CN101446827B - Process fault analysis device of process industry system and method therefor - Google Patents
Process fault analysis device of process industry system and method therefor Download PDFInfo
- Publication number
- CN101446827B CN101446827B CN2008102321321A CN200810232132A CN101446827B CN 101446827 B CN101446827 B CN 101446827B CN 2008102321321 A CN2008102321321 A CN 2008102321321A CN 200810232132 A CN200810232132 A CN 200810232132A CN 101446827 B CN101446827 B CN 101446827B
- Authority
- CN
- China
- Prior art keywords
- analysis
- failure
- information
- node
- fault
- 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.)
- Expired - Fee Related
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
Landscapes
- Test And Diagnosis Of Digital Computers (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to a process industry system fault analysis device based on complex network theories and a method therefor. The device comprises a system structure information base, a man-machine interaction module, a process data management module, a system network characteristics analysis module and a process fault analysis module. The process fault analysis device and the method provided by the invention can be employed for identifying the key parts of the process system, reducing the process monitoring variables and solving the problem that the selection of monitoring point position depends on people's knowledge or experience; meanwhile, the invention makes the best of real-time data information of the process industry process system, thereby ensuring monitoring over the process fault is more instantaneous and accurate; and the utilization of the system domain knowledge improves the fault recognition and separation capability of the traditional PCA monitoring method.
Description
Technical field
The present invention relates to process industry system, the process fault analysis device and the method for particularly a kind of process industry system based on Complex Networks Theory.
Background technology
In process industry, because the continuous expansion of industrial process scale, the increase day by day of complicacy, the security of production system and reliability requirement also improve day by day, and production run safety, reliable, non-fault ground stable operation have become a vital task of modern industry.For this reason, in system's operational process, need in time detect fault or unusual generation, and fault type is judged and source of trouble location, elimination adverse effect factor.Traditional process exception monitoring method can be divided three classes: based on resolve, based on knowledge and based on the method for data-driven.Be based upon on the strict mathematics model basis based on the method for resolving, as Kalman filter, parameter estimation, methods such as equivalent space; Method based on knowledge is mainly set up according to qualutative model, as fault tree (FTA), decision tree (DT) etc.; Based on the method for data mainly is process data to gather, excavates the information that implies in the data by various data processing and analytical approach, and then instructs production run, as multivariate statistical method, cluster analysis, spectrum analysis etc.Because no matter the production system in the process industry is a whole factory or an independent productive unit, all is a big system, obtain relatively difficulty of strict mathematics model and detailed systematic knowledge, therefore, based on resolve and be restricted based on the method for knowledge; And, there is a large amount of instrument and meters in the typical modern process, big system can produce lot of data, so obtained using widely based on aspects such as exception monitoring in process industry of the analytical approach of data-driven, fault diagnosises.
Yet, in actual applications, the matter of utmost importance that exists based on the method for data-driven is exactly the selection problem of data monitoring point, i.e. hundreds of monitoring site to existing in the system, which some position is necessary with important to process monitoring, in fact, the technician only pays close attention to important monitoring site in the part experience understanding usually; Moreover owing to do not use the procedures system domain knowledge based on the method for data-driven, identification that it is unusual and separating power will be restricted (though it is unusual to know that certain variable produces, being difficult to determine its basic reason); In addition,,, will produce more noise information undoubtedly, increase the weight of the computation burden of malfunction monitoring and analysis if all processes variable is monitored to large-scale complicated system.Therefore, how to select to reflect procedures system monitored parameters, reduce the monitored parameters number, the identification and the separating power that improve the fault of process have significant application value for safety, the even running of assurance process industry system.
In fact, the production system in the process industry is transmitted by various media (fluid, electric power, signal), and discrete apparatus is connected into a complex electromechanical systems network interrelated, that highly be coupled.There is correlativity between the data variable in the system, has strong coupling, on global configuration, present complicated latticed form.Therefore, the major technique thinking of support of the present invention is: utilize numerous variablees to have the characteristics of latticed form, and the topological property that situational variables has at network, and on this basis important process variable is selected; On this basis, in conjunction with traditional data-driven method,, the systematic procedure fault is monitored and discerned as pca method (PCA).Based on above technical thought, a kind of process industry system trouble analysis device and method that the applicant proposes based on Complex Networks Theory, can solve the problems referred to above that face in the complex process system trouble analysis, reduce and optimizing process monitored parameters number, improve monitoring, the separating power of the system failure.
Summary of the invention
The object of the present invention is to provide a kind of process industry systematic procedure fail analysis device and method based on Complex Networks Theory, so that can select to numerous variablees that large-scale process industry system exists, can discern the crucial significant variable that influences system failure monitoring, reduce process monitoring variable number; Simultaneously, make full use of real time data information in the process flow industry process system, coupling system structure connection information, failure mode information etc. improve unusually or the separation recognition capability of fault.
