CN105784366A - Wind turbine generator bearing fault diagnosis method under variable speed - Google Patents

Abstract

The invention discloses a wind turbine generator bearing fault diagnosis method under the variable speed. According to the method, a rotation angle change curve is drafted according to the bearing rotation speed; equal angle division for the rotation angle change curve is carried out, and an equal angle re-sampling time sequence is determined; interpolation for a bearing vibration signal is carried out according to the time sequence, a random Gauss white noise sequence is added to an angle domain vibration signal, and the signal added with the white noise is processed by utilizing an empirical mode decomposition (EMD) algorithm to acquire multiple sets of IMF; a kurtosis value of each IMF component is calculated; the IMF with the largest kurtosis value is selected and taken as a sensitive IMF; Hilbert envelope demodulation for the sensitive IMF is carried out to obtain an envelope signal, the envelope signal is processed by utilizing Fourier transform to obtain order envelope spectrum of the sensitive IMF, the fault characteristic frequency is extracted, and fault diagnosis on the wind turbine generator bearing is realized. The method is advantaged in that influence of rotation speed change on vibration signal analysis can be eliminated, and accuracy and validity of fault diagnosis can be improved.

Description

A kind of Wind turbines Method for Bearing Fault Diagnosis under variable speed

Technical field

The invention belongs to mechanical equipment fault detection technique field, relate to the Wind turbines Method for Bearing Fault Diagnosis under a kind of variable speed.

Background technology

Owing to wind field is predominantly located in the area that circumstance complication is severe, often by the impact of extreme weather.Along with the increase of unit accumulated running time, unit parts are constantly aging, very easily break down.Many positions such as main shaft on Wind turbines, driftage, change oar, electromotor, gear-box are equipped with bearing, and bearing fault occupies significantly high ratio in unit fault.In order to reduce the downtime of Wind turbines, reduce the maintenance cost of unit, the important parts of bearings of Wind turbines is carried out status monitoring necessary.Owing to the service condition of Wind turbines bearing is severe, stress is complicated, and is often subject to the impact of alternate load, shock loading so that bearing becomes one of common parts easily broken down of Wind turbines.Current analysis of vibration signal technology is the Main Means of rotating machinery condition monitoring and fault diagnosis, but there is significant difference in Wind turbines and other large rotating machinery equipment, blower fan is change at running medium speed, its vibration signal often shows the characteristic of non-stationary so that traditional method for diagnosing faults is difficult to the accurate differentiation of malfunction when variable speed.

Owing to wind generator system becomes increasingly complex, the parts part comprised also gets more and more so that its vibration signal frequency content is sufficiently complex, is difficult to detect its fault characteristic information.Wind turbines vibration signal in variable speed process is generally non-stationary signal, traditional frequency-domain analysis method is merely able to process stationary signal, but it is the bearing vibration signal belonging to non-stationary signal is then almost helpless, it is impossible to effectively to diagnose the fault that actual Wind turbines occurs.

Summary of the invention

It is an object of the invention to provide the Wind turbines Method for Bearing Fault Diagnosis under a kind of variable speed, solve blower fan to change its vibration signal at running medium speed and often show the characteristic of non-stationary so that traditional method for diagnosing faults is difficult to the problem accurately differentiated of malfunction when variable speed.

The technical solution adopted in the present invention is to carry out according to following steps:

Step 1: gather the vibration signal of bearing, the rotating speed of synchronous acquisition blower fan main shaft；

Step 2: draw out corner change curve according to rotating speed；

Step 3: corner change curve is angularly divided, it is determined that the angularly time series of resampling；

Step 4: bearing vibration signal is interpolated according to time series, obtains the angular domain vibration signal of bearing；

Step 5: add random Gaussian white noise sequence in angular domain vibration signal, obtain signal, utilize empirical mode decomposition EMD algorithm to obtain one group of IMF the signal adding white noise；

Step 6: add different white noise sequences every time, repeat step 5, until reaching the times N set；

Step 7: calculate the average that n times decompose each IMF obtained, as decomposing final result；

Step 8: utilize the average of each IMF, calculates the kurtosis value of each IMF component respectively；

Step 9: relatively each IMF component kurtosis value K_iSize, filter out the IMF component c that kurtosis value is maximum_maxT () is as sensitive IMF；

Step 10: sensitive IMF is carried out Hilbert envelope demodulation and obtains envelope signal, utilizes Fourier transform pairs envelope signal to carry out processing the Order Envelope Spectrum Analysis obtaining sensitive IMF, thus extracting fault characteristic frequency, it is achieved the fault diagnosis of Wind turbines bearing.

Further, the step of described empirical mode decomposition EMD algorithm is as follows:

Step 1: extract signalAll of local maximum and minimum point；

Step 2: utilize cubic spline interpolation to connect local maximum and local minizing point respectively, construct signalCoenvelope line and lower envelope line；

Step 3: calculate the meansigma methods of upper and lower envelope, by signalDeduct meansigma methods and obtain h (t)；

Step 4: judge that can h (t) meet the condition of intrinsic mode function, if meeting, h (t) is first intrinsic mode function component IMF1 of signal x (t)；If being unsatisfactory for, using h (t) as new signal, repeat step 1～4, until meeting the condition of intrinsic mode function；

Step 5: by IMF1 from signalIn separate, obtain signal discrepance, it is judged that can discrepance continue to decompose, and if could; discrepance would be repeated step 1～5 as new signal, if could not; EMD would decompose and terminates, thus by signalIt is decomposed into one group of IMF.

Further, the condition of described intrinsic mode function includes two aspects: the number of (1) extreme point and the number of zero crossing must equal or maximum differences one；(2) at any time, Local modulus maxima the meansigma methods of the coenvelope line formed and the lower envelope line formed by local minizing point is equal to zero, and namely upper and lower envelope is relative to time shaft Local Symmetric.

Further, described step 10 to realize step as follows:

Step 1: sensitive IMF component is carried out Hilbert conversion,

$y\left(t\right)=c_{max}\left(t\right)*\frac{1}{\mathrm{\π}t}$

In formula: y (t) is hubert transformed signal, c_maxT () is sensitive IMF component, * is convolution symbol；

Step 2: ask for envelope signal z (t) of sensitive IMF according to following formula:

$z\left(t\right)=\sqrt{{\[c_{max}\left(t\right)\]}^2+y^2\left(t\right)};$

Step 3: gained envelope signal z (t) carries out Fourier transformation, obtains the Order Envelope Spectrum Analysis of sensitive IMF；

Step 4: extracting the frequency content that in sensitive IMF Order Envelope Spectrum Analysis, amplitude is prominent, contrasting with the theoretical fault characteristic frequency of bearing, thus diagnosing the bearing fault of Wind turbines；

The wherein theoretical fault characteristic frequency f of bearing inner race_iComputing formula be:

$f_i=\frac{1}{2}(1+\frac{d}{D}cos\mathrm{\β})f_{r}Z$

The theoretical fault characteristic frequency f of bearing outer ring_oComputing formula be:

$f_o=\frac{1}{2}(1-\frac{d}{D}cos\mathrm{\β})f_{r}Z$

In formula, f_rTurning frequently for axle, D is bearing pitch diameter, and d is rolling element diameter, and β is the contact angle of bearing, and Z is rolling element quantity.

The invention has the beneficial effects as follows by the time domain vibration signal of Wind turbines bearing is carried out angular domain resampling, the time-domain signal of non-stationary can be converted to stable angular domain signal, the rotation speed change impact on analysis of vibration signal can be eliminated, improve accuracy and the effectiveness of fault diagnosis.

Accompanying drawing explanation

Fig. 1 is Method for Bearing Fault Diagnosis flow chart；

Fig. 2 is integrated empirical mode decomposition (EEMD) algorithm flow chart；

Fig. 3 (a) is outer ring fault time-domain signal；

Fig. 3 (b) is outer ring fault angular domain resampling signal；

Fig. 4 is bearing outer ring fault-signal EEMD decomposition result；

Fig. 5 (a) is the envelope of C1 component；

Fig. 5 (b) is the envelope spectrum of C1 component.

Detailed description of the invention

Below in conjunction with detailed description of the invention, the present invention is described in detail.

It is illustrated in figure 1 the Wind turbines Method for Bearing Fault Diagnosis schematic flow sheet under variable speed of the present invention, specifically comprises the following steps that

One, fault signature extracts

1 angular domain resampling

Step 1: install acceleration transducer, vibration signal x (t) of Real-time Collection bearing on Wind turbines bearing, and install speed probe, rotating speed n (t) of synchronous acquisition blower fan main shaft on Wind turbines.

Step 2: utilize least-square fitting approach to be fitted rotating speed n (t) obtaining rotating speed fitting function n^*(t)。

Step 3: to rotating speed fitting function n^*T () is integrated, can obtain variation relation θ (t) of angle of eccentricity and time, and draw out corner change curve according to θ (t), and integral formula is as follows:

θ (t)=∫ n^*(t)dt

Step 4: corner change curve is angularly divided, it is determined that the angularly time series t of resampling_k。

Step 5: according to time series t_kBearing vibration signal x (t) is interpolated, finally gives the angular domain vibration signal x of bearing^*(t)。

2 integrated empirical mode decomposition EEMD

Utilize the integrated empirical mode decomposition (EEMD) improved to angular domain vibration signal x^*T () processes.The flow chart of the integrated empirical mode decomposition EEMD algorithm improved in the present invention is shown in Fig. 2, and the concrete decomposition step of algorithm is as follows:

Step 1: at angular domain vibration signal x^*T () adds random Gaussian white noise sequence, obtain signal

Step 2: the signal of white noise will be addedEmpirical mode decomposition (EMD) is utilized to obtain one group of IMF.

Step 3: every time add different white noise sequences, repeat the above steps 1 and step 2, until reaching the times N set.

Step 4: calculate the average c that n times decompose each IMF obtained_iT (), decomposes final result as EEMD.

Wherein, the step of empirical mode decomposition (EMD) algorithm is as follows:

Step 1: extract signalAll of local maximum and minimum point.

Step 2: utilize cubic spline interpolation to connect local maximum and local minizing point respectively, construct signalCoenvelope line and lower envelope line.

Step 3: calculate the meansigma methods of upper and lower envelope, by signalDeduct meansigma methods and obtain h (t).

Step 4: judge that can h (t) meet the condition of intrinsic mode function, if meeting, h (t) is first intrinsic mode function component IMF1 of signal x (t)；If being unsatisfactory for, using h (t) as new signal, repeat step 1～4, until meeting the condition of intrinsic mode function.The condition of intrinsic mode function mainly includes two aspects: the number of (1) extreme point and the number of zero crossing must equal or maximum differences one；(2) at any time, Local modulus maxima the meansigma methods of the coenvelope line formed and the lower envelope line formed by local minizing point is equal to zero, and namely upper and lower envelope is relative to time shaft Local Symmetric.

Step 5: by IMF1 from signalIn separate, obtain signal discrepance, it is judged that can discrepance continue to decompose, and if could; discrepance would be repeated step 1～5 as new signal, if could not; EMD would decompose and terminates, thus by signalIt is decomposed into one group of IMF.

The 3 sensitive IMF of screening

After angular domain vibration signal is decomposed into one group of IMF component by EEMD algorithm, calculate the kurtosis value K of each IMF component respectively_i, computing formula is as follows:

$K_i=E{\left(c_i\right(t)-{\mathrm{\μ}}_i)}^4/{{\mathrm{\σ}}_i}^4$

In formula: c_iT () decomposes the i-th rank IMF component of gained, K for EEMD_iFor c_iThe kurtosis value of (t) component, μ_iFor c_iThe meansigma methods of (t) component, σ_iFor c_iT the variance of () component, E represents the operative symbol asking for expected value.

When bearing is in normal operating condition, the amplitude distribution of vibration signal is similar to normal distribution, and its kurtosis value approximates 3.When there is more impact composition in signal, when namely comprising more fault message, the kurtosis of signal can substantially become big.Kurtosis value is bigger, impacts the proportion shared by composition the more in signal, and fault characteristic information also more easily extracts.Therefore, each IMF component kurtosis value K is compared_iSize, filter out the IMF component c that kurtosis value is maximum_maxT () is as sensitive IMF.

4 Hilbert envelope demodulation

Filtering out sensitive IMF component c_maxAfter (t), sensitive IMF is carried out Hilbert (Hilbert) envelope demodulation and obtains envelope signal, Fourier transform pairs envelope signal is utilized to carry out processing the Order Envelope Spectrum Analysis obtaining sensitive IMF, thus extracting fault characteristic frequency, it is achieved the fault diagnosis of Wind turbines bearing.Specifically comprise the following steps that

Step 1: sensitive IMF component is carried out Hilbert conversion,

$y\left(t\right)=c_{max}\left(t\right)*\frac{1}{\mathrm{\π}t}$

In formula: y (t) is hubert transformed signal, c_maxT () is sensitive IMF component, * is convolution symbol.

Step 2: ask for envelope signal z (t) of sensitive IMF according to following formula,

$z\left(t\right)=\sqrt{{\[c_{max}\left(t\right)\]}^2+y^2\left(t\right)}$

Step 3: gained envelope signal z (t) is carried out Fourier transformation, can obtain the Order Envelope Spectrum Analysis of sensitive IMF.

Step 4: extracting the frequency content that in sensitive IMF Order Envelope Spectrum Analysis, amplitude is prominent, contrasting with the theoretical fault characteristic frequency of bearing, thus diagnosing the bearing fault of Wind turbines.

The wherein theoretical fault characteristic frequency f of bearing inner race_iComputing formula be:

$f_i=\frac{1}{2}(1+\frac{d}{D}cos\mathrm{\β})f_{r}Z$

The theoretical fault characteristic frequency f of bearing outer ring_oComputing formula be:

$f_o=\frac{1}{2}(1-\frac{d}{D}cos\mathrm{\β})f_{r}Z$

In formula, f_rTurning frequently for axle, D is bearing pitch diameter, and d is rolling element diameter, and β is the contact angle of bearing, and Z is rolling element quantity.

Two, sample calculation analysis checking

Vibration data is the bearing vibration signal of actual measurement.Experiment bearing used is 6025-2RS type deep groove ball bearing, and lesion size is 0.018mm, and the fault degree of depth is 0.028cm.Sensor sample rate f_sFor 12KHz, sampling number is 3600.Rolling bearing is mainly made up of outer ring, inner ring and rolling element three part, adopts spark technology to be processed to simulation outer ring local damage at bearing outer ring, and the fault degree of setting belongs to slighter degree.

Fig. 3 is the time-domain diagram of the Wind turbines variable speed process middle (center) bearing outer ring fault vibration signal that Wind turbines bearing outer ring fault-signal figure, Fig. 3 (a) measure for sensor.Outer ring fault-signal is carried out angular domain resampling, obtains its angular domain signal such as shown in Fig. 3 (b).Bearing outer ring fault angle territory signal carrying out integrated empirical modal (EEMD) and decomposes 8 intrinsic mode function (IMF) components of acquisition, the time domain waveform of each IMF component is as shown in Figure 4.Calculating the kurtosis value of 8 IMF components of gained respectively, result shows that the kurtosis value of C1 component is maximum, chooses C1 component as sensitive IMF component according to kurtosis criterion.C1 component carrying out Hilbert (Hilbert) envelope demodulation and extracts the envelope signal of C1 component, Fig. 5 is fault diagnosis result.As shown in Fig. 5 (a).Recycling Fourier transform pairs envelope signal carries out spectrum analysis, obtains the envelope spectrum of C1 component such as shown in Fig. 5 (b).Can be seen that from Fig. 5 (b) amplitude of outer ring fault characteristic frequency (105Hz) and two frequencys multiplication (209Hz), frequency tripling (314Hz) substantially highlights, outer ring fault signature is extracted well.It is shown that the method for the invention can efficiently extract bearing fault characteristics, the fault diagnosis for Wind turbines bearing provides foundation.

It is also an advantage of the present invention that:

(1) what adopt is vibration data, and analysis of vibration signal method is a kind of effective condition detection method, is particularly well-suited to the rotating machineries such as bearing.

(2) the Wind turbines bearing vibration signal in variable speed process is non-stationary signal, the effect that traditional analysis method cannot obtain.Angular domain resampling is a kind of effective ways analyzing Non-stationary vibration signal, it is possible to eliminate the rotation speed change impact on analysis of vibration signal.

(3) integrated empirical mode decomposition can suppress the interference that noise causes effectively, improves accuracy and the effectiveness of fault diagnosis.

(4) by the time domain vibration signal of Wind turbines bearing is carried out angular domain resampling, it is possible to the time-domain signal of non-stationary is converted stable angular domain signal to, thus being further analyzed.

(5) bearing signal decomposition can be several intrinsic mode function components by integrated empirical mode decomposition adaptively, thus being separated by the modulation intelligence relevant with fault.

(6) envelope signal corresponding with shock pulse being attached on high-frequency vibration signal is extracted in envelope demodulation, thus being more concentrated at faults information.

The above is only the better embodiment to the present invention, not the present invention is done any pro forma restriction, every any simple modification embodiment of above done according to the technical spirit of the present invention, equivalent variations and modification, belong in the scope of technical solution of the present invention.

