CN109213121A - A kind of fan braking system card cylinder method for diagnosing faults - Google Patents
A kind of fan braking system card cylinder method for diagnosing faults Download PDFInfo
- Publication number
- CN109213121A CN109213121A CN201810888737.XA CN201810888737A CN109213121A CN 109213121 A CN109213121 A CN 109213121A CN 201810888737 A CN201810888737 A CN 201810888737A CN 109213121 A CN109213121 A CN 109213121A
- Authority
- CN
- China
- Prior art keywords
- signal
- evidence
- layer
- output
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0259—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the response to fault detection
- G05B23/0262—Confirmation of fault detection, e.g. extra checks to confirm that a failure has indeed occurred
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a kind of fan braking system card cylinder method for diagnosing faults, including step 1, acquisition brake clearance-time response curve data；2, data prediction；3, fault feature vector is extracted from brake clearance-time response curve using wavelet packet；4, training parameter is determined according to the dimension of input signal；5, the resulting input signal sample of step 2 is divided into training set and test set；6, network training is carried out according to the resulting training parameter of step 4, obtains fuzzy neural network model；7, fuzzy neural network model optimizes, and tests fuzzy neural network model；8, it is merged as independent evidence with D-S evidence theory using the recognition result of each fuzzy neural network, obtains comprehensive diagnostic result；9: output diagnostic result.The invention has the advantages that carrying out the card cylinder fault diagnosis of fan braking system using multiple technologies means, the accuracy rate of fault diagnosis is improved.
Description
Technical field
The invention belongs to technical field of wind power generation, and in particular to a kind of fan braking system card cylinder method for diagnosing faults.
Background technique
The reliability of fan braking system is improved, usually there are two types of methods: first is that improving the reliability of braking system；Second is that
The operating status of braking system detect and then to its fault diagnosis.In terms of improving reliability, due to blower braking system
It is typically all caused by the coefficient of friction significant decrease of brake shoe material that system, which breaks down, and the factor for reducing coefficient of friction mainly has lock
Watt temperature rise, slip velocity and pressure variation.In three factors, the influence of wherein temperature rise is maximum, since blower is in load shape
It runs at high speed under state, shoe temperature can steeply rise in braking process, reduction coefficient of friction, and the change meeting of coefficient of friction
Braking system performance is had a huge impact.There are also the analytic expressions that document is deduced brake shoe temperature rise at present, but parse
Formula be it is nonlinear, also considerably complicated in structure, effect is poor in practical applications.
In terms of fault diagnosis, the failure of fan braking system is always the difficult point of Fault Diagnosis of Fan, with big data
Theoretical development, excavates signal from the bulk information in information system, with artificial intelligence carry out the method for signal fused at
For fan braking system fault diagnosis new direction.The result of current artificial intelligence fault diagnosis, diagnosis is unsatisfactory, accurately
Rate is low, and rate of false alarm is high.
Many field failure examples show that the oil cylinder piston card cylinder of braking system is a common failure.Based on this, originally
Invention proposes a kind of method based on artificial intelligence, diagnoses to the braking system card cylinder failure of blower.The present invention is made
Fuzzy neural network is exactly the product that fuzzy theory is combined with neural network, it summarizes neural network and fuzzy theory
The advantages of, integrate study, association, identification, information processing.
Summary of the invention
The technical problem to be solved by the invention is to provide a kind of fan braking system card cylinder method for diagnosing faults, it will
Fuzzy neural network is engaged with D-S evidence, accurate detection blower fan operating status, improves fan braking system card cylinder fault diagnosis
Accuracy rate.
