Keywords: Concrete, Coupling, Matched Filtering, Transducer Array, Ultrasound
This paper was presented at the International Symposium Non-DestructiveTesting in Civil Engineering (NDT-CE) 26.-28.09.1995 in Berlin. NDT-CE, Full Program or the Ultrasound Part

Table of contents
1. INTRODUCTION
2. DEVICES
   2.1 TRANSDUCERS
   2.2 SIGNAL GENERATION AND -ACQUISITION
3. SIGNAL ANALYSIS
4. DETECTION OF TARGETS OF INTEREST
5. EXPERIMENTAL WORK
   5.1 MEASUREMENT OF THICKNESS
   5.2 DETECTION OF VOIDS
   5.3 DETECTION OF HONEYCOMBS
   5.4 DETECTION AND EVALUATION OF TENDON DUCTS
6. CONCLUSIONS AND REMARKS ON IMPROPER FILLED DUCTS

1 INTRODUCTION

The development of non-destructive testing methods utilizing ultrasound pulse echo technique is the intention of a research project started in September 1992 at Technical University of Darmstadt in co-operation with Instrumentation Innovation Ltd. It is sponsored by Deutscher Ausschuß fiir Stahlbeton, Deutscher Betonverein E. V. and industrial partners. The aims of the investigations are to provide methods for the inspection of concrete buildings in order to prove their safety and to give a non-destructive testing tool for quality control into the civil engineers hand. The targets of interest are the measurement of thickness of structures which are only accessible from one side, the detection of cracks, voids and honeycombing. Due to several damages of prestressed structures there is considerable interest to prove whether a tendon duct is properly filled with grout or not. The grout pumped into the void between the prestressing tendon and the duct has to connect the surrounding concrete to the tendon mechanically and to protect the steel against corrosion. The absence of grout in some parts of the duct which may be tolerable from the mechanical point of view may become harmful because under unfavourable conditions there is a danger of sudden structural collapse without pre warning if the tendons corrode. Therefore the detection of improper filled ducts is of great interest in non-destructive evaluation of such structures. It is clear that the detection of voids in a metallic duct surrounded by the inhomogeneous material concrete, which is reinforced by several layers of reinforcement meshs is not a simple task.

2 DEVICES

2.1 TRANSDUCERS In most cases of ultrasonic concrete testing in pulse echo technique the received signals are contaminated with noise caused by the random arrangements of aggregates. This grain noise can be reduced by making use of spatial averaging. Utilizing large aperture transducers is the simplest way of averaging. Another advantage of such transducers is the narrow beam in emission and reception which simplifies the signal interpretation. Dividing the large aperture into a transducer array allows the performing of different averaging strategies. The type of signal analysis applied here requires a large bandwidth with an upper limit of approximately 300 kHz the wavelength of about 1.3 cm then corresponds to the average diameter of aggregates in most common concrete types.

figure 1: Transducer array

Such an array consisting of 19 single elements is shown in figure 1. Every element may be used as a transmitter or a receiver. Their sensitivity against compression waves is much greater in comparison to their sensitivity against shear waves. Such a behaviour is advantageous because the contribution of different wave modes to the ultrasonic signal in combination with grain noise makes useful signal evaluation nearly impossible. Especially for measurements with one transducer acting as a transmitter and another as receiver the degradation of surface waves which consist of a bigger shear than compression component is advantageous. The array is attached to the concrete surface with fast setting mortar as a coupling agent. After its hardening the transducers are held in place so the handling becomes comfortable for the operator especially if measurements are taken over head. Furthermore the acoustic impedance of fast setting mortar is nearly equal to those of concrete and there is only a negligible loss by the entry of the ultrasonic wave into the concrete volume.

2.2 SIGNAL GENERATION AND -ACQUISITION
Conventional devices for ultrasonic testing are generating a short electrical pulse to excite the transducer. With such a technique the resulting ultrasonic signal depends mainly on the transfer function of the transducer. For the signal processing described below it is necessary to create ultrasound pulses of a special form with well known frequency content. For that reason the transmitted signal must be produced by a signal generator.

The generation of electrical drive signals is realized by the use of a PC plug-in board which contains a programmable memory to store the desired waveforms and a digital-to-analog converter with a resolution of 12 bits. Due to the output of the D/A-converter is only 10 V peak to peak it is necessary to make use of a power amplifier. Its maximum amplitude is 200 V. To protect it against overheating it is only switched on during a signal is transmitted, The transducer array is controlled by a 20 channel multiplexer which is able to switch the transducers in an user defined sequence. One special property of the multiplexer is the ability to switch preamplifier transmitted signals of several hundreds of volts as well as very low received signals attenuated by concrete. The internal isolation of the multiplexer is sufficient for switching such different voltages with a tolerable electrical breakthrough, This breakthrough can be compensated by signal analysis described below. The time of switching is about 1 microsecond. Standard transmitter receiver combinations are offered in the control program but it is also possible to produce new ones interactively during run-time.

The signal collected by the receiving transducer is converted by an analog-to-digital converter plug-in board which writes the received signals into its memory. It has a resolution of 12 bits and a maximum conversion rate of 10 MHz. In order to keep the data this record is stored on hard disk or on another medium after collection. With this data acquisition system it is possible to collect the data of a transducer array in approximately 1 second.


figure 2: hardware configuration

3 SIGNAL ANALYSIS

There are two kind of echoes in the received signal of an ultrasonic back-scattering test performed on concrete. One kind of echo is caused by sources which are relevant for the detection of defects. The other is caused by back scattered waves at concrete aggregates. This information might be useful if the state and kind of concrete itself are of interest. The noise like contribution of back scattered waves at aggregates becomes extremely harmful for defectoscopy. Therefore this part of the signal is often called grain noise.

A method to optimize the signal to noise ratio is given by conjugate filtering [1]. This operation convolutes the received signal with a time-reversed version of the transmitted signal. In the ideal case in absence of attenuation, noise and grain noise the received back wall echo is a time shifted version of the transmitted waveform and the conjugate filter forms the time shifted auto-correlation function of this waveform. It is clear that only the ratio between the useful signal and thermal noise is maximized. Due to the coherent character of grain noise the conjugate filter is not able to improve the ratio between the useful signal and grain noise. Detectibility and time resolution of echoes depend on the sharpness and the magnitude of their corresponding peaks in the processed signal. The use of test signals with narrow auto-correlation functions in combination with small side lobes is advantageous. This technique is called pulse compression [1]

In order to simplify the interpretation the envelope of the filters output is formed. The maxima of the resulting function correspond to time shifts of best relation between the transmitted and the received signal. As mentioned above echoes from concrete aggregates and from other targets produce such maxima as well.
The transducer array enables the operator to examine a circular area of 250 mm diameter. The recorded data can be processed by two strategies [2].

The first is simply combining the records by superposition and forming the average. Changes in phase over the examined area belonging to the same range and time of flight are leading to cancellation. Echoes produced from defects which are assumed to be extended over a greater volume in the same range are amplified and the probability of detection increases. Both effects are helpful for the localization of back walls extended defects which are nearly parallel to the surface of the test object. Since the method makes use of the signals phase it is called phase sensitive detection. The resulting data set is than treated by conjugate filtering and envelope detection.

The second strategy of combining the records of adjacent receivers named phase insensitive detection is to perform superposition and averaging after conjugate filtering and envelope detection of every received data set. This method preserves the whole back scattered signal energy. That means scattering from all secondary sources and also multiple scattering form the resulting function. Because this function contains on the kind and the condition of concrete it can lie used to determine material properties.

4 DETECTION OF TARGETS OF INTEREST

Due to the attenuation which is mainly caused by multiple scattering at aggregates it is often hard to decide whether an echo is caused by a defect or by some aggregates. In most cases the operator is interested in the detection of echoes introduced by defects. The phase insensitive measurements can be considered as a scattering signature of the examined kind of concrete. By means of accomplishing a great number of spatial averages using this processing strategy the curve becomes smoother. This curve can be used as a reference which contains the back scattered energy of the applied signal from the individual type of concrete which is considered as a material property. In order to use it as a threshold the sensitivity can be adjusted by a weight factor. The measurements from the positions of interests are processed by the phase sensitive algorithm and then compared with the reference. If a peak of the phase sensitive curve exceeds the reference level at an instant it must be expected that there is a scatterer which can not be a normal part of concrete. The occurrence of such a peak can not be explained by arguments concerning the inhomogenous structure of the material. Due to the frequency dependence of back scattering at aggregates it is necessary to compute a reference for each type of transmitted pulse.

Another method to identify significant peaks is the comparison of the phase sensitive and phase insensitive processed signal computed with the same raw data. If there is an uncanceled peak in the phase sensitive curvature and the same peak occurs in the phase insensitive processed signal, it must be supposed that there is an unusual strong scatterer in the corresponding depth. Due to the extent of the transducer array it is very unlike that this peak is caused by grain noise and the presence of some interface or defect can be concluded with great certainty. The height of the phase sensitive processed peak is always less than that occurring in the phase insensitive curve. This phenomenon must be explained with little phase differences between the single measurements. Such differences may be introduced by some unregularity of the thickness of the coupling agent and by the tolerances of the transducers.

Acoustical and electrical breakthrough discriminates the signal during the pulse is send. By using pulses possessing a large bandwidth even redected signals within the duration of transmission may be used for interpretation. The pulse compression discussed above compresses the breakthrough energy in the same manner like the energy of the received echoes. If the bandwidth of the transmitted signal is sufficient the breakthrough is compressed into the time of flight through the transducers face plate and coupling agent which is not subject of interpretation.

5 EXPERIMENTAL WORK

5.1 MEASUREMENT OF THICKNESS
In the following the results of thickness measurements at a specimen made of concrete with a compression strength of 45 N/mm² and with maximum aggregate size 16 mm. The dimensions of the block were 0.6 m x 0.7 m x 1.0 m.
The signal used was a linear frequency modulated pulse which is often called chirp signal. It was enveloped by a sin-function and had a duration of 100 as. Its centre frequency was 125 kHz with a bandwidth of 150 kHz.
In order to convert the time scaling of the test results into units of depth a transmission test has been carried out. The measured sound velocity was 4261 m/s. A reference measurement according to section 4 was recorded over the longest available distance of 1 m and processed by the phase insensitive strategy.

The result of the thickness measurement over the distance of 0.6 m is shown in figure 3. The result processed with the phase sensitive algorithm (thick line) is showing good conformity to those computed with the phase insensitive one (thin line). As mentioned above the phase insensitive results are always higher than the phase sensitive ones. Only the peaks belonging to the distance of the back wall are exceeding the reference which is shown as a dotted line.

figure 3: backwall echo from 60 cm

The result of the thickness measurement in the 1 m direction of the block is shown in figure 4. The reference measurement was carried out at the same face of the block. It is clear that the phase insensitive result arid the reference are the same and the only available criterion for the significance of a peak is the conformity between phase sensitive (thick line) and phase insensitive (thin line) processed results like it is described in section 4. The only peak which fulfills this is the peak at l.01 m which is clearly caused by the back wall echo.

figure 4:backwall echo frown 100 cm

5.2 DETECTION OF VOIDS
The experiments described here were performed with specimens containing cylindrical holes. Due to the geometry of the defect the problem of finding voids is considerably more difficult compared with the measurement of thickness. The main reason for that are the losses by scattering of the incoming wave to all directions. In the following one example of the test series is presented.

