CN1865980A - Ultrasonic detection method for near surface defect - Google Patents

Abstract

The related ultrasonic detection method for near-surface defect comprises: building mathematic model, using the time-domain sequence of near-surface defect echo to obtain its energy sequence; then, extracting the slow-variable part as the signal low-frequency part to compare with echo without defect. This invention uses clear principle, has high efficiency, reduces detection leakage rate, and fit to wide application.

Description

The supersonic detection method of near surface flaw

Technical field:

The present invention relates to a kind of supersonic detection method of near surface flaw, is a kind of pulse ultrasonic wave that is specially adapted to when detecting test specimen, the method for discrimination of near surface flaw.This method has provided the theoretical foundation of its enforcement from the angle of digital signal processing, the process of enforcement and to the implementation result of an exemplary special case.

Background technology:

The detection of near surface flaw is a tradition and typical research topic in Non-Destructive Testing.The detection method of near surface flaw is a lot, such as, pulse ultrasonic wave reflectometry, magnaflux, eddy current detection method, magnetic are remembered detection method, Magnetic Flux Leakage Inspecting method, magnetic flaw detection ink detection method, are climbed ripple detection method, surface wave detection method and thermography method etc.These methods generally all have tested object and test environment requirement separately, do not have a kind of universal method that can be used for any test occasion.This also is the several different methods and the reason of depositing.

In impulse ultrasound reflection detection method, the defective of close medium interface is submerged in the echoed signal, is difficult to effectively separate the existence that causes measuring the blind area.Considering from the angle of signal time domain, is exactly that signal is more approaching at the due in of time domain, and a signal does not also finish, and another signal arrives.In the Ultrasonic Detection of defective, this phenomenon occurs and mainly contain following two kinds of situations.First kind of situation be, the pulse ultrasonic wave of sensor emission is coupled to the signal that receiving circuit produces also not to be finished, and the ultrasonic echo of near surface flaw just arrives.At this moment, amplifying circuit is operate as normal not as yet, flaw echoes is diminished, and two signal aliasings together, causes near surface flaw to detect.Second kind of situation is, because the ultrasound wave of coupling when measuring distance has guaranteed not launch in the echo, but, interface echo and near surface flaw echo are superimposed, less or defective when nearer, also is difficult to find the existence of defective from the surface from detected time domain echo when the relative interface echo of flaw echo.In actual detected, this phenomenon may occur in the upper surface of test specimen to be checked, also may occur in lower surface.No matter which kind of situation all can cause the recall rate of defective is descended.Generally speaking, first kind of situation avoided taking place as far as possible.The method that solves is to increase the pulse ultrasonic wave of sensor emission and the effective travel on surface.Such as, the transmitting terminal at wafer when Design of Sensor increases medium, or adds medium between sensor and test specimen, also can change incident angle etc.

For second kind of situation, i.e. the situation of flaw echo and dielectric surface echo stack is difficult to by the change of hardware or means of testing two waveforms be separated.At present, ways of addressing this issue mainly is a software approach, promptly realizes the separation of signal by certain digital signal processing algorithm, thereby realizes the detection near surface flaw.According to data (C.Fritsch, A.Veca, Detecting small flaws near the interface inpulse-echo, Ultrasonics, Vol.42,2004, pp.797-801), mainly contain convolution method, Xi Er baud conversion method, cepstrum spectrometry, impulse compression method etc. at present.The several method of introducing in the data all has conforming separately, and result of use is unsatisfactory.

Summary of the invention:

The objective of the invention is at the deficiencies in the prior art, a kind of near surface flaw Ultrasonic Detection new method based on digital signal processing is provided.Promptly, from the interface echo of test specimen, effectively isolate flaw indication from the angle of signal Processing.To reduce the omission of this class defective.

For realizing such purpose, the present invention has set up the ultrasonic echo mathematical model of near surface flaw theoretically, and derived have, echoed signal exists during no near surface flaw difference.Simultaneously, the concrete steps of implementing this method have also been set forth in this invention, and demonstrate with special case.

The present invention includes following concrete steps:

Set up mathematical model, mainly comprise:

1) having set up the mathematical model y (t) of near surface flaw impulse ultrasound echoed signal, explained with the form of signal envelope, convolution, phase shift and shock response, is the theoretical foundation of employed method in the invention;

2) ultrasound echo signal is modulated, demodulation obtains the envelope y of ultrasound echo signal _f(t), it equal the envelope of reference signal with by the envelope sum of defective modulation signal;

Reference signal is obtained, and mainly comprises:

3) use the test operating mode identical, the defect-free surface echo is detected, obtain the ultrasonic time domain echoed signal of defect-free surface, be called reference signal with actual detected;

4) by actual test macro and test specimen to be checked, obtain the ultrasound echo signal time series, with time series pointwise square, be mapped to energy space again, obtain its corresponding energy sequence;

5) adopt wave digital lowpass filter, energy signal is carried out demodulation, filter the high frequency composition, obtain the envelope of signal, be called the reference signal envelope;

Near surface flaw detects, and mainly comprises:

6) under identical test operating mode, record actual ultrasound echo signal, by 4)-5) step obtain the measured signal envelope;

7) with measured signal envelope and reference signal envelope relatively, promptly take pointwise to subtract each other after, take absolute value, and amplitude normalization, obtain residual signal; During zero defect, residual signal is zero, and during defectiveness, residual signal is not equal to zero, thereby makes the judgement that has or not near surface flaw.

Described wave digital lowpass filter is ripple FIR low-pass filters such as 22 rank.

Description of drawings:

Fig. 1 is a FB(flow block) of the present invention

The convolution principle synoptic diagram that Fig. 2 derives for theoretical model of the present invention

Fig. 3 is the zero defect reference signal; (a) time-domain signal; (b) energy signal; (c) signal envelope

Fig. 4 is near surface flaw echoed signal and result (); (a) time-domain signal; (b) energy signal; (c) signal envelope

Fig. 5 is near surface flaw echoed signal and result (two); (a) energy signal; (b) signal envelope; (c) envelope relatively; (d) residual signal

Embodiment:

Below in conjunction with drawings and Examples technical scheme of the present invention is further described.

1) the ultrasonic echo model of near surface flaw

In the impulse ultrasound of near surface flaw detects, mainly contain the stack that two kinds of situations can cause echoed signal.The one, the little defective of nearly inside surface, the 2nd, the little defective of nearly outside surface.The present invention with the little defective of inside surface as analytic target.If during zero defect, have only inside surface echo and outside surface echo (disregarding repeatedly reflection) in the reflection echo in the test specimen of measuring point correspondence.If there is defective in the test specimen, then as long as defective can form effective reflecting body, reflection echo signal (not considering that here noise is to the problem of flooding of weak echo signal and the influence of clutter) can appear between inside surface echoed signal and outside surface echoed signal then.Method about the processing of this flaw echoes is a lot.But, as defective very near the surface, cause flaw echo and surface echo to be superimposed, from the time-domain signal that collects, have only the surfaces externally and internally echo, between two surface echos, do not have the appearance of tangible echo, so be easy to defective is failed to judge.And from the angle of frequency domain, because two place's echoes have the identical frequency domain distribution of being close to, so also be difficult to do to distinguish effectively.

In order to derive the separation algorithm of signal, set up the mathematical model of ultrasonic reflection echoed signal under the signal stack situation.If the ultrasound echo signal y (t) that records is:

y(t)＝y ₁(t)+y ₂(t)+n(t) (1)

Wherein, y ₁(t) be surface echo; y ₂(t) be flaw echo; N (t) is a noise.

For the narrow-band impulse Ultrasonic Detection, the signal x (t) that establishes sensor emission is:

$x\left(t\right)=e^{i{w}_{0}t}g\left(t\right)---\left(2\right)$

Wherein, e ^Iw0tBe harmonic signal; G (t) is a pulse envelope.

Then, surface echo and flaw echo can be expressed as (wouldn't count noise) respectively:

$y_1\left(t\right)=x\left(t\right)\⊗h_1\left(t\right)=e^{{\mathrm{iw}}_{0}t}g\left(t\right)\⊗h_1\left(t\right)---\left(3\right)$

$y_2\left(t\right)=x\left(t\right)\⊗h_2\left(t\right)=e^{{\mathrm{iw}}_{0}t}g\left(t\right)\⊗h_2\left(t\right)---\left(4\right)$

In the formula,  represents convolution algorithm; h ₁(t) be that the surface is to the hyperacoustic impulse response of incident; h ₂(t) be that defective is to the hyperacoustic impulse response of incident.

In actual detected because that the emission ultrasound wave arrives the surface is different with the stroke of defective, the phase place that transmits that arrives two targets so scarcely with, suppose that the time difference of correspondence is t ₀Simultaneously, because the part that transmits is by surface reflection, the stroke of adding the two is also different, and the amplitude of signal that then arrives two targets is also different, if the amplitude of the ultrasonic signal on surface is incided in normalization, the amplitude that can be set to signal when reaching defective is a.Then, detected echoed signal can be expressed as:

$y\left(t\right)=e^{{\mathrm{iw}}_{0}t}\×[g\left(t\right)\⊗h_1\left(t\right)+\mathrm{ag}(t-t_0)\⊗h_2\left(t\right)]+n\left(t\right)---\left(5\right)$

In pulse ultrasonic wave detected, target echo signal can be regarded as the convolution of test macro, travel path and objective body.As shown in Figure 2.

If there is not defective, then the surface reflection echo can be expressed as:

y ₁(t)＝x(t)h _pfh _sfδ(t)h _sbh _tb (6)

If defectiveness, then the reflection echo of defective can be expressed as:

y ₂(t)＝x(t)h _pfh _sfh ₂h _sbh _tb (7)

In the formula, x (t) is an input signal; h _PfIt is the shock response of sensor testing system forward direction; h _SfIt is the forward direction shock response of travel path; h ₁Be flawless shock response, can be taken as δ (t); h ₂It is the shock response of defective; h _SbBe the back of test macro to shock response; h _TbBe the back of travel path to shock response.

If the test operating mode is constant, then can be with formula (6) substitution formula (7), and consider time delay and amplitude,

y ₂(t)＝ay ₁(t)h ₂δ(t-t ₀) (8)

Formula (8) shows that a flaw indication can be expressed as the convolution of defect-free surface echo and defective shock response.At this moment, y (t) is expressed as:

$y\left(t\right)=y_1\left(t\right)+y_2\left(t\right)=y_1\left(t\right)\⊗[\mathrm{\δ}\left(t\right)+{\mathrm{ah}}_2\⊗\mathrm{\δ}(t-t_0)]+n\left(t\right)$

$=e^{{\mathrm{iw}}_{0}t}\×[g\left(t\right)\⊗\mathrm{\δ}\left(t\right)+\mathrm{ag}\left(t\right)\⊗h_2\⊗\mathrm{\δ}(t-t_0)]+n\left(t\right)---\left(9\right)$

$=e^{{\mathrm{iw}}_{0}t}\×[g\left(t\right)+\mathrm{ag}(t-t_0)\⊗h_2]+n\left(t\right)$

Because hyperacoustic frequency of operation was generally several MHz during impulse ultrasound detected, frequency is higher, and the pairing frequency of its pulse envelope is generally lower.By frequency-domain analysis, can know and contain radio-frequency component (ultrasound wave frequency of operation) and low frequency part (frequency of pulse envelope) in the signal echoed signal.In the echoed signal, the abundant information that contains reflecting body in the envelope of signal, such as, contain reflected signal in the differential of envelope in the size of ascent stage climbing speed, the fall off rate size of decline stage signal etc., it also is a tolerance of signal oscillating amplitude size; The duration of envelope has been reflected the length of signal oscillating time etc.In actual detected, the energy of noise generally concentrates on high frequency region, and the radio-frequency component in the signal is vulnerable to interference of noise.Under the situation of the distortion of disregarding the echoed signal medium frequency, the signal in the formula (9) is carried out filtering, filter the radio-frequency component in the echoed signal, the echoed signal after then handling can be expressed as:

y _f(t)＝g(t)+bg(t-t ₀)h ₂(t) (10)

In the formula, b is a range coefficient.In the formula (10), filter delay is summed up in the point that among the variable t.

From formula (10) as can be seen, the envelope of near surface flaw ultrasound echo signal equals the envelope of original signal and is modulated later envelope sum by defective.If there is not the existence of defect waves in the echo, then second on following formula the right is zero, at this moment, has only the envelope of surface echo in the echo envelope.In the superincumbent derivation of equation, the coefficient of original signal envelope front and phase shift have been carried out normalized.

2) the detection implementation algorithm of near surface flaw

If the signals sampling frequency is f _s, sampling interval is T _s, f _s=1/T _sSignal y (t) obtains discrete series { y (n) } after sampling, n=1, and 2 ..., N.N is a number of data points.Because ultrasound echo signal y (t) ∈ is L ²(R), obtain the conventional burst:

e(n)＝y ²(n)，n＝1，2，…，N (11)

In order accurately to obtain the envelope of signal, adopt wave digital lowpass filter, sequence e (n) is carried out low-pass filtering.Filtered signal is exactly the envelope of original signal, i.e. low frequency part.It differs a constant multiple with original signal on amplitude, its phase place also has certain delay, will pay attention in Filter Design.Through filtered signal is exactly signal y _f(t) discrete series y _f(n).

In order to detect sequences y _f(n) whether contain flaw indication in, by test specimen flawless ultrasound echo signal under identical test operating mode is utilized and processing signals y as the reference signal _fThe envelope of signal when (t) identical algorithm obtains zero defect is equivalent to the corresponding discrete series of g (t).Then, maximum amplitude to discrete signal carries out normalized, the phase differential that the subtraction signal processing links causes, just can obtain a residual signal, theoretically, when zero defect existed in the echo, the envelope of echo signal was identical with the envelope of reference signal, at this moment, remaining burst should be zero; When defectiveness exists in the signal, the envelope of echo signal and the envelope of reference signal have bigger difference, remaining signal is non-vanishing, and from formula (10) as can be known, residual signal is the impulse response and the convolution that has changed the reference signal envelope of amplitude and phase place of defective.

Lowpass digital filter adopts ripple FIR wave filters such as 22 rank, and design process is finished by Matlab is auxiliary.

3) special case is implemented and checking

For said method is verified, provide one in the invention and implemented special case, utilize the normal probe sonac that sample is tested and analyzed.

At first obtain reference signal, it is the echoed signal of inside surface during zero defect in the test specimen, as shown in Figure 3.Wherein, (a) be the time domain waveform of surface echo; (b) be its energy signal; (c) be that the signal envelope that obtains was equivalent to the g (t) in the formula (10) after it passed through low-pass filter.

Then, two groups of near surface flaws are detected.Wherein, defect reflection face and surface among Fig. 4 differ far away, can tell the existence of defective from its time-domain signal (Fig. 4 (a)) and energy signal (Fig. 4 (b)).Certainly, the signal after the processing (Fig. 4 (c)) also is easy to find out the existence of defective.Single advantage of from the displaying of Fig. 4, also not seeing new method, but can illustrate that at least new method also is suitable for the separable signal of time domain, can not produce erroneous judgement to it.But, if the reinforcement of signal high-frequency noises can make the separatrix of two target echo signals in the time domain approach fuzzy, bringing certain difficulty to detection, signal after treatment then can be not influenced.

In order to represent the advantage of this method, reduced the distance of defective in the research with the surface, at this moment, be difficult on from the echoed signal that measures tell and whether had defective, from its energy signal (Fig. 5 (a)), also be difficult to make differentiation, Fig. 5 (b) is its envelope, and the envelope of it and reference signal has certain difference as can be seen from the envelope, but not obvious (Fig. 5 (c)).For this reason, at the normalization amplitude information and deduct (signal among Fig. 5 (c) is through the signal after handling like this) after the phase shift that each processing links causes, from echo signal, deducted reference signal, the result who obtains gets its absolute value, its result is shown in Fig. 5 (d), therefrom can obviously find out the difference of it and flawless reference signal, description defect exists, and overlapped signal has been carried out effective separation.

Claims (2)

1, a kind of near surface flaw supersonic detection method comprises following concrete steps:

Set up mathematical model, mainly comprise:

1) having set up the mathematical model y (t) of near surface flaw impulse ultrasound echoed signal, explained with the form of signal envelope, convolution, phase shift and shock response, is the theoretical foundation of employed method in the invention;

2) ultrasound echo signal is modulated, demodulation obtains the envelope y of ultrasound echo signal _f(t), it equal the envelope of reference signal with by the envelope sum of defective modulation signal;

Reference signal is obtained, and mainly comprises:

3) use the test operating mode identical, the defect-free surface echo is detected, obtain the ultrasonic time domain echoed signal of defect-free surface, be called reference signal with actual detected;

4) by actual test macro and test specimen to be checked, obtain the ultrasound echo signal time series, with time series pointwise square, be mapped to energy space again, obtain its corresponding energy sequence;

5) adopt wave digital lowpass filter, energy signal is carried out demodulation, filter the high frequency composition, obtain the envelope of signal, be called the reference signal envelope;

Near surface flaw detects, and mainly comprises:

6) under identical test operating mode, record actual ultrasound echo signal, by 4)-5) step obtain the measured signal envelope;

7) with measured signal envelope and reference signal envelope relatively, promptly take pointwise to subtract each other after, take absolute value, and amplitude normalization, obtain residual signal; During zero defect, residual signal is zero, and during defectiveness, residual signal is not equal to zero, thereby makes the judgement that has or not near surface flaw.

2, described wave digital lowpass filter is ripple FIR low-pass filters such as 22 rank.

Images