CN103713628B - Fault diagnosis method based on signed directed graph and data constitution - Google Patents

Fault diagnosis method based on signed directed graph and data constitution Download PDF

Info

Publication number
CN103713628B
CN103713628B CN201310753722.XA CN201310753722A CN103713628B CN 103713628 B CN103713628 B CN 103713628B CN 201310753722 A CN201310753722 A CN 201310753722A CN 103713628 B CN103713628 B CN 103713628B
Authority
CN
China
Prior art keywords
data
fault
reconstruct
state measurement
node
Prior art date
Application number
CN201310753722.XA
Other languages
Chinese (zh)
Other versions
CN103713628A (en
Inventor
王毓
魏岩
张峰华
杨煜普
Original Assignee
上海交通大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 上海交通大学 filed Critical 上海交通大学
Priority to CN201310753722.XA priority Critical patent/CN103713628B/en
Publication of CN103713628A publication Critical patent/CN103713628A/en
Application granted granted Critical
Publication of CN103713628B publication Critical patent/CN103713628B/en

Links

Abstract

The invention discloses a fault diagnosis method based on a signed directed graph and data constitution. The fault diagnosis method includes the steps of firstly, collecting normal data to conduct offline training, decomposing pre-processed data through a PCA, and solving a control limit of an SPE; secondly, setting up an SDG model according to a system flow graph, and determining parameters of all variables V-mask after setting the missing alarm rate, the false alarm rate and the detection offset; thirdly, collecting process data of system unknown state in real time, monitoring a CUSUM of all the variables and the SPE of a sample, if the SPE exceeds the control limit, diagnosing that a fault happens to a system, determining an effective node through the CUSUM, and conducting reconstitution on data in all compatible path directions by searching for all possible compatible paths, wherein the direction with the largest fault isolation index is taken as a real fault propagation direction, the starting node in the direction is taken as a causal variable of the fault, and an event which causes the node to be abnormal is taken as the root cause of the fault.

Description

Method for diagnosing faults based on the reconstruct of signed digraph data
Technical field
The present invention relates to the fault diagnosis system field of multivariate complication system, more particularly, to one kind are based on signed digraph The method for diagnosing faults of data reconstruct.
Background technology
Constantly develop towards large-scale, direction that is intelligent and complicating with industrial process control system, ask safely Topic is increasingly becoming one of subject matter of everybody care.One of core component as Process Control System, dynamical system Fault detection and diagnosis (fdd) technology be exactly in order to adapt to industrial system to improve reliability and reduce accident risk needs And what formation and development was got up.In the past few decades, troubleshooting issue has obtained the extensive concern of Chinese scholars, gushes Reveal the various methods with regard to fault detect and isolation.These methods can be divided into qualitative method and quantitative analyses two on the whole Big class.Wherein, in quantitative analysis method, it is to pay close attention to most methods in recent years based on the method for data-driven.
In the method for diagnosing faults based on data-driven, it is to enjoy in recent years based on the method for diagnosing faults of multivariate statistics One of method of concern.Traditional is to differentiate which variable makes corresponding statistic surpass based on the multivariate statistical method of contribution plot Go out the most popular method of normal value, the variable that those have maximum contribution value to statistic is considered as cause fault former Dependent variable.But the maximum deficiency of the method is, contribution margin easily transfers to its dependent variable from a variable, and that is, contribution margin is maximum Variable be not necessarily the basic reason causing fault.Additionally, traditional is not accounted for based on the multivariate statistical method of contribution plot To fault propagation problem in systems, therefore it is difficult to detect the basic reason causing fault.
Therefore, those skilled in the art is devoted to developing a kind of fault diagnosis based on the reconstruct of signed digraph data Method, it is considered to fault propagation problem in systems, is diagnosed to be the basic reason causing fault effectively.
Content of the invention
In view of the drawbacks described above of prior art, the technical problem to be solved is to provide one kind to have based on symbol To the method for diagnosing faults of figure data reconstruct, the method uses square prediction error (spe) and accumulation and (cusum) statistic Carry out fault detect, by being reconstructed to the sample data of all compatible path direction of sdg when fault occurs, after reconstruct The maximum direction of residual error change is considered as the direction of propagation of fault, and start node in this direction is for becoming the reason causing trouble Amount, is diagnosed to be the basic reason causing fault effectively.
For achieving the above object, the invention provides a kind of based on signed digraph data reconstruct fault diagnosis side Method, comprises the steps:
Step one: a number of multivariate normal data in acquisition system running, and normal to described multivariate Data carries out pretreatment, as known state measurement data;
Step 2: pca decomposition is carried out to known state measurement data described in step one, by the measurement of described known state Data is decomposed into pivot part and residual error portion, and obtains the control of the square prediction error of the measurement data of described known state Limit;
Step 3: the architectural characteristic according to described system and response characteristic, set up the oriented graphical diagram of described system (sdg), the node of described oriented graphical diagram is system univariate parameter, and sets rate of false alarm, rate of failing to report parameter, according to system Sdg determines the parameter value of each variable v-mask of described known state measurement data;
Step 4: a number of multivariable process data in real-time acquisition system running, and to described multivariate Process data carries out pretreatment, as unknown state measurement data, counts described unknown state measurement data accumulation and cusum system Evaluation and the square prediction error spe of described unknown state measurement data, if spe exceed step 2 control limit then it represents that System malfunctions;
Step 5: if system does not break down, repeat step four, if system malfunctions, double by v-mask Whether effectively arm judges the node of described sdg, and for the described process data variable more than underarm on v, node symbol is respectively "+" and "-" number;
Step 6: after determining all node states, search for all of compatible path in described sdg, i.e. adjacent node symbol It is multiplied for positive path;
Step 7: in all effective node direction and described compatible path direction, described unknown state measurement data is entered Line reconstruction, direction maximum for reconstruct index is just set to the propagation path of fault, judges rising on the propagation path of described fault The basic reason that beginning node occurs for causing trouble.
Described multivariate normal data pre-treatment step in the better embodiment of the present invention, in described step one For: first the described a number of multivariate normal data of collection is deducted the average of described multivariate normal data, then Variance divided by described multivariate normal data.
In another better embodiment of the present invention, in described step 2, pca method is chosen according to eigenvalue contribution rate Pivot is it is desirable to contribution rate is more than 85%.
In the better embodiment of the present invention, in described step 2, square prediction error control limit computing formula isWherein (1- α) × 100% is the described confidence level controlling limit, HereλiFor described pretreated described multivariate normal data sample The i-th big eigenvalue of this covariance matrix.
In another better embodiment of the present invention, described in described step 3, the parameter of v-mask is:And h=d*k, wherein k is the slope of v, and d is nearest sampled point with a distance from v fixed point, h For nearest sampled point with a distance from the upper underarm of v, α is rate of false alarm, and β is rate of failing to report, and δ is the side-play amount (sample that can detect The multiple of standard deviation), σxStandard deviation for sample.
In the better embodiment of the present invention, in described step 7 if reconstruct direction is exactly the propagation path of fault, The square prediction error of the unknown state measurement data after so reconstructing should obtain the reduction of maximum, the subtracting of square prediction error Shown in little degree such as formula (1):
η 2 = y s p e - y j s p e y s p e = f j 2 | | ζ ~ | | 2 y s p e - - - ( 1 )
Its span is wherein y between 0~1speFor reconstruct before sample square prediction error,After reconstruct The square prediction error of unknown state measurement data,For the projection in residual error space for the ζ, that is,C is the master of measurement First space projection matrix, in formula (1)The η making2Maximum direction is just the propagation path of fault, propagates The basic reason that start node on path then occurs for fault.
What the present invention provided considers fault in systems based on the method for diagnosing faults of signed digraph data reconstruct Propagation problem, by sdg data reconstruct combine, entered by the sample on compatible paths all to sdg when fault occurs Line reconstruction, the wherein maximum direction of reconstruct index are considered as the direction of propagation of physical fault, and the initial section on the direction of propagation Point is considered as the basic reason variable causing fault.By the architectural characteristic of system is combined with data-driven method, make Obtaining the more traditional contribution drawing method of the present invention has more preferable diagnosis effect to the basic reason causing fault.
Technique effect below with reference to design, concrete structure and generation to the present invention for the accompanying drawing is described further, with It is fully understood from the purpose of the present invention, feature and effect.
Brief description
Fig. 1 is the signed digraph of the te simulating experimental system of a preferred embodiment of the present invention;
Fig. 2 is the Troubleshooting Flowchart of a preferred embodiment of the present invention;
Fig. 3 is the test data square prediction error figure of the te procedure fault 1 of a preferred embodiment of the present invention;
Fig. 4 is Fault Isolation index η of a preferred embodiment of the present invention2Block diagram.
Specific embodiment
A kind of method for diagnosing faults based on the reconstruct of signed digraph data, is instructed offline by gathering normal data Practice, as known state measurement data, by pca, pretreated data is decomposed, and then obtain the control of spe Limit.Then according to system flow, set up sdg model, after setting rate of failing to report, rate of false alarm, detecting side-play amount, determine each variable v- The parameter of mask.Then the process data of real-time acquisition system unknown state, as unknown state measurement data, to each variable Cusum and sample spe is monitored, if spe has exceeded control limit, expression system be there occurs fault, united by cusum afterwards Metering determines effective node, by searching for all possible compatible path, the data on all compatible path directions is carried out with weight Structure, the maximum direction of its Fault Isolation index, is real fault propagation direction, start node in this direction is considered as The reason fault variable, and lead to the abnormal event of this node to be considered as the basic reason producing fault.
Fig. 1 is the signed digraph of fault diagnosis field typical emulation platform te process, and the node of in figure is a list Variable, the parameter of certain part in expression system, solid line represents active influence, and that is, start node increases, the section that solid arrow points to Point also increases, and dotted line is just contrary, represents node in figure circle,
Method for diagnosing faults flow process based on the reconstruct of signed digraph data is as shown in Fig. 2 specifically comprise the following steps that
Off-line modeling process:
1st, data acquisition.Acquisition system process a number of multivariate normal number under system normal operating condition According to.
2nd, data prediction, by data center's nondimensionalization.As known state measurement data.Changeable collect Amount normal data lines up matrix xn×m, wherein n is the number of collecting sample, and m is variable number, for xn×mIn each arrange to Amount xi(i=1 ... m), by formulaObtain the newly vectorial y that average is that 0 variance is 1i, whereinAnd siIt is respectively xiAverage and variance, such matrix xn×mIt is changed into matrix y after pretreatmentn×m.
3rd, pretreated data is carried out with pca decomposition, and calculates the spe of fault detect controlling limit.First obtain association side Difference matrix ytThe eigenvalue of y and individual features vector λiAnd pi(i=1 ... m), pivot number k being taken is according to eigenvalue summation 96% is set, and the corresponding characteristic vector of the big eigenvalue of front k constitutes principal component space, and other m-k eigenvalue is corresponding Characteristic vector constitutes residual error space.Thus sample vector y (m dimensional vector) can be decomposed intoWherein pivot PartRepresent sample vector y in principal component space spProjection, that is,Residual error portionRepresent sample vector y In residual error space srProjection, that is,Wherein i is unit matrix.The square prediction error formula of sample y isControl is limited toWherein (1- α) × 100% For confidence level, and have:
g s p e = θ 2 θ 1 , h s p e = θ 1 2 θ 2
HereλiIth feature value for covariance matrix.
4th, set up system symbol Directed Graph Model, set the whether abnormal parameters of detection node variable.According to system Procedure graph, draws the signed digraph of system, as shown in Figure 1.In order to during diagnosis, whether decision node is effective, here Whether normal range is exceeded come the cusum statistic of detection node variable by v-mask.V-mask is to separate in cusum table Go out the effective tool of abnormal data, its shape is in opening v font to the left, and its summit and nearest sampled point are in same level On line, both distances are d, and underarm is the slope line that slope is k.Data in v word all represents normal data, otherwise for extremely Data.The major parameter of v-mask can determine according to formula below:
k = δσ x 2
d = 2 δ 2 l n ( 1 - β α )
H=d*k
Wherein k is the slope of v, and d is nearest sampled point with a distance from v fixed point, and the sampled point that h is nearest is upper and lower from v The distance of arm.α is rate of false alarm;β is rate of failing to report;δ is the side-play amount (multiple of sample standard deviation) that can detect;σxFor sample Standard deviation.
Inline diagnosis process:
5th, real-time acquisition system process data, as unknown state measurement data.
6th, calculate the square prediction error spe of real time data, and it is controlled limit to be compared with spe, without super Cross control limit, return the next process data of 5 collections.
7 if it exceeds spe controls limit, and the accumulation counting each variable sentenced with cusum statistic and by the both arms of v-mask Whether the node variable of disconnected sdg is effective node variable, for the variable more than underarm on v, node symbol be respectively "+" and “-”.The computing formula of accumulation and cusum statistic is:Wherein m is number of samples,For institute's test sample This average,The average of i-th variable that step 1 calculates.
8th, search for all possible compatible path in sdg, that is, adjacent node symbol is multiplied for positive path in this path On, if the symbol of sdg be "+", correspondence direction vector analog value be 1, if the symbol of sdg be "-", correspondence direction The analog value of vector is -1, otherwise for 0.For example, the respective nodes detecting the 1st variable and the 3rd variable be "+" and "-", Other nodes are all 0 and the 1st variable and the 3rd variable on compatible path, then the direction vector on this path corresponds to position The value put is exactly corresponding node state 1, -1 or 0, unitization after obtain the direction vector in this path and be[1,0,- 1,0…0].
9th, in all effective node direction and compatible path direction, unknown state measurement data is surveyed and be reconstructed.Reconstruct Purpose be exactly will measure sample along reconstruct direction move, until from principal component space near with a distance from, if reconstruct direction be exactly The direction of propagation of fault, then the square prediction error of the unknown state measurement data after mobile should obtain the reduction of maximum. Shown in the reduction degree such as formula (1) of square prediction error:
η 2 = y s p e - y j s p e y s p e = f j 2 | | ζ ~ | | 2 y s p e - - - ( 1 ) ,
Its span is wherein y between 0~1speFor reconstruct before sample square prediction error,For reconstruct after not Know the square prediction error of state measurement data,For the projection in residual error space for the ζ, that is,In formula (1)The η making2Maximum direction is just the propagation path of fault, and the start node on propagation path is then fault The basic reason occurring.
Te is an emulation platform, and it simulates 20 faults, and Fig. 3 is te process (when a certain fault occurs) test data Square prediction error figure, in figure dotted line is to control limit.
Fig. 4 is Fault Isolation index η after reconstruct on all effective nodes and compatible path2Block diagram, abscissa Represent the numbering in effective node and compatible path, wherein numbering 1~12 is effective node, and numbering 13~15 is compatible road Footpath, is respectively as follows:
x a ↓ &doublerightarrow; m v 3 ↑
m v 10 ↓ &doublerightarrow; t 21 ↑
x c ↑ &doublerightarrow; p 7 ↑ &doublerightarrow; p 13 ↑ &doublerightarrow; p 16 ↑
As can be known from Fig. 4, the Fault Isolation index on this compatible path of numbering 15 is maximum, and therefore this path is fault Propagation path, i.e. xc change leads to greatly reactor pressure p7 to increase, thus leading to separator pressure p13 and desorbing pressure tower p16 to increase Greatly, due to only faulty 1 can make that xc component becomes big and xb component is constant in 20 faults of te process, can thus be concluded that it is fault 1 there occurs.
The preferred embodiment of the present invention described in detail above.It should be appreciated that those of ordinary skill in the art is no Need creative work just can make many modifications and variations according to the design of the present invention.Therefore, all technology in the art It is available that personnel pass through logical analysis, reasoning, or a limited experiment under this invention's idea on the basis of existing technology Technical scheme, all should be in the protection domain being defined in the patent claims.

Claims (6)

1. a kind of method for diagnosing faults based on the reconstruct of signed digraph data is it is characterised in that comprise the steps:
Step one: a number of multivariate normal data in acquisition system running, and to described multivariate normal data Carry out pretreatment, as known state measurement data;
Step 2: pca decomposition is carried out to known state measurement data described in step one, described known state measurement data is divided Solve as pivot part and residual error portion, and obtain the control limit of the square prediction error of described known state measurement data;
Step 3: the architectural characteristic according to described system and response characteristic, set up the oriented graphical diagram (sdg) of described system, institute The node stating oriented graphical diagram is system univariate parameter, and sets rate of false alarm, rate of failing to report parameter, and the sdg according to system determines The parameter value of each variable v-mask of described known state measurement data;
Step 4: a number of multivariable process data in real-time acquisition system running, and to described multivariable process Data carries out pretreatment, as unknown state measurement data, counts described unknown state measurement data accumulation and cusum statistical value And the square prediction error spe of described unknown state measurement data, if spe exceeds the control limit of step 2 then it represents that system Break down;
Step 5: if system does not break down, repeat step four, if system malfunctions, sentenced by the both arms of v-mask Effectively whether, for the described unknown state measurement data more than underarm on v-mask, node symbol divides the node of disconnected described sdg Not Wei "+" and "-" number;
Step 6: after determining all node states, search for all of compatible path in described sdg, that is, adjacent node symbol is multiplied For positive path;
Step 7: weight is carried out to described unknown state measurement data on all effective node direction and described compatible path direction Structure, direction maximum for reconstruct index is just set to the propagation path of fault, judges the initial section on the propagation path of described fault The basic reason that point occurs for causing trouble.
2. the method for diagnosing faults based on the reconstruct of signed digraph data as claimed in claim 1 is it is characterised in that described Described multivariate normal data pre-treatment step in step one is: first will be normal for the described a number of multivariate of collection Data deducts the average of described multivariate normal data, then divided by the variance of described multivariate normal data.
3. the method for diagnosing faults based on the reconstruct of signed digraph data as claimed in claim 1 is it is characterised in that described In step 2, pca method chooses pivot it is desirable to contribution rate is more than 85% according to eigenvalue contribution rate.
4. the method for diagnosing faults based on the reconstruct of signed digraph data as claimed in claim 1 is it is characterised in that described In step 2, square prediction error control limit computing formula isWhereinFor hspeCard side Distribution, (1- α) × 100% is the described confidence level controlling limit,HereλiFor described pretreated described multivariate normal data sample covariance matrix I-th big eigenvalue, k is the pivot number of pca, and m is variable number.
5. the method for diagnosing faults based on the reconstruct of signed digraph data as claimed in claim 1 is it is characterised in that described The parameter value of v-mask described in step 3 is:And h=d*k, wherein k is the oblique of v Rate, d is nearest sampled point with a distance from v fixed point, h be nearest sampled point with a distance from the upper underarm of v, α is rate of false alarm, β For rate of failing to report, δ is the side-play amount that can detect, and described side-play amount is the multiple of sample standard deviation, σxStandard deviation for sample.
6. the method for diagnosing faults based on the reconstruct of signed digraph data as claimed in claim 1 is it is characterised in that described In the step 7 if maximum direction of reconstruct index is exactly the propagation path of fault, then the unknown state measurement data after reconstruct Square prediction error should obtain maximum reduction, the computing formula of the reduction degree of square prediction error is:
η 2 = y s p e - y j s p e y s p e = f j 2 | | ζ ~ | | 2 y s p e
Its span is wherein y between 0~1speFor reconstruct before sample square prediction error,For reconstruct after unknown The square prediction error of state measurement data, f then represents fault size, and ζ represents the direction vector in compatible path,For ζ residual The projection of difference space, that is,C is the principal component space projection matrix of measurement,Make η2Maximum side To the propagation path being just fault, the basic reason that the start node on the propagation path of fault then occurs for fault.
CN201310753722.XA 2013-12-31 2013-12-31 Fault diagnosis method based on signed directed graph and data constitution CN103713628B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310753722.XA CN103713628B (en) 2013-12-31 2013-12-31 Fault diagnosis method based on signed directed graph and data constitution

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310753722.XA CN103713628B (en) 2013-12-31 2013-12-31 Fault diagnosis method based on signed directed graph and data constitution

Publications (2)

Publication Number Publication Date
CN103713628A CN103713628A (en) 2014-04-09
CN103713628B true CN103713628B (en) 2017-01-18

Family

ID=50406679

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310753722.XA CN103713628B (en) 2013-12-31 2013-12-31 Fault diagnosis method based on signed directed graph and data constitution

Country Status (1)

Country Link
CN (1) CN103713628B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104182642B (en) * 2014-08-28 2017-06-09 清华大学 A kind of fault detection method based on rarefaction representation
EP3048613B1 (en) * 2015-01-20 2019-06-12 ABB Schweiz AG Method for analysis of plant disturbance propagations
CN104794013B (en) * 2015-03-20 2018-03-13 百度在线网络技术(北京)有限公司 Alignment system running status, the method and device for establishing system running state model
CN104793604B (en) * 2015-04-10 2017-05-17 浙江大学 Principal component tracking based industrial fault monitoring method and application thereof
CN104965506B (en) * 2015-06-09 2017-12-05 南京航空航天大学 One kind is based on adjustable parameter Distributed Flight Control System real-time fault diagnosis method
CN105634796A (en) * 2015-12-22 2016-06-01 山西合力创新科技有限公司 Network device failure prediction and diagnosis method
CN105974356B (en) * 2016-07-22 2019-02-05 国网浙江省电力公司电力科学研究院 A kind of automatic power-measuring calibrating pipeline stall diagnostic method
CN106933097A (en) * 2017-05-15 2017-07-07 青岛科技大学 A kind of Fault Diagnosis for Chemical Process method based on multi-level optimization PCC SDG
CN109187060B (en) * 2018-07-31 2019-10-18 同济大学 The detection of train speed sensor abnormal signal and axis locking method for diagnosing faults
CN109947076B (en) * 2019-03-14 2020-06-02 华中科技大学 Industrial process fault diagnosis method based on Bayesian information criterion

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3953717A (en) * 1973-09-10 1976-04-27 Compagnie Honeywell Bull (Societe Anonyme) Test and diagnosis device
JP2000075923A (en) * 1998-09-03 2000-03-14 Yamatake Corp Abnormality diagnostic device
CN1655082A (en) * 2005-01-27 2005-08-17 上海交通大学 Non-linear fault diagnosis method based on core pivot element analysis
CN101995880A (en) * 2010-10-28 2011-03-30 中国石油化工股份有限公司 System for diagnosing and testing abnormal operating condition during petrochemical process
CN102004486A (en) * 2010-09-26 2011-04-06 中国石油化工股份有限公司 Hybrid fault diagnosis method based on qualitative signed directed graph in petrochemical process

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3953717A (en) * 1973-09-10 1976-04-27 Compagnie Honeywell Bull (Societe Anonyme) Test and diagnosis device
JP2000075923A (en) * 1998-09-03 2000-03-14 Yamatake Corp Abnormality diagnostic device
CN1655082A (en) * 2005-01-27 2005-08-17 上海交通大学 Non-linear fault diagnosis method based on core pivot element analysis
CN102004486A (en) * 2010-09-26 2011-04-06 中国石油化工股份有限公司 Hybrid fault diagnosis method based on qualitative signed directed graph in petrochemical process
CN101995880A (en) * 2010-10-28 2011-03-30 中国石油化工股份有限公司 System for diagnosing and testing abnormal operating condition during petrochemical process

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
PCA_SDG在TEP多源故障诊断中的应用;田娟 等;《软件》;20121231;第33卷(第1期);第58-60页 *
SWE_IPCA方法在传感器故障诊断中的应用;王通 等;《仪器仪表学报》;20130831;第34卷(第8期);第1842-1846页 *
基于PCA和SDG的传感器故障诊断方法研究及应用;高东 等;《系统仿真学报》;20110331;第23卷(第3期);第567-573页 *
基于改进SDG的电站热力系统故障诊断方法研究;曹文亮 等;《中国电机工程学报》;20051231;第25卷(第23期);第124-128页 *

Also Published As

Publication number Publication date
CN103713628A (en) 2014-04-09

Similar Documents

Publication Publication Date Title
CN103983453B (en) A kind of executing agency of aero-engine and the differentiating method of sensor fault diagnosis
CN102944418B (en) Wind turbine generator group blade fault diagnosis method
CN102221651B (en) Fault on-line diagnosis and early warning method of flameproof dry-type transformer for mine
CN103699698B (en) A kind of being based on improves Bayesian rail transit fault identification method and system
CN102663412B (en) Power equipment current-carrying fault trend prediction method based on least squares support vector machine
CN104595170B (en) A kind of air compressor machine monitoring and diagnosis system and method for self-adaptive kernel gauss hybrid models
CN105204496B (en) The method and system of EMUs air brake control system sensor fault diagnosis
CN102411106B (en) Fault monitoring method and device for power transformer
CN106199305B (en) Underground coal mine power supply system dry-type transformer insulation health state evaluation method
CN104102773B (en) A kind of equipment fault early-warning and state monitoring method
CN103245861B (en) A kind of transformer fault diagnosis method based on Bayesian network
CN102208028B (en) Fault predicting and diagnosing method suitable for dynamic complex system
CN105279365B (en) For the method for the sample for learning abnormality detection
CN105846780B (en) A kind of photovoltaic module method for diagnosing faults based on decision-tree model
CN102496069B (en) Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP)
CN105095963B (en) A kind of Precise Diagnosis and the method for prediction wind-tunnel facilities failure
CN102934312B (en) Energy production system and control thereof
CN101251564B (en) Method for diagnosis failure of power transformer using extension theory and rough set theory
CN102497024B (en) Intelligent warning system based on integer programming
CN102063109B (en) Neural network-based subway train fault diagnosis device and method
CN102779230B (en) State analysis and maintenance decision judging method of power transformer system
CN103996077A (en) Electric equipment fault forecasting method based on multi-dimension time sequence
CN107179503A (en) The method of Wind turbines intelligent fault diagnosis early warning based on random forest
CN105044559B (en) The subregion method for diagnosing faults of grounding net of transformer substation
CN105004498A (en) Vibration fault diagnosis method of hydroelectric generating set

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
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20170118

Termination date: 20191231