CN105806613A - Planetary gear case fault diagnosis method based on order complexity - Google Patents

Abstract

The invention belongs to the technical field of rotation machinery fault diagnosis, especially relates to a planetary gear case fault diagnosis method based on order complexity, and is specially applied to a fault diagnosis method of a wind power planetary gear case under a variable working condition. The method comprises the following steps: preprocessing vibration signals; dividing faults of a planetary gear train into two kinds; obtaining a fault feature parameter; analyzing the vibration signals under a variable working condition of a planetary gear case; and performing instance verification discovery. According to the invention, the influence of the variable working condition can be avoided by use of an order analysis method, and the Lempel-Ziv complexity of angle-domain signals after order analysis processing is more stationary than that of time-domain signals. According to the Lempel-Ziv complexity of the vibration angle-domain signals of the planetary gear case, local faults and distribution faults of planetary gears can be identified, through simulation and analysis of actual cases, the effectiveness of the method is verified, and at the same time, under the condition of lack of fault data, the complexity can also be taken as the feature parameter for determining whether there is a fault or not.

Description

A kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity

Technical field

The invention belongs to rotary machinery fault diagnosis technical field, particularly relate to a kind of based on rank than the planet tooth of complexity Roller box method for diagnosing faults, the method for diagnosing faults of the wind power planetary gear case being particularly well-suited under variable working condition.

Background technology

Epicyclic gearbox has that gear ratio is big, bearing capacity strong, transmission efficiency high, is widely used in wind-force and sends out In the large complicated plant equipment such as electricity, helicopter.Severe working environment, usually causes the fault of critical component in gear-box to send out Raw.For the fault diagnosis research of fixed axis gear case achieved with preliminary effect, but epicyclic gearbox is different from fixed center The conventional gearbox that axle rotates, the gear movement in epicyclic gearbox is typical compound motion, and its vibratory response ratio is traditional Gear-box is increasingly complex, and it is bigger that the diagnosis of fault and identification difficulty compare conventional gearbox, conventional gearbox method for diagnosing faults It is difficult to directly indiscriminately imitate the fault diagnosis being applied to epicyclic gearbox.

Vibration signal is mainly analyzed by the fault diagnosis of gear-box, and vibration signal can describe gear vibration well Fault message, but owing to gear-box is under the running status of variable working condition, be directly analyzed existing certain to vibration signal Error.On the basis of the analysis of complexity of conventional gearbox and the vibration signal of bearing, for epicyclic gearbox variable working condition Service condition, the present invention uses order ratio analysis and EMD empirical mode decomposition method that vibration signal carries out pretreatment, analyzes row Complexity under star gear-box difference running status, and then distinguish its malfunction.By the vibration to wind energy turbine set fault unit Signal is analyzed, and demonstrates effectiveness of the invention.

Summary of the invention

For the deficiencies in the prior art, the present invention provides a kind of variable working condition epicyclic gearbox Analysis on Fault Diagnosis method.

The technical scheme is that and be achieved in that:

A kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, comprise the steps:

(1) vibration signal pretreatment: according to Computed order tracking technology, will be in the wind power planetary gear case sensor of variable working condition The vibration signal gathered carries out pretreatment, and time-domain signal non-linear, non-stationary is converted into the angular domain letter with stationarity Number, it is to avoid use hardware mode to realize the expensive cost of equiangular sampling；During nonstationary vibration based on linear interpolation method The nonstationary vibration time-domain signal that constant duration is sampled, than reconfiguration technique, is converted into and has smooth performance by the rank of territory signal Angular domain vibration signal, it is ensured that property complete cycle of epicyclic gearbox vibration angular domain signal, it is to avoid the shadow of variable parameter operation environment Ring；

(2) fault of planet circular system is divided into two classes: distributed faults and local fault；To the distributed faults of planet circular system and The calculating of the characteristic frequency of local fault is analyzed；With magnificent sharp SL1500 wind power planetary gear case as object of study, to two-stage Planet circular system, the vibration signal of one-level parallel stage epicyclic gearbox emulate；

(3) Fault characteristic parameters: Lempel-Ziv complexity is in rolling bearing and the fault diagnosis of simple parallel level gear In it has been shown that the concept of Lempel-Ziv complexity can be carried out point as the Fault characteristic parameters of bearing failure diagnosis Analysis, further confirms that Lempel-Ziv complexity can be as the characteristic parameter of fault diagnosis；

(4) vibration signal under planetary planet gear case variable working condition is analyzed, finds Non-stationary vibration signal Lempel-Ziv complexity can fluctuate, but the angular domain signal after the proportion sampling processing of rank can effectively prevent and exchange work The condition impact on Lempel-Ziv complexity；Thus can verify, can be as fault diagnosis characteristic parameter under variable working condition； Lempel-Ziv complexity can be as the characteristic parameter of the state recognition of the epicyclic gearbox under bad working environments environment；In step (1), on the basis of (2), use rank than the complexity epicyclic gearbox diagnosing malfunction to labyrinth, calculate different faults Rank under pattern, than complexity, by observing the situation of change than complexity of the rank corresponding to each fault, find that complexity is compared on rank Can be as the characteristic parameter of epicyclic gearbox Fault Pattern Recognition；This step utilize rank than multiplicity, by wind-powered electricity generation planet tooth The fault signature rank that roller box is under different conditions calculate than complexity, it is achieved the identification of epicyclic gearbox fault mode；

(5) case verification finds, epicyclic gearbox rank under different faults pattern are relatively bigger than complexity difference, Ke Yizuo For the evaluation of epicyclic gearbox state, in the case of lacking sufficient fault data, it is also possible to as gear-box health status Assessment；Through this step case verification, utilize rank than multiplicity, wind power planetary gear case malfunction can be identified, lacking In the case of few fault data, it is also possible to as the assessment parameter of epicyclic gearbox health status.

Vibration signal pretreatment described in step (1): be to gear planetary wheel case vibration signal pretreatment；At epicyclic gearbox Under the working environment of variable working condition, its vibration signal gathered is non-stationary signal, and the fault diagnosis of gear-box is produced very by this Big difficulty, uses rank than the method for resampling, vibration signal to be carried out angular domain resampling, by the time domain vibration signal of non-stationary It is converted into the angular domain signal of steady or quasi-smooth performance, it is simple to the analysis to vibration signal；The core of order ratio analysis technology is Obtain the constant angle increment sampled data of relative reference axis, it is therefore desirable to can accurately obtain the moment of order sampling and corresponding base Quasi-rotating speed, i.e. realizes order tracking technique；Common Computed order tracking method have hardware order tracking technique method, calculate Computed order tracking method and based on The Computed order tracking method of instantaneous Frequency Estimation；The resampling using calculating Computed order tracking method to realize vibration signal calculates, by heavily adopting The vibration signal relevant with rotating speed produced during mechanical speed change can be efficiently separated out, simultaneously to unrelated with rotating speed by sample Signal play certain inhibitory action, redistribute the sampling interval of signal by angle position, rotation speed change pair can be rejected The impact of signal randomness；The most all containing multiple interference component in actual gear-box vibration signal, this allows for The extraction of its fault signature becomes relatively difficult；EMD method can be adaptive by arbitrarily according to the local time-varying characteristics of signal One sophisticated signal is decomposed into a series of component, is reconstructed signal by correlation coefficient rule, rejects in primary signal Interference component；Use order ratio analysis and EMD decomposition method effectively vibration signal can be carried out pretreatment, the letter to next step Number analysis is got ready；

Due to the coupling between the complexity of operating condition and parts, the vibration data of Wind turbines has obvious non-thread Property and non-stationary property, it is therefore necessary to use vibration data preconditioning technique unit vibration data are carried out preposition process；This Bright based on rank than the wind generating set vibration data processing method of resampling technique, the method can efficiently solve Wind turbines and shake The variable speed problem of dynamic signal, can be converted into the vibration time-domain signal with non-stationary characteristic and have shaking of smooth performance Dynamic angular domain signal, its ultimate principle is angularly to adopt using software algorithm to be converted into by the time-domain signal that constant duration is sampled Angular domain vibration signal under original mold formula, the angular domain vibration signal after conversion i.e. ensure that the verity of original time domain signal, has tool Having had characteristic positive period, the vibrational feature extracting extracted on the basis of angular domain vibration signal can reflect wind-powered electricity generation accurately and effectively Malfunction under the conditions of unit variable parameter operation；

Its rank are more as follows than the key step of resampling technique:

1. assume that the vibration data that vibration measuring point obtains is { x₁,x₂,x₃,…x_n-2,x_n-1,x_n, first by between such time Every the time-domain signal that sampling obtains, it is converted into the angular domain signal of constant duration sampling, during this calculating, assumes rotating speed in short-term In comprised according to the mode speed change of constant angular, the vibration period that when using constant duration sampling, frequency of vibration is minimum Sampling number the most, for ensure reconstruction signal will not distortion, using counting as interpolation of comprising of lowest vibration frequency cycle Standard value reference value, rotate a circle when this reference value is converted into equiangular sampling counting of being comprised；This is counted and is isogonism During degree sampling, rotor every turn should gather volume and counts, rotating speed the fastest every turn gather count the fewest, therefore rotating speed is with reference to should be with the highest Rotating speed is reference；Assume that the low-limit frequency composition vibrating measuring point is f₀, gear box ratio is the n (gear ratio of three gear stages It is respectively n₁,n₂,n₃), according to the maximum speed requirement of generator side, during normal power generation, the range of speeds of main shaft is n_min～n_max, Then using high speed shaft rotating speed as reference rotation velocity, its value is n_ck=n₁×n₂×n₃×n_max, it is calculated the rank angular domain than resampling The interpolation of signal reconstruction is with reference to counting as n=n_ck*f_o/ 60, every turn of point gathered when i.e. representing employing equiangular sampling mode Number should be n, it is assumed that comprise in the analysis of vibration signal cycle always counts as N, then the angular domain sequence redefined is: Theta (N)=0:2 π/n:2 π (N-1)/n；Use linear interpolation method that Theta (N) inserts the angular domain letter under constant duration sampling In number, it is thus achieved that the vibration angular domain signal under equiangular sampling mode；

