Technische Universität Dresden: Qucosa
Not a member yet
16344 research outputs found
Sort by
A Virtualization Layer for the Management of Hardware Tasks in Embedded Reconfigurable Systems
No full textVirtualisierung ermöglicht die gleichzeitige Ausführung mehrerer Gastbetriebssysteme mit unterschiedlichen Anforderungen, z.B. an Echtzeit oder Sicherheit, auf derselben Hardwareplattform. Der Einsatz von Field Programmable Gate Arrays (FPGAs) bietet eingebetteten Systemen die Möglichkeit, einen Teil der Funktionalität in Form von Hardwaretasks zu beschleunigen. Allerdings sind FPGAs nicht für die gemeinsame Nutzung durch mehrere Betriebssysteme konzipiert. Daher stellt diese Arbeit eine FPGA-Virtualisierungsschicht vor, die auf die spezifischen Bedürfnisse des jeweiligen eingebetteten Systems zugeschnitten werden kann. Um eine Zugriffskontrolle für Hardware-Tasks zu erreichen, bietet sie einen POSIX-Thread-basierten Frontend/Backend-Treibermechanismus und einen Memory-Management Unit-basierten Mechanismus in Hardware. Die dynamische Scheduling-Strategie berücksichtigt harte und weiche Echtzeit-Anforderungen von Hardware-Tasks in Kombination mit Tasks ohne Echtzeit-Anforderungen. Als Kommunikationsinfrastruktur werden busbasierte Systeme und Networks-on-Chip (NoCs) unterstützt. Zwei plattformabhängige Synchronisationsstrategien verbessern die Energie-Effizienz für Hardware-Tasks. Auf Flash-basierten FPGAs werden Leerlaufphasen verlängert, um den Schlafmodus Flash*Freeze optimal einzusetzen. Bei SRAM-basierten FPGAs werden Tasks synchronisiert, um die Spannung möglichst niedrig skalieren zu können.:1. Introduction
1.1 Motivation
1.2 Challenges and Research Questions
1.3 Contributions
2. Background and Related Work
2.1 Virtualization of Computing Systems
2.2 Virtualization of Reconfigurable Systems
2.3 Related Work on FPGA Virtualization
3 Virtualization Services of L4ReC
3.1 Virtualization Model
3.2 Basic Services of L4ReC
4 Scheduling of Independent Hardware Tasks
4.1 Overview
4.2 Design Methodology
4.3 Implementation
4.4 Schedulability Analysis
4.5 Evaluation
4.6 Summary and Discussion
5 Scheduling of Chained Hardware Tasks
5.1 Overview
5.2 Design Methodology
5.3 Implementation
5.4 Evaluation
5.5 Summary and Discussion
6 Enhancement of Energy-Efficiency
6.1 Overview
6.2 Solution for Flash-based FPGAs
6.3 Solution for SRAM-based FPGAs
6.4 Summary and Discussion
7 Conclusion and Outlook
7.1 Conclusion
7.2 Future WorkVirtualization of embedded systems plays an increasingly important role. It allows guest operating systems with different requirements regarding, e.g., real-time or safety to run separated from each other on the same hardware platform. The consolidation of multiple systems onto the same platform reduces the hardware footprint, increases reliability and scalability, and the abstraction of the underlying hardware leads to a faster time-to-market. The utilization of Field Programmable Gate Arrays (FPGAs) offers embedded systems the opportunity to accelerate part of the functionality in form of hardware tasks executed on hardware accelerators. Hardware tasks enable to exploit the FPGA’s parallelism, reconfigurability, and energy saving potential. Nevertheless, FPGAs are not conceptualized for shared usage. So, FPGA virtualization poses the challenge to adhere with data and performance isolation. The limited FPGA area and high reconfiguration latencies impede the shared usage of hardware accelerators. Energy restricted embedded systems face the challenge that certain mechanisms for power reduction can only be applied to the entire FPGA and so are hardly applicable when several accelerators are used concurrently by different tasks.
This work addresses the research questions how to conquer the limitations of embedded reconfigurable systems, how to ensure isolation and how to meet varying requirements of diverse applications. We present an FPGA virtualization layer that can be tailored to the specific needs of the respective embedded system. In order to give tasks executed in guest operating systems access to hardware accelerators, we depict a POSIX Threads-based frontend/backend driver mechanism and a mechanism that controls access via a custom memory management unit. Communication via AXI4-Lite protocol as well as AXI4-Stream protocol is supported.
In embedded systems, often small and efficient real-time operating systems, e.g., for safety-critical control systems, run next to general purpose operating systems that offer consumer functionality. Thus, our virtualization layer provides a dynamic scheduling strategy that considers hard, soft, and no real-time requirements of hardware tasks. It masks reconfiguration latencies and so increases task throughput and FPGA resource utilization. Further, different communication infrastructure is supported. Next to bus-based systems, Networks-on-Chip (NoCs) are in particular beneficial for communication intense systems. Our virtualization layer enables the shared use of chained hardware accelerators in a NoC. Task priorities are respected by dynamically relocating hardware tasks with lower priorities if necessary. In order to improve energy efficiency, we provide two platform dependent synchronization strategies for hardware tasks. By prolonging idle phases, we leverage the low power mode “Flash*Freeze” for flash-based FPGAs. On SRAM-based FPGAs, our strategy increases the profit drawn by voltage scaling without relying on idle phases.
The evaluation reveals that the scheduling strategy for bus-based architectures considering hard, soft, and no real-time requirements reaches a 193% higher FPGA utilization than an earliest deadline first-based strategy, as well as a reduction of deadline misses by 3%. The scheduling strategy for NoC-based architectures manages to process 22.6% more tasks than an approach that grants exclusive access to the processing elements in the NoC. The synchronization strategy for flash-based FPGAs reduces the energy consumption on average by 33.5% compared to a strategy that does not employ the Flash*Freeze mode and by 22.5% compared to a strategy that uses the Flash*Freeze mode, but no synchronization. The synchronization strategy for SRAM-based FPGAs achieves an average power saving by 16.3% compared to a strategy that scales voltage without synchronizing hardware tasks. The algorithms targeting mixed criticality task systems can be combined with the proposed energy saving strategies. Thus, the proposed FPGA virtualization layer tackles the challenges of embedded multi-tenant systems and enables them to take advantage of the synergisms that emerge between hardware and software tasks that are created in a hardware/software co-design.:1. Introduction
1.1 Motivation
1.2 Challenges and Research Questions
1.3 Contributions
2. Background and Related Work
2.1 Virtualization of Computing Systems
2.2 Virtualization of Reconfigurable Systems
2.3 Related Work on FPGA Virtualization
3 Virtualization Services of L4ReC
3.1 Virtualization Model
3.2 Basic Services of L4ReC
4 Scheduling of Independent Hardware Tasks
4.1 Overview
4.2 Design Methodology
4.3 Implementation
4.4 Schedulability Analysis
4.5 Evaluation
4.6 Summary and Discussion
5 Scheduling of Chained Hardware Tasks
5.1 Overview
5.2 Design Methodology
5.3 Implementation
5.4 Evaluation
5.5 Summary and Discussion
6 Enhancement of Energy-Efficiency
6.1 Overview
6.2 Solution for Flash-based FPGAs
6.3 Solution for SRAM-based FPGAs
6.4 Summary and Discussion
7 Conclusion and Outlook
7.1 Conclusion
7.2 Future Wor
Non-Reciprocity in Active Silicon Microring Resonators
No full textThis thesis addresses non-reciprocity in silicon ring resonators, using the scheme of time-varying modulation of the refractive index. The proposed structure is a silicon ring resonator with two small time-modulated regions. First, the theoretical fundamentals are presented and analyzed. Based on the theoretical considerations, proper design for the structure is presented. Finally, the designed ring resonator is experimentally characterized in order to prove the described concepts.
