CN109916628A - Based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy - Google Patents
Based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy Download PDFInfo
- Publication number
- CN109916628A CN109916628A CN201910273110.8A CN201910273110A CN109916628A CN 109916628 A CN109916628 A CN 109916628A CN 201910273110 A CN201910273110 A CN 201910273110A CN 109916628 A CN109916628 A CN 109916628A
- Authority
- CN
- China
- Prior art keywords
- signal
- fault
- entropy
- time series
- component
- 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
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
Based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy, it is related to digital processing field, to solve the problems, such as that the method for diagnosing faults of existing bearing vibration signal still has that feature extraction separability is not strong, fault identification accuracy rate is low and fault severity level is analyzed insufficient, the present invention includes step 1: obtaining the bearing vibration sample of signal collection under different faults type, different faults degree；Step 2；It obtains optimal PR component and carries out subsequent characteristics extraction；Step 3；Obtain the fault feature vector under different faults type, different faults degree；Step 4；Feature vector is input in random forest grader；Step 5: obtaining rolling bearing fault type and fault severity level.The feature vector that the present invention extracts has good separability, has stronger failure-description ability, and average recognition accuracy reaches 99.25%.It the composite can be widely applied to bearing failure diagnosis field.
Description
Technical field
It is specially a kind of based on the axis for improving multiple dimensioned amplitude perception arrangement entropy the present invention relates to digital processing field
Hold feature extracting method.
Background technique
Rolling bearing is one of most common component in rotating machinery, but due to factors such as abrasion, fatigue, burn into overloads
Influence, rolling bearing easily breaks down during the work time, influences mechanical equipment overall performance.Therefore, rolling bearing
Fault diagnosis and severity analysis to guarantee mechanical equipment reliability of operation and formulate corresponding maintenance policy with important meaning
Justice.
The position of rolling bearing fault and severity will lead to the impact characteristics of vibration signal there are notable differences, therefore
Fault diagnosis technology based on vibration signal becomes one of the important research direction of current scrolling bearing monitoring abnormal state.Rolling
The essence of dynamic bearing fault diagnosis is a mode identification procedure, mainly includes feature extraction and failure modes.But due to rolling
Dynamic bearing vibration signal has the characteristics that non-linear and non-stationary while easy by a variety of extraneous factors in operational process
Interference, noise is relatively low, and bearing fault characteristics is caused to be difficult to be effectively extracted, and influences the accuracy rate of fault diagnosis result.
In consideration of it, related scholar has carried out a large amount of research work for rolling bearing fault diagnosis, and achieve certain
Research achievement.It describes in existing literature and is realized using set empirical mode decomposition (EEMD) to bearing vibration signal
Adaptive decomposition, and the intrinsic mode letter comprising main bearings status information is determined using the method for kurtosis value combination related coefficient
Number, recycles its singular value as feature vector, realizes the multiple faults point to rolling bearing by hypersphere multi-class support vector machine
Class.However, EEMD can not be fully solved the modal overlap problem of EMD, the intrinsic mode function of kurtosis value combination related coefficient
Back-and-forth method can lose split bearing fault message, and the nuclear parameter of hypersphere multi-class support vector machine is chosen and optimization is excessively complicated,
Increase the difficulty of practical application.It is described in another document and decomposes (LMD) algorithm to rolling bearing vibration using local mean value
Dynamic signal is pre-processed, and multi-scale entropy (MSE) is recycled to extract fault feature vector, finally constructs BP neural network classifier
Realize fault type recognition.But MSE, during the coarse of time series, the sequence length after coarse can be with ruler
It spends the increase of the factor and shortens.When scale factor is larger, multiple dimensioned entropy has unstability, and then influences feature extraction
Validity.The fault signature extracted in bearing vibration signal using multiple dimensioned arrangement entropy is described in an also document,
And feature selecting is carried out using LaplacianScore algorithm, then realize that fault type is known by support vector machines (SVM)
Not.However, the feature extraction based on arrangement entropy has ignored influence of the element magnitude to entropy in time series, can to extract
Feature has biggish randomness, influences fault identification accuracy rate.Therefore, the existing failure based on bearing vibration signal is examined
Disconnected method still has that feature extraction separability is not strong, fault identification accuracy rate is low and fault severity level analysis is insufficient etc. and asks
Topic.
Summary of the invention
The purpose of the present invention is: for existing bearing vibration signal method for diagnosing faults still have feature extraction can
Divide the problems such as property is not strong, fault identification accuracy rate is low and fault severity level analysis is insufficient, proposes a kind of based on the more rulers of improvement
Spend the Fault Diagnosis of Roller Bearings of amplitude perception arrangement entropy.
The present invention adopts the following technical scheme that realization: based on the rolling bearing fault for improving multiple dimensioned amplitude perception arrangement entropy
Diagnostic method, comprising the following steps:
Step 1: the bearing vibration signal under known different faults type, different faults degree is obtained, and is formed
Bearing vibration sample of signal collection under different faults type, different faults degree；
Step 2；For each of sample set vibration signal, intrinsic time Scale Decomposition is carried out, is obtained a series of solid
There is rotation PR component, and therefrom chooses optimal PR component and carry out subsequent characteristics extraction；
Step 3；Include using improving multiple dimensioned amplitude perception arrangement entropy and extracting under optimal PR component different time scales
The feature of bearing vibration signal forms the fault feature vector under different faults type, different faults degree, the improvement
The obtaining step of multiple dimensioned amplitude perception arrangement entropy are as follows:
Step 3 one: assuming that time series to be analyzed is { x1,x2,...,xN, one is generated using improved coarse process
The new coarse time series of groupWherein,
Step 3 two: it for each time scale factor τ and Embedded dimensions d, calculates separatelyIn it is every
The amplitude perception arrangement entropy of a time series, and its average value is defined as to improve multiple dimensioned amplitude perception arrangement entropy,
Wherein, AAPE is amplitude perception arrangement entropy；
Step 4；By carrying out feature extraction to bearing vibration sample of signal collection, bearing vibration signal is formed
This feature vector is input in random forest grader by fault signature collection.