2. the vibration time-domain signal of constant duration collection is converted into corresponding angular domain vibration signal, due to Wind turbines The function of pitch-controlled system, Wind turbines rotation speed change is relatively mild, can false wind group of motors rotating speed be equal angular acceleration Speed change, wherein generating unit speed signal can obtain from generator speed encoder, it is therefore assumed that the unit in the time of each second turns Speed is equal angular acceleration change；If the speed of mainshaft data that vibration measuring point obtained in ten minutes are n_a, within lower ten minutes, obtain Speed of mainshaft data be n_b, corresponding angular velocity is respectively ω_bAnd ω_a, therefore the set drive chain vibration in this time period is surveyed Point reference rotation velocity change curve is expressed as:

(ω_b-ω_a)·t+ω_a (1)

This angular velocity curve is integrated in time scale obtain angle formula:

$\mathrm{\θ}={\∫}_{t_1}^{t_2}\[\left({\mathrm{\ω}}_b-{\mathrm{\ω}}_a\right)\·t+{\mathrm{\ω}}_a\]dt---\left(2\right)$

By formula (1), constant duration sampling is obtained time domain vibration signal and be converted into the angular domain signal of correspondence；

3. the method using linear interpolation carries out interpolation to the angular domain signal of constant duration, and its per revolution is gathered Count step 1. in calculated acquisition；The angle of constant duration acquisition angle territory signal is counted according to interpolation and carries out drawing Point, try to achieve the angle coordinate { θ of the point needing to carry out interpolation₁,θ₂,…,θ_n-1,θ_n, these points are sampled angle at constant duration Territory signal enterprising line linearity interpolation, it is thus achieved that { x_θ1,x_θ2,…x_θn-1,x_θn, by formula [x_t(i-1)-x_θ(i)]×[x_t(i)-x_θ(i)] ≤ 0 search is positioned at the actual value of interpolation point both sides, by linear interpolation formula

$x_{\mathrm{\θ}\left(i\right)}=\frac{x_{t\left(i\right)}\·\left(\mathrm{\θ}\right(i)-t(i-1))}{t\left(i\right)-t\left(i-1\right)}---\left(3\right)$

The vertical coordinate of coordinate points is inserted, it is thus achieved that angularly angular domain signal { x after calculating linear interpolation_θ1,x_θ2,…x_θn-1, x_θn}；So far we achieve constant duration sample time domain signal to equiangular sampling angular domain letter by the 1-3 step in step 2 Number conversion, the angular domain signal after conversion has smooth performance and characteristic complete cycle；

The most achieve the non-stationary time domain vibration signal under the conditions of variable working condition to flat by rank than resampling technique The conversion of steady angular domain vibration signal, vibration signal now has had characteristic complete cycle, to the angular domain with characteristic complete cycle Vibration signal extracts fault pre-alarming and whether diagnosis index can occur or development degree by faults accurately and efficiently；But want It is noted that in angular domain sampling, with equal angle sampling frequency S_fCorresponding for order sample frequency S_o(Sampling Frequency of Order tracking), i.e. the reference axis angularly data gathered that often rotate a circle are counted, in order to protect The integrity of information in card primary signal, order is sampled as time-domain sampling, is required for meeting sampling thheorem: order sampling is frequently Rate have to be larger than the highest order composition O_maxTwice, i.e.

S_o＞ O_max(4)；

The spectrogram obtained after angularly signal makees Fourier transformation to resampling is order spectrum；Equiangular sampling is utilized to believe Number and order spectrum carry out each parameter in the analysis of angular domain signal and the time domain of signal, frequency-domain analysis and have and close one to one System；

Vibration angular domain signal after resampling usually comprises noise, in order to reduce the interference of this composition, select herein based on EMD method carries out noise reduction process, and the method can reduce low-frequency disturbance effectively highlight high frequency intrinsic vibration with cancelling noise, it is easy to Fault characteristic frequency is extracted in sophisticated signal；It is a kind of adaptive signal decomposition method that EMD decomposes, and its advantage shows themselves in that (1) basic function automatically generates: can draw from EMD catabolic process, and the method, according to the feature of primary signal, adaptive is chosen Optimum basic function, and need to be pre-selected basic function by wavelet decomposition, the first selection of basic function is cumbersome, catabolic process Middle basic function is once selected can not be changed；(2) there is adaptive filtering characteristic；(3) adaptive multiresolution；EMD divides IMF after solution contains frequency content from high to low；Owing to EMD catabolic process existing end effect, interpolation error, excessively The situations such as decomposition, cause decomposition result to there may be pseudo-component, and primary signal are unrelated；May be containing with fault frequently in pseudo-component The frequency that rate overlaps, it should the pseudo-component rejection trying every possible means to impact these falls；Can be had by cross-correlation coefficient criterion Effect identifies pseudo-component, it is achieved method is the cross-correlation coefficient calculating IMF component with primary signal, be can recognize that by the size of coefficient Pseudo-component；

Component with the cross-correlation coefficient of primary signal S is:

${\mathrm{\ρ}}_{s,{\hat{c}}_j}=max\left(R_{s,{\hat{c}}_j}\right(\mathrm{\τ}))/max\left(R_s\right(\mathrm{\τ}\left)\right)---(5);$

Wherein, R_s(τ) it is the autocorrelation coefficient of original signal；

Pseudo-component can be rejected by cross-correlation coefficient criterion, then be reconstructed signal, reduce the interference of noise；Through rank It is angular domain stationary signal than the signal after resampling and EMD noise reduction process.

The distributed faults to planet circular system and the calculating of the characteristic frequency of local fault described in step (2) are analyzed:

Wind turbines gear speedup case various structures, gear ratio is big, for reducing the size of gear-box, generally planetary gear Structure, is analyzed a certain Wind turbines epicyclic gearbox here, its wind power planetary gear box structure be two-stage planetary gear, one Level parallel gears structure；

The gear for two stage planetary gear train of Wind turbines epicyclic gearbox and parallel stage gear structure；

Epicyclic gearbox is different from traditional gear-box, and its structure is made up of sun gear, planetary gear, gear ring and planet carrier； Generally, gear ring maintains static, and sun gear rotates around the central axis of self, and planetary gear not only rotation is also about the sun Wheel revolution；Planetary gear both engaged with sun gear, engaged with gear ring again；In epicyclic gearbox, the compound motion of multiple gears causes The complexity of vibration signal；In the monitoring of vibration, sensor is typically mounted on gear ring or the casing that is attached thereto to gather and shakes Dynamic signal, the meshing point of sun gear-planetary gear and planetary gear-gear ring Meshing Pair rotates change to the position of sensor with planet carrier Change so that the vibration transfer path between meshing point to sensor changes, and the bang path of this time-varying is to vibration-testing Signal produces amplitude-modulating modulation effect；

Gear distress is generally divided into two classes: distributed fault and local fault；Planet circular system generation different faults type and During abort situation, its vibration signal model and fault characteristic frequency can be different, and vibration signal when there is fault is carried out Analysis can realize the diagnosis to fault；

Wind power planetary gear case is generally divided into planet circular system, parallel stage gear, for planet circular system and parallel stage gear, its Fault can be divided into local fault and distributed fault；In single-pinion planetary gear case, sun gear planetary gear and planetary gear gear ring The meshing frequency of two kinds of Meshing Pair is identical；Generally gear ring maintains static, and sun gear, planetary gear and planet carrier rotate, in this feelings Under condition, meshing frequency:

f_m=f_cZ_r=(f_s ^(r)-f_c)Z_s(6)；

In formula: Z_rAnd Z_sIt is respectively gear ring and the number of teeth of sun gear；f_mFor meshing frequency；f_cSpeed for planet carrier； f_s ^(r)Absolute speed for sun gear；

Too star-wheel local fault characteristic frequency is:

$f_s=\frac{f_m}{Z_s}N---\left(7\right);$

In formula: f_mFor meshing frequency；Z_sFor the sun gear number of teeth；N is planetary gear quantity；

Planetary gear local fault characteristic frequency is:

$f_p=2\frac{f_m}{Z_p}---\left(8\right);$

In formula: f_mFor meshing frequency；Z_pFor the planetary gear number of teeth；Gear ring local fault characteristic frequency is:

$f_r=\frac{f_m}{Z_r}N---\left(9\right);$

In formula: f_mFor meshing frequency；Z_rFor the gear ring number of teeth；

In epicyclic gearbox, the distributed fault characteristic frequency of various gears is equal to gear opposing rows carrier (sun gear and tooth Circle fault) or the speed of gear ring (planetary gear fault)；The meshing frequency f of known epicyclic gearbox_mTooth with certain gear Number Z_g, then this gear opposing rows carrier (sun gear and gear ring fault) or the speed of gear ring (planetary gear fault):

f_g=f_g/Z_g(10)；

Then the characteristic frequency of sun gear, planetary gear and gear ring distributed fault is respectively as follows:

f_s'=f_m/Z_s(11)；

f_p'=f_m/Z_p(12)；

f_r'=f_m/Z_r(13)；

In formula: f_mFor meshing frequency；f_s'、f_p'、f_r'Sun gear, planetary gear and the characteristic frequency of gear ring distributed fault；Z_s For the sun gear number of teeth；Z_pThe number of teeth for planetary gear；Z_rFor the gear ring number of teeth；

