Analytische Chemie
Filtern
Erscheinungsjahr
- 2023 (16) (entfernen)
Dokumenttyp
- Zeitschriftenartikel (9)
- Vortrag (4)
- Beitrag zu einem Tagungsband (2)
- Posterpräsentation (1)
Schlagworte
- Additive manufacturing (3)
- Residual stress (3)
- Automatic segmentation (2)
- Computed tomography (2)
- Dice score (2)
- Large Scale Facilities (2)
- Multi-phase materials (2)
- Neutron Diffraction (2)
- Röntgen-Refraktion (2)
- Synthetic Training Data (2)
Organisationseinheit der BAM
- 8 Zerstörungsfreie Prüfung (16)
- 8.5 Röntgenbildgebung (16)
- 5 Werkstofftechnik (4)
- 5.1 Mikrostruktur Design und Degradation (2)
- 9 Komponentensicherheit (2)
- 9.6 Additive Fertigung metallischer Komponenten (2)
- 5.3 Polymere Verbundwerkstoffe (1)
- 5.4 Multimateriale Fertigungsprozesse (1)
- 6 Materialchemie (1)
- 6.1 Oberflächen- und Dünnschichtanalyse (1)
Paper des Monats
- ja (1)
Eingeladener Vortrag
- nein (4)
The greatest challenge when using deep convolutional neural networks (DCNNs) for automatic segmentation of microstructural X-ray computed tomography (XCT) data is the acquisition of sufficient and relevant data to train the working network. Traditionally, these have been attained by manually annotating a few slices for 2D DCNNs. However, complex multiphase microstructures would presumably be better segmented with 3D networks. However, manual segmentation labeling for 3D problems is prohibitive. In this work, we introduce a method for generating synthetic XCT data for a challenging six-phase Al–Si alloy composite reinforced with ceramic fibers and particles. Moreover, we propose certain data augmentations (brightness, contrast, noise, and blur), a special in-house designed deep convolutional neural network (Triple UNet), and a multi-view forwarding strategy to promote generalized learning from synthetic data and therefore achieve successful segmentations. We obtain an overall Dice score of 0.77. Lastly, we prove the detrimental effects of artifacts in the XCT data on achieving accurate segmentations when synthetic data are employed for training the DCNNs. The methods presented in this work are applicable to other materials and imaging techniques as well. Successful segmentation coupled with neural networks trained with synthetic data will accelerate scientific output.
The overview of the activity of Federal Institute for Material Research and Testing (BAM, Belin, Germany) in the field of additively manufacturing material characterization will be presented. The research of our group is focused on the 3D imaging of AM materials by means of X-ray Computed Tomography at the lab and at synchrotron, and the residual stress characterization by diffraction (nondestructive technique). Also, two successful research project in collaboration with CAM2, Sweden are presented.
Regardless of the experimental care practiced in acquiring X-ray computed tomography (XCT) data, artifacts might still exist, such as noise and blur. This is typical for fast XCT data acquisitions (e.g., in-situ investigations), or low-dose XCT. Such artifacts can complicate subsequent analysis of the data. Digital filters can moderately cure extensive artifacts. The selection of filter type, intensity, and order of application is not always straight forward. To tackle these problems, a complete sequential multilevel, multi-scale framework: BAM SynthCOND, employing newly designed deep convolutional neural networks (DCNNs), was formulated. Although data conditioning with neural networks is not uncommon, the main complication is that completely artifact-free XCT data for training do not exist. Thus, training data were acquired from an in-house developed library (BAM SynthMAT) capable of generating synthetic XCT material microstructures. Some novel DCNN architectures were introduced (2D/3D ACEnet_Denoise, 2D/3D ACEnet_Deblur) along with the concept of Assertive Contrast Enhancement (ACE) training, which boosts the performance of neural networks trained with continuous loss functions. The proposed methodology accomplished very good generalization from low resemblance synthetic training data. Indeed, denoising, sharpening (deblurring), and even ring artifact removal performance were achieved on experimental post-CT scans of challenging multiphase Al-Si Metal Matrix Composite (MMC) microstructures. The conditioning efficiencies were: 92% for combined denoising/sharpening, 99% for standalone denoising, and 95% for standalone sharpening. The results proved to be independent of the artifact intensity. We believe that the novel concepts and methodology developed in this work can be directly applied on the CT projections prior to reconstruction, or easily be extended to other imaging techniques such as: Microscopy, Neutron Tomography, Ultrasonics, etc.
