Failure detection system for chip tantalum electrolytic capacitor

Abstract

The invention relates to the technical field of electronic component detection, in particular to a failure detection system for a chip-type tantalum electrolytic capacitor, which comprises the steps of analyzing operation parameters of the chip-type tantalum electrolytic capacitor in a historical time period, judging whether the chip-type tantalum electrolytic capacitor is failed or not, identifying the chip-type tantalum electrolytic capacitor which is not failed temporarily, and determining multi-dimensional time sequence data; analyzing the multi-dimensional time series data, outputting a comprehensive coefficient, and determining retention parameters in the operation parameters; classifying all the reserved parameters according to the operation state, and outputting a cluster; according to the retention parameters in the cluster, a failure detection model of the chip tantalum electrolytic capacitor is constructed, and the residual failure time under different operation states is predicted in combination with an optimization target. The residual failure time of the chip tantalum electrolytic capacitor under different operation states is accurately predicted, and the residual failure time of the chip tantalum electrolytic capacitor under different operation states is accurately predicted according to the retention parameters monitored in real time. And outputting the real-time monitored remaining failure time when the retention parameter reaches the operating parameter threshold value.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic component detection, and particularly relates to a failure detection system for chip tantalum electrolytic capacitors. Background Art

[0002] Chip tantalum electrolytic capacitors have been widely used in electronic devices due to their advantages such as small size, large capacitance, and stable performance. However, during long-term operation, affected by various factors such as voltage, current, and temperature, chip tantalum electrolytic capacitors are prone to failure problems, and once they fail, it may lead to a decline in the performance of the entire electronic device or even a malfunction.

[0003] Existing chip tantalum electrolytic capacitors have many deficiencies in failure detection. On the one hand, only a single operating parameter is concerned, such as only monitoring voltage or equivalent series resistance, which cannot comprehensively reflect the actual operating state of the capacitor and is prone to misjudgment or missed judgment. On the other hand, the analysis of the change of operating parameters over time is lacking, and parameters that have little effect on failure prediction cannot be effectively removed, resulting in low accuracy and effectiveness of the prediction model.

[0004] Therefore, we propose a failure detection system for chip tantalum electrolytic capacitors. Summary of the Invention

[0005] The purpose of the present invention is to provide a failure detection system for chip tantalum electrolytic capacitors to solve at least one of the above-mentioned problems in the prior art.

[0006] The present invention provides a failure detection system for chip tantalum electrolytic capacitors, including:

[0007] A data acquisition and analysis module: analyzing the operating parameters of chip tantalum electrolytic capacitors in a historical period, determining whether the chip tantalum electrolytic capacitors fail, identifying chip tantalum electrolytic capacitors that have not yet failed, and determining multi-dimensional time series data based on the operating parameters of the chip tantalum electrolytic capacitors that have not yet failed;

[0008] A retained parameter determination module: outputting a comprehensive coefficient by performing correlation and discreteness analysis on the multi-dimensional time series data, and determining the retained parameters in the operating parameters;

[0009] A retained parameter classification module: classifying all retained parameters according to the operating states of chip tantalum electrolytic capacitors that have not yet failed, and outputting clustering clusters;

[0010] A failure detection module: constructing a failure detection model for chip tantalum electrolytic capacitors based on the retained parameters in the clustering clusters, and predicting the remaining failure time of chip tantalum electrolytic capacitors in different operating states in combination with the optimization objective.

[0011] As a further technical solution of the present invention: The specific process of identifying the non-failed chip tantalum electrolytic capacitor is as follows:

[0012] Analyze the operating parameters of the chip tantalum electrolytic capacitor within a historical period, and output a parameter abnormality characterization value; if the parameter abnormality characterization value is less than the parameter abnormality characterization threshold, the chip tantalum electrolytic capacitor has not failed yet.

[0013] As a further technical solution of the present invention: The process of obtaining the parameter abnormality characterization value is as follows:

[0014] Set a historical period, collect the operating parameters of the chip tantalum electrolytic capacitor within the historical period; if the operating parameter is greater than or equal to the operating parameter threshold, generate a parameter abnormality signal; extract the duration of the parameter abnormality signal and perform a ratio process with the duration corresponding to the historical period to obtain the parameter abnormality characterization value.