With the planet carrier of primary planet train that is connected with main shaft as reference rotation velocity, at different levels each in epicyclic gearbox The local fault of individual gear and the feature rank ratio of distributed faults calculate；

Here signal is that matlab emulates signal, ignores the shadow between gear-box middle (center) bearing and each gear in signal Ring, it is assumed that the vibration effect between each gearbox drive level does not exists, the method is carried out simulating, verifying for planet circular system at Vibration signal model list of references when normal, distributed faults, local fault, the most just repeats no more；

Thus, epicyclic gearbox is at different levels in emulation when being in nominal situation, and its vibration signal model is:

x₁(t)=A [1-cos (2 π 3f_c1·t)]cos(2π·f_m1·t+θ₁)

+B·[1-cos(2π·3f_c2·t)]·cos(2π·f_m2·t+θ₂)+C·cos(2π·f_m3·t+θ₃) (14)

During epicyclic gearbox primary planet train sun gear generation local fault, its vibration signal model is:

During epicyclic gearbox primary planet train sun gear occurrence and distribution fault, its vibration signal model is:

In formula (9)～(11): x₁(t)、x₂(t)、x₃T () is that epicyclic gearbox is in normal primary planet train sun gear There is vibration signal sequence when local fault, distributed faults；T is time series；θ₁、θ₂、θ₃、φ、For initial phase；f_m1、 f_m2、f_m3For meshing frequencies at different levels；f_c1、f_c2Speed for I and II planet carrier；Absolute rotation for one-level sun gear Turn frequency；f_s1、f_s1'For characteristic frequency when one-level sun gear generation local fault and distributed faults；A, B, C are immeasurable three cardinal guides Number, each state duration of epicyclic gearbox can be different, the most no longer describes in detail；Each vibration signal employing frequency is 8192HZ。

Fault characteristic parameters described in step (3): often make its frequency form when epicyclic gearbox breaks down and each Becoming framing, frequently characteristic to change, so that time domain waveform occurs distortion in various degree, this distortion is also epicyclic gearbox The gear distress directly reflection in time-domain signal；When the waveform of signal, spectrum structure change, the complexity of its signal Also change；Jiang Jiandong Lempel-Ziv Complexity Measurement qualitative assessment large-sized unit running status；Hong have studied Lempel-Ziv complexity index is used for assessing faulty bearings faulted condition；Practice have shown that, this index is to weigh finite time The efficient tool of sequence complexity；

The Lempel-Ziv complexity of sequence can obtain by through time circulation, calculates process as follows:

(1) S is initialized_v,0={ }, Q₀={ }, C_N(0)=0, r=1；OrderDue to Q_rIt is not belonging to S_v,r-1, Then C_N(r)=C_N(r-1), Q_r={ }, r=r+1；

(2) orderJudge Q_rWhether belong toIf belonging to, then C_N(r)=C_N(r-1), r =r+1, repetitive process (2)；

(3) if being not belonging to, then C_N(r)=C_N(r-1)+1, Q_r={ }, r=r+1, repetitive process (2)；

The such as segmented version of sequence { 0000 } is { 0 000 }, C_N=2；Symbol sebolic addressing { 010101 } segmented version is { 01 0101 }, C_NIt is just 3；Examine for the comparability strict to data Considering, Lempel and Ziv further provides a kind of normalization reference formula, value is defined between [0-1], and detailed algorithm is:

$0\≤C_{nN}=\frac{C_N\left(N\right)}{C_{UL,N}}\≤1---\left(12\right)$

$C_{UL,N}=\underset{N\→\mathrm{\∞}}{\lim }{C}_N\left(N\right)=\underset{N\→\mathrm{\∞}}{\lim }\frac{N}{(1-\mathrm{\β}){\log }_{k}N}\≈\frac{N}{{\log }_{k}N}---\left(13\right)$

Wherein: k is middle S_NThe number of element is (for binary system S_NSequence, k=2)；According to Lempel-Ziv product complexity theory Understanding with document, the complexity of a sequence is the biggest, illustrates that its interpolation operation is the most, and the intension component of sequence is the most, retouches Needed for stating given symbol sebolic addressing minimum, mutually different segmented version is the most, given sequence periodicity is the most weak, occur new in The speed of culvert pattern is the fastest；The physical significance of Lempel-Ziv complexity is that it reflects a time series along with length There is the speed of new model in the increase of degree；Complexity is the biggest, and the new change that data occur within length of window period in time is described Changing the most, the speed that new change occurs is the fastest, shows that the data variation in this period is unordered and complexity；Otherwise complexity is more Little, then the speed of explanation generation new change is the slowest, and data variation is regular, the strongest；Therefore, vibration signal Lempel-Ziv complexity index can objectively respond out the situation that system mode changes.

The identification realizing epicyclic gearbox fault mode described in step (4): under epicyclic gearbox normal condition Variable speed vibration signal is analyzed, a length of 20s of vibration signal, and its vibration signal is variable speed signal, the width of vibration signal Value decreases with the reduction of rotating speed with frequency；According to variable speed vibration signal through rank than resampling and EMD decomposition method Angular domain stationary signal after process, draws, angular domain signal is more steady compared to time-domain signal, and the sampling interval of signal more advises Rule；

The complexity that epicyclic gearbox vibrates time-domain signal and angular domain signal calculates；Time-domain signal calculates Length be respectively the signal length of 3s and 1.5s, the length calculating angular domain signal is respectively the signal length of sum, logical Cross calculate unlike signal length Lempel-Ziv complexity numerical value it can be seen that the complexity numerical fluctuations of time-domain signal relatively Greatly, the most steadily, and the complexity numerical value of angular domain signal reaches unanimity, more stable, illustrates, the complexity of angular domain signal is compared The running status of epicyclic gearbox more can be reflected in time-domain signal；

By order ratio analysis method, the steady angular domain signal under epicyclic gearbox difference running status is carried out Lempel- The calculating of Ziv complexity, is corresponding in turn to that epicyclic gearbox is normal, gear ring distributed faults, planetary gear distributed faults, too from left to right Sun wheel distributed faults, gear ring local fault, planetary gear local fault, the Lempel-Ziv complexity change of sun gear local fault Trend, draws concrete numerical value,

As can be seen from the results, the vibration of epicyclic gearbox normal condition is the simplest, and when local fault occurs, it is non- Linearly compare bigger, compared to distributed faults, there is higher Lempel-Ziv value；By Lempel-Ziv complexity numerical value, energy Enough running statuses to epicyclic gearbox effectively identify.

Case verification described in step (5) finds: carry out certain sharp SL1500 wind turbine gearbox vibration data of wind energy turbine set China Analyzing, the employing frequency of vibration signal is 5120Hz, and signal length is 6s, and three segment signals are respectively gear-box normal condition, tooth Vibration signal under box high speed level gear wear condition and gearbox high-speed level gear tooth breakage state, each section of vibration signal is in Different rotating speeds, time-domain signal is steady not, by vibration time-domain signal is carried out rank than resampling and the pre-place of EMD decomposition method Reason, calculates the Lempel-Ziv complexity of signal under three states；

Owing to lacking the service data under each running status of gear-box, it is difficult to each malfunction of gear-box is known Not, but Lempel-Ziv numerical value from above three running statuses it can be seen that, Lempel-Ziv complexity can conduct Characterize the characteristic parameter of running state of gear box；Simultaneously in the case of planet gearbox fault data abundance, can be to planet The fault of gear divides, it is achieved the interval division to the Lempel-Ziv complexity under each fault mode.

Advantages of the present invention and having the beneficial effect that:

The present invention uses order ratio analysis method can avoid the impact of variable working condition, the angular domain letter after order ratio analysis processes Number Lempel-Ziv complexity relatively time-domain signal more steady, Lempel-Ziv complexity index is used for evaluating different faults shape The change of state dynamical system, it is a kind of method of speed characterizing and occurring new model in time series, this method at first by Lempel and Ziv proposes, and is therefore named as Lempel-Ziv complexity, and it is affected less by variable working condition, can avoid exchanging work The impact of condition.The Lempel-Ziv complexity of epicyclic gearbox vibration angular domain signal can be to planetary gear local fault and distribution Fault is identified, and by emulation and the analysis of real case, demonstrates effectiveness of the invention, is lacking fault data simultaneously In the case of the characteristic parameter that whether can also occur as fault.

Accompanying drawing explanation

Fig. 1 vibration data rank proportion sampling flow chart

Fig. 2 is planetary gear box structure schematic drawing in the present invention；

Fig. 3 is epicyclic gearbox tach signal and vibration time-domain signal in the present invention；

Fig. 4 is epicyclic gearbox vibration angular domain signal in the present invention；

Fig. 5 is the Lempel-Ziv complexity numerical value of time-domain signal and angular domain signal under normal condition in the present invention；

Fig. 6 is the Lempel-Ziv change in value trend of different conditions in the present invention；

Fig. 7 is the vibration time-domain signal of three running statuses of epicyclic gearbox in the present invention；

Fig. 8 is epicyclic gearbox fault diagnosis flow scheme based on Lempel-Ziv complexity in the present invention.

Detailed description of the invention

The present invention be a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, comprise the steps:

(1) vibration signal pretreatment.According to Computed order tracking technology, the wind power planetary gear case sensor of variable working condition will be in The vibration signal gathered carries out pretreatment, and time-domain signal non-linear, non-stationary is converted into the angular domain letter with stationarity Number, it is to avoid use hardware mode to realize the expensive cost of equiangular sampling.During nonstationary vibration based on linear interpolation method The nonstationary vibration time-domain signal that constant duration is sampled, than reconfiguration technique, is converted into and has smooth performance by the rank of territory signal Angular domain vibration signal, it is ensured that property complete cycle of epicyclic gearbox vibration angular domain signal, it is to avoid the shadow of variable parameter operation environment Ring.

(2) fault of planet circular system is divided into two classes: distributed faults and local fault.To the distributed faults of planet circular system and The calculating of the characteristic frequency of local fault is analyzed.With magnificent sharp SL1500 wind power planetary gear case as object of study, to two-stage Planet circular system, the vibration signal of one-level parallel stage epicyclic gearbox emulate.

(3) Fault characteristic parameters.Lempel-Ziv complexity is in rolling bearing and the fault diagnosis of simple parallel level gear In it has been shown that the concept of Lempel-Ziv complexity can be carried out point as the Fault characteristic parameters of bearing failure diagnosis Analysis, further confirms that Lempel-Ziv complexity can be as the characteristic parameter of fault diagnosis.

(4) vibration signal under planetary planet gear case variable working condition is analyzed, finds Non-stationary vibration signal Lempel-Ziv complexity can fluctuate, but the angular domain signal after the proportion sampling processing of rank can effectively prevent and exchange work The condition impact on Lempel-Ziv complexity.Thus can verify, can be as fault diagnosis characteristic parameter under variable working condition. Lempel-Ziv complexity can be as the characteristic parameter of the state recognition of the epicyclic gearbox under bad working environments environment.In step (1), on the basis of (2), use rank than the complexity epicyclic gearbox diagnosing malfunction to labyrinth, calculate different faults Rank under pattern, than complexity, by observing the situation of change than complexity of the rank corresponding to each fault, find that complexity is compared on rank Can be as the characteristic parameter of epicyclic gearbox Fault Pattern Recognition.This step utilize rank than multiplicity, by wind-powered electricity generation planet tooth The fault signature rank that roller box is under different conditions calculate than complexity, it is possible to achieve the knowledge of epicyclic gearbox fault mode Not.

(5) case verification finds, epicyclic gearbox rank under different faults pattern are relatively bigger than complexity difference, Ke Yizuo For the evaluation of epicyclic gearbox state, in the case of lacking sufficient fault data, it is also possible to as gear-box health status Assessment.Through this step case verification, utilize rank than multiplicity, wind power planetary gear case malfunction can be identified, lacking In the case of few fault data, it is also possible to as the assessment parameter of epicyclic gearbox health status.

Detailed being explained as follows is made below for the inventive method:

Step 1, gear planetary wheel case vibration signal pretreatment.Epicyclic gearbox is under the working environment of variable working condition, and it is adopted The vibration signal integrated is as non-stationary signal, and this produces the biggest difficulty to the fault diagnosis of gear-box, uses rank than resampling Method carries out angular domain resampling to vibration signal, and the time domain vibration signal of non-stationary is converted into the angle of steady or quasi-smooth performance Territory signal, it is simple to the analysis to vibration signal.The core of order ratio analysis technology is to obtain the constant angle increment of relative reference axis Sampled data, it is therefore desirable to can accurately obtain the moment of order sampling and corresponding reference rotation speed, i.e. realize order tracking technique.Common Computed order tracking method have hardware order tracking technique method, calculate Computed order tracking method and Computed order tracking method based on instantaneous Frequency Estimation Deng.The resampling using calculating Computed order tracking method to realize vibration signal calculates, and can be produced by resampling during mechanical speed change The raw vibration signal relevant with rotating speed efficiently separates out, the signal unrelated with rotating speed plays certain suppression simultaneously and makees With, redistribute the sampling interval of signal by angle position, the rotation speed change impact on signal randomness can be rejected.Actual The most all containing multiple interference component in gear-box vibration signal, this extraction allowing for its fault signature becomes to compare Difficulty.Any one sophisticated signal adaptive can be decomposed into a series of according to the local time-varying characteristics of signal by EMD method Component, is reconstructed signal by correlation coefficient rule, rejects the interference component in primary signal.Use order ratio analysis with EMD decomposition method effectively can carry out pretreatment to vibration signal, and the signal analysis to next step is got ready.

Due to the coupling between the complexity of operating condition and parts, the vibration data of Wind turbines has obvious non-thread Property and non-stationary property, it is therefore necessary to use vibration data preconditioning technique unit vibration data are carried out preposition process.This Bright based on rank than the wind generating set vibration data processing method of resampling technique, the method can efficiently solve Wind turbines and shake The variable speed problem of dynamic signal, can be converted into the vibration time-domain signal with non-stationary characteristic and have shaking of smooth performance Dynamic angular domain signal, its ultimate principle is angularly to adopt using software algorithm to be converted into by the time-domain signal that constant duration is sampled Angular domain vibration signal under original mold formula, the angular domain vibration signal after conversion i.e. ensure that the verity of original time domain signal, has tool Having had characteristic positive period, the vibrational feature extracting extracted on the basis of angular domain vibration signal can reflect wind-powered electricity generation accurately and effectively Malfunction under the conditions of unit variable parameter operation.

Its rank are more as follows than the key step of resampling technique:

1. assume that the vibration data that vibration measuring point obtains is { x₁,x₂,x₃,…x_n-2,x_n-1,x_n, first by between such time Every the time-domain signal that sampling obtains, it is converted into the angular domain signal of constant duration sampling, during this calculating, assumes rotating speed in short-term In comprised according to the mode speed change of constant angular, the vibration period that when using constant duration sampling, frequency of vibration is minimum Sampling number the most, for ensure reconstruction signal will not distortion, using counting as interpolation of comprising of lowest vibration frequency cycle Standard value reference value, rotate a circle when this reference value is converted into equiangular sampling counting of being comprised.This is counted and is isogonism During degree sampling, rotor every turn should gather volume and counts, rotating speed the fastest every turn gather count the fewest, therefore rotating speed is with reference to should be with the highest Rotating speed is reference.Assume that the low-limit frequency composition vibrating measuring point is f₀, gear box ratio is the n (gear ratio of three gear stages It is respectively n₁,n₂,n₃), according to the maximum speed requirement of generator side, during normal power generation, the range of speeds of main shaft is n_min～n_max, Then using high speed shaft rotating speed as reference rotation velocity, its value is n_ck=n₁×n₂×n₃×n_max, it is calculated the rank angular domain than resampling The interpolation of signal reconstruction is with reference to counting as n=n_ck*f_o/ 60, every turn of point gathered when i.e. representing employing equiangular sampling mode Number should be n, it is assumed that comprise in the analysis of vibration signal cycle always counts as N, then the angular domain sequence redefined is: Theta (N)=0:2 π/n:2 π (N-1)/n；Use linear interpolation method that Theta (N) inserts the angular domain letter under constant duration sampling In number, it is thus achieved that the vibration angular domain signal under equiangular sampling mode.

2. the vibration time-domain signal of constant duration collection is converted into corresponding angular domain vibration signal, due to Wind turbines The function of pitch-controlled system, Wind turbines rotation speed change is relatively mild, can false wind group of motors rotating speed be equal angular acceleration Speed change, wherein generating unit speed signal can obtain from generator speed encoder, it is therefore assumed that the unit in the time of each second turns Speed is equal angular acceleration change.If the speed of mainshaft data that vibration measuring point obtained in ten minutes are n_a, within lower ten minutes, obtain Speed of mainshaft data be n_b, corresponding angular velocity is respectively ω_bAnd ω_a, therefore the set drive chain vibration in this time period is surveyed Point reference rotation velocity change curve is expressed as:

(ω_b-ω_a)·t+ω_a (1)

This angular velocity curve is integrated in time scale obtain angle formula:

$\mathrm{\θ}={\∫}_{t_1}^{t_2}\[\left({\mathrm{\ω}}_b-{\mathrm{\ω}}_a\right)\·t+{\mathrm{\ω}}_a\]dt---\left(2\right)$

By formula (1), constant duration sampling is obtained time domain vibration signal and be converted into the angular domain signal of correspondence.

3. the method using linear interpolation carries out interpolation to the angular domain signal of constant duration, and its per revolution is gathered Count step 1. in calculated acquisition.The angle of constant duration acquisition angle territory signal is counted according to interpolation and carries out drawing Point, try to achieve the angle coordinate { θ of the point needing to carry out interpolation₁,θ₂,…,θ_n-1,θ_n, these points are sampled angle at constant duration Territory signal enterprising line linearity interpolation, it is thus achieved that { x_θ1,x_θ2,…x_θn-1,x_θn, by formula [x_t(i-1)-x_θ(i)]×[x_t(i)-x_θ(i)] ≤ 0 search is positioned at the actual value of interpolation point both sides, by linear interpolation formula

$x_{\mathrm{\θ}\left(i\right)}=\frac{x_{t\left(i\right)}\·\left(\mathrm{\θ}\right(i)-t(i-1))}{t\left(i\right)-t\left(i-1\right)}---\left(3\right)$

The vertical coordinate of coordinate points is inserted, it is thus achieved that angularly angular domain signal { x after calculating linear interpolation_θ1,x_θ2,…x_θn-1, x_θn}.So far we achieve constant duration sample time domain signal to equiangular sampling angular domain letter by the 1-3 step in step 2 Number conversion, the angular domain signal after conversion has smooth performance and characteristic complete cycle, and rank are than the stream of resampling angular domain signal reconstruction Journey is shown in Fig. 1.

The most achieve the non-stationary time domain vibration signal under the conditions of variable working condition to flat by rank than resampling technique The conversion of steady angular domain vibration signal, vibration signal now has had characteristic complete cycle, to the angular domain with characteristic complete cycle Vibration signal extracts fault pre-alarming and whether diagnosis index can occur or development degree by faults accurately and efficiently.But want It is noted that in angular domain sampling, with equal angle sampling frequency S_fCorresponding for order sample frequency S_o(Sampling Frequency of Order tracking), i.e. the reference axis angularly data gathered that often rotate a circle are counted, in order to protect The integrity of information in card primary signal, order is sampled as time-domain sampling, is required for meeting sampling thheorem: order sampling is frequently Rate have to be larger than the highest order composition O_maxTwice, i.e.

S_o＞ O_max (4)

The spectrogram obtained after angularly signal makees Fourier transformation to resampling is order spectrum.Equiangular sampling is utilized to believe Number and order spectrum carry out each parameter in the analysis of angular domain signal and the time domain of signal, frequency-domain analysis and have and close one to one System, as shown in table 2.

Vibration angular domain signal after resampling usually comprises noise, in order to reduce the interference of this composition, select herein based on EMD method carries out noise reduction process, and the method can reduce low-frequency disturbance effectively highlight high frequency intrinsic vibration with cancelling noise, it is easy to Fault characteristic frequency is extracted in sophisticated signal.It is a kind of adaptive signal decomposition method that EMD decomposes, and its advantage shows themselves in that (1) basic function automatically generates: can draw from EMD catabolic process, and the method, according to the feature of primary signal, adaptive is chosen Optimum basic function, and need to be pre-selected basic function by wavelet decomposition, the first selection of basic function is cumbersome, catabolic process Middle basic function is once selected can not be changed；(2) there is adaptive filtering characteristic；(3) adaptive multiresolution.EMD divides IMF after solution contains frequency content from high to low.Owing to EMD catabolic process existing end effect, interpolation error, excessively The situations such as decomposition, cause decomposition result to there may be pseudo-component, and primary signal are unrelated.May be containing with fault frequently in pseudo-component The frequency that rate overlaps, it should the pseudo-component rejection trying every possible means to impact these falls.Can be had by cross-correlation coefficient criterion Effect identifies pseudo-component, it is achieved method is the cross-correlation coefficient calculating IMF component with primary signal, be can recognize that by the size of coefficient Pseudo-component.

Component with the cross-correlation coefficient of primary signal S is:

${\mathrm{\ρ}}_{s,{\hat{c}}_j}=max\left(R_{s,{\hat{c}}_j}\right(\mathrm{\τ}))/max\left(R_s\right(\mathrm{\τ}\left)\right)---\left(5\right)$

Wherein, R_s(τ) it is the autocorrelation coefficient of original signal.

Pseudo-component can be rejected by cross-correlation coefficient criterion, then be reconstructed signal, reduce the interference of noise.Through rank It is angular domain stationary signal than the signal after resampling and EMD noise reduction process.

Step 2, fault characteristic frequency calculates.Wind turbines gear speedup case various structures, gear ratio is big, for reducing gear The size of case, generally planetary gear construction, a certain Wind turbines epicyclic gearbox is analyzed by the present invention, its wind-powered electricity generation planet Gear box structure is two-stage planetary gear, one-level parallel gears structure, and its structure is as shown in Figure 2.

The gear for two stage planetary gear train of Wind turbines epicyclic gearbox and parallel stage gear structure, its structural parameters are as shown in table 2.

Epicyclic gearbox is different from traditional gear-box, and its structure is made up of sun gear, planetary gear, gear ring and planet carrier. Generally, gear ring maintains static, and sun gear rotates around the central axis of self, and planetary gear not only rotation is also about the sun Wheel revolution.Planetary gear both engaged with sun gear, engaged with gear ring again.In epicyclic gearbox, the compound motion of multiple gears causes The complexity of vibration signal.In the monitoring of vibration, sensor is typically mounted on gear ring or the casing that is attached thereto to gather and shakes Dynamic signal, the meshing point of sun gear-planetary gear and planetary gear-gear ring Meshing Pair rotates change to the position of sensor with planet carrier Change so that the vibration transfer path between meshing point to sensor changes, and the bang path of this time-varying is to vibration-testing Signal produces amplitude-modulating modulation effect.

Gear distress is generally divided into two classes: distributed fault and local fault.Planet circular system generation different faults type and During abort situation, its vibration signal model and fault characteristic frequency can be different, and vibration signal when there is fault is carried out Analysis can realize the diagnosis to fault, planet circular system distributed faults and local fault characteristic frequency with vibration signal model in ginseng Examine document has and tell about in detail, the most just repeat no more.

Wind power planetary gear case is generally divided into planet circular system, parallel stage gear, for planet circular system and parallel stage gear, its Fault can be divided into local fault and distributed fault.In single-pinion planetary gear case, sun gear planetary gear and planetary gear gear ring The meshing frequency of two kinds of Meshing Pair is identical.Generally gear ring maintains static, and sun gear, planetary gear and planet carrier rotate, in this feelings Under condition, meshing frequency:

f_m=f_cZ_r=(f_s ^(r)-f_c)Z_s(6)；

In formula: Z_rAnd Z_sIt is respectively gear ring and the number of teeth of sun gear；f_mFor meshing frequency；f_cSpeed for planet carrier； f_s ^(r)Absolute speed for sun gear.

Sun gear local fault characteristic frequency is:

$f_s=\frac{f_m}{Z_s}N---\left(7\right);$

In formula: f_mFor meshing frequency；Z_sFor the sun gear number of teeth；N is planetary gear quantity.

Planetary gear local fault characteristic frequency is:

$f_p=2\frac{f_m}{Z_p}---\left(8\right);$

In formula: f_mFor meshing frequency；Z_pFor the planetary gear number of teeth.Gear ring local fault characteristic frequency is:

$f_r=\frac{f_m}{Z_r}N---\left(9\right);$

In formula: f_mFor meshing frequency；Z_rFor the gear ring number of teeth.

In epicyclic gearbox, the distributed fault characteristic frequency of various gears is equal to gear opposing rows carrier (sun gear and tooth Circle fault) or the speed of gear ring (planetary gear fault).The meshing frequency f of known epicyclic gearbox_mTooth with certain gear Number Z_g, then this gear opposing rows carrier (sun gear and gear ring fault) or the speed of gear ring (planetary gear fault):

f_g=f_m/Z_g(10)；

Then the characteristic frequency of sun gear, planetary gear and gear ring distributed fault is respectively as follows:

f_s'=f_m/Z_s(11)；

f_p'=f_m/Z_p(12)；

f_r'=f_m/Z_r(13)；

In formula: f_mFor meshing frequency；f_s'、f_p'、f_r'Sun gear, planetary gear and the characteristic frequency of gear ring distributed fault；Z_s For the sun gear number of teeth；Z_pThe number of teeth for planetary gear；Z_rFor the gear ring number of teeth.

With the planet carrier of primary planet train that is connected with main shaft as reference rotation velocity, at different levels each in epicyclic gearbox The local fault of individual gear and the feature rank ratio of distributed faults calculate, as shown in table 3.

The signal of the present invention is that matlab emulates signal, ignores the shadow between gear-box middle (center) bearing and each gear in signal Ring, it is assumed that the vibration effect between each gearbox drive level does not exists, and the method for the present invention is carried out simulating, verifying for planet Train is in normally, distributed faults, local fault time vibration signal model list of references, the most just repeat no more.

Thus, it is possible to emulation epicyclic gearbox is at different levels when being in nominal situation, its vibration signal model is:

x₁(t)=A [1-cos (2 π 3f_c1·t)]cos(2π·f_m1·t+θ₁)

+B·[1-cos(2π·3_fc2·t)]·cos(2π·_fm2·t+θ₂)+C·cos(2π·_fm3·t+θ₃) (14)

During epicyclic gearbox primary planet train sun gear generation local fault, its vibration signal model is:

During epicyclic gearbox primary planet train sun gear occurrence and distribution fault, its vibration signal model is:

In formula (14)～(16): x₁(t)、x₂(t)、x₃T () is that epicyclic gearbox is in normal primary planet train sun gear There is vibration signal sequence when local fault, distributed faults；T is time series；θ₁、θ₂、θ₃、φ、For initial phase；f_m1、 f_m2、f_m3For meshing frequencies at different levels；f_c1、f_c2Speed for I and II planet carrier；Absolute rotation for one-level sun gear Turn frequency；f_s1、f_s1'For characteristic frequency when one-level sun gear generation local fault and distributed faults；A, B, C are immeasurable three cardinal guides Number, each state duration of epicyclic gearbox can be different, the most no longer describes in detail.Each vibration signal employing frequency is 8192HZ。

Step 3, Fault characteristic parameters.Often make its frequency composition when epicyclic gearbox breaks down and respectively become framing, frequently Characteristic changes, so that time domain waveform occurs distortion in various degree, this distortion is also epicyclic gearbox gear distress Directly reflection in time-domain signal.When the waveform of signal, spectrum structure change, the complexity of its signal also becomes Change.Jiang Jiandong Lempel-Ziv Complexity Measurement qualitative assessment large-sized unit running status.It is multiple that Hong have studied Lempel-Ziv Miscellaneous degree index is used for assessing faulty bearings faulted condition.Practice have shown that, this index is to weigh finite time sequence complexity Efficient tool.

The Lempel-Ziv complexity of sequence can obtain by through time circulation, calculates process as follows:

(1) S is initialized_v,0={ }, Q₀={ }, C_N(0)=0, r=1；OrderDue to Q_rIt is not belonging to S_v,r-1, Then C_N(r)=C_N(r-1), Q_r={ }, r=r+1；

(2) orderJudge Q_rWhether belong toIf belonging to, then C_N(r)=C_N(r-1), r =r+1, repetitive process (2)；

(3) if being not belonging to, then C_N(r)=C_N(r-1)+1, Q_r={ }, r=r+1, repetitive process (2).

