Fault detection method and system for diode

Abstract

The invention relates to the field of diode fault analysis, in particular to a fault detection method and system for a diode, and the detection method comprises the steps: obtaining application parameters of each detection and analysis target in a detection and analysis main body, so as to determine an analysis difference state, and the analysis difference state is determined based on application difference parameters and application characteristic parameters; determining whether to perform feature difference analysis or feature application analysis on the detection analysis subject based on the analysis difference state; when feature difference analysis is carried out, whether feature integration processing or feature isolation processing is carried out on the corresponding detection and analysis target is determined according to the evaluation association index of each detection and analysis target, so that a training execution data set of each detection and analysis target is obtained; during feature applicable analysis, whether applicable optimization is carried out or not is determined according to the feature evaluation crossover index of the detection and analysis subject, so that optimized extraction parameters are determined. According to the invention, the effectiveness of the fault detection model on diode fault detection in the actual operation process is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of diode fault analysis, in particular to a fault detection method and system for diodes. BACKGROUND

[0002] With the continuous development of deep learning technology, its use for fault detection of diodes can effectively reduce the manpower consumption and timeliness of the fault detection process. However, in the actual fault detection process, there are differences between the fault characteristics of diodes of different categories and different application scenarios. Building a fault prediction model for diodes in different situations increases the burden on data analysis in practical applications. Therefore, how to ensure effective identification of unique abnormal characteristics of diodes in different application scenarios while improving the universality of the constructed fault prediction model is a problem that needs to be solved by those skilled in the art.

[0003] Chinese Patent Application Publication No. CN120387075A discloses a fault prediction method, device, computer equipment and medium. The method includes: first, the I-V characteristics of the bypass diode are corrected by introducing a series of equivalent resistors, and a non-invasive parameter estimation method is used to determine the parameters in the photovoltaic module model; then, the data simulated by the photovoltaic module model is filtered by an information gain algorithm to obtain training data, and a fault prediction model is trained based on the training data; then, the target photovoltaic module is predicted by combining the autoencoder and the fault prediction model; subsequently, the parameters in the photovoltaic module model are fault-inverted to optimize the parameters in the photovoltaic module model; finally, a life prediction model is established based on the electrical data simulated by the optimized photovoltaic module model to accurately predict the life of the photovoltaic module. However, the above-mentioned scheme has the following defects: it fails to consider the unique fault characteristics of different diodes in different application scenarios during actual fault detection or prediction, resulting in poor adaptability and effectiveness of the actual acquired fault prediction model to the actual application scenario. SUMMARY

[0004] Therefore, the present application provides a fault detection method and system for diodes to overcome the problem that the prior art fails to consider the unique fault characteristics of different diodes in different application scenarios during actual fault detection or prediction, resulting in poor adaptability and effectiveness of the actual acquired fault prediction model to the actual application scenario.

[0005] To achieve the above-mentioned purpose, the present application provides a fault detection method for diodes, comprising: Obtaining application parameters of each detection analysis target in the detection analysis subject to determine an analysis difference state of the detection analysis subject, the analysis difference state being determined based on an application difference parameter and an application characteristic parameter; Determining whether to perform characteristic difference analysis or characteristic application analysis on the detection analysis subject based on the analysis difference state of the detection analysis subject; When performing characteristic difference analysis, determining whether to perform characteristic integration processing or characteristic isolation processing on the corresponding detection analysis target according to an evaluation correlation index of each detection analysis target to obtain a training execution data set of each detection analysis target, the evaluation correlation index being determined according to a fault characteristic evaluation parameter of each detection analysis target; When performing characteristic application analysis, determining whether to perform application optimization according to a characteristic evaluation cross index of the detection analysis subject to determine an optimization extraction parameter.

[0006] Further, the analysis difference state includes a first analysis difference state and a second analysis difference state; The detection analysis subject in the first analysis difference state is a detection analysis subject with an application difference parameter greater than a preset application difference parameter or an application characteristic parameter greater than a preset application characteristic parameter; The detection analysis subject in the second analysis difference state is a detection analysis subject with an application difference parameter less than or equal to a preset application difference parameter and an application characteristic parameter less than or equal to a preset application characteristic parameter.

[0007] Further, when the detection analysis subject is in the first analysis difference state, performing characteristic difference analysis on the detection analysis subject; The evaluation correlation index of any detection analysis target is determined according to a fault characteristic evaluation parameter corresponding to each fault characteristic.

