Health condition detection system of electronic component

Abstract

The invention discloses an electronic component health condition detection system, which comprises an information acquisition module, an information analysis module, a simulation test module, a model construction module, a knowledge graph module, a health condition level module, a detection report module and a data management module, inputting the to-be-detected feature information of the electronic component into the health condition prediction model to obtain a prediction result; inputting the prediction result into an electronic component knowledge graph to obtain a health state analysis result; according to the method, historical data and test data of the electronic component are analyzed in a multi-dimensional mode based on the knowledge graph technology, factors related to abnormity of the electronic component are accurately found, the health state trend of the electronic component is predicted through artificial intelligence, the health state of the electronic component can be analyzed in time and comprehensively, and the detection precision is improved.

Description

Technical Field

[0001] The invention belongs to the field of electronic components, and in particular relates to a health status detection system for electronic components. Background Art

[0002] With the development of electronic technology, electronic components are increasingly used in various electronic products. The health of these components is directly related to the performance and reliability of the entire electronic product. Therefore, it is particularly important to accurately detect the health of electronic components.

[0003] Traditional electronic component inspection methods mainly rely on manual visual inspection, which is inefficient and greatly affected by the subjective factors of the inspectors. With the advancement of technology, automated inspection technology has gradually become mainstream, including machine vision systems, artificial intelligence (AI) and deep learning technology, as well as laser scanning and 3D imaging technology.

[0004] Chinese Patent Publication No.: CN118379298A discloses an electronic component detection method. The present invention discloses an electronic component detection method, which belongs to the field of component detection technology, and includes the following steps: S1, obtaining a surface processing image of the component, and extracting the absolute defect area of ​​the surface processing image; S2, generating an absolute defect value of the surface processing image based on the absolute defect area; S3, generating the detection result of the electronic component based on the absolute defect value of the surface processing image. In the present invention, the electronic component detection method can delineate the absolute defect area of ​​the image when the IC device is completed, that is, a large range is first determined in the surface processing image to determine whether the electronic component has defects; after obtaining the absolute defect area, a characteristic parameter is generated for the area, that is, the absolute defect value, and the final detection result is generated by the absolute defect value, which is conducive to timely discovering the defects of the IC device and improving the reliability and life of the electronic components.

[0005] However, in actual use, electronic components are affected by multi-dimensional and uncertain factors, and the health status of electronic components cannot be assessed simply by extracting the absolute defect area of ​​the surface processing image of the component. Under specific circumstances affected by multiple factors, electronic components may malfunction and fail to operate. Failure to provide early warning of abnormal problems will pose a hidden danger to the safe use of electronic products that rely on electronic components. Summary of the invention

[0006] In order to solve the problem that the influencing factors of electronic components in the above-mentioned prior art are multi-dimensional and uncertain, and the health status of electronic components cannot be evaluated simply by extracting the absolute defect area of ​​the surface processing image of the component, when under specific circumstances affected by multiple factors, the electronic components may become abnormal and unable to operate, and early warning cannot be given in the face of abnormal problems, posing hidden dangers to the safe use of electronic products that rely on electronic components. The present invention proposes a health status detection system for electronic components to improve the above-mentioned problems.

[0007] The specific application is as follows:

[0008] A health status detection system for electronic components, comprising an information collection module, an information analysis module, a simulation test module, a model construction module, a knowledge graph module, a health status level module, a detection report module and a data management module;

[0009] The information collection module is used to collect status monitoring information of abnormal electronic components, wherein the status monitoring information includes status monitoring information corresponding to when the abnormal electronic components occur abnormally during actual use, and the status monitoring information includes operation status information of the electronic components and signals generated during operation;

[0010] The information analysis module is used to determine the detection feature information according to the acquired state monitoring information, perform normalization preprocessing on the detection feature information by feature engineering, and then extract features from the detection feature information to generate detection features;

[0011] The simulation test module is used to control the electronic component detection device to perform a health condition test on the electronic component to be tested according to the detection characteristics, and generate a test detection characteristic;

[0012] The model building module builds a health status prediction model according to the detection characteristics and the experimental detection characteristics;

[0013] The knowledge graph module inputs the feature information to be detected into the health status prediction model to obtain a prediction result, and inputs the prediction result into a pre-built electronic component knowledge graph to obtain a health status analysis value and analysis result of the electronic component;

[0014] The health status level module sorts the health status analysis values ​​output by the electronic component knowledge graph, takes samples of 15%, 50%, and 80% percentiles as three mean points, and outputs three health levels: low risk, medium risk, and high risk. The low risk does not require treatment, and the medium risk and high risk require detailed inspection reports.

