ABSTRACT

This paper deals with an innovative technique, which improves the pulse echo testing of concrete by using
1. specially designed ultrasonic probes
2. pulse compression technique
3. ultrasonic wave detection by laser doppler interferometer
4. random speckle modulation
5. space time signal processing methods
The pulse echo technique is carried out by sending frequency modulated chirp signals and performing a cross correlation between the received and the transmitted signal. In combination with the application of recently available ultrasonic concrete probes as transmitter, this leads to an improvement of the signal to noise ratio. A laser doppler interferometer, equipped with a random speckle modulator, is used as detector of the ultrasound. Finally, the data sets can be processed with various methods, involving the time signals of several space points. Examples are the space averaging and the synthetic aperture focussing technique (SAFT).
The advantage of the suggested technique is demonstrated by practical measurements. The improvement compared to standard laser interferometric measurements will increase the feasibility of laser interferometric detection for non-destructive testing in civil engineering.

Table of contents
• INTRODUCTION
• SIGNALS WITH A HIGH TIME-BANDWIDTH PRODUCT
• LASER VIBROMETER IMPROVED BY SPECKLE MODULATION
• PRACTICAL REALISATION AND RESULTS
• CONCLUSION
• ACKNOWLEDGEMENTS
• REFERENCES

INTRODUCTION

The difficulties of non destructive testing of concrete by ultrasonic methods result from the strong signal attenuation, caused mainly by scattering at the inhomogeneities of concrete. Besides decreasing the transmitted ultrasonic signal, it leads to strong coherent noise. This noise can mask even large backwall echoes totally. Scattering diminishes for decreasing frequencies. Therefore, rather low frequencies in the range of about 100 kHz must de used.
There has been some progress in resent time. Highly efficient transducers has been developed and are available now. Pulse-compression is known to improve the signal to noise ratio (2) in situations with a high insertion loss. It has been applied to ultrasonic testing of concrete successfully (7).
It is well known from ultrasonic testing in other fields, that space averaging techniques suppress the coherent back-scattering noise effectively (8). For applying such methods, signals from a large number of measurement points are necessary. They can be obtained in a convenient way by using laser interferometric detection. In (3), the interferometric detection is performed in a 2D scanning aperture and the measured data are reconstructed by the 3D-SAFT algorithm (4) involving itself spatial averaging.
Till now the main problem of laser interferometry has been the insufficient sensitivity, resulting in a low signal to noise (S/N) ratio . This is a special problem in an automatic scanning mode, because no fine focussing can be performed, resulting in large incoherent (electronic) noise of the interferometer. Thus, improvement of the signal to noise ratio is crucial for application of laser interferometric detection and space averaging techniques. We improve the S/N ratio by several methods. On the one hand we increase the signal by applying the very efficient probes mentioned above as transmitter and including the pulse compression technique in our system and on the other hand we reduce the laser vibrometer noise by implementation a random speckle modulation technique (5). By this method we get signals of tolerable incoherent noise appropriate for signal processing to suppress the additional coherent scattering noise.

SIGNALS WITH A HIGH TIME-BANDWIDTH PRODUCT

The influence of attenuation can be reduced by increasing the signal energy transmitted into the system. A common way is, to increase the amplitude of the ultrasonic pulse. This makes high demands on the used electronic amplifier and ultrasonic transducer. Furthermore, human safety demands should be borne in mind by working with signal amplitudes not exceeding thousand Volts. The alternative is, to use signals with a high time bandwidth product. All those signals with an auto-correlation function close to the delta pulse are suitable. Applying a matched filter for such a signal the auto correlation function of the transmitted signal results, delayed for signal runtime in specimen. Such a filter is called correlation- or pulse-compression filter (1). Suitable signals for use with pulse compression techniques are frequency-modulated-signals (so called fm-chirps), pseudo-stochastic signals, which are m-sequences or Barker-codes, or white noise. In the suggested method fm-chirp signals are used for transmission. The matched filter is applied by the cross-correlation between the received and the transmitted signal.
The whole energy of the long duration waveform will be compressed in a very short pulse. The amplitude gain (AG) is expressed by
AG = ½ time bandwidth (2). That means, to obtain the same amplitude difference between signal and noise compared to a corresponding pulse excitation, the amplitude of the transmitted chirp signal can be reduced by the factor AG. The pulse width after correlation becomes smaller by increasing the time-bandwidth product (2).


In figure 1 the common pulse excitation technique is compared with the pulse compression technique. For convenience a simple simulation is shown. The system consisting of a transmitting probe, the propagation medium (concrete) and the receiving probe is represented by a delay of 250 µs, a damping of 50 dB and addition of 0.25 mV (rms) noise. The transmitted chirp has a bandwidth of 400 kHz and a length of 250 µs. This yields an amplitude gain of AG = 7. In accordance with that a chirp with an amplitude of only 70 V leads to the same signal to noise ratio after correlation as the pulse with amplitude AG * 70Volts ~ 500Volts. Chirp-signals are best suitable for ultrasonic testing in concrete. They can be generated easily by arbitrary waveform generators. The frequency-spectrum of the chirp can be adapted to the transfer function of the system (figure 2). Consequently, a maximum of ultrasonic energy is brought into the system .

Fig. 1. principle of the pulse compression technique


Fig. 2. linear frequency-modulated chirp signal x(t) and its frequency spectrum X(f)

Broadband transducers are required for chirp signals. The sensitivity of piezoelectric sensors is much lower for broadband signals than for narrowband signals. So the advantage of piezoelectric sensors is diminished in this case. We used an alternative broadband-sensor: the laser interferometer.

LASER VIBROMETER IMPROVED BY SPECKLE MODULATION

The main advantage of optical detection by laser interferometry is the ability of working without contact to the specimen. Thus, laser can be used on rough or awkward shaped surfaces or on areas, which are difficult to reach (bridges etc.). This frequently occurs in civil engineering. Scanning areas on the surface of concrete with a laser interferometric sensor might be an effective method for non-destructive evaluation (NDE). On the other hand laser interferometry has not yet become a practicable tool in NDE, due to the fact that its sensitivity is insufficient in most practical applications. The signal to noise ratio in manufacturers specification cannot be maintained on ordinary scattering surfaces. The surface of concrete is rough and the coherence of laser light will result in interference speckles at random distribution. When optimising the measurement only for one single point on the surface, it is possible to adjust the laser by changing the focus, so that the photo-detector is covered by a bright speckle. In scanning mode normally the photo-detector is partially covered by random distributed bright speckles. Thus, the effective sensitivity inherently fluctuates. Although the absolute sensitivity of laser interferometer remains the same, its noise level increases significantly, when dark speckle are encountered. This adverse effect can be eliminated almost completely by the random speckle modulation technique suggested in (5). The idea is very simple. Changing the focus by a small amount moves the speckle pattern in an unpredictable way. Doing this slowly enough and observing the interferometer signal amplitude, the ultrasonic measurement can be taken, when the interferometer noise is low. This will be repeated at each measurement point. We modified a Polytec heterodyne scanning laser vibrometer PSV 100. Instead of changing the focus as described in (5), we let the scanning mirrors vibrate with small amplitudes. The deflection angle is smaller then 0.04°. That is a deflection of less than 0.7 mm at a distance of 1 m between laser vibrometer and surface. It can be assured, that a bright speckle falls on the photodiode for approximately 1 ms by a modulation-frequency of about 50 Hz. This is sufficient to trigger an ultrasonic transmitter and detect the response before the bright speckle disappears. Figure 4 shows selected signals of a measurement. It can be seen, that for bright speckles (large signal amplitude) the noise is low. In the example shown there is enough time to trigger two measurements. The improvement by using the random speckle modulation is shown in figure 5. We carried out a line-scan (ultrasonic B-scan) over a specimen by moving the laser beam in a range of 50 mm. An ultrasonic transducer is used to generate the pulse having a centre frequency of 100 kHz. Without random speckle modulation some of the A-scans are very noisy, while others gave a good signal. The random speckle modulation increases the average signal to noise ratio by 3 to 5.

Fig. 3. scheme of an ultrasonic laser interferometric measurement including random speckle modulation Fig. 4. principle of the random speckle modulation; when a bright speckle occurs, it triggers a measurement (lower trace), the noise in the signal channel is low for this time interval (upper trace) Fig. 5. laser detected ultrasonic line-scan, left without, right with random speckle modulation

PRACTICAL REALISATION AND RESULTS

The experimental setup is shown in figure 6. It consists of a personal computer, an arbitrary waveform generator followed by a power amplifier, an ultrasonic transducer (Krautkraemer G0,2-R1) and a Polytec laser interferometer with the developed random speckle modulator. The PC controls the process via IEEE 488.2 and performs the cross correlation and data storage. The arbitrary waveform generator provides the chirp signal. The waveform can be easily varied for different applications by PC-controlling the generator. The maximum output voltage of the gated power amplifier is 400 Vpp with a signal period of maximum 500 µs and repetition rates lower than 50 Hz. That is quite enough for most applications. For digitising the received data we use an oscilloscope with a signal averaging option. The sampled data can be read via IEEE- connection into the PC, where the correlation is carried out off -line.

Fig. 6. scheme of the realised measurement system

The scan angle of the laser vibrometer is ( 20° in each direction. The theoretical number of points is 4096 in each direction, this corresponds to 1.6 million points. But it takes approximately 1 to 5 seconds measure time for each point depending on the number of averaged signals. Currently, scan areas up to 100 x 100 points are practicable, which takes less than 3 hours data aquisition. Our arrangement is merely an easy test by combining standard devices. The speed can be considerable increased by using fast DSP-boards for digitising and calculating the correlation. In order to test our method we scanned an area of a concrete specimen, which includes some defined voids in various depths. The test specimen (2000x1500x700mm) has a maximum aggregate size of 16 mm. Figure 7 shows the plan of the specimen and the scanned area together with the transducer positions. The scanned area was 500x300mm with a raster size of 5 mm. The time duration of the transmitted chirp signal with a frequency range of 30 kHz to 150 kHz was about 500 µs. Five signals were averaged at each point. The measurement took approximately 1.5 s per point.

Fig. 7. concrete specimen; the duct, some artificial defects, the measurement fields and the positions of the ultrasonic probes are indicated; the thickness of the plate is 700 mm

By looking at the received data (see selected single A-scan in figure 8a) the Rayleigh wave signal is very strong. This is a special problem of any point sensor and can be reduced by using linear or square aperture. A simple averaging over a line with 100 points is applied in figure 8b. Now the Rayleigh wave indication is suppressed together with the coherent scattering noise. Thus, the backwall echo appears clearly now.

Fig. 8. left: a single A-scan from the data set of the lower scan area, the Raleigh wave (peak at 170 mm) is very dominant, whereas backwall is invisible;
right: averaging over a linear aperture with 100 supporting points, the backwall appears clearly

In figures 9 and 10 the results of reconstruction by means of 3D-SAFT are shown. The backwall can be identified in both images without doubt. Figure 9 represents the results of the lower part of the specimen (figure 7). The voids in the depth of 450 mm and 540 mm are clearly recognisable. The backwall exists in the reconstructed data, but its visibility is suppressed in this image to show detected voids by drawing iso-lines in suitable amplitude. In a depth of 450 mm is a void, filled with a rigid foam quader of 50x50x50 mm3. The other void is a plastic-tube with a length of 50 mm and a diameter of 50 mm filled with gravel. Figure 10 illustrates another measurement, now in the upper scan area of the test specimen (figure 7). A mortar filled duct in a depth of 250 mm with a gravel filled void behind (350 mm) having a lateral size of 200 mm could be located. The reconstructed backwall seems to have a gap. This can be easy interpreted as a shadowed region because of the duct. Additional measurements involving different positions of the transmitting probe may clarify this effect further.

Fig. 9. 3D-SAFT reconstruction of measured data set, left: 2D-projections, right: 3D-visualisation, in the right figure backwall is made invisible for presentation. Fig.10. 3D-SAFT reconstruction of data set from upper scan area. left: 2D-projections, right: 3D-visualisation In the depth of 250 mm appears the duct and behind of them in 350 mm a gravel void.

CONCLUSION

The first results demonstrate the efficiency of ultrasonic testing with laser interferometric detection. In co-operation with pulse compression and random speckle modulation the signal to noise ratio can be reduced so that laser interferometric detection becomes practicable. The technique has to be developed to increase the processing speed. The described technique may help to improve the acceptance for ultrasonic methods in civil engineering. The 3D-SAFT reconstruction is able to increase the signal to noise ratio further and leads to a 3D-image of reflector distributions inside the specimen.
Furthermore, the presented technique can be used as an advanced tool for experimental verification of modelling results concerning the propagation of ultrasonic waves in concrete (6). This is subject of current investigations.

ACKNOWLEDGEMENTS

The results described in this paper are part of the work supported by the "Deutsche Forschungsgemeinschaft" in the grant No Ko 1386/1-1, which is grateful acknowledged. The authors would also like to thank the "Bundesanstalt für Straßenwesen" for giving the opportunity to use the concrete specimen and for their support of our measurements.

REFERENCES

Authors

Bernd Köhler*, G. Hentges*, W. Mueller
Fraunhofer Institute
for Non-Destructive Testing (IZFP)*, IZFP
Branch Lab Dresden
Krueger-Strasse 22
01326 Dresden Germany
email: koehler@eadq.izfp.fhg.de
Homepage http://mm.fhg.de/depts/izfp-e.html

Paper presented at the Conference: NDT in Civil Engineering ´97, Proc. pp 123-134, .
See also the Report about the confernce.