CN101251446A - Method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform - Google Patents

Abstract

The invention relates to a rubbing sound emission signal denoising method based on discrete fraction cosine transform. The invention comprises the technical proposal that: a sound emission signal is extracted from an emission sensor of a friction point and superimposed with noise; the signal comprising noise is sampled, quantized and subframed and added with windows to obtain a discrete time sequence; the sequence is carried out discrete fraction Fourier transform with the fraction a to obtain F<a>(k), then the signal after being filtered is carried out the discrete fraction Fourier transform with the fraction 2-a and the combination of frames; the signal after being denoised is output. The rubbing sound emission signal denoising method solves the disadvantages that a sound emission source has diversity, transient property and a noise source has complexity and can not gain satisfactory effect. The rubbing sound emission signal denoising method can obtain good denoising effect to the rubbing sound emission signal under the environment of a plurality of nonstationary noise, strengthen the sound emission signal, has robustness to the diversity and transient property of the sound emission source and the complexity of the noise source, effectively solves the problem of the large calculation amount in the prior method and has wide applicability.

Description

Method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform

Technical field

The present invention relates to the noise-reduction method of bump-scrape acoustic emission signal, particularly based on the method for denoising bump-scrape acoustic emission signal of discrete fraction cosine transform.

Background technology

Acoustic emission is a kind of approaches and methods of rotating machinery bump-scrape fault detect.But the noise problem that acoustic emission faced is very serious, and the polyphyly strong noise pollution that especially brings under rotating machinery running environment makes acoustic emission signal be submerged in the noise and can't reflect the time of day of equipment.Extracting with discerning useful acoustical signal from noise is acoustic emission key issue in application content, also is that present great majority can not effectively carry out the technical bottleneck that acoustic emission detects.Therefore, noise reduction process is the focus and the difficult point of acoustic emission signal research always.

Before the present invention, the noise-reduction method of bump-scrape acoustic emission signal commonly used mainly is to utilize wavelet transformation to come acoustic emission signal is carried out noise reduction process, it is to come detected noise signal by the catastrophe point of seeking wavelet transformation, its defective is that sudden change information may be flooded by noise when noise is serious, thus the error of causing; Also with good grounds acoustic emission mode, waveform transfer mode by useful acoustical signal, and distribute cancelling noise with the different wave feature and the frequency band of noise, but, when useful acoustical signal and the mutual aliasing of noise band were serious, mode acoustic emission method was difficult to prove effective, in addition in the rotor-support-foundation system structure, sound-source signal distorts in transmittance process seriously, and the rub pattern of acoustical signal of being difficult to crash is discerned.In acoustical signal noise reduction process system, it also is a kind of effective ways that solve noise pollution that useful acoustical signal is strengthened, and the signal Enhancement Method of for example utilizing discrete cosine transform (DCT) is exactly the focus of studying at present.Why DCT has good noiseproof feature, be because the situation of change that it only comes signal acquisition by the variation that detects the whole amplitude of signal in each moment window, rather than remove to catch sign mutation with the same order of magnitude of noise, therefore to insensitive for noise, on strengthening, voice obtained good effect.Yet, because the complicacy of diversity, transient state and the noise source of acoustic emission source is utilized DCT to carry out the acoustic emission signal noise reduction process and still can not be obtained satisfied effect.

Summary of the invention

Purpose of the present invention just is to overcome the defective of above-mentioned prior art, the noise-reduction method of design, a kind of bump-scrape acoustic emission signal based on discrete fraction cosine transform of development.

Technical scheme of the present invention is:

Based on the method for denoising bump-scrape acoustic emission signal of discrete fraction cosine transform, its major technique step is:

(1) extraction of acoustic emission signal and pre-service:

(1-1) extract acoustic emission signal near first calibrate AE sensor of friction point or the second sound emission sensor that is positioned at the shaft coupling other end by bumping the acoustic emission experiment platform that rubs;

(1-2) superimposed noise on continuous bump-scrape acoustic emission signal;

(1-3) signal that contains noise is sampled, quantize, divide frame, windowing obtains discrete-time series;

(2) be the discrete fraction Fourier transform of a to this discrete-time series as mark, obtain F ^a(k);

$F^a\left(k\right)=\frac{1}{3}\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}[1+\exp \left(\frac{2\mathrm{\&pi;}(a-n)i}{3}\right)+\exp \left(\frac{4\mathrm{\&pi;}(a-n)i}{3}\right)]F^n$

When n=0, F ⁿBe time series k itself; When n=1, F ⁿDiscrete cosine transform for time series k; When n=2, F ⁿDiscrete inverse cosine conversion for time series k;

(3) calculate acoustic emission signal spectrum λ _s(k) and noise spectrum λ _n(k), and to F ^a(k) carry out filtering: (3-1) λ _s(k) estimator is: λ _s(k)=β _Sλ _s(k-1)+(1-β _S) max{F ^a(k) ²-λ _n(k), 0}.λ _n(k) estimator in quiet sound section is: λ _n(k)=β _Nλ _n(k-1)+(1-β _N) F (k), at the estimator of acoustic emission signal section be: λ _n(k)=λ _n(k-1);

(3-2) wave filter is output as: ${\hat{S}}^a\left(k\right)=\frac{{\mathrm{\&lambda;}}_s\left(k\right)}{{\mathrm{\&lambda;}}_s\left(k\right)+{\mathrm{\&lambda;}}_n\left(k\right)}{F}^a\left(k\right);$

(4) filtered signal is carried out the discrete fraction Fourier transform that mark is 2-a, and carry out the merging of frame, the signal behind the output noise reduction.

In described method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform, the acoustic emission signal that first calibrate AE sensor that more approaches true sound-source signal is extracted is good usually.

In described method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform, described sample frequency is 2MHz, and the A/D precision during quantification is 12, and the frame length of every frame is 512 sampled points when dividing frame, 50% frame is overlapping, and what select during windowing is Hamming window.

In described method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform, described noise can be white noise, workshop noise, pink colour noise, Volvo automobile noise etc.

In described method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform, described mark a=0.7～1.1 or a=1.8～2.3.

Advantage of the present invention and effect are:

1. all can obtain noise reduction preferably to the bump-scrape acoustic emission signal under the multiple nonstationary noise environment, thereby strengthen acoustic emission signal.And for the diversity of acoustic emission source, transient state and complicated noise source all have robustness.

The computation complexity of this method be in the same order of magnitude based on the noise reduction algorithm of discrete cosine transform, under the prerequisite that guarantees noise reduction, solved the big problem of classic method calculated amount effectively.

3. applicability is extensive, and this noise reduction algorithm can be applied to the enhancing and the noise reduction of acoustic emission signal, also can be applied to acoustic emission signal identification simultaneously and wait other field.

Other advantages of the present invention and effect will continue to describe below.

Description of drawings

Fig. 1---bump the acoustic emission experiment platform synoptic diagram that rubs.

Fig. 2---bump the acoustic emission waveform figure that rubs continuously.

Fig. 3---based on the denoising bump-scrape acoustic emission signal system architecture diagram of discrete fraction cosine transform.

Fig. 4---the performance chart of denoising bump-scrape acoustic emission signal effect under the different mark a.

The performance comparison result synoptic diagram of method for denoising bump-scrape acoustic emission signal under Fig. 5---4 kinds of noise circumstances.

Embodiment

Below in conjunction with drawings and Examples, technical solutions according to the invention are further elaborated.

One. the extraction of acoustic emission signal and pre-service

1. the extraction of acoustic emission signal

This test adopts 3 supportings 2 to stride rotor-support-foundation system, as shown in Figure 1.Three bearings, promptly clutch shaft bearing 2, second bearing 6, the 3rd bearing 8 all are the hydrodynamic lubrication sliding bearings, friction point 4 is near motor 1, the first calibrate AE sensor 3 close friction points 4.Second sound emission sensor 7 is positioned at the other end of shaft coupling 5.By the acoustic emission signal waveform that first calibrate AE sensor 3 and second sound emission sensor 7 extract, can be used for analyzing the decay and the discontinuous medium coupling of the process back signal distortion situation of acoustic signal propagation.Setting sample frequency in the experiment is 2MHz, and the bump-scrape acoustic emission signal waveform as shown in Figure 2 continuously.

2. the pre-service of acoustic emission signal

Stack white Gaussian noise and nonstationary noise (noise source is provided by the Dutch RSRE research centre under the Britain TNO perception association, comprises the pink colour noise, Volvo automobile noise and workshop noise) on continuous bump-scrape acoustic emission signal.Signals and associated noises is sampled, quantize, divide frame, windowing.Wherein sample frequency is 2MHz, and the A/D precision during quantification is 12, and the frame length of every frame is 512 sampled points when dividing frame, and 50% frame is overlapping, can select Hamming window during windowing.

Two. discrete fraction Fourier transform (DFCT)

By time series k, can be in the hope of its discrete Fourier transformation DCT (k) and discrete fourier inverse transformation IDCT (k).On the basis of discrete cosine transform, introduce the thought of fraction transformation.The purpose of fraction transformation is to seek a kind of information of carrying time domain and frequency domain simultaneously, thereby better describes pending information.So we construct one-period is 3 discrete fraction cosine transform (DFCT):

$F^a\left(k\right)=A_0^{a}x+A_1^a{F}^1\left(k\right)+A_2^a{F}^2\left(k\right)$ (formula 1)

Wherein, F ⁰(k)=and x, F ¹(k)=and DCT (x), F ²(k)=IDCT (x) (formula 2)

By the theory and the characteristic of fractional order operator, and the character of discrete Fourier transformation, following formula is carried out abbreviation, can find out and calculate F ^a(k) universal expression formula:

$F^a\left(k\right)=\frac{1}{3}\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}[1+\exp \left(\frac{2\mathrm{\&pi;}(a-n)i}{3}\right)+\exp \left(\frac{4\mathrm{\&pi;}(a-n)i}{3}\right)]F^n$ (formula 3)

When n=0, F ⁿBe time series k itself; When n=1, F ⁿDiscrete cosine transform for time series k; When n=2, F ⁿDiscrete inverse cosine conversion for time series k.

Three. spectrum is estimated, filtering

Fig. 3 is the denoising bump-scrape acoustic emission signal system architecture diagram based on discrete fraction cosine transform.Among the figure, λ _s(k)=E{|S ^a(k) | ², λ _n(k)=E{|N ^a(k) | ²Be respectively the spectrum and the power spectrum of acoustic emission signal.λ _s(k) estimator is:

λ _a(k)=β _Sλ _s(k-1)+(1-β _S) max{F ^a(k) ²-λ _n(k), 0} (formula 4)

λ _n(k) estimator in quiet sound section is: λ _n(k)=β _Nλ _n(k-1)+(1-β _N) F ¹(k), (formula 5)

λ _n(k) estimator in the acoustic emission signal section is: λ _n(k)=λ _n(k-1); (formula 6)

To be signal F after the discrete fraction Fourier transform of a through mark ^a(k) send into wave filter, wave filter is output as:

${\hat{S}}^a\left(k\right)=\frac{{\mathrm{\&lambda;}}_s\left(k\right)}{{\mathrm{\&lambda;}}_s\left(k\right)+{\mathrm{\&lambda;}}_n\left(k\right)}{F}^a\left(k\right)$ (formula 7)

Four. filtered signal is carried out the discrete fraction Fourier transform that mark is 2-a, and carry out the merging of frame, the signal behind the output noise reduction.

Five. performance evaluation

Fig. 4 has provided the performance curve of denoising bump-scrape acoustic emission signal effect under the different mark a, and output signal-to-noise ratio (SNR) is big more, and noise reduction is good more.In the time of can finding a=1, output signal-to-noise ratio is not a maximum, when a=0.7～1.1 and a=1.8～2.3, can obtain noise reduction preferably, and when the a=0.4 or the a=2.6 left and right sides, system performance is the poorest.

Fig. 5 has provided measured discrete cosine transform and based on the performance of mark discrete cosine transform noise reduction algorithm relatively.As can be seen, the DFCT method will generally be better than noise reduction algorithm based on standard DCT for the noise reduction of the acoustic emission signal under 4 kinds of noise circumstances.

Aspect the complexity of algorithm, to compare with the DCT method, the DFCT method has increased 3N multiplication and 2N sub-addition, and its computation complexity will be a little more than the DCT method, but they are in same magnitude.

By above analysis and performance evaluation, can find, be better than the noise reduction algorithm of measured discrete cosine transform for the noise reduction of acoustic emission signal based on the discrete fraction Fourier transform method, and the computation complexity of DFCT algorithm and DCT algorithm are in the same order of magnitude, are the effective ways of bump-scrape acoustic emission signal being carried out noise reduction process.

The scope that the present invention asks for protection is not limited only to the description of this embodiment.

Claims (5)

1. based on the method for denoising bump-scrape acoustic emission signal of discrete fraction cosine transform, the steps include:

(1) extraction of acoustic emission signal and pre-service:

(1-1) extract acoustic emission signal near first calibrate AE sensor of friction point or the second sound emission sensor that is positioned at the shaft coupling other end by bumping the acoustic emission experiment platform that rubs;

(1-2) superimposed noise on continuous bump-scrape acoustic emission signal;

(1-3) signal that contains noise is sampled, quantize, divide frame, windowing obtains discrete-time series;

(2) be the discrete fraction Fourier transform of a to this discrete-time series as mark, obtain F ^a(k);

$F^a\left(k\right)=\frac{1}{3}\underset{n=0}{\overset{2}{\mathrm{\&Sigma;}}}[1+\exp \left(\frac{2\mathrm{\&pi;}(a-n)i}{3}\right)+\exp \left(\frac{4\mathrm{\&pi;}(a-n)i}{3}\right)]F^n$

When n=0, F ⁿBe time series k itself; When n=1, F ⁿDiscrete cosine transform for time series k; When n=2, F ⁿDiscrete inverse cosine conversion for time series k;

(3) calculate acoustic emission signal spectrum λ _s(k) and noise spectrum λ _n(k), and to F ^a(k) carry out filtering: (3-1) λ _s(k) estimator is: λ _s(k)=β _Sλ _s(k-1)+(1-β _S) max{F ^a(k) ²-λ _n(k), 0}; λ _n(k) estimator in quiet sound section is: λ _n(k)=β _Nλ _n(k-1)+(1-β _N) F (k), at the estimator of acoustic emission signal section be: λ _n(k)=λ _n(k-1);

(3-2) wave filter is output as: ${\hat{S}}^a\left(k\right)=\frac{{\mathrm{\&lambda;}}_s\left(k\right)}{{\mathrm{\&lambda;}}_s\left(k\right)+{\mathrm{\&lambda;}}_n\left(k\right)}{F}^a\left(k\right);$

(4) filtered signal is carried out the discrete fraction Fourier transform that mark is 2-a, and carry out the merging of frame, output signal.

2. the method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform according to claim 1 is characterized by, and the acoustic emission signal of extracting from first calibrate AE sensor in the step (1-1) is good.

3. the method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform according to claim 1, it is characterized by, described sample frequency is 2MHz, A/D precision during quantification is 12, the frame length of every frame is 512 sampled points when dividing frame, 50% frame is overlapping, and what select during windowing is Hamming window.

4. the method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform according to claim 1 is characterized by, and described noise can be white noise, workshop noise, pink colour noise, Volvo automobile noise.

5. the method for denoising bump-scrape acoustic emission signal based on discrete fraction cosine transform according to claim 1 is characterized in that mark a=0.7～1.1 or a=1.8～2.3.

Images