Assessing the effect of sample orientation on dimensional X-ray computed tomography through experimental and simulated data

This paper evaluates the dimensional accuracy of industrial X-ray computed tomography (CT) measurements in function of the sample's orientation within the measurement volume. Experimental measurements are contrasted against predictions from simulations that, based on Radon-space analysis and computation of X-ray traversal-path lengths through ray-casting, estimate the optimal orientation of samples that would minimize both cone-beam distortions and beam hardening artifacts in CT data reconstructions. The simulated data were obtained with the Dreamcaster, a tool for processing and visual analysis of industrial CT datasets. For the experimental data, the tilt angles from the CT rotation axis that resulted in the smallest deviations between dimensional CT data and reference measurements-obtained from contact coordinate measuring machines (CMMs)-were determined for each sample. Tilt angles that resulted in large dimensional measurement errors were also determined. In most of the cases, the data observed in the experimental measurements agree with the trends predicted by simulated data. For slender workpieces (with high aspect ratios), three basic zones of axial angular tilt were identified: a 'green zone' from 10 to 35 degrees in which deviations between CT and CMM are minimum, a 'red zone' from 0 to 5 degrees and from 65 to 90 degrees in which the largest deviations are measured, and 'yellow zones' between the two. When optimal orientations from the 'green zone' are used for scanning, the dimensional deviations of CT data from reference measurements are under 10 mu m.