CN108388232A - A kind of operational mode fault monitoring method of crude oil desalting process - Google Patents
A kind of operational mode fault monitoring method of crude oil desalting process Download PDFInfo
- Publication number
- CN108388232A CN108388232A CN201810231182.1A CN201810231182A CN108388232A CN 108388232 A CN108388232 A CN 108388232A CN 201810231182 A CN201810231182 A CN 201810231182A CN 108388232 A CN108388232 A CN 108388232A
- Authority
- CN
- China
- Prior art keywords
- operational mode
- monitoring method
- fault monitoring
- potential function
- sample point
- 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
-
- 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
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
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 present invention provides a kind of operational mode fault monitoring method of crude oil desalting process, belongs to modern process flow industry process monitoring field.This method utilizes the mass data in process operation, it is proposed a kind of fault detection technique based on operational mode, the concept of operational mode is introduced into industrial control field, the operational mode of industrial process is built by feature selecting and feature extraction, it recycles potential function diagnostic method to classify the operational mode of acquisition, judges whether system is in malfunction.Since operational mode has contained more procedural informations, it more can comprehensively and accurately reaction system behavior and operating status, to improve the accuracy rate of modern process industry fault detect, this is of great significance for abundant and evolution control theory, there are important theory value and realistic meaning simultaneously for the competitiveness etc. of the energy consumption of the process of reduction, optimization operating cost and raising product, has wide practical use surely in process industry.
Description
Technical field
The invention belongs to modern process flow industry process monitoring fields, are related to a kind of crude oil desalting process of process flow industry process
Operational mode fault monitoring method, what it is suitable for production fields such as oil, chemical industry, metallurgy, electric power, medicine, food processings is
Mode fault of uniting monitors.
Background technology
Contain a large amount of inorganic salts such as NaCl, CaCl in crude oil2、MgCl2Deng, in the refining process of oil, these chlorinations
The presence of object will lead to blocking and the pollution and a series of problems, such as the reduction of catalyst life of the burn into pipeline of equipment,
Therefore desalting processing is carried out for being essential for petroleum refining process to crude oil.In addition, due to petroleum refining process
It is highly integrated, the failure of desalination processes may lead to the paralysis of entire petroleum refining process, or even cause major accident,
It is therefore desirable to the operating statuses to the process to be monitored.Traditional course monitoring method is divided into three classes:Based on mechanism model
Method, Knowledge based engineering method and the method based on data-driven.Wherein, the fault detection method based on mechanism model relies on
In the accuracy of mechanism model, therefore affect its application effect;Knowledge based engineering fault detection method need not establish system
Accurate mathematical model, it is only necessary to system action be judged according to industrial knowledge and knowhow and carry out fault detect, obtained
It is extensively studied and applies, however the excessive dependence due to such methods to knowledge experience, cause its versatility and transfer ability
It is poor.And based on the method for data-driven compared to it is traditional fault detect is carried out based on mechanism model and knowledge for, neither
The mathematical model of system is needed, does not also depend on industrial knowledge and knowhow, this method is by excavating in industrial process data
Carry out the normal and malfunction of expression system in information, there is preferable versatility.
In numerous methods of the fault detect based on data-driven, studies and what application was most is multivariate statistical procedure prison
The key of control technology, the technology is multivariate projection and multivariate control chart.Wherein multivariate projection process will measure obtained higher-dimension original
The beginning variable space projects in the latent space being made of a small number of pivot variables, at present using more multivariate projection method have it is main at
Analysis, offset minimum binary and Independent component analysis etc..Multivariate control chart is then to be based on this spatial alternation, in new vector
Construction and counting statistics figureofmerit are used for fault detect in space, and the common statistics figureofmerit of tradition has the statistics in projector space
Statistic in amount and residual error space.In fact, no matter multivariate control chart is using which kind of statistics figureofmerit, essence is all a kind of
Discrimination standard or decision-making foundation, according to the relationship of statistics figureofmerit calculated value and statistic norm controlling limit, to current work
Which kind of state (failure is normal) be in and carries out differentiation and decision for industry automated system.
In recent years, it with the continuous development of data mining and machine learning techniques, has emerged many new quick, intelligent
Differentiation and decision-making technique, such as neural network therein and support vector machines, two kinds of machine learning algorithms are much studied
Person is used for the fault detect of industrial automation system.In addition potential function (Potential function) is used as a kind of function admirable
Nonlinear discriminant method, there is training speed rapidly and efficiently and accurately and reliably classification capacity, therefore the present invention utilized
The running mass data of journey proposes a kind of fault detection technique based on operational mode, and the concept of operational mode is introduced work
Industry control field is built the operational mode of industrial process by feature selecting and feature extraction, recycles potential function diagnostic method pair
The operational mode of acquisition is classified, and judges whether system is in malfunction.Since operational mode has contained more processes
Information, more can comprehensively and accurately reaction system behavior and operating status, to improve the standard of modern process industry fault detect
True rate, this is of great significance for abundant and evolution control theory, energy consumption, optimization simultaneously for the process of reduction
Operating cost and the competitiveness etc. for improving product have important theory value and realistic meaning, have surely in process industry wide
General application prospect.
Invention content
The present invention is intended to provide a kind of operational mode fault monitoring method of crude oil desalting process.By the concept of operational mode
Process industry control field is introduced, the operational mode of industrial process is built by feature selecting and feature extraction, recycles gesture letter
Number diagnostic method classifies to the operational mode of acquisition, distinguishes the normal condition and accident condition of production process.
The technical solution adopted by the present invention:
A kind of operational mode fault monitoring method of crude oil desalting process, is divided into three parts, and first part is initial number
According to acquisition and pretreatment, second part is the structure of operational mode vector, and Part III is using potential function diagnostic method to structure
The operational mode vector built is classified;It is as follows:
Step 1:The acquisition and pretreatment of primary data
The primitive character information during crude oil desalting is acquired, by rejecting abnormalities data, filling missing data, corrects mistake
Accidentally data and the mode of alignment of data carry out the pretreatment of data;Primitive character information during acquisition crude oil desalting includes breast
The density equally correlated process variable of agent content, temperature, pressure, Cl contents, water.
Step 2:The structure of operational mode vector
(1) variables choice is carried out to the data after being pre-processed in step 1, pick out it is effective, can reflect system action
With the characteristic of process status, while invalid feature is rejected, tentatively to reduce the feature space dimension of systematic procedure information;Institute
The variables choice stated, the method used can be the method for exhaustion, Monte Carlo analysis, heuristic search method or Lasso regularization methods etc.
Method;
(2) characteristic that step (1) obtains is converted using converter technique, obtains comprehensive characteristics, and build fortune
Row pattern vector X=[X1,X2,…,Xn], it is ensured that the correlation between characteristic is eliminated, reduces systematic procedure information again
Feature space dimension;The converter technique refers to the skills such as pivot analysis, K mean cluster, Bayes's classification or potential function differentiation
Art;
Step 3:Classified to the operational mode vector of structure using potential function diagnostic method
(3) potential function H (X, X are usedm) come express step (2) structure operational mode vector X=[X1,X2,…,Xn], it obtains
With sample point XmCentered on, XmPotential energy distribution situation in surrounding space at any point X, wherein n is the number of sample point, m=
1,2 ... and n } it is m-th of sample point;
Wherein:Belong to ω1The potential energy of class sample point is positive value;ω will be belonged to2The potential energy value of class sample point, which is multiplied by -1, becomes negative
Value;ω1In the area of space of class sample point cluster, to all ω1The potential energy distribution of class sample point is overlapped, and is obtained one " high
Peak ";ω2In the area of space of class sample point cluster, to all ω2The potential energy distribution of class sample point is overlapped, and obtains one
" low ebb ";
(4) initial build potential function H is set0(X)=0, accumulation potential function is updated using iterative algorithm formula (1):
Hk+1(X)=Hk(X)+rk+1H(X,Xm+1) (1)
Wherein, k is iterations, rk+1To correct term coefficient, rk+1Calculation formula such as formula (2):
(5) the accumulation potential function iterative algorithm in step (4) is executed, until accumulation potential function can be just to all sample points
Really classification, obtains final accumulation potential function H (X);
(6) as H (X) < 0, then the operating status of crude oil desalting process is in fault mode;As H (X)>When 0, then crude oil
The operating status of desalination processes is in normal mode.
In the step 1, the mode of rejecting abnormalities data is Pauta criterion, and the mode for filling missing data is mean value
Completion method.
When in the step (2) using pivot analysis technology, lowest accumulated contribution rate is T=80%.
