Fast Estimation of Optimal Specimen Placements in 3D X-ray Computed Tomography

摘要

3D X-ray computed tomography (3DXCT) is increasingly used in industry as a method for quality control and nondestructive testing. More recently a further demanding application area was found in metrology. All these application areas share the need, that the highest possible accuracy and precision is required from every scanning result. In this context, the issue of errors and distortions introduced by artefacts is critical. Picking the optimal placement of a specimen on the rotary plate leads to a reduction of artefacts and an improvement of the overall quality in the resulting dataset. However, finding optimal and stable placements of complex specimens is tedious, time-consuming and therefore expensive. In this work a tool for 3DXCT systems was developed, which estimates the optimal placement of a specimen using its 3D geometrical model prior to a real scan. This geometrical model is usually available either as a CAD model or obtained from a reference scan of a different modality. The proposed method allows the determination of potentially good or bad placements of the specimen on the rotary plate as well as the identification of regions of the specimen, where most of the artefacts are likely to appear. A specimen's placement is defined by its orientation on the rotary plate. Besides the penetration lengths of the X-rays through the specimen also the placement stability and the corresponding Radon space representation are considered in the analysis. The GPU-based ray casting is used to simulate the scanning procedure and to calculate the penetration lengths of the rays. The Radon space analysis facilitates the identification of critical faces, which will be inaccurately represented in the XCT reconstruction data. In order to estimate the amount of data lost in Radon space every triangle of the 3D geometrical model is investigated. Additionally, a feature-selection functionality is provided, in order to constrain the analysis on critical features or areas of interest. The results are visually represented in 3D views, in order to depict areas, which are estimated to suffer the most from artefacts. Additionally results are visually presented by linked views, allowing visual analysis, comparison and exploration. A stability widget depicts the robustness of the placement with respect to parameter variations. The results of applying the tool on a complex real world component are demonstrated in detail. The calculated optimal placement is tested versus the initial placement of the specimen. All evaluations are performed using commercially available software tools. In order to verify the optimality of the found placement, initial and optimal placements are tested using variance comparisons.