Keywords: Concrete, Impact-Echo, Radar, 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
With the aim to solve these and similar problems research efforts have in the past years
focused on the development of the pulse-echo technique (The term pulse-echo is used in a broader sense here ). In this context the pulse-echo technique includes the following methods:
When some of the established methods used in other areas of materials testing are applied to concrete testing, certain difficulties are encountered as concrete is an inhomogeneous, porous building material. However, in the recent years great efforts have been made to adapt such methods to testing concrete and some of them are being applied successfully.
Numerous companies and institutes are working on such projects employing, however, different types of test specimens and different test objects. Hence no comparable data are available on the specific advantages and limitations and also on the accuracy of these methods. What makes it even more complicated is the fact that concrete is a highly variable product, even when the main production parameters, such as maximum Aggregate size and cement mix are identical. There are quite a few more parameters, such as moisture content and pore size, which greatly influence the propagation of microwaves and elastic waves.
Especially with regard to the ultrasonic pulse-echo methods we examined different test setups and evaluation techniques for testing concrete, i.e. direct evaluation of the echo signals, signal filtering and algorithms using the principle of synthetic aperture.
In order to be able to make reliable statements on the advantages and limitations of these methods, we decided to carry out comparative tests on identical test specimens with the methods which we have partly modified and improved. Additionally, the propagation, especially of acoustic waves in concrete were simulated to substantiate the test results and solve specific problems.
As the test specimens described in the literature often involve too many problems which render the interpretation of the results difficult, simple test specimens were designed for the first test series described here.
Below the objectives of the investigations and the test specimens are described, the methods are outlined (chapter 2) and the results are compiled (chapter 3.1). In chapter 3.2 the results are compared, partly using simulation techniques. This is only a short presentation; for further information the reader is referred to publications which are partly included in the symposium proceedings.
Figure 1: Plan of the concrete specimens. For the results discussed in the text note, that the x-axis leads to the top and the y-axis to the right, so that the z-axis is positive under the surface. (30k)
On these specimens, the following test series involved the
2: Part of research project from Deutscher Ausschuß für Stahlbeton and Deutscher Betonverein (Authors: Prof. Dr. Ing. P. Grübel, Dr. rer. nat. O. Kroggel, Dipl.-Ing. R. Jansohn, Dipl.-Ing. M. Ratmann)
For locating the metallic duct and for thickness determination, the antenna has been moved in parallel tracks over the surface of the specimen. Figure 2 shows the radar scans of specimen 1 measured at surface B (far from the duct). The amplitudes illustrate the strength of the reflections. From the dielectric constant and the transmission time, a concrete thickness of (508 +/- 10) mm is calculated. The concrete cover on the duct is (340 +/- 10) mm. The lateral position of the duct can be determined with an accuracy of +/- 10 mm. The localization of the duct at different points is shown in Figure 10.
Figure 2: Radar B-scan for specimen 1, measured from surface B. Note, that the y-axis is turned in comparison to figure 1. In the region of intense duct reflections the backwall echo is sup- pressed. (13k)
The measurement results of both types of specimens with different Aggregate size do not differ significantly. Because of electromagnetic shielding the Radar measurements are not suitable to detect voids in the metallic ducts. For more details sec .
In order to overcome the specific problems of concrete, a specially developed low frequency
flaw detector for construction site testing of strongly attenuating (sound scattering) materials
was applied in combination with low frequency probes especially optimized for this field of
application. Its A-scan display enables, even with concrete, the evaluation of the signal amplitudes and times of flight as with common ultrasonic material testing .
Most of the test tasks set here could be solved for the specimen 1 (8 mm maximum Aggregate
size) with satisfactory results. Figure 4 shows the first and second backwall echo from the 500
mm wall (digital reading T = 498.1 mm) measured with a single element probe from the
With this equipment and with the application of the usual pulse echo method it looks possible to carry out tests on concrete with admixtures larger than 8 mm; for the specimen 2 (32 mm maximum Aggregate size) the echo signal is measurable, but the signal to noise ratio is not yet satisfying.
Figure 4: Wall thickness measurement for specimen 1 using the pulse echo method and A-scan evaluation.
|The tests were performed with an array consisting of seven transducers. The fast setting mortar used as coupling agent is well matched to the acoustic properties of concrete. Another advantage is that the transducer array is held in the desired position without any additional device. The equipment used is described in detail in these proceedings . Measurements were made at all transmitter/receiver-combination of adjacent transducers. This means a total of 24 single shots which cover an area of approximately 150 mm in diameter. The signals are processed by spatial averaging to suppress the effects of scattering at Aggregate particles. Signal processing was done by matched filtering and calculation of the envelope by utilizing the complex analytic signal. Figure 5 shows two A-scans at two locations of the specimen 1, surface A. The solid line shows the result over the void 2 of the tendon duct. The peak at 104 mm indicates the depth of the duct surface. An explanation for the slight deviation of the depth measurement is the fact that the array covers an area so that the geometrical situations are not exactly the same for all single measurements. The backwall echo at 505 mm (grey line) is taken at a point without duct and mesh reinforcement and is in good agreement with the true thickness of the specimen.|
Figure 5: Specimen 1 Face A measured by means of ultrasonic pulse array technique.
|By manually moving the US-IE transducer along the concrete surface (specimen 1, surface A) B-scans were generated on-line. This permits a lateral localisation of the duct within ñ 5mm. In figure 6 a B-scan measured along a vertical axis is shown. This was repeated along 16 axes of the surface, so that the localisation of the duct could be measured precisely, these points are included in figure 10. The position in the depth was interpreted from the simultaneous measured A-scans in the range from 102 to 78 mm depending on the site. The measurements were also possible from the reinforced half of the specimen, but yet no difference between void and filled duct could be detected. Further details are described in [ 11].||
Figure 6: (8k) |
B-scan, measured at specimen 1 surface A along a vertical axis. The reflections from the duct begin at 50 µs, the hyperbolic form can be noted, resulting from the scan.
|Ultrasonic data is recorded in single transducer pulse echo technique along a linear aperture at
regular spacing. A SAFT-imaging algorithm  is employed to reconstruct a 2-dimensional
section of the material below the aperture. The resulting images, usually B-scans and demodulated SAFT-images, are plotted by means of the analysis software.|
Figure 7 shows the reconstruction of an unfilled post-tensioning duct. The bright spot indicates the part of the ducts surface which is illuminated by the acoustic waves. The duct is easily distinguished from background noise originating from Aggregate scattering and reconstruction artefacts. In contrast, the image of a filled duct exhibits a smaller amplitude relative to back- ground noise (not shown here). Aggregate size, sound path, and transducers contact to surface are most important to influence the quality of the images. The accuracy of the localisation of an object depends on the knowledge of the exact pulse velocity which in turn depends on the length of the transit path due to dispersion. The values in Table 1 are found without dispersion correction. For more detailed information on this imaging method see  in these proceedings.
| Figure 7: (13k)|
LSAF-R image of the metal duct calculated from a B-scan at specimen 1 (surface A). The B-scan was measured along the x-axis between 1 and K (horizontal axis in this figure, compare figure 1). 0 means point I/K-5/6, negative values show to the direction to the top of the specimen.
A large number of data is required for applying the 3D-SAFT algorithms. They are produced in our investigations by using a scanning laser Doppler vibrometer as an ultrasonic detector (typical number of dots 1000, screen width 10 mm). Another advantage of this technique is that apart from applying a reflective coating no special preparations are required even for very rough surfaces. As an example we demonstrate the location of void 2 measured from face B of specimen 1 (figure 8). The diameter of the void is about 20% of the aperture diameter so that the void and the tendon duct can easily be identified when the thickness-sensitive algorithm is applied.
The three-dimensional image is produced according to the principle of the three-dimensional SAFT reconstruction . The intensity maximum correlates with the void position and indicates, that it is shifted to the right (figure 8b). No signals are received from the filled tendon duct so that the unfilled area can easily be identified.
Figure 8a and 8 b:
Evaluation of a set of 975 A-scans (point width 10 mm) over void 2 (measured from side B, see figure 1). A fixed emitter position (ultra- sound centre frequency 80 kHz) and a scanning laser Doppler interferometer as receiver were used. Left: Evaluation by means of thickness sensitive algorithm, the duct and backwall echo appear. Right: Reconstruction by means of 3D-SAFT, sectional view at z = 330 mm. The void is clearly separate from the filled duct and appears to be shifted to the right.
|Fig. 8a||Fig 8b|
Figure 9 compares EL-FT-SAFI' images in the pressure wave mode along a vertical line, which have been computed from EFIT simulated data for various scattering objects: solid metal cylinder, hollow metal tube, metal tube filled with cement, metal tube filled with cement and exhibiting delaminations. The solid metal cylinder is identified by strong reflections from the top surface and the backwall, the backwall signal is missing for the hollow tube; it appears again for the cement filled tube, but the top surface is lower in amplitude. The delamination causes the backwall to disappear, the result is nearly compatible to the hollow cylinder.
Figure 9: Comparison of EL-FT-SAFT images in the pressure wave mode along a vertical line.
Results for the location of the duct from surface A (specimen 1). The numbers correspond to the group number used in the text. The result shows that the duct has not exactly the same position as it was given by construction (see figure 1).
With regard to the thickness measurement the axial resolution determined by the different methods is given in parentheses (3). The additional error introduced when variation of velocities occur, can be disregarded here as the velocities were determined separately for each method at the point of interest. The data of the concrete cover thickness over the tendon duct in the area without mesh reinforcement are compared in Table 1 tot void 2. They differ from the nominal value (100 mm from surface A, 3 10 mm from surface B) only when the impact-echo method (no. 21) was used and the SAFT reconstruction technique (nos. 8, 9) was applied and when the thickness measurements were carried out from surface A. Depth measurements of the areas with mesh reinforcement could not be determined by all groups. The radar measurements of surface A produced interfering signals from reinforcing bars and the tendon duct. Although the measurement of the depth is affected by the reinforcing bars the ultrasonic B-scan reconstructions (no. 5), however, permit an accurate determination of the depth. Out results indicate that the determination of the geometry of the tendon duct involves principally no problems. Further investigations are, however, needed to optimise the techniques in relation to the effects of different mesh sizes and number of reinforcing mats. The location of unfilled areas in a tendon duct could so far only be determined adequately by means of the 3D-SAFT technique no. 8) or by the L-SAFT (no. 7) technique by comparing the signal intensity. The difficulties encountered can be explained by the EFIT-simulation technique:
It is inevitable that delaminations occur in the filled areas of the tendon duct as the cement
shrinks. The acoustic waves are reflected from these delaminations as they are from other
voids. When the stress wave patterns reflected from an unfilled area of a tendon duct and a tendon duct with delaminations are reconstructed by EFIT, it becomes evident that the ultrasonic
intensity of reflections from an unfilled tendon duct are slightly higher (figure 9) than those of
from a duct with delaminations, in our case calculated here the dominated areas of the tendon
duct reach 50%. The results reveal clearly that only the image reconstructions offer the possibility to differentiate these areas clearly. In how far the different backscatter signals of spot
measurements supply useful results - as was the case with the impact-echo technique - must
further be investigated.
( 3. The axial resolution is determined here as follows: Half width at half maximum (FWHM/2) for gaussian like distributions of intensity vs depth and the slope time for systems who work with the pulse edge.)
Table 1: Over-view for the main results. The no. corresponds to the method-number in the text.
|Thickness (axial resolution [mm]b)||Detection of the Duct (Depth [mm])||Distinction void / filled duct|
|near (Surface A)||far (Surface B)|
|AS 8 c||AS 32||no reb. d||rebars||no reb.||rebars|
|1||Radar||B-scan||508 (20)||504 (20)||90||90 e||340||340||impossiblef|
|2||Impact Echo|| Frequency|
|3||UT||single element||A-scan||498(24)||uncertain||N. R.||N. R.||N. R.||N. R.||N. R.|
|sep. transm. and receiver||A-scan||505 (24)||uncertain||105||105||290||305||N. R.|
|4||Array||A-scan||504 (21)||N. R.||115||96||341||---||N. R.|
|5||single element||B-scan||508 (15)||N. R.||97||90 - 87||---||---||N. R.|
|6||single element|| B-scan, |
| 500 |
|uncertain||89||89||329||---||from surface A|
|7||2D-Aperture|| Phase shifted|
|500 (40)||N. R.||140||---||330||---||N. R.|
|8||3D-SAFT||500(50)||---||120||---||330||---||from surface B|
The cooperation will be continued as the results obtained so far have shown that such comparative investigations do not only strengthen the confidence in one's own findings but are a valuable encouragement for further developments. Theoretical simulations proved to be very useful to understand better the processes involved and will definitely promote future investigations.
|Top to this page|
|| UTonline ||