WO2012121599A1

WO2012121599A1 - Method and system for non-destructive testing

Info

Publication number: WO2012121599A1
Application number: PCT/NL2012/050139
Authority: WO
Inventors: Rudolf SPRIK; Robbert Jan DE BOER
Original assignee: Universiteit Van Amsterdam
Priority date: 2011-03-07
Filing date: 2012-03-07
Publication date: 2012-09-13
Also published as: NL2006345C2

Abstract

A testing system with a control unit (10) connected to a plurality of sending and or receiving transducers (5, 6). In combination, the control unit (10) and transducers (5, 6) are arranged to receive a current signal originating from the sample, to calculate a fidelity index η by correlating a tail part of the current signal with a reference signal, and to compare the fidelity index η with a threshold value η_thr

Description

Method and system for non-destructive testing

Field of the invention

The present invention relates to a non-destructive testing method for a sample, comprising receiving a current signal originating from the sample.

Prior art

American patent publication US-A-6,234,035 describes a method for non- destructive testing using ultrasonic inspection using a focused wave device. A focus lens is coupled to a transducer to focus an ultrasonic signal on an area to be inspected, and a stop is placed in the focus lens to block selected ultrasonic waves. Echo responses of waves not blocked are processed. The detection system works only in a very local area, in case of testing of large surfaces the detector needs to be repositioned and measurements need to be repeated.

Summary of the invention

The present invention seeks to provide a non-destructive testing method and system allowing to accurately and reliably test samples. The samples can have any form, having large surface structures and/or or large bulk material dimensions.

According to the present invention, a method according to the preamble defined above is provided, the method further comprising calculating a fidelity index η by- correlating a tail part of the current with a reference signal, and comparing the fidelity index η with a threshold value. By using only the tail part of the current signal and reference signal, the signal components comprising the information necessary to detect possible faults in the material of the sample are effectively used, allowing a reliable and robust non-destructive testing of samples in any form, e.g. having large area surfaces or bulk material parts. The present invention embodiments allow testing at discrete time instances while no permanent monitoring is required, by using a comparison between the reference signal and the current signal detected at the time of measurement.

Short description of drawings The present invention will be discussed in more detail below, using a number of exemplary embodiments, with reference to the attached drawings, in which

Fig, 1 shows a schematic diagram of a non-destructive testing system according to an embodiment of the present invention;

Fig. 2 shows an example of an current signal obtained using the present invention embodiments; and

Fig. 3 shows a top view of a test arrangement with one sending transducer and two receiving transducers according to a further embodiment of the present inv ention. Detailed description of exemplary embodiments

The present invention embodiments provide a manner to identify changes in complex materials, such as composite materials, using (ultrasonic) sound, based on signal analysis of only a part of a complex current (echo) signal detected in a sample. Using the present embodiments, samples having a larger surface can be tested as compared to present day non-destructive testing techniques.

In many complex materials used in large surfaces, such as composite parts in aircraft and the like, but also in large building structures such as bridges, damage is sometimes difficult to discover. The damage may be present in inner layers of the composite material, not visible from the outside, while still the structural properties of the part are influenced. Composite materials used in large surface elements of a construction include, but are not limited to carbon fiber material, glass fiber material, aluminum cored composite plates, carbon fiber reinforced materials, etc.

Tn general, the present invention embodiments relate to a non-destructive testing method for a sample (e.g. a plate type sample), comprising receiving a current signal (e.g. using an ultrasonic transducer), calculating a fidelity index by correlating a tail pail of the current signal with a reference signal, and comparing the fidelity index with a threshold value. The current signal is a actual signal, obtained at the time of

measurement.

The method embodiments may be implemented using a system as shown schematically in the block diagram of Fig. 1 . A sample 1 to be tested, in general a plate type sample having a major surface, is provided with a sending transducer 5 and three receiving transducers 6 in the embodiment shown. In the material of the sample 1, a fault 2 is present, e.g. a crack in one of the lay ers of carbon fiber material. The sending and receiving transducers 5, 6 are e.g. piezoelectric transducers. The transducers 5, 6 may also be other types of suitable transducers, such as optic vibrometers (operating on the principle of laser reflections), or fiber optic transducers.

The sending transducer 5 is connected to a control unit 10 via a digital-to-analog converter (DAC) 14 and an amplifier 3. Each of the receiving transducers 6 is connected to the control unit 10 via associated amplifiers 11 and a bank of analog-to- digital converters (ADC) 12. The control unit 10 is furthermore connected to an input/output unit 15.

The sending transducer 5 is present in embodiments using an active testing method, and is used to inject an elastic wave into the sample (e.g. an ultrasonic wave or an acoustic wave).

In embodiments using a passive testing method, the sending transducer 5 is not present, and environmental noise or other source of ultrasonic waves is used to obtain a signal from the receiving transducers 6. In various embodiments, the minimum number of receiving transducers 6 is one (active embodiments only) or more than one (active and passive embodiments).

The sending transducer 5 or noise from the environment of the sample 1 to be tested, causes an acoustic wave to travel through the sample 1. The form and consistency of the sample 1 causes multiple reflections at the interfaces in the sample 1 (internal and external interfaces), but also other effects are playing a role, such as multiple scattering

(at imperfections in the material of the sample 1, or at boundaries of the sample 1, and results thereof including reverberation), transient effects, absorption, dynamic changes, non-linear responses, etc. Also, the presence (or absence) of a fault 2 will have a specific influence on the acoustic waves travelling through the sample 1. The receiving transducers 6 are used to pick up the (current) echo signal(s), after which the current signal is preprocessed in amplifiers 11 and ADC's 12, and further processed in control unit 10.

It has been found that it is advantageous not to use the entire received current signal, but only a part of it. An example of a typical detected signal originating from one of the receiving transducers, in response to a Gaussian

input signal injected into the sample 1 using the sending transducer 5 is shown in Fig. 2. A Gaussian pulse is understood herein as a pulse with a carrier frequency and a Gaussian envelope, see also further below). The first part (1 ) of the signal only represents noise, which can be used to identify the noise level. The actual start of the signal is the first arrival of an echo (2) which part is not useful and not used in the present invention embodiments. At the tail part (3) or rear part of the current (echo) signal, valuable information can be determined using the present invention

embodiments. The echo of the Gauss input signal will be more dispersed, and comprise information over the pathway travelled through the sample 1. The echo signal is a complex response of the sample 1, is reproducible and comprises information about the sample J . Thus, this third part (3) of the current signal can be expl oited for testing and characterization purposes.

In general wordings, the present invention embodiments comprise receiving a current signal from the plate type sample 1 (using an ultrasonic transducer or receiving transducer 6), calculating a fidelity index η by correlating a tail part of the current signal with a reference signal, and comparing the fidelity index η with a threshold value. The fidelity index η will be 1 if the conditions in the sample 1 between the recording of the reference signal and the recording of the current signal have not changed. A value of the fidelity index η substantially lower than 1, e.g. lower than 0.9 may be an indication of a fault in the material of the sample 1. It is noted that due to signal noise, signal processing and limitations of equipment used, the fidelity index η will in practice always be somewhat less than I ,

In a specific embodiment, the fidelity index η is calculated according to

m x M( )^■ C(t + τ))

η≡ .

^max._T{R(f)■ R t + τ))■ max^C t ■ C(t + x)} wherein R(t) is the reference signal and C(t) is the current signal, and τ is a time shift for calculating the cross correlation of the reference sign al R and current si gnal C and the auto correlations of the reference signal R and current signal C. The operator max_T is the maximum value and is used to compensate for drift and jitter in the time axis.

The denominator in this formula is a normalization factor comprising the two auto correlations of the reference signal R(t) and current signal C(t), as a result of which the ideal fidelity index η will range in value from 0 to 1.

The signal processing is implemented using the control unit 10. In an active mode, the control unit 10 is arranged to provide a suitable signal to the sending transducer 5. Furthermore, the control unit 10 is arranged to record and process the current signal(s) obtained from the receiving transducer(s) 6, A previous recorded echo signal from the same samplel (or an equivalent sample 1) is stored as the reference signal R(t) in an embodiment. The earlier actual measurement on the same sample I may involve a time lapse which is large (e.g. from manufactuiing date of sample) or relatively short (e.g. at set intervals, e.g. for health monitoring). The earlier actual measurement may also involve a reference sample, used for quality control by comparing products manufactured to the reference sample.

Finally, the control unit 10 is arranged to provide the calculations necessary for the determination of the fidelity index, using cross correlation and auto correlation techniques. Finally, the control unit 10 is arranged to compare the fidelity index to a predetermined threshold value, and to provide comparison results, e.g. using the input/output unit 15.

The current signal C is in a further embodiment a weighted average of multiple received echo signals from the same receiving transducer 6, which allows to obtain a more robust and reliable measurement.

If multiple receiving transducers 6 are positioned on the sample 1 to be tested, the individual fidelity index may be calculated in the control unit 10 for each receiving transducer 6, and the overall fidelity index η is e.g. a weighted average of all individual fidelity indices.

In more general terms, the following embodiments can be described when using multiple sending transducers 5 and multiple receiving transducers 6.

Tn active mode M sending transducers 5 transmit an ultrasonic signal and N receiving transducers 6 receive the associated (ultra)sonic responses. The

transmitters 5 are sending a pulse P^m(t) with m = l... in sequence and each response is recorded on the N available receivers 6. Hence, at most M pulses need to be transmitted and a total of N signal traces i?™¹ (m = I ...M; n = 1...N ) are available for further analysis.

A Gaussian pulse P^m(t) = P₀ exp(-(( - to)/ ²) cos(2%f(t- to)) with amplitude PQ centred around t with a width Δ and carrier frequency /is used as transmitted pulse shape from transmitter position m. The Gaussian pulse has a limited width in time (~ A) and in frequency (-1/Δ) and is represented using a sampling frequency ₃ > (i.e. a sampling time dt = \/F_s). Alternative pulse shapes are wavelets, chirped signals, or programmed

(Gaussian) noise. Optimized use of different choices for each P^m(t) and/or concurrently transmitting pulses can enhance performance even further.

The signal response on transducer n due to pulse m is recorded with a sampling rate F_s and is represented here by R" with i the point index for the discrete recording time axis h = i.dt running from / = 1 to the total of recorded points per trace i = i_m-

The signai-to-noise performance can be enhanced by averaging repeated recordings of iT^u\ over a collection time where no changes in the response due to defect formation etc. occurs. Thus, the current signal may be an averaged signal using multiple (but synchronized) sample periods.

To test for system integrity a comparison of recorded signals is made using a normalized correlation method. Required is a reference signal ?"¹" obtained under conditions where the system is functioning correctly. The reference signal can e.g. be obtained during periodic maintenance or recorded during operation.

Whenever a test of system integrity is needed the current signal C^ms is recorded and compared with R™¹; to obtain a fidelity parameter η™¹ for each transmitter-receiver pair mn. From the total recorded signals only the interval between i\ and h. k + δ/^' is used in the calculation. The fidelity parameter η™¹ is defined by correlating and normalizing the reference and the current signal as:

The normalized correlation expressed in this equation compensates for timing jitter between the two recorded signals by searching for the maximum value as function of correlation time delay τ. Further improvement is possible by e.g. interpolation of the time axis to obtain higher temporal resolution. The calculations necessary to evaluate η"¹¹¹ are well suited for parallel processing on e.g. DSP based computing boards by fast Fourier transform techniques (a specific embodiment of the control unit 10).

Correlation techniques are not sensitive to the conversion bit depth of the DAC and ADC components 12, 14 used to generate and record digitized signals. However, the signals should cover the available dynamic range.

The overall fidelity parameter η is e.g. calculated as the (weighted) sum of η¹™ over all transmitter-receiver pairs mn. The fidelity parameter η is by definition 1 for identical reference and comparison signals. Changes in the signals due to e.g. defects, surface contamination etc. make η < 1.

From tests during the actual implementation and deployment the appropriate threshold value associated with a failure condition can be determined. Experiments under laboratory conditions on small composite panels indicate when η < with = 0.9 considerable damage has occurred. In further embodiment, the threshold value is defined as = 0.95 to obtain a higher reliability of detecting ail possible faults or damages, even very small damages. In an even further embodiment, the threshold value is defined as rjthr = 0.8 to obtain a higher allowed degree of damage or pollution in the sample I tested before providing an alarm.

The optimum choice for i_h δ/^', dt and the pulse parameters as given for P^m(t) above depends on the particular implementation in tested systems. For example larger z^'i selects a part starting i .dt later in the current signal and includes signals that propagated over longer distances in the tested structure of sample 1. Larger δ/^' enliances the relative changes i η and extends the low frequency resolution to Sampling time 8t sets the maximum frequency that can be detected in the signal. During implementation the number of detectors 6, relative distances and frequency parameters can be optimized to obtain maximum reliability of the test.

A further group of embodiments are related to passive test methods, wherein external noise sources are used, or even thermal noise. In passive mode only N signals are recorded on the receivers 6 and no active signals are used. The noise from the environment is the basis of the detection procedure. From the literature is known that by correlating the N recorded signals an estimate of the response function similar to the function between transmitter-detector pars can be obtained.

In the passive mode signals R^nai (reference signal) and "ⁿ (current signal) used in the calculation of η are replaced by

and

respectively Here Rⁿ; is the reference signal and " the comparison signal recorded over the same time interval. The spectral properties of the noise sources determines the usable spectral range of the detected signals. Filtering the recorded signals helps to concentrate on the most prominent features in the spectra.

In a further embodiment, also cross correlation functions are used of signals betwe n using the formulas:

and

The passively received reference and current signals result from the noise

(which is considered similar at the moment when measuring the reference signal and when measuring the current signal).

In a further embodiment of the present invention, the tail part of the current signal has a starting moment of more than double the transition time of an acoustic (e.g. ultrasonic) wave in the plate type sample. The transition time is dependent on the material, the frequency range of the acoustic wave, and further parameters such as temperature, load, etc. By applying this precondition, the direct signal (first arrival) and a first reflection will not be included in the correlation calculations, and only the relevant part of the current signal for this invention embodiments will be included. Advantages are that the method is (more or less) independent from the sensor location. In the control unit 10, the starting time i_\ can be fine-tuned for specific types of sampl es 1. As an example, the tail part of the current signal has a starting moment of more than 1.4ms (after the direct signal), and in experiments good results have been achieved using a start time of between 4-5 ms. It is noted that for other types of structures, the settings used for the starting moment may differ.

The signal content of the current signal and reference signal has a bandwidth limited to e.g. between 50 kHz and 1.5 MHz in a further embodiment. This range is mostly determined by the types of transducers used. When using certain types of transducers even frequencies lower than 50 kHz may be used. In the control unit 10, the frequency content of the current signal is also related to the sampling frequency 6t. In ihe active mode of operation, a Gaussian pulse is used as described above, of which the center frequency is e.g. 225kHz, 747kHz, or 210kHz),

In a further embodiment, a correlation window size is larger than HKlus, and is e.g. about 200, 400, 800 or 1600μ8. In the multiple sending and receiving transducer embodiment as described above, this correspond to a choice of 61, which can be effected in the control unit 10. A very good result in experiments was obtained for a correlation window size of more than 800μ8.

Experiments have been conducted to provide a proof of concept and insight on parameters affecting the performance of the present invention embodiments. In Fig. 3, the set-up of the test is shown schematically. The test sample 1 is a carbon fiber composite plate (355mm by 157 mm and a thickness of 2.93mm). The test sample was manually damaged twice with a mechanical mill during the experiment, resulting in faults 2 in the form of carves on the plate I as indicated in Fig. 3. One sending transducer 5 is positioned near a first fixture 3, and receiving transducers 6 are positioned on the other positions indicated. A force may be applied to the plate 1 using a second fixture 4 to simulate load conditions to the plate 1.

In the experimental set-up, an a digital oscilloscope was used to average collected current signals, wherein the averaging amount was determined by collection time and burst rate of the input signal. The digital oscilloscope takes the function of the ADC 12 in the block diagram of Fig. 1, and an arbitrary wave generator (AWG) is used to take the function of DAC 14. A personal computer controlling the oscilloscope and AWG is used to take the function of processor unit 10 in Fig.1. A burst rate of 7 Hz was used, and for the frequencies of 210 kHz, 747 kHz and 1225 kHz collection times of 3s, 5s and 6s were used, resulting in an averaging of 21, 35 and 42 times. An output sample of 6 ms was captured with 30,000 sample points.

The experiments demonstrated that indeed the fidelity index as calculated dropped after the first damage, and dropped even further after applying the second damage. It was also demonstrated that the result was independent from the location of the receiving transducer 6.

For quality control, measurements were conducted using various stresses on the plate 1, applied to the second fixture 4. in this type of testing, a reference signal is not available (or even needed), and an external factor (load) is used as changing parameter, as opposed to fault detection where a reference signal is used. The fidelity index was calculated as function of load, with and without damage. The results showed that indeed a damaged plate 1 has a higher fidelity index drop rate as compared to the undamaged plate 1.

As discussed above, an input signal is injected in the plate type sample 1 to obtain the current signal, e.g. using M piezoelectric transducers 5, wherein M>1. This relates to the active modes of operation, where one or more input signals can be used (of the same or different shape, size and frequency). In one specific embodiment, he input signal is a Gaussian pulse, as discussed above, however, as an alternative a wavelet, a chirped signal, or a programmed (Gaussian) noise signal may be used.

In both the active and passive modes of operation, the current signal is obtained using N piezoelectric transducers 6 in a further embodiment, wherein N>1.

The present invention method embodiments can be applied advantageously for e.g. health monitoring or quality control of composite parts. For the health monitoring application, the reference signal is obtained earlier in time by recording the current signal in a new, approved part. During its lifetime control signals can be obtained as current signal to check for any mechanical changes.

The quality control application makes use of the fact that the output signal is affected by local mechanical stresses. Under the same amount of external loading, parts with imperfections will show different, higher, mechanical stresses. Therefore, the fidelity index is expected to decrease faster under a fixed external mechanical load in the presence of imperfections.

To implement this aspect of the present invention, the reference signal and current signal each comprise a series of signals recorded under varying load conditions of the plate type sample. Furthermore, the control unit 10 may implement further calculations for determining a fidelity index drop rate.

Possible application areas for the embodiments described include, but are not limited to:

- condition monitoring of composite material parts during use and/or under (high) load (e.g. landing gear parts during flight); Possible, life time can be enlarged, design can be adapted to minimize over-dimensioning the part, or the inspection interval can be enlarged or even deleted.

- quality control in fabrication: production units can be compared to a 'standard' unit, allowing to detect possible inclusions and fractures in composite materials. - determining impact or damage in a thermoplast material tail section of an aeroplane.

The present invention embodiments have been described above with reference to a number of exemplary embodiments as shown in the drawings. Modifications and alternative implementations of some parts or elements are possible, and are included in the scope of protection as defined in the appended claims.

Claims

1. A non-destructive testing method for a sample, comprising

receiving a current signal originating from the sample,

calculating a fidelity index η by correlating a tail part of the current signal with a reference signal, and comparing the fidelity index η with a threshold value η^.

2. The non-destructive testing method according to claim 1, wherein the fidelity index η is cal culated according to

_ max^{R(t) · C(t + τ))

wherein R(t) is the reference signal and C(t) is the current signal, τ is a time shift for calculating the cross correlation of the reference signal R(t) and current signal C(t) and the auto correlations of the reference signal R(t) and current signal C(t), and the operator max denotes the maximum value,

3. The non-destructive testing method according to claim 1 , wherein a plurality of m sending transducers and a plurality of n receiving transducers is attached to the sample, the method further comprising calculating an individual fidelity parameter rj^mn associated with a pair of a sending transducer and a receiving transducer according to

wherein

max_x is the maximum value as function of correlation time delay τ.

4. The non-destructive testing method according to any one of claims 1-3, wherein the reference signal Is a stored signal.

5. The non-destructive testing method according to any one of claims 1-4, wherein the tail part of the current signal has a starting moment of more than double the transition time of an elastic wave in the sample.

6. The non-destructive testing method according to claim 5, wherein the tail part of the current signal has a starting moment of more than 1.4ms,

7. The non-destructive testing method according to any one of claims 1-6, wherein an input signal is injected in the sample to obtain the current signal.

8. The non-destructive testing method according to claim 7, wherein the input signal is a Gaussian pulse.

9. The non-destructive testing method according to any one of claims 1-8, wherein the current signal is obtained using N receiving transducers.

10. The non-destructive testing method according to any one of claims 1-9, wherein the current signal is an averaged signal using multiple sample periods.

11. The non-destructive testing method according to any one of claims 1-10, wherein the reference signal and current signal each comprise a series of signals recorded under varying load conditions of the sample.

12. The non-destructive testing method according to claim 11 , further comprising determining a fidelity index drop rate.

13. The non-destructive testing method according to any one of claims 1-12, wherein the sample is a plate type sample comprises a layered material.

14. A testing system comprising a control unit connected to a plurality of sending and or receiving transducers, wherein the control unit is arranged to implement the method according to any one of claims 1-13.