Die Brechung von Röntgenstrahlen (Röntgenrefraktion) an Grenzflächen zwischen Materialien unterschiedlicher Dichte ist analog zur Ablenk-ung von sichtbarem Licht an z.B. Glasoberflächen. Es gibt jedoch zwei wesentliche Unterschiede:
a) konvexe Grenzflächen verursachen Divergenz (d.h. der Brechungsindex n ist kleiner als 1), und
b) die Ablenkungswinkel sind sehr klein, und reichen von einigen Bogensekunden bis zu einigen Bogenminuten (d.h. n ist nahe bei 1);
Wie auch bei sichtbarem Licht ist die Ablenkungsrichtung der Röntgenstrahlen abhängig von der Orientierung der durchstrahlten Grenzfläche. Aufgrund dieser Eigenschaften eignen sich Röntgenrefraktionsmethoden hervorragend für:
a) die Erkennung und Quantifizierung von Defekten wie Poren und Mikrorissen und
b) die Bewertung von Porosität und Partikeleigenschaften wie Orientierung, Größe und räumliche Verteilung.
Wir zeigen die Anwendung der Röntgenrefraktionsradiographie (2,5D Technik) und der -tomographie (3D Technik) für die Untersuchung verschiedener Probleme in der Werkstoffwissenschaft und -technologie:
a) Sintern von SiC-Grünkörpern
b) Porositätsanalyse in Dieselpartikelfiltersilikaten
c) Faser-Matrix-Haftung in Metall- und Polymermatrixverbundwerkstoffen
d) Mikrorissbildung in Glaskeramik.
Wir zeigen, dass der Einsatz von Röntgenrefraktionsmethoden quantitative Ergebnisse liefert, die direkt als Parameter in Werkstoffmodellen verwendet werden können.
X-ray refraction is analogous to visible light deflection by matter; it occurs at boundaries between different media. The main difference between visible light and X-rays is that in the latter case deflection angles are very small, from a few seconds to a few minutes of arc (i.e., the refraction index n is near to 1). Importantly, deflection of X-rays is also sensitive to the orientation of the object boundaries. These features make X-ray refraction techniques extremely suitable to a) detect defects such as pores and microcracks and quantify their densities in bulk (not too heavy) materials, and b) evaluate porosity and particle properties such as orientation, size, and spatial distribution (by mapping). While X-ray refraction techniques cannot in general image single defects, they can detect objects with size above a few wavelengths of the radiation.
Such techniques, especially at the Synchrotron BESSY II, Berlin, Germany, can be used in-situ, i.e. when the specimen is subjected to temperatures or external loads.
The use of X-ray refraction analysis yields quantitative information, which can be directly input in kinetics, mechanical and damage models.
We hereby show the application of non-destructive X-ray refraction radiography (SXRR, 2D mapping also called topography) to problems in additive manufacturing:
1) Porosity analysis in PBF-LM-Ti64. Through the use of SXRR, we could not only map the (very sparse) porosity distribution between the layers and quantify it, but also classify, and thereby separate, the filled porosity (unmolten powder) from the keyhole and gas pores (Figure 1).
2) In-situ heat treatment of laser powder bed fusion PBF-LM-AlSi10Mg to monitor microstructure and porosity evolution as a function of temperature (Figure 2). By means of SXRR we indirectly observed the initial eutectic Si network break down into larger particles as a function of increasing temperature. We also could detect the thermally induced porosity (TIP). Such changes in the Si-phase morphology upon heating is currently only possible using scanning electron microscopy, but with a much smaller field-of-view. SXRR also allows observing the growth of some individual pores, usually studied via X-ray computed tomography, but again on much smaller fields-of-view.
Our results show the great potential of in-situ SXRR as a tool to gain in-depth knowledge of the defect distribution and the susceptibility of any material to thermally induced damage and/or microstructure evolution over statistically relevant volumes.
In Leichtbauanwendungen sind Materialien mit geringen Dichten und hohen Festigkeiten gefragt. Insbesondere Faser-Kunststoff-Verbunde (FKV) erfüllen diese Anforderungen und werden zunehmend für Strukturbauteile eingesetzt. Die Oberflächen moderner Rotorblätter von Windkraftanlagen werden aus FKV gefertigt. Das Versagen dieser Materialien hat massiven Einfluss auf die Einsatzfähigkeit der gesamten Windkraftanlage. Daher ist die Kenntnis des Versagensverhaltens dieser Komponenten unter statischer und zyklischer Belastung von großem Interesse, um die Betriebsbeanspruchung abzuschätzen. Im Idealfall wird die Schadensentwicklung unter Betriebslast mit zerstörungsfreien in-situ-Prüfverfahren ermittelt. Hier berichten wir über die Beobachtung der Rissentwicklung in Carbon-Faserverstärkten Kunststoffen (CFK) unter kontinuierlicher Zugbelastung durch in-situ Synchrotron-Röntgenrefraktionsradiographie. Eine selbst entwickelte elektromechanische Zugprüfmaschine mit einem Kraftbereich bis zu 15 kN wurde in den Strahlengang an der BAMline (BESSY II) integriert. Da in herkömmlichen (Absorptions-) Radiographien Defekte wie Zwischenfaserbrüche oder Faser-Matrix-Enthaftung keinen ausreichenden Kontrast verursachen, wird zur Kontrastanhebung die Röntgenrefraktion benutzt. Hier wird das sogenannte Diffraction Enhanced Imaging (DEI) angewandt, um Primär- und gebrochene Strahlung mit Hilfe eines Analysatorkristalls zu trennen. Diese Technik ermöglicht schnelle Messungen mit einem Gesichtsfeld von einigen Quadratmillimetern (hier: 14 mm × 7.2 mm) und ist ideal für in-situ Untersuchungen. Die Bildgebung und der Zugversuch erfolgen mit einer Bildfrequenz von 0.7 / s und einer Dehnrate von 0.00055 / s . Bei 0°/90° Fasergelegen treten die ersten Zwischenfaserrisse bei 380 MPa (Dehnung 0.8 %) auf. Vor dem Versagen bei ca. 760 MPa (Dehnung 2.0 %) beobachten wir die Ausbildung eines nahezu äquidistanten Rissmusters (Risse in ca. 1 mm Abstand), das sich im geschädigten Zustand über die gesamte Probe erstreckt.
We propose a novel iterative segmentation algorithm (i.S.Sy.Da.T.A: Iterative Segmentation Synthetic Data Training Algorithm) employing Deep Convolutional Neural Networks and synthetic training data for X-ray tomographic reconstructions of complex microstructures. In our method, we reinforce the synthetic training data with experimental XCT datasets that were automatically segmented in the previous iteration. This strategy produces better segmentations in successive iterations. We test our algorithm with experimental XCT re constructions of a 6-phase Al-Si Matrix Composite reinforced with ceramic fibers and particles. We perform the analysis in 3D with a special network architecture that demonstrates good generalization with synthetic training data. We show that our iterative algorithm returns better segmentations compared to the standard single training approach. More specifically, phases possessing similar attenuation coefficients can be better segmented: for Al2O3 fibers, SiC particles, and Intermetallics, we see an increase of the Dice score with respect to the classic approach: from 0.49 to 0.54, from 0.66 to 0.72, and from 0.55 to 0.66 respectively. Furthermore, the overall Dice score increases from 0.77 to 0.79. The methods presented in this work are also applicable to other materials and imaging techniques.
The microstructure of an apatite-wollastonite (code name AP40) glass-ceramic is analyzed in this study by combining 2D microscopy, phase analysis, X-ray absorption and synchrotron X-ray refraction computed tomography (XCT and SXRCT, respectively). It is shown that this combination provides a useful toolbox to characterize the global microstructure in a wide scale range, from sub-micrometer to millimeter. The material displays a complex microstructure comprising a glassy matrix with embedded fluorapatite and wollastonite small crystals. In this matrix, large (up to 200 μm) spike-shaped structures are distributed. Such microstructural features are oriented around a central sphere, thereby forming a structure resembling a sea urchin. A unique feature of SXRCT, in contrast to XCT, is that internal interfaces are visualized; this allows one to show the 3D distribution of these urchins with exceptionally good contrast. Furthermore, it is revealed that the spike-shaped structures are not single crystals, but rather composed of sub-micrometric crystals, which are identified as fluorapatite and diopside phases by SEM-EDX analysis.
For how trivial or provocative it can sound, the best neutron spectrometer in the world does not produce science and technology by itself. By definition of “Materials Science”, neutron scattering data on engineering materials must be used as a tool to understand, and even tailor, materials performance. In order for this to happen, neutron data need to be
1. Acquired under the most relevant condition possible
2. Coupled to other experimental techniques
3. Capitalized by means of proper simulations and data analysis
Point 1- calls for an intense use and the development of top-notch of in-situ techniques; Point 2- means that the sole use of neutron data will not lead to any solution of a global problem; All points above hint to the fact that access to neutron sources is not routine, and therefore it is imperative to search ways to make neutron data rentable and sustainable for the material science and industrial research community.
In this presentation, and based on two examples, we will show a couple of strategies to combine neutron data with other experiments, and with theoretical models to raise the validity of experiments to the level of problem-solving. As one might imagine, these are only a few among the almost infinite combinations possible to help improving material properties, performance, and safety, i.e., ripe for everyday use.
Two biopolyol-based foams derived from banana leaves (BL) or stems (BS) were produced, and their compression mechanical behavior and 3D microstructure were characterized. Traditional compression and in situ tests were performed during 3D image acquisition using X-ray microtomography. A methodology of image acquisition, processing, and analysis was developed to discriminate the foam cells and measure their numbers, volumes, and shapes along with the compression steps.
The two foams had similar compression behaviors, but the average cell volume was five times larger for the BS foam than the BL foam. It was also shown that the number of cells increased with increasing compression while the average cell volume decreased. Cell shapes were elongated and did not change with compression. A possible explanation for these characteristics was proposed based on the possibility of cell collapse. The developed methodology will facilitate a broader study of biopolyol-based foams intending to verify the possibility of using these foams as green alternatives to the typical petrol-based foams.