NDT.net • May 2004 • Vol. 9 No.05
CT-IP 2003 Proceedings

Experiences in using a 320 kV micro focus X-ray tube and a large sized amorphous silicon detector for 3D-CT

Y. Onel, B. Illerhaus, J. Goebbels

Abstract

A novel 3D-CT scanner was set up at BAM to extend the applicability of the X-ray transmission microtomography (ľ-CT) to specimens up to 30 cm in diameter. The scanner consists of a bipolar 320 kV micro focus tube and a flat panel detector of amorphous silicon. Minimisation of scattered radiation, control of room and system temperature, and corrective preprocessing on the projection data are necessary to achieve expectable CT images. Statistical evaluations to show the limits and results of CT investigations are reported.

1. System description

Fig 1: 320 kV microfocus X-ray tube (left), 4–axes specimen micropositioner (middle), flat panel detector array (right)

To extend the applicability of ľ-CT to larger and dense objects BAM has set up a further 3D-CT system additional to the existing ones (1,2). We use a bipolar microfocus X-ray tube (3) with a maximum voltage of 320 kV (Fig. 1 left). Therein a standard microfocus tube with –200 kV is used on the one side. In place of a grounded anode target an oil-cooled target with +120 kV is used. In comparison to a 200 kV tube the penetrable material thickness is increased by 50%. The focal spot size depends only on the power of the electron that impinges the target. The high voltage applied to the target limits the distance between target and tube window to 35 mm. Thus the maximal geometric magnification of the system is reduced to 1:30.

The flat panel detector (4) is based on an array of 1024x1024 pixels incorporating amorphous silicon photodiodes and thin-film transistors offering a 40x40 cm 2 active area at a pixel pitch of 400 ľm (Fig. 1 right). This indirect detection array was coupled to a commercial scintillating screen “Lanex Fast”. The exposure time for a single projection varies between 0.5 s and 8 s. Usually the acquisition time of a 3D scan is less than an hour. The acquisition electronics convert the analog pixel data to a 16-bit quantity. The detector system provides a high signal-to-noise ratio (SNR) of 3000 so that its SN characteristics are determined by the quantum efficiency only. The 3D-CT scanner consists of a 4-axes specimen micropositioner controlling specimen height, horizontal translation transverse to the x-ray beam and specimen rotation (Fig. 1 middle). The fourth axis moves the specimen linearly between the x-source and flat panel detector to provide variable magnification.

The smallest possible voxel size of the this CT scanner is about 35 ľm which is suitable for scanning of objects between 1 to 30 cm in diameter. The limiting factors are the maximum penetrable material thickness of 170 mm aluminum or 45 mm steel and secondly the maximum available number of pixels in width of 1000 as usual, or 2000 when doubling the field of view.

2. Applications

2.1. Analysis of the internal damage of concrete caused by frost attack

The aim of this investigation was to test the detectability of frost-induced cracks and to obtain the spatial distribution of moisture inside of concrete disk specimens. For the given standard size of the specimen, about 100 mm in diameter, the best possible resolution was required. Based on the outer diameter a minimal resolution of 65 ľm can be achieved due to the experimental setup, which results in a projection image of 1650 pixels in width, thus exceeding the pixel format of the detector. Due to his design the 3D-CT scanner is much easier to handle then an image intensifier. Thus an easy way of enlarging the field-of-view (for comparison see (5)) is provided by shifting the detector laterally. Projection images of up to 2000 pixels in width are practicable.

Fig 2: Concrete disk specimen Fig 3: CT of the disk specimen after wet freezing and thawing cycles: Left dry sample, middle moistened sample, right moisture distribution (difference tomogram)

As an example here Fig. 2 shows tomograms of a concrete disk sample that was damaged internally as a consequence of exposure to several wet freezing and thawing cycles. The sample was first investigated in the dry state (Fig. 3 left). After moisturized the same probe was measured again (Fig. 3 centre). Finally, both tomograms were brought into alignment and a third tomogram was calculated which shows the difference between both tomograms. This tomogram (Fig. 3 right) shows only the 3D water distribution as positive densities. Cracks with no water were not found. Beside some enrichments of the water content in cracks the water distribution is homogeneous out side the stones. The broken stone at the right side of the sample shows the same water content as the surrounding area (~10%) and is thus not visible in the difference image.

2.2. Characterisation of electron-beam-weld

Fig 4: Electron-beam-weld probe cut from a thick walled copper canister Fig 5: Overview of all defects in a volume of 203 mm3 detected in the EB-weld, shown in two perpendicular directions

Fig. 4 shows a probe with an electron-beam-weld, which was cut (perpendicular) from a canister with a thick wall. The penetrated thickness of copper was about 30 mm. The measurement was done with 320 kV, 0.1 mA and a pre-filter of 1.5 mm tin with the minimal possible voxel size of 35 ľm. Fig. 5 shows all detected voids within the investigated volume from two orthogonal directions. To obtain these images a ‘Maximum-Minimum’ operation was performed over all pixels along two directions. In the right image a large area of lack of fusion is found. Voids with a volume of 703 ľm3 were verified.

2.3. Characterisation of oxidative degradation of CC-SiC composites

The ceramic composite material C/C-SiC is used as heat shield for temperatures of up to 2700 °C. The infiltration of silicon into carbon matrix shall protect the C/C tissue, which is already oxidised in the range of 600 - 900 °C, otherwise mass loss and a reduction in stability will occur. Figure 6 shows as example a horizontal section of the tomogram from one out of five rods. The experimental series varied the oxidation time at 650 °C. Figure 6 shows a tomogram before, left top, and after eight hours, left bottom, of thermal treatment. In the centre the image of the difference is shown, the higher the mass reduction, the bright the grey level. The graph at the right side illustrates the mass reduction depending on oxidation time, evaluated from the CT images.

