CN108052092A - A kind of subway electromechanical equipment abnormal state detection method based on big data analysis - Google Patents
A kind of subway electromechanical equipment abnormal state detection method based on big data analysis Download PDFInfo
- Publication number
- CN108052092A CN108052092A CN201711371666.8A CN201711371666A CN108052092A CN 108052092 A CN108052092 A CN 108052092A CN 201711371666 A CN201711371666 A CN 201711371666A CN 108052092 A CN108052092 A CN 108052092A
- Authority
- CN
- China
- Prior art keywords
- data
- electromechanical equipment
- parameter
- subway electromechanical
- big data
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
- G05B23/0245—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a qualitative model, e.g. rule based; if-then decisions
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a kind of subway electromechanical equipment abnormal state detection method based on big data analysis, from data evolutionary process, the abnormality detection of the brand-new angle of data correlation realization subway electromechanical equipment.The potential feature of historical data is excavated by time series models and adaptive neural network, and the dynamic rule of data versus time is represented with transition probability sequence.For the monitoring data of multidimensional, simplify the correlativity between each parameter with unsupervised clustering, so as to avoid the problem that correlation is difficult to determine between parameter.It proposes abnormality detection system, and is allowed to be suitable for subway electromechanical device state monitoring data flow, realize Rapid Detection abnormal in data flow.
Description
Technical field
The invention belongs to the neutral net equipment fault technical fields under big data environment, refer specifically to for a kind of big data ring
The inspection of subway electromechanical equipment abnormal state based on self organizing neural network (self-organized maps, SOM) algorithm under border
Survey method.
Background technology
Subway electromechanical equipment can be subject to overload, overvoltage, built-in electrical insulation aging, natural environment in actual moving process
Etc. anomalous events influence, these abnormal operating conditions can cause the generation of equipment deficiency, failure, therefore to subway electromechanical equipment shape
State is carried out abnormality detection with very strong necessity.In the actual O＆M of equipment, the equipment component of triangular web is all based on greatly
Information detects exception using simple threshold values determination method.This traditional threshold determination has limitation, one side equipment letter
It ceases utilization rate and state evaluation accuracy is all relatively low, be on the other hand difficult to detect by the Hidden fault and fault category of equipment,
And the fixed threshold in related specifications standard is difficult to the otherness of bonding apparatus operating condition.
Due to subway electromechanical equipment status data (including on-line monitoring, live detection, the preventive trial data etc.) scale of construction
Greatly, the characteristics of wide variety, big data technology can be introduced into unit exception detection, fully excavates the exception of status data
Information.Nearly 2 years big data technologies are quickly grown in internet, finance, logistics field, embody high social value, and
The starting stage is in track transportation industry big data technology, big data analysis technology passes through the association between finding facility information
Relation provides brand-new solution method and thinking to provide unit exception detection accuracy.
Subway electromechanical equipment state change has complexity, randomness and periodic feature.Monitor subway electromechanical equipment
Abnormal state has the non-linear and uncertain of height, and at the same time correlation is stronger, the common method master of such problem
There are forecast of regression model method and the machine learning method using neutral net as representative.
The autoregression model (auto-regressive, AR) of time series is suitable for many industrial process, its main feature is that AR
Systematic memory is strong, and the behavior at moment the past, this low dynamic phase in equipment running process are depended in the value of time t
Symbol.The variation of subway electromechanical equipment a part of quantity of state in normal course of operation is smaller, such as pulling force, earth current, these
State quantity data belongs to stationary sequence, can be directly fitted with AR；Another part quantity of state is in periodicity change, but changes width
Value is little, such as oil temperature, environment temperature, can be also fitted after removing its date periodicity by AR.Due to diving for subway electromechanical equipment
Volt property fault progression is slow, therefore when equipment is in abnormality, the parameter monitored is often without departing from directive/guide or regulation
Limit value, so as to be difficult to discover.It is single for the online monitoring data without departing from quantity of state limit value according to conclusions
Purely its abnormality can not be detected with AR models.
Neutral net (Neural Networks) is a highly complex non-linear dynamic learning system, as to multiple
Miscellaneous nonlinear system approaches device, has self study, self-organizing and generalization ability, there is very big advantage in prediction field.From group
It is by unsupervised learning method to knit neutral net (self organized maps, SOM) operation principle, allows each god of competition layer
It is matched through member by competing with input pattern, finally only there are one the victor that neuron becomes competition, this acquisition nerves
The input of member just represents the classification to input pattern.Due to not containing desired output in the training sample of unsupervised learning, do not have
Any priori, therefore suitable for data volume is big, the Condition Monitoring Data without label.
The content of the invention
Above-mentioned the deficiencies in the prior art are directed to, it is an object of the invention to provide a kind of subways based on big data analysis
Electromechanical equipment abnormal state detection method has to solve subway electromechanical equipment abnormal state detection method traditional in the prior art
There is a limitation, facility information utilization rate and state evaluation accuracy are all relatively low and be difficult to detect by the Hidden fault of equipment
And the problem of fault category.
In order to achieve the above objectives, the technical solution adopted by the present invention is as follows：
A kind of subway electromechanical equipment abnormal state detection method based on big data analysis of the present invention, including walking as follows
Suddenly：
1) electromechanical equipment status data is extracted from subway electromechanical equipment historical state data storehouse, and is conducted into big data
Storage system；
2) subway electromechanical equipment historical state data is read from above-mentioned big data storage system；
3) above-mentioned subway electromechanical equipment historical state data is substituted into AR models and SOM models is trained, Ran Houjian
Vertical multidimensional equipment operation characteristic parameter；
4) anomalous discrimination is carried out to real-time monitoring of equipment data according to above-mentioned multidimensional equipment operation characteristic parameter.
Preferably, the device history status data extracted in the step 1) includes：Temperature, voltage, electric current, pressure,
Passenger flow data or transaction data.
Preferably, the step 1) specifically further includes：When data are extracted, incremental number is periodically read from database
According to, and write big data storage system.
Preferably, specifically also included in the step 3)：
31) be directed to each parameter historical data, by AR models, SOM algorithms calculate transition probability matrix X1,
X2,…,Xn}；
32) historical data of all parameters is directed to, multidimensional time-series are clustered by DBSCAN algorithms, by history
Data are polymerized to m cluster.
Preferably, specifically included in the step 4)：The electromechanical equipment real-time status data stream of on-line monitoring is substituted into step
It is rapid 31) in transition probability matrix obtain the transition probability sequence of each parameter, and judge whether the data at each time point belong to step
It is rapid 32) in m cluster.
Preferably, the step 4) specifically also includes：
A. when the transition probability sequence of each parameter is all there is no 0 value, and the data at each time point belong to m cluster in data flow
In 1 when, then there is no abnormal for the segment data；
B. it is worth when the transition probability sequence of each parameter exists less than 30, and less than the number in three sampling periods in data flow
According to m cluster is not belonging to, then exists in the segment data less than 3 noise spots, belong to sensor abnormality；
C. when the transition probability sequence of k parameter, there are one section of 0 value, k<N, and more than 3 sampling periods in data flow
Data are not belonging to m cluster, then judge that abnormal operating condition occurs in equipment；
D. to abnormal operating condition in above-mentioned steps c, according to 0 point of the mistake of parameter transition probability sequence, unit exception is judged
The time of origin of state.
Beneficial effects of the present invention：
The present invention is based on self organizing neural network (self-organized maps, SOM) technology, there is self study, adaptive
The advantages that answering, is fault-tolerant under big data storage system endlessly mass data training, establishes equipment state model parameter.
The invention has the advantages that can the historical data and current data of multi-dimensional state amount be combined, and it can realize the abnormal inspection of dynamic data
It surveys, it is higher compared with traditional threshold determination method accuracy rate.
The present invention is directed to the monitoring data of multidimensional, simplifies the correlativity between each parameter with unsupervised clustering,
So as to avoid the problem that correlation is difficult to determine between parameter.It proposes abnormality detection system, and is allowed to be suitable for subway electromechanical equipment
Condition Monitoring Data stream realizes Rapid Detection abnormal in data flow.
Description of the drawings
Fig. 1 is the functional block diagram of electromechanical equipment abnormal state detection method of the present invention.
Fig. 2 is the functional block diagram of the equipment state abnormality detection model based on AR and SOM.
Specific embodiment
For the ease of the understanding of those skilled in the art, the present invention is made further with reference to embodiment and attached drawing
Bright, the content that embodiment refers to not is limitation of the invention.
