Capacitor self-healing short circuit detection method and system based on high-energy pulse cleaning

Abstract

The invention discloses a capacitor self-healing short-circuit detection method and system based on high-energy pulse cleaning, and relates to the technical field of electrical tests.The capacitor self-healing short-circuit detection method comprises the steps that firstly, environment simulation excitation with the stress level lower than the rated working condition is adopted to replace traditional high-voltage electric stress testing, and potential damage to a complete capacitor in the detection process is avoided; the method solves the inherent contradiction between the detection sensitivity and the product safety for a long time, realizes non-destructive detection, can safely perform deep quality screening on the precision capacitor which is subjected to secondary packaging, and can perform deep quality screening on the precision capacitor through synchronous acquisition and deep fusion analysis of electric, acoustic and thermal multi-source response signals. According to the method, a multi-dimensional feature vector capable of comprehensively representing a defect state is constructed, the defect recognition precision and reliability are improved, accurate diagnosis and targeted repair are integrated into a closed-loop self-healing system, and full-process automation from problem finding to problem solving is achieved.

Description

Technical Field

[0001] This invention relates to the field of electrical testing technology, specifically to a capacitor self-healing short-circuit detection method and system based on high-energy pulse cleaning. Background Technology

[0002] Aluminum electrolytic capacitors, as a key electronic component, are widely used in power filtering, energy storage, and coupling circuits of various electronic devices due to their advantages such as large capacitance per unit volume and relatively low cost. To improve the environmental tolerance and extend the service life of these products, especially in high-reliability applications, a secondary encapsulation process is often used after mounting on the mounting plate, involving resin injection to enhance sealing and slow down the evaporation of the internal electrolyte. Rigorous quality testing of these final-encapsulated capacitors, particularly checking for potential micro-short circuits, is a crucial step in ensuring their performance and reliability.

[0003] Existing technologies, such as patent applications with publication numbers CN119129409A, CN115032567B, and CN116930619A related to capacitor fault detection, show that existing methods for detecting capacitor short-circuit faults mainly include DC resistance testing, insulation resistance testing, and withstand voltage testing. DC resistance testing primarily determines the presence of a low-impedance hard short circuit by measuring the resistance between the capacitor's terminals, but its sensitivity is insufficient for detecting high-impedance or transient micro-short circuits. Insulation resistance and withstand voltage tests assess insulation performance by applying a high DC voltage to the capacitor and observing its leakage current or whether breakdown occurs. Furthermore, there are attempts to use pulse energy to eliminate micro-defects, i.e., applying an energy pulse to a capacitor that may have a short circuit, attempting to ablate or vaporize the short circuit point.

[0004] However, the aforementioned existing technologies have significant technical drawbacks. First, high-voltage testing methods are somewhat destructive; the applied strong electric field may cause irreversible damage to intact dielectric oxide films, and could even amplify a minor, repairable defect into a permanent hard short circuit, leading to the damage and scrapping of good products. Second, most of these methods can only provide a binary judgment result of pass or fail, failing to provide a refined diagnosis of the nature, location, and severity of defects, and lacking the ability to provide guidance for subsequent process improvements or repairs. For pulse scavenging methods, the application of energy is often non-targeted, making energy control difficult. Insufficient energy may lead to scavenging failure, while excessive energy may damage surrounding normal structures. The process lacks effective monitoring and verification, making reliability difficult to guarantee. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the present invention aims to provide a capacitor self-healing short-circuit detection method and system based on high-energy pulse cleaning. By employing environmental simulation excitation and multi-source signal fusion analysis, it can achieve non-destructive detection and targeted self-healing repair of hidden short-circuit points.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The first aspect of the present invention provides a capacitor self-healing short circuit detection method based on high-energy pulse cleaning, wherein S1, based on the acquired environmental test parameters, an environmental simulation excitation is applied to the capacitor under test to induce a characteristic response from the potential defects inside the capacitor under test.

[0007] S2. During the application of the environmental simulation excitation, the multi-source response signals of the capacitor under test are simultaneously acquired and fused to generate a defect activity feature vector.

[0008] S3. Based on the defect activity feature vector, determine whether there is a self-healing hidden short circuit point, and generate defect location information and activity level.

[0009] S4. In response to the determination that there is a self-healing latent short circuit point, based on the defect location information and the activity level, generate targeted repair energy parameters. If it is determined to be a non-self-healing latent short circuit point, then directly execute S7 to generate the final state of the capacitor under test.

[0010] S5. Based on the targeted repair energy parameters, inject targeted repair energy into the target area of ​​the capacitor under test to eliminate the hidden short circuit point.

[0011] S6. After injecting the targeted repair energy, perform a self-healing verification test on the capacitor under test to generate verification results.

[0012] S7. Based on the verification results, determine the final state of the capacitor under test.

[0013] A second aspect of the present invention provides a system for performing the capacitor self-healing short-circuit detection method based on high-energy pulse cleaning as described in the present invention, comprising: an environment simulation module for generating and applying the environment simulation excitation.

[0014] A multi-source signal acquisition module is used to synchronously acquire the multi-source response signals.

[0015] The signal processing and fusion module is used to receive the multi-source response signal and generate the defect activity feature vector.

[0016] The intelligent determination module is used to receive the defect activity feature vector and generate the defect location information and activity level.

[0017] The targeted energy generation and control module is used to generate and inject the targeted repair energy based on the defect location information and the activity level.

[0018] The central control and verification module is used to coordinate and control the above modules and perform the self-healing verification test to determine the final state.

[0019] The beneficial effects of the present invention are as follows: (1) The present invention surpasses the traditional detection technology by using a new paradigm of detection and repair with low stress and multi-physical field coupling. It adopts a mild environmental simulation excitation instead of the traditional high voltage electric stress test, avoids potential damage to intact capacitors during the detection process, solves the inherent contradiction between detection sensitivity and product safety in the long term, realizes non-destructive testing, and can safely perform in-depth quality screening on precision capacitors that have completed secondary packaging.

[0020] (2) This invention constructs a multi-dimensional feature vector that can comprehensively characterize the defect state by synchronously acquiring and deeply fusing and analyzing multi-source response signals of electricity, sound, and heat, thereby improving the accuracy and reliability of defect identification. This multi-dimensional diagnostic method can accurately distinguish between active short-circuit points with repair value and hard short-circuit points that should be scrapped directly, and provides their precise location and activity level information, elevating the detection results from the traditional pass / fail status to a new level of intelligent condition assessment and repair decision-making.

[0021] (3) This invention integrates precise diagnosis and targeted repair into a closed-loop self-healing system, realizing full-process automation from problem discovery to problem solving. It can customize and inject repair energy according to the activity level of the defect, and use the capacitor's own encapsulation structure to complete safe local repair. This not only improves the product yield, but also upgrades quality control from passive screening and elimination to proactive quality improvement and lifespan assurance, thereby enhancing the long-term reliability of the product from the source. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0024] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 As shown, the first aspect of the present invention provides a capacitor self-healing short circuit detection method based on high-energy pulse cleaning, comprising: S1, applying an environmental simulation excitation to the capacitor under test based on the acquired environmental test parameters to induce a characteristic response from potential defects inside the capacitor under test.

[0027] In a specific embodiment of the present invention, before the simultaneous acquisition of the multi-source response signal of the capacitor under test, the method further includes: arranging an acoustic wave sensor array on the surface of the resin encapsulation layer of the capacitor under test for acquiring the acoustic response signal.

[0028] An infrared thermal imager is placed around the capacitor under test to acquire the thermal response signal.

[0029] The acoustic sensor array, the infrared thermal imager, and the electrical measurement unit are synchronously triggered and calibrated to ensure that the timestamps of the multi-source response signals are synchronized.

[0030] It is important to note that since subsequent defect determination heavily relies on the precise temporal correspondence between the electrical, acoustic, and thermal signals, it is crucial to ensure that the operating cycles of all acquisition devices are completely consistent. The calibration process is typically achieved through a central synchronization signal generator, which simultaneously sends a trigger signal to the circuitry used to measure the electrical response signal, the data acquisition card for the acoustic sensor array, and the infrared thermal imager. All devices begin acquiring data simultaneously upon receiving the trigger signal, assigning a uniform, high-precision timestamp to each frame of data. This synchronized trigger calibration ensures that the electrical, acoustic, and thermal images at any given time point are strictly corresponding during subsequent analysis, thus laying a solid foundation for the effective fusion of multi-source response signals.

[0031] In a specific embodiment of the present invention, the environmental test parameters include the starting temperature, the ending temperature, the temperature change rate, and the number of cycles. In order to actively and gently induce potential latent defects inside the capacitor under test, this method adopts a composite environmental simulation excitation, which includes controlling the temperature of the environment in which the capacitor under test is located to cycle between the starting temperature and the ending temperature according to the temperature change rate, so as to form a thermal vibration environment.

[0032] It should be noted that this thermal vibration environment can cause microscopic mechanical stress in different materials inside the capacitor, such as electrode foil, dielectric film, and encapsulating resin, due to the difference in their coefficients of thermal expansion and contraction. This stress can slightly disturb potential defects, such as tiny metal burrs, and change their contact state.

[0033] During thermal vibration, a preset first voltage, i.e., a low-voltage DC bias, is continuously applied to the capacitor under test. At the same time, a high-energy pulse is instantaneously applied to the capacitor under test through a pulse generation module, with one high-energy pulse applied per cycle.

[0034] It should be noted that the preset low-voltage DC bias is much lower than the rated operating voltage of the capacitor. The preset low-voltage DC bias is set based on the design rated voltage of the capacitor and is 10%-30% of its rated voltage, excluding the rated voltage itself. The setting logic is that the low voltage range can avoid irreversible damage to the normal dielectric layer of the capacitor caused by high voltage, and can also form a local electric field concentration in the latent defects such as tiny metal burrs and microcracks in the dielectric layer by continuously applying a DC electric field. Combined with the mechanical stress generated by the thermal vibration environment, it can further promote the microscopic expansion and feature manifestation of defects, without interfering with the targeted nature of defect induction. For example, if the design rated voltage of the capacitor under test is 500V DC, the preset low-voltage DC bias can be set to 80V DC, 50V DC, or 120V DC.

[0035] The high-energy pulse is specifically a pulse voltage with an energy density higher than that of a low-voltage DC bias but with controlled total energy. The instantaneous power density of the pulse is sufficient to activate potential defects, but its duration is extremely short, so that the total energy integral injected into the capacitor is far below the threshold that causes thermal breakdown of the dielectric. Thus, it exhibits high-energy cleaning on a microscopic level while maintaining low-stress characteristics on a macroscopic level.

[0036] While a low-voltage DC bias alone is insufficient to cause any damage to a healthy medium, its presence is crucial. When thermal vibration alters the contact state at a defect point, the low-voltage DC bias creates a weak but detectable leakage current path, transforming a previously hidden, non-conductive physical defect into an electrical signal that can be detected by external circuitry. Therefore, this combination of thermal vibration and low-voltage bias constitutes an environmental simulation stimulus for inducing defect manifestation.

[0037] It should be noted that the environmental testing is specifically based on the design goals of the capacitor under test, while also adapting to the requirements for inducing latent defects. The design goals of the capacitor clearly define its expected operating environment, such as the design operating temperature range, the tolerance rate of temperature change, and the operating cycle requirements corresponding to the design service life. This provides a basic range for setting the starting temperature, ending temperature, and temperature change rate of this invention. On this basis, combined with the requirement of actively and gently inducing latent defects, the parameters are specifically optimized, such as adjusting the number of cycles, to ensure that potential defects are fully manifested, avoiding defect-induced failure or damage to the normal structure of the capacitor due to excessive deviation of parameters from the design goals. This ultimately forms the environmental testing parameters.

[0038] For example, regarding the starting and ending temperatures: based on the rated operating temperature range in the capacitor design objectives, extend it by 5-10°C towards its tolerance limit. This covers extreme operating conditions where potential defects are likely to appear, while avoiding damage to the normal structure caused by exceeding the tolerance threshold of the capacitor material.

[0039] Regarding the rate of temperature change: set it to 0.8-1.2 times the rate of temperature change under operating conditions in the design target to ensure that the micro-stress caused by temperature fluctuations is concentrated on latent defects such as micro-cracks and burrs, rather than inducing new non-defect damage.

[0040] Regarding the number of cycles: The termination condition is that no new defect signals are detected after 3 consecutive cycles. The initial number of cycles is set to 1 / 3 of the number of cycles corresponding to the service life in the design target. If no defect signal is detected, the number of cycles is gradually increased until the defect manifestation requirement is met or 2 / 3 of the design number of cycles is reached, so as to avoid excessive cycling and damage to the capacitor.

