CN115329848A - Equipment fault diagnosis method and device, electronic equipment and computer storage medium - Google Patents
Equipment fault diagnosis method and device, electronic equipment and computer storage medium Download PDFInfo
- Publication number
- CN115329848A CN115329848A CN202210886381.2A CN202210886381A CN115329848A CN 115329848 A CN115329848 A CN 115329848A CN 202210886381 A CN202210886381 A CN 202210886381A CN 115329848 A CN115329848 A CN 115329848A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- data set
- sequence data
- library
- pattern
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The application provides a method and a device for diagnosing equipment faults, electronic equipment and a computer storage medium, wherein the method comprises the following steps: determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set; classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set; determining a first pattern library based on the recognizable time sequence data set, and labeling the unrecognizable time sequence data set based on an expert experience library to obtain a second pattern library; and determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library. According to the equipment fault diagnosis method provided by the embodiment of the application, the fault diagnosis mode base is determined through the time sequence abnormity detection model and the time sequence classification model in combination with the expert experience base, and the accuracy of fault diagnosis is improved.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for diagnosing a device fault, an electronic device, and a computer storage medium.
Background
In the process of equipment fault diagnosis, the main fault diagnosis method is a fault diagnosis method based on mainstream artificial intelligence. However, the fault diagnosis method based on the mainstream artificial intelligence has the following technical defects: first, the algorithm module has insufficient detection capability for abnormal signals that do not appear in the training set. Second, the classification capability of the fault classification model is not scientifically evaluated. Third, the time-series characteristics of the current signal are not considered for the feature extraction of the current signal. Therefore, the fault diagnosis is carried out by simply applying artificial intelligence, and the problems that the classification capability is not scientifically evaluated, the classification category is incomplete, the special data is ignored and the like exist, so that the accuracy of fault diagnosis is low.
Disclosure of Invention
The application provides a method and a device for diagnosing equipment faults, electronic equipment and a computer storage medium, aiming at improving the accuracy of fault diagnosis.
In a first aspect, the present application provides an apparatus fault diagnosis method, including:
determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
In one embodiment, the labeling the unidentifiable time-series data set based on an expert experience library to obtain a second pattern library, including:
determining preset sample data with the highest abnormal degree in the abnormal point table;
matching a plurality of fault modes with highest similarity in a fault data mode library according to a time sequence matching algorithm and by combining the preset sample data;
marking the unidentifiable time sequence data set based on the expert experience library by combining a plurality of fault modes to obtain the type mode and the mode number of the unidentifiable abnormal time sequence data set and the type mode and the mode number of the unidentifiable normal time sequence data set;
and determining the second pattern library according to the type pattern and the pattern number of the unidentifiable abnormal time sequence data set and the type pattern and the pattern number of the unidentifiable normal time sequence data set.
The detecting the device data set through the time sequence anomaly detection model to obtain an abnormal time sequence data set and a normal time sequence data set, including:
determining the statistical characteristics of each path of time sequence data in the equipment data set, and determining a statistical characteristic table according to the statistical characteristics of each path of time sequence data;
calculating the abnormal score of each path of time sequence data by combining a statistical learning abnormal detection algorithm with the statistical feature table;
and determining the abnormal time sequence data set and the normal time sequence data set according to the abnormal score of each path of the time sequence data.
Classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set, wherein the method comprises the following steps:
classifying the abnormal time sequence data set through the time sequence classification model to obtain an unidentifiable abnormal time sequence data set and an identifiable abnormal time sequence data set;
classifying the normal time sequence data set through the time sequence classification model to obtain an unidentifiable normal time sequence data set and an identifiable normal time sequence data set;
determining the unrecognized abnormal time series data set and the unrecognized normal time series data set as the unrecognized time series data set;
determining the identifiable anomalous timing data set and the identifiable normal timing data set as the identifiable timing data set.
The determining a first library of patterns based on the identifiable time-series dataset comprises:
determining a first type of pattern and pattern number thereof for the identifiable abnormal timing data set and determining a second type of pattern and pattern number thereof for the identifiable normal timing data set;
determining the first pattern library based on the first type pattern and its pattern number, and the second type pattern and its pattern number.
After determining the failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, the method further comprises:
retraining the time sequence anomaly detection model and the time sequence classification model through the anomaly time sequence data set and the normal time sequence data set at preset intervals to obtain a retrained time sequence anomaly detection model and a retrained time sequence classification model;
verifying the retrained time sequence abnormity detection model and the retrained time sequence classification model through a newly acquired equipment data set to obtain a verification result;
and if the verification result meets the preset verification requirement, determining a fault diagnosis mode library of the newly acquired equipment data set based on the retrained time sequence abnormity detection model and the retrained time sequence classification model.
After determining the failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, the method further comprises:
determining the mode type of each mode in the second mode library, and determining whether a new fault mode exists in the second mode library according to the mode type;
and if the new fault mode exists in the second mode library, updating the marking type and the marking strategy of the expert experience library according to the mode type and the new fault mode.
