No abstract

X-ray computed tomography (CT) imaging for industrial applications is limited to certain physical conditions to be fulfilled. The size of the measuring object and the accumulated wall thickness are two fundamental conditions. An omission of these conditions by not capturing object attenuation information by the x-ray detector leads to missing data in the 3D reconstruction process and results as a consequence in image degradation and artifacts. Conventional industrial x-ray CT is based on cone-beam projections and circular or helical scanning trajectories using linear axis and a rotary (lift) table. For many inspection tasks on big-sized or unusually shaped objects the physical limits for obtaining a sufficient high image quality are reached very quickly when using conventional CT systems. Industrial six-axis robots offer much more flexibility with respect to the conditions mentioned earlier and can overcome the limitations of conventional scanners. In the present work we characterized an industrial six-axis robot in its working space following ISO 9283 in terms of pose accuracy and pose repeatability. These results are then used to simulate faulty scanning trajectories in terms of pose deviations where a single robot is used as an object manipulator to rotate virtual specimens on a circular trajectory resulting in different (faulty) reconstruction datasets. These datasets are evaluated visually and by using performance parameters and geometrical features in order to determine the reproduction fidelity (performance) of a one arm robot-based CT system depending on different pose errors. With the results obtained it was shown that a robot-based CT system of type B (in our classification scheme) using one robot as object manipulator should be able to reach a spatial resolution power in the range of the voxel size (in our case 200 µm) and smaller (neglecting effects from focal spot size, detector unsharpness from x-ray to light conversation and scatter radiation) if systematic pose errors are compensated using appropriate calibration methods.

In this research work we investigated the accuracy of a standard industrial robot. We wanted to find out, how accurate an X-Ray Computed Tomography (CT) scan can be performed when using such a robot as a manipulator. The accuracy was measured using a laser-interferometer. The measured deviations were used to run an X-Ray simulation via Fraunhofer EZRT’s Scorpius X-Lab. Metrological analysis was performed as a measure for the quality of the simulated CT-scan. The metrological deviations reflect the feasible accuracy of a CT-scan in a real CT-setup.

The market for extruded profiles for different applications (automotive, construction sector) is a big part of the plastics processing industry. Currently, the whole production process runs almost without quality checks. Only random tests are done to check the profile for its geometry, dimensions and defects. For extruded pipes and other rotation-symmetric products, there are special inline-inspection systems. However, none of these systems is capable of checking the whole inner and outer geometry of an arbitrarily formed profile. Due to this, we evaluate a concept for a multi-source-CT system, which is able to inspect the whole geometry of extruded profiles during the production process. In this work, we evaluate a concept and build a prototype system to validate the functional capability. Considerable attention is paid to the use of commercially available components because of the cost-effectiveness of the whole system. On the whole, the system comprises 16 x-ray-sources and two polygon-shaped detectors. The components are separated into two rings, each with eight x-ray sources and one detector. The rings are arranged around the extrusion line so that extruded profile traverses through the rings during the extrusion. The whole system is semi-static, which means the components are rigid relative to each other, but the system makes an eccentric movement perpendicular to the direction of extrusion. Due to this, the number of projection angles increases from 16 to about 420 projection angles, which is sufficient for reconstruction. To further increase the sampling rate, three x-ray-sources are pulsed simultaneously. Therefore, we are able to do a full 3D scan of extruded profiles with an extrusion feed of up 100 mm/sec.