A kind of subway electromechanical equipment abnormal state detection based on big data analysis shown in reference picture 1, Fig. 2, of the invention
Method applied to the differentiation of subway electromechanical equipment abnormality, includes the following steps：
S1：Electromechanical equipment status data is extracted from subway electromechanical equipment historical state data storehouse, and is conducted into big number
According to storage system；
S2：Subway electromechanical equipment historical state data is read from above-mentioned big data storage system；
S3：For the historical data of each parameter, by AR models, SOM algorithms calculate transition probability matrix X1,
X2,…,Xn}；
S4：For the historical data of all parameters, the multidimensional time-series are clustered by DBSCAN algorithms, will be gone through
History data are polymerized to m cluster；
S5：The real-time stream of on-line monitoring is substituted into above-mentioned transition probability matrix and obtains the transition probability sequence of each parameter
Row, and judge whether the data at each time point belong to m above-mentioned cluster；
Result in S5 carries out abnormality detection to data stream, and abnormal logic detection is as follows：
A. when the transition probability sequence of each parameter is all there is no 0 value, and the data at each time point belong to m cluster in data flow
In 1 when, then there is no abnormal for the segment data；
B. when the transition probability sequence of each parameter is there are 0 value of a few (is less than 3), and a small number of time points in data flow
The data of (be less than three sampling periods) are not belonging to m cluster, then in the segment data there are a few (less than 3) noise spot,
Belong to sensor abnormality；
C. when the transition probability sequence of k parameter, there are one section of 0 value, k<N, and a long time point (is more than 3 in data flow
A sampling period) data be not belonging to m cluster, then judge that abnormal operating condition occurs in equipment；
D. to abnormal operating condition in above-mentioned steps c, according to 0 point of the mistake of parameter transition probability sequence, unit exception is judged
The time of origin of state.
It is as follows to electromechanical equipment status data AR Model of First fitting formula in the above method：
Wherein, xtFor the time series of online monitoring data；etFor normal distribution sequence, therefore, xtObey N (μ, σ2)
Normal distribution, wherein μ and σ meet following relation：
μ=μe/(1-α) (2)
σ2=(α2μ2+λ2+μe 2)/(1-α2) (3)
In the above method, the probability density function belonging to neuron represents as follows：God is represented by single order transition probability P
Through the correlativity between member, the single order transition probability in AR (n) models between neuron is P [Ct+1|, Ct..., Ct-n+1], by
This single order transition probability that can obtain AR (1) model is P [Ct+1|Ct].Due to representing x in formula (1)tNormal Distribution, then xt's
Probability-distribution function available standards normal distyribution functions represents, xtDistribution function represent：
In formula, a=(CI+CI+1)/2And b=(CI+CI-1)/2；
Sequence C={ C1, C2..., CNOutput node as SOM；
In the above method, the transition probability between neuron represents as follows:
In the above method, the problem of for multidimensional Parameter Fusion, the present invention passes through density-based algorithms
(density-based spatial clustering of applications with noise, DBSCAN) is to multidimensional
Online monitoring data is clustered, the characteristics of online monitoring data amount can be made full use of big and by the complexity between each parameter
Correlativity simplifies.
There are many concrete application approach of the present invention, and the above is only the preferred embodiment of the present invention, it is noted that for
For those skilled in the art, without departing from the principle of the present invention, several improvement can also be made, this
A little improve also should be regarded as protection scope of the present invention.
Claims (6)
1. a kind of subway electromechanical equipment abnormal state detection method based on big data analysis, which is characterized in that including walking as follows
Suddenly：
1) electromechanical equipment status data is extracted from subway electromechanical equipment historical state data storehouse, and is conducted into big data storage
System；
2) subway electromechanical equipment historical state data is read from above-mentioned big data storage system；
3) above-mentioned subway electromechanical equipment historical state data is substituted into AR models and SOM models is trained, then established more
Tie up equipment operation characteristic parameter；
4) anomalous discrimination is carried out to real-time monitoring of equipment data according to above-mentioned multidimensional equipment operation characteristic parameter.
2. the subway electromechanical equipment abnormal state detection method according to claim 1 based on big data analysis, feature
It is, the device history status data extracted in the step 1) includes：Temperature, voltage, electric current, pressure, passenger flow data or
Transaction data.
3. the subway electromechanical equipment abnormal state detection method according to claim 1 or 2 based on big data analysis, special
Sign is that the step 1) specifically further includes：When data are extracted, incremental data is periodically read from database, and is write
Big data storage system.