Claims (4)

1. the Wind turbines Method for Bearing Fault Diagnosis under a variable speed, it is characterised in that carry out according to following steps:

Step 1: gather the vibration signal of bearing, the rotating speed of synchronous acquisition blower fan main shaft；

Step 2: draw out corner change curve according to rotating speed；

Step 3: corner change curve is angularly divided, it is determined that the angularly time series of resampling；

Step 4: bearing vibration signal is interpolated according to time series, obtains the angular domain vibration signal of bearing；

Step 5: add random Gaussian white noise sequence in angular domain vibration signal, obtain signal, utilize empirical mode decomposition EMD algorithm to obtain one group of IMF the signal adding white noise；

Step 6: add different white noise sequences every time, repeat step 5, until reaching the times N set；

Step 7: calculate the average that n times decompose each IMF obtained, as decomposing final result；

Step 8: utilize the average of each IMF, calculates the kurtosis value of each IMF component respectively；

Step 9: relatively each IMF component kurtosis value K_iSize, filter out the IMF component c that kurtosis value is maximum_maxT () is as sensitive IMF；

Step 10: sensitive IMF is carried out Hilbert envelope demodulation and obtains envelope signal, utilizes Fourier transform pairs envelope signal to carry out processing the Order Envelope Spectrum Analysis obtaining sensitive IMF, thus extracting fault characteristic frequency, it is achieved the fault diagnosis of Wind turbines bearing.

2. the Wind turbines Method for Bearing Fault Diagnosis under a kind of variable speed described in claim 1, it is characterised in that: the step of described empirical mode decomposition EMD algorithm is as follows:

Step 1: extract signalAll of local maximum and minimum point；

Step 2: utilize cubic spline interpolation to connect local maximum and local minizing point respectively, construct signalCoenvelope line and lower envelope line；

Step 3: calculate the meansigma methods of upper and lower envelope, by signalDeduct meansigma methods and obtain h (t)；

Step 4: judge that can h (t) meet the condition of intrinsic mode function, if meeting, h (t) is first intrinsic mode function component IMF1 of signal x (t)；If being unsatisfactory for, using h (t) as new signal, repeat step 1～4, until meeting the condition of intrinsic mode function；

Step 5: by IMF1 from signalIn separate, obtain signal discrepance, it is judged that can discrepance continue to decompose, and if could; discrepance would be repeated step 1～5 as new signal, if could not; EMD would decompose and terminates, thus by signalIt is decomposed into one group of IMF.

3. the Wind turbines Method for Bearing Fault Diagnosis under a kind of variable speed described in claim 2, it is characterised in that: the condition of described intrinsic mode function includes two aspects: the number of (1) extreme point and the number of zero crossing must equal or maximum differences one；(2) at any time, Local modulus maxima the meansigma methods of the coenvelope line formed and the lower envelope line formed by local minizing point is equal to zero, and namely upper and lower envelope is relative to time shaft Local Symmetric.

4. the Wind turbines Method for Bearing Fault Diagnosis under a kind of variable speed described in claim 1, it is characterised in that: described step 10 to realize step as follows:

Step 1: sensitive IMF component is carried out Hilbert conversion,

$y\left(t\right)=c_{max}\left(t\right)*\frac{1}{\mathrm{\π}t}$

In formula: y (t) is hubert transformed signal, c_maxT () is sensitive IMF component, * is convolution symbol；

Step 2: ask for envelope signal z (t) of sensitive IMF according to following formula:

$z\left(t\right)=\sqrt{{\[c_{max}\left(t\right)\]}^2+y^2\left(t\right)};$

Step 3: gained envelope signal z (t) carries out Fourier transformation, obtains the Order Envelope Spectrum Analysis of sensitive IMF；

Step 4: extracting the frequency content that in sensitive IMF Order Envelope Spectrum Analysis, amplitude is prominent, contrasting with the theoretical fault characteristic frequency of bearing, thus diagnosing the bearing fault of Wind turbines；

The wherein theoretical fault characteristic frequency f of bearing inner race_iComputing formula be:

$f_i=\frac{1}{2}(1+\frac{d}{D}cos\mathrm{\β})f_{r}Z$

The theoretical fault characteristic frequency f of bearing outer ring_oComputing formula be:

$f_o=\frac{1}{2}(1-\frac{d}{D}cos\mathrm{\β})f_{r}Z$

In formula, f_rTurning frequently for axle, D is bearing pitch diameter, and d is rolling element diameter, and β is the contact angle of bearing, and Z is rolling element quantity.