Step 5: test set is inputted in random forest grader, obtains the rolling bearing fault type and event of test set
Hinder severity.
The invention has the following beneficial effects: the multiple dimensioned amplitude perception arrangement entropy fault signatures of improvement in the present invention to extract
Method has good fault severity level descriptive power, and the feature vector of extraction has good separability；
Improving multiple dimensioned amplitude perception arrangement entropy improves coarse process in multiscale analysis, and is perceived using amplitude
Arrange the entropy characteristic sensitive to signal amplitude and frequency variation, calculate the AAPE value under different time scales and composition characteristic to
Amount has stronger failure-description ability；
The Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy and RF can accurately identify failure
On the basis of type, further fault severity level is analyzed, it is average in the case where fault severity level is complex
Recognition accuracy reaches 99.25%.
Detailed description of the invention
Fig. 1 is fault identification and fault severity level analysis flow chart diagram.
Fig. 2 is the 0 bearing vibration signal graph for loading lower different faults severity.
Fig. 3 is the 0 bearing vibration signal graph for loading lower different faults severity.
Fig. 4 is the 0 bearing vibration signal graph for loading lower different faults severity.
Fig. 5 is the 0 multiple dimensioned amplitude perception arrangement entropy feature clustering figure of improvement for loading lower different faults type.
Fig. 6 is the 0 multiple dimensioned amplitude perception arrangement entropy feature clustering figure of improvement for loading lower different faults severity.
Specific embodiment
Specific embodiment 1: illustrating present embodiment below with reference to Fig. 1.Present embodiment is based on improving multiple dimensioned
The Fault Diagnosis of Roller Bearings of amplitude perception arrangement entropy, comprising the following steps:
Step 1: the bearing vibration signal under known different faults type, different faults degree is obtained, and is formed
Bearing vibration sample of signal collection under different faults type, different faults degree；
Step 2；For each of sample set vibration signal, intrinsic time Scale Decomposition is carried out, is obtained a series of solid
There is rotation PR component, and therefrom chooses optimal PR component and carry out subsequent characteristics extraction；
Step 3；Include using improving multiple dimensioned amplitude perception arrangement entropy and extracting under optimal PR component different time scales
The feature of bearing vibration signal forms the fault feature vector under different faults type, different faults degree, the improvement
The obtaining step of multiple dimensioned amplitude perception arrangement entropy are as follows:
Step 3 one: assuming that time series to be analyzed is { x1,x2,...,xN, one is generated using improved coarse process
The new coarse time series of groupWherein,J represents yi,1、
yi,2In 1,2；
Step 3 two: it for each time scale factor τ and Embedded dimensions d, calculates separatelyIn it is every
The amplitude perception arrangement entropy of a time series, and its average value is defined as to improve multiple dimensioned amplitude perception arrangement entropy,
Wherein, AAPE is amplitude perception arrangement entropy；
Step 4；By carrying out feature extraction to bearing vibration sample of signal collection, bearing vibration signal is formed
This feature vector is input in random forest grader by fault signature collection.
Step 5: test set is inputted in random forest grader, obtains the rolling bearing fault type and event of test set
Hinder severity.
Firstly, the present invention extracts the best intrinsic of bearing vibration signal by intrinsic time Scale Decomposition (ITD) method
Rotational component highlights the feature of fault-signal not of the same race；Then, multiple dimensioned amplitude perception arrangement entropy (IMAAPE) of improvement is utilized
The characteristic sensitive to fault-signal amplitude and frequency variation calculates the amplitude perception arrangement entropy under different time scales as feature
Vector, while the coarse process in multiscale analysis is improved, improve the stability of fault signature extraction；Finally, utilizing
Fault signature collection constructs random forest multi-categorizer, can be achieved with by simple parameter selection to rolling bearing different faults class
The identification of type and severity analysis, have stronger generalization ability.
Method for diagnosing faults proposed by the present invention can be realized rolling bearing inner ring, outer ring and ball fault identification and failure
Severity analysis, the fault identification and fault severity level analysis process of this method are as shown in Figure 1.
Improved coarse process described in the text is the prior art, and source is Azami H, Escudero
J.Improved multiscale permutation entropy for biomedical signal analysis:
Interpretation and application to electroencephalogram recordings[J]
.Biomedical Signal Processing and Control,2016,23:28-41。
Embodiment:
The rolling bearing fault data set that the present invention selects U.S. Xi Chu university bearing data center to provide is to being proposed
Method for diagnosing faults carries out experimental verification.Using SKF bearing as research object, data set is acquired by acceleration transducer for experiment
The normally bearing vibration signal under four kinds of (NM), inner ring failure (IR), outer ring failure (OR) and ball failure (B) states, sampling
Frequency is 12KHz；For three kinds of fault types, choosing fault diameter respectively is tri- kinds of differences of 7mils, 14mils and 21mils
Fault severity level carry out data acquisition.0 loads the time domain wave of the bearing vibration signal of lower different faults severity
Shape is as shown in Figures 2 and 3, it is seen that the variation and the change of the amplitude and frequency of its vibration signal of fault type and fault severity level
Change related.Table 1 show fault type and severity composition in experiment sample, and it includes 10 kinds different that experiment sample, which has altogether,
Rolling bearing health status.Every kind of bearing data are not divided into multiple data samples overlappingly, contain N=in each sample
1024 sampled points, the experimental data set that lower 50 samples of every kind of health status of composition are constituted.Wherein, using every kind of health status
Lower 10 samples are as training set, and 40 samples are as test set.
