CN107525674B - Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving - Google Patents
Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving Download PDFInfo
- Publication number
- CN107525674B CN107525674B CN201710392666.XA CN201710392666A CN107525674B CN 107525674 B CN107525674 B CN 107525674B CN 201710392666 A CN201710392666 A CN 201710392666A CN 107525674 B CN107525674 B CN 107525674B
- Authority
- CN
- China
- Prior art keywords
- frequency
- crestal line
- signal
- probability distribution
- turn
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
Frequency estimation method and detection device are turned based on crestal line probability distribution and the instantaneous of local fluctuation characteristic the invention discloses a kind of, method from vibration signal the following steps are included: isolate low frequency region, and resonance band is identified using quickly spectrum kurtosis method, to realize band separation, and signal low frequency region and resonance band time-frequency distributions feature are enhanced using amplitude cumulative square algorithm；It searches for signal low frequency region respectively using peak search algorithm and turns synchronous crestal line frequently with resonance band, pre-estimation goes out to turn frequency information, to resonance band crestal line synchronization process；Using the abnormal crestal line deciding field method based on probability distribution, to determine crestal line invalid position, improve the result merged to the crestal line that low-frequency range and resonance envelope signal are extracted；It establishes and is based on localised waving characteristic exception section fusion criterion, introduce standard deviation and indicate crestal line fluctuation characteristic as statistical indicator, merge index as abnormal data section to evaluate fusion results, realize and turn the accurate estimation of frequency.
Description
Technical field
The invention belongs to the technical fields of variable speed rolling bearing fault diagnosis, more particularly to a kind of crestal line probability that is based on to divide
Cloth and the instantaneous of local fluctuation characteristic turn frequency estimation method and detection device.
Background technique
Rolling bearing is that most common spare part in common use is also one of the components being most easily lost in all kinds of rotating machineries.Rotation
The failure about 30% of favourable turn tool is due to caused by bearing fault.Therefore, the health status of bearing is detected extremely heavy
It wants.Vibration of the bearing under failure excitation often exists in the form of transient state characteristic, it is effectively extracted and studied can
Accurate evaluation bearing operation conditions is bearing failure diagnosis key.In actual motion environment, the variable working condition such as speed change, varying load
The often normality of bearing operation.Therefore, carry out rolling bearing transient state characteristic under variable speed operating condition and extract research with practical meaning
It is also extremely challenging while adopted.
Order tracking technique is one of variable working condition equipment fault diagnosis common analysis.Hardware order tracking technique and calculate order with
Track is two kinds traditional based on tachometer Order Tracking, but the cost of revolving speed acquisition device and installation restrict these methods
Use scope.Many scholars are based on signal time frequency analysis and propose relevant transient speed recognition methods in recent years.Such as Hunan
Peng Fuqiang of university etc. turns frequency using the estimation of line frequency modulation small echo path tracing algorithm is instantaneous；Wang of Beijing University of Chemical Technology etc. is used
Wigner-Will transformation carries out time frequency analysis to bearing vibration signal, is then based on opposite mutual information principle to fault signature frequency
Spectrum carries out extracted in self-adaptive；Shi of Ottawa university etc. gradually demodulates transformation using broad sense and increases time-frequency with synchronous extruding algorithm
Aggregation improves the extraction accuracy of instantaneous frequency；Jacek Urbanek of AGH university etc. is coarse by time-frequency distributions first
Estimation turns frequently, and then counterweight sampling bandpass filtered signal instead samples the accurate instantaneous turn of frequency of acquisition.The above research is extraction variable-speed motor
Tool equipment rotary speed information provides new approach.It should be pointed out that although letter can be improved in some new Time-Frequency Analysis Methods
The readability of number time-frequency representation, but the problems such as it is complicated that algorithm can be brought to calculate, and timeliness is not good enough.And Short Time Fourier Transform,
Although the Time-Frequency Analysis Methods such as wavelet transformation can simply and rapidly obtain the time-frequency distributions feature of signal, because of Heisenberg
The reasons such as the uncertain principle of lattice and target time-frequency crestal line energy are faint, the time-frequency representation for causing this kind of Time-Frequency Analysis Method to obtain
Aggregation difference and peak value searching scheduling algorithm in time-frequency representation result to turn frequency information extraction not accurate enough.
Summary of the invention
In view of simple, the rapid charater of STFT analysis method, it is suitable for Engineering Signal and analyzes, mentioned to overcome based on STFT
The defect of rotary speed information method is taken, the present invention will establish the crestal line fusion criterion based on probability distribution and local fluctuation characteristic, mention
It is a kind of out that frequency estimation method is instantaneously turned based on crestal line probability distribution and local fluctuation characteristic, with realize fast and accurately from
STFT, which is analyzed, obtains equipment revolving speed in result.
