The aim of this research is to investigate the performance of a bi-stable device using a single active element and to design a higher order all optical binary delta-sigma modulator (BΔΣM). A Delta sigma modulator has two important components that require enhancement to achieve robust modulation. The first component is the integrator which accumulates the error and at the same time leaks it. Here, the integrator is a single ring laser consisting of a semiconductor optical amplifier (SOA) and a filter to allow the light frequency of interest into the ring. The other component is the bi-stable device (called Schmitt trigger) that switches either ON (1) or OFF (0). There are different novel approaches to developing a bi-stable circuit. First, the coupled two ring lasers where each ring suppresses each other. Second, a novel idea that considered as a bi-stable device with single active element to achieve reduced power and reduce cost. This type of circuit is merged ring lasers with using single SOA. This system is modeled and its bistability hysteretic characteristics is investigated. The first bi-stable device is used to construct an all optical BΔΣM with 1st, 2nd and 3rd -order approaches. It performs better when the SOA bulk device is replaced by multi-quantum well (MQW) SOA.

The rising fuel prices, environmental concerns over the emission of greenhouse gases, and the limited availability of fossil fuels led to the current focus on developing alternative fuel sources that are sustainable and environmentally benign. Lignocellulosic biomass, due to its high carbon value, abundance and for being greenhouse gas neutral, is a promising alternative energy resource. Fast pyrolysis of lignocellulosic biomass produces high energy density liquid fuel, called bio-oil, which has the potential as transportation fuel. But, crude bio-oils are chemically complex liquids with high oxygen contents (40 % oxygen content), high viscosity, low pH, low thermal stability, and poor heating values (20 MJ/Kg). Therefore, bio-oils must be substantially upgraded (de-oxygenated) to highly stable, non-corrosive, and high calorific value liquid fuels prior to their use as transportation fuels. This research was conducted to investigate the efficiency of various acid catalysts in upgrading (cracking) the oxygenated pine wood pyrolysis vapors to high quality liquid fuel. Initial catalyst screening studies proved that zeolite acidity and pore structure is essential for effective cracking of pyrolysis vapors. Low space velocities and moderate temperatures were found to be favorable for the deoxygenation of pyrolysis vapors. Various zeolites were tested, of which HZSM-5 with low Si/Al ratio was found to be an effective cracking catalyst. But the use of zeolites resulted in poor liquid yields. Zeolites were promoted with transition metal ions in order to inhibit the secondary cracking reactions occurring on Brönsted acid sites. The metal-promoted biunctional catalysts were found to be the most effective catalysts, among all the catalysts employed in this research, in promoting hydrocarbon forming reactions without adversely affecting the liquid yields. Catalyst coking was unavoidable but the addition of metal ions to zeolites lowered the extent of coking. TG analysis of used catalysts indicated that the catalysts can be regenerated by calcining at 600-650 °C.

Random disruptions resulting in loss of functionality in service legs or intermodal terminals of transportation networks are an inevitable part of operations, and considering the crucial role of aforementioned networks, it is prudent to strive towards avoiding high-consequence disruption events. The magnitude of the negative impact of a disruption is dependent on component criticality; therefore, limited resources of disruption mitigation should be assigned to the infrastructure with the highest priority. However, categorizing the service legs and terminals based on their actual post-disruption impact is computationally heavy and inefficient. We propose a methodology based on the combination of a bi-objective hazmat shipment planning optimization model and machine learning to identify critical infrastructure more efficiently. The proposed methodology is applied to part of CSX Corporation’s intermodal rail-truck network in the United States as a realistic size problem instance, in order to gain managerial insight and to evaluate the performance of the methodology. / Thesis / Master of Science (MSc)

Whether it is designing and controlling super-efficient high speed transport systems or understanding environmental fluid flows, a key question that arises is: what state does the fluid take and why? An answer to this question lies in understanding the hydrodynamic stability properties of the flow as a function of parameters. While much work has been done in this area in the past, there are many open questions that need to be addressed. Here we study the effect of spatial domain size, number of modes, non-hermitianness and non-normality on the finite time and infinite time stability properties of a standing, viscous shock flow problem. It has been shown that the above problems are not only non-normal but also non-hermitian, when the base flow has shear. The eigenvalue problems corresponding to infinite spatial domain, finite spatial domain, Forward and L2 adjoint problems are solved exactly by converting the linear partial differential equations into nonlinear Riccati equations. In the finite domain case, the full time dependent solutions are obtained analytically using bi-orthogonal basis functions. In the infinite domain case, the point spectrum of the forward operator is shown to be unbounded and that of the adjoint operator to be empty. In the unbounded case, the spectrum fills the entire area on one side of a parabola in the complex plane and is connected. As the fluid viscosity decreases the width of the parabola increases and in the limit of zero viscosity covers almost entire left half plane(LHP). On the other hand, as the fluid viscosity increases the width of parabola decreases and in the limit of infinite viscosity becomes negative real axis, which is the spectrum of heat equation. The spectrum of adjoint problem is empty for all values of the viscosity and prescribed velocity. In the finite spatial domain case, the point spectrum lies in the open left half plane for all Reynolds numbers and hence asymptotically stable. The results obtained showed that perturbations grow substantially large for finite time before they decay at large times. It is also found that retainig right number of modes is crucial for observing transient growth phenomena. Finally, the linear results are compared with the nonlinear finite amplitude simulation results. The relevance of current results to other fluid flows is presented.

京都大学 / 新制・課程博士 / 博士(理学) / 甲第24393号 / 理博第4892号 / 新制||理||1699(附属図書館) / 京都大学大学院理学研究科数学・数理解析専攻 / (主査)教授 長谷川 真人, 教授 牧野 和久, 准教授 照井 一成 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

The aim of this work is to develop efficient named entity recognition from the given text that in turn improves the performance of the systems that use natural language processing (NLP). The performance of IoT-based devices such as Alexa and Cortana significantly depends upon an efficient NLP model. To increase the capability of the smart IoT devices in comprehending the natural language, named entity recognition (NER) tools play an important role in these devices. In general, the NER is a two-step process that initially the proper nouns are identified from text and then classify them into predefined categories of entities such as person, location, measure, organization and time. NER is often performed as a subtask while processing natural languages which increases the accuracy level of a NLP task. In this paper, we propose deep neural network architecture for named entity recognition for the resource-scarce language Hindi, based on convolutional neural network (CNN), bidirectional long short-term memory (Bi-LSTM) neural network and conditional random field (CRF). In the proposed approach, initially, we use skip-gram word2vec model and GloVe model to represent words in semantic vectors which are further used in different deep neural network-based architectures. In the proposed approach, we use character- and word-level embedding to represent the text that includes information at fine-grained level. Due to the use of character-level embeddings, the proposed model is robust for the out-of-vocabulary words. Experimental results show that the combination of Bi-LSTM, CNN and CRF algorithms performs better as compared to the other baseline methods such as recurrent neural network, long short-term memory and Bi-LSTM individually.

Social oro och sociala riskfaktorer är något som finns i alla samhällen. Social oro är olika typer av utfall eller angrepp på det som samhällets institutioner anser skyddsvärt. Sociala riskfaktorer är däremot de negativa omständigheterna som påverkar att social oro överhuvudtaget uppstår. För att besvara studiens frågeställning: ”Hur kan socioekonomiska och sociodemografiska indikatorer för arbete mot social oro på kommunal nivå, visualiseras i en dashboard-prototyp?” är det framför allt viktigt att ta reda på vilka indikatorer som faktiskt skall ingå i en sådan prototyp samt ta reda på vilka behov kommunerna har och belysa hur en dashboard kan uppfylla dessa behov. Frågeställningen besvaras med en kvalitativ metodansats bestående av datainsamlingsmetoden intervjuer samt en mindre omfattande litteraturundersökning på det material som ingått i litteraturgranskningen innehållande olika indikatorer. Intervjuerna genomfördes för att komma i direkt kontakt med kommunerna och undersöka vilka behov som finns i arbetet i dagsläget, vilka indikatorer som mäts i dagsläget och utöver det ta reda på andra kringliggande frågor som vilken data som finns tillgänglig och till exempel vilka förbättringsmöjligheter som respondenterna tycker sig se i det nuvarande arbetet. Detta för att kunna identifiera ifall en dashboard skulle kunna uppfylla dessa behov. En mindre omfattande litteraturundersökning genomfördes parallellt för att ta reda på vilka indikatorer som förekom i flest antal unika källor i litteraturgranskningen för att på det sättet kunna styrka trovärdigheten om generaliserbarheten för resultatet explicit gällande framtagning av indikatorer. Detta eftersom antalet respondenter ansågs vara för få för att skapa trovärdighet om generaliserbarhet i denna fråga. För att analysera kvalitativa data användes analysmetoden innehållsanalys vilken går ut på att korta ner innehåll till dess innersta kärna genom att identifiera mönster i innehållet som sedan tilldelas olika teman. Mönster och teman kallas i studien undertema och grundtema. För att kunna presentera resultatet av litteraturundersökningen på ett begripligt sätt gjordes innehållsanalysen i Excell. Enskilda indikatorer som ingått i studiens litteraturgranskning kategoriserades där i undertema och grundtema och data visualiserades genom användandet av Tableau. Resultatet av studien är ett antal indikatorer som tillsammans anses utgöra det mest grundläggande behovet av mätningar hos kommuner för att skapa en överblick av sociala riskfaktorer och social oro. Resultatet inkluderar också vilken typ av dashboard layout och vilket syfte dashboarden skall ha inom kommuner och varför. Dessutom skapades en dashboard-prototyp utefter de slutsatser studien kommit fram till.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第22757号 / 工博第4756号 / 新制||工||1744(附属図書館) / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 五十嵐 晃, 教授 高橋 良和, 准教授 古川 愛子 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

No description available.

Involving end-users in system development processes has proven to be critically important for a successful system. This is because a successful system must fulfill the needs of end-users and generate value for them. However, involving the end-users can be challenging and require a large amount of resources. This paper intends to study how end-users are involved in the development processes, whether the needs are met and what challenges participation encounter within a public organization. The methods of choice for data collection are interviews with developers and analysis-owners, as well as a survey for the end-users. This study's result shows that end-users are primarily involved through the analysis-owner, who acts as a representative for the end users. The study also shows that there are many respondents who are satisfied with Power BI in general, that it is easy to use and is useful in their profession. The challenges that were noticed are time resources, development methods and evaluation. Our ambition is that this study can be used for further development within the studied organization.