[0008] Further, performing characteristic integration processing on the detection analysis target with an evaluation correlation index greater than a preset evaluation correlation index, and determining whether to perform data fusion extraction according to a time sequence correlation parameter of each characteristic integration combination; Performing data fusion extraction on the characteristic integration combination with a time sequence correlation parameter greater than a preset time sequence correlation parameter to obtain training execution data of the characteristic integration combination, and determining a data extraction scale according to the time sequence correlation parameter; The data extraction scale and the time sequence correlation parameter of the corresponding characteristic integration combination are in a positive correlation relationship.

[0009] Further, the setting mode of the characteristic integration combination is determined based on a fault concurrency parameter of the detection analysis target; Determining a characteristic integration combination of the detection analysis target with a fault concurrency parameter greater than a preset fault concurrency parameter based on a concurrency time sequence correlation index; Determine the feature set combination of the detection analysis target whose fault concurrent parameter is less than or equal to the preset fault concurrent parameter based on the evaluation coincidence index.

[0010] Further, perform feature isolation processing on the detection analysis target whose evaluation correlation index is less than or equal to the preset evaluation correlation index, perform data fusion extraction on each fault feature, and determine the data extraction scale of each fault feature according to the evaluation correlation index; The data extraction scale and the evaluation correlation index of the corresponding fault feature are in a positive correlation relationship.

[0011] Further, when the detection analysis subject is in a two-class analysis difference state or completes feature difference analysis, perform feature applicable analysis on the detection analysis subject. The feature evaluation cross index is determined based on the fault feature evaluation parameters of each detection analysis target in the detection analysis subject.

[0012] Further, when the feature evaluation cross index of the detection analysis subject is greater than the preset feature cross evaluation index, perform applicable optimization on the detection analysis subject. Perform multi-scale data extraction on each optimization extraction parameter. The optimization extraction parameter is the fault feature evaluation parameter whose evaluation cross index is greater than the preset evaluation cross index.

[0013] Further, the application difference parameter and the application feature parameter are determined based on the evaluation value range of each fault feature evaluation parameter of the detection analysis target existing in the detection analysis subject.

[0014] The application also provides a fault detection system for a diode, comprising: A subject evaluation module is used to obtain the application parameters of each detection analysis target in the detection analysis subject, so as to determine the analysis difference state in which the detection analysis subject is located, and the analysis difference state is determined based on the application difference parameter and the application feature parameter; An analysis execution module is connected with the subject evaluation module, and is used to determine whether to perform feature difference analysis or feature applicable analysis on the detection analysis subject based on the analysis difference state in which the detection analysis subject is located; A difference analysis module is connected with the analysis execution module, and is used to determine whether to perform feature integration processing or feature isolation processing on the corresponding detection analysis target according to the evaluation correlation index of each detection analysis target, so as to obtain the training execution data set of each detection analysis target, and the evaluation correlation index is determined according to the fault feature evaluation parameter of each detection analysis target; An applicable analysis module is connected with the analysis execution module, and is used to determine whether to perform applicable optimization according to the feature evaluation cross index of the detection analysis subject, so as to determine the optimization extraction parameter.

[0015] Compared with the prior art, the application has the beneficial effect that the application determines the analysis difference state of the detection analysis subject according to the wide range and difference of the fault feature evaluation parameter range of the diode in the specific application process of the detection analysis subject, and determines the targeted data acquisition mode for the detection analysis subject, so that the training data used for fault detection model training is more in line with the needs of the actual situation, thereby improving the effectiveness and timeliness of the fault detection model for the diode in the actual work process.

[0016] Further, in the application, feature difference analysis is performed on detection analysis subjects in a first analysis difference state, and since the detection analysis targets of such detection analysis subjects have relatively wide ranges of fault feature evaluation parameters or relatively large differences between fault feature evaluation parameters in the actual application process, it is necessary to ensure effective extraction of the unique difference features of each detection analysis target in the training process of the fault detection model for each detection analysis target, so as to ensure effective identification of the unique abnormal features of the diode in different application scenarios.

[0017] Further, in the application, when performing feature difference analysis, it is determined whether to perform feature integration processing or feature isolation processing on the corresponding detection analysis target according to the evaluation correlation index of each detection analysis target, the evaluation correlation index represents the cross degree between fault feature evaluation parameters of different fault features in the fault detection process of the detection analysis target, and the specific data extraction process is determined according to the evaluation correlation index, so that the specific extraction process of the training data is more in line with the actual fault detection needs of the detection analysis target, and the application improves the effectiveness of the fault detection model for the diode in the actual work process.

[0018] Further, in the application, feature applicability analysis is performed on detection analysis subjects in a second analysis difference state or after feature difference analysis, and since the detection analysis targets of such detection analysis subjects have relatively small ranges of fault feature evaluation parameters and relatively small differences between fault feature evaluation parameters in the actual application process, or have completed the optimization of the targeted data extraction of each fault feature, it is necessary to further determine the application range of the fault feature evaluation parameters, and to determine whether to ensure the comprehensiveness of the obtained data features through multi-scale extraction, the application improves the universality of the constructed fault prediction model for the actual work process, thereby improving the timeliness of fault detection. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a schematic diagram of the diode fault detection method of the application; Figure 2 This is a flowchart illustrating how the present invention determines the analytical difference state based on application difference parameters and application characteristic parameters. Figure 3 This is a flowchart illustrating the present invention for determining whether to perform feature difference analysis or feature applicability analysis on the detection and analysis subject based on the analysis difference state of the detection and analysis subject. Figure 4 This is a module connection diagram of the fault detection system for diodes according to the present invention. Detailed Implementation

[0020] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0022] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0023] Please see Figures 1 to 3 As shown, the present invention provides a fault detection method for diodes, comprising: The application parameters of each detection and analysis target within the detection and analysis subject are obtained to determine the analysis difference state of the detection and analysis subject. The analysis difference state is determined based on the application difference parameters and application characteristic parameters. Determine whether to perform feature difference analysis or feature applicability analysis on the subject of detection and analysis based on the analysis difference status of the subject of detection and analysis. When performing feature difference analysis, the evaluation correlation index of each detection and analysis target determines whether feature integration processing or feature isolation processing should be performed for the corresponding detection and analysis target, so as to determine the training execution data set of each detection and analysis target. The evaluation correlation index is determined based on the fault feature evaluation parameters of each detection and analysis target. When performing feature applicability analysis, the cross-index of the feature evaluation of the subject of detection and analysis is used to determine whether to perform applicability optimization, so as to determine the optimization extraction parameters. The optimization extraction parameters are determined based on the evaluation cross-index.

[0024] This invention optimizes the selection process of training data in the training phase of a diode fault detection model. It improves the effectiveness of the training results by optimizing the acquisition process of training data during the actual training process. This is achieved through effective extraction of both general and specific features, combined with targeted extraction based on actual application scenarios. During training data extraction, the detection data for each diode has been cleaned of noise interference and abnormal data caused by malfunctions in the detection equipment. The application objects of the diodes requiring fault detection are designated as the detection and analysis subjects, and corresponding data are stored within the detection and analysis subjects. In several detection and analysis targets, all of which are diodes used in the actual circuit of the detection and analysis subject, the categories of possible fault characteristics of the detection and analysis targets include, but are not limited to: forward conduction failure, reverse cutoff failure, switching characteristic degradation, overheating aging, poor pin contact, and package damage. Each fault characteristic has at least one fault characteristic evaluation parameter to assess whether the fault characteristic exists. The categories of fault characteristic evaluation parameters include, but are not limited to: forward voltage drop, forward current, forward turn-on time, reverse leakage current, reverse recovery time, reverse breakdown voltage, peak surge current, peak surge voltage, junction temperature, temperature coefficient, vibration acceleration, and pin resistance. This invention utilizes several training optimization records. Each training optimization record records at least one instance of the training data extraction process for the model training phase used to detect faults in diodes within the analysis subject. This process includes application difference parameters, application feature parameters, evaluation correlation index, time-series correlation parameters, fault concurrency parameters, feature evaluation cross-index, and evaluation cross-index. Each training optimization record also has a corresponding pass / fail marker. The pass / fail marker records whether the effectiveness and adaptability of the fault detection model for diodes within the analysis subject meet the user's requirements. It is understood that the user can determine whether the effectiveness and adaptability of the fault detection model for diodes within the analysis subject meet the user's requirements based on self-defined indicators.

[0025] Specifically, the analytical difference states include a first type of analytical difference state and a second type of difference state; The detection and analysis subject in the first type of analysis difference state is the detection and analysis subject whose application difference parameter is greater than the preset application difference parameter or whose application feature parameter is greater than the preset application feature parameter; The detection and analysis subject in the second type of analysis difference state is the detection and analysis subject whose application difference parameter is less than or equal to the preset application difference parameter and whose application feature parameter is less than or equal to the preset application feature parameter.

