CN106649438A - Time series data unexpected fault detection method - Google Patents
Time series data unexpected fault detection method Download PDFInfo
- Publication number
- CN106649438A CN106649438A CN201610814543.6A CN201610814543A CN106649438A CN 106649438 A CN106649438 A CN 106649438A CN 201610814543 A CN201610814543 A CN 201610814543A CN 106649438 A CN106649438 A CN 106649438A
- Authority
- CN
- China
- Prior art keywords
- parameter
- data
- feature
- character
- storehouse
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a time series data unexpected fault detection method. Different situations presented by parameters are studied through known data, and a parameter information library and an event feature library are established; a relation between the combined features of different samples and the time series system fault status is analyzed; a data preprocessing module, a TK-Means clustering module, a feature library generation module and a detection and judgment module are included. The method comprises the steps that after data mining operations such as multi-feature extraction and clustering are conducted on all parameter values, feature characters of all the parameters at the same time are combined to establish the event feature library; when it is observed that no appeared feature combination exists in the event feature library in real-time data, it is judged that an unexpected event occurs, and the unexpected event is a potential unexpected fault.
Description
Technical field
The invention belongs to Intelligent Information Processing and field of computer technology, and in particular to a kind of unexpected failure of time series data
Detection method.
Background technology
Complicated sequential system is made up of many parts, and each part has many observed parameters again.In general, part
The different situation of each parameter can present the ruuning situation of system entirety or local.Need the difference presented to given data
Situation is learnt, it is established that affair character storehouse.Then, the combinations of features pattern and sequential system for analyzing different event is run
The relation of situation.Thus the current operation conditions of sequential system can in real time be reflected according to the assemblage characteristic of current data.
If detecting the integrated mode not occurred in affair character storehouse, then be exactly to there occurs unexpected event.It is unexpected
Event may represent emerging situation, unknown fault signature before more likely represent.Existing method is mostly being based on
The method of reasoning by cases (i.e. Case Based Reasoning, CBR) is analyzed to data attribute, by it with case library in
Case characteristic is carried out contrasting and then determines concrete failure；Or for new unexpected case decision error, or incapability is
Power.It is often relatively difficult that existing method compiles case library, needs to consume a large amount of manpowers.
The content of the invention
Object of the present invention is to provide a kind of unexpected fault detection method of time series data, can be used in real-time detection
Unexpected event or failure, and associated ancillary information is provided, help system user goes to judge what the unexpected event was reflected
System operation situation.
To reach above-mentioned purpose, the technical solution used in the present invention is：
A kind of unexpected fault detection method of time series data, realizing the system of the method includes data preprocessing module, TK-
Means cluster modules, feature database generation module, detection determination module, it is comprised the concrete steps that：
1) first, using data preprocessing module elimination of burst noise is carried out to initial data, file and normalizing is processed at equal intervals
Change is processed, and obtains standardized data, and extracts related multiple features vector；
2) secondly, characteristic vector is clustered using TK-Means cluster modules, is then based on cluster result, will be original
Data are expressed as feature string；
3) and then, using master data of the feature database generation module to each parameter in data prediction and cluster process
In recorded parameter information storehouse, including maximum, minimum of a value, the cluster center vector situation corresponding with characteristic character of each parameter,
Then generating one is used to record the occurrence number of all parameter combination features and the affair character storehouse of frequency；
4) it is last, its characteristic vector is obtained by detecting that determination module is real-time to new data, and to the multiple ginsengs of synchronization
The combination of number feature is contrasted with data in feature database, if the assemblage characteristic has been recorded in feature database, then be exactly
Known mode, is otherwise exactly to detect unexpected pattern, and unexpected pattern is exactly potential unexpected failure.
Described data preprocessing module " elimination of burst noise process " refers to the invalid outlier of deletion, retains virtual value；Data are entered
Row processes ensure that the time interval in continuous time section between any two data point is identical at equal intervals；Data are through at equal intervals
It is normalized after process and eliminates impact of the different dimensions to result so as to obtain standardized data.
