Capacitor pin health status monitoring method, system and storage medium

Abstract

The present invention discloses a method, system, and storage medium for monitoring the health status of capacitor pins, which relates to the field of health management technology. The method includes: applying a multi-frequency excitation signal to the capacitor pins for data acquisition, obtaining pin impedance spectrum data and synchronously capturing a thermal distribution image data set; performing multimodal feature fusion on the pin impedance spectrum data and the thermal distribution image data, extracting a degradation feature vector, and predicting the pin service life based on the vector to generate a remaining life prediction value; dividing the health status level according to the predicted value, triggering an automatic sorting instruction for sorting and identification, and generating a health monitoring result. This application solves the technical problem that traditional monitoring methods rely on a single electrical or thermal parameter, resulting in one-sided health status assessment and low remaining life prediction accuracy. It achieves the technical effect of achieving more accurate capacitor pin health monitoring and life prediction through multimodal data fusion, improving monitoring accuracy, and optimizing equipment maintenance and health management efficiency.

Description

Technical Field

[0001] The present invention relates to the field of health management technology, and in particular to a method, system and storage medium for monitoring the health status of capacitor pins. Background Art

[0002] As an important electronic component, capacitors are widely used in industries such as electricity, communications, and automotive electronics. The health status of capacitor pins, as an important interface part of the capacitor, directly affects the service life and reliability of the capacitor. Since capacitors are exposed to complex environments such as high frequency, high temperature, and strong current for a long time, capacitor pins are prone to aging, damage, and other faults, which in turn affect the normal operation of the equipment. In the existing technology, most monitoring methods rely on a single physical parameter or detection method, such as electrical performance testing and temperature monitoring, but these methods are difficult to comprehensively and accurately evaluate the degradation status of capacitor pins and cannot meet the requirements of reliability and high efficiency monitoring in practical applications. Therefore, there is an urgent need for a health status monitoring method that integrates multimodal data and has high-precision life prediction capabilities to achieve a technological breakthrough in capacitor pin health management. Summary of the Invention

[0003] The present invention provides a capacitor pin health status monitoring method, system, and storage medium to address the technical problems of traditional monitoring methods relying on a single electrical or thermal parameter, resulting in one-sided health status assessment and low remaining life prediction accuracy. This technology achieves more accurate capacitor pin health monitoring and life prediction through multimodal data fusion, improving monitoring accuracy and optimizing equipment maintenance and health management efficiency.

[0004] In a first aspect, the present invention provides a method for monitoring the health status of capacitor pins, wherein the method comprises:

[0005] Apply a multi-frequency excitation signal to the capacitor pin for data acquisition to obtain a pin impedance spectrum data set, perform thermal imaging synchronous capture on the capacitor pin, and generate a thermal distribution image data set; perform multimodal feature fusion on the pin impedance spectrum data set and the thermal distribution image data set to extract the pin degradation feature vector, perform service life prediction on the capacitor pin according to the pin degradation feature vector, and generate a multi-level remaining life prediction value; divide the capacitor pin into multiple health status levels based on the multi-level remaining life prediction value, trigger automatic sorting instructions according to the multiple health status levels for sorting identification, and generate a pin health monitoring result; wherein, generating a multi-level remaining life prediction value includes: extracting the pin impedance spectrum data set and the thermal distribution image data set through frequency domain and spatiotemporal feature extraction , obtaining a first eigenvector and a second eigenvector and performing electro-thermal feature coupling to obtain a pin degradation feature vector; performing a dual-branch feature analysis using the pin degradation feature vector to obtain a dual-branch feature and performing a survival analysis to predict the multi-level remaining life prediction value; wherein, generating the pin degradation feature vector includes: extracting first-dimensional data from the pin impedance spectrum data set based on the first eigenvector, and extracting second-dimensional data from the thermal distribution image data set based on the second eigenvector; synchronizing the first-dimensional data and the second-dimensional data, and using the synchronized data to map the pin impedance spectrum data set to the thermal space coordinate system of the thermal distribution image data set, obtaining a same-dimensional fusion feature matrix and performing dimensionality reduction analysis to generate the pin degradation feature vector.

[0006] In a second aspect, the present invention further provides a capacitor pin health status monitoring system, wherein the capacitor pin health status monitoring system comprises:

[0007] Data acquisition module: applies multi-frequency excitation signals to capacitor pins for data acquisition, obtains pin impedance spectrum data set, performs thermal imaging synchronous capture on capacitor pins, and generates thermal distribution image data set; life prediction module: performs multimodal feature fusion on the pin impedance spectrum data set and the thermal distribution image data set, extracts pin degradation feature vectors, predicts the service life of capacitor pins according to the pin degradation feature vectors, and generates multi-level remaining life prediction values; sorting and identification module: divides multiple health status levels based on the multi-level remaining life prediction values, triggers automatic sorting instructions according to the multiple health status levels for sorting and identification, and generates pin health monitoring results; wherein, the life prediction module is used to: based on the pin impedance spectrum data set and the thermal distribution image data set, Through frequency domain and spatiotemporal feature extraction, a first eigenvector and a second eigenvector are obtained and electro-thermal feature coupling is performed to obtain a pin degradation feature vector; a dual-branch feature analysis is performed using the pin degradation feature vector to obtain a dual-branch feature and a survival analysis is performed to predict the multi-level remaining life prediction value; wherein, the life prediction module is used to: extract first-dimensional data from the pin impedance spectrum data set based on the first eigenvector, and extract second-dimensional data from the thermal distribution image data set based on the second eigenvector; synchronize the first-dimensional data and the second-dimensional data, and use the synchronized data to map the pin impedance spectrum data set to the thermal space coordinate system of the thermal distribution image data set, obtain a same-dimensional fusion feature matrix and perform dimensionality reduction analysis to generate a pin degradation feature vector.

[0008] In a third aspect, the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the capacitor pin health status monitoring method provided by the present invention.

