CN106548191A - Continuous process fault detection method based on collection nucleation locality preserving projections - Google Patents
Continuous process fault detection method based on collection nucleation locality preserving projections Download PDFInfo
- Publication number
- CN106548191A CN106548191A CN201610847885.8A CN201610847885A CN106548191A CN 106548191 A CN106548191 A CN 106548191A CN 201610847885 A CN201610847885 A CN 201610847885A CN 106548191 A CN106548191 A CN 106548191A
- Authority
- CN
- China
- Prior art keywords
- data
- statistic
- continuous process
- detection method
- fault detection
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Probability & Statistics with Applications (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
Continuous process fault detection method based on collection nucleation locality preserving projections, it is related to a kind of continuous process fault detection method with regard to kernel parameter selection, including using the historical data under normal condition as modeling data training set, kernel mapping is carried out to modeling data using gaussian kernel function, non-linear continuous process modeling and fault detect is carried out with EKLPP methods.First training data matrix is standardized.Nuclear matrix is converted thereof into KLPP methods and replaces data to be tested, each testing result is changed into using Bayesian decision-makings the Probability Forms for breaking down, it is whether whether normal to judge the time data more than control limit according to integrated statistic.If integrated statistic exceedes control limit, the time data is failure.If testing indicate that system malfunctions, need staff to find out situation, remove the dangerous condition.Present invention efficiently solves for different failures, the problem that applicable parameter also can be different.
Description
Technical field
The present invention relates to a kind of non-linear procedure failure testing method, more particularly to a kind of to process different faults nuclear parameter
The partial fault detection method of selection.
Background technology
Chemical process increasingly embodies nonlinear feature as one of industrial key areas, tradition
Global fault's detection method can only retain the global information of initial data, destroy its partial structurtes, how effectively to extract
In production process, the local message of initial data carries out monitoring to non-linear process becomes the important interior of fault detection technique research
Hold.
Core locality preserving projections method has obvious advantage for the partial structurtes problem for processing holding nonlinear data.
But, people would generally use gaussian kernel function when using traditional core locality preserving projections method processing data, and this causes
The performance of the method is subject to nuclear parametercImpact.So far, the foundation that the selection of nuclear parameter does not determine how is carried out,
Generally chosen using empirical method.It is obvious that the performance of fault detect depends on the selection of nuclear parameter.If therefore
Barrier is known, then the nuclear parameter of the suitable failure of selection carries out fault detect and can be obtained by optimum detection result.But production
During failure be not usually unique, it is faulty that the nuclear parameter for thus selecting is not necessarily applied to institute, for different
Failure, applicable parameter can be significantly different.Therefore, find a kind of effective method nuclear parameter is reasonably selected to seem outstanding
For important.
The content of the invention
It is an object of the invention to provide a kind of based on collection nucleation locality preserving projections (Ensemble Kernel
Locality Preserving Projections, EKLPP) continuous process fault detection method.The method effectively can be carried
The partial structurtes information of original nonlinear data in industrial processes is taken, by choosing the gaussian kernel function with different parameters
Setting up multiple core locality preserving projections (Kernel Locality Preserving Projections, KLPP) model carries out event
Barrier detection, the Probability Forms for being converted into breaking down by the testing result of each model using Bayesian decision-makings, then by integrated
Each testing result is combined by learning method, effectively solves the problems, such as kernel parameter selection.
The purpose of the present invention is achieved through the following technical solutions：
Based on the continuous process fault detection method of collection nucleation locality preserving projections, methods described includes procedure below：
Fault detect is carried out using traditional KLPP methods and would generally be subject to nuclear parametercImpact, collect nucleation locality preserving projections
The purpose of method is exactly to choose the nuclear parameter suitable for different faults.First, the history number under the conditions of collection normal production operation
According to collection as training sample, training sample is standardized.Then, choose the gaussian kernel function with different nuclear parameters
Kernel mapping is carried out to the data matrix after standardization, the nuclear matrix of training data is obtained.Each nuclear moment tried to achieve with training data
Battle array is replaced initial data and carries out LPP projections in nuclear space respectively, obtains projection matrix, sets up a series of sub- KLPP models.Most
Afterwards, the SPE statistics and T of each submodel are calculated2Statistic, determines the control of two statistics of each submodel using Density Estimator
System limit.Batch to be detected is being standardized with after kernel mapping, projects to respectively and multiple inspections are obtained on different submodels
Survey result.Recycle Bayesian decision-makings that each testing result is changed into the Probability Forms for breaking down.Finally, by integrated
Each testing result is combined by learning method, and by by itself and control limitαIt is compared to carry out fault detect.
The described continuous process fault detection method based on collection nucleation locality preserving projections, the modeling process are just included
Normal state model and Fault Model.For the gaussian kernel function that the data decimation in industrial process has different nuclear parameters enters
Row kernel mapping, replaces primitive modeling data using the nuclear matrix after conversion.Multiple models are set up with LPP, and calculates different moulds
The control limit of the statistic and statistic of type.
The described continuous process fault detection method based on collection nucleation locality preserving projections, the Fault Model will
Newly arrive and enter line translation with different kernel functions after time data pretreatment, by the data projection after conversion to different sub- KLPP
Its statistic testing result is calculated on model, each testing result is processed using Bayesian decision-makings and integrated study method,
With control limitαFault detect is carried out relatively.
Advantages of the present invention with effect is：
1. the present invention effectively can carry out Nonlinear feature extraction and dimensionality reduction to initial data.Due to the production spy of continuous process
Point, most of data all have non-linear.Nonlinear data is processed with collection nucleation locality preserving projections method, first
Initial data is projected to by higher dimensional space by nonlinear mapping, then recycles locality preserving projections method to carry out information retrieval
And dimensionality reduction.
2. the present invention effectively maintains the partial structurtes of initial data, improves the utilization rate of effective information.Traditional
Global Algorithm would generally cause data neighborhood structural damage, the method for the present invention to belong to local algorithm.It is suitable by choosing
Projecting direction causes this adjacency information matrix of preferably can keeping intact between the sample after dimensionality reduction.In simple terms, for
Former sample concentrates adjacent sample still adjacent in the projected, the sample distant with former sample collection still keep in the projected compared with
Remote distance, effectively maintains the partial structurtes of initial data.
3. the present invention solves Kernel-parameter Selection Problem of traditional KLPP methods when different faults are processed.By choosing
Gaussian kernel function with different nuclear parameters is set up multiple submodels and carries out fault detect, using Bayesian decision-makings by each detection
As a result the Probability Forms for breaking down are changed into, each testing result is combined by integrated study method, is efficiently solved
For different failures, the problem that applicable parameter also can be different.
Description of the drawings
Fig. 1 is the core algorithm flow chart of the present invention.
Specific embodiment
With reference to embodiment, the present invention is described in detail.
The present invention is that continuous process data are carried out with pretreatment with gaussian kernel function, extracts the non-linear letter of initial data
Breath.On the basis of pretreatment, the partial structurtes of initial data are kept using locality preserving projections.By choosing with different IPs
The gaussian kernel function of parameter solves the problems, such as that parameter selects to affect failure detection result, sets up multiple submodels, utilizes
Bayesian decision-makings and integrated study method will carry out continuous process fault detect after the combination of each testing result.This technology solves biography
Kernel parameter selection identical problem of the system KLPP methods when different faults are processed.
Fault detection technique based on collection nucleation locality preserving projections：In order to carry out procedure fault detection, need using
Then new data are detected by the normal data modeling known.The present invention is using the normal historical data for collecting as building
The training set of modulus evidence, is modeled using EKLPP methods and fault detect.After training set is carried out pretreatment, core is constructed
Centralization matrix, carries out extracting the non-linear local message of modeling data with KLPP methods.Choose a series of with different IPs
The gaussian kernel function of parameter sets up multiple KLPP submodels, and calculates the statistic and its control limit of each model.For new-comer
After moment sample carries out pretreatment, kernel mapping is carried out to which with KLPP methods, by the matrix projection after conversion to each submodel
On, each testing result is changed into the Probability Forms for breaking down using Bayesian decision-makings, by integrated study method by each inspection
Survey result to be combined, whether the control limit for whether exceeding modeling according to integrated statistic is normal to judge the data at the moment.
Software system：In order to realize fault detect, MALTAB software programming of the present invention using MathWorks companies
Exploitation, the continuous process data to collecting in industrial process are detected, when the statistic of testing data is prescribed a time limit more than control,
Then the time data is failure, i.e. system malfunctions, needs staff to find out situation in time, removes the dangerous condition.
The present invention is made up of following two parts：Normal condition model and Fault Model.It is normal in industrial process
Historical data, extracts the non-linear local message of initial data by KLPP.Choose a series of Gausses with different nuclear parameters
Kernel function carries out kernel mapping, replaces primitive modeling data using the data after conversion, sets up multiple submodels with KLPP methods,
The control limit of each two statistic of submodel is determined using Density Estimator.
For time data of newly arriving, after pretreatment is carried out, non-linear local message is extracted first with KLPP and obtain core change
Matrix is changed, kernel matrix is projected to and on each submodel, is obtained statistic testing result, using Bayesian decision-makings by each inspection
Survey result and change into the Probability Forms for breaking down, each testing result is combined by integrated study method, and by itself and control
System limitαIt is compared, it is whether whether normal to judge the time data more than control limit according to integrated statistic.If integrated system
More than control limit, then the time data is failure for metering；Otherwise it is normal, realizes based on collection nucleation locality preserving projections
Continuous process fault detect.
The present invention extracts the local message of nonlinear data using core locality preserving projections, and maximized utilization data have
Effect information.At the same time, Kernel-parameter Selection Problem of traditional KLPP methods when different faults are processed is solved, non-thread is improve
Property procedure fault detection degree of accuracy.
Claims (3)
1. based on collection nucleation locality preserving projections continuous process fault detection method, it is characterised in that methods described include with
Lower process：
Using the normal data for collecting in process of production as the training set of modeling data, place is standardized to training data
Reason, extracts the non-linear local message of modeling data using KLPP methods；Choose a series of gaussian kernel with different nuclear parameters
Function, sets up multiple submodels with KLPP methods, and calculates the control limit of the statistic and statistic of each model；To newly arriving
Moment sample carry out pretreatment after, convert thereof into nuclear matrix with KLPP methods and replace data to be tested, and project to each
On individual submodel, each testing result is changed into the Probability Forms for breaking down using Bayesian decision-makings, by integrated study
Each testing result is combined by method, and by which and controls limitαBe compared, according to integrated statistic whether exceed control limit come
Judge whether the time data is normal；If integrated statistic exceedes control limit, the time data is failure；Just otherwise it is
Normal；If testing indicate that system malfunctions, then need staff to find out situation in time, remove the dangerous condition.
2. according to claim 1 based on the continuous process fault detection method for collecting nucleation locality preserving projections, its feature
It is that described modeling includes normal condition model and Fault Model；For the data in industrial process, selection has
The gaussian kernel function of different nuclear parameters carries out kernel mapping, replaces primitive modeling data using the data after conversion；Set up with LPP
Multiple models, and calculate the control limit of the statistic and statistic of different models.
3. according to claim 1 based on the continuous process fault detection method for collecting nucleation locality preserving projections, its feature
It is that the Fault Model is that time data of newly arriving carries out entering line translation with different kernel functions again after pretreatment,
Its statistic testing result will be calculated on data projection after conversion to different sub- KLPP models, using Bayesian decision-makings and
Integrated study method carries out fault detect after each testing result is combined.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610847885.8A CN106548191B (en) | 2016-09-23 | 2016-09-23 | Continuous process fault detection method based on collection nucleation locality preserving projections |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610847885.8A CN106548191B (en) | 2016-09-23 | 2016-09-23 | Continuous process fault detection method based on collection nucleation locality preserving projections |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106548191A true CN106548191A (en) | 2017-03-29 |
| CN106548191B CN106548191B (en) | 2019-04-05 |
Family
ID=58369444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610847885.8A Expired - Fee Related CN106548191B (en) | 2016-09-23 | 2016-09-23 | Continuous process fault detection method based on collection nucleation locality preserving projections |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106548191B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765587A (en) * | 2019-09-30 | 2020-02-07 | 北京化工大学 | Complex petrochemical process fault diagnosis method based on dynamic regularization judgment local retention projection |
| CN114611606A (en) * | 2022-03-07 | 2022-06-10 | 安徽理工大学 | Fault detection method based on nuclear hybrid space projection |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104483962A (en) * | 2014-11-20 | 2015-04-01 | 沈阳化工大学 | Intermittent process online fault detection method based on multi-direction orthogonal locality preserving projections |
| CN105511445A (en) * | 2015-12-01 | 2016-04-20 | 沈阳化工大学 | Multi-modal process fault detection method based on local neighbor standardization matrix |
-
2016
- 2016-09-23 CN CN201610847885.8A patent/CN106548191B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104483962A (en) * | 2014-11-20 | 2015-04-01 | 沈阳化工大学 | Intermittent process online fault detection method based on multi-direction orthogonal locality preserving projections |
| CN105511445A (en) * | 2015-12-01 | 2016-04-20 | 沈阳化工大学 | Multi-modal process fault detection method based on local neighbor standardization matrix |
Non-Patent Citations (3)
| Title |
|---|
| DENG XIAOGANG 等: "Sparse Kernel Locality Preserving Projection and Its Application in Nonlinear Process Fault Detection", 《CHINESE JOURNAL OF CHEMICAL ENGINEERING》 * |
| LIJIA LUO 等: "Nonlinear process monitoring based on kernel global-local preserving projections", 《JOURNAL OF PROCESS CONTROL》 * |
| 韩敏 等: "基于改进核主成分分析的故障检测与诊断方法", 《化 工 学 报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765587A (en) * | 2019-09-30 | 2020-02-07 | 北京化工大学 | Complex petrochemical process fault diagnosis method based on dynamic regularization judgment local retention projection |
| CN114611606A (en) * | 2022-03-07 | 2022-06-10 | 安徽理工大学 | Fault detection method based on nuclear hybrid space projection |
| CN114611606B (en) * | 2022-03-07 | 2024-06-18 | 安徽理工大学 | Fault detection method based on nuclear hybrid space projection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106548191B (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mohammadpourfard et al. | A statistical unsupervised method against false data injection attacks: A visualization-based approach | |
| CN110909811A (en) | OCSVM (online charging management system) -based power grid abnormal behavior detection and analysis method and system | |
| Wilson et al. | Deep learning-aided cyber-attack detection in power transmission systems | |
| Karimipour et al. | Intelligent anomaly detection for large-scale smart grids | |
| CN107358347A (en) | Equipment cluster health state evaluation method based on industrial big data | |
| CN106341414A (en) | Bayesian network-based multi-step attack security situation assessment method | |
| CN106888205A (en) | A kind of non-intrusion type is based on the PLC method for detecting abnormality of power consumption analysis | |
| CN109501834A (en) | A kind of point machine failure prediction method and device | |
| CN101771582B (en) | Safety monitoring correlation analysis method based on state machine | |
| CN109522192A (en) | A kind of prediction technique of knowledge based map and complex network combination | |
| CN109470946B (en) | Power generation equipment fault detection method and system | |
| CN106203519A (en) | Fault pre-alarming algorithm based on taxonomic clustering | |
| CN106850558A (en) | Intelligent electric meter abnormal state detection method based on seaconal model time series | |
| CN104486406A (en) | Layered resource monitoring method based on cloud data center | |
| US11709486B2 (en) | Fault signal recovery system and method | |
| CN112632535A (en) | Attack detection method and device, electronic equipment and storage medium | |
| CN106548191A (en) | Continuous process fault detection method based on collection nucleation locality preserving projections | |
| CN106292499A (en) | The safe related function method and device of secondary equipment of intelligent converting station O&M operation object | |
| CN102981096A (en) | Power grid fault identification method based on WAMS (Wide Area Measurement System) time sequence decomposition | |
| CN106338977A (en) | Nonlinear process fault detection method based on differential locality preserving projection (DLPP) | |
| CN112285484B (en) | Power system fault diagnosis information fusion method and device based on deep neural network | |
| CN103529337A (en) | Method for recognizing nonlinear correlation between equipment failures and electric quantity information | |
| CN107423156A (en) | Fault pre-alarming algorithm based on taxonomic clustering | |
| CN106054859B (en) | The double-deck integrated form industrial process fault detection method based on amendment type independent component analysis | |
| CN112083275A (en) | Distribution network fault type identification method and system |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190405 Termination date: 20210923 |