CN106710653A - Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit - Google Patents
Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit Download PDFInfo
- Publication number
- CN106710653A CN106710653A CN201611101988.6A CN201611101988A CN106710653A CN 106710653 A CN106710653 A CN 106710653A CN 201611101988 A CN201611101988 A CN 201611101988A CN 106710653 A CN106710653 A CN 106710653A
- Authority
- CN
- China
- Prior art keywords
- matrix
- data
- pivot
- time data
- real time
- 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
- G21—NUCLEAR PHYSICS; NUCLEAR ENGINEERING
- G21D—NUCLEAR POWER PLANT
- G21D3/00—Control of nuclear power plant
- G21D3/04—Safety arrangements
- G21D3/06—Safety arrangements responsive to faults within the plant
-
- 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
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E30/00—Energy generation of nuclear origin
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Plasma & Fusion (AREA)
- General Engineering & Computer Science (AREA)
- High Energy & Nuclear Physics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a real-time data abnormal diagnosis method for monitoring operation of a nuclear power unit. The method comprises the following steps: selecting a plurality of important variables for monitoring operation of the nuclear power unit to form a real-time data anomaly diagnosis variable set, establishing an abnormal diagnosis model through the principal component analysis and the independent element analysis technology, determining a control limit used for judging whether an abnormal event occurs in the real-time operation data or not through the nuclear density estimation technology; meanwhile setting an automatic updating mechanism of the model to ensure the precision of the model. The method disclosed by the invention overcomes data model 'ill-condition' problems caused by the characteristics of the actual industrial data such as serious correlation and redundancy, random noise, mode variability, intrinsic nonlinearity and the like, 'false alarm' and 'missed alarm' can be effectively avoided. The method is suitable for real-time monitoring of the nuclear power unit and early warning and diagnosis of the ''abnormal event or accident'', and is beneficial to improving the safety of the nuclear power unit.
Description
Technical field
Field adjustable and comercial operation field before starting the present invention relates to nuclear power generating sets, more particularly to it is a kind of for nuclear power machine
The real time data abnormality diagnostic method of group operation monitoring.
Background technology
It is possible that unexpected situation, such as chaser, jump heap, primary Ioops leakage during nuclear power generating sets run.At this moment need
Operation, maintenance, debugging efforts personnel are by process data analysis failure cause, failure judgement trend etc..At present, global newly-built core
Group of motors employs advanced master control room design and overall application digitizes I&C system (abbreviation DCS), to realize full factory's technique
The centralized watch of system, flexible operating and control defencive function.However, accounting for the independent intellectual of absolute leading position with regard to the current country
For property right CPR1000 nuclear power generating sets, although DCS realizes perfect control and protection task, the depth under DCS data platforms
Layer data is excavated, and especially the anomaly analysis relevant with nuclear plant safety, failure (accident) diagnosis and risk management and control function are non-
Often it is short of.Therefore, based on perfect DCS system nuclear power generating sets are carried out with the abnormity diagnosis of real time data, so that early warning may be sent out
Raw failure and accident, the security for improving nuclear power generating sets, it has also become necessary and urgent technological means.
In recent ten years, academia and engineering circles think that this has carried out many researchs and trial, but due to it is following it is difficult into
Work(application is simultaneously few:(1) this kind of method generally requires real time data abnormity diagnosis model, and the foundation of the model is based on largely
Real time data sample.Because actual industrial data have serious correlation, redundancy feature, cause and set up by such data
Model more or less exist " morbid state ";(2) actual industrial system typically has multi-mode operation, random noise influence serious
The characteristics of, this just determines that data exception diagnostic model can not be unalterable, must possess according to operational mode change and noise feelings
The ability that condition is automatically updated;(3) decision criteria for data exception " event " is extremely important, is to avoid " false alarm ", " fail to report
The key of police ".By this kind of decision criteria is determined by the dynamic property for being diagnosed system more, in the past by system performance simplification
The method of reason is difficult to be competent at naturally.
The content of the invention
The purpose of the present invention is to solve the shortcomings of the prior art, there is provided a kind of reality for nuclear power generating sets operation monitoring
When data exception diagnostic method.
The purpose of the present invention is realized with step by the following technical programs:One kind runs monitoring for nuclear power generating sets
Real time data abnormality diagnostic method, specifically include following steps:
(1) multiple significant variables relevant with failure are selected to constitute real time data abnormity diagnosis variables set, on this basis,
The service data of the significant variable of online acquisition nuclear power generating sets production process, obtains setting up the original instruction of data exception diagnostic model
Practice data set X0;
(2) standardization, normalized before being modeled to resulting original training data collection X0, obtain actual answering
Modeling training dataset X;
(3) " albefaction " treatment is carried out to training dataset, i.e., pivot point is carried out to resulting modeling training dataset X
Analysis, and pivot number is determined according to crosscheck method, obtain pivot matrix T and matrix of loadings P;
(4) the pivot matrix T and matrix of loadings P obtained according to step (3), set up real time data different with independent component analysis
Normal diagnostic model;
(5) real time data abnormity diagnosis model is updated and is tieed up by the block formula recursion alignment technique of restricted memory
Shield;
(6) on the basis of the real time data abnormity diagnosis model set up, cuclear density is determined using Density Estimator method
Estimation function equation, determines that the control whether " anomalous event " occurs is limited by Density Estimator functional equation
(7) the service data collection of one group of current significant variable of Real-time Collection and it is standardized, the standardization detection for obtaining
Data set is Xnew;
(8) it is X to integrate detection datanewSubstitute into step 4 real time data abnormity diagnosis model, obtain current detection data
Independent variable matrix Snew;
(9) using the Density Estimator functional equation and 95% or 99% confidence level in step 6, independent variable matrix is substituted into
Snew, obtainWhereinIt is Density Estimator functional equation, H is bandwidth matrices;WillWith normal operation
Scope control is limitedIt is compared, ifThen represent process units in normal operation range, otherwise
Represent category unusual condition or generate failure.
Further, the step (4) is specially:The pivot matrix T and load moment obtained according to step (3) pivot analysis
Battle array P, carries out independent component analysis, to ask for the fault diagnosis model that independent entry solution closes matrix form equation
Wherein, S is independent variable matrix, and W is the pseudoinverse of transformation matrix,It is that the solution based on pivot analysis closes matrix;Independent entry number takes
It is pivot number k;
Further, the detailed process of the independent component analysis is as follows:
(4.1) pivot matrix T and matrix of loadings P is taken, and makes a=1, a is iterative steps;
(4.2) the initial column vector w for taking that norm is 1 is appointeda(0), if a >=2, make
Wherein Wa-1=[w1,w2,…wa-1], and make normalized wi(0)=wi(0)/||wi(0) | |, it is internal layer iteration to make b=0, b
Step number;
(4.3) to waIt is iterated renewal, wa(b+1)=wa(b)-μ[E{ta(wa(b)Tta)3}-3wa(b)], in formula:μ=
0.1 or 0.01 is learning rate, and expectation E therein carries out estimation acquisition by sample value.
(4.4) to wa(b+1) decomposition is carried out to projectAnd carry out normalizing
Change treatment wa(b+1)=wa(b+1)/||wa(b+1)||。
(4.5) if | wa(b+1)Twa(b) | ≈ 1, then continue step (4.6), otherwise, make b=b+1 and go to step
(4.3)。
(4.6) w is madea=wa(b+1), a=a+1, then goes to step (4.2) and iterates to calculate always to a=k.
(4.7) W=[w that will be obtained1w2……wk] P that obtains with step (3) byConstitute solution and close matrix, i.e.,
Fault diagnosis model is obtained
Further, the step (5) is specially:If original modeling training dataset is { Xj, j is current iteration step number,
Fresh data collection is { X after running after a whilej+1, then { XjBy the reality that is obtained after pivot analysis and independent component analysis
When data exception diagnostic model beWhereinkjRespectively former modeling training dataset
XjAverage, variance and pivot number, PjIt is to XjThe matrix of loadings obtained after pivot analysis, WjTo be obtained after independent component analysis
Transformation matrix pseudoinverse.According to fresh data collection { Xj+1AndReal time data after being updated
Abnormity diagnosis modelWherein Pj+1=Pj+αPΔFP(||Xj+1-Xj||,γP,Pj),
Wj+1=Wj+αWΔFW(||Xj+1-Xj||,γW,Wj), Δ FP,ΔFWIt is block formula recursion correction function, γP,γWFor correction is forgotten
The factor, αP,αWIt is weight coefficient,K is respectively average, variance and the pivot number of fresh data collection,.
