Refine
Year of publication
- 2018 (90) (remove)
Document Type
- Doctoral Thesis (51)
- Article (28)
- Conference Proceeding (5)
- Master's Thesis (3)
- Preprint (3)
Language
- English (90) (remove)
Has Fulltext
- yes (90)
Keywords
- Visualization (3)
- Evaluation (2)
- classification (2)
- iron (2)
- machine learning (2)
- 1D-CFD (1)
- 2D-CFD (1)
- ADAS (1)
- Addukt (1)
- Algorithmic Differentiation (1)
Faculty / Organisational entity
- Kaiserslautern - Fachbereich Informatik (18)
- Kaiserslautern - Fachbereich Mathematik (18)
- Kaiserslautern - Fachbereich Maschinenbau und Verfahrenstechnik (13)
- Kaiserslautern - Fachbereich Biologie (11)
- Kaiserslautern - Fachbereich Elektrotechnik und Informationstechnik (7)
- Kaiserslautern - Fachbereich Physik (7)
- Kaiserslautern - Fachbereich Sozialwissenschaften (6)
- Kaiserslautern - Fachbereich Chemie (5)
- Kaiserslautern - Fachbereich Wirtschaftswissenschaften (3)
- Kaiserslautern - Fachbereich Raum- und Umweltplanung (2)
Postmortem Analysis of Decayed Online Social Communities: Cascade Pattern Analysis and Prediction
(2018)
Recently, many online social networks, such as MySpace, Orkut, and Friendster, have faced inactivity decay of their members, which contributed to the collapse of these networks. The reasons, mechanics, and prevention mechanisms of such inactivity decay are not fully understood. In this work, we analyze decayed and alive subwebsites from the Stack Exchange platform. The analysis mainly focuses on the inactivity cascades that occur among the members of these communities. We provide measures to understand the decay process and statistical analysis to extract the patterns that accompany the inactivity decay. Additionally, we predict cascade size and cascade virality using machine learning. The results of this work include a statistically significant difference of the decay patterns between the decayed and the alive subwebsites. These patterns are mainly cascade size, cascade virality, cascade duration, and cascade similarity. Additionally, the contributed prediction framework showed satisfactorily prediction results compared to a baseline predictor. Supported by empirical evidence, the main findings of this work are (1) there are significantly different decay patterns in the alive and the decayed subwebsites of the Stack Exchange; (2) the cascade’s node degrees contribute more to the decay process than the cascade’s virality, which indicates that the expert members of the Stack Exchange subwebsites were mainly responsible for the activity or inactivity of the Stack Exchange subwebsites; (3) the Statistics subwebsite is going through decay dynamics that may lead to it becoming fully-decayed; (4) the decay process is not governed by only one network measure, it is better described using multiple measures; (5) decayed subwebsites were originally less resilient to inactivity decay, unlike the alive subwebsites; and (6) network’s structure in the early stages of its evolution dictates the activity/inactivity characteristics of the network.
Optical Character Recognition (OCR) system plays an important role in digitization of data acquired as images from a variety of sources. Although the area is very well explored for Latin languages, some of the languages based on Arabic cursive script are not yet explored. It is due to many factors: Most importantly are the unavailability of proper data sets and complexities posed by cursive scripts. The Pashto language is one of such languages which needs considerable exploration towards OCR. In order to develop such an OCR system, this thesis provides a pioneering study that explores deep learning for the Pashto language in the field of OCR.
The Pashto language is spoken by more than $50$ million people across the world, and it is an active medium both for oral as well as written communication. It is associated with rich literary heritage and contains huge written collection. These written materials present contents of simple to complex nature, and layouts from hand-scribed to printed text. The Pashto language presents mainly two types of complexities (i) generic w.r.t. cursive script, (ii) specific w.r.t. Pashto language. Generic complexities are cursiveness, context dependency, and breaker character anomalies, as well as space anomalies. Pashto specific complexities are variations in shape for a single character and shape similarity for some of the additional Pashto characters. Existing research in the area of Arabic OCR did not lead to an end-to-end solution for the mentioned complexities and therefore could not be generalized to build a sophisticated OCR system for Pashto.
The contribution of this thesis spans in three levels, conceptual level, data level, and practical level. In the conceptual level, we have deeply explored the Pashto language and identified those characters, which are responsible for the challenges mentioned above. In the data level, a comprehensive dataset is introduced containing real images of hand-scribed contents. The dataset is manually transcribed and has the most frequent layout patterns associated with the Pashto language. The practical level contribution provides a bridge, in the form of a complete Pashto OCR system, and connects the outcomes of the conceptual and data levels contributions. The practical contribution comprises of skew detection, text-line segmentation, feature extraction, classification, and post-processing. The OCR module is more strengthened by using deep learning paradigm to recognize Pashto cursive script by the framework of Recursive Neural Networks (RNN). Proposed Pashto text recognition is based on Long Short-Term Memory Network (LSTM) and realizes a character recognition rate of $90.78\%$ on Pashto real hand-scribed images. All these contributions are integrated into an application to provide a flexible and generic End-to-End Pashto OCR system.
The impact of this thesis is not only specific to the Pashto language, but it is also beneficial to other cursive languages like Arabic, Urdu, and Persian e.t.c. The main reason is the Pashto character set, which is a superset of Arabic, Persian, and Urdu languages. Therefore, the conceptual contribution of this thesis provides insight and proposes solutions to almost all generic complexities associated with Arabic, Persian, and Urdu languages. For example, an anomaly caused by breaker characters is deeply analyzed, which is shared among 70 languages, mainly use Arabic script. This thesis presents a solution to this issue and is equally beneficial to almost all Arabic like languages.
The scope of this thesis has two important aspects. First, a social impact, i.e., how a society may benefit from it. The main advantages are to bring the historical and almost vanished document to life and to ensure the opportunities to explore, analyze, translate, share, and understand the contents of Pashto language globally. Second, the advancement and exploration of the technical aspects. Because, this thesis empirically explores the recognition and challenges which are solely related to the Pashto language, both regarding character-set and the materials which present such complexities. Furthermore, the conceptual and practical background of this thesis regarding complexities of Pashto language is very beneficial regarding OCR for other cursive languages.
This research explores the development of web based reference software for
characterisation of surface roughness for two-dimensional surface data. The reference software used for verification of surface characteristics makes the evaluation methods easier for clients. The algorithms used in this software
are based on International ISO standards. Most software used in industrial measuring
instruments may give variations in the parameters calculated due to numerical changes in
calculation. Such variations can be verified using the proposed reference software.
The evaluation of surface roughness is carried out in four major steps: data capture, data
align, data filtering and parameter calculation. This work walks through each of these steps
explaining how surface profiles are evaluated by pre-processing steps called fitting and
filtering. The analysis process is then followed by parameter evaluation according to DIN EN
ISO 4287 and DIN EN ISO 13565-2 standards to extract important information from the
profile to characterise surface roughness.
The growing computational power enables the establishment of the Population Balance Equation (PBE)
to model the steady state and dynamic behavior of multiphase flow unit operations. Accordingly, the twophase
flow
behavior inside liquid-liquid extraction equipment is characterized by different factors. These
factors include: interactions among droplets (breakage and coalescence), different time scales due to the
size distribution of the dispersed phase, and micro time scales of the interphase diffusional mass transfer
process. As a result of this, the general PBE has no well known analytical solution and therefore robust
numerical solution methods with low computational cost are highly admired.
In this work, the Sectional Quadrature Method of Moments (SQMOM) (Attarakih, M. M., Drumm, C.,
Bart, H.-J. (2009). Solution of the population balance equation using the Sectional Quadrature Method of
Moments (SQMOM). Chem. Eng. Sci. 64, 742-752) is extended to take into account the continuous flow
systems in spatial domain. In this regard, the SQMOM is extended to solve the spatially distributed
nonhomogeneous bivariate PBE to model the hydrodynamics and physical/reactive mass transfer
behavior of liquid-liquid extraction equipment. Based on the extended SQMOM, two different steady
state and dynamic simulation algorithms for hydrodynamics and mass transfer behavior of liquid-liquid
extraction equipment are developed and efficiently implemented. At the steady state modeling level, a
Spatially-Mixed SQMOM (SM-SQMOM) algorithm is developed and successfully implemented in a onedimensional
physical spatial domain. The integral spatial numerical flux is closed using the mean mass
droplet diameter based on the One Primary and One Secondary Particle Method (OPOSPM which is the
simplest case of the SQMOM). On the other hand the hydrodynamics integral source terms are closed
using the analytical Two-Equal Weight Quadrature (TEqWQ). To avoid the numerical solution of the
droplet rise velocity, an analytical solution based on the algebraic velocity model is derived for the
particular case of unit velocity exponent appearing in the droplet swarm model. In addition to this, the
source term due to mass transport is closed using OPOSPM. The resulting system of ordinary differential
equations with respect to space is solved using the MATLAB adaptive Runge–Kutta method (ODE45). At
the dynamic modeling level, the SQMOM is extended to a one-dimensional physical spatial domain and
resolved using the finite volume method. To close the mathematical model, the required quadrature nodes
and weights are calculated using the analytical solution based on the Two Unequal Weights Quadrature
(TUEWQ) formula. By applying the finite volume method to the spatial domain, a semi-discreet ordinary
differential equation system is obtained and solved. Both steady state and dynamic algorithms are
extensively validated at analytical, numerical, and experimental levels. At the numerical level, the
predictions of both algorithms are validated using the extended fixed pivot technique as implemented in
PPBLab software (Attarakih, M., Alzyod, S., Abu-Khader, M., Bart, H.-J. (2012). PPBLAB: A new
multivariate population balance environment for particulate system modeling and simulation. Procedia
Eng. 42, pp. 144-562). At the experimental validation level, the extended SQMOM is successfully used
to model the steady state hydrodynamics and physical and reactive mass transfer behavior of agitated
liquid-liquid extraction columns under different operating conditions. In this regard, both models are
found efficient and able to follow liquid extraction column behavior during column scale-up, where three
column diameters were investigated (DN32, DN80, and DN150). To shed more light on the local
interactions among the contacted phases, a reduced coupled PBE and CFD framework is used to model
the hydrodynamic behavior of pulsed sieve plate columns. In this regard, OPOSPM is utilized and
implemented in FLUENT 18.2 commercial software as a special case of the SQMOM. The dropletdroplet
interactions
(breakage
and
coalescence)
are
taken
into
account
using
OPOSPM,
while
the
required
information
about
the
velocity
field
and
energy
dissipation
is
calculated
by
the
CFD
model.
In
addition
to
this,
the proposed coupled OPOSPM-CFD framework is extended to include the mass transfer. The
proposed framework is numerically tested and the results are compared with the published experimental
data. The required breakage and coalescence parameters to perform the 2D-CFD simulation are estimated
using PPBLab software, where a 1D-CFD simulation using a multi-sectional gird is performed. A very
good agreement is obtained at the experimental and the numerical validation levels.
The transfer of substrates between to enzymes within a biosynthesis pathway is an effective way to synthesize the specific product and a good way to avoid metabolic interference. This process is called metabolic channeling and it describes the (in-)direct transfer of an intermediate molecule between the active sites of two enzymes. By forming multi-enzyme cascades the efficiency of product formation and the flux is elevated and intermediate products are transferred and converted in a correct manner by the enzymes.
During tetrapyrrole biosynthesis several substrate transfer events occur and are prerequisite for an optimal pigment synthesis. In this project the metabolic channeling process during the pink pigment phycoerythrobilin (PEB) was investigated. The responsible ferredoxin-dependent bilin reductases (FDBR) for PEB formation are PebA and PebB. During the pigment synthesis the intermediate molecule 15,16-dihydrobiliverdin (DHBV) is formed and transferred from PebA to PebB. While in earlier studies a metabolic channeling of DHBV was postulated, this work revealed new insights into the requirements of this protein-protein interaction. It became clear, that the most important requirement for the PebA/PebB interaction is based on the affinity to their substrate/product DHBV. The already high affinity of both enzymes to each other is enhanced in the presence of DHBV in the binding pocket of PebA which leads to a rapid transfer to the subsequent enzyme PebB. DHBV is a labile molecule and needs to be rapidly channeled in order to get correctly further reduced to PEB. Fluorescence titration experiments and transfer assays confirmed the enhancement effect of DHBV for its own transfer.
More insights became clear by creating an active fusion protein of PebA and PebB and comparing its reaction mechanism with standard FDBRs. This fusion protein was able to convert biliverdin IXα (BV IXα) to PEB similar to the PebS activity, which also can convert BV IXα via DHBV to PEB as a single enzyme. The product and intermediate of the reaction were identified via HPLC and UV-Vis spectroscopy.
The results of this work revealed that PebA and PebB interact via a proximity channeling process where the intermediate DHBV plays an important role for the interaction. It also highlights the importance of substrate channeling in the synthesis of PEB to optimize the flux of intermediates through this metabolic pathway.
The extraction kinetics of polyphenols, which are leached from red vine leaves, are studied and evaluated using a laboratory robot and nonconventional processing techniques such as ultrasonic (US)-, microwave (MW)-, and pulsed electric field (PEF)-assisted extraction processes. The robotic high-throughput screening reveals optimal extraction conditions at a pH value of 2.5, a temperature of 56 °C, and a solvent mixture of methanol:water:HCl of 50:49:1 v/v/v. Nonconventional processing techniques, such as MW- and US-assisted extraction, have the fastest kinetics and produce the highest polyphenol yield. The non-conventional techniques yield is 2.29 g/L (MW) resp. 2.47 g/L (US) for particles that range in size from 450 to 2000 µm and 2.20 g/L (MW) resp. 2.05 g/L (US) for particles that range from 2000 to 4000 µm. PEF has the lowest yield of polyphenols with 0.94 g/L (450–2000 µm), resp. 0.64 g/L (2000–4000 µm) in comparison to 1.82 g/L (2000 to 4000 µm) in a standard stirred vessel (50 °C). When undried red vine leaves (2000 to 4000 µm) are used the total phenol content is 1.44 g/L with PEF.
Epoxy belongs to a category of high-performance thermosetting polymers which have been used extensively in industrial and consumer applications. Highly cross-linked epoxy polymers offer excellent mechanical properties, adhesion, and chemical resistance. However, unmodified epoxies are prone to brittle fracture and crack propagation due to their highly crosslinked structure. As a result, epoxies are normally toughened to ensure the usability of these materials in practical applications.
This research work focuses on the development of novel modified epoxy matrices, with enhanced mechanical, fracture mechanical and thermal properties, suitable to be processed by filament winding technology, to manufacture composite based calender roller covers with improved performance in comparison to commercially available products.
In the first stage, a neat epoxy resin (EP) was modified using three different high functionality epoxy resins with two type of hardeners i.e. amine-based (H1) and anhydride-based (H2). Series of hybrid epoxy resins were obtained by systematic variation of high functionality epoxy resin contents with reference epoxy system. The resulting matrices were characterized by their tensile properties and the best system was chosen from each hardener system i.e. amine and anhydride. For tailored amine based system (MEP_H1) 14 % improvement was measured for bulk samples similarly, for tailored anhydride system (MEP_H2) 11 % improvement was measured when tested at 23 °C.
Further, tailored epoxy systems (MEP_H1 and MEP_H2) were modified using specially designed block copolymer (BCP), and core-shell rubber nanoparticles (CSR). Series of nanocomposites were obtained by systematic variation of filler contents. The resulting matrices were extensively characterized qualitatively and quantitatively to reveal the effect of each filler on the polymer properties. It was shown that the BCP confer better fracture properties to the epoxy resin at low filler loading without losing the other mechanical properties. These characteristics were accompanied by ductility and temperature stability. All composites were tested at 23 °C and at 80 °C to understand the effect of temperature on the mechanical and fracture properties.
Examinations on fractured specimen surfaces provided information about the mechanisms responsible for reinforcement. Nanoparticles generate several energy dissipating mechanisms in the epoxy, e.g. plastic deformation of the matrix, cavitation, void growth, debonding and crack pinning. These were closely related to the microstructure of the materials. The characteristic of the microstructure was verified by microscopy methods (SEM and AFM). The microstructure of neat epoxy hardener system was strongly influenced by the nanoparticles and the resulting interfacial interactions. The interaction of nanoparticles with a different hardener system will result in different morphology which will ultimately influence the mechanical and fracture mechanical properties of the nanocomposites. Hybrid toughening using a combination of the block-copolymer / core-shell rubber nanoparticles and block copolymer / TiO2 nanoparticles has been investigated in the epoxy systems. It was found out that addition of rigid phase with a soft phase recovers the loss of strength in the nanocomposites caused by a softer phase.
In order to clarify the relevant relationships, the microstructural and mechanical properties were correlated. The Counto’s, Halpin-Tsai, and Lewis-Nielsen equations were used to calculate the modulus of the composites and predicted modulus fit well with the measured values. Modeling was done to predict the toughening contribution from block copolymers and core-shell rubber nanoparticles. There was good agreement between the predicted values and the experimental values for the fracture energy.
The core muscles play a central role in stabilizing the head during headers in soccer. The objective of this study was to examine the influence of a fatigued core musculature on the acceleration of the head during jump headers and run headers. Acceleration of the head was measured in a pre-post-design in 68 soccer players (age: 21.5 ± 3.8 years, height: 180.0 ± 13.9 cm, weight: 76.9 ± 8.1 kg). Data were recorded by means of a telemetric 3D acceleration sensor and with a pendulum header. The treatment encompassed two exercises each for the ventral, lateral, and dorsal muscle chains. The acceleration of the head between pre- and post-test was reduced by 0.3 G (p = 0.011) in jump headers and by 0.2 G (p = 0.067) in run headers. An additional analysis of all pretests showed an increased acceleration in run headers when compared to stand headers (p < 0.001) and jump headers (p < 0.001). No differences were found in the sub-group comparisons: semi-professional vs. recreational players, offensive vs. defensive players. Based on the results, we conclude that the acceleration of the head after fatiguing the core muscles does not increase, which stands in contrast to postulated expectations. More tests with accelerated soccer balls are required for a conclusive statement.
Fast Internet content delivery relies on two layers of caches on the request path. Firstly, content delivery networks (CDNs) seek to answer user requests before they traverse slow Internet paths. Secondly, aggregation caches in data centers seek to answer user requests before they traverse slow backend systems. The key challenge in managing these caches is the high variability of object sizes, request patterns, and retrieval latencies. Unfortunately, most existing literature focuses on caching with low (or no) variability in object sizes and ignores the intricacies of data center subsystems.
This thesis seeks to fill this gap with three contributions. First, we design a new caching system, called AdaptSize, that is robust under high object size variability. Second, we derive a method (called Flow-Offline Optimum or FOO) to predict the optimal cache hit ratio under variable object sizes. Third, we design a new caching system, called RobinHood, that exploits variances in retrieval latencies to deliver faster responses to user requests in data centers.
The techniques proposed in this thesis significantly improve the performance of CDN and data center caches. On two production traces from one of the world's largest CDN AdaptSize achieves 30-91% higher hit ratios than widely-used production systems, and 33-46% higher hit ratios than state-of-the-art research systems. Further, AdaptSize reduces the latency by more than 30% at the median, 90-percentile and 99-percentile.
We evaluate the accuracy of our FOO analysis technique on eight different production traces spanning four major Internet companies.
We find that FOO's error is at most 0.3%. Further, FOO reveals that the gap between online policies and OPT is much larger than previously thought: 27% on average, and up to 43% on web application traces.
We evaluate RobinHood with production traces from a major Internet company on a 50-server cluster. We find that RobinHood improves the 99-percentile latency by more than 50% over existing caching systems.
As load imbalances grow, RobinHood's latency improvement can be more than 2x. Further, we show that RobinHood is robust against server failures and adapts to automatic scaling of backend systems.
The results of this thesis demonstrate the power of guiding the design of practical caching policies using mathematical performance models and analysis. These models are general enough to find application in other areas of caching design and future challenges in Internet content delivery.
Optimal control of partial differential equations is an important task in applied mathematics where it is used in order to optimize, for example, industrial or medical processes. In this thesis we investigate an optimal control problem with tracking type cost functional for the Cattaneo equation with distributed control, that is, \(\tau y_{tt} + y_t - \Delta y = u\). Our focus is on the theoretical and numerical analysis of the limit process \(\tau \to 0\) where we prove the convergence of solutions of the Cattaneo equation to solutions of the heat equation.
We start by deriving both the Cattaneo and the classical heat equation as well as introducing our notation and some functional analytic background. Afterwards, we prove the well-posedness of the Cattaneo equation for homogeneous Dirichlet boundary conditions, that is, we show the existence and uniqueness of a weak solution together with its continuous dependence on the data. We need this in the following, where we investigate the optimal control problem for the Cattaneo equation: We show the existence and uniqueness of a global minimizer for an optimal control problem with tracking type cost functional and the Cattaneo equation as a constraint. Subsequently, we do an asymptotic analysis for \(\tau \to 0\) for both the forward equation and the aforementioned optimal control problem and show that the solutions of these problems for the Cattaneo equation converge strongly to the ones for the heat equation. Finally, we investigate these problems numerically, where we examine the different behaviour of the models and also consider the limit \(\tau \to 0\), suggesting a linear convergence rate.