The such as segmented version of sequence { 0000 } is { 0 000 }, C_N=2.Symbol sebolic addressing { 010101 } segmented version is { 01 0101 }, C_NIt is just 3.Examine for the comparability strict to data Considering, Lempel and Ziv further provides a kind of normalization reference formula, value is defined between [0-1], and detailed algorithm is:

$0\≤C_{nN}=\frac{C_N\left(N\right)}{C_{UL,N}}\≤1---\left(17\right)$

$C_{UL,N}=\underset{N\→\mathrm{\∞}}{\lim }{C}_N\left(N\right)=\underset{N\→\mathrm{\∞}}{\lim }\frac{N}{(1-\mathrm{\β}){\log }_{k}N}\≈\frac{N}{{\log }_{k}N}---\left(18\right)$

Wherein: k is middle S_NThe number of element is (for binary system S_NSequence, k=2).According to Lempel-Ziv product complexity theory Understanding with document, the complexity of a sequence is the biggest, illustrates that its interpolation operation is the most, and the intension component of sequence is the most, retouches Needed for stating given symbol sebolic addressing minimum, mutually different segmented version is the most, given sequence periodicity is the most weak, occur new in The speed of culvert pattern is the fastest.The physical significance of Lempel-Ziv complexity is that it reflects a time series along with length There is the speed of new model in the increase of degree.Complexity is the biggest, and the new change that data occur within length of window period in time is described Changing the most, the speed that new change occurs is the fastest, shows that the data variation in this period is unordered and complexity.Otherwise complexity is more Little, then the speed of explanation generation new change is the slowest, and data variation is regular, the strongest.Therefore, vibration signal Lempel-Ziv complexity index can objectively respond out the situation that system mode changes.

Step 4, the identification of fault mode.Variable speed vibration signal under epicyclic gearbox normal condition is analyzed, The a length of 20s of vibration signal, its tach signal and vibration signal are fig. 3, it is shown that vibration signal is variable speed letter Number, the amplitude of vibration signal decreases with the reduction of rotating speed with frequency.Fig. 4 is that variable speed vibration signal is adopted through rank proportion Angular domain stationary signal after sample and the process of EMD decomposition method, it can be seen that angular domain signal is compared to time-domain signal more Steadily, the sampling interval of signal more rule.

The complexity that epicyclic gearbox vibrates time-domain signal and angular domain signal calculates.Time-domain signal calculates Length be respectively the signal length of 3s and 1.5s, the length calculating angular domain signal is respectively the signal length of sum, logical Cross calculate unlike signal length Lempel-Ziv complexity numerical value it can be seen that the complexity numerical fluctuations of time-domain signal relatively Greatly, the most steadily, and the complexity numerical value of angular domain signal reaches unanimity, more stable, it may be said that bright, the complexity of angular domain signal The running status of epicyclic gearbox more can be reflected, as shown in Figure 5 compared to time-domain signal.

By order ratio analysis method, the steady angular domain signal under epicyclic gearbox difference running status is carried out Lempel- The calculating of Ziv complexity, Fig. 6 is corresponding in turn to that epicyclic gearbox is normal from left to right, the distribution of gear ring distributed faults, planetary gear therefore Barrier, the Lempel-Ziv complexity of sun gear distributed faults, gear ring local fault, planetary gear local fault, sun gear local fault Degree variation tendency, table 4 is its concrete numerical value.

As can be seen from the results, the vibration of epicyclic gearbox normal condition is the simplest, and when local fault occurs, it is non- Linearly compare bigger, compared to distributed faults, there is higher Lempel-Ziv value.By Lempel-Ziv complexity numerical value, can Effectively identify with the running status to epicyclic gearbox.

Step 5, case verification.Certain sharp SL1500 wind turbine gearbox vibration data of wind energy turbine set China is analyzed, vibration letter Number employing frequency be 5120Hz, signal length is 6s, and three segment signals are respectively gear-box normal condition, gearbox high-speed level tooth Vibration signal under foot wheel abrasion state and gearbox high-speed level gear tooth breakage state, each section of vibration signal is in different rotating speeds, time Territory signal is steady, as shown in Figure 7 not.By vibration time-domain signal is carried out rank than resampling and the pre-place of EMD decomposition method Reason, calculates the Lempel-Ziv complexity of signal under three states, and its numerical value is as shown in table 5.

Owing to lacking the service data under each running status of gear-box, it is difficult to each malfunction of gear-box is known Not, but Lempel-Ziv numerical value from above three running statuses it can be seen that, Lempel-Ziv complexity can conduct Characterize the characteristic parameter of running state of gear box.Simultaneously in the case of planet gearbox fault data abundance, can be to planet The fault of gear divides, it is achieved the interval division to the Lempel-Ziv complexity under each fault mode.According to this Invention, epicyclic gearbox fault diagnosis flow scheme based on Lempel-Ziv complexity is as shown in Figure 8.

Table 1 time domain, angular domain relativity.

Table 2 wind power planetary gear box structure parameter.

Table 3 epicyclic gearbox fault characteristic frequency calculates.

The Lempel-Ziv complexity of table 4 epicyclic gearbox different conditions.

Lempel-Ziv complexity under three states of table 5 epicyclic gearbox.

Normally	Level gear wear at a high speed	Level gear tooth breakage at a high speed
			0.2628	0.4023	0.5625

Claims (6)

1. one kind based on rank than the epicyclic gearbox method for diagnosing faults of complexity, it is characterised in that: comprise the steps:

(1) vibration signal pretreatment: according to Computed order tracking technology, the wind power planetary gear case sensor being in variable working condition is adopted The vibration signal of collection carries out pretreatment, time-domain signal non-linear, non-stationary is converted into the angular domain signal with stationarity, keeps away Exempt from the expensive cost using hardware mode to realize equiangular sampling；Nonstationary vibration time-domain signal based on linear interpolation method Rank than reconfiguration technique, the nonstationary vibration time-domain signal that constant duration is sampled is converted into and there is the angular domain of smooth performance shakes Dynamic signal, it is ensured that property complete cycle of epicyclic gearbox vibration angular domain signal, it is to avoid the impact of variable parameter operation environment；

(2) fault of planet circular system is divided into two classes: distributed faults and local fault；Distributed faults and local to planet circular system The calculating of the characteristic frequency of fault is analyzed；With magnificent sharp SL1500 wind power planetary gear case as object of study, to two-stage planet Train, the vibration signal of one-level parallel stage epicyclic gearbox emulate；

(3) Fault characteristic parameters: Lempel-Ziv complexity in the fault diagnosis of rolling bearing and simple parallel level gear Through showing, the concept of Lempel-Ziv complexity can be analyzed as the Fault characteristic parameters of bearing failure diagnosis, enters One step confirms that Lempel-Ziv complexity can be as the characteristic parameter of fault diagnosis；

(4) vibration signal under planetary planet gear case variable working condition is analyzed, finds the Lempel-of Non-stationary vibration signal Ziv complexity can fluctuate, but the angular domain signal after the proportion sampling processing of rank can effectively prevent variable working condition pair The impact of Lempel-Ziv complexity；Thus can verify, can be as fault diagnosis characteristic parameter under variable working condition；Lempel- Ziv complexity can be as the characteristic parameter of the state recognition of the epicyclic gearbox under bad working environments environment；In step (1), (2) On the basis of, use rank than the complexity epicyclic gearbox diagnosing malfunction to labyrinth, calculate under different faults pattern Rank than complexity, by observing the situation of change than complexity of the rank corresponding to each fault, find that rank can be made than complexity Characteristic parameter for epicyclic gearbox Fault Pattern Recognition；This step utilize rank than multiplicity, to wind power planetary gear case Fault signature rank under different conditions calculate than complexity, it is achieved the identification of epicyclic gearbox fault mode；

(5) case verification finds, epicyclic gearbox rank under different faults pattern are relatively bigger than complexity difference, can be as row The evaluation of star gear-box state, in the case of lacking sufficient fault data, it is also possible to as the assessment of gear-box health status； Through this step case verification, utilize rank than multiplicity, wind power planetary gear case malfunction can be identified, lack fault In the case of data, it is also possible to as the assessment parameter of epicyclic gearbox health status.

The most according to claim 1 a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, its feature exists In: the vibration signal pretreatment described in step (1):

It is to gear planetary wheel case vibration signal pretreatment；Epicyclic gearbox is under the working environment of variable working condition, and what it gathered shakes Dynamic signal is non-stationary signal, and this produces the biggest difficulty to the fault diagnosis of gear-box, uses the rank method pair than resampling Vibration signal carries out angular domain resampling, and the time domain vibration signal of non-stationary is converted into the angular domain letter of steady or quasi-smooth performance Number, it is simple to the analysis to vibration signal；The core of order ratio analysis technology is to obtain the constant angle increment sampling of relative reference axis Data, it is therefore desirable to can accurately obtain the moment of order sampling and corresponding reference rotation speed, i.e. realize order tracking technique；Common rank Comparison-tracking method has hardware order tracking technique method, calculates Computed order tracking method and Computed order tracking method based on instantaneous Frequency Estimation；Use Calculate Computed order tracking method realize vibration signal resampling calculate, by resampling can by during mechanical speed change produce with turn The relevant vibration signal of speed efficiently separates out, the signal unrelated with rotating speed is played certain inhibitory action simultaneously, by turning Angle Position redistributes the sampling interval of signal, can reject the rotation speed change impact on signal randomness；Actual gear-box The most all containing multiple interference component in vibration signal, this extraction allowing for its fault signature becomes relatively difficult； Any one sophisticated signal adaptive can be decomposed into a series of component according to the local time-varying characteristics of signal by EMD method, By correlation coefficient rule, signal is reconstructed, rejects the interference component in primary signal；Order ratio analysis is used to decompose with EMD Method effectively can carry out pretreatment to vibration signal, and the signal analysis to next step is got ready；

Due to the coupling between the complexity of operating condition and parts, the vibration data of Wind turbines have the most non-linear and Non-stationary property, it is therefore necessary to use vibration data preconditioning technique that unit vibration data are carried out preposition process；Base of the present invention The ratio wind generating set vibration data processing method of resampling technique in rank, the method can efficiently solve wind generating set vibration letter Number variable speed problem, the vibration time-domain signal with non-stationary characteristic can be converted into and there is the angle of throw of smooth performance Territory signal, its ultimate principle is will to use software algorithm that the time-domain signal that constant duration is sampled is converted into equiangular sampling mould Angular domain vibration signal under formula, the angular domain vibration signal after conversion i.e. ensure that the verity of original time domain signal, is provided with Positive period characteristic, on the basis of angular domain vibration signal extract vibrational feature extracting can reflect Wind turbines accurately and effectively Malfunction under the conditions of variable parameter operation；

Its rank are more as follows than the key step of resampling technique:

1. assume that the vibration data that vibration measuring point obtains is { x₁,x₂,x₃,…x_n-2,x_n-1,x_n, first this constant duration is adopted The time-domain signal that sample obtains, is converted into the angular domain signal of constant duration sampling, in assuming the rotating speed short time during this calculating According to the mode speed change of constant angular, what the vibration period that when using constant duration sampling, frequency of vibration is minimum was comprised adopts Number of samples is the most, for ensure reconstruction signal will not distortion, using counting as the standard of interpolation of comprising of lowest vibration frequency cycle Value reference value, rotate a circle when this reference value is converted into equiangular sampling counting of being comprised；This is counted to be and angularly adopts During sample, rotor every turn should gather volume and counts, rotating speed the fastest every turn gather count the fewest, therefore rotating speed is with reference to should be with maximum speed For reference；Assume that the low-limit frequency composition vibrating measuring point is f₀, gear box ratio is that (gear ratio of three gear stages is respectively for n For n₁,n₂,n₃), according to the maximum speed requirement of generator side, during normal power generation, the range of speeds of main shaft is n_min～n_max, then with High speed shaft rotating speed is as reference rotation velocity, and its value is n_ck=n₁×n₂×n₃×n_max, it is calculated the rank angular domain signal than resampling The interpolation of reconstruct is with reference to counting as n=n_ck*f_o/ 60, i.e. represent use equiangular sampling mode time every turn gathered count should For n, it is assumed that comprise in the analysis of vibration signal cycle always counts as N, then the angular domain sequence redefined is: Theta (N)= 0:2π/n:2π(N-1)/n；Linear interpolation method is used to be inserted by Theta (N) in the angular domain signal under constant duration sampling, Obtain the vibration angular domain signal under equiangular sampling mode；

2. the vibration time-domain signal of constant duration collection is converted into corresponding angular domain vibration signal, owing to Wind turbines becomes oar The function of system, Wind turbines rotation speed change is relatively mild, can false wind group of motors rotating speed be equal angular acceleration speed change, Wherein generating unit speed signal can obtain from generator speed encoder, it is therefore assumed that the generating unit speed in the time of each second is equal Angular acceleration change；If the speed of mainshaft data that vibration measuring point obtained in ten minutes are n_a, the main shaft of acquisition in lower ten minutes Rotary speed data is n_b, corresponding angular velocity is respectively ω_bAnd ω_a, the therefore set drive chain vibration measuring point reference in this time period Rotation speed change curve table is shown as:

(ω_b-ω_a)·t+ω_a (1)

This angular velocity curve is integrated in time scale obtain angle formula:

$\mathrm{\θ}={\∫}_{t_1}^{t_2}\[\left({\mathrm{\ω}}_b-{\mathrm{\ω}}_a\right)\·t+{\mathrm{\ω}}_a\]dt---\left(2\right)$

By formula (1), constant duration sampling is obtained time domain vibration signal and be converted into the angular domain signal of correspondence；

3. the method using linear interpolation carries out interpolation to the angular domain signal of constant duration, the point that its per revolution is gathered Number step 1. in calculated acquisition；The angle of constant duration acquisition angle territory signal is counted according to interpolation and divides, Try to achieve the angle coordinate { θ of the point needing to carry out interpolation₁,θ₂,…,θ_n-1,θ_n, to these points at constant duration sampling angular domain letter Number enterprising line linearity interpolation, it is thus achieved that { x_θ1,x_θ2,…x_θn-1,x_θn, by formula [x_t(i-1)-x_θ(i)]×[x_t(i)-x_θ(i)]≤0 is searched Rope is positioned at the actual value of interpolation point both sides, by linear interpolation formula

$x_{\mathrm{\θ}\left(i\right)}=\frac{x_{t\left(i\right)}\·\left(\mathrm{\θ}\right(i)-t(i-1))}{t\left(i\right)-t(i-1)}---\left(3\right)$

The vertical coordinate of coordinate points is inserted, it is thus achieved that angularly angular domain signal { x after calculating linear interpolation_θ1,x_θ2,…x_θn-1,x_θn}；So far We achieve the conversion to equiangular sampling angular domain signal of the constant duration sample time domain signal by the 1-3 step in step 2, Angular domain signal after conversion has smooth performance and characteristic complete cycle；

The most achieve the non-stationary time domain vibration signal under the conditions of variable working condition to steady angle by rank than resampling technique The conversion of territory vibration signal, vibration signal now has had characteristic complete cycle, vibrates the angular domain with characteristic complete cycle Whether signal extraction fault pre-alarming and diagnosis index can occur or development degree by faults accurately and efficiently；But it is noted that , in angular domain sampling, with equal angle sampling frequency S_fCorresponding for order sample frequency S_o(Sampling frequency of Order tracking), i.e. the reference axis angularly data gathered that often rotate a circle are counted, in order to ensure to believe in primary signal The integrity of breath, order is sampled as time-domain sampling, is required for meeting sampling thheorem: order sample frequency have to be larger than the highest Order composition O_maxTwice, i.e.

S_o＞ O_max(4)；

The spectrogram obtained after angularly signal makees Fourier transformation to resampling is order spectrum；Utilize equiangular sampling signal and Each parameter that order spectrum is carried out in the analysis of angular domain signal and the time domain of signal, frequency-domain analysis has relation one to one；

Vibration angular domain signal after resampling usually comprises noise, in order to reduce the interference of this composition, selects herein based on EMD side Method carries out noise reduction process, and the method can reduce low-frequency disturbance and effectively highlight high frequency intrinsic vibration with cancelling noise, it is easy in complexity Signal extracts fault characteristic frequency；It is a kind of adaptive signal decomposition method that EMD decomposes, and its advantage shows themselves in that (1) base letter Number automatically generate: can draw from EMD catabolic process, the method according to the feature of primary signal, adaptive choose optimum Basic function, and need to be pre-selected basic function by wavelet decomposition, the first selection of basic function is cumbersome, base letter in catabolic process Number is once selected can not be changed；(2) there is adaptive filtering characteristic；(3) adaptive multiresolution；After EMD decomposes IMF contains frequency content from high to low；Owing to EMD catabolic process existing end effect, interpolation error, excessive decomposition etc. Situation, causes decomposition result to there may be pseudo-component, and primary signal is unrelated；Pseudo-component may overlap containing with failure-frequency Frequency, it should the pseudo-component rejection trying every possible means to impact these falls；Can effectively be identified by cross-correlation coefficient criterion Pseudo-component, it is achieved method is the cross-correlation coefficient calculating IMF component with primary signal, by the recognizable puppet point of the size of coefficient Amount；

Component with the cross-correlation coefficient of primary signal S is:

${\mathrm{\ρ}}_{s,{\hat{c}}_j}=max\left(R_{s,{\hat{c}}_j}\right(\mathrm{\τ}\left)\right)/max\left(R_s\right(\mathrm{\τ}\left)\right)---\left(5\right);$

Wherein, R_s(τ) it is the autocorrelation coefficient of original signal；

Pseudo-component can be rejected by cross-correlation coefficient criterion, then be reconstructed signal, reduce the interference of noise；Through rank proportion Signal after sampling and EMD noise reduction process is angular domain stationary signal.

The most according to claim 1 a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, its feature exists In: the distributed faults to planet circular system and the calculating of the characteristic frequency of local fault described in step (2) are analyzed:

Wind turbines gear speedup case various structures, gear ratio is big, for reducing the size of gear-box, generally planetary gear knot Structure, is analyzed a certain Wind turbines epicyclic gearbox here, and its wind power planetary gear box structure is two-stage planetary gear, one-level Parallel gears structure；

The gear for two stage planetary gear train of Wind turbines epicyclic gearbox and parallel stage gear structure；

Epicyclic gearbox is different from traditional gear-box, and its structure is made up of sun gear, planetary gear, gear ring and planet carrier；Typically In the case of, gear ring maintains static, and sun gear rotates around the central axis of self, and planetary gear not only rotation is also about sun gear public Turn；Planetary gear both engaged with sun gear, engaged with gear ring again；In epicyclic gearbox, the compound motion of multiple gears causes vibration The complexity of signal；In the monitoring of vibration, sensor is typically mounted on gear ring or the casing that is attached thereto and gathers vibration letter Number, the meshing point of sun gear-planetary gear and planetary gear-gear ring Meshing Pair is rotationally-varying with planet carrier to the position of sensor, Meshing point is changed to the vibration transfer path between sensor, and the bang path of this time-varying is to vibration-testing signal Produce amplitude-modulating modulation effect；

Gear distress is generally divided into two classes: distributed fault and local fault；Planet circular system generation different faults type and fault During position, its vibration signal model and fault characteristic frequency can be different, and vibration signal when there is fault is analyzed The diagnosis to fault can be realized；

Wind power planetary gear case is generally divided into planet circular system, parallel stage gear, for planet circular system and parallel stage gear, its fault Local fault and distributed fault can be divided into；In single-pinion planetary gear case, sun gear planetary gear and planetary gear gear ring two kinds The meshing frequency of Meshing Pair is identical；Generally gear ring maintains static, and sun gear, planetary gear and planet carrier rotate, in this case, Meshing frequency:

$f_m=f_c{Z}_r=(f_s^{\left(r\right)}-f_c)Z_s---\left(6\right);$

In formula: Z_rAnd Z_sIt is respectively gear ring and the number of teeth of sun gear；f_mFor meshing frequency；f_cSpeed for planet carrier； Absolute speed for sun gear；

Too star-wheel local fault characteristic frequency is:

$f_s=\frac{f_m}{Z_s}N---\left(7\right);$

In formula: f_mFor meshing frequency；Z_sFor the sun gear number of teeth；N is planetary gear quantity；

Planetary gear local fault characteristic frequency is:

$f_p=2\frac{f_m}{Z_p}---\left(8\right);$

In formula: f_mFor meshing frequency；Z_pFor the planetary gear number of teeth；Gear ring local fault characteristic frequency is:

$f_r=\frac{f_m}{Z_r}N---\left(9\right);$

In formula: f_mFor meshing frequency；Z_rFor the gear ring number of teeth；

In epicyclic gearbox, the distributed fault characteristic frequency of various gears is equal to gear opposing rows carrier (sun gear and gear ring event Barrier) or the speed of gear ring (planetary gear fault)；The meshing frequency f of known epicyclic gearbox_mTooth number Z with certain gear_g, Then this gear opposing rows carrier (sun gear and gear ring fault) or the speed of gear ring (planetary gear fault):

f_g=f_g/Z_g(10)；

Then the characteristic frequency of sun gear, planetary gear and gear ring distributed fault is respectively as follows:

f_s'=f_m/Z_s(11)；

f_p'=f_m/Z_p(12)；

f_r'=f_m/Z_r(13)；

In formula: f_mFor meshing frequency；f_s'、f_p'、f_r'Sun gear, planetary gear and the characteristic frequency of gear ring distributed fault；Z_sFor too Sun tooth number；Z_pThe number of teeth for planetary gear；Z_rFor the gear ring number of teeth；

With the planet carrier of primary planet train that is connected with main shaft as reference rotation velocity, to each tooth at different levels in epicyclic gearbox The local fault of wheel and the feature rank ratio of distributed faults calculate；

Here signal is that matlab emulates signal, ignores the impact between gear-box middle (center) bearing and each gear in signal, false If the vibration effect between each gearbox drive level does not exists, the method is carried out simulating, verifying and is just in for planet circular system Often, distributed faults, local fault time vibration signal model list of references, the most just repeat no more；

Thus, epicyclic gearbox is at different levels in emulation when being in nominal situation, and its vibration signal model is:

x₁(t)=A [1-cos (2 π 3f_c1·t)]cos(2π·f_m1·t+θ₁)

+B·[1-cos(2π·3f_c2·t)]·cos(2π·f_m2·t+θ₂)+C·cos(2π·f_m3·t+θ₃) (14)

During epicyclic gearbox primary planet train sun gear generation local fault, its vibration signal model is:

During epicyclic gearbox primary planet train sun gear occurrence and distribution fault, its vibration signal model is:

In formula (9)～(11): x₁(t)、x₂(t)、x₃T () is that epicyclic gearbox is in the generation of normal primary planet train sun gear Vibration signal sequence when local fault, distributed faults；T is time series；θ₁、θ₂、θ₃、φ、For initial phase；f_m1、f_m2、 f_m3For meshing frequencies at different levels；f_c1、f_c2Speed for I and II planet carrier；Absolute rotary frequency for one-level sun gear Rate；f_s1、f_s1'For characteristic frequency when one-level sun gear generation local fault and distributed faults；A, B, C are dimensionless constant, OK Each state duration of star gear-box can be different, the most no longer describes in detail；Each vibration signal uses frequency to be 8192HZ.

The most according to claim 1 a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, its feature exists In: the Fault characteristic parameters described in step (3):

Often make its frequency composition when epicyclic gearbox breaks down and respectively become framing, frequently characteristic to change, so that time There is distortion in various degree in domain waveform, this distortion is also the most anti-in time-domain signal of epicyclic gearbox gear distress Reflect；When the waveform of signal, spectrum structure change, the complexity of its signal also changes；Jiang Jiandong Lempel- Ziv Complexity Measurement qualitative assessment large-sized unit running status；Hong have studied Lempel-Ziv complexity index for assessing event Barrier damage of the bearing state；Practice have shown that, this index is to weigh the efficient tool of finite time sequence complexity；

The Lempel-Ziv complexity of sequence can obtain by through time circulation, calculates process as follows:

(1) S is initialized_v,0={ }, Q₀={ }, C_N(0)=0, r=1；OrderDue to Q_rIt is not belonging to S_v,r-1, then C_N (r)=C_N(r-1), Q_r={ }, r=r+1；

(2) orderJudge Q_rWhether belong toIf belonging to, then C_N(r)=C_N(r-1), r=r+ 1, repetitive process (2)；

(3) if being not belonging to, then C_N(r)=C_N(r-1)+1, Q_r={ }, r=r+1, repetitive process (2)；

The such as segmented version of sequence { 0000 } is { 0 000 }, C_N=2；Symbol sebolic addressing { 010101 } segmented version is { 01 0101 }, C_NIt is just 3；Examine for the comparability strict to data Considering, Lempel and Ziv further provides a kind of normalization reference formula, value is defined between [0-1], and detailed algorithm is:

$0\≤C_{nN}=\frac{C_N\left(N\right)}{C_{UL,N}}\≤1---\left(12\right)$

$C_{UL,N}=\underset{N\→\mathrm{\∞}}{\lim }{C}_N\left(N\right)=\underset{N\→\mathrm{\∞}}{\lim }\frac{N}{(1-\mathrm{\β}){\log }_{k}N}\≈\frac{N}{{\log }_{k}N}---\left(13\right)$

Wherein: k is middle S_NThe number of element is (for binary system S_NSequence, k=2)；According to Lempel-Ziv product complexity theory and literary composition Offering and understand, the complexity of a sequence is the biggest, illustrates that its interpolation operation is the most, and the intension component of sequence is the most, and description is given Needed for determining symbol sebolic addressing minimum, mutually different segmented version is the most, given sequence periodicity is the most weak, new connotation mould occurs The speed of formula is the fastest；The physical significance of Lempel-Ziv complexity is that it reflects a time series along with length Increase the speed that new model occurs；Complexity is the biggest, illustrates that new change that data occur within length of window period in time is more Many, the speed that new change occurs is the fastest, shows that the data variation in this period is unordered and complexity；Otherwise complexity is the least, Then the speed of explanation generation new change is the slowest, and data variation is regular, the strongest；Therefore, the Lempel-of vibration signal Ziv complexity index can objectively respond out the situation that system mode changes.

The most according to claim 1 a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, its feature exists In: the identification realizing epicyclic gearbox fault mode described in step (4):

It is that the variable speed vibration signal under epicyclic gearbox normal condition is analyzed, a length of 20s of vibration signal, its vibration Signal is variable speed signal, and the amplitude of vibration signal decreases with the reduction of rotating speed with frequency；According to variable speed vibration letter Angular domain stationary signal number after rank process than resampling and EMD decomposition method, show, angular domain signal is compared to time-domain signal More steady, the sampling interval of signal more rule；

The complexity that epicyclic gearbox vibrates time-domain signal and angular domain signal calculates；Time-domain signal carries out the length calculated Degree is respectively the signal length of 3s and 1.5s, and the length calculating angular domain signal is respectively the signal length of sum, by meter Calculate the Lempel-Ziv complexity numerical value of unlike signal length it can be seen that the complexity numerical fluctuations of time-domain signal is relatively big, no Enough steadily and the complexity numerical value of angular domain signal reaches unanimity, more stable, illustrates, the complexity of angular domain signal compared to time Territory signal more can reflect the running status of epicyclic gearbox；

By order ratio analysis method, the steady angular domain signal under epicyclic gearbox difference running status is carried out Lempel-Ziv again The calculating of miscellaneous degree, is corresponding in turn to that epicyclic gearbox is normal, gear ring distributed faults, planetary gear distributed faults, sun gear from left to right Distributed faults, gear ring local fault, planetary gear local fault, the Lempel-Ziv complexity change of sun gear local fault become Gesture, draws concrete numerical value,

As can be seen from the results, the vibration of epicyclic gearbox normal condition is the simplest, and when local fault occurs, it is non-linear Compare bigger, compared to distributed faults, there is higher Lempel-Ziv value；By Lempel-Ziv complexity numerical value, it is possible to right The running status of epicyclic gearbox effectively identifies.

The most according to claim 1 a kind of based on rank than the epicyclic gearbox method for diagnosing faults of complexity, its feature exists In: the case verification described in step (5) finds:

Being analyzed certain sharp SL1500 wind turbine gearbox vibration data of wind energy turbine set China, the employing frequency of vibration signal is 5120Hz, signal length is 6s, and three segment signals are respectively gear-box normal condition, gearbox high-speed level gear wear condition and tooth Vibration signal under box high speed level gear tooth breakage state, each section of vibration signal is in different rotating speeds, and time-domain signal is steady not, By vibration time-domain signal being carried out rank than resampling and the pretreatment of EMD decomposition method, calculate signal under three states Lempel-Ziv complexity；

Owing to lacking the service data under each running status of gear-box, it is difficult to each malfunction of gear-box is identified, But Lempel-Ziv numerical value from above three running statuses it can be seen that, Lempel-Ziv complexity can be as sign The characteristic parameter of running state of gear box；Simultaneously in the case of planet gearbox fault data abundance, can be to planetary gear Fault divide, it is achieved the interval division to the Lempel-Ziv complexity under each fault mode.