In order to achieve the above object, the present invention provides a kind of technical solutions: one kind being based on crestal line probability distribution and part
Fluctuation characteristic instantaneously turns frequency estimation method, includes the following steps:
Step 1: low frequency region is isolated from vibration signal, and resonance band is identified using quickly spectrum kurtosis method,
To realize band separation, and square algorithm is added up respectively to the time-frequency distributions of signal low frequency region and resonance band using amplitude
Feature is enhanced；
Step 2: searching for signal low frequency region respectively using peak search algorithm with resonance band and turn synchronous crestal line frequently, respectively
Pre-estimation goes out to turn frequency information, and to resonance band crestal line synchronization process；
Step 3: being changed using the abnormal crestal line deciding field method based on probability distribution with determining the invalid position of crestal line
The result that the kind crestal line extracted to low-frequency range and resonance envelope signal is merged；
Step 4: establishing the abnormal section fusion criterion based on localised waving characteristic, introduce standard deviation and come as statistical indicator
It indicates crestal line fluctuation characteristic, merges index as abnormal data section to evaluate fusion results, realize the accurate estimation for turning frequency.
Further, step 1 specifically includes the following steps:
Step 1.1: to vibration signal x (t) (t ∈ [0, tn]) band separation is carried out, low-frequency range selection range is [0, f0],
f0500Hz is taken, conventional rotary shaft can be covered and turn frequency and its frequency multiplication information x1(t)；
Step 1.2: resonance band takes [fa, fb].Selection for resonance band can theoretically be joined by bearing arrangement
Number is calculated, however operation needs certain priori knowledge in this way, and difficulty is larger, here, introducing classical quick spectrum kurtosis
Method accurately and efficiently identifies resonance band signal x2(t)；
Step 1.3: low-band signal x1(t) STFT can be expressed as
In formula (1), τ indicate time shift, ω indicate frequency, h (t) is that center is located at τ=0, be highly 1, limited width when
Window function observes signal x by h (t)1(t) part is x1(t) h (t), h*(t-τ)ejωtIt is the basic function of STFT,It is the result of STFT；
It similarly can be in the hope of resonance band signal x2(t) STFT is expressed as
Step 1.4: it is assumed that signal x1It (t) is a cosine signal, if x1(t)=Acos (2 π f0T), then formula (1) can table
It is shown as
In formula (2), A indicates amplitude, f0Indicate that frequency, formula (2) show that in STFT time-frequency distributions, energy concentrates on ω
=2 π f0Frequency band on, then, for a harmonic signalIts time-frequency distributions is represented by
In formula (3), M is harmonic component number, signal x1(t) energy concentrates on ω1=2 π f0, ω2=4 π f0..., ωM
=2M π f0At equal Frequency points,
The amplitude of k-th of the component harmonic signal squared results that add up are
In formula (4), k ∈ [1, M],To carry out cumulative square of amplitude treated time frequency point amplitude to time-frequency distributions,(τ on k-th of component harmonic signalm, ωn) the enhanced amplitude of time frequency point is expressed as
Thus low-band signal x is obtained1(t) result is after cumulative square of time-frequency enhancing of amplitude
Similarly available resonance band signal x2(t) the cumulative enhanced result of square time-frequency of amplitude
Further, step 2 specifically includes the following steps:
Step 2.1: low-frequency range being extracted using peak search algorithm search and turns the synchronous crestal line of frequency, peak search algorithm is specific
Are as follows:
In formula (6), τ indicates the time, and f indicates to turn frequency；P (τ, f) indicates parameter when energy is maximized, and Δ f is frequency
Search range, (τi, fi) indicate in τiThe instantaneous frequency f at momenti, τ0Indicate the time that peak value searching is carried out under global frequencies
Section, τiIndicate the time that peak value searching is carried out under local frequencies range；
But by bearing institute in systems there may be unbalanced shaft, misalign the problems such as, in time-frequency distributions with to turn frequency same
Step ground crestal line information occurs frequently now at double, wherein the maximum synchronous crestal line of the energy extracted is not necessarily and turns frequency curve, it is therefore, low
The crestal line of frequency extraction is possible to turn frequency or its frequency multiplication, when extraction low-frequency range turns frequency crestal line, known need to turn frequency place substantially
Range, to be corrected to result.So far the instantaneous frequency profile P of low-frequency range can be extractedx1(τ, f)；
The instantaneous frequency profile P of resonance band can similarly be extractedx2(τ, f).
Step 2.2: low-band signal is being melted with the synchronous crestal line information frequently that turns that resonance and demodulation envelope signal extracts
When conjunction, it is required that instantaneous frequency profile is consistent, therefore resonance band synchronizing of crestal line need to be handled: by resonance band
Crestal line Px2(τ, f) turns frequency calibration information P with low frequency regionx1Synchronizing is converted on the basis of (τ, f), is for result after conversion
In formula (7), k is to synchronize parameter.Parameter k can be determined according to registration R
The value of k meets R minimalization, and 0 < k < K, K generally takes 5~6 times of fault signature order.
