CN104316277B - Air-tightness monitoring method based on sound detection Yu Blind Signal Separation - Google Patents

Abstract

The invention discloses a kind of equipment for Checking Air Tightness of Pressure Vessel based on sound detection and Blind Signal Separation and method, the present invention is directed to compressor, the air-tightness monitoring of the equipment such as air accumulator is big with air pressure in detection application, inconvenience pin problem, Gas leak position cannot be positioned, None-identified problem when there is multiple spot gas leakage, utilize the turbulent flow sounding effect produced during revealing gas, diverse location at tested container pastes multiple highly sensitive sonic transducers, and acoustical signal is carried out conditioning amplification, blind signal separation is carried out with after microprocessor collection, by comparison of wave shape and signal intensity comparison, the identification carrying out multiple spot gas leakage positions with Gas leak position, and use light flash, pipe and warn, LCD screen shows signal waveform, Gas leak position is counted with gas leakage.The present invention can be used for the on-line monitoring that leaks gas for a long time of the equipment such as pipeline, container, can also be used for products air tightness quality testing, and accuracy is high, simple operation, cheap.

Description

Air-tightness monitoring method based on sound detection Yu Blind Signal Separation

[technical field]

The present invention relates to container, compressor, pipeline, tire etc. and external environment condition and there is the air-tightness of equipment of pressure differential, gas leakage detection and on-line monitoring field, particularly to a kind of air-tightness monitoring method based on sound detection and Blind Signal Separation and device.

[background technology]

Air-tightness is a very important parameter of the equipment such as container, compressor, and conventional air-tightness detection method includes pressure application, discharge method and bubble method.Pressure application puts into sonic transducer, and inflation of pressurizeing in a reservoir, is then sealed by container.After a period of time, if container internal gas pressure reduces, then showing to there is gas leakage, decrease speed is the fastest, shows to leak gas the most serious.Hydrometry i.e. fills an effusion meter between tested container and a hermetic container, when there is leakage in tested container, and the gas flow tested container in hermetic container, then can judge to exist the order of severity of gas leakage by flow value.Bubble method is to be placed in water vat by the tested container of inflation, judges whether to leak gas by having seen whether that bubble comes up.If bubble is big and many, illustrate that gas leakage is serious, if bubble-free produces, then explanation not gas leakage.Additionally, there is a method in which trowelling method of crying, namely trowelling suds etc. outside tested container, see if there is bubble and produce.

Bubble method can determine that Gas leak position, gas leakage are counted, but cannot Quantitative Monitoring air leakage, perusal time length, easily produce fatigue.In order to overcome this problem, in recent years, person skilled develops machine vision method, i.e. use the method for image recognition to observe size and the quantity of bubble, so that it is determined that air leakage.Machine vision method requires to use high-speed camera, and wants three faces to image, the most relatively costly；It is less that tonometry is applicable to container, the situation that pressure is bigger, but relatively big at container, and when leakage rate is less, owing to rate of pressure change is too slow, resolution is the lowest.And, this method needs to be placed in container pressure sonic transducer, is extracted by holding wire, and this itself influences whether bubble-tight measurement.First, the limited pressure that lead terminal can bear, this influences whether the stress test upper limit of container, and secondly binding post itself exists air-tightness problem；Hydrometry also needs to a constant voltage equipment, operates comparatively laborious；Additionally, both approaches the most not can determine that Gas leak position, multiple spot gas leakage can not be determined whether there is.Trowelling method can determine that Gas leak position equally, and gas leakage is counted, but troublesome poeration, and cannot monitor in real time, owing to there is the probability of blast in the air-leakage test of high-pressure bottle, also with higher danger.

In recent years, some new gas leakage method of testings are occurred in that.Including supercritical ultrasonics technology, helium mass spectrography.Ultrasound wave uses the wide audio frequency characteristics of turbulent flow, uses whether sonac test equipment leaks gas.This method is convenient, cheap, directly perceived, can on-line monitoring and accuracy be good for a long time；Helium mass spectrography precision is high, but high cost.Although both approaches is accurate, but still cannot determine Gas leak position, multiple spot gas leakage can not be determined whether there is.

[summary of the invention]

It is an object of the invention to, for the deficiency of existing air-leakage test technology, it is provided that a kind of accurately, easy to use, can realize air leakage point location and multiple spot gas leakage identifies, air-tightness monitoring method based on sound detection and Blind Signal Separation and device.

To achieve these goals, the technical solution used in the present invention is:

Equipment for Checking Air Tightness of Pressure Vessel based on sound detection Yu Blind Signal Separation, including the multiple sonic transducers being arranged in outside tested container, the outfan of each sonic transducer is connected with the input of Multi-path synchronous sampler by corresponding signal condition amplifying circuit respectively, the outfan of Multi-path synchronous sampler is connected with the input of microprocessor, and the outfan of microprocessor is connected with audible-visual annunciator and display screen respectively.

The present invention is further improved by, and multiple sonic transducers are pasted onto outside tested container the position being susceptible to gas leakage；If the position that gas leakage probability is high cannot be determined, then multiple sonic transducers are evenly distributed on outside tested container.

The present invention is further improved by, and the amplification ratio of the signal condition amplifying circuit of each road sonic transducer is adjusted to unanimously, and i.e. institute's voicing sensor and same gas leakage occurring source are when identical, and the amplitude of modulate circuit output signal is the most identical.

The present invention is further improved by, and microprocessor synchronizes to sample multi-Channel Acoustic sensor, and sample frequency is not less than 50kHz.

Detection method based on sound detection Yu the equipment for Checking Air Tightness of Pressure Vessel of Blind Signal Separation, comprises the following steps that

1) judging that each road sonic transducer signal is strong and weak, line frequency analysis of spectrum of going forward side by side, if each road sonic transducer signal is the most weak, i.e. signal amplitude is in noise range, spectrum distribution width, trends towards white noise, then show generation of not leaking gas, do not do any alert process, otherwise enter step 2)；

2) each road sonic transducer signal is carried out correlation analysis two-by-two, it is judged that the concordance of each road sonic transducer signal waveform, if each road sonic transducer signal waveform is basically identical, only one of which air leakage point is described, and this air leakage point is located proximate at the sonic transducer that waveforms amplitude is relatively large, now, point on the basis of the sonic transducer position that three signal amplitudes are relatively large, with amplitude as weights, carry out position weighting, ask for occurring the position (x of air leakage point, y, z), wherein

X=(x₁ρ₁+x₂ρ₂+x₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1a)

Y=(y₁ρ₁+y₂ρ₂+y₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1b)

Z=(z₁ρ₁+z₂ρ₂+z₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1c)

In formula: (x₁,y₁,z₁) it is the maximum sonic transducer position of signal amplitude, ρ₁For this sonic transducer signal amplitude；(x₂,y₂,z₂) it is the sonic transducer position that signal amplitude the 2nd is big, ρ₂For this sonic transducer signal amplitude；(x₃,y₃,z₃) it is the sonic transducer position that signal amplitude the 3rd is big, ρ₃For this sonic transducer signal amplitude；

If waveform is inconsistent, illustrates to there are 2 or more than 2 air leakage points, enter step 3)；

3) each road sonic transducer signal is carried out blind signal separation, Hou Ge road sonic transducer signal will be separated according to step 2) process, carry out correlation analysis respectively, determine there are how many air leakage points, and air leakage point is positioned；

4) if finding to exist gas leakage, then start audible-visual annunciator and report to the police, and show that gas leakage is counted and Gas leak position on a display screen, show the sonic transducer signal waveform of air leakage point, to represent gas leakage degree simultaneously.

The present invention is further improved by, step 3) in, Blind Signal Separation method realizes as follows:

31) estimating the distribution of sonic transducer output signal, distinguishing it is super-Gaussian distribution, or secondary Gauss distribution, and the method for differentiation is as follows:

(1) taking one section of tested container gas leakage acoustical signal no less than 50 cycles, the distribution of statistics tested container gas leakage acoustical signal numerical value, transverse axis is signal numerical value, and the longitudinal axis is the value quantity of signal numerical value；

(2) with Gaussian function exp (-(v-m)/d²) distribution of step (1) tested container gas leakage acoustical signal numerical value is carried out function approximation, ask for its average m and root-mean-square d, wherein, v is the acoustical signal value of sonic transducer output, and exp () is the natural Exponents with e as the end；

(3) if the peak height of the actual distribution ratio of peak Gaussian approximation function curve of signal is the highest, then it is super-Gaussian distribution, if low, then it is time Gauss distribution；

32) mixed model making sonic transducer system is:

X=Ws (2)

In formula: x is the output signal sequence of acoustic sensor array, x=[x₁,…,x_n]^T, x_iThe output of expression i-th sonic transducer, i=1,2, n, represent total n sonic transducer；S is the input signal of acoustic sensor array, namely measured signal, s=[s₁,…,s_m]^T, s_jRepresenting jth measured signal, this measured signal is the sound that certain air leakage point causes, or external noise, j=1,2, m, represent total m measured signal, and m≤n；W represents that (n × m) ties up hybrid matrix, and

For the Blind Signal Separation of said system, the relational expression being defined below so that y=cs；

Y=Hx (4)

In formula: y represents the output vector that m ties up, y=[y₁,…,y_m]^T, y_iRepresenting the signal that i-th blind separation obtains, the signal that this blind separation obtains is the approximation of the acoustical signal that certain air leakage point causes, or the approximation of certain external noise signals；H is that (m × n) ties up matrix；C is scalar constant；The recursive algorithm of its natural water surface coatings deconvolution is:

H (k+1)=H (k)+μ (k) [H (k)-f (y (k)) u^*T(k)] (5a)

U (k)=H (k) y (k) (5b)

In formula: * is complex value adjoint operator；μ (k) represents Learning Step vector；F (y)=[f₁(y₁),…,f_m(y_m)]^T,f_i(y_i) select according to the following rules:

In formula: tanh () is tan.

The present invention is further improved by, step 2) in, the concordance of each road sonic transducer waveform realizes as follows:

(1) first to two sonic transducer signal x₁And x₂Carry out auto-correlation computation, ask for its autocorrelation coefficient r_{x1 x1}And r_{x2 x2}:

$r_{x1x1}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_1\left(i\right)x_1\left(i\right)---\left(7\right)$

$r_{x2x2}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_2\left(i\right)x_2\left(i\right)---\left(8\right)$

In formula: x₁(i) and x₂I () represents sonic transducer signal x respectively₁And x₂I-th gather data, N represents the number of data points acquired, the signal period number of the collection that it is corresponding be no less than 10；

(2) then two sonic transducer signals are carried out computing cross-correlation, ask for its cross-correlation coefficient r_x1x2:

$r_{x1x2}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_1\left(i\right)x_2\left(i\right)---\left(9\right)$

IfWherein δ_xFor the noise power in sonic transducer output signal, then it is assumed that the waveform of two signals is consistent, otherwise, inconsistent.

The present invention is further improved by, step 2) in, if there is the air leakage point of more than 2, its location positioning uses method of correlation to determine, i.e. asks for the kth signal y after separating_kWith jth sonic transducer output signal x_jBetween correlation coefficient:

$r_{xjyk}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_j\left(i\right)y_k\left(i\right)---\left(10\right)$

In formula: correlation coefficient r_xjykThen characterize the distance of kth leaking source signal and jth sonic transducer, then to all of r of kth signal after separating_xjykBe ranked up, take three maximum, and use corresponding r_xjykReplace the ρ in formula (1)₁、ρ₂And ρ₃, and the position of this air leakage point is asked for by formula (1).

Compared with prior art, the present invention has a following beneficial effect:

Present invention equipment for Checking Air Tightness of Pressure Vessel based on sound detection Yu Blind Signal Separation includes the multiple sonic transducers being arranged in outside tested container, the outfan of each sonic transducer is connected with the input of Multi-path synchronous sampler by corresponding signal condition amplifying circuit respectively, the outfan of Multi-path synchronous sampler is connected with the input of microprocessor, and the outfan of microprocessor is connected with audible-visual annunciator and LCDs respectively；Present invention container air-tightness detection method based on sound detection Yu Blind Signal Separation, the method utilizes the turbulent flow sounding effect produced during revealing gas, diverse location at tested container pastes multiple highly sensitive sonic transducers, and acoustical signal is carried out conditioning amplification, with microprocessor to multi-Channel Acoustic transducing signal synchronous acquisition, then carry out correlation analysis and carry out comparison of wave shape and signal intensity comparison and blind signal separation, the identification carrying out multiple spot gas leakage again positions with Gas leak position, if there is gas leakage, then start audible-visual annunciator light flash, pipe and warn, and on LCD screen, show signal waveform, gas leakage is counted and Gas leak position.

[accompanying drawing explanation]

Fig. 1 is air-leakage test schematic diagram based on sound detection Yu Blind Signal Separation；

Fig. 2 is three different gas leakage acoustical signals output waveform figures when testing respectively；

Fig. 3 is the gas leakage distribution of acoustical signal, Gauss distribution, secondary Gauss distribution and super-Gaussian distribution curve chart；

Fig. 4 is three diverse location sonic transducers output signal diagrams under three different gas leakage acoustical signals excitations；

Fig. 5 is the signal waveforms after separating signal.

[detailed description of the invention]

With specific embodiment, the present invention is described in further details below in conjunction with the accompanying drawings.

Fig. 1 is air-leakage test schematic diagram based on sound detection Yu Blind Signal Separation.Owing to gas tank is cylindrical, the corner of gas tank is weld, and generally at this weld easily due to rosin joint, or local welding is not sufficiently strong, causes occurring under high-pressure situations crack, thus leaks air.Therefore, in order to reduce sonic transducer quantity, outside each side of gas tank, paste sonic transducer with bottom surface.If it is apparent that the gas tank lower left corner leaks air, then No. 6 and No. 7 modulate circuit output signals maximums, because acoustical signal can decay along with the increase of propagation distance, the signal intensity of closest sonic transducer institute perception is the strongest.

So that institute's voicing sensor has identical output signal in identical distance to identical acoustical signal, the modulate circuit amplification ratio of each road sonic transducer is adjusted to unanimously, i.e. institute's voicing sensor and same gas leakage occurring source are when identical, and the amplitude of modulate circuit output signal is the most identical.Output waveform when three different gas leakage acoustical signals are tested respectively is as shown in Figure 2.

Taking the tested container gas leakage acoustical signal in one period of 100 cycle, the distribution of statistical signal numerical value, transverse axis is signal numerical value, and the longitudinal axis is the value quantity of this numerical value of signal；With Gaussian function exp (-(v-m)/d²) this distribution is carried out function approximation, ask for its average m and root-mean-square d, v is the acoustical signal value of sonic transducer output here, and exp () is the natural Exponents with e as the end；If the peak height of the actual distribution ratio of peak Gaussian approximation function curve of signal is the highest, then it is super-Gaussian distribution, if low, then it is time Gauss distribution, secondary gaussian sum super-Gaussian distribution curve is with Gaussian distribution curve as shown in Figure 3.Substantially, acoustical signal is mostly super-Gaussian distribution；

Finally, microprocessor synchronizes to sample multi-Channel Acoustic sensor, and sample frequency is 100kHz, and the sampling period is more than 50 cycles.Three signals perceived in seven sonic transducers are as shown in Figure 4.As can be seen from Figure 4, wherein No. three sonic transducer signal amplitudes are stronger, and well beyond the noise level of signal amplitude, spectrum distribution is narrower, show that there is gas leakage situation occurs, and then can be analyzed processing as follows:

Three road signals are carried out correlation analysis two-by-two, finds the most inconsistent.This point directly observes waveform it can also be seen that three not all right have obvious difference between any two.Leak gas at two accordingly, it is possible to exist.

Sound wave is pressed super-Gaussian and is processed, and signal is carried out Blind Signal Separation, obtains signal as shown in Figure 5.Signal in Fig. 5 is done again correlation analysis, finds the most inconsistent.It practice, from fig. 5, it can be seen that three waveforms are the most different, but and the basic one_to_one corresponding of Fig. 2.This is because the signal of Fig. 4 is exactly these three signal syntheses.This also illustrates, Blind Signal Separation method has efficiently separated three composite signals, it is thus achieved that be distributed independent component.

Three road signals are done Fourier transformation and carries out spectrum analysis, find that the 2nd road signal spectrum in Fig. 5 is wider, in the range of the half of sample frequency, spectral decay is the least, therefore, the 2nd road signal actually be mainly noise, this point directly observe the 2nd road signal in Fig. 5 it can also be seen that, this signal does not has much rules substantially, consistent with white noise signal waveform.Therefore, thus can determine that and leak gas at existence two；

With y in Fig. 5₁As a example by calculate the 1st separate after signal and the correlation coefficient of three sonic transducer output signals:

$r_{x1y1}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_1\left(i\right)y_1\left(i\right)$

$r_{x2y1}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_2\left(i\right)y_1\left(i\right)$

$r_{x3y1}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_2\left(i\right)y_1\left(i\right)$

With correlation coefficient r_xjykFor weights, carry out position weighting, ask for its general location:

X=(x₁r_x1y1+x₂r_x2y1+x₃r_x3y1)/(r_x1y1+r_x2y1+r_x3y1) (1a)

Y=(y₁r_x1y1+y₂r_x2y1+y₃r_x3y1)/(r_x1y1+r_x2y1+r_x3y1) (1b)

Z=(z₁r_x1y1+z₂r_x2y1+z₃r_x3y1)/(r_x1y1+r_x2y1+r_x3y1) (1c)

In formula: (x₁,y₁,z₁) it is the maximum sonic transducer position of signal amplitude；(x₂,y₂,z₂) it is the sonic transducer position that signal amplitude the 2nd is big；(x₃,y₃,z₃) it is the sonic transducer position that signal amplitude the 3rd is big, ρ₃For this sonic transducer signal amplitude；(x, y, z) be waveform be y₁There is position (noting, signal is italic, and position coordinates is roman) in gas leakage signal.

Start sound and light alarm, and show that gas leakage is counted and Gas leak position on a display screen, and display waveform, to represent gas leakage degree.

Claims (7)

1. container air-tightness detection method based on sound detection Yu Blind Signal Separation, it is characterised in that detection side The detection device that method uses includes the multiple sonic transducers being arranged in outside tested container, each sonic transducer Outfan is connected with the input of Multi-path synchronous sampler by corresponding signal condition amplifying circuit respectively, many The outfan of road synchronized sampler is connected with the input of microprocessor, the outfan of microprocessor respectively with sound Light crossing-signal is connected with display screen；

This detection method comprises the following steps that

1) judge that each road sonic transducer signal is strong and weak, line frequency analysis of spectrum of going forward side by side, if each road sonic transducer signal The most weak, i.e. signal amplitude is in noise range, spectrum distribution width, trends towards white noise, then show not leak Gas occurs, and does not do any alert process, otherwise enters step 2)；

2) each road sonic transducer signal is carried out correlation analysis two-by-two, it is judged that each road sonic transducer signal waveform Concordance, if each road sonic transducer signal waveform is basically identical, illustrates only one of which air leakage point, and this leakage Gas point is located proximate at the sonic transducer that waveforms amplitude is relatively large, now, relative with three signal amplitudes Point on the basis of bigger sonic transducer position, with amplitude as weights, carries out position weighting, asks for leaking gas Point position (x, y, z), wherein,

X=(x₁ρ₁+x₂ρ₂+x₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1a)

Y=(y₁ρ₁+y₂ρ₂+y₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1b)

Z=(z₁ρ₁+z₂ρ₂+z₃ρ₃)/(ρ₁+ρ₂+ρ₃) (1c)

In formula: (x₁,y₁,z₁) it is the maximum sonic transducer position of signal amplitude, ρ₁For this sonic transducer signal width Value；(x₂,y₂,z₂) it is the sonic transducer position that signal amplitude the 2nd is big, ρ₂For this sonic transducer signal amplitude； (x₃,y₃,z₃) it is the sonic transducer position that signal amplitude the 3rd is big, ρ₃For this sonic transducer signal amplitude；

If waveform is inconsistent, illustrates to there are 2 or more than 2 air leakage points, enter step 3)；

3) each road sonic transducer signal is carried out blind signal separation, Hou Ge road sonic transducer letter will be separated Number according to step 2) process, carry out correlation analysis respectively, determine there are how many air leakage points, and gas leakage is clicked on Row location；

4) if finding to exist gas leakage, then start audible-visual annunciator and report to the police, and show that gas leakage is counted on a display screen And Gas leak position, show the sonic transducer signal waveform of air leakage point, to represent gas leakage degree simultaneously.

Container air-tightness detection method based on sound detection Yu Blind Signal Separation the most according to claim 1, It is characterized in that, step 3) in, Blind Signal Separation method realizes as follows:

31) estimating the distribution of sonic transducer output signal, distinguishing it is super-Gaussian distribution, or secondary Gauss divides Cloth, the method for differentiation is as follows:

(1) one section of tested container gas leakage acoustical signal no less than 50 cycles, statistics tested container gas leakage are taken The distribution of acoustical signal numerical value, transverse axis is signal numerical value, and the longitudinal axis is the value quantity of signal numerical value；

(2) with Gaussian function exp (-(v-m)/d²) to step (1) tested container gas leakage acoustical signal numerical value distribution Carrying out function approximation, ask for its average m and root-mean-square d, wherein, v is the acoustical signal value of sonic transducer output, Exp () is the natural Exponents with e as the end；

(3) if the peak height of the actual distribution ratio of peak Gaussian approximation function curve of signal is the highest, then divide for super-Gaussian Cloth, if low, then it is time Gauss distribution；

32) mixed model making sonic transducer system is:

X=Ws (2)

In formula: x is the output signal sequence of acoustic sensor array, x=[x₁,…,x_n]^T, x_iRepresent that i-th sound passes The output of sensor, i=1,2, n, represent total n sonic transducer；S is the input signal of acoustic sensor array, Namely measured signal, s=[s₁,…,s_m]^T, s_jRepresenting jth measured signal, this measured signal is certain air leakage point The sound caused, or external noise, j=1,2, m, represent total m measured signal, and m≤n； W represents that (n × m) ties up hybrid matrix, and

For the Blind Signal Separation of said system, the relational expression being defined below so that y=cs；

Y=Hx (4)

In formula: y represents the output vector that m ties up, y=[y₁,…,y_m]^T, y_iRepresent the letter that i-th blind separation obtains Number, the signal that this blind separation obtains is the approximation of the acoustical signal that certain air leakage point causes, or certain The approximation of external noise signals；H is that (m × n) ties up matrix；C is scalar constant；Its natural water surface coatings uncoiling Long-pending recursive algorithm is:

H (k+1)=H (k)+μ (k) [H (k)-f (y (k)) u^*T(k)] (5a)

U (k)=H (k) y (k) (5b)

In formula: * is complex value adjoint operator；μ (k) represents Learning Step vector；F (y)=[f₁(y₁),…,f_m(y_m)]^T,f_i(y_i) select according to the following rules:

In formula: tanh () is tan.

Container air-tightness detection side based on sound detection Yu Blind Signal Separation the most according to claim 1 Method, it is characterised in that step 2) in, the concordance of each road sonic transducer waveform realizes as follows:

(1) first to two sonic transducer signal x₁And x₂Carry out auto-correlation computation, ask for its autocorrelation coefficient r_x1x1And r_x2x2:

$r_{x1x1}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_1\left(i\right)x_1\left(i\right)---\left(7\right)$

$r_{x2x2}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_2\left(i\right)x_2\left(i\right)---\left(8\right)$

In formula: x₁(i) and x₂I () represents sonic transducer signal x respectively₁And x₂I-th gather data, N represents The number of data points acquired, the signal period number of the collection that it is corresponding is no less than 10；

(2) then two sonic transducer signals are carried out computing cross-correlation, ask for its cross-correlation coefficient r_x1x2:

$r_{x1x2}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_1\left(i\right)x_2\left(i\right)---\left(9\right)$

IfWherein δ_xFor the noise power in sonic transducer output signal, Then think that the waveform of two signals is consistent, otherwise, inconsistent.

Container air-tightness detection method based on sound detection Yu Blind Signal Separation the most according to claim 1, It is characterized in that, step 2) in, if there is the air leakage point of more than 2, its location positioning uses method of correlation true Fixed, i.e. ask for the kth signal y after separating_kWith jth sonic transducer output signal x_jBetween correlation coefficient:

$r_{xjyk}=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{x}_j\left(i\right)y_k\left(i\right)---\left(10\right)$

In formula: correlation coefficient r_xjykThen characterize the distance of kth leaking source signal and jth sonic transducer, Then to all of r of kth signal after separating_xjykBe ranked up, take three maximum, and use corresponding r_xjyk Replace the ρ in formula (1)₁、ρ₂And ρ₃, and the position of this air leakage point is asked for by formula (1).

Container air-tightness detection method based on sound detection Yu Blind Signal Separation the most according to claim 1, It is characterized in that: in detection device, multiple sonic transducers are pasted onto outside tested container the position being susceptible to gas leakage Put；If the position that gas leakage probability is high cannot be determined, then multiple sonic transducers are evenly distributed on tested container Outside.

Container air-tightness detection method based on sound detection Yu Blind Signal Separation the most according to claim 1, It is characterized in that: the amplification ratio of the signal condition amplifying circuit of detection device Zhong Ge road sonic transducer is adjusted to one Causing, i.e. institute's voicing sensor and same gas leakage occurring source are when identical, the width of modulate circuit output signal It is worth the most identical.

Container air-tightness detection method based on sound detection Yu Blind Signal Separation the most according to claim 1, It is characterized in that: in detection device, microprocessor synchronizes to sample multi-Channel Acoustic sensor, and frequency of sampling Rate is not less than 50kHz.