Rolling Bearing Fault Character extracts experiment:
Execute Rolling Bearing Fault Character extract before, need to using intrinsic time Scale Decomposition (ITD) to vibration signal into
Row pretreatment further protrudes the morphological features such as the intrinsic instantaneous amplitude of signal, frequency.It is illustrated in figure 4 fault diameter 7mils
Fault vibration signal decomposition is 5 PR components and 1 monotonic trend component by lower ball failure ITD decomposition result, ITD.By dividing
Result is solved as it can be seen that the highest optimal PR component of kurtosis value includes the main amplitude and frequecy characteristic of description ball failure.
The fault type and severity of 1 experiment sample of table form
Tab.1Compositionoffaulttypesandseverityinexperimentalsamples
Bearing vibration signal is carried out after ITD is decomposed using multiple dimensioned amplitude perception arrangement entropy (IMAAPE) is improved
Feature extraction.To optimal PR component carry out IMAAPE feature extraction, setting Embedded dimensions be d=4, time delay l=1, when
Between scale be τ=20, regulation coefficient A=0.5.IMAAPE feature extraction is carried out to the vibration signal sample of test set, obtains 20
The bearing vibration signal fault feature vector of dimension.It is illustrated in figure 5 the IMAAPE feature of different faults type under 0 load
Dendrogram, can be seen that feature extracting method proposed by the present invention from preceding 3 dimensions of feature vector can be preferably to just
Often, inner ring failure, outer ring failure and ball failure are described, and feature vector has stronger cluster.As Fig. 6 show 0
The IMAAPE feature clustering figure of lower different faults severity is loaded, preceding 2 dimensions can be seen that this hair in selected characteristic vector
The feature extracting method of bright proposition also has preferable cluster property to the feature extraction result of different faults severity.
The performance of 2 Rolling Bearing Fault Character extraction algorithm of table compares
Tab.2Performancecomparisonofrollingbearingfaultfeatureextractionalgo
rithm
In order to illustrate the performance of IMAAPE bearing vibration signal characteristic extracting methods, the present invention by IMAAPE with it is existing
The effect of Rolling Bearing Fault Character extracting method compares.Experiment is using 0 load lower inner ring failure, outer ring failure and ball
Each 40 samples of failure are analyzed, and experimental result is as shown in table 2.
Under different faults type cases, the between class distance average value of feature vector is bigger, indicates that feature extracting method extracts
Different faults type feature difference it is bigger；The inter- object distance average value of feature vector is got under different faults type cases
It is small, indicate that the feature difference for the same fault type that feature extracting method extracts is smaller.As can be seen from Table 2, IMAAPE's is averaged
Between class distance, which is greater than, improves multiple dimensioned arrangement entropy (IMPE) and fine compound multiple dimensioned arrangement entropy (RCMPE), but it is more to be less than improvement
Scale Sample Entropy (IMSE) improves multiple dimensioned fuzzy entropy (IMFE) and fine compound multiple dimensioned Sample Entropy (RCMSE), and
The average inter- object distance of IMAAPE is minimum in all feature extracting methods.The experimental result illustrates the axis of rolling that IMAAPE is extracted
Holding fault signature has preferable cluster property.In addition, as shown in Table 2, when calculating identical sampling number, IMAAPE's
It is average time-consuming to be the smallest in all feature extracting methods, with preferable real-time.
3 Rolling Bearing Fault Character extraction algorithm performance of table compares
Tab.3Performancecomparisonofrollingbearingfaultfeatureextractionalgo
rithm
The separability of the IMAAPE fault signature extracting method proposed in order to further illustrate the present invention, the present invention are sharp respectively
The feature extracting method described in table 2 is combined with random forest grader, and setting CART decision tree quantity is 50, to rolling
40 test samples of every kind of classification under 10 kinds of bearing different health status are analyzed, and experimental result is as shown in table 3.As it can be seen that
Compared with current different Rolling Bearing Fault Character extracting methods, IMAAPE fault signature extracting method tool proposed by the present invention
There is better fault severity level descriptive power, the feature vector of extraction has higher separability.
The experiment of rolling bearing fault type identification
It mutually ties to verify proposed by the present invention based on improving amplitude perception arrangement entropy (IMAAPE) with random forest (RF)
The performance of the Method for Bearing Fault Diagnosis of conjunction, to 40 test samples of every kind of classification under 10 kinds of rolling bearing different health status
Experimental verification is carried out, the results are shown in Table 4.As it can be seen that the Fault Diagnosis of Roller Bearings proposed can efficiently identify it is normal,
Inner ring failure, outer ring failure, ball failure, and fault severity level can be more efficiently analyzed, rate of false alarm is lower, average
Recognition accuracy is up to 99.25%.
The discrimination for the Fault Diagnosis of Roller Bearings that table 4 proposes
Tab.4 Identification rate of the proposed rolling bearing fault
diagnosis method
Performance of the method in rolling bearing fault diagnosis is proposed in order to further illustrate the present invention, it will be proposed by the present invention
Fault Diagnosis of Roller Bearings is compared with existing method, and experimental result is as shown in table 5.As can be seen that the present invention proposes
Method can be realized rolling bearing fault type identification, and can further analyze bearing fault severity level.It is tight in failure
In the case that weight degree is single, fault type can be accurately identified；In the case where fault severity level is complex, still have
There is relatively high mean failure rate discrimination.