[0026] The values ​​of the preset application difference parameters and preset application feature parameters can be determined by the user based on the actual working scenario. For example, the user can set them based on the training optimization records. The higher the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject, the smaller the values ​​of the preset application difference parameters and preset application feature parameters. A method for determining the value of the preset application difference parameters is provided, in which the maximum value of the application difference parameters of the detection and analysis subject in the second-class analysis difference state in the training optimization records that meet the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset application difference parameter. A method for determining the value of the preset application feature parameters is provided, in which the maximum value of the application feature parameters of the detection and analysis subject in the second-class analysis difference state in the training optimization records that meet the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset application feature parameter.

[0027] Specifically, when the detection and analysis subject is in a state of difference in analysis, feature difference analysis is performed on the detection and analysis subject; The evaluation correlation index of any of the detection and analysis targets is determined based on the fault feature evaluation parameters corresponding to each fault feature.

[0028] If the detection and analysis subject is in a state of difference, it indicates that the fault feature evaluation parameters of the detection and analysis targets within the subject have a relatively wide range or large differences under normal working conditions during actual application. This indicates that the training process of the model for fault detection of each detection and analysis target within the subject needs to ensure the effective extraction of the unique difference features of each detection and analysis target. The correlation index is used to characterize the degree of correlation between the fault feature evaluation parameters on which the evaluation process of various types of fault features of each detection and analysis target is based. In this way, the acquisition method of training data for fault features corresponding to different detection and analysis targets can be optimized in a targeted manner to ensure the effectiveness of the training results of the fault detection model. For a single detection and analysis target, the evaluation correlation index n represents the number of categories of fault characteristics present in the target being detected and analyzed. The number of categories of associated fault features existing in the detection and analysis target. For a single fault feature existing in the detection and analysis target, if the fault feature evaluation parameter corresponding to the fault feature is the same as the fault feature evaluation parameter corresponding to other fault features existing in the detection and analysis target, then the fault feature is recorded as an associated fault feature.

[0029] Specifically, feature integration processing is performed on detection and analysis targets whose evaluation correlation index is greater than the preset evaluation correlation index, and whether data fusion extraction is performed is determined based on the temporal correlation parameters of each feature integration combination. Data fusion extraction is performed on feature ensembles with temporal correlation parameters greater than preset temporal correlation parameters to obtain training execution data for such feature ensembles, and the data extraction scale is determined based on the temporal correlation parameters. The data extraction scale and the corresponding time-series correlation parameter of the feature integration combination are positively correlated.

[0030] For a single detection and analysis target, if the evaluation correlation index is greater than the preset evaluation correlation index, it indicates that there is an overlap between the fault feature evaluation parameters of different fault features in the fault detection process of the detection and analysis target. Therefore, feature integration processing is required for the detection and analysis target to ensure that the acquired training data can effectively improve the fault detection model after training and effectively identify each fault feature, thereby improving data processing efficiency. When obtaining the training execution data set of the detection and analysis target, training data is extracted for each feature integration combination. The set of training execution data extracted from each feature integration combination or fault features that do not form a feature integration combination is recorded as the training execution data set of the detection and analysis target. For a single feature integration combination, the temporal correlation parameter is the maximum value of the concurrency parameters among the fault features present in the feature integration combination. If the temporal correlation parameter of the feature set combination is greater than the preset temporal correlation parameter, it indicates that the occurrence times of the fault features included in the feature integration combination are relatively concentrated, which in turn indicates that the occurrence times of the abnormal data of the corresponding fault feature evaluation parameters are relatively concentrated and correlated. By performing data fusion extraction on the feature integration combination, it is ensured that the training process of the fault detection model can effectively retain the correlation between fault features. The set of training execution data corresponding to each fault feature included in the feature integration combination is used as input to execute the training of the fault detection model. Fault features within the feature integration combination whose temporal correlation parameter is less than or equal to the preset temporal correlation parameter, as well as fault features that do not form a feature integration combination, are isolated. The values ​​of the preset evaluation correlation index and the preset time-series correlation parameter can be determined by the user according to the actual working scenario. For example, the user can set them based on the training optimization records. The higher the user's requirements for the effectiveness and adaptability of the diode fault detection model within the detection and analysis subject, the higher the value of the preset evaluation correlation index. A method for determining the value of the preset evaluation correlation index is provided, in which the training optimization records for feature difference analysis of the detection and analysis subject are recorded as difference reference records. The minimum value of the evaluation correlation index of the detection and analysis target that meets the user's requirements for the effectiveness and adaptability of the diode fault detection model within the detection and analysis subject is recorded as the preset evaluation correlation index. A method for determining the value of the preset time-series correlation parameter is provided, in which the minimum value of the time-series correlation parameter of the feature integration combination extracted by data fusion in the difference reference records that meets the user's requirements for the effectiveness and adaptability of the diode fault detection model within the detection and analysis subject is recorded as the preset time-series correlation parameter.

[0031] Specifically, the setting method of the feature integration combination is determined based on the fault concurrency parameters of the detection and analysis target; Feature integration combination based on concurrent time series correlation index to determine the detection and analysis target where the fault concurrency parameter is greater than the preset fault concurrency parameter; Feature integration combination based on the evaluation overlap index to determine the detection and analysis target whose fault concurrency parameter is less than or equal to the preset fault concurrency parameter.

[0032] Specifically, for a single detection and analysis target, the fault concurrency parameter is the proportion of the number of concurrent fault feature categories existing in the detection and analysis target to the total number of fault feature categories existing in the corresponding detection and analysis target. For any two fault features, if the concurrency parameter between the two fault features is less than the preset concurrency parameter, then the two fault features are mutually recorded as each other's concurrent fault features. The concurrency parameter is the average of the interval between each occurrence time of the two fault features. The preset fault concurrency parameter and its value can be determined by the user according to the actual working scenario. For example, the user can set it based on the training optimization record. A method for determining the value of the preset fault concurrency parameter is provided, where the training optimization record that determines the feature integration combination based on the concurrent time series correlation index is recorded as the concurrent reference record, and the minimum value of the fault concurrency parameter in the concurrent reference record that meets the user's requirements for the effectiveness and adaptability of the fault detection model of the diode in the detection and analysis subject is recorded as the preset fault concurrency parameter. A value for the preset concurrency parameter is provided, where the preset concurrency parameter is 2 minutes. The feature integration combination is a set of several fault features corresponding to the detection and analysis target. For a single detection and analysis target, if the fault concurrency parameter is greater than the preset fault concurrency parameter, it indicates that the fault features of the detection and analysis target have a high degree of concurrency. In this case, the temporal correlation between the fault features existing in the determined feature integration combination is guaranteed. The concurrency temporal correlation index of the determined feature integration combination is greater than the preset concurrency temporal correlation index. For a single feature integration combination, the concurrency temporal correlation index is the average of the concurrency parameters between each fault feature in the feature integration combination and other fault features in the feature integration combination. If the fault concurrency parameter is less than or equal to the preset fault concurrency parameter, it indicates that the concurrency between the fault features of the detection and analysis target is low. In this case, the degree of overlap of the evaluation parameters between the fault features in the determined feature integration combination is guaranteed. The evaluation overlap index of the determined feature integration combination is greater than the preset evaluation overlap index. For a single feature integration combination, the evaluation overlap index is the average value of the evaluation overlap parameters between each fault feature in the fault features in the feature integration combination and other fault features in the feature integration combination. For any two fault features, the evaluation overlap parameter = the number of categories of fault feature evaluation parameters that both fault features correspond to / the number of categories of different fault feature evaluation parameters that the two fault features correspond to. The values ​​of the preset concurrent timing correlation index and the preset evaluation overlap index can be determined by the user according to the actual working scenario. For example, the user can set them based on the training optimization records. The higher the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject, the larger the value of the preset concurrent timing correlation index and the larger the value of the preset evaluation overlap index. A method for determining the value of the preset concurrent timing correlation index is provided, in which the minimum value of the concurrent timing correlation index in the concurrent reference records that meets the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset concurrent timing correlation index. A method for determining the value of the preset evaluation overlap index is provided, in which the training optimization records that determine the feature integration combination based on the evaluation overlap index are recorded as overlap reference records, and the minimum value of the evaluation overlap index in the overlap reference records that meets the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset evaluation overlap index.

[0033] Specifically, for detection and analysis targets whose evaluation correlation index is less than or equal to the preset evaluation correlation index, feature isolation processing is performed, data fusion and extraction are performed for each fault feature, and the data extraction scale of each fault feature is determined according to the evaluation correlation index. The data extraction scale is positively correlated with the corresponding fault feature assessment correlation index.

[0034] Specifically, for a single detection and analysis target, if the evaluation correlation index is less than or equal to the preset evaluation correlation index, it indicates that the degree of crossover between the evaluation parameters of different fault features in the fault detection process for that detection and analysis target is small. Therefore, feature isolation processing is performed for this detection and analysis target to ensure that the acquired training data can effectively improve the effectiveness of the trained fault detection model in identifying each fault feature. When obtaining the training execution data set for this detection and analysis target, data fusion extraction is performed separately for each fault feature to obtain the training data on which the corresponding fault feature is based. The set of training execution data obtained for each fault feature is denoted as the training execution data set, and the training execution data corresponding to each fault feature is used as input to train the fault detection model. For a single fault feature, based on sliding... When obtaining the training execution data corresponding to the fault feature using the moving window method, the data extraction scale corresponding to the fault feature is determined based on the evaluation correlation index. The data extraction scale includes extraction window parameters and window sliding parameters. The extraction window parameters are the window size corresponding to the extraction of data obtained from the evaluation parameters of each category of fault features corresponding to the fault feature. The window sliding parameters are the window movement step size corresponding to the extraction of data obtained from the evaluation parameters of each category of fault features corresponding to the fault feature. The units of measurement for both the extraction window parameters and the window sliding parameters are time units. The evaluation correlation index is the number of different categories of fault feature evaluation parameters corresponding to the fault feature. How to perform data extraction based on the sliding window method is a topic already known to those skilled in the art and will not be elaborated here.

[0035] Specifically, when the detection and analysis subject is in a state of second-class analysis difference or has completed feature difference analysis, feature applicability analysis is performed on the detection and analysis subject; The feature evaluation cross-index is determined based on the fault feature evaluation parameters of each detection and analysis target within the detection and analysis subject.

[0036] Specifically, when the feature evaluation cross index of the detection and analysis subject is greater than the preset feature cross evaluation index, applicable optimization is performed on the detection and analysis subject. Data extraction is performed based on multi-scale data extraction and various optimized extraction parameters; The optimized extraction parameters are the fault feature evaluation parameters where the evaluation cross-index is greater than the preset evaluation cross-index.

[0037] When the detection and analysis subject is in the second-class analysis difference state or has completed the feature difference analysis, it indicates that the range of fault feature evaluation parameters of the detection and analysis subject in actual application is relatively small and the difference between fault feature evaluation parameters is small, or that the optimization of targeted data extraction for each fault feature has been completed. At this time, feature applicability analysis is performed on the detection and analysis subject so that the acquired training data can improve the universality of the trained fault detection model for the abnormal identification of fault feature evaluation parameters of various categories. The feature evaluation cross index , This refers to the number of categories of evaluation parameters for different fault characteristics corresponding to each detection and analysis target within the detection and analysis subject. The number of different categories of cross-evaluation parameters corresponding to each detection and analysis target within the detection and analysis subject. For a single fault feature evaluation parameter, if the fault feature evaluation parameter is used to evaluate the fault features of multiple detection and analysis targets, then the fault feature evaluation parameter is recorded as a cross-evaluation parameter. The number of all detection and analysis targets corresponding to the fault feature evaluation parameter used for the fault feature evaluation is recorded as the evaluation cross-index of the fault feature evaluation parameter. If the feature evaluation cross-index of the detection and analysis subject is greater than the preset feature cross-evaluation index, it indicates that there is a large degree of cross-interaction among the parameters used by the detection and analysis subject to analyze each fault feature. This indicates that there are many categories of fault feature evaluation parameters that can perform the identification of multiple fault features. The cross-index of the various fault feature evaluation parameters involved in the fault detection of each detection and analysis target corresponding to the detection and analysis subject is evaluated. For a single fault feature evaluation parameter whose evaluation cross-index is greater than the preset evaluation cross-index, i.e., the optimized extraction parameter, data extraction is performed based on multiple data extraction scales to obtain data features at multiple scales. This improves data processing efficiency while ensuring the applicability of the evaluation of each fault feature. During multi-scale data extraction, the window size and sliding step size of the sliding window are determined from small to large to determine multiple data extraction scales. The number of determined data extraction scales is positively correlated with the evaluation cross-index of the corresponding optimized extraction parameter, ensuring that the acquired data can meet the needs of different time scales. The preset feature cross-evaluation index and the preset evaluation cross-index can be determined by the user according to the actual working scenario. For example, the user can set them based on training optimization records. The higher the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject, the smaller the value of the preset feature cross-evaluation index. A method for determining the value of the preset feature cross-evaluation index is provided, in which the training optimization records for feature applicability analysis of the detection and analysis subject are recorded as applicable reference records. The minimum value of the feature cross-evaluation index of the detection and analysis subject in the applicable reference records that meet the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset feature cross-evaluation index. A method for determining the value of the preset evaluation cross-index is provided, in which the minimum value of the evaluation cross-index of the optimized extracted parameters in the applicable reference records that meet the user's requirements for the effectiveness and adaptability of the fault detection model of diodes within the detection and analysis subject is recorded as the preset evaluation cross-index.

[0038] Specifically, the application difference parameters and the application feature parameters are determined based on the evaluation value range of various fault feature evaluation parameters of the detection and analysis targets existing within the detection and analysis subject.

[0039] Wherein, the application difference parameter is the average of the range difference index of each detection and analysis target within the detection and analysis subject for each fault feature evaluation parameter. For a single fault feature evaluation parameter, the evaluation value range of each detection and analysis subject for that fault feature evaluation parameter is obtained, and the range difference index is... , This represents the average of the reference values ​​for the evaluation range of this fault characteristic assessment parameter by various testing and analysis entities. The standard deviation of the reference values ​​for the evaluation value range of each detection and analysis subject for this fault feature evaluation parameter is defined as follows: the maximum value within the evaluation value range is the maximum value allowed for this fault feature evaluation parameter under normal operating conditions; the minimum value within the evaluation value range is the minimum value allowed for this fault feature evaluation parameter under normal operating conditions; the reference value for the evaluation value range is the median value existing within the evaluation value range; and the applied feature parameter is the average of the range coverage index of each detection and analysis target within the detection and analysis subject for each fault feature evaluation parameter. For a single fault feature evaluation parameter, the range coverage index is defined as follows: , This represents the absolute value of the difference between the maximum and minimum values ​​within the range of evaluation values ​​for this fault characteristic parameter by each testing and analysis entity. The maximum value within the range of evaluation values ​​for each fault characteristic evaluation parameter by each testing and analysis entity is the maximum value. How to determine the evaluation value range of each fault characteristic evaluation parameter is a topic easily understood by those skilled in the art and will not be elaborated here.

[0040] Please see Figure 4 As shown, Figure 4 The present invention provides a fault detection system for diodes, comprising: (The diagram shows the module connection diagram for a diode fault detection system.) The subject evaluation module is used to obtain the application parameters of each detection and analysis target within the detection and analysis subject in order to determine the analysis difference state of the detection and analysis subject. The analysis difference state is determined based on the application difference parameters and application characteristic parameters. An analysis execution module, which is connected to the subject evaluation module, is used to determine whether to perform feature difference analysis or feature applicability analysis on the subject based on the analysis difference status of the subject being analyzed. The difference analysis module, which is connected to the analysis execution module, is used to determine whether to perform feature integration processing or feature isolation processing on the corresponding detection and analysis target based on the evaluation correlation index of each detection and analysis target, so as to obtain the training execution data set of each detection and analysis target. The evaluation correlation index is determined based on the fault feature evaluation parameters of each detection and analysis target. An applicable analysis module, which is connected to the analysis execution module, is used to evaluate the cross-index based on the characteristics of the detection and analysis subject to determine whether to perform applicable optimization, so as to determine the optimized extraction parameters.

[0041] This invention does not limit the specific structure of the main evaluation module, analysis execution module, difference analysis module, and applicable analysis module. They themselves and each unit therein can be composed of logic components, including field-programmable components, computers, or microprocessors in computers, but they need to be able to implement the above methods.

[0042] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A fault detection method for diodes, characterized in that, include: The application parameters of each detection and analysis target within the detection and analysis subject are obtained to determine the analysis difference state of the detection and analysis subject. The analysis difference state is determined based on the application difference parameters and application characteristic parameters. Determine whether to perform feature difference analysis or feature applicability analysis on the subject of detection and analysis based on the analysis difference status of the subject of detection and analysis. When performing feature difference analysis, the evaluation correlation index of each detection and analysis target determines whether to perform feature integration processing or feature isolation processing for the corresponding detection and analysis target in order to obtain the training execution data set of each detection and analysis target. The evaluation correlation index is determined based on the fault feature evaluation parameters of each detection and analysis target. When performing feature applicability analysis, the cross-index of the feature evaluation of the subject of detection and analysis is used to determine whether to perform applicability optimization, so as to determine the optimal extraction parameters.

2. The fault detection method for diodes according to claim 1, characterized in that, The analytical difference states include a first-class analytical difference state and a second-class difference state; The detection and analysis subject in the first type of analysis difference state is the detection and analysis subject whose application difference parameter is greater than the preset application difference parameter or whose application feature parameter is greater than the preset application feature parameter; The detection and analysis subject in the second type of analysis difference state is the detection and analysis subject whose application difference parameter is less than or equal to the preset application difference parameter and whose application feature parameter is less than or equal to the preset application feature parameter.

3. The fault detection method for diodes according to claim 1, characterized in that, When the detection and analysis subject is in a state of difference in analysis, feature difference analysis is performed on the detection and analysis subject. The evaluation correlation index of any of the detection and analysis targets is determined based on the fault feature evaluation parameters corresponding to each fault feature.

4. The fault detection method for diodes according to claim 3, characterized in that, For detection and analysis targets whose evaluation correlation index is greater than the preset evaluation correlation index, feature integration processing is performed, and whether data fusion extraction is performed is determined based on the temporal correlation parameters of each feature integration combination. Data fusion extraction is performed on feature ensembles with temporal correlation parameters greater than preset temporal correlation parameters to obtain training execution data for such feature ensembles, and the data extraction scale is determined based on the temporal correlation parameters. The data extraction scale and the corresponding time-series correlation parameter of the feature integration combination are positively correlated.

5. The fault detection method for diodes according to claim 4, characterized in that, The configuration of the feature integration combination is determined based on the fault concurrency parameters of the detection and analysis target; Feature integration combination based on concurrent time series correlation index to determine the detection and analysis target where the fault concurrency parameter is greater than the preset fault concurrency parameter; Feature integration combination based on the evaluation overlap index to determine the detection and analysis target whose fault concurrency parameter is less than or equal to the preset fault concurrency parameter.

6. The fault detection method for diodes according to claim 3, characterized in that, For detection and analysis targets whose evaluation correlation index is less than or equal to the preset evaluation correlation index, feature isolation processing is performed, data fusion and extraction are performed for each fault feature, and the data extraction scale of each fault feature is determined according to the evaluation correlation index. The data extraction scale is positively correlated with the corresponding fault feature assessment correlation index.

7. The fault detection method for diodes according to claim 1, characterized in that, When the detection and analysis subject is in a state of second-class analysis difference or has completed feature difference analysis, feature applicability analysis is performed on the detection and analysis subject. The feature evaluation cross-index is determined based on the fault feature evaluation parameters of each detection and analysis target within the detection and analysis subject.

8. The fault detection method for diodes according to claim 7, characterized in that, When the feature evaluation cross index of the detection and analysis subject is greater than the preset feature cross evaluation index, applicable optimization is performed on the detection and analysis subject. Multi-scale data extraction is performed for each optimized extraction parameter; The optimized extraction parameters are the fault feature evaluation parameters where the evaluation cross-index is greater than the preset evaluation cross-index.

9. The fault detection method for diodes according to claim 1, characterized in that, The application difference parameters and application feature parameters are determined based on the evaluation value range of various fault feature evaluation parameters of the detection and analysis targets existing within the detection and analysis subject.

10. A fault detection system applying the fault detection method for diodes according to claims 1 to 9, characterized in that, include: The subject evaluation module is used to obtain the application parameters of each detection and analysis target within the detection and analysis subject in order to determine the analysis difference state of the detection and analysis subject. The analysis difference state is determined based on the application difference parameters and application characteristic parameters. An analysis execution module, which is connected to the subject evaluation module, is used to determine whether to perform feature difference analysis or feature applicability analysis on the subject based on the analysis difference status of the subject being analyzed. The difference analysis module, which is connected to the analysis execution module, is used to determine whether to perform feature integration processing or feature isolation processing on the corresponding detection and analysis target based on the evaluation correlation index of each detection and analysis target, so as to obtain the training execution data set of each detection and analysis target. The evaluation correlation index is determined based on the fault feature evaluation parameters of each detection and analysis target. An applicable analysis module, which is connected to the analysis execution module, is used to evaluate the cross-index based on the characteristics of the detection and analysis subject to determine whether to perform applicable optimization, so as to determine the optimized extraction parameters.

Images