The Largest Open Access Portal of Nondestructive Testing (NDT)

Within this work, optimisation advances of an at-line X-ray computed tomography (XCT) inspection system at Nemak Linz are presented. The ZEISS VoluMax 1500 G2 is operated together with an ABB loading robot for automated inspection of serial parts as wells as prototypes of aluminium casted cylinder heads produced for the automotive industry. Typical scanning times for these high absorbing specimens are ranging from 40 to 400 seconds. The defect analysis software provided by the manufacturer is used to automatically identify and classify defects of serial parts and decide between okay and not okay parts. The analysis software consists of anomaly detection, defect classification and a segmentation based on deep learning that requires an initial teaching phase. The detection seems robust, but the artificial manufacturing of parts with critical defect types and sizes at specific locations and their experimental verification is challenging and very time consuming. This work presents an exemplary verification for one selected specimen and for one particular defect type with different sizes. Besides real also simulated data with virtual defects are evaluated using the anomaly detection software. Advantages of numerical simulation in combination with the presented anomaly detection are discussed in detail.

Multiscale X-ray computed tomography (XCT) usually includes scanning an entire part at lower resolution. Subsequently, a smaller sample is cut out of the respective specimen and scanned at a higher resolution. Accordingly, in this work typical XCT resolutions ranging from (135 µm)³ voxel size down to (250 nm)³ are presented. Using different XCT-modes such as region of interest (ROI) scans or laminography (XCL) modes this multiscale approach is also possible without cutting the sample in smaller pieces. We show that cracks with a width between 122 and 56 µm can be seen at a relatively low resolution of (135 µm)³ on one example of a carbon fibre reinforced polymer (CFRP) sample. With XCL voxel sizes down to (0.75 µm)³ can be reached, showing clear structures in the range of 16 µm. Main disadvantage of XCL is that only a view layers and no full 3D-microstructure can be represent well. Using a XCT resolution in the range of (2 µm)³ voxel size for CFRPs may lead to miss interpretation in regards of porosity because of propagation based phase contrast effects. High-resolution ROI-XCT scans at (250 nm)³ shows, that epoxy rich areas between individual C-fibres smaller than 6 µm are leading to this dark grey values, easily misinterpreted as voids. Multimodal XCT data was generated using a Talbot-Lau Grating Interferometer (TLGI) XCT to obtain modalities such as dark-field contrast and differential phase contrast in addition to standard attenuation contrast. In one example it is shown that metal-artefacts in CFRP issued by a Cu-mesh can be significantly reduced by TLGI-XCT. This provides improved image quality and the possibility to segment voids close to metallic components. For easier interpretation and a better understanding of material features the open source software open_iA was used with new implemented visualisation approaches for multimodal and multiscale data-visualisation.

Short glass fibre reinforced polymers (SGFR) are increasingly used in automotive industry for the replacement of metals since they provide a significant reduction of weight. Parts with anisotropic fibre orientation were produced during the injection moulding process. Local microstructure parameters such as fibre length and orientation are decisive for the damage behaviour of the injection moulded parts. Damage type and propagation investigation are necessary for better understanding of material behaviour under stress. X-ray computed tomography (XCT) is a suitable non-destructive method to characterise internal defects in SGFR specimens. With an interrupted in-situ XCT tensile test three dimensional damage propagation in SGFR can be observed. In this study an interrupted in-situ computed tomography technique was used to quantify the main damage mechanism in polyamide specimens with a glass fibre content (PAGF) of 15 and 30 wt%. Double notched adapted miniature tensile specimens were cut from a plate in two orientations, 0° and 90° relatively to the melt flow. Therefore the tensile force was applied parallel and perpendicular to the expected main fibre orientation (0° respectively 90°). After every force step, a XCT scan with a voxel size of (2 µm)^3 was performed to detect the internal defects. Four different defect types were classified: matrix fractures, fibre pull-outs, fibre/matrix debondings and fibre fractures. Fibre pull-out was the dominating defect type. Defects were mainly induced in the shear region of the 0° specimens. The core region showed nearly no defects at the last load step before breakage. Summing up the results show that visualisation and quantification of the defect volume and mechanism at certain load steps was possible with this method. Clear differences between damage induction of the 0° and 90° specimens were observed.

The sufficiency of bone fixation via ceramic or metallic implants largely depends on the osseointegration at the bone-implant interface and is of major importance in trauma, orthopedic, and maxillofacial surgery. Preclinical imaging via microcomputed tomography (XCT) is the state of the art solution to study the ingrowth of mineralized tissue at the bone-implant interface and to investigate the influence of innovative materials on the regeneration of bone tissue. However, a major challenge in XCT imaging is impaired image quality due to metallic objects such as implants and orthopedic screws as they can cause severe image artefacts characterized by bright and dark streaks due to scattering and beam hardening effects. These artefacts can severely reduce image quality, impeding the quantification of trabecular structures and therefore reducing the diagnostic value of the data. In this paper we present multi-modal image data acquired using Talbot-Lau grating interferometer (TLGI) XCT visualizing the interface between surgical screws and artificial trabecular bone. Our findings reveal that differential phase contrast (DPC) is less prone to metal artefacts improving the visualization of microstructure in bone adjacent to metal components. For this purpose, we produced polymeric foam structures imitating the microstructure of human bone to systematically investigate the advantages of DPC imaging for multi-component applications. We created a realistic set-up of four synthetic foam cylinders produced with different filler materials representing artificial bone. Two to four titanium screws (thread diameter 1.2 mm, length 16 mm) were applied into the artificial bone parallel to each other. The specimens were investigated using a TLGI XCT system (SkyScan 1294) to extract information in relation to absorption contrast (AC), DPC, and dark-field contrast (DFC). In AC, metal streaking artefacts are severe, covering microstructure between the screws completely. DPC images in return show almost no streaking artefacts, which enhances image segmentation and volume rendering.

Faserverstärkte Kunststoffe sind aus vielen industriellen Bereichen, wie beispielsweise der Fahrzeugindustrie, Verpackungsindustrie, Bauwirtschaft oder auch der Freizeitindustrie, nicht mehr wegzudenken. Die steigenden Anforderungen an die Materialien, hinsichtlich mechanischer und physikalischer Eigenschaften, haben es notwendig gemacht sich mit der Mikrostruktur genauer zu beschäftigen. Dazu sind zerstörungsfreie Verfahren wie Röntgen-Computertomografie (CT) von großem Vorteil. Mittels CT können eigenschaftsbestimmende Faktoren, wie z.B. Faserorientierung und Faserlängenverteilung, ermittelt werden. Die Kenntnis solcher Faktoren ermöglicht es einerseits, die Herstellungsparameter zu optimieren und andererseits Modelsimulationen für die Vorhersage mechanischer Eigenschaften zu generieren. In dieser Arbeit werden verschiedene spritzgegossene faserverstärkte Kunststoffe mittels Computertomografie untersucht und in weiterer Folge die Faserorientierung, sowie die Faserlängenverteilung, mit der hauseigenen Fasercharakterisierungssoftware analysiert. Es werden Kunststoffe aus Polypropylen verstärkt mit Glasfaser, Kohlenstofffaser, Polymerfaser oder Zellulosefaser untersucht. Unterschiede in der Dichte bzw. dem Röntgenabsorptionskoeffizienten führen zu unterschiedlich großem Kontrast zwischen Fasern und Matrix. Der Kontrast zwischen den Polymerfasern und der Kunststoffmatrix ist beispielsweise eher schwach, was wiederum Probleme bei der Bestimmung der Faserlängenverteilung bereiten kann. Weiters haben auch Fasergehalt, Faserdurchmesser und Fasergeometrie einen Einfluss auf die Datenqualität. Die Röntgen-Computertomografie-Scans werden bei einer Voxelgröße von bis zu (1 µm)^3 durchgeführt. Die erforderlichen Messzeiten, um eine ausreichende Datenqualität für eine softwarebasierte Faseranalyse zu erreichen, liegen dabei im Bereich von drei bis zwölf Stunden. Anhand der dreidimensionalen Datensätze lassen sich Füllstoffeigenschaften, aber auch Inhomogenitäten der Matrix, bestimmen. Im Rahmen dieser Arbeit werden neben Schnittbildern und Ergebnissen der Faseranalysen, ein Vergleich der verschiedenen Faserkunststoffverbundsysteme präsentiert, sowie die Vorteile und Limitierungen von Computertomografie für die Fasercharakterisierung dieser Materialien diskutiert.