In a second aspect, the present application provides an apparatus for diagnosing device failure, comprising:
the anomaly detection module is used for determining the acquired equipment data set and detecting the equipment data set through a time sequence anomaly detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
the identification classification module is used for classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
the pattern labeling module is used for determining a first pattern library based on the identifiable time sequence data set and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
a pattern library determining module, configured to determine a failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, so as to perform device failure diagnosis through the failure diagnosis pattern library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
In a third aspect, the present application further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the method for diagnosing the device failure according to the first aspect is implemented.
In a fourth aspect, the present application also provides a non-transitory computer-readable storage medium comprising a computer program which, when executed by the processor, implements the device fault diagnosis method of the first aspect.
In a fifth aspect, the present application further provides a computer program product comprising a computer program, which when executed by the processor, implements the device failure diagnosis method of the first aspect.
According to the equipment fault diagnosis method, the device, the electronic equipment and the computer storage medium, in the equipment fault diagnosis process, artificial intelligence of the time sequence abnormity detection model and the time sequence classification model is taken as a core, the expert experience library is introduced as an intermediate link of the whole equipment fault diagnosis, and auxiliary marking is carried out on the marking process of the abnormal time sequence data set, so that the identification effect of the artificial intelligence on small sample volume special data is increased, the fault diagnosis mode library is accurately determined, and the fault diagnosis accuracy is improved.
Drawings
In order to more clearly illustrate the technical solutions of the present application, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic flow chart diagram of a method for diagnosing equipment failure provided by the present application;
FIG. 2 is a schematic structural diagram of a device failure diagnosis apparatus provided in the present application;
fig. 3 is a schematic structural diagram of an electronic device provided in the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely with reference to the accompanying drawings in the present application, and it is obvious that the described embodiments are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The device fault diagnosis method, apparatus, electronic device, and computer storage medium provided by the present application are described with reference to fig. 1 to 3. FIG. 1 is a schematic flow chart diagram of a method for diagnosing equipment failure provided by the present application; fig. 2 is a schematic structural diagram of the device fault diagnosis apparatus provided in the present application; fig. 3 is a schematic structural diagram of an electronic device provided in the present application.
While a logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order than that shown or described.
The embodiment of the present application takes an electronic device as an execution subject for example, and the embodiment of the present application takes a fault diagnosis system as one of the expression forms of the electronic device, which is not limited.
Referring to fig. 1, fig. 1 is a schematic flow chart of a device fault diagnosis method provided in the present application. The equipment fault diagnosis method provided by the embodiment of the application comprises the following steps:
and S10, determining the acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set.
It should be noted that the time sequence anomaly detection model and the time sequence classification model in the embodiment of the present application both belong to artificial intelligence, and a main application scenario of the embodiment of the present application is intelligent operation and maintenance oriented to smart city rail transit construction, and is applicable to automated operation and maintenance of all power supply equipment, and in order to clearly express an implementation process, a specific device takes a switch device as an example; the device data includes, but is not limited to, current timing data and voltage timing data, and for clarity of description, specific device data is exemplified by the current timing data. Further, the time sequence anomaly detection model and the time sequence classification model in the embodiment of the application are trained off line and are obtained by training based on sample data labeled by an expert experience base.
It can therefore be understood that the fault diagnosis system needs to collect equipment data of the switch machine equipment to obtain a current time sequence data set of the switch machine equipment. Further, the fault diagnosis system inputs the current time sequence data set into a time sequence abnormality detection model which is trained offline for abnormality detection, judges whether abnormal current time sequence data exist in the current time sequence data set acquired by the switch machine equipment, and determines abnormal current time sequence data in the current time sequence data set so as to obtain an abnormal time sequence data set and a normal time sequence data set, specifically, as described in step S101 to step S103.
Further, the description of steps S101 to S103 is as follows:
step S101, determining the statistical characteristics of each path of time sequence data in the equipment data set, and determining a statistical characteristic table according to the statistical characteristics of each path of time sequence data;
step S102, calculating the abnormal score of each path of time sequence data by combining a statistical learning abnormal detection algorithm with the statistical feature table;
step S103, determining the abnormal time sequence data set and the normal time sequence data set according to the abnormal score of each path of the time sequence data.
It should be noted that the purpose of the time series abnormality detection model is to calculate the degree of abnormality of each current curve in the offline database. The time sequence abnormity detection algorithm in the time sequence abnormity detection model mainly comprises two types, wherein the first type is an unsupervised abnormity time sequence detection algorithm based on statistical characteristics, the second type is an abnormity detection algorithm based on supervised deep learning, and in order to clearly express the implementation process, the unsupervised abnormity time sequence detection algorithm based on statistical characteristics is taken as an example for the time sequence abnormity detection algorithm in the time sequence abnormity detection model.
