·Table of Contents
·Industrial Plants and Structures
Sensitive Detection of planar Defects by a Mechanised Radiometric Weld Inspection SystemB. Redmer, U. Ewert
Federal Institute for Materials Research and Testing (BAM), Berlin, Germany
A.V. Likhachov, V.V. Pickalov
Institute of Theoretical and Applied Mechanics, Novosibirsk, Russia
Federal Institute for Materials Research and Testing (BAM), Berlin, Germany
Tsinghua University, Bejing, China
Institute of Introscopy, Tomsk, Russia
A highly sensitive radiometric line camera was assembled together with an ultrasound-manipulator (US-manipulator) and an X-ray tube to investigate this technique for improvement of the radiation technique. Circumferential welds were scanned and an improved image quality was obtained for pipes with about 10-20 mm wall thickness in a reasonable measurement time. Significant improvements of the image quality were achieved for testing of water filled pipes. The technique can be extended by multi angle inspection to improve the detectibility and enable depth evaluation of cracks and lack of fusion. Tomosynthesis permits the calculation of three dimensional presentations from multi-angle projections. This improves the evaluation of indications, like cracks, lack of fusion and undercuts.
Fig 1: View of a) the complete line scanner and b) of the working principle of the unit.|
|Fig 2: Comparison of a digitised radiograph (upper image) and a line camera image (lower image) of a water filled pipe weld.|
The applied principle of line scanning yields several advantages in comparison to the film radiography. The radiation direction corresponds to the central projection technique, which always enables radial penetration. All radial flaws are detected with maximum contrast. The special slit collimator of the camera reduces the intensity of the scattered radiation significantly. The well adjusted system yields therefore better image quality than film radiography for pipes with thick walls and especially for water filled pipes (fig. 2). The developed system provides radiometric images with high image quality for a penetrated steel thickness of 4 - 46 mm (2 - 23 mm wall thickness) in a comparable time to manual radiography. Due to the limited resolution of the applied camera of about 160 mm (measured with double wire IQI by EN 462-5) the optimum application range covers wall thicknesses of 10-25 mm from the present point of view. The testing of higher wall thicknesses is also possible but requires the modification of the used line camera.
The description of the line camera design is outlined in [2-4]. After a few modifications the camera could be operated near the physical sensitivity limit. The spectra of different X-ray tubes and g-sources were calculated to simulate the detector properties and the expected sensitivities in dependence on the wall thickness of steel and the measurement time [2, 5].Only X-ray tubes enable the application of the line scan technique in a reasonable time, which is comparable with the manual radiography technique and its image quality. The line camera yields a sensitivity of about 90% (81% DQE - Detective Quantum Efficiency) of the calculated physical limit up to about 35 mm penetrated steel thickness at 225 keV.
The above described technique of measurement enables the depth evaluation of indications by means of the laminography and/or planar tomography. Laminography describes the method and geometry for measurement and reconstruction. It is applied if the complete access to all sides of the object is limited. Due to the limitation of the applicable projection angles the laminography is considered as "limited data"-technique. Tomosynthesis describes the variety of the algorithms for reconstruction from limited number of projections.
Fig. 3 shows the principle of the application of the multi-angle technique. The X-ray tube is shifted in the direction to the pipe axis step by step. For each step a full 360o scan is taken. In contrast to the X-ray tube the line camera is not shifted parallel to the pipe axis. This algorithm corresponds to the coplanar translational laminography  transformed to a cylinder coordinate system for the inspection of pipes. If the X-ray tube is shifted parallel to the pipe axis and additionally, corresponding to fig. 3, each step is combined with an angular shift (pre-scan and post-scan), the principle of coplanar rotational laminography can be applied . Due to this analogy between the configuration of multi-angle technique and laminography the three-dimensional (3D) reconstruction of the weld is possible. Corresponding to fig. 3 either all measured radiometric scans can be measured or (and) the weld can be inspected plane by plane from the reconstructed 3D-data set.
|Fig 3: Principle of laminographic inspection of circumferential welds. Camera and tube are turned around the pipe with a phase shift of 180°. The tube is shifted step by step in direction of the pipe axis for each orbital scan.|
The modified filtered backprojection was found as the most effective reconstruction method regarding the reconstruction time and the image quality. Also good results were achieved with the average - method combined with high / low pass filter  and the symmetry analysis on the basis of a covariance - analysis . As example fig. 4 shows the reconstruction of a crack indication in a austenitic pipe by means of the average - method.
|Fig 4: 3D-Surface image of a crack.|
|Fig 5: Principle of planar tomography. The X-ray tube is shifted along the surface parallel to the pipe axis and several hundred projections are taken for reconstruction.||Fig 6: Reconstructed profile of a circumferential weld of an austenitic pipe after oscillation|
Planar tomography provides undistorted reconstruction results for cracks and lack of fusion if the defect plane is oriented inside the inspection angle range. Artefacts arise for volumetric indications, only due to the missing projection angle problem. The used reconstruction is based on a modified filtered back projection. This algorithm is a typical analysis method and is applied for cross-sectional analysis of preselected regions of interest. The measurement time for one set of projections amounts up to 6 minutes per cross-section depending on the wall thickness (8...20 mm). The reconstruction time is about one minute on a PC. Fig. 6 shows a typical indication of a crack in an austenitic pipe weld.
|Fig 7: Surface indicators on a pipe wall.The cross section and surface profile is calculated along the line.|
The figures 8a and 8b present the reconstructed cross section of the pipe wall in a defined reconstruction space. Triangular artefacts are recognised in fig. 8a above and below the indicator marks, which make it difficult to determine the surface profile. These artefacts are reduced (fig. 8b) by the application of a high pass filter on the basis of a 1-dimensional median filter (equ. 1).
|Imagefiltered = Imagerreconstructed - Imagereconstructed and median filter||(1)|
The wall thickness, respectively the depth of an indication (e.g. crack) relatively to the surface profile, can be determined qualitatively and quantitatively by the measurement of the distances between the inner and outer surface.
|Fig 8: Results of the reconstructed cross section area before (a) and after (b) high pass median filtering.|
Tomosynthesis with special image pre-processing enables to enhance the detectibility of cracks and to extract the information about their shapes, structure and depth. Special statistical approaches for the enhancement of accuracy are investigated now. The consideration of these possibilities is beyond the limits of this paper. Furthermore, there exist the possibility to enhance the tomosynthesis by consideration of a priory knowledge in a post iterative process. Interesting first results have been obtained by .
|© AIPnD , created by NDT.net|||Home| |Top||