Temporal coupled mode theory (TCMT) is used as a tool to understand the operation principle of non-reciprocity in the ring structure, and also to design and optimize the proposed devices based on this characteristic. According to the TCMT, the non reciprocity is caused by the cross-coupling between the counter-rotating modes of the ring, which breaks their degeneracy. The TCMT is used to find the angular distance, phase difference, and width of the two time-modulated regions on the ring resonator that breaks the degeneracy of the clockwise and counter-clockwise modes and hence leads to non-reciprocity in the ring resonator. In the next step, TCMT equations are solved to find the modes of this time-modulated ring. For this purpose, two methods of harmonic balance method (HBM) and transfer matrix method (TMM) are used. Thus, degeneracy breaking is demonstrated from the modal analysis. Furthermore, the results of the modal analysis enable us to explain the underlying operation principle of this structure via two counter-acting mechanisms of the separation between the resonance frequencies of counter-rotating modes and energy transfer to the side harmonic. These two mechanisms are related to the amplitude of the cross-coupling and self-coupling terms in the TCMT equations, respectively.
In Chapter 3, we present the design and simulation of the proposed devices, based on the time-modulated ring resonator. First, a magnetic-free, compact, and complementary metal-oxide semiconductor (CMOS)-compatible optical isolator is proposed based on the non-reciprocal time-modulated ring side coupled to one waveguide. In the next step, using the results of the TCMT model, a realistic design guideline based on a CMOS-compatible silicon photonics foundry process is presented. We show that there is a trade-off between the two aforementioned parameters of frequency splitting and energy transfer to the side harmonics, to obtain maximum isolation. We show that we can enhance isolation by tuning the modulation phase difference. Robustness of the design is investigated through a sensitivity analysis for the fabrication variations in the distance and width of the modulated regions. The second proposed structure is an optical circulator which is realized by placing the designed time-modulated ring resonator in an add-drop configuration, and by placing a Bragg reflector in one of the ports.
Finally, in Chapter 4 we present, to the best of our knowledge, the first experimental demonstration of non-reciprocity in a compact active silicon photonic ring resonator with time-modulated regions, fabricated with a CMOS-compatible commercial foundry. We demonstrate symmetry breaking of counter-rotating modes in an active silicon photonic ring resonator by applying phase-shifted RF signals to only two small p-i-n junctions on the ring, without employing traveling-wave modulation in the whole structure. By reversing the polarity of the RF phase difference (e.g. (45°,-45°) asymmetric phases) opposite resonance wavelengths are obtained, with a 16-dB contrast between the transmissions of the asymmetric phases and a low insertion loss of 0.6 dB under 27 dBm RF power. We achieve the highest ratio of the asymmetric transmission to the insertion loss among the state-of-the-art silicon non-reciprocal integrated optical structures based on time varying modulation. The non-reciprocal ring can be used as a magnetic-free, low-loss, compact, and CMOS-compatible integrated optical isolator that does not require magnetic bias or high-intensity input light, as has been done in the other approaches. Moreover, the results of this work can be used to realize a novel method for RF phase demodulation by converting a phase-modulated RF signal into amplitude-modulated light
The German Transmission Grid Expansion in longterm perspective – What is the impact of Renewable Integration?
No full textThe German power sector will face a profound change from conventional power generation to an energy system with a high share of intermittent renewable energy sources. The following analysis addresses the future transmission grid design for Germany in 2030 in a context of different renewable integration schemes. Therefore a DC load flow model is set-up, which deals with the trade-off between grid expansion and regional renewable integration. Two scenarios concerning different renewable feed-in design options are considered: curtailment of renewables only at negative prices in the height of their individual levelized costs of electricity, and curtailment at marginal costs (tending to zero). To cope with the large-scale network capacity expansion problem, Benders Decomposition and DC-load flow approximation are applied. The results show that the different renewable integration schemes have different effects on the AC grid expansions in contrast to High-Voltage-DirectCurrent (HVDC) connections
Herkunftsanalyse und geotektonischer Rahmen der Varisziden im Kellerwald, im Lahn-Dill-Gebiet und in der Lindener Mark (südöstliches Rheinisches Schiefergebirge)
No full textDas südöstliche Rheinische Schiefergebirge beinhaltet als Teil der Rhenoherzynischen Zone innerhalb der Zentraleuropäischen Varisziden verschiedene geologische Einheiten, die oft auf engstem Raum stark verschuppt und untereinander nach NW überschoben sind. Sie galten bis in die 1970er Jahre als autochthon, obwohl sie sich trotz gleichen Alters deutlich in ihrem stratigraphischen Aufbau und auch in ihrer Faziesentwicklung unterscheiden.
Um diese Einheiten deutlicher zu differenzieren, die Herkunft ihrer Sedimente zu klären und die geotektonische Entwicklung des Gebietes zu verstehen, wurden im Lahn-Dill-Gebiet, im Kellerwald und in der Lindener Mark 38 Ordovizische, Devonische und Unterkarbonische Sandsteine und Grauwacken wissenschaftlich untersucht. Dabei wurden die in den siliziklastischen Sedimenten enthaltenen detritischen Zirkone mittels moderner radiometrischer Methoden (U-Pb-Th LA–ICP–MS und Lu-Hf LA-MC-ICP-MS) sowie morphometrischer Vermessungen analysiert. Jede Probe lieferte dabei eine bestimmte Altersverteilung der Zirkonpopulationen anhand derer Rückschlüsse auf das jeweilige Liefergebiet möglich sind.
Ein unterdevonischer Sandstein aus der nördlich ans Untersuchungsgebiet anschließenden Siegen-Antiklinale stellte eine Referenzprobe für diejenigen siliziklastische Sedimente dar, die einen Bezug zum einstigen Großkontinent Laurussia haben. Diese Probe deutet auf Ablagerungen hin, die hauptsächlich von Baltika und in geringem Maße von Avalonia geschüttet wurden. Die größten Peaks liegen in diesem Fall zwischen 1,0 und 1,6 Milliarden Jahre (Ga), also im Bereich des Mesoproterozoikums, und ~400-450 Millionen Jahre (Ma). Vergleichbare Ergebnisse traten vor allem bei den unter- bis oberdevonischen autochthonen Ablagerungen der Dill-Eder- und Lahn-Mulde sowie bei den Rhenoherzynischen Deckeneinheiten (Kammquarzit-Decke, Frankenbach-Decke, Bicken-Ense- und Wildestein-Decke) auf.
Die Cluster der U-Pb-Alter der allochthonen Einheiten wie der Gießen-Decke sowie der heute neu definierten Hörre-Decke, Steinhorn-Decke, Lohra-Decke und Lindener Mark-Decke deuten dagegen auf eine Provenance aus der Saxothuringischen oder Armorikanischen Zone hin. Zirkonpopulationen dieser Zonen sind repräsentativ für ein Gondwana gebundenes Hinterland und sind charakterisiert durch Alter zwischen ~530-700 Ma sowie Alter aus dem Paläoproterozoikum und Archaikum (~1,8-2.2 Ga, ~2.5-2.7 Ga und 3.0-3-4 Ga). Im Bereich des Mesoproterozoikums weisen sie jedoch eine deutliche Lücke auf. Dies unterscheidet sie deutlich von den Sedimenten, welche von Baltika geliefert wurden. Damit können für das Untersuchungsgebiet zwei sich stark voneinander unterscheidende Liefergebiete nachgewiesen werden. Allerdings finden sich derartige Altersverteilungen auch in den unterkarbonischen Einheiten des Autochthons, wo sie vermutlich die bereits einsetzende Flyschsedimentation des gerade entstandenen Variszischen Gebirges darstellen und somit den Abtrag der Deckeneinheiten widerspiegeln. Außerdem zeichnen sich diese Einheiten durch das Fehlen typischer Gesteine eines Intraplattenvulkanismus aus. Lediglich an der Basis der Gießen-Decke treten Metabasalte mit einer Signatur eines Mittelozeanischen Rückens (MORB-Typ-Affinität) als tektonische Schuppen auf.
Nach Auswertung aller Analysen, einer umfangreichen Literaturrecherche und der Interpretation der Geländedaten ergab sich das Bild eines Deckenstapels aus Rhenoherzynsichen Decken und Saxothuringischen Ferndecken, wobei die Liefergebiete dieser Decken auf zwei verschiedenen Kontinenten lagen, Gondwana und Laurussia. Die Decken wurden während der Kollision dieser zwei Großkontinente auf die autochthonen Einheiten des Rheinischen Schiefergebirges aufgeschoben und miteinander verschuppt.
Das geotektonische Modell zur Zusammenführung dieser Liefergebiete geht davon aus, dass der Rheische Ozean zunächst vermutlich über eine nordwärts gerichtete Subduktion unter den südlichen Teil Avalonias im Devon geschlossen wurde. Durch die steile Subduktion seiner ozeanischen Platte öffnete sich ein Rhenoherzynisches back-arc-Becken. Die Öffnung dieses Beckens ging einher mit der Dehnung und Absenkung der kontinentalen Kruste Nord-Avalonias (südliches Schelfgebiet von Laurussia), einem intensiven Intraplattenvulkanismus und der Spaltung von Avalonia in einen nördlichen und einen südlichen Teil. In der finalen Phase wurde das Rhenoherzynische Becken durch eine südwärts gerichtete Subduktion geschlossen, wobei beide Großkontinente kollidierten.:Abbildungsverzeichnis 4
Tabellenverzeichnis 18
Abkürzungsverzeichnis 19
Zusammenfassung / Abstract 20
1. Einleitung und Zielstellung 24
2. Der großgeologische Rahmen 28
3. Das südöstliche Rheinische Schiefergebirge 35
3.1 Lahn-Dill-Gebiet 35
3.2 Kellerwald 39
3.3 Struktur-Einheiten im Lahn-Dill-Gebiet und im Kellerwald 42
3.3.1 Dill-, Eder- und Lahn-Mulde 42
3.3.2 Hörre-Zone 46
3.3.3 Gießen-Decke 49
3.3.4 Lohra-Kehna-Einheiten 50
3.3.5 Bicken–Ense- und Wildestein-Schuppe 51
3.3.5.1 Kammquarzit-Schuppenzone 52
3.3.6 Frankenbacher Schuppenzone 53
3.4 Die Lindener Mark 54
3.5 Vulkanismus im Untersuchungsgebiet 57
3.6 Siegen-Antiklinale 60
4. Material und Methoden 63
4.1 Probenahme 65
4.1.1 Proben aus dem Lahn-Dill-Gebiet 66
4.1.1.1 Bisch-2 und Bisch-4 66
4.1.1.2 Bott-1 67
4.1.1.3 Eckel-2 67
4.1.1.4 Eln-1 und Eln-2 68
4.1.1.5 Haig-1 69
4.1.1.6 Herm-1 70
4.1.1.7 Heu-2 71
4.1.1.8 Hob-1 und Kamm-1 72
4.1.1.9 HP-1 73
4.1.1.10 Kirch-1 73
4.1.1.11 Nau-1 und Nau-2 74
4.1.1.12 Weip-1 75
4.1.1.13 Will-1 75
4.1.1.14 45441 76
4.1.1.15 45443 76
4.1.1.16 45444 77
4.1.2 Proben aus dem Kellerwald 79
4.1.2.1 KE-1 79
4.1.2.2 KE-2 79
4.1.2.3 KE-3 80
4.2.1.4 KE-4 80
4.2.1.5 KE-5a 80
4.2.1.6 KE-6 81
4.2.1.7 KE-8 82
4.2.1.8 KE-9c 83
4.1.2.9 KE-11 84
4.1.2.10 KE-12 84
4.1.2.11 KE-14 85
4.1.2.12 KE-15 85
4.1.2.13 KE-16 86
4.1.2.14 KE-17 87
4.1.2.15 KE-18 87
4.1.2.16 KE-19 88
4.1.3 Proben aus der Lindener Mark 90
4.1.3.1 L1-LM2 90
4.1.3.2 L2-LM1 91
4.2 Probenaufbereitung 91
4.3 Rasterelektronenmikroskopie 94
4.4 Analytik mittels LA-ICP-MS 95
4.4.1 U-Pb-Analytik an Zirkonen 95
4.4.2 Das Th-U-Verhältnis bei Zirkonen 98
4.4.3 Analytik der Lu-Hf-Isotope an Zirkonen 99
4.5 Zirkonmorphometrie 101
4.6 Dünnschliffanalytik 101
5. Ergebnisse und deren Deutung 103
5.1 U-Pb-Analytik an Zirkonen und REM-Analytik 103
5.2 Th/U-Verhältnisse der gemessenen Zirkone 110
5.3 εHf-Werte und Modell-Alter 113
5.4 Zirkonmorphometrie 116
5.5 Dünnschliffanalytik 118
6. Interpretation 119
6.1 Modellentwicklung zum Kollisionsgeschehen 119
6.2 Neustrukturierung der geologischen Einheiten 133
6.2.1 Hörre-Decke und Kammquarzit-Decke 133
6.2.2 Neudefinition der Lohra-Decke 136
6.2.3 Neudefinition der Steinhorn-Decke 139
6.2.4 Neudefinition der Frankenbach-Decke 142
6.2.5 Neudefinition der Lindener Mark-Decke 143
6.2.6 Auswirkung auf die Geologischen Karten Hessens 143
7. Ausblick 145
8. Literaturverzeichnis 146
9. Anhang 17
Microstructure Reconstruction from Differentiable Descriptors - A General Framework and Adapted Algorithms
No full textAccelerating materials engineering by computer-based workflows is a major goal in academia and industry. In the context of random heterogeneous media, the utility of numerical simulation and homogenization methods hinges on the underlying computational domains. For this purpose, microstructure reconstruction algorithms aim at generating realistic realizations of these media.
In this work, a general framework and various algorithms for descriptor-based microstructure reconstruction are presented. The key idea is to establish and exploit the differentiability of the underlying statistical description in order to massively accelerate the reconstruction process. As a basis, microstructure reconstruction is formulated as an optimization problem for 2D phase indicator functions and solved by gradient-based methods. This idea is extended to 3D by regarding it as a coupled reconstruction of individual slices and introducing multigrid and post-processing methods for increased efficiency and reduced noise. The subsequent extension to crystallographic orientation is centered around the symmetrization of the differentiable descriptors and the post-processing. Finally, three adapted reconstruction algorithms are presented, namely a descriptor-based neural cellular automaton, a 3D extension of the Portilla-Simoncelli algorithm and a method dedicated specifically to ellipsoidal inclusions. They can be regarded as special cases of the general framework, where the broad applicability is sacrificed in favor of a drastically improved performance.:1) Introduction
I) Theory and Overview
2) Characterizing microstructure morphology by statistical descriptors
3) Microstructure reconstruction
II) Developed framework for optimization-based microstructure reconstruction
4) Introducing differentiable descriptors to optimization-based reconstruction
5) Reconstructing 3D volumes from 2D slices by differentiable descriptors
6) Reconstructing orientation maps
III) Adapted methods for specific applications
7) Learning structure-generating partial differential equation systems from descriptors
8) Fast reconstruction of 3D structures using the Portilla-Simoncelli algorithm
9) Reconstructing structures with ellipsoidal inclusions from analytical descriptors
IV) Conclusion
10) Concluding remarks
Bibliography
A) Numerical simulation and homogenization
B) Descriptor developments
C) Additional information on numerical experimentsDie Materialentwicklung durch computergestützte Workflows zu beschleunigen ist ein wichtiges Ziel in Wissenschaft und Industrie. Im Kontext heterogener Medien mit stochastischer Mikrostruktur hängt die Anwendbarkeit numerischer Simulations- und Homogenisierungsmethoden von den zugrundeliegenden Rechengebieten ab. Daher zielen Mikrostrukturrekonstruktions\-algorithmen darauf ab, realistische Realisierungen jener Medien zu erzeugen.
In dieser Arbeit werden ein allgemeines Framework und verschiedene Algorithmen zur deskriptorbasierten Mikrostrukturrekonstruktion vorgestellt. Die Kernidee besteht im Herstellen und Ausnutzen der Differenzierbarkeit der zugrundeliegenden statistischen Beschreibung, um die Rekonstruktion massiv zu beschleunigen. Dazu wird die Mikrostrukturrekonstruktion als ein Optimierungsproblem für 2D-Phasenindikatorfunktionen formuliert und gradientenbasiert gelöst. Diese Idee wird auf 3D erweitert, indem sie als gekoppelte Rekonstruktion einzelner Schichten betrachtet wird und Multi-Grid und Post-Processing zur Effizienzsteigerung und Rauschreduzierung eingeführt werden. Die anschließende Erweiterung auf kristallographische Orientierung behandelt vorrangig die Symmetrisierung der differenzierbaren Deskriptoren und das Post-Processing. Schließlich werden drei angepasste Rekonstruktionsalgorithmen vorgestellt, nämlich ein deskriptorbasierter neuronaler zellulärer Automat, eine 3D-Erweiterung des Portilla-Simoncelli-Algorithmus und eine Methode, die speziell auf ellipsoidische Einschlüsse zugeschnitten ist. Diese drei Methoden können als Spezialfälle des allgemeinen Frameworks betrachtet werden, bei denen die weniger breite Anwendbarkeit einer drastisch verbesserten Leistung gegenübersteht.:1) Introduction
I) Theory and Overview
2) Characterizing microstructure morphology by statistical descriptors
3) Microstructure reconstruction
II) Developed framework for optimization-based microstructure reconstruction
4) Introducing differentiable descriptors to optimization-based reconstruction
5) Reconstructing 3D volumes from 2D slices by differentiable descriptors
6) Reconstructing orientation maps
III) Adapted methods for specific applications
7) Learning structure-generating partial differential equation systems from descriptors
8) Fast reconstruction of 3D structures using the Portilla-Simoncelli algorithm
9) Reconstructing structures with ellipsoidal inclusions from analytical descriptors
IV) Conclusion
10) Concluding remarks
Bibliography
A) Numerical simulation and homogenization
B) Descriptor developments
C) Additional information on numerical experiment
Soziotechnische Systeme - Symbiose von Mensch & Maschine: Online-Content zum interaktiven Whitepaper KORESIL
Get PDFAn der Schnittstelle zwischen Mensch und Maschine sind soziotechnische Systeme als eben jene Strukturen definiert, bei denen es zu einer bilateralen Interaktion zwischen Mensch und Technik kommt. Im Kontext produzierender Unternehmen definiert der Begriff das zielführende Zusammenwirken von sozialen Systemen bzw. Individuen, z. B. der Fachkraft am Fließband, und technischen Systemen, wie dem Fließband selbst, einem Coboter oder Computern, welche die menschliche Arbeitsleistung unterstützen.
Neueste soziotechnische Systeme bieten große Potentiale für die Arbeitswelt von morgen und ermöglichen Effizienz- und Qualitätssteigerungen, die ein Reshoring von Fertigung in Hochlohnländer möglich machen. Eine erhöhte Resilienz der Lieferketten ist dabei ein positiver Nebeneffekt. Zugleich können gesundheitsgefährdende Tätigkeiten entschärft und so die physische wie psychische Gesundheit der Arbeitskräfte geschützt werden.
Im vorliegenden Whitepaper wird in diesem Zusammenhang insbesondere die Augmented Reality Technologie fokussiert, die durch die Überlagerung visueller Informationen im realen Sichtfeld der Nutzenden geprägt ist und vollkommen neue Möglichkeiten für Ausbildungs-, Entwicklungs- und Produktionsumgebungen bietet. Es werden entsprechende Anwendungsszenarien vorgestellt, Vor- und Nachteile der Technologie diskutiert und Potentiale für die Zukunft skizziert.:Soziotechnische Systeme 1
Augmented Reality und ihre Einsatzgebiete 4
Ergonomie in Augmented Reality Anwendungen 9
KORESIL-Anwendungsforschung zu Augmented Reality 14
KORESIL-Studienergebnisse zur Nutzung von Augmented Reality 24
Executive Summary 28
Literatur 3
Psychophysiological reactions during the trauma-film paradigm and their predictive value for intrusions
Get PDFBackground: Adequate adaptation of the autonomic nervous system (ANS) is crucial in potentially life-threatening situations. The defence cascade provides a descriptive model of progressing dominant physiological reactions in such situations, including cardiovascular parameters and body mobility. The empirical evidence for this model is scarce, and the influence of physiological reactions in this model for predicting trauma-induced intrusions is unresolved.
Objectives: Using a trauma-film paradigm, we aimed to test physiological reactions to a highly stressful film as an analogue to a traumatic event along the defence cascade model. We also aimed to examine the predictive power of physiological activity for subsequent intrusive symptoms.
Method: Forty-seven healthy female participants watched a stressful and a neutral film in randomized order. Heart rate (HR), heart rate variability (HRV), and body sway were measured. Participants tracked frequency, distress, and quality of subsequent intrusions in a diary for 7 consecutive days.
Results: For the stressful film, we observed an initial decrease in HR, followed by an increase, before the HR stabilized at a high level, which was not found during the neutral film. No differences in HRV were observed between the two films. Body sway and trembling frequency were heightened during the stressful film. Neither HR nor HRV predicted subsequent intrusions, whereas perceived distress during the stressful film did.
Conclusions: Our results suggest that the physiological trauma-analogue response is characterized by an orientation response and subsequent hyperarousal, reaching a high physiological plateau. In contrast to the assumptions of the defence cascade model, the hyperarousal was not followed by downregulation. Potential explanations are discussed. For trauma-associated intrusions in the subsequent week, psychological distress during the film seems to be more important than physiological distress. Understanding the interaction between physiological and psychological responses during threat informs the study of ANS imbalances in mental disorders such as post-traumatic stress disorder
Label‑free multiphoton microscopy enables histopathological assessment of colorectal liver metastases and supports automated classifcation of neoplastic tissue
Get PDFAs the state of resection margins is an important prognostic factor after extirpation of colorectal liver metastases, surgeons aim to obtain negative margins, sometimes elaborated by resections of the positive resection plane after intraoperative frozen sections. However, this is time consuming and results sometimes remain unclear during surgery. Label-free multimodal multiphoton microscopy (MPM) is an optical technique that retrieves morpho-chemical information avoiding all staining and that can potentially be performed in real-time. Here, we investigated colorectal liver metastases and hepatic tissue using a combination of three endogenous nonlinear signals, namely: coherent anti-Stokes Raman scattering (CARS) to visualize lipids, two-photon excited fluorescence (TPEF) to visualize cellular patterns, and second harmonic generation (SHG) to visualize collagen fibers. We acquired and analyzed over forty thousand MPM images of metastatic and normal liver tissue of 106 patients. The morphological information with biochemical specificity produced by MPM allowed discriminating normal liver from metastatic tissue and discerning the tumor borders on cryosections as well as formalin-fixed bulk tissue. Furthermore, automated tissue type classification with a correct rate close to 95% was possible using a simple approach based on discriminant analysis of texture parameters. Therefore, MPM has the potential to increase the precision of resection margins in hepatic surgery of metastases without prolonging surgical intervention
Promise not fulfilled: FinTech, data privacy, and the GDPR
Get PDFThis article analyzes how the General Data Protection Regulation (GDPR) has affected the privacy practices of FinTech firms. We study the content of 276 privacy statements respectively before and after the GDPR became binding. Using text analysis methods, we find that the readability of the privacy statements has decreased. The texts of privacy statements have become longer and use more standardized language, resulting in worse user comprehension. This calls into question whether the GDPR has achieved its original goal—the protection of natural persons regarding the transparent processing of personal data. We also link the content of the privacy statements to FinTech-specific determinants. Before the GDPR became binding, more external investors and a higher legal capital were related to a higher quantity of data processed and more transparency, but not thereafter. Finally, we document mimicking behavior among FinTech industry peers with regard to the data processed and transparency
SetQuence & SetOmic: Deep set transformers for whole genome and exome tumour analysis
Get PDFIn oncology, Deep Learning has shown great potential to personalise tasks such as tumour type classification, based on per-patient omics data-sets. Being high dimensional, incorporation of such data in one model is a challenge, often leading to one-dimensional studies and, therefore, information loss. Instead, we first propose relying on non-fixed sets of whole genome or whole exome variant-associated sequences, which can be used for supervised learning of oncology-relevant tasks by our Set Transformer based Deep Neural Network, SetQuence. We optimise this architecture to improve its efficiency. This allows for exploration of not just coding but also non-coding variants, from large datasets. Second, we extend the model to incorporate these representations together with multiple other sources of omics data in a flexible way with SetOmic. Evaluation, using these representations, shows improved robustness and reduced information loss compared to previous approaches, while still being computationally tractable. By means of Explainable Artificial Intelligence methods, our models are able to recapitulate the biological contribution of highly attributed features in the tumours studied. This validation opens the door to novel directions in multi-faceted genome and exome wide biomarker discovery and personalised treatment among other presently clinically relevant tasks