It is realized the technical problem to be solved by the present invention is to technical solution in this way, comprising the following steps: step
1, acquire data: from a large amount of information about fan braking system operating status contained in SCADA information system, depth is dug
Excavate brake clearance-time signal；
Step 2, data prediction: including removing the singular point in signal data, to signal noise silencing, data normalization processing
And output signal similarization is handled；
The extraction of step 3, fault characteristic value: using wavelet packet extracted from brake clearance-time signal fault signature to
Amount；
Step 4 determines training parameter according to the dimension of input signal；Training parameter has hidden layer number of nodes, display intermediate
As a result period, maximum number of iterations, learning rate and error threshold；The dimension of input signal corresponds to input layer number；
The resulting input signal sample of step 2 is divided into training set and test set by step 5, should be comprising normal in sample
Data include fault data again；
Step 6 carries out network training according to the resulting training parameter of step 4 and training set data, obtains fuzznet
Network model；
Step 7, fuzzy neural network model optimization, test fuzzy neural network model: by fuzzy neural network model
Output valve is compared with ideal output (i.e. actually measured data or the value of oneself setting), is obtained error, is carried out error
Anti-pass, the parameter in more new model obtain optimal fuzzy neural network model until error is met the requirements；With test set data
Examine the performance of network；
Step 8, D-S evidence theory fusion: the output valve of fuzzy neural network is sent into Decision fusion module, utilizes each mould
The recognition result of paste neural network is merged as independent evidence with D-S evidence theory, obtains comprehensive diagnosis knot
Fruit；
Step 9: output diagnostic result: the result of diagnosis being sent into diagnostic system and is analyzed, and is in by final result
In present human-computer interaction interface.
Since the present invention is using the data inside SCADA information system, existing data are taken full advantage of, with big data
Theory is braked based on the viewpoint of information fusion around blower using tools such as wavelet theory, fuzzy neural network, evidence theories
The card cylinder failure of system oil cylinder carries out fault diagnosis, and the data of acquisition have obtained fully utilizing by neural network, D-S evidence
The method of theory fusion increases the fault-tolerance of some signal, and the present invention carries out fan braking system using multiple technologies means
Card cylinder fault diagnosis, improves the accuracy rate of fault diagnosis.
Compared with prior art, the solution have the advantages that:
1, the present invention has effectively achieved the extraction to fault characteristic value using wavelet theory.
2, the present invention data strong to correlation are handled, and accelerate the convergent speed of neural network.
3, it makes full use of the data in system to be trained fuzzy neural network, has obtained optimal model, then again
It is secondary to be tested using test set data to model, it is ensured that the accuracy of model.
4, the card cylinder failure of fan braking system is diagnosed by merging multiple sensor signals, further improves card cylinder
The accuracy rate of fault diagnosis provides feasibility for the intelligent health management of Wind turbines, the safe and stable of power grid may be implemented
And efficient operation.
Detailed description of the invention
Detailed description of the invention of the invention is as follows:
Fig. 1 is fault method flow chart of the invention；
Fig. 2 is wavelet packet signal decomposition figure in the present invention；
Fig. 3 is fuzzy neural network topology diagram in the present invention；
Fig. 4 is fuzzy neural network model optimization and test flow chart in the present invention；
Fig. 5 is the fault diagnosis frame diagram of D-S evidence theory in the present invention.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples:
Currently, there is various kinds of sensors detection blower fan operating status, be fitted in wind field CMS (condition monitoring system) and
SCADA (data acquisition and supervisor control) the two systems, it is various that the two systems can acquire in real time generating set
Data.As shown in Figure 1, the present invention the following steps are included:
Step 1, acquisition data: what is contained in depth excavation SCADA information system is a large amount of about fan braking system operation
The information of state is acquired two eddy current sensor signals by A/D collection plate respectively, obtains brake clearance-time letter
Number；
Brake clearance-time signal is mainly used for subsequent characteristic vector pickup.
Step 2, data prediction: including removing the singular point in signal data, to signal noise silencing, data normalization processing
And by the processing of output signal similarization, similarization processing is when inputting identical and output bias very little, by output data
It is adjusted to identical；
The singular point of data-signal just refers to emergent bigger or much smaller than previous moment and later moment in time signal
Signal, since the running environment of blower is more severe, sensor may be because extraneous interference or some sensor failure, that
Data measured by this sensor may will be wrong, it is necessary to which the data for measuring it are rejected, otherwise can serious shadow
Ring final result.
The principle for picking out singular point is: the standard deviation by calculating input sample signal, according still further to statistical probability theory
Data by deviation greater than 3 σ are rejected.
Input quantity has different physical significances and different dimensions, and the data after normalization all change between [0,1],
There is status of equal importance to make respectively to input component in network training.Therefore data normalized must could be used for pre-
Survey problem.The formula of data normalization processing is as follows:
In formula, x*For the value after data normalization, xiFor i-th of data of signal, xmax、xminRespectively sample data is most
Big value and minimum value.
Similarization processing refers to when inputting identical and output bias very little, and output data is adjusted to identical；
When the processing of output signal similarization, according to setting output bias threshold value is required, when output bias is met the requirements, just
Output data can be adjusted to identical, what is done so is the convergence rate in order to accelerate operation.
