Micro-CT for crack detection in concrete

Josef Feuerecker, 04.02.2014

Artikel auf Deutsch

Micro-computed tomography (micro-CT) using x-rays or gamma rays is a non-destructive testing method which provides a local image of the object using radiography and the differences in the absorption of radiation. A volume rendering of the specimen is produced by rotating the specimen and by overlapping the images taken using a computer.

Historical development

X-rays, which were first discovered in 1895 by Wilhelm Conrad Röntgen (1845-1923), established the basis for all future developments in this testing method. The first usable developments towards computed tomography emerged in 1969 and were primarily developed for medical applications. The first machine invented by Godfrey Hounsfield consisted of a tube to transmit x-rays and a detector to measure the intensity of the emergent x-rays. The translational motion of the detector and tube at a fixed distance to each other enabled the layer under investigation to be scanned. Only much later did scientists recognise that this technology was also suitable for industrial applications. The major difference between current CT machines used in industry and those used in medicine is that the tube and detector move around the object in the case of medical equipment, whereas the test object is moved in the case of industrial machines. The spectrum of the materials to be tested is also considerably wider in industrial applications. The resolution of the images generated has improved in the wake of further technical developments in CT machines. The term micro CT is used to indicate that the voxel size (volume element) is in the double-digit micrometre range.

Fundamental physical principles

The fundamental physical principles of this testing method are those of x-ray tomography (see X-ray tomography). The x-rays used consist of electromagnetic waves in the wavelength range from 1.0 to 10 nanometres that propagate rectilinearly. When the x-rays hit matter, they are attenuated. This attenuation can be affected by many factors. On the one hand, it is affected by the material, particularly its bulk density and chemical composition. On the other hand, the thickness of test material is a factor. Another influence factor is the energy of the radiation. This can be changed by the voltage of the X-ray tube or by the gamma emitter used. The emergent attenuated rays can then be measured during radiography. The attenuation can be expressed using the following formula:

$//<![CDATA[ \begin{array}{l}I = I_0 * e^{-\mu x}\end{array} //]]>$

where:

• $//<![CDATA[ \begin{array}{l}I_0\end{array} //]]>$: the intensity of the primary energy source in [keV]

• $//<![CDATA[ \begin{array}{l}μ\end{array} //]]>$: the absorption coefficient of the material in [$//<![CDATA[ \begin{array}{l}\frac{1}{m}\end{array} //]]>$]

• $//<![CDATA[ \begin{array}{l}x\end{array} //]]>$: the distance travelled in [m]

• $//<![CDATA[ \begin{array}{l}I\end{array} //]]>$: the attenuated radiation over the distance $//<![CDATA[ \begin{array}{l}x\end{array} //]]>$ in [keV]

The radiation that emerges from the test object is allocated an absorption number according to its attenuation and each is expressed using different grey scales. Individual materials can be differentiated from each other owing to the differences in attenuation of radiation, thus producing a projection of a solid on a plane. In computer tomography, however, many such images are created from various perspectives and then overlapped using a computer-based algorithm, thus creating an image of the volume structure of the specimen. As manually evaluating the featured crack involves considerable effort, a suitable algorithm is needed to carry this out efficiently and with the desired precision.

Application

A desired specimen is placed on the test bench during the test procedure. Measuring the specimen is mainly limited by the radiation energy that can be generated. This is because concrete exhibits a relatively low half-value thickness. This means that only a few centimetres can be penetrated when the radiant power is low. The test bench rotates from the start of measurement. In contrast with medical applications, this is easier to achieve mechanically. The emitter is positioned on one side of the machine, while the sensor is positioned on the opposite side and measures the radiation that emerges. It is therefore possible to completely radiograph the specimen layer for layer, thus providing a volume rendering. Crack detection in concrete is advantageous because the attenuation coefficients of air and concrete, indicators of the attenuation of radiation, are very different. In the case of a radiation energy of 10 MeV, the attentuation coefficient of concrete is $//<![CDATA[ \begin{array}{l}μ=0,0537 [cm^{-1}]\end{array} //]]>$ and that of air is $//<![CDATA[ \begin{array}{l}μ=2,62∗10^{−5} [cm^{−1}]\end{array} //]]>$. For this reason, the layer boundaries from concrete to the crack (filled with air) can be measured easily and the crack can be shown clearly. However, problems may occur due to the largely inhomogeneous composition of concrete. Two radiation sources with differing radiation energies are used in order to solve this problem. In this regard, the layer boundaries enable improved recognition of extremely similar materials such as hydrated cement and mineral aggregate. However, the systematic evaluation of data sets is relatively complex. The algorithms used need to reliably rule out layer boundaries between the rock and cement matrix or small air voids as potential cracks. Suitable algorithms are, for example, the evaluation of Hesse matrix eigenvalues combined with a percolation. Another process is what is referred to as template matching. According to the report, ^[1] this process is particularly suitable as detected cracks can be portrayed better and are loaded with fewer artefacts. Another step towards improving results is post processing through testing of the surroundings. As cracks are mainly flat objects and comprise several connected voxels, a crack that has been previously identified incorrectly can be removed by testing the surroundings. If there are not enough crack voxels in the surroundings, this this area will not be assigned to the crack and thus removed. In particular, this improves the resolution of the peripheral areas.

Further evaluations are necessary in order to make more accurate statements about the cracks. First of all, this includes what is called embedding of cracks. This means that an analysis identifies the surrounding material (concrete or granular material) that the crack runs through. For this purpose, the surrounding grey scale is determined for each crack voxel. Using template matching, it is also possible to obtain a template index that creates a normal vector from the largest grey scale differences of the adjoining voxels. Using this index, the orientation of the crack can be determined and the surrounding material can be investigated. Additional crack characteristics are crack size, crack opening and crack surface. Crack size is the total of all voxels that are part of the crack. Crack opening is the average opening of the crack and is derived from interpretations of polygonal cracking. For example, this is the result of a crack profile. The crack surface is created using polygonal interpretations. Triangular surfaces are generated for this purpose and then evaluated. The surface of all the triangular surface results in the surface of the crack.

Literature

1. Paetsch, O.; Baum, D.; Ehrig, K.; Meinel, D.; Prohaska, S.: Vergleich automatischer 3-D-Risserkennungsmethoden für die quantitative Analyse der Schadensentwicklung in Betonproben mit Computer Tomographie. Berlin, 2012.
2. Grosse, C. U.: Grundlagen der Zerstörungsfreien Prüfung. Skript zur Vorlesung der TU München. München, 2013.
3. Versuch 9: Radioaktivität. Accessed 18.01.2014.
4. Untersuchung moderner Verfahren zur Berechnung der Abschirmung ionisierender Strahlung. Accessed 10.01.2014.
5. Computertomographie (CT). Accessed 20.01.2014.