In order to realize above-mentioned task, the present invention takes following technical solution:
A kind of process industry systematic procedure fail analysis device based on Complex Networks Theory, this device comprises:
(1) human-computer interaction module, be used to realize the mutual of user and process fault analysis system, import the relevant information of analyzed object, comprise contact details and analysis result etc. between technological process information, process measuring point variable information, variable, and can revise system architecture information bank, process data and fault analysis flow process.
(2) process data administration module is used for the real time data that system's operational process produces is stored and managed, and comprises DC S control system data and artificial spot check data etc.
(3) system architecture information bank is used to store, the data structuring model relevant information of management system node structure information and fault mode;
(4) grid specificity analysis module is used for procedures system is carried out network struction and system is carried out the network characteristic analysis.
(5) process fault analysis module, the implementation system trouble analysis, and the workflow of analyzing managed;
Process data administration module of the present invention, system architecture information bank, grid specificity analysis module are connected with the process fault analysis module respectively, and human-computer interaction module is connected with process fault analysis module, system architecture information bank respectively.
Information spinner in the described system architecture information bank will comprise: the failure cause in neighbour's matrix of descriptive system network model, network topology characteristic and the fault mode, associated nodes, associated nodes sign, the order of severity and maintenance measure information, and the failure mode information employing is the storage mode of introducer with the associated nodes sign.
Above-mentioned process industry systematic procedure failure analysis methods based on Complex Networks Theory specifically may further comprise the steps:
Step 1: process industry grid specificity analysis.According to system's industrial flow information, process measuring point variable information etc. procedures system is carried out STRUCTURE DECOMPOSITION, with monitored parameters meshed network formal description system, by system is carried out the network characteristic analysis, find out the key variables node set that to system dynamics behavior has material impact according to network characteristic; The menu of key variables node is present: if a carries out status monitoring to these key nodes, then the information of obtaining can reflect the entire system state; B, if key node is in expectation state, then entire system also is in expectation state.Obtain the data structuring model of system architecture information, this data structuring model comprises node descriptor, failure mode information etc., and is stored in the system architecture information bank.
Step 2: based on system's pivot analysis (PCA) model construction and the monitoring of key node collection.Key node collection constructing system data owner meta analysis (PCA) monitoring model that utilizes data in the procedures system even running and analysis to obtain utilizes the real time data of process operation that the systematic procedure fault is monitored.
Step 3: obtain variable node collection (Fault Sets-FS) and qualitative failure-description information thereof; If break down, then obtain variable node set (Fault Sets-FS) and qualitative failure-description information thereof with failure symptom;
Step 4: based on the failure reason analysis of associated nodes sign with the observation sign; After obtaining failure symptom, the fault mode that contrast has been set up can carry out failure reason analysis, determines maintenance measure.
Step 5: analysis result is shown by human-computer interaction module.
Described process industry grid specificity analysis comprises the steps:
(a) according to system's industrial flow information, process measuring point variable information procedures system is carried out STRUCTURE DECOMPOSITION,, and adopt neighbour's matrix form to be stored in the system architecture information bank with monitored parameters meshed network formal description system;
(b) by neighbour's matrix is calculated according to the network characteristic definition, obtain cluster coefficient, feature path, the node number of degrees, node Jie and count the network characteristic value, and be stored in the system architecture information bank, for providing foundation from network perspective recognition system and subsequent analysis;
(c) determine node importance degree w according to network characteristic
i, obtain the key variables set of node of reflection entire system behavior, this information stores is in the system architecture information bank;
(d) failure mode information is obtained; According to historical failure data establishing system failure mode information, comprise failure cause, associated nodes, associated nodes sign, the order of severity and maintenance measure.
The described pivot analysis PCA of system monitoring method based on the key node collection, at first, the key variables collection that utilizes the process data under the stable state non-fault situation and obtain, the principal component model under the structure normal condition, promptly construct pivot and get sub matrix and pivot matrix of loadings, E is a residual matrix; Then, for a new sample, construct corresponding statistic T
2With Q and corresponding control limit threshold values thereof
And Q
αThe behavior of monitoring system process.
Based on the failure cause analysis method of associated nodes sign with the observation sign, according to the associated nodes sign information in the fault mode, determine the initial reason set, explain the failure symptom that all observes if there is a kind of reason, then this reason is the final analysis result, and single failure may take place illustrative system; If can not all explain the observation sign, then record explanation sign is counted the reason of nES maximum as one of final possible cause; Remaining reason set is repeated above step, enough all explain the failure symptom of observation up to the energy collecting of gained reason.
Adopt of the present invention based on Complex Networks Theory process industry systematic procedure fail analysis device and method can the identification process system key position, reduce the process monitoring variable, solved the defective that the knowledge that relies on the people in the past or experience are selected monitoring site; Simultaneously, make full use of real time data information in the process flow industry process system, make that the monitoring of procedure fault is timely more and accurate; Moreover, owing to utilized system's domain knowledge, improved the Fault Identification and the separating power of conventional P CA monitoring method.
Description of drawings
Fig. 1 is the structural representation of device of the present invention.
Fig. 2 is the data structuring model figure of system architecture information bank;
Fig. 3 is a workflow diagram of the present invention;
Fig. 4 is the fault analysis synoptic diagram;
Fig. 5 is an embodiment of the invention subsystem process flow diagram;
Fig. 6 is this grid illustraton of model (part);
The present invention is described in further detail below in conjunction with accompanying drawing.
Embodiment
At first notions such as the Complex Networks Theory that relates among the present invention, network topology characteristic, node strength of association are done following simple introduction and definition:
Pivot analysis (principal component analysis): be called for short PCA, it is a kind of technology commonly used in the process monitoring, the basic thought of pca method is exactly under the situation that keeps procedural information variable quantity as much as possible, the manifold that the variable that has correlativity each other by is formed is carried out dimensionality reduction to obtain the process of mutual each other incoherent characteristic signal (pivot signal), promptly use the dynamic change of the pivot characterization process data matrix of less dimensions.It is by the process statistics amount T of structure based on process pivot characteristic signal subspace information
2With the statistic Q of residual information subspace information, determine its control limit, and then implementation procedure is carried out process monitoring.When monitoring system exception,, carry out fault diagnosis by contrasting existing fault mode by making up contribution plot or residual identification fault variable.
Traditional PCA method has the following disadvantages in actual applications: a, and the monitoring performance depends on the quality of process data, and too much variable not only increases computation burden, also introduces noise information easily, need select process variable; B, a little less than Fault Identification and the diagnosis capability, though contribution plot or residual can provide and the closely-related variable of fault, changing big variable is not failure cause usually, therefore also needs technician's professional knowledge.The present invention will propose a covering device and method and overcome above problem on the basis of Complex Networks Theory.
Complex Networks Theory (complex network theory): a large amount of complication system that exists in the real world, if ignore true individual concrete physical significance in the system, it is abstracted into a node v, contact represents that with the limit e of company between the point then complication system just can be described with network model between the individuality.From in form, in fact network comprises the figure G on series of points, limit, and (N), wherein N is the node number for V, E to be designated as G=.The Complex Networks Theory of Xing Qiing is a kind of scientific theory of exploring the complication system universal feature on this basis, and it thinks the system topological characteristic to the behavior important influence of system, as fault relay in the transmission in crowd's net, the electrical network etc.At present, Complex Networks Theory has obtained fast development in multiple ambit.
Network characteristic: be used for portraying the statistical property of complex network structures, as degree, feature path, cluster coefficients, Jie's number etc., apparatus of the present invention system mainly uses correlation properties and is defined as follows:
Degree: the node number of degrees are meant the limit number that connects this node, are designated as k.The number of degrees to all nodes are averaged, and promptly obtain the average number of degrees of network, are designated as K.
Distance: expression connect any 2 node i that communicate, j shortest path through the number on limit, be designated as D (i, j).
The feature path: all nodes of network between the mean value of distance.It is one and describes the characteristic parameter of arbitrary node to a distance from overall angle
The cluster coefficient: cluster coefficients is to weigh the characteristic parameter of neighboring node contact tight ness rating, and the cluster coefficients expression formula of whole network is as follows, and wherein: in the formula, n is a number of network node, k
iThe number of degrees (fillet number) of expression node i, t
iExisting fillet number between the adjacent node of expression node i.
Jie's number: node Jie number is meant the number of the shortest path of this node of process in the network, is designated as node Jie and counts S.
Below, with reference to accompanying drawing the specific embodiment of the present invention is illustrated
(1) structure of device of the present invention is formed
With reference to shown in Figure 1, the industrial system process fault analysis device comprises computing machine and the trouble analysis system that is arranged on this computing machine, and trouble analysis system comprises: human-computer interaction module, process fault analysis module, process data administration module, grid specificity analysis module and system architecture information bank.Can adopt computer memory that system architecture information, process data and fault analysis flow process are stored, and adopting IO interface to connect keyboard, external memory storage and display, the structural network model that generates in the analytic process, network characteristic information, malfunctioning node collection (Fault Sets-FS) and analysis result etc. can adopt the form of man-machine interaction to express in display.
Human-computer interaction module: be used to realize the mutual of user and process fault analysis system, comprise the relevant information input of analyzed object, comprise contact details between technological process information, process measuring point variable information, variable, and the output demonstration of analysis result etc., and can revise system architecture information bank, process data and fault analysis flow process.
Grid specificity analysis module: this module is that the pith of apparatus of the present invention carries out network struction to procedures system and the grid characteristic is calculated and analyzed, the key variables set of node (Crucial Variable Sets-CVS) of identification reflection entire system behavior is as the analytic target of subsequent process malfunction monitoring analysis.
The process fault analysis module: this module is the core of apparatus of the present invention, and on the basis of systematic procedure data and grid specificity analysis, the implementation system failure is monitored and the analysis of causes, and the workflow of safety analysis is managed;
System architecture information bank: system architecture information, failure mode information, network characteristic information etc. are stored and manage by relational database.
The process data administration module is used for the real time data that system's operational process produces is stored and managed, and comprises DC S control system data and artificial spot check data etc.
(2) structure of the data structuring model of system architecture information bank
With reference to shown in Figure 2, this model is stored in the system information storehouse, and information spinner will comprise neighbour's matrix, network topology characteristic, fault mode of descriptive system network model etc.Neighbour's matrix is used for the descriptive system network, is designated as { a
Ij, node i, when there is annexation in j, a
IjValue is 1; Otherwise be 0.Network topology characteristic degree of having characteristic, Jie count characteristic, cluster coefficient etc.
Mainly comprise failure cause, associated nodes, associated nodes sign, the order of severity and maintenance measure information in the failure mode information.The a variety of causes that fault mode takes place comprises directly causing fault or causing physics or chemical process, the misoperation etc. that make equipment deficiency develop into fault, and valve is bonding, controller inefficacy, pump leakage, flow rate disturbance etc.Associated nodes is meant the variable node that may have influence on after certain fault takes place.The associated nodes sign is meant that the sign of back variable node takes place certain fault, because in the process industry system, the variable of monitoring all has upper and lower threshold values when design, therefore can determine the associated nodes feature with the qualitative description that departs from threshold values.As when the pump generation leakage failure, will make other variable nodes T (temperature) surpass last threshold values, P (pressure) reduces, then available T (+), P (-) expression.The associated nodes feature also can be obtained (as accumulative total and control chart CUSUM etc.) by the training of historical failure data, also can determine according to industrial knowledge and operating personnel's experience.Obviously, this qualitative description ratio based on systematic knowledge is more simple based on the quantitative test of data.For ease of the consequent malfunction analysis, it is the storage mode of introducer that the fault mode pattern information that apparatus of the present invention relate to adopts with the associated nodes sign, as:
C
0(+)：FEED1—CCH，V1—FCH，Bump1—EF；
This fault mode expressing information is: positive deviation (sign) takes place in variable C0; Reason may be that charging 1 (FEED1) composition increases (CH) or valve 1 (V1) and lost efficacy and be in high-order bonding fault (FH), pump 1 (Bump1) break down (EF);
The order of severity refers to severity of consequence that fault mode produces.Maintenance measure refers to that the analyst should point out and estimate the measure that those can be used for eliminating or alleviating fault probability of happening and influence, comprises preventive maintenance, maintenance and change in design etc.
(3) based on the workflow of the process industry systematic procedure fault analysis of Complex Networks Theory
With reference to shown in Figure 3, comprise the steps: based on the process industry systematic procedure fault analysis of Complex Networks Theory
Step 1: process industry grid specificity analysis, according to system's industrial flow information, process measuring point variable information etc. procedures system is carried out STRUCTURE DECOMPOSITION, with monitored parameters meshed network formal description system, by system is carried out the network characteristic analysis, obtain the key variables node set, obtain the data structuring model of system architecture information, this data structuring model comprises node descriptor, failure mode information etc., and is stored in the system architecture information bank.This step specifically comprises:
(a) according to system's industrial flow information, process measuring point variable information etc. procedures system is carried out STRUCTURE DECOMPOSITION, with monitored parameters meshed network formal description system.Company limit between variable node i, the j represents that there is the influence relation in they, owing to mainly pay close attention to the topological property of variable node network at this, therefore only needs qualitative description to go out whether there is contact between the variable.This part can be set up according to process chart, system operation handbook and technician's knowledge.After obtaining the system variable network model, for ease of subsequent analysis, describe, and be stored in the system architecture information bank with neighbour's matrix.
(b) the variable node network characteristic is analyzed; By neighbour's matrix is calculated according to the network characteristic definition, can obtain network characteristic values such as cluster coefficient, feature path, the node number of degrees, node Jie number, and be stored in the system architecture information bank.Wherein, cluster coefficient, feature path can help us to be familiar with the overall permanence of network, and, feature path less (with respect to random network with interstitial content) big as the cluster coefficient illustrate that then network has the worldlet characteristic; And characteristics such as the node number of degrees, node Jie number will be the foundations that we carry out the key variables node analysis.
(c) the key variables node set is obtained; For a complication system, there is the complex effects relation between each variable, traditional analytical approach or carry out quantitative test from strict mathematics model, or analyze qualitatively from the experimental knowledge aspect, seldom this influence is analyzed from grid characteristic angle.Apparatus of the present invention are mainly paid close attention to the node number of degrees of network and node several two class features that are situated between.At first, judge whether network has the worldlet characteristic; Then, in the network with worldlet characteristic, the number of degrees of network node play an important role to association between variables, and the number of degrees of certain node are big more, and the internuncial pathway of its correspondence is just many more.Considering simultaneously why the worldlet network has unique geometric properties, is to have introduced the cause that the minute quantity long-range connects (distal edge) because the reconnect procedure on limit is a network.In complex network, when certain node has long-range and connects, other nodes between shortest path usually can be preferential promptly node Jie number is high more through this category node, this node status in network is important more.This two classes combined factors is considered, definition w
iWeigh the key variables node:
Wherein, w
k, w
sCorrespond respectively to the weight of the node number of degrees and node Jie number.W with all nodes
iSort by size w
iBig promptly is the key variables node of identification, and resulting variable node collection can be used as the key variables set of node (Crucial Variable Sets-CVS) of reflection entire system behavior.This information stores is in the system architecture information bank.
(d) failure mode information is obtained; According to historical failure data establishing system failure mode information, comprise failure cause, associated nodes, node diagnostic, the order of severity and maintenance measure etc.Associated nodes herein promptly is the key variables set of node of determining in (c).
Step 2: based on the process monitoring of pivot analysis (PCA); After obtaining the key variables set of node, just can monitor the behavior of total system by monitoring these variablees.System's operational process is monitored with pivot analysis method (PCA) at this;
At first utilize the process data under the stable state non-fault situation, the principal component model under the structure normal condition; If X ∈ is R
N * m(m observational variable, n sampling number) is one section sample under the normal operating condition; It is carried out standardization (average is 0, and variance is 1) obtain X ∈ R
N * m, and to covariance matrix X
TX/ (n-1) carries out svd and obtains:
Wherein
Be the estimated value of X, T
k∈ R
N * kAnd P
k∈ R
M * kBe respectively pivot and get sub matrix and pivot matrix of loadings, E is a residual matrix.
Then, for a new sample x
New∈ R
1 * m, construct corresponding statistic T
2With Q and corresponding control limit threshold values thereof
And Q
αThe behavior of monitoring system process.Statistic T for example
2And corresponding control limit threshold values
Can determine by following formula:
D wherein
λ=diag (λ
1, λ
2λ
k) be the variance matrix of a preceding k pivot, F
α(k, n-k) for degree of confidence is α, degree of freedom is respectively the higher limit of the F distribution of k and n-k, the acquisition of can tabling look-up.Q and corresponding control limit threshold values Q thereof
αStructure no longer describe in detail, referring to pertinent literature.T for a certain moment sample gained
2And Q, if
Perhaps Q＜Q
α, then declarative procedure has fault to take place; Otherwise, controlled process.
Step 3:, then obtain associated nodes set (FaultSets-FS) and qualitative failure-description information thereof with failure symptom if break down; For a certain moment sample,, get (+) if just be offset then this variable sign, otherwise get (-), as P according to known variables bound threshold values (can suitably dwindle the bound threshold values)
A(+) expression sign is " pressure variations P
AJust be offset ", also can obtain this sign information by univariate statistics.Simultaneously, the appearance of considering sign is a time series events, and sequencing is promptly arranged, and therefore can repeatedly obtain failure symptom monitoring in fault generation back a period of time.
Step 4: based on the failure reason analysis of associated nodes sign with the observation sign;
After obtaining failure symptom, the fault mode that contrast has been set up can carry out failure reason analysis.
Step 5: the analysis result and the order of severity, maintenance measure show.
(4) based on the failure cause analysis method of associated nodes sign with the observation sign;
With reference to shown in Figure 4, after obtaining the failure symptom that observes,, determine the initial reason set according to the associated nodes sign information in the fault mode; If exist a kind of reason can explain the failure symptom that all observes, then this reason is the final analysis result, and single failure may take place illustrative system; If can not all explain the observation sign, then record explains that the maximum reason of sign number (nES) is as one of final possible cause; Remaining reason set is repeated above the analysis, enough all explain the failure symptom of observation up to the energy collecting of gained reason.Obviously, the method that provides of apparatus of the present invention can be analyzed system's single fault, multiple failure.
(5) with reference to shown in Figure 5, lubricating oil system is to provide lubricating oil to supply with when being mainly compressor train work, and the normal operation of assurance system is.Workflow is: be stored in the antirust steam turbine oil of N46 in the fuel tank 70.00, be forced into 1.6MPa through oil pump 70.41, after condenser E1 cooling, filter F 1 is filtered, and by direct-operated regulator 70.60 oil pressure is controlled at 0.95MPa, after be divided into two-way.One tunnel each bearing and jiggering device through being sent to steam turbine, compressor, gearbox done lubricated using; One the tunnel is sent to Steam Turhine Adjustment mechanism makes speed governing hydraulic oil, returns fuel tank after each road oil return converges.
With reference to shown in Figure 6, this system has 62 nodes, as space is limited, provides partial view at this, and wherein the literal variable in the circle identifies and nodal scheme, and in LT103a7, LT103 is the variable sign, and a7 is the network node label, i.e. the 7th node.Table 1 is according to pressing w after the network characteristic analysis
iPreceding 20 variablees in value size ordering back, with this as the key variables collection (CVS) that obtains; Table 2 is sign and reason example in the fault mode pattern information that makes up;
Table 1 is pressed w
iPreceding 20 system nodes of value size ordering
| Ordering | Nodal scheme | Node identification | w i | Node is described |
| 1. | 27 | PI111 | 2.0222 | Lubrication pressure |
| 2. | 7 | LT103 | 1.5048 | Oil tank liquid level |
| 3. | 60 | TI101 | 1.4252 | The fuel tank temperature |
| 4. | 26 | 70.61 | 1.2662 | Direct-operated regulator pressure |
| 5. | 20 | 70.60 | 1.2662 | Direct-operated regulator pressure |
| 6. | 59 | TI107 | 1.1798 | The oil temperature |
| 7. | 24 | FI111 | 1.1252 | Flow of lubrication |
| 8. | 11 | PI105 | 1.1233 | The main pump top hole pressure |
| 9. | 14 | PI109 | 0.96443 | 70.34 valve place oil pressure |
| 10. | 15 | PI104 | 0.89226 | The stand-by pump top hole pressure |
| 11. | 18 | FI104 | 0.86457 | Stand-by pump oil flow |
| 12. | 9 | FI105 | 0.76457 | Main oil pump oil flow |
| 13. | 21 | PI7009 | 0.75806 | 70.09 valve goes out pressure |
| 14. | 5 | FI106 | 0.71887 | The accident oil pressure |
| 15. | 45 | TIA7632 | 0.58004 | Air compressor machine front end bearing bush temperature |
| 16. | 50 | TIA7635 | 0.56703 | Supercharger front end bearing bush temperature |
| 17. | 47 | TIA7640 | 0.56269 | Wheel box bearing shell temperature |
| 18. | 29 | PI76081 | 0.52668 | Air compressor machine front end lubrication pressure |
| 19. | 34 | PI76088 | 0.51367 | Supercharger front end lubrication pressure |
| 20. | 31 | PI76084 | 0.5115 | Gear-box lubricating oil is pressed |
Table 2 sign and reason example
| The variable sign | Reason |
| PI111(-) | FI111—FCH，PI109—PCL，V70.61—EF； |
| LT103(+) | FI111—FCH，FI7009—FCH；V70.07—EF，V70.10-EF |
After obtaining key variables collection and dependent failure pattern information, can at the key variables collection utilize based on data pivot analysis (PCA) method system operation situation monitor, noting abnormalities, (obtaining variable observation sign) is back to be analyzed reason, is provided results such as reason, maintenance measure at last by human-computer interaction module.
Procedures system more complicated in the actual production, node are numerous, therefore can obtain more key variables collection earlier, utilize technician's experimental knowledge to do further on this basis and accept or reject; Moreover, when the structure fault mode, except can from servicing manual, process historical data, obtaining, the analysis ability that also depends on the technician to systematic procedure, in order to improve the comprehensive of failure cause, at this failure mode information is carried out Open Management, be convenient to add and revise.
Claims (1)
1. the process fault analysis method of a process industry system is characterized in that, comprises the following steps:
Step 1: process industry grid specificity analysis, according to system's industrial flow information, process measuring point variable information procedures system is carried out STRUCTURE DECOMPOSITION, with monitored parameters meshed network formal description system, by system is carried out the network characteristic analysis, determine that according to network characteristic to system dynamics behavior has the key variables node set of material impact; Obtain the data structuring model of system architecture information, this data structuring model comprises node descriptor, failure mode information, and is stored in the system architecture information bank;
Step 2: based on system's pivot analysis (PCA) process monitoring of key node collection, utilize the real time data in the process system operation, key node collection constructing system data owner meta analysis (PCA) monitoring model that analysis is obtained, the systematic procedure fault is monitored;
Step 3: variable node collection and qualitative failure-description information thereof are obtained, if break down, then obtain variable node set and qualitative failure-description information thereof with failure symptom;
Step 4: based on the failure reason analysis of associated nodes sign with the observation sign, after obtaining failure symptom, the fault mode that contrast has been set up can carry out failure reason analysis, determines maintenance measure;
Step 5: analysis result is shown by human-computer interaction module,
Described process industry grid specificity analysis comprises the steps:
(a) according to system's industrial flow information, process measuring point variable information procedures system is carried out STRUCTURE DECOMPOSITION,, and adopt neighbour's matrix form to be stored in the system architecture information bank with monitored parameters meshed network formal description system;
(b) by neighbour's matrix is calculated according to the network characteristic definition, obtain cluster coefficient, feature path, the node number of degrees, node Jie and count the network characteristic value, and be stored in the system architecture information bank, for providing foundation from network perspective recognition system and subsequent analysis;
(c) determine node importance degree w according to network characteristic
i, obtain the key variables set of node of reflection entire system behavior, this information stores is in the system architecture information bank;
(d) failure mode information is obtained; According to historical failure data establishing system failure mode information, comprise failure cause, associated nodes, associated nodes sign, the order of severity and maintenance measure,
Described system's pivot analysis (PCA) monitoring method based on the key node collection, at first, the key variables collection that utilizes the process data under the stable state non-fault situation and obtain, the principal component model under the structure normal condition promptly constructs pivot and gets sub matrix and pivot matrix of loadings; Then, for a new sample, construct corresponding statistic T
2With Q and corresponding control limit threshold values T thereof
α 2And Q
αThe behavior of monitoring system process, described based on the failure cause analysis method of associated nodes sign with the observation sign, according to the associated nodes sign information in the fault mode, determine the initial reason set, explain the failure symptom that all observes if there is a kind of reason, then this reason is the final analysis result, and single failure may take place illustrative system; If can not all explain the observation sign, then record explanation sign is counted the reason of maximum as one of final possible cause; Remaining reason set is repeated above step, enough all explain the failure symptom of observation up to the energy collecting of gained reason.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102321321A CN101446827B (en) | 2008-11-06 | 2008-11-06 | Process fault analysis device of process industry system and method therefor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008102321321A CN101446827B (en) | 2008-11-06 | 2008-11-06 | Process fault analysis device of process industry system and method therefor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101446827A CN101446827A (en) | 2009-06-03 |
| CN101446827B true CN101446827B (en) | 2011-06-22 |
Family
ID=40742531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008102321321A Expired - Fee Related CN101446827B (en) | 2008-11-06 | 2008-11-06 | Process fault analysis device of process industry system and method therefor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101446827B (en) |
Families Citing this family (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096821B (en) * | 2010-04-15 | 2014-05-07 | 西安理工大学 | Number plate identification method under strong interference environment on basis of complex network theory |
| CN102681473A (en) * | 2012-04-01 | 2012-09-19 | 中南大学 | Fault detecting method for sulfur flotation process on basis of texture unit distribution |
| CN102736546B (en) * | 2012-06-28 | 2014-06-04 | 西安交通大学 | State monitoring device of complex electromechanical system for flow industry and method |
| US10095813B2 (en) * | 2013-11-18 | 2018-10-09 | The Boeing Company | Safety analysis of a complex system using component-oriented fault trees |
| US9483342B2 (en) * | 2014-03-20 | 2016-11-01 | Siemens Aktiengesellschaft | Supporting failure mode and effects analysis |
| CN105893726B (en) * | 2014-11-21 | 2019-07-30 | 张琳 | A kind of monitoring method of characteristic signal monitoring point |
| CN105721193B (en) * | 2014-12-05 | 2020-04-28 | 方正国际软件(北京)有限公司 | System information monitoring method and device |
| CN104503441B (en) * | 2014-12-22 | 2017-02-22 | 北京化工大学 | Process fault monitoring method based on improved dynamic visible graph |
| CN105067295A (en) * | 2015-07-20 | 2015-11-18 | 柳州好顺科技有限公司 | Engineering machine fault alarm device |
| CN106572494A (en) * | 2015-10-08 | 2017-04-19 | 中兴通讯股份有限公司 | Base station fault detection method and device |
| CN105955241B (en) * | 2016-06-03 | 2018-09-14 | 北京科技大学 | A kind of quality fault localization method based on joint data-driven production process |
| CN106384130A (en) * | 2016-09-22 | 2017-02-08 | 宁波大学 | Fault detection method based on data multi-neighbor-local-feature embedding |
| CN107196805A (en) * | 2017-06-01 | 2017-09-22 | 国网天津市电力公司 | Power information system failure auxiliary locator |
| CN107633309B (en) * | 2017-09-22 | 2018-08-31 | 合肥工业大学 | It is a kind of complexity former maintenance policy determine method and system |
| CN107992014A (en) * | 2018-01-17 | 2018-05-04 | 珠海市长陆工业自动控制系统股份有限公司 | Process industry production line production playback and fault analysis diagnosis system |
| CN108388221B (en) * | 2018-03-21 | 2019-06-21 | 江南大学 | A kind of multilayered schema monitoring method of process flow industry process |
| CN110173872A (en) * | 2019-04-22 | 2019-08-27 | 广东美博制冷设备有限公司 | The misoperation detection method and device of air conditioner |
| CN110213087B (en) * | 2019-05-16 | 2020-08-25 | 北京航空航天大学 | Complex system fault positioning method based on dynamic multilayer coupling network |
| CN110109435B (en) * | 2019-05-22 | 2020-07-24 | 杭州电子科技大学 | Online monitoring method for improving two-step subspace division |
| CN110716539B (en) * | 2019-11-26 | 2021-03-16 | 宁波市轨道交通集团有限公司 | Fault diagnosis and analysis method and device |
| CN111308991B (en) * | 2020-03-19 | 2021-10-01 | 中国大唐集团科学技术研究院有限公司华东电力试验研究院 | Coal mill operation fault identification method and application |
| CN112116198B (en) * | 2020-08-04 | 2023-06-20 | 西安交通大学 | Data-driven process industrial state perception network key node screening method |
| CN112130487A (en) * | 2020-09-02 | 2020-12-25 | 珠海格力电器股份有限公司 | Equipment fault early warning method and device |
| CN112464038B (en) * | 2020-10-30 | 2024-03-29 | 深圳市安云信息科技有限公司 | Fault association tree generation method and device |
| CN114353261A (en) * | 2021-12-13 | 2022-04-15 | 广东电网有限责任公司 | Air conditioning unit fault analysis method and device, terminal device and storage medium |
| CN115993807B (en) * | 2023-03-23 | 2023-06-09 | 日照鲁光电子科技有限公司 | Production monitoring optimization control method and system for silicon carbide |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1987697A (en) * | 2006-12-22 | 2007-06-27 | 浙江大学 | Fault diagnostic system and method for under industrial producing process small sample condition |
| CN101196743A (en) * | 2007-12-26 | 2008-06-11 | 西安交通大学 | Dynamoelectric system safety analyzing device and method based on cause-effect network model |
-
2008
- 2008-11-06 CN CN2008102321321A patent/CN101446827B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1987697A (en) * | 2006-12-22 | 2007-06-27 | 浙江大学 | Fault diagnostic system and method for under industrial producing process small sample condition |
| CN101196743A (en) * | 2007-12-26 | 2008-06-11 | 西安交通大学 | Dynamoelectric system safety analyzing device and method based on cause-effect network model |
Non-Patent Citations (2)
| Title |
|---|
| 姜洪权等.基于复杂网络理论的流程工业系统安全性分析.西安交通大学学报.2007,41(7),806-810. * |
| 李果等.基于小世界拓扑模型的复杂系统安全分析.机械工程学报.2008,44(5),86-91. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101446827A (en) | 2009-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101446827B (en) | Process fault analysis device of process industry system and method therefor | |
| CN112052979B (en) | Equipment spare part demand prediction system based on fault prediction and health management | |
| CN102736546B (en) | State monitoring device of complex electromechanical system for flow industry and method | |
| CN109829468A (en) | Civil aircraft Fault Diagnosis of Complex System method based on Bayesian network | |
| CN109948860A (en) | A kind of mechanical system method for predicting residual useful life and system | |
| CN103793859B (en) | A kind of wind power plant operation monitoring and event integrated evaluating method | |
| CN101529355A (en) | On-line monitoring and diagnostics of a process using multivariate statistical analysis | |
| CN108051709A (en) | Transformer state online evaluation analysis method based on artificial intelligence technology | |
| CN107153929A (en) | Gearbox of wind turbine fault monitoring method and system based on deep neural network | |
| CN102262690A (en) | Modeling method of early warning model of mixed failures and early warning model of mixed failures | |
| CN101187803B (en) | Ammonia converter production optimization method based on data excavation technology | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| CN111412579A (en) | Air conditioning unit fault type diagnosis method and system based on big data | |
| CN101799674A (en) | Method for analyzing service state of numerical control equipment | |
| CN103020438A (en) | Aero-engine reliability monitoring method based on mixed weibull distribution | |
| CN117196159A (en) | Intelligent water service partition metering system based on Internet big data analysis | |
| CN111896244B (en) | Fully mechanized mining equipment PHM system | |
| CN106203637A (en) | A kind of mobile system health evaluating method based on mahalanobis distance and health index | |
| CN106607461A (en) | Hydraulic AGC fault diagnosis expert system for cold-rolling mill | |
| CN116756909A (en) | Early warning diagnosis system of thermal power plant based on data model and mechanism model | |
| CN117032120A (en) | Integrated intelligent cloud control system and control method for air compression station | |
| KR20220089853A (en) | Method for Failure prediction and prognostics and health management of renewable energy generation facilities using machine learning technology | |
| CN105302476B (en) | A kind of reliability data online acquisition for nuclear power plant equipment analyzes storage system and its storage method | |
| CN113114748A (en) | Intelligent prediction system based on data analysis | |
| KR101896442B1 (en) | System, Server and Method for monitoring wind plant with ICT fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110622 Termination date: 20131106 |