Specifically, the fault diagnosis system extracts statistical characteristics of each path of current time sequence data in the current time sequence data set through a time sequence abnormity detection model, wherein the statistical characteristics include, but are not limited to, mean statistical characteristics, standard deviation statistical characteristics, entropy characteristic statistical characteristics and period same-cycle ratio statistical characteristics. Further, the fault diagnosis system forms a statistical characteristic table according to the mean statistical characteristic, the standard deviation statistical characteristic, the entropy statistical characteristic and the period same-cycle ratio statistical characteristic of each path of current time sequence data. Further, the fault diagnosis system calculates the abnormal score of each path of current time sequence data by combining a statistical learning abnormal detection algorithm with a statistical feature table, wherein the statistical learning abnormal detection algorithm includes, but is not limited to, an isolated Forest (Isolation Forest) and a Local Outlier Factor (Local Outlier Factor).
Further, the fault diagnosis system compares the abnormal score of each path of current time sequence data with a preset critical score, wherein the preset critical score is set according to actual conditions. Further, the fault diagnosis system determines all current time series data, of which the abnormal score is greater than or equal to a preset critical score, as abnormal current time series data, thereby obtaining an abnormal time series data set (abnormal current time series data set). Meanwhile, the fault diagnosis system determines all current time sequence data with the abnormal score smaller than a preset critical score as normal current time sequence data, so that a normal time sequence data set (a normal current time sequence data set) is obtained.
According to the method and the device, the abnormal score of each path of current time sequence data is calculated, and then each path of current time sequence data is detected according to the abnormal score, so that the abnormal time sequence data set and the normal time sequence data set are accurately determined, and a data basis is provided for improving the accuracy of fault diagnosis.
And S20, classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set.
Further, after the fault diagnosis system obtains an abnormal current time sequence data set and a normal current time sequence data set through an offline trained time sequence abnormality detection model, the abnormal current time sequence data set and the normal current time sequence data set are input into the offline trained time sequence classification model, the abnormal current time sequence data set and the normal current time sequence data set are classified through the time sequence classification model, and the belonged categories of the abnormal current time sequence data set and the normal current time sequence data set are determined, wherein the belonged categories comprise an unidentifiable category and an identifiable category, the identifiable category is a category which can be directly identified, the unidentifiable category is a category which cannot be directly identified, and auxiliary identification needs to be carried out through an auxiliary tool.
Thus, the fault diagnosis system classifies all identifiable current timing data into one class, resulting in an identifiable current timing data set. Meanwhile, the fault diagnosis system classifies all unidentifiable current time sequence data into one type to obtain an unidentifiable current time sequence data set, which is specifically described in step S201 to step S204.
Further, the description of step S201 to step S204 is as follows:
step S201, classifying the abnormal time sequence data set through the time sequence classification model to obtain an unidentifiable abnormal time sequence data set and an identifiable abnormal time sequence data set;
step S202, classifying the normal time sequence data set through the time sequence classification model to obtain an unidentifiable normal time sequence data set and an identifiable normal time sequence data set;
step S203, determining the unrecognizable abnormal time sequence data set and the unrecognizable normal time sequence data set as the unrecognizable time sequence data set;
step S204, the identifiable abnormal time sequence data set and the identifiable normal time sequence data set are determined as the identifiable time sequence data set.
Specifically, the fault diagnosis system classifies all abnormal current time sequence data in the abnormal current time sequence data set through the time sequence classification model, and distinguishes unidentifiable abnormal current time sequence data and identifiable abnormal current time sequence data in the abnormal current time sequence data set. Further, the fault diagnosis system classifies all the identifiable abnormal current time sequence data to obtain an identifiable abnormal current time sequence data set. Meanwhile, the fault diagnosis system classifies all unidentifiable abnormal current time sequence data to obtain an unidentifiable abnormal current time sequence data set.
Further, the fault diagnosis system classifies all normal current time sequence data in the normal current time sequence data set through the time sequence classification model, and distinguishes unidentifiable normal current time sequence data and identifiable normal current time sequence data in the normal current time sequence data set. And classifying all the identifiable normal current time sequence data by the fault diagnosis system to obtain an identifiable normal current time sequence data set. Meanwhile, the fault diagnosis system classifies all unidentifiable normal current time sequence data to obtain an unidentifiable normal current time sequence data set.
Further, the fault diagnosis system determines the unidentifiable abnormal current time sequence data set and the unidentifiable normal current time sequence data set as unidentifiable current time sequence data sets. Meanwhile, the fault diagnosis system determines the identifiable abnormal current time sequence data set and the identifiable normal current time sequence data set as identifiable current time sequence data sets.
According to the method and the device, the unrecognizable time sequence data set and the recognizable time sequence data set are accurately classified through the time sequence classification model, and a data basis is provided for improving the accuracy of fault diagnosis.
It should be further noted that the fault diagnosis system further includes an integrated pattern mining algorithm module, and the integrated pattern mining algorithm module mainly solves the pattern mining problem of the abnormal sample. First, the explanation of the pattern in the embodiment of the present application refers to the process of generating a regular change and self-repeating pattern on an object or event. The pattern in the current timing refers to a series of current signals having the same characteristics, and a series of pattern information exists regardless of the normal current timing or the abnormal current timing.
According to the embodiment of the application, the mode information in the normal sample and the abnormal sample can be effectively extracted by using an integrated clustering and redundant clustering merging strategy. The current time sequence mode acquired by the integrated mode mining algorithm module can be used as supplement of mode information of a fault mode library, so that the system faces the fault type which cannot be predicted in an experiment. The number of the patterns obtained by the integrated pattern mining algorithm module can be used as a classification basis of the classification algorithm. Therefore, the integrated pattern mining algorithm module and the online application pattern algorithm module jointly form a time sequence abnormity detection model and a time sequence classification model.
Further, on the basis of the previous labeling result of the online application mode algorithm module and the result of the integrated mode mining algorithm module, the online application algorithm model is trained. In practical applications, the line application algorithm model includes: a time sequence abnormity detection model and a time sequence classification model.
For the time sequence anomaly detection model, the algorithms adopted by the time sequence anomaly detection model are an XGBOD anomaly detection algorithm, an anomaly detection algorithm based on deep learning and an unsupervised anomaly detection algorithm based on Isolation Forest, and the integrated algorithms have stable and robust effect and can effectively detect the time sequence of the anomaly.
For the time-series classification model, the purpose of the time-series classification model is to correctly classify the time-series current signals into their corresponding failure modes. The time sequence classification model is a template matching algorithm based on distance measurement, a time sequence classification algorithm based on LSTM and GRU and a time sequence classification algorithm based on time sequence statistical characteristics and LightGBM.
Step S30, determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
and S40, determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library.
Further, for the identifiable current timing data set, the fault diagnosis system determines patterns and numbers thereof of all identifiable current timing data in the identifiable current timing data set, and determines a first pattern library according to the patterns and numbers thereof of all identifiable current timing data, as described in step S201 to step S202.
Further, for the unidentifiable current time sequence data set, the fault diagnosis system needs to push the unidentifiable current time sequence data set to the Web front end, and the unidentifiable current time sequence data set is labeled through an expert experience library in the Web front end to determine the modes and the numbers of all unidentifiable current time sequence data in the unidentifiable current time sequence data set.
Further, the fault diagnosis system determines a second pattern library according to the patterns and the numbers of all the unidentifiable current time series data, as described in step S203 to step S206. Wherein, the expert experience library is a summary experience library of technicians in the concrete engineering implementation process.
Further, the expert experience base can also understand that for the unidentifiable current time sequence data set, the fault diagnosis system needs to push the unidentifiable current time sequence data set to the Web front end, and a field technician performs field marking on the pushed unidentifiable current time sequence data set at the Web front end to determine the modes and the numbers of all unidentifiable current time sequence data in the unidentifiable current time sequence data set. Further, the fault diagnosis system determines a second pattern library according to the patterns of all the unidentifiable current time sequence data and the numbers thereof.
Further, the fault diagnosis system determines a mode library formed by the automatically identified first mode library and a second mode library marked by the expert experience library as a fault diagnosis mode library of the current time sequence data set, and performs equipment fault diagnosis on the switch machine equipment through the fault diagnosis mode library.
According to the equipment fault diagnosis method provided by the embodiment of the application, in the equipment fault diagnosis process, the artificial intelligence of the time sequence abnormity detection model and the time sequence classification model is taken as a core, the expert experience base is introduced to serve as an intermediate link of the whole equipment fault diagnosis, the auxiliary marking is carried out on the marking process of the abnormal time sequence data set, the identification effect of the artificial intelligence on the special data with small sample size is increased, the fault diagnosis mode base is accurately determined, and the fault diagnosis accuracy is improved.
Further, the description of steps S301 to S306 is as follows:
step S301, determining a first type mode and a mode number thereof of the identifiable abnormal time sequence data set, and determining a second type mode and a mode number thereof of the identifiable normal time sequence data set;
step S302, determining the first mode library based on the first type mode and the mode number thereof, and the second type mode and the mode number thereof;
step S303, determining preset sample data with the highest abnormal degree in the abnormal score table;
step S304, matching a plurality of fault modes with highest similarity in a fault data mode base according to a time sequence matching algorithm and in combination with the preset sample data;
step S305, labeling the unidentifiable time sequence data set based on an expert experience base and in combination with a plurality of fault modes to obtain the type mode and the mode number of the unidentifiable abnormal time sequence data set and the type mode and the mode number of the unidentifiable normal time sequence data set;
step S306, determining the second pattern library according to the type pattern and the pattern number of the unrecognizable abnormal time sequence data set and the type pattern and the pattern number of the unrecognizable normal time sequence data set.
Specifically, for an identifiable current timing dataset, the fault diagnosis system determines a first type of pattern and pattern number thereof in the identifiable current timing dataset that identifies an abnormal current timing dataset, and determines a second type of pattern and pattern number thereof that identifies a normal current timing dataset. Further, the fault diagnosis system determines a first pattern library based on all the first type patterns and the pattern numbers thereof, and all the second type patterns and the pattern numbers thereof.
In one embodiment, the identifiable abnormal current timing sequence data set comprises identifiable abnormal current timing sequence data 1 and identifiable abnormal current timing sequence data 2, the type mode of the identifiable abnormal current timing sequence data 1 is an abnormal mode, and the mode number is 1; the type mode of the abnormal current time sequence data 2 can be identified as an abnormal mode, and the mode number is 2; the recognizable normal current time sequence data set comprises recognizable normal current time sequence data 1 and recognizable normal current time sequence data 2, the type mode of the recognizable normal current time sequence data 1 is a normal mode, and the mode number is 1; the type mode of the normal current timing data 2 can be identified as a normal mode, and the mode number is 2. Thus, the first pattern library is { abnormal pattern 1; an abnormal mode 2; a normal mode 1; normal mode 2}.
It should be further noted that the expert experience base labeling is performed based on a front-end labeling framework for active learning, and the front-end labeling framework for active learning is a framework of a pool-based active learning method and a Top-K template search algorithm set, so as to solve the problem that a huge time sequence is difficult to label one by one. The main objectives for the front-end markup framework for active learning are also: firstly, sample data injected by an actively-learned front-end labeling framework can train a time sequence anomaly detection model and a time sequence classification model; secondly, the sample data injected by the front-end marking framework after active learning can also be used for scientifically evaluating the effectiveness of a subsequent series of algorithms and frameworks.
Therefore, the fault diagnosis system firstly determines the preset strip sample data with the highest abnormal degree in the abnormal score table through a correlation algorithm, wherein the number of the preset strips is set according to the actual situation.
Further, the fault diagnosis system matches a plurality of Top-k fault modes with the highest similarity in a fault data mode base according to a time sequence matching algorithm and preset sample data. Further, the fault diagnosis system pushes a plurality of fault modes and the unrecognizable current time sequence data set to a Web front end, the unrecognizable current time sequence data set is marked through an expert experience library in the Web front end, the type mode and the mode number of the unrecognizable abnormal current time sequence data set are determined, and the type mode and the mode number of the unrecognizable normal current time sequence data set are determined, wherein the Web front end is the front end of JavaScript, HTML and CSS. Further, the fault diagnosis system determines a second pattern library according to the type pattern and the pattern number of the unrecognizable abnormal current time sequence data set and the type pattern and the pattern number of the unrecognizable normal current time sequence data set, and specifically refers to the step of the first pattern library.
Further, the expert experience base can also understand that the fault diagnosis system pushes a plurality of fault modes and unidentifiable current time sequence data sets to the Web front end, and field technicians mark the pushed unidentifiable current time sequence data sets on the Web front end on site to determine the type mode and the mode number of the unidentifiable abnormal current time sequence data set and the type mode and the mode number of the unidentifiable normal current time sequence data set. Further, the fault diagnosis system determines a second pattern library according to the type pattern and the pattern number of the unrecognizable abnormal current time sequence data set and the type pattern and the pattern number of the unrecognizable normal current time sequence data set, and specifically refers to the step of the first pattern library.
According to the method and the device, the artificial intelligence of the time sequence abnormity detection model and the time sequence classification model is taken as a core, the expert experience base is introduced to serve as an intermediate link of the whole equipment fault diagnosis, and the labeling process of the abnormal time sequence data set is labeled in an auxiliary mode, so that the identification effect of the artificial intelligence on the small sample size special data is increased, the fault diagnosis mode base is determined accurately, and the fault diagnosis accuracy is improved.
Further, after determining the second pattern library, it is necessary to determine whether a new pattern is generated in the second pattern library, specifically referring to step S50 to step S60.
Step S50, determining the mode type of each mode in the second mode library, and determining whether a new fault mode exists in the second mode library according to the mode type;
and S60, if the new fault mode exists in the second mode library, updating the labeling type and the labeling strategy of the expert experience library according to the mode type and the new fault mode.
Further, the fault diagnosis system determines the mode type of each mode in the second mode library, and determines whether a new fault mode exists in the second mode library according to the mode type. If it is determined that a new fault mode exists in the second mode base, the fault diagnosis system updates the annotation type and the annotation strategy of the expert experience base according to the mode type of the new fault mode and the new fault mode, wherein the annotation strategy can be understood as an annotation tool, namely the fault diagnosis system updates the annotation type and the annotation tool of the expert experience base.
That is, it can be understood that, if it is determined that a new fault mode exists in the second mode library, the operation and maintenance technician creates a new annotation type and a new annotation tool, and annotates the new fault mode through the new annotation type and the new annotation tool.
According to the embodiment of the application, the mode types of all the modes in the second mode library are detected in real time, and the marking types and the marking strategies of the expert experience library are updated in real time according to the mode types, so that the mode types and the mode numbers of all the unidentifiable abnormal current time sequence data are accurately marked, an accurate fault diagnosis mode library is obtained, and the accuracy of fault diagnosis is improved.
Further, after the fault diagnosis pattern library is generated, the time-series abnormality detection model and the time-series classification model need to be updated at intervals, specifically referring to step S70 to step S90.
Step S70, retraining the time sequence abnormal detection model and the time sequence classification model through the abnormal time sequence data set and the normal time sequence data set at preset time intervals to obtain a retrained time sequence abnormal detection model and a retrained time sequence classification model;
step S80, verifying the retrained time sequence abnormity detection model and the retrained time sequence classification model through a newly acquired equipment data set to obtain a verification result;
and S90, if the verification result meets the preset verification requirement, determining a fault diagnosis mode library of the newly acquired equipment data set based on the retrained time sequence abnormity detection model and the retrained time sequence classification model.
It should be noted that, as a function of time, on the one hand, wear of the components of the switch machine equipment inevitably occurs; on the other hand, the field overhaul of the operation and maintenance personnel will also cause the current signal curve generated by the switch machine equipment to change. Therefore, due to the wear of the mechanical structure or the phenomenon that the current time sequence acquired by the sensor is different from the current time sequence accumulated by the previous system, the failure of the time sequence abnormality detection model and the time sequence classification model for the current signal is finally caused. Such a phenomenon is called a Concept Drift phenomenon (Concept Drift) in the field of intelligent operation and maintenance. The learning algorithm cannot be applied to new data through concepts learned by historical data, so that the model is invalid.
Further, to the problem of concept drift: and retraining the time sequence anomaly detection model and the time sequence classification model through the historical abnormal current time sequence data set and the historical normal current time sequence data set at intervals, so as to obtain the retrained time sequence anomaly detection model and the retrained time sequence classification model, wherein the preset time length is determined by technicians.
Further, the fault diagnosis system collects current time sequence data of new switch machine equipment, and verifies the retrained time sequence abnormity detection model and the retrained time sequence classification model through the current time sequence data of the new switch machine equipment to obtain a verification result, wherein the verification result can be in line with a preset verification requirement or not in line with the preset verification requirement.
If the verification result is determined to meet the preset verification requirement, the fault diagnosis system determines a new fault diagnosis mode library of the current time sequence data of the new switch machine equipment based on the retrained time sequence abnormality detection model and the retrained time sequence classification model, which is described with reference to steps S10 to S40 specifically, and is not repeated here, so that equipment fault diagnosis is performed on the current time sequence data of the new switch machine equipment through the new fault diagnosis mode library.
If the verification result is determined to be not in accordance with the preset verification requirement, the fault diagnosis system circulates the historical abnormal current time sequence data set and the historical normal current time sequence data set, and trains the time sequence abnormal detection model and the time sequence classification model until the obtained verification result is in accordance with the preset verification requirement.
According to the method and the device, the time sequence abnormity detection model and the time sequence classification model are updated through the historical abnormal current time sequence data set and the historical normal current time sequence data set, and the accuracy of the time sequence abnormity detection model and the time sequence classification model is guaranteed, so that the accuracy of a fault diagnosis mode library is guaranteed, and the accuracy of fault diagnosis is improved.
Further, the device fault diagnosis apparatus provided by the present application is described below, and the device fault diagnosis apparatus and the device fault diagnosis method may be referred to in correspondence with each other.
As shown in fig. 2, fig. 2 is a schematic structural diagram of an apparatus fault diagnosis device provided in the present application, and the apparatus fault diagnosis device includes:
an anomaly detection module 201, configured to determine an acquired device data set, and detect the device data set through a time sequence anomaly detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
the identification classification module 202 is configured to classify the abnormal time series data set and the normal time series data set through a time series classification model to obtain an unidentifiable time series data set and an identifiable time series data set;
the pattern labeling module 203 is used for determining a first pattern library based on the identifiable time sequence data set and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
a pattern library determining module 204, configured to determine a fault diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, so as to perform device fault diagnosis through the fault diagnosis pattern library.
Further, the anomaly detection module 201 is further configured to:
determining the statistical characteristics of each path of time sequence data in the equipment data set, and determining a statistical characteristic table according to the statistical characteristics of each path of time sequence data;
calculating the abnormal score of each path of time sequence data by combining a statistical learning abnormal detection algorithm with the statistical feature table;
and determining the abnormal time sequence data set and the normal time sequence data set according to the abnormal score of each path of the time sequence data.
Further, the recognition classification module 202 is further configured to:
classifying the abnormal time sequence data set through the time sequence classification model to obtain an unrecognizable abnormal time sequence data set and an identifiable abnormal time sequence data set;
classifying the normal time sequence data set through the time sequence classification model to obtain an unidentifiable normal time sequence data set and an identifiable normal time sequence data set;
determining the unrecognized abnormal time series data set and the unrecognized normal time series data set as the unrecognized time series data set;
determining the identifiable anomalous timing data set and the identifiable normal timing data set as the identifiable timing data set.
Further, the mode labeling module 203 is further configured to:
determining preset sample data with the highest abnormal degree in the abnormal point table;
matching a plurality of fault modes with highest similarity in a fault data mode library according to a time sequence matching algorithm and by combining the preset sample data;
marking the unidentifiable time sequence data set based on the expert experience library by combining a plurality of fault modes to obtain the type mode and the mode number of the unidentifiable abnormal time sequence data set and the type mode and the mode number of the unidentifiable normal time sequence data set;
and determining the second pattern library according to the type pattern and the pattern number of the unidentifiable abnormal time sequence data set and the type pattern and the pattern number of the unidentifiable normal time sequence data set.
Further, the mode labeling module 203 is further configured to:
determining a first type of pattern and pattern number thereof for the identifiable abnormal timing data set and determining a second type of pattern and pattern number thereof for the identifiable normal timing data set;
determining the first pattern library based on the first type pattern and its pattern number, and the second type pattern and its pattern number.
Further, the device failure diagnosis apparatus further includes: an update module to:
retraining the time sequence anomaly detection model and the time sequence classification model through the anomaly time sequence data set and the normal time sequence data set at preset intervals to obtain a retrained time sequence anomaly detection model and a retrained time sequence classification model;
verifying the retrained time sequence abnormity detection model and the retrained time sequence classification model through a newly acquired equipment data set to obtain a verification result;
and if the verification result meets the preset verification requirement, determining a fault diagnosis mode library of the newly acquired equipment data set based on the retrained time sequence abnormity detection model and the retrained time sequence classification model.
Further, the update module is further configured to:
determining the mode type of each mode in the second mode library, and determining whether a new fault mode exists in the second mode library according to the mode type;
and if the new fault mode exists in the second mode library, updating the marking type and the marking strategy of the expert experience library according to the mode type and the new fault mode.
The specific embodiment of the device fault diagnosis apparatus provided in the present application is substantially the same as the embodiments of the device fault diagnosis method described above, and details are not described herein.
Fig. 3 illustrates a physical structure diagram of an electronic device, and as shown in fig. 3, the electronic device may include: a processor (processor) 310, a communication Interface (communication Interface) 320, a memory (memory) 330 and a communication bus 340, wherein the processor 310, the communication Interface 320 and the memory 330 communicate with each other via the communication bus 340. The processor 310 may invoke logic instructions in the memory 330 to perform a device fault diagnosis method comprising:
determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
In addition, the logic instructions in the memory 330 may be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
In another aspect, the present application also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, which when executed by a computer, enable the computer to perform the method for diagnosing the equipment failure provided by the above methods, the method comprising:
determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
In yet another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program that, when executed by a processor, is implemented to perform the method for diagnosing equipment failure provided above, the method comprising:
determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. An apparatus fault diagnosis method, comprising:
determining an acquired equipment data set, and detecting the equipment data set through a time sequence abnormity detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
determining a first pattern library based on the identifiable time sequence data set, and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
determining a fault diagnosis mode library of the equipment data set based on the first mode library and the second mode library so as to carry out equipment fault diagnosis through the fault diagnosis mode library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data labeled by the expert experience base.
2. The equipment fault diagnosis method according to claim 1, wherein the labeling the unidentifiable time-series data set based on an expert experience library to obtain a second pattern library comprises:
determining preset sample data with the highest abnormal degree in the abnormal point table;
matching a plurality of fault modes with highest similarity in a fault data mode library according to a time sequence matching algorithm and by combining the preset sample data;
marking the unidentifiable time sequence data set based on the expert experience library by combining a plurality of fault modes to obtain the type mode and the mode number of the unidentifiable abnormal time sequence data set and the type mode and the mode number of the unidentifiable normal time sequence data set;
and determining the second pattern library according to the type pattern and the pattern number of the unidentifiable abnormal time sequence data set and the type pattern and the pattern number of the unidentifiable normal time sequence data set.
3. The method according to claim 1, wherein the detecting the device data set by the time series anomaly detection model to obtain an anomaly time series data set and a normal time series data set comprises:
determining the statistical characteristics of each path of time sequence data in the equipment data set, and determining a statistical characteristic table according to the statistical characteristics of each path of time sequence data;
calculating the abnormal score of each path of time sequence data by combining a statistical learning abnormal detection algorithm with the statistical feature table;
and determining the abnormal time sequence data set and the normal time sequence data set according to the abnormal score of each path of the time sequence data.
4. The method of claim 1, wherein the classifying the abnormal time-series data set and the normal time-series data set by a time-series classification model to obtain an unrecognized time-series data set and a recognizable time-series data set comprises:
classifying the abnormal time sequence data set through the time sequence classification model to obtain an unidentifiable abnormal time sequence data set and an identifiable abnormal time sequence data set;
classifying the normal time sequence data set through the time sequence classification model to obtain an unidentifiable normal time sequence data set and an identifiable normal time sequence data set;
determining the unrecognizable abnormal time series data set and the unrecognizable normal time series data set as the unrecognizable time series data set;
determining the identifiable anomalous timing data set and the identifiable normal timing data set as the identifiable timing data set.
5. The device fault diagnostic method of claim 4, wherein said determining a first pattern library based on the identifiable time-series data set comprises:
determining a first type of pattern and pattern number thereof for the identifiable abnormal temporal data set, and determining a second type of pattern and pattern number thereof for the identifiable normal temporal data set;
determining the first pattern library based on the first type pattern and its pattern number, and the second type pattern and its pattern number.
6. The device failure diagnosis method according to claim 1, wherein after determining the failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, further comprising:
retraining the time sequence anomaly detection model and the time sequence classification model through the anomaly time sequence data set and the normal time sequence data set at preset intervals to obtain a retrained time sequence anomaly detection model and a retrained time sequence classification model;
verifying the retrained time sequence abnormity detection model and the retrained time sequence classification model through a newly acquired equipment data set to obtain a verification result;
and if the verification result meets the preset verification requirement, determining a fault diagnosis mode library of the newly acquired equipment data set based on the retrained time sequence abnormity detection model and the retrained time sequence classification model.
7. The method according to any one of claims 1 to 6, wherein after determining the failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, the method further comprises:
determining the mode type of each mode in the second mode library, and determining whether a new fault mode exists in the second mode library according to the mode type;
and if the new fault mode exists in the second mode library, updating the marking type and the marking strategy of the expert experience library according to the mode type and the new fault mode.
8. An apparatus fault diagnosis device, characterized by comprising:
the anomaly detection module is used for determining the acquired equipment data set and detecting the equipment data set through a time sequence anomaly detection model to obtain an abnormal time sequence data set and a normal time sequence data set;
the identification classification module is used for classifying the abnormal time sequence data set and the normal time sequence data set through a time sequence classification model to obtain an unidentifiable time sequence data set and an identifiable time sequence data set;
the pattern labeling module is used for determining a first pattern library based on the identifiable time sequence data set and labeling the unidentifiable time sequence data set based on an expert experience library to obtain a second pattern library;
a pattern library determining module, configured to determine a failure diagnosis pattern library of the device data set based on the first pattern library and the second pattern library, so as to perform device failure diagnosis through the failure diagnosis pattern library;
and the time sequence abnormity detection model and the time sequence classification model are obtained by training based on sample data marked by the expert experience base.
9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the device failure diagnosis method according to any one of claims 1 to 7 when executing the computer program.
10. A non-transitory computer-readable storage medium including a computer program, wherein the computer program, when executed by a processor, implements the device fault diagnosis method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210886381.2A CN115329848A (en) | 2022-07-26 | 2022-07-26 | Equipment fault diagnosis method and device, electronic equipment and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210886381.2A CN115329848A (en) | 2022-07-26 | 2022-07-26 | Equipment fault diagnosis method and device, electronic equipment and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115329848A true CN115329848A (en) | 2022-11-11 |
Family
ID=83920436
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210886381.2A Pending CN115329848A (en) | 2022-07-26 | 2022-07-26 | Equipment fault diagnosis method and device, electronic equipment and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115329848A (en) |
-
2022
- 2022-07-26 CN CN202210886381.2A patent/CN115329848A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111210024B (en) | Model training method, device, computer equipment and storage medium | |
| CN113282461B (en) | Alarm identification method and device for transmission network | |
| EP2905665B1 (en) | Information processing apparatus, diagnosis method, and program | |
| Ross et al. | Nonparametric monitoring of data streams for changes in location and scale | |
| CN111435366A (en) | Equipment fault diagnosis method and device and electronic equipment | |
| US11971959B2 (en) | Data test method, electronic device and storage medium | |
| CN106570513A (en) | Fault diagnosis method and apparatus for big data network system | |
| CN112036505A (en) | Method and device for determining equipment state of turnout switch machine and electronic equipment | |
| US20160369777A1 (en) | System and method for detecting anomaly conditions of sensor attached devices | |
| CN111459700A (en) | Method and apparatus for diagnosing device failure, diagnostic device, and storage medium | |
| JP2019185422A (en) | Failure prediction method, failure prediction device, and failure prediction program | |
| US20180164794A1 (en) | Methods and Systems for Discovery of Prognostic Subsequences in Time Series | |
| CN113269042B (en) | Intelligent traffic management method and system based on driving vehicle violation identification | |
| CN113590451B (en) | Root cause positioning method, operation and maintenance server and storage medium | |
| CN112905371B (en) | Software change checking method and device based on heterogeneous multi-source data anomaly detection | |
| CN115269314A (en) | Transaction abnormity detection method based on log | |
| CN112529109A (en) | Unsupervised multi-model-based anomaly detection method and system | |
| CN109308411B (en) | Method and system for hierarchically detecting software behavior defects based on artificial intelligence decision tree | |
| CN114118295A (en) | Anomaly detection model training method, anomaly detection device and medium | |
| CN111898661A (en) | Method and device for monitoring working state of turnout switch machine | |
| CN109242131A (en) | Track switch information processing method and device | |
| CN110728768B (en) | Switch operation and maintenance method and system | |
| CN113670611A (en) | Bearing early degradation evaluation method, system, medium and electronic equipment | |
| CN115329848A (en) | Equipment fault diagnosis method and device, electronic equipment and computer storage medium | |
| CN113722239B (en) | Airborne embedded software quality detection method, device, medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |