CN110377007A - Fault Diagnosis for Chemical Process method based on pivot analysis - Google Patents
Fault Diagnosis for Chemical Process method based on pivot analysis Download PDFInfo
- Publication number
- CN110377007A CN110377007A CN201910664045.1A CN201910664045A CN110377007A CN 110377007 A CN110377007 A CN 110377007A CN 201910664045 A CN201910664045 A CN 201910664045A CN 110377007 A CN110377007 A CN 110377007A
- Authority
- CN
- China
- Prior art keywords
- new
- statistic
- data
- singular value
- matrix
- 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
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Automatic Analysis And Handling Materials Therefor (AREA)
Abstract
The Fault Diagnosis for Chemical Process method based on pivot analysis that the invention discloses a kind of, include the following steps: to decompose the industrial process sample data matrix under treated nominal situation, obtain singular value matrix, later, probability principal component model is constructed to singular value matrix E, and single statistic statistical analysis process is established, meanwhile, obtain statistic boundary；The sample data in industrial process under real-time running state is acquired, and calculates corresponding statistic STnew；Later, by STnewIt is compared with the statistic boundary calculated, if STnewAbove statistics boundary, then determine to break down in operational process.The Fault Diagnosis for Chemical Process method based on pivot analysis solves existing principle component analysis and is difficult to handle hiding data, is difficult to obtain the defect of accurate conclusion, simultaneously, compared to existing Probabilistic Principal Component Analysis method, the application can preferably handle the rough error data occurred in chemical process.
Description
Technical field
The present invention relates to Fault Diagnosis for Chemical Process fields, specifically provide a kind of chemical process event based on pivot analysis
Hinder diagnostic method.
Background technique
With deepening continuously for the fault detection method research based on data-driven, traditional Multivariable Statistical Process Control
Method shows certain defect, is only capable of realizing that malfunction monitoring accurately and timely, such as pivot analysis are only capable of under given conditions
Detection process data obey linear, Gaussian Profile process；Independent pivot is more sensitive for nongausian process.
Therefore, a kind of new Fault Diagnosis for Chemical Process method based on pivot analysis is studied, to improve fault diagnosis rate,
As people's urgent problem to be solved.
Summary of the invention
In consideration of it, the purpose of the present invention is to provide a kind of Fault Diagnosis for Chemical Process method based on pivot analysis, with
Defect present in traditional pivot analysis and Probabilistic Principal Component Analysis is solved, the accuracy rate of fault detection is increased.
Present invention provide the technical scheme that the Fault Diagnosis for Chemical Process method based on pivot analysis, including walk as follows
It is rapid:
S1: the industrial process sample data under nominal situation is handled, and by sparse optimal model to data
Matrix is decomposed, and singular value matrix is obtained, and later, constructs probability principal component model to singular value matrix E, and establish single statistics
Statistical analysis process is measured, meanwhile, obtain statistic boundary；
S2: the sample data in acquisition industrial process under real-time running state, and calculate corresponding statistic STnew；
S3: by ST obtained in S2newWith statistic boundary obtained in S1It is compared, if STnewIn statistical circles
LimitTop then determines to break down in operational process, and alarms immediately, otherwise, continues to execute S2.
It is preferred that S1 includes the following steps:
S11: the industrial process sample data under acquisition nominal situation constructs training dataset X ∈ Rm×n, wherein m was
Journey sample number, n are process variable number；
S12: collected sample data is normalized, the data matrix Y after being normalized, wherein return
The data mean value in data matrix Y after one change is 0, variance 1；
S13: the data matrix Y after normalization is decomposed by sparse optimal model, obtains singular value matrix E；
S14: W, σ of singular value matrix E are calculated by EM algorithm2, and to singular value matrix E building as shown in formula (1)
Probability principal component model:
E=Wy+ μ+ε (1)
In formula, y is to hide vector, and W is factor loading matrix, and μ indicates the mean value of singular value matrix E, ε Representative errors item,
Wherein, y and ε Gaussian distributed, y meet standardized normal distribution, i.e. y~N (0,1), error term ε obey the normal state that variance is σ
Distribution, i.e. ε~N (0, σ2I)；
S15: according to statistic boundary of formula (2) the Counting statistics amount ST in the case where confidence level is α
In formula,For statistic boundary, yiFor the hiding vector in collecting sample, n is process variable number.