Further, the step (6) is specially:The independent metadata matrix obtained by diagnostic model is X0=[x1,
x2,…xn]T, n is independent entry number, xnIt is n-th independent entry, then the kernel estimates of density function f (x) areWherein, i=1,2 ... n, H are bandwidth matrices, and | H | is the row of H
Column, K () is kernel function, andBy above-mentioned Density Estimator function
The control limit for characterizing normal operation range can be tried to achieve with 95% or 99% confidence level
The present invention compared with prior art, with having the beneficial effect that:
(1) method of the present invention is based on the modeling principle of data-driven, avoids the process mechanism analysis of complexity so that existing
Field is easy to implement, is particularly suited for higher-dimension and the very abundant industrial occasions of process data as similar nuclear power generating sets;
(2) abnormity diagnosis model is set up present invention employs the independent component analysis technology based on multivariate statistical projection method principle,
Eliminate significantly due to actual industrial data have serious correlation, redundancy and there is asking for " morbid state " in the data model that brings
Topic, makes model accuracy improve many compared with other methods;
(3) determine present invention employs Density Estimator technology accurate for judging the judgement whether " anomalous event " occurs
Then, the probability of " false alarm " and " failing to report police " generation is effectively reduced;
(4) the block formula recursion alignment technique invention introduces restricted memory data abnormity diagnosis are carried out it is on-line automatic more
Newly, it is to avoid model degradation, the defect of precise decreasing brought by production model or equipment state change;
(5) method of the present invention is used for the actual motion of monitor in real time nuclear power generating sets, and the security that can produce nuclear power is obtained
To raising.
Brief description of the drawings
Fig. 1 is a typical nuclear power generating sets device structure schematic diagram.
Fig. 2 is nuclear power generating sets real time data exception method flow diagram of the invention;In figure, reactor 1, control rod and driving
It is mechanism 2, primary Ioops system 3, voltage-stablizer 4, main coolant pump 5, steam generator 6, secondary coolant circuit system 7, steam turbine generator 8, solidifying
Vapour device 9, feed pump 10, circulation 11, cooling water source 12;
Specific embodiment
Below by a typical case study on implementation and with reference to accompanying drawing, specific embodiment of the invention is elaborated, this
The purpose and effect of invention will be apparent from.
Fig. 1 is a device structure schematic diagram for typical nuclear power generating sets, and it is by reactor 1, control rod and drive mechanism
2nd, primary Ioops system 3, voltage-stablizer 4, main coolant pump 5, steam generator 6, secondary coolant circuit system 7, steam turbine generator 8, condenser
9th, feed pump 10, circulation 11 and cooling water source 12 are constituted, wherein, the important monitoring relevant with nuclear power generating sets security becomes
Amount is mostly distributed in 1-8 equipment.
To realize being carried out continuously for nuclear power production, the generally operation using DCS to nuclear power generating sets equipment is controlled by and grasps
Make.Because loop is complicated, equipment is numerous, especially worked under high temperature, high pressure, radiation environment for a long time, produce production abnormal and
Equipment fault is can hardly be avoided.
Embodiment
The multiple some significant variables relevant with various possible breakdowns of selection constitute real time data abnormity diagnosis variables set.
On the basis of this, the operation service data (typically needing continuous acquisition more than 72 hours) of online acquisition nuclear power generating sets production process is obtained
To the original training data collection X0 for setting up data exception diagnostic model.The abnormity diagnosis variables set of selection mainly includes:Power network is born
Radioactive concentration, main steaming in lotus, reactor nucleus power, steam turbine generator rotating speed, containment sump water level, containment in air
Vapour radioactivity, primary Ioops coolant pressure, reactor pressure vessel water level, boiler pressure, steam generator water level, one
Loop flow, primary Ioops coolant temperature, voltage-stablizer water level, condenser exhaust gas radioactive concentration, voltage-stablizer pressure, 1-5# steamers
Machine bearing watt shakes, 1-5# steam turbine bearing axles shake, 6-8# dynamo bearings watt shake, 6-8# generator shafts shake, thrust bearing displacement,
1-5# steam turbine bearings temperature, 6-8# dynamo bearings temperature, thrust bearing temperature, high pressure cylinder delivery temperature, low pressure (LP) cylinder exhaust
Temperature, main steam temperature, main steam pressure, main steam flow, main frame lubricating oil pressure, EH oil pressures, main frame lubricant level,
Cold hydrogen temperature of condenser vacuum, generator unit stator temperature, generator etc..
Offline foundation and on-line maintenance, " anomalous event or accident " including real time data abnormity diagnosis model it is online pre-
Alert and two stages of diagnosis.
First stage:The offline foundation of real time data abnormity diagnosis model and on-line maintenance function are comprised the following steps:
Step A1:On the basis of to nuclear power generating sets operation mechanism and conventional troubleshooting report analysis, it is determined that with it is various
The relevant some significant variables of possible breakdown constitute real time data abnormity diagnosis variables set, and continuous collecting is for a period of time (more than 48
Hour) real-time running data, composition sets up the original training data collection X0 of data exception diagnostic model;
Step A2:Standardization before being modeled to resulting original training data collection X0, becomes each process
The average of amount is that 0, variance is 1, thus obtains the training dataset X of practical application.Standardized method is:
Mean value computation:
Variance is calculated:
Normalization is calculated:
In formula, N is the number of samples of data set.
Step A3:With covariance matrix singular value decomposition principle, pivot is carried out to resulting modeling training dataset X
Analysis, and determined to meet the pivot number of certain variance sign rate according to crosscheck technology, obtain pivot matrix T and load moment
Battle array P.The step for be referred to as modeling " albefaction " treatment of training dataset, specific method is:Computation modeling training dataset X first
Covariance matrix ∑X, then to ∑XSingular value decomposition is carried out, tactic characteristic root λ by size is obtained1,λ2,λ3……
With corresponding characteristic vector p1,p2,p3..., the 3rd, the singular value decomposition to modeling training dataset X carries out variance tribute
Offer the crosscheck of rate, determine characteristic root number (that is, pivot number) k that information is dominant, the 4th, before choosing k feature to
Amount, construction matrix of loadings P=[p1p2……pk]T, finally, ask for pivot matrix T=[t1t2……tk]T=XP.
Step A4:Because actual industrial data mostly have serious correlation, redundancy feature, cause by such data
More or less there is " morbid state " in the model of foundation.Therefore, with the independent component analysis technology based on multivariate statistical projection method principle
Abnormity diagnosis model is set up, i.e., the pivot matrix T and matrix of loadings P for being obtained according to pivot analysis further carry out independent entry point
Analysis, to ask for the fault diagnosis model that independent entry solution closes matrix form equationWherein, S is
Independent variable matrix, W is the pseudoinverse of transformation matrix,It is that the solution based on pivot analysis closes matrix.Due to having carried out pivot analysis
" albefaction " is processed, and independent entry number can simply be taken as pivot number k.The detailed process of independent component analysis is:
(4.1) pivot matrix T and matrix of loadings P is taken, and makes a=1, a is iterative steps;
(4.2) the initial column vector w for taking that norm is 1 is appointeda(0), if a >=2, make
Wherein Wa-1=[w1,w2,…wa-1], and make normalized wi(0)=wi(0)/||wi(0) | |, make b=0;
(4.3) to waIt is iterated renewal, wa(b+1)=wa(b)-μ[E{ta(wa(b)Tta)3}-3wa(b)], in formula:μ=
0.1 or 0.01 is learning rate, and expectation E therein carries out estimation acquisition by sample value.
(4.4) to wa(b+1) decomposition is carried out to projectAnd carry out normalizing
Change treatment wa(b+1)=wa(b+1)/||wa(b+1)||。
(4.5) if | wa(b+1)Twa(b) | ≈ 1, then continue step (4.6), otherwise, make b=b+1 and go to step
(4.3)。
(4.6) w is madea=wa(b+1), a=a+1, then goes to step (4.2) and iterates to calculate always to a=k.
(4.7) W=[w that will be obtained1w2……wk] P that obtains with step (3) byConstitute solution and close matrix, obtain final product
Fault diagnosis model is arrived
Step A5:The characteristics of typically influenceing serious with multi-mode operation, random noise due to actual industrial system, this is just
Determine that data exception diagnostic model must possess the ability automatically updated according to operational mode change and noise situations.Therefore,
When line operation is applied, the renewal and maintenance of abnormity diagnosis model should be carried out according to real-time sampling data regular (being set to 96 hours),
To ensure the long-time accuracy of model, i.e., using fresh sampled data to Current Diagnostic model
Calculating is corrected by the block formula recursion fault diagnosis model alignment technique of restricted memory.Specially:If original modeling training number
It is { X according to collectionj, j is current iteration step number, and fresh data collection is { X after running after a whilej+1, then { XjBy pivot
Analyze and be with the real time data abnormity diagnosis model obtained after independent component analysisWherein
σX0,j,kjThe X of respectively former modeling training datasetjAverage, variance and pivot number, PjIt is to XjObtained after pivot analysis
Matrix of loadings, WjIt is the pseudoinverse of transformation matrix obtained after independent component analysis.According to fresh data collection { Xj+1AndReal time data abnormity diagnosis model after being updatedWherein
Pj+1=Pj+αPΔFP(||Xj+1-Xj||,γP,Pj),Wj+1=Wj+αWΔFW(||Xj+1-Xj||,γW,Wj), Δ FP,ΔFWIt is block
Formula recursion correction function, γP,γWIt is correction forgetting factor, αP,αWIt is weight coefficient,σX0,j+1, k is respectively fresh number
According to the average of collection, variance and pivot number.It can be seen that, it is only accumulated when new data and is arrived certain based on block formula data first
Scale is just processed when forming data block, and in terms of being iterated in itself instead of data by data block in the recursive process
Calculate, greatly reduce amount of calculation.Secondly, it depends on the information that limited latest data block is provided all the time, often increases by one
It is necessary to remove an old data block, that is, the data for influenceing model are newest some data blocks to new data block all the time.Finally, limit
The forgetting factor of fixed memory is introduced into data block queue, and the time is more long, smaller (the data block letter of current moment of data block confidence level
Spend for 1), the data block reliability of adjacent moment presses exponential damping.
Second stage:The on-line early warning and diagnostic function of " anomalous event or accident " are comprised the following steps:
Step B1:Decision criteria for data " anomalous event " is extremely important, avoids " false alarm ", " failing to report police "
It is crucial.In the present invention, set up in the first stage on the basis of real time data abnormity diagnosis model, it is true using Density Estimator technology
It is fixed to be used to judge the criterion (control limit) whether " anomalous event " occursSpecially:The independent entry come by diagnostic model
Data matrix is X=[x1,x2,…xn]T, n is independent entry number, xnIt is the kernel estimates of n-th independent entry then density function f (x)
ForWherein i=1,2 ... n, H are bandwidth matrices, and | H | is the determinant of H, K
() is kernel function, andThe probability density of kernel function estimation actually may be used
With regard as the segmentation gradient constituted on sample point plus and, kernel function K () determine segmentation gradient shape and bandwidth then
Determine its width.More common K () is gaussian kernel function, and optional form is
After giving sample and have chosen kernel function form, Density Estimator functionDepending on the selection of H, optional diagonal matrixThe normal operation of sign can be tried to achieve by above-mentioned Density Estimator function and 95% or 99% confidence level
The control limit of scope
Step B2:When needing to carry out current operating condition " anomalous event " diagnosis, one group of Real-time Collection currently runs
Data set is simultaneously standardized, and it is X that the standardization detection data for obtaining integratesnew。
Step B3:It is X that detection data is integratednewReal time data abnormity diagnosis model is substituted into, the only of current detection data is obtained
Vertical variable matrix Snew。
Step B4:Using the Density Estimator functional equation in step B1 and 95% or 99% confidence level, independent entry is substituted into
Matrix SnewCan calculateAnd control to limit with normal operation rangeCompare, ifThen table
Show process units in normal operation range, otherwise represent category unusual condition or generate failure.
Step B5:Once being judged as that " anomalous event " occurs, then start " anomalous event " early warning treatment.
More than be a specific, complete implementation process of the invention, the example be used for illustrating usage of the invention and
It is non-to limit the invention.Any change for carrying out within the scope of the invention as claimed, belongs to of the invention
Protection domain.
Claims (5)
1. it is a kind of to run the real time data abnormality diagnostic method for monitoring for nuclear power generating sets, it is characterised in that to specifically include as follows
Step:
(1) multiple significant variables relevant with failure are selected to constitute real time data abnormity diagnosis variables set, on this basis, online
The service data of the significant variable of nuclear power generating sets production process is gathered, obtains setting up the original training number of data exception diagnostic model
According to collection X0;
(2) standardization, normalized before being modeled to resulting original training data collection X0, obtain practical application
Modeling training dataset X;
(3) " albefaction " treatment is carried out to training dataset, i.e., pivot analysis is carried out to resulting modeling training dataset X, and
Pivot number is determined according to crosscheck method, pivot matrix T and matrix of loadings P is obtained;
(4) the pivot matrix T and matrix of loadings P obtained according to step (3), sets up real time data and examines extremely with independent component analysis
Disconnected model;
(5) real time data abnormity diagnosis model is updated and is safeguarded by the block formula recursion alignment technique of restricted memory;
(6) on the basis of the real time data abnormity diagnosis model set up, Density Estimator is determined using Density Estimator method
Functional equation, determines that the control whether " anomalous event " occurs is limited by Density Estimator functional equation
(7) the service data collection of one group of current significant variable of Real-time Collection and it is standardized, the standardization detection data for obtaining
It is X to integratenew;
(8) it is X to integrate detection datanewSubstitute into step 4 real time data abnormity diagnosis model, obtain the only of current detection data
Vertical variable matrix Snew;
(9) using the Density Estimator functional equation and 95% or 99% confidence level in step 6, independent variable matrix S is substituted intonew, obtain
ArriveWhereinIt is Density Estimator functional equation, H is bandwidth matrices;WillWith normal operation range
Control limitIt is compared, ifThen represent that process units in normal operation range, is otherwise represented
Category unusual condition generates failure.
2. real time data abnormality diagnostic method according to claim 1, it is characterised in that the step (4) is specially:Root
According to pivot matrix T and matrix of loadings P that step (3) pivot analysis are obtained, independent component analysis are carried out, square is closed to ask for independent entry solution
The fault diagnosis model of battle array equation formWherein, S is independent variable matrix, and W is transformation matrix
Pseudoinverse,It is that the solution based on pivot analysis closes matrix;Independent entry number is taken as pivot number k.
3. real time data abnormality diagnostic method according to claim 2, it is characterised in that the independent component analysis it is specific
Process is as follows:
(4.1) pivot matrix T and matrix of loadings P is taken, and makes a=1, a is iterative steps;
(4.2) the initial column vector w for taking that norm is 1 is appointeda(0), if a >=2, make
Wherein Wa-1=[w1,w2,…wa-1], and make normalized wi(0)=wi(0)/||wi(0) | |, it is internal layer iteration to make b=0, b
Step number;
(4.3) to waIt is iterated renewal,In formula:μ=
0.1 or 0.01 is learning rate, and expectation E therein carries out estimation acquisition by sample value.
(4.4) to wa(b+1) decomposition is carried out to projectAnd it is normalized place
Reason wa(b+1)=wa(b+1)/||wa(b+1)||。
(4.5) if | wa(b+1)Twa(b) | ≈ 1, then continue step (4.6), otherwise, make b=b+1 and go to step (4.3).
(4.6) w is madea=wa(b+1), a=a+1, then goes to step (4.2) and iterates to calculate always to a=k.
(4.7) W=[w that will be obtained1w2……wk] P that obtains with step (3) byConstitute solution and close matrix, that is, obtain
Fault diagnosis model
4. real time data abnormality diagnostic method according to claim 1, it is characterised in that the step (5) is specially:If
Original modeling training dataset is { Xj, j is current iteration step number, and fresh data collection is { X after running after a whilej+1, then
{XjBe by the real time data abnormity diagnosis model obtained after pivot analysis and independent component analysis
WhereinRespectively former modeling training dataset XjAverage, variance and pivot number, PjIt is to XjPivot analysis
The matrix of loadings for obtaining afterwards, WjIt is the pseudoinverse of transformation matrix obtained after independent component analysis.According to fresh data collection { Xj+1AndReal time data abnormity diagnosis model after being updated
Wherein Pj+1=Pj+αPΔFP(||Xj+1-Xj||,γP,Pj),Wj+1=Wj+αWΔFW(||Xj+1-Xj||,γW,Wj), Δ FP,ΔFW
It is block formula recursion correction function, γP,γWIt is correction forgetting factor, αP,αWIt is weight coefficient,It is respectively new
The average of fresh data set, variance and pivot number.
5. real time data abnormality diagnostic method according to claim 1, it is characterised in that the step (6) is specially:
The independent metadata matrix obtained by diagnostic model is X0=[x1,x2,…xn]T, n is independent entry number, xnFor n-th it is only
Vertical unit, then the kernel estimates of density function f (x) areWherein, i=1,2 ... n,
H is bandwidth matrices, and | H | is the determinant of H, and K () is kernel function, andBy
Above-mentioned Density Estimator function and 95% or 99% confidence level can try to achieve the control limit for characterizing normal operation range
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611101988.6A CN106710653A (en) | 2016-12-05 | 2016-12-05 | Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611101988.6A CN106710653A (en) | 2016-12-05 | 2016-12-05 | Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106710653A true CN106710653A (en) | 2017-05-24 |
Family
ID=58935590
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611101988.6A Pending CN106710653A (en) | 2016-12-05 | 2016-12-05 | Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106710653A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109447187A (en) * | 2018-12-25 | 2019-03-08 | 中南大学 | Method of Motor Fault Diagnosis and system |
CN109525453A (en) * | 2018-11-02 | 2019-03-26 | 长沙学院 | Networking CPS method for detecting abnormality and system based on node dependence |
CN109581923A (en) * | 2018-12-05 | 2019-04-05 | 深圳中广核工程设计有限公司 | Nuclear power station feed water control system |
CN110415844A (en) * | 2019-06-21 | 2019-11-05 | 杭州安脉盛智能技术有限公司 | Control rod drive mechanism performance estimating method and system based on movement crucial sequential of fault |
CN110887664A (en) * | 2019-12-04 | 2020-03-17 | 中国船舶工业系统工程研究院 | Method and device for establishing bearing fault identification model |
CN111602411A (en) * | 2018-02-27 | 2020-08-28 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and program |
CN112446436A (en) * | 2020-12-11 | 2021-03-05 | 浙江大学 | Anti-fuzzy unmanned vehicle multi-target tracking method based on generation countermeasure network |
CN112684778A (en) * | 2020-12-24 | 2021-04-20 | 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) | Steam generator water supply system diagnosis method based on multi-source information reinforcement learning |
CN116300774A (en) * | 2023-05-23 | 2023-06-23 | 蓝星智云(山东)智能科技有限公司 | Intermittent process visual monitoring method based on principal component analysis and nuclear density estimation |
CN116828413A (en) * | 2023-08-24 | 2023-09-29 | 长河信息股份有限公司 | Heterogeneous multi-source data signal receiving and transmitting system and method |
CN117834389A (en) * | 2024-03-04 | 2024-04-05 | 中国西安卫星测控中心 | Fault analysis method based on abnormal communication service characteristic element matrix |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008040812A (en) * | 2006-08-07 | 2008-02-21 | Keio Gijuku | Metabolism network estimation method and metabolism network estimation device |
WO2009046185A2 (en) * | 2007-10-02 | 2009-04-09 | The Regents Of The University Of Michigan | Method and apparatus for clustering and visualization of multicolor cytometry data |
CN101738435A (en) * | 2009-11-30 | 2010-06-16 | 浙江大学 | Dynamic fault diagnostic method of polymer aggregation in gas-solid fluidized bed reactor |
CN101825876A (en) * | 2009-11-30 | 2010-09-08 | 浙江大学 | Portable fault diagnosis tester applied to petrochemical production device |
CN103868692A (en) * | 2014-03-18 | 2014-06-18 | 电子科技大学 | Rotary machine fault diagnosis method based on kernel density estimation and K-L divergence |
CN104062968A (en) * | 2014-06-10 | 2014-09-24 | 华东理工大学 | Continuous chemical process fault detection method |
-
2016
- 2016-12-05 CN CN201611101988.6A patent/CN106710653A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008040812A (en) * | 2006-08-07 | 2008-02-21 | Keio Gijuku | Metabolism network estimation method and metabolism network estimation device |
WO2009046185A2 (en) * | 2007-10-02 | 2009-04-09 | The Regents Of The University Of Michigan | Method and apparatus for clustering and visualization of multicolor cytometry data |
CN101738435A (en) * | 2009-11-30 | 2010-06-16 | 浙江大学 | Dynamic fault diagnostic method of polymer aggregation in gas-solid fluidized bed reactor |
CN101825876A (en) * | 2009-11-30 | 2010-09-08 | 浙江大学 | Portable fault diagnosis tester applied to petrochemical production device |
CN103868692A (en) * | 2014-03-18 | 2014-06-18 | 电子科技大学 | Rotary machine fault diagnosis method based on kernel density estimation and K-L divergence |
CN104062968A (en) * | 2014-06-10 | 2014-09-24 | 华东理工大学 | Continuous chemical process fault detection method |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11422914B2 (en) | 2018-02-27 | 2022-08-23 | Omron Corporation | Metadata generation apparatus, metadata generation method, and program |
CN111602411B (en) * | 2018-02-27 | 2022-07-22 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and storage medium |
CN111602411A (en) * | 2018-02-27 | 2020-08-28 | 欧姆龙株式会社 | Metadata generation device, metadata generation method, and program |
CN109525453A (en) * | 2018-11-02 | 2019-03-26 | 长沙学院 | Networking CPS method for detecting abnormality and system based on node dependence |
CN109525453B (en) * | 2018-11-02 | 2021-01-01 | 长沙学院 | Networked CPS (control performance Standard) anomaly detection method and system based on node dependency relationship |
CN109581923B (en) * | 2018-12-05 | 2021-10-15 | 深圳中广核工程设计有限公司 | Water supply control system for nuclear power station |
CN109581923A (en) * | 2018-12-05 | 2019-04-05 | 深圳中广核工程设计有限公司 | Nuclear power station feed water control system |
CN109447187B (en) * | 2018-12-25 | 2021-06-15 | 中南大学 | Motor fault diagnosis method and system |
CN109447187A (en) * | 2018-12-25 | 2019-03-08 | 中南大学 | Method of Motor Fault Diagnosis and system |
CN110415844A (en) * | 2019-06-21 | 2019-11-05 | 杭州安脉盛智能技术有限公司 | Control rod drive mechanism performance estimating method and system based on movement crucial sequential of fault |
CN110887664B (en) * | 2019-12-04 | 2021-05-25 | 中国船舶工业系统工程研究院 | Method and device for establishing bearing fault identification model |
CN110887664A (en) * | 2019-12-04 | 2020-03-17 | 中国船舶工业系统工程研究院 | Method and device for establishing bearing fault identification model |
CN112446436A (en) * | 2020-12-11 | 2021-03-05 | 浙江大学 | Anti-fuzzy unmanned vehicle multi-target tracking method based on generation countermeasure network |
CN112684778A (en) * | 2020-12-24 | 2021-04-20 | 武汉第二船舶设计研究所(中国船舶重工集团公司第七一九研究所) | Steam generator water supply system diagnosis method based on multi-source information reinforcement learning |
CN116300774A (en) * | 2023-05-23 | 2023-06-23 | 蓝星智云(山东)智能科技有限公司 | Intermittent process visual monitoring method based on principal component analysis and nuclear density estimation |
CN116300774B (en) * | 2023-05-23 | 2023-08-08 | 蓝星智云(山东)智能科技有限公司 | Intermittent process visual monitoring method based on principal component analysis and nuclear density estimation |
CN116828413A (en) * | 2023-08-24 | 2023-09-29 | 长河信息股份有限公司 | Heterogeneous multi-source data signal receiving and transmitting system and method |
CN116828413B (en) * | 2023-08-24 | 2023-12-08 | 长河信息股份有限公司 | Heterogeneous multi-source data signal receiving and transmitting system and method |
CN117834389A (en) * | 2024-03-04 | 2024-04-05 | 中国西安卫星测控中心 | Fault analysis method based on abnormal communication service characteristic element matrix |
CN117834389B (en) * | 2024-03-04 | 2024-05-03 | 中国西安卫星测控中心 | Fault analysis method based on abnormal communication service characteristic element matrix |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106710653A (en) | Real-time data abnormal diagnosis method for monitoring operation of nuclear power unit | |
CN111582392B (en) | Multi-working-condition health state online monitoring method for key components of wind turbine generator | |
CN106682814B (en) | Wind turbine generator fault intelligent diagnosis method based on fault knowledge base | |
CN111597682B (en) | Method for predicting remaining life of bearing of gearbox of wind turbine | |
JP2021064370A (en) | Method and system for semi-supervised deep abnormality detection for large-scale industrial monitoring system based on time-series data utilizing digital twin simulation data | |
CN108680358A (en) | A kind of Wind turbines failure prediction method based on bearing temperature model | |
CN112487910A (en) | Fault early warning method and system for nuclear turbine system | |
CN111720296B (en) | Steam feed pump equipment abnormal state early warning method based on self-encoder | |
CN115453356B (en) | Power equipment operation state monitoring and analyzing method, system, terminal and medium | |
CN109356798B (en) | Wind driven generator gearbox state monitoring method based on co-integration analysis | |
CN107728059A (en) | A kind of pitch-controlled system state evaluating method | |
CN108506171A (en) | A kind of large-scale half direct-drive unit cooling system for gear box fault early warning method | |
CN116771610A (en) | Method for adjusting fault evaluation value of variable pitch system of wind turbine | |
CN118188342A (en) | Fan-oriented fault early warning and life prediction method | |
CN113313365A (en) | Degradation early warning method and device for primary air fan | |
DE102008037532A1 (en) | Automatic detection and reporting of wear of internal turbine components | |
CN112489841A (en) | Water level fault-tolerant control method for steam generator of nuclear power unit | |
EP1189126B1 (en) | Method for monitoring an installation | |
CN116467592A (en) | Production equipment fault intelligent monitoring method and system based on deep learning | |
CN110298132A (en) | A kind of turbine-generator units bearing throw trend forecasting method | |
CN115687864A (en) | XGboost-KDE-based fan main bearing fault early warning method | |
CN113323819A (en) | Fan safety chain fault analysis method and system based on fuzzy expert system | |
Tingting et al. | Early warning method for power station auxiliary failure considering large-scale operating conditions | |
CN113361840A (en) | Assessment method and system for abandoned wind electric quantity in wind power plant and outside plant | |
CN112855459A (en) | Early warning method for bearing fault of fan generator |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
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: 20170524 |