The different Fault Diagnosis of Roller Bearings discriminations of table 5 compare
Tab.5Comparisonofidentificationratesofdifferentfaultdiagnosismethods
forrolli ngbearings
Conclusion
1) rolling bearing fault signal decomposition can be steadily one group of PR component by ITD, wherein optimal PR component can
Prominent rolling bearing fault signal it is main when-frequency characteristic, be convenient for consequent malfunction feature extraction；
2) IMAAPE improves the coarse process in multiscale analysis, and using amplitude perception arrangement entropy to signal amplitude
The sensitive characteristic with frequency variation, calculates the AAPE value and composition characteristic vector under different time scales, has stronger failure
Descriptive power；
3) Fault Diagnosis of Roller Bearings based on IMAAPE and RF can on the basis of accurately identifying fault type,
Further fault severity level is analyzed, in the case where fault severity level is complex, average recognition accuracy reaches
To 99.25%.
The method that the present invention studies is only applicable to fault type recognition under rolling bearing fixed load at present and failure is tight
Weight degree analyzing.In order to further enhance the generalization ability of the Fault Diagnosis of Roller Bearings, subsequent research emphasis is to become
The fault type of rolling bearing and fault severity level are analyzed under loading condition.
Specific embodiment 2: present embodiment is the further explanation to specific embodiment one, present embodiment with
The difference of specific embodiment one is in the step 2, if XtFor known signal to be analyzed, definitionIt is mentioned for baseline
Operator is taken,X can be extractedtMiddle background signalAnd obtain corresponding intrinsic rotational componentSignal XtPoint
Xie Wei
The intrinsic time key step of Scale Decomposition algorithm is as follows:
Step 2 one: assuming that { τk, k=1,2 ... } indicate signal XtLocal extremum, default τ0=0；
Step 2 two: in section [0, τk] in define LtAnd Ht, and XtThe t ∈ [0, τ in sectionk+2], at continuous threshold interval
(τk,τk+1] in extraction background signal LtIt indicates are as follows:
Wherein,
In formula, α is linear scaling factor, extracts intrinsic rotational component amplitude, 0 < α < 1 for adjusting；
Step 2 three: according to formula (2) and formula (3), intrinsic rotational component HtIt can indicate are as follows:
Wherein,Inherently to rotate extraction operator, 3 intrinsic rotational components of formula；
Step 2 four: by background signal LtStep 2 one to 23 is repeated as the input signal decomposed next time, is obtained
A series of PR components, the termination condition of decomposition are background signal LtBecome dull or is less than some preset value；
After the decomposition of intrinsic time Scale Decomposition, time series XtIt is broken down into a series of PR components and a dullness becomes
Gesture component, the kurtosis value of signal can effectively describe the pulse characteristic of signal, and kurtosis value is higher, and the pulse characteristics that signal is included are got over
Abundant, obtaining the maximum PR component of kurtosis value is best intrinsic rotational component, and calculating process is as described below:
Wherein, KiIndicate the kurtosis value of i-th of PR component, n indicates length of time series, UiFor the normalizing of i-th of PR component
Change kurtosis value, m is the number of PR component, and best intrinsic rotational component chooses UiCorresponding PR component when for maximum value,
It is the sum of the biquadratic of all data points in above-mentioned intrinsic rotation (PR) component, i refers to i-th of PR component.
Specific embodiment 3: present embodiment is the further explanation to specific embodiment one, present embodiment with
The difference of specific embodiment one is the obtaining step of improved coarse process in the step 3 one are as follows:
Step 3 is one by one: assuming that the time series { X that a length is Ni}={ x1,x2,...,xN, utilize scale factor τ
=1,2 ..., n carries out coarse to sequence, and coarse process is shown below,
Wherein,Indicate the new time series obtained after coarse when scale factor is τ；
Step 3 one or two: calculating the sample entropy of the new time series of each coarse, obtains the n under different time scales
A multiple dimensioned entropy describes the signal characteristics of original time series；
Step 3 one or three: improving the coarse process of multi-scale entropy, and improved coarse time series indicates
For
Wherein,
Specific embodiment 4: present embodiment is the further explanation to specific embodiment one, present embodiment with
The difference of specific embodiment one is that the obtaining step of the amplitude perception arrangement entropy of time series in the step 3 two is；
Step 321: assuming that given length is the time series x={ x of N1,x2,...,xN, for each time point
Signal x is embedded into d dimension space and obtains reconstruct vector by t,Its
In, d and l respectively indicate Embedded dimensions and time delays；
Step 3 two or two: each vector is pressedThe size of middle element carries out ascending order arrangement, i.e.,
Wherein j*Indicate element in reconstruct vectorIn order, be embedded in
When dimension is d, d is shared！It kind puts in order, i-th kind puts in order and be denoted as πiEach put in order πiProbability of occurrence indicates are as follows:
Wherein, f (πi) indicate statistical arrangement sequence πiThe function of frequency of occurrence, wheneverInner element puts in order as πi
When, f (πi) just add 1, it arranges entropy and is defined as follows shown in formula,
Step 3 two or three: assuming thatInitial value be 0, for time seriesT gradually increases to N-d+1 from 1
During, whenever occur putting in order forWhen,It will be updated,
The weight of deviation, generally takes 0.5 between adjustment signal amplitude mean value and amplitude；
Step 3 two or four: in entire time seriesThe probability of appearanceAre as follows:
The amplitude perception arrangement entropy of time series indicates are as follows:
Specific embodiment 5: present embodiment is the further explanation to specific embodiment one, present embodiment with
The difference of specific embodiment one is the random forest grader specific steps in the step 4 are as follows: assuming that random forest is classified
Device is by multiple decision tree { hj(x,Θk), k=1,2 ..., n } composition, { Θk, k=1,2 ..., n } indicate mutually indepedent and same
The random vector of distribution；The training sample set representations of random forest grader
D={ (x1,y1),(x2,y2),...,(xN,yN), xi=(xi,1,...,xi,p)TIndicate i-th of training sample xiTool
There are p characteristic value, yiIndicate training sample xiCorresponding label；N times Bootstrap sampling is carried out to training sample set D, obtains n
A subsample Bootstrap Dj(j=1,2 ..., n)；For each subsample Dj, construct decision-tree model hj(x) (general choosing
With CART decision tree), it is final to obtain by one group of decision tree { h1(x),h2(x),…,hk(x) } decision tree classifier formed；It is right
In a new test sample, is voted by n decision tree, obtain final class of the classification as test sample of most polls
Not, categorised decision is as follows:
Wherein, hj(x) jth decision tree is represented, it is 1 when there is this number duration in set that I (), which is indicative function, no
Then value is 0；Y indicates class label yiThe target variable of composition.