Further preferably, S2 includes the following steps:
S21: the sample data in acquisition industrial process under real-time running state；
S22: collected data are done into normalized, the sample data after being normalized；
S23: the sample data after normalization is decomposed by sparse optimal model, obtains singular value matrix Enew,
And singular value matrix E is calculated by EM algorithmnewWnew、σnew 2、μnew；
S24: being calculated using the probability principal component model that has constructed in S1 and hide vector y, later, using formula (2) calculate with
Its corresponding statistic STnew, it is used for malfunction monitoring.
Fault Diagnosis for Chemical Process method provided by the invention based on pivot analysis, including model and online is established offline
Fault detection two large divisions, wherein establish in model process offline, firstly, extracting affine space by sparse optimal model
Interior unusual Value Data, secondly, probability principal component model is established on unusual Value Data, finally, establishing single statistic statistical
Analysis process can avoid so that obeying a kind of measurement according to the statistic of the probability principal component model of singular value matrix building because of volume
Degree leads to the problem of more results and can not obtain accurate conclusion, meanwhile, hiding data, online fault detection have been handled well
Cheng Zhong, firstly, the sample data in acquisition industrial process under real-time running state, and calculate corresponding statistic STnew,
Later, by STnewIt is compared with the statistic boundary calculated, if STnewAbove statistics boundary, then operational process is determined
Fault detection is realized in middle failure.
Fault Diagnosis for Chemical Process method provided by the invention based on pivot analysis, has the beneficial effect that: not only solving
Existing principal component model is difficult to handle the defect of hiding data, also solves the statistic of existing principal component model because obeying two amounts
It is difficult to obtain the defect of accurate conclusion when the two conflicts caused by degree, meanwhile, compared to existing Probabilistic Principal Component Analysis
Method can preferably handle the rough error data occurred in chemical process.
Specific embodiment
The present invention is further explained below in conjunction with specific embodiment, but the not limitation present invention.
The Fault Diagnosis for Chemical Process method based on pivot analysis that the present invention provides a kind of, includes the following steps:
S1: the industrial process sample data under nominal situation is handled, and by sparse optimal model to data
Matrix is decomposed, and singular value matrix is obtained, and later, constructs probability principal component model to singular value matrix E, and establish single statistics
Statistical analysis process is measured, meanwhile, obtain statistic boundary；
S11: the industrial process sample data under acquisition nominal situation constructs training dataset X ∈ Rm×n, wherein m was
Journey sample number, n are process variable number；
S12: collected sample data is normalized, the data matrix Y after being normalized, wherein return
The data mean value in data matrix Y after one change is 0, variance 1；
S13: the data matrix Y after normalization is decomposed by sparse optimal model, obtains singular value matrix E；
S14: W, σ of singular value matrix E are calculated by EM algorithm2, and to singular value matrix E building as shown in formula (1)
Probability principal component model:
E=Wy+ μ+ε (1)
In formula, y is to hide vector, and W is factor loading matrix, and μ indicates the mean value of singular value matrix E, ε Representative errors item,
Wherein, y and ε Gaussian distributed, y meet standardized normal distribution, i.e. y~N (0,1), error term ε obey the normal state that variance is σ
Distribution, i.e. ε~N (0, σ2I)；
S15: according to statistic boundary of formula (2) the Counting statistics amount ST in the case where confidence level is α
In formula,For statistic boundary, yiFor the hiding vector in collecting sample, n is process variable number；
In the step, firstly, the unusual Value Data in affine space is extracted by sparse optimal model, secondly, in surprise
Probability principal component model is established on different Value Data, finally, establishing single statistic statistical analysis process, not only solves existing pivot
Model is difficult to handle the defect of hiding data, caused by also solving the statistic of existing principal component model because obeying two amounts degree
It is difficult to obtain the defect of accurate conclusion when the two conflicts, meanwhile, it, can be more preferable compared to existing Probabilistic Principal Component Analysis method
Processing chemical process in the rough error data that occur.
S2: the sample data in acquisition industrial process under real-time running state, and calculate corresponding statistic STnew；
S21: the sample data in acquisition industrial process under real-time running state；
S22: collected data are done into normalized, the sample data after being normalized；
S23: the sample data after normalization is decomposed by sparse optimal model, obtains singular value matrix Enew,
And singular value matrix E is calculated by EM algorithmnewWnew、σnew 2、μnew；
S24: being calculated using the probability principal component model that has constructed in S1 and hide vector y, later, using formula (2) calculate with
Its corresponding statistic STnew, it is used for malfunction monitoring；
S3: by ST obtained in S2newWith statistic boundary obtained in S1It is compared, if STnewIn statistical circles
LimitTop is (i.e.:), then determine to break down in operational process, and alarm immediately, otherwise, continues to hold
Row S2.
Embodiments of the present invention are elaborated above, but present invention is not limited to the embodiments described above,
Those of ordinary skill in the art within the scope of knowledge, can also make various without departing from the purpose of the present invention
Variation.
Claims (3)
1. the Fault Diagnosis for Chemical Process method based on pivot analysis, which comprises the steps of:
S1: the industrial process sample data under nominal situation is handled, and by sparse optimal model to data matrix
It is decomposed, obtains singular value matrix, later, probability principal component model is constructed to singular value matrix E, and establish single statistic system
Analytic process is counted, meanwhile, obtain statistic boundary；
S2: the sample data in acquisition industrial process under real-time running state, and calculate corresponding statistic STnew；
S3: by ST obtained in S2newWith statistic boundary obtained in S1It is compared, if STnewIn statistics boundaryTop then determines to break down in operational process, and alarms immediately, otherwise, continues to execute S2.
2. the Fault Diagnosis for Chemical Process method described in accordance with the claim 1 based on pivot analysis, it is characterised in that: S1 includes
Following steps:
S11: the industrial process sample data under acquisition nominal situation constructs training dataset X ∈ Rm×n, wherein m is process sample
This number, n are process variable number；
S12: collected sample data is normalized, the data matrix Y after being normalized, wherein normalization
The data mean value in data matrix Y afterwards is 0, variance 1；
S13: the data matrix Y after normalization is decomposed by sparse optimal model, obtains singular value matrix E；
S14: W, σ of singular value matrix E are calculated by EM algorithm2, and the probability as shown in formula (1) is constructed to singular value matrix E
Principal component model:
E=Wy+ μ+ε (1)
In formula, y is to hide vector, and W is factor loading matrix, and μ indicates the mean value of singular value matrix E, ε Representative errors item, wherein
Y and ε Gaussian distributed, y meet standardized normal distribution, i.e. y~N (0,1), error term ε obey the normal distribution that variance is σ,
That is ε~N (0, σ2I)；
S15: according to statistic boundary of formula (2) the Counting statistics amount ST in the case where confidence level is α
In formula,For statistic boundary, yiFor the hiding vector in collecting sample, n is process variable number.
3. the Fault Diagnosis for Chemical Process method based on pivot analysis according to claim 2, it is characterised in that: S2 includes
Following steps:
S21: the sample data in acquisition industrial process under real-time running state；
S22: collected data are done into normalized, the sample data after being normalized；
S23: the sample data after normalization is decomposed by sparse optimal model, obtains singular value matrix Enew, and lead to
It crosses EM algorithm and calculates singular value matrix EnewWnew、σnew 2、μnew；
S24: being calculated using the probability principal component model constructed in S1 and hide vector y, and later, it is right with it to be calculated using formula (2)
The statistic ST answerednew, it is used for malfunction monitoring.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910664045.1A CN110377007A (en) | 2019-07-23 | 2019-07-23 | Fault Diagnosis for Chemical Process method based on pivot analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910664045.1A CN110377007A (en) | 2019-07-23 | 2019-07-23 | Fault Diagnosis for Chemical Process method based on pivot analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110377007A true CN110377007A (en) | 2019-10-25 |
Family
ID=68254925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910664045.1A Pending CN110377007A (en) | 2019-07-23 | 2019-07-23 | Fault Diagnosis for Chemical Process method based on pivot analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110377007A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112051839A (en) * | 2020-09-17 | 2020-12-08 | 中国计量大学 | Process monitoring and fault diagnosis method based on tree structure sparsity |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104793604A (en) * | 2015-04-10 | 2015-07-22 | 浙江大学 | Principal component tracking based industrial fault monitoring method and application thereof |
-
2019
- 2019-07-23 CN CN201910664045.1A patent/CN110377007A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104793604A (en) * | 2015-04-10 | 2015-07-22 | 浙江大学 | Principal component tracking based industrial fault monitoring method and application thereof |
Non-Patent Citations (1)
| Title |
|---|
| 马超: "基于稀疏表示和概率主元分析的化工过程故障检测与识别", 《中国优秀硕士学位论文全文数据库 工程科技I辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112051839A (en) * | 2020-09-17 | 2020-12-08 | 中国计量大学 | Process monitoring and fault diagnosis method based on tree structure sparsity |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109459993A (en) | A kind of process flow industry process online adaptive Fault monitoring and diagnosis method | |
| Rotem et al. | Ethylene compressor monitoring using model‐based PCA | |
| CN100480926C (en) | Industrial process fault diagnosis system and method based on wavelet analysis | |
| CN108665119B (en) | Water supply pipe network abnormal working condition early warning method | |
| JP6606003B2 (en) | Process diagnostic apparatus, process diagnostic method, and computer program | |
| CN112414694B (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN113344133B (en) | Method and system for detecting abnormal fluctuation of time sequence behaviors | |
| CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
| CN105334823B (en) | Industrial process fault detection method based on the linear dynamic system model for having supervision | |
| WO2021114320A1 (en) | Wastewater treatment process fault monitoring method using oica-rnn fusion model | |
| CN103926919B (en) | Industrial process fault detection method based on wavelet transformation and Lasso function | |
| US20220027230A1 (en) | Predicting Equipment Fail Mode from Process Trace | |
| CN111340110A (en) | Fault early warning method based on industrial process running state trend analysis | |
| CN108257365A (en) | A kind of industrial alarm designs method based on global nonspecific evidence dynamic fusion | |
| CN109298633A (en) | Chemical production process fault monitoring method based on adaptive piecemeal Non-negative Matrix Factorization | |
| CN112904810A (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
| CN115561416A (en) | Method and device for detecting inlet water quality of sewage treatment plant in real time | |
| CN114580572B (en) | Abnormal value identification method and device, electronic equipment and storage medium | |
| CN109978059B (en) | Early warning method for tower flushing faults of primary distillation tower in crude oil distillation process | |
| CN110377007A (en) | Fault Diagnosis for Chemical Process method based on pivot analysis | |
| CN201017232Y (en) | Industry process non-linearity failure diagnosis device based on fisher | |
| CN114295162A (en) | Environmental monitoring system based on data acquisition | |
| CN201035376Y (en) | Failure diagnosis device under small sample conditional in the process of manufacturing production | |
| CN201017233Y (en) | Manufacturing production process failure diagnosis device based on wavelet analyzing | |
| CN112288126B (en) | Sampling data abnormal change online monitoring and diagnosing method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191025 |