Atualmente, no meio jurídico, há um reconhecimento implícito de que as provas materiais necessitam de embasamento científico para alcançar a autenticidade imprescindível ao estabelecimento da convicção dos magistrados. A natureza de determinados exames, como a classificação de substâncias proibidas, demandam a utilização de técnicas e saberes oriundos das ciências naturais e tecnológicas. O trabalho pericial deve ser pautado pela cientificidade, com a aplicação de conhecimentos de diversas áreas, dentre as quais está incluída a estatística forense. Neste trabalho foram utilizadas ferramentas estatísticas para avaliar a qualidade e a contribuição dos laudos periciais para os casos envolvendo substâncias ilícitas e correlacionar o conteúdo destes documentos com a sentença judicial. Numa primeira etapa foram analisadas as informações contidas em laudos toxicológicos de drogas, com o intuito de quantificar a qualidade e importância que eles poderiam fornecer em um processo. Para isso foram analisados 1008 documentos oficiais de diversas jurisdições, divididos em 504 conjuntos de laudos preliminares e definitivos do mesmo caso forense A intenção foi apreciar um conjunto heterogêneo de documentos para possibilitar uma melhor análise. A quantificação foi apreciada através de equações empíricas elaboradas. A validação do método ocorreu por análise de dados multivariados. A metodologia empregada demonstrou-se bastante robusta. A segunda fase do trabalho foi aplicar o resultado dos exames da etapa precedente e correlacionar com a decisão judicial. Para tanto, foram esmiuçadas 167 sentenças proferidas em primeira instância e que continham os laudos elencados na primeira fase. A ferramenta utilizada foi a inferência Bayesiana. Os resultados apontaram que os laudos periciais sempre foram essenciais neste tipo de procedimento julgatório. A qualidade dos documentos produzidos encontrava-se entre boa e ótima, avalizada pelo parâmetro \"relevância do laudo\". Alguns aspectos nos documentos poderiam ser aperfeiçoados, como, por exemplo, a inserção de fotografias do material apreendido e/ou imagens alusivas às análises laboratoriais. Estes estudos permitiram estabelecer um valor de corte para a quantificação da qualidade dos laudos, a partir do qual houve 100% de concordância entre o laudo direcionado e a sentença, para casos de condenação onde o suspeito foi considerado traficante. Por fim, a metodologia proposta apresentou potencial promissor e possibilidade de ser utilizada em outros tipos de casos forenses, como, por exemplo, homicídios, suicídios e outros. / There is an implicit recognition in the current legal scenario that material evidences require scientific support in order to achieve the authenticity that the magistrates need for making decisions. The nature of certain exams, such as classification of prohibited substances, requires the use of techniques and knowledge from natural sciences and technology. The forensic work must rely on scientific methods and apply knowledge from several areas, including forensic statistics. The present work used statistic tools to evaluate the quality and the contribution of forensic reports about illegal substances; the goal is to correlate the content of these documents with the court ruling. In the first part we analyzed the information from toxicology reports on drugs, aiming at the quantification of the importance they might bear to court proceedings. We have parsed 1008 official documents from several jurisdictions, divided into 504 sets of preliminary and final reports from the same case. The objective was to evaluate a heterogeneous document set for a better analysis. The quantification was determined from elaborate empiric equations. The validation of the method was performed by multivariate data analysis. The methodology used in the present work has proved very robust. The second part was the application of the results from the previous part and correlation to the court ruling. We have thoroughly examined 167 rulings at first instance that contained the reports cited in the first part. We have used Bayesian inference, and the results indicated that forensic reports were always required in this type of court proceeding. The quality of the documents was considered good or excellent, as stated in the parameter \"relevance of the report\". Some aspects could be improved, for instance, images of collected material evidence or laboratory analytical procedures could be included. These studies allowed establishing a cut-off value for the quantification of the report quality, from which a 100% agreement between the report and the court decision was achieved, in cases where the suspect was found guilty. Finally, the proposed methodology in this work showed a good potential and could be used in other kinds of forensic cases, such as homicide, suicide and other forensic investigations.

This study was aimed at exploring data analysis techniques for generating accurate estimates of the loss in quality of fresh fruits, vegetables and cut flowers in chilled supply chains based on data from advanced sensors. It was motivated by the recent interest in the application of advanced sensors, by emerging concepts in quality controlled logistics, and by the desire to minimise quality losses during transport and storage of the produce. Cut roses were used in this work although the findings will also be applicable to other produce. The literature has reported that whilst temperature was considered to be the most critical post-harvest factor, others such as growing conditions could also be important in the senescence of cut roses. Kinetic modelling was the most commonly used modelling approach for shelf life predictions of foods and perishable produce, but not for estimating vase life (VL) of cut flowers, and so this was explored in this work along with multiple linear regression (MLR) and partial least squares (PLS). As the senescence of cut roses is not fully understood, kinetic modelling could not be implemented directly. Consequently, a novel technique, called Kinetic Linear System (KLS), was developed based on kinetic modelling principles. Simulation studies of shelf life predictions for tomatoes, mushrooms, seasoned soybean sprouts, cooked shrimps and other seafood products showed that the KLS models could effectively replace the kinetic ones. With respect to VL predictions KLS, PLS and MLR were investigated for data analysis from an in-house experiment with cut roses from Cookes Rose Farm (Jersey). The analysis concluded that when the initial and final VLs were available for model calibration, effective estimates of the post-harvest loss in VL of cut roses could be obtained using the post-harvest temperature. Otherwise, when the initial VLs were not available, such effective estimates could not be obtained. Moreover, pre-harvest conditions were shown to correlate with the VL loss but the correlation was too weak to produce or improve an effective estimate of the loss. The results showed that KLS performance was the best while PLS one could be acceptable; but MLR performance was not adequate. In another experiment, boxes of cut roses were transported from a Kenyan farm to a UK distribution centre. Using KLS and PLS techniques, the analysis showed that the growing temperature could be used to obtain effective estimates of the VLs at the farm, at the distribution centre and also the in-transit loss. Further, using post-harvest temperature would lead to a smaller error for the VL at the distribution centre and the VL loss. Nevertheless, the estimates of the VL loss may not be useful practically due to the excessive relative prediction error. Overall, although PLS had a slightly smaller prediction error, KLS worked effectively in many cases where PLS failed, it could handle constraints while PLS could not.In conclusion, KLS and PLS can be used to generate effective estimates of the post-harvest VL loss of cut roses based on post-harvest temperature stresses recorded by advanced sensors. However, the estimates may not be useful practically due to significant relative errors. Alternatively, pre-harvest temperature could be used although it may lead to slightly higher errors. Although PLS had slightly smaller errors KLS was more robust and flexible. Further work is recommended in the objective evaluations of product quality, alternative non-linear techniques and dynamic decision support system.

Die vorliegende Arbeit befaßt sich mit instationären, uni- oder multivariaten Zeitreihen, die bei der Beobachtung komplexer nichtlinearer dynamischer Systeme entstehen und sich der Modellierung durch ein globales Modell entziehen. In vielen natürlichen oder gesellschaftlichen Prozessen kann man jedoch wiederkehrende Phänomene beobachten, die von deren Rhythmen beeinflußt sind; ebenso lassen sich in technischen Prozessen beispielsweise aufgrund einer bedarfsorientierten Steuerung wiederholte, aber nicht periodische Verhaltensweisen ausmachen. Für solche Systeme und Zeitreihen wird deshalb vorgeschlagen, eine partielle Modellierung durch mehrere lokale Modelle vorzunehmen, die wiederkehrende Phänomene in Form zeitlich begrenzter Muster beschreiben. Um den Unwägbarkeiten dieser und sich anschließender Aufgabenstellungen Rechnung zu tragen, werden in dieser Arbeit durchgehend unscharfe Ansätze zur Modellierung von Mustern und ihrer Weiterverarbeitung gewählt und ausgearbeitet. Die Aufgabenstellung der Erkennung von Mustern in fortlaufenden Zeitreihen wird dahingehend verallgemeinert, daß unvollständige, sich noch in Entwicklung befindliche Musterinstanzen erkannt werden können. Basierend auf ebendieser frühzeitigen Erkennung kann der Verlauf der Zeitreihe -- und damit das weitere Systemverhalten -- lokal prognostiziert werden. Auf Besonderheiten und Schwierigkeiten, die sich aus der neuartigen Aufgabe der Online-Erkennung von Mustern ergeben, wird jeweils vermittels geeigneter Beispiele eingegangen, ebenso die praktische Verwendbarkeit des musterbasierten Vorhersageprinzips anhand realer Daten dokumentiert. / This dissertation focuses on non-stationary multivariate time series stemming from the observation of complex nonlinear dynamical systems. While one global model for such systems and time series may not always be feasible, we may observe recurring phenomena (patterns) in some of these time series. These phenomena might, for example, be caused by the rhythms of natural or societal processes, or a demand-oriented control of technical processes. For such systems and time series a partial modelling by means of multiple local models is being proposed. To cope with the intrinsic uncertainties of this task, fuzzy methods and models are being used throughout this work. Means are introduced for modelling and recognition of patterns in multivariate time series. Based on a novel method for the early recognition of incomplete patterns in streaming time series, a short-time prediction becomes feasible. Peculiarities and intrinsic difficulties of an online recognition of incomplete patterns are being discussed with the help of suitable examples. The usability of the pattern-based prediction approach is being demonstrated by means of real-world data.

