weight value of 1.0 is assigned to current intervals in the linear region. All valid current intervals are traversed, and the average differential inductance value measured in real-time for each interval is extracted and compared with the corresponding reference inductance value in the factory fingerprint curve. The difference between the two is calculated and squared. The squared result is multiplied by the corresponding weighting coefficient. All weighted squared differences are summed, and finally, the square root of the sum is taken to obtain a scalar value reflecting the overall degree of deviation. The weighted Euclidean distance algorithm is then used to calculate the curve difference. The formula for the weighted Euclidean distance algorithm is: ,in, Indicates the degree of difference in the curves. Indicates the total number of current ranges. This represents the k-th current interval being calculated. This represents the weighting coefficient for the k-th current interval. This represents the average differential inductance value of the real-time inductor-current characteristic curve in the k-th current interval. This represents the reference inductance value of the preset factory fingerprint curve in the k-th current range.

[0026] In this embodiment, the step of extracting the resonant center frequency from the spectrum specifically includes: processing the input-side common-mode current using a power spectral density analysis algorithm to obtain the common-mode current spectrum, identifying local maxima points of the common-mode current spectrum within a preset frequency band to obtain a set of resonant peaks, and using a spectral centroid algorithm to track the position of the maximum energy peak in the set of resonant peaks to determine the resonant center frequency.

[0027] Specifically, the sampling rate of the high-frequency current transformer is set to above 10MHz to obtain a certain length of input-side common-mode current time-domain sequence. This time-domain sequence is windowed using a Hanning window function to reduce spectral leakage. The time-domain signal is converted into a frequency-domain signal using a fast Fourier transform. The square of the frequency-domain signal amplitude is calculated and divided by the frequency resolution to obtain the power spectral density distribution. The power spectral density curve is scanned within a pre-defined frequency band of interest, such as 150kHz to 30MHz. All local maxima are located by finding points where the first derivative is zero and the second derivative is negative. Peak points with amplitudes exceeding the background noise floor by more than 6dB are selected to form a set of resonant peaks. For the main resonant peak with the largest energy in the set, all frequency points within a 3dB bandwidth to the left and right of its peak are extracted. The weighted average of these frequency points is calculated with amplitude as the weight, thereby accurately locking the energy centroid position of the resonant peak. The spectral centroid algorithm is used to track the position of the largest energy peak in the set of resonant peaks to determine the resonant center frequency.

[0028] In this embodiment, the resonant center frequency is associated with the parasitic parameters of the EMI filter, and the correlation relationship satisfies: ,in, Indicates the resonant center frequency. Represents pi (π). This represents the inductance value of the common-mode choke in an EMI filter. This indicates the capacitance value of the Y capacitor in the EMI filter.

[0029] Specifically, an equivalent circuit model of the EMI filter is established, and the common-mode choke is simplified to an inductor. The Y-capacitor to ground is simplified to a capacitor. Ignoring the effects of line resistance and other minor parasitic parameters, and based on the principle of LC series resonance, an expression for the natural frequency at which the circuit resonates is derived. This expression reveals the direct influence mechanism of changes in physical component parameters on frequency characteristics; that is, a decrease in inductance or capacitance will directly cause the resonant frequency to shift towards higher frequencies. Mathematical equations are then constructed based on this physical law. ,in, Indicates the resonant center frequency. Represents pi (π). This represents the inductance value of the common-mode choke in an EMI filter. This indicates the capacitance value of the Y capacitor in the EMI filter.

[0030] In this embodiment, the step of determining that the EMI filter has an aging fault specifically includes: monitoring the trend of frequency drift; if the frequency drift shows a monotonically increasing trend and the frequency drift is greater than a preset drift threshold, it is determined that the capacitance of the Y capacitor in the EMI filter has decayed.

[0031] Specifically, a time-series buffer is established, and the resonant center frequency calculated each time is recorded hourly to form a historical frequency trend line. A reference frequency value is set, which is the average value of the resonant center frequencies measured in the 24 hours before the initial operation of the equipment. The percentage change of the current measured frequency relative to the reference frequency is calculated. At the same time, linear regression analysis is performed on the frequency data of the most recent 30 consecutive sampling points, and the slope of the regression line is calculated. If the slope is positive and the correlation coefficient test shows a significant linear growth trend, and the calculated frequency drift exceeds a preset drift threshold, for example, 5% of the reference frequency, this 5% comes from the boundary of the capacitance tolerance range, indicating that the effective plate area of ​​the metallization film inside the Y capacitor is reduced due to the self-healing effect, which in turn causes an irreversible decrease in capacitance. Considering the above conditions, if the frequency drift shows a monotonically increasing trend and the frequency drift is greater than the preset drift threshold, it is determined that the capacitance of the Y capacitor in the EMI filter has decayed.

[0032] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for predicting switching faults in power supply cabinets, characterized in that, Includes the following steps: Collect the inductor terminal voltage and inductor current of the inductors in the power cabinet, as well as the input common-mode current of the EMI filter in the power cabinet; The real-time differential inductance value is calculated based on the current change rate of the inductor terminal voltage and the inductor current. The inductor current is divided into several current intervals, and the average differential inductance value in each current interval is statistically analyzed to construct a real-time inductor-current characteristic curve. Calculate the curve difference between the real-time inductor-current characteristic curve and the preset factory fingerprint curve, and determine the type of physical damage to the inductor based on the curve difference. Spectral analysis is performed on the input-side common-mode current to extract the resonant center frequency in the spectrum and calculate the frequency drift of the resonant center frequency relative to the preset reference frequency. If the frequency drift exceeds a preset drift threshold, the EMI filter is determined to have an aging fault.

2. The method for predicting switching faults in a power supply cabinet according to claim 1, characterized in that, The steps for calculating the real-time differential inductance value specifically include: The rising slope of the inductor current during the conduction of the switching transistor is obtained, and the rising slope is used as the rate of change of the current. The real-time differential inductance value is calculated using the differential inductance calculation formula; The formula for calculating the differential inductance is: ,in, This represents the real-time differential inductance value. This represents the voltage at the inductor terminals. This indicates the upward slope.

3. The method for predicting switching faults in a power supply cabinet according to claim 1, characterized in that, The steps for constructing a real-time inductor-current characteristic curve include: Set the current segmentation step size, and divide the numerical range of the inductor current into several current intervals according to the current segmentation step size; The average differential inductance value is obtained by summing all the real-time differential inductance values ​​falling within each current range and calculating the average value. Using the aforementioned current ranges as the horizontal axis and the corresponding average differential inductance value as the vertical axis, the real-time inductance-current characteristic curve is generated by fitting.

4. The method for predicting switching faults in a power supply cabinet according to claim 3, characterized in that, The steps for determining the type of physical damage to the inductor specifically include: Determine the low current range and the high current range from the aforementioned current ranges; If the average differential inductance value of the real-time inductance-current characteristic curve in the low current range is lower than the inductance value of the preset factory fingerprint curve in the same range, it is determined that the inductor has a crack fault with increased core air gap. If the inductance drop inflection point of the real-time inductance-current characteristic curve in the high current range is earlier than the inflection point of the preset factory fingerprint curve, it is determined that the inductor has a core saturation characteristic degradation fault.

5. The method for predicting switching faults in a power supply cabinet according to claim 1, characterized in that, The steps for calculating curve difference include: The curve difference degree is calculated using the weighted Euclidean distance algorithm; The formula for the weighted Euclidean distance algorithm is as follows: ,in, This indicates the degree of difference between the curves. This indicates the total number of the current ranges. This represents the k-th current interval being calculated. This represents the weighting coefficient for the k-th current interval. This represents the average differential inductance value of the real-time inductance-current characteristic curve in the k-th current interval. This represents the reference inductance value of the preset factory fingerprint curve in the k-th current range.

6. The method for predicting switching faults in a power cabinet according to claim 1, characterized in that, The steps for extracting the resonant center frequency from the spectrum specifically include: The input-side common-mode current is processed using a power spectral density analysis algorithm to obtain the common-mode current spectrum; Identify the local maxima of the common-mode current spectrum within a preset frequency band to obtain a set of resonance peaks; The position of the maximum energy peak in the set of resonance peaks is traced using the spectral centroid algorithm to determine the resonance center frequency.

7. The method for predicting switching faults in a power supply cabinet according to claim 6, characterized in that, The resonant center frequency is related to the parasitic parameters of the EMI filter, and the correlation relationship satisfies: ,in, Indicates the resonant center frequency, Represents pi (π). This represents the inductance value of the common-mode choke in the EMI filter. This indicates the capacitance value of the Y capacitor in the EMI filter.

8. The method for predicting switching faults in a power supply cabinet according to claim 1, characterized in that, The steps for determining that the EMI filter has an aging fault specifically include: Monitor the changing trend of the frequency drift; If the frequency drift shows a monotonically increasing trend and the frequency drift is greater than the preset drift threshold, it is determined that the capacitance of the Y capacitor in the EMI filter has decayed.

9. The system for predicting switching faults in a power cabinet according to any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire the inductor terminal voltage and inductor current of the inductors in the power cabinet, as well as the input common-mode current of the EMI filter in the power cabinet. The feature construction module is used to calculate the real-time differential inductance value based on the current change rate of the inductor terminal voltage and the inductor current, divide the inductor current into several current intervals, count the average differential inductance value in each current interval, and construct a real-time inductor-current feature curve. The damage determination module is used to calculate the curve difference between the real-time inductor-current characteristic curve and the preset factory fingerprint curve, and to determine the physical damage type of the inductor based on the curve difference. The spectrum analysis module is used to perform spectrum analysis on the input-side common-mode current, extract the resonant center frequency in the spectrum, and calculate the frequency drift of the resonant center frequency relative to the preset reference frequency. An aging fault determination module is used to determine that the EMI filter has an aging fault if the frequency drift exceeds a preset drift threshold.

Images