CN104123438A - Method for recognizing second noise transmission channel model - Google Patents

Abstract

The invention provides a method for recognizing a second noise transmission channel model. White noise is simulated and generated according to a standard white noise generating method and transmitted to air through a loudspeaker, and an error sensor detects loudspeaker sound wave signals which are attenuated by air; the white noise is filtered through a filter to generate signals; the initial parameter value of the filter is set as zero; the signals generated after the white noise is filtered through the filter are calculated; error between the signals detected by the error sensor and the filtered signals is calculated; the error is judged, and if the error does not meet the preset requirement, the coefficient of the filter is updated, and the calculation of the filtered signals is continued; if the error meets the preset requirement, calculation is stopped, and the coefficient of the filter is output. Compared with a traditional method, with the method, a more accurate noise transmission model can be obtained, the method is very suitable for recognizing the second channel model between the loudspeaker and the error sensor in active noise control design, and a basis is provided for achieving a good denoising effect in active noise control.

Description

Noise transfer second channel identification Method

Technical field

The present invention relates to iron and steel metallurgical industry Noise Control field, be specifically related to a kind of noise transfer second channel identification Method.

Background technology

In the industrial circles such as iron and steel, metallurgy, due to the operation of various big machineries, particularly, such as the operation of the high-energy noise sources such as pressure reducing valve of blast furnace group, cause the head of district of the steel plant phase among serious noise pollution.Owing to can causing staff's health and psychology to sustain damage for a long time, so in engineering, often adopt initiatively denoising and two kinds of method noise decrease energy of passive denoising in this high intensity noise pollutes.In active noise control technique, in order to reach good denoising effect, need identification loudspeaker to error receiving end noise transfer second channel model before, classic method is often difficult to identification and obtains noise transfer model accurately.

Summary of the invention

The technical problem to be solved in the present invention is: a kind of noise transfer second channel identification Method is provided, can obtains model parameter more accurately.

The present invention addresses the above problem taked technical scheme to be: a kind of noise transfer second channel identification Method, is characterized in that: it comprises the following steps:

1) making the sampling period is T _s, according to the white noise production method simulation of standard, generate white noise y (k), y (k) to be propagated in air through loudspeaker, error pick-up detects the signal r ' of loudspeaker sound wave after attenuation of air (k); Signal r (k)=C (z) y (k) of white noise y (k) after filter filtering, C (z) represents the transport function between analogue noise source and error pick-up, expression formula is C (z)=c ₀+ c ₁z ^-1+ c ₂z ^-2+ ... + c _nz ^-n, wherein symbol " z " represents shifting function, n is filter order, c ₀, c ₁..., c _nfor filter coefficient;

2) establish the initial value of filter parameter and be 0;

3) calculate the signal r (k) of white noise after filter filtering, c wherein _i(k) represent filter coefficient c _ivalue when the k time sampling;

4) error e (k) of the signal r ' that error of calculation sensor detects (k) and between filtered signal r (k), e (k)=r ' (k)-r (k);

5) size of judgement e (k):

If e (k) does not reach preset requirement, upgrade filter coefficient, make c _i(k+1)=c _i(k)+μ * e (k) * y (k-i), i=0 wherein, 1,2 ..., n-1, μ is iteration step length, μ satisfies condition p wherein _ypower for white noise y (k); Return to step 3);

If e (k) reaches preset requirement, stop calculating output filter coefficient.

Press such scheme, step 5) in, if e (k)>=0.1 upgrades filter coefficient, make c _i(k+1)=c _i(k)+μ * e (k) * y (k-i), i=0 wherein, 1,2 ..., n-1, μ is iteration step length, μ satisfies condition p wherein _ypower for white noise y (k); Return to step 3);

If e (k) is <0.1, stop calculating output filter coefficient.

Beneficial effect of the present invention is: compare with classic method, the inventive method can obtain noise transfer model more accurately, while being applicable to very much active noise controlling design, the second channel model between loudspeaker and error-detecting sensor is carried out to identification, for active noise controlling obtains good denoising effect, provide the foundation.

Accompanying drawing explanation

Fig. 1 is identification principle figure of the present invention;

Fig. 2 is embodiment 1 median filter parameter c ₀convergence curve;

Fig. 3 is embodiment 1 median filter parameter c ₁convergence curve;

Fig. 4 is embodiment 1 median filter parameter c ₂convergence curve;

Fig. 5 is embodiment 1 median filter parameter c ₃convergence curve;

Fig. 6 is embodiment 1 median filter parameter c ₄convergence curve;

Fig. 7 is embodiment 1 median filter parameter c ₅convergence curve;

Fig. 8 is embodiment 1 median filter parameter c ₆convergence curve;

Fig. 9 is embodiment 1 median filter parameter c ₇convergence curve;

Figure 10 is embodiment 1 median filter parameter c ₈convergence curve;

Figure 11 is embodiment 1 median filter parameter c ₉convergence curve;

Figure 12 is embodiment 1 median filter parameter c ₁₀convergence curve;

Figure 13 is error e in embodiment 1 (k) change curve.

Embodiment

Below in conjunction with instantiation and accompanying drawing, the present invention will be further described.

On some directions of propagation, sound is along straight line to propagating at a distance, and first sensor of foremost is original noise transducer, and centre is loudspeaker, the error pick-up of finally placing.Wherein original noise transducer be for check noise source along the direction of propagation noise signal on original noise transducer position, and loudspeaker is the sound wave offseting with noisiness sending by active noise control technique, last error pick-up is for detecting the device of original noise sound wave and loudspeaker sound wave, because the sound wave that original signal and loudspeaker send is cancelled each other in air, therefore actual what receive is the acoustic signals that original noise and loudspeaker sound wave superpose mutually to error pick-up, it can check the denoising performance of noise active control technology good and bad.Because noise is to the so-called first passage of travel path of error pick-up, so the sound wave just loudspeaker being sent is called second channel to the travel path of error pick-up.

A kind of noise transfer second channel identification Method that the present embodiment provides, it comprises the following steps:

1) according to accompanying drawing 1, set up noise transfer second channel, making the sampling period is T _s, according to the white noise production method simulation of standard, generate white noise y (k), y (k) to be propagated in air through loudspeaker, error pick-up detects the signal r ' of loudspeaker sound wave after attenuation of air (k); Signal r (k)=C (z) y (k) of white noise y (k) after filter filtering, C (z) represents the transport function between analogue noise source and error pick-up, expression formula is C (z)=c ₀+ c ₁z ^-1+ c ₂z ^-2+ ... + c _nz ^-n, wherein symbol " z " represents shifting function, n is filter order, c ₀, c ₁..., c _nfor filter coefficient;

2) establish the initial value of filter parameter and be 0;

3) calculate the signal r (k) of white noise after filter filtering, c wherein _i(k) represent filter coefficient c _ivalue when the k time sampling;

4) error e (k) of the signal r ' that error of calculation sensor detects (k) and between filtered signal r (k), e (k)=r ' (k)-r (k);

5) size of judgement e (k):

If e (k) does not reach preset requirement (judging e (k)>=0.1 o'clock in the present embodiment), upgrade filter coefficient, make c _i(k+1)=c _i(k)+μ * e (k) * y (k-i), i=0 wherein, 1,2 ..., n-1, μ is iteration step length, μ satisfies condition p wherein _ypower for white noise y (k); Return to step 3);

If e (k) reaches preset requirement (while judging e (k) <0.1 in the present embodiment), stop calculating output filter coefficient.

Principle of work of the present invention is: C in accompanying drawing 1 (z) represents noise second channel transport function, and the application regards noise second channel model as a finite impulse wave filter, and its expression-form is C (z)=c ₀+ c ₁z ^-1+ c ₂z ^-2+ ... + c _nz ^-n, wherein symbol " z " represents shifting function, n is filter order, c ₀, c ₁..., c _nfor filter coefficient; Y (k) represents white noise, and r (k) is the filtered signal of device after filtering, and wherein k represents signal sampling sequence number; R ' (k) represents that error pick-up receives signal, and e (k) represents the difference of r ' (k) and between r (k).

While supposing that noise is propagated in air, transport function between from loudspeaker to error pick-up is H (z), so in accompanying drawing 1 white noise y (k) after the attenuation of air error pick-up detect the signal obtain become r ' (k)=H (z) y (k), and signal r (k)=C (z) y (k), if while representing that the error of r ' (k) and between r (k) is zero, wave filter C (z) is noise second channel TRANSFER MODEL.

If r ' (k)=C (z) y (k), C (z) is noise second channel transport function accurately, so above formula can be converted to following form:

$r^{\&prime;}\left(k\right)=\left[\begin{array}{ccccc}y\left(k\right)& y(k-1)& y(k-2)& \&CenterDot;\&CenterDot;\&CenterDot;& y(k-n)\end{array}\right]\left[\begin{array}{c}{c}_0\\ c_1\\ c_2\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ c_n\end{array}\right],$ Wherein y (k-n) represents the random signal values of the k-n time sampling, if order $c=\left[\begin{array}{c}{c}_0\\ c_1\\ c_2\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ c_n\end{array}\right],$ There is following linear equation:

In above-mentioned linear equation, parameter c is unknown number, and according to stochastic approximation optimum theory, the iterative computation formula that can obtain parameter c is:

In above formula, symbol " T " representing matrix is asked transposition computing, due to therefore being launched into matrix form is:

$\left[\begin{array}{c}{c}_0(k+1)\\ c_1(k+1)\\ c_2(k+1)\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ c_n(k+1)\end{array}\right]=\left[\begin{array}{c}{c}_0\left(k\right)\\ c_1\left(k\right)\\ c_2\left(k\right)\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ c_n\left(k\right)\end{array}\right]+\mathrm{\&mu;}\&times;\left[\begin{array}{c}y\left(k\right)\\ y(k-1)\\ y(k-2)\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ y(k-n)\end{array}\right]\&times;e\left(k\right)$

For each element in vectorial c, can obtain its iteration expression formula and be:

c _i(k+1)＝c _i(k)+μ×e(k)×y(k-i)

I ∈ in above formula [0 n].

The a certain active noise controlling system of take is example, and wherein loudspeaker distance error signal transducer is 20 meters, and for the second channel model between identification loudspeaker and error pick-up, it is T that present case makes the sampling period _s=0.00001 second, according to step 1) generate white noise y (k), through loudspeaker, in air, propagate.If second channel model is 11 rank, the initial parameter of designing filter is 0, i.e. c ₀=0, c ₁=0 ..., c ₁₁=0.According to step 3)-5) iterative computation filter parameter progressively, after the calculating in 0.1 second, error e (k) <0.1, the filter parameter that identification obtains is respectively c ₀=0.0147475299960828, c ₁=0.0237478779311739, c ₂=0.0653174737962928, c ₃=0.124706748500808, c ₄=0.179643922151532, c ₅=0.20221317008156, c ₆=0.179643922151532, c ₇=0.124706748500808, c ₈=0.0653174737962928, c ₉=0.0237478779311739, c ₁₀=0.00474752999608277, each parameter iteration convergence process curve is as shown in accompanying drawing 2-12, and error e (k) change curve as shown in Figure 13.

Above embodiment is only for calculating thought of the present invention and feature are described, its object is to make those skilled in the art can understand content of the present invention and implement according to this, and protection scope of the present invention is not limited to above-described embodiment.So the disclosed principle of all foundations, equivalent variations or the modification that mentality of designing is done, all within protection scope of the present invention.

Claims (2)

1. a noise transfer second channel identification Method, is characterized in that: it comprises the following steps:

1) making the sampling period is T _s, according to the white noise production method simulation of standard, generate white noise y (k), y (k) to be propagated in air through loudspeaker, error pick-up detects the signal r ' of loudspeaker sound wave after attenuation of air (k); Signal r (k)=C (z) y (k) of white noise y (k) after filter filtering, C (z) represents the transport function between analogue noise source and error pick-up, expression formula is C (z)=c ₀+ c ₁z ^-1+ c ₂z ^-2+ ... + c _nz ^-n, wherein symbol " z " represents shifting function, n is filter order, c ₀, c ₁..., c _nfor filter coefficient;

2) establish the initial value of filter parameter and be 0;

3) calculate the signal r (k) of white noise after filter filtering, c wherein _i(k) represent filter coefficient c _ivalue when the k time sampling;

4) error e (k) of the signal r ' that error of calculation sensor detects (k) and between filtered signal r (k), e (k)=r ' (k)-r (k);

5) size of judgement e (k):

If e (k) does not reach preset requirement, upgrade filter coefficient, make c _i(k+1)=c _i(k)+μ * e (k) * y (k-i), i=0 wherein, 1,2 ..., n-1, μ is iteration step length, μ satisfies condition p wherein _ypower for white noise y (k); Return to step 3);

If e (k) reaches preset requirement, stop calculating output filter coefficient.

2. noise transfer second channel identification Method according to claim 1, is characterized in that: step 5) in, if e (k)>=0.1 upgrades filter coefficient, make c _i(k+1)=c _i(k)+μ * e (k) * y (k-i), i=0 wherein, 1,2 ..., n-1, μ is iteration step length, μ satisfies condition p wherein _ypower for white noise y (k); Return to step 3);

If e (k) is <0.1, stop calculating output filter coefficient.