Specific embodiment 6: present embodiment is the further explanation to specific embodiment two, present embodiment with
The difference of specific embodiment two is α=0.5 in the step 2 two.
It should be noted that specific embodiment is only the explanation and illustration to technical solution of the present invention, it cannot be with this
Limit rights protection scope.What all claims according to the present invention and specification were made is only locally to change, Reng Yingluo
Enter in protection scope of the present invention.
Claims (6)
1. based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy, it is characterised in that including following step
It is rapid:
Step 1: the bearing vibration signal under known different faults type, different faults degree is obtained, and forms difference
Bearing vibration sample of signal collection under failure mode, different faults degree；
Step 2；For each of sample set vibration signal, intrinsic time Scale Decomposition is carried out, obtains a series of intrinsic rotations
Turn PR component, and therefrom chooses optimal PR component and carry out subsequent characteristics extraction；
Step 3；The rolling for including under optimal PR component different time scales is extracted using multiple dimensioned amplitude perception arrangement entropy is improved
The feature of bearing vibration signal forms the fault feature vector under different faults type, different faults degree, described to improve more rulers
Spend the obtaining step of amplitude perception arrangement entropy are as follows:
Step 3 one: assuming that time series to be analyzed is { x1,x2,...,xN, it is new using one group of improved coarse process generation
Coarse time seriesWherein,
Step 3 two: it for each time scale factor τ and Embedded dimensions d, calculates separatelyIn each time
The amplitude perception arrangement entropy of sequence, and its average value is defined as to improve multiple dimensioned amplitude perception arrangement entropy,
Wherein, AAPE is amplitude perception arrangement entropy；
Step 4；By carrying out feature extraction to bearing vibration sample of signal collection, bearing vibration signal fault is formed
This feature vector is input in random forest grader by feature set.
Step 5: test set is inputted in random forest grader, rolling bearing fault type and the failure for obtaining test set are tight
Weight degree.
2. the Fault Diagnosis of Roller Bearings according to claim 1 based on the multiple dimensioned amplitude perception arrangement entropy of improvement,
It is characterized by: in the step 2, if XtFor known signal to be analyzed, definitionFor baseline extraction operator,It can
Extract XtMiddle background signalAnd obtain corresponding intrinsic rotational componentSignal XtIt is decomposed into
The intrinsic time key step of Scale Decomposition algorithm is as follows:
Step 2 one: assuming that { τk, k=1,2 ... } indicate signal XtLocal extremum, default τ0=0；
Step 2 two: in section [0, τk] in define LtAnd Ht, and XtThe t ∈ [0, τ in sectionk+2], in continuous threshold interval (τk,
τk+1] in extraction background signal LtIt indicates are as follows:
Wherein,
In formula, α is linear scaling factor, extracts intrinsic rotational component amplitude, 0 < α < 1 for adjusting；
Step 2 three: according to formula (2) and formula (3), intrinsic rotational component HtIt can indicate are as follows:
Wherein,Inherently to rotate extraction operator；
Step 2 four: by background signal LtStep 2 one to 23 is repeated as the input signal decomposed next time, is obtained a series of
PR component, the termination condition of decomposition are background signal LtBecome dull or is less than some preset value；
After the decomposition of intrinsic time Scale Decomposition, time series XtIt is broken down into a series of PR components and a monotonic trend point
Amount, the kurtosis value of signal can effectively describe the pulse characteristic of signal, and kurtosis value is higher, and the pulse characteristics that signal is included are richer
Richness, obtaining the maximum PR component of kurtosis value is best intrinsic rotational component, and calculating process is as described below:
Wherein, KiIndicate the kurtosis value of i-th of PR component, n indicates length of time series, UiFor the normalization peak of i-th of PR component
Angle value, m are the numbers of PR component, and best intrinsic rotational component chooses UiCorresponding PR component when for maximum value.
3. the Fault Diagnosis of Roller Bearings according to claim 1 based on the multiple dimensioned amplitude perception arrangement entropy of improvement,
It is characterized in that, in the step 3 one improved coarse process obtaining step are as follows:
Step 3 is one by one: assuming that the time series { X that a length is Ni}={ x1,x2,...,xN, using scale factor τ=1,
2 ..., n carries out coarse to sequence, and coarse process is shown below,
Wherein,Indicate the new time series obtained after coarse when scale factor is τ；
Step 3 one or two: calculating the sample entropy of the new time series of each coarse, and n obtained under different time scales are more
Scale entropy describes the signal characteristics of original time series；
Step 3 one or three: improving the coarse process of multi-scale entropy, and improved coarse time series is expressed as
Wherein,
4. the Fault Diagnosis of Roller Bearings according to claim 1 based on the multiple dimensioned amplitude perception arrangement entropy of improvement,
It is characterized in that, the obtaining step of the amplitude perception arrangement entropy of time series is in the step 3 two；
Step 321: assuming that given length is the time series X={ x of N1,x2,...,xN, it, will for each time point t
Signal x is embedded into d dimension space and obtains reconstruct vector,Wherein, d
Embedded dimensions and time delays are respectively indicated with l；
Step 3 two or two: each vector is pressedThe size of middle element carries out ascending order arrangement, i.e.,
Wherein j*Indicate element in reconstruct vectorIn order, in Embedded dimensions
When for d, d is shared！It kind puts in order, i-th kind puts in order and be denoted as πiEach put in order πiProbability of occurrence indicates are as follows:
Wherein, f (πi) indicate statistical arrangement sequence πiThe function of frequency of occurrence, wheneverInner element puts in order as πiWhen, f
(πi) just add 1, it arranges entropy and is defined as follows shown in formula,
Step 3 two or three: assuming thatInitial value be 0, for time seriesThe process that t gradually increases to N-d+1 from 1
In, whenever occur putting in order forWhen,It will be updated,
The weight of deviation, generally takes 0.5 between adjustment signal amplitude mean value and amplitude；
Step 3 two or four: in entire time seriesThe probability of appearanceAre as follows:
The amplitude perception arrangement entropy of time series indicates are as follows:
5. the Fault Diagnosis of Roller Bearings according to claim 1 based on the multiple dimensioned amplitude perception arrangement entropy of improvement,
It is characterized in that random forest grader specific steps described in step 4 are as follows: assuming that random forest grader is by multiple decisions
Set { hj(x,Θk), k=1,2 ..., n } composition, { Θk, k=1,2 ..., n indicate it is mutually indepedent and with distribution it is random to
Amount；Training sample set representations the D={ (x of random forest grader1,y1),(x2,y2),...,(xN,yN), xi=(xi,1,...,
xi,p)TIndicate i-th of training sample xiWith p characteristic value, yiIndicate training sample xiCorresponding label；To training sample set D
N times Bootstrap sampling is carried out, the n subsample Bootstrap D is obtainedj(j=1,2 ..., n)；For each subsample Dj, structure
Build decision-tree model hj(x), final to obtain by one group of decision tree { h1(x),h2(x),…,hk(x) } decision tree classification formed
Device；The test sample new for one is voted by n decision tree, obtains the classifications of most polls as test sample most
Whole classification, categorised decision are as follows:
Wherein, hj(x) jth decision tree is represented, I () is indicative function, i.e., is 1 when there is this number duration in set, otherwise value is
0；Y indicates class label yiThe target variable of composition.
6. the Fault Diagnosis of Roller Bearings according to claim 2 based on the multiple dimensioned amplitude perception arrangement entropy of improvement,
It is characterized in that α=0.5 in the step 2 two.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910273110.8A CN109916628B (en) | 2019-04-04 | 2019-04-04 | Rolling bearing fault diagnosis method based on improved multi-scale amplitude perception permutation entropy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910273110.8A CN109916628B (en) | 2019-04-04 | 2019-04-04 | Rolling bearing fault diagnosis method based on improved multi-scale amplitude perception permutation entropy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109916628A true CN109916628A (en) | 2019-06-21 |
| CN109916628B CN109916628B (en) | 2020-11-06 |
Family
ID=66968862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910273110.8A Active CN109916628B (en) | 2019-04-04 | 2019-04-04 | Rolling bearing fault diagnosis method based on improved multi-scale amplitude perception permutation entropy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109916628B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672326A (en) * | 2019-09-29 | 2020-01-10 | 上海联影智能医疗科技有限公司 | Bearing fault detection method and computer readable storage medium |
| CN110866519A (en) * | 2019-12-24 | 2020-03-06 | 安徽工业大学 | Rolling bearing fault diagnosis method based on Fourier decomposition and multi-scale arrangement entropy partial mean value |
| CN110929673A (en) * | 2019-12-02 | 2020-03-27 | 吉林松江河水力发电有限责任公司 | Transformer winding vibration signal identification method based on ITD (inverse discrete cosine transformation) permutation entropy and CGWO-SVM (Carrier-support vector machine) |
| CN110991422A (en) * | 2019-12-25 | 2020-04-10 | 安徽工业大学 | Rolling bearing fault diagnosis method based on multi-element time-shifting multi-scale permutation entropy |
| CN111060314A (en) * | 2019-11-23 | 2020-04-24 | 五邑大学 | Fault diagnosis method and test simulation device for rolling bearing of motor train unit |
| CN111103139A (en) * | 2019-12-31 | 2020-05-05 | 福州大学 | Rolling bearing fault diagnosis method based on GRCMSE and manifold learning |
| CN111122162A (en) * | 2019-12-25 | 2020-05-08 | 杭州电子科技大学 | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy |
| CN111397868A (en) * | 2020-02-27 | 2020-07-10 | 广西电网有限责任公司电力科学研究院 | Breaker fault analysis method based on aggregation empirical mode decomposition algorithm |
| CN111476323A (en) * | 2020-06-01 | 2020-07-31 | 合肥工业大学 | Bearing fault classification method and system |
| CN111723701A (en) * | 2020-06-08 | 2020-09-29 | 西安交通大学 | Underwater target identification method |
| CN112149845A (en) * | 2020-09-23 | 2020-12-29 | 山东通维信息工程有限公司 | Intelligent operation and maintenance method based on big data and machine learning |
| CN112444395A (en) * | 2020-11-15 | 2021-03-05 | 华东交通大学 | CMWPE and SaE-ELM based locomotive wheel pair bearing fault diagnosis method |
| CN112819029A (en) * | 2020-12-31 | 2021-05-18 | 杭州朗阳科技有限公司 | Novel machine learning method for diagnosing gearbox fault based on multi-scale permutation entropy |
| CN116316706A (en) * | 2023-05-08 | 2023-06-23 | 湖南大学 | Oscillation positioning method and system based on complementary average inherent time scale decomposition |
| CN116358864A (en) * | 2023-06-01 | 2023-06-30 | 西安因联信息科技有限公司 | Method and system for diagnosing fault type of rotary mechanical equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104849050A (en) * | 2015-06-02 | 2015-08-19 | 安徽工业大学 | Rolling bearing fault diagnosis method based on composite multi-scale permutation entropies |
| CN105971901A (en) * | 2016-05-03 | 2016-09-28 | 北京航空航天大学 | Centrifugal pump fault diagnosis method based on complete ensemble empirical mode decomposition and random forest |
| CN107101813A (en) * | 2017-04-26 | 2017-08-29 | 河北工业大学 | A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal |
| US20170255878A1 (en) * | 2016-03-07 | 2017-09-07 | Micron Technology, Inc. | Space efficient random forests implementation utilizing automata processors |
| CN108760300A (en) * | 2018-04-19 | 2018-11-06 | 西安工业大学 | A method of intelligent fault diagnosis being carried out to it according to bearing vibration signal |
| CN109540523A (en) * | 2018-11-27 | 2019-03-29 | 中国舰船研究设计中心 | A kind of ship centrifugal pump Fault Diagnosis of Roller Bearings |
-
2019
- 2019-04-04 CN CN201910273110.8A patent/CN109916628B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104849050A (en) * | 2015-06-02 | 2015-08-19 | 安徽工业大学 | Rolling bearing fault diagnosis method based on composite multi-scale permutation entropies |
| US20170255878A1 (en) * | 2016-03-07 | 2017-09-07 | Micron Technology, Inc. | Space efficient random forests implementation utilizing automata processors |
| CN105971901A (en) * | 2016-05-03 | 2016-09-28 | 北京航空航天大学 | Centrifugal pump fault diagnosis method based on complete ensemble empirical mode decomposition and random forest |
| CN107101813A (en) * | 2017-04-26 | 2017-08-29 | 河北工业大学 | A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal |
| CN108760300A (en) * | 2018-04-19 | 2018-11-06 | 西安工业大学 | A method of intelligent fault diagnosis being carried out to it according to bearing vibration signal |
| CN109540523A (en) * | 2018-11-27 | 2019-03-29 | 中国舰船研究设计中心 | A kind of ship centrifugal pump Fault Diagnosis of Roller Bearings |
Non-Patent Citations (3)
| Title |
|---|
| 仲太生 等: ""基于多尺度排列熵的自调心双列滚动轴承故障诊断"", 《机械设计与制造工程》 * |
| 瞿金秀 等: ""基于多尺度排列熵和支持向量机的轴承故障诊断"", 《煤矿机械》 * |
| 郑近德 等: ""多尺度排列熵及其在滚动轴承故障诊断中的应用"", 《中国机械工程》 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110672326A (en) * | 2019-09-29 | 2020-01-10 | 上海联影智能医疗科技有限公司 | Bearing fault detection method and computer readable storage medium |
| CN110672326B (en) * | 2019-09-29 | 2021-12-21 | 上海联影智能医疗科技有限公司 | Bearing fault detection method and computer readable storage medium |
| CN111060314B (en) * | 2019-11-23 | 2021-10-26 | 五邑大学 | Fault diagnosis method and test simulation device for rolling bearing of motor train unit |
| CN111060314A (en) * | 2019-11-23 | 2020-04-24 | 五邑大学 | Fault diagnosis method and test simulation device for rolling bearing of motor train unit |
| CN110929673A (en) * | 2019-12-02 | 2020-03-27 | 吉林松江河水力发电有限责任公司 | Transformer winding vibration signal identification method based on ITD (inverse discrete cosine transformation) permutation entropy and CGWO-SVM (Carrier-support vector machine) |
| CN110866519A (en) * | 2019-12-24 | 2020-03-06 | 安徽工业大学 | Rolling bearing fault diagnosis method based on Fourier decomposition and multi-scale arrangement entropy partial mean value |
| CN110866519B (en) * | 2019-12-24 | 2023-04-28 | 安徽工业大学 | Rolling bearing fault diagnosis method based on Fourier decomposition and multiscale permutation entropy deviation value |
| CN110991422A (en) * | 2019-12-25 | 2020-04-10 | 安徽工业大学 | Rolling bearing fault diagnosis method based on multi-element time-shifting multi-scale permutation entropy |
| CN110991422B (en) * | 2019-12-25 | 2023-05-26 | 安徽工业大学 | Rolling bearing fault diagnosis method based on multi-element time-shifting multi-scale permutation entropy |
| CN111122162A (en) * | 2019-12-25 | 2020-05-08 | 杭州电子科技大学 | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy |
| CN111103139A (en) * | 2019-12-31 | 2020-05-05 | 福州大学 | Rolling bearing fault diagnosis method based on GRCMSE and manifold learning |
| WO2021135630A1 (en) * | 2019-12-31 | 2021-07-08 | 福州大学 | Rolling bearing fault diagnosis method based on grcmse and manifold learning |
| CN111397868A (en) * | 2020-02-27 | 2020-07-10 | 广西电网有限责任公司电力科学研究院 | Breaker fault analysis method based on aggregation empirical mode decomposition algorithm |
| CN111397868B (en) * | 2020-02-27 | 2022-02-08 | 广西电网有限责任公司电力科学研究院 | Breaker fault analysis method based on aggregation empirical mode decomposition algorithm |
| CN111476323B (en) * | 2020-06-01 | 2023-05-12 | 合肥工业大学 | Bearing fault classification method and system |
| CN111476323A (en) * | 2020-06-01 | 2020-07-31 | 合肥工业大学 | Bearing fault classification method and system |
| CN111723701A (en) * | 2020-06-08 | 2020-09-29 | 西安交通大学 | Underwater target identification method |
| CN112149845A (en) * | 2020-09-23 | 2020-12-29 | 山东通维信息工程有限公司 | Intelligent operation and maintenance method based on big data and machine learning |
| CN112444395A (en) * | 2020-11-15 | 2021-03-05 | 华东交通大学 | CMWPE and SaE-ELM based locomotive wheel pair bearing fault diagnosis method |
| CN112819029A (en) * | 2020-12-31 | 2021-05-18 | 杭州朗阳科技有限公司 | Novel machine learning method for diagnosing gearbox fault based on multi-scale permutation entropy |
| CN116316706A (en) * | 2023-05-08 | 2023-06-23 | 湖南大学 | Oscillation positioning method and system based on complementary average inherent time scale decomposition |
| CN116316706B (en) * | 2023-05-08 | 2023-07-21 | 湖南大学 | Oscillation positioning method and system based on complementary average inherent time scale decomposition |
| CN116358864A (en) * | 2023-06-01 | 2023-06-30 | 西安因联信息科技有限公司 | Method and system for diagnosing fault type of rotary mechanical equipment |
| CN116358864B (en) * | 2023-06-01 | 2023-08-29 | 西安因联信息科技有限公司 | Method and system for diagnosing fault type of rotary mechanical equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109916628B (en) | 2020-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109916628A (en) | Based on the Fault Diagnosis of Roller Bearings for improving multiple dimensioned amplitude perception arrangement entropy | |
| Zheng et al. | Rolling bearing fault detection and diagnosis based on composite multiscale fuzzy entropy and ensemble support vector machines | |
| Fu et al. | Fault feature selection and diagnosis of rolling bearings based on EEMD and optimized Elman_AdaBoost algorithm | |
| Yan et al. | Deep regularized variational autoencoder for intelligent fault diagnosis of rotor–bearing system within entire life-cycle process | |
| Wang et al. | Fault diagnosis of a rolling bearing using wavelet packet denoising and random forests | |
| CN111797567B (en) | Bearing fault classification method and system based on deep learning network | |
| CN114090396B (en) | Cloud environment multi-index unsupervised anomaly detection and root cause analysis method | |
| CN105275833A (en) | CEEMD (Complementary Empirical Mode Decomposition)-STFT (Short-Time Fourier Transform) time-frequency information entropy and multi-SVM (Support Vector Machine) based fault diagnosis method for centrifugal pump | |
| Bai et al. | Research on feature selection for rotating machinery based on Supervision Kernel Entropy Component Analysis with Whale Optimization Algorithm | |
| CN109583386A (en) | A kind of intelligent rotating mechanical breakdown depth network characterization discrimination method | |
| CN105258789B (en) | A kind of extracting method and device of vibration signal characteristics frequency band | |
| CN105910823A (en) | Rolling bearing fault diagnosis method | |
| CN107451760B (en) | Rolling bearing fault diagnosis method based on time window slip limited Boltzmann machine | |
| CN115221930B (en) | Fault diagnosis method for rolling bearing | |
| CN103291600A (en) | Fault diagnosis method for hydraulic pump based on EMD-AR (empirical mode decomposition-auto-regressive) and MTS (mahalanobis taguchi system) | |
| Zheng et al. | Refined time-shift multiscale normalised dispersion entropy and its application to fault diagnosis of rolling bearing | |
| CN110458240A (en) | A kind of three-phase bridge rectifier method for diagnosing faults, terminal device and storage medium | |
| CN113532829A (en) | Reciprocating compressor fault diagnosis method based on improved RCMDE | |
| CN114417699A (en) | Pump valve fault detection method | |
| CN111259949A (en) | Fault identification model construction method, model and identification method for aircraft environmental control system | |
| CN116186590A (en) | Bearing fault diagnosis method based on data layer feature fusion and convolutional neural network | |
| Pranesh et al. | Vibration-based brake health prediction using statistical features–A machine learning framework | |
| Liu et al. | Fault diagnosis method of rolling bearing based on the multiple features of LMD and random forest | |
| CN108414228B (en) | Based on averagely more granularity decision rough sets and NNBC Method for Bearing Fault Diagnosis | |
| CN116429426A (en) | Bearing fault diagnosis method, device and medium for multi-domain feature fusion |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221102 Address after: Room 103-1, Yizi Building, No. 23, Sandadi Road, Xiangfang District, Harbin City, Heilongjiang Province, 150000 Patentee after: Harbin Kesu Intelligent Technology Co.,Ltd. Address before: 150080 No. 52, Xuefu Road, Nangang District, Heilongjiang, Harbin Patentee before: HARBIN University OF SCIENCE AND TECHNOLOGY |