Anomaly detection algorithm configuration method and device, electronic equipment and storage medium

A technology of abnormal detection and configuration method, applied in the field of automobile production lines, can solve the problems of multiple manpower and material resources, consumption, high cost of automobile production lines, etc., and achieve the effect of reducing manpower and material resources and reducing costs.

Pending Publication Date: 2021-11-26
SIEMENS FACTORY AUTOMATION ENG
5 Cites 0 Cited by

AI-Extracted Technical Summary

Problems solved by technology

Since the automobile production line includes various types of equipment, different types of equipment require different anomaly detection algorithms, and the same type of equipment with different operating conditions also requires different anomaly detection algorithms. Therefore, when realizing predictive maintenance of automobile production lines, users It i...
View more

Method used

In the embodiment of the present application, according to the abnormality detection result output by the abnormality detection algorithm, the calculation abnormality detection algorithm detects the first abnormality frequency that the equipment appears abnormal in the current detection cycle, and calculates the abnormality detection algorithm in the historical detection cycle When the second abnormal frequency of equipment abnormality is detected, when the difference between the first abnormal frequency and the second abnormal frequency is less than the preset frequency difference threshold, it is determined that the frequency of abnormality detected by the abnormality detection algorithm is stable; otherwise, the abnormality detection is determined The algorithm detects that the abnormal frequency of the equipment is unstable, and the abnormal frequency of the equipment is detected by the abnormal detection algorithm in different detection cycles, so as to determine whether the abnormal frequency of the equipment detected by the abnormal detection algorithm is stable, so as to ensure the accuracy of the detection results. In turn, it can ensure the accuracy of anomaly detection algorithms and improve the accuracy of predictive maintenance on automotive production lines.
In the embodiment of the present application, by presetting the step size value of adjusting the parameters of the abnormality detection algorithm, after determining the trend of adjusting the parameters of the abnormality detection algorithm, adding the step size value to the parameters of the abnormality detection algorithm Or reduce the step value to gradually adjust the parameters of the anomaly detection algorithm, and gradually improve the accuracy of the anomaly detection algorithm to detect equipment anomalies, so as to realize automatic configuration of the anomaly detection algorithm and reduce the manpower for deploying predictive maintenance of automobile production lines cost.
In the embodiment of the present application, the variable information is determined according to the operating data of the equipment, after the variable information is input as the abnormal detection algorithm configured by the equipment, the abnormal detection result output by the abnormal detection algorithm is obtained, and the abnormal detection algorithm is judged according to the abnormal detection result Detect whether the abnormal frequency of the equipment is stable, and adjust the parameters of the abnormal detection algorithm if it is unstable. During the normal operation of the equipment, if the anomaly detection algorithm can accurately detect whether the equipment is abnormal, the frequency of equipment abnormalities detected by the anomaly detection algorithm should be stable, so it can be detected whether the frequency of equipment anomalies is stable according to the anomaly detection algorithm , to determine whether the anomaly detection algorithm can accurately detect whether the device is abnormal, and then when it is determined that the anomaly detection algorithm cannot accurately detect whether the device is abnormal, automatically adjust the parameters of the anomaly detection algorithm until the anomaly detection algorithm can accurately detect whether the device is abnormal . It can be seen that the user only needs to configure the corresponding type of anomaly detection algorithm for the equipment in the automobile production line, and the parameters of the anomaly detection algorithm can be automatically configured according to the actual operation of the equipment, so that the configuration of anomaly detection for the equipment in the automobile production line can be reduced The manpower and material resources consumed by the algorithm can reduce the cost of predictive maintenance of the automobile production line.
Through the electronic equipment of this embodiment, the variable information is determined according to the operating data of the equipment, after the variable information is input as the abnormal detection algorithm of the equipment configuration, the abnormal detection result output by the abnormal detection algorithm is obtained, and the abnormal detection is judged according to the abnormal detection result The algorithm detects whether the abnormal frequency of the equipment is stable, and if it is unstable, adjust the parameters of the abnormal detection algorithm. During the normal operation of the equipment, if...
View more

Abstract

The invention provides a configuration method and device of an anomaly detection algorithm, electronic equipment and a storage medium, the configuration method of the anomaly detection algorithm is used for configuring the anomaly detection algorithm of equipment in predictive maintenance of an automobile production line, and the configuration method of the anomaly detection algorithm comprises the following steps: obtaining operation data of the equipment; determining variable information of the equipment according to the operation data, wherein the variable information is directly obtained from the operation data or obtained by processing the operation data; inputting the variable information into the anomaly detection algorithm to obtain an anomaly detection result output by the anomaly detection algorithm; according to the anomaly detection result, judging whether the anomaly frequency of the equipment detected by the anomaly detection algorithm is stable or not; and if the abnormal frequency of the equipment detected by the abnormal detection algorithm is unstable, adjusting parameters of the abnormal detection algorithm. According to the scheme, the predictive maintenance cost of the automobile production line can be reduced.

Application Domain

Character and pattern recognitionTechnology management +2

Technology Topic

EngineeringAlgorithm +4

Image

  • Anomaly detection algorithm configuration method and device, electronic equipment and storage medium
  • Anomaly detection algorithm configuration method and device, electronic equipment and storage medium
  • Anomaly detection algorithm configuration method and device, electronic equipment and storage medium

Examples

  • Experimental program(13)

Example Embodiment

[0129] Example one
[0130] figure 1 The present application is a flowchart of an abnormality detection method for configuring an algorithm according to a first embodiment 100, such as figure 1 , The configuration of the abnormality detection algorithm 100 comprises steps of:
[0131] 101, operating data acquisition equipment.
[0132] Automobile production line comprising a plurality of the need for predictive maintenance equipment, device configuration corresponding predictive maintenance anomaly detection algorithms for each desired parameter error detection algorithm but may not apply to the actual operating conditions of the apparatus, is this anomaly detection algorithm parameters need to be adjusted. Automobile production lines require predictive maintenance device may be individual components, such as bearings, may be a device composed of a plurality of parts, such as high-speed lift.
[0133] For automobile production line needs to be predictive maintenance of equipment, operating data acquisition process of the operation of the device, the acquired operating data can reflect the operating status of the device, may be different for different types of operational data of the required equipment. The acquired operating data of a data device may be a vibration, audio data, speed data, temperature data and the like.
[0134] 102, operating data determining apparatus according to variable information, wherein said variable information directly from the operating data or by the data processing operation is obtained.
[0135] Variable information is input information abnormality detection algorithm, after the variable information input to the abnormality detection method, the abnormality detection algorithm can be based on the variable information to determine whether the device is abnormal. Variable information can be acquired from the operation data of the device directly, or by operating data of the device obtained by processing, such as variable information temperature sound frequency device producing sound amplitude, a device is the vibration frequency generated by the device speed device or the like.
[0136] 103, the variable information input to the abnormality detection algorithm, obtaining abnormality detection result output from the abnormality detection algorithm.
[0137] Anomaly detection algorithms for variable information detecting apparatus according to whether there is an abnormality, after the variable information input to the abnormality detection method, the abnormality detection algorithm outputs an abnormality detection result, the abnormality detection result for the presence of abnormal or absence of an abnormality indication device apparatus. For example, the abnormality detection result is output from the abnormality detection algorithm exception code, an error code of a characterization device is abnormal, abnormal characterization device code 0 no abnormalities exist.
[0138] 104, according to the abnormality detection result, the frequency anomaly detection algorithm determines an abnormality occurs in the apparatus is stable.
[0139] Since the result of the abnormality detection device characterized by abnormal presence or absence of an abnormality, abnormality detection result and therefore the abnormality detection algorithm is repeatedly output, it is possible to determine the frequency anomaly occurs in the equipment a certain time period, in turn, may determine the frequency of occurrence of an abnormal device is stable. Incidentally, the determination device according to the abnormality detection result of the frequency of occurrence of an abnormality, refers to an abnormal detection algorithm to the device frequency abnormality occurs, since the process parameters anomaly detection algorithms for adjusting, the abnormality detection algorithm may not be completely accurate whether there is abnormality detecting apparatus, so that the frequency anomaly frequency of actual occurrence of an abnormality determining device according to an abnormality detection result of the device occurs may vary.
[0140] 105, if the abnormality detecting apparatus detects an algorithm of frequency instability occurrence of an abnormality, the abnormality detection algorithm parameters to adjust.
[0141] If the abnormality detection algorithm detects an occurrence frequency instability apparatus abnormality, the abnormality detection described algorithm can not accurately detect whether the device is abnormal, i.e., the actual operating conditions and parameters of the device abnormality detection algorithm does not match, it is possible anomaly detection algorithm the parameter is adjusted so that the actual operating conditions and parameters of the device matches the anomaly detection algorithm, thereby enabling the abnormality detection algorithm can detect the presence of a device abnormality. After the anomaly detection algorithm parameters to adjust, the above steps 101-105 loop until the apparatus detects the abnormality detection algorithm abnormal occurrence frequency stability, i.e., the abnormality detection method can accurately detect whether the device is abnormal.
[0142] In the present application embodiment, determining the variable information of the operating data of the device, after the input variable information abnormality detection algorithm device configured to obtain the abnormality detection result of the abnormality detection algorithm output determines abnormality detection algorithm an apparatus according to the abnormality detection result abnormal frequency is stable, if the stability of the abnormality detection algorithm parameters to adjust. Frequency during normal operation, if the abnormality detection method capable of accurately detecting device is abnormal, the abnormality detection algorithm to detect an apparatus abnormality frequency should be stable, it is possible to detect an apparatus abnormality based on the abnormality detection algorithm is stable , it is determined that the abnormality detection algorithm is able to accurately detect whether the device is abnormal, and further in determining abnormality detection algorithm can not accurately detect whether the device is abnormal, the automatic parameters anomaly detection algorithm are adjusted until the abnormality detection method capable of accurately detecting device is abnormal . Thus, the user only needs to configure the automobile production line anomaly detection algorithm corresponding to the device type, the anomaly detection algorithm parameters may be automatically configured according to the actual operation of the device, the device configuration can be reduced to automobile production line abnormality detection when the algorithm consumed human and material resources, reduce the cost of auto production line of predictive maintenance.

Example Embodiment

[0143] Example 2
[0144] On the basis of the configuration method for providing abnormality detection algorithm according to an embodiment of the, if it is determined the abnormality detection algorithm to detect an apparatus occurrence frequency stability abnormality can be detected in the abnormality detection algorithm to the equipment malfunction occurs, transmits alarm information to a user, by the the user equipment further confirm whether an abnormality actually occurred, and further to adjust the parameters of the abnormality detection algorithm according to the confirmation result of the user, thereby improving the accuracy of abnormality detection algorithm to detect an abnormality of the device.
[0145] figure 2 The present application is a diagram showing the abnormality detection method for adjusting parameters of an algorithm according to a second embodiment 200, as figure 2 , The data processing program 202 according to the configuration information 201 acquired operating data device, the data handler 202 of the data processing operation, to obtain variable information 203, variable information 203 is stored in the database 204, the algorithm runs the program from the database 205 204 reading variable information 203, and the read variable information 203 input to the abnormality detection algorithm, obtaining abnormality detection algorithm 206 outputs an abnormality detection result, stable frequency determining program 207 detects an abnormality detection algorithm according to the abnormality detection apparatus 206 determines the result of abnormal frequency is stable. When the frequency stability determination program 207 determines a frequency anomaly detection algorithm to detect an apparatus abnormal stable, if the abnormality detection result 206 characterization device abnormality, event logger 212 transmits alarm information 208 to a user, event logging program 212 receiving the user after labeling for alarm information 208 feedback information 209, the sample data 209 and 210 according to the tagging information variable information 203, and the formed sample data 210 stored in the database 204.
[0146]When the abnormality detection device detects the occurrence frequency algorithm stable abnormality, the abnormality detection algorithms have been described capable of accurately detecting device is abnormal, this time, if the abnormality detection result of abnormal characterization device 206, sending the corresponding alarm information to the user 208, to confirm whether the user equipment is really abnormal, then the user receives feedback information label 209, wherein the callout 209 is used to indicate the variable information corresponding to the state of the device 203 is operating normally or an abnormal operation, and according to variable and tagging information 209 form information 203 sample data 210, 210 and the sample data stored in the database 204. Sample data 210 includes positive and negative samples, the same as the confirmation result of the positive samples abnormality detection algorithm of the detection result of the user, the negative samples for different confirmation result of the detection result of the user's anomaly detection algorithm, the sample data 210 stored in the database , 210 can be analyzed by the sample data in the database, the anomaly detection algorithm to determine the device abnormality detection accuracy, facilitate the anomaly detection algorithms were evaluated and optimized.
[0147] In one possible implementation, if the label information indicates variable information corresponding to the apparatus operates abnormally, the abnormal label information further includes type information, the anomaly category classification information for indicating the device abnormality occurs, i.e., the abnormality detection algorithm apparatus abnormal, abnormal user equipment does confirm and verify the category of devices that appear abnormal. In this case, according to the abnormality type information and variable information including label information, the classification of the anomaly previously constructed models to optimize the train, wherein the variable information for classification model abnormality determination device according abnormality occurring category.
[0148] Specifically, when the device information indicates abnormal label, after the sample data stored in the database, run the algorithm program 205 reads 210 data from the sample database 204, according to the read sample data 210 includes variable information and abnormal categories information, abnormal classification model to optimize training.
[0149] After when the abnormality detection algorithm to equipment malfunction after sending alarm information to the user, the user receives the alarm information to confirm whether the device actually abnormal, and the feedback corresponding annotation information based on the confirmation result to confirm the variable information corresponding to the device operating status abnormal or normal running operation, thereby generating a sample data including variable information and annotation information, i.e., the sample data comprises variable information device and the operating state of the device is located at variable information (normal operation or abnormal operation). If the user equipment does confirm abnormal, not only the user feedback information label indicating abnormal device, the exception category label information further includes information indicating the device abnormality occurring, it may further include sample data based on the variable information and the exception type information, for abnormal classification model to optimize the training, so that unusual classification model capable of the type of exception appears to more accurately identify the device based on variable information, which is not only able to detect whether the device is abnormal, but also be able to accurately determine the device type of exception appears to realize car pinpoint abnormal production line, improving the user experience.
[0150] In one possible implementation, the algorithm determines the abnormality detecting apparatus according to variable information abnormality occurs, but the user does not confirm the unusual equipment, i.e. label information indicating abnormal device does not appear, then generate the sample data including variable information and annotation information the sample data after the anomaly detection algorithm parameters are adjusted so that the abnormality detection device algorithm does not appear abnormal based on the detection result of the variable information device status.
[0151] In the embodiment of the present application, such as figure 2 , The algorithm runs the program 205 will send the abnormality detection result of the program 206 determines 207 the frequency stability, frequency stability determination program 207 determines abnormal detection algorithm 206 detects an abnormality detection apparatus according to the result of the frequency abnormality is stable, the program determines if the frequency stability algorithm 207 determines that the abnormality detection device detects the abnormal frequency instability occurs, send feedback information 205 to the arithmetic operation program 211, the algorithm runs the program 205 to adjust the parameters of the feedback information 211 according to the abnormality detection algorithm.
[0152] Incidentally, the user can 202 sends the configuration to the data processing program by the user experience (User Experience, UX) tool information 201, event logger 212 can send alarm information 208 via the user experience tool to the user, via the user experience tool to the event recording program 212 transmits the information denoted by 209. The user experience can be a tool HMI (Human Machine Interaction, HMI).
[0153] It is further noted that, as figure 2 , The data processing program 202 comprises a data adapter 2021, data of the adapter 2021 may acquire data of different protocols to be able to collect various types of operational data, adapted to different types of field data to ensure the anomaly detection algorithm provided by the embodiment of the present application the configuration has strong applicability. Configuration information may be configured by data adapter 201 of access rules 2021, 2021 can be such that the data adapter according to the rules set by the subscription data. After data adapter 2021 to the operating data acquisition device, the data handler 202 may purge data 201 and the data processing operation data according to the configuration information 203 to obtain the variable information.

Example Embodiment

[0154] Example three
[0155] exist figure 1 The abnormality detection method illustrated basic configuration of the algorithm 100, the step 104 determines that an abnormality detection algorithm device is abnormal occurrence frequency stability, the device can detect an abnormality occurs a difference frequency abnormality detection algorithm in different detection periods, abnormality detection algorithm to determine an abnormality occurrence frequency stability of the device. image 3 300 is a flowchart according to a third embodiment of the present application frequency stability abnormality determination method, such as image 3 , The stability of the frequency of occurrence of abnormality determination method 300 comprises the steps of:
[0156] 301, according to the abnormality detection result, calculating a first anomaly frequency, wherein the frequency of the first abnormal abnormality detection algorithm detects abnormality in the current detection period for characterizing the device appears.
[0157] Predetermined detection period, at the end of the current detection period, the number of device abnormality detected by abnormality occurs in the current detection period detection algorithm, it calculates a frequency anomaly detection algorithm device abnormality occurs in the current detection period, as a first abnormality frequency.
[0158] 302, obtaining a second anomaly frequency, wherein the second frequency anomaly detection algorithm detects abnormality in the abnormality history period detecting apparatus for characterizing occurs.
[0159] History of the detection result whether the device is abnormal, the abnormality detection algorithm is calculated in accordance with the history of the abnormality detection algorithm in the frequency detection period of the abnormal apparatus, a second abnormality probability.
[0160] It should be understood, the current detection period and historical detection period, the abnormality detection algorithm corresponding to the device with the same operating conditions, such as the production of the same model car in a current detection period and detection period history, automobile production line. Further, in the current detection period and historical detection period, once the abnormality detection algorithm outputs an abnormality detection result based on at least the variable information.
[0161] 303, a first abnormality determining the difference frequency and the second frequency is less than a predetermined abnormal frequency difference threshold, if it is Y, step 304, if no N, step 305 is performed.
[0162] After obtaining the first and second abnormality frequency anomaly frequency, calculate the difference of the first frequency and the second abnormality abnormal frequency, and determines whether the difference between the first frequency and the second abnormality anomaly frequency is less than a preset frequency difference threshold . If the difference between the first frequency and the second abnormal abnormality frequency less than the frequency difference threshold value, the frequency anomaly detection algorithm described an apparatus abnormality is not greater fluctuation and thus step 304 is performed. If the difference between the first frequency and the second abnormality anomaly frequency is a frequency equal to or greater than the threshold difference, the algorithm described abnormality detection device detects the presence of large fluctuations in the frequency of occurrence of abnormality, and thus step 305 is performed.
[0163] Incidentally, since the first abnormal abnormality frequency may be greater than the second frequency, may also be smaller than the second anomaly frequency, so the difference between a first frequency and the second abnormality abnormality frequency means that the first abnormal abnormality frequency and the second frequency the absolute value of the difference.
[0164] 304, to determine the frequency anomaly detection algorithm stabilizing device abnormality occurs, and ends the current process.
[0165] When the difference between the first frequency and the second abnormal abnormality frequency less than the frequency difference threshold value, an abnormality detection algorithm to detect abnormal frequency of occurrence of the apparatus greater fluctuation does not occur, and thus the detection algorithm determines abnormality occurrence frequency stabilization apparatus abnormality.
[0166] 305, it is determined that the abnormality detection algorithm to detect abnormal frequency instability of the device.
[0167] When the difference between the first frequency and the second abnormality anomaly frequency is equal to or greater than a threshold frequency difference, an apparatus abnormality detection algorithm appears there is a large fluctuation frequency abnormality, abnormality detection algorithm further determines an abnormality occurrence frequency apparatus not Stablize.
[0168] Embodiment, the abnormality detection result output from the abnormality detection algorithm calculates abnormality detection algorithm in the current detection period of the first abnormal abnormality frequency device, and calculates the abnormality detection algorithm in the detection period history apparatus embodiment in the present application the second anomaly frequency abnormality occurs, the abnormality when the difference between the first frequency and the second frequency is an abnormal frequency difference smaller than a preset threshold, the algorithm determines the abnormality detection device detects the abnormality occurrence frequency stability, or to determine an abnormality detection algorithm abnormal frequency instability of the device, different detection periods by the abnormality detection algorithm to detect abnormal changes the frequency of the device fails, the algorithm determines the abnormality detection device detects an abnormal frequency is stable, to ensure the accuracy of test results, and thus to ensure the accuracy of anomaly detection algorithms to improve the accuracy of auto production line predictive maintenance.

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.

Similar technology patents

Alarm for detecting faults of overhead lines of distribution network

InactiveCN101446619AObvious beneficial effectSave manpower and material resources
Owner:上海艾晋电力科技有限公司 +2

Classification and recommendation of technical efficacy words

  • Save manpower and material resources
  • low cost

Automatic feeding assembling and transmitting device for electronic elements

InactiveCN102612311AAutomate operationSave manpower and material resources
Owner:深圳市中科创安科技有限公司

System and method for transmitting wireless digital service signals via power transmission lines

ActiveUS7929940B1reduce bandwidth requirementlow cost
Owner:NEXTEL COMMUNICATIONS

Plastic waveguide-fed horn antenna

InactiveUS20100214185A1low cost
Owner:RGT UNIV OF CALIFORNIA

System and method for determination of position

InactiveUS20090149202A1low costreduce requirement
Owner:STEELE CHRISTIAN

Adaptive antenna optimization network

InactiveUS6961368B2low costminimal space
Owner:ERICSSON INC
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products