Fig 6: Determination of oxidative degradation of CC-SiC composites

2.4. Characterisation of aluminium castings

The 3D-CT scanner is capable to detect some tens of X-photons if 17 cm of aluminum is penetrated at the maximum beam energy of 320 kV. If an homogenous specimen showing widely varying material thicknesses up to detection limit is scanned, as in the case of an engine cylinder head, the resultant CT image shows severe reconstruction defects in the density distribution. Especially the indication of material in the space free of material degrades the image quality. This phenomena disallows the further evaluation of a degraded 3D tomogram for the purpose of surface rendering (Fig. 7 left).

Fig 7: Improvement of 3D-CT by the corrective preprocessing of projection images. Left: Original state, lower density within the sample, indication of material in the space free of material, surface rendering not feasible. Right: CT image after corrective preprocessing, reduced density deviation within the sample, well defined free space, surface rendering now feasible.

Several causes are responsible for these phenomena: A projection image of a 2D detector shows a long-range blurring. In this case each pixel is influenced from all pixels within a radius of 100 pixels. After penetration of 17 cm of aluminium a linear attenuation factor of 300 is expected, but the measurement gives only a factor of 13. This effect is called simple as the “back glow” of the sample. In addition to this diffusion of light in the phosphor screen the specimen causes heavy scattering and hardening of the X-beam.

This problem can be solved, when the physical processes up to the data acquisition are mathematically described and corresponding algorithms for the compensation of the negative influences are introduced. These algorithms have to be applied to each projection image prior to the CT reconstruction. A digitized projection image blurred by light scattering can be restored by deconvolution with the inverse point spread function (PSF) of the detector system. The PSF depends on the beam energy. Unfortunately the experimental determination of the proper PSF is linked to great efforts. In order to bypass this problem, mathematically formulated models replace the experimental PSF. In summary this process requires first a PSF modelling tool capable of generating suitable synthetic PSF, secondly a deconvolution algorithm capable of optimally recovering attenuation data form a degraded digital radiograph. The attenuation of a polyenergetic X-radiation during the penetration of an object occurs depending on the beam energy differently, so that the spectrum of the primary radiation is shifted to higher energy. This effect called beam hardening can be eliminated to a great extent by means of a correction function. If detector and source characteristics are know, this can be calculated but in most cases some probes of different thickness will be measured to produce a correction function. The both last phenomena are apparently counteracting to each other; therefore the artefacts are disappearing under specific conditions.

In a tomographic set up the distance between object and detector is large enough to approach the detected scattered radiation as a signal distributed uniformly on the detector. After this additional offset to zero is determined it can be subtracted from each projection image in a suitable way. As the pixel values can be vary close to noise level a dynamic noise reduction should be applied during preprocessing of the projection image in order to avoid zero crossings.

Figure 7 at right shows the resultant CT image after all corrections have been applied to a tomographic data set which would otherwise lead to a CT image as on the left side of Figure 7. The lines drawn in on the CT images are illustrated as grey level

profiles in the graphs below the images. Due to the corrective preprocessing in the space free of material density values are indicated that are less than 10 percent of the maximum density. The now measurable wall thicknesses were compared with measurements at the real object. A difference of 1 to 2 pixels mostly underestimated was found.

3. Enhancement of projection images

Since the beginning of the 3D-CT in the 1990es area detectors have been used as image sensors. When a correction algorithm is applied to projection images to minimize the geometrical pillow distortion of an image intensifier device the reconstructed CT image displays the specimen of interest in proper geometry, but the image is disturbed by severe density errors. Nevertheless inclusions, voids and cracks are still visible. Since the introduction of large size area detectors on the basis of amorphous silicon and through the wide spreading use in medicine, these detectors are also employed in industrial radiography. Some of the commercial available flat panel detectors provide high SNR and high dynamic range of 16-bit. Both features are key requirements when designing a CT scanner. But in practice some additional measures have to be taken. The corrective preprocessing of the projection images is presented in chapter 2.4. Furthermore, the minimization of scattered radiation from parts placed in the field of X-rays, as well as an additional radiation shield around the x-ray source should be considered. Finally the detector system should be temperature stabilised at +/- 0.5 °K, a lowering of temperature is useful.

4. Summarise and outlook

The described CT system provides an accurate and fast 3D tomographic analyse. The scanning time is 45 min in average, the minimal spatial resolution is 35 ľm and projection images up to 2000 pixels in width may be acquired. A further improvement in the object size can be achieved by using more efficient X-ray converters and higher X-ray energies.
These investigations were partially supported by the DFG in Germany.

References

  1. Riesemeier, H., Goebbels, J., Illerhaus, B., Onel, Y., Reimers, P., “ 3-D Mikrocomputertomograph für die Werkstoffentwicklung und Bauteilprüfung”, DGZfP Berichtsband 37 (1993) p. 280
  2. Illerhaus, B., Goebbels, J., Reimers, P., “The Principle of computerised Tomography and its Application in the Reconstruction of Hidden Surfaces of Art”, 4 th Int. Conf. NDT of Works of Art, DGZfP Berichtsband 45,1 (1994) p. 41
  3. www.yxlon.de
  4. www.perkinelmer.com
    (optoelectronics.perkinelmer.com / Downloads / rid1640.pdf)
  5. Illerhaus, B., Goebbels, J., Riesemeier, H., Staiger, H., “Correction techniques for detector systems in 3D-CT” . Proceedings of SPIE Vol. 3152, (1997)

© NDT.net |Top|