Further, in step 2.1, τ is determined0There are two ways to period: first is that rule of thumb estimating；It is another
A is exactly that global search is first carried out within the entire period, at the beginning of finding no end effect influence, then this
Period before moment is as τ0。
Further, in step 2.1, the moment curve after peak value searching is discrete results, can use linear interpolation
Discrete curve is changed into full curve.
Further, step 3 specifically includes the following steps:
Step 3.1: result P is extracted to low-frequency range and resonance envelope signal crestal linex1(τ, f) withDifference is carried out,
Δ C (τ) is obtained,
In formula (9),WithRespectively in τiMoment Px1(τ, f) andFrequency values,
Theoretically if the crestal line extracted is effectively as a result, the value of so Δ C (τ) is 0, actually all due to two crestal lines
There are certain invalid position, the value of Δ C (τ) will appear nonzero value, therefore the position of abnormal crestal line is judged by Δ C (τ) value；
Step 3.2: there is exception in crestal line, can occur in the form of abnormal section, and abnormal section is to merge section P (τK,
τL), it is determined by a pair of of rising edge and failing edge, the rising edge τ in fusion section is defined according to the information of Δ C (τ)KAnd failing edge
τLPosition
τK=τn, if (Δ C (τn)≤t)&(ΔC(τn+1) > t) (10),
τL=τn, if (Δ C (τn) > t) & (Δ C (τn+1)≤t) (11),
In formula (10) and formula (11), t is the threshold value for determining abnormal area；
Step 3.3: come threshold value t by the way of probability distribution statistical,
T=Δ C (τ) s.t. max (pdf (Δ C (τ))) (12),
In formula (12), pdf (Δ C (τ)) indicates the probability distribution of Δ C (τ), and max (pdf (Δ C (τ))) indicates probability statistics
Maximum value, namely showing that threshold value t takes the crestal line difference frequency of probability statistics maximum value position is threshold value,
If blend curve time value τ=[0, τe], for special fusion section P (0, τl)、P(τk, τe) equally need to distinguish
Meet formula (10) and formula (11), abnormal data section is as shown in dash area in attached drawing 1.
Further, step 4 specifically includes the following steps:
Step 4.1: being directed to every crestal line, the standard deviation in abnormal crestal line section is smaller, then this crestal line is different
Result in normal section turns frequency estimated value closer to true, therefore turns frequency after abnormal crestal line section fusion and be estimated as P (τ, f)
Step 4.2: for non-abnormal data section, due to Px1(τ, f) withDifference is smaller, therefore taking mean value is to turn
Frequency estimation P (τ, f),
So far, instantaneously turn frequency estimated result after obtaining data fusion.
The present invention also provides another technical solutions: it is a kind of using above-mentioned based on crestal line probability distribution and localised waving
Feature instantaneously turns the detection device of frequency estimation method, it include main shaft, the first bearing being connected on main shaft and second bearing,
The first bearing seat that is connected with first bearing, the second bearing seat being connected with second bearing, driving spindle rotation motor,
One end be connected to the motor connect and the other end be connected to a computer and be used to adjust motor output revolving speed frequency converter, first axle
Hold and be equipped with pitting fault, the acceleration transducer for measuring vibration signal is installed in first bearing seat, acceleration transducer with
The input terminal of one data collector is connected, and the output end of the data collector is connected to a computer.
Further, at least one the quality disk for being provided with load is connected on main shaft, at least one quality disk is located at first
Between bearing and second bearing.
Further, acceleration transducer is mounted on the top of first bearing seat.
Further, main shaft is connected with the output shaft of motor by shaft coupling.
By using above-mentioned technical proposal, one kind proposed by the invention is based on crestal line probability distribution and local fluctuation characteristic
Instantaneously turn frequency estimation method and detection device, it is sharp first to the vibration signal of rolling bearing under collected variable speed operating condition
The time-frequency locality of signal low-frequency range and resonance band is improved with cumulative square of strategy of amplitude；Then to the low of mechanical fault signals
Frequency range, resonance band time-frequency crestal line are extracted and are synchronized to estimate to a turn frequency；Finally two class revolving speed correlations of extraction are believed
Breath is merged, and the abnormal section turned in frequency is estimated with correction, when carrying out crestal line fusion by crestal line probability distribution to exception
Fusion section is positioned, and is merged later by establishing the fusion criterion of localised waving feature to abnormal fusion section, real
Now turn the accurate acquisition of frequency.
Compared with the prior art, advantage possessed by the present invention is as follows:
1, the invention is therefrom to extract rotary speed information based on signal itself, avoids the installation of tachometer, solves not side
Just the occasion for installing sensor is difficult to the problem of obtaining rotary speed information, while decreasing the application cost of tachometer, therefore should
Invent the new direction that there is good engineering application value and variable speed bearing failure diagnosis to develop.
2, since Heisenberg does not know principle, the method for estimating rotating speed of time-frequency distributions is directly based upon often because time-frequency is assembled
Property it is poor so that using peak value searching scheduling algorithm to turning not accurate enough when frequency information extracts in time-frequency distributions result, should
The time-frequency distributions aggregation that the amplitude superposition square strategy that invention proposes can obtain is more excellent.
3, due to turning the energy of frequency information at any time and non-uniform Distribution, the frequency information that turns of energy lower period is easy to
It is submerged in strong noise background, in conjunction with amplitude, the superposition square strategy present invention establishes probability distribution and local fluctuation characteristic simultaneously
Fusion criterion, the two is merged, original crestal line recognition result is improved, improve turn frequency estimation an accuracy.
The above description is only an overview of the technical scheme of the present invention, in order to better understand the technical means of the present invention,
And can be implemented in accordance with the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention and the accompanying drawings.
Detailed description of the invention
Fig. 1 is the step 3.3 for instantaneously turning frequency estimation method the present invention is based on crestal line probability distribution and local fluctuation characteristic
Middle abnormal data interval diagram；
Fig. 2 is the detection device for instantaneously turning frequency estimation method the present invention is based on crestal line probability distribution and local fluctuation characteristic
Structural schematic diagram；
Fig. 3 is the detection device for instantaneously turning frequency estimation method the present invention is based on crestal line probability distribution and local fluctuation characteristic
Mechanical assembly structure schematic diagram；
Fig. 4 is the flow chart for instantaneously turning frequency estimation method the present invention is based on crestal line probability distribution and local fluctuation characteristic；
Fig. 5 a is experimental signal resonance band identification in the present embodiment: inner ring faulty bearings vibration signal；Fig. 5 b is inner ring
Faulty bearings vibration signals spectrograph；Fig. 5 c is SK algorithm decomposition result；
Fig. 6 a is experimental signal time-frequency representation enhancing in the present embodiment: low-frequency range time-frequency representation；Fig. 6 b is low-frequency range time-frequency
Feature enhancement results；Fig. 6 c is resonance band time-frequency representation；(d) resonance band time-frequency characteristics enhance result；
Fig. 7 a is the instantaneous frequency profile that low-frequency range is extracted in the present embodiment；Fig. 7 b is that resonance band is extracted in the present embodiment
Instantaneous frequency profile；Fig. 7 c is that instantaneous frequency synchronizes in the present embodiment；
Fig. 8 a is positioning fusion section in the present embodiment: Δ C (τ)；Fig. 8 b is Δ C (τ) amplitude probability point in the present embodiment
Cloth result；Fig. 8 c is abnormal fusion compartmental results in the present embodiment；
Fig. 9 is to turn frequency estimated result based on time-frequency characteristics enhancing and the instantaneous of information fusion in the present embodiment.
Figure label are as follows:
1, first bearing；2, second bearing；3, first bearing seat；4, second bearing seat；5, main shaft；6, motor；7, accelerate
Spend sensor；8, shaft coupling；10, data collector；9, frequency converter；11, computer；12, quality disk.
Specific embodiment
With reference to the accompanying drawings and examples, specific embodiments of the present invention will be described in further detail.Implement below
Example is not intended to limit the scope of the invention for illustrating the present invention.
Referring to attached drawing 2 and attached drawing 3, frequency is turned based on crestal line probability distribution and the instantaneous of local fluctuation characteristic in the present embodiment
The detection device of estimation method, it includes main shaft 5, the first bearing being connected on main shaft 51 and second bearing 2 and first bearing
1 be connected first bearing seat 3, be connected with second bearing 2 second bearing seat 4, driving spindle 5 rotate motor 6, one end
It is connected with motor 6 and the other end is connected with computer 11 and the frequency converter 9 of the output revolving speed for adjusting motor 6, first
Bearing 1 is equipped with pitting fault, is equipped with the acceleration transducer 7 for measuring vibration signal, acceleration sensing in first bearing seat 3
Device 7 is connected with the input terminal of a data collector 10, and the output end of the data collector 10 is connected with computer 11.
Connection is provided at least one quality disk 12 of load on main shaft 5, at least one quality disk 12 is located at first bearing
Between second bearing.In the present embodiment, there are two quality disks 12 for connection.
Main shaft 5 is connected with the output shaft of motor 6 by shaft coupling 8.In the present embodiment, motor 6 use SIEMENS, 3
~, 2.0HP.
Since the signal-to-noise ratio that acceleration transducer 7 is mounted on the collected signal of different location can be different, the present embodiment
In in order to obtain the higher data of signal-to-noise ratio, acceleration transducer 7 is mounted on to the top of first bearing seat 3.Acceleration sensing
Device 7 uses PCB ICP 353C03.
The variation of 9 frequency of frequency converter is controlled by 11 pre-set programs of computer, the frequency for adjusting frequency converter 9 can be adjusted
The output revolving speed of motor 6, to realize the vibration signal under acquisition different rotating speeds or variable speed.
Data collector 10 receives the data that are exported of acceleration transducer 7, and data are pre-processed (including A/D
Conversion, rectification, amplification, filtering etc.), the electric signal of acquisition is then sent to computer 11 and is recorded, shown and further located
Reason.In the present embodiment, data collector 10 uses NIcDAQ-9234.
First bearing 1 and second bearing 2 are all made of two-row ball bearing.1207 EKTN9/C3 of model SKF, every row's ball
Number Z=15, rolling element diameter d=8.7mm, contact angle α=0 °, bearing pitch diameter D=53.5mm.By spark technology
It is provided with the pitting fault that diameter is 0.9mm on the inner ring of one bearing 1, sample frequency fs=25.6kHz, sampling time t are set
=10s, axis turn frequency fc and fluctuate in 15-25Hz range.
Referring to attached drawing 4, one of the present embodiment is estimated based on the instantaneous frequency that turns of crestal line probability distribution and local fluctuation characteristic
Meter method, includes the following steps:
Step 1: low frequency region is isolated from vibration signal, and resonance band is identified using quickly spectrum kurtosis method,
To realize band separation.And square algorithm is added up respectively to the time-frequency distributions of signal low frequency region and resonance band using amplitude
Feature is enhanced.
Step 2: searching for signal low frequency region respectively using peak search algorithm with resonance band and turn synchronous crestal line frequently, respectively
Pre-estimation goes out to turn frequency information, and to resonance band crestal line synchronization process.
Step 3: being changed using the abnormal crestal line deciding field method based on probability distribution with determining the invalid position of crestal line
The result that the kind crestal line extracted to low-frequency range and resonance envelope signal is merged.
Step 4: establishing the abnormal section fusion criterion based on localised waving characteristic, introduce standard deviation and come as statistical indicator
It indicates crestal line fluctuation characteristic, merges index as abnormal data section to evaluate fusion results, realize the accurate estimation for turning frequency.
In a kind of highly preferred embodiment, step 1 specifically includes the following steps:
Step 1.1: the bearing vibration signal of acquisition inner ring failure is as shown in Figure 5 a, the frequency spectrum of corresponding signal such as Fig. 5 b institute
Show, low-band signal is obtained by low-pass filtering, filtering frequency range is [0,500Hz].
Step 1.2: resonance band use is quickly composed high and steep algorithm and is extracted.The is concentrated on by resonance band known to Fig. 5 c
7.6 layers [15600,16000Hz] section, to the resonance band envelope demodulation.
Step 1.3: low-band signal x is acquired by formula (1)1(t) and resonance band signal x2(t) STFT resultWithThe STFT time-frequency representation result of low-frequency range and resonance band is as shown in figs 6 a and 6 c.
Step 1.4: low-band signal x is acquired by formula (2) to formula (5)1(t) and resonance band signal x2(t) amplitude is tired
Result after adding square time-frequency to enhanceWithResult such as Fig. 6 b and Fig. 6 d institute after cumulative square of enhancing of amplitude
Show, it can be seen that the noise contribution after time-frequency enhancing in signal has obtained effective inhibition.
In a kind of highly preferred embodiment, step 2 specifically includes the following steps:
Step 2.1: instantaneous frequency distilling is carried out to time-frequency spectrum shown in Fig. 6 b and Fig. 6 d using peak search algorithm.According to
Approximate range where turning frequency is corrected the crestal line result of extraction, finally obtains the synchronous letter of turn frequency as shown in figs. 7 a and 7b
Cease Px1、Px2.There is extraordinary wave by the frequency estimation results that turn that low-frequency range and resonance envelope signal obtain in circle instruction in Fig. 7 a
It is dynamic, it is primarily due to due to noise jamming, and the energy of corresponding time-frequency crestal line is faint, it is accurate to effective crestal line information to cause to fail
It extracts.
Step 2.2: using to Px2Synchronizing processing, obtains synchronization crestal line as shown in Figure 7 c
Further, the step 3 specifically includes the following steps:
Step 3.1: obtaining low-frequency range as shown in Figure 8 a using formula (9) and the difference DELTA of the crestal line for envelope signal extraction of resonating
C(τ).The probability distribution result of Δ C (τ) is as shown in Figure 8 b.
Step 3.2 is to step 3.3: obtaining threshold value t=0.2081 according to formula (12).Using formula (10) and formula (11) obtain as
Abnormal fusion section shown in Fig. 8 c, wherein the abnormal interval Δ C (τ) for the label irised out is much larger than threshold value t, the i.e. corresponding circle of sensation
Between there is significant abnormal estimation condition.
Further, the step 4 specifically includes the following steps:
Step 4.1: abnormal fusion section being merged using fusion criterion shown in formula (13) and formula (14).
Step 4.2: for non-abnormal data section, due to Px1WithDifference is smaller, therefore taking mean value is to turn frequency to estimateIt obtains turning to turn frequently as shown in Figure 9 to estimate frequently.Compare P and Px1、It can be found that turning frequency estimation results
Px1、P can be effectively corrected after carrying out crestal line fusionx1、Middle exception turns frequency wave phenomenon, and what is merged turns frequency estimation more
Close to legitimate reading.
A kind of instantaneous turn of frequency estimation side based on crestal line probability distribution and local fluctuation characteristic proposed through the invention
Method and detection device, to the vibration signal of rolling bearing under collected variable speed operating condition, first with the cumulative square plan of amplitude
Slightly improve the time-frequency locality of signal low-frequency range and resonance band；When then to the low-frequency range of mechanical fault signals, resonance band
Frequency crestal line is extracted and is synchronized to estimate to a turn frequency；Finally two class revolving speed relevant informations of extraction are merged, with school
The abnormal section turned in frequency is just being estimated, abnormal fusion section is being determined by crestal line probability distribution when carrying out crestal line fusion
Position, later merges abnormal fusion section by establishing the fusion criterion of localised waving feature, realizes and turns accurately obtaining for frequency
It takes.
The above is only a preferred embodiment of the present invention, it is not intended to restrict the invention, it is noted that for this skill
For the those of ordinary skill in art field, without departing from the technical principles of the invention, can also make it is several improvement and
Modification, these improvements and modifications also should be regarded as protection scope of the present invention.
Claims (11)
1. a kind of turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic, it is characterised in that: including such as
Lower step:
Step 1: low frequency region is isolated from vibration signal, and resonance band is identified using quickly spectrum kurtosis method, thus
Realize band separation, and using amplitude add up square algorithm respectively to the time-frequency distributions feature of signal low-frequency range and resonance band into
Row enhancing；
Step 2: searching for signal low-frequency range respectively using peak search algorithm with resonance band and turn synchronous crestal line frequently, respectively pre-estimation
Turn frequency information out, and to resonance band crestal line synchronization process；
Step 3: the abnormal crestal line deciding field method based on probability distribution is utilized, to determine the invalid position of crestal line, improvement pair
The result that the crestal line that low-frequency range and resonance envelope signal extract is merged；
Step 4: establishing the abnormal section fusion criterion based on localised waving characteristic, introduce standard deviation and indicated as statistical indicator
Crestal line fluctuation characteristic merges index as abnormal data section to evaluate fusion results, realizes the accurate estimation for turning frequency.
2. according to claim 1 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic,
Be characterized in that: the step 1 includes:
Step 1.1: to vibration signal x (t) (t ∈ [0, tn]) band separation is carried out, low-frequency range selection range is [0, f0], f0It takes
500Hz；
Step 1.2: resonance band takes [fa,fb], it introduces and composes kurtosis method quickly to identify resonance band signal x2(t)；
Step 1.3: low-band signal x1(t) STFT can be expressed as
In formula (1), τ indicates time shift, and ω indicates frequency, and it is highly 1, the when window letter of limited width that h (t), which is that center is located at τ=0,
Number, observes signal x by h (t)1(t) part is x1(t) h (t), h*(t-τ)ejωtIt is the basic function of STFT,
STFT as a result,
It similarly can be in the hope of resonance band signal x2(t) STFT is expressed as
Step 1.4: it is assumed that signal x1It (t) is a cosine signal, if x1(t)=Acos (2 π f0T), then formula (1) is represented by
In formula (2), A indicates amplitude, f0Indicate that frequency, formula (2) show that in STFT time-frequency distributions, energy concentrates on the π of ω=2 f0
Frequency band on, then, for a harmonic signalIts time-frequency distributions is represented by
In formula (3), M is harmonic component number, signal x1(t) energy concentrates on ω1=2 π f0, ω2=4 π f0..., ωM=2M
πf0At equal Frequency points,
The amplitude of k-th of the component harmonic signal squared results that add up are
In formula (4), k ∈ [1, M],To carry out cumulative square of amplitude treated time frequency point amplitude to time-frequency distributions,(τ on k-th of component harmonic signalm,ωn) the enhanced amplitude of time frequency point is expressed as
Thus low-band signal x is obtained1(t) result is after cumulative square of time-frequency enhancing of amplitudeSimilarly obtain resonance frequency
Segment signal x2(t) the cumulative enhanced result of square time-frequency of amplitude
3. according to claim 1 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic,
Be characterized in that: the step 2 includes:
Step 2.1: low-frequency range being extracted using peak search algorithm search and turns the synchronous crestal line of frequency, the peak search algorithm are as follows:
In formula (6), τ indicates the time, and f indicates to turn frequency；P (τ, f) indicates parameter when energy is maximized, and △ f is frequency search
Range, (τi,fi) indicate in τiThe instantaneous frequency f at momenti, τ0Indicate the period that peak value searching is carried out under global frequencies, τi
Indicate the time that peak value searching is carried out under local frequencies range,
So far the instantaneous frequency profile P of low-frequency range can be extractedx1(τ,f)；
The instantaneous frequency profile P of resonance band can similarly be extractedx2(τ,f)；
Step 2.2: to the processing of resonance band synchronizing of crestal line: by resonance band crestal line Px2(τ, f) turns frequency with low frequency region
Calibration information Px1Synchronizing is converted on the basis of (τ, f), is for result after conversion
In formula (7), k is to synchronize parameter, and parameter k can be determined according to registration R:
The value of k meets R minimalization, and 0 < k < K, K generally take 5~6 times of fault signature order.
4. according to claim 3 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic,
It is characterized in that: in the step 2.1, determining τ0There are two ways to period: first is that rule of thumb estimating；The other is first
Global search is carried out within the entire period, at the beginning of finding no end effect influence, then before this moment
Period is as τ0。
5. according to claim 3 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic,
Be characterized in that: in the step 2.1, the moment curve after peak value searching is discrete results, using linear interpolation by discrete curve
It is changed into full curve.
6. according to claim 1 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic,
Be characterized in that: the step 3 includes:
Step 3.1: result P is extracted to low-frequency range and resonance envelope signal crestal linex1(τ, f) withDifference is carried out, △ is obtained
C (τ),
In formula (9),WithRespectively in τiMoment Px1(τ, f) andFrequency values,
The position of abnormal crestal line is judged by △ C (τ) value；
Step 3.2: there is exception in crestal line, can occur in the form of abnormal section, and abnormal section is to merge section P (τK,τL),
It is determined by a pair of of rising edge and failing edge, the rising edge τ in fusion section is defined according to the information of △ C (τ)KWith failing edge τLPosition
It sets:
τK=τn,if(△C(τn)≤t)&(△C(τn+1) > t) (10),
τL=τn,if(△C(τn)>t)&(△C(τn+1)≤t) (11),
In formula (10) and formula (11), t is the threshold value for determining abnormal area；
Step 3.3: come threshold value t by the way of probability distribution statistical,
T=△ C (τ) s.t.max (pdf (△ C (τ))) (12),
In formula (12), pdf (△ C (τ)) indicates the probability distribution of △ C (τ), and max (pdf (△ C (τ))) indicates that probability statistics are maximum
Value, namely showing that threshold value t takes the crestal line difference frequency of probability statistics maximum value position is threshold value,
If blend curve time value τ=[0, τe], for special fusion section P (0, τl)、P(τk,τe) need to equally meet respectively
Formula (10) and formula (11).
7. a kind of instantaneous turn of frequency estimation side based on crestal line probability distribution and local fluctuation characteristic according to claim 1
Method, it is characterised in that: the step 4 includes:
Step 4.1: being directed to every crestal line, the standard deviation in abnormal crestal line section is smaller, then this crestal line is in exceptions area
Interior result turns frequency estimated value closer to true, therefore a turn frequency is estimated as P (τ, f) after abnormal crestal line section fusion:
Step 4.2: for non-abnormal data section, due to Px1(τ, f) withDifference is smaller, therefore taking mean value is to turn frequency to estimate
It counts P (τ, f):
So far, instantaneously turn frequency estimated result after obtaining data fusion.
8. a kind of turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic using described in claim 1
Detection device, it is characterised in that: it include main shaft, the first bearing being connected on the main shaft and second bearing, with it is described
First bearing seat that first bearing is connected, the second bearing seat being connected with the second bearing, the driving main axis
Motor, one end is connected with the motor and the other end is connected to a computer and is used to adjust the output revolving speed of the motor
Frequency converter, the first bearing is equipped with pitting fault, is equipped in the first bearing seat and measures vibration signal
Acceleration transducer, the acceleration transducer are connected with the input terminal of a data collector, the data collector it is defeated
Outlet is connected with the computer.
9. the inspection according to claim 8 that instantaneously turn frequency estimation method based on crestal line probability distribution and local fluctuation characteristic
Survey device, it is characterised in that: connection is provided at least one quality disk of load, at least one described matter on the main shaft
Disk is measured between first bearing and second bearing.
10. according to claim 8 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic
Detection device, it is characterised in that: the acceleration transducer is mounted on the top of the first bearing seat.
11. according to claim 8 turn frequency estimation method based on crestal line probability distribution and the instantaneous of local fluctuation characteristic
Detection device, it is characterised in that: the main shaft is connected with the output shaft of motor by shaft coupling.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710392666.XA CN107525674B (en) | 2017-05-27 | 2017-05-27 | Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710392666.XA CN107525674B (en) | 2017-05-27 | 2017-05-27 | Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107525674A CN107525674A (en) | 2017-12-29 |
| CN107525674B true CN107525674B (en) | 2019-06-28 |
Family
ID=60748676
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710392666.XA Active CN107525674B (en) | 2017-05-27 | 2017-05-27 | Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107525674B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388839A (en) * | 2018-01-26 | 2018-08-10 | 电子科技大学 | A kind of strong fluctuation of speed feature extracting method based on second order sync extraction transformation |
| CN109488630B (en) * | 2018-11-13 | 2023-07-14 | 上海金艺检测技术有限公司 | Centrifugal fan rotor misalignment fault diagnosis method based on harmonic relative index |
| CN109668733B (en) * | 2018-12-21 | 2020-07-28 | 苏州大学 | Variable-speed bearing fault diagnosis method based on decomposition of variational nonlinear mode |
| CN110174270A (en) * | 2019-06-10 | 2019-08-27 | 苏州大学 | Multi-source time-frequency crestal line extracting method |
| CN110617964A (en) * | 2019-07-29 | 2019-12-27 | 中国铁道科学研究院集团有限公司城市轨道交通中心 | Synchronous compression transformation order ratio analysis method for fault diagnosis of rolling bearing |
| CN110907162B (en) * | 2019-12-13 | 2021-10-15 | 北京天泽智云科技有限公司 | Rotating machinery fault feature extraction method without tachometer under variable rotating speed |
| CN111458122B (en) * | 2020-04-08 | 2022-03-29 | 苏州大学 | Rotary machine fault diagnosis method based on matching enhancement time-frequency representation |
| CN111707528B (en) * | 2020-07-16 | 2023-05-26 | 上海宇航系统工程研究所 | Kinetic environment time-frequency conversion method based on accumulated damage equivalence |
| CN111879508B (en) * | 2020-07-28 | 2022-06-10 | 无锡迈斯德智能测控技术有限公司 | Method and device for estimating instantaneous rotating speed of rotating machine based on time-frequency transformation and storage medium |
| CN112362343A (en) * | 2020-10-28 | 2021-02-12 | 华南理工大学 | Distributed fault feature extraction method for gearbox under variable rotating speed based on frequency modulation dictionary |
| CN114492539B (en) * | 2022-02-21 | 2023-04-28 | 西南交通大学 | Bearing fault detection method and device, electronic equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101666677B (en) * | 2009-09-25 | 2011-05-11 | 北京工业大学 | Method for extracting feature information of weak faults of low-speed heavy-duty equipment |
| CN102607845B (en) * | 2012-03-05 | 2014-09-10 | 北京工大智源科技发展有限公司 | Bearing fault characteristic extracting method for redundantly lifting wavelet transform based on self-adaptive fitting |
| CN105241666B (en) * | 2015-09-21 | 2017-10-20 | 华南理工大学 | A kind of Rolling Bearing Fault Character extracting method theoretical based on sparse signal representation |
| CN105547698B (en) * | 2015-12-31 | 2017-11-14 | 新疆金风科技股份有限公司 | The method for diagnosing faults and device of rolling bearing |
-
2017
- 2017-05-27 CN CN201710392666.XA patent/CN107525674B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107525674A (en) | 2017-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107525674B (en) | Turn frequency estimation method and detection device based on crestal line probability distribution and localised waving | |
| CN106769033B (en) | Variable speed rolling bearing fault recognition methods based on order envelope time-frequency energy spectrum | |
| CN108195587B (en) | Motor rolling bearing fault diagnosis method and system | |
| CN107941510B (en) | Extracting method based on the angularly Rolling Bearing Fault Character of dual sampling | |
| CN104655380B (en) | A kind of rotating machinery fault signature extracting method | |
| CN105784366A (en) | Wind turbine generator bearing fault diagnosis method under variable speed | |
| Saidi et al. | The use of SESK as a trend parameter for localized bearing fault diagnosis in induction machines | |
| CN109855874B (en) | Random resonance filter for enhancing detection of weak signals in vibration assisted by sound | |
| Wang et al. | Bearing fault diagnosis of direct-drive wind turbines using multiscale filtering spectrum | |
| CN110763462B (en) | Time-varying vibration signal fault diagnosis method based on synchronous compression operator | |
| CN109946081A (en) | A kind of method for diagnosing faults under variable speed when rolling bearing skidding | |
| CN105388012A (en) | Order tracking method based on nonlinear frequency modulation wavelet transformation | |
| CN102353500B (en) | Extraction method of unbalanced signal for dynamic balance measurement | |
| WO2020057066A1 (en) | Fault diagnosis method employing enhanced modulation signal bispectrum analysis for rolling bearing | |
| CN108444704A (en) | A kind of Fault Diagnosis of Rolling Element Bearings method | |
| Liu et al. | An online bearing fault diagnosis technique via improved demodulation spectrum analysis under variable speed conditions | |
| Lin et al. | A review and strategy for the diagnosis of speed-varying machinery | |
| Zhang et al. | Doppler correction using short-time MUSIC and angle interpolation resampling for wayside acoustic defective bearing diagnosis | |
| CN108120598A (en) | Square phase-couple and the bearing incipient fault detection method for improving bispectrum algorithm | |
| CN105738061A (en) | Image analysis method of vibration signal | |
| Ding et al. | Multiple instantaneous frequency ridge based integration strategy for bearing fault diagnosis under variable speed operations | |
| CN106198008B (en) | A kind of motor bearings outer ring fault identification method | |
| Shi et al. | A dual-guided adaptive decomposition method of fault information and fault sensitivity for multi-component fault diagnosis under varying speeds | |
| CN114486263B (en) | Noise reduction demodulation method for vibration signal of rolling bearing of rotary machine | |
| CN116380451A (en) | Variable-rotation-speed gearbox fault diagnosis method based on multi-harmonic time-frequency ridge line enhancement extraction |
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 |