CN109115491A - A kind of evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis - Google Patents

A kind of evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis Download PDF

Info

Publication number
CN109115491A
CN109115491A CN201810939939.2A CN201810939939A CN109115491A CN 109115491 A CN109115491 A CN 109115491A CN 201810939939 A CN201810939939 A CN 201810939939A CN 109115491 A CN109115491 A CN 109115491A
Authority
CN
China
Prior art keywords
evidence
input
sample
value
sum
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810939939.2A
Other languages
Chinese (zh)
Other versions
CN109115491B (en
Inventor
徐晓滨
张德清
高海波
盛晨兴
侯平智
蒋鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
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 Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN201810939939.2A priority Critical patent/CN109115491B/en
Publication of CN109115491A publication Critical patent/CN109115491A/en
Application granted granted Critical
Publication of CN109115491B publication Critical patent/CN109115491B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/028—Acoustic or vibration analysis

Landscapes

  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • General Physics & Mathematics (AREA)
  • Complex Calculations (AREA)

Abstract

The invention discloses a kind of evidence fusion methods of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis.This method obtains the frequency domain character signal of the positions such as different faults mode bottom base and bracket first；The reference value that input feature vector is determined by K mean algorithm calculates the similarity distribution of sample；The cultellation statistical form of relationship between construction reflection input and fault mode, and it is converted to the evidence matrix table of input；The classification capacity and totality uncertainty of input information source are calculated based on rough set theory and comentropy；Determine the reliability and evidence weight of input information source；The evidence of input sample vector activation is merged using evidential reasoning rule and determines fault mode according to fusion results.Shafting propulsion system mechanical breakdown mode can be effectively estimated by the vibration signal that the sensor being mounted on ship obtains in this method, and at low cost, precision is high, realizes the real-time detection and Precise Diagnosis of Electrical Propulsion Ship shafting propulsion system mechanical breakdown.

Description

A kind of evidence fusion of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis Method
Technical field
The present invention relates to a kind of evidence fusion methods of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis, belong to Marine mechanical equipment condition monitoring and fault diagnosis technical field.
Background technique
Electrical Propulsion Ship shafting propulsion system mechanical equipment is equipment particularly important in marine system, is responsible for transmission boat Motoricity, working condition are mutually linked up with concerning shipping safety, and with economic benefit.The work of marine shafting propulsion system mechanical equipment Make that environment is more severe, accelerates the decline of equipment performance.The failure of equipment often causes a series of chain reactions, finally leads Cause systematic entirety that can decline, even equipment collapse and systemic breakdown, therefore to Electrical Propulsion Ship shafting propulsion system machine Tool fault diagnosis research be very it is necessary to it is significant.
Through investigating, domestic inspection and repair of ship technology also rests on the stage of periodic inspection at present, in maintenance process not only There are serious wasting of resources phenomenons, and there are some potential safety problemss.To guarantee equipment reliability service, production safety is ensured Property and economy, accurately judging to unit exception and malfunction is particularly important.Condition monitoring and fault diagnosis The use of technology can make service engineer monitor the operation irregularity of marine shafting propulsion system mechanical equipment, discovery danger in time And the failure of equipment safety operation, and necessary history run data is provided to the periodic maintenance of equipment, this is for reducing equipment Maintenance cost and promote its working efficiency and safety play the role of it is vital.System is promoted for modernization marine shafting System mechanical equipment can arrange the vibration signal of intensive vibrating sensor acquisition equipment in its each key position, from these The fault characteristic information extracted in vibration signal can reflect the various failures of equipment.Two problems are faced at this time, first is that needing A kind of simple and easy mode in engineering is found, the fault signature monitoring data of magnanimity are analyzed and processed, so that it is mentioned The diagnostic evidence of confession is objective credible；Second is that how to realize the synthesis of fault characteristic information.Usually provided using single sensor Characteristic information cannot consersion unit completely failure, the information for needing multiple sensors to provide merged to promote the essence of diagnosis Accuracy.
Summary of the invention
The purpose of the present invention is to propose to a kind of evidence fusions of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis Method, by the failure for installing the acquisition of vibration displacement sensor on the pedestal and bracket of Electrical Propulsion Ship shafting propulsion system Characteristic information determines the reference value of input feature vector first with K mean algorithm, then special from magnanimity failure based on input reference The likelihood reliability that each failure occurs is counted in sign data and obtains diagnostic evidence, recycles rough set theory and information entropy theory assessment The objective reliability and uncertainty of evidence, then the reliability factor and weight of conclusion evidence, are finally advised using evidential reasoning It then merges the evidence of input sample vector activation and determines fault mode according to fusion results, it is special that the method overcome single source failures Reference ceases the shortcomings that diagnostic techniques, at low cost, and precision is high, and realizes the real-time detection and Precise Diagnosis of failure.
The evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis proposed by the present invention, including with Under each step:
(1) Electrical Propulsion Ship shafting propulsion system mechanical breakdown set Θ={ F is set1,…,Fi,…,FN, FiIt represents I-th of failure in failure collection Θ, i=1,2 ..., N, N are failure number；Setting is mounted on the positions such as pedestal and bracket Vibration displacement sensor obtain position time domain vibration acceleration signal be { S1(r),…Sm(r),…SM(r) }, motor With the rotational speed of 150r/min-200r/min, the time domain vibration acceleration signal of 8s is acquired every time, under every kind of fault mode N times are acquired respectively, then are acquired sum=N*n times altogether, sampling number r=1,2 ..., sum, M are number of probes.
(2) the time domain vibration acceleration signal { S that will be sampled every time in step (1)1(r),…Sm(r),…SM(r) } it carries out Fast Fourier Transform (FFT) is transformed to corresponding frequency-region signal, and the amplitude for then choosing 1 times of fundamental frequency, 2 times of fundamental frequencies and 3 times of fundamental frequencies is made For fault characteristic signals { x1(r),…xj(r),…xJ(r) }, fault characteristic signals number J=3*M；By { x1(r),…xj (r),…xJ(r) } it is expressed as sample set U={ [x1(r),…xj(r),…xJ(r)] | r=1,2 ..., sum }, wherein x (r) =[x1(r),…xj(r),…xJIt (r)] is a sample vector.
(3) utilize K mean algorithm by the input feature vector signal x of each signal source in this sum vector samplej(r) it presses K aggregate of data { X is divided into according to sequence from small to large1,j,...Xk,j,...,XK,j, the corresponding cluster centre of aggregate of data press from It is small to being ordered as greatlySet characteristic it is discrete after value set V={ 1 ... t ..., K }, often One aggregate of data corresponds to a discrete value t in V, all input feature vector signal x inside aggregate of dataj(r) corresponding aggregate of data Discrete value t, the specific steps are as follows:
(3-1) is in input feature vector signal xjUnder data acquisition system { xj(1),...xj(r),...,xj(sum) } it is selected at random in Take K dataRespectively as several cluster { X1,j,...Xk,j,...,XK,jCenter.
(3-2) is for remaining data xj(r), its distance d for arriving each center is calculatedk(r), t=1,2 ... K, and The value is distributed to the number cluster X where nearest centerk,j。
(3-3) recalculates each several cluster centers
Wherein | Xk,j| represent number cluster Xk,jElement number.
(3-4) if each center no longer changes, i.e. clustering criteria function convergence obtains dividing the aggregate of data completed, each Aggregate of data corresponds to a discrete value in V；Otherwise, step (3-2) and step (3-3) are repeated.
(4) A is setj={ a1,j,a2,j,...ak,j,...,aK+1,j,aK+2,jIt is input feature vector signal xj(r) input ginseng Value set is examined, { a1,j, aK+2,jIt is respectively input feature vector signal xj(r) minimum value and maximum value, { a2,j,...,aK+1,jIt is step Suddenly input feature vector signal x in (3)jAccording to the cluster centre arranged from small to large
(5) by each of sum sample vector input feature vector signal xj(r) it encloses corresponding failure mode and becomes two First sample is to (xj(r),Fi)；It is respectively the form about reference value similarity with the variation of qualitative information conversion method, and constructs institute There is binary sample to the statistical form for carrying out cultellation in the form of similarity, the specific steps are as follows:
(5-1) binary sample is to (xj(r),Fi) input value xj(r) reference value a is matchedk,jSimilarity be distributed as
UT(xj(r))={ (ak,j,vk,j) | j=1 ..., J；K=1 ..., K+2 } (2a)
Wherein
vk',j=0 k'=1 ..., K+2, k' ≠ k, k+1 (2c)
vk,jIndicate input value xj(r) reference value a is matchedk,jSimilarity.
(5-2) according to step (5-1), binary sample is to (xj(r),Fi) form of similarity distribution can be converted into (vk,j,vk+1,j), wherein vk,jIndicate binary sample to (xj(r),Fi) in input value match reference value ak,j, while end value is Fi Similarity.
(5-3) according to step (5-1) and (5-2), by all binary samples in sample set U to the shape for being converted into similarity Formula can construct all binary samples to the statistical form for carrying out cultellation in the form of similarity with them, as shown in table 1 below, wherein bi,k, indicate all input value xj(r) reference value a is matchedk,jAnd failure mode is FiBinary sample to (xj(r),Fi) similar The sum of degree,Indicate that all end values are FiBinary sample to the sum of similarity,Indicate all Input value xj(r) reference value a is matchedk,jBinary sample to the sum of similarity, and have
1 binary sample of table is to (xj(r),Fi) cultellation statistical form
(6) according to the cultellation statistical form in step (5), it can get and work as input value xj(r) reference value a is takenk,jWhen, end value For FiReliability be
And haveIt then can define and correspond to reference value ak,jEvidence be
Therefore, evidence matrix table as shown in Table 2 can be constructed and carry out description information source input value xj(r) and result FiBetween Relationship.
Table 2 inputs information source xjEvidence matrix table
(7) objective reliability factor RS is definedjDescription input information source xjThe objective capability for differentiating failure mode, is specifically obtained Take that steps are as follows:
All samples of (7-1) after step (3) constitute the decision sheet form for rough set processing, such as 3 institute of table Show, input feature vector signal is conditional attribute at this time, and fault mode is decision attribute.
Decision table of the table 3 after sliding-model control
U x1 … xj … xJ D
x(1) t1,1 … t1,j … t1,J F1
… … … … … …
x(r) tr,1 … tr,j … tr,J Fi
… … … … … … …
x(sum) tsum,1 … tsum,j … tsum,J FN
Discrete value t in decision tabler,jThe decision attribute of ∈ V, D representative sample.
(7-2) can construct the equivalence relation of decision attribute D according to the decision table after discretization
RD=(x, y) ∈ U × U | D (x)=D (y) } (5)
Known equivalence relation can derive equivalence class [x]D=y ∈ U | (x, y) ∈ RD, finally by fault mode D to sample The division of this set U are as follows:
U/RD={ [x]D,x∈U} (6)
U/R can be more intuitively expressed asD={ Y1,...,Yi,...,YN, YiIt is the subset of sample set U respectively, and YiIn Sample decision attribute is all identical；Input information source xjAbout YiUpper approximate and lower aprons be respectively
(7-3) is obtained by step (7-2)Withxj (Yi), input information source x can be calculated by following formulajIt is objective reliable Sex factor
Wherein
Card () represents set element number.
(8) the reliability R of evidence is definedjDescription input information source xjAssess the ability of failure mode；Define the weight of evidence WjEvidence e is describedjCompared to the relative importance of other evidences, specific obtaining step is as follows:
(8-1) defines information source xjThe overall uncertainty of reliability is
The objective reliability factor RS that (8-2) is obtained according to step (6)j, the information source x of (8-1) acquisitionjTotality it is not true Fixed degree can be calculated input information source x by following formulajReliability
(8-3) sets evidence ejWeight WjEqual to corresponding reliability Rj, this is because the higher evidence of reliability ought to It is corresponding with higher evidence weight, it is clear that according to the actual situation, the weight of evidence can use the optimization method instruction of data-driven Practice.
(9) any one group of input sample vector x (r)=[x in sample set is given1(r),…xj(r),…xJ(r)], root The input information source reliability R that the input evidence matrix table and step (8) obtained according to step (6) obtainsjWith evidence weight Wj, can It is out of order type F using evidential reasoning rule-based reasoningi, the specific steps are as follows:
(9-1) is for input value xj(r), the section [a of certain two reference value composition is necessarily fallen intok,j,ak+1,j], at this time The corresponding evidence of the two reference valuesWithIt is activated, then input value xj(r) evidence can be by reference value evidenceWith It is obtained in the form of weighted sum
ej={ (Fi,pi,j), i=1 ..., N } (11a)
(9-2) obtains x using formula (11a) and formula (11b)1(r) and x2(r) evidence e1And e2, pass through evidential reasoning rule They are merged, obtaining fusion results is
O={ (Fi,pi,e(2)), i=1 ..., N } (12a)
(9-3) using step (9-1) and step (9-2) recursively find out the fusion results of J evidence for O (x (r))= {(Fi,pi,e(J)), i=1 ..., N }, the maximum p of valuei,e(J)Corresponding FiThe fault type as really occurred.
The evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis proposed by the present invention, according to adopting The input feature vector signal of collection, the reference value of input feature vector is determined by K mean algorithm；Sample is obtained using qualitative information conversion method This is to the comprehensive similarity about input and result reference value, and the cultellation for constructing reflection input reference and end value relationship is united Count table；The corresponding evidence of each reference value is obtained according to the table, constructs evidence matrix table；According to input information source objective classification ability The objective reliability of input information source is obtained with rough set theory；The totality for obtaining each signal source based on comentropy is uncertain Degree determines the reliability and evidence weight of input information source；The evidence for obtaining each group of input sample vector of sample set, utilizes card It it is theorized that rule obtains fusion results, therefrom reasoning obtains marine shafting propulsion system failure mode.It compiles according to the method for the present invention The program (translation and compiling environment LabView, C++ etc.) of system can be run on monitoring computer, and combination sensor, data collector Equal hardware form on-line monitoring system, configure on ship, to realize Electrical Propulsion Ship shafting propulsion system mechanical equipment Real-time state monitoring and fault diagnosis.
Detailed description of the invention
Fig. 1 is the flow diagram of the method for the present invention；
Fig. 2 is that Electrical Propulsion Ship shafting propulsion system simulated experiment platform data are adopted in the embodiment using the method for the present invention Collecting system structure chart；
Fig. 3 (a)-Fig. 3 (e) is the K mean value number fasciation of 5 fault signatures in the embodiment of the present invention into result figure.
Specific implementation method
A kind of evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis proposed by the present invention, Flow diagram is as shown in Figure 1, include following steps:
(1) Electrical Propulsion Ship shafting propulsion system mechanical breakdown set Θ={ F is set1,…,Fi,…,FN, FiIt represents I-th of failure in failure collection Θ, i=1,2 ..., N, N are failure number；Setting is mounted on the positions such as pedestal and bracket Vibration displacement sensor obtain position time domain vibration acceleration signal be { S1(r),…Sm(r),…SM(r) }, motor With the rotational speed of 150r/min-200r/min, the time domain vibration acceleration signal of 8s is acquired every time, under every kind of fault mode N times are acquired respectively, then are acquired sum=N*n times altogether, sampling number r=1,2 ..., sum, M are number of probes.
(2) the time domain vibration acceleration signal { S that will be sampled every time in step (1)1(r),…Sm(r),…SM(r) } it carries out Fast Fourier Transform (FFT) is transformed to corresponding frequency-region signal, and the amplitude for then choosing 1 times of fundamental frequency, 2 times of fundamental frequencies and 3 times of fundamental frequencies is made For fault characteristic signals { x1(r),…xj(r),…xJ(r) }, fault characteristic signals number J=3*M；By { x1(r),…xj (r),…xJ(r) } it is expressed as sample set U={ [x1(r),…xj(r),…xJ(r)] | r=1,2 ..., sum }, wherein x (r) =[x1(r),…xj(r),…xJIt (r)] is a sample vector.
(3) utilize K mean algorithm by the input feature vector signal x of each signal source in this sum vector samplej(r) it presses K aggregate of data { X is divided into according to sequence from small to large1,j,...Xk,j,...,XK,j, the corresponding cluster centre of aggregate of data press from It is small to being ordered as greatlySet characteristic it is discrete after value set V={ 1 ... t ..., K }, often One aggregate of data corresponds to a discrete value t in V, all input feature vector signal x inside aggregate of dataj(r) corresponding aggregate of data Discrete value t, the specific steps are as follows:
(3-1) is in input feature vector signal xjUnder data acquisition system { xj(1),...xj(r),...,xj(sum) } it is selected at random in Take K dataRespectively as several cluster { X1,j,...Xk,j,...,XK,jCenter.
(3-2) is for remaining data xj(r), its distance d for arriving each center is calculatedk(r), t=1,2 ... K, and The value is distributed to the number cluster X where nearest centerk,j。
(3-3) recalculates each several cluster centers
Wherein | Xk,j| represent number cluster Xk,jElement number.
(3-4) if each center no longer changes, i.e. clustering criteria function convergence obtains dividing the aggregate of data completed, each Aggregate of data corresponds to a discrete value in V；Otherwise, step (3-2) and step (3-3) are repeated.
(4) A is setj={ a1,j,a2,j,...ak,j,...,aK+1,j,aK+2,jIt is input feature vector signal xj(r) input ginseng Value set is examined, { a1,j, aK+2,jIt is respectively input feature vector signal xj(r) minimum value and maximum value, { a2,j,...,aK+1,jIt is step Suddenly input feature vector signal x in (3)jAccording to the cluster centre arranged from small to large
For the ease of the understanding to input reference, illustrate here.It is assumed that being measured respectively under 6 kinds of fault modes 100 sample vectors constitute sum=600 group sample set, share 5 input feature vector signal { x1(r),x2(r),x3(r),x4 (r),x5(r) }, by step (3) K mean algorithm by x1(here with x1For illustrate) 600 input feature vector signal values gather and be 4 aggregates of data, can obtain the corresponding cluster centre of each aggregate of data is { 0.056,0.201,0.2702,0.3426 }, input feature vector Signal x1(r) minimum value and maximum value is respectively 0.0404 and 0.417, then input feature vector signal x1(r) input reference collection Close A1={ 0.0404,0.056,0.201,0.2702,0.3426,0.417 }.
(5) by each of sum sample vector input feature vector signal xj(r) it encloses corresponding failure mode and becomes two First sample is to (xj(r),Fi)；It is respectively the form about reference value similarity with the variation of qualitative information conversion method, and constructs institute There is binary sample to the statistical form for carrying out cultellation in the form of similarity, the specific steps are as follows:
(5-1) binary sample is to (xj(r),Fi) input value xj(r) reference value a is matchedk,jSimilarity be distributed as
UT(xj(r))={ (ak,j,vk,j) | j=1 ..., J；K=1 ..., K+2 } (2a)
Wherein
vk',j=0 k'=1 ..., K+2, k' ≠ k, k+1 (2c)
vk,jIndicate input value xj(r) reference value a is matchedk,jSimilarity.
(5-2) according to step (5-1), binary sample is to (xj(r),Fi) form of similarity distribution can be converted into (vk,j,vk+1,j), wherein vk,jIndicate binary sample to (xj(r),Fi) in input value match reference value ak,j, while end value is Fi Similarity.
(5-3) according to step (5-1) and (5-2), by all binary samples in sample set U to the shape for being converted into similarity Formula can construct all binary samples to the statistical form for carrying out cultellation in the form of similarity with them, as shown in table 1 below, wherein bi,k,Indicate all input value xj(r) reference value a is matchedk,jAnd failure mode is FiBinary sample to (xj(r),Fi) similar The sum of degree,Indicate that all end values are FiBinary sample to the sum of similarity,Indicate all Input value xj(r) reference value a is matchedk,jBinary sample to the sum of similarity, and have
1 binary sample of table is to (xj(r),Fi) cultellation statistical form
In order to deepen to binary sample to (xj(r),Fi) similarity understanding, it is assumed here that a sample vector [(x1 (r),F1)]=[0.0532, F1], the input reference set of step (4) example hypothesis is continued to use, can be obtained by formula (2a)-(2c) defeated Enter value x1(r) similarity for matching reference value is v1,1=0.18 and v2,1=0.82；And then it can get sample to (x1(r),F1) Similarity is distributed (v1,1,v2,1)=(0.18,0.82).
Binary sample is to (x in order to facilitate understandingj(r),Fi) cultellation statistical form, continue to use sample set in step (4) with With reference to value set, all sum=600 binary samples of sample set are obtained to (x according to step (5-1) and (5-2)1(r),Fi) Similarity distribution, cultellation statistical form can be constructed, as shown in table 4 below
4 binary sample of table is to (x1(r),Fi) cultellation statistical form
(6) according to the cultellation statistical form in step (5), it can get and work as input value xj(r) reference value a is takenk,jWhen, end value For FiReliability be
And haveIt then can define and correspond to reference value ak,jEvidence be
Therefore, evidence matrix table as shown in Table 2 can be constructed and carry out description information source input value xj(r) and result FiBetween Relationship；
Table 2 inputs information source xjEvidence matrix table
Continue to continue to use input feature vector signal x in step (5)1Cultellation statistical form deepen to evidence matrix table shown in upper table Understanding.According to table 4, input value x can be obtained by formula (3) and formula (4)1(r) reference value a is taken2,1Corresponding evidence is when=0.0404
Similarly, the corresponding evidence of other reference values can be sought, then input x can be constructed1Evidence matrix table, such as table Shown in 5
Table 5 inputs information source x1Evidence matrix table
(7) objective reliability factor RS is definedjDescription input information source xjThe objective capability for differentiating failure mode, is specifically obtained Take that steps are as follows:
All samples of (7-1) after step (3) constitute the decision sheet form for rough set processing, such as 3 institute of table Show, input feature vector signal is conditional attribute at this time, and fault mode is decision attribute；
Decision table of the table 3 after sliding-model control
U x1 … xj … xJ D
x(1) t1,1 … t1,j … t1,J F1
… … … … … …
x(r) tr,1 … tr,j … tr,J Fi
… … … … … … …
x(sum) tsum,1 … tsum,j … tsum,J FN
Discrete value t in decision tabler,jThe decision attribute of ∈ V, D representative sample.
(7-2) can construct the equivalence relation of decision attribute D according to the decision table after discretization
RD=(x, y) ∈ U × U | D (x)=D (y) } (5)
Known equivalence relation can derive equivalence class [x]D=y ∈ U | (x, y) ∈ RD, finally by fault mode D to sample The division of this set U are as follows:
U/RD={ [x]D,x∈U} (6)
U/R can be more intuitively expressed asD={ Y1,...,Yi,...,YN, YiIt is the subset of sample set U respectively, and YiIn Sample decision attribute is all identical；Input information source xjAbout YiUpper approximate and lower aprons be respectively
(7-3) is obtained by step (7-2)Withxj (Yi), input information source x can be calculated by following formulajIt is objective reliable Sex factor
Wherein
Card () represents the element number of set.
(8) the reliability R of evidence is definedjDescription input information source xjAssess the ability of failure mode；Define the weight of evidence WjEvidence e is describedjCompared to the relative importance of other evidences, specific obtaining step is as follows:
(8-1) defines information source xjThe overall uncertainty of reliability is
The objective reliability factor RS that (8-2) is obtained according to step (6)j, the information source x of (8-1) acquisitionjTotality it is not true Fixed degree can be calculated input information source x by following formulajReliability:
(8-3) sets evidence ejWeight WjEqual to corresponding reliability Rj, this is because the higher evidence of reliability ought to It is corresponding with higher evidence weight, it is clear that according to the actual situation, the weight of evidence can use the optimization method instruction of data-driven Practice.
(9) any one group of input sample vector x (r)=[x in sample set is given1(r),…xj(r),…xJ(r)], root The input information source reliability R that the input evidence matrix table and step (8) obtained according to step (6) obtainsjWith evidence weight Wj, can It is out of order type F using evidential reasoning rule-based reasoningi, the specific steps are as follows:
(9-1) is for input value xj(r), the section [a of certain two reference value composition is necessarily fallen intok,j,ak+1,j], at this time The corresponding evidence of the two reference valuesWithIt is activated, then input value xj(r) evidence can be by reference value evidenceWithWith The form of weighted sum obtains
ej={ (Fi,pi,j), i=1 ..., N } (11a)
(9-2) obtains x using formula (11a) and formula (11b)1(r) and x2(r) evidence e1And e2, pass through evidential reasoning rule They are merged, obtaining fusion results is
O={ (Fi,pi,e(2)), i=1 ..., N } (12a)
(9-3) using step (9-1) and step (9-2) recursively find out the fusion results of J evidence for O (x (r))= {(Fi,pi,e(J)), i=1 ..., N }, the maximum p of valuei,e(J)Corresponding FiThe fault type as really occurred.
Below in conjunction with attached drawing, the embodiment of the method for the present invention is discussed in detail:
The flow chart of the method for the present invention is as shown in Figure 1, core is: acquisition different faults mode bottom base and bracket etc. The temporal signatures signal of position；The data of acquisition are passed through into Fast Fourier Transform (FFT), acquisition samples corresponding frequency domain character every time； Input feature vector signal is divided into multiple aggregates of data and determines corresponding several cluster centers and discrete value；It is determined according to several cluster centers defeated Enter the reference value set of characteristic signal；It calculates the similarity of binary sample pair and constructs the cultellation statistics of all binary samples pair Table；The evidence matrix table of input information source is constructed according to cultellation statistical form；The visitor of input information source is determined using rough set principle See reliability；Input information source x is calculated based on comentropyjOverall uncertainty, and determine input information source reliability and Weight；Finally using the evidence of evidential reasoning rule fusion input sample vector activation and from fusion results reasoning fault mode.
Below in conjunction with the data instance that marine shafting propulsion system simulated experiment platform in Fig. 2 acquires, the present invention is discussed in detail Each step of method.
1, the acquisition and pretreatment of experimental data
Vibration displacement sensor is placed in acquisition on the pedestal and backing positions of shafting propulsion system simulated experiment platform to turn Sub- vibration signal is respectively provided with failure " normal " (F1), " rotor unbalance " (four kinds of different degrees of F2,F3,F4,F5), " rotor Misalign " (F6), then failure collection is Θ={ F1,F2,F3,F4,F5,F6}；Motor is with the rotational speed of 180r/min, fundamental frequency 1X For 3Hz, the time domain vibration acceleration signal of 8s is acquired every time, acquires 210 times, is then acquired altogether respectively under every kind of fault mode 1260 times, 100 groups of total sum=600 group samples are randomly selected from every kind of fault mode as training sample set, remaining sample Test for inference pattern；The time domain vibration acceleration signal sampled every time carries out Fast Fourier Transform (FFT), is transformed to corresponding Frequency-region signal, when a fault has occurred, the increase situation of frequency and its amplitude that different failures is shown is also different, failure Vibrational energy mostly concentrate on 1X~3X, so it is special as failure to choose 1 times of fundamental frequency, the amplitude of 2 times of fundamental frequencies and 3 times of fundamental frequencies Reference number；Here 5 fault characteristic signals { x are chosen1,x2,x3,x4,x5As final input feature vector signal, it can wait until sample This set U={ [x1(r),x2(r),x3(r),x4(r),x5(r)] | r=1,2 ..., sum }, wherein x (r)=[x1(r),x2 (r),x3(r),x4(r),x5It (r) is a sample vector.
2, multiple aggregates of data are divided into fault signature using K mean value and determine corresponding several cluster centers and discrete value
According to the distribution of characteristic measurements, for each fault signature xjIt is divided into 4 aggregates of data and sets 4 discrete values Ordinate is added for this 600 groups of input feature vector signals in order to intuitively indicate the number cluster after dividing in V={ 1,2,3,4 }, from 0.01 to 6 are sequentially increased, and Fig. 3 (a)-Fig. 3 (e) illustrates the number cluster after dividing.Input feature vector signal { x1,x2,x3,x4,x5? Cluster centre be respectively { 0.056,0.201,0.2702,0.3426 }, { 0.0276,0.1513,0.2215,0.2828 }, { 0.0127,0.0257,0.1247,0.1444 }, { 0.0051,0.0097,0.0144,0.0196 } and 0.031,0.0426, 0.0604,0.0751}。
3, the selection of input feature vector signal reference value
Input feature vector signal x1Input reference set A1=0.0404,0.056,0.201,0.2702,0.3426, 0.417}；Input feature vector signal x2Input reference set A2=0.0193,0.0276,0.1513,0.2215,0.2828, 0.333}；Input feature vector signal x3Input reference set A3=0.0063,0.0127,0.0257,0.1247,0.1444, 0.1687}；Input feature vector signal x4Input reference set A4=0.0032,0.0051,0.0097,0.0144,0.0196, 0.0223}；Input feature vector signal x5Input reference set A5=0.0209,0.031,0.0426,0.0604,0.0751, 0.0946}。
4, binary sample is obtained to (xj(r),Fi) similarity form about reference value, binary sample is constructed to (xj(r), Fi) cultellation statistical form
Using all binary samples in the method for the present invention step (5) acquisition sum=600 group training sample set to (xj (r),Fi) similarity distribution, construct the cultellation statistical form such as table 1 in the method for the present invention step (5) shown in, input binary sample To (x1(r),Fi)、(x2(r),Fi)、(x3(r),Fi)、(x4(r),Fi) and (x5(r),Fi) cultellation statistical form respectively such as following table 6, shown in table 7, table 8, table 9 and table 10
6 binary sample of table is to (x1(r),Fi) cultellation statistical form
7 binary sample of table is to (x2(r),Fi) cultellation statistical form
8 binary sample of table is to (x3(r),Fi) cultellation statistical form
9 binary sample of table is to (x4(r),Fi) cultellation statistical form
10 binary sample of table is to (x5(r),Fi) cultellation statistical form
5, step (6) seeks input x according to the method for the present inventionjThe corresponding evidence of each reference value, and construct evidence matrix table
Step (5) obtains each input f according to the method for the present inventioniCultellation statistical form after, according to the step of the method for the present invention Suddenly (6) obtain input xjThe corresponding evidence of each reference value, and then construct input xjEvidence matrix table, as the following table 11, table 12, Shown in table 13, table 14 and table 15
Table 11 inputs x1Evidence matrix table
Table 12 inputs x2Evidence matrix table
Table 13 inputs x3Evidence matrix table
Table 14 inputs x4Evidence matrix table
Table 15 inputs x5Evidence matrix table
6, step (7) obtains the objective reliability factor RS for inputting information source according to the method for the present inventionj, detailed process is such as Under:
All samples after step of the present invention (3) constitute the decision sheet form for rough set processing, such as table 16 Shown, input feature vector signal is conditional attribute at this time, and fault mode is decision attribute；
Decision table of the table 16 after sliding-model control
U x1 x2 x3 x4 x5 D
x(1) 3 4 1 4 4 F1
x(2) 3 4 1 3 3 F1
x(3) 3 4 4 3 4 F1
… … … … … … …
x(sum) tsum,1 … tsum,j … tsum,J Fi
Input information source x can be calculated according to the formula (8a), formula (8b) and formula (8c) of the method for the present invention step (7)1、x2、x3,、 x4And x5Objective reliability factor be respectively
7, step (8) obtains the reliability for inputting information source according to the method for the present invention, and detailed process is as follows:
The formula (9) of step (8) can calculate the overall uncertainty of 5 information source reliabilities and be according to the method for the present invention
So utilizing (10) formula to can be obtained the reliability of input information source is respectively R1=0.9135, R2=0.8953, R3 =0.9360, R4=0.7803 and R5=0.7426.
8, according to the method for the present invention in step (9) reasoning test sample set every group of sample fault mode
Such as sample vector x (r)=[x1(r),x2(r),x3(r),x4(r),x5(r)]=[0.0532,0.034, 0.0139,0.0154,0.046], step (5) can obtain sample input x according to the method for the present invention1(r) with similarity v1,1= 0.18 and v2,1=0.82 activation evidenceWithInput x2(r) with similarity v2,2=0.9486 and v3,2=0.0514 activation card According toWithInput x3(r) with similarity v2,3=0.9113 and v3,3=0.0887 activation evidenceWithInput x4(r) with phase Like degree v44=0.8208 and v54=0.1792 activation evidenceWithInput x5(r) with similarity v3,5=0.8099 and v4,5= 0.1901 activation evidenceWith(11) formula of step (9) obtains e according to the method for the present invention1=[0.7844,0.1881, 0.0207,0.006,0,0], e2=[0.5623,0.0923,0.0058,0.0043,0,0.3353], e3=[0.2695, 0.1925,0.0602,0.1741,0.3028,0], e4=[0.2925,0.248,0.0689,0.2121,0.1785,0], e5= [0.2731,0.1092,0.0485,0.1106,0.2311,0.2275] then utilizes the evidential reasoning of formula step (9) formula (12) Fusion rule can obtain fusion results are as follows:
O (x (r))={ (F1,0.9245),(F2,0.0471),(F3,0.0023),(F4,0.0077),(F5,0.0131), (F6, 0.0052) } fault mode can be obtained as F1。
Likewise it is possible to calculate the fault mode of all test samples, and then it can get the diagnosis of test sample set just The fault diagnosis result of true rate, test sample is as shown in table 17, and really quasi- rate has reached 98.9% to total failare, reaches general diagnostic system System quasi- rate requirement really.
The fault diagnosis result of 17 test sample of table
Fault mode F1 F2 F3 F4 F5 F6
True quasi- rate 100% 97.3% 99.1% 98.2% 99.1% 100%

Claims (1)

1. a kind of evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis, it is characterised in that this method The following steps are included:
(1) Electrical Propulsion Ship shafting propulsion system mechanical breakdown set Θ={ F is set1,…,Fi,…,FN, FiRepresenting fault I-th of failure in set Θ, i=1,2 ..., N, N are failure number；Set the vibration being mounted on pedestal and backing positions The time domain vibration acceleration signal that displacement sensor obtains position is { S1(r),…Sm(r),…SM(r) }, motor with The rotational speed of 150r/min-200r/min acquires the time domain vibration acceleration signal of 8s every time, divides under every kind of fault mode Not Cai Ji n times, then altogether acquire sum=N*n times, sampling number r=1,2 ..., sum, M are number of probes；
(2) the time domain vibration acceleration signal { S that will be sampled every time in step (1)1(r),…Sm(r),…SM(r) } it carries out quick Fourier transformation is transformed to corresponding frequency-region signal, then chooses the amplitude of 1 times of fundamental frequency, 2 times of fundamental frequencies and 3 times of fundamental frequencies as event Hinder characteristic signal { x1(r),…xj(r),…xJ(r) }, fault characteristic signals number J=3*M；By { x1(r),…xj(r),…xJ (r) } it is expressed as sample set U={ [x1(r),…xj(r),…xJ(r)] | r=1,2 ..., sum }, wherein x (r)=[x1 (r),…xj(r),…xJIt (r)] is a sample vector；
(3) utilize K mean algorithm by the input feature vector signal x of each signal source in this sum vector samplej(r) according to from small K aggregate of data { X is divided into big sequence1,j,...Xk,j,...,XK,j, the corresponding cluster centre of aggregate of data is by from small to large It is ordered asSet characteristic it is discrete after value set V={ 1 ... t ..., K }, each number A discrete value t in V is corresponded to according to cluster, all input feature vector signal x inside aggregate of dataj(r) the corresponding aggregate of data is discrete Value t, the specific steps are as follows:
(3-1) is in input feature vector signal xjUnder data acquisition system { xj(1),...xj(r),...,xj(sum) } K are randomly selected in DataRespectively as several cluster { X1,j,...Xk,j,...,XK,jCenter；
(3-2) is for remaining data xj(r), its distance d for arriving each center is calculatedk(r), t=1,2 ... K, and should Value is distributed to the number cluster X where nearest centerk,j；
(3-3) recalculates each several cluster centers
Wherein | Xk,j| represent number cluster Xk,jElement number；
(3-4) if each center no longer changes, i.e. clustering criteria function convergence obtains dividing the aggregate of data completed, each data Cluster corresponds to a discrete value in V；Otherwise, step (3-2) and step (3-3) are repeated；
(4) A is setj={ a1,j,a2,j,...ak,j,...,aK+1,j,aK+2,jIt is input feature vector signal xj(r) input reference Set, { a1,j, aK+2,jIt is respectively input feature vector signal xj(r) minimum value and maximum value, { a2,j,...,aK+1,jIt is step (3) input feature vector signal x injAccording to the cluster centre arranged from small to large
(5) by each of sum sample vector input feature vector signal xj(r) it encloses corresponding failure mode and becomes binary sample This is to (xj(r),Fi)；It is respectively the form about reference value similarity with the variation of qualitative information conversion method, and constructs all two First sample is to the statistical form for carrying out cultellation in the form of similarity, the specific steps are as follows:
(5-1) binary sample is to (xj(r),Fi) input value xj(r) reference value a is matchedk,jSimilarity be distributed as
UT(xj(r))={ (ak,j,vk,j) | j=1 ..., J；K=1 ..., K+2 } (2a)
Wherein
vk',j=0 k'=1 ..., K+2, k' ≠ k, k+1 (2c)
vk,jIndicate input value xj(r) reference value a is matchedk,jSimilarity；
(5-2) according to step (5-1), binary sample is to (xj(r),Fi) form (v of similarity distribution can be converted intok,j, vk+1,j), wherein vk,jIndicate binary sample to (xj(r),Fi) in input value match reference value ak,j, while end value is FiPhase Like degree；
(5-3) according to step (5-1) and (5-2), by all binary samples in sample set U to the form for being converted into similarity, All binary samples can be constructed to the statistical form for carrying out cultellation in the form of similarity with them, as shown in table 1 below, wherein bi,k,Table Show all input value xj(r) reference value a is matchedk,jAnd failure mode is FiBinary sample to (xj(r),Fi) similarity With,Indicate that all end values are FiBinary sample to the sum of similarity,Indicate all inputs Value xj(r) reference value a is matchedk,jBinary sample to the sum of similarity, and have
1 binary sample of table is to (xj(r),Fi) cultellation statistical form
(6) according to the cultellation statistical form in step (5), it can get and work as input value xj(r) reference value a is takenk,jWhen, end value Fi's Reliability is
And haveIt then can define and correspond to reference value ak,jEvidence be
Therefore, evidence matrix table as shown in Table 2 can be constructed and carry out description information source input value xj(r) and result FiBetween pass System；
Table 2 inputs information source xjEvidence matrix table
(7) objective reliability factor RS is definedjDescription input information source xjThe objective capability of failure mode is differentiated, it is specific to obtain step It is rapid as follows:
All samples of (7-1) after step (3) constitute the decision sheet form for rough set processing, as shown in table 3, this When input feature vector signal be conditional attribute, fault mode is decision attribute；
Decision table of the table 3 after sliding-model control
U x1 … xj … xJ D x(1) t1,1 … t1,j … t1,J F1 … … … … … … x(r) tr,1 … tr,j … tr,J Fi … … … … … … … x(sum) tsum,1 … tsum,j … tsum,J FN
Discrete value t in decision tabler,jThe decision attribute of ∈ V, D representative sample；
(7-2) can construct the equivalence relation of decision attribute D according to the decision table after discretization
RD=(x, y) ∈ U × U | D (x)=D (y) } (5)
Known equivalence relation can derive equivalence class [x]D=y ∈ U | (x, y) ∈ RD, finally by fault mode D to sample set Close the division of U are as follows:
U/RD={ [x]D,x∈U} (6)
More intuitively it is expressed as U/RD={ Y1,...,Yi,...,YN, YiIt is the subset of sample set U respectively, and YiMiddle sample decision Attribute is all identical；Input information source xjAbout YiUpper approximate and lower aprons be respectively
(7-3) is obtained by step (7-2)Withxj (Yi), input information source x can be calculated by following formulajObjective reliability because Son
Wherein
Card () represents set element number；
(8) the reliability R of evidence is definedjDescription input information source xjAssess the ability of failure mode；Define the weight W of evidencejIt retouches State evidence ejCompared to the relative importance of other evidences, specific obtaining step is as follows:
(8-1) defines information source xjThe overall uncertainty of reliability is
The objective reliability factor RS that (8-2) is obtained according to step (6)j, the information source x of (8-1) acquisitionjOverall uncertainty, Input information source x can be calculated by following formulajReliability
(8-3) sets evidence ejWeight WjEqual to corresponding reliability Rj, this is because the higher evidence of reliability ought to correspond to There is higher evidence weight, the weight of evidence can use the optimization method training of data-driven；
(9) any one group of input sample vector x (r)=[x in sample set is given1(r),…xj(r),…xJ(r)], according to step Suddenly the input information source reliability R that the input evidence matrix table and step (8) that (6) obtain obtainjWith evidence weight Wj, available Evidential reasoning rule-based reasoning is out of order type Fi, the specific steps are as follows:
(9-1) is for input value xj(r), the section [a of certain two reference value composition is necessarily fallen intok,j,ak+1,j], at this time this two The corresponding evidence of a reference valueWithIt is activated, then input value xj(r) evidence can be by reference value evidenceWithWith weighting The form of sum obtains
ej={ (Fi,pi,j), i=1 ..., N } (11a)
(9-2) obtains x using formula (11a) and formula (11b)1(r) and x2(r) evidence e1And e2, by evidential reasoning rule to it Merged, obtaining fusion results is
O={ (Fi,pi,e(2)), i=1 ..., N } (12a)
(9-3) is O (x (r))={ (F using the fusion results that step (9-1) and step (9-2) recursively find out J evidencei, pi,e(J)), i=1 ..., N }, the maximum p of valuei,e(J)Corresponding FiThe fault type as really occurred.
CN201810939939.2A 2018-10-16 2018-10-16 Evidence fusion method for mechanical fault diagnosis of shafting propulsion system of electrically-propelled ship Active CN109115491B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810939939.2A CN109115491B (en) 2018-10-16 2018-10-16 Evidence fusion method for mechanical fault diagnosis of shafting propulsion system of electrically-propelled ship

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810939939.2A CN109115491B (en) 2018-10-16 2018-10-16 Evidence fusion method for mechanical fault diagnosis of shafting propulsion system of electrically-propelled ship

Publications (2)

Publication Number Publication Date
CN109115491A true CN109115491A (en) 2019-01-01
CN109115491B CN109115491B (en) 2020-04-24

Family

ID=64853225

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810939939.2A Active CN109115491B (en) 2018-10-16 2018-10-16 Evidence fusion method for mechanical fault diagnosis of shafting propulsion system of electrically-propelled ship

Country Status (1)

Country Link
CN (1) CN109115491B (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109765786A (en) * 2019-01-25 2019-05-17 杭州电子科技大学 A kind of electricity based on evidence filtering is pushed boat machine shaft imbalance fault detection method
CN109974837A (en) * 2019-04-18 2019-07-05 杭州电子科技大学 A kind of Ship Structure damnification recognition method of Process Based
CN110057581A (en) * 2019-04-29 2019-07-26 杭州电子科技大学 Rotary machinery fault diagnosis method based on interval type reliability rule-based reasoning
CN110146279A (en) * 2019-05-21 2019-08-20 杭州电子科技大学 A kind of marine shafting imbalance fault diagnostic method based on vector evidential reasoning
CN110196165A (en) * 2019-04-29 2019-09-03 杭州电子科技大学 A kind of rotating machinery ball bearing method for diagnosing faults based on K-means cluster and evidential reasoning
CN110414407A (en) * 2019-07-23 2019-11-05 武汉理工大学 Propeller for vessels volume gas method of identification based on evidential reasoning rule and adaptive boosting
CN112001124A (en) * 2020-08-27 2020-11-27 杭州电子科技大学 ER rule-based identification method for key functional units of ship electric propulsion system
CN112016471A (en) * 2020-08-27 2020-12-01 杭州电子科技大学 Rolling bearing fault diagnosis method under incomplete sample condition
CN112686279A (en) * 2021-02-08 2021-04-20 杭州电子科技大学 Gearbox fault diagnosis method based on K-means clustering and evidence fusion
CN113269057A (en) * 2021-05-07 2021-08-17 杭州电子科技大学 Motor rotor fault prediction method based on confidence rule base reasoning
CN114326652A (en) * 2021-11-23 2022-04-12 中国船舶工业集团公司第七0八研究所 Marine water jet propeller fault diagnosis system
CN115391399A (en) * 2022-08-24 2022-11-25 厦门美亚亿安信息科技有限公司 Method and system for collecting service system data

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2110778B1 (en) * 2008-04-14 2013-07-31 Honeywell International Inc. Determining corrective actions using a geometrically-based determination of sufficient confidence
CN104331543A (en) * 2014-10-17 2015-02-04 中国船舶重工集团公司第七一二研究所 Fault diagnostic expert system for marine electrical propulsion system and establishing method thereof
CN104408322A (en) * 2014-12-08 2015-03-11 杭州电子科技大学 Rotating mechanical device fault diagnosis method capable of synthesizing multisource fault probability likelihood credibility
CN105923014B (en) * 2016-04-27 2018-01-02 杭州电子科技大学 A kind of track transition Amplitude Estimation method based on evidential reasoning rule
CN108528475A (en) * 2018-04-13 2018-09-14 杭州电子科技大学 A kind of track transition fault alarm method based on multi-level fusion

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2110778B1 (en) * 2008-04-14 2013-07-31 Honeywell International Inc. Determining corrective actions using a geometrically-based determination of sufficient confidence
CN104331543A (en) * 2014-10-17 2015-02-04 中国船舶重工集团公司第七一二研究所 Fault diagnostic expert system for marine electrical propulsion system and establishing method thereof
CN104408322A (en) * 2014-12-08 2015-03-11 杭州电子科技大学 Rotating mechanical device fault diagnosis method capable of synthesizing multisource fault probability likelihood credibility
CN105923014B (en) * 2016-04-27 2018-01-02 杭州电子科技大学 A kind of track transition Amplitude Estimation method based on evidential reasoning rule
CN108528475A (en) * 2018-04-13 2018-09-14 杭州电子科技大学 A kind of track transition fault alarm method based on multi-level fusion

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
徐晓滨 等: "基于证据推理规则的信息融合故障诊断方法", 《控制理论与应用》 *

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109765786A (en) * 2019-01-25 2019-05-17 杭州电子科技大学 A kind of electricity based on evidence filtering is pushed boat machine shaft imbalance fault detection method
CN109974837A (en) * 2019-04-18 2019-07-05 杭州电子科技大学 A kind of Ship Structure damnification recognition method of Process Based
CN109974837B (en) * 2019-04-18 2022-05-20 杭州电子科技大学 Ship structure damage identification method based on rule reasoning
CN110196165A (en) * 2019-04-29 2019-09-03 杭州电子科技大学 A kind of rotating machinery ball bearing method for diagnosing faults based on K-means cluster and evidential reasoning
CN110057581B (en) * 2019-04-29 2020-10-09 杭州电子科技大学 Rotary machine fault diagnosis method based on interval type credibility rule reasoning
CN110057581A (en) * 2019-04-29 2019-07-26 杭州电子科技大学 Rotary machinery fault diagnosis method based on interval type reliability rule-based reasoning
CN110146279A (en) * 2019-05-21 2019-08-20 杭州电子科技大学 A kind of marine shafting imbalance fault diagnostic method based on vector evidential reasoning
CN110414407A (en) * 2019-07-23 2019-11-05 武汉理工大学 Propeller for vessels volume gas method of identification based on evidential reasoning rule and adaptive boosting
CN110414407B (en) * 2019-07-23 2022-12-23 武汉理工大学 Ship propeller air entrainment identification method based on evidence reasoning rule and self-adaptive lifting
CN112001124B (en) * 2020-08-27 2023-09-05 杭州电子科技大学 ER rule-based identification method for key functional units of ship electric propulsion system
CN112001124A (en) * 2020-08-27 2020-11-27 杭州电子科技大学 ER rule-based identification method for key functional units of ship electric propulsion system
CN112016471A (en) * 2020-08-27 2020-12-01 杭州电子科技大学 Rolling bearing fault diagnosis method under incomplete sample condition
CN112016471B (en) * 2020-08-27 2024-02-23 杭州电子科技大学 Rolling bearing fault diagnosis method under incomplete sample condition
CN112686279A (en) * 2021-02-08 2021-04-20 杭州电子科技大学 Gearbox fault diagnosis method based on K-means clustering and evidence fusion
CN112686279B (en) * 2021-02-08 2024-03-22 杭州电子科技大学 Gear box fault diagnosis method based on K-means clustering and evidence fusion
CN113269057A (en) * 2021-05-07 2021-08-17 杭州电子科技大学 Motor rotor fault prediction method based on confidence rule base reasoning
CN113269057B (en) * 2021-05-07 2024-06-07 杭州电子科技大学 Motor rotor fault prediction method based on confidence rule base reasoning
CN114326652A (en) * 2021-11-23 2022-04-12 中国船舶工业集团公司第七0八研究所 Marine water jet propeller fault diagnosis system
CN115391399A (en) * 2022-08-24 2022-11-25 厦门美亚亿安信息科技有限公司 Method and system for collecting service system data

Also Published As

Publication number Publication date
CN109115491B (en) 2020-04-24

Similar Documents

Publication Publication Date Title
CN109115491A (en) A kind of evidence fusion method of Electrical Propulsion Ship shafting propulsion system mechanical fault diagnosis
CN104697767B (en) Rotor system fault diagnosis method and device based on vibration analysis
Lei et al. EEMD method and WNN for fault diagnosis of locomotive roller bearings
CN106226074B (en) Rotary machinery fault diagnosis method based on convolutional neural networks and small echo gray-scale map
CN102520697B (en) Onsite information preprocessing method of remote cooperative diagnosis
Kankar et al. Vibration-based fault diagnosis of a rotor bearing system using artificial neural network and support vector machine
CN109613428A (en) It is a kind of can be as system and its application in motor device fault detection method
CN104408322B (en) Rotating mechanical device fault diagnosis method capable of synthesizing multisource fault probability likelihood credibility
CN103048133B (en) Bayesian network-based rolling bearing fault diagnosis method
CN102797671A (en) Fault detection method and device of reciprocating compressor
CN104897403A (en) Self-adaption fault diagnosis method based on permutation entropy (PE) and manifold-based dynamic time warping (MDTW)
CN109033612A (en) A kind of Diagnosis Method of Transformer Faults based on vibration noise and BP neural network
CN101660969A (en) Intelligent fault diagnosis method for gear box
CN105866250A (en) Vibration-based identification method for cracks of ventilating vane
CN104112072A (en) Operating modal parameter identification method for principal component analysis on basis of wavelet threshold denoising
CN109932179A (en) A kind of rolling bearing fault testing method based on the reconstruct of DS Adaptive spectra
CN109211546A (en) Rotary machinery fault diagnosis method based on noise reduction autocoder and incremental learning
CN109029995A (en) Bearing apparatus method for monitoring state based on cluster and multilayer autoencoder network
CN109253872A (en) A kind of rotor operation state monitoring method based on CEEMDAN
CN105956514A (en) Helicopter rotor abnormity detecting method driven by vibration data
CN113283292B (en) Method and device for diagnosing faults of underwater micro-propeller
CN106762343A (en) The diagnostic method of the hydraulic generator set thrust bearing failure based on online data
CN110057581A (en) Rotary machinery fault diagnosis method based on interval type reliability rule-based reasoning
CN111239249A (en) Ventilator blade crack fault diagnosis method based on Hilbert-Huang transform
CN116257747A (en) Early abnormality detection method for rotor system of fusion kurtosis and convolution self-encoder

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant