CN103389341B - windmill blade crack detection method - Google Patents

Abstract

A kind of windmill blade crack detection method, be characterized in: (1) installs calibrate AE sensor on pneumatic equipment blades, and the acoustic emission signal received is passed to acoustic emission acquisition system, determine the sample frequency of signal, sampling length, frequency filtering; (2) bandwidth parameter of Morlet wavelet basis function is optimized based on Shannon Wavelet Entropy, obtain the Morlet wavelet basis function mated with running crack and crack initiation sound emission signal characteristic, then the code reassignment scale spectrum calculating acoustic emission signal judges crackle state; (3) extended mode of crack fault is then judged according to the time-frequency characteristics parameter of extracted Crack Acoustic Emission Signal.The present invention can detect the state of blade cracks dynamic expansion rapidly and accurately, ensures the security of pneumatic equipment blades and high efficiency, extends pneumatic equipment blades serviceable life, reduces maintenance cost.

Description

Windmill blade crack detection method

Technical field

The present invention relates to a kind of detection method of blade of wind-driven generator, particularly relate to a kind of windmill blade crack detection method detecting blade of wind-driven generator surface microcrack.

Background technology

Blade is the critical component that wind energy conversion system obtains wind energy, the effects such as centrifugal force, bending stress, aerodynamic force, thermal stress are not only born when running in harsh physical environment, also be subject to the erosion of frost sleet and the destruction of Lightning strike, pneumatic equipment blades very easily causes the cracking of girder, skin, adhesives when running continuously under heavy load jumpy, is out of shape, comes off and ruptures, serious threat is brought to the safe operation of whole unit, even cause the generation of major accident, blade cracks becomes ubiquitous potential safety hazard in large-scale wind power field.

The approximate time of current overseas utilization finite element simulation and Method of Fracture Mechanics measurable pneumatic equipment blades generation crackle and residual life.But because these methods cannot determine faint, the unstable signal characteristic that pneumatic equipment blades initial crack produces; and pneumatic equipment blades long-term operation under the operating modes such as bending moment of being everlasting, driftage, shutdown, startup; complicated random load distribution and severe field meteorological condition add fracturing mechanics difficulty in computation; therefore need to find a kind of new windmill blade crack detection method, thus reach the effect of monitoring blade injury state and prediction damage trend.

Acoustic emission testing technology has greater advantage in pneumatic equipment blades crack detection, but pneumatic equipment blades Crack Acoustic Emission Signal is through amplifying, filtering, isolation transmission, after type conversion and long-distance sand transport, be submerged in the mixed signal of collection, owing to being subject to travel path, glass fiber reinforced plastics composite material attribute, the impact of the factors such as sensors coupled agent characteristic, wind energy conversion system Crack Acoustic Emission Signal shows significantly non-linear and non-stationary property, cause analyzing and extract acoustic emission source characteristic parameter to have difficulties, adopt traditional filtering method often cannot recover the feature of original signal.And traditional acoustic emission testing technology thinks that signal is propagated with a certain fixed speed, and the compound substance of pneumatic equipment blades has stronger anisotropy, and the evaluation mechanism causing conventional acoustic emission parameters corresponding with acoustic emission source state is difficult to distinct by this.

Summary of the invention

Object of the present invention is just to overcome prior art above shortcomings, is difficult to the problem of extraction and pneumatic equipment blades special material character, proposes a kind of new windmill blade crack detection method for microcrack sound emission signal characteristic.This detection method is by extracting Crack Acoustic Emission Signal characteristic parameter, and then the development trend of quantitative Diagnosis pneumatic equipment blades crackle is carried out according to the characteristic quantity extracted, and the relation mechanism of clear and definite crackle state and FEATURE PARAMETERS OF ACOUSTIC EMISSION thus, both the difficult problem that complicated Mechanics Calculation solves had been avoided, also overcome the problem that classical signal disposal route cannot extract multiple coupled signal minutiae, and can the evaluation mechanism of distinct crackle state and characteristic parameter, reach early detection pneumatic equipment blades crack fault and accurately can judge the object of crackle state.

The technical scheme that the present invention provides is: this windmill blade crack detection method, has been characterized in following steps:

(1) first on pneumatic equipment blades, calibrate AE sensor is installed, and the acoustic emission signal received is passed to acoustic emission acquisition system (employing be the PCI-2 acoustic emission acquisition system of PAC company of the U.S.), determine the acquisition parameters such as the sample frequency of signal, sampling length, frequency filtering;

(2) bandwidth parameter of Morlet wavelet basis function is then optimized based on Shannon Wavelet Entropy, obtain the Morlet wavelet basis function mated with running crack and crack initiation sound emission signal characteristic, the code reassignment scale spectrum calculating acoustic emission signal again judges crackle state, is namely optimized and the Program extraction pneumatic equipment blades running crack of code reassignment scale spectrum signal transacting and the time-frequency characteristics parameter of crack initiation acoustic emission signal by Shannon Wavelet Entropy;

(3) extended mode of crack fault is then judged according to the time-frequency characteristics parameter of extracted Crack Acoustic Emission Signal.

The program computation of the shannon Wavelet Entropy optimization in step of the present invention (2) and code reassignment scale spectrum signal transacting has following steps:

(1). extract wavelet coefficient according to formula (5)

$W_x(a,b;\mathrm{\ψ})={\∫}_{-\∞}^{\∞}x\left(t\right){\mathrm{\ψ}}_{a,b}\left(t\right)\mathrm{dt}---\left(5\right)$

(2). according to formula (9), the coefficient processing of wavelet transformation is become a probability distribution sequence p _i

$p_i=\frac{\left|W_x(a_i,t)\right|}{\underset{j=1}{\overset{M}{\mathrm{\Σ}}}\left|W_x(a_j,t)\right|}---\left(9\right)$

W _x(a, t) is wavelet coefficient values, and M is that wavelet coefficient gets scale parameter not of the same race;

(3). calculate shannon Wavelet Entropy according to formula (10), namely

$H\left(p\right)=-\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\left(p_i\lg p_i\right),\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{p}_i=1---\left(10\right)$

(4). make fc=1, draw shannon Wavelet Entropy and different f _bthe relation curve of parameter.In relation curve when shannon Wavelet Entropy is minimum, corresponding f _bcontrol the optimum bandwidth parameter of wavelet shape exactly;

(5). by optimum f _bsubstitution formula (8) determines optimum Morlet wavelet basis function, and wavelet basis function now mates most with the acoustic emission signal collected, that is:

$\mathrm{\ψ}\left(t\right)=\frac{1}{\sqrt{f_b\mathrm{\π}}}\exp (-t^2/f_b)\exp \left(i2\mathrm{\π}{f}_{c}t\right)---\left(8\right)$

F in formula _bbe the optimum bandwidth parameter that step (4) calculates, for the time frequency resolution of Balanced multi-wavelet, determine the speed degree of oscillating waveform decay.F _cbe centre frequency, determine the oscillation frequency of wavelet shapes, general desirable f _c=1;

(6). according to formula (11)--formula (14) calculates code reassignment scale spectrum, namely

${\mathrm{RSG}}_x(\hat{a},\hat{b};\mathrm{\ψ})={\∫}_{-\∞}^{\∞}{\∫}_{-\∞}^{\∞}{(\hat{a}/a)}^{2}S{G}_x(a,b;\mathrm{\ψ})$

$\mathrm{\δ}(\hat{b}-b^{\′}(a,b))\mathrm{\δ}(\hat{a}-a^{\′}(a,b))\mathrm{dadb}---\left(11\right)$

$b^{\′}(a,b)=b-\mathrm{Re}\left\{a\frac{W_x(a,b;{\mathrm{\ψ}}^{\′})W_x^{*}(a,b;\mathrm{\ψ})}{{\left|W_x(a,b;\mathrm{\ψ})\right|}^2}\right\}---\left(12\right)$

$\frac{{\mathrm{\ω}}_0}{a^{\′}(a,b)}=\frac{{\mathrm{\ω}}_0}{a}+\mathrm{Im}\left\{\frac{W_x(a,b;\widehat{\mathrm{\ψ}})W_x^{*}(a,b;\mathrm{\ψ})}{2\mathrm{\πa}{\left|W_x(a,b;\mathrm{\ψ})\right|}^2}\right\}---\left(13\right)$

ψ′(t)＝tψ(t)， $\widehat{\mathrm{\ψ}}\left(t\right)=\frac{\mathrm{d\ψ}\left(t\right)}{\mathrm{dt}}---\left(14\right);$

(7). draw;

(8). terminate.

Calculated examples in the present invention is (for the pneumatic equipment blades of glass fiber reinforced plastics composite material, total length 1000mm, covering average thickness is 6mm, and mean breadth is 65mm) that carry out under experimental conditions.Because pneumatic equipment blades detection method of the present invention has good time-frequency locality and anti-interference, therefore, it is possible to extract crack fault feature to greatest extent, according to the energy distribution of processing signals, frequency composition and pulse after the type can judging crackle, thus determine the extended mode of pneumatic equipment blades crack fault.

When after change experiment condition (as blade material, imposed load condition etc.), the Crack Acoustic Emission Signal collected will change thereupon, and therefore the time-frequency figure of processing signals also will change.But adopt this method still can Optimization of Wavelet basis function and calculate code reassignment scale spectrum, last cancelling noise extracts the time-frequency characteristics of Crack Acoustic Emission Signal, that is: energy distribution, frequency composition and pulse are after parameter.Stress relief energy due to crack initiation is comparatively large but spreading rate slow, and therefore the energy distribution of acoustic emission signal is comparatively even, without high-frequency component and two subpulses after length.And the impact discharging stress during Crack Extension is strong, therefore has that radio-frequency component, energy distribution are uneven, pulse is after shorter.

When engineer applied is of the present invention, generally acoustic emission acquisition system long term monitoring pneumatic equipment blades is all installed on wind energy conversion system, according to the pneumatic equipment blades acoustic emission signal that actual acquisition arrives, by signal processing method of the present invention, As time goes on constantly longitudinal comparison extract the time-frequency characteristics of acoustic emission signal, the state of Crack Extension can be judged.

The present invention by installing calibrate AE sensor and passing to signal acquiring system on wind energy conversion system leaf, the acoustic emission signal gathered is processed based on Optimization of Wavelet code reassignment scale spectrum, this signal processing method has higher time-frequency locality and the jamproof characteristic of restraint speckle, can effectively disturb by stress release treatment, reflect the time-frequency characteristics of pneumatic equipment blades Crack Acoustic Emission Signal exactly, for monitoring blade state and prediction degeneration provide effective means.

The present invention is directed to pneumatic equipment blades Crack Acoustic Emission Signal, bandwidth parameter is optimized by minimum shannon entropy method, thus determine the wavelet scale spectrum basis function of applicable crack initiation and running crack, the time frequency resolution overcoming Traditional Wavelet analytical approach can not reach best defect simultaneously, and restrained effectively noise.

Small echo code reassignment scale spectrum method of the present invention is applicable to non-linear, the non-stationary impact signal of process, can judge which kind of state crackle belongs to by the time-frequency characteristics parameter processing rear signal, avoid the difficult problem that conventional acoustic emission parameters and acoustic emission source state are difficult to set up the mechanism evaluated, therefore workable, the diagnosis of the method fast, precision is high.

Compared with prior art, beneficial effect of the present invention is: the today of constantly maximizing at wind energy conversion system, and the monitoring system of domestic enterprise to pneumatic equipment blades does not have ripe product, and main dependence on import, Some Domestic software also exists larger gap compared with abroad.Meanwhile, the design environment of external product is different from domestic wind field, cannot improve the operational reliability of wind energy conversion system in rugged surroundings.The present invention can detect the state of blade cracks dynamic expansion rapidly and accurately, ensures the security of pneumatic equipment blades and high efficiency, extends pneumatic equipment blades serviceable life, reduces maintenance cost.The present invention is simultaneously effective equally to the crack detection of the compound substance such as fiberglass, carbon fiber, be particularly suitable for the main equipment blade surface crack detection such as steam turbine, wind energy conversion system, fan blower, greatly can reduce the testing cost of various kinds of equipment blade, economic benefit is obvious.

Accompanying drawing explanation

Fig. 1 is running crack acoustic emission signal.Fig. 2 is crack initiation acoustic emission signal.

Fig. 3 is the shannon entropy of running crack acoustic emission signal and the relation curve of bandwidth parameter.

Fig. 4 is the shannon entropy of crack initiation acoustic emission signal and the relation curve of bandwidth parameter.

Fig. 5 is the program computation block diagram of the optimization of shannon Wavelet Entropy and code reassignment scale spectrum signal transacting.

Fig. 6 is the Optimization of Wavelet code reassignment scale spectrum of running crack acoustic emission signal.

Fig. 7 is the Optimization of Wavelet code reassignment scale spectrum of crack initiation acoustic emission signal.

Embodiment

Content of the present invention is described further below in conjunction with Figure of description and embodiment:

This windmill blade crack detection method, first on pneumatic equipment blades, calibrate AE sensor is installed, and the acoustic emission signal received is passed to the PCI-2 acoustic emission acquisition system of PAC company of the U.S., determine the acquisition parameters such as the sample frequency of signal, sampling length, frequency filtering; Then the code reassignment scale spectrum basis function bandwidth parameter of the acoustic emission signal of running crack and crack initiation is calculated based on Shannon entropy theory, obtain the Morle wavelet basis function of this two types Crack Acoustic Emission Signal the most applicable, calculate code reassignment scale spectrum with the wavelet basis function after optimizing, thus extract the time-frequency characteristics of pneumatic equipment blades acoustic emission signal; Then according to the position of these characteristic parameters and sensor, the extended mode of crack fault is judged according to experiment value.Concrete steps of the present invention are:

1) factor of first analyzing influence crack tip stress intensity factor.

Because pneumatic equipment blades is glass fiber reinforced plastics composite material, classical fracture theory the crack tip stress field of inapplicable this compound substance, therefore crack detection technology in the past all ignores the plastic zone impact of crackle, if the σ in plastic zone, fine crack dead ahead _r, then I mode-Ⅲ crack stress strength factor K _ibe expressed as:

$K_I={\mathrm{\σ}}_r\sqrt{\mathrm{\πL}/2}---\left(1\right)$

The present invention considers has plastic yield near crack tip, and effective dimensions L is larger than full-size(d) for its crackle, and the alternate load of pneumatic equipment blades strains relevant with plastic zone, therefore the K of amendment type (1) _ias follows:

$K_{\mathrm{IS}}=A{\mathrm{\σ}}_r^{\∞}\sqrt{\mathrm{\π}(L+\mathrm{\λ})/2}---\left(2\right)$

Coefficient A in formula (2) is relevant with loading speed and crack shape.Wherein, plastic zone size is:

$\mathrm{\λ}=\frac{1}{2\mathrm{\π}}{\left(\frac{K_I}{{\mathrm{\σ}}_{\mathrm{rs}}}\right)}^2---\left(3\right)$

Formula (3) is substituted into formula (2) can obtain:

$K_{\mathrm{IS}}=A{\mathrm{\σ}}_r^{\∞}\sqrt{\mathrm{\πL}/2}\sqrt{[1+{\left(\frac{{\mathrm{\σ}}_r^{\∞}}{{\mathrm{\σ}}_{\mathrm{rs}}}\right)}^2]}---\left(4\right)$

Formula (4) illustrates, crack tip strain rule is the function of loading speed, i.e. revised K _iSrelevant with plastic strain and crack size.Particularly under the effect of complex environment power, the stress field of crack tip no longer linearly increases rule with the increase of external applied load, and namely the spectral range of AE signal is not presenting simple linear relationship along with the rule of load change.When the spectrum signature of the loading history determined corresponding to state of crack growth and AE signal, the relation mechanism between AE signal waveform and acoustic emission source Stress transmit just can be set up.Therefore for the blade cracks of other material, as long as produce from driving source and propagation medium two angle analysis stress and propagate and the associating of AE signal characteristic, the detection method of the present invention's proposition just has universality.

2) for the pneumatic equipment blades of glass fiber reinforced plastics composite material, total length 1000mm, covering average thickness is 6mm, and mean breadth is 65mm.Root of blade is fastening, and at distance root 700mm place imposed load, loading frequency is 10Hz, measures load with statical strain indicator at load(ing) point.Indoor temperature is 20 DEG C, and whole process affects the environmental factor of blade cracks expansion without other.Acquisition system is made up of calibrate AE sensor, prime amplifier, main amplifier, digital signal simulator and record display instrument, the signal that sensor exports, through to amplify and after digital conversion, to be sent in display by data collecting card and to store by acoustic emission signal.Because quality and the signal source feature of image data, transmission path, crack propagation rate, multiple solutions material character are relevant with test macro performance, the impact of ground unrest before therefore gathering, should be reduced as far as possible.Specific practice comprises: single-frequency noise and fluctuation noise are eliminated by methods such as detected parameters setting, sensor localization calibrations.Utilize pencil-lead fracture stimulation source repeatedly to calibrate sensor before experiment, prediction channel noise level, sensitivity adjustment, propagation attenuation is measured, wave speed measurement.Determine the technical parameter of the signals collecting such as the sample frequency of For Blade Crack Fault signal, sampling length, frequency filtering and detection.The crack initiation of the pneumatic equipment blades gathered and the acoustic emission signal of stable running crack are respectively as shown in Figure 1 and Figure 2.

3) AE signal u (x, t) can be decomposed into the modulation signal of a reflection group velocity and the exponential type signal of a reflection phase velocity.It is generally acknowledged, what play a major role when stress wave is propagated in thin-slab construction is single order flexural wave and single order expansion ripple, the AE signal now gathered due to mode and frequency dispersion few, process is relatively easy.But for this kind of compound substance of pneumatic equipment blades, often adopt high frequency sensors, and pneumatic equipment blades belongs to rod component, thickness is not little relative to its width, now u (x, t) is the signal after high-order flexural wave and high-order expansion wave-wave superpose.Because expansion wave velocity is fast, discrete little, the ripple of different frequency is propagated with same speed.And flexural wave speed is slow, propagate with frequency harmonics form, expand frequency range, the feature extraction for AE signal brings very large difficulty.Therefore the present invention adopts wavelet analysis method, carries out feature extraction according to the different time-varying characteristics that different frequency component in AE signal has.The wavelet transformation of AE signal u (x, t) is defined as follows:

$W_t(a,b,\mathrm{\ψ})={\∫}_{-\∞}^{+\∞}u(x,t){\mathrm{\ψ}}_{a,b}\left(t\right)\mathrm{dt}---\left(5\right)$

In formula, the basis function (morther wavelet) that Ψ a, b (t) are wavelet transformation, parameter a and b is called translation parameters and scale parameter.The expansion of crackle is that caused by energy is constantly accumulated, all sound emission process are burst procedures, and thus when processing AE signal, way the most frequently used at present makes discretize to parameter by power series, i.e. a=2j, b=k ²j (j=0,1,2...; K=...-2 ,-1,0,1,2...), then wavelet transformation basis function can be write as:

ψ _j，k＝2 ^-j/2ψ(2 ^-jt-k)(6)

Because wavelet transformation does not lose any information, conversion is energy conservation, and therefore following formula is set up

$<x\left(t\right),x\left(t\right)>={\&Integral;\left|x\left(t\right)\right|}^2\mathrm{dt}$

$=\frac{1}{C_{\mathrm{\&psi;}}}\&Integral;a^{-2}\mathrm{da}\&Integral;{\left|W_x(a,b;\mathrm{\&psi;})\right|}^2\mathrm{db}---\left(7\right)$

4) for modulation signal and the exponential type signal of AE signal, and crackle AE signal has abrupt transients, this just requires that wavelet basis has compact sup-port, and improve partial analysis ability at frequency domain and there is rapid decay characteristic, when processing Crack Extension AE signal, prioritizing selection has the wavelet basis of compact sup-port and rapid decay.Because acoustic emission signal has the profile of similar Gauss function in a frequency domain, and Morlet small echo also has similar characteristics at frequency domain, and possess Time-Frequency Localization characteristic and good time frequency compactly supported property simultaneously, therefore the present invention adopts Morlet small echo as the morther wavelet of wavelet transformation.Morlet small echo is the negative exponential function under Gaussian envelope:

$\mathrm{\&psi;}\left(t\right)=\frac{1}{\sqrt{f_b\mathrm{\&pi;}}}\exp (-t^2/f_b)\exp \left(i2\mathrm{\&pi;}{f}_{c}t\right)---\left(8\right)$

F in formula _bbe the bandwidth parameter controlling wavelet shape, for the time frequency resolution of Balanced multi-wavelet, determine the speed degree of oscillating waveform decay.F _cbe centre frequency, determine the oscillation frequency of wavelet shapes.If f _cbe taken as constant, in certain scope, change f _b, the wavelet shapes of the most applicable analytic signal can be obtained.

5) the bandwidth parameter fb of the method Optimization of Wavelet basis function of shannon entropy is utilized.The size of Shannon entropy reflects the homogeneity of probability distribution, and least uniform probability distribution has maximum entropy, and when Shannon Wavelet Entropy is minimum, corresponding morther wavelet is exactly the small echo mated most with characteristic component.Therefore, the f making Wavelet Entropy minimum _bthe form parameter of Morlet small echo can be optimized.First, the coefficient processing of wavelet transformation is become a probability distribution sequence p _ias shown in the formula:

$p_i=\frac{\left|W_x(a_i,t)\right|}{\underset{j=1}{\overset{M}{\mathrm{\&Sigma;}}}\left|W_x(a_j,t)\right|}---\left(9\right)$

W _x(a, t) is wavelet coefficient values, and M is that wavelet coefficient gets scale parameter not of the same race, generally gets M=1024.

By p _ithe entropy calculated is called Shannon Wavelet Entropy, is defined as:

When shannon Wavelet Entropy is minimum, corresponding wavelet basis function is exactly the small echo mated most with AE signal characteristic composition.As can be drawn from Figure 3 as optimum f _bhave minimum Wavelet Entropy when=5.1, the Morlet basis function obtained thus is to calculate the small echo code reassignment scale spectrum of running crack acoustic emission signal.F after can optimizing from Fig. 4 _b=3.1, the Morlet basis function of crack initiation can be determined.

6) limit by Heisenberg uncertainty principle restricts, | W _x(a, b; ψ) | ²/ C _ψa ²it is nonsensical for accurately regarding upper certain any the energy of plane (a, b) as.But according to (7) formula, it is regarded as the energy density function of (a, b) in plane so long.That is | W _x(a, b; ψ) | ²Δ a Δ b/C _ψa ²give centered by yardstick a and time b, yardstick is spaced apart Δ a, and the time interval is the energy of Δ b.Definition SG _x(a, b; ψ)=| W _x(a, b; ψ) | ²for wavelet scale spectrum, a persevering spectrogram determining relative bandwidth can be regarded as, can the Time-Frequency Information of reflected signal, ψ (t) center angular frequency is ω ₀, (ω ₀/ a _i, a _it _c+ b _i) be SG _x(a, b; ψ) at t _cthe geometric center in moment represents the mean value of this local energy density.But crackle AE signal local energy distribution is not geometry symmetry, and in this local, energy barycenter is energy barycenter represents the energy distribution of this local better than geometric center of gravity.

The present invention adopts code reassignment scale spectrum energy barycenter is described, as shown in the formula:

${\mathrm{RSG}}_x(\hat{a},\hat{b};\mathrm{\&psi;})={\&Integral;}_{-\&infin;}^{\&infin;}{\&Integral;}_{-\&infin;}^{\&infin;}{(\hat{a}/a)}^2{\mathrm{SG}}_x(a,b;\mathrm{\&psi;})$

$\mathrm{\&delta;}(\hat{b}-b^{\&prime;}(a,b))\mathrm{\&delta;}(\hat{a}-a^{\&prime;}(a,b))\mathrm{dadb}---\left(11\right)$

$b^{\&prime;}(a,b)=b-\mathrm{Re}\left\{a\frac{W_x(a,b;{\mathrm{\&psi;}}^{\&prime;})W_x^{*}(a,b;\mathrm{\&psi;})}{{\left|W_x(a,b;\mathrm{\&psi;})\right|}^2}\right\}---\left(12\right)$

$\frac{{\mathrm{\&omega;}}_0}{a^{\&prime;}(a,b)}=\frac{{\mathrm{\&omega;}}_0}{a}+\mathrm{Im}\left\{\frac{W_x(a,b;\widehat{\mathrm{\&psi;}})W_x^{*}(a,b;\mathrm{\&psi;})}{2\mathrm{\&pi;a}{\left|W_x(a,b;\mathrm{\&psi;})\right|}^2}\right\}---\left(13\right)$

ψ′(T)＝tψ(t)， $\widehat{\mathrm{\&psi;}}\left(t\right)=\frac{\mathrm{d\&psi;}\left(t\right)}{\mathrm{dt}}---\left(14\right)$

As wavelet basis function and f _bwhen selecting suitable, in code reassignment scale spectrum, signal characteristic composition assembles the energy distribution becoming amplitude in time scale plane, then diffuses in time scale plane with the dissimilar energy of wavelet basis.

As shown in Figure 5, this calculation procedure has following steps to the program computation block diagram of the optimization of shannon Wavelet Entropy and code reassignment scale spectrum signal transacting:

(1). extract wavelet coefficient according to formula (5);

(2). according to formula (9), the coefficient processing of wavelet transformation is become a probability distribution sequence p _i;

(3). calculate shannon Wavelet Entropy according to formula (10);

(4). make fc=1, draw the relation curve of shannon Wavelet Entropy and different fb parameter;

(5). optimum fb is substituted into formula (8) and determines optimum Morlet morther wavelet;

(6). according to formula (11)--formula (14) calculates code reassignment scale spectrum;

(7). draw; (8) terminate.

7) code reassignment scale spectrum is calculated with the wavelet basis function after optimization, obtain dissimilar crack composition, thus the noise eliminated in non-stationary acoustic emission signal, excavate signal characteristic parameter, the small echo code reassignment scale spectrum of acoustic emission signal when finally obtaining Crack Extension and germinate, respectively as shown in Figure 6, Figure 7.As shown in Figure 6, the composition containing 150kHz relative high frequency in running crack signal, main energetic still concentrates on 100kHz place, and radio-frequency component and the primary energy content time interval are 0.03ms; As shown in Figure 7, be all the pulse signal of about 100kHz in crack initiation signal, and two parts of signals energy distribution is comparatively even, the time interval is 0.07ms.Difference can identify running crack and crack initiation thus.

This is that therefore the energy distribution of acoustic emission signal is comparatively even because the stress relief energy of crack initiation is comparatively large but spreading rate slow, without high-frequency component and two subpulses after length.And the impact discharging stress during Crack Extension is strong, therefore has that radio-frequency component, energy distribution are uneven, pulse is after shorter.

Claims (1)

1. a windmill blade crack detection method, is characterized in that including following steps:

(1) first on pneumatic equipment blades, calibrate AE sensor is installed, and the acoustic emission signal received is passed to acoustic emission acquisition system, determine the acquisition parameter of signal: sample frequency, sampling length, frequency filtering;

(2) bandwidth parameter of Morlet wavelet basis function is then optimized based on Shannon Wavelet Entropy, obtain the Morlet wavelet basis function mated with running crack and crack initiation sound emission signal characteristic, the code reassignment scale spectrum calculating acoustic emission signal again judges crackle state, is namely optimized and the Program extraction pneumatic equipment blades running crack of code reassignment scale spectrum signal transacting and the time-frequency characteristics parameter of crack initiation acoustic emission signal by Shannon Wavelet Entropy;

(3) extended mode of crack fault is then judged according to the time-frequency characteristics parameter of extracted Crack Acoustic Emission Signal;

Shannon Wavelet Entropy optimization in wherein said step (2) and the program computation of code reassignment scale spectrum signal transacting have following steps:

1). extract wavelet coefficient according to formula (5)

$W_x(a,b;\mathrm{\&psi;})={\&Integral;}_{-\mathrm{\&infin;}}^{\mathrm{\&infin;}}x\left(t\right){\mathrm{\&psi;}}_{a,b}\left(t\right)dt---\left(5\right)$

2). according to formula (9), the coefficient processing of wavelet transformation is become a probability distribution sequence p _i

$p_i=\frac{\left|W_x(a_i,t)\right|}{\underset{j=1}{\overset{M}{\&Sigma;}}\left|W_x(a_j,t)\right|}---\left(9\right)$

W _x(a, t) is wavelet coefficient values, and M is that wavelet coefficient gets scale parameter not of the same race;

3). calculate shannon Wavelet Entropy according to formula (10), namely

$H\left(p\right)=-\underset{i=1}{\overset{n}{\&Sigma;}}\left(p_i{\mathrm{lgp}}_i\right),\underset{i=1}{\overset{n}{\&Sigma;}}{p}_i=1---\left(10\right)$

4). make fc=1, draw shannon Wavelet Entropy and different f _bthe relation curve of parameter, in relation curve when shannon Wavelet Entropy is minimum, corresponding f _bcontrol the optimum bandwidth parameter of wavelet shape exactly;

5). by optimum f _bsubstitution formula (8) determines optimum Morlet wavelet basis function, and wavelet basis function now mates most with the acoustic emission signal collected, that is:

$\mathrm{\&psi;}\left(t\right)=\frac{1}{\sqrt{f_b\mathrm{\&pi;}}}\exp (-t^2/f_b)\exp \left(i2{\mathrm{\&pi;f}}_{c}t\right)---\left(8\right)$

F in formula _bbe the optimum bandwidth parameter that step (4) calculates, for the time frequency resolution of Balanced multi-wavelet, determine the speed degree of oscillating waveform decay, f _cbe centre frequency, determine the oscillation frequency of wavelet shapes, get f _c=1;

6). according to formula (11)--formula (14) calculates code reassignment scale spectrum, namely

${\mathrm{RSG}}_x(\hat{a},\hat{b};\mathrm{\&psi;})={\&Integral;}_{\mathrm{\&infin;}}^{-\mathrm{\&infin;}}{\&Integral;}_{-\mathrm{\&infin;}}^{\mathrm{\&infin;}}{(\hat{a}/a)}^2{\mathrm{SG}}_x(a,b;\mathrm{\&psi;})$

$\mathrm{\&delta;}(\hat{b}-b^{\&prime;}(a,b\left)\right)\mathrm{\&delta;}(\hat{a}-a^{\&prime;}(a,b\left)\right)dadb---\left(11\right)$

$b^{\&prime;}(a,b)=b-\mathrm{Re}\left\{a\frac{W_x(a,b;{\mathrm{\&psi;}}^{\&prime;})W_x^{*}(a,b;\mathrm{\&psi;})}{|W_x(a,b;\mathrm{\&psi;}){|}^2}\right\}---\left(12\right)$

$\frac{{\mathrm{\&omega;}}_0}{a^{\&prime;}(a,b)}=\frac{{\mathrm{\&omega;}}_0}{a}+\mathrm{Im}\left\{\frac{W_x(a,b;\widehat{\mathrm{\&psi;}})W_x^{*}(a,b;\mathrm{\&psi;})}{2\mathrm{\&pi;}a|W_x(a,b;\mathrm{\&psi;}){|}^2}\right\}---\left(13\right)$

${\mathrm{\&psi;}}^{\&prime;}\left(t\right)=t\mathrm{\&psi;}\left(t\right),\widehat{\mathrm{\&psi;}}\left(t\right)=\frac{d\mathrm{\&psi;}\left(t\right)}{dt}---\left(14\right);$

7). draw;

8). terminate.