Following growing public awareness of the danger from hurricanes and tremendous demands for analysis of loss, many researchers have conducted studies to develop hurricane damage analysis methods. Although researchers have identified the significant indicators, there currently is no comprehensive research for identifying the relationship among the vulnerabilities, natural disasters, and economic losses associated with individual buildings. To address this lack of research, this study will identify vulnerabilities and hurricane indicators, develop metrics to measure the influence of economic losses from hurricanes, and visualize the spatial distribution of vulnerability to evaluate overall hurricane damage. This paper has utilized the Geographic Information System (GIS) to facilitate collecting and managing data, and has combined vulnerability factors to assess the financial losses suffered by Texas coastal counties. A multiple linear regression method has been applied to develop hurricane economic damage predicting models. To reflect the pecuniary loss, insured loss payment was used as the dependent variable to predict the actual financial damage and ratio. Geographical vulnerability indicators, built environment vulnerability indicators, and hurricane indicators were all used as independent variables. Accordingly, the models and findings may possibly provide vital references for government agencies, emergency planners, and insurance companies hoping to predict hurricane damage.

Ground motions caused by earthquakes may be strong enough to cause destruction of infrastructure and possibly casualties. If such past destructive earthquakes are analysed, the gained information could be used to develop earthquake warning systems that predicts and possibly reduce the damage potential of further earthquakes. The Swedish National Seismic Network (SNSN) runs an automated early warning system that attempts to predict the damage of an earthquake that just got recorded, and forward the predictions to relevant government agencies. The predictions are based on, e.g. earthquake magnitude, source depth and an estimate of the size of affected human population. The purpose of this thesis is to introduce an additional parameter: earthquake intensity, which is a measure of the intensity with which the ground shakes. Based on this, a new earthquake hazard scheme, the Intensity Based Earthquake Classification (IBEC) scheme, is created. This scheme suggests alternate methods, relative to SNSN, of how earthquake classifications can be made. These methods will use an intensity database established by modelling scenario earthquakes in the open-source software ShakeMap by the U.S. Geological Survey. The database consists of scenarios on the intervals: 4.0 ≤ Mw ≤ 9.0 and 10 ≤ depth ≤ 150 kilometre, and covers the whole intensity scale, Modified Mercalli Intensity, 1.0 ≤ Imm ≤ 10.0. The IBEC classification scheme also enabled the creation of the 'Population-to-Area' criterion. It improves prediction of earthquakes that struck isolated cities, located in e.g. valleys in large mountainous areas and deserts. Even though such earthquakes are relatively uncommon, once they occur, they may cause great damage as many cities in such regions around the world often are less developed regarding resistance to ground motions. / Markrörelser orsakade av jordbävningar kan va starka nog att skada vår infrastruktur och orsaka dödsoffer. Genom att analysera forna destruktiva jordbävningar och utveckla program som försöker att förutsäga deras inverkan så kan den potentiella skada minskas. Svenska Nationella Seismiska Nätet (SNSN) driver ett automatiserat tidigt varningssystem som försöker förutsäga skadorna som följer en jordbävning som precis spelats in, och vidarebefodra denna information till relevanta myndigheter. Förutsägelserna är baserade på, t.ex. jordbävnings-magnitud och djup samt uppskattning av mänsklig population i det påverkade området. Syftet med denna avhandlingen är att introducera ytterligare en parameter: jordbävnings-intensitet, som är ett mått av intensiteten i markrörelserna. Baserat på detta skapas ett jordbävnings-schema kallat Intensity Based Earthquake Classification (IBEC). Detta schema föreslår alternativa metoder, relativt SNSN, för hur jordbävnings-klassificering kan göras. Dessa metoder använder sig av en intensitets-databas etablerad genom modellering av jordbävning-scenarios i open source-\linebreak programmet ShakeMap, skapat av U.S. Geological Survey. Databasen består av scenarior över intervallen 4.0 ≤ Mw ≤ 9.0 och 10 ≤ djup ≤ 150 kilometer, vilka täcker hela intensitetsskalan, Modified Mercalli Intensity, 1.0 ≤ Imm ≤ 10.0. IBECs klassificeringsschema har även möjliggjort skapandet av "Population-mot-Area"-kriteriet. Detta förbättrar förutsägelsen av jordbävningar som träffar isolerade städer, placerade i t.ex. dalgångar i stora bergskjedjor och öknar. Även om denna typ av jordbävningar är relativt ovanliga så orsakar dom ofta enorm skada då sådana här städer ofta är mindre utvecklade rörande byggnaders motstånd mot markrörelser.

This master thesis investigates the use of machine learning for predicting and assessing the risk of railway vehicle component failures. Data used for failure prediction often comes with limitations due to the complex nature of maintenance or sometimes requires investments for the extraction of information. Instead of real-time data, historical data and failure timestamps, easily accessed by organisations, are examined to see if they have the potential to contribute to a more effective maintenance strategy. Datasets used in maintenance often contain censored data and to overcome this problem survival machine learning models were also examined. Therefore both traditional machine learning models and survival machine learning models were evaluated and compared based on their C-index value. The results demonstrate that the survival machine learning models, which incorporate the risk and time-to-event aspects of the data, performed better than the traditional ones regarding the risk ranking of components. Random survival forest had the best result, and a ranking of important features. These findings indicate that there is a potential for survival machine learning, applied to existing historical data used for risk assessment for components failure.

Submitted by Cássia Santos (cassia.bcufg@gmail.com) on 2017-06-29T12:25:26Z No. of bitstreams: 2 Dissertação - Geovanne Caetano Gomes - 2017.pdf: 4000475 bytes, checksum: 9ebf09d429582174011cbfd00331c5d4 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-07-10T14:15:56Z (GMT) No. of bitstreams: 2 Dissertação - Geovanne Caetano Gomes - 2017.pdf: 4000475 bytes, checksum: 9ebf09d429582174011cbfd00331c5d4 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-07-10T14:15:56Z (GMT). No. of bitstreams: 2 Dissertação - Geovanne Caetano Gomes - 2017.pdf: 4000475 bytes, checksum: 9ebf09d429582174011cbfd00331c5d4 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2017-05-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Among the several deleterious actions may attack concrete elements, is the alkali-aggregate reaction (AAR), which affects, mainly, structures of dams, bridges and foundations, where the alkali-silica reaction (ASR) is the most common. One of the main challenges regarding the prediction of this phenomenon is the development of models that may predict damage specific for this reaction, which constitute the theme of this research. Firstly, a systematic literature review was conducted with respect to the models developed, with the organization and classification of the data found, presenting a clear and detailed state-of-art. Therefore, the studies published in journals in the last five years (2012-2016) were selected, in order to conduct their categorization regarding the scale and nature of the analysis, type of modeling, and the software necessary to execute the simulations, besides the summarizing, grouping and analysis of the information concerning the input data necessary to the execution of each modeling, as well as the results generated by each one of them. The models which do not predict damage, i.e. general models that simulate the ASR, were investigated to verify their contribution to a better understanding of the chemical and physical processes that occur in the concrete affect by the reaction. Finally, it was verified that the models analyzed are based on different theories and methods of analyses, demanding distinct input data and generating heterogeneous output data, which are meticulously explained in this paper. / Dentre as várias ações deletérias que podem atacar elementos de concreto tem-se a reação álcali-agregado (RAA), a qual afeta principalmente as estruturas de barragens, pontes e fundações, sendo a reação do tipo álcali-sílica (RAS) a mais recorrente nelas. Um dos principais desafios no que tange à predição desse fenômeno é o desenvolvimento de modelos de previsão de dano específicos dessa reação, constituindo-se o tema da presente pesquisa. A priori, executou-se uma revisão sistemática da literatura a respeito dos modelos desenvolvidos, com a organização e classificação dos dados encontrados, apresentando-se o estado da arte de forma clara e detalhada. Em seguida, foram elencados os trabalhos publicados em periódicos indexados nos últimos cinco anos (2012-2016), executando-se a categorização dos modelos quanto à escala e natureza de análise, tipo de modelagem, e softwares necessários para executar as simulações, além da sintetização, agrupamento e análise de informações concernentes aos dados de entrada necessários para a execução de cada modelação, bem como dos resultados gerados por elas. Para os modelos que não preveem dano, isto é, modelos gerais que simulam a RAS, investigou-se sua contribuição para o melhor entendimento dos processos químico-físicos que ocorrem no concreto afetado por ela. Verificou-se, assim, que os modelos analisados são pautados em diferentes teorias e métodos de análise, demandando dados de entrada distintos e gerando dados de saída heterogêneos, os quais são discriminados minuciosamente neste trabalho.