The present invention utilizes the mass data in process operation, proposes a kind of fault detection technique based on operational mode, will
The concept of operational mode introduces industrial control field, and the operational mode of industrial process is built by feature selecting and feature extraction,
It recycles potential function diagnostic method to classify the operational mode of acquisition, judges whether system is in malfunction.Due to operation
Pattern has contained more procedural informations, more can comprehensively and accurately reaction system behavior and operating status, to improve the modern times
The accuracy rate of process industry fault detect, this is of great significance for abundant and evolution control theory, simultaneously for drop
The energy consumption of low process, the competitiveness etc. for optimizing operating cost and improving product have important theory value and reality meaning
Justice has wide practical use surely in process industry.
Description of the drawings
Fig. 1 is implementation steps flow diagram.
Fig. 2 is characterized spatial distribution map.
Fig. 3 is function schematic diagram.
Fig. 4 is accumulation potential function schematic diagram.
Fig. 5 is that crude oil desalting process accumulates potential function front view.
Fig. 6 is crude oil desalting process operation pattern classification figure.
Specific implementation mode
Technical scheme of the present invention is clearly described with reference to technical solution and attached drawing.
Embodiment 1
As shown in Figure 1, specific implementation step and algorithm are as follows:
Step 1:Demulsification agent content, temperature, pressure, Cl contents, the density of water etc. during selected Desalting and Dewatering from Crude Oil
The related process variable procedural information original as system acquires altogether 1680 samples, wherein it is normal in system to have 960
When collect, remaining 720 sample collects in the case where system is in malfunction.According to Pauta criterion, divide
Other normal data set and fault data collection in step 1 carries out abnormality value removing pretreatment, and the missing values generated after rejecting are then
It is filled according to mean value completion method, finally data is standardized.
Step 2:Principal component analysis, setting lowest accumulated contribution rate T=are carried out to the data after being pre-processed in step 1
80%, it selectes the first two comprehensive characteristics and constitutes operational mode, all sample points are as shown in Figure 2 in the distribution of feature space.
Step 3:To the operational mode vector X=[X built1,X2,…,Xn], wherein n is the number of sample point, uses gesture
Function H (X, Xm) express with sample point XmCentered on, the potential energy distribution situation in surrounding space at any point X, wherein m=
1,2 ... and n } it is m-th of sample point.In two class problems, belong to the first kind (ω1Class) sample point potential energy be positive value, can
With by the second class (ω2Class) sample point potential energy value be multiplied by -1 become negative value, in this way in ω1The area of space of class sample point cluster
In, one " peak " will be formed by being overlapped to the potential energy distribution of all sample points, similarly to ω2The potential energy distribution of class sample point
One " low ebb " will be formed after space is overlapped, Fig. 3 is potential function schematic diagram, and Fig. 4 is accumulation potential function schematic diagram.
Step 4:Assuming that initial build potential function is H0(X)=0, accumulation potential function is carried out more using following iterative algorithm
Newly:
Hk+1(X)=Hk(X)+rk+1H(X,Xm+1)
Wherein, k represents iterations, rk+1To correct term coefficient, calculation formula is as follows:
Step 5:The accumulation potential function iterative algorithm in step 4 is executed, until accumulation potential function can to all sample points
Correct classification, obtains final accumulation potential function H (X).Fig. 5 is the front view of crude oil desalting process accumulation potential function.
Step 6:If H (X)<0, then the operating status of crude oil desalting process be in fault mode;If H (X)>0, then crude oil is de-
The operating status of salt process is in normal mode.Fig. 6 is the classification chart of crude oil desalting process operation pattern.
Claims (10)
1. a kind of operational mode fault monitoring method of crude oil desalting process, which is characterized in that steps are as follows:
Step 1:The acquisition and pretreatment of primary data
The primitive character information during crude oil desalting is acquired, by rejecting abnormalities data, filling missing data, corrects error number
According to the pretreatment for carrying out data with the mode of alignment of data;
Step 2:The structure of operational mode vector
(1) data after being pre-processed to step 1 carry out variables choice, pick out it is effective, can reflect system action and process
The characteristic of state, while invalid feature is rejected, tentatively to reduce the feature space dimension of systematic procedure information;
(2) characteristic that step (1) obtains is converted using converter technique, obtains comprehensive characteristics, and build operation mould
Formula vector X=[X1,X2,…,Xn], it is ensured that the correlation between characteristic is eliminated, reduces the feature of systematic procedure information again
Space dimensionality;
Step 3:Classified to the operational mode vector of structure using potential function diagnostic method
(3) potential function H (X, X are usedm) come express step (2) structure operational mode vector X=[X1,X2,…,Xn], it obtains with sample
This XmCentered on, XmPotential energy distribution situation in surrounding space at any point X, wherein n be sample point number, m=1,
2 ... n } it is m-th of sample point;
Wherein, belong to ω1The potential energy of class sample point is positive value, belongs to ω2The potential energy value of class sample point, which is multiplied by -1, becomes negative value;
In ω1In the area of space of class sample point cluster, to all ω1The potential energy distribution of class sample point is overlapped, and obtains one
" peak ";
In ω2In the area of space of class sample point cluster, to all ω2The potential energy distribution of class sample point is overlapped, and obtains one
" low ebb ";
(4) initial build potential function H is set0(X)=0, accumulation potential function is updated using iterative algorithm formula (1):
Hk+1(X)=Hk(X)+rk+1H(X,Xm+1) (1)
Wherein, k is iterations, rk+1To correct term coefficient, rk+1Calculation formula such as formula (2):
(5) the accumulation potential function iterative algorithm in step (4) is executed, until accumulation potential function can correctly divide all sample points
Class obtains final accumulation potential function H (X);
(6) as H (X) < 0, then the operating status of crude oil desalting process is in fault mode;As H (X)>When 0, then crude oil desalting
The operating status of process is in normal mode.
2. operational mode fault monitoring method according to claim 1, which is characterized in that the primitive character packet
Include emulsion content, temperature, pressure, Cl contents, the density of water.
3. operational mode fault monitoring method according to claim 1 or 2, which is characterized in that the rejecting abnormalities data
Mode be Pauta criterion, fill missing data mode be mean value completion method.
4. operational mode fault monitoring method according to claim 1 or 2, which is characterized in that the variables choice is adopted
Method is the method for exhaustion, Monte Carlo analysis, heuristic search method or Lasso regularization methods.
5. operational mode fault monitoring method according to claim 3, which is characterized in that the variables choice uses
Method be the method for exhaustion, Monte Carlo analysis, heuristic search method or Lasso regularization methods.
6. the operational mode fault monitoring method according to claims 1 or 2 or 5, which is characterized in that the converter technique
Refer to that pivot analysis, K mean cluster, Bayes's classification or potential function differentiate.
7. operational mode fault monitoring method according to claim 3, which is characterized in that the converter technique refers to master
Meta analysis, K mean cluster, Bayes's classification or potential function differentiate.
8. operational mode fault monitoring method according to claim 4, which is characterized in that the converter technique refers to master
Meta analysis, K mean cluster, Bayes's classification or potential function differentiate.
9. operational mode fault monitoring method according to claim 6, which is characterized in that when using pivot analysis technology,
Lowest accumulated contribution rate is set as T=80%.
10. operational mode fault monitoring method according to claim 7 or 8, which is characterized in that use pivot analysis technology
When, lowest accumulated contribution rate is set as T=80%.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810231182.1A CN108388232B (en) | 2018-03-20 | 2018-03-20 | Method for monitoring operation mode fault in crude oil desalting process |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810231182.1A CN108388232B (en) | 2018-03-20 | 2018-03-20 | Method for monitoring operation mode fault in crude oil desalting process |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108388232A true CN108388232A (en) | 2018-08-10 |
CN108388232B CN108388232B (en) | 2020-07-24 |
Family
ID=63067827
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810231182.1A Active CN108388232B (en) | 2018-03-20 | 2018-03-20 | Method for monitoring operation mode fault in crude oil desalting process |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108388232B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109270907A (en) * | 2018-10-24 | 2019-01-25 | 中国计量大学 | A kind of process monitoring and method for diagnosing faults based on the decomposition of stratified probability density |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101860347A (en) * | 2010-06-02 | 2010-10-13 | 天津大学 | Stochastic resonance signal recovery method based on signal classification |
CN103335844A (en) * | 2013-06-24 | 2013-10-02 | 中国计量学院 | Fault detection method for adaptive stochastic resonance bearing |
CN105021399A (en) * | 2015-06-26 | 2015-11-04 | 长安大学 | Feature extraction method based on single-channel signal blind-separation rolling bearing |
CN105547717A (en) * | 2015-12-04 | 2016-05-04 | 哈尔滨工程大学 | Diesel engine lubricating system fault diagnosis method based on Bayes network |
CN106019082A (en) * | 2016-05-26 | 2016-10-12 | 上海电力学院 | Fault line detection method for DG-containing power distribution network based on transient zero sequence current |
CN106203325A (en) * | 2016-07-07 | 2016-12-07 | 燕山大学 | Based on strengthening the Weak fault signal characteristic extracting methods cascading multistable accidental resonance |
CN106843195A (en) * | 2017-01-25 | 2017-06-13 | 浙江大学 | Based on the Fault Classification that the integrated semi-supervised Fei Sheer of self adaptation differentiates |
-
2018
- 2018-03-20 CN CN201810231182.1A patent/CN108388232B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101860347A (en) * | 2010-06-02 | 2010-10-13 | 天津大学 | Stochastic resonance signal recovery method based on signal classification |
CN103335844A (en) * | 2013-06-24 | 2013-10-02 | 中国计量学院 | Fault detection method for adaptive stochastic resonance bearing |
CN105021399A (en) * | 2015-06-26 | 2015-11-04 | 长安大学 | Feature extraction method based on single-channel signal blind-separation rolling bearing |
CN105547717A (en) * | 2015-12-04 | 2016-05-04 | 哈尔滨工程大学 | Diesel engine lubricating system fault diagnosis method based on Bayes network |
CN106019082A (en) * | 2016-05-26 | 2016-10-12 | 上海电力学院 | Fault line detection method for DG-containing power distribution network based on transient zero sequence current |
CN106203325A (en) * | 2016-07-07 | 2016-12-07 | 燕山大学 | Based on strengthening the Weak fault signal characteristic extracting methods cascading multistable accidental resonance |
CN106843195A (en) * | 2017-01-25 | 2017-06-13 | 浙江大学 | Based on the Fault Classification that the integrated semi-supervised Fei Sheer of self adaptation differentiates |
Non-Patent Citations (1)
Title |
---|
李志星: "基于强噪声背景下随机共振的微弱故障诊断方法研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109270907A (en) * | 2018-10-24 | 2019-01-25 | 中国计量大学 | A kind of process monitoring and method for diagnosing faults based on the decomposition of stratified probability density |
CN109270907B (en) * | 2018-10-24 | 2020-07-28 | 中国计量大学 | Process monitoring and fault diagnosis method based on hierarchical probability density decomposition |
Also Published As
Publication number | Publication date |
---|---|
CN108388232B (en) | 2020-07-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106555788B (en) | Application based on the deep learning of Fuzzy Processing in hydraulic equipment fault diagnosis | |
CN105279365B (en) | For the method for the sample for learning abnormality detection | |
El-Midany et al. | A proposed framework for control chart pattern recognition in multivariate process using artificial neural networks | |
CN110309886B (en) | Wireless sensor high-dimensional data real-time anomaly detection method based on deep learning | |
CN109740859A (en) | Transformer condition evaluation and system based on Principal Component Analysis and support vector machines | |
CN107272667A (en) | A kind of industrial process fault detection method based on parallel PLS | |
Yacout | Fault detection and diagnosis for condition based maintenance using the logical analysis of data | |
Charbonnier et al. | A weighted dissimilarity index to isolate faults during alarm floods | |
CN111931819A (en) | Machine fault prediction and classification method based on deep learning | |
CN109389325B (en) | Method for evaluating state of electronic transformer of transformer substation based on wavelet neural network | |
CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
CN108090657A (en) | Oil & Gas Storage facility risk assessment based on Xiu Hate control theories and probabilistic neural network manages system and method with on-line early warning | |
CN110009126B (en) | Online alarm analysis method based on fusion of PLS model and PCA contribution degree | |
CN108020781A (en) | A kind of circuit breaker failure diagnostic method | |
CN106950945A (en) | A kind of fault detection method based on dimension changeable type independent component analysis model | |
CN112192318B (en) | Machining tool state monitoring method and system | |
CN111126820A (en) | Electricity stealing prevention method and system | |
Li et al. | Safety control modeling method based on Bayesian network transfer learning for the thickening process of gold hydrometallurgy | |
Aslam et al. | Anomaly detection using explainable random forest for the prediction of undesirable events in oil wells | |
Xu et al. | Wear particle classification using genetic programming evolved features | |
CN111488939A (en) | Model training method, classification method, device and equipment | |
Cuentas et al. | An SVM-GA based monitoring system for pattern recognition of autocorrelated processes | |
CN108388232A (en) | A kind of operational mode fault monitoring method of crude oil desalting process | |
Ren et al. | Spatial-temporal associations representation and application for process monitoring using graph convolution neural network | |
CN112016597B (en) | Depth sampling method based on Bayesian unbalance measurement in machine learning |
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 |