58491 research outputs found
Sort by
Modeling and evaluation of thermo-mechanical properties of open-cell ceramic foams for metal melt filtration
No full textOpen-cell ceramic foams are used in metal melt filtration processes to clean and calm the liquid melt. Due to the high temperatures and pressure of the melt, thermo-mechanical stresses occur in the filter structures, which require a corresponding evaluation of strength, deformation, and failure. The ceramic materials used no longer behave elastically and brittle at operating temperatures of up to 1650 ◦ C, but exhibit viscoplastic behavior. Experimental investigations of the deformation of filter structures during the filtration process are difficult or even impossible, which is why simulation methods are used to investigate the filtration process and the filter loading.
The filters considered in this work are manufactured using a replica process in which a ceramic slurry is applied to an open-cell polyurethane foam, which is dried and fired in a thermal process. Real filter structures consist of a network of several thousand struts with varying geometries. Direct numerical simulation of these geometries is possible in principle, but it is very complex and expensive, which is why homogenization methods are used. Representative volume elements of the ceramic foams are generated and analyzed using the finite element method. The micro-macro relations determined in the process are mapped using corresponding continuum mechanical models. These models allow the evaluation of the thermo-mechanical behavior of filter materials and filter structures.
This thesis provides a critical overview of methods for generating, characterizing, and homogenizing foam structures. The generation of realistic foam structures is carried out using various methods from the fields of mathematics and mechanics and is described in detail. Analytical and data-driven approaches are used for the actual homogenization. The analytical approaches use adaptations of continuum mechanical models from the field of granular media. The data-driven approaches use neural networks, which replace or supplement hard-to-describe thermodynamic potentials used in material modeling. Both approaches can be used in a developed general framework for the modeling of any porous structures.
As a result of the research and modeling work carried out, generic and real foams are compared in terms of their topological and geometrical properties. It is discussed how local geometrical variations of foam structures affect the macroscopic behavior, considering different thermo-mechanical properties such as elasticity, viscoplasticity, and fracture strength. The developed homogenization concepts are compared with each other and with other concepts from the scientific literature and evaluated with respect to their accuracy, flexibility, and efficiency. Finally, possible further developments and applications are discussed.:1. Introduction
1.1. Motivation and objectives
1.2. Structure of the thesis
2. State of the art research
2.1. Integration of sub-project B05 into the CRC 920
2.2. Manufacture of open-cell foam structures
2.2.1. Schwartzwalder process
2.2.2. Additive manufacturing
2.2.3. Additional coatings
2.2.4. Bulk material properties
2.3. Characterization of open-cell foam structures
2.3.1. Topological and geometrical characteristics
2.3.2. Thermo-mechanical characteristics
2.3.3. Fluid dynamical characteristics
2.4. Modeling of open-cell foam structures
2.4.1. Geometrical models of foams
2.4.2. Direct numerical simulation
2.4.3. Homogenization approaches
2.4.4. Data-driven and machine-learning approaches
2.4.5. Constitutive models for open-cell foam structures
3. Modeling of open-porous ceramic foams
3.1. Foam surfaces of strut networks based on implicit functions
3.2. Sphere packings and Laguerre tessellations
3.3. Surface evolver, dry foams, wet foams, and foam froth
3.4. Voxel models and isosurfaces of foams
3.5. Finite element model
3.5.1. Models with structural elements
3.5.2. Unstructured tetrahedral meshes
3.5.3. Structured meshes
3.6. Generating foam structures using FoamGUI
3.7. Homogenized constitutive models
3.7.1. Scale bridging, meso and micro models
3.7.2. Effective elastic properties
3.7.3. Elastic limit surfaces
3.7.4. Effective yield surfaces
3.7.5. Modified Ehlers model
3.7.6. Constitutive model for viscoplastic behavior
3.7.7. Constitutive framework for plastic behavior
3.7.8. General return algorithm
3.7.9. Application to the phenomenological models
3.7.10. Hybrid models
3.7.11. Neural networks
3.7.12. Data sampling for the neural network training
3.7.13. Parameter identification for the modified Ehlers model
4. Results
4.1. Geometrical foam models
4.1.1. Foam models based on implicit functions
4.1.2. Foam models based on sphere packings
4.2. Effective thermo-mechanical properties
4.2.1. Geometry dependent elastic properties
4.2.2. Yield and failure surfaces
4.2.3. Fracture mechanical properties
4.2.4. Fracture mechanical properties for thermo-shock loading
4.2.5. Visco-plastic properties
4.2.6. Effective plastic properties
5. Conclusions & Discussio
Evaluation von Dünnschnittgewebekulturen des Ovarialkarzinoms als Testsystem für ein individuelles Therapieansprechen
No full text
CFD-gestützte Untersuchung und verbesserte Auswertung von Hochtemperaturexperimenten
No full textDie vorliegende Dissertationsschrift widmet sich der CFD-gestützten Untersuchung von Hochtemperaturversuchen, die für das Verständnis der Wärme- und Stoffübertragung im Labormaßstab von entscheidender Bedeutung sind. Mithilfe der CFD-Modellierung werden komplexe Temperatur- und Stoffverteilungen in Versuchsanlagen detailliert analysiert. Dies umfasst ein Modell zur Abbildung des Wärme- und Stoffübergangs im HITECOM-Reaktor und damit zur Identifizierung kritischer Betriebsbedingungen, ein Modell zur Analyse der Partikelbewegung im KIVAN-Reaktor unter inhomogenen Temperatur- und Strömungsbedingungen sowie die modellgestützte Analyse der Stofftransportlimitierungen in einer DMT-Thermowaage. Die Kombination von CFD-Modellierung und experimentellen Daten unterstützt die Entwicklung präziser kinetischer Modelle und verbessert die Vorhersage des Stoffumsatzes. Diese Ergebnisse tragen wesentlich zur Weiterentwicklung und Optimierung von Hochtemperaturprozessen in verschiedenen Industriezweigen bei
Ich will das selbst können!: Selbstständigkeitsförderung von Kindern mit Schlaganfall in einem Sozialpädiatrischen Zentrum
No full textEin Schlaganfall bei Kindern ist zwar selten, wird aber oft von zahlreichen kognitiven und physischen Einschränkungen begleitet, die nicht selten Kinder an den Rollstuhl binden. Insbesondere dann werden regelmäßige Therapiebesuche ein lebenslanger Begleiter. Oftmals finden diese Therapien in Sozialpädiatrischen Zentren (SPZ) statt, die häufig nicht kindgerecht und barrierefrei gestaltet sind. Die unterschiedlichen Barrieren, die sich in Sozialpädiatrischen Zentren für diese Kinder auftun, stellen eine besondere Herausforderung dar.
Diese Forschungsarbeit untersucht durch eine Analyse die Barrieren eines SPZs in Bezug auf die Selbstständigkeit von Kindern. [... aus dem Text
Evaluation of solar irradiances simulated by the Integrated Forecasting System of ECMWF using airborne observations in the Arctic
No full textPrognosen basierend auf numerischen Wettervorhersagemodellen sind in der Arktis, die im Vergleich zum Rest der Welt durch eine verstärkte Erwärmung charakterisiert ist, besonders unsicher. Oberflächeneigenschaften von offenem Wasser und Meereis im Arktischen Ozean sowie Wolken und ihre Eigenschaften sind Hauptquellen der Unsicherheit in der Vorhersage des Strahlungsenergiehaushalts und dem damit verbundenen Heizen und Kühlen der Atmosphäre. In dieser Arbeit wird ein systematischer Vergleich von Flugzeugmessungen der Arctic CLoud Observations Using airborne measurements during polar Day (ACLOUD) Messkampagne mit
Strahlungstransfersimulationen mithilfe des ecRad Strahlungsschemas vom European Centre for Medium-Range Weather Forecasts (ECMWF) durchgeführt.
Der erste Teil dieser Arbeit evaluiert die solaren Bestrahlungsstärken, die das Integrated Forecasting System (IFS) des ECMWF vorhergesagt hat. Während eine zufriedenstellende Übereinstimmung zwischen den IFS Simulationen und den Messungen in der Abwärtsrichtung festgestellt wird, weichen die aufwärtsgerichteten solaren Bestrahlungsstärken, die von der Ozeanoberfläche und den Wolken reflektiert werden, signifikant voneinander ab. Über Meereis unterschätzt das IFS
die Beobachtungen um 35 Wm−2 und über offenem Ozean überschätzt es sie um 28 Wm−2. Um mögliche Gründe für diese Abweichung zu identifizieren, wird eine umfassende Sensitivitätsanalyse durchgeführt, die verschiedene Oberflächeneigenschaften sowie makro- und mikrophysikalische Wolkeneigenschaften umfasst. Diese Sensitivitätsstudie wird realisiert, indem operationelle Parametrisierungen und prognostische Wolkenvariablen in ecRad durch ACLOUD Beobachtungen ersetzt werden. Die fehlerhafte Repräsentation der Meereisalbedo wird mit über 50 % als die Hauptursache für die Abweichung zwischen Simulationen und Beobachtungen identifiziert. Weitere signifikante Beiträge zur Abweichung über Meereis liefern die zu geringe Variabilität der Wolkentröpfchenanzahlkonzentration und der zu geringe Flüssigwasserpfad im IFS. Über offenem Ozean sind die signifikantesten Beiträge mit einem zu hohen Flüssigwasserpfad und der fehlerhaften Repräsentation des Bewölkungsanteils im IFS verbunden.
Der zweite Teil dieser Arbeit untersucht den Einfluss der Eisoptikparametrisierung auf die von Mischphasenwolken reflektierte solare Bestrahlungsstärke im IFS. Ein satellitenbasiertes Retrievalprodukt des Eiswasserpfades wird miteinbezogen, um ACLOUD Messungen der bandintegrierten solaren Bestrahlungsstärke mit Strahlungstransfersimulationen zu vergleichen, die verschiedene Eisoptikparametrisierungen anwenden. Die (relativ) größten Differenzen der aufwärtsgerichteten solaren Bestrahlungsstärke treten in den Wellenlängenbändern im Nahinfrarot auf und tragen dazu bei, dass die Simulationen die Beobachtungen um 10–13 Wm−2 unterschätzen. Diese Unterschätzung wird verringert, aber nicht vollständig kompensiert, indem Eisoptikparametrisierungen angewendet werden, die eine Mischung von Wuchsformen und eine Oberflächenrauigkeit beinhalten. Nicht-operationelle Eisoptikparametrisierungen erhöhen die reflektierte solare Bestrahlungsstärke für von Eis dominierten und von Wasser dominierten Wolkenobergrenzen jeweils um 35–68 Wm−2 und 3–7 Wm−2.:1 Introduction
1.1 Arctic weather and climate
1.2 Mixed-phase clouds
1.3 Ice optics parameterization
1.4 Previous model evaluations
1.4.1 Sea ice albedo
1.4.2 Cloud fraction
1.4.3 Liquid and ice water path
1.4.4 Cloud droplet number concentration
1.4.5 Ice optics parameterization
1.4.6 Radiative flux densities
1.5 Objectives and Outline
2 Definitions
2.1 Radiometric quantities
2.2 Cloud properties
2.2.1 Micro- and macrophysical properties
2.2.2 Single-scattering optical properties
2.2.3 Volumetric optical properties
2.3 Radiative transfer equation
3 Observations
3.1 Airborne measurement campaign
3.1.1 Radiation measurements
3.1.2 Cloud remote sensing
3.1.3 Cloud in situ observations
3.2 Satellite observations
4 Radiative transfer simulations
4.1 Input from the Integrated Forecasting System
4.2 ecRad scheme
4.3 libRadtran
5 Evaluation strategy
5.1 Comparison approach
5.2 Surface classification
5.3 Solver comparison and selection
5.4 Handling of scale mismatch
5.5 Comparison metrics
6 Comparison of measured and simulated broadband solar irradiances
6.1 Reference comparison
6.2 Sensitivity studies
6.2.1 Sea ice albedo
6.2.2 Cloud fraction
6.2.3 Macro- and microphysical cloud properties
6.2.3.1 Ice water path
6.2.3.2 Liquid water path
6.2.3.3 Cloud droplet number concentration
6.2.4 Interactions between properties
6.3 Additional effects
6.3.1 Liquid water path variability
6.3.2 Three-dimensional radiative effects
7 Impact of ice optics parameterization
7.1 Case study
7.1.1 Regional ecRad simulations
7.1.2 Radar reflectivity approach
7.1.3 Satellite cloud retrieval approach
7.1.4 Comparison of measured and simulated band-integrated solar irradiances
7.2 Statistical analysis
7.2.1 Upward solar irradiance bias
7.2.2 Influence of cloud phase profile
8 Summary, conclusions and outlook
8.1 Comparison of measured and simulated broadband irradiances
8.2 Ice optics parameterization
8.3 Outlook
References
List of symbols
List of abbreviations
List of figures
List of tablesForecasts based on numerical weather prediction models are particularly uncertain in the Arctic, which is characterized by enhanced warming compared to the rest of the globe. Surface characteristics of open water or sea ice in the Arctic Ocean as well as clouds and their properties are major sources of uncertainty in forecasting the radiative energy budget and the resulting heating and cooling of the atmosphere. In this thesis, a systematic comparison of airborne observations collected during the Arctic CLoud Observations Using airborne measurements during polar Day (ACLOUD) campaign with radiative transfer simulations using the ecRad radiation scheme by the European Centre for Medium-Range Weather Forecasts (ECMWF) is performed.
The first part of this thesis evaluates the solar irradiances simulated by the Integrated Forecasting System (IFS) of ECMWF. While a satisfactory agreement between the IFS simulations and observations is found in the downward direction, upward solar irradiances reflected by the surface and clouds significantly disagree. Above sea ice, the IFS underestimates the observations by 35 Wm−2 and above open ocean, it overestimates the measurements by 28 Wm−2. To identify possible causes for the discrepancy, a comprehensive sensitivity analysis covering various surface and macro- and microphysical cloud properties is carried out. This sensitivity study is
realized by replacing operational parameterizations and prognostic cloud variables in ecRad with ACLOUD observations. The misrepresentation of the sea ice albedo is identified as the primary contributor to the simulation-measurement bias, accounting for more than 50 % of it. Further significant contributors above sea ice are the too-low variability in cloud droplet number concentration and the too-low liquid water path in the IFS. Above open ocean, the most significant bias contributions are linked to a too-high liquid water path and a cloud fraction misrepresentation in the IFS.
The second part of this thesis investigates the impact of the ice optics parameterization on the reflected solar irradiance by mixed-phase clouds in IFS. A satellite-based ice water path retrieval product is included to compare ACLOUD observations of band-integrated solar irradiance with radiative transfer simulations which apply different ice optics parameterizations. The (relatively) largest differences in upward solar irradiance occur for the near-infrared wavelength bands and
contribute to the simulations underestimating the observations by 10–13 Wm−2. This underestimation is reduced but not fully compensated by applying ice optics parameterizations which include a mixture of ice crystal habits and surface roughness. Non-operational ice optics parameterizations increase the reflected solar irradiance for ice-dominated and liquid-dominated cloud tops during ACLOUD by 35–68 Wm−2 and 3–7 Wm−2, respectively.:1 Introduction
1.1 Arctic weather and climate
1.2 Mixed-phase clouds
1.3 Ice optics parameterization
1.4 Previous model evaluations
1.4.1 Sea ice albedo
1.4.2 Cloud fraction
1.4.3 Liquid and ice water path
1.4.4 Cloud droplet number concentration
1.4.5 Ice optics parameterization
1.4.6 Radiative flux densities
1.5 Objectives and Outline
2 Definitions
2.1 Radiometric quantities
2.2 Cloud properties
2.2.1 Micro- and macrophysical properties
2.2.2 Single-scattering optical properties
2.2.3 Volumetric optical properties
2.3 Radiative transfer equation
3 Observations
3.1 Airborne measurement campaign
3.1.1 Radiation measurements
3.1.2 Cloud remote sensing
3.1.3 Cloud in situ observations
3.2 Satellite observations
4 Radiative transfer simulations
4.1 Input from the Integrated Forecasting System
4.2 ecRad scheme
4.3 libRadtran
5 Evaluation strategy
5.1 Comparison approach
5.2 Surface classification
5.3 Solver comparison and selection
5.4 Handling of scale mismatch
5.5 Comparison metrics
6 Comparison of measured and simulated broadband solar irradiances
6.1 Reference comparison
6.2 Sensitivity studies
6.2.1 Sea ice albedo
6.2.2 Cloud fraction
6.2.3 Macro- and microphysical cloud properties
6.2.3.1 Ice water path
6.2.3.2 Liquid water path
6.2.3.3 Cloud droplet number concentration
6.2.4 Interactions between properties
6.3 Additional effects
6.3.1 Liquid water path variability
6.3.2 Three-dimensional radiative effects
7 Impact of ice optics parameterization
7.1 Case study
7.1.1 Regional ecRad simulations
7.1.2 Radar reflectivity approach
7.1.3 Satellite cloud retrieval approach
7.1.4 Comparison of measured and simulated band-integrated solar irradiances
7.2 Statistical analysis
7.2.1 Upward solar irradiance bias
7.2.2 Influence of cloud phase profile
8 Summary, conclusions and outlook
8.1 Comparison of measured and simulated broadband irradiances
8.2 Ice optics parameterization
8.3 Outlook
References
List of symbols
List of abbreviations
List of figures
List of table
Matrix functions in multiplex network analysis
No full textThe fast and accurate approximation of matrix functions and their action on vectors is a core problem in numerical linear algebra. Prominently, the matrix exponential fueled research on matrix functions due to applications in differential equations. Krylov subspace methods are among the most efficient numerical methods for the approximation of the action of a matrix function on a vector. They produce (near-) optimal polynomial or rational approximations for a wide class of matrix functions by interpolating the scalar function at approximate eigenvalues of the matrix. Their computational efficiency hinges on fast matrix-vector products in the polynomial and the choice of suitable poles as well as fast linear system solves in the rational case. Complex systems from various scientific disciplines can be modeled by graphs or networks recording pairwise interactions between a finite set of entities. These complex networks possess natural linear algebraic representations and many fundamental problems in network science can be formulated as classical numerical linear algebra problems such as eigenvalue problems, linear systems, or matrix function expressions. In recent years, generalized network models such as multiplex networks, which record different types of relationships or interactions between the same entities in different layers have received considerable attention due to their increased flexibility in modeling complex real-world phenomena. Multiplex networks can be represented by structured matrices and the generalization of well-studied single-layer network methods to the multiplex case is an ongoing endeavor in the network science community. In this thesis, we consider several methods for the analysis of structural and dynamical network properties that rely on matrix function expressions and generalize them from single-layer to multiplex networks. We discuss centrality measures, the solution of stiff systems of non-linear differential equations with exponential Runge--Kutta integrators, as well as un- and semi-supervised community detection---putting a special focus on the efficient numerical treatment of the problems. Besides leveraging standard Krylov techniques, we present novel combinations of computational methods and advance state-of-the-art methodology. For instance, we prove an a-posteriori error estimate for rational Krylov approximations of the action of the matrix exponential on vectors, which enables a novel adaptive rational Krylov procedure with approximately constant iteration numbers and a near-linear scaling of the runtime with respect to the problem size. We present highly scalable linear-complexity methods for all considered problems, which allows insights into large-scale multiplex networks from social, transport, and imaging applications
