International Symposium (NDT-CE 2003)Non-Destructive Testing in Civil Engineering 2003
|Start > Contributions >Lectures > Ultrasonic 2:|
Recent Progress in Scanning Laser Detection of Pulse-Echo Ultrasound in ConcreteBernd Koehler, Fraunhofer-IZFP/EADQ Dresden, Germany;
Scanning laser detection of ultrasound has proven to be a very powerful tool to detect echoes from in-homogeneities in concrete structures (e.g. back-wall, gravel pockets, voids, tendon ducts, etc). This results from the use of a large synthetic aperture as well as from the possibility to apply synthetic aperture focusing technique (SAFT). The most important disadvantage of the method so far is the very long measurement time according to the relatively bad signal-to-noise ratio (SNR) of laser detection. In this paper, the improvement of the underlying measurement strategy is discussed and the resulting reduction of acquisition time is estimated. Measurement of ultrasonic waves traveling along a cross section of a sample allows to visualize the scattering process explicitly. This reveals the influence of interfaces between aggregates and cement matrix, which are not perfect in each case. Implications for detection of injection faults by impulse-echo and impact-echo techniques are discussed shortly.
Compared to the application of elastic waves for non-destructive testing (NDT) of metals, NDT in civil engineering (CE) is a very challenging task. This is due to the strong heterogeneity of typical materials like concrete, leading to multiple scattering and strong attenuation. Even in 1991, the first attempts to use ultrasonic pulse echo methods for NDT of concrete structures were still considered unpromising . Testing of strongly heterogeneous materials by elastic waves faces two problems. First, the attenuation of the coherent wave must be sufficiently low and second, the coherent back-scattering from the heterogeneity must be small compared to the sought signal from the structure. Applying low frequencies of 100 kHz down to impact echo frequencies of 10 kHz usually ensures sufficiently low attenuation. Thus thick samples up to meter range can be penetrated. Suppression of back-scattering noise can be achieved by focussing to the sensitive volume range. This can be very efficiently realized by scanning over a synthetic aperture combined with synthetic aperture focussing technique (SAFT). For the first time in CE, it was introduced by the Bundesanstalt für Materialforschung BAM . It could be improved  using signals with high time-bandwidth product (HTB) and a special optical technique: the speckle modulation. So an imaging quality for artificial scatterers and parts of a tendon duct could be achieved which has not been reachable by other methods yet. Up to now, this technique bears a significant disadvantage. A large number of averagings is necessary to get a sufficient signal-to-noise ratio (SNR), leading to huge measurement times. The first goal of this paper is to indicate some possible improvements in this respect. Numerical modeling of wave propagation helped to elucidate the physical principles of the ultrasonic pulse-echo method. The study focused on the influence of aggregates, e.g. gravel and sand, and pores ,  on the propagation of elastic waves. Surprisingly, the simulation results revealed that the pores exert a substantial influence on wave scattering and attenuation compared to the influence exerted by much larger aggregates. Currently, the influence of interfaces e.g. between the aggregates and the cement matrix is unclear. Lacking any in-depth knowledge, modeling supposes continuity of appropriate stress and displacement components at the boundaries. Precise non-contact measurements of sound fields taken at surfaces of concrete samples helped to clarify the situation. The second goal of this paper is to demonstrate this.
2. Measurement set-up
The kernel of the experimental set-up is a PSV 100/300 Polytec Scanning Vibrometer supplemented by an OVD 30 displacement decoder. Scanning vibrometer detect the particle displacement or the velocity component normal to the surface sequentially along a grid of points. Vibrometer output is given by a series of voltagetime signals proportional to displacement/velocity and corresponding to the sequence of grid points. The hardware and software of commercial Polytec Scanning Vibrometers are optimized for modal analysis applications. Signals are digitized only for lower frequencies not covering the whole bandwidth (20 MHz) of the OVD 30. The software in the PSV 100 stores averaged data only in the form of complex spectra. This allows resonance modes to be visualized but no transient propagation of waves after pulse excitation. MUSIC (Modular Ultrasonic Imaging in Concrete) has been developed to overcome these limitations. Its main components are a PC, an arbitrary waveform generator, a high voltage amplifier and a digital scope (DSO). The PC controls the arbitrary waveform generator and permits excitation of the transducer by user-designed signals. It steers the motion of the vibrometer mirrors and collects signals from the DSO. If required, it also computes a cross-correlation with the excitation signal and stores the results. It turned out that visualization of the raw data is necessary as a very important early validation of the measurement. Running under LabView, the software module ViewLasus was developed. It allows extraction of A-scans for all measurement points, B-scans along selected scan lines and C-scans (snapshots of the displacements). Animation of the wave motion is performed by displaying a sequence of Cscans. Additionally, evaluation of the signal-to-noise ratio (SNR) is possible both for each single scan point and averaged for the whole scan area. A simple preprocessing can be carried out. Single points of high SNR can be removed manually. Alternatively, all points exceeding some SNR level can be set to zero. Reductions in the data set by thinning out the number of scan points or cutting of the scan area is supported too.
3. Significant conditions for the laser detection of ultrasonic waves
3.1 Applicability of retro-reflex foil
The optical properties of the scanned surface strongly influence the interferometer noise. The more scattered light is caught by the interferometer, the lower the noise. As known, retro-reflex foil is advantageous with respect to high back-scattering of light. But it inherits also a drawback. Our experience gained for the high frequency (some MHz) range shows that this foil does not follow the surface motion instantaneously, there are time delays and phase shifts instead. Before its wide use in sound detection at concrete samples, it has to be cleared how good this foil follows the surface motion. The frequency range of interest is below 500 kHz. Another critical point is a possibly imperfect contact between foil and surface, e.g. due to forming small blisters in the process of foil application. A wave of a frequency f = 500 kHz and traveling along a specially prepared surface was visualized with ViewLasus (Figure 1). A part of the surface remained untreated, another part was colored with bright paint and a third one was covered with retro-reflex foil (Scotchlite 7610 of 3M). Figure 2 shows that the untreated region does not deliver any relevant data due to high noise. The waves in the painted and foil regions show no difference. Especially at the transition of the two fields, there is no phase difference or time delay which would mark a sign of a distinct motion of the foil compared to that of the surface. So the applied foil proves to be usable at least up to 500 kHz, enough for all applications in concrete. Of course, the quality of the gluing depends on the quality of the surface and the skill of the operator. This has to be tested for each prepared surface separately.
3.2 Signal to noise ratio for different surface qualities
ViewLasus was applied to evaluate the signal-to-noise (SNR) ratio in the following
way. The time interval before the wave reaches the scan area is taken. During that
period there is only noise at all scan points. This noise is calculated as rms (root
mean square) value and divided by the signal maximum magnitude at the same scan
point. This value is then averaged over the whole scan area (typically 400 points).
Noise and therefore the SNR is influenced by various factors. Figure 3 summarizes
the influence of some of them, which were determined in a series of experiments.
Excitation was exerted by a single square pulse of 600 V and of a duration of
2.5 ľs. The exciting transducer was a Krautkraemer G0,2R1 probe. The surface of
the sample was pure ground, additionally painted and equipped with retro-reflex foil.
Scattering of the obtained values indicates the difficulty with all vibrometer measurements:
the signal-to-noise ratio of a given measurement depends on a number of
factors which are difficult to be controlled in detail. Nevertheless, some trends are
3.3 Influence of excitation on SNR, sine chirp versus square wave chirp
Signals with high time-bandwidth product can be used to increase the SNR via an
increase of the transmitted energy. For the purpose of vibrometric detection of ultrasound,
this was applied in MUSIC  for the first time. Up to now only sine chirp signals
were used as HTB signals. But the output of power amplifiers for sine chirps is
limited mainly due to the internal power loss, which must be dissipated in an appropriate
way. To further enhance the signal-to-noise ratio, a special transmission amplifier
was developed and built which amplifies binary signals, e.g. square chirps, msequences
and Golay codes. Especially square chirps were tested. The following
3.4 Estimation of possible reduction of measurement time
Improved signal-to-noise ratio in a single signal reduces the number of repetitions necessary for averaging and therefore the total measurement time. The degree will be estimated in the following: If vs is the voltage value of the signal and vn the rms voltage of the noise, then the SNR of a single (not averaged) measurement is given as SNR0 = vs / vn. Let SNR be the signal-to-noise ratio necessary for successful use of the data, Nx * Ny the number of scan points, Ts the length of the transmitted signal, Tr the repetition time and d the duty ratio = Ts / Tr . Then it follows
The reduction of the time needed for one complete scan can be estimated according to the above consideration as more than two orders of magnitude (Table 2). This huge value cannot be realized fully because the transfer of data from the acquisition tool (DSO) to the PC needs some fixed time per measuring point summing up to a value which cannot be neglected for fast scans. Nevertheless it was possible to reduce the typical time spans for the data acquisition from the order of 10 h (example given in ) to about 1 h. Figure 4 shows a result of SAFT processing of raw data, obtained with this new technique. It should be mentioned that up to now injection faults can not be detected in this way.
4. Contributions to the wave propagation phenomena in concrete
The capability of ViewLasus to visualize transient vibration patterns is applied to two further topics: visualization of the transient vibration pattern of exciting transducers and experimental study of some aspects of wave propagation in concrete. Music and ViewLasus together make it easy to visualize the transient vibration pattern of ultrasonic transducers at their aperture. For demonstration the Krautkraemer G=0,2 R1 probe was used. Figure 5 gives a snapshot of a longer movie. There is clear evidence that the transducer acts like a piston source. Deviations from the piston shape of the displacements are small and concentrated at the rim of the excitation surface. Careful determination of the displacement field allows the positions of the electrical contacts in the transducer to be identified. These types of measurements can now be carried out routinely. They give valuable information about the excitation pattern which is important as input for modeling of sound propagation. For an experimental study of wave propagation, a small concrete sample (9 x 9 x 5 cmł) was used. The transducer was placed near an edge of it and the ultrasound was detected at the adjoining free surface. The transducer was excited by a pulse 3 ľs in length at a voltage of 1000 V. It produces broadband ultrasonic pulses with a bandwidth ranging from 50 kHz to 200 kHz.
The upper left picture in Figure 6 shows an optical image of the scanned sample surface. The aggregate structure is clearly visible and some pores can also be seen. The other images in Figure 6 provide a series of snapshots of the wave propagation as a function of time. These gray-scale copies of color images show a substantial scattering of the excited wave. Additionally, careful examination of the original color images reveals several discontinuities in the measured displacements. One such discontinuity is also shown in the gray-scale images of Figure 6 (see arrows). It apparently coincide with a boundary between an aggregate and the cement matrix. Since (out-of-plane) displacement (which is given in the images) is always continuous at ideal boundaries, there must occur at least partial delamination or a weak boundary at this position. It is very likely that such delaminations are typical for concrete. If so, this should also be considered in the simulations because it will considerably increase scattering and attenuation. Scattering and attenuation might be too low from current modeling as also implied by direct comparison (Figure 7). In this case, the experimental snapshot is compared to a snapshot of a 3D simulation with a model sample. Its aggregate size distribution (grating A12) and porosity (1.5%) are similar to that of the experimental sample. The center frequency of the excitation pulse in simulation is 200 kHz which corresponds to the upper frequency limit of the experiment. The samples (9 x 10 cm˛ and 9 x 5 cm˛) are given at the same magnification in Figure 7. The scattering is considerably lower in the modeling scenario, with closed waveforms being visible even after several reflections, whereas in the experiment virtually diffuse energy transport already takes place just before the first reflection of the pulse.
5. Summary and outlook
Laser detection of ultrasonic waves over a synthetic aperture together with SAFT reconstruction proved to be powerful for NDT of concrete. The discussed measures for its improvement are able to overcome its most serious disadvantage - the large time necessary for data acquisition. Now it is possible to take one scan in about 1 h. For the first time, visualization of the wave propagation in concrete showed explicitly that the assumptions made for modeling have to be re-examined thoroughly. Discontinuities identified in the sound field between aggregates and cement matrix can only be explained by weak interfaces or small delaminations. It can be assumed that additionally to aggregates and pores - a further cause for the strong scattering of elastic waves in concrete has been found. Future work is being planned to verify this. It is envisaged to directly compare modeling and measurement results at the same realization of heterogeneity. For that the distribution of aggregates and pores in the sample has to be determined explicitly and put into the model. Preliminary tests showed that X-ray computer tomography is able to yield the data necessary i.e. a map of the aggregate and pore configuration of the sample. Probably, one reason for the lack of evidence of injection faults in experimental pulse-echo and impact-echo data is just the presence of such weak interfaces also between concrete and tendon ducts and between tendon ducts and injection mortar. To prove this conclusion, the sound field measurements will be extended from ultrasonic to impact excitation and the samples will also include sectioned tendon ducts.
This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) within the scope of the Forschergruppe FOR384, and this is gratefully acknowledged.