[0041] The technical advantage of this step lies in its innovative use of a low-stress, non-destructive defect induction method, avoiding the product damage risks associated with traditional high-voltage testing. By leveraging the differences in the thermophysical properties of materials, the gentle physical quantity of temperature change is transformed into an effective means of activating internal defects. The synchronous application of a low-voltage DC bias acts like a sensitive probe, instantly converting the microscopic physical changes induced by thermal vibration into measurable electrical signals, achieving an efficient conversion from physical disturbance to electrical response. This thermo-electric synergistic excitation method can trick latent defects that are dormant under normal temperature and pressure into actively exposing themselves, improving the sensitivity and coverage of detection, especially for potential fault points that may worsen due to temperature changes in the product's future actual operating environment, providing pre-screening capabilities. This changes the concept of defect detection from passive investigation to active induction, laying the foundation for truly non-destructive and efficient early defect diagnosis.

[0042] S2. During the application of the environmental simulation excitation, the multi-source response signals of the capacitor under test are simultaneously acquired and fused to generate a defect activity feature vector.

[0043] In a specific embodiment of the present invention, the multi-source response signal includes an electrical response signal, an acoustic response signal, and a thermal response signal; wherein, fusing the multi-source response signals to generate a defect activity feature vector includes: analyzing the electrical response signal and extracting a first feature set.

[0044] It should be noted that the analysis of the electrical response signal includes time-domain analysis and frequency-domain analysis. Time-domain analysis aims to directly observe the change of the current waveform over time, focusing on capturing current spikes or fluctuations caused by changes in the resistance of internal defects induced by thermal cycling, thereby extracting the pulsation amplitude and frequency of the leakage current in the first feature set. Frequency-domain analysis, for example through Fast Fourier Transform, can decompose complex current waveforms into components of different frequencies, helping to identify periodic noise or defect-specific frequency signals. In addition, correlation analysis is performed between the extracted current pulsation events and the synchronously recorded temperature change curves to calculate the correlation coefficient between them. This coefficient can effectively distinguish between random noise and defect responses strongly correlated with thermal excitation, enriching the first feature set. The current pulsation event refers to the instantaneous, non-stationary change in the resistance of local latent defects inside the capacitor, such as microcracks in the dielectric layer, tiny metal burrs, and non-conductive impurity particles, after being activated by energy impact under the environment of high-energy pulse applied to the capacitor under test and simultaneous thermal excitation, which forms instantaneous current spikes or periodic current fluctuations in the leakage current signal of the capacitor.

[0045] The acoustic response signal is processed to extract a second feature set.

[0046] It should be noted that the acoustic response signals captured by the acoustic wave sensor can reveal minute discharges or material stress releases that may occur when defects such as burrs exist inside the capacitor under electrothermal stress. These events propagate outward as high-frequency elastic waves, or acoustic emission signals. By processing the acquired acoustic response signals, such as filtering and threshold detection, these high-frequency acoustic emission events representing internal micro-discharges or material stress releases can be accurately identified. Statistical analysis of the energy peak value, frequency per unit time, peak signal amplitude, and pulse width of these events constitutes the second feature set.

[0047] The thermal response signal is analyzed to extract the third feature set.

[0048] It should be noted that the thermal response signal recorded by an infrared thermal imager, if there is a continuously discharging short circuit point inside, will become a tiny heat source. Although this heat source is very weak, the heat it generates will be conducted to the surface through the encapsulation material. By analyzing a continuous sequence of thermal images using image processing algorithms, stable micro-regions of temperature difference, distinct from the surrounding background, can be identified on the surface of the capacitor package under test, i.e., temperature anomalous points. Extracting the location of these temperature anomalous points in the thermal image, their temperature difference amplitude with the background, and the fluctuation range of this temperature difference over time constitutes the third feature set. The fluctuation range of the temperature difference over time can be specifically calculated by selecting a continuous sequence of thermal images of 10-30 seconds as the analysis window, and calculating the fluctuation range of the temperature difference amplitude within this window (fluctuation range = (maximum value - minimum value) / average value × 100%).

[0049] The first feature set, the second feature set, and the third feature set are time-synchronized and sampled and normalized based on a unified system clock, and then correlated and aligned in the time dimension to generate the defect activity feature vector. The defect activity feature vector is used to characterize the multiphysics coupling response mode of the potential defect.

[0050] It should be noted that the above method replaces traditional high-voltage testing with a synergistic induction mode of low-voltage DC bias, instantaneous high-energy pulse, and thermal vibration. This innovatively combines electrical, acoustic, and thermal signals for simultaneous acquisition and deep fusion, generating a multi-dimensional feature vector that comprehensively characterizes defect activity. The introduction of the instantaneous high-energy pulse plays a crucial synergistic role: on the one hand, its instantaneous energy impact can precisely activate tiny, active, latent short-circuit points that are difficult to detect under conventional low-voltage conditions, while gently cleaning up existing tiny short-circuit points to prevent defect signals from being masked; on the other hand, the synergistic effect of the high-energy pulse, low-voltage bias, and thermal vibration further amplifies the differences in multi-physics responses in the defect region, such as electric field concentration, enhanced stress release, and improved heat generation efficiency at the defect location, thereby improving the characteristic identification of electrical, acoustic, and thermal signals.

[0051] This fusion analysis approach enables the system to accurately identify and locate minute, active, hidden short circuits that are easily overlooked in conventional testing, under extremely low electrical stress conditions without causing any damage to the capacitor. It utilizes the inherent correlation between multi-physics field signals to construct a unique defect fingerprint, improving the sensitivity and reliability of the detection and avoiding misjudgments or missed detections caused by a single information dimension. Simultaneously, this method transforms the resin encapsulation layer, originally a protective barrier, into an effective medium for transmitting internal acoustic and thermal information, achieving synergistic effects between the detection technology and the product packaging structure. It resolves the contradiction between sensitivity and safety in traditional non-destructive testing methods, providing a new technical path to improve the yield and long-term reliability of repackaged capacitors.

[0052] S3. Based on the defect activity feature vector, determine whether there is a self-healing hidden short circuit point, and generate defect location information and activity level.

[0053] In a specific embodiment of the present invention, determining whether there is a self-healing hidden short circuit based on the defect activity feature vector and generating defect location information and activity level includes: inputting the defect activity feature vector into the defect determination model to calculate the probability that the defect activity feature vector belongs to each defect category, thereby obtaining the defect classification probability distribution.

[0054] It's important to note that the defect identification model is not a simple threshold comparator, but rather a machine learning classifier built upon multiphysics response patterns. A machine learning classifier is one that is pre-trained using a large number of known states, including defect-free, self-healing short-circuit, and hard short-circuit capacitors, to learn and master the complex correspondence between different defect states and multiphysics signal patterns, thus acquiring the ability to intelligently identify and classify defects. When a new defect-active feature vector is input, the defect identification model uses its learned knowledge to calculate the probability that the defect-active feature vector belongs to each preset defect category. These preset defect categories mainly include three types: self-healing active short-circuit points, non-self-healing hard short-circuit points, and defect-free states.

[0055] Based on the defect classification probability distribution, it is determined whether the self-healing hidden short circuit point exists, and a defect judgment conclusion is generated.

[0056] It should be noted that the specific method for determining whether the self-healing hidden short circuit point exists is as follows: if the probability of a self-healing active short circuit point exceeds a preset first threshold, then it is confirmed that there is a self-healing hidden short circuit point inside the capacitor under test.

[0057] In response to the defect determination conclusion that there is a self-healing hidden short circuit, the location information of the temperature difference abnormal point contained in the third feature set in the defect active feature vector is analyzed to generate the defect location information.

[0058] The activity level is calculated and generated by combining the values ​​of the first feature set and the second feature set in the defect activity feature vector.

[0059] It should be noted that the specific calculation method of the activity level is as follows: the measured values ​​of each parameter in the first feature set and the second feature set are normalized, and the influence weights of each parameter on the defect activity in the database are weighted and summed to obtain a feature comprehensive score. Based on the preset activity level and feature comprehensive score mapping relationship table, the activity level is mapped to obtain the activity level. The influence weights of each parameter on the defect activity are, for example, a leakage voltage pulsation amplitude weight of 0.3 and an acoustic signal amplitude weight of 0.2. The influence weights are specifically determined by prior experimental data statistics and stored in the database. The preset activity level and feature comprehensive score mapping relationship table is as follows: a feature comprehensive score ≥ 0.7 is a first-level activity level, 0.3-0.7 is a second-level activity level, and < 0.3 is a third-level activity level, corresponding to the activity level of the generated defect.

[0060] The technical advantage of this step lies in its ability to accurately classify and deeply analyze the internal latent defect state of capacitors by introducing an intelligent judgment model based on multi-physics modes. Compared to traditional methods that rely on a single physical quantity for judgment, this invention can identify the complex coupling relationships between electrical, acoustic, and thermal signals, improving diagnostic accuracy, especially in effectively distinguishing between active short-circuit points with repair value and hard short-circuit points that should be scrapped directly. This refined classification capability is the prerequisite and foundation for subsequent self-healing operations, avoiding the erroneous scrapping of repairable products and preventing ineffective repair attempts on hard short-circuit points. Simultaneously, by backtracking and extracting location and activity information from the fused feature vectors, the target of subsequent repair operations is clearly defined, and the energy delivery is precise, transforming energy impact into targeted minimally invasive surgery. This provides crucial decision support for the successful implementation of the entire self-healing detection method, improving the level of intelligent production and product yield.

[0061] S4. In response to the determination that there is a self-healing latent short circuit point, based on the defect location information and the activity level, generate targeted repair energy parameters. If it is determined to be a non-self-healing latent short circuit point, then directly execute S7 to generate the final state of the capacitor under test.

[0062] In a specific embodiment of the present invention, the targeted repair energy parameters include the frequency, amplitude, duration, and waveform of the energy injection; generating the targeted repair energy parameters includes: acquiring the electrical parameters of the capacitor under test.

[0063] The electrical parameters include the rated voltage, nominal capacitance, equivalent series resistance, insulation resistance, and dielectric loss tangent of the capacitor under test. These parameters are the basic electrical properties of the capacitor before it leaves the factory or is tested, and can be directly obtained through conventional electrical testing equipment.

[0064] Based on the activity level, the reference amplitude and duration of energy injection are determined based on the mapping table between the activity level and the reference amplitude and duration.

[0065] It should be noted that defects with higher activity levels may be more stubborn or slightly larger, thus requiring a relatively higher reference amplitude or a slightly longer processing time to ensure effective removal; conversely, small defects with lower activity levels correspond to milder energy parameters to avoid energy excess.

[0066] The mapping table between the activity level and the reference amplitude and duration was established in advance through experimental calibration. The experiment selected capacitor samples containing standard defects of various activity levels, fixing the low-voltage DC bias, thermal excitation cycle, and the timing of high-energy pulse application, adjusting only the reference amplitude and total duration of energy injection. Multiple rounds of experiments were repeated for each set of parameters. The acceptance criterion was complete defect cleaning and no irreversible damage to the capacitor dielectric layer. The optimal parameter combination corresponding to each activity level sample was recorded. After stability verification and elimination of discrete data, the precise correspondence between each activity level and the reference amplitude and duration was determined, ultimately forming a directly callable mapping table. Those skilled in the art can reproduce the mapping table establishment process based on the above steps.

[0067] For example, if the defect activity level is Level 1, the corresponding energy injection reference amplitude is set to 28% of the rated voltage of the capacitor under test, and the duration is set to 65ms; if it is Level 2, the reference amplitude is set to 20% of the rated voltage, and the total duration is set to 100ms; if it is Level 3, the reference amplitude is set to 12% of the rated voltage, and the total duration is set to 150ms. The reference amplitude decreases with decreasing activity level, while the duration increases with decreasing activity level, adapting to the energy requirements of different active defects. This ensures that highly active defects are effectively treated while preventing low-activity defects from being damaged by excessive energy.

[0068] By combining the electrical parameters of the capacitor under test with the estimated resistive characteristics of the hidden short circuit point, a specific frequency range is selected.

[0069] The estimated resistive characteristics refer to the range of short-circuit resistance values, temperature coefficient of resistance, and resistance fluctuation characteristics with voltage that are estimated in advance based on common causes of latent short circuits in capacitors, such as conduction by metal burrs and conduction by microcracks in the dielectric layer, combined with a database of defects in similar capacitors.

[0070] The specific method for selecting a specific frequency range is a combination of table lookup and simulation verification: a reference table is pre-established to correlate capacitor electrical parameters, estimated resistive characteristics at short-circuit points, and the optimal excitation frequency range. The reference table is generated by fitting a large amount of experimental data from defect detection of similar capacitors. In practical applications, the obtained electrical parameters and estimated resistive characteristics of the capacitor to be tested are substituted into the reference table to obtain a preliminary specific frequency range. Then, the electric field response intensity of the defect region at this frequency is simulated using circuit simulation software, and the final specific frequency range is determined after fine-tuning the frequency boundaries.

[0071] It should be noted that, according to the principle of parallel current division in circuits, to ensure that the repair current primarily flows through the resistive short-circuit point rather than the capacitive capacitor body, a signal with a capacitive reactance much greater than the short-circuit resistance, such as a low-frequency band (<100Hz) or a predominantly DC component, should be selected. When a low-frequency or DC pulse signal of this specific frequency is applied, the current will naturally choose the path of lowest impedance. The purpose of this step is to find a specific frequency range at which the normal capacitor dielectric exhibits high capacitive reactance, acting like an insulator, while the short-circuit path exhibits low impedance. When an AC signal of this specific frequency is applied, the current will naturally choose the path of lowest impedance, allowing most of the energy to selectively couple to the hidden short-circuit point, achieving targeted energy delivery.

[0072] Based on the reference amplitude, total duration, and specific frequency range, the waveform of the energy injection is determined in combination with preset waveform rules to form the targeted repair energy parameters.

[0073] The waveform is not a simple AC signal, but is designed to be a low-frequency signal or pulsed DC signal with a frequency below 100Hz superimposed on the low-voltage DC bias that is still being applied by the ambient simulation excitation.

[0074] It should be noted that the preset logic of the basic waveform type rules is as follows: Based on a large amount of experimental data and circuit simulation results of targeted repair of defects in similar capacitors, a general rule is pre-written, which prioritizes the selection of a square wave superimposed on the low-voltage DC bias continuously applied during the environmental simulation excitation stage as the default waveform. This superimposed square wave has concentrated energy and a steep rising edge, which can quickly and accurately act on the defect area, and its energy utilization rate is higher than that of a sine wave. The duty cycle of the square wave is preset to 50%. At the same time, a sine wave superimposed on the same low-voltage DC bias is set as the alternative waveform, and its initial phase is preset to 0°. The above basic rules do not need to be repeated for conventional capacitors under test and can be directly called.

[0075] For scenarios requiring special presets: Special presets are only needed for certain types of capacitors under test, such as high-voltage large-capacity capacitors, thin-film dielectric capacitors, and high-sensitivity precision capacitors. The specific operation involves combining the electrical parameters of the capacitor under test and the estimated resistive characteristics of the defects, and matching the corresponding fine-tuning rules from a pre-established capacitor model-defect type-waveform parameter lookup table. For example, the square wave duty cycle can be adjusted to 30%-40%, or switched to a pulsed sine wave. The superimposed low-voltage DC bias can be kept constant or finely adjusted according to the capacitor's withstand voltage characteristics to ensure that the superimposed energy injection waveform accurately matches the capacitor characteristics, avoiding damage to the dielectric layer during the repair process.

[0076] Meanwhile, considering the impact of the temperature state of the current environmental simulation excitation on the repair efficiency, the energy parameters are fine-tuned by matching the fine-tuning coefficient corresponding to the current temperature state according to the temperature state-fine-tuning ratio mapping table. The temperature state-fine-tuning ratio mapping table is calibrated based on the defect-targeted repair experimental data of multiple sets of capacitors of different models. The experiment selects capacitor samples containing defects of various activity levels, and the criteria for judgment are complete defect repair and no damage to the dielectric layer. The parameter values ​​corresponding to the optimal repair effect are determined after statistical analysis.

[0077] For example, the fine-tuning rules are as follows: when the temperature reaches the extreme value of the thermal excitation cycle and the rate of change is 0, i.e., the temperature is in a steady state, the energy parameters are fine-tuned to increase the reference amplitude by 6%, shorten the total duration by 12%, and keep the frequency range unchanged.

[0078] When the temperature is in the heating / cooling stage and the rate of change is ≥5℃ / min, i.e., the temperature is in a dynamic change state, the energy parameters are fine-tuned to reduce the reference amplitude by 10%, extend the total duration by 18%, and shift the frequency to the lower frequency range by 8%.

[0079] When the ambient simulation excitation start temperature is below 25°C, i.e., in low-temperature start-up state, the superimposed low-voltage DC bias will be increased by 4%.

[0080] The technical advantage of this step lies in achieving precise guidance and on-demand allocation of repair energy, abandoning the one-size-fits-all, crude energy application method of traditional pulse cleaning. By actively transforming passively received defect activity level information into refined control of the four key dimensions of repair energy, this method ensures that each repair operation is highly targeted. In particular, the strategy of frequency selection using impedance differences applies physical principles to the targeted energy allocation, allowing energy to accurately hit the defect and avoid affecting innocent healthy areas. This ability to customize repair solutions based on individual defect characteristics not only improves the success rate of repair but, more importantly, reduces collateral damage to intact parts of the capacitor during the repair process, ensuring stable and reliable product performance after repair. This seamless connection from diagnosis to prescription is a key link in achieving efficient and safe self-healing, reflecting a high degree of intelligence and refined process control.

[0081] S5. Based on the targeted repair energy parameters, inject targeted repair energy into the target area of ​​the capacitor under test to eliminate the hidden short circuit point.

[0082] In a specific embodiment of the present invention, the injected targeted repair energy includes: maintaining the temperature conditions in the environmental simulation excitation so that the resin encapsulation layer of the capacitor under test is in a heat preservation state.

[0083] Based on the targeted repair energy parameters, a low-frequency signal or pulsed DC signal with a frequency below 100Hz is generated and applied to the capacitor under test.

[0084] Specifically, the test fixture is applied to the two electrodes of the capacitor under test.

[0085] The heat capacity and low thermal conductivity of the resin encapsulation layer are used to locally restrict and accumulate the Joule heat generated by the targeted repair energy. At the same time, the thermal conductivity path of the electrolyte inside the capacitor under test is used for synergistic thermal management, so as to achieve local melting or oxidation insulation of the metal bridge at the hidden short circuit point.

[0086] It should be noted that, since the frequency selection process has already prioritized the flow of the AC signal energy through the low-resistance hidden short-circuit point, most of the current will converge at this defect after the signal is applied, generating Joule heat due to the resistance. At this point, a key synergistic effect of the present invention becomes apparent: the full utilization of the physical properties of the resin encapsulation layer of the capacitor under test. This resin material typically has high heat capacity and low thermal conductivity, meaning it can store a significant amount of heat without easily dissipating it. Therefore, the Joule heat generated at the short-circuit point is locally confined and accumulated by this natural insulation layer, causing the temperature at the defect point to rise rapidly in a short time. Simultaneously, synergistic thermal management is implemented on the electrolyte heat conduction path inside the capacitor to achieve controlled thermal melting of the hidden short-circuit point. As a liquid, the electrolyte has relatively good thermal conductivity, effectively carrying away some heat from the other side of the short-circuit point, preventing excessive heat diffusion into the depths of the capacitor core and causing irreversible damage. This combination of heat generation from the encapsulation layer and temperature control from the electrolyte ultimately eliminates short-circuit paths by using locally concentrated heat to melt and passivate tiny metal burrs that could cause short circuits, or to oxidize and ablate conductive impurities, all within a relatively controllable temperature range. The energy and temperature of the entire process are controlled within safe thresholds that will not damage the surrounding intact oxide film and encapsulation structure.

[0087] The technical advantage of this step lies in its innovative transformation of the capacitor's encapsulation structure and internal material properties into active participants in achieving a precise and safe repair process, forming a microscopic intelligent thermal management system. By utilizing the synergistic effect between the thermal insulation effect of the resin encapsulation layer and the thermal conductivity effect of the electrolyte, it achieves efficient focusing and safe control of energy at internal micro-defects without the use of complex external heating or cooling devices. This not only simplifies the structure of the repair device and reduces costs, but more importantly, by utilizing the synergistic physical mechanism of the encapsulation layer restricting heat diffusion and the electrolyte assisting in heat dissipation, it achieves efficient focusing and safe control of energy at internal micro-defects without the use of complex external heating or cooling devices. The resulting localized thermal melting or oxidation effect ensures the effectiveness of defect removal while protecting the overall structure and long-term performance of the capacitor. This is a physical process that achieves high success rate and high reliability self-healing repair, fully demonstrating the effect of multi-factor synergy in the technical solution.

[0088] S6. After injecting the targeted repair energy, perform a self-healing verification test on the capacitor under test to generate verification results.

[0089] In a specific embodiment of the present invention, the self-healing verification test of the capacitor under test includes: before performing the environmental simulation excitation, acquiring a set of baseline signals of the capacitor under test under a reference environmental state.

[0090] The environmental simulation excitation and the targeted repair energy are removed, so that the capacitor under test is restored to the baseline environmental state.

[0091] It should be noted that the restoration of the capacitor under test to the reference environmental state refers to the natural state after the external excitation is removed, allowing for a certain range of physical parameter drift, but the electrical behavior should return to stability.

[0092] Under the aforementioned baseline environmental conditions, a set of verification signals are acquired for the capacitor under test.

[0093] Specifically, under the reference environmental conditions, a preset low-voltage DC bias for verification is applied to the capacitor under test. This bias is the same as the DC bias used during the initial defect induction, with the aim of evaluating the capacitor's electrical performance under mild and consistent conditions. Simultaneously with this bias, signal acquisition is restarted, but this time the focus is primarily on two key indicators: the steady-state leakage current signal and the background acoustic emission signal.

[0094] The verification signal is compared with the baseline signal to generate the verification result.

[0095] It should be noted that if the steady-state leakage current signal has returned to normal and remains stable, and the defect characteristic events in the background acoustic emission signal have completely disappeared, then the repair is considered successful, and a self-healing success verification result is generated, indicating that the latent defect has been repaired. Conversely, if the leakage current is still too large or unstable, or if the acoustic emission of the defect characteristic can still be detected, then a self-healing failure verification result is generated, indicating that the latent defect has not been completely repaired, and that the repair operation has not achieved the expected results.

[0096] It should also be noted that the method for determining whether the steady-state leakage current signal has returned to normal and remained stable is as follows: the repaired steady-state leakage current value is compared with the baseline leakage current signal in the baseline signal collected under the reference environment. If the repaired value falls within the fluctuation range of the baseline leakage current signal of the preset ratio, such as ±5%, and remains unchanged for a preset time, such as 30s, without significant fluctuation, it is determined that it has returned to normal and remained stable.

[0097] The method for determining the complete disappearance of defect feature events in the background acoustic emission signal is as follows: the acoustic emission frequency range of the defect feature events corresponding to the hidden short circuit point is pre-calibrated, such as 100kHz-500kHz. The repaired background acoustic emission signal is compared with the baseline acoustic emission signal in the baseline signal. If the amplitude of the repaired signal in the characteristic frequency range is lower than the baseline background noise amplitude, and no such feature event is detected for a continuous preset time, the defect feature event is determined to have completely disappeared.

[0098] The technical advantage of this step lies in establishing an objective, reliable, and highly confident closed-loop verification mechanism. It assesses the effectiveness of the repair by comparing the repaired state with the original healthy baseline state, avoiding potential misjudgments based on a single measurement. This verification logic ensures the uniformity and impartiality of the evaluation standards. Simultaneously, the verification process not only focuses on the restoration of electrical parameters but also monitors the silent state of acoustic signals, achieving dual confirmation from both electrical and acoustic dimensions, further enhancing the reliability of the verification results. This multi-dimensional, baseline-based verification method provides conclusive evidence of the success of self-healing, ensuring that only truly repaired and stable products are deemed qualified. This provides a solid guarantee for the effectiveness of the entire self-healing short-circuit detection method and the quality of the final output product, forming an indispensable link in the complete process from diagnosis to treatment to rehabilitation assessment.

[0099] In a specific embodiment of the present invention, after generating the verification result, the method further includes: in response to the verification result being a self-healing failure, obtaining the current defect activity feature vector.

[0100] The verification results of the self-healing failure and the current defect activity feature vector are stored as learning samples in the sample database.

[0101] The judgment threshold of the defect judgment model is periodically calibrated using the accumulated data in the sample database, and based on the verification results of the self-healing failure, it is determined that the hidden short circuit point has evolved into a hard short circuit point that cannot be self-healed, and the final state of the capacitor under test is directly updated.

[0102] It should be noted that the specific dimensions of the defect activity feature vector are: steady-state leakage current amplitude and fluctuation coefficient, frequency and amplitude of defect feature acoustic emission signal, quantification value of defect activity level, ambient temperature during detection, and electrical parameters of the capacitor under test. Using this new information, the internal parameters in the defect judgment model are automatically updated. For example, the judgment threshold used to distinguish different defect categories can be adjusted: the threshold for the corresponding category can be fluctuated up or down by 5%-8% based on the deviation ratio between the defect activity feature vector in failed cases and the original judgment threshold; or the classification boundary of the machine learning classifier can be fine-tuned: using gradient descent with a learning rate of 0.01 and 50 iterations, the classification boundary is iteratively optimized based on new learning samples. In this way, the defect judgment model can learn from failed experiences, and its judgment will be more accurate when encountering similar defect patterns again.

[0103] Based on the verification results of the self-healing failure, it is determined that the latent short circuit point has evolved into an irreversible hard short circuit point, and the final state of the capacitor under test is directly updated, which is a reassessment of the current state of the capacitor under test. Based on the verification results of the self-healing failure, the various parameters of the capacitor after the repair failure are re-analyzed, and it is determined whether the latent short circuit point has undergone a change in nature during the repair attempt. For example, if the steady-state leakage current amplitude of the capacitor under test suddenly rises to more than 10 times the baseline signal and shows no downward trend within 10 seconds, and at the same time the background sound emission signal shows a continuous strong pulse signal in the hard short circuit characteristic frequency range such as 800kHz-1200kHz, then it is determined that the defect has evolved into an irreversible and more serious irreversible hard short circuit point due to energy injection. If such a judgment is made, the final state of the capacitor under test is updated, that is, the defect has evolved into an irreversible hard short circuit point, and it is directly marked as an unrepairable scrap, and any further repair attempts are terminated.

[0104] The technical advantage of this step lies in introducing intelligent feedback and adaptive learning capabilities into the entire inspection method, transforming it from a fixed linear process into a self-improving dynamic closed-loop system. Through learning from failed cases, the defect judgment model continuously evolves, and its diagnostic accuracy increases with the number of processed samples, demonstrating the system's high level of intelligence and long-term optimization potential. Simultaneously, the mechanism for re-evaluating the product status after failure ensures timely loss mitigation, avoiding wasted time and resources on products beyond repair and improving the efficiency of the entire production process. This mechanism not only enhances the accuracy of individual inspections but, more importantly, strengthens the robustness and long-term reliability of the entire system, making this invention not merely an inspection tool but a self-evolving intelligent quality control platform.

[0105] S7. Based on the verification results, determine the final state of the capacitor under test.

[0106] The final state of the capacitor under test includes the latent defect being repaired, the latent defect not being completely repaired, and the latent defect evolving into an irreversible hard short circuit.

[0107] This invention proposes a novel technical principle for capacitor quality inspection and repair. Its core idea is to transform passive detection into active induction, single diagnosis into multi-dimensional fusion, and repair into precise self-healing. This principle abandons the traditional detection method of directly applying high-voltage electrical stress. Instead, it first constructs an environmental simulation excitation based on the capacitor's own electrical parameters and preset environmental test parameters, simulating the harsh operating conditions it might encounter in the future. This excitation is not a single electrical signal, but a gentle perturbation of thermo-electric coupling, aimed at inducing latent defects that are dormant under normal conditions to actively manifest. During the excitation process, it does not only focus on the electrical signal, but simultaneously captures the capacitor's subtle responses from multiple physical dimensions, including electrical, acoustic, and thermal aspects, forming multi-source response signals. Subsequently, advanced signal processing technology deeply fuses these signals from different sources but with inherent connections, generating a defect activity feature vector that comprehensively and three-dimensionally describes the defect state. Based on this multi-dimensional vector and a preset intelligent defect judgment model, it can accurately determine whether there are repairable hidden short circuit points and simultaneously output their spatial location and activity level. Once a repairable target is identified, based on this precise information, targeted repair energy parameters specifically designed for that defect are generated and reverse-engineered. Energy is then injected into the target area in a highly controlled manner, utilizing the physical properties of the package itself to collaboratively eliminate the defect. After repair, all excitation is removed, allowing the capacitor to return to normal operation. A rigorous self-healing verification test is then performed under this baseline state, and the final product status is determined based on the verification results.

[0108] The beneficial technical effects of this solution are as follows: First, it resolves the sharp contradiction between sensitivity and safety in traditional testing methods. By replacing high-voltage impact with low-stress induction, non-destructive testing is achieved, protecting intact capacitors from test damage. Second, through the fusion analysis of multi-source signals, the accuracy and reliability of defect identification are improved, enabling the discovery and precise diagnosis of early, latent defects that traditional single electrical testing methods cannot reach, thus shifting the quality control checkpoint forward. More importantly, this invention integrates diagnosis and treatment, creating a closed-loop self-healing process that can repair a large number of products that would otherwise be considered scrap but are actually repairable, improving production yield and economic efficiency. This method utilizes the synergistic effect of the packaging structure and internal materials to achieve precise control of repair energy, ensuring the safety and efficiency of the repair process. Finally, through a complete induction-diagnosis-repair-verification process, this invention can not only eliminate irreparable defective products but also repair and verify recoverable products, eliminating the risk of early product failure and improving overall reliability and service life.

[0109] Reference Figure 2 As shown, a second aspect of the present invention provides a system for performing the capacitor self-healing short-circuit detection method based on high-energy pulse cleaning as described in the present invention, comprising: an environment simulation module for generating and applying the environment simulation excitation.

[0110] A multi-source signal acquisition module is used to synchronously acquire the multi-source response signals.

[0111] The signal processing and fusion module is used to receive the multi-source response signal and generate the defect activity feature vector.

[0112] The intelligent determination module is used to receive the defect activity feature vector and generate the defect location information and activity level.

[0113] The targeted energy generation and control module is used to generate and inject the targeted repair energy based on the defect location information and the activity level.

[0114] The central control and verification module is used to coordinate and control the above modules and perform the self-healing verification test to determine the final state.

[0115] It should also be noted that the various threshold settings described in this invention, including the steady-state leakage current judgment threshold, the acoustic emission defect characteristic frequency band threshold, the temperature change rate judgment threshold, and the energy parameter fine-tuning ratio threshold, are specifically based on historical operating data of capacitor latent defect detection and repair, typical defect evolution failure cases, and experimental test results of a large number of different capacitor models. Simultaneously, combined with the experience of experts in the field of power electronic equipment, the thresholds are calibrated by comprehensively considering factors such as the actual operating environment of the capacitor under test, the design life and performance indicators of the capacitor dielectric layer and electrode foil, and the safety threshold of the targeted repair energy for the capacitor body. This has a solid scientific basis and practical operability. Furthermore, the relevant threshold calibration techniques in the prior art are relatively mature and will not be elaborated upon here. The mapping tables and model parameters are all calibrated and optimized based on a large amount of experimental data. For example, the mapping table between activity level and repair parameters is obtained through systematic testing on standard defect samples to ensure repair effectiveness and safety. The defect judgment model, if employing a convolutional neural network structure, takes a feature vector fused from multiple sources as input and outputs a defect category probability. The training data comes from a measured signal library of capacitors in known states.