[0009] The present invention discloses a method, system, and storage medium for monitoring the health status of capacitor pins, including: applying a multi-frequency excitation signal to the capacitor pins for data acquisition to obtain a pin impedance spectrum data set, synchronously capturing the capacitor pins by thermal imaging, and generating a thermal distribution image data set; performing multimodal feature fusion on the pin impedance spectrum data set and the thermal distribution image data set to extract a pin degradation feature vector, predicting the service life of the capacitor pins based on the pin degradation feature vector, and generating a multi-level remaining life prediction value; dividing the capacitor pins into multiple health status levels based on the multi-level remaining life prediction value, triggering automatic sorting instructions according to the multiple health status levels for sorting and identification, and generating a pin health monitoring result. The method, system, and storage medium for monitoring the health status of capacitor pins disclosed in the present invention solve the technical problems of traditional monitoring methods relying on a single electrical or thermal parameter, resulting in one-sided health status assessment and low remaining life prediction accuracy, and achieve the technical effect of more accurate capacitor pin health monitoring and life prediction through multimodal data fusion, improving monitoring accuracy, and optimizing equipment maintenance and health management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 Schematic diagram of the flow of the capacitor pin health status monitoring method of the present invention.

[0011] Figure 2 Schematic diagram of the structure of the capacitor pin health status monitoring system of the present invention.

[0012] Description of the accompanying drawings: data acquisition module 11, life prediction module 12, sorting and identification module 13. DETAILED DESCRIPTION

[0013] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the drawings.

[0014] Example 1, as Figure 1 : is a flow chart of a method for monitoring the health status of capacitor pins according to the present invention, wherein the method for monitoring the health status of capacitor pins includes:

[0015] A multi-frequency excitation signal is applied to the capacitor pins for data acquisition to obtain a pin impedance spectrum dataset. Thermal imaging of the capacitor pins is synchronously captured to generate thermal distribution image data.

[0016] Specifically, during the capacitor pin health monitoring process, a multi-frequency excitation signal is first applied to the capacitor pin. This excitation signal covers multiple frequency bands to fully stimulate the capacitor pin's response characteristics. This multi-frequency excitation signal enables acquisition of impedance data from the capacitor pin at different frequencies, resulting in an impedance spectrum dataset. This impedance spectrum dataset provides the basis for subsequent pin health analysis. By analyzing the impedance spectrum data, changes in the pin's electrical characteristics can be assessed, reflecting degradation or failure trends at different frequencies. Simultaneously, an infrared thermal imager is used to expose the capacitor pin under specific excitation signals. By capturing thermal radiation data from the pin area, thermal distribution image data is generated. This thermal distribution image data reflects temperature variations on the pin surface. Abnormal temperature variations are often associated with electrical degradation or localized failures. The simultaneously acquired thermal imaging data allows intuitive observation of thermal changes in the pin, further enabling assessment of its health status. Ultimately, the impedance spectrum dataset and thermal distribution image data are processed and analyzed as multimodal data, providing comprehensive data support for capacitor pin health assessment, lifespan prediction, and fault diagnosis.

[0017] In some embodiments, applying a multi-frequency excitation signal to a capacitor pin to collect data and obtain a pin impedance spectrum dataset includes:

[0018] A gradient frequency band analysis is performed based on the capacitor pins, and a step frequency sweep sequence is set; the step frequency sweep sequence is traversed to perform feature identification according to the frequency band interval to determine multiple characteristic frequency points; a non-contact impedance analyzer is started to apply a sinusoidal excitation voltage signal according to the multiple characteristic frequency points to obtain complex impedance response data; a multi-dimensional impedance analysis is performed on the multiple characteristic frequency points based on the complex impedance response data to construct a three-dimensional impedance matrix; a pin space reference point is defined based on the capacitor pins, and the three-dimensional impedance matrix is ​​mapped according to the multiple characteristic frequency points according to the pin space reference point to obtain a pin impedance spectrum data set.

[0019] Specifically, first, based on the frequency response characteristics of the capacitor pins, a gradient frequency band analysis is performed in combination with actual business needs, and a step sweep sequence is set. In this sequence, the frequency range is divided into three different frequency bands, the first frequency band is 1kHz to 100kHz, the second frequency band is 100kHz to 1MHz, and the third frequency band is 1MHz to 10MHz. Through this segmentation method, various ranges from low frequency to high frequency can be covered, and the response characteristics of the capacitor pins at different frequencies can be obtained. Subsequently, the set step sweep sequence is traversed, and features are identified in each frequency band according to the frequency band interval to determine multiple characteristic frequency points. These characteristic frequency points serve as key reference values, representing the response characteristics of the pins at different frequencies. Selecting these frequency points helps to accurately capture the changes in the electrical characteristics of the capacitor pins, thereby providing key data for subsequent analysis. After determining the characteristic frequency points, the non-contact impedance analyzer is activated and sinusoidal excitation voltage signals are applied at the determined characteristic frequency points. This allows the non-contact impedance analyzer to obtain complex impedance response data at each characteristic frequency point. This complex impedance response data includes the real and imaginary impedances of the pin at these frequencies. Based on this collected complex impedance response data, a multidimensional impedance analysis is then performed for each characteristic frequency point. During this process, the frequency value of each characteristic frequency point is obtained, the square root of the sum of the real and imaginary impedances is taken to obtain the impedance magnitude, and the inverse tangent of the ratio of the imaginary to real impedance is calculated to obtain the phase angle. By using the frequency value, impedance magnitude, and phase angle as each dimension, a three-dimensional impedance matrix is ​​constructed. This three-dimensional impedance matrix details the electrical behavior of the capacitor pin at each frequency. Finally, based on the pin's defined spatial reference point, the resulting three-dimensional impedance matrix is ​​spatially mapped to the characteristic frequency points. This mapping process maps the impedance data to the pin's spatial coordinates, ensuring that the data at each frequency point matches the physical location of the capacitor pin. In this way, the impedance spectrum dataset of the pin was finally obtained, providing comprehensive and accurate electrical data support for subsequent health status assessment and life prediction.

[0020] Table 1: Example of a three-dimensional impedance matrix:

[0021]

[0022] Table 1 shows an example of a three-dimensional impedance matrix, displaying the impedance magnitude and phase angle data for capacitor pins at different frequencies. Each data point corresponds to the pin response characteristics at a specific frequency, including the impedance magnitude (unit: Ω) and phase angle (unit: °). This data is obtained by applying multi-frequency excitation signals to the capacitor pins, collecting and analyzing the complex impedance response. This provides a comprehensive understanding of the capacitor pin's electrical characteristics, supporting subsequent health monitoring and life prediction.

[0023] In some embodiments, thermal imaging is performed on capacitor pins to generate a thermal distribution image dataset, and the method includes:

[0024] The infrared thermal imager is triggered to expose the capacitor pins according to the multiple characteristic frequency points to capture thermal radiation image data of the pin area; temperature rise is identified based on the thermal radiation image data, and the temperature rise area is identified to perform temperature difference calculation to obtain temperature difference gradient data; thermal storage indexing is performed according to the multiple characteristic frequency points based on the temperature difference gradient data to construct a thermal parameter matrix; thermal distribution detection is performed based on the thermal parameter matrix and the thermal radiation image data to generate a thermal distribution image data set.

[0025] Specifically, the system first triggers an infrared thermal imager to expose the capacitor pins at multiple predetermined characteristic frequency points. During this process, since the capacitor pins may experience varying temperature variations under different frequency excitations, the infrared thermal imager captures thermal radiation image data of the pin area. This thermal radiation image data clearly demonstrates the temperature distribution on the pin surface, particularly any hotspots or temperature anomalies. Based on this captured thermal radiation image data, the system then analyzes the temperature variations to identify areas of higher temperature, known as temperature rise zones. These temperature rise zones are often associated with degradation or localized failure within the capacitor pins. The system then calculates the temperature difference within these temperature rise zones to generate temperature gradient data. The temperature gradient is a key indicator that describes temperature variations between different regions, reflecting uneven thermal distribution and potential failure points. Each temperature gradient data point is then associated with a corresponding frequency point. This data is then organized and stored according to frequency characteristics using a thermal storage index, forming a thermal parameter matrix. This matrix contains the temperature difference information at each frequency point, providing the foundation for subsequent thermodynamic analysis. Then, combining the thermal parameter matrix with the thermal radiation image data, a pseudo-color mapping technique is used to detect thermal distribution and construct a thermal distribution map. For example, a blue-red gradient is used to represent the temperature range (e.g., blue corresponds to 25–30°C, red corresponds to 35–40°C). Isotherms (e.g., 30°C, 35°C, 40°C) are superimposed in the temperature rise region, and temperature gradient arrows are labeled (pointing to the high temperature area). Finally, the thermal distribution map for each characteristic frequency point is stored in a collection to form a thermal distribution image dataset. This thermal distribution image dataset intuitively displays the temperature changes on the pin surface, helping to further understand the pin health, identify potential failure risks, and provide a basis for decision-making.

[0026] The pin impedance spectrum dataset and the thermal distribution image dataset are subjected to multimodal feature fusion to extract pin degradation feature vectors, and the service life of the capacitor pins is predicted based on the pin degradation feature vectors to generate a multi-level remaining life prediction value.

[0027] Specifically, during the capacitor pin health status monitoring process, frequency domain feature extraction is first performed on the pin impedance spectrum dataset, and spatiotemporal feature analysis is performed on the thermal distribution image dataset to obtain the first eigenvector and the second eigenvector. These two eigenvectors are then used to couple the two datasets to extract the degradation feature vector of the capacitor pin. This pin degradation feature vector integrates the changes in the electrical performance and thermal characteristics of the pin and can accurately reflect the degradation state of the pin under different operating conditions. Subsequently, a multi-level remaining life prediction value is generated by performing dual-branch feature analysis and survival analysis on the pin degradation feature vector. This multi-level remaining life prediction value can be divided into multiple levels according to different degradation stages to more accurately assess the health status of the pin. Each level represents the remaining usage time of the pin from its current state to complete failure, helping to predict the remaining life of the pin under different operating environments.

[0028] In some embodiments, generating a multi-level remaining life prediction value includes:

[0029] Based on the pin impedance spectrum dataset and the thermal distribution image dataset, a first eigenvector and a second eigenvector are obtained through frequency domain and spatiotemporal feature extraction, and electric-thermal feature coupling is performed to obtain a pin degradation feature vector. A dual-branch feature analysis is performed using the pin degradation feature vector to obtain a dual-branch feature and a survival analysis is performed to predict the multi-level remaining life prediction value.

[0030] Specifically, frequency domain feature extraction is first performed based on the obtained pin impedance spectrum dataset. This process extracts all frequency domain features from the pin impedance spectrum dataset, including impedance amplitude and phase angle. These extracted frequency domain features are then concatenated according to a preset template vector to generate a first eigenvector. This first eigenvector reflects the electrical performance variations of the pin at different frequencies, providing fundamental data for subsequent degradation analysis. Furthermore, spatial features are extracted from the thermal image dataset, including the location of the temperature rise region, temperature gradient, and texture features (calculated using a gray-level co-occurrence matrix, such as contrast, correlation, and uniformity). The temperature rise time and temperature rise rate are also extracted from the thermal image dataset. These spatiotemporal features are then concatenated according to a preset template to generate a second eigenvector. This second eigenvector describes the temperature variation pattern on the pin surface, providing thermal degradation information and further revealing the health status of the pin. Subsequently, the pin impedance spectrum dataset and the thermal image dataset are coupled for electrical-thermal features based on the extracted first and second eigenvectors. During this process, data from different dimensions is synchronized and then subjected to dimensionality reduction to form a comprehensive pin degradation feature vector that comprehensively reflects the overall health of the capacitor pins. Based on this generated pin degradation feature vector, a dual-branch feature analysis is then performed. This dual-branch feature analysis involves synchronously inputting the pin degradation feature vector into a spatial-temporal dual-branch network. The spatial branch of the network processes the spatial characteristics of the pins, while the temporal branch of the network processes the temporal evolution of the degradation characteristics, thereby extracting spatial structural features and temporal features. The obtained spatial structural and temporal features are then combined using a weighted approach to construct a dual-branch feature. Survival analysis is then performed on this dual-branch feature. This survival analysis calculates the survival probability of the capacitor pins, divides them into multiple survival intervals, and plots survival probability curves based on the survival intervals. Finally, based on the survival probability curves, a multi-level remaining life prediction is performed for the capacitor pins, generating multi-level remaining life prediction values ​​that provide data support for subsequent pin health assessment and life warning.

[0031] In some embodiments, generating a pin degradation feature vector includes:

[0032] First dimensional data is extracted from the pin impedance spectrum dataset based on the first eigenvector, and second dimensional data is extracted from the thermal distribution image dataset based on the second eigenvector. The first dimensional data and the second dimensional data are synchronized, and the synchronized data is used to map the pin impedance spectrum dataset to the thermal space coordinate system of the thermal distribution image dataset, obtain a same-dimensional fusion feature matrix, perform dimensionality reduction analysis, and generate a pin degradation feature vector.

[0033] Specifically, first, based on the first eigenvector, the first dimension of the three-dimensional impedance matrix is ​​extracted from the pin impedance spectrum dataset. The three-dimensional impedance matrix typically consists of frequency, impedance amplitude, and phase angle. By comparing the frequency domain features associated with the first eigenvector with the data in the three-dimensional impedance matrix, the corresponding frequency points of the frequency domain features are obtained from the three-dimensional impedance matrix. This data serves as the first dimension, providing the basis for subsequent feature fusion. Subsequently, based on the second eigenvector, the second dimension of the thermal parameter matrix is ​​extracted from the thermal distribution image dataset. The thermal parameter matrix describes the temperature distribution at each point on the pin surface. The extracted second dimension data is the coordinates of each point in the thermal parameter matrix, which can reflect the thermal distribution of the pin. Next, using the characteristic frequency point as a synchronization reference, the first and second dimension data are synchronized. This allows the first and second dimension data belonging to the same characteristic frequency point to be aligned, thereby determining the third dimension data. This third dimension data is generated by synchronizing the frequency and thermal features and represents the combined changes in electrical and thermal performance. Then, based on the third-dimensional data, the multiple characteristic frequency points in the pin impedance spectrum dataset are mapped to the thermal space coordinate system of the thermal distribution image dataset. This means that the frequency points in the impedance spectrum data are mapped to the spatial coordinates in the thermal image, combining the electrical characteristics of each frequency point with the thermal characteristics of the corresponding location. This results in a co-dimensional fused feature matrix, which contains data on frequency, temperature, and their interactions, and can more accurately describe the overall health of the capacitor pins. Finally, a dimensionality reduction analysis is performed on the co-dimensional fused feature matrix. The goal of dimensionality reduction analysis is to reduce the dimensionality of the feature matrix to remove redundant feature information and retain the most important features that effectively represent the pin degradation state. Dimensionality reduction is typically performed using mathematical methods such as principal component analysis (PCA), linear discriminant analysis (LDA), and t-SNE. For example, PCA first calculates the covariance matrix of the co-dimensional fused feature matrix, then uses eigenvalue decomposition to find the eigenvectors. PCA selects the principal components with the largest variance as the main features after dimensionality reduction, ultimately generating a new low-dimensional space. These principal components represent the most informative parts of the data. This process compresses redundant features in the co-dimensional fused feature matrix into a few principal components. By sequentially splicing the extracted principal components into a long feature vector, different dimensionality reduction results are merged together to form a pin degradation feature vector. This pin degradation feature vector will be used for health status assessment and life prediction of capacitor pins, providing an important basis for life prediction and health assessment of capacitor pins.

[0034] In some embodiments, performing a dual-branch feature analysis on the capacitor pin based on the pin degradation feature vector to obtain a dual-branch feature includes:

[0035] The pin degradation feature vector is synchronized to a space-time dual-branch network, which includes a space branch and a time branch: S1: convolution processing is performed on the pin degradation feature vector based on the space branch to extract the pin crack extension feature map; S2: spatial structure reorganization is performed based on the pin crack extension feature map to determine the spatial structure feature; S3: frequency domain evolution is performed on the pin degradation feature vector based on the time branch to extract the degradation rate feature; S4: degradation timing analysis is performed based on the degradation rate feature to determine the degradation timing feature; feature interaction is performed between the spatial structure feature and the degradation timing feature to generate a feature interaction tensor; an attention mechanism is introduced to perform weighted analysis in combination with the feature interaction tensor to construct a covariate matrix, and the covariate matrix is ​​added to the dual-branch feature.

[0036] Specifically, the pin degradation feature vector is first synchronized to the spatial-temporal dual-branch network. The spatial-temporal dual-branch network consists of a spatial branch and a temporal branch. The spatial branch is built based on a convolutional neural network and is trained in advance through steps such as forward propagation, loss calculation, backpropagation, and parameter optimization. It is mainly responsible for processing spatial features. The temporal branch is built based on a long short-term memory network and also extracts and undergoes the same training as mentioned above. It is mainly responsible for processing time-related features. After the spatial-temporal dual-branch network receives the pin degradation feature vector, it will copy the pin degradation feature vector and pass it to the spatial branch and the temporal branch respectively. In the spatial branch, based on the pin degradation feature vector, the spatial branch will first use a convolution layer to perform convolution processing, convolving the spatial data in the pin degradation feature vector with the convolution kernel. After the convolution is completed, the pin crack growth feature map is extracted from the spatial distribution of the pin. This pin crack growth feature map can reflect the crack growth process caused by degradation on the pin surface and provide degradation information at the spatial level. After extracting the crack growth feature map, it is passed to subsequent layers (such as pooling layers and fully connected layers) for spatial structural reorganization. Spatial structural reorganization analyzes the distribution and evolution patterns of pin surface cracks based on the crack growth feature map, thereby converting the crack growth feature map into more structured information and extracting the spatial structural characteristics of the pin cracks. This spatial structural feature can reveal the growth trend, distribution density, and possible failure points of the pin surface cracks. In the time series branch, the LSTM layer is used to perform frequency domain evolution on the pin degradation feature vector. That is, through the LSTM memory mechanism, the long-term dependencies in the input data are captured, and the degradation rate characteristics of each time step are derived. These degradation rate characteristics can reveal the frequency components of the pin degradation process and help analyze the changing trends of the pin health status over time. After extracting the degradation rate features, they are passed to the subsequent LSTM layer for degradation timing analysis, identifying key timing features of the degradation process, such as the acceleration, stabilization, and recovery phases. These degradation timing features provide temporal information about the pin degradation process, helping to predict changes in the pin's health over time. The extracted spatial structure features are then subjected to feature interaction with the degradation timing features. This interaction combines the spatial structure features and the degradation timing features to create a joint feature vector. The outer product of the joint feature vector is then calculated to generate an interaction tensor. This interaction tensor integrates information from both spatial and temporal dimensions, comprehensively describing the dynamic process and spatial distribution characteristics of pin degradation, and capturing the coupling relationship between spatial structure and temporal dynamics.Next, the attention mechanism is introduced. This attention mechanism is primarily used to assign different weights to different components in the feature interaction tensor. Common attention mechanisms include self-attention and additive attention. Taking self-attention as an example, the interaction tensor is input into the attention mechanism, and an importance score is calculated for each feature in the interaction tensor. This process is typically implemented using a query vector (Query), a key vector (Key), and a value vector (Value). The query vector typically comes from the current input features, while the key and value vectors are generated from the features in the interaction tensor. The attention score for each feature is calculated by calculating the similarity between the query vector and the key vector (such as the dot product or other similarity metrics). The resulting score is then normalized (typically using the softmax function) to obtain the attention weight for each feature. Finally, the input interaction tensor is weighted by the attention weights to obtain weighted features. The calculated weighted features are the covariance matrix, which contains the relationships and weights between all features. This covariance matrix can reflect the interactions between different features and is used in subsequent prediction tasks. After obtaining the covariate matrix, it will be added to the dual-branch features to enrich the input features of subsequent life prediction, thereby improving the accuracy of pin health assessment.

[0037] In some embodiments, performing survival analysis based on the dual-branch signature to determine the multi-level remaining life expectancy prediction value comprises:

[0038] Based on the covariate matrix, a risk analysis of the capacitor pins is performed to calculate a baseline risk factor; the conditional survival probability is calculated according to the baseline risk factor, and a plurality of survival intervals are divided according to the conditional survival probability; the survival time is marked based on the plurality of survival intervals, and a survival probability curve is drawn; the capacitor pins are reverse matched according to the survival probability curve, and data synchronization is performed according to the matching result to generate the multi-level remaining life prediction value.

[0039] Specifically, a risk analysis of capacitor pins is first performed based on the obtained covariate matrix. The covariate matrix contains information on multiple dimensions of pin degradation characteristics, such as electrical performance and thermal characteristics. By taking a weighted sum of the eigenvalues ​​in the covariate matrix, a baseline risk factor is calculated. The baseline risk factor reflects the overall health risk level of the capacitor pins in their current state. A higher baseline risk factor indicates that the pins are likely to be in a more severe state of degradation and have a higher risk of failure. Subsequently, based on the calculated baseline risk factor, the conditional survival probability is further calculated. The conditional survival probability is the probability that the capacitor pins will remain functional over a future period of time, given their current health state. Based on the Cox proportional hazards model or other appropriate survival analysis model, the system uses the baseline risk factor to calculate the survival probability of the capacitor pins over a certain future time period. The conditional survival probabilities at different time points t are then obtained to reflect the pins' survival capabilities over different time periods. After obtaining the conditional survival probability, multiple survival probability thresholds are set based on the survival probability distribution. These survival probability thresholds are then used to create multiple survival intervals. For example, a high survival probability interval (survival probability greater than 90%) indicates that the pin is healthy and has a high survival probability. A medium survival probability interval (survival probability between 50% and 90%) indicates that the pin is moderately healthy, with some degradation possible but not reaching the level of failure. A low survival probability interval (survival probability less than 50%) indicates that the pin is in poor health, has a high failure risk, and has a low survival probability. Each survival interval is then assigned a survival time range. Each survival interval represents the health of the pin within a certain timeframe. Therefore, time labels are associated with the survival probability of each interval. High survival probability intervals are labeled with longer survival times, such as 5 years or longer; medium survival probability intervals are labeled with shorter survival times, such as 2 to 5 years; and low survival probability intervals are labeled with shorter survival times, such as 1 year or less. After the survival time is marked, a curve showing the change in survival probability over time is drawn based on the survival probability data for each survival interval. Specifically, the horizontal axis is set to time, and the vertical axis is set to survival probability. Points are then plotted on the time axis based on the survival intervals and corresponding survival probabilities. These points are then fitted using the least squares method to create a survival probability curve. This survival probability curve shows the trend of the pin's survival probability over time throughout its entire lifecycle, helping to accurately predict the pin's remaining lifespan. Reverse matching is then performed based on the plotted survival probability curve. Reverse matching involves calculating a multi-level remaining lifespan prediction based on the corresponding relationship between the current health status of the capacitor pin and the survival probability curve. This process generates an accurate lifespan prediction for the pin by matching the current pin's risk level with the conditional survival probability.Finally, based on the matching results, data synchronization is performed to ensure that all health assessment information is updated and generate accurate multi-level remaining life prediction values. These prediction values ​​can further provide support for equipment management and maintenance scheduling, help extend the service life of equipment, and reduce the risk of failure.

[0040] Based on the multi-level remaining life prediction values, multiple health status levels are divided, and automatic sorting instructions are triggered according to the multiple health status levels to perform sorting identification and generate pin health monitoring results.

[0041] Specifically, based on the generated multi-level RLL prediction values, the health status of the capacitor pins is first classified. Each pin is assigned a different health status level based on its RLL value, typically including healthy, warning, and failed. This classification provides a more accurate health assessment for each pin, facilitating subsequent maintenance and management. Once the health status level is determined, corresponding automatic sorting instructions are triggered based on the health status level. These instructions are used to sort and identify the capacitor pins during production or repair. Specifically, when a pin's health status is healthy, it is marked as qualified and sent to the next stage of normal use or storage. A pin with a warning health status triggers a warning flag, indicating the need for further inspection, repair, or monitoring. A pin with a failed health status triggers a failed flag, marking it unusable and entering the scrapping process. This automatic sorting and identification process enables real-time intelligent management based on the health status of capacitor pins, enabling more efficient maintenance, optimizing resource allocation, and preventing unqualified pins from entering the production line, ensuring capacitor quality and reliability and reducing potential failure risks.

[0042] In some embodiments, the method includes dividing the health status into multiple levels based on the multi-level remaining life prediction values, triggering automatic sorting instructions to perform sorting and marking according to the multiple health status levels, and generating pin health monitoring results. The method includes:

[0043] A mutation analysis is performed based on the multi-level remaining life prediction value to extract the life sudden drop prediction value; an abnormal growth calculation is performed on the capacitor pin based on the life sudden drop prediction value to obtain the abnormal growth rate; an abnormal growth rate interval of the capacitor pin is set, the abnormal growth rate interval has a minimum critical value and a maximum critical value, and the abnormal growth rate interval is a threshold interval from the minimum critical value to the maximum critical value; when the abnormal growth rate is less than the minimum critical value of the abnormal growth rate interval, a first health status level is generated, an automatic sorting instruction is triggered to perform health mark sorting, and a pin health qualified monitoring result is generated; when the abnormal growth rate is within the abnormal growth rate interval, a second health status level is generated, an automatic sorting instruction is triggered to perform warning mark sorting, and a pin health repair monitoring result is generated; when the abnormal growth rate is greater than the maximum critical value of the abnormal growth rate interval, a third health status level is generated, an automatic sorting instruction is triggered to perform failure mark sorting, and a pin health scrap monitoring result is generated.

[0044] Specifically, based on the obtained multi-level RLL prediction values, a sliding window difference method is used to calculate the difference between adjacent RLL prediction values. If a difference exceeds a preset threshold, a sudden change in the current RLL prediction value is flagged and used as the lifespan drop prediction value. This lifespan drop prediction value reflects the critical point at which the pin may experience abnormal degradation, indicating a sudden deterioration in the health of the pin. Subsequently, based on the extracted lifespan drop prediction value, the new lifespan drop prediction value is subtracted from the previous multi-level RLL prediction value and divided by the time interval to obtain the abnormal growth rate. This abnormal growth rate can help identify abnormal changes in the pin at a specific stage and promptly detect potential faults. Next, an abnormal growth rate range is set for the capacitor pin. The upper and lower limits of this range are the minimum and maximum thresholds, respectively. This abnormal growth rate range is set based on historical data and reflects the range of normal and abnormal degradation during the pin's health degradation process. The minimum threshold corresponds to the minimum normal growth rate during the pin degradation process, while the maximum threshold corresponds to the maximum acceptable normal degradation rate. After completing the calculation of the abnormal growth rate and setting the threshold, the corresponding health status level will be generated according to the different situations of the abnormal growth rate, and the automatic sorting instruction will be triggered. Specifically, when the abnormal growth rate of the pin is less than the minimum critical value of the abnormal growth rate interval, it means that the degradation of the pin is within the normal range and the health is good. Therefore, the first health status level will be generated, and the automatic sorting instruction will be triggered to perform health identification sorting. At this time, the capacitor pin is marked as healthy and qualified, and the pin health qualified monitoring result is generated, entering the normal use or storage stage. When the abnormal growth rate of the pin is within the abnormal growth rate interval, it means that the degradation rate of the pin is near the upper limit of the normal range, and there is a certain risk of degradation. Therefore, the second health status level is generated, and the automatic sorting instruction is triggered to perform early warning identification sorting, and generate the pin health repair monitoring result, which indicates that the pin can still be used, but requires further inspection or regular maintenance to prevent further degradation. When the abnormal growth rate of a pin exceeds the maximum critical value of the abnormal growth rate range, it indicates that the pin's degradation rate has exceeded the normal range and may pose a serious failure risk. At this time, the system generates a third health status level, triggering automatic sorting instructions to sort out failure marks and generate pin health scrap monitoring results. The pin is marked as unqualified and enters the scrapping process to prevent it from further use in the device. Through this process, the health status of capacitor pins can be monitored in real time, and automatic sorting and early warning processing are carried out based on changes in abnormal growth rates, ensuring the reliability and safety of the equipment, reducing potential failure risks, and extending the equipment's service life.

[0045] In summary, the capacitor pin health status monitoring method provided by the present invention has the following technical effects:

[0046] A multi-frequency excitation signal is applied to the capacitor pins for data acquisition to obtain a pin impedance spectrum dataset, and thermal imaging of the capacitor pins is synchronously captured to generate a thermal distribution image dataset. Multimodal feature fusion is performed on the pin impedance spectrum dataset and the thermal distribution image dataset to extract pin degradation feature vectors. The service life of the capacitor pins is predicted based on the pin degradation feature vectors to generate multi-level remaining life prediction values. Multiple health status levels are divided based on the multi-level remaining life prediction values, and automatic sorting instructions are triggered according to the multiple health status levels for sorting and identification, generating pin health monitoring results, thereby achieving the technical effect of more accurate capacitor pin health monitoring and life prediction through multimodal data fusion, improving monitoring accuracy, and optimizing equipment maintenance and health management efficiency.

[0047] Example 2, as Figure 2 This is a schematic diagram of the structure of the capacitor pin health status monitoring system of the present invention. For example, Figure 1 The flow chart of the capacitor pin health status monitoring method of the present invention can be shown as follows: Figure 2 The structure shown is implemented.

[0048] Based on the same concept as the capacitor pin health status monitoring method in the above embodiment, the present invention also provides a capacitor pin health status monitoring system comprising:

[0049] Data acquisition module 11: applies a multi-frequency excitation signal to the capacitor pins for data acquisition, obtains a pin impedance spectrum dataset, and synchronously captures the capacitor pins with thermal imaging to generate a thermal distribution image dataset; life prediction module 12: performs multimodal feature fusion on the pin impedance spectrum dataset and the thermal distribution image dataset, extracts the pin degradation feature vector, predicts the service life of the capacitor pins based on the pin degradation feature vector, and generates a multi-level remaining life prediction value; sorting and identification module 13: divides the capacitor pins into multiple health status levels based on the multi-level remaining life prediction values, triggers automatic sorting instructions according to the multiple health status levels for sorting and identification, and generates pin health monitoring results. Among them, the life prediction module 12 is used to: based on the pin impedance spectrum dataset and the thermal distribution image dataset, obtain the first eigenvector and the second eigenvector through frequency domain and spatiotemporal feature extraction and perform electro-thermal feature coupling to obtain the pin degradation feature vector; use the pin degradation feature vector to perform dual-branch feature analysis to obtain dual-branch features and perform survival analysis to predict the multi-level remaining life prediction value; wherein, the life prediction module 12 is used to: extract first dimensional data from the pin impedance spectrum dataset based on the first eigenvector, and extract second dimensional data from the thermal distribution image dataset based on the second eigenvector; synchronize the first dimensional data and the second dimensional data, and use the synchronized data to map the pin impedance spectrum dataset to the thermal space coordinate system of the thermal distribution image dataset, obtain the same-dimensional fusion feature matrix and perform dimensionality reduction analysis to generate the pin degradation feature vector.

[0050] In some embodiments, the data acquisition module 11 includes:

[0051] A gradient frequency band analysis is performed based on the capacitor pins, and a step frequency sweep sequence is set; the step frequency sweep sequence is traversed to perform feature identification according to the frequency band interval to determine multiple characteristic frequency points; a non-contact impedance analyzer is started to apply a sinusoidal excitation voltage signal according to the multiple characteristic frequency points to obtain complex impedance response data; a multi-dimensional impedance analysis is performed on the multiple characteristic frequency points based on the complex impedance response data to construct a three-dimensional impedance matrix; a pin space reference point is defined based on the capacitor pins, and the three-dimensional impedance matrix is ​​mapped according to the multiple characteristic frequency points according to the pin space reference point to obtain a pin impedance spectrum data set.

[0052] In some embodiments, the data acquisition module 11 includes:

[0053] The infrared thermal imager is triggered to expose the capacitor pins according to the multiple characteristic frequency points to capture thermal radiation image data of the pin area; temperature rise is identified based on the thermal radiation image data, and the temperature rise area is identified to perform temperature difference calculation to obtain temperature difference gradient data; thermal storage indexing is performed according to the multiple characteristic frequency points based on the temperature difference gradient data to construct a thermal parameter matrix; thermal distribution detection is performed based on the thermal parameter matrix and the thermal radiation image data to generate a thermal distribution image data set.

[0054] In some embodiments, the lifespan prediction module 12 includes:

[0055] The pin degradation feature vector is synchronized to a space-time dual-branch network, which includes a space branch and a time branch: S1: convolution processing is performed on the pin degradation feature vector based on the space branch to extract the pin crack extension feature map; S2: spatial structure reorganization is performed based on the pin crack extension feature map to determine the spatial structure feature; S3: frequency domain evolution is performed on the pin degradation feature vector based on the time branch to extract the degradation rate feature; S4: degradation timing analysis is performed based on the degradation rate feature to determine the degradation timing feature; feature interaction is performed between the spatial structure feature and the degradation timing feature to generate a feature interaction tensor; an attention mechanism is introduced to perform weighted analysis in combination with the feature interaction tensor to construct a covariate matrix, and the covariate matrix is ​​added to the dual-branch feature.

[0056] In some embodiments, the lifespan prediction module 12 includes:

[0057] Based on the covariate matrix, a risk analysis of the capacitor pins is performed to calculate a baseline risk factor; the conditional survival probability is calculated according to the baseline risk factor, and a plurality of survival intervals are divided according to the conditional survival probability; the survival time is marked based on the plurality of survival intervals, and a survival probability curve is drawn; the capacitor pins are reverse matched according to the survival probability curve, and data synchronization is performed according to the matching result to generate the multi-level remaining life prediction value.

[0058] In some embodiments, the sorting identification module 13 includes:

[0059] A mutation analysis is performed based on the multi-level remaining life prediction value to extract the life sudden drop prediction value; an abnormal growth calculation is performed on the capacitor pin based on the life sudden drop prediction value to obtain the abnormal growth rate; an abnormal growth rate interval of the capacitor pin is set, the abnormal growth rate interval has a minimum critical value and a maximum critical value, and the abnormal growth rate interval is a threshold interval from the minimum critical value to the maximum critical value; when the abnormal growth rate is less than the minimum critical value of the abnormal growth rate interval, a first health status level is generated, an automatic sorting instruction is triggered to perform health mark sorting, and a pin health qualified monitoring result is generated; when the abnormal growth rate is within the abnormal growth rate interval, a second health status level is generated, an automatic sorting instruction is triggered to perform warning mark sorting, and a pin health repair monitoring result is generated; when the abnormal growth rate is greater than the maximum critical value of the abnormal growth rate interval, a third health status level is generated, an automatic sorting instruction is triggered to perform failure mark sorting, and a pin health scrap monitoring result is generated.

[0060] In a third embodiment, the present invention further provides a computer-readable storage medium that can be used to store software programs, computer executable programs, and modules, such as program instructions / modules corresponding to the capacitor pin health status monitoring method in the embodiment of the present invention, thereby implementing the above-mentioned capacitor pin health status monitoring method.

[0061] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention and are all included in the scope of protection of the present invention.

Claims

1. A method for monitoring the health status of capacitor pins, characterized in that: The method comprises: Apply multi-frequency excitation signals to the capacitor pins for data acquisition to obtain a pin impedance spectrum dataset, and synchronously capture thermal imaging of the capacitor pins to generate a thermal distribution image dataset; Performing multimodal feature fusion on the pin impedance spectrum dataset and the thermal distribution image dataset to extract pin degradation feature vectors, predicting the service life of capacitor pins based on the pin degradation feature vectors, and generating multi-level remaining service life prediction values; Dividing the pins into multiple health status levels based on the multi-level remaining life prediction values, triggering automatic sorting instructions to perform sorting marking according to the multiple health status levels, and generating pin health monitoring results; Among them, generating multi-level remaining life prediction values ​​includes: Based on the pin impedance spectrum dataset and the thermal distribution image dataset, a first eigenvector and a second eigenvector are obtained by extracting frequency domain and spatiotemporal features, and then electrical-thermal feature coupling is performed to obtain a pin degradation feature vector; Performing a dual-branch feature analysis using the pin degradation feature vector to obtain a dual-branch feature and performing a survival analysis to predict the multi-level remaining life prediction value; The generation of pin degradation feature vectors includes: extracting first dimensional data from the pin impedance spectrum dataset based on the first eigenvector, and extracting second dimensional data from the thermal distribution image dataset based on the second eigenvector; The first dimensional data and the second dimensional data are synchronized, and the synchronized data is used to map the pin impedance spectrum dataset to the thermal space coordinate system of the thermal distribution image dataset, obtain a same-dimensional fusion feature matrix, perform dimensionality reduction analysis, and generate a pin degradation feature vector.

2. The capacitor pin health status monitoring method according to claim 1, wherein: Applying a multi-frequency excitation signal to the capacitor pins to collect data and obtain a pin impedance spectrum data set, the method includes: Perform gradient frequency band analysis based on the capacitor pin and set the step sweep sequence; Traversing the step frequency sweep sequence to perform feature identification according to frequency band intervals to determine multiple feature frequency points; Starting the non-contact impedance analyzer to apply a sinusoidal excitation voltage signal according to the plurality of characteristic frequency points to obtain complex impedance response data; Performing multidimensional impedance analysis on the multiple characteristic frequency points based on the complex impedance response data to construct a three-dimensional impedance matrix; A pin space reference point is defined based on the capacitor pin, and the three-dimensional impedance matrix is ​​mapped according to the plurality of characteristic frequency points according to the pin space reference point to obtain a pin impedance spectrum data set.

3. The capacitor pin health status monitoring method according to claim 2, wherein: Capturing capacitor pins with thermal imaging to generate a thermal distribution image dataset, the method includes: triggering an infrared thermal imager to expose the capacitor pins according to the multiple characteristic frequency points, and capturing thermal radiation image data of the pin area; Performing temperature rise identification based on the thermal radiation image data, marking the temperature rise area, performing temperature difference calculation, and obtaining temperature difference gradient data; Performing thermal storage indexing according to the temperature gradient data and the plurality of characteristic frequency points to construct a thermal parameter matrix; Thermal distribution detection is performed based on the thermal parameter matrix and the thermal radiation image data to generate a thermal distribution image data set.

4. The capacitor pin health status monitoring method according to claim 1, wherein: Performing a dual-branch feature analysis on the capacitor pin based on the pin degradation feature vector to obtain a dual-branch feature, the method comprising: Synchronize the pin degradation feature vector to a space-time dual-branch network, where the space-time dual-branch network includes a space branch and a time branch: S1: performing convolution processing on the pin degradation feature vector based on the spatial branch to extract a pin crack growth feature map; S2: performing spatial structural reorganization based on the pin crack propagation characteristic map to determine spatial structural characteristics; S3: performing frequency domain evolution on the pin degradation feature vector based on the timing branch to extract degradation rate features; S4: performing degradation timing analysis based on the degradation rate characteristics to determine degradation timing characteristics; Performing feature interaction between the spatial structure feature and the degraded temporal feature to generate a feature interaction tensor; An attention mechanism is introduced to perform weighted analysis in combination with the feature interaction tensor to construct a covariate matrix, which is then added to the dual-branch feature.

5. The capacitor pin health status monitoring method according to claim 4, wherein: Performing survival analysis based on the dual-branch feature to determine the multi-level remaining life prediction value includes: Performing capacitor pin risk analysis based on the covariate matrix and calculating a baseline risk factor; Calculating the conditional survival probability according to the baseline risk factor, and dividing the survival intervals into multiple intervals according to the conditional survival probability; Marking the survival time based on the multiple survival intervals and drawing a survival probability curve; Reverse matching is performed on capacitor pins according to the survival probability curve, and data synchronization is performed according to the matching result to generate the multi-level remaining life prediction value.

6. The capacitor pin health status monitoring method according to claim 1, wherein: Dividing a plurality of health status levels based on the multi-level remaining life prediction values, triggering automatic sorting instructions to perform sorting marking according to the plurality of health status levels, and generating a pin health monitoring result, the method includes: Performing mutation analysis based on the multi-level remaining life prediction values ​​to extract a sudden drop in life prediction value; Performing abnormal growth calculation on the capacitor pins according to the life sudden reduction prediction value to obtain an abnormal growth rate; Setting an abnormal growth rate interval of the capacitor pin, wherein the abnormal growth rate interval has a minimum critical value and a maximum critical value, and the abnormal growth rate interval is a threshold value interval from the minimum critical value to the maximum critical value; When the abnormal growth rate is less than the minimum critical value of the abnormal growth rate interval, a first health status level is generated, an automatic sorting instruction is triggered to perform health identification sorting, and a pin health qualified monitoring result is generated; When the abnormal growth rate is within the abnormal growth rate range, a second health status level is generated, an automatic sorting instruction is triggered to perform early warning mark sorting, and a pin health repair monitoring result is generated; When the abnormal growth rate is greater than the maximum critical value of the abnormal growth rate interval, a third health status level is generated, an automatic sorting instruction is triggered to perform failure mark sorting, and a pin health scrapping monitoring result is generated.

7. Capacitor pin health status monitoring system, characterized in that, For implementing the capacitor pin health status monitoring method according to any one of claims 1 to 6, the system comprises: Data acquisition module: applies multi-frequency excitation signals to the capacitor pins to collect data, obtains the pin impedance spectrum dataset, and synchronously captures thermal imaging of the capacitor pins to generate a thermal distribution image dataset; Life prediction module: performs multimodal feature fusion on the pin impedance spectrum dataset and the thermal distribution image dataset to extract the pin degradation feature vector, predicts the service life of the capacitor pin based on the pin degradation feature vector, and generates a multi-level remaining life prediction value; Sorting and identification module: divides the pins into multiple health status levels based on the multi-level remaining life prediction values, triggers automatic sorting instructions for sorting and identification according to the multiple health status levels, and generates pin health monitoring results; Among them, the life prediction module is used to: Based on the pin impedance spectrum dataset and the thermal distribution image dataset, a first eigenvector and a second eigenvector are obtained through frequency domain and spatiotemporal feature extraction, and electrical-thermal feature coupling is performed to obtain a pin degradation feature vector; a dual-branch feature analysis is performed using the pin degradation feature vector to obtain a dual-branch feature and a survival analysis is performed to predict the multi-level remaining life prediction value; Among them, the life prediction module is used to: First dimensional data is extracted from the pin impedance spectrum dataset based on the first eigenvector, and second dimensional data is extracted from the thermal distribution image dataset based on the second eigenvector. The first dimensional data and the second dimensional data are synchronized, and the synchronized data is used to map the pin impedance spectrum dataset to the thermal space coordinate system of the thermal distribution image dataset, obtain a same-dimensional fusion feature matrix, perform dimensionality reduction analysis, and generate a pin degradation feature vector.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the capacitor pin health status monitoring method according to any one of claims 1 to 6 is implemented.

Images