CN102937522A - Composite fault diagnosis method and system of gear case - Google Patents
Composite fault diagnosis method and system of gear case Download PDFInfo
- Publication number
- CN102937522A CN102937522A CN2012103153118A CN201210315311A CN102937522A CN 102937522 A CN102937522 A CN 102937522A CN 2012103153118 A CN2012103153118 A CN 2012103153118A CN 201210315311 A CN201210315311 A CN 201210315311A CN 102937522 A CN102937522 A CN 102937522A
- Authority
- CN
- China
- Prior art keywords
- bank
- gear case
- filters
- gear
- combined failure
- 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
Images
Abstract
The invention discloses a composite fault diagnosis method and system of a gear case. The composite fault diagnosis method comprises firstly measuring and storing vibration signals of the gear case; then adopting 1/4 sampling translation dual tree complex wavelet transform for resolving; and finally extracting a plurality of fault characteristics by adopting an energy operator demodulation method in a plurality of sub-band signals obtained by resolving, and further identifying a composite fault mode. The composite fault diagnosis method is fused with a complementary characteristic using dual tree complex wavelets and energy operator demodulation, and the obtained gear case is high in recognition capability of composite fault characteristic extraction. By means of speed capability of parallel realization of the dual tree complex wavelet transform and the energy operator demodulation method, partial polling and on-line monitoring of the gear case in a working condition can be completely applied, sudden accidents can be avoided, and the composite fault diagnosis method and system can be applicable to gear case portions in different models.
Description
Technical field
The present invention relates to the gearbox fault detection field, be specifically related to a kind of gear case combined failure diagnostic method and system.
Background technology
Gear case is because its ratio of gear is fixed, and driving torque is large, and compact conformation becomes gear parts commonly used, in various plant equipment, is widely used.Move because gear is everlasting at a high speed, under the severe environmental conditions such as heavy duty and thump, very easily wear and tear, the various faults such as tired, broken teeth and crackle, and further bring out other fault, thereby cause tremendous economic loss.Therefore, the running status of gear case is monitored and identified in time to the fault of its generation, there is important engineering significance.
In the gear manufacture due to gear in service and alignment error, peel off, driving source that the fault such as crackle can directly become vibration, it is the cycle that these driving sources all be take the revolution of gear shaft, contains gyrofrequency and the frequency multiplication thereof of this axle in Gearbox vibration signal.Therefore the vibration signal that hinders gear often shows as the modulation of gyrofrequency to meshing frequency, forms centered by meshing frequency the sideband be spacedly distributed on spectrogram.In fact the signal that participates in modulation also comprises the frequency multiplication of meshing frequency, and due to the acting in conjunction of Frequency And Amplitude Modulation, the frequency spectrum finally formed shows as a series of asymmetric sideband group centered by meshing frequency and each harmonic thereof.The core of Fault Diagnosis of Gear Case is the analyzing and processing of the vibration signal to gathering, yet current gear box fault diagnosis method is mainly for single fault mode, as gear tooth breakage, spot corrosion, tooth root crackle and etc. manufacture and the alignment error of gear.Diagnostic method for single gear distress has frequency-domain analysis, cepstrum analysis and narrowband demodulation analysis etc.In the frequency domain diagnosis of gear, be the source of trouble information reflected by sideband, sideband feature difference corresponding to different gear distresses diagnosed.The sideband that local faults such as the peeling off of gear, tooth root crackle or part broken teeth causes is wider and smooth, and the sideband exponent number is many and amplitude is big or small evenly; The sideband that the distribution faults such as gear spot corrosion cause, its side frequency exponent number is few and concentrate on the both sides of meshing frequency and frequency multiplication thereof; The modulation sideband that the faults such as gear is uneven, misalign, machinery is loosening cause is asymmetric.Yet, because the side frequency composition often has instability, under actual working environment, especially when several faults and while depositing, the variation of side frequency will present resultant effect, its Changing Pattern is difficult to describe by above-mentioned a certain typical case, so the analysis of traditional frequency domain sideband is difficult to identify a plurality of faults of gear case.Tradition cepstrum method another kind of common method in Gear Fault Diagnosis, cepstrum is subject to the impact of delivering path very little, can will in original spectrogram, become the sideband of family to be reduced to a single spectral line, can identify the periodic structure in complicated spectrogram, separation and extraction goes out the periodic component of signal, but, when the Gearbox vibration signal signal to noise ratio (S/N ratio) is low, it is helpless that cepstrum analysis often seems.In addition, when the gear multiple faults exists, utilize a single spectral line in the scramble spectrogram can't accurately judge gear distress type and failure cause.The narrowband demodulation technology is widely used in Gear Fault Diagnosis, and it is frequency centered by certain single order gear mesh frequency, by choosing suitable bandwidth, vibration signal is carried out to demodulation analysis after bandpass filtering, obtains the gear distress characteristic information.If centre frequency and bandwidth were chosen not at that time, may cause the false judgment to the gear running status.Patent of the present invention proposes a kind of novel diagnostic method for the feature extraction of gear case combined failure.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of gear case combined failure diagnostic method and system, it is by simple vibration measurement, the employing dual-tree complex wavelet transform efficiently, decomposition obtains a plurality of subband complex signals reliably, and then adopt each subband signal of energy operator demodulation method demodulation, extract gear correlated characteristic identifying and diagnosing.
For addressing the above problem, the present invention is achieved by the following scheme:
A kind of gear case combined failure of the present invention diagnostic method, comprise the steps:
(1) adopt the acceleration vibration transducer to pick up the gear case vibration signal, this acceleration transducer is arranged on gearbox input shaft end cap to be measured;
(2) the gearbox of wind turbine vibration signal gathered is carried out to 1/4 sampling translation dual-tree complex wavelet transform, obtain a plurality of sub-band division signals of signal;
At first, construct 1/4 sampling translation dual-tree complex wavelet transform two tree bank of filters used, the bank of filters that the ground floor of this pair of tree bank of filters decomposes use is identical with the bank of filters that original dual-tree complex wavelet decomposes ground floor, and follow-up high level decomposes bank of filters used and all adopts and postpone 1/4 sampling structure, the later layer bank of filters postpones 1/4 sampling period on the basis of last layer bank of filters;
Secondly, utilize two bank of filters of setting of construct, adopt two wavelet transforms arranged side by side to the gear case vibration signal the picked up decomposition that walks abreast, obtain a plurality of sub-band division signals that gear case vibrates; Above-mentioned a plurality of sub-band division signal includes a low frequency sub-band decomposed signal and a series of high-frequency sub-band decomposed signal; Wherein the real part of each sub-band division signal and imaginary part consist of first wavelet transform and second wavelet transform respectively;
(3) a plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic and carry out identifying and diagnosing.
It is bi-orthogonal filter that the ground floor of two tree bank of filters described in above-mentioned steps (2) decomposes the bank of filters of using.
It is linear phase Q translate filter that the high level of two tree bank of filters described in above-mentioned steps (2) decomposes the bank of filters of using.
Described in above-mentioned steps (2), the decomposition number of plies of described two tree bank of filters is between 3～5 layers.
Described in above-mentioned steps (3), the method for energy operator demodulation is:
In formula, x
i(t) mean a pending sub-band division signal, ψ means energy operator.
Adopt artificial and mode that computing machine combines in above-mentioned steps (3) to the gear combined failure correlated characteristic that extracts and carry out identifying and diagnosing,
The gear combined failure correlated characteristic that extracts and pre-stored known gears combined failure feature in the property data base of computing machine are compared; When the gear combined failure correlated characteristic extracted is identical with known gears combined failure feature, Computer Automatic Recognition goes out the concrete fault mode of this gear case to be measured; When the gear combined failure correlated characteristic extracted is not identical with known gears combined failure feature, the gear combined failure correlated characteristic that computing machine can't be differentiated this is stored, under storage can't automatic discrimination the concrete fault mode of gear case need to adopt artificial investigation mode could progressively go out to judge the concrete fault mode of gear case.
In above-mentioned steps (3), the gear combined failure correlated characteristic that can't differentiate for this, after the artificial investigation mode of employing has been determined the concrete fault mode of gear case, need the concrete fault mode extension storage of the gear case of this gear combined failure correlated characteristic and correspondence to the property data base of computing machine.
A kind of gear case combined failure of the present invention diagnostic system, it mainly consists of acceleration vibration transducer, 1/4 sampling translation dual-tree complex wavelet transform module, energy operator demodulation process module, pattern recognition module and property data base;
The acceleration vibration transducer, be arranged on gearbox input shaft end cap to be measured, for picking up the gear case vibration signal;
1/4 sampling translation dual-tree complex wavelet transform module, comprise 2 two tree bank of filters arranged side by side, for the gearbox of wind turbine vibration signal to gathering, carries out 1/4 sampling translation dual-tree complex wavelet transform, obtains a plurality of sub-band division signals of signal;
Above-mentioned each two tree bank of filters comprise ground floor analysis filter bank and follow-up high-rise analysis filter bank, this ground floor analysis filter bank is identical with the bank of filters that original dual-tree complex wavelet decomposes ground floor, and follow-up high level decomposes bank of filters used and all adopts and postpone 1/4 sampling structure, the later layer bank of filters postpones 1/4 sampling period on the basis of last layer bank of filters;
Energy operator demodulation process module; A plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic;
Pattern recognition module, in the gear combined failure correlated characteristic that energy operator demodulation process module is extracted and property data base, known gears combined failure feature compares, and identifies the concrete fault mode of gear case to be measured.
It is bi-orthogonal filter that the ground floor of above-mentioned two tree bank of filters decomposes the bank of filters of using.
It is linear phase Q translate filter that the high level of above-mentioned two tree bank of filters decomposes the bank of filters of using.
The decomposition number of plies of above-mentioned two tree bank of filters is between 3～5 layers.
Compared with prior art, the present invention has following features:
1, overcome a gear case combined failure diagnosis difficult problem, utilize the characteristics such as the approximate translation invariance of dual-tree complex wavelet transform and analyticity, different faults is decomposed in the different sub-band signal of vibration signal, then the fast demodulation method by energy operator accurately identifies the inner a plurality of fault signature compositions of damage gear case, by artificial or computing machine, carrys out the failure judgement pattern;
2, merge the complementary characteristic that utilizes dual-tree complex wavelet and energy operator demodulation; Dual-tree complex wavelet transform has approximate analysis and translation invariant feature, and these all are very beneficial for the combined failure feature extraction, and the gear case combined failure feature extraction recognition capability of acquisition is high;
3, fast operation, by means of the Parallel Implementation of dual-tree complex wavelet transform and the rapidity of energy operator demodulating algorithm, the method can be applied to duty lower tooth roller box fully and partly patrol and examine and on-line monitoring, avoids sudden accident to occur;
4, can be suitable for the gear case part of different model, extensively promote the use of in the equipment that can extensively adopt at gear casees such as wind power generating set.
The accompanying drawing explanation
The diagnostic method process flow diagram that Fig. 1 is a kind of gear case combined failure;
The decomposition and reconstruction schematic diagram that Fig. 2 is a kind of dual-tree complex wavelet transform;
Fig. 3 is gear case vibration signal time domain waveform;
Fig. 4 is gear case vibration signal power spectrum;
Fig. 5 is gear case vibration signal cepstrum;
Fig. 6 is the ground floor detail signal that gear case vibration signal dual-tree complex wavelet decomposes;
Fig. 7 (a) and (b) be respectively second layer approximation signal and the energy operator demodulation spectra thereof that gear case vibration signal dual-tree complex wavelet transform decomposes;
Fig. 8 is maintenance rear gear box vibration signal;
The ground floor detail signal that Fig. 9 is gear case vibration signal dual-tree complex wavelet transform;
Figure 10 is the second layer approximation signal that gear case vibration signal dual-tree complex wavelet decomposes.