The measurements discussed here were made at a concrete block with the dimensions 0.7 m x 0.5 m x 0.36 m. This block contains a drilling in the centre of the area 0.7 m x 0.36 m of 10 cm diameter. The transmitted signal chosen was a ? sin enveloped, linear frequency modulated pulse with a centre frequency of 125 kHz, a bandwidth of 150 kHz and a pulse duration of 100 µs. A transmission test yielded to a sound velocity of 4042 m/s.

The transducers were placed directly over the hole so the distance to the edge of it was 13 cm. Figure 5 shows the result of the test made in that position. In a depth of 13.4 cm a peak is visible in the phase sensitive (bold line) and phase insensitive signal (thin line) as well.

figure 5: void detected at 13.4 cm

5.3 DETECTION OF HONEYCOMBS
The echoes caused by honeycombs do not differ from those produced at voids. Due to the fact that ultrasound is sensitive against interfaces honeycombs are also reflecting a part of ultrasonic energie.

5.4 DETECTION AND EVALUATION OF TENDON DUCTS
The experiments concerning the detection and evaluation of tendon ducts were performed on several specimens containing metallic ducts which were in some sections filled with grout and in others unfilled. The measurements presented here were done on a specimen with six built-in ducts. Their diameter was 50 mm and their distance to the surface 90 mm. The specimens size was 0.5 m x 0.5 m x 0.4 m and it was made of concrete with a compression strength of 45 N/mm² and with a maximum aggregate size of 16 mm. The ultrasonic compression wave speed was determined in transmission to be 4600 m/s.

Preliminary tests showed that the evaluation of echo amplitudes is difficult because the scattering behaviour of the examined concrete differs from place to place. This is an important observation for the evaluation of the state of ducts.

In order to begin with the most simple case only two states were examined. The first was "completely unfilled" the second was "perfectly filled with grout". Due to the statements made above only positions with nearly the same geometry were chosen. That means the spatial averaging process was performed with data sets from transducers with the same angle of view to the duct. The average was done with 36 sets of raw data. By means of the large amount of data the influence of grain noise was minimized. The results of the measurements are shown in figure 6. The processed signal corresponding to the measurement over the filled duct is shown as a solid line the one over the unfilled as a dashed line. It can be seen that the empty duct produces an echo which is 2.24 times higher compared to that produced by the filled one. Furthermore it can be seen that the weak back wall echoes are nearly equal.

figure 6:figure 6: measurements over unfilled and filled duct

It must be concluded that the distinction between filled and unfilled ducts utilizing ultrasound is principally possible There is a need of further investigations concerning the technique of detection and the definition of harmful voids for the tendons. The question to which extent voids are tolerable naturally affects the development of non-destructive testing tools and methods.

6 CONCLUSIONS AND REMARKS ON IMPROPER FILLED DUCTS It is possible to measure the thickness of concrete structures and to detect tendon ducts, voids and honeycombs with the ultrasound impulse echo technique. The question wether a duct is properlly filled or not can only be answered by spatial averaging utilizing a larger amount of single measurements. Echoes caused at ducts where the mortar lost the bond to the ducts inner surface are not to distinguish from those generated at completly unfilled ones. Figure 7 shows four stages of imperfect filling. Because ultrasound is sensitive against interfaces in all stages an ultrasonic test would yield in nearly the same result. It is obvious that only the last stage becomes harmful for the tendons corrosion protection.

figure 7: different stages of imperfect filling

REFERENCES

1. LYNN, P. A.: An Introduction to the Analysis and Processing of Signals. 3rd edition, Hemisphere Publishing Corporation, New York Washington Philadelphia London, 1989
2. BAUR, E.: Einführung in die Radartechnik. B. G. Teubner, Stuttgart, 1985
3. Bilfinger + Berger Bau AG, W. Franz GmbH, E. Heitkamp GmbH, Ph. Holzmann AG, Imbau Industrielles Batten GmbH, Wayss & Freytag AG