Sustainability has gained a lot of attention in the field of research. Researchers and consumers both prioritize sustainability and environmental issues over previously dominant materials, such as plastic. Packaging and disposable items that used to be made of plastic have largely been replaced with paper. Unfortunately, paper does not perform as well as plastic regarding barrier properties against grease, oxygen, or water vapor. Barrier properties are an important factor when choosing packaging material for food, among other things, as they help maintain the shelf life of the product. In order to improve the properties of the paper packaging and expand its use, the paper is coated with a polymer. However, the polymer contributes to challenges in the recycling of the products as some of the polymer attaches to the fibers, causing difficulties in the separation of each material. Small fragments of plastic may end up in the material streams and the recycled pulp due to the existing challenges in completely removing plastic from cellulosic substrates during recycling. This thesis analyzes the possibilities of identifying and classifying plastic fragments of polyethylene (PE) and polyvinyl alcohol (PVOH) in recycled paper sheets using near-infrared spectroscopy together with multivariate data analysis. The purpose of the work is to develop models that can identify possible residues that may appear in recycled products from various industries. Paper sheets of two different grammages and six different compositions of recycled fiber and virgin fiber were created and scanned by NIR, with and without plastic film under the sheets. The scans were used to develop classification models to identify and categorize scans not included in the calibration data set. The performance of the models was tested by applying them to images of sheets of paper with plastic fragments of different sizes and different type underneath. The results indicated potential in the method. The prediction of the paper sheets with a lower grammage was mostly correct, whereas the classification of polyethylene showed the best performance. There was some noise in the prediction of the plastic fragments, regardless of the grammage of the paper. The noise may be due to a wide variation in the calibration data set since it consisted of paper sheets of six different compositions. A large part of the noise was incorrectly classified as polyvinyl alcohol, which can be due to differences in the manufacturing process of the plastic films. The conclusion of the thesis is that it is feasible to identify and categorize plastic fragments of polyethylene and polyvinyl alcohol in recycled paper sheets with a certain margin of error. It can be stated that the method shows promise, but further research and development in the field is required to build models that can be applied to a wider range of samples.