Described TK-Means cluster modules adopt application number：2015105516228, a kind of time series data off-note
Digging system and method, wherein, ' a ' represents the feature of most frequent appearance, and ' b ' represents the secondary feature for frequently occurring, and the rest may be inferred；
Probability of occurrence less than given threshold value 0.02 feature be represented as '’；Empty data, i.e., non-record data it is interval or deleted
Interval, then represented with ' # '.
Described feature database generation module generates a parameter information storehouse and an affair character storehouse：
The foundation in described parameter information storehouse includes：
2-1) take a pending parameter；
The record of the parameter 2-2) is created in parameter information storehouse；
2-3) record the maximum and minimum value tag of the parameter；
2-4) record the mean parameter, variance, wavelet coefficient, fourier coefficient, the maximum of frequecy characteristic and minimum of a value；
2-5) obtain the result of the parameter TK-Means cluster；
2-6) cluster result records each cluster center vector of the parameter and its corresponding characteristic character；
2-7) judge currently whether processed all of parameter；If untreated complete all parameters, step 2-1 weight is gone to
Said process is performed again；If having processed all parameters, execution step 2-8)；
2-8) output result, undated parameter information bank content；Generation parameter information bank process terminates；
Process is set up in described affair character storehouse：
The characteristic character sequence of all parameters of 3-1) aliging；
3-2) recording feature storehouse statistics beginning and ending time, window number, parameter composition；
Current character position i 3-3) is put for 0；
Whether current character position i 3-4) is judged less than sequence length, if current character position i is less than sequence length,
Then execution step 3-5), execution step 3-10 if current character position i is more than or equal to sequence length)；
3-5) take characteristic character of all parameters on i positions and be combined into feature mode t；
3-6) judge feature mode t whether Already in affair character storehouse；If Already in event is special for pattern t
In levying storehouse, then execution step 3-7), feature mode t is not present in affair character storehouse, then execution step 3-8)；
3-7) occurrence number of feature mode t is added 1；
3-8) occurrence number of feature mode t is set to 1；
3-9) current character position i is added 1；Then step 3-4 is gone to) above-mentioned circulation is repeated until current character position
I is put more than or equal to sequence length；
The frequency of occurrences of all patterns 3-10) is updated, i.e., with the occurrence number of the pattern divided by sequence length；
3-11) output result, update event feature database content；Generate affair character storehouse process to terminate.
The detection of described detection determination module judges to comprise the following steps：
4-1) take data of all parameters in current window；
4-2) take characteristic vector of the parameter current in current window；
4-3) calculate Euclidean distance d of the parameter attribute vector and each cluster center vector in the parameter information storehousei；
4-4) ask for diIn minimum of a value d；
D 4-5) is judged whether less than given threshold value δ, δ=0.2)；If d<δ, then execution step 4-6), otherwise
Execution step 4-7)；
Wherein, h (w) represents the characteristic character of current window w, and T represents the set of the parameter attribute character, diRepresent w with
Certain cluster center vector ciBetween Euclidean distance, δ represents given threshold value；
4-6) parameter current is apart from the characteristic character corresponding to d in the characteristic character of current window；
4-7) parameter current be designated as in the characteristic character of current window "”；
4-8) judge whether all parameters have been processed；If parameter it is untreated it is complete if execution step 4-9), if institute
There is parameter all to process, then execution step 4-10)；
4-9) take next parameter；Then go to step and repeat 4-2) to 4-8) all processed to all parameters；
4-10) all parameters are combined into a feature mode V in the characteristic character of current window；
Whether 4-11) judge in affair character storehouse comprising current signature pattern V；
If 4-11) not including current signature pattern V, execution step 4-12 in affair character storehouse), if event is special
Levy and contained in storehouse current signature pattern V, then execution step 4-14)
4-12) current window occurs in that a unexpected event N, and its feature mode is V；
The relevant information of the unexpected event N 4-13) is recorded and exports, unexpected fault detect decision process terminates；
4-14) current window occurs in that known event K, i.e. event corresponding to feature database middle mold Formula V；
4-14) export the relevant information of known event K；It is unexpected that fault detect decision process terminate.
Relative to prior art, the present invention can learn to expand affair character storehouse automatically using Data Mining Strategy, as long as
There is mass data to can be achieved with self-adapting type study, it is not necessary to manual intervention.Fault mode is carried out with the combination of characteristic character
High abstraction, is capable of detecting when new unknown unexpected event/fault mode, but also while have recorded new events/event
The corresponding feature mode of barrier.The inventive method not only provides detailed features information and structural information to analyze new failure, and
And also make the ability that system possesses automatic augmented features storehouse.
Description of the drawings
Fig. 1 is this method module frame figure.
Fig. 2 is the flow chart of generation parameter information bank.
Fig. 3 is the flow chart for generating affair character storehouse.
Fig. 4 is the flow chart for detecting determination module.
Fig. 5 is an example parameter information bank.
Fig. 6 is an example event feature database.
Fig. 7 is the example results that real-time detection judgement is carried out to new data.
Specific embodiment
Present invention achieves a kind of unexpected fault detection method of time series data, the method learns number using historical data
According to essential characteristic and combinations of features pattern and it is saved in information bank and feature database, then feature is extracted to data to be tested,
And check its combinations of features whether with existing patterns match in feature database.If feature mode to be detected and existing pattern in storehouse
Matching, then what data to be tested occurred is exactly known event, has otherwise been detected by unexpected event.Unexpected event is exactly
Potential unexpected failure, typically reflects system operation exception or failure.
With reference to Fig. 1, realizing the system of the present invention includes data preprocessing module 1-1, TK-Means cluster module 1-2, spy
Storehouse generation module 1-3, detection determination module 1-4 are levied, it is comprised the concrete steps that：
1) first, elimination of burst noise carried out to initial data using data preprocessing module 1-1, file processed at equal intervals and is returned
One change is processed, and obtains standardized data, and extracts related multiple features vector；
" elimination of burst noise process " refers to the invalid outlier of deletion, retains virtual value；Data are processed at equal intervals and is ensured even
Time interval in the continuous time period between any two data point is identical；Data are normalized place after processing at equal intervals
Reason eliminates impact of the different dimensions to result so as to obtain standardized data.
2) secondly, characteristic vector is clustered using TK-Means cluster module 1-2, is then based on cluster result, will
Initial data is expressed as feature string；
Using application number：2015105516228, the digging system and method for a kind of time series data off-note, wherein,
' a ' represents the feature of most frequent appearance, and ' b ' represents the secondary feature for frequently occurring, and the rest may be inferred；Probability of occurrence is less than given threshold value
0.02 feature be represented as '’；Empty data, i.e., the interval or deleted interval of non-record data, then with ' # ' come table
Show.
3) and then, it is basic in data prediction and cluster process to each parameter using feature database generation module 1-3
Data recorded in parameter information storehouse, including maximum, minimum of a value, the cluster center vector feelings corresponding with characteristic character of each parameter
Condition, then generating one is used to record the occurrence number of all parameter combination features and the affair character storehouse of frequency；
Referring to Fig. 2, the foundation in described parameter information storehouse includes：
2-1) take a pending parameter；
The record of the parameter 2-2) is created in parameter information storehouse；
2-3) record the maximum and minimum value tag of the parameter；
2-4) record the mean parameter, variance, wavelet coefficient, fourier coefficient, the maximum of frequecy characteristic and minimum of a value；
2-5) obtain the result of the parameter TK-Means cluster；
2-6) cluster result records each cluster center vector of the parameter and its corresponding characteristic character；
2-7) judge currently whether processed all of parameter；If untreated complete all parameters, step 2-1 weight is gone to
Said process is performed again；If having processed all parameters, execution step 2-8)；
2-8) output result, undated parameter information bank content；Generation parameter information bank process terminates；
Referring to Fig. 3, process is set up in described affair character storehouse：
The characteristic character sequence of all parameters of 3-1) aliging；
3-2) recording feature storehouse statistics beginning and ending time, window number, parameter composition；
Current character position i 3-3) is put for 0；
Whether current character position i 3-4) is judged less than sequence length, if current character position i is less than sequence length,
Then execution step 3-5), execution step 3-10 if current character position i is more than or equal to sequence length)；
3-5) take characteristic character of all parameters on i positions and be combined into feature mode t；
3-6) judge feature mode t whether Already in affair character storehouse；If Already in event is special for pattern t
In levying storehouse, then execution step 3-7), feature mode t is not present in affair character storehouse, then execution step 3-8)；
3-7) occurrence number of feature mode t is added 1；
3-8) occurrence number of feature mode t is set to 1；
3-9) current character position i is added 1；Then step 3-4 is gone to) above-mentioned circulation is repeated until current character position
I is put more than or equal to sequence length；
The frequency of occurrences of all patterns 3-10) is updated, i.e., with the occurrence number of the pattern divided by sequence length；
3-11) output result, update event feature database content；Generate affair character storehouse process to terminate.
4) it is last, its characteristic vector is obtained by detecting that determination module 1-4 is real-time to new data, and it is multiple to synchronization
The combination of parameter attribute is contrasted with data in feature database, if the assemblage characteristic has been recorded in feature database, then just
It is known mode, is otherwise exactly to detect unexpected pattern, unexpected pattern is exactly potential unexpected failure.
Referring to Fig. 4, detect that the detection of determination module 1-4 judges to comprise the following steps：
4-1) take data of all parameters in current window；
4-2) take characteristic vector of the parameter current in current window；
4-3) calculate Euclidean distance d of the parameter attribute vector and each cluster center vector in the parameter information storehousei；
4-4) ask for diIn minimum of a value d；
D 4-5) is judged whether less than given threshold value δ, δ=0.2)；If d<δ, then execution step 4-6), otherwise perform step
Rapid 4-7)；
Wherein, h (w) represents the characteristic character of current window w, and T represents the set of the parameter attribute character, diRepresent w with
Certain cluster center vector ciBetween Euclidean distance, δ represents given threshold value；
4-6) parameter current is apart from the characteristic character corresponding to d in the characteristic character of current window；
4-7) parameter current be designated as in the characteristic character of current window "”；
4-8) judge whether all parameters have been processed；If parameter it is untreated it is complete if execution step 4-9), if institute
There is parameter all to process, then execution step 4-10)；
4-9) take next parameter；Then go to step and repeat 4-2) to 4-8) all processed to all parameters；
4-10) all parameters are combined into a feature mode V in the characteristic character of current window；
Whether 4-11) judge in affair character storehouse comprising current signature pattern V；
If 4-11) not including current signature pattern V, execution step 4-12 in affair character storehouse), if event is special
Levy and contained in storehouse current signature pattern V, then execution step 4-14)
4-12) current window occurs in that a unexpected event N, and its feature mode is V；
The relevant information of the unexpected event N 4-13) is recorded and exports, unexpected fault detect decision process terminates；
4-14) current window occurs in that known event K, i.e. event corresponding to feature database middle mold Formula V；
4-14) export the relevant information of known event K；It is unexpected that fault detect decision process terminate.
With reference to Fig. 5, it is a parameter information storehouse example.In figure 0 data below represent the parameter maximum and
Minimum of a value is interval, for normalized.1 data below represents the maximum and minimum that the parameter extracts characteristic vector
Value is interval, for the normalized of characteristic vector.2 data below represent the characteristic character that the Parameter Clustering is generated, and
Cluster center vector corresponding to it, the i.e. characteristic vector of this feature character.
With reference to Fig. 6, it is an affair character storehouse example.The combination feelings of multiple parameters characteristic character are have recorded in the storehouse
Condition, in the number of times occurred comprising window sum, time started, end time, combinations of features pattern and feature mode and frequency etc.
Hold.
With reference to Fig. 7, it is the example results that real-time detection judgement is carried out to new data.Including, combinations of features occur when
Between, combinations of features mode contents, this feature combination be known event pattern or unknown event schema.If there is known event mould
Formula, then show the number of times and probability of event schema appearance.If there is unknown event schema, report finds unexpected event.
Claims (5)
1. the unexpected fault detection method of a kind of time series data, it is characterised in that：The system for realizing the method is located in advance including data
Reason module (1-1), TK-Means cluster modules (1-2), feature database generation module (1-3), detection determination module (1-4), its tool
Body step is；
1) first, using data preprocessing module (1-1) elimination of burst noise is carried out to initial data, file and normalizing is processed at equal intervals
Change is processed, and obtains standardized data, and extracts related multiple features vector；
2) secondly, characteristic vector is clustered using TK-Means cluster modules (1-2), cluster result is then based on, by original
Beginning data are expressed as feature string；
3) and then, using basic number of the feature database generation module (1-3) to each parameter in data prediction and cluster process
According to recorded in parameter information storehouse, including maximum, minimum of a value, the cluster center vector feelings corresponding with characteristic character of each parameter
Condition, then generating one is used to record the occurrence number of all parameter combination features and the affair character storehouse of frequency；
4) it is last, its characteristic vector is obtained by detecting that determination module (1-4) is real-time to new data, and to the multiple ginsengs of synchronization
The combination of number feature is contrasted with data in feature database, if the assemblage characteristic has been recorded in feature database, then be exactly
Known mode, is otherwise exactly to detect unexpected pattern, and unexpected pattern is exactly potential unexpected failure.
2. the unexpected fault detection method of time series data according to claim 1, it is characterised in that：Described data prediction
Module (1-1) " elimination of burst noise process " refers to the invalid outlier of deletion, retains virtual value；Data are processed at equal intervals and is ensured even
Time interval in the continuous time period between any two data point is identical；Data are normalized place after processing at equal intervals
Reason eliminates impact of the different dimensions to result so as to obtain standardized data.
3. the unexpected fault detection method of time series data according to claim 1, it is characterised in that：Described TK-Means gathers
Generic module (1-2) adopts application number：2015105516228, the digging system and method for a kind of time series data off-note, its
In, ' a ' represents the feature of most frequent appearance, and ' b ' represents the secondary feature for frequently occurring, and the rest may be inferred；Probability of occurrence is less than given
The feature of threshold value 0.02 be represented as '’；Empty data, i.e., the interval or deleted interval of non-record data, then with ' # ' come
Represent.
4. the unexpected fault detection method of time series data according to claim 1, it is characterised in that：Described feature database is generated
Module (1-3) generates a parameter information storehouse and an affair character storehouse：
The foundation in described parameter information storehouse includes：
2-1) take a pending parameter；
The record of the parameter 2-2) is created in parameter information storehouse；
2-3) record the maximum and minimum value tag of the parameter；
2-4) record the mean parameter, variance, wavelet coefficient, fourier coefficient, the maximum of frequecy characteristic and minimum of a value；
2-5) obtain the result of the parameter TK-Means cluster；
2-6) cluster result records each cluster center vector of the parameter and its corresponding characteristic character；
2-7) judge currently whether processed all of parameter；If untreated complete all parameters, go to step 2-1 and repeat to hold
Row said process；If having processed all parameters, execution step 2-8)；
2-8) output result, undated parameter information bank content；Generation parameter information bank process terminates；
Process is set up in described affair character storehouse：
The characteristic character sequence of all parameters of 3-1) aliging；
3-2) recording feature storehouse statistics beginning and ending time, window number, parameter composition；
Current character position i 3-3) is put for 0；
3-4) judge that current character position i, whether less than sequence length, if current character position i is less than sequence length, holds
Row step 3-5), execution step 3-10 if current character position i is more than or equal to sequence length)；
3-5) take characteristic character of all parameters on i positions and be combined into feature mode t；
3-6) judge feature mode t whether Already in affair character storehouse；If pattern t Already in affair character storehouse
In, then execution step 3-7), feature mode t is not present in affair character storehouse, then execution step 3-8)；
3-7) occurrence number of feature mode t is added 1；
3-8) occurrence number of feature mode t is set to 1；
3-9) current character position i is added 1；Then step 3-4 is gone to) above-mentioned circulation is repeated until current character position i
More than or equal to sequence length；
The frequency of occurrences of all patterns 3-10) is updated, i.e., with the occurrence number of the pattern divided by sequence length；
3-11) output result, update event feature database content；Generate affair character storehouse process to terminate.
