Indiana University-Purdue University Indianapolis (IUPUI) / Powder bed fusion (PBF) process is a metal additive manufacturing process, which can build parts with any complexity from a wide range of metallic materials. Research in the PBF process predominantly focuses on the impact of a few parameters on the ultimate properties of the printed part. The lack of a systematic approach to optimizing the process parameters for a better performance of given material results in a sub-optimal process limiting the potential of the application. This process needs a comprehensive study of all the influential parameters and their impact on the mechanical and microstructural properties of a fabricated part. Furthermore, there is a need to develop a quantitative system for mapping the material properties and process parameters with the ultimate quality of the fabricated part to achieve improvement in the manufacturing cycle as well as the quality of the final part produced by the PBF process. To address the aforementioned challenges, this research proposes a framework to optimize the process for 316L stainless steel material. This framework characterizes the influence of process parameters on the microstructure and mechanical properties of the fabricated part using a series of experiments. These experiments study the significance of process parameters and their variance as well as study the microstructure and mechanical properties of fabricated parts by conducting tensile, impact, hardness, surface roughness, and densification tests, and ultimately obtain the optimum range of parameters. This would result in a more complete understanding of the correlation between process parameters and part quality. Furthermore, the data acquired from the experiments are employed to develop an intelligent parameter suggestion multi-layer feedforward (FF) backpropagation (BP) artificial neural network (ANN). This network estimates the fabrication time and suggests the parameter setting accordingly to the user/manufacturers desired characteristics of the end-product. Further, research is in progress to evaluate the framework for assemblies and complex part designs and incorporate the results in the network for achieving process repeatability and consistency.

A hospital readmission can occur due to insufficient treatment or the emergence of an underlying disease that was not apparent at the initial hospital stay. The unplanned readmission rate is often viewed as an indicator of the health system performance and may reflect the quality of clinical care provided during hospitalization. Readmissions have also been reported to account for a significant portion of inpatient care expenditures. In an effort to improve treatment quality, clinical outcomes, and hospital operating costs, we present machine learning methods for identifying and predicting potentially preventable readmissions (PPR). In the first part of the thesis, we use logistic regression, extreme gradient boosting, and neural network to predict 30-day unplanned readmissions. In the second part, we apply association rule analysis to assess the clinical association between initial admission and readmission, followed by employing counterfactual analysis to identify potentially preventable readmissions. This comprehensive analysis can assist health care providers in targeting interventions to effectively reduce preventable readmissions. / Thesis / Master of Science (MSc)

Livestock grazing is an important part for the livelihood of a large part of the world’s population. While in some areas of the world water accessibility is often taken for granted, in arid regions this can be a limited resource. The central Asian country of Mongolia is one of the countries that sees excessive livestock grazing in an arid region. Nomadic pastoralism is widespread to ensure access to fresh water sources as well as pastures. These movement patterns are under ongoing research, but so far, the impact of water accessibility on these movement patterns has not been investigated. Specifically in the Great Gobi B Strictly Protected Area, pastoralists rely on water obtained from a variety of sources like lakes, rivers, and wells. I analysed camp usage and the availability of water for pastoralists to uncover current movement patterns and how these are influenced by water. I want to gain insight on how the distance to the closest water source influences camp usage and how this changes with seasonality and the size of the herd. For this I conducted Kruskal-Wallis and two-sided Pearson tests respectively. A potential overlap between wildlife and livestock was investigated since this might be a potential conflict and further leading to the spread of diseases. Furthermore, I modelled precipitation and temperature until the year 2050 to spot a potential redistribution of water in an already arid region. There was no correlation found between the seasons nor herd size and the distance to the closest and second closest water source. In my analysis, the overlap between wildlife and livestock was found only to a minor extend. Precipitation and temperature in the region were found to change only to a marginal degree from 2023 to 2050. Due to this, I identified the most important water sources and camps based on the analysisof the current data.

[pt] Na geotecnia, os parâmetros de resistência do solo são essenciais para qualquer projeto. Os ensaios de campo e laboratório são essenciais, mas ainda enfrentam muitas limitações práticas e financeiras. Além disso, métodos tradicionais, apoiados em relações empíricas ou teóricas, frequentemente não conseguem abranger a complexidade comportamental do solo. Diante disso, destaca-se a necessidade de explorar alternativas para superar essas barreiras. Neste contexto, a inteligência artificial surge como uma abordagem inovadora. Este estudo propõe um modelo preditivo para analisar a curva tensão-deslocamento em ensaios de cisalhamento direto e tensão-deformação em ensaios de cisalhamento simples (Direct Simple Shear - DSS) em areia. Após coletar e digitalizar dados de diversas fontes acadêmicas, formou-se uma base experimental robusta para treinar três algoritmos de Machine Learning (ML): Support Vector Regression (SVR), Random Forest (RF) e Feedforward Neural Network (FNN). Foram realizadas análises comparativas dos modelos, com foco particular na avaliação de métricas de desempenho e curvas dos ensaios de validação. O RF destacou-se por sua precisão e confiabilidade. Embora os modelos SVR e FNN tenham demonstrado utilidade, o RF emergiu como o mais eficaz. Este resultado reforça a viabilidade dos modelos de ML, particularmente o RF, como ferramentas valiosas para engenheiros geotécnicos e pesquisadores na previsão do comportamento de areias, mesmo com um conjunto de dados limitado. / [en] In geotechnics, soil resistance parameters are essential for any project. Field and laboratory tests are essential, but still face many practical and financial limitations. Moreover, traditional methods, relying on empirical or theoretical relationships, often fail to encompass the soil s behavioral complexity. In light of this, there is a highlighted need to explore alternatives to overcome these barriers. In this context, artificial intelligence emerges as an innovative approach. This study proposes a predictive model to analyze the stress-displacement curve in direct shear tests and stress-strain in Direct Simple Shear (DSS) in sand. After collecting and digitizing data from various academic sources, a robust experimental base was formed to train three Machine Learning (ML) algorithms: Support Vector Regression (SVR), Random Forest (RF), and Feedforward Neural Network (FNN). Comparative analyses of the models were conducted, with a particular focus on the evaluation of performance metrics and validation test curves. RF stood out for its accuracy and reliability. Although the SVR and FNN models demonstrated utility, RF emerged as the most effective. This result reinforces the viability of ML models, particularly RF, as valuable tools for geotechnical engineers and researchers in predicting the behavior of sands, even with a limited data set.

Comparison of Dyna-CLUE and Land Change Modeler software for predictive modelling in the suburban area of Prague Abstract The objective of this work was to predict the development of the suburban area of Prague, using Dyna- CLUE and Land Change Modeler (LCM) software, and based on the results compare the capabilities of both these software tools. In this work I used time series of land cover data obtained by the department of applied geoinformatics and cartography, local plans of the municipalities, and information about soil protection provided by the Research Institute for Soil and Water Conservation. Based on these data, a predicted land cover map for 2020 was created in both software tools. The results were then compared and showed that the models respect the restriction of development in predetermined areas. In accordance with local plans, new residential development was properly allocated. For commercial development, the requirements were not completely fulfilled. It is evident that both models are able to create a correct map of future land cover based on specified requirements. However, the instability of LCM and the necessity of using other software while working with Dyna- CLUE somewhat complicated the work. Keywords: Dyna-CLUE, Land Change Modeler, predictive modelling, land cover, residential...

Simulations of various hydration levels of lamellar phase 23:2 Diyne PC were performed, and subsequent, serial docking simulations of a tyrosine monomer were replicated for each system in both hydrated and dehydrated states.<br>The goal was to evaluate how hydration impacts self-assembly and crystallization on the surface, and<br>whether or not these simulations, when run sequentially, could determine the answer. It was discovered that hydrated and dehydrated surfaces behave differently, and that<br>headgroup orientation plays a role in the initial docking and self-assembly process of the tyrosine monomer. It was also determined that potential energy as a sole metric<br>for determining whether or not a specific conformation of intermolecular orientation is not entirely useful, and docking scores are likely useful metrics in discriminating between conformations with identical potential energy values. <br>

Big data applications have become increasingly popular with the emerge of cloud computing and the explosion of artificial intelligence. Hence, the increasing adoption of data-hungry machines and services is driving the need for more power to keep the datacenters of the world running. It has become crucial for large IT companies such as Google, Facebook, Amazon etc. to monitor the energy efficiency of their datacenters’ facilities and take actions on optimization of these heavy consumers of electricity. This master thesis work proposes several predictive models to forecast PUE (Power Usage Effectiveness), regarded as the industry-de-facto metric for measuring datacenter’s IT power efficiency. This approach is a novel capacity management technique to predict and monitor the environment in order to prevent future disastrous events, which are strictly unacceptable in datacenter’s business.

Au cours de ces dernières décennies, les avionneurs n’ont cessé d’augmenter la puissance électrique embarquée à bord des avions. Cette intensification de l’usage de l’électricité, dans le but de rationaliser les énergies secondaires de l’avion (pneumatique, hydraulique, mécanique) constitue le fondement du concept de l’avion plus électrique. Une des contreparties de l’augmentation du nombre de charges électriques réside dans le fait qu’elles doivent fonctionner dans le même environnement électromagnétique, ce qui engendre des problèmes de compatibilité. Cette discipline a été traitée jusqu’à présent en fin de développement d’un système, avant l’étape de la certification et de l’intégration sur avion. La prise en compte de ces contraintes dès la phase de conception, via l’estimation des perturbations électromagnétiques conduites et rayonnées par simulation, peut permettre d’importants gains de temps et de coûts en réduisant les phases d’essais. La première étape de ce projet de recherche est la mise en place d’une approche de modélisation compatible avec les processus d’optimisation. Il est alors indispensable de prendre en compte l’ensemble des sous-systèmes qui composent la chaîne électromécanique, à savoir les RSILs, les câbles, le convertisseur et le moteur. L’approche de modélisation choisie est de type directe ; elle consiste à représenter la chaîne électromécanique dans la base de mode commun par des quadripôles. Ce modèle générique permet d’estimer les courants de mode commun directement dans le domaine fréquentiel en différents points du système. Par ailleurs, afin d’être compétitif vis-à-vis des autres vecteurs d’énergie présents sur avion, la densité de puissance des systèmes électriques doit être drastiquement augmentée. L’introduction des semi conducteurs grands gaps à base de Carbure de Silicium (SiC) permet de contribuer à l’augmentation de la densité de puissance des électroniques de puissance. Cependant, dans ces travaux de thèse, nous veillons à la non régression des performances au niveau système et notamment vis-à-vis de l’impact des émissions électromagnétiques conduites de mode commun. Une fois les modèles en émission établis, diverses solutions de filtrage sont étudiées : filtrage passif externe et interne. Une démarche d’optimisation multi-objectifs (masse, pertes) et multi contraintes (qualité réseau, stabilité, CEM, thermique, etc.) est proposée. Des études de sensibilité mettent en évidence les variables de conception ayant le plus d’impact sur les émissions conduites. Cette approche permet le dimensionnement optimal des composants de l’onduleur (module de puissance, dissipateur, filtres de mode commun et de mode différentiel, paramètres de la commande rapprochée). Les résultats obtenus grâce à l’algorithme génétique employé permettent de construire des courbes de tendance utiles pour l’aide au dimensionnement. / Over the last decades, aircraft manufacturers have not ceased to increase the electrical power on board aircrafts. This intensification of the use of electricity, in order to rationalize the secondary energies of the aircraft, lays the foundation for the concept of the More Electric Aircraft (MEA). One of the counterparts to increasing the number of the electrical loads is that they must operate in the same electromagnetic environment, which creates compatibility issues. This discipline has been treated so far at the end of the development of a system, before the stage of certification and aircraft integration. Taking into account these constraints from the design phase, via the estimationof conducted and radiated electromagnetic disturbances by simulation, significant time and costs savings could be achieved by reducing the test phases. The first step of this research project is the implementation of a modeling approach suitable with optimization processes. It is then essential to take into account all subsystems that form the electromechanical drive, namely the LISNs, the cables, the power converter and the electric motor. The modeling approach chosen is of the direct type; it consists of representing the electromechanical chain in the common mode base by two ports networks. This generic model allows us to estimate common mode currents directly in the frequency domain at different locations. Besides, one of the main challenges associated to MEA is thus to drastically increase the power density of electrical power systems, without compromising on reliability. The development of new Wide Bandgap (WBG) semiconductor technologies made of Silicon Carbide, can significantly increase efficiency, performance and power density of adjustable speed electrical power drive systems. Nevertheless, due to their higher switching speed and voltage overshoot, WBG semiconductors used in power converters of an electromechanical chain may have some drawbacks when it comes to ElectroMagnetic Interference. Understanding the switching behavior of WBG components is necessary in order to keep switching speed and overvoltage at a reasonable level. In this PhD thesis, we ensure that the introduction of this emerging technology does not lead to a regression of performance at system level. Once we establish the conducted emissions models, different filtering solutions have been used: external and internal passive filters. An optimization dedicated to the resolution of a multi-objectives problem (mass, losses) and multi-constraints (quality, stability, EMC, thermal, etc.) in order to minimize the mass of the converter is accomplished. Sensitivity studies led to the identification of the design variables which have the biggest impacts on conducted emissions. This tool allows the optimal sizing of the inverter’s components (power module, heat sink, common mode and differential mode filters, close control parameters). The results obtained thanks to the use of a genetic algorithm make it possible to develop trend curvesfor an inverter sizing.

Machine learning is becoming a part of everyday life and has an indisputable impact across large array of industries. In the financial industry, this impact lies particularly in predictive modelling. The goal of this thesis is to describe the basic principles of artificial intelligence and its subset, machine learning. The most widely used machine learning techniques are outlined both in a theoretical and a practical way. As a result, four models were assembled within the thesis. Results and limitations of each model were discussed and these models were also mutually compared based on their individual per- formance. The evaluation was executed on a real world dataset, provided by Home Credit company. Final performance of machine learning methods, measured by the KS and GINI metrics, was either very comparable or even worse than the performance of a traditional logistic regression. Still, the problem may lie in an insu cient dataset, in the improper data prepara- tion, or in inappropriately used algorithms, not necessarily in the models themselves.

Seabirds are wide-ranging apex-predators and useful bio-indicators of marine systems. Nevertheless, changes are occurring in the marine environment, and seabirds require protection from the deleterious effects of climate change, fisheries, pollution, offshore development, introduced predators and invasive species. The UK supports internationally important populations of seabirds but also has vast wind and wave resources, therefore understanding how seabirds use the marine environment is vital in order to quantify the potential consequences of further exploiting these resources. In this thesis I first describe the range of wave energy converting devices operational or in development in the UK, and review the potential threats and benefits these developments may have for marine birds. I then synthesise data from colony-based surveys with detailed information on population dynamics, foraging ecology and near-colony behaviour, to develop a projection model that identifies important at-sea areas for breeding seabirds. These models show a positive spatial correlation with one of the most intensive at-sea seabird survey datasets, and provide qualitatively similar findings to existing tracking data. This approach has the potential to identify overlap with offshore energy developments, and could be developed to suit a range of species or whole communities and provide a theoretical framework for the study of factors such as colony size regulation. The non-breeding period is a key element of the annual cycle of seabirds and conditions experienced during one season may carry-over to influence the next. Understanding behaviour throughout the annual cycle has implications for both ecological theory and conservation. Bio-logging can provide detailed information on movements away from breeding colonies, and the analysis of stable isotope ratios in body tissues can provide information on foraging during the non-breeding period. I combine these two approaches to describe the migration strategies of northern gannets Morus bassanus breeding at two colonies in the north-west Atlantic, revealing a high degree of both winter site fidelity and dietary consistency between years. These migratory strategies also have carry-over effects with consequences for both body condition and timing of arrival on the breeding grounds. Finally, I investigate the threats posed to seabirds and other marine predators during the non-breeding period by collating information on the distributions of five different species of apex predator wintering in the Northwest African upwelling region. I describe the threat of over-fishing and fisheries bycatch to marine vertebrates in this region, and highlight the need for pelagic marine protected areas to adequately protect migratory animals throughout the annual cycle. In summary, the combination of colony-based studies, bio-logging, stable isotope analysis and modelling techniques can provide a comprehensive understanding of the interactions between individuals and the marine environment over multiple spatial and temporal scales.