CN108955855A - Vibration signal characteristics extracting method, monitoring method and device towards rotating machinery - Google Patents
Vibration signal characteristics extracting method, monitoring method and device towards rotating machinery Download PDFInfo
- Publication number
- CN108955855A CN108955855A CN201810586819.9A CN201810586819A CN108955855A CN 108955855 A CN108955855 A CN 108955855A CN 201810586819 A CN201810586819 A CN 201810586819A CN 108955855 A CN108955855 A CN 108955855A
- Authority
- CN
- China
- Prior art keywords
- vibration signal
- rotating machinery
- period
- vibration
- group
- 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
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H1/00—Measuring characteristics of vibrations in solids by using direct conduction to the detector
- G01H1/003—Measuring characteristics of vibrations in solids by using direct conduction to the detector of rotating machines
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H1/00—Measuring characteristics of vibrations in solids by using direct conduction to the detector
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01P—MEASURING LINEAR OR ANGULAR SPEED, ACCELERATION, DECELERATION, OR SHOCK; INDICATING PRESENCE, ABSENCE, OR DIRECTION, OF MOVEMENT
- G01P3/00—Measuring linear or angular speed; Measuring differences of linear or angular speeds
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a kind of vibration signal characteristics extracting method, monitoring method and device towards rotating machinery.Wherein, vibration signal characteristics extracting method includes step (1) building code book: using collected vibration signal as training sample, training sample is clustered, one group of coded word is constituted by the cluster mass center of every one kind and cluster variance, and then construct code book；Step (2) divides the period: carrying out period division to the vibration signal of real-time monitoring；Step (3) calculates related figure；Step (4) calculates entropy: carrying out dimension-reduction treatment to obtained related graphic sequence using entropy, obtains the feature of reaction machine performance operation trend.Related figure is applied in characteristic extraction procedure by it, during carrying out feature extraction by the amplitude of vibration signal, it is also contemplated that the timing information of original vibration signal, making the feature extracted includes more effective informations.
Description
Technical field
The invention belongs to rotating machinery field more particularly to a kind of vibration signal characteristics extraction sides towards rotating machinery
Method, monitoring method and device.
Background technique
In recent years, rotating machinery develops towards complication, enlarged direction, and application is also more and more extensive, and any one
The appearance of minor issue is likely to lead to serious consequence, therefore being monitored to the operating status of rotating machinery is very must
It wants.Wherein, being monitored to the revolving speed of rotating machinery is a major issue, is supervised by the revolving speed to rotating machinery
It surveys, can be realized the monitoring of running state of rotating machine, suffer from the fields such as machining, digital control and answer extensively
With.
It is had the following problems towards rotating machinery rotation speed monitoring:
(1) in practical applications, since by rotating machinery size, the limitation of installation site and environment is directly utilized
It is highly difficult that speedometer, which is monitored the revolving speed of rotating machinery,.
(2) carrying out analysis by the vibration signal generated to rotation speed change can be realized rotating machinery rotation speed monitoring, at present
Common analysis of vibration signal method has frequency-domain analysis and two kinds of time-domain analysis.
(2.1) although the frequency-domain analysis method based on Fourier transformation is widely used, when collected
When data amplitude is smaller or collected data volume is smaller, frequency-domain analysis is not applicable.
(2.2) time-domain analysis generally comprises two parts: feature extraction and decision.Wherein feature extraction is in detection algorithm
It is played a very important role in design, and has important influence to final testing result.Current existing temporal signatures
Extracting method has root-mean-square value, and variance, kurtosis, the degree of bias, peak value etc., these feature extracting methods are all to only rely on vibration signal
Amplitude carry out feature extraction, ignore the timing information of vibration signal.
Summary of the invention
In order to solve the deficiencies in the prior art, the first object of the present invention is to provide a kind of vibration towards rotating machinery
Signal characteristic extracting methods consider the timing information of original vibration signal, and making the feature extracted includes more to have
Imitate information.
A kind of vibration signal characteristics extracting method towards rotating machinery of the invention, comprising:
Step (1) constructs code book: using collected vibration signal as training sample, training sample clustered, by
The cluster mass center and cluster variance of every one kind constitute one group of coded word, and then construct code book；
Step (2) divides the period: carrying out period division to the vibration signal of real-time monitoring；
Step (3) calculates related figure, specifically includes:
Step (3.1): one group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
Step (3.2): any to choose two groups of coded words in code book；
Step (3.3): a vibration data in every group of vibration data is calculated to one group chosen using gaussian kernel function
The mapping value of coded word and another vibration data to another group of coded word chosen mapping value, by the two mapping values
It is multiplied, the mapping value result of product of all groups of vibration datas that finally add up again obtains the correlation of the current period of vibration signal
Figure；
Step (3.4): to the vibration signal for continuously monitoring obtained each period, step (3.2)~step is repeated
(3.3), number related graphic sequence identical to number of cycles is obtained；
Step (4) calculates entropy: carrying out dimension-reduction treatment to obtained related graphic sequence using entropy, it is mechanical to obtain reaction
The feature of state operation trend.
Further, in the step (1) building code book, training sample is clustered using K-means algorithm.
K-means algorithm is input cluster number k, and the database comprising n data object, output meet variance
A kind of algorithm of minimum sandards k cluster.K-means algorithm receives input quantity k；Then n data object is divided into k
Cluster to meet cluster obtained: the object similarity in same cluster is higher；And the object phase in different clusters
It is smaller like spending.
It should be noted that K- can also be used other than clustering using K-means algorithm to training sample
MEDOIDS algorithm or CLARANS algorithm cluster training sample.
Further, in the step (2) segmentation period, the vibration of real-time monitoring is believed using time series analysis
Number carry out period division, detailed process, comprising:
Step (2.1): based on the vibration signal newly monitored, one group of candidate's vibration signal segment A is determined；
Step (2.2): any to choose with identical as the starting point for selecting vibration signal segment, end point is in the choosing
A candidate segment B in vibration signal segment calculates candidate segment B and newly monitors using dynamic time warping method
The matching loss of vibration signal；
Step (2.3): with the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Step (2.4): period division is carried out to vibration signal according to optimal period.
The present invention using optimal period to vibration signal carry out period division, can quickly and accurately to vibration signal into
Row feature extraction.
It should be noted that period division can also be carried out using vibration signal of the predetermined period to real-time monitoring.
The second object of the present invention is to provide a kind of vibration signal characteristics extraction element towards rotating machinery, considers
The timing information for having arrived original vibration signal, making the feature extracted includes more effective informations.
A kind of vibration signal characteristics extraction element towards rotating machinery of the invention, including vibration signal characteristics extract
Processor, the vibration signal characteristics extraction processor, comprising:
Construct code book module, be configured as: using collected vibration signal as training sample, to training sample into
Row cluster constitutes one group of coded word by the cluster mass center of every one kind and cluster variance, and then constructs code book；
Divide cycle module, be configured as: period division is carried out to the vibration signal of real-time monitoring；
Related module is calculated, is configured as:
One group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
It is any to choose two groups of coded words in code book；
The reflecting to the one group of coded word chosen of a vibration data in every group of vibration data is calculated using gaussian kernel function
Value and another vibration data are penetrated to the mapping value for another group of coded word chosen, the two mapping values are multiplied, finally
Add up the mapping value result of product of all groups of vibration datas again, obtains the related figure of the current period of vibration signal；
Obtain number related graphic sequence identical to vibration signal number of cycles.
Entropy module is calculated, is configured as: dimension-reduction treatment being carried out to obtained related graphic sequence using entropy, is obtained
React the feature of machine performance operation trend.
Further, in the building code book module, training sample is clustered using K-means algorithm.
K-means algorithm is input cluster number k, and the database comprising n data object, output meet variance
A kind of algorithm of minimum sandards k cluster.K-means algorithm receives input quantity k；Then n data object is divided into k
Cluster to meet cluster obtained: the object similarity in same cluster is higher；And the object phase in different clusters
It is smaller like spending.
It should be noted that K- can also be used other than clustering using K-means algorithm to training sample
MEDOIDS algorithm or CLARANS algorithm cluster training sample.
Further, in the segmentation cycle module, using time series analysis to the vibration signal of real-time monitoring into
The row period divides, and the segmentation cycle module is also configured to
Based on the vibration signal newly monitored, one group of candidate's vibration signal segment A is determined；
Any to choose with identical as the starting point for selecting vibration signal segment, end point selects vibration signal piece described
A candidate segment B in section calculates candidate segment B and the vibration signal that newly monitors using dynamic time warping method
Matching loss；
With the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Period division is carried out to vibration signal according to optimal period.
The present invention using optimal period to vibration signal carry out period division, can quickly and accurately to vibration signal into
Row feature extraction.
It should be noted that period division can also be carried out using vibration signal of the predetermined period to real-time monitoring.
The third object of the present invention is to provide a kind of method for monitoring operation states of rotating machinery, uses time-domain analysis
Method the vibration signal of rotating machinery is analyzed, acquisition reaction machine performance operation trend feature, recycle Europe
Formula distance carries out abnormality degree measurement to feature, finally carries out decision to obtained abnormality degree using hypothesis testing, can be quickly quasi-
Really monitor the operating status of rotating machinery.
A kind of method for monitoring operation states of rotating machinery of the invention, comprising:
It extracts characterization step: being obtained using the vibration signal characteristics extracting method described above towards rotating machinery
React the feature of machine performance operation trend；
Calculate abnormality degree step: the feature of the reaction machine performance operation trend based on acquisition, using Euclidean distance to it
Carry out abnormality degree measurement；
Analysis of Policy Making step: decision is carried out to obtained abnormality degree using hypothesis testing.
Further, in the Analysis of Policy Making step, rotating machinery is examined using 3 σ principles in hypothesis testing
Whether operating status changes generation.
Where it is assumed that examining is to infer a kind of overall method by sample according to certain assumed condition in mathematical statistics.
The specific practice is: making certain it is assumed that being denoted as H0 to the totality studied according to the needs of problem；Suitable statistic is chosen, this
The selection of a statistic will make when assuming that H0 is set up, and be distributed as known；By the sample surveyed, statistic is calculated
Value, and tested according to previously given significance, make refusal or receive to assume the judgement of H0.
Hypothesis testing is an important content in Sampling Deduction.It is to make an overall objective according to raw data to be
It is no be equal to some numerical value, a certain stochastic variable whether obey certain probability distribution it is assumed that then using sample data use
Certain statistical method calculates the statistic in relation to examining, and according to certain principle of probability, judges to estimate with lesser risk
Count value and overall numerical value (or estimation distribution and actual distribution) whether there is significant difference, if should receive null hypothesis
A kind of method of inspection of selection.
The fourth object of the present invention is to provide a kind of monitoring running state device of rotating machinery, uses time-domain analysis
Method the vibration signal of rotating machinery is analyzed, acquisition reaction machine performance operation trend feature, recycle Europe
Formula distance carries out abnormality degree measurement to feature, finally carries out decision to obtained abnormality degree using hypothesis testing, can be quickly quasi-
Really monitor the operating status of rotating machinery.
A kind of monitoring running state device of rotating machinery of the invention, comprising:
Vibration signal characteristics extraction element towards rotating machinery described above；
The vibration signal characteristics extraction element is also connected with status monitoring processor, the status monitoring processor packet
It includes:
Calculate abnormality degree module, be configured as: the feature of the reaction machine performance operation trend based on acquisition utilizes
Euclidean distance carries out abnormality degree measurement to it；
Analysis of Policy Making module, is configured as: carrying out decision to obtained abnormality degree using hypothesis testing.
Further, in the Analysis of Policy Making module, rotating machinery is examined using 3 σ principles in hypothesis testing
Whether operating status changes generation.
Where it is assumed that examining is to infer a kind of overall method by sample according to certain assumed condition in mathematical statistics.
The specific practice is: making certain it is assumed that being denoted as H0 to the totality studied according to the needs of problem；Suitable statistic is chosen, this
The selection of a statistic will make when assuming that H0 is set up, and be distributed as known；By the sample surveyed, statistic is calculated
Value, and tested according to previously given significance, make refusal or receive to assume the judgement of H0.
Hypothesis testing is an important content in Sampling Deduction.It is to make an overall objective according to raw data to be
It is no be equal to some numerical value, a certain stochastic variable whether obey certain probability distribution it is assumed that then using sample data use
Certain statistical method calculates the statistic in relation to examining, and according to certain principle of probability, judges to estimate with lesser risk
Count value and overall numerical value (or estimation distribution and actual distribution) whether there is significant difference, if should receive null hypothesis
A kind of method of inspection of selection.
Compared with prior art, the beneficial effects of the present invention are:
(1) in the present invention, related figure is applied in characteristic extraction procedure, is carried out in the amplitude by vibration signal special
During sign is extracted, it is also contemplated that the timing information of original vibration signal, make the feature extracted include it is more effectively
Information.
(2) in the present invention, it is analyzed using vibration signal of the method for time-domain analysis to rotating machinery, acquisition
The feature of machine performance operation trend is reacted, Euclidean distance is recycled to carry out abnormality degree measurement to feature, finally using hypothesis inspection
It tests and decision is carried out to obtained abnormality degree, can rapidly and accurately monitor the operating status of rotating machinery.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, the application's
Illustrative embodiments and their description are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is a kind of vibration signal characteristics extracting method flow chart towards rotating machinery of the invention.
Fig. 2 is the acquisition signal schematic representation that the present invention is applied in rotating machinery rotation speed monitoring.
Fig. 3 is the feature extraction schematic diagram that the present invention is applied in rotating machinery rotation speed monitoring.
Fig. 4 is vibration signal characteristics extraction processor structural schematic diagram of the invention.
Fig. 5 is a kind of flow chart of the method for monitoring operation states of rotating machinery of the invention.
Fig. 6 is the abnormality degree schematic diagram that the present invention is applied in rotating machinery rotation speed monitoring.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless
Otherwise indicated, all technical and scientific terms used herein has and the application person of an ordinary skill in the technical field
Normally understood identical meanings.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular shape
Formula be also intended to include plural form, additionally, it should be understood that, when in the present specification use term "comprising" and/or
When " comprising ", existing characteristics, step, operation, device, component and/or their combination are indicated.
Fig. 1 is a kind of vibration signal characteristics extracting method flow chart towards rotating machinery of the invention.
As shown in Figure 1, a kind of vibration signal characteristics extracting method towards rotating machinery of the invention, including step (1)
~step (4).Specifically:
Step (1) constructs code book: using collected vibration signal as training sample, training sample clustered, by
The cluster mass center and cluster variance of every one kind constitute one group of coded word, and then construct code book.
In the step (1) building code book, training sample is clustered using K-means algorithm.
K-means algorithm is input cluster number k, and the database comprising n data object, output meet variance
A kind of algorithm of minimum sandards k cluster.K-means algorithm receives input quantity k；Then n data object is divided into k
Cluster to meet cluster obtained: the object similarity in same cluster is higher；And the object phase in different clusters
It is smaller like spending.
It should be noted that K- can also be used other than clustering using K-means algorithm to training sample
MEDOIDS algorithm or CLARANS algorithm cluster training sample.