5. the unexpected fault detection method of time series data according to claim 1, it is characterised in that：Described detection judges mould
The detection of block (1-4) judges to comprise the following steps：
4-1) take data of all parameters in current window；
4-2) take characteristic vector of the parameter current in current window；
4-3) calculate Euclidean distance d of the parameter attribute vector and each cluster center vector in the parameter information storehousei；
4-4) ask for diIn minimum of a value d；
D 4-5) is judged whether less than given threshold value δ, δ=0.2)；If d<δ, then execution step 4-6), otherwise
Execution step 4-7)；
Wherein, h (w) represents the characteristic character of current window w, and T represents the set of the parameter attribute character, diRepresent w and certain cluster
Center vector ciBetween Euclidean distance, δ represents given threshold value；
4-6) parameter current is apart from the characteristic character corresponding to d in the characteristic character of current window；
4-7) parameter current be designated as in the characteristic character of current window "”；
4-8) judge whether all parameters have been processed；If parameter it is untreated it is complete if execution step 4-9), if all ginsengs
Number has all been processed, then execution step 4-10)；
4-9) take next parameter；Then go to step and repeat 4-2) to 4-8) all processed to all parameters；
4-10) all parameters are combined into a feature mode V in the characteristic character of current window；
Whether 4-11) judge in affair character storehouse comprising current signature pattern V；
If 4-11) not including current signature pattern V, execution step 4-12 in affair character storehouse), if affair character storehouse
In contained current signature pattern V, then execution step 4-14)
4-12) current window occurs in that a unexpected event N, and its feature mode is V；
The relevant information of the unexpected event N 4-13) is recorded and exports, unexpected fault detect decision process terminates；
4-14) current window occurs in that known event K, i.e. event corresponding to feature database middle mold Formula V；
4-14) export the relevant information of known event K；It is unexpected that fault detect decision process terminate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610814543.6A CN106649438A (en) | 2016-09-09 | 2016-09-09 | Time series data unexpected fault detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610814543.6A CN106649438A (en) | 2016-09-09 | 2016-09-09 | Time series data unexpected fault detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106649438A true CN106649438A (en) | 2017-05-10 |
Family
ID=58851989
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610814543.6A Pending CN106649438A (en) | 2016-09-09 | 2016-09-09 | Time series data unexpected fault detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106649438A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052599A (en) * | 2017-12-12 | 2018-05-18 | 清华大学 | A kind of method and apparatus of the time series data storage of supported feature inquiry |
| CN109117374A (en) * | 2018-08-20 | 2019-01-01 | 浪潮电子信息产业股份有限公司 | A kind of method and system of automatic acquisition failure |
| CN110032490A (en) * | 2018-12-28 | 2019-07-19 | ***股份有限公司 | Method and device thereof for detection system exception |
| WO2020019403A1 (en) * | 2018-07-26 | 2020-01-30 | 平安科技（深圳）有限公司 | Electricity consumption abnormality detection method, apparatus and device, and readable storage medium |
| CN113176962A (en) * | 2021-04-14 | 2021-07-27 | 北京中大科慧科技发展有限公司 | Machine room IT equipment fault accurate detection method and system for data center |
| CN115455451A (en) * | 2022-09-23 | 2022-12-09 | 蔡玉娟 | Encrypted data grading decryption method and AI system |
-
2016
- 2016-09-09 CN CN201610814543.6A patent/CN106649438A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108052599A (en) * | 2017-12-12 | 2018-05-18 | 清华大学 | A kind of method and apparatus of the time series data storage of supported feature inquiry |
| WO2020019403A1 (en) * | 2018-07-26 | 2020-01-30 | 平安科技（深圳）有限公司 | Electricity consumption abnormality detection method, apparatus and device, and readable storage medium |
| CN109117374A (en) * | 2018-08-20 | 2019-01-01 | 浪潮电子信息产业股份有限公司 | A kind of method and system of automatic acquisition failure |
| CN109117374B (en) * | 2018-08-20 | 2021-10-22 | 浪潮电子信息产业股份有限公司 | Method and system for automatically acquiring fault |
| CN110032490A (en) * | 2018-12-28 | 2019-07-19 | ***股份有限公司 | Method and device thereof for detection system exception |
| CN113176962A (en) * | 2021-04-14 | 2021-07-27 | 北京中大科慧科技发展有限公司 | Machine room IT equipment fault accurate detection method and system for data center |
| CN113176962B (en) * | 2021-04-14 | 2024-05-07 | 北京中大科慧科技发展有限公司 | Computer room IT equipment fault accurate detection method and system for data center |
| CN115455451A (en) * | 2022-09-23 | 2022-12-09 | 蔡玉娟 | Encrypted data grading decryption method and AI system |
| CN115455451B (en) * | 2022-09-23 | 2023-10-24 | 北京国联视讯信息技术股份有限公司 | Encrypted data hierarchical decryption method and AI system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106649438A (en) | Time series data unexpected fault detection method | |
| CN107070943B (en) | Industrial internet intrusion detection method based on flow characteristic diagram and perceptual hash | |
| WO2018076571A1 (en) | Method and system for detecting abnormal value in lte network | |
| CN103761173A (en) | Log based computer system fault diagnosis method and device | |
| CN112529341B (en) | Drilling well leakage probability prediction method based on naive Bayesian algorithm | |
| CN108062484A (en) | A kind of classification stage division based on data sensitive feature and database metadata | |
| CN109308411B (en) | Method and system for hierarchically detecting software behavior defects based on artificial intelligence decision tree | |
| CN102509001B (en) | Method for automatically removing time sequence data outlier point | |
| CN106600115A (en) | Intelligent operation and maintenance analysis method for enterprise information system | |
| CN109684374B (en) | Method and device for extracting key value pairs of time series data | |
| CN103227734A (en) | Method for detecting abnormity of OpenStack cloud platform | |
| US20160255109A1 (en) | Detection method and apparatus | |
| CN111191720B (en) | Service scene identification method and device and electronic equipment | |
| CN111027615A (en) | Middleware fault early warning method and system based on machine learning | |
| CN106446081B (en) | The method for excavating time series data incidence relation based on variation consistency | |
| CN109740044B (en) | Enterprise transaction early warning method based on time series intelligent prediction | |
| CN106327323A (en) | Bank frequent item mode mining method and bank frequent item mode mining system | |
| CN107239694A (en) | A kind of Android application permissions inference method and device based on user comment | |
| CN106650446A (en) | Identification method and system of malicious program behavior, based on system call | |
| CN110263944A (en) | A kind of multivariable failure prediction method and device | |
| CN110059126B (en) | LKJ abnormal value data-based complex correlation network analysis method and system | |
| CN106598997A (en) | Method and device for computing membership degree of text subject | |
| CN106156179A (en) | A kind of information retrieval method and device | |
| CN108920694B (en) | Short text multi-label classification method and device | |
| WO2024027487A1 (en) | Health degree evaluation method and apparatus based on intelligent operations and maintenance scene |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170510 |