The extraction of step 3, fault characteristic value: using wavelet packet extracted from brake clearance-time signal fault signature to
Amount；
The step of extracting fault characteristic value is as follows:
Step 31 carries out wavelet decomposition to brake clearance-time signal；
Wavelet packet signal decomposition figure is as shown in Figure 2: carrying out three layers of WAVELET PACKET DECOMPOSITION to signal, extracts third layer respectively from low
Frequency arrives the signal characteristic of 8 frequency contents of high frequency；In Fig. 2, (0,0) node on behalf original signal S, (1,1) node on behalf small echo
Wrap the first layer low frequency coefficient decomposed, the first layer high frequency coefficient of (1,2) node on behalf WAVELET PACKET DECOMPOSITION, (2,1) and (2,2) section
Point represents the second layer low frequency coefficient of WAVELET PACKET DECOMPOSITION, the second layer high frequency system of (2,3) and (2,4) node on behalf WAVELET PACKET DECOMPOSITION
Number, (3,1) ... (3,8) node indicate the third layer first of WAVELET PACKET DECOMPOSITION to the coefficient of the 8th node.
WAVELET PACKET DECOMPOSITION coefficient is reconstructed in step 32, extracts the signal of each frequency range；With S31Indicate x31(x31It is
The signal of first node of WAVELET PACKET DECOMPOSITION third layer) reconstruction signal, S32Indicate x32(x32It is WAVELET PACKET DECOMPOSITION third layer
The signal of two nodes) reconstruction signal, other and so on；Then original signal S can be indicated are as follows: S=S31+S32+S33+S34+
S35+S36+S37+S38
Step 33, the gross energy for seeking each band signal；Since input signal is a random signal, output signal is also
One random signal；If reconstruction signal S3j(j=1,2 ..., 8) corresponding energy is E3j(j=1,2 ..., 8), then have
In formula: x3j(j=1,2 ..., 8) indicates reconstruction signal S3jDiscrete point amplitude.
Step 34, construction feature vector, the construction of feature vector are as follows:
T=[E31,E32,E33,E34,E35,E36,E37,E38]
When energy is larger, E3j(j=1,2 ..., 8) is usually a biggish numerical value, is brought to subsequent data analysis
Some inconveniences enable so feature vector T is normalized
T '=[E31/E,E32/E,E33/E,E34/E,E35/E,E36/E,E37/E,E38/E]
Vector T ' for normalization after feature vector.
Step 4 determines training parameter according to the dimension of input signal；Training parameter has hidden layer number of nodes, display intermediate
As a result period, maximum number of iterations, learning rate and error threshold；The dimension of input signal corresponds to input layer number；
The determination formula of hidden layer number of nodes:
In formula: m is node in hidden layer, and n is input layer number, and l is output layer number of nodes, and α is normal between 1-10
Number.
Error threshold requires to choose according to engineering available accuracy.
The resulting input signal sample of step 2 is divided into training set and test set by step 5, should in input signal sample
It again include fault data comprising normal data.Input signal sample be exactly with each sensor measure about blower braking system
The data for operating status of uniting.
Step 6 carries out network training according to the resulting training parameter of step 4 and training set data, obtains fuzznet
Network model；
The structure of fuzzy neural network is input layer, degree of membership respectively as shown in figure 3, fuzzy neural network haves three layers altogether
Function generation layer and output layer, input layer one share d neuron, and subordinating degree function generation layer has h neuron, and output layer has
G neuron, X indicate input data, y indicate output as a result, input variable subordinating degree function is selected as Gaussian function, Gaussian
Degree of membership letterμ and σ respectively indicate mean value and standard value,Indicate the input of a-th of node of kth layer,
The output for indicating a-th of node of kth layer, is subordinate to angle valueConnection weight between first and second layer is normal
Number 1, action function are fuzzy membership functions μbcFuzzy relation between (), i.e. b-th of factor and c-th of index is subordinate to letter
Number；wvFor the weight coefficient of second, third interlayer.
Each layer treatment process of fuzzy neural network is as follows:
If f-th of neuron input and output of input layer are respectively xfAnd Of, p-th of neuron of subordinating degree function generation layer be defeated
Entering output is respectivelyWithQ-th of neuron input and output of output layer are respectivelyWithThen
The output of step 61, input layer is equal to its input value, i.e.,
Of=xf,
F is input layer serial number, f=1,2 ..., d；
Step 62, subordinating degree function layer:
In formula: mrpAnd σrpRespectively indicate the equal of the Gaussian subordinating degree function of p-th of fuzzy set of r-th of input quantity
Value and standard deviation, rrpIt is r-th of factor to the degree of membership of p-th of index, they are all the adjustable parameters of network；
Step 63, output layer:
In formula:For the output of q-th of neuron of fuzzy neural network,For p-th of nerve of subordinating degree function generation layer
The output of member, wpqFor p-th of neuron of the second layer and q-th of third layer interneuronal weight coefficient.
The fuzzy neural network has three classes adjustable parameter: one kind is parameter of regularity, they are the power systems of second, third interlayer
Number；Second class adjustable parameter is the mean value m of Gaussian subordinating degree function respectivelyrpAnd standard deviation sigmarp, they are located at the section of the second layer
In point, the parameter of input subordinating degree function is represented；Third class is evaluation result output layer.
Step 7, fuzzy neural network model optimization, test fuzzy neural network model；
By the output valve of fuzzy neural network model and ideal output (i.e. actually measured data or oneself setting
Value) it is compared, it obtains error, carries out error-duration model, the parameter in more new model obtains optimal until error is met the requirements
Fuzzy neural network model；With the performance of test set data detection network.
As shown in figure 4, fuzzy neural network model optimization and test the following steps are included:
Step 71, input sample；
Step 72, input parameter and determining error threshold；
Step 73 compares the reality output of above-mentioned network and ideal output, is obtained using secondary cost function defeated
Error E out；
If the given t sample from training set, input sampleD is the mind that input layer has altogether
Through member, sample is exportedThe neuron that g has altogether for output layer, s=1,2 ..., t,For s
The desired output signal of u-th of neuron of a sample,For the network output signal of s-th of sample, u-th of neuron, then
Step 74 judges whether error obtained above is less than error threshold, if it is thening follow the steps 76, otherwise executes
Step 75；
Step 75, error are greater than error threshold and then carry out error-duration model, from third layer to the layer-by-layer back transfer of first layer；Benefit
With blur method come the adjustable parameter w of corrective networksv, mrp, σrp, each parameter correction values are as follows:
The more new algorithm of each customized parameter is as follows:
wv(T+1)=wv(T)+η1Δwv(T)
mrp(T+1)=mrp(T)+η2Δmrp(T)
σrp(T+1)=σrp(T)+η3Δσrp(T)
In formula, wv(T) and wvIt (T+1) is the parametric function for updating front and back；η1, η2, η3The learning rate of respectively each parameter；T
Discrete-time variable.
In the case where output error is greater than error threshold, continue the parameter in training pattern until error is less than threshold value
Only；
Step 76, error are less than error threshold, have obtained optimal fuzzy neural network model；
Step 77, with the performance of test set data detection network.
Step 8, D-S evidence theory fusion: the output valve of fuzzy neural network is normalized, decision is then fed into and melts
Block is molded, removes data test signal using multiple sensors, the data that a sensor measures are to a fuzzy neural network, so
It afterwards using the recognition result of each fuzzy neural network as independent evidence, is merged, is obtained comprehensive with D-S evidence theory
Diagnostic result.
Above-mentioned evidence is modified using entropy principle, reduces the conflicting between each evidence.As shown in figure 5, D-S evidence melts
Close the following steps are included:
Step 81, evidence entropy modification: for having F evidence, N kind fault type, mbaIndicate a articles evidence to b kind failure
Basic probability assignment function, that is, BPA value, then evidence entropy are as follows:
Its inverse is normalized to obtain the weight w of each evidencea:
The then total weighted average of original BPA value
Wherein
Each original BPA value of a articles evidence and total weighted averageDeviation be denoted as:
εa=(ε1a,ε2a,…,εba,…,εNa), wherein a=1,2 ..., F
NoteFor maWithBetween population deviation degree, utilize δaOriginal evidence is adjusted, after modification
Evidence be denoted as m 'a, then:
m′a=mba-εbaδa, wherein a=1,2 ..., F, b=1,2 ..., N
Step 82 is merged the modified evidence of step 81 using D-S evidence theory, the fusion formula of G evidence
It is as follows:
In formula: m (A) indicates that the fused basic probability assignment function of failure A, G indicate sensor, that is, evidence body
Number, K indicate the conflict factor, and K is bigger to illustrate that the conflicting between an evidence is bigger, conversely, then smaller, m1、m2、…mGIndicate each
The basic probability assignment function of a evidence, A1、A2、…AnIndicate the failure collection in each evidence comprising failure A；
Step 83 makes final diagnosis decision using relevant Decision rule；
IfFailure mould in three kinds of failure domains (fault mode that braking system is likely to occur, indicated with Θ)
Formula meets:
m(A)≥ε1
m(A)-m(B)≥ε2
In formula, ε1, ε2For preset threshold value, then A is judgement as a result, empirically, here ε1=0.6, ε2=
0.3, m (A), m (B), m (C) indicate the basic probability assignment function of tri- kinds of fault modes of A, B, C.
Step 9, output diagnostic result；
The result of diagnosis is sent into diagnostic system and is analyzed, and final result is presented in human-computer interaction interface.
The step carrying out practically process is: prediction result being sent into information bank and is saved, and is delivered to inference machine, the inference machine is to information
Prediction result in library is analyzed, and is matched the information in knowledge base, is obtained corresponding failure cause, and the failure cause is defeated
Out into interpreter, is explained and be presented in human-computer interaction interface accordingly, the available corresponding reasoning process of user.
Claims (8)
1. a kind of fan braking system card cylinder method for diagnosing faults, characterized in that the following steps are included:
Step 1, acquisition data: a large amount of information about fan braking system operating status contained from SCADA information system
In, depth excavates brake clearance-time signal；
Step 2, data prediction: including removal signal data in singular point, to signal noise silencing, data normalization processing and
By the processing of output signal similarization；
The extraction of step 3, fault characteristic value: fault feature vector is extracted from brake clearance-time signal using wavelet packet；
Step 4 determines training parameter according to the dimension of input signal；Training parameter has hidden layer number of nodes, display intermediate result
Period, maximum number of iterations, learning rate and error threshold；The dimension of input signal corresponds to input layer number；
The resulting input signal sample of step 2 is divided into training set and test set by step 5, should include in input signal sample
Normal data includes fault data again；
Step 6 carries out network training according to the resulting training parameter of step 4 and training set data, obtains fuzzy neural network mould
Type；
Step 7, fuzzy neural network model optimization, test fuzzy neural network model: by the output of fuzzy neural network model
Value is compared with ideal output, obtains error, carries out error-duration model, the parameter in more new model, until error is met the requirements,
Obtain optimal fuzzy neural network model；With the performance of test set data detection network；
Step 8, D-S evidence theory fusion: the output valve of fuzzy neural network is sent into Decision fusion module, utilizes each fuzzy mind
Recognition result through network is merged as independent evidence with D-S evidence theory, obtains comprehensive diagnostic result；
Step 9: output diagnostic result: the result of diagnosis being sent into diagnostic system and is analyzed, and final result is presented on
In human-computer interaction interface.
2. fan braking system card cylinder method for diagnosing faults according to claim 1, characterized in that in step 2, data
The formula of normalized is as follows:
In formula, x*For the value after data normalization, xiFor i-th of data of signal, xmax、xminThe respectively maximum value of sample data
And minimum value.
3. fan braking system card cylinder method for diagnosing faults according to claim 2, characterized in that in step 3, extract
The step of fault characteristic value, is as follows:
Step 31 carries out wavelet decomposition to brake clearance-time signal；
Three layers of WAVELET PACKET DECOMPOSITION are carried out to signal, extract the signal characteristic of third layer 8 frequency contents from low to high respectively；
WAVELET PACKET DECOMPOSITION coefficient is reconstructed in step 32, extracts the signal of each frequency range；With S31Indicate x31Reconstruct letter
Number, x31It is the signal of first node of WAVELET PACKET DECOMPOSITION third layer；S32Indicate x32Reconstruction signal, x32It is WAVELET PACKET DECOMPOSITION
The signal of three layers of second node, other and so on；Then original signal S can be indicated are as follows: S=S31+S32+S33+S34+S35+
S36+S37+S38；
Step 33, the gross energy for seeking each band signal；
Since input signal is a random signal, output signal is also a random signal；If reconstruction signal S3jIt is corresponding
Energy is E3j, then have:
In formula: x3jIndicate reconstruction signal S3jDiscrete point amplitude, j=1,2 ..., 8；
Step 34, construction feature vector, the construction of feature vector are as follows:
T=[E31,E32,E33,E34,E35,E36,E37,E38]
When energy is larger, E3jA usually biggish numerical value, is normalized feature vector T, enables
T '=[E31/E,E32/E,E33/E,E34/E,E35/E,E36/E,E37/E,E38/E]
Vector T ' for normalization after feature vector.
4. fan braking system card cylinder method for diagnosing faults according to claim 3, characterized in that in step 4, hide
The determination formula of node layer number:
In formula: m is node in hidden layer, and n is input layer number, and l is output layer number of nodes, and α is the constant between 1-10.
5. fan braking system card cylinder method for diagnosing faults according to claim 4, characterized in that in step 6, obscure
Each layer treatment process of neural network is as follows:
If f-th of neuron input and output of input layer are respectively xfAnd Of, p-th of neuron of subordinating degree function generation layer inputs defeated
It is respectively outWithQ-th of neuron input and output of output layer are respectivelyWithThen
The output of step 61, input layer is equal to its input value:
Of=xf
F is input layer serial number, and f=1,2 ..., d, d is the neuron population of input layer；
Step 62, subordinating degree function layer:
In formula, mrpAnd σrpRespectively indicate the mean value and mark of the Gaussian subordinating degree function of p-th of fuzzy set of r-th of input quantity
Poor, the r of standardrpIt is r-th of factor to the degree of membership of p-th of index, they are all the adjustable parameters of network；
Step 63, output layer:
In formula:For the output of q-th of neuron of fuzzy neural network,For p-th of neuron of subordinating degree function generation layer
Output, wpqFor p-th of neuron of the second layer and q-th of third layer interneuronal weight coefficient.
6. fan braking system card cylinder method for diagnosing faults according to claim 5, characterized in that in step 7, described
The calculating formula of error E are as follows:
In formula, t is the total sample number of self-training collection, and s is sample serial number, and s=1,2 ..., t, g is the neuron population of output layer,
U is output layer neuron serial number；For the desired output signal of s-th of sample, u-th of neuron,For s-th of sample u
The network output signal of a neuron.
7. fan braking system card cylinder method for diagnosing faults according to claim 6, characterized in that in step 7, update
Parameter in model has: the weight coefficient w of second, third interlayerv, the mean value m of Gaussian subordinating degree functionrpAnd standard deviation sigmarp,
Each parameter correction values are as follows:
The more new algorithm of each parameter is as follows:
wv(T+1)=wv(T)+η1Δwv(T)
mrp(T+1)=mrp(T)+η2Δmrp(T)
σrp(T+1)=σrp(T)+η3Δσrp(T)
In formula, wv(T) and wvIt (T+1) is the parametric function for updating front and back；η1, η2, η3The learning rate of respectively each parameter；
T discrete-time variable.
8. fan braking system card cylinder method for diagnosing faults according to claim 7, characterized in that in step 8, described
D-S evidence fusion the following steps are included:
Step 1), evidence entropy modification: for having F evidence, N kind fault type, mbaIndicate a articles evidence to the BPA of b kind failure
It is worth, then evidence entropy are as follows:
Its inverse is normalized to obtain the weight w of each evidencea:
The then total weighted average of original BPA value
In formula,
Each original BPA value of a articles evidence and total weighted averageDeviation be denoted as:
εa=(ε1a,ε2a,…,εba,…,εNa)
NoteFor maWithBetween population deviation degree, utilize δaTo adjust original evidence, modified card
According to being denoted as m 'a, then:
m′a=mba-εbaδa
Step 2) merges the modified evidence of step 1) using D-S evidence theory, and the fusion formula of G evidence is such as
Under:
In formula: m (A) indicates that the fused basic probability assignment function of failure A, G indicate sensor, that is, evidence body number, K
Indicate the conflict factor, m1、m2、…mGIndicate the basic probability assignment function of each evidence, A1、A2、…AnIt indicates in each evidence
Failure collection comprising failure A；
Step 3) makes final diagnosis decision using relevant Decision rule；
Θ indicates the fault mode that braking system is likely to occur, ifFault mode in three kinds of failure domains meets:
m(A)≥ε1
m(A)-m(B)≥ε2
In formula, ε1, ε2For preset threshold value, then A is judgement as a result, m (A), m (B), m (C) indicate tri- kinds of failures of A, B, C
The BPA value of mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810888737.XA CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810888737.XA CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109213121A true CN109213121A (en) | 2019-01-15 |
| CN109213121B CN109213121B (en) | 2021-05-18 |
Family
ID=64988180
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810888737.XA Active CN109213121B (en) | 2018-08-07 | 2018-08-07 | Method for diagnosing clamping cylinder fault of fan braking system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109213121B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110617960A (en) * | 2019-10-12 | 2019-12-27 | 华北电力大学 | Wind turbine generator gearbox fault diagnosis method and system |
| CN110826690A (en) * | 2019-10-10 | 2020-02-21 | 深圳供电局有限公司 | Equipment state identification method and system and computer readable storage medium |
| CN111497812A (en) * | 2019-01-31 | 2020-08-07 | 天津所托瑞安汽车科技有限公司 | Vehicle braking system monitoring method based on platform |
| CN111781435A (en) * | 2019-04-04 | 2020-10-16 | 中车唐山机车车辆有限公司 | Fault detection method and device for four-quadrant rectifier |
| CN111985820A (en) * | 2020-08-24 | 2020-11-24 | 深圳市加码能源科技有限公司 | FNN and DS fusion-based fault identification method for charging operation management system |
| CN112101457A (en) * | 2020-09-15 | 2020-12-18 | 湖南科技大学 | PMSM demagnetization fault diagnosis method based on torque signal fuzzy intelligent learning |
| CN112798956A (en) * | 2020-12-31 | 2021-05-14 | 江苏国科智能电气有限公司 | Wind turbine generator fault diagnosis method based on multi-resolution time sequence cyclic neural network |
| CN113790890A (en) * | 2021-09-10 | 2021-12-14 | 南京航空航天大学 | Wavelet packet decomposition weight fuzzy entropy and ELM-based bearing fault classification method and device |
| CN113963456A (en) * | 2020-07-20 | 2022-01-21 | 核工业理化工程研究院 | Method and system for analyzing operation data of multiple high-speed rotating devices |
| CN114690038A (en) * | 2022-06-01 | 2022-07-01 | 华中科技大学 | Motor fault identification method and system based on neural network and storage medium |
| WO2023082427A1 (en) * | 2021-11-11 | 2023-05-19 | 烟台杰瑞石油服务集团股份有限公司 | Fault diagnosis method and apparatus, diagnosis model construction method, and device and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2447888A1 (en) * | 2010-10-27 | 2012-05-02 | Honeywell International Inc. | System and method for determining fault diagnosability of a health monitoring system |
| CN105223809A (en) * | 2015-07-10 | 2016-01-06 | 沈阳工业大学 | The synchronous control system of the fuzzy neural network compensator of H type platform and method |
| CN205240976U (en) * | 2015-10-16 | 2016-05-18 | 日立电梯（中国）有限公司 | Arresting gear fault detection system |
| CN106779063A (en) * | 2016-11-15 | 2017-05-31 | 河南理工大学 | A kind of hoist braking system method for diagnosing faults based on RBF networks |
-
2018
- 2018-08-07 CN CN201810888737.XA patent/CN109213121B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2447888A1 (en) * | 2010-10-27 | 2012-05-02 | Honeywell International Inc. | System and method for determining fault diagnosability of a health monitoring system |
| CN105223809A (en) * | 2015-07-10 | 2016-01-06 | 沈阳工业大学 | The synchronous control system of the fuzzy neural network compensator of H type platform and method |
| CN205240976U (en) * | 2015-10-16 | 2016-05-18 | 日立电梯（中国）有限公司 | Arresting gear fault detection system |
| CN106779063A (en) * | 2016-11-15 | 2017-05-31 | 河南理工大学 | A kind of hoist braking system method for diagnosing faults based on RBF networks |
Non-Patent Citations (2)
| Title |
|---|
| 王峰: "网络环境下矿井提升机智能故障诊断关键技术研究", 《中国优秀博士学位论文全文数据库工程科技Ⅰ辑》 * |
| 王致杰、等: "基于证据理论的综合故障诊断方法研究", 《上海电机学院学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111497812A (en) * | 2019-01-31 | 2020-08-07 | 天津所托瑞安汽车科技有限公司 | Vehicle braking system monitoring method based on platform |
| CN111781435A (en) * | 2019-04-04 | 2020-10-16 | 中车唐山机车车辆有限公司 | Fault detection method and device for four-quadrant rectifier |
| CN110826690A (en) * | 2019-10-10 | 2020-02-21 | 深圳供电局有限公司 | Equipment state identification method and system and computer readable storage medium |
| CN110617960A (en) * | 2019-10-12 | 2019-12-27 | 华北电力大学 | Wind turbine generator gearbox fault diagnosis method and system |
| CN113963456A (en) * | 2020-07-20 | 2022-01-21 | 核工业理化工程研究院 | Method and system for analyzing operation data of multiple high-speed rotating devices |
| CN111985820A (en) * | 2020-08-24 | 2020-11-24 | 深圳市加码能源科技有限公司 | FNN and DS fusion-based fault identification method for charging operation management system |
| CN112101457A (en) * | 2020-09-15 | 2020-12-18 | 湖南科技大学 | PMSM demagnetization fault diagnosis method based on torque signal fuzzy intelligent learning |
| CN112101457B (en) * | 2020-09-15 | 2023-11-17 | 湖南科技大学 | PMSM demagnetizing fault diagnosis method based on torque signal fuzzy intelligent learning |
| CN112798956A (en) * | 2020-12-31 | 2021-05-14 | 江苏国科智能电气有限公司 | Wind turbine generator fault diagnosis method based on multi-resolution time sequence cyclic neural network |
| CN113790890A (en) * | 2021-09-10 | 2021-12-14 | 南京航空航天大学 | Wavelet packet decomposition weight fuzzy entropy and ELM-based bearing fault classification method and device |
| WO2023082427A1 (en) * | 2021-11-11 | 2023-05-19 | 烟台杰瑞石油服务集团股份有限公司 | Fault diagnosis method and apparatus, diagnosis model construction method, and device and medium |
| US11886178B2 (en) | 2021-11-11 | 2024-01-30 | Yantai Jereh Oilfield Services Group Co., Ltd. | Fault diagnosis method, method for building fault diagnosis model, equipment, device and medium |
| CN114690038A (en) * | 2022-06-01 | 2022-07-01 | 华中科技大学 | Motor fault identification method and system based on neural network and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109213121B (en) | 2021-05-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109213121A (en) | A kind of fan braking system card cylinder method for diagnosing faults | |
| Cheng et al. | Study on planetary gear fault diagnosis based on entropy feature fusion of ensemble empirical mode decomposition | |
| CN106338406B (en) | The on-line monitoring of train traction electric drive system and fault early warning system and method | |
| CN104712542B (en) | A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults | |
| Lindemann et al. | Anomaly detection and prediction in discrete manufacturing based on cooperative LSTM networks | |
| US8306931B1 (en) | Detecting, classifying, and tracking abnormal data in a data stream | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| CN105825271B (en) | Satellite failure diagnosis and prediction method based on evidential reasoning | |
| CN110163075A (en) | A kind of multi-information fusion method for diagnosing faults based on Weight Training | |
| CN109818798A (en) | A kind of wireless sensor network intruding detection system and method merging KPCA and ELM | |
| CN112101431A (en) | Electronic equipment fault diagnosis system | |
| Silva et al. | Artificial immune systems applied to fault detection and isolation: A brief review of immune response-based approaches and a case study | |
| CN107976934B (en) | A kind of oil truck oil and gas leakage speed intelligent early-warning system based on wireless sensor network | |
| Lima et al. | An artificial immune system with continuous-learning for voltage disturbance diagnosis in electrical distribution systems | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| CN116684878B (en) | 5G information transmission data safety monitoring system | |
| CN115587309A (en) | Method, device and equipment for extracting key features of short-circuit resistance of transformer | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| Lei et al. | Fault diagnosis of rotating machinery based on a new hybrid clustering algorithm | |
| Desmet et al. | Leak detection in compressed air systems using unsupervised anomaly detection techniques | |
| CN112798290B (en) | Abnormal state monitoring method of gas turbine based on spectrum reconstruction error | |
| CN117768235A (en) | Real-time flow monitoring alarm system based on Internet of things | |
| Liu et al. | A divide and conquer approach to anomaly detection, localization and diagnosis | |
| AU2021413008A1 (en) | Train compartment vibration monitoring method, vibration signal feature library establishment method and vibration signal feature library application method | |
| Calado et al. | FDI approach to the DAMADICS benchmark problem based on qualitative reasoning coupled with fuzzy neural networks |
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 |