[0015] As a further technical solution of the present invention: The process of obtaining the multi-dimensional time series data is as follows:

[0016] Divide the historical period into several sub-periods, extract the operating parameters of the sub-periods, and combine the operating parameters of all sub-periods into multi-dimensional time series data.

[0017] As a further technical solution of the present invention: The process of obtaining the comprehensive coefficient is as follows:

[0018] Perform correlation and discreteness analysis on the multi-dimensional time series data, output the correlation coefficient and the discreteness coefficient, and perform a summation process on the correlation coefficient and the discreteness coefficient to obtain the comprehensive coefficient.

[0019] As a further technical solution of the present invention: The process of obtaining the correlation coefficient is as follows:

[0020] Construct a time series of the operating parameter, calculate the mean value of the time series of the operating parameter, extract the characteristic value corresponding to the operating parameter in the time series, construct a parameter series of the operating parameter, and calculate the mean value of the parameter series of the operating parameter;

[0021] Calculate the correlation coefficient between the operating parameter and time in combination with the Pearson correlation coefficient calculation formula.

[0022] As a further technical solution of the present invention: The process of obtaining the discreteness coefficient is as follows:

[0023] According to the mean value of the parameter series of the operating parameter, calculate the standard deviation of the parameter series corresponding to the operating parameter, and further calculate the discreteness coefficient of the operating parameter.

[0024] As a further technical solution of the present invention: The specific process of classifying all retained parameters according to the operating state of the non-failed chip tantalum electrolytic capacitor is as follows:

[0025] Numerically encode the retained parameters and perform normalization processing. Analyze the retained parameters after normalization processing, output the truncation distance, start with each data point as a separate clustering cluster, and gradually merge the clustering clusters according to the principle of the closest distance. During the merging process, when the distance between clustering clusters is greater than the truncation distance, stop the merging. The different clustering clusters finally obtained correspond to different operating states of the chip tantalum electrolytic capacitor.

[0026] As a further technical solution of the present invention: the way to obtain the truncation distance is:

[0027] According to the retained parameters after normalization processing, calculate the Euclidean distance between sample point parameters;

[0028] Sort the Euclidean distances between all sample point parameters from largest to smallest to obtain a distance sorting table, and extract the minimum value of the top 2% distances in the distance sorting table, which is denoted as the truncation distance.

[0029] As a further technical solution of the present invention: the specific process of predicting the remaining failure time under different operating states is:

[0030] Extract the retained parameters in a clustering cluster. Based on an artificial neural network, divide the retained parameters into a training set and a validation set, and use the training set to train the model;

[0031] Through iterative optimization until the loss no longer decreases, obtain a failure detection model for the chip tantalum electrolytic capacitor after training is completed;

[0032] Use the trained model, take the operating parameter threshold as the input, output the failure time point when the operating parameter threshold is reached, and perform a difference operation between the failure time point and the current time point to obtain the remaining failure time.

[0033] The beneficial effects of the present invention:

[0034] 1. The present invention analyzes the operating parameters of the chip tantalum electrolytic capacitor in the historical period, judges whether the chip tantalum electrolytic capacitor fails, identifies the chip tantalum electrolytic capacitor that has not yet failed, and determines multi-dimensional time series data; among them, the operating parameters of the chip tantalum electrolytic capacitor include voltage, current, equivalent series resistance, and ambient temperature; analyze the multi-dimensional time series data, output a comprehensive coefficient, and determine the retained parameters in the operating parameters; the purpose of the present invention to analyze the chip tantalum electrolytic capacitor that has not yet failed and eliminate the features with a comprehensive coefficient lower than the comprehensive coefficient threshold is to better reflect the degradation characteristics of the chip tantalum electrolytic capacitor. Screening a large number of operating parameters in the multi-dimensional time series data of the chip tantalum electrolytic capacitor according to the comprehensive coefficient helps to remove the operating parameters with poor effects on the failure prediction of the chip tantalum electrolytic capacitor, providing reliable support for improving the accuracy and effectiveness of the subsequent prediction model.

[0035] 2. The present invention classifies all reserved parameters according to the operating status, where the operating status types include: normal operation, mild degradation, and severe degradation; extracts the reserved parameters in the clustering clusters, constructs a failure detection model for the chip tantalum electrolytic capacitor based on the reserved parameters in the clustering clusters, and predicts the remaining failure time under different operating statuses in combination with the optimization objective; the present invention divides the reserved parameters in the multi-dimensional time series data according to the operating status of the chip tantalum electrolytic capacitor, analyzes the influence of the reserved parameters on the capacitor failure under different operating statuses, outputs the influence coefficients under different operating statuses, and accurately predicts the remaining failure time of the chip tantalum electrolytic capacitor under different operating statuses according to the influence coefficients under different operating statuses. By constructing a failure detection model based on the influence coefficients, multiple reserved parameters can be comprehensively considered to form a relatively comprehensive failure detection system, and then the remaining failure time when the reserved parameters monitored in real time reach the operating parameter threshold can be output according to the reserved parameters monitored in real time, realizing early warning of the failure of the chip tantalum electrolytic capacitor. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0037] Figure 1 It is a system block diagram of a failure detection system for a chip tantalum electrolytic capacitor according to an embodiment of the present invention;

[0038] Figure 2 It is a flowchart of a failure detection method for a chip tantalum electrolytic capacitor according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] Embodiment 1

[0041] As Figure 1 shown, a failure detection system for a chip tantalum electrolytic capacitor provided by an embodiment of the present invention specifically includes:

[0042] Data acquisition and analysis module: analyze the operating parameters of chip tantalum electrolytic capacitors in the historical period, determine whether the chip tantalum electrolytic capacitors have failed, identify the chip tantalum electrolytic capacitors that have not failed, and determine multi-dimensional time series data based on the operating parameters of the chip tantalum electrolytic capacitors that have not failed;

[0043] Among them, the operating parameters of chip tantalum electrolytic capacitors include voltage, current, equivalent series resistance, and ambient temperature;

[0044] A historical period is set, the duration of the historical period is T, the value of T is preset by a person skilled in the art, the end time of the historical period is the current time, recorded as TE, so the start time of the historical period is (TE-T), and the operating parameters of the chip tantalum electrolytic capacitor are collected during the historical period;

[0045] It can be understood that the voltage and current of the chip tantalum electrolytic capacitor are collected using an oscilloscope, the equivalent series resistance of the chip tantalum electrolytic capacitor is collected using an ESR meter, and the ambient temperature of the chip tantalum electrolytic capacitor is collected using a temperature sensor;

[0046] Compare the operating parameters with the operating parameter thresholds. The specific process is as follows:

[0047] If the operating parameter is greater than or equal to the operating parameter threshold, a parameter abnormality signal is generated;

[0048] If the operating parameter is less than the operating parameter threshold, a parameter normal signal is generated;

[0049] The duration of the parameter abnormality signal is extracted and compared with the duration corresponding to the historical period to obtain the parameter abnormality representation value;

[0050] In some embodiments, the parameter anomaly characterization value is compared to a parameter anomaly characterization threshold:

[0051] If the parameter abnormality characterization value is greater than or equal to the parameter abnormality characterization threshold, it means that the chip tantalum electrolytic capacitor has failed and maintenance personnel need to be arranged to replace the chip tantalum electrolytic capacitor;

[0052] If the parameter abnormality characterization value is less than the parameter abnormality characterization threshold, it means that the chip tantalum electrolytic capacitor has not failed yet;

[0053] It is understandable that the parameter anomaly characterization value can reflect the overall operating condition of the chip tantalum electrolytic capacitor within a historical period. Since it synthesizes the time proportion of parameter anomalies, it can more comprehensively evaluate whether the capacitor has truly shown signs of performance degradation or failure. For some capacitors with relatively low parameter anomaly characterization values, even if the operating parameters occasionally exceed the threshold, there is no need to replace them immediately, thus avoiding unnecessary resource waste, reducing the maintenance cost of the equipment, and improving the utilization efficiency of resources. For example, although some capacitors occasionally exhibit parameter anomalies, it is found through calculating the parameter anomaly characterization value that they are still within the normal range. In this case, it is safe and feasible to continue using the capacitor without replacement. Through the screening of the parameter anomaly characterization value, maintenance resources can be concentrated on those capacitors that truly need to be replaced, improving the efficiency of maintenance work;

[0054] Based on the fact that the chip tantalum electrolytic capacitor has not failed yet, divide the historical period into several sub-periods at equal time intervals, extract the operating parameters of the sub-periods, and combine the operating parameters of all sub-periods into multi-dimensional time series data;

[0055] Retention parameter determination module: By performing correlation and discreteness analysis on the multi-dimensional time series data, output a comprehensive coefficient to determine the retention parameters among the operating parameters;

[0056] S21. Construct the time series T of the i-th operating parameter i =[t i1 , t i2 ,..., t in , extract the eigenvalue corresponding to the i-th operating parameter in the time series, and construct the parameter sequence φ i =[φ i1 , φ i2 ,..., φ in ;

[0057] According to the time series T i =[t i1 , t i2 ,..., t in of the i-th operating parameter, calculate the mean value of the time series of the i-th operating parameter where t ij represents the sampling time of the j-th time point in the i-th operating parameter, j = 1, 2,..., n, and n represents the total number of operating parameters in the parameter sequence. The specific calculation formula is:

[0058]

[0059] According to the parameter sequence φ i =[φ i1 , φ i2,..., φ in , the mean value of the parameter sequence of the i-th operating parameter is calculated where φ ij represents the eigenvalue at the j-th time point in the i-th operating parameter, j = 1, 2,..., n, and n represents the total number of operating parameters in the parameter sequence. The specific calculation formula is:

[0060]

[0061] It can be understood that the total number of operating parameters is the same as the total number of time points in the time series and they are in one-to-one correspondence;

[0062] Furthermore, the correlation coefficient r between the i-th operating parameter and time is calculated i , and the specific calculation formula is:

[0063]

[0064] It can be understood that the correlation coefficient characterizes the linear correlation between the operating parameter and time. The range of the correlation coefficient is [0, 1]. The larger the absolute value of the correlation coefficient, the stronger the correlation; the closer the absolute value of the correlation coefficient is to 0, the weaker the correlation. If the calculated correlation coefficient is large, it means that the correlation between the i-th operating parameter and time is strong, that is, the i-th operating parameter has a relatively obvious change law over time. On the contrary, if the calculated correlation coefficient is small, it means that the change of the i-th operating parameter is not closely related to time;

[0065] S22. According to the mean value of the parameter sequence of the i-th operating parameter the standard deviation σ of the parameter sequence corresponding to the i-th operating parameter is calculated i , and the specific calculation formula is:

[0066]

[0067] Furthermore, the coefficient of dispersion D of the i-th operating parameter is calculated i , and the specific calculation formula is:

[0068]

[0069] where the correction coefficient α is set to prevent the denominator in the coefficient of dispersion calculation formula from being 0;

[0070] It can be understood that D i reflects the degree of dispersion of the characteristics in the i-th operating parameter. The larger the calculated coefficient of dispersion, the smaller the coefficient of dispersion of the i-th operating parameter, the greater the degree of dispersion of the operating parameter, and the more dispersed the data distribution. On the contrary, the larger the coefficient of dispersion, the more concentrated the data distribution;

[0071] S23. Sum the correlation coefficient and the discreteness coefficient to obtain a comprehensive coefficient.

[0072] In some embodiments, compare the comprehensive coefficient with a comprehensive coefficient threshold. The specific process is as follows:

[0073] If the comprehensive coefficient is greater than the comprehensive coefficient threshold, record the corresponding operating parameter as a retained parameter.

[0074] If the comprehensive coefficient is less than the comprehensive coefficient threshold, record the corresponding operating parameter as a non-retained parameter.

[0075] It can be understood that the purpose of excluding the retained parameters with a comprehensive coefficient lower than the comprehensive coefficient threshold is to better reflect the degradation characteristics of the chip tantalum electrolytic capacitor. Screening a large number of operating parameters in the multi-dimensional time series data of the chip tantalum electrolytic capacitor through the comprehensive coefficient helps to remove the operating parameters with poor effects on the failure prediction of the chip tantalum electrolytic capacitor, providing reliable support for improving the accuracy and effectiveness of the subsequent prediction model.

[0076] Exemplarily, assume that the historical time period T = 10 hours, which is divided into 10 sub-time periods (data is collected once per hour), and the data of the operating parameters is as follows:

[0077] Sub-period Voltage (V) Current (A) ESR (Ω) Temperature (°C) 1 5.0 0.2 0.1 25 2 5.1 0.3 0.12 25 3 5.2 0.2 0.15 26 4 5.3 0.4 0.18 25 5 5.4 0.3 0.2 26 6 5.5 0.5 0.25 25 7 5.6 0.4 0.3 26 8 5.7 0.6 0.35 25 9 5.8 0.5 0.4 26 10 5.9 0.7 0.45 25

[0078] Taking voltage as an example, the time series T = [1, 2,..., 10], and the mean of the time series

[0079] The parameter sequence φ of voltage 电压 = [5.0, 5.1, 5.1, 5.2, 5.3, 5.4, 5.5, 5.6, 5.7, 5.8, 5.9], and the mean of the voltage parameter sequence The correlation coefficient r 电压 ≈ 1;

[0080] The discreteness coefficient D 电压 ≈ 0.0527;

[0081] For voltage

[0082] Assume that the comprehensive coefficient threshold of voltage is 0.8. Since the comprehensive coefficient of voltage is greater than the comprehensive coefficient threshold of voltage, voltage is a retained parameter.

[0083] The technical solution of this embodiment is: analyzing the operating parameters of chip tantalum electrolytic capacitors within a historical period, determining whether the chip tantalum electrolytic capacitors have failed, identifying chip tantalum electrolytic capacitors that have not failed yet, and determining multidimensional time series data based on the operating parameters of chip tantalum electrolytic capacitors that have not failed yet; outputting a comprehensive coefficient by performing correlation and discreteness analysis on the multidimensional time series data, and determining the retained parameters in the operating parameters; the present invention analyzes chip tantalum electrolytic capacitors that have not failed yet, and eliminates the characteristics whose comprehensive coefficients are lower than the comprehensive coefficient threshold in order to better reflect the degradation characteristics of chip tantalum electrolytic capacitors, and screens the numerous operating parameters in the multidimensional time series data of chip tantalum electrolytic capacitors according to the comprehensive coefficients, which helps to remove the operating parameters that have a poor effect on the failure prediction of chip tantalum electrolytic capacitors, and provides reliable support for the subsequent improvement of the accuracy and effectiveness of the prediction model.

[0084] Embodiment 2

[0085] like Figure 1 As shown, an embodiment of the present invention provides a failure detection system for a chip tantalum electrolytic capacitor, specifically comprising:

[0086] Retention parameter classification module: classifies all retention parameters according to the operating status of the chip tantalum electrolytic capacitors that have not failed, and outputs clusters, where the operating status types include: normal operation, slight degradation, and severe degradation;

[0087] S31, numerically encoding the reserved parameters and performing normalization processing;

[0088] By normalizing the formula: Calculate the normalized operating parameter X nor , where X is the original eigenvalue, X min and X max are the minimum and maximum values of the feature respectively;

[0089] S32, calculate the pth sample point parameter x according to the retained parameters after normalization processing p =(x p1 , x p2 , …, x pm ) and the qth sample point parameter x q =(x q1 , x q2 , …, x qm ) between pq , the specific formula is:

[0090]

[0091] Among them, m represents the number of types of retained parameters in the sample point parameters, x pk and xpk respectively represent the k-th eigenvalue of the p-th sample point and the q-th sample point;

[0092] Exemplarily, assume that the parameter of the p-th sample point after normalization is x p =(x p1 , x p2 , x p3 )(corresponding to voltage, current, and ambient temperature respectively), and the parameter of the q-th sample point is x q =(x q1 , x q2 , x q3 ), then the Euclidean distance is:

[0093] Sort the Euclidean distances d pq between all sample point parameters from largest to smallest to obtain a distance sorting table, and extract the minimum value of the top 2% distances in the distance sorting table, denoted as the truncation distance d c ;

[0094] S34. Start with each data point as a separate clustering cluster, and gradually merge the clustering clusters according to the principle of the closest distance. During the merging process, when the distance between clustering clusters is greater than the truncation distance, stop merging. The finally obtained different clustering clusters correspond to different operating states of the chip tantalum electrolytic capacitor;

[0095] Failure detection module: Construct a failure detection model for the chip tantalum electrolytic capacitor based on the retained parameters in the clustering cluster, and predict the remaining failure time of the chip tantalum electrolytic capacitor in different operating states in combination with the optimization objective;

[0096] Extract the retained parameters in a clustering cluster. Based on the artificial neural network, divide the retained parameters into a training set and a validation set according to the ratio of 8:2, and use the training set to train the model;

[0097] Adopt minimizing the loss function as the optimization objective:

[0098]

[0099] where Cr represents the r-th operating state, Nr represents the number of samples in this operating state, represents the cross-entropy loss in the r-th operating state; M is the number of types of operating states, and λ is the regularization coefficient;

[0100] It can be understood that the first term in the minimizing loss function is the weighted sum of the cross-entropy losses in each operating state, and the second term is the regularization term of the loss difference between different operating states. The purpose is to make the performance of the model more balanced in different operating states, and the optimization objective is to make L reg minimized;

[0101] Through iterative optimization until the loss no longer decreases, a failure detection model for the chip tantalum electrolytic capacitor after training is obtained;

[0102] Using the trained model, taking the operating parameter threshold as the input, outputting the failure time point when the operating parameter threshold is reached, and taking the difference between the failure time point and the current time point to obtain the remaining useful life;

[0103] It can be understood that, compared with the prior art, by dividing the retained parameters in the multi-dimensional time series data according to the operating state, accurately predicting the remaining useful life of the chip tantalum electrolytic capacitor under different operating states, constructing a failure detection model, multiple retained parameters can be comprehensively considered to form a relatively comprehensive failure detection system. Furthermore, according to the real-time monitored retained parameter values, the remaining useful life when the real-time monitored retained parameters reach the operating parameter threshold can be output, realizing early warning of the failure of the chip tantalum electrolytic capacitor;

[0104] The technical solution of this embodiment is: classifying all retained parameters according to the operating state of the chip tantalum electrolytic capacitor that has not failed yet, and outputting clustering clusters. Among them, the types of operating states include: normal operation, mild degradation, and severe degradation; constructing a failure detection model for the chip tantalum electrolytic capacitor according to the retained parameters in the clustering clusters, and predicting the remaining useful life of the chip tantalum electrolytic capacitor under different operating states in combination with the optimization objective; by dividing the retained parameters in the multi-dimensional time series data according to the operating state of the chip tantalum electrolytic capacitor, accurately predicting the remaining useful life of the chip tantalum electrolytic capacitor under different operating states, constructing a failure detection model, multiple retained parameters can be comprehensively considered to form a relatively comprehensive failure detection system. Furthermore, according to the real-time monitored retained parameters, the remaining useful life when the real-time monitored retained parameters reach the operating parameter threshold can be output, realizing early warning of the failure of the chip tantalum electrolytic capacitor.

[0105] Embodiment III

[0106] As Figure 1 shown, a failure detection method for a chip tantalum electrolytic capacitor provided by an embodiment of the present invention specifically includes:

[0107] Step 1: Analyze the operating parameters of the chip tantalum electrolytic capacitor in the historical period, judge whether the chip tantalum electrolytic capacitor fails, identify the chip tantalum electrolytic capacitor that has not failed yet, and determine multi-dimensional time series data based on the operating parameters of the chip tantalum electrolytic capacitor that has not failed yet;

[0108] Among them, the operating parameters of the chip tantalum electrolytic capacitor include voltage, current, equivalent series resistance, and ambient temperature;

[0109] Set a historical period. The duration of the historical period is T, and the value of T is pre-set by those skilled in the art. The end time of the historical period is the current moment, denoted as TE. Therefore, the start time of the historical period is (TE - T). During the historical period, collect the operating parameters of the chip tantalum electrolytic capacitor;

[0110] Compare the operating parameters with the operating parameter thresholds:

[0111] If the operating parameter is greater than or equal to the operating parameter threshold, generate a parameter anomaly signal;

[0112] If the operating parameter is less than the operating parameter threshold, generate a parameter normal signal;

[0113] Extract the duration of the parameter anomaly signal and perform a ratio process with the duration corresponding to the historical period to obtain a parameter anomaly characterization value;

[0114] Compare the parameter anomaly characterization value with the parameter anomaly characterization threshold:

[0115] If the parameter anomaly characterization value is greater than or equal to the parameter anomaly characterization threshold, it indicates that the chip tantalum electrolytic capacitor has failed, and maintenance personnel need to be arranged to replace the chip tantalum electrolytic capacitor;

[0116] If the parameter anomaly characterization value is less than the parameter anomaly characterization threshold, it indicates that the chip tantalum electrolytic capacitor has not failed yet;

[0117] Based on the fact that the chip tantalum electrolytic capacitor has not failed yet, divide the historical period into several sub-periods at equal time intervals, extract the operating parameters of the sub-periods, and combine the operating parameters of all sub-periods into multi-dimensional time series data;

[0118] Step 2: Output a comprehensive coefficient by performing correlation and discreteness analysis on the multi-dimensional time series data, and determine the retained parameters in the operating parameters;

[0119] Construct the time series T of the i-th operating parameter i = [t i1 , t i2 ,..., t in , extract the characteristic values corresponding to the i-th operating parameter in the time series, and construct the parameter sequence φ of the i-th operating parameter i = [φ i1 , φ i2 ,..., φ in ;

[0120] According to the time series T of the i-th operating parameter i = [t i1 , t i2 ,..., t in, the time series mean of the i-th operating parameter is calculated

[0121] According to the parameter sequence φ of the i-th operating parameter i =[φ i1 , φ i2 ,..., φ in , the parameter sequence mean of the i-th operating parameter is calculated

[0122] Furthermore, the correlation coefficient r between the i-th operating parameter and time is calculated i ; According to the parameter sequence mean of the i-th operating parameter the parameter sequence standard deviation σ corresponding to the i-th operating parameter is calculated i ;

[0123] Furthermore, the discreteness coefficient D of the i-th operating parameter is calculated i , and the specific calculation formula is:

[0124]

[0125] Among them, the correction coefficient α is set to prevent the denominator in the discreteness coefficient calculation formula from being zero;

[0126] The correlation coefficient and the discreteness coefficient are summed to obtain a comprehensive coefficient;

[0127] The comprehensive coefficient is compared with the comprehensive coefficient threshold:

[0128] If the comprehensive coefficient is greater than the comprehensive coefficient threshold, the corresponding operating parameter is recorded as a reserved parameter;

[0129] If the comprehensive coefficient is less than the comprehensive coefficient threshold, the corresponding operating parameter is recorded as a non-reserved parameter;

[0130] Step 3: Classify all the reserved parameters according to the operating status of the non-failed chip tantalum electrolytic capacitors, and output the clustering clusters;

[0131] Numerically encode the reserved parameters and perform normalization processing; According to the normalized reserved parameters, calculate the Euclidean distance d between the parameter x of the p-th sample point p =(x p1 , x p2 , …, x pm ) and the parameter x of the q-th sample point q =(x q1 , x q2 , …, x qm ); pq ;

[0132] The Euclidean distance d between all sample point parameters pq is sorted from largest to smallest to obtain a distance sorting table, and the minimum value of the first 2% of the distances in the distance sorting table is extracted and denoted as the truncation distance d c ;

[0133] Start with each data point as a separate clustering cluster, and gradually merge the clustering clusters according to the principle of the closest distance. During the merging process, when the distance between clustering clusters is greater than the truncation distance, stop merging. The different clustering clusters finally obtained correspond to different operating states of the chip tantalum electrolytic capacitor;

[0134] Step 4: Construct a failure detection model for the chip tantalum electrolytic capacitor based on the retained parameters in the clustering cluster, and predict the remaining failure time of the chip tantalum electrolytic capacitor under different operating states in combination with the optimization objective;

[0135] Extract the retained parameters in a clustering cluster. Based on the artificial neural network, divide the retained parameters into a training set and a validation set in a ratio of 8:2, and use the training set to train the model;

[0136] Adopt minimizing the loss function as the optimization objective:

[0137]

[0138] where Cr represents the r-th operating state, Nr represents the number of samples in this operating state, represents the cross-entropy loss in the r-th operating state; M is the number of types of operating states, and λ is the regularization coefficient;

[0139] Through iterative optimization until the loss no longer decreases, obtain the failure detection model of the trained chip tantalum electrolytic capacitor;

[0140] Use the trained model, take the operating parameter threshold as the input, output the failure time point when the operating parameter threshold is reached, and subtract the current time point from the failure time point to obtain the remaining failure time.

[0141] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0142] The above has described a specific embodiment of the present invention in detail, but the content described is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. Any equivalent changes and improvements made within the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A failure detection system for chip tantalum electrolytic capacitors, characterized in that, include: Data acquisition and analysis module: analyze the operating parameters of chip tantalum electrolytic capacitors in the historical period, determine whether the chip tantalum electrolytic capacitors have failed, identify the chip tantalum electrolytic capacitors that have not failed, and determine multi-dimensional time series data based on the operating parameters of the chip tantalum electrolytic capacitors that have not failed; Retention parameter determination module: By performing correlation and discreteness analysis on multi-dimensional time series data, outputting comprehensive coefficients, and determining the retention parameters in the operating parameters; Retained parameter classification module: classifies all retained parameters according to the operating status of the chip tantalum electrolytic capacitors that have not failed, and outputs clusters; Failure detection module: Based on the retained parameters in the clustering clusters, a failure detection model for chip tantalum electrolytic capacitors is constructed, and the remaining failure time of chip tantalum electrolytic capacitors under different operating conditions is predicted in combination with the optimization objectives.

2. The failure detection system for a chip tantalum electrolytic capacitor according to claim 1, characterized in that, The specific process of identifying chip tantalum electrolytic capacitors that have not yet failed is as follows: Analyze the operating parameters of chip tantalum electrolytic capacitors in the historical period and output abnormal parameter characterization values; If the parameter abnormality characterization value is less than the parameter abnormality characterization threshold, the chip tantalum electrolytic capacitor has not failed yet.

3. The failure detection system for a chip tantalum electrolytic capacitor according to claim 2, characterized in that, The process of obtaining the parameter abnormality characterization value is as follows: Set a historical period, and collect the operating parameters of the chip tantalum electrolytic capacitor within the historical period; if the operating parameter is greater than or equal to the operating parameter threshold, generate a parameter abnormality signal; The duration of the parameter anomaly signal is extracted and ratioed with the duration corresponding to the historical period to obtain the parameter anomaly characterization value.

4. A failure detection system for a chip tantalum electrolytic capacitor according to claim 3, characterized in that, The process of acquiring multidimensional time series data is as follows: The historical period is divided into several sub-periods, the operating parameters of the sub-periods are extracted, and the operating parameters of all sub-periods are combined into multi-dimensional time series data.

5. The failure detection system for a chip tantalum electrolytic capacitor according to claim 4, wherein, The process of obtaining the comprehensive coefficient is: Perform correlation and discreteness analysis on multidimensional time series data, output the correlation coefficient and discreteness coefficient, sum the correlation coefficient and discreteness coefficient to obtain the comprehensive coefficient.

6. The failure detection system for a chip tantalum electrolytic capacitor according to claim 5, characterized in that, The process of obtaining the correlation coefficient is: Construct a time series of operating parameters, calculate the mean of the time series of operating parameters, extract the eigenvalues corresponding to the operating parameters in the time series, construct a parameter series of operating parameters, and calculate the mean of the parameter series of operating parameters; The correlation coefficient between the operating parameters and time is calculated by combining the Pearson correlation coefficient calculation formula.

7. The failure detection system for a chip tantalum electrolytic capacitor according to claim 6, wherein The process of obtaining the discrete coefficient is: According to the parameter sequence mean of the operating parameters, the parameter sequence standard deviation corresponding to the operating parameters is calculated, and then the discrete coefficient of the operating parameters is calculated.

8. A failure detection system for a chip tantalum electrolytic capacitor according to claim 5, characterized in that, The specific process of classifying all retained parameters according to the operating status of the chip tantalum electrolytic capacitors that have not failed is as follows: The retained parameters are numerically encoded and normalized, the retained parameters after normalization are analyzed, the cutoff distance is output, each data point is taken as a separate cluster, and the clusters are gradually merged according to the principle of the closest distance. During the merging process, when the distance between clusters is greater than the cutoff distance, the merging is stopped, and the different clusters finally obtained correspond to different operating states of the chip tantalum electrolytic capacitor.

9. A failure detection system for a chip tantalum electrolytic capacitor according to claim 8, characterized in that, The cutoff distance is obtained as follows: According to the retained parameters after normalization, the Euclidean distance between the sample point parameters is calculated; Sort the Euclidean distances between all sample point parameters from largest to smallest to obtain a distance sorting table, and extract the minimum value of the top 2% distances in the distance sorting table, which is denoted as the truncation distance.

10. The failure detection system for a chip tantalum electrolytic capacitor according to claim 9, characterized in that, The specific process of predicting the remaining failure time under different operating states is as follows: Extract the retained parameters in a clustering cluster. Based on an artificial neural network, divide the retained parameters into a training set and a validation set, and use the training set to train the model; Through iterative optimization until the loss no longer decreases, obtain a failure detection model for the trained chip tantalum electrolytic capacitor; Use the trained model, take the operating parameter threshold as the input, output the failure time point when the operating parameter threshold is reached, and perform a difference operation between the failure time point and the current time point to obtain the remaining failure time.

Images