Specifically, step (1) building code book includes:
Step (1.1): will on rotating machinery collected vibration signal xi(i=1,2 ..., m) are as training sample
This collection D={ x1,x2,…,xm}；Wherein, m is positive integer；
Step (1.2): setting clusters number as k, and k sample is randomly choosed from training sample set D as initial k
Mass center: { μ1,μ2,…,μk, each mass center is corresponding a kind of, then a shared k class, is denoted as { Q1,Q2,…,Qk}；
Step (1.3): for i=1,2 ..., m calculates sample xiWith each mass center μj, the distance of (j=1,2 ..., k)
di,j:
di,j=| | xi-μj||2 2
By xiIt is divided into the smallest di,jCorresponding classification QjIn, and update Qj=Qj∪{xi}；
Step (1.4): to QjIn all sample points recalculate mass center:
Step (1.5): iterative step (1.3) step (1.4) the two steps, until new mass center and the protoplasm heart are equal
Or be less than and formulate threshold value, calculate every one kind CjIn all samples variance:
The mass center and variance of each class can be obtained by above-mentioned calculating, the mass center and variance of every one kind constitute one group
Coded word, a shared k class, then a shared k group coded word, this k group coded word constitute code book.
Step (2) divides the period: carrying out period division to the vibration signal of real-time monitoring.
Specifically, in the step (2) segmentation period, using time series analysis to the vibration signal of real-time monitoring
Carry out period division, detailed process, comprising:
Step (2.1): based on the vibration signal X={ x newly monitored1,x2,…,xZ, determine one group of candidate's vibration signal
Segment A=Y={ y1,…,yl}；Wherein Z is positive integer；l∈[la,lb].Wherein laAnd lbRule of thumb estimation obtains and is
Positive integer.
Step (2.2): any to choose with identical as the starting point for selecting vibration signal segment, end point is in the choosing
A candidate segment B in vibration signal segment calculates candidate segment B and newly monitors using dynamic time warping method
The matching loss of vibration signal；
For a candidate segment B=Y={ y1,…,ylv, lv∈[la,lb], v ∈ [a, b] is advised using dynamic time
Whole (DTW) calculates it and the vibration signal X={ x that newly monitors1,x2,…,xZMatching loss, the formula of matching loss
Are as follows:
Dv(Y, X)=d (yi,xj)
Wherein:
d(y1,x1)=| y1-x1|；
d(y1,x2)=| y1-x2|；
d(y2,x1)=| y2-x1|；
Wherein | yi,xj| it is yiWith xjDifference.
To v=a ..., b, each candidate segment B=Y={ y1,…,ylvMatching loss is computed repeatedly, obtain { Dv,
V=a ..., b find the smallest DvCorresponding l, as optimal period T.
Step (2.3): with the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Step (2.4): period division is carried out to vibration signal according to optimal period.
The present invention using optimal period to vibration signal carry out period division, can quickly and accurately to vibration signal into
Row feature extraction.
It should be noted that period division can also be carried out using vibration signal of the predetermined period to real-time monitoring.
Step (3) calculates related figure, specifically includes:
Step (3.1): one group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
Step (3.2): any to choose two groups of coded words in code book；
Step (3.3): a vibration data in every group of vibration data is calculated to one group chosen using gaussian kernel function
The mapping value of coded word and another vibration data to another group of coded word chosen mapping value, by the two mapping values
It is multiplied, the mapping value result of product of all groups of vibration datas that finally add up again obtains the correlation of the current period of vibration signal
Figure；
Step (3.4): to the vibration signal for continuously monitoring obtained each period, step (3.2)~step is repeated
(3.3), obtain number with periodicity identical phase graphic sequence；
Such as: the number x for being Δ t for time interval in a cycle dividedtAnd xt+Δt, using gaussian kernel function,
It is calculated separately to i-th group of coded word (μi,σi 2) and jth group coded word (μj,σj 2) mapping, obtain Wi,tAnd Wj,t+Δt, Gauss
Kernel function calculation formula is as follows:
And calculate Wi,tAnd Wj,t+ΔtProduct；
The number x for being Δ t to each pair of time interval in n-th of period that length is TtAnd xt+Δt, t ∈ (1,2 ..., T-
Δ t) carries out gaussian kernel function calculating, obtains t W of T- Δi,tAnd Wj,t+ΔtProduct these product accumulations are obtained into correlation
Scheme Cn(i,j；Δ t), Cn(i,j；The calculation formula of Δ t) is as follows:
Assuming that there is k group coded word, i.e. i=1,2 ..., k, j=1,2 ..., k, then CnIt is the matrix of a k × k；
By the related figure C of k × knEach of element by row it is end to end, write as one have k2Capable column vector
C'n；
Assuming that there is N number of period, repeat the above steps to each period, obtain N number of related figure, by this N correlation figure according to
K is obtained by column arrangement according to cycle sequences2The dependency graph representation of row N column, { C'1,C'2,…,C'n,…C'N}。
Step (4) calculates entropy: carrying out dimension-reduction treatment to obtained related graphic sequence using entropy, it is mechanical to obtain reaction
The feature of state operation trend, detailed process include:
Step (4.1): for the k being calculated2The n-th column C' in the related figure of row N columnn, its entropy is calculated, formula is such as
Under
Wherein
Wherein N is related figure columns, that is, number of cycles；N=1,2 ..., N represent a certain column of related figure, i.e., certain
A cycle, m=1,2 ..., k2, represent certain a line of related figure；Q is a constant, pmnIndicate the m in n-th of period
Shared specific gravity of the related map values cmn of number in n-th of period；
Step (4.2): it for each period, repeats step (4.1), obtains { e1,…,en,…eN, as extract
Feature.
The vibration signal characteristics that an engine mission rotation speed change has been shown below extract example, and experimental result is such as
Shown in Fig. 3.Example is simply introduced below:
Original signal is gearbox vibration signal, and engine initial speed is 300rpm, and then revolving speed becomes 400rpm,
Collected vibration signal mentions gearbox as shown in Fig. 2, collected vibration signal is carried out feature according to above-mentioned steps in real time
It takes, the feature extracted is as shown in Figure 3.
In the present invention, related figure is applied in characteristic extraction procedure, feature is carried out in the amplitude by vibration signal and mentions
During taking, it is also contemplated that the timing information of original vibration signal, making the feature extracted includes more effectively letters
Breath.
The present invention also provides a kind of vibration signal characteristics extraction element towards rotating machinery, considers original
The timing information of vibration signal, making the feature extracted includes more effective informations.
A kind of vibration signal characteristics extraction element towards rotating machinery of the invention, including vibration signal characteristics extract
Processor.As shown in figure 4, the vibration signal characteristics extraction processor, comprising:
(1) code book module is constructed, is configured as: using collected vibration signal as training sample, to training sample
It is clustered, one group of coded word is constituted by the cluster mass center of every one kind and cluster variance, and then construct code book.
Specifically, in the building code book module, training sample is clustered using K-means algorithm.
K-means algorithm is input cluster number k, and the database comprising n data object, output meet variance
A kind of algorithm of minimum sandards k cluster.K-means algorithm receives input quantity k；Then n data object is divided into k
Cluster to meet cluster obtained: the object similarity in same cluster is higher；And the object phase in different clusters
It is smaller like spending.
It should be noted that K- can also be used other than clustering using K-means algorithm to training sample
MEDOIDS algorithm or CLARANS algorithm cluster training sample.
(2) divide cycle module, be configured as: period division is carried out to the vibration signal of real-time monitoring.
Specifically, it in the segmentation cycle module, is carried out using vibration signal of the time series analysis to real-time monitoring
Period divides, and the segmentation cycle module is also configured to
Based on the vibration signal newly monitored, one group of candidate's vibration signal segment A is determined；
Any to choose with identical as the starting point for selecting vibration signal segment, end point selects vibration signal piece described
A candidate segment B in section calculates candidate segment B and the vibration signal that newly monitors using dynamic time warping method
Matching loss；
With the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Period division is carried out to vibration signal according to optimal period.
The present invention using optimal period to vibration signal carry out period division, can quickly and accurately to vibration signal into
Row feature extraction.
It should be noted that period division can also be carried out using vibration signal of the predetermined period to real-time monitoring.
(3) related module is calculated, is configured as:
One group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
It is any to choose two groups of coded words in code book；
The reflecting to the one group of coded word chosen of a vibration data in every group of vibration data is calculated using gaussian kernel function
Value and another vibration data are penetrated to the mapping value for another group of coded word chosen, the two mapping values are multiplied, finally
Add up the mapping value result of product of all groups of vibration datas again, obtains the related figure of the current period of vibration signal；
Obtain number related graphic sequence identical to vibration signal number of cycles.
(4) entropy module is calculated, is configured as: dimension-reduction treatment being carried out to obtained related graphic sequence using entropy, is obtained
To the feature of reaction machine performance operation trend.
In the present invention, related figure is applied in characteristic extraction procedure, feature is carried out in the amplitude by vibration signal and mentions
During taking, it is also contemplated that the timing information of original vibration signal, making the feature extracted includes more effectively letters
Breath.
Fig. 5 is a kind of flow chart of the method for monitoring operation states of rotating machinery of the invention.
As shown in figure 5, a kind of method for monitoring operation states of rotating machinery of the invention, comprising:
(1) extract characterization step: using as shown in Figure 1 the vibration signal characteristics extracting method towards rotating machinery come
Obtain the feature of reaction machine performance operation trend.
(2) calculate abnormality degree step: the feature of the reaction machine performance operation trend based on acquisition utilizes Euclidean distance
Abnormality degree measurement is carried out to it.
Feature { the e extracted based on method as shown in Figure 11,…,en,…eN, calculate en-1And enBetween it is poor exhausted
Abnormality degree is used as to value, i.e.,
sn=| en-en-1|
Successively to { e1,…,en,…eNIn two neighboring number as above calculated, obtain abnormality degree { s2,…,sn,…
sN}。
(3) decision Analysis of Policy Making step: is carried out to obtained abnormality degree using hypothesis testing.
In the Analysis of Policy Making step, using 3 in hypothesis testingσPrinciple examines the operating status of rotating machinery
Whether generation is changed.
For obtained abnormal angle value { s2,…,sn,…sN, change has been checked whether using 3 σ principles in hypothesis testing
Change and occur, it is assumed that examines as follows
H0: | sn-μ'j| 3 σ ' of <j, do not change appearance；
H1: | sn-μ'j|≥3σ'j', change and occur.
Wherein, μ 'jWith σ 'jRespectively data { sj, j=2,3 ..., the average value and standard deviation of n-1；H0It indicates n-th
A period is without exception, H1It indicates to change generation n-th of period.
Where it is assumed that examining is to infer a kind of overall method by sample according to certain assumed condition in mathematical statistics.
The specific practice is: making certain it is assumed that being denoted as H0 to the totality studied according to the needs of problem；Suitable statistic is chosen, this
The selection of a statistic will make when assuming that H0 is set up, and be distributed as known；By the sample surveyed, statistic is calculated
Value, and tested according to previously given significance, make refusal or receive to assume the judgement of H0.
Hypothesis testing is an important content in Sampling Deduction.It is to make an overall objective according to raw data to be
It is no be equal to some numerical value, a certain stochastic variable whether obey certain probability distribution it is assumed that then using sample data use
Certain statistical method calculates the statistic in relation to examining, and according to certain principle of probability, judges to estimate with lesser risk
Count value and overall numerical value (or estimation distribution and actual distribution) whether there is significant difference, if should receive null hypothesis
A kind of method of inspection of selection.
The vibration signal characteristics that an engine mission rotation speed change has been shown below extract example:
Original signal is gearbox vibration signal, and engine initial speed is 300rpm, and then revolving speed becomes 400rpm,
Collected vibration signal mentions gearbox as shown in Fig. 2, collected vibration signal is carried out feature according to above-mentioned steps in real time
Decision is taken and statisticallys analyze, the feature extracted is as shown in figure 3, Fig. 6 is the abnormal angle value for the feature extracted, in Fig. 6
Labeled as the position for the rotation speed change point that this method detects.
As shown in figure 5, this method includes feature extraction and statistical analysis decision two parts.Wherein, feature extraction is divided into again
Study stage and test phase.The study stage is to be gathered using (K-means) to collected characteristic of rotating machines vibration signal
Class obtains the process of code book；Test phase is to carry out the period point to new collected vibration signal using dynamic time warping
It cuts；Related figure is carried out using the Gaussian kernel vibration signal good to period divisions to calculate, and obtained related figure is carried out with entropy
Dimension-reduction treatment obtains the process that can reflect the feature of machine performance trend.Finally, for statistical analysis for obtained feature
And decision, realize the monitoring to running state of rotating machine.
In the present invention, it is analyzed using vibration signal of the method for time-domain analysis to rotating machinery, the reaction of acquisition
The feature of machine performance operation trend recycles Euclidean distance to carry out abnormality degree measurement to feature, finally uses hypothesis testing pair
Obtained abnormality degree carries out decision, can rapidly and accurately monitor the operating status of rotating machinery.
The present invention also provides a kind of monitoring running state devices of rotating machinery.
A kind of monitoring running state device of rotating machinery of the invention, comprising:
Vibration signal characteristics extraction element towards rotating machinery；
The vibration signal characteristics extraction element is also connected with status monitoring processor, the status monitoring processor packet
It includes:
Calculate abnormality degree module, be configured as: the feature of the reaction machine performance operation trend based on acquisition utilizes
Euclidean distance carries out abnormality degree measurement to it；
Analysis of Policy Making module, is configured as: carrying out decision to obtained abnormality degree using hypothesis testing.
Specifically, in the Analysis of Policy Making module, the fortune of rotating machinery is examined using 3 σ principles in hypothesis testing
Whether row state changes generation.
Where it is assumed that examining is to infer a kind of overall method by sample according to certain assumed condition in mathematical statistics.
The specific practice is: making certain it is assumed that being denoted as H0 to the totality studied according to the needs of problem；Suitable statistic is chosen, this
The selection of a statistic will make when assuming that H0 is set up, and be distributed as known；By the sample surveyed, statistic is calculated
Value, and tested according to previously given significance, make refusal or receive to assume the judgement of H0.
Hypothesis testing is an important content in Sampling Deduction.It is to make an overall objective according to raw data to be
It is no be equal to some numerical value, a certain stochastic variable whether obey certain probability distribution it is assumed that then using sample data use
Certain statistical method calculates the statistic in relation to examining, and according to certain principle of probability, judges to estimate with lesser risk
Count value and overall numerical value (or estimation distribution and actual distribution) whether there is significant difference, if should receive null hypothesis
A kind of method of inspection of selection.
In the present invention, it is analyzed using vibration signal of the method for time-domain analysis to rotating machinery, the reaction of acquisition
The feature of machine performance operation trend recycles Euclidean distance to carry out abnormality degree measurement to feature, finally uses hypothesis testing pair
Obtained abnormality degree carries out decision, can rapidly and accurately monitor the operating status of rotating machinery.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects to the present invention
The limitation of range, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art
Member does not need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (10)
1. a kind of vibration signal characteristics extracting method towards rotating machinery characterized by comprising
Step (1) constructs code book: using collected vibration signal as training sample, clustering to training sample, by each
The cluster mass center and cluster variance of class constitute one group of coded word, and then construct code book；
Step (2) divides the period: carrying out period division to the vibration signal of real-time monitoring；
Step (3) calculates related figure, specifically includes:
Step (3.1): one group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
Step (3.2): any to choose two groups of coded words in code book；
Step (3.3): a vibration data in every group of vibration data is calculated to the one group of coding chosen using gaussian kernel function
The two mapping values are multiplied by the mapping value of word and another vibration data to the mapping value for another group of coded word chosen,
The mapping value result of product of the last all groups of vibration datas that add up again obtains the related figure of the current period of vibration signal；
Step (3.4): it to the vibration signal for continuously monitoring obtained each period, repeats step (3.2)~step (3.3), obtains
To number related graphic sequence identical to periodicity；
Step (4) calculates entropy: carrying out dimension-reduction treatment to obtained related graphic sequence using entropy, obtains reaction machine performance fortune
The feature of row trend.
2. a kind of vibration signal characteristics extracting method towards rotating machinery as described in claim 1, which is characterized in that in institute
It states in step (1) building code book, training sample is clustered using K-means algorithm.
3. a kind of vibration signal characteristics extracting method towards rotating machinery as described in claim 1, which is characterized in that in institute
It states in step (2) segmentation period, carries out period division, specific mistake using vibration signal of the time series analysis to real-time monitoring
Journey, comprising:
Step (2.1): based on the vibration signal newly monitored, one group of candidate's vibration signal segment A is determined；
Step (2.2): any to choose with identical as the starting point for selecting vibration signal segment, end point is vibrated in the choosing to be believed
A candidate segment B in number segment, the vibration signal for utilizing dynamic time warping method to calculate candidate segment B and newly monitor
Matching loss；
Step (2.3): with the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Step (2.4): period division is carried out to vibration signal according to optimal period.
4. a kind of vibration signal characteristics extraction element towards rotating machinery, which is characterized in that extracted including vibration signal characteristics
Processor, the vibration signal characteristics extraction processor, comprising:
Code book module is constructed, is configured as: using collected vibration signal as training sample, training sample being gathered
Class constitutes one group of coded word by the cluster mass center of every one kind and cluster variance, and then constructs code book；
Divide cycle module, be configured as: period division is carried out to the vibration signal of real-time monitoring；
Related module is calculated, is configured as:
One group of vibration data is formed by two vibration datas of prefixed time interval in a cycle；
It is any to choose two groups of coded words in code book；
Using gaussian kernel function calculate a vibration data in every group of vibration data to choose one group of coded word mapping value,
And the two mapping values are multiplied, finally add up again to the mapping value for another group of coded word chosen by another vibration data
The mapping value result of product of all groups of vibration datas obtains the related figure of the current period of vibration signal；
Obtain number related graphic sequence identical to vibration signal number of cycles；
Entropy module is calculated, is configured as: dimension-reduction treatment being carried out to obtained related graphic sequence using entropy, obtains reaction machine
The feature of tool state operation trend.
5. a kind of vibration signal characteristics extraction element towards rotating machinery as claimed in claim 4, which is characterized in that in institute
It states in building code book module, training sample is clustered using K-means algorithm.
6. a kind of vibration signal characteristics extraction element towards rotating machinery as claimed in claim 4, which is characterized in that in institute
It states in segmentation cycle module, carries out period division, the segmentation week using vibration signal of the time series analysis to real-time monitoring
Phase module is also configured to
Based on the vibration signal newly monitored, one group of candidate's vibration signal segment A is determined；
Any to choose with identical as the starting point for selecting vibration signal segment, end point is selected in vibration signal segment described
One candidate segment B calculates candidate segment B using dynamic time warping method and the matching of the vibration signal newly monitored is damaged
Consumption；
With the period of candidate segment B corresponding to matching loss minimum, as optimal period；
Period division is carried out to vibration signal according to optimal period.
7. a kind of method for monitoring operation states of rotating machinery characterized by comprising
It extracts characterization step: being mentioned using the vibration signal characteristics as claimed in any one of claims 1-3 towards rotating machinery
Method is taken to obtain the feature of reaction machine performance operation trend；
Calculate abnormality degree step: the feature of the reaction machine performance operation trend based on acquisition carries out it using Euclidean distance
Abnormality degree measurement；
Analysis of Policy Making step: decision is carried out to obtained abnormality degree using hypothesis testing.
8. a kind of method for monitoring operation states of rotating machinery as claimed in claim 7, which is characterized in that in the decision point
It analyses in step, examines whether the operating status of rotating machinery changes generation using 3 σ principles in hypothesis testing.
9. a kind of monitoring running state device of rotating machinery characterized by comprising
The vibration signal characteristics extraction element towards rotating machinery as described in any one of claim 4-6；
The vibration signal characteristics extraction element is also connected with status monitoring processor, and the status monitoring processor includes:
Calculate abnormality degree module, be configured as: based on acquisition reaction machine performance operation trend feature, using it is European away from
Abnormality degree measurement is carried out to it；
Analysis of Policy Making module, is configured as: carrying out decision to obtained abnormality degree using hypothesis testing.
10. the monitoring running state device of rotating machinery as claimed in claim 9, which is characterized in that in the Analysis of Policy Making
In module, examine whether the operating status of rotating machinery changes generation using 3 σ principles in hypothesis testing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810586819.9A CN108955855B (en) | 2018-06-08 | 2018-06-08 | Vibration signal feature extraction method, monitoring method and device for rotary machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810586819.9A CN108955855B (en) | 2018-06-08 | 2018-06-08 | Vibration signal feature extraction method, monitoring method and device for rotary machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108955855A true CN108955855A (en) | 2018-12-07 |
| CN108955855B CN108955855B (en) | 2020-05-08 |
Family
ID=64493498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810586819.9A Active CN108955855B (en) | 2018-06-08 | 2018-06-08 | Vibration signal feature extraction method, monitoring method and device for rotary machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108955855B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109655267A (en) * | 2019-01-07 | 2019-04-19 | 江苏大学 | A kind of automobile hub bearing fault signature extracting method |
| CN111929044A (en) * | 2020-07-15 | 2020-11-13 | 西门子工厂自动化工程有限公司 | Method, apparatus, computing device and storage medium for monitoring device status |
| CN111985380A (en) * | 2020-08-13 | 2020-11-24 | 山东大学 | Bearing degradation process state monitoring method, system, equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2102776C1 (en) * | 1995-06-19 | 1998-01-20 | Юрий Петрович Кострыгин | Vibroseismic survey method |
| CN105675320A (en) * | 2016-01-06 | 2016-06-15 | 山东大学 | Method for real time monitoring mechanical system operation status on the basis of acoustic signal analysis |
| CN106599271A (en) * | 2016-12-22 | 2017-04-26 | 江苏方天电力技术有限公司 | Emission monitoring time series data abnormal value detection method for coal-fired unit |
| CN107610464A (en) * | 2017-08-11 | 2018-01-19 | 河海大学 | A kind of trajectory predictions method based on Gaussian Mixture time series models |
-
2018
- 2018-06-08 CN CN201810586819.9A patent/CN108955855B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2102776C1 (en) * | 1995-06-19 | 1998-01-20 | Юрий Петрович Кострыгин | Vibroseismic survey method |
| CN105675320A (en) * | 2016-01-06 | 2016-06-15 | 山东大学 | Method for real time monitoring mechanical system operation status on the basis of acoustic signal analysis |
| CN106599271A (en) * | 2016-12-22 | 2017-04-26 | 江苏方天电力技术有限公司 | Emission monitoring time series data abnormal value detection method for coal-fired unit |
| CN107610464A (en) * | 2017-08-11 | 2018-01-19 | 河海大学 | A kind of trajectory predictions method based on Gaussian Mixture time series models |
Non-Patent Citations (1)
| Title |
|---|
| HOJAT HEIDARI BAFROUI: "Application of wavelet energy and Shannon entropy for feature", 《NEUROCOMPUTING》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109655267A (en) * | 2019-01-07 | 2019-04-19 | 江苏大学 | A kind of automobile hub bearing fault signature extracting method |
| CN111929044A (en) * | 2020-07-15 | 2020-11-13 | 西门子工厂自动化工程有限公司 | Method, apparatus, computing device and storage medium for monitoring device status |
| CN111929044B (en) * | 2020-07-15 | 2023-08-08 | 西门子工厂自动化工程有限公司 | Method, apparatus, computing device and storage medium for monitoring device status |
| CN111985380A (en) * | 2020-08-13 | 2020-11-24 | 山东大学 | Bearing degradation process state monitoring method, system, equipment and storage medium |
| CN111985380B (en) * | 2020-08-13 | 2022-07-05 | 山东大学 | Bearing degradation process state monitoring method, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108955855B (en) | 2020-05-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Metric-based meta-learning model for few-shot fault diagnosis under multiple limited data conditions | |
| Zheng et al. | Generalized composite multiscale permutation entropy and Laplacian score based rolling bearing fault diagnosis | |
| Wang et al. | A novel weighted sparse representation classification strategy based on dictionary learning for rotating machinery | |
| Zheng et al. | Generalized refined composite multiscale fuzzy entropy and multi-cluster feature selection based intelligent fault diagnosis of rolling bearing | |
| Zeng et al. | Maximum margin classification based on flexible convex hulls for fault diagnosis of roller bearings | |
| CN108955855A (en) | Vibration signal characteristics extracting method, monitoring method and device towards rotating machinery | |
| Nayak et al. | Study of distance metrics on k-nearest neighbor algorithm for star categorization | |
| CN111122162B (en) | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy | |
| Dong et al. | Intelligent fault diagnosis of rolling bearings based on refined composite multi-scale dispersion q-complexity and adaptive whale algorithm-extreme learning machine | |
| CN109584232A (en) | Equipment use state on-line monitoring method, system and terminal based on image recognition | |
| Li et al. | Refined composite variable-step multiscale multimapping dispersion entropy: A nonlinear dynamical index | |
| Dai et al. | Complexity–entropy causality plane based on power spectral entropy for complex time series | |
| Dhandapani et al. | A novel bearing faults detection method using generalized Gaussian distribution refined composite multiscale dispersion entropy | |
| Kyrkou et al. | Imaging time-series for NILM | |
| Ma et al. | Underdetermined blind source separation based on source number estimation and improved sparse component analysis | |
| CN116467141A (en) | Log recognition model training, log clustering method, related system and equipment | |
| Yang et al. | Change detection in rotational speed of industrial machinery using Bag-of-Words based feature extraction from vibration signals | |
| Akcan et al. | Diagnosing bearing fault location, size, and rotational speed with entropy variables using extreme learning machine | |
| Li et al. | Two-dimensional composite multi-scale time–frequency reverse dispersion entropy-based fault diagnosis for rolling bearing | |
| CN117606801A (en) | Cross-domain bearing fault diagnosis method based on multi-characterization self-adaptive network | |
| WO2001035268A1 (en) | Improved spectrum analysis method and apparatus | |
| CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion | |
| CN110096708A (en) | A kind of determining method and device of calibration collection | |
| Mao et al. | Bearing fault diagnosis based on multi-task learning | |
| JP2019133462A (en) | Detection program, detection method and detection device |
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 |