[0015] The detection report module is used to generate a detailed health detection report when the health level is medium risk or high risk;

[0016] The data management module is used to store and manage the data generated during the health status detection process of the electronic components and the corresponding health status detection results.

[0017] Further, the information collection module includes an access submodule and a collection submodule;

[0018] The access submodule is connected to the electronic component that has an abnormality, and is used to obtain status monitoring information from the status monitoring terminal when the electronic component has an abnormality or a period of time before the abnormality occurs;

[0019] The acquisition submodule is used for the sensor to actively collect status monitoring information of abnormal electronic components;

[0020] Among them, the status monitoring information of abnormal electronic components includes the operating status information of the electronic components and the signals generated during operation; the operating status information includes the current curve, voltage curve, temperature curve, real-time current value, real-time voltage value, and real-time temperature value; the signals generated during operation represent the signal information generated by the electronic components during the working process.

[0021] Furthermore, the information analysis module includes a data preprocessing submodule and a feature extraction submodule;

[0022] The data preprocessing submodule preprocesses the state monitoring information to obtain a preprocessed data set, including performing data screening processing on the state monitoring information, wherein data that meets a preset sample standard is extracted from the state monitoring information to form first preprocessed data;

[0023] Eliminate erroneous data from the first preprocessed data, and convert data of the same indicator in different formats into data of the same format to obtain second preprocessed data;

[0024] Screening out records with missing values ​​in a single record in the second preprocessed data and filling the missing values ​​to obtain third preprocessed data; the missing values ​​are filled by Lagrange interpolation method;

[0025] Normalizing the data of different indicators in the third preprocessed data to obtain preprocessed data sets as detection feature information;

[0026] The feature extraction submodule is used to perform feature extraction based on the detection feature information to generate detection features, wherein the detection features include electrical features, signal features and temperature features; wherein the electrical features include the voltage, current, resonant frequency and impedance of electronic components; and the signal features include sound signals and vibration signals, including at least the vibration and noise of fans and transformers.

[0027] Furthermore, the specific implementation process of the signal feature is:

[0028] L1: Obtain historical test parameters, historical sound signal test data, and historical vibration signal test data of electronic components tested in normal working conditions and under different fault types through sensors, and mark the fault severity of the historical test parameters, historical sound signal test data, and historical vibration signal test data tested under different fault types;

[0029] L2: Draw a waveform signal curve graph for the historical sound signal test data and the historical vibration signal test data, calculate the signal frequency fluctuation information of the historical sound signal test data and the historical vibration signal test data according to the waveform signal curve graph, determine the abnormal signal position according to the signal frequency fluctuation information, perform waveform extraction on the abnormal signal position in the waveform signal curve graph, and determine the noise type existing in the historical sound signal test data and the historical vibration signal test data;

[0030] L3: Determine a signal noise removal method for the historical sound signal test data and the historical vibration signal test data according to the existing noise type and the noise removal type of each denoising method, and perform abnormal processing on the historical sound signal test data and the historical vibration signal test data according to the signal noise removal method;

[0031] L4: Divide the historical sound signal test data and the historical vibration signal test data after the abnormal processing into N equal parts of sound signal data blocks and vibration signal data blocks, convert the N into binary representation, and determine the binary bit length of N;

[0032] L5: Based on wavelet transform and wavelet function, CWT continuous wavelet transform operation is performed on the sound signal data block and the vibration signal data block, the sound signal data block and the vibration signal data block are scaled and translated, and the historical sound signal test data and the historical vibration signal test data are converted into frequency domain signals to obtain the historical sound signal frequency domain data and the historical vibration signal frequency domain data;

[0033] L6: Plotting the historical sound signal frequency domain data and the historical vibration signal frequency domain data into a sound signal spectrum diagram and a vibration signal spectrum diagram;

[0034] L7: Calculate the peak frequency, frequency bandwidth, center frequency, and spectrum energy of the sound signal spectrum diagram and the vibration signal spectrum diagram to obtain the sound signal spectrum characteristics and the vibration signal spectrum characteristics.

[0035] Further, the simulation test module includes an electrical simulation submodule and a vibration simulation submodule;

[0036] The electrical simulation submodule is used to perform a simulation test on the electronic components according to the obtained electrical characteristics, so that the tested electronic components operate in a state corresponding to the electrical characteristics, and generate test electrical characteristics; wherein the operation control device is electrically connected to the tested electronic components, and is used to adjust the voltage, current, resonant frequency, and impedance parameters of the tested electronic components according to the electrical characteristics; wherein the operation control device includes circuit simulation software, hardware simulation platform, and load simulation equipment;

[0037] The vibration simulation submodule is used to generate test signal characteristics based on the obtained signal characteristics so that the test electronic components are affected by external forces corresponding to the signal characteristics; wherein the external force influence adopts the HBK LDS vibration test system.

[0038] Furthermore, the detection characteristic information includes a continuous and stable fixed state and a changing state that changes with time; wherein, when the detection characteristic information is in a changing state, the detection characteristic information includes the detection characteristic information corresponding to each detection moment, so that the simulation test module can dynamically adjust and control the corresponding simulation parameters according to the detection characteristic information.

[0039] Furthermore, the specific implementation process of the health status prediction model is as follows:

[0040] Using regression analysis algorithm, a benchmark health assessment model is constructed, and the importance of each influencing factor of the detection characteristics and experimental detection characteristics is evaluated through standardized regression coefficients, and the independent influencing factors of electronic components are screened out to form an independent variable data set;

[0041] The independent variable data set is trained through the XGBoost model, where the independent variables used to train the model are the variables in the independent variable data set. After multiple iterative training, the optimal health status prediction model is obtained.

[0042] Furthermore, the process of constructing the electronic component knowledge graph includes:

[0043] Construct an electronic component knowledge graph based on technical documents, product specifications, user manuals, maintenance records, and patent literature related to electronic components;

[0044] The knowledge graph DeepKE is constructed, and the knowledge graph ontology is initialized in a hierarchical manner through ChatGPT retrieval, and the topics are divided into entities, relationships, attributes and events. The topics are processed through natural language processing technology;

[0045] Based on the end-to-end entity relationship extraction of the CNN-LSTMs model, a network-like knowledge structure is formed, and attribute information of specific entities is collected from different information sources. The relationship pattern between entity attributes and attribute values ​​is mined, and the event pattern between entity attributes and attribute values ​​is mined, thereby locating attribute names and attribute values ​​in the text;

[0046] Deep neural networks are used to parse time-series, high-dimensional, heterogeneous historical test data, explore the correlation between abnormal data performance, and then deduce the logical rules behind it. An intelligent engine is built based on causal logic, conditional logic, sequential logic, and concurrent logic. The engine is used to build a knowledge graph of electronic components. The historical test data includes detection features extracted by the information analysis module and experimental detection features extracted by the simulation test module.

[0047] Furthermore, the detection report module also includes a statistical unit;

[0048] The statistical unit is used to collect statistics based on the health status detection results obtained from the electronic components to be tested, and generate a health status detection report for the electronic components to be tested.

[0049] The beneficial effects of the health status detection system of an electronic component of the present invention are as follows:

[0050] The present invention collects the state monitoring information of abnormal electronic components through the information collection module, wherein the state monitoring information includes the state monitoring information correspondingly obtained when the abnormal electronic components occur abnormally during actual use, and the state monitoring information includes the operation state information of the electronic components and the signals generated during operation; the information analysis module is used to determine the detection feature information according to the acquired state monitoring information, and use feature engineering to perform normalization preprocessing on the detection feature information, and then extract the features in the detection feature information to generate the detection feature; the simulation test module is used to control the electronic component detection device to perform a health test on the test electronic component according to the detection feature, and generate the test detection feature; the model construction module constructs a health prediction model according to the detection feature and the test detection feature; the knowledge graph module inputs the feature information to be detected into the health prediction model to obtain the prediction result, and inputs the prediction result into the pre-constructed electronic component knowledge graph to obtain the health status analysis value and analysis result of the electronic component. The present invention analyzes the historical data and test data of electronic components in multiple dimensions based on the knowledge graph technology, accurately finds the factors related to the abnormality of the electronic components, and uses artificial intelligence to predict the health status trend of the electronic components, which can timely and comprehensively analyze the health status of the electronic components and improve the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A system architecture diagram of a health status detection system for electronic components of the present invention;

[0052] Figure 2 This is a diagram of an information collection module of a health status detection system for electronic components of the present invention;

[0053] Figure 3 This is an information analysis module diagram of a health status detection system for electronic components of the present invention;

[0054] Figure 4 This is a diagram of a simulation test module of a health status detection system for electronic components of the present invention. DETAILED DESCRIPTION

[0055] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Example 1

[0057] First, the technical terms involved in the embodiments of the present application are explained.

[0058] (1) Knowledge Graph

[0059] A knowledge graph is a structured semantic knowledge base that uses graphics to express the relationships between entities. These entities can be devices, places, organizations, concepts, etc., while relationships describe the various connections between entities, such as "belongs to", "is located in", "created by", etc. Knowledge graphs are often used to enhance the understanding capabilities of search engines, provide richer information retrieval, and play a role in recommendation systems, natural language processing, data mining and other fields.

[0060] Figure 1 This is a system architecture diagram of a health status detection system for electronic components of the present invention.

[0061] Its system includes information collection module, information analysis module, simulation test module, model building module, knowledge graph module, health status level module, test report module and data management module;

[0062] The information collection module is used to collect status monitoring information of abnormal electronic components, wherein the status monitoring information includes status monitoring information corresponding to when the abnormal electronic components occur abnormally during actual use, and the status monitoring information includes operation status information of the electronic components and signals generated during operation;

[0063] The information analysis module is used to determine the detection feature information according to the acquired state monitoring information, perform normalization preprocessing on the detection feature information by feature engineering, and then extract features from the detection feature information to generate detection features;

[0064] The simulation test module is used to control the electronic component detection device to perform a health condition test on the electronic component to be tested according to the detection characteristics, and generate a test detection characteristic;

[0065] The model building module builds a health status prediction model according to the detection characteristics and the experimental detection characteristics;

[0066] The knowledge graph module inputs the feature information to be detected into the health status prediction model to obtain a prediction result, and inputs the prediction result into a pre-built electronic component knowledge graph to obtain a health status analysis value and analysis result of the electronic component;

[0067] The health status level module sorts the health status analysis values ​​output by the electronic component knowledge graph, takes samples of 15%, 50%, and 80% percentiles as three mean points, and outputs three health levels: low risk, medium risk, and high risk. The low risk does not require treatment, and the medium risk and high risk require detailed inspection reports.

[0068] The detection report module is used to generate a detailed health detection report when the health level is medium risk or high risk;

[0069] The data management module is used to store and manage the data generated during the health status detection process of the electronic components and the corresponding health status detection results.

[0070] Further, the information collection module includes an access submodule and a collection submodule;

[0071] The access submodule is connected to the electronic component that has an abnormality, and is used to obtain status monitoring information from the status monitoring terminal when the electronic component has an abnormality or a period of time before the abnormality occurs;

[0072] The acquisition submodule is used for the sensor to actively collect status monitoring information of abnormal electronic components;

[0073] Among them, the status monitoring information of abnormal electronic components includes the operating status information of the electronic components and the signals generated during operation; the operating status information includes the current curve, voltage curve, temperature curve, real-time current value, real-time voltage value, and real-time temperature value; the signals generated during operation represent the signal information generated by the electronic components during the working process.

[0074] Specifically, monitoring of current curve, voltage curve and temperature curve:

[0075] Using high-precision sensors and data acquisition systems, the current, voltage and temperature changes of electronic components under different working conditions are recorded in real time. These curves can reflect the behavior patterns of components in normal operation and possible failure.

[0076] Recording of real-time current value, real-time voltage value, and real-time temperature value:

[0077] Through the monitoring system connected to the electronic components, the current, voltage and temperature values ​​are obtained in real time. These real-time data are crucial for quickly identifying abnormalities and responding;

[0078] Analysis of signals generated during operation:

[0079] Signal information generated by electronic components during operation, including vibration, electromagnetic interference (EMI), and noise, can be identified through advanced signal processing techniques and algorithms to identify abnormal signal characteristics that may indicate faults;

[0080] The advanced signal processing techniques and algorithms include but are not limited to the following algorithms:

[0081] Empirical Mode Decomposition (EMD): EMD is an adaptive method for decomposing a complex signal into a series of intrinsic mode functions (IMFs), each of which has different frequency characteristics and can represent different parts of the signal;

[0082] Ensemble empirical mode decomposition (EEMD): EEMD is a variant of EMD that solves the problem of modal aliasing by adding white noise to the signal, then performing EMD on multiple such signals and finally taking the average;

[0083] Variational Mode Decomposition (VMD): VMD is a non-recursive method that decomposes the signal into modes with different center frequencies by solving a variational problem.

[0084] Singular Spectrum Analysis (SSA): SSA is a method used to analyze time series data. It extracts signals representing different components of the original time series by constructing a trajectory matrix and decomposing and reconstructing it.

[0085] Short-time Fourier transform (STFT): STFT is a method of local spectrum analysis of a signal, which analyzes how the frequency components of a signal change over time by applying Fourier transform on a local window of the signal;

[0086] Adaptive filter: Adaptive filters can automatically adjust their parameters according to the characteristics of the input signal, such as the least mean square (LMS) algorithm and the recursive least squares (RLS) algorithm, which are often used for noise elimination and signal prediction;

[0087] High-order spectrum analysis: High-order spectrum analysis is a method to study the statistical characteristics of signals, which can reveal the nonlinearity and non-Gaussian characteristics of signals;

[0088] Principal Component Analysis (PCA): PCA is a statistical method used to transform possibly correlated variables into a set of linearly uncorrelated variables with a smaller number of values ​​through an orthogonal transformation;

[0089] Wavelet Packet Decomposition (WPD): WPD is an extension of wavelet transform, which recursively decomposes the signal to obtain sub-band signals of different frequency bands;

[0090] Maximum Overlap Discrete Wavelet Transform (MODWT): MODWT is a variant of wavelet transform that does not perform downsampling, retains all signal information, and is suitable for accurate signal analysis.

[0091] Data collection and processing:

[0092] Use automated test equipment ATE and data acquisition cards DAQ to collect operating data of components. These devices can provide dynamic testing capabilities and perform tests such as IC testing, memory testing, and RF testing.

[0093] Further, the information analysis module includes a data preprocessing submodule and a feature extraction submodule;

[0094] The data preprocessing submodule preprocesses the state monitoring information to obtain a preprocessed data set, including performing data screening processing on the state monitoring information, wherein data that meets a preset sample standard is extracted from the state monitoring information to form first preprocessed data;

[0095] Eliminate erroneous data from the first preprocessed data, and convert data of the same indicator in different formats into data of the same format to obtain second preprocessed data;

[0096] Screening out records with missing values ​​in a single record in the second preprocessed data and filling the missing values ​​to obtain third preprocessed data; the missing values ​​are filled by Lagrange interpolation method;

[0097] Normalizing the data of different indicators in the third preprocessed data to obtain preprocessed data sets as detection feature information;

[0098] The feature extraction submodule is used to perform feature extraction based on the detection feature information to generate detection features, wherein the detection features include electrical features, signal features and temperature features; wherein the electrical features include the voltage, current, resonant frequency and impedance of electronic components; and the signal features include sound signals and vibration signals, including at least the vibration and noise of fans and transformers.

[0099] Furthermore, the specific implementation process of the signal feature is:

[0100] L1: Obtain historical test parameters, historical sound signal test data, and historical vibration signal test data of electronic components tested in normal working conditions and under different fault types through sensors, and mark the fault severity of the historical test parameters, historical sound signal test data, and historical vibration signal test data tested under different fault types;

[0101] L2: Draw a waveform signal curve graph for the historical sound signal test data and the historical vibration signal test data, calculate the signal frequency fluctuation information of the historical sound signal test data and the historical vibration signal test data according to the waveform signal curve graph, determine the abnormal signal position according to the signal frequency fluctuation information, perform waveform extraction on the abnormal signal position in the waveform signal curve graph, and determine the noise type existing in the historical sound signal test data and the historical vibration signal test data;

[0102] L3: Determine a signal noise removal method for the historical sound signal test data and the historical vibration signal test data according to the existing noise type and the noise removal type of each denoising method, and perform abnormal processing on the historical sound signal test data and the historical vibration signal test data according to the signal noise removal method;

[0103] L4: Divide the historical sound signal test data and the historical vibration signal test data after the abnormal processing into N equal parts of sound signal data blocks and vibration signal data blocks, convert the N into binary representation, and determine the binary bit length of N;

[0104] L5: Based on wavelet transform and wavelet function, CWT continuous wavelet transform operation is performed on the sound signal data block and the vibration signal data block, the sound signal data block and the vibration signal data block are scaled and translated, and the historical sound signal test data and the historical vibration signal test data are converted into frequency domain signals to obtain the historical sound signal frequency domain data and the historical vibration signal frequency domain data;

[0105] L6: Plotting the historical sound signal frequency domain data and the historical vibration signal frequency domain data into a sound signal spectrum diagram and a vibration signal spectrum diagram;

[0106] L7: Calculate the peak frequency, frequency bandwidth, center frequency, and spectrum energy of the sound signal spectrum diagram and the vibration signal spectrum diagram to obtain the sound signal spectrum characteristics and the vibration signal spectrum characteristics.

[0107] Further, the simulation test module includes an electrical simulation submodule and a vibration simulation submodule;

[0108] The electrical simulation submodule is used to perform a simulation test on the electronic components according to the obtained electrical characteristics, so that the tested electronic components operate in a state corresponding to the electrical characteristics, and generate test electrical characteristics; wherein the operation control device is electrically connected to the tested electronic components, and is used to adjust the voltage, current, resonant frequency, and impedance parameters of the tested electronic components according to the electrical characteristics; wherein the operation control device includes circuit simulation software, hardware simulation platform, and load simulation equipment;

[0109] The vibration simulation submodule is used to generate test signal characteristics based on the obtained signal characteristics so that the test electronic components are affected by external forces corresponding to the signal characteristics; wherein the external force influence adopts the HBK LDS vibration test system.

[0110] Furthermore, the detection characteristic information includes a continuous and stable fixed state and a changing state that changes with time; wherein, when the detection characteristic information is in a changing state, the detection characteristic information includes the detection characteristic information corresponding to each detection moment, so that the simulation test module can dynamically adjust and control the corresponding simulation parameters according to the detection characteristic information.

[0111] Furthermore, the specific implementation process of the health status prediction model is as follows:

[0112] Using regression analysis algorithm, a benchmark health assessment model is constructed, and the importance of each influencing factor of the detection characteristics and experimental detection characteristics is evaluated through standardized regression coefficients, and the independent influencing factors of electronic components are screened out to form an independent variable data set;

[0113] The independent variable data set is trained through the XGBoost model, where the independent variables used to train the model are the variables in the independent variable data set. After multiple iterative training, the optimal health status prediction model is obtained.

[0114] Furthermore, the process of constructing the electronic component knowledge graph includes:

[0115] Construct an electronic component knowledge graph based on technical documents, product specifications, user manuals, maintenance records, and patent literature related to electronic components;

[0116] The knowledge graph DeepKE is constructed, and the knowledge graph ontology is initialized in a hierarchical manner through ChatGPT retrieval, and the topics are divided into entities, relationships, attributes and events. The topics are processed through natural language processing technology;

[0117] Based on the end-to-end entity relationship extraction of the CNN-LSTMs model, a network-like knowledge structure is formed, and attribute information of specific entities is collected from different information sources. The relationship pattern between entity attributes and attribute values ​​is mined, and the event pattern between entity attributes and attribute values ​​is mined, thereby locating attribute names and attribute values ​​in the text;

[0118] Deep neural networks are used to parse time-series, high-dimensional, heterogeneous historical test data, explore the correlation between abnormal data performance, and then deduce the logical rules behind it. An intelligent engine is built based on causal logic, conditional logic, sequential logic, and concurrent logic. The engine is used to build a knowledge graph of electronic components. The historical test data includes detection features extracted by the information analysis module and experimental detection features extracted by the simulation test module.

[0119] Specifically, under the premise that the above-mentioned detection characteristics and experimental detection characteristics obey the normal distribution and satisfy the homogeneity of variance, this embodiment also needs to perform single factor analysis, including current curve scoring, voltage curve scoring, temperature curve scoring, and the signal generated during operation represents the vibration signal information generated by the electronic components during the charging process. The statistically significant variables are preliminarily screened out through the T check, and these variables are included in the multiple linear regression model. The stepwise regression method is used for variable screening, and the standardized characteristic variables and partial regression square sums are calculated one by one. The Q check at the significance level α is as follows:

[0120]

[0121] Where D is the partial regression sum of squares, t is the relationship coefficient matrix, n is the number of all variables, j is the number of selected variables, and m is the number of standardized variables. Repeat this method, each time selecting one from the independent variables that have not been introduced into the regression model, until no variables are introduced after verification.

[0122] Furthermore, the detection report module also includes a statistical unit;

[0123] The statistical unit is used to collect statistics based on the health status detection results obtained from the electronic components to be tested, and generate a health status detection report for the electronic components to be tested.

[0124] The beneficial effects of the health status detection system of an electronic component of the present invention are as follows:

[0125] The present invention collects state monitoring information of abnormal electronic components through an information collection module, wherein the state monitoring information includes state monitoring information corresponding to the abnormal electronic components when the abnormal electronic components occur in the actual use process, and the state monitoring information includes the operation state information of the electronic components and the signals generated during the operation; the information analysis module is used to determine the detection feature information according to the acquired state monitoring information, and use feature engineering to perform normalization preprocessing on the detection feature information, and then extract the features in the detection feature information to generate the detection feature; the simulation test module is used to control the electronic component detection device to perform a health test on the test electronic component according to the detection feature, and generate the test detection feature; the model construction module constructs a health prediction model according to the detection feature and the test detection feature; the knowledge graph module inputs the feature information to be detected into the health prediction model to obtain the prediction result, and inputs the prediction result into the pre-constructed electronic component knowledge graph to obtain the health status analysis value and analysis result of the electronic component. The present invention analyzes the historical data and test data of electronic components in multiple dimensions based on the knowledge graph technology, accurately finds the factors related to the abnormality of the electronic components, and uses artificial intelligence to predict the health status trend of the electronic components, which can timely and comprehensively analyze the health status of the electronic components and improve the detection accuracy.

[0126] The present invention and its embodiments are described above, which is not restrictive. What is shown in the accompanying drawings is only one of the embodiments of the present invention, and the actual content is not limited to this. In short, if ordinary technicians in this field are inspired by it and do not deviate from the purpose of the invention, they can creatively design structural methods and embodiments similar to the technical scheme, which should all fall within the scope of protection of the present invention.

Claims

1. A health status detection system for electronic components, characterized in that: It includes information collection module, information analysis module, simulation test module, model building module, knowledge graph module, health status level module, test report module and data management module; The information collection module is used to collect status monitoring information of abnormal electronic components, wherein the status monitoring information includes status monitoring information corresponding to when the abnormal electronic components occur abnormally during actual use, and the status monitoring information includes operation status information of the electronic components and signals generated during operation; The information analysis module is used to determine the detection feature information according to the acquired state monitoring information, perform normalization preprocessing on the detection feature information by feature engineering, and then extract features from the detection feature information to generate detection features; The simulation test module is used to control the electronic component detection device to perform a health condition test on the electronic component to be tested according to the detection characteristics, and generate a test detection characteristic; The model building module builds a health status prediction model according to the detection characteristics and the experimental detection characteristics; The knowledge graph module inputs the feature information to be detected into the health status prediction model to obtain a prediction result, and inputs the prediction result into a pre-built electronic component knowledge graph to obtain a health status analysis value and analysis result of the electronic component; The health status level module sorts the health status analysis values ​​output by the electronic component knowledge graph, takes samples of 15%, 50%, and 80% percentiles as three mean points, and outputs three health levels: low risk, medium risk, and high risk. The low risk does not require treatment, and the medium risk and high risk require detailed inspection reports. The detection report module is used to generate a detailed health detection report when the health level is medium risk or high risk; The data management module is used to store and manage the data generated during the health status detection process of the electronic components and the corresponding health status detection results.

2. The electronic component health status detection system according to claim 1, characterized in that: The information collection module includes an access submodule and a collection submodule; The access submodule is connected to the electronic component that has an abnormality, and is used to obtain status monitoring information from the status monitoring terminal when the electronic component has an abnormality or a period of time before the abnormality occurs; The acquisition submodule is used for the sensor to actively collect status monitoring information of abnormal electronic components; Among them, the status monitoring information of abnormal electronic components includes the operating status information of the electronic components and the signals generated during operation; the operating status information includes the current curve, voltage curve, temperature curve, real-time current value, real-time voltage value, and real-time temperature value; the signals generated during operation represent the signal information generated by the electronic components during the working process.

3. The electronic component health status detection system according to claim 2, characterized in that: The information analysis module includes a data preprocessing submodule and a feature extraction submodule; The data preprocessing submodule preprocesses the state monitoring information to obtain a preprocessed data set, including performing data screening processing on the state monitoring information, wherein data that meets a preset sample standard is extracted from the state monitoring information to form first preprocessed data; Eliminate erroneous data from the first preprocessed data, and convert data of the same indicator in different formats into data of the same format to obtain second preprocessed data; Screening out records with missing values ​​in a single record in the second preprocessed data and filling the missing values ​​to obtain third preprocessed data; the missing values ​​are filled by Lagrange interpolation method; Normalizing the data of different indicators in the third preprocessed data to obtain preprocessed data sets as detection feature information; The feature extraction submodule is used to perform feature extraction based on the detection feature information to generate detection features, wherein the detection features include electrical features, signal features and temperature features; wherein the electrical features include the voltage, current, resonant frequency and impedance of electronic components; and the signal features include sound signals and vibration signals, including at least the vibration and noise of fans and transformers.

4. The electronic component health status detection system according to claim 3, characterized in that: The specific implementation process of the signal feature is as follows: L1: Obtain historical test parameters, historical sound signal test data, and historical vibration signal test data of electronic components tested in normal working conditions and under different fault types through sensors, and mark the fault severity of the historical test parameters, historical sound signal test data, and historical vibration signal test data tested under different fault types; L2: Draw a waveform signal curve graph for the historical sound signal test data and the historical vibration signal test data, calculate the signal frequency fluctuation information of the historical sound signal test data and the historical vibration signal test data according to the waveform signal curve graph, determine the abnormal signal position according to the signal frequency fluctuation information, perform waveform extraction on the abnormal signal position in the waveform signal curve graph, and determine the noise type existing in the historical sound signal test data and the historical vibration signal test data; L3: Determine a signal noise removal method for the historical sound signal test data and the historical vibration signal test data according to the existing noise type and the noise removal type of each denoising method, and perform abnormal processing on the historical sound signal test data and the historical vibration signal test data according to the signal noise removal method; L4: Divide the historical sound signal test data and the historical vibration signal test data after the abnormal processing into N equal parts of sound signal data blocks and vibration signal data blocks, convert the N into binary representation, and determine the binary bit length of N; L5: Based on wavelet transform and wavelet function, CWT continuous wavelet transform operation is performed on the sound signal data block and the vibration signal data block, the sound signal data block and the vibration signal data block are scaled and translated, and the historical sound signal test data and the historical vibration signal test data are converted into frequency domain signals to obtain the historical sound signal frequency domain data and the historical vibration signal frequency domain data; L6: Plotting the historical sound signal frequency domain data and the historical vibration signal frequency domain data into a sound signal spectrum diagram and a vibration signal spectrum diagram; L7: Calculate the peak frequency, frequency bandwidth, center frequency, and spectrum energy of the sound signal spectrum diagram and the vibration signal spectrum diagram to obtain the sound signal spectrum characteristics and the vibration signal spectrum characteristics.

5. The electronic component health status detection system according to claim 4, characterized in that: The simulation test module includes an electrical simulation submodule and a vibration simulation submodule; The electrical simulation submodule is used to perform a simulation test on the electronic components according to the obtained electrical characteristics, so that the tested electronic components operate in a state corresponding to the electrical characteristics, and generate experimental test electrical characteristics; The operation control device is electrically connected to the test electronic components and is used to adjust the voltage, current, resonant frequency and impedance parameters of the test electronic components according to the electrical characteristics; the operation control device includes circuit simulation software, hardware simulation platform and load simulation equipment; The vibration simulation submodule is used to generate a test signal feature based on the obtained signal feature so that the test electronic component is affected by an external force corresponding to the signal feature; The external force influence is tested by HBK LDS vibration test system.

6. The electronic component health status detection system according to claim 5, characterized in that: The detection characteristic information includes a continuous and stable fixed state and a changing state that changes with time; wherein, when the detection characteristic information is in a changing state, the detection characteristic information includes the detection characteristic information corresponding to each detection moment, so that the simulation test module can dynamically adjust and control the corresponding simulation parameters according to the detection characteristic information.

7. The electronic component health status detection system according to claim 6, characterized in that: The specific implementation process of the health status prediction model is as follows: Using regression analysis algorithm, a benchmark health assessment model is constructed, and the importance of each influencing factor of the detection characteristics and experimental detection characteristics is evaluated through standardized regression coefficients, and the independent influencing factors of electronic components are screened out to form an independent variable data set; The independent variable data set is trained through the XGBoost model, where the independent variables used to train the model are the variables in the independent variable data set. After multiple iterative training, the optimal health status prediction model is obtained.

8. The electronic component health status detection system according to claim 7, characterized in that: The construction process of the electronic component knowledge graph includes: Construct an electronic component knowledge graph based on technical documents, product specifications, user manuals, maintenance records, and patent literature related to electronic components; The knowledge graph DeepKE is constructed, and the knowledge graph ontology is initialized in a hierarchical manner through ChatGPT retrieval, and the topics are divided into entities, relationships, attributes and events. The topics are processed through natural language processing technology; Based on the end-to-end entity relationship extraction of the CNN-LSTMs model, a network-like knowledge structure is formed, and attribute information of specific entities is collected from different information sources. The relationship pattern between entity attributes and attribute values ​​is mined, and the event pattern between entity attributes and attribute values ​​is mined, thereby locating attribute names and attribute values ​​in the text; Deep neural networks are used to parse time-series, high-dimensional, heterogeneous historical test data, explore the correlation between abnormal data performance, and then deduce the logical rules behind it. An intelligent engine is built based on causal logic, conditional logic, sequential logic, and concurrent logic. The engine is used to build a knowledge graph of electronic components. The historical test data includes detection features extracted by the information analysis module and experimental detection features extracted by the simulation test module.

9. The electronic component health status detection system according to claim 8, characterized in that: The detection report module also includes a statistical unit; The statistical unit is used to collect statistics based on the health status detection results obtained from the electronic components to be tested, and generate a health status detection report for the electronic components to be tested.

Images