[0116] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0118] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0119] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A capacitor self-healing short-circuit detection method based on high-energy pulse cleaning, characterized in that, include: S1. Based on the acquired environmental test parameters, apply environmental simulation excitation to the capacitor under test to induce characteristic responses from potential defects inside the capacitor under test. S2. During the application of the environmental simulation excitation, the multi-source response signals of the capacitor under test are simultaneously acquired and fused to generate a defect activity feature vector. S3. Based on the defect activity feature vector, determine whether there is a self-healing hidden short circuit point, and generate defect location information and activity level. S4. In response to the determination that there is a self-healing latent short circuit point, based on the defect location information and the activity level, generate targeted repair energy parameters. If it is determined to be a non-self-healing latent short circuit point, then directly execute S7 to generate the final state of the capacitor under test. S5. Based on the targeted repair energy parameters, inject targeted repair energy into the target area of ​​the capacitor under test to eliminate the hidden short circuit point. S6. After injecting the targeted repair energy, perform a self-healing verification test on the capacitor under test to generate verification results. S7. Based on the verification results, determine the final state of the capacitor under test.

2. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 1, characterized in that, Before synchronously acquiring the multi-source response signal of the capacitor under test, the method further includes: An array of acoustic wave sensors is arranged on the surface of the resin encapsulation layer of the capacitor under test to collect the acoustic response signal. An infrared thermal imager is placed around the capacitor under test to acquire the thermal response signal; The acoustic sensor array, the infrared thermal imager, and the electrical measurement unit are synchronously triggered and calibrated to ensure that the timestamps of the multi-source response signals are synchronized.

3. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 1, characterized in that, The environmental test parameters include the starting temperature, ending temperature, temperature change rate, and number of cycles; the environmental simulation stimulus includes: The temperature of the environment in which the capacitor under test is located is controlled to cycle between the starting temperature and the ending temperature according to the temperature change rate, so as to form a thermal vibration environment. During thermal vibration, a preset first voltage is continuously applied to the capacitor under test, and a high-energy pulse is instantaneously applied to the capacitor under test through a pulse generation module, with one high-energy pulse applied per cycle.

4. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 1, characterized in that, The multi-source response signals include electrical response signals, acoustic response signals, and thermal response signals; wherein, fusing the multi-source response signals to generate a defect activity feature vector includes: The electrical response signal is analyzed to extract a first feature set; The acoustic response signal is processed to extract the second feature set; The thermal response signal is analyzed to extract the third feature set; The first feature set, the second feature set, and the third feature set are correlated and aligned in the time dimension to generate the defect activity feature vector, which is used to characterize the multiphysics coupling response mode of the potential defect.

5. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 4, characterized in that, The step of determining whether a self-healing hidden short circuit exists based on the defect activity feature vector, and generating defect location information and activity level, includes: The defect activity feature vector is input into the defect determination model to calculate the probability that the defect activity feature vector belongs to each defect category, thereby obtaining the defect classification probability distribution. Based on the defect classification probability distribution, determine whether the self-healing hidden short circuit point exists and generate a defect judgment conclusion. In response to the defect determination conclusion that there is a self-healing hidden short circuit point, the location information of the temperature difference abnormal point contained in the third feature set in the defect active feature vector is analyzed to generate the defect location information. The activity level is calculated and generated by combining the values ​​of the first feature set and the second feature set in the defect activity feature vector.

6. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 5, characterized in that, The targeted repair energy parameters include the frequency, amplitude, duration, and waveform of the energy injection; The generated targeted repair energy parameters include: Obtain the electrical parameters of the capacitor under test; Based on the activity level, the reference amplitude and duration of energy injection are determined based on the mapping table between the activity level and the reference amplitude and duration. By combining the electrical parameters of the capacitor under test with the estimated resistive characteristics of the hidden short circuit point, a specific frequency range is selected; Based on the reference amplitude, total duration, and specific frequency range, the waveform of the energy injection is determined in combination with preset waveform rules to form the targeted repair energy parameters.

7. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 1, characterized in that, The injected targeted repair energy includes: Maintain the temperature conditions in the environmental simulation excitation so that the resin encapsulation layer of the capacitor under test is kept in a thermal state; Based on the targeted repair energy parameters, a low-frequency signal or a pulsed DC signal is generated and applied to the capacitor under test; The heat capacity and low thermal conductivity of the resin encapsulation layer are used to locally restrict and accumulate the Joule heat generated by the targeted repair energy. At the same time, the thermal conductivity path of the electrolyte inside the capacitor under test is used for synergistic thermal management, so as to achieve local melting or oxidation insulation of the metal bridge at the hidden short circuit point.

8. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 1, characterized in that, The self-healing verification test of the capacitor under test includes: Before performing the environmental simulation excitation, a set of baseline signals of the capacitor under test under reference environmental conditions is acquired; Remove the environmental simulation excitation and the targeted repair energy to restore the capacitor under test to the baseline environmental state; Under the aforementioned baseline environmental conditions, a set of verification signals are acquired for the capacitor under test. The verification signal is compared with the baseline signal to generate the verification result.

9. The capacitor self-healing short-circuit detection method based on high-energy pulse cleaning according to claim 8, characterized in that, After generating the verification result, the method further includes: In response to the verification result indicating self-healing failure, the current defect activity feature vector is obtained; The verification results of the self-healing failure and the current defect activity feature vector are stored as learning samples in the sample database; The judgment threshold of the defect judgment model is periodically calibrated using the accumulated data in the sample database, and based on the verification results of the self-healing failure, it is determined that the hidden short circuit point has evolved into a hard short circuit point that cannot be self-healed, and the final state of the capacitor under test is directly updated.

10. A system for performing the capacitor self-healing short-circuit detection method based on high-energy pulse cleaning as described in any one of claims 1-9, characterized in that, include: The environment simulation module is used to generate and apply the aforementioned environment simulation stimuli; A multi-source signal acquisition module is used to synchronously acquire the multi-source response signals; The signal processing and fusion module is used to receive the multi-source response signal and generate the defect activity feature vector. The intelligent determination module is used to receive the defect activity feature vector and generate the defect location information and activity level. The targeted energy generation and control module is used to generate and inject the targeted repair energy based on the defect location information and the activity level. The central control and verification module is used to coordinate and control the above modules and perform the self-healing verification test to determine the final state.

Images