Embodiment
The fault of wind-driven generator wheel-box of below take is example, and the present invention is described in detail:
Referring to Fig. 1, a kind of gear case combined failure diagnostic method, comprise the steps:
(1) adopt the acceleration vibration transducer to pick up the gear case vibration signal, this acceleration transducer is arranged on gearbox input shaft end cap to be measured.
In the present embodiment, measuring wind power generating set output shaft rotating speed is that 495r/min(turns frequency for 8.25Hz), the pinion wheel number of teeth of gear case is 20, and the gear wheel number of teeth is 151, and ratio of gear is 0.132, and it is 8.25Hz frequently that pinion wheel turns.The gear case of blower meshing frequency is 165Hz.In the fan operation process, the gear case vibration is violent, adopts the acceleration vibration transducer to pick up the gear case vibration signal, and acceleration transducer is arranged on the gearbox input shaft end cap.For the rotating speed of the some axles of prototype gear case need to be installed velocity gauge or shaft encoder on corresponding axis.In view of the gear case of blower rotating speed is not high, sample frequency is set to 2000Hz, and Fig. 3 is gear case vibration signal time domain waveform.Simply from vibration signal time domain waveform observation, can find out more in a jumble, can not get useful diagnostic message.
Signal Power Spectrum Analysis is to search failure messages such as having or not obvious modulation from spectrogram, and power spectrumanalysis is mainly by identification frequency conversion information, the different sideband feature that different gear distresses is corresponding.The gear distress pattern substantially can be divided into two classes: a class is local fault, such as the peeling off of gear, tooth root crackle or part broken teeth etc.; Another kind of is distributed fault, as uneven as gear wear, gear, gear misaligns, machinery becomes flexible etc.The sideband that local fault causes is wider and smooth, and the sideband exponent number is many and amplitude is big or small evenly.The sideband that gear wear distributed fault causes, its side frequency exponent number is few and concentrate on the both sides of meshing frequency and frequency multiplication thereof; The modulation sideband that gear is uneven, misalign, the loosening distributed fault of machinery causes is asymmetric.The power spectrum that Fig. 4 is the gear case vibration signal, gear mesh frequency and higher hamonic wave thereof appear in spectrogram, and have the sideband composition of 8.25Hz in meshing frequency and higher hamonic wave both sides thereof, this frequency turns and frequently equates with pinion wheel, shows that the reason of gear case high vibration is relevant with pinion wheel.Although the spectrum analysis of gear case vibration signal provides some useful Fault Diagnosis of Gear Case information, but spectrum analysis is limited to the analysis ability of non-stationary signal, can not discloses well gear crack, local fatigue and the non-stationary signal feature that local fault causes such as peel off.
Cepstrum is a conventional art of Gear Fault Diagnosis, and this technology is still more effective for gear single failure pattern, but diagnosis has limitation for combined failure.The cepstrum that Fig. 5 is the gear case vibration signal, in cepstrum, obvious peak value appears in the 0.121s place, and 0.121s turns frequently with pinion wheel that 8.25Hz is corresponding, it is relevant with pinion wheel that cepstral analysis also shows that gear case vibrates.But traditional power spectrum sideband analysis and cepstrum analysis all can't accurately judge the gear case high vibration caused due to the pinion wheel distributed faults or due to the pinion wheel local fault.
2, the gearbox of wind turbine vibration signal gathered is carried out to 1/4 sampling translation dual-tree complex wavelet transform, obtain a plurality of sub-band division signals of signal.
2.1, construct 1/4 sampling translation dual-tree complex wavelet transform two tree bank of filters used, the bank of filters that wherein the ground floor decomposition is used is decomposed ground floor with original dual-tree complex wavelet and is constructed identical, and follow-up high-rise decomposition bank of filters used all adopts delay 1/4 to sample structure, i.e. later layer bank of filters 1/4 sampling period of delay on the basis of last layer bank of filters.
2.2, utilize the bank of filters of pair setting construct, adopt two wavelet transforms arranged side by side to the gear case vibration signal the picked up decomposition that walks abreast, obtain a plurality of sub-band division signals of gear case vibration, a plurality of sub-band division signals comprise a low frequency sub-band decomposed signal and a series of high-frequency sub-band decomposed signal; Wherein the real part of each sub-band division signal and imaginary part consist of first wavelet transform and second wavelet transform respectively.
Wavelet transform (DWT) is due to the representation of a kind of effective base that a certain signal is provided, and operation efficiency is higher, thereby is widely used.But there is shortcomings such as block overlap of frequency bands effect, Gipps effect and translation variability in DWT.Dual-tree complex wavelet transform (Dual-tree Complex Wavelet Transform, DTCWT) be a kind of New Wavelet Transform method with many good characteristics that development in recent years is got up, there is the advantages such as approximate translation invariance, directional selectivity (processing for two dimensional image) and approximate analysis.So-called analyticity refers to that two tree wavelet filter frequency spectrums almost do not have the negative frequency composition.The translation invariance of DTCWT is very beneficial for the extraction of shock characteristic, and the characteristic of approximate analysis can effectively be eliminated the block overlap of frequency bands effect, therefore is very beneficial for combined failure and extracts.
1/4 translation dual-tree complex wavelet arthmetic statement is as follows:
The coefficient g of dual-tree complex wavelet two groups of wave filters used
0And h (n)
0(n) to meet following formula
g
0(n)=h
0(N-1-n)①
In formula, N is even number, means to have h
0(n) length of that group wave filter of coefficient.In such cases known by above formula, the amplitude condition of two groups of wave filters is accurately to meet, and has following relation between the phase place of two groups of wave filters
∠G
0(e
jω)=-∠H
0(e
jω)-(N-1)ω②
It is very simple that DTCWT realizes, due to its adopt two parallel and use different low passes and the DWT of Hi-pass filter, its decomposition and reconstruction is as shown in Figure 2; Therefore, DTCWT can relatively realize with concurrent program, will greatly improve its analysis efficiency.In its decomposable process, two real wavelet transformations adopt two groups of different wave filters, and each group all meets respectively the perfect reconstruction condition, and the co-design of two groups of wave filters guarantees that whole conversion is approximate analysis.
If ψ
hAnd ψ (t)
g(t) mean respectively two real-valued small echos, the correspondingly φ that DTCWT adopts
hAnd φ (t)
g(t) be its scaling function.
Since DTCWT consists of two DWT, so according to wavelet theory, above the wavelet coefficient of DWT
With scale coefficient
Can obtain according to inner product operation
Here, l is scale factor, and J is maximum decomposition scale.Similarly set below
With
Coefficient, can by will be 3. with 4. in ψ
hAnd φ (t)
h(t) transposing is ψ
gAnd φ (t)
g(t) after, obtain.Final DTCWT output coefficient of wavelet decomposition is to obtain according to two tree combinations,
The wavelet coefficient length obtained like this, along with deeply can reducing by half gradually of decomposing, if will obtain the decomposition result isometric with original signal, can adopt down the wavelet coefficient list that two formulas mean to prop up restructing algorithm
Wherein, m and n mean the length of wave filter, and span depends on two tree wavelet filter actual used.In the present embodiment, m=13 when ground floor decomposes, n=19; In follow-up each layer decomposes, m=n=14.For the selection of the number of plies J decomposed, similar with the wavelet transformation of other types, what generally adopt is exactly to decompose 3～5 layers.If decomposing the number of plies is 4 layers, to decompose the number of plies corresponding be the 2nd layer to the 4th layer decomposition to so-called follow-up high level.
Talk about publicly knownly, wavelet decomposition and restructing algorithm can adopt the fast algorithm of Mallat.Top " real tree " braning factor two yardstick l in upper figure and the coefficient between l+1
With
There is following relation
H wherein
0With h
1" real tree " wavelet transformation low pass and Hi-pass filter used above meaning respectively, and
With
It is its reconfigurable filter.Similarly, below, " empty tree " braning factor can be obtained by following formula
Similarly, g
0With g
1" empty tree " wavelet transformation low pass and Hi-pass filter used below meaning respectively, and
With
It is its reconfigurable filter.
In the present embodiment, described two tree bank of filters of decomposing except ground floor are all to select the follow-up high-rise number of plies of decomposing of 14 rank linear phase Q translate filter of said method structure.Two tree bank of filters of certainly, decomposing except ground floor also can be selected the use of the linear phase Q translate filter of 16 other orders such as grade as follow-up high-rise decomposition.Because all having advantages of dual-tree complex wavelet, decomposes by the different wave filter adopted, and therefore very micro-on the impact of analysis result.
Two tree wave filters of above-mentioned structure are applicable to the DTCWT conversion of decomposing except ground floor, are also that the bank of filters that the ground floor decomposition is used is the condition that does not meet half sampling delay.In fact ground floor decomposes wave filter used and can select leeway very large, such as the biorthogonal filtering that can select orthogonal filter and other orders.In the present embodiment, the DTCWT ground floor decomposes all employings (13, 19) bi-orthogonal filter of rank near symmetrical, concrete low pass and Hi-pass filter coefficient are respectively h0=[-0.001758, 0, 0.02227,-0.04688,-0.04824, 0.2969, 0.5555, 0.2969,-0.04824,-0.04687, 0.02227, 0,-0.001758], g0=[-0.00007063, 0, 0.001342,-0.001883,-0.007157, 0.02386, 0.05564,-0.05169,-0.2998, 0.5594,-0.2998,-0.5169, 0.05564, 0.02386,-0.007157,-0.001883, 0.001342, 0,-0.00007063].
3, a plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic and carry out identifying and diagnosing.
During the Gear Fault Diagnosis process, broken teeth for gear, rippling, the faults such as tooth root crackle, it is characterized in that in certain the sub-band division signal decomposed life period is spaced apart fault gear place axle and turns impact frequently and (do not rotate a circle, in damage, the gear place can excite an impact), and the picture gear wear, the faults such as installation misaligns, feature is not very obvious, common way is that to need the people be whether check that fault gear place axle in the demodulation frequency spectrum of its certain sub-band division signal turns 2 frequencys multiplication frequently larger (because during normal gear engagement than fundamental frequency amplitude, the amplitude of fundamental frequency will be higher than two frequencys multiplication).Some fault mode of gear, it is more difficult that the computing machine automatic decision implements.Because any one (or a plurality of) sub-band division signal all may be used to failure judgement, therefore need to be analyzed all sub-band division signals, then judge which subband signal the inside has fault characteristic information to exist.Often be present in different subband signal compositions for the combined failure feature, need especially each subband signal is done and analyzed and extract feature.In the present embodiment, for the Accurate Diagnosis gearbox fault, adopt dual-tree complex wavelet transform to carry out 3 layers of explication de texte to the gear case vibration signal.
The energy operator demodulation is a kind of modulation signal demodulation method, and the method principle is simple, easily realization, is very suitable for online fault diagnosis.But, for multicomponent signal, the method is difficult to extract a certain characteristic component.Therefore, the energy operator demodulating process is after dual-tree complex wavelet decomposes.Taking full advantage of on the good minute frequency band characteristic basis of dual-tree complex wavelet, carry out follow-up demodulation analysis.The all sub-band division signals that respectively decomposition obtained in the present embodiment carry out the energy operator demodulation process, to extract gear combined failure correlated characteristic; Above-mentioned energy operator means with ψ, to a sub-band division signal x
i(t) carrying out concrete demodulation method is
Restituted signal is carried out to analysis of spectrum, can extract correlated characteristic, such as shock characteristic frequency (judgement turns frequently corresponding with that root turning axle) and harmonic amplitude size (judgement has or not wearing and tearing) etc.
Any one sub-band division signal all may be used to failure judgement, so our exactly all sub-band division signals being analyzed of can doing, and then judges which subband signal the inside has fault characteristic information to exist.Often be present in different subband signal compositions for the combined failure feature, need especially each subband signal is done to analysis.After extracting gear combined failure correlated characteristic, can judge by artificial or computing machine the concrete fault mode of gear case.
Due to the complicacy of gearbox fault, so the present invention adopts method artificial and that computing machine matches to judge the concrete fault mode of gear case.At first, the gear combined failure correlated characteristic that extracts and pre-stored known gears combined failure feature in the property data base of computing machine are compared, determine thus the concrete fault mode of gear case to be measured.The known gears combined failure feature of above-mentioned characteristic library storage is to obtain through long-term experiment in early stage, the corresponding a kind of gearbox fault of each gear combined failure feature.In the later stage of gear case on-line monitoring process, above-mentioned these faults of online gear case vibration signal after treatment can be compared like this.If, when a certain known gears combined failure feature in the data characteristics database of the gear combined failure correlated characteristic extracted and computing machine is identical, can judge the fault type that this gear case to be measured occurs; And if all known gears combined failure features in the data characteristics database of the gear combined failure correlated characteristic extracted and computing machine are when all identical, the gear combined failure correlated characteristic that this can't be differentiated is stored, under storage can't automatic discrimination the concrete fault mode of gear case at first need to adopt artificial investigation mode could progressively go out to judge the concrete fault mode of gear case.Secondly, the gear combined failure correlated characteristic that can't differentiate for this, after the artificial investigation mode of employing has been determined the concrete fault mode of gear case, need the concrete fault mode extension storage of this gear combined failure correlated characteristic and corresponding gear case to the property data base of computing machine, with pattern and the type of the gear combined failure that expands Computer Automatic Recognition.
In the present embodiment, the dual-tree complex wavelet decomposition result is analyzed, detected and have or not the periodic shock feature; And adopt energy operator to carry out demodulation, detection of gear fault signature on the demodulation spectrogram.Fig. 6 is that dual-tree complex wavelet transform decomposes the ground floor detail signal, and as can be seen from the figure, the periodic shock signal that detail signal will be hidden in vibration signal clearly discloses out.The cycle that impact signal occurs is 0.121 second, and the frequency of occurrences is 8.25Hz, and the frequency of occurrences of impact signal is frequently consistent with turning of pinion wheel.Show thus, there are the local faults such as local fatigue is peeled off, crackle in gearbox pinion.Fig. 7 (a) and (b) be that dual-tree complex wavelet transform decomposes second layer approximation signal and energy operator demodulation spectrogram thereof.Can find out significantly in signal and have the which amplitude modulation phenomenon in approximation signal, approximation signal is carried out to the energy operator demodulation analysis, can find out that on the demodulation spectrogram its modulating frequency is 8.25Hz, and the higher hamonic wave of modulating frequency appearred, gearbox fault modulation source frequency turns 8.25Hz frequently with pinion wheel and equates, the which amplitude modulation phenomenon causes because there is the improper distributed fault of assembling in pinion wheel.Infer thus, gearbox pinion exists combined failure simultaneously: local fatigue is peeled off fault and is distributed the improper fault of assembling.
Shut down maintenance find the pinion wheel tooth root exist local fatigue peel off and the gear assembling improper, ressemble gear case, the start after vibration obviously reduces, Fig. 8 is maintenance rear gear box vibration signal.Vibration signal after maintenance is utilized to the dual-tree complex wavelet transform analysis, the ground floor detail signal that Fig. 9 is gear case vibration signal dual-tree complex wavelet transform, characterizing pinion wheel in detail signal exists the recurrent pulse of local fault feature still to exist, Figure 10 is the second layer approximation signal that gear case vibration signal dual-tree complex wavelet decomposes, and in approximation signal, modulation phenomenon disappears.Therefore, with traditional spectrum analysis, with cepstrum analysis, compare, associating dual-tree complex wavelet transform and energy operator demodulation method can be identified in gear case the local fault that caused by fatigue flake etc. and by the assembling distributed fault caused such as improper simultaneously.
Designed a kind of gear case combined failure diagnostic system according to said method, it mainly consists of acceleration vibration transducer, 1/4 sampling translation dual-tree complex wavelet transform module, energy operator demodulation process module, pattern recognition module and property data base.
The acceleration vibration transducer, be arranged on gearbox input shaft end cap to be measured, for picking up the gear case vibration signal.
1/4 sampling translation dual-tree complex wavelet transform module, comprise 2 two tree bank of filters arranged side by side, for the gearbox of wind turbine vibration signal to gathering, carries out 1/4 sampling translation dual-tree complex wavelet transform, obtains a plurality of sub-band division signals of signal.Above-mentioned each two tree bank of filters comprise ground floor analysis filter bank and follow-up high-rise analysis filter bank, this ground floor analysis filter bank is identical with the bank of filters that original dual-tree complex wavelet decomposes ground floor, and follow-up high level decomposes bank of filters used and all adopts and postpone 1/4 sampling structure, the later layer bank of filters postpones 1/4 sampling period on the basis of last layer bank of filters.In the present invention, the decomposition number of plies of described two tree bank of filters is between 3～5 layers.It is bi-orthogonal filter that the ground floor of wherein said two tree bank of filters decomposes the bank of filters of using.It is linear phase Q translate filter that the high level of described two tree bank of filters decomposes the bank of filters of using.
Energy operator demodulation process module; A plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic.
Pattern recognition module, in the gear combined failure correlated characteristic that energy operator demodulation process module is extracted and property data base, known gears combined failure feature compares, and identifies the concrete fault mode of gear case to be measured.
Above-described embodiment, it is only the specific case that purpose of the present invention, technical scheme and beneficial effect are further described, the present invention not is defined in this, in detecting with the gear case combined failure that can be applied in aerogenerator other field in addition as the present invention.All any modifications of making, be equal to replacement, improvement etc., within all being included in protection scope of the present invention within scope of disclosure of the present invention.
Claims (11)
1. a gear case combined failure diagnostic method, is characterized in that comprising the steps:
(1) adopt the acceleration vibration transducer to pick up the gear case vibration signal, this acceleration transducer is arranged on gearbox input shaft end cap to be measured;
(2) the gearbox of wind turbine vibration signal gathered is carried out to 1/4 sampling translation dual-tree complex wavelet transform, obtain a plurality of sub-band division signals of signal;
At first, construct 1/4 sampling translation dual-tree complex wavelet transform two tree bank of filters used, the bank of filters that the ground floor of this pair of tree bank of filters decomposes use is identical with the bank of filters that original dual-tree complex wavelet decomposes ground floor, and follow-up high level decomposes bank of filters used and all adopts and postpone 1/4 sampling structure, the later layer bank of filters postpones 1/4 sampling period on the basis of last layer bank of filters;
Secondly, utilize two bank of filters of setting of construct, adopt two wavelet transforms arranged side by side to the gear case vibration signal the picked up decomposition that walks abreast, obtain a plurality of sub-band division signals that gear case vibrates; Above-mentioned a plurality of sub-band division signal includes a low frequency sub-band decomposed signal and a series of high-frequency sub-band decomposed signal; Wherein the real part of each sub-band division signal and imaginary part consist of first wavelet transform and second wavelet transform respectively;
(3) a plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic and carry out identifying and diagnosing.
2. a kind of gear case combined failure diagnostic method according to claim 1 is characterized in that: it is bi-orthogonal filter that the ground floors of two tree bank of filters described in step (2) decompose the bank of filters of using.
3. a kind of gear case combined failure diagnostic method according to claim 1 is characterized in that: it is linear phase Q translate filter that the high levels of two tree bank of filters described in step (2) decompose the bank of filters of using.
4. according to the described a kind of gear case combined failure diagnostic method of any one in claim 1～3, it is characterized in that: described in step (2), the decomposition number of plies of described two tree bank of filters is between 3～5 layers.
5. a kind of gear case combined failure diagnostic method according to claim 1, it is characterized in that: described in step (3), the method for energy operator demodulation is:
In formula, x
i(t) mean a pending sub-band division signal, ψ means energy operator.
6. a kind of gear case combined failure diagnostic method according to claim 1 is characterized in that: adopt artificial and mode that computing machine combines in step (3) to the gear combined failure correlated characteristic that extracts and carry out identifying and diagnosing,
The gear combined failure correlated characteristic that extracts and pre-stored known gears combined failure feature in the property data base of computing machine are compared; When the gear combined failure correlated characteristic extracted is identical with known gears combined failure feature, Computer Automatic Recognition goes out the concrete fault mode of this gear case to be measured; When the gear combined failure correlated characteristic extracted is not identical with known gears combined failure feature, the gear combined failure correlated characteristic that computing machine can't be differentiated this is stored, under storage can't automatic discrimination the concrete fault mode of gear case need to adopt artificial investigation mode could progressively go out to judge the concrete fault mode of gear case.
7. a kind of gear case combined failure diagnostic method according to claim 6, it is characterized in that: in step (3), the gear combined failure correlated characteristic that can't differentiate for this, after the artificial investigation mode of employing has been determined the concrete fault mode of gear case, need the concrete fault mode extension storage of the gear case of this gear combined failure correlated characteristic and correspondence to the property data base of computing machine.
8. a gear case combined failure diagnostic system is characterized in that: it mainly consists of acceleration vibration transducer, 1/4 sampling translation dual-tree complex wavelet transform module, energy operator demodulation process module, pattern recognition module and property data base;
The acceleration vibration transducer, be arranged on gearbox input shaft end cap to be measured, for picking up the gear case vibration signal;
1/4 sampling translation dual-tree complex wavelet transform module, comprise 2 two tree bank of filters arranged side by side, for the gearbox of wind turbine vibration signal to gathering, carries out 1/4 sampling translation dual-tree complex wavelet transform, obtains a plurality of sub-band division signals of signal;
Above-mentioned each two tree bank of filters comprise ground floor analysis filter bank and follow-up high-rise analysis filter bank, this ground floor analysis filter bank is identical with the bank of filters that original dual-tree complex wavelet decomposes ground floor, and follow-up high level decomposes bank of filters used and all adopts and postpone 1/4 sampling structure, the later layer bank of filters postpones 1/4 sampling period on the basis of last layer bank of filters;
Energy operator demodulation process module; A plurality of sub-band division signals that one by one decomposition obtained carry out the energy operator demodulation process, extract gear combined failure correlated characteristic;
Pattern recognition module, in the gear combined failure correlated characteristic that energy operator demodulation process module is extracted and property data base, known gears combined failure feature compares, and identifies the concrete fault mode of gear case to be measured.
9. a kind of gear case combined failure diagnostic system according to claim 8 is characterized in that: it is bi-orthogonal filter that the ground floors of described two tree bank of filters decompose the bank of filters of using.
10. a kind of gear case combined failure diagnostic system according to claim 8 is characterized in that: it is linear phase Q translate filter that the high levels of described two tree bank of filters decompose the bank of filters of using.
11. the described a kind of gear case combined failure diagnostic system of any one according to Claim 8～10 is characterized in that: the decomposition number of plies of described two tree bank of filters is between 3～5 layers.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210315311.8A CN102937522B (en) | 2012-08-30 | 2012-08-30 | Composite fault diagnosis method and system of gear case |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210315311.8A CN102937522B (en) | 2012-08-30 | 2012-08-30 | Composite fault diagnosis method and system of gear case |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102937522A true CN102937522A (en) | 2013-02-20 |
| CN102937522B CN102937522B (en) | 2015-01-21 |
Family
ID=47696433
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210315311.8A Active CN102937522B (en) | 2012-08-30 | 2012-08-30 | Composite fault diagnosis method and system of gear case |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102937522B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353396A (en) * | 2013-06-24 | 2013-10-16 | 西安交通大学 | Gear case fault diagnosis method based on non-timescale short-time phase demodulation |
| CN103499437A (en) * | 2013-09-11 | 2014-01-08 | 西安交通大学 | Rotary machine fault detection method of dual-tree complex wavelet transformation with adjustable quality factors |
| CN104458250A (en) * | 2014-12-02 | 2015-03-25 | 上海电机学院 | Intelligent gearbox fault diagnosis method |
| CN106092564A (en) * | 2016-06-06 | 2016-11-09 | 电子科技大学 | The gear failure diagnosing method demodulated based on ESMD and energy operator |
| CN106092879A (en) * | 2016-06-07 | 2016-11-09 | 西安向阳航天材料股份有限公司 | Explosion clad pipe bonding state detection method based on vibratory response information |
| CN106127184A (en) * | 2016-07-05 | 2016-11-16 | 上海电机学院 | A kind of gear case of blower method for diagnosing faults |
| CN106404159A (en) * | 2016-06-14 | 2017-02-15 | 北京航天控制仪器研究所 | Continuous vibration transmission spectrum determination method of rocket sled test |
| CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
| CN107525671A (en) * | 2017-07-28 | 2017-12-29 | 中国科学院电工研究所 | A kind of wind-powered electricity generation driving-chain combined failure character separation and discrimination method |
| CN107907324A (en) * | 2017-10-17 | 2018-04-13 | 北京信息科技大学 | A kind of Fault Diagnosis of Gear Case method composed based on DTCWT and order |
| CN109106345A (en) * | 2018-06-27 | 2019-01-01 | 北京中欧美经济技术发展中心 | Pulse signal characteristic detection method and device |
| CN109374293A (en) * | 2018-10-29 | 2019-02-22 | 桂林电子科技大学 | A kind of gear failure diagnosing method |
| CN109765052A (en) * | 2019-01-21 | 2019-05-17 | 福州大学 | Epicyclic gearbox Incipient Fault Diagnosis method based on GOA-ASR |
| CN110333071A (en) * | 2019-06-28 | 2019-10-15 | 华北电力大学 | A kind of mechanical oscillation signal processing method using narrowband Cepstrum Transform |
| CN111256965A (en) * | 2020-01-20 | 2020-06-09 | 郑州轻工业大学 | Multi-scale information fusion stacked sparse self-coding rotary machine fault diagnosis method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
| CN101634605A (en) * | 2009-04-10 | 2010-01-27 | 北京工业大学 | Intelligent gearbox fault diagnosis method based on mixed inference and neural network |
| CN101738314A (en) * | 2009-12-14 | 2010-06-16 | 江苏省现代企业信息化应用支撑软件工程技术研发中心 | Antibody concentration-based gear failure diagnosing method |
| CN102570979A (en) * | 2011-12-20 | 2012-07-11 | 重庆大学 | Iterative Teager energy operator demodulation method and system |
-
2012
- 2012-08-30 CN CN201210315311.8A patent/CN102937522B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101634605A (en) * | 2009-04-10 | 2010-01-27 | 北京工业大学 | Intelligent gearbox fault diagnosis method based on mixed inference and neural network |
| CN101587017A (en) * | 2009-06-19 | 2009-11-25 | 湖南大学 | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum |
| CN101738314A (en) * | 2009-12-14 | 2010-06-16 | 江苏省现代企业信息化应用支撑软件工程技术研发中心 | Antibody concentration-based gear failure diagnosing method |
| CN102570979A (en) * | 2011-12-20 | 2012-07-11 | 重庆大学 | Iterative Teager energy operator demodulation method and system |
Non-Patent Citations (8)
| Title |
|---|
| KINGSBURY N G: "A dual-tree complex wavelet transform with improved orthogonality and symmetry properties", 《INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 * |
| SELESNICK I W 等: "《The dual-tree complex wavelet transform》", 《IEEE SIGNAL PROCESSING MAGAZINE》 * |
| 张德祥等: "《基于经验模式分解和Teager能量谱的齿轮箱故障诊断》", 《振动与冲击》 * |
| 王衍学: "《机械故障监测诊断的若干新方法及其应用研究》", 《中国博士学位论文全文数据库 工程科技II辑》 * |
| 田昊等: "《基于盲源分离的齿轮箱复合故障诊断研究》", 《兵工学报》 * |
| 艾树峰: "《基于双树复小波变换的轴承故障诊断研究》", 《中国机械工程》 * |
| 苏文胜等: "《双树复小波域隐Markov树模型降噪及在机械故障诊断中的应用》", 《振动与冲击》 * |
| 高立新等: "《基于CBR和RBR混合推理的齿轮箱智能诊断技术》", 《北京工业大学学报》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353396A (en) * | 2013-06-24 | 2013-10-16 | 西安交通大学 | Gear case fault diagnosis method based on non-timescale short-time phase demodulation |
| CN103353396B (en) * | 2013-06-24 | 2015-11-25 | 西安交通大学 | Based on the Fault Diagnosis of Gear Case method without markers phase demodulating in short-term |
| CN103499437A (en) * | 2013-09-11 | 2014-01-08 | 西安交通大学 | Rotary machine fault detection method of dual-tree complex wavelet transformation with adjustable quality factors |
| CN103499437B (en) * | 2013-09-11 | 2016-02-24 | 西安交通大学 | The rotating machinery fault detection method of adjustable quality factor dual-tree complex wavelet transform |
| CN104458250A (en) * | 2014-12-02 | 2015-03-25 | 上海电机学院 | Intelligent gearbox fault diagnosis method |
| CN106092564A (en) * | 2016-06-06 | 2016-11-09 | 电子科技大学 | The gear failure diagnosing method demodulated based on ESMD and energy operator |
| CN106092879A (en) * | 2016-06-07 | 2016-11-09 | 西安向阳航天材料股份有限公司 | Explosion clad pipe bonding state detection method based on vibratory response information |
| CN106092879B (en) * | 2016-06-07 | 2019-07-12 | 西安向阳航天材料股份有限公司 | Explosion clad pipe bonding state detection method based on vibratory response information |
| CN106404159A (en) * | 2016-06-14 | 2017-02-15 | 北京航天控制仪器研究所 | Continuous vibration transmission spectrum determination method of rocket sled test |
| CN106404159B (en) * | 2016-06-14 | 2019-03-12 | 北京航天控制仪器研究所 | A kind of determining method of Rocket sled test continuous shaking transmitting spectrum |
| CN106127184A (en) * | 2016-07-05 | 2016-11-16 | 上海电机学院 | A kind of gear case of blower method for diagnosing faults |
| CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
| CN106874833B (en) * | 2016-12-26 | 2021-05-28 | 中国船舶重工集团公司第七一0研究所 | Vibration event pattern recognition method |
| CN107525671A (en) * | 2017-07-28 | 2017-12-29 | 中国科学院电工研究所 | A kind of wind-powered electricity generation driving-chain combined failure character separation and discrimination method |
| CN107907324A (en) * | 2017-10-17 | 2018-04-13 | 北京信息科技大学 | A kind of Fault Diagnosis of Gear Case method composed based on DTCWT and order |
| CN109106345A (en) * | 2018-06-27 | 2019-01-01 | 北京中欧美经济技术发展中心 | Pulse signal characteristic detection method and device |
| CN109374293A (en) * | 2018-10-29 | 2019-02-22 | 桂林电子科技大学 | A kind of gear failure diagnosing method |
| CN109374293B (en) * | 2018-10-29 | 2020-07-24 | 珠海市华星装备信息科技有限公司 | Gear fault diagnosis method |
| CN109765052A (en) * | 2019-01-21 | 2019-05-17 | 福州大学 | Epicyclic gearbox Incipient Fault Diagnosis method based on GOA-ASR |
| CN110333071A (en) * | 2019-06-28 | 2019-10-15 | 华北电力大学 | A kind of mechanical oscillation signal processing method using narrowband Cepstrum Transform |
| CN110333071B (en) * | 2019-06-28 | 2021-09-10 | 华北电力大学 | Mechanical vibration signal processing method using narrowband cepstrum transformation |
| CN111256965A (en) * | 2020-01-20 | 2020-06-09 | 郑州轻工业大学 | Multi-scale information fusion stacked sparse self-coding rotary machine fault diagnosis method |
| CN111256965B (en) * | 2020-01-20 | 2022-03-11 | 郑州轻工业大学 | Multi-scale information fusion stacked sparse self-coding rotary machine fault diagnosis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102937522B (en) | 2015-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102937522B (en) | Composite fault diagnosis method and system of gear case | |
| CN101435799B (en) | Failure diagnosis method and apparatus of hydroturbine based on acoustic emission technology | |
| CN102798529B (en) | Method and system for diagnosing bearing faults of large-size wind turbine bearing | |
| Bajric et al. | Feature extraction using discrete wavelet transform for gear fault diagnosis of wind turbine gearbox | |
| Hu et al. | A new wind turbine fault diagnosis method based on ensemble intrinsic time-scale decomposition and WPT-fractal dimension | |
| CN103499437B (en) | The rotating machinery fault detection method of adjustable quality factor dual-tree complex wavelet transform | |
| CN102269655B (en) | Method for diagnosing bearing fault | |
| Yan et al. | Harmonic wavelet-based data filtering for enhanced machine defect identification | |
| CN102243140B (en) | Mechanical equipment state monitoring method based on sub-band signal analysis | |
| CN104677623B (en) | A kind of blade of wind-driven generator fault acoustics in place diagnostic method and monitoring system | |
| Strömbergsson et al. | Bearing monitoring in the wind turbine drivetrain: A comparative study of the FFT and wavelet transforms | |
| CN106124982A (en) | Automatic expert's resultant fault diagnostic system of a kind of Wind turbines and diagnostic method | |
| CN105928702B (en) | Variable working condition box bearing method for diagnosing faults based on form PCA | |
| CN105760839A (en) | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine | |
| CN101587017A (en) | Gear fault diagnosis method based on part mean decomposition cycle frequency spectrum | |
| CN107796611A (en) | A kind of warning system for being used to detect wind-driven generator abnormal work | |
| CN103439110A (en) | Early-stage weak fault diagnosis method of rolling bearing | |
| CN105784353A (en) | Fault diagnosis method for gear case of aerogenerator | |
| CN110333071B (en) | Mechanical vibration signal processing method using narrowband cepstrum transformation | |
| CN110514437A (en) | A kind of epicyclic gearbox vibration signal separation method and system for fault diagnosis | |
| CN105445022A (en) | Planetary gear fault diagnosis method based on dual-tree complex wavelet transform-entropy feature fusion | |
| CN102507186A (en) | Characteristic parameter-based method for condition monitoring and fault identification of planetary gearbox | |
| CN106055734A (en) | Method for failure diagnosis of wind power gearbox | |
| CN103234748A (en) | Klingelnberg bevel gear fault diagnosis method based on sensitive IMF (instinct mode function) components | |
| CN113806893A (en) | Fan state monitoring and fault diagnosis method and system based on industrial internet |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |