Refine
Year of publication
Document Type
- Doctoral Thesis (14)
- Article (5)
- Bachelor Thesis (1)
- Monograph/Edited Volume (1)
Is part of the Bibliography
- yes (21) (remove)
Keywords
- Simulation (21) (remove)
Institute
- Institut für Geowissenschaften (4)
- Institut für Physik und Astronomie (4)
- Institut für Biochemie und Biologie (2)
- Institut für Chemie (2)
- Institut für Umweltwissenschaften und Geographie (2)
- Wirtschaftswissenschaften (2)
- Fachgruppe Betriebswirtschaftslehre (1)
- Hasso-Plattner-Institut für Digital Engineering GmbH (1)
- Hasso-Plattner-Institut für Digital Engineering gGmbH (1)
- Institut für Informatik und Computational Science (1)
Volatile supply and sales markets, coupled with increasing product individualization and complex production processes, present significant challenges for manufacturing companies. These must navigate and adapt to ever-shifting external and internal factors while ensuring robustness against process variabilities and unforeseen events. This has a pronounced impact on production control, which serves as the operational intersection between production planning and the shop- floor resources, and necessitates the capability to manage intricate process interdependencies effectively. Considering the increasing dynamics and product diversification, alongside the need to maintain constant production performances, the implementation of innovative control strategies becomes crucial.
In recent years, the integration of Industry 4.0 technologies and machine learning methods has gained prominence in addressing emerging challenges in production applications. Within this context, this cumulative thesis analyzes deep learning based production systems based on five publications. Particular attention is paid to the applications of deep reinforcement learning, aiming to explore its potential in dynamic control contexts. Analysis reveal that deep reinforcement learning excels in various applications, especially in dynamic production control tasks. Its efficacy can be attributed to its interactive learning and real-time operational model. However, despite its evident utility, there are notable structural, organizational, and algorithmic gaps in the prevailing research. A predominant portion of deep reinforcement learning based approaches is limited to specific job shop scenarios and often overlooks the potential synergies in combined resources. Furthermore, it highlights the rare implementation of multi-agent systems and semi-heterarchical systems in practical settings. A notable gap remains in the integration of deep reinforcement learning into a hyper-heuristic.
To bridge these research gaps, this thesis introduces a deep reinforcement learning based hyper- heuristic for the control of modular production systems, developed in accordance with the design science research methodology. Implemented within a semi-heterarchical multi-agent framework, this approach achieves a threefold reduction in control and optimisation complexity while ensuring high scalability, adaptability, and robustness of the system. In comparative benchmarks, this control methodology outperforms rule-based heuristics, reducing throughput times and tardiness, and effectively incorporates customer and order-centric metrics. The control artifact facilitates a rapid scenario generation, motivating for further research efforts and bridging the gap to real-world applications. The overarching goal is to foster a synergy between theoretical insights and practical solutions, thereby enriching scientific discourse and addressing current industrial challenges.
The degradation of polymers is described by mathematical models based on bond cleavage statistics including the decreasing probability of chain cuts with decreasing average chain length. We derive equations for the degradation of chains under a random chain cut and a chain end cut mechanism, which are compared to existing models. The results are used to predict the influence of internal molecular parameters. It is shown that both chain cut mechanisms lead to a similar shape of the mass or molecular mass loss curve. A characteristic time is derived, which can be used to extract the maximum length of soluble fragments l of the polymer. We show that the complete description is needed to extract the degradation rate constant k from the molecular mass loss curve and that l can be used to design polymers that lose less mechanical stability before entering the mass loss phase.
We investigate models for incremental binary classification, an example for supervised online learning. Our starting point is a model for human and machine learning suggested by E.M.Gold.
In the first part, we consider incremental learning algorithms that use all of the available binary labeled training data in order to compute the current hypothesis. For this model, we observe that the algorithm can be assumed to always terminate and that the distribution of the training data does not influence learnability. This is still true if we pose additional delayable requirements that remain valid despite a hypothesis output delayed in time. Additionally, we consider the non-delayable requirement of consistent learning. Our corresponding results underpin the claim for delayability being a suitable structural property to describe and collectively investigate a major part of learning success criteria. Our first theorem states the pairwise implications or incomparabilities between an established collection of delayable learning success criteria, the so-called complete map. Especially, the learning algorithm can be assumed to only change its last hypothesis in case it is inconsistent with the current training data. Such a learning behaviour is called conservative.
By referring to learning functions, we obtain a hierarchy of approximative learning success criteria. Hereby we allow an increasing finite number of errors of the hypothesized concept by the learning algorithm compared with the concept to be learned. Moreover, we observe a duality depending on whether vacillations between infinitely many different correct hypotheses are still considered a successful learning behaviour. This contrasts the vacillatory hierarchy for learning from solely positive information.
We also consider a hypothesis space located between the two most common hypothesis space types in the nearby relevant literature and provide the complete map.
In the second part, we model more efficient learning algorithms. These update their hypothesis referring to the current datum and without direct regress to past training data. We focus on iterative (hypothesis based) and BMS (state based) learning algorithms. Iterative learning algorithms use the last hypothesis and the current datum in order to infer the new hypothesis.
Past research analyzed, for example, the above mentioned pairwise relations between delayable learning success criteria when learning from purely positive training data. We compare delayable learning success criteria with respect to iterative learning algorithms, as well as learning from either exclusively positive or binary labeled data. The existence of concept classes that can be learned by an iterative learning algorithm but not in a conservative way had already been observed, showing that conservativeness is restrictive. An additional requirement arising from cognitive science research %and also observed when training neural networks is U-shapedness, stating that the learning algorithm does diverge from a correct hypothesis. We show that forbidding U-shapes also restricts iterative learners from binary labeled data.
In order to compute the next hypothesis, BMS learning algorithms refer to the currently observed datum and the actual state of the learning algorithm. For learning algorithms equipped with an infinite amount of states, we provide the complete map. A learning success criterion is semantic if it still holds, when the learning algorithm outputs other parameters standing for the same classifier. Syntactic (non-semantic) learning success criteria, for example conservativeness and syntactic non-U-shapedness, restrict BMS learning algorithms. For proving the equivalence of the syntactic requirements, we refer to witness-based learning processes. In these, every change of the hypothesis is justified by a later on correctly classified witness from the training data. Moreover, for every semantic delayable learning requirement, iterative and BMS learning algorithms are equivalent. In case the considered learning success criterion incorporates syntactic non-U-shapedness, BMS learning algorithms can learn more concept classes than iterative learning algorithms.
The proofs are combinatorial, inspired by investigating formal languages or employ results from computability theory, such as infinite recursion theorems (fixed point theorems).
Precipitation and land use in terms of livestock grazing have been identified as two of the most important drivers structuring the vegetation composition of semi-arid and arid savannas. Savanna research on the impact of these drivers has widely applied the so-called plant functional type (PFT) approach, grouping the vegetation into two or three broad types (here called meta-PFTs): woody plants and grasses, which are sometimes divided into perennial and annual grasses. However, little is known about the response of functional traits within these coarse types towards water availability or livestock grazing. In this study, we extended an existing eco-hydrological savanna vegetation model to capture trait diversity within the three broad meta-PFTs to assess the effects of both grazing and mean annual precipitation (MAP) on trait composition along a gradient of both drivers. Our results show a complex pattern of trait responses to grazing and aridity. The response differs for the three meta-PFTs. From our findings, we derive that trait responses to grazing and aridity for perennial grasses are similar, as suggested by the convergence model for grazing and aridity. However, we also see that this only holds for simulations below a MAP of 500 mm. This combined with the finding that trait response differs between the three meta-PFTs leads to the conclusion that there is no single, universal trait or set of traits determining the response to grazing and aridity. We finally discuss how simulation models including trait variability within meta-PFTs are necessary to understand ecosystem responses to environmental drivers, both locally and globally and how this perspective will help to extend conceptual frameworks of other ecosystems to savanna research.
Die Bienaymé-Galton-Watson Prozesse können für die Untersuchung von speziellen und sich entwickelnden Populationen verwendet werden. Die Populationen umfassen Individuen, welche sich identisch, zufällig, selbstständig und unabhängig voneinander fortpflanzen und die jeweils nur eine Generation existieren. Die n-te Generation ergibt sich als zufällige Summe der Individuen der (n-1)-ten Generation. Die Relevanz dieser Prozesse begründet sich innerhalb der Historie und der inner- und außermathematischen Bedeutung. Die Geschichte der Bienaymé-Galton-Watson-Prozesse wird anhand der Entwicklung des Konzeptes bis heute dargestellt. Dabei werden die Wissenschaftler:innen verschiedener Disziplinen angeführt, die Erkenntnisse zu dem Themengebiet beigetragen und das Konzept in ihren Fachbereichen angeführt haben. Somit ergibt sich die außermathematische Signifikanz. Des Weiteren erhält man die innermathematische Bedeutsamkeit mittels des Konzeptes der Verzweigungsprozesse, welches auf die Bienaymé-Galton-Watson Prozesse zurückzuführen ist. Die Verzweigungsprozesse stellen eines der aussagekräftigsten Modelle für die Beschreibung des Populationswachstums dar. Darüber hinaus besteht die derzeitige Wichtigkeit durch die Anwendungsmöglichkeit der Verzweigungsprozesse und der Bienaymé-Galton-Watson Prozesse innerhalb der Epidemiologie. Es werden die Ebola- und die Corona-Pandemie als Anwendungsfelder angeführt. Die Prozesse dienen als Entscheidungsstütze für die Politik und ermöglichen Aussagen über die Auswirkungen von Maßnahmen bezüglich der Pandemien. Neben den Prozessen werden ebenfalls der bedingte Erwartungswert bezüglich diskreter Zufallsvariablen, die wahrscheinlichkeitserzeugende Funktion und die zufällige Summe eingeführt. Die Konzepte vereinfachen die Beschreibung der Prozesse und bilden somit die Grundlage der Betrachtungen. Außerdem werden die benötigten und weiterführenden Eigenschaften der grundlegenden Themengebiete und der Prozesse aufgeführt und bewiesen. Das Kapitel erreicht seinen Höhepunkt bei dem Beweis des Kritikalitätstheorems, wodurch eine Aussage über das Aussterben des Prozesses in verschiedenen Fällen und somit über die Aussterbewahrscheinlichkeit getätigt werden kann. Die Fälle werden anhand der zu erwartenden Anzahl an Nachkommen eines Individuums unterschieden. Es zeigt sich, dass ein Prozess bei einer zu erwartenden Anzahl kleiner gleich Eins mit Sicherheit ausstirbt und bei einer Anzahl größer als Eins, die Population nicht in jedem Fall aussterben muss. Danach werden einzelne Beispiele, wie der linear fractional case, die Population von Fibroblasten (Bindegewebszellen) von Mäusen und die Entstehungsfragestellung der Prozesse, angeführt. Diese werden mithilfe der erlangten Ergebnisse untersucht und einige ausgewählte zufällige Dynamiken werden im nachfolgenden Kapitel simuliert. Die Simulationen erfolgen durch ein in Python erstelltes Programm und werden mithilfe der Inversionsmethode realisiert. Die Simulationen stellen beispielhaft die Entwicklungen in den verschiedenen Kritikalitätsfällen der Prozesse dar. Zudem werden die Häufigkeiten der einzelnen Populationsgrößen in Form von Histogrammen angebracht. Dabei lässt sich der Unterschied zwischen den einzelnen Fällen bestätigen und es wird die Anwendungsmöglichkeit der Bienaymé-Galton-Watson Prozesse bei komplexeren Problemen deutlich. Histogramme bekräftigen, dass die einzelnen Populationsgrößen nur endlich oft vorkommen. Diese Aussage wurde von Galton aufgeworfen und in der Extinktions-Explosions-Dichotomie verwendet. Die dargestellten Erkenntnisse über das Themengebiet und die Betrachtung des Konzeptes werden mit einer didaktischen Analyse abgeschlossen. Die Untersuchung beinhaltet die Berücksichtigung der Fundamentalen Ideen, der Fundamentalen Ideen der Stochastik und der Leitidee „Daten und Zufall“. Dabei ergibt sich, dass in Abhängigkeit der gewählten Perspektive die Anwendung der Bienaymé-Galton-Watson Prozesse innerhalb der Schule plausibel ist und von Vorteil für die Schüler:innen sein kann. Für die Behandlung wird exemplarisch der Rahmenlehrplan für Berlin und Brandenburg analysiert und mit dem Kernlehrplan Nordrhein-Westfalens verglichen. Die Konzeption des Lehrplans aus Berlin und Brandenburg lässt nicht den Schluss zu, dass die Bienaymé-Galton-Watson Prozesse angewendet werden sollten. Es lässt sich feststellen, dass die zugrunde liegende Leitidee nicht vollumfänglich mit manchen Fundamentalen Ideen der Stochastik vereinbar ist. Somit würde eine Modifikation hinsichtlich einer stärkeren Orientierung des Lehrplans an den Fundamentalen Ideen die Anwendung der Prozesse ermöglichen. Die Aussage wird durch die Betrachtung und Übertragung eines nordrhein-westfälischen Unterrichtsentwurfes für stochastische Prozesse auf die Bienaymé-Galton-Watson Prozesse unterstützt. Darüber hinaus werden eine Concept Map und ein Vernetzungspentagraph nach von der Bank konzipiert um diesen Aspekt hervorzuheben.
Faced with the increasing needs of companies, optimal dimensioning of IT hardware is becoming challenging for decision makers. In terms of analytical infrastructures, a highly evolutionary environment causes volatile, time dependent workloads in its components, and intelligent, flexible task distribution between local systems and cloud services is attractive. With the aim of developing a flexible and efficient design for analytical infrastructures, this paper proposes a flexible architecture model, which allocates tasks following a machine-specific decision heuristic. A simulation benchmarks this system with existing strategies and identifies the new decision maxim as superior in a first scenario-based simulation.
The business problem of having inefficient processes, imprecise process analyses, and simulations as well as non-transparent artificial neuronal network models can be overcome by an easy-to-use modeling concept. With the aim of developing a flexible and efficient approach to modeling, simulating, and optimizing processes, this paper proposes a flexible Concept of Neuronal Modeling (CoNM). The modeling concept, which is described by the modeling language designed and its mathematical formulation and is connected to a technical substantiation, is based on a collection of novel sub-artifacts. As these have been implemented as a computational model, the set of CoNM tools carries out novel kinds of Neuronal Process Modeling (NPM), Neuronal Process Simulations (NPS), and Neuronal Process Optimizations (NPO). The efficacy of the designed artifacts was demonstrated rigorously by means of six experiments and a simulator of real industrial production processes.
Research on novel and advanced biomaterials is an indispensable step towards their applications in desirable fields such as tissue engineering, regenerative medicine, cell culture, or biotechnology. The work presented here focuses on such a promising material: polyelectrolyte multilayer (PEM) composed of hyaluronic acid (HA) and poly(L-lysine) (PLL). This gel-like polymer surface coating is able to accumulate (bio-)molecules such as proteins or drugs and release them in a controlled manner. It serves as a mimic of the extracellular matrix (ECM) in composition and intrinsic properties. These qualities make the HA/PLL multilayers a promising candidate for multiple bio-applications such as those mentioned above. The work presented aims at the development of a straightforward approach for assessment of multi-fractional diffusion in multilayers (first part) and at control of local molecular transport into or from the multilayers by laser light trigger (second part).
The mechanism of the loading and release is governed by the interaction of bioactives with the multilayer constituents and by the diffusion phenomenon overall. The diffusion of a molecule in HA/PLL multilayers shows multiple fractions of different diffusion rate. Approaches, that are able to assess the mobility of molecules in such a complex system, are limited. This shortcoming motivated the design of a novel evaluation tool presented here.
The tool employs a simulation-based approach for evaluation of the data acquired by fluorescence recovery after photobleaching (FRAP) method. In this approach, possible fluorescence recovery scenarios are primarily simulated and afterwards compared with the data acquired while optimizing parameters of a model until a sufficient match is achieved. Fluorescent latex particles of different sizes and fluorescein in an aqueous medium are utilized as test samples validating the analysis results. The diffusion of protein cytochrome c in HA/PLL multilayers is evaluated as well.
This tool significantly broadens the possibilities of analysis of spatiotemporal FRAP data, which originate from multi-fractional diffusion, while striving to be widely applicable. This tool has the potential to elucidate the mechanisms of molecular transport and empower rational engineering of the drug release systems.
The second part of the work focuses on the fabrication of such a spatiotemporarily-controlled drug release system employing the HA/PLL multilayer. This release system comprises different layers of various functionalities that together form a sandwich structure. The bottom layer, which serves as a reservoir, is formed by HA/PLL PEM deposited on a planar glass substrate. On top of the PEM, a layer of so-called hybrids is deposited. The hybrids consist of thermoresponsive poly(N-isopropylacrylamide) (PNIPAM) -based hydrogel microparticles with surface-attached gold nanorods. The layer of hybrids is intended to serve as a gate that controls the local molecular transport through the PEM–solution-interface. The possibility of stimulating the molecular transport by near-infrared (NIR) laser irradiation is being explored.
From several tested approaches for the deposition of hybrids onto the PEM surface, the drying-based approach was identified as optimal. Experiments, that examine the functionality of the fabricated sandwich at elevated temperature, document the reversible volume phase transition of the PEM-attached hybrids while sustaining the sandwich stability. Further, the gold nanorods were shown to effectively absorb light radiation in the tissue- and cell-friendly NIR spectral region while transducing the energy of light into heat. The rapid and reversible shrinkage of the PEM-attached hybrids was thereby achieved. Finally, dextran was employed as a model transport molecule. It loads into the PEM reservoir in a few seconds with the partition constant of 2.4, while it spontaneously releases in a slower, sustained manner. The local laser irradiation of the sandwich, which contains the fluorescein isothiocyanate tagged dextran, leads to a gradual reduction of fluorescence intensity in the irradiated region.
The release system fabricated employs renowned photoresponsivity of the hybrids in an innovative setting. The results of the research are a step towards a spatially-controlled on-demand drug release system that paves the way to spatiotemporally controlled drug release.
The approaches developed in this work have the potential to elucidate the molecular dynamics in ECM and to foster engineering of multilayers with properties tuned to mimic the ECM. The work aims at spatiotemporal control over the diffusion of bioactives and their presentation to the cells.
Existing factories face multiple problems due to their hierarchical structure of decision making and control. Cyber-physical systems principally allow to increase the degree of autonomy to new heights. But which degree of autonomy is really useful and beneficiary? This paper differentiates diverse definitions of autonomy and approaches to determine them. Some experimental findings in a lab environment help to answer the question raised in this paper.
The warm water geothermal reservoir below the village of Waiwera in New Zealand has been known by the native Maori for centuries. Development by the European immigrants began in 1863. Until the year 1969, the warm water flowing from all drilled wells was artesian. Due to overproduction, water up to 50 A degrees C now needs to be pumped to surface. Further, between 1975 and 1976, all warm water seeps on the beach of Waiwera ran dry. Within the context of sustainable water management, hydrogeological models must be developed as part of a management plan. Approaches of varying complexity have been set-up and applied since the 1980s. However, none of the models directly provide all results required for optimal water management. Answers are given simply to parts of the questions, nonetheless improving resource management of the geothermal reservoir.