Die in dieser Arbeit vorgestellten Studien untersuchen verschiedene Ansätze für die Analyse von spektroskopischen Daten für die Erforschung anderer Himmelskörper. Der Fokus lag hierbei auf der laserinduzierten Plasmaspektroskopie (LIBS, engl. laser-induced breakdown spectroscopy), aber auch Daten der Raman-Spektroskopie wurden analysiert. Das erste extraterrestrisch eingesetzte LIBS Instrument, ChemCam, auf dem Mars Science Laboratory (MSL) der NASA untersucht die Marsoberfläche seit 2012 und weitere Missionen mit LIBS und Raman Instrumenten zum Mars sind geplant. Neben analytischen Ansätzen wurden statistische Methoden, die als multivariate Datenanalysen (MVA) bekannt sind, verwendet und evaluiert. In dieser Arbeit werden insgesamt vier Studien vorgestellt. In der ersten Studie wurde die Normalisierung von LIBS Daten mit Plasmaparametern, also der Plasmatemperatur und der Elektronendichte, untersucht. In der zweiten Studie wurden LIBS Messungen unter Vakuumbedingungen im Hinblick auf den Ionisierungsgrad des Plasmas untersucht. In der dritten Studie wurden MVA Methoden wie die Hauptkomponentenanalyse (PCA) und die partielle Regression kleinster Quadrate (PLS-R) zur Identifizierung und Quantifizierung von Halogenen mittels molekularer Emissionen angewandt. Die Ergebnisse sind vielversprechend, da es möglich war z.B. Chlor in einem ausgewählten Konzentrationsbereich zu quantifizieren. In der letzten Studie wurden LIBS-Daten mit komplementären Raman-Daten von Mars relevanten Salzen in einem low-level Datenfusionsansatz kombiniert. Es wurden MVA Methoden angewandt und auch Konzepte der high-level Datenfusion implementiert. Mit der low-level LIBS und Raman Datenfusion konnten im Vergleich zu den einzelnen Techniken mehr Salze richtig identifiziert werden. Der Gewinn durch die low-level Datenfusion ist jedoch vergleichsweise gering und für konkrete Missionen müssen individuelle und angepasste Strategien für die gemeinsame Analyse von LIBS und Raman-Daten gefunden werden. / The studies presented in this thesis investigate different data analysis approaches for mainly laser-induced breakdown spectroscopy (LIBS) and also Raman data in the context of planetary in-situ exploration. Most studies were motivated by Mars exploration due to the first extraterrestrially employed LIBS instrument ChemCam on NASA's Mars Science Laboratory (MSL) and further planned LIBS and Raman instruments on upcoming missions to Mars. Next to analytical approaches, statistical methods known as multivariate data analysis (MVA) were applied and evaluated. In this thesis, four studies are presented in which LIBS and Raman data analysis strategies are evaluated. In the first study, LIBS data normalization with plasma parameters, namely the plasma temperature and the electron density, was studied. In the second study, LIBS measurements in vacuum conditions were investigated with a focus on the degree of ionization of the LIBS plasma. In the third study, the capability of MVA methods such as principal component analysis (PCA) and partial least squares regression (PLS-R) for the identification and quantification of halogens by means of molecular emissions was tested. The outcomes are promising, as it was possible to distinguish apatites and to quantify chlorine in a particular concentration range. In the fourth and last study, LIBS data was combined with complementary Raman data in a low-level data fusion approach using MVA methods. Also, concepts of high-level data fusion were implemented. Low-level LIBS and Raman data fusion can improve identification capabilities in comparison to the single datasets. However, the improvement is comparatively small regarding the higher amount of information in the low-level fused data and dedicated strategies for the joint analysis of LIBS and Raman data have to be found for particular scientific objectives.

A global transformation is being fueled by unprecedented growth in the quality, quantity, and number of different parameters in environmental data through the convergence of several technological advances in data collection and modeling. Although these data hold great potential for helping us understand many complex and, in some cases, life-threatening environmental processes, our ability to generate such data is far outpacing our ability to analyze it. In particular, conventional environmental data analysis tools are inadequate for coping with the size and complexity of these data. As a result, users are forced to reduce the problem in order to adapt to the capabilities of the tools. To overcome these limitations, we must complement the power of computational methods with human knowledge, flexible thinking, imagination, and our capacity for insight by developing visual analysis tools that distill information into the actionable criteria needed for enhanced decision support. In light of said challenges, we have integrated automated statistical analysis capabilities with a highly interactive, multivariate visualization interface to produce a promising approach for visual environmental data analysis. By combining advanced interaction techniques such as dynamic axis scaling, conjunctive parallel coordinates, statistical indicators, and aerial perspective shading, we provide an enhanced variant of the classical parallel coordinates plot. Furthermore, the system facilitates statistical processes such as stepwise linear regression and correlation analysis to assist in the identification and quantification of the most significant predictors for a particular dependent variable. These capabilities are combined into a unique geovisual analytics system that is demonstrated via a pedagogical case study and three North Atlantic tropical cyclone climate studies using a systematic workflow. In addition to revealing several significant associations between environmental observations and tropical cyclone activity, this research corroborates the notion that enhanced parallel coordinates coupled with statistical analysis can be used for more effective knowledge discovery and confirmation in complex, real-world data sets.

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

O uso assertivo e eficiente das estratégias de aprendizagem depende, muitas vezes, da compreensão e consideração de aspectos psicológicos e motivacionais. O adequado emprego de estratégias de aprendizagem se reflete no desempenho acadêmico, no domínio de construtos e modelos e no amadurecimento crítico e científico. A presente tese defende que há uma relação entre as estratégias de aprendizagem autorregulada e as estratégias de aprendizagem autodeterminada predominantes em alunos de mestrado e doutorado em Contabilidade. O estudo se justifica, porquanto, porque além de inaugurar uma linha de pesquisa ainda inédita no contexto da Contabilidade Humana, seus resultados destacam um original entendimento da relação da aprendizagem com a regulação e a motivação pessoal. Tem como objetivo principal apresentar diagnóstico, dimensões e correlações das estratégias de aprendizagem autorregulada e aprendizagem autodeterminada de alunos de programas de pós-graduação stricto sensu em Contabilidade no Brasil. Participaram do survey 516 respondentes, sendo 383 mestrandos e 133 doutorandos. Foram aplicados dois instrumentos psicométricos: Self-Regulated Learning Strategies (SRLS) e Motivated Strategies for Learning Questionnaire (MSLQ). O modelo operacional de pesquisa delineou a formulação de oito hipóteses, sendo que a primeira delas sustenta a defesa da tese, enquanto as demais defendem a influência das variáveis idade, gênero, tipo de curso, estágio no curso, tipo de instituição de graduação, nota do curso atribuída pela Capes e graus de instrução dos pais nos níveis de Self-Regulated Learning (SRL) e Self-Determination Theory (SDT). A partir da análise multivariada dos dados, os resultados corroboraram a tese e a influência do gênero no nível de SRL. A metaconclusão desta tese ratifica os estudos referenciados, confirmando que a aprendizagem pode ser dominada e controlada pelo indivíduo, ao se adotar estratégias individuais de regulação e motivação. Uma importante contribuição desta pesquisa consiste em oferecer conclusões empíricas que podem ajudar docentes, discentes, pesquisadores, instituições de ensino e programas de pós-graduação a compreender mais sistematicamente os aspectos da aprendizagem autorregulada e da aprendizagem autodeterminada que caracterizam o aluno de Contabilidade. Limitações importantes deste estudo podem ser vistas como oportunidades para pesquisas futuras: a amostra envolve um público específico, a pesquisa survey pode apresentar vieses de método comum e a baixa participação de alunos de mestrado profissional. Estudos futuros poderão adotar outras estratégias metodológicas e/ou envolver amostras mais diversificadas ou em maior lastro temporal / Assertive and efficient use of learning strategies often depends of the understanding and consideration of psychological and motivational aspects. Appropriate use of learning strategies is reflected in the academic performance, in the appropriation of constructs and models and in the critical and scientific maturity. This dissertation argues that there is a relationship between predominating self-regulated learning strategies and self-determined learning strategies in accounting master\'s and doctorate students. The study can be justified in view of, apart from inaugurating a research line within the context of Human Accounting, their results highlight a unique understanding of the relationship of learning with regulation and personal motivation. Its main goal is to present a diagnosis, the dimensions and the correlations of self-regulated learning and self-determined learning strategies of graduate Accounting students in Brazil. Five hundred and sixteen respondents participated in the survey, comprising 383 master\'s and 133 doctoral students. Two psychometric instruments were applied: the Self-Regulated Learning Strategies (SRLS) and the Motivated Strategies for Learning Questionnaire (MSLQ). The operating model research outlined the formulation of eight hypotheses, being that the first of them supports the thesis, while the others investigate the influence in the levels of Self-Regulated Learning (SRL) and Self-Determination Theory (SDT) of age, gender, type of course, stage in the course, type of undergraduate institution (public or private), grade attributed by Capes to the course and parental formