CN117093944A - Time sequence data template self-adaptive abnormal mode identification method and system - Google Patents
Time sequence data template self-adaptive abnormal mode identification method and system Download PDFInfo
- Publication number
- CN117093944A CN117093944A CN202311175826.7A CN202311175826A CN117093944A CN 117093944 A CN117093944 A CN 117093944A CN 202311175826 A CN202311175826 A CN 202311175826A CN 117093944 A CN117093944 A CN 117093944A
- Authority
- CN
- China
- Prior art keywords
- data
- identified
- abnormal
- mode
- time sequence
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 215
- 238000000034 method Methods 0.000 title claims abstract description 96
- 230000005856 abnormality Effects 0.000 claims abstract description 101
- 238000013441 quality evaluation Methods 0.000 claims abstract description 49
- 230000008569 process Effects 0.000 claims description 35
- 238000001514 detection method Methods 0.000 claims description 32
- 238000003909 pattern recognition Methods 0.000 claims description 31
- 230000000630 rising effect Effects 0.000 claims description 24
- 230000011218 segmentation Effects 0.000 claims description 17
- 230000000737 periodic effect Effects 0.000 claims description 11
- 230000003044 adaptive effect Effects 0.000 claims description 9
- 238000009825 accumulation Methods 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 6
- 238000012567 pattern recognition method Methods 0.000 claims description 6
- 230000006978 adaptation Effects 0.000 claims description 5
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 238000001303 quality assessment method Methods 0.000 claims description 3
- 208000022821 personality disease Diseases 0.000 claims 1
- 230000036541 health Effects 0.000 abstract description 6
- 238000003745 diagnosis Methods 0.000 abstract description 5
- 230000008859 change Effects 0.000 description 9
- 238000012423 maintenance Methods 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000007726 management method Methods 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 208000024891 symptom Diseases 0.000 description 4
- 230000002776 aggregation Effects 0.000 description 3
- 238000004220 aggregation Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 208000035851 morphological anomaly Diseases 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Evolutionary Computation (AREA)
- Pure & Applied Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The application discloses a method and a system for identifying an abnormal mode of self-adaption of a time sequence data template, wherein the method for identifying the abnormal mode of self-adaption of the time sequence data template comprises the following steps: performing quality evaluation on the time sequence data to be identified, continuously accumulating abnormal judgment on the time sequence data to be identified if the quality evaluation result is no abnormality, defining a mode to be identified according to the time sequence data to be identified and setting search information if the abnormal judgment result is true abnormality; performing self-adaptive abnormal mode identification on the time sequence data to be identified to obtain an identification abnormal mode; dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode recognition on the unrecognized data segment to obtain a new recognition abnormal mode, and merging or dividing the new recognition abnormal mode according to the recognition abnormal mode to obtain a merging mode or a dividing mode. The method can be suitable for identifying the abnormal mode of time sequence data in the industrial fault diagnosis and health management scene.
Description
Technical Field
The application relates to the technical field of abnormal pattern recognition of time sequence data in industrial fault diagnosis and health management scenes, in particular to a method and a system for recognizing an abnormal pattern by self-adapting a time sequence data template.
Background
Sources of time series data in an industrial PHM (Prognostics Health Management, fault diagnosis and health management) scenario include: DCS (distributed control system) and SCADA (Supervisory Control And Data Acquisition, data acquisition and monitoring control system) and the like, the time sequence data has the characteristics of large data volume, poor data quality, high real-time performance, high dimensionality, high complexity and the like, and the time sequence data comprises the problems of periodical change, trend change, jump, noise, drift and the like.
Currently, anomaly detection of time series data in industrial PHM scenes has been widely used, and common ones are: a method of detecting dot abnormality and a method of detecting aggregation abnormality, for example: the anomaly detection method based on statistics, clustering, density, machine learning, deep learning and rules cannot reflect the development process or morphological change of data, so that the anomaly/fault process of industrial equipment cannot be mapped from the data angle, and phenomena caused by non-equipment anomalies/faults can be identified, so that false alarms are generated. In an industrial scene, besides the point of attention abnormality or aggregation abnormality, the morphology/mode changes of the time sequence data of the equipment are required to be paid attention to, the morphology/mode often contains the fault symptoms of the equipment besides representing the changes of working conditions and load information of some equipment, the phenomenon is usually represented as the changes of the morphology/mode of a sequence to be detected or a subsequence thereof, the specific morphology/mode can be roughly divided into trend abnormality (upward or downward), step abnormality (ascending or descending), periodic abnormality and the like, and the abnormal changes of the morphology/mode of the equipment corresponding to the time sequence data are identified through analysis and modeling of the morphology/mode of the time sequence data, so that the reliability and the safety of the equipment are improved, and the maintenance cost and the production risk are reduced.
Currently, in the pattern recognition of the existing time series data, common methods for the dimension reduction representation of the time series data are as follows: PAA (Piecewise Aggregate Approximation, piecewise aggregated approximation), SAX (Symbolic Aggregate Approximation, fitting approximation aggregation) and its extended symbolized representation methods, transform-based SVD (Singular Value Decomposition ) and/or DTW (Dynamic Time Warping, dynamic time warping) methods, piecewise polynomial fitting and/or least squares fitting based methods, and time-ordered data similarity measurement methods. The time domain similarity refers to an observation result of a certain potential identical curve in a time dimension, and common methods include: nearest neighbor (1 NN), euclidean distance, and Dynamic Time Warping (DTW). Shape similarity refers to the fact that time sequences of the same class can be distinguished by some identical subsequences or shapes, common methods are: shapelets (time series) and its expansion algorithm. The change similarity refers to a sequence with stronger correlation, essentially belongs to a classification and clustering algorithm based on a model, and common methods are as follows: SVM (Support Vector Machine ). However, most of the existing methods for pattern recognition of time series data are applied to the fields of finance and automobiles, because of the characteristics of high and high complexity of data in industrial scenes, when the known form and template are used for performing abnormal pattern recognition in a time sequence, special attention is required to select and adjust parameters such as data length, numerical scale, time window and sliding window, so that the existing methods for pattern recognition of time series data cannot accurately and efficiently recognize abnormal forms/patterns of industrial equipment data, namely: the method is not suitable for the abnormal pattern recognition of time sequence data in the industrial fault diagnosis and health management scenes.
Disclosure of Invention
The application aims to provide a time sequence data template self-adaptive abnormal pattern recognition method and system, which can be suitable for recognition of abnormal patterns of time sequence data in industrial fault diagnosis and health management scenes and have the technical effects of high recognition efficiency and high accuracy.
In order to achieve the above object, the present application provides a method for identifying abnormal patterns by self-adapting a time-series data template, comprising the following steps: s1: performing quality evaluation on the time sequence data to be identified to generate a quality evaluation result, if the current frequency of the quality evaluation item of the time sequence data to be identified is greater than a preset frequency threshold value, the generated quality evaluation result is abnormal, the process is ended, and a data quality abnormality prompt is sent; if the current frequency of the quality evaluation item of the time sequence data to be identified is smaller than or equal to a preset frequency threshold value, the generated quality evaluation result is abnormal, and S2 is executed; wherein the quality assessment item comprises at least: data drift, data interruption, data overrun, and data repetition value; s2: performing continuous accumulation abnormality judgment on the time sequence data to be identified, generating an abnormality judgment result, if continuous multipoint abnormality exists in the time sequence data to be identified, determining that the generated abnormality judgment result is true abnormality, and executing S3; if discrete individual points in the time sequence data to be identified are abnormal, the generated abnormal judgment result is false abnormal, the process is ended, and an abnormal trip point prompt of the data is sent; if no point abnormality exists in the time sequence data to be identified, the generated abnormality judgment result is no abnormality, and the process is ended; s3: defining a mode to be identified according to the time sequence data to be identified, and setting search information, wherein the mode to be identified at least comprises: trend rising mode, trend falling mode, step rising mode, step falling mode and periodic variation mode; the search information includes at least: the set searching parameters and the initialized minimum distance set; s4: performing self-adaptive abnormal pattern recognition on the time sequence data to be recognized according to the pattern to be recognized and the search information to obtain a recognition abnormal pattern; s5: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
As above, the sub-steps of defining the pattern to be recognized according to the time series data to be recognized and setting the search information are as follows: s31: constructing a plurality of data form templates according to the time sequence data to be identified, describing forms in each data form template, and obtaining a mode to be identified after the description is completed; s32: setting search parameters of a sliding window, initializing a minimum distance set, and taking the set search parameters and the initialized minimum distance set as search information; wherein, the search parameter at least comprises: minimum window length, maximum window length, and sliding window step size.
As described above, the sub-steps of performing adaptive abnormal pattern recognition on the time series data to be recognized according to the pattern to be recognized and the search information to obtain the recognition abnormal pattern are as follows: s41: judging the length of the time sequence data to be identified according to the search parameters, ending the flow if the length of the time sequence data to be identified is smaller than the minimum window length, and not carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the minimum window length and less than the maximum window length, taking the window length from the minimum window length to the length of the time sequence data to be identified as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the maximum window length, taking the window length from the minimum window length to the maximum window length as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; s42: taking time sequence data to be identified under the current window length of the current window as current window data, acquiring a data maximum value and a data minimum value of the current window data, taking the current window length as the time scale of the current window, taking the data maximum value and the data minimum value as the data scale of the current window, and dynamically generating a mode to be identified as a dynamic template under the current window according to the time scale and the data scale, wherein the dynamic template at least comprises: trend rising templates, trend falling templates, step rising templates, step falling templates and periodic variation templates; s43: smoothing the current window data to obtain smoothed data; s45: calculating Euclidean distances between the smoothed data and the dynamic template, and recording all Euclidean distances into an distance set; s46: judging the Euclidean distance set by using the initialized minimum distance set, if any Euclidean distance smaller than the distance value in the initialized minimum distance set exists in the Euclidean distance set, iterating the initialized minimum distance set by using the Euclidean distance set to obtain an identification minimum distance set, completing sliding window search, and taking the category of a dynamic template corresponding to the minimum value in the identification minimum distance set as an identification abnormal mode of the time sequence data to be identified; if any Euclidean distance smaller than the distance value in the initialized minimum distance set does not exist in the Euclidean distance set, the initialized minimum distance set is not utilized to iterate the initialized minimum distance set, sliding window searching is completed by taking the initialized minimum distance set as the identified minimum distance set, and the category of the dynamic template corresponding to the minimum value in the identified minimum distance set is taken as the identified abnormal mode of the time sequence data to be identified.
As described above, the number of unidentified data segments is two, and the two unidentified data segments are respectively subjected to adaptive abnormal pattern recognition to obtain a new recognition abnormal pattern, and the new recognition abnormal pattern is combined or divided according to the recognition abnormal pattern of the identified data segments to obtain a combined pattern or a divided pattern.
As above, wherein the time series data to be identified is divided into an identified data segment and at least one unidentified data segment; carrying out self-adaptive abnormal mode recognition on the unrecognized data segment to obtain a new recognition abnormal mode, merging or dividing the new recognition abnormal mode according to the recognition abnormal mode of the recognized data segment, and obtaining a merging mode or dividing mode by the following substeps: s51: dividing the time sequence data to be identified to obtain a plurality of data segments, wherein the plurality of data segments at least comprise: a recognized data segment and at least one unidentified data segment, wherein the recognized data segment is current window data; s52: carrying out self-adaptive abnormal mode identification on the unrecognized data segment according to the mode to be identified and the search information to obtain a new identification abnormal mode; s53: judging whether the new identification abnormal mode is the same as the identification abnormal mode of the identified data segment, if so, carrying out mode combination to obtain a combined mode, and obtaining a new data length; if the data segments are different, performing mode segmentation to obtain a segmentation mode and a new unidentified data segment; s54: judging the data length of the new unidentified data segment according to the preset template length, and ending the flow if the new data length is smaller than the template length; if the new data length is greater than or equal to the template length, S51 is performed.
As described above, steps S41 to S46 are re-executed with the unrecognized data segment as new time series data to be identified, the recognition abnormal pattern is obtained, and the recognition abnormal pattern is used as a new recognition abnormal pattern.
As above, wherein the preset template length is equal to the minimum window length.
As described above, two or more data segments having the same identification abnormality pattern are combined into one data segment, and a plurality of data segments having different identification abnormality patterns are divided into different data segments.
The application also provides an abnormal pattern recognition system of the self-adapting time sequence data template, which comprises the following steps: the system comprises a data quality evaluation module, an abnormality detection module, a mode identification module and a mode segmentation and merging module; wherein, the data quality evaluation module: the method comprises the steps of performing quality evaluation on time sequence data to be identified, and generating a quality evaluation result, wherein the quality evaluation result is abnormal or free of abnormality; when the quality evaluation result is that no abnormality exists, the time sequence data to be identified is sent to an abnormality detection module; when the quality evaluation result is abnormal, ending the flow and sending a data quality abnormality prompt; an abnormality detection module: continuously accumulating anomaly judgment on the time sequence data to be identified to generate an anomaly judgment result, wherein the anomaly judgment result is true anomaly, false anomaly or no anomaly; when the abnormality judgment result is true abnormality, the time sequence data to be identified is sent to a mode identification module; when the abnormality judgment result is false abnormality, ending the flow and sending an abnormality jump point prompt for the data; ending the flow when the abnormality judgment result is no abnormality; a mode identification module: defining a mode to be identified according to the time sequence data to be identified, and setting search information; performing self-adaptive abnormal pattern recognition on the time sequence data to be recognized according to the pattern to be recognized and the search information to obtain a recognition abnormal pattern; mode segmentation and merging module: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
As described above, the abnormality detection module is provided with an abnormality detection method, and continuous accumulation abnormality determination is performed on the time-series data to be identified by the abnormality detection method, so as to generate an abnormality determination result.
The beneficial effects achieved by the application are as follows:
(1) The method and the system for identifying the abnormal pattern of the self-adaptive time sequence data template are suitable for identifying the abnormal pattern of various time sequence data in a general industrial scene.
(2) The self-adaptive abnormal pattern recognition method of the time sequence data template utilizes the time scale and the data scale (namely, numerical scale) of the time sequence data to be recognized to dynamically generate a dynamic template self-adaptive to the time scale and the data scale of the time sequence data to be recognized, and the minimum Euclidean distance between the time sequence data to be recognized and the dynamic template is obtained by combining continuous cumulative evaluation, sliding window search and greedy matching methods.
(3) The method for identifying the abnormal pattern of the self-adaptive time sequence data template also carries out pattern segmentation and/or combination on the residual data of the time sequence data to be identified, further improves the identification accuracy and can accurately and efficiently identify the abnormal form/pattern of the industrial equipment data.
(4) The self-adaptive abnormal mode recognition method of the time sequence data template optimizes the abnormal detection and improves the stability of the abnormal detection.
(5) In the traditional scheme, the common method for carrying out pattern recognition by using templates can only repeatedly define the method of the templates for each group of data to carry out pattern recognition, the general practical engineering is not preferable, two unresolved problems exist, one is how to repeatedly use multiple groups of data by using a candidate template and how to repeatedly update the multiple groups of data, and the other is how to adapt to each other by using data templates with different numerical scales.
(6) In the sliding window searching process, the self-adaptive abnormal mode recognition method of the time sequence data template provides a matching method based on greedy ideas, and optimizes the data storage space and the traversing efficiency.
(7) Aiming at the characteristics of time sequence data (such as large data quantity, poor data quality, high real-time performance, high dimensionality, high complexity and the like) in an industrial scene, the application can accurately and efficiently identify the abnormal mode by a template self-adaption method, discover equipment fault symptoms and other abnormal conditions as soon as possible, and repair and maintain as soon as possible according to the discovered equipment faults, thereby reducing downtime and maintenance cost, improving the production efficiency and service life of equipment, and reducing maintenance cost and production risk.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings for a person having ordinary skill in the art.
FIG. 1 is a schematic diagram of an embodiment of an abnormal pattern recognition system with adaptive timing data templates;
FIG. 2 is a flow chart of one embodiment of a method for abnormal pattern recognition for temporal data template adaptation;
FIG. 3 is a schematic diagram of one embodiment of gearbox vibration peak data;
FIG. 4 is a schematic diagram of a process for calculating Euclidean distance between five types of dynamic templates and a sliding window which are dynamically generated;
FIG. 5 is a final matching result of step process identification by the abnormal pattern identification method of time sequence data template adaptation.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
As shown in fig. 1, the present application provides an abnormal pattern recognition system with adaptive time series data templates, comprising: a data quality evaluation module 1, an anomaly detection module 2, a pattern recognition module 3, and a pattern segmentation and merging module 4.
Wherein, data quality evaluation module 1: the method comprises the steps of performing quality evaluation on time sequence data to be identified, and generating a quality evaluation result, wherein the quality evaluation result is abnormal or free of abnormality; when the quality evaluation result is that no abnormality exists, the time sequence data to be identified is sent to an abnormality detection module; and when the quality evaluation result is abnormal, ending the flow and sending a data quality abnormality prompt.
Abnormality detection module 2: continuously accumulating anomaly judgment on the time sequence data to be identified to generate an anomaly judgment result, wherein the anomaly judgment result is true anomaly, false anomaly or no anomaly; when the abnormality judgment result is true abnormality, the time sequence data to be identified is sent to a mode identification module; when the abnormality judgment result is false abnormality, ending the flow and sending an abnormality jump point prompt for the data; and when the abnormality judgment result is no abnormality, ending the flow.
Pattern recognition module 3: defining a mode to be identified according to the time sequence data to be identified, and setting search information; and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified according to the mode to be identified and the search information to obtain an identification abnormal mode.
Mode segmentation and merging module 4: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
Further, an anomaly detection method is arranged in the anomaly detection module 2, continuous accumulation anomaly judgment is carried out on the time sequence data to be identified through the anomaly detection method, an anomaly judgment result is generated, if continuous multipoint anomalies exist in the time sequence data to be identified, the generated anomaly judgment result is true anomalies, and the time sequence data to be identified is sent to the pattern identification module; if discrete individual points in the time sequence data to be identified are abnormal, the generated abnormal judgment result is false abnormal, the process is ended, and an abnormal trip point prompt of the data is sent; if no point abnormality exists in the time sequence data to be identified, the generated abnormality judgment result is no abnormality, and the process is ended.
Specifically, the anomaly detection method set in the anomaly detection module 2 may be a z-score (a method for detecting anomalies in parameters in a one-dimensional or low-dimensional feature space), a box-line graph method, a random forest and LOF (Local outlier factor, local anomaly factor) algorithm, but is not limited to the z-score, the box-line graph method, the random forest and LOF algorithms, and the anomaly detection method added with continuous cumulative evaluation is preferred in the present application, namely: the abnormality detection method can judge whether continuous multipoint abnormality occurs in the time sequence data to be identified.
In an industrial scene, a trip point on data often appears, but the occurrence of the trip point cannot indicate that equipment fails/is abnormal, and false alarm generated by the abnormal trip point often gives wrong maintenance guidance to operation and maintenance personnel. Therefore, the continuous accumulation abnormal judgment is carried out on the time sequence data to be identified by adopting the abnormal detection method added with the continuous accumulation evaluation, so that the judgment accuracy can be improved, and the wrong maintenance guidance for operation and maintenance personnel is avoided.
Optimization of the anomaly detection method refers to the addition of a continuous cumulative assessment, namely: and if the time sequence data to be identified needs to have continuous multipoint abnormity, the group of data is considered to be truly abnormal, otherwise, the data is prompted to have trip point prompt.
As shown in fig. 2, the present application further provides a method for identifying abnormal patterns by self-adapting a time sequence data template, which includes the following steps:
s1: performing quality evaluation on the time sequence data to be identified to generate a quality evaluation result, if the current frequency of the quality evaluation item of the time sequence data to be identified is greater than a preset frequency threshold value, the generated quality evaluation result is abnormal, the process is ended, and a data quality abnormality prompt is sent; if the current frequency of the quality evaluation item of the time sequence data to be identified is smaller than or equal to a preset frequency threshold value, the generated quality evaluation result is abnormal, and S2 is executed; wherein the quality assessment item comprises at least: data drift, data interruption, data overrun, and data repetition.
Specifically, the length of the time sequence data to be identified is. Time sequence data to be identified->Inputting the time series data to a data quality evaluation module, wherein the data quality evaluation module is used for identifying the time series data +.>Quality evaluation is carried out on the quality evaluation items of (2) if the time series data to be identified are +>If the current frequency of the quality evaluation item of (a) is greater than a preset frequency threshold value, the time sequence data to be identified is represented by +.>The data quality problem exists, the generated quality evaluation result is abnormal, the process is ended, and a data quality abnormality prompt is sent; if time series data to be identified->If the current frequency of the quality evaluation items of (a) is less than or equal to a preset frequency threshold value, the time series data to be identified is represented +.>No data quality problem exists, the generated quality evaluation result is no abnormality,s2 is performed. The time sequence data to be identified is the data to be evaluated.
S2: performing continuous accumulation abnormality judgment on the time sequence data to be identified, generating an abnormality judgment result, if continuous multipoint abnormality exists in the time sequence data to be identified, determining that the generated abnormality judgment result is true abnormality, and executing S3; if discrete individual points in the time sequence data to be identified are abnormal, the generated abnormal judgment result is false abnormal, the process is ended, and an abnormal trip point prompt of the data is sent; if no point abnormality exists in the time sequence data to be identified, the generated abnormality judgment result is no abnormality, and the process is ended.
Specifically, the abnormality detection module judges the time sequence data to be identifiedWhether or not there is continuity +.>Point anomaly, if present, indicates time-series data +.>For true abnormality, generating abnormality judgment result as true abnormality, and identifying time sequence data +.>And sending the data to a pattern recognition module. If discrete individual points in the time sequence data to be identified are abnormal, the generated abnormal judgment result is false abnormal, the process is ended, and an abnormal trip point prompt of the data is sent; if no point abnormality exists in the time sequence data to be identified, the generated abnormality judgment result is no abnormality, and the process is ended.
Wherein,the specific value of (2) depends on the actual scenario.
S3: defining a mode to be identified according to the time sequence data to be identified, and setting search information, wherein the mode to be identified at least comprises: trend rising mode, trend falling mode, step rising mode, step falling mode and periodic variation mode; the search information includes at least: the set search parameters and the initialized minimum distance set.
Further, the substeps of defining a pattern to be recognized according to the time sequence data to be recognized and setting search information are as follows: s31: and constructing a plurality of data form templates according to the time sequence data to be identified, describing the form in each data form template, and obtaining the mode to be identified after the description is completed.
Specifically, a data form template is constructed according to an actual scene, a plurality of forms are set, the forms are described according to field knowledge, experience or interested data forms, and a mode to be identified is obtained after the description is completed. The morphology and/or the morphology number of the time series data in different scenes are different. The actual scene of the present application is an industrial scene, but is not limited to an industrial scene, and the present application is preferably an industrial scene. The time sequence data in the industrial scene at least comprises: five common morphological anomalies, trend up, trend down, step up, step down and periodic change, are preferred in the present application: and taking trend rising, trend falling, step rising, step falling and periodic variation as modes to be identified. In the process of generating the pattern to be identified, the description of the pattern is the rough pattern of the data pattern, the initial template matched with the pattern can be generated without refining to the specific time, numerical scale and other parameter values, various data can be universal, the initial template is called after the time sequence data to be identified is received, and various data pattern templates are constructed according to various patterns of the time sequence data to be identified.
Furthermore, the pattern to be identified can be continuously updated iteratively according to the actual requirement.
S32: setting search parameters of a sliding window, initializing a minimum distance set, and taking the set search parameters and the initialized minimum distance set as search information; wherein, the search parameter at least comprises: minimum window length, maximum window length, and sliding window step size.
Specifically, the minimum window length in the sliding window setting processMaximum window length->Sliding window step sizeWherein the minimum window length +.>Maximum window length->And sliding window step +.>The specific value of (2) is defined according to the actual scenario. Minimum window length +.>Is the minimum length of the recognizable pattern.
Wherein the initialized minimum distance setIs infinite.
S4: and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified according to the mode to be identified and the search information to obtain an identification abnormal mode.
Further, according to the pattern to be identified and the search information, performing adaptive abnormal pattern identification on the time sequence data to be identified, and obtaining the abnormal pattern to be identified comprises the following substeps:
s41: judging the length of the time sequence data to be identified according to the search parameters, ending the flow if the length of the time sequence data to be identified is smaller than the minimum window length, and not carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the minimum window length and less than the maximum window length, taking the window length from the minimum window length to the length of the time sequence data to be identified as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the maximum window length, taking the window length from the minimum window length to the maximum window length as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal pattern identification on the time sequence data to be identified.
S42: taking time sequence data to be identified under the current window length of the current window as current window data, acquiring a data maximum value and a data minimum value of the current window data, taking the current window length as the time scale of the current window, taking the data maximum value and the data minimum value as the data scale of the current window, and dynamically generating a mode to be identified as a dynamic template under the current window according to the time scale and the data scale, wherein the dynamic template at least comprises: trend rising template, trend falling template, step rising template, step falling template and periodic variation template.
Specifically, the optimal solution at the current window length is defined as infinity. The trend rising template, the trend falling template, the step rising template, the step falling template and the period change template are dynamic templates matched with the time scale and the data scale of the current window. In the sliding window searching process, the data of the time sequence data to be identified under the current window length of the current window is the current window data。
S43: and carrying out smoothing treatment on the current window data to obtain smoothed data.
Specifically, the current window data is traversed by a sliding window traversing methodSmoothing by polynomial fitting enables the current window data +. >Becomes smoother and more regular, removing the current window data +.>Noise and unnecessary fluctuations in the frequency spectrum. Calculating the Euclidean distance using the obtained smoothed data can make the pattern recognition module more robustThe true signal is of interest, rather than noise, thereby improving the accuracy and robustness of the pattern recognition module.
Further, the smoothing processing is performed on the current window data by adopting a local least squares fitting method to obtain smoothed data, but the method is not limited to the local least squares fitting method, and can be other smoothing methods, and the method is preferably the local least squares fitting method.
S45: and calculating Euclidean distances between the smoothed data and the dynamic template, and recording all Euclidean distances into a Euclidean distance set.
Specifically, similarity calculation is carried out on the smoothed data and five types of dynamic templates, namely a trend rising template, a trend falling template, a step rising template, a step falling template and a period change template, so as to obtain Euclidean distances between the smoothed data and each type of dynamic template, and all Euclidean distances are recorded into an Euclidean distance set。
Furthermore, if all the Euclidean distances obtained by calculation are required to be saved in the sliding window traversal process, the process data are very large, and the algorithm execution efficiency and the storage space are affected, so that the application adopts a greedy matching method in each window length of the current window length, namely: and in the process of calculating the Euclidean distance through sliding window traversal, all values before the iteration of the Euclidean distance set with the minimum Euclidean distance under the current window are taken, and then the obtained Euclidean distance set is a solution with the minimum Euclidean distance in the whole sliding window traversal process.
S46: judging the Euclidean distance set by using the initialized minimum distance set, if any Euclidean distance smaller than the distance value in the initialized minimum distance set exists in the Euclidean distance set, iterating the initialized minimum distance set by using the Euclidean distance set to obtain an identification minimum distance set, completing sliding window search, and taking the category of a dynamic template corresponding to the minimum value in the identification minimum distance set as an identification abnormal mode of the time sequence data to be identified; if any Euclidean distance smaller than the distance value in the initialized minimum distance set does not exist in the Euclidean distance set, the initialized minimum distance set is not utilized to iterate the initialized minimum distance set, sliding window searching is completed by taking the initialized minimum distance set as the identified minimum distance set, and the category of the dynamic template corresponding to the minimum value in the identified minimum distance set is taken as the identified abnormal mode of the time sequence data to be identified.
Specifically, the minimum distance set which is iteratively initialized by using the Euclidean distance set is a greedy matching process of the minimum distance of iteration. And in the recognition minimum distance set obtained by matching the smoothed data with the multiple types of dynamic templates, the type of the dynamic template with the minimum Euclidean distance is the abnormal mode recognized by the time sequence data to be recognized.
S5: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
Furthermore, the number of the unidentified data segments is determined according to practical situations, the application preferably carries out self-adaptive abnormal mode identification on the two unidentified data segments to obtain a new identification abnormal mode, and carries out combination or segmentation on the new identification abnormal mode according to the identification abnormal mode of the identified data segments to obtain a combination mode or a segmentation mode.
Further, the time sequence data to be identified is divided into an identified data segment and at least one unidentified data segment; carrying out self-adaptive abnormal mode recognition on the unrecognized data segment to obtain a new recognition abnormal mode, merging or dividing the new recognition abnormal mode according to the recognition abnormal mode of the recognized data segment, and obtaining a merging mode or dividing mode by the following substeps:
s51: dividing the time sequence data to be identified to obtain a plurality of data segments, wherein the plurality of data segments at least comprise: a recognized data segment and at least one unidentified data segment, wherein the recognized data segment is current window data.
Specifically, for a group of complex time sequence data to be identified containing multiple modes, through the mode identification process of step S4, a data segment most similar to one of multiple dynamic templates (i.e. the current window data with the smallest euclidean distance) is already obtained, and after the data segment is removed from the data to be evaluated as an identified data segment, at least two remaining unidentified data segments are generated as unidentified data segments.
S52: and carrying out self-adaptive abnormal mode identification on the unrecognized data segment according to the mode to be identified and the search information to obtain a new identification abnormal mode.
Specifically, as an embodiment, the steps S41 to S46 are re-executed with the unidentified data segment as new time series data to be identified, the identified abnormal pattern is obtained, and the identified abnormal pattern is used as a new identified abnormal pattern.
S53: judging whether the new identification abnormal mode is the same as the identification abnormal mode of the identified data segment, if so, carrying out mode combination to obtain a combined mode, and obtaining a new data length; if the data segments are different, pattern segmentation is performed to obtain a segmentation pattern and a new unidentified data segment is obtained.
Specifically, if the new recognition abnormal mode is the same as the recognition abnormal mode of the recognized data segment, the recognition result obtained by the current pattern recognition process is the same as the recognition result obtained by the previous pattern recognition process, and pattern merging is performed. If the new recognition abnormal mode is different from the recognition abnormal mode of the recognized data segment, the recognition result obtained by the current mode recognition flow is different from the recognition result obtained by the last mode recognition flow, and the mode segmentation is performed.
Further, two or more data segments with the same identification abnormal pattern are combined into one data segment, and a plurality of data segments with different identification abnormal patterns are divided into different data segments.
S54: judging the data length of the new unidentified data segment according to the preset template length, and ending the flow if the new data length is smaller than the template length; if the new data length is greater than or equal to the template length, S51 is performed.
Specifically, the specific value of the preset template length is determined according to the actual situation, and the preset template length is the minimum identification length.
As one example, the preset template length is equal to the minimum window length。
Specifically, the actual engineering may generate an abnormal phenomenon that the characteristic of the abnormal/fault of the equipment does not exceed the threshold value compared with the baseline data, the abnormal phenomenon is often not far from the distribution or distance of the baseline data, the form/mode of the abnormal phenomenon often includes the fault symptom of the equipment besides representing the change of some equipment working conditions and load information, the abnormal phenomenon is usually represented as the change of a sub-sequence mode, and the specific mode can be at least divided into: trend rising anomaly, trend falling anomaly, step rising anomaly, step falling anomaly, and periodic variation anomaly. According to field experience, the data form of interest is set to be five morphological anomalies of trend rising, trend falling, step rising, step falling and periodic variation.
As one embodiment, fig. 3 is an abnormal data of a vibration peak of a section of gearbox identified by a device after abnormality detection, where the abnormal data includes 720 data points, and a section of step process exists in the 720 data points, and the step process is identified by using an abnormal pattern identification method adaptive to a time sequence data template.
First, the minimum window length is set to 120 points, the maximum window length is 360 points, the sliding window step length is 12 points, the process of calculating Euclidean distance between five dynamically generated dynamic templates and the sliding window is shown in fig. 4, wherein a is an adaptive dynamic template generated according to the data maximum value and the data minimum value of the data segment of the current window, and b is the data segment generated in the sliding window process.
As shown in fig. 5, c represents the original data, d represents the original data segment under the sliding window, e represents the upward step template in the five types of dynamic templates is most similar to the data segment under the window, namely: the part with the smallest Euclidean distance. In the process of calculating the Euclidean distance between the current window data and five types of self-adaptive dynamic templates (namely, a trend rising template, a trend falling template, a step rising template, a step falling template and a period change template), the minimum Euclidean distance exists in the 9 th traversal with the window length of 144 points is identified, namely: euclidean_distance=2.54.
The beneficial effects achieved by the application are as follows:
(1) The method and the system for identifying the abnormal pattern of the self-adaptive time sequence data template are suitable for identifying the abnormal pattern of various time sequence data in a general industrial scene.
(2) The self-adaptive abnormal pattern recognition method of the time sequence data template utilizes the time scale and the data scale (namely, numerical scale) of the time sequence data to be recognized to dynamically generate a dynamic template self-adaptive to the time scale and the data scale of the time sequence data to be recognized, and the minimum Euclidean distance between the time sequence data to be recognized and the dynamic template is obtained by combining continuous cumulative evaluation, sliding window search and greedy matching methods.
(3) The method for identifying the abnormal pattern of the self-adaptive time sequence data template also carries out pattern segmentation and/or combination on the residual data of the time sequence data to be identified, further improves the identification accuracy and can accurately and efficiently identify the abnormal form/pattern of the industrial equipment data.
(4) The self-adaptive abnormal mode recognition method of the time sequence data template optimizes the abnormal detection and improves the stability of the abnormal detection.
(5) In the traditional scheme, the common method for carrying out pattern recognition by using templates can only repeatedly define the method of the templates for each group of data to carry out pattern recognition, the general practical engineering is not preferable, two unresolved problems exist, one is how to repeatedly use multiple groups of data by using a candidate template and how to repeatedly update the multiple groups of data, and the other is how to adapt to each other by using data templates with different numerical scales.
(6) In the sliding window searching process, the self-adaptive abnormal mode recognition method of the time sequence data template provides a matching method based on greedy ideas, and optimizes the data storage space and the traversing efficiency.
(7) Aiming at the characteristics of time sequence data (such as large data quantity, poor data quality, high real-time performance, high dimensionality, high complexity and the like) in an industrial scene, the application can accurately and efficiently identify the abnormal mode by a template self-adaption method, discover equipment fault symptoms and other abnormal conditions as soon as possible, and repair and maintain as soon as possible according to the discovered equipment faults, thereby reducing downtime and maintenance cost, improving the production efficiency and service life of equipment, and reducing maintenance cost and production risk.
While preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the scope of the application be interpreted as including the preferred embodiments and all alterations and modifications that fall within the scope of the application. It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the present application and the technical equivalents thereof, the present application is also intended to include such modifications and variations.
Claims (10)
1. The abnormal pattern recognition method for the self-adaption of the time sequence data template is characterized by comprising the following steps of:
s1: performing quality evaluation on the time sequence data to be identified to generate a quality evaluation result, if the current frequency of the quality evaluation item of the time sequence data to be identified is greater than a preset frequency threshold value, the generated quality evaluation result is abnormal, the process is ended, and a data quality abnormality prompt is sent; if the current frequency of the quality evaluation item of the time sequence data to be identified is smaller than or equal to a preset frequency threshold value, the generated quality evaluation result is abnormal, and S2 is executed; wherein the quality assessment item comprises at least: data drift, data interruption, data overrun, and data repetition value;
S2: performing continuous accumulation abnormality judgment on the time sequence data to be identified, generating an abnormality judgment result, if continuous multipoint abnormality exists in the time sequence data to be identified, determining that the generated abnormality judgment result is true abnormality, and executing S3; if discrete individual points in the time sequence data to be identified are abnormal, the generated abnormal judgment result is false abnormal, the process is ended, and an abnormal trip point prompt of the data is sent; if no point abnormality exists in the time sequence data to be identified, the generated abnormality judgment result is no abnormality, and the process is ended;
s3: defining a mode to be identified according to the time sequence data to be identified, and setting search information, wherein the mode to be identified at least comprises: trend rising mode, trend falling mode, step rising mode, step falling mode and periodic variation mode; the search information includes at least: the set searching parameters and the initialized minimum distance set;
s4: performing self-adaptive abnormal pattern recognition on the time sequence data to be recognized according to the pattern to be recognized and the search information to obtain a recognition abnormal pattern;
s5: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
2. The abnormal pattern recognition method of a time series data template adaptation according to claim 1, wherein the sub-steps of defining a pattern to be recognized according to the time series data to be recognized and setting search information are as follows: s31: constructing a plurality of data form templates according to the time sequence data to be identified, describing forms in each data form template, and obtaining a mode to be identified after the description is completed;
s32: setting search parameters of a sliding window, initializing a minimum distance set, and taking the set search parameters and the initialized minimum distance set as search information; wherein, the search parameter at least comprises: minimum window length, maximum window length, and sliding window step size.
3. The abnormal pattern recognition method of the time series data template adaptation according to claim 2, wherein the sub-steps of performing the self-adaptation abnormal pattern recognition on the time series data to be recognized according to the pattern to be recognized and the search information, and obtaining the recognition abnormal pattern are as follows:
s41: judging the length of the time sequence data to be identified according to the search parameters, ending the flow if the length of the time sequence data to be identified is smaller than the minimum window length, and not carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the minimum window length and less than the maximum window length, taking the window length from the minimum window length to the length of the time sequence data to be identified as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified; if the length of the time sequence data to be identified is greater than the maximum window length, taking the window length from the minimum window length to the maximum window length as the current window length, searching a sliding window according to the current window length, executing S42, and carrying out self-adaptive abnormal mode identification on the time sequence data to be identified;
S42: taking time sequence data to be identified under the current window length of the current window as current window data, acquiring a data maximum value and a data minimum value of the current window data, taking the current window length as the time scale of the current window, taking the data maximum value and the data minimum value as the data scale of the current window, and dynamically generating a mode to be identified as a dynamic template under the current window according to the time scale and the data scale, wherein the dynamic template at least comprises: trend rising templates, trend falling templates, step rising templates, step falling templates and periodic variation templates;
s43: smoothing the current window data to obtain smoothed data;
s45: calculating Euclidean distances between the smoothed data and the dynamic template, and recording all Euclidean distances into an distance set;
s46: judging the Euclidean distance set by using the initialized minimum distance set, if any Euclidean distance smaller than the distance value in the initialized minimum distance set exists in the Euclidean distance set, iterating the initialized minimum distance set by using the Euclidean distance set to obtain an identification minimum distance set, completing sliding window search, and taking the category of a dynamic template corresponding to the minimum value in the identification minimum distance set as an identification abnormal mode of the time sequence data to be identified; if any Euclidean distance smaller than the distance value in the initialized minimum distance set does not exist in the Euclidean distance set, the initialized minimum distance set is not utilized to iterate the initialized minimum distance set, sliding window searching is completed by taking the initialized minimum distance set as the identified minimum distance set, and the category of the dynamic template corresponding to the minimum value in the identified minimum distance set is taken as the identified abnormal mode of the time sequence data to be identified.
4. The method for identifying an abnormal pattern adaptively to a time-series data template according to claim 3, wherein the number of unidentified data segments is two, the two unidentified data segments are respectively subjected to adaptive abnormal pattern identification to obtain a new identified abnormal pattern, and the new identified abnormal pattern is combined or divided according to the identified abnormal pattern of the identified data segments to obtain a combined pattern or a divided pattern.
5. The method for adaptive anomaly pattern recognition of a time series data template according to claim 4, wherein the time series data to be recognized is divided into a recognized data segment and at least one unidentified data segment; carrying out self-adaptive abnormal mode recognition on the unrecognized data segment to obtain a new recognition abnormal mode, merging or dividing the new recognition abnormal mode according to the recognition abnormal mode of the recognized data segment, and obtaining a merging mode or dividing mode by the following substeps:
s51: dividing the time sequence data to be identified to obtain a plurality of data segments, wherein the plurality of data segments at least comprise: a recognized data segment and at least one unidentified data segment, wherein the recognized data segment is current window data;
s52: carrying out self-adaptive abnormal mode identification on the unrecognized data segment according to the mode to be identified and the search information to obtain a new identification abnormal mode;
S53: judging whether the new identification abnormal mode is the same as the identification abnormal mode of the identified data segment, if so, carrying out mode combination to obtain a combined mode, and obtaining a new data length; if the data segments are different, performing mode segmentation to obtain a segmentation mode and a new unidentified data segment;
s54: judging the data length of the new unidentified data segment according to the preset template length, and ending the flow if the new data length is smaller than the template length; if the new data length is greater than or equal to the template length, S51 is performed.
6. The method for identifying an abnormal pattern according to claim 5, wherein the unidentified data segment is used as new time series data to be identified, steps S41-S46 are re-executed to obtain an identified abnormal pattern, and the identified abnormal pattern is used as a new identified abnormal pattern.
7. The method for identifying an abnormal pattern according to claim 5, wherein the preset template length is equal to a minimum window length.
8. The method for identifying an abnormal pattern according to claim 5, wherein two or more data segments having the same abnormal pattern are combined into one data segment, and a plurality of data segments having different abnormal patterns are divided into different data segments.
9. An abnormal pattern recognition system for time series data template adaptation, comprising: the system comprises a data quality evaluation module, an abnormality detection module, a mode identification module and a mode segmentation and merging module;
wherein, the data quality evaluation module: the method comprises the steps of performing quality evaluation on time sequence data to be identified, and generating a quality evaluation result, wherein the quality evaluation result is abnormal or free of abnormality; when the quality evaluation result is that no abnormality exists, the time sequence data to be identified is sent to an abnormality detection module; when the quality evaluation result is abnormal, ending the flow and sending a data quality abnormality prompt;
an abnormality detection module: continuously accumulating anomaly judgment on the time sequence data to be identified to generate an anomaly judgment result, wherein the anomaly judgment result is true anomaly, false anomaly or no anomaly; when the abnormality judgment result is true abnormality, the time sequence data to be identified is sent to a mode identification module; when the abnormality judgment result is false abnormality, ending the flow and sending an abnormality jump point prompt for the data; ending the flow when the abnormality judgment result is no abnormality;
a mode identification module: defining a mode to be identified according to the time sequence data to be identified, and setting search information; performing self-adaptive abnormal pattern recognition on the time sequence data to be recognized according to the pattern to be recognized and the search information to obtain a recognition abnormal pattern;
Mode segmentation and merging module: dividing the time sequence data to be identified into an identified data segment and at least one unidentified data segment; and carrying out self-adaptive abnormal mode identification on the unrecognized data segment to obtain a new identification abnormal mode, and merging or dividing the new identification abnormal mode according to the identification abnormal mode of the identified data segment to obtain a merging mode or a dividing mode.
10. The abnormal pattern recognition system according to claim 9, wherein an abnormality detection method is provided in the abnormality detection module, and continuous cumulative abnormality judgment is performed on the time-series data to be recognized by the abnormality detection method, thereby generating an abnormality judgment result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311175826.7A CN117093944A (en) | 2023-09-13 | 2023-09-13 | Time sequence data template self-adaptive abnormal mode identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311175826.7A CN117093944A (en) | 2023-09-13 | 2023-09-13 | Time sequence data template self-adaptive abnormal mode identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117093944A true CN117093944A (en) | 2023-11-21 |
Family
ID=88771874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311175826.7A Pending CN117093944A (en) | 2023-09-13 | 2023-09-13 | Time sequence data template self-adaptive abnormal mode identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117093944A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117972318A (en) * | 2024-04-02 | 2024-05-03 | 山东万洋石油科技有限公司 | Pulse waveform testing method and system for high-speed self-generating shear valve |
-
2023
- 2023-09-13 CN CN202311175826.7A patent/CN117093944A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117972318A (en) * | 2024-04-02 | 2024-05-03 | 山东万洋石油科技有限公司 | Pulse waveform testing method and system for high-speed self-generating shear valve |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112527788B (en) | Method and device for detecting and cleaning abnormal value of transformer monitoring data | |
| US9779361B2 (en) | Method for learning exemplars for anomaly detection | |
| US20150220847A1 (en) | Information Processing Apparatus, Diagnosis Method, and Program | |
| US11860608B2 (en) | Industrial equipment operation, maintenance and optimization method and system based on complex network model | |
| US20180164794A1 (en) | Methods and Systems for Discovery of Prognostic Subsequences in Time Series | |
| CN110969556A (en) | Method and device for detecting river water quality abnormity by machine learning multi-dimension multi-model fusion | |
| CN117093944A (en) | Time sequence data template self-adaptive abnormal mode identification method and system | |
| Lyu et al. | Remaining useful life estimation with multiple local similarities | |
| Calvo-Bascones et al. | A collaborative network of digital twins for anomaly detection applications of complex systems. Snitch Digital Twin concept | |
| CN117131110B (en) | Method and system for monitoring dielectric loss of capacitive equipment based on correlation analysis | |
| Wang et al. | Fault detection and identification using a Kullback-Leibler divergence based multi-block principal component analysis and Bayesian inference | |
| CN111832880B (en) | Computer-implemented determination of quality indicators for ongoing production lot runs | |
| CN112529053A (en) | Short-term prediction method and system for time sequence data in server | |
| CN116610938A (en) | Method and equipment for detecting unsupervised abnormality of semiconductor manufacture in curve mode segmentation | |
| CN116108371A (en) | Cloud service abnormity diagnosis method and system based on cascade abnormity generation network | |
| Gao et al. | A novel fault detection model based on vector quantization sparse autoencoder for nonlinear complex systems | |
| CN112241751B (en) | Anomaly detection method and system, and computer readable storage medium | |
| Yu et al. | A hybrid learning-based model for on-line monitoring and diagnosis of out-of-control signals in multivariate manufacturing processes | |
| KR102486463B1 (en) | Method and Apparatus for Real Time Fault Detection Using Time series data According to Degradation | |
| Yasenjiang et al. | Fault Diagnosis and Prediction of Continuous Industrial Processes Based on Hidden Markov Model‐Bayesian Network Hybrid Model | |
| CN113377630A (en) | Universal KPI anomaly detection framework implementation method | |
| CN115249059A (en) | Model training and abnormal data analysis method and device and computer storage medium | |
| Garg et al. | Machine learning-based abnormality detection approach for vacuum pump assembly line | |
| Manca et al. | Explainable AI for Industrial Alarm Flood Classification Using Counterfactuals | |
| Olena et al. | EXTRACTION OF PRECEDENTS USING THE NEAREST NEIGHBOR METHOD |
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 |