CN103033154A - Autoregression spectrum analysis method for improving ultrasonic testing time resolution - Google Patents

Abstract

The invention relates to an autoregression spectrum analysis method for improving ultrasonic testing time resolution. The method comprises that ultrasonic echo signals y(t) are collected specific to to-be-tested liquid; under the condition that testing parameters are not changed, the thickness of the to-be-tested liquid is set to be zero, the ultrasonic echo signals are collected and are used as reference signals h(t) in a Wiener filtering process; the ultrasonic echo signals y(t) are subjected to Wiener filtering, an impulse response frequency spectrum E(w) of the end face of a liquid layer is obtained; an autoregression model order K is chosen, the autoregression coefficient under the order is calculated with an R-W method; frequency spectrum reconstruction is carried out by choosing a frequency attenuation window with a certain width; a reconstructed frequency spectrum is subjected to inverse Fourier transform so that a time-domain signal x(t) is obtained, according to the wave crest position and the wave trough position of the time-domain signal x(t), the position of echo on the upper surface and the lower surface of the liquid layer, and accordingly the transmission time of sound wave in the liquid layer is determined, and the thickness of the liquid layer can be calculated by combining the transmission time with the transmission speed of the sound wave in the liquid layer.

Description

A kind of Autoregressive Spectrum Analysis method for improving the Ultrasonic Detection temporal resolution

Technical field

The present invention relates to a kind of Autoregressive Spectrum Analysis method for improving the ultrasound detection signal temporal resolution.

Background technology

As one of five large conventional Dynamic Non-Destruction Measurements, ultrasonic technology can be used for the measurement of thickness of dielectric layers.Based on ultrasonic propagation time in the medium and velocity of propagation, Conventional Ultrasound reflection echo technology can realize the measurement of large scale thickness of dielectric layers.But when thickness of dielectric layers hour, the upper and lower surperficial ultrasonic reflection echo of dielectric layer will be superimposed, be difficult to directly obtain ultrasound wave by the travel-time at interface, liquid level both sides from time domain waveform, thereby can't directly utilize poor method of Conventional Ultrasound echo time to realize the film dielectric layer thickness measure.For example the oil film thickness in the hydrostatic slideway generally arrives between the hundreds of micron at tens microns, and oil film thickness has a significant impact for every mechanical property of hydrostatic slideway.Realize that the thickness measure of thin oil film just must improve the temporal resolution of ultrasound echo signal.

In order to overcome the stack of thin layer liquid level upper and lower surface ultrasonic reflection echo time domain, the ripple bag is difficult to the difficult points such as differentiation, can adopt the high frequency ultrasound probe, reduces the pulsewidth of measuring system.The reduction of the raising of survey frequency and system's pulsewidth can both improve the measuring accuracy of measuring system, and the reflection echo of originally stack is made a distinction, and improves the precision that the thin layer liquid thickness is measured.But the cost of manufacture of the cost of manufacture of measuring system hardware circuit and high frequency ultrasound probe all will improve greatly, so its widespread use has been subject to very big restriction.

Summary of the invention

The object of the present invention is to provide a kind of high precision, the easy to operate Autoregressive Spectrum Analysis method that is used for improving the Ultrasonic Detection temporal resolution.On the basis of the survey frequency that need not to improve measuring system, reduction measuring system pulsewidth, the method is by processing the ultrasound echo signal of measuring, the method that is integrated application Wiener filtering and Autoregressive Spectrum Analysis is processed signal, improve the temporal resolution of echoed signal, the liquid level upper and lower surface ultrasonic reflection echo area that originally is superimposed is separated, calculate the thin layer liquid thickness according to ultrasonic propagation time in the liquid level and velocity of propagation again.Therefore, the method can improve the temporal resolution of ultrasound detection signal, and can realize the thickness measure of thin layer liquid, and not only measuring accuracy is high, and easy operating.

A kind of Autoregressive Spectrum Analysis method for improving the Ultrasonic Detection temporal resolution that the present invention proposes, its ultimate principle is:

For thin layer liquid thickness measuring system, in time domain, the ultrasound echo signal y (t) that measures can be expressed as the convolution of the impulse response u (t) of the impulse response e (t) of liquid level end face and measuring system, consider system noise n (t), y (t) can be expressed as:

y(t)-u(t)*e(t) （4）

U (t) is called the distortion small echo of measuring system, and by following formula (1) as can be known, the distortion small echo of measuring system and system noise signal all can exert an influence to the ultrasound echo signal y (t) that measures.E (t) is the impulse response of liquid level end face, embodiment liquid level characteristic information that can be complete.Therefore wish to pass through signal processing method filtering measuring system distortion small echo and system noise to the impact of measuring-signal, obtain e (t), calculate liquid layer thickness by e (t).In the actual signal processing procedure, can gather suitable signal and be similar to the distortion small echo that replaces measuring system as the reference signal.

Wiener filtering is as a kind of deconvolution filtering method, can filtering measuring system distortion small echo and system noise on the impact of measuring-signal, and can reach good signal treatment effect.S filter is processed signal in frequency domain, and calculates the impulse response of liquid level end face by following formula:

$E\left(\mathrm{\ω}\right)=\frac{Y\left(\mathrm{\ω}\right)H*\left(\mathrm{\ω}\right)}{{\left|H\left(\mathrm{\ω}\right)\right|}^2+Q}---\left(1\right)$

When between two steel test blocks, not having liquid, gather the lower surface echo of steel test block, with this signal as the reference signal h in the Wiener filtering process (t).H (ω) is the frequency spectrum of reference signal, and H* (ω) is the conjugate complex number of H (ω), and Y (ω) is the frequency spectrum of the ultrasound echo signal that measures, and Q has reflected the noise situations of measuring-signal.Usually Q can be taken as | H (ω) | ²1%.

E (ω) is done inversefouriertransform can obtain time-domain signal e (t), use the method for Autoregressive Spectrum Analysis and can further process e (t), signal to noise ratio (S/N ratio) and the temporal resolution of e (t) are further enhanced.

The signal spectrum equivalence that Wiener filtering is obtained is autoregressive process, and the method for Autoregressive Spectrum Analysis is also processed in frequency domain.

At first need select the exponent number of autoregressive model, model order is high can to improve the precision that signal is processed, but for fairly simple ultrasound echo signal, the raising of model order can't significantly improve the signal processing accuracy can reduce arithmetic speed on the contrary.Measure ultrasound echo signal for oil film thickness and process, Autoregressive is taken as 20, can satisfy the accuracy requirement that signal is processed fully.

Next is used the Burger method and calculates autoregressive coefficient under this exponent number.

Select at last the high part signal of Signal-to-Noise as known signal, use the signal on autoregressive model prediction known signal both sides, namely select the frequency window of one fixed width corresponding to frequency domain, frequency spectrum within the frequency window as known frequency spectrum, is used the frequency spectrum of both sides outside the autoregressive model prediction frequency window.Choosing of frequency window can exert an influence to the signal processing, for this thin layer liquid thickness measuring method, through many experiments selection-6db decay window.The width of supposing frequency window corresponds to and is m-n, according to the frequency spectrum of signal outside the following two formula predict frequency windows.

${\hat{x}}_p=-\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{a}_i^{*}{x}_{p+i},p=\mathrm{1,2},...,m-1---\left(2\right)$

${\hat{x}}_q=-\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{a}_i{x}_{q-i},q=n+1,...,N---\left(3\right)$

Wherein m is the left margin subscript of frequency window, and n is the right margin subscript of frequency window.K is the exponent number of autoregressive model, a _iBe autoregressive coefficient,

A _iConjugate complex number, The frequency spectrum in left side outside the frequency window, Be the frequency spectrum on right side outside the frequency window, N is sampling number.

The frequency spectrum that keeps E (ω) to be positioned within the frequency window is constant, respectively with calculating

Replacement E (ω) is positioned at left side frequency spectrum outside window, uses

Replacement E (ω) is positioned at right side frequency spectrum outside window.Thereby obtain the frequency spectrum X (ω) of reconstruct.

The frequency spectrum X (ω) that reconstruct is obtained carries out the time-domain signal after inversefouriertransform can obtain processing.Velocity of propagation in liquid level can be determined liquid layer thickness according to the Wave crest and wave trough position in the time-domain signal and ultrasound wave.

Technical scheme of the present invention is as follows:

The thin layer liquid thickness experiments of measuring system that uses among the present invention comprises computing machine, oscillograph, impulse ejection/receiving instrument, sensor, oil lamella, steel test specimen, clearance gauge, as shown in Figure 1.What adopted by impulse ejection/receiving instrument 3 stimulus sensor 4(sensors 4 is the 5MHz longitudinal wave probe that olympus company produces) produce ultrasonic signal, and by sensor 4 received ultrasonic signals, gather and the storage ultrasonic signal by oscillograph 2, at last this signal is transferred to computing machine 1, use MATLAB numerical analysis platform the echo data that receives is processed.Oil film 5 thickness between the steel test specimen 6 are set by clearance gauge 7.Concrete measuring process is:

1) for the liquid level between upper strata steel test block and the lower floor's steel test block, uses measuring system excitation ultrasound wave and gather ultrasound echo signal y (t);

2) under the condition that does not change measurement parameter, with liquid layer thickness zero setting, use measuring system excitation ultrasound wave and also gather ultrasound echo signal, with this signal as the reference signal h in the Wiener filtering process (t);

3) ultrasound echo signal y (t) is carried out Wiener filtering, respectively y (t) and h (t) are done inversefouriertransform, obtain H (ω) and Y (ω), application formula 1 is calculated the impulse response E (ω) of liquid level end face

4) selected Autoregressive is 20, and application Burger method is calculated the autoregressive coefficient under this exponent number

5) selection-6db decay window, this window is applied to the E (ω) that step 3 obtains, and the part that E (ω) is positioned at window is as known frequency spectrum, according to the autoregressive coefficient of calculating in the step 4, the frequency spectrum on application formula 2 and formula 3 predict frequency window both sides is reconstructed frequency spectrum

6) frequency spectrum of reconstruct carried out inversefouriertransform, obtain time-domain signal x (t) after treatment

7) according to the Wave crest and wave trough position among the time-domain signal x (t) after processing, determine the position of liquid level upper and lower surface echo, determine the travel-time of sound wave in liquid level with this.Can calculate the thickness of liquid level in conjunction with the velocity of propagation of sound wave in liquid

Beneficial effect

(1) need not to increase on the basis of measuring system cost, by the ultrasound echo signal that gathers is processed, improve the temporal resolution of echoed signal, realized the thickness measure to thin layer liquid, solved the problem that echoed signal adjacent wave bag is difficult to distinguish.(2) measuring system can be carried out Real-time Collection, and the parameter in the signal processing does not need to make very large change, can tackle the situation that liquid layer thickness changes, so that the measurement of thin layer liquid thickness has very large dirigibility.

Description of drawings

Fig. 1 is a kind of Autoregressive Spectrum Analysis method thin layer liquid thickness measuring system schematic diagram for improving the Ultrasonic Detection temporal resolution of the present invention;

Fig. 2 is a kind of Autoregressive Spectrum Analysis method flow diagram for improving the Ultrasonic Detection temporal resolution of the present invention;

Fig. 3 uses the original time-domain signal that the present invention measures 350 μ m water film thickness

Fig. 4 is the time-domain signal of using after the present invention processes 350 μ m water film thickness

Wherein: 1-computing machine; 2-oscillograph; 3-pulse generation/receiving instrument; 4-sensor; 5-oil lamella; 6-steel test specimen; 7-clearance gauge.

Embodiment

1) for the ease of operation and enforcement, carry out experiment measuring with water layer as liquid level, the steel specimen thickness is 10mm, and clearance gauge is put between the upper and lower steel test specimen, constructs 350 μ m thickness water layers.Sample frequency is set as 100MHz, uses above-mentioned thin layer liquid thickness measuring system excitation and gathers ultrasound echo signal y (t);

2) under the condition that does not change measurement parameter, with water film thickness zero setting, gather the upper surface echo of lower floor's steel test block.With this signal as the reference signal h in the Wiener filtering process (t);

3) ultrasound echo signal y (t) is carried out Wiener filtering, respectively y (t) and h (t) are done inversefouriertransform, obtain H (ω) and Y (ω), application formula 1 is calculated the impulse response E (ω) of moisture film end face;

4) selected Autoregressive is 20, and application Burger method is calculated the autoregressive coefficient under this exponent number;

5) selection-6db decay window, this window is applied to the E (ω) that step 3 obtains, and the part that E (ω) is positioned at window is as known frequency spectrum, according to the autoregressive coefficient of calculating in the step 4, the frequency spectrum on application formula 2 and formula 3 predict frequency window both sides is reconstructed frequency spectrum

6) frequency spectrum of reconstruct carried out inversefouriertransform, obtain time-domain signal x (t) after treatment

7) the wave trough position point a of the time-domain signal x (t) after the processing, crest location point b, the respectively position of corresponding moisture film upper surface, lower surface echo.Show such as Fig. 4: sound wave is 213-262 at the corresponding a-b of the number of data points of water transmission, and sample frequency is 100Mhz, and the corresponding time interval is 49 μ s, and the theoretical velocity of wave of compressional wave in water is 1473m/s.Experiment determines that water layer thickness is 360.15 μ m as calculated, and relative error is 2.9%

Under laboratory condition, carried out again many experiments research for the water layer 150 μ m-450 μ m of different-thickness, the order of accuarcy of experimental result is all within the error allowed band, so the method can realize measuring requirement fully.

Claims (3)

1. Autoregressive Spectrum Analysis method that be used for to improve the Ultrasonic Detection temporal resolution, based on the measuring system that is formed by computing machine, oscillograph, impulse ejection/receiving instrument, sensor, steel test specimen, impulse ejection/receiving instrument stimulus sensor produces ultrasonic signal, ultrasonic signal passes upper strata steel test specimen, testing liquid, reflects through lower floor's steel test specimen, reflected signal passes through sensor, impulse ejection/receiving instrument, oscillograph successively, finally transfer to computing machine, it is characterized in that may further comprise the steps:

1) for the liquid level between upper strata steel test block and the lower floor's steel test block, uses measuring system excitation ultrasound wave and gather ultrasound echo signal y (t);

2) under the condition that does not change measurement parameter, with liquid layer thickness zero setting, use measuring system excitation ultrasound wave and also gather ultrasound echo signal, with this signal as the reference signal h in the Wiener filtering process (t);

3) ultrasound echo signal y (t) is carried out Wiener filtering, be specially respectively y (t) and h (t) are done inversefouriertransform, corresponding Y (ω) and the H (ω) of obtaining, application formula 1 is calculated the impulse response frequency spectrum E (ω) of liquid level end face

$E\left(\mathrm{\ω}\right)=\frac{Y\left(\mathrm{\ω}\right)H*\left(\mathrm{\ω}\right)}{{\left|H\left(\mathrm{\ω}\right)\right|}^2+Q}---\left(1\right)$

Wherein Q is the noise effect factor, and the noise level of reaction signal is taken as | H (ω) | ²1%

H* (ω) is the conjugate complex number of H (ω);

4) selected autoregressive model exponent number K, application Burger method is calculated the autoregressive coefficient under this exponent number;

5) the frequency decay window of selection one fixed width, this window is applied to the E (ω) that step 3 obtains, and the part that intercepting E (ω) is positioned at frequency window is as known frequency spectrum, according to the autoregressive coefficient that calculates in the step 4, the frequency spectrum of the both sides outside application formula 2 and the formula 3 predict frequency windows

${\hat{x}}_q=-\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{a}_i{x}_{q-i},q=n+1,...,N---\left(3\right)$

${\hat{x}}_q=-\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{a}_i{x}_{q-i},q=n+1,...,N---\left(3\right)$

Wherein m is the left margin subscript of frequency window, and n is the right margin subscript of frequency window, and k is the exponent number of autoregressive model, a _iBe autoregressive coefficient,

A _iConjugate complex number,

The frequency spectrum in left side outside the frequency window,

The frequency spectrum on right side outside the frequency window,

Be

I the data point in right side, Be

I the data point in left side, N is sampling number;

6) frequency spectrum that keeps E (ω) to be positioned within the frequency window is constant, respectively with calculating in the step 5

Replacement E (ω) is positioned at left side frequency spectrum outside window, uses

Replacement E (ω) is positioned at right side frequency spectrum outside window, thereby obtains the frequency spectrum X (ω) of reconstruct;

7) X (ω) is carried out inversefouriertransform and can obtain time-domain signal x (t), x (t) is the ultrasound wave time-domain signal of temporal resolution after being improved;

8) according to the Wave crest and wave trough position of the time-domain signal x (t) that obtains, determine the position of liquid level upper and lower surface echo, determine the travel-time of sound wave in liquid level with this, can calculate the thickness of liquid level in conjunction with the velocity of propagation of sound wave in liquid level.

2. a kind of Autoregressive Spectrum Analysis method Autoregressive for improving the ultrasound detection signal temporal resolution as claimed in claim 1 is chosen as 20.

3. the width of a kind of Autoregressive Spectrum Analysis method frequency decay window for improving the ultrasound detection signal temporal resolution as claimed in claim 1 is chosen as-6db.

Images