The Largest Open Access Portal of Nondestructive Testing (NDT)

The comparison of many 3D volumes to find subtle differences is tedious, time-consuming and error-prone. Previously we presented Dynamic Volume Lines [1], a novel tool for the interactive visual analysis and comparison of ensembles of 3D volumes, which are linearized using Hilbert spacing-filling curve and represented as 1D line plots. In this paper we further demonstrate the usefulness and capabilities of our method by conducting a detailed visual analysis and evaluation of an artificial specimen from simulated 3D X-Ray Computed Tomography (XCT) and a real-world XCT titanium alloy specimen.

CT data of industrially produced cast metal parts are often afflicted with artefacts due to complex geometries ill-suited for the scanning process. Simple global threshold-based porosity detection algorithms usually fail to deliver meaningful results. Other adaptive methods can handle image artefacts, but require long preprocessing times. This makes an efficient analysis workflow infeasible. We propose an alternative approach for analyzing and visualizing volume defects in a fully interactive manner, where analyzing volumes becomes more of an interactive exploration instead of time-consuming parameter guessing interrupted by long processing times. Our system is based on a highly efficient GPU implementation of a segmentation algorithm for porosity detection. The runtime is on the order of seconds for a full volume and parametrization is kept simple due to a single threshold parameter. A fully interactive user interface comprised of multiple linked views allows to quickly identify defects of interest, while filtering out artefacts even in noisy areas.

This work introduces methods for analyzing the three imaging modalities delivered by Talbot-Lau grating interferometry X-ray computed tomography (TLGI-XCT). The first problem we address is providing a quick way to show a fusion of all three modalities. For this purpose the tri-modal transfer function widget is introduced. The widget controls a mixing function that uses the output of the transfer functions of all three modalities, allowing the user to create one customized fused image. A second problem prevalent in processing TLGI-XCT data is a lack of tools for analyzing the segmentation process of such multimodal data. We address this by providing methods for computing three types of uncertainty: From probabilistic segmentation algorithms, from the voxel neighborhoods as well as from a collection of results. We furthermore introduce a linked views interface to explore this data. The techniques are evaluated on a TLGI-XCT scan of a carbon-fiber reinforced dataset with impact damage. We show that the transfer function widget accelerates and facilitates the exploration of this dataset, while the uncertainty analysis methods give insights into how to tweak and improve segmentation algorithms for more suitable results.

Industrial 3D X-ray computed tomography (XCT) has been adopted in the industries as a high-resolution non-destructive testing method. XCT enables resolving individual fibers, which is the basis for calculating the fiber orientation of each individual fiber and fiber orientation tensors, respectively. In this work, we present an interactive visualization tool for comparing the orientation tensor information between simulation and real-world XCT data or between two real-world XCT data. Based on the orientation tensor data we calculate three different similarity measures (degree of orientation, cosine similarity and tensor similarity) to characterize the data. The calculated measures are presented with heat maps to show the fiber orientation correlations between simulation and real-world XCT data. In addition, we use superquadric tensor glyphs to visualize and compare the fiber orientations of the simulation data and the XCT data. This type of glyphs, introduced by Kindlmann et. al., convey tensor variables by mapping the tensor eigenvectors and eigenvalues to the orientation and shape of a geometric primitive. The superquadric tensor glyphs and the heat maps of the similarity measures are shown layer by layer and can be overlaid onto each other. We demonstrate our tool by using an injection molded short glass fiber-reinforced polymer specimen.

To develop and optimize of advanced composite materials such as glass fiber reinforced polymers (GFRPs) for a specific application area is an important topic. To inspect mechanical properties of GFRPs, material engineers use interrupted in situ tensile tests. During these tests, a test specimen is scanned multiple times in an industrial computed tomography (CT) scanner under various loads, starting from no load until the final fracture of the specimen. In this work we focus on the final step of the interrupted in situ tensile test, which is scanned when the specimen is completely losing its structural integrity in the final fracture zone. The defects occurring in the subsequent loading stages merge and ultimately form the final fracture. For this reason, conventional techniques tend to generate error prone final fracture regions or surfaces and thus require more advanced algorithms for extraction. The main contribution of this paper is found in the comparison of different techniques for extracting the final fracture. In the comparison we outline advantages and drawbacks of the presented techniques relative to each other.

Multi-material components (MMCs) containing metal and plastic belong to one of the most commonly produced category of industrial objects. Industrial 3D X-ray computed tomography (3DXCT) is particularly suited for the examination of MMCs as it allows capturing external and internal interfaces of each individual material of the fully assembled part. 3DXCT enables further material analysis, e.g., exploration of material properties such as porosity and detection of internal defects. The presence of highly absorbing metal within low absorbing plastics strongly impacts the 3DXCT analysis as it causes severe artifacts. Dark-band or streaking artifacts may distort neighboring areas with lower-density materials in the reconstructed volume and may hinder dimensional measurement tasks. Furthermore, it is highly challenging to reduce streaking and dark-band artifacts using the reconstructed volume without having any additional knowledge about the specimens materials and its geometry. In t.....

Dieses Paper präsentiert eine neuartige Methode für Multimaterialbauteile, um mit Hilfe von Dual Energy CT reproduzierbare Oberflächenmodelle zu extrahieren. Anwendungsszenario ist hierbei die dimensionelle Messtechnik in der hochauflösenden 3D Röntgencomputertomographie (3DCT). Die präsentierte Methode zielt darauf ab, die Vorteile von zwei unterschiedlichen CT Messungen eines Bauteils zu verbinden. Einerseits wird eine hochauflösende MikroCT Messung durchgeführt, um Details des Bauteils mit hoher Präzision zu erfassen. Andererseits wird eine Hoch-Energie MakroCT Messung verwendet um die Struktur des Bauteils möglichst artefaktfrei aufzunehmen. Wir stellen in dieser Arbeit eine Prozesskette vor, die Algorithmen für die Bildfusion und die lokale Oberflächenextraktion verbindet: Im ersten Schritt wird das in den Daten inhärente Rauschen minimiert. Nach einer Registrierung der einzelnen Datensätze werden die Daten in einem speziellen Datenfusionsschritt kombiniert. Die Struktur wird dabei vom Hoch-Energie Datensatz und präzise Details aus der MikroCT Messung übernommen. Im finalen Schritt wird ein verbessertes Oberflächenmodell des Bauteils mit einer lokal adaptiven Methode generiert.

In der präsentierten Arbeit wird das Ziel verfolgt, mittels Dual Viewing optimale Röntgencomputertomographiedaten für die weitere Verarbeitung zu erzeugen. Das Hauptaugenmerk liegt dabei auf der Reduktion von durchstrahlungsrichtungsabhängigen Artefakten im resultierenden Datensatz. Das vorgestellte Verfahren wird in drei Schritte unterteilt: Der erste Schritt besteht darin, das Messobjekt zweimal mit unterschiedlicher Orientierung im Röntgenstrahl zu tomographieren. Die unterschiedliche Ausrichtung bewirkt eine unterschiedliche Ausprägung der direktionalen Artefakte im Datensatz. Im zweiten Schritt werden die Datensätze mittels einer schnellen Grobregistrierung und einer präzisen Feinregistrierung aufeinander ausgerichtet. Nach der Registrierung werden die beiden Datensätze fusioniert. Dabei werden unter Annahme normalverteilter Wahrscheinlichkeitsdichtefunktionen der einzelnen Materialien die Posteriorwahrscheinlichkeiten für jedes Voxel der beiden Datensätze mittels des Bayes Theorems berechnet. Mit Hilfe der Posteriorwahrscheinlichkeiten kann im Schritt der Datenfusion auf das wahrscheinlichste Material in jedem Voxel geschlossen werden. Somit wird ein Großteil der Artefakte entfernt und die Stärken jedes Datensatzes berücksichtigt. Die Evaluierung der präsentierten Methode wird an ausgewählten Objekten mittels verschiedener Visualisierungsmethoden demonstriert.

Variance comparison is a common means of quality assurance in modern engineering. It is used to compare the measured geometry of a specimen with reference geometry data. For providing the geometry information especially of internal features, 3D computed tomography (3D-CT) is increasingly used. The drawback of this method: Measurement results of typical 3D-CT systems with cone beam x-ray sources are affected by material, geometry and position of the specimen in the ray. To create a surface model of a 3D-CT dataset for variance comparison usually a single threshold is specified to distinguish between air and material, and a surface model is extracted along the selected greyvalue. But this approach deforms the created surface in the area of artifacts and even holes in the surface may emerge. Furthermore this method is a "trial & error" method and therefore it is not objective. This paper describes a robust method for creating reproducible and exact object surfaces of distorted volume datasets for variance comparison in 3D-CT. We propose a pipeline model, using common 3D image processing filters to create the surface model. In particular, our pipeline uses gradient information for detecting the object contour: A prefiltering step with an edge preserving diffusion filter reduces noise without blurring the edges in the dataset. Instead of specifying a single threshold and creating a surface model, a gradient image is extracted for further data processing. To extract a fully connected binary volume, a watershed segmentation filter is applied to the gradient image. Finally the surface is constructed, using elastic surface nets, which smooth aliasing effects of the binary volume dataset. We conclude that it is possible to extract reproducible surface models of distorted volumes with moderate computational effort. Our pipeline is quite robust concerning most common artifact types like noise-induced streaks, aliasing, partial volume and scattered radiation effects, and after all the pipeline produces exact results, which is of great importance for variance comparison tasks.