4. the subway electromechanical equipment abnormal state detection method according to claim 1 based on big data analysis, feature
It is, is specifically also included in the step 3)：
31) be directed to each parameter historical data, by AR models, SOM algorithms calculate transition probability matrix X1, X2 ...,
Xn}；
32) historical data of all parameters is directed to, multidimensional time-series are clustered by DBSCAN algorithms, by historical data
It is polymerized to m cluster.
5. the subway electromechanical equipment abnormal state detection method according to claim 4 based on big data analysis, feature
It is, is specifically included in the step 4)：The transfer electromechanical equipment real-time stream of on-line monitoring substituted into step 31) is general
Rate matrix obtains the transition probability sequence of each parameter, and judges that the m whether data at each time point belong in step 32) is a poly-
Class.
6. the subway electromechanical equipment abnormal state detection method according to claim 5 based on big data analysis, feature
It is, the step 4) specifically also includes：
A. when the transition probability sequence of each parameter is all there is no 0 value, and the data at each time point belong in m cluster in data flow
At 1, then there is no abnormal for the segment data；
B. it is worth when the transition probability sequence of each parameter exists less than 30, and the data in three sampling periods is less than not in data flow
Belong to m cluster, then exist in the segment data less than 3 noise spots, belong to sensor abnormality；
C. when the transition probability sequence of k parameter, there are one section of 0 value, k<N, and more than the data in 3 sampling periods in data flow
M cluster is not belonging to, then judges that abnormal operating condition occurs in equipment；
D. to abnormal operating condition in above-mentioned steps c, according to 0 point of the mistake of parameter transition probability sequence, unit exception state is judged
Time of origin.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711371666.8A CN108052092B (en) | 2017-12-19 | 2017-12-19 | Subway electromechanical equipment state anomaly detection method based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711371666.8A CN108052092B (en) | 2017-12-19 | 2017-12-19 | Subway electromechanical equipment state anomaly detection method based on big data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108052092A true CN108052092A (en) | 2018-05-18 |
| CN108052092B CN108052092B (en) | 2020-02-21 |
Family
ID=62133800
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711371666.8A Active CN108052092B (en) | 2017-12-19 | 2017-12-19 | Subway electromechanical equipment state anomaly detection method based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108052092B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108666915A (en) * | 2018-05-22 | 2018-10-16 | 江苏容源电力设备有限公司 | Buried change electric appliance security maintenance method |
| CN110502724A (en) * | 2019-08-09 | 2019-11-26 | 国网山西省电力公司 | Equipment state prediction method and terminal device based on self organizing neural network |
| CN110738255A (en) * | 2019-10-15 | 2020-01-31 | 和尘自仪(嘉兴)科技有限公司 | device state monitoring method based on clustering algorithm |
| CN111768082A (en) * | 2020-06-02 | 2020-10-13 | 广东电网有限责任公司 | Power equipment state evaluation method based on big data analysis |
| CN112131069A (en) * | 2019-06-24 | 2020-12-25 | 中船重工特种设备有限责任公司 | Equipment operation monitoring method and system based on clustering |
| CN113034733A (en) * | 2021-03-29 | 2021-06-25 | 南京格物智能科技有限公司 | Detection method and detection device for blockage of subway traction motor filter screen |
| CN117171596A (en) * | 2023-11-02 | 2023-12-05 | 宝鸡市兴宇腾测控设备有限公司 | Online monitoring method and system for pressure transmitter |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103714383A (en) * | 2014-01-09 | 2014-04-09 | 北京泰乐德信息技术有限公司 | Rail transit fault diagnosis method and system based on rough set |
| CN103914735A (en) * | 2014-04-17 | 2014-07-09 | 北京泰乐德信息技术有限公司 | Failure recognition method and system based on neural network self-learning |
| CN104091070A (en) * | 2014-07-07 | 2014-10-08 | 北京泰乐德信息技术有限公司 | Rail transit fault diagnosis method and system based on time series analysis |
| CN104317778A (en) * | 2014-10-30 | 2015-01-28 | 国家电网公司 | Massive monitoring data based substation equipment fault diagnosis method |
-
2017
- 2017-12-19 CN CN201711371666.8A patent/CN108052092B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103714383A (en) * | 2014-01-09 | 2014-04-09 | 北京泰乐德信息技术有限公司 | Rail transit fault diagnosis method and system based on rough set |
| CN103914735A (en) * | 2014-04-17 | 2014-07-09 | 北京泰乐德信息技术有限公司 | Failure recognition method and system based on neural network self-learning |
| CN104091070A (en) * | 2014-07-07 | 2014-10-08 | 北京泰乐德信息技术有限公司 | Rail transit fault diagnosis method and system based on time series analysis |
| CN104317778A (en) * | 2014-10-30 | 2015-01-28 | 国家电网公司 | Massive monitoring data based substation equipment fault diagnosis method |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108666915A (en) * | 2018-05-22 | 2018-10-16 | 江苏容源电力设备有限公司 | Buried change electric appliance security maintenance method |
| CN112131069A (en) * | 2019-06-24 | 2020-12-25 | 中船重工特种设备有限责任公司 | Equipment operation monitoring method and system based on clustering |
| CN112131069B (en) * | 2019-06-24 | 2023-07-21 | 中船重工特种设备有限责任公司 | Equipment operation monitoring method and system based on clustering |
| CN110502724A (en) * | 2019-08-09 | 2019-11-26 | 国网山西省电力公司 | Equipment state prediction method and terminal device based on self organizing neural network |
| CN110502724B (en) * | 2019-08-09 | 2023-04-07 | 国网山西省电力公司 | Equipment state prediction method based on self-organizing neural network and terminal equipment |
| CN110738255A (en) * | 2019-10-15 | 2020-01-31 | 和尘自仪(嘉兴)科技有限公司 | device state monitoring method based on clustering algorithm |
| CN111768082A (en) * | 2020-06-02 | 2020-10-13 | 广东电网有限责任公司 | Power equipment state evaluation method based on big data analysis |
| CN113034733A (en) * | 2021-03-29 | 2021-06-25 | 南京格物智能科技有限公司 | Detection method and detection device for blockage of subway traction motor filter screen |
| CN117171596A (en) * | 2023-11-02 | 2023-12-05 | 宝鸡市兴宇腾测控设备有限公司 | Online monitoring method and system for pressure transmitter |
| CN117171596B (en) * | 2023-11-02 | 2024-01-23 | 宝鸡市兴宇腾测控设备有限公司 | Online monitoring method and system for pressure transmitter |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108052092B (en) | 2020-02-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108052092A (en) | A kind of subway electromechanical equipment abnormal state detection method based on big data analysis | |
| Hsieh et al. | Unsupervised online anomaly detection on multivariate sensing time series data for smart manufacturing | |
| Malhotra et al. | Long short term memory networks for anomaly detection in time series. | |
| Syamala et al. | Machine learning-integrated IoT-based smart home energy management system | |
| CN108418841A (en) | Next-generation key message infrastructure network Security Situation Awareness Systems based on AI | |
| CN108897954A (en) | Wind turbines temperature pre-warning method and its system based on BootStrap confidence calculations | |
| CN104657792A (en) | Early warning method based on rule engine and intelligent prediction | |
| Yang et al. | Autoregressive coefficient-invariant control chart pattern recognition in autocorrelated manufacturing processes using neural network ensemble | |
| CN113360679B (en) | Fault diagnosis method based on knowledge graph technology | |
| Li et al. | Internet of things assisted condition‐based support for smart manufacturing industry using learning technique | |
| CN102013148A (en) | Multi-information fusion fire hazard detection method | |
| CN111382494A (en) | System and method for detecting anomalies in sensory data of industrial machines | |
| Su et al. | Nonlinear compensation algorithm for multidimensional temporal data: A missing value imputation for the power grid applications | |
| Qin et al. | Remaining useful life prediction for rotating machinery based on optimal degradation indicator | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| Tang et al. | Prediction of bearing performance degradation with bottleneck feature based on LSTM network | |
| Sun et al. | Lightweight bidirectional long short-term memory based on automated model pruning with application to bearing remaining useful life prediction | |
| CN115481726A (en) | Industrial robot complete machine health assessment method and system | |
| Filios et al. | An agnostic data-driven approach to predict stoppages of industrial packing machine in near future | |
| CN113984389A (en) | Rolling bearing fault diagnosis method based on multi-receptive-field and improved capsule map neural network | |
| Fan et al. | Luad: A lightweight unsupervised anomaly detection scheme for multivariate time series data | |
| Bhoomika et al. | Time series forecasting and point anomaly detection of sensor signals using lstm neural network architectures | |
| Merkt | Predictive models for maintenance optimization: an analytical literature survey of industrial maintenance strategies | |
| Raghuvamsi et al. | A novel deep learning architecture for distribution system topology identification with missing PMU measurements | |
| Sun et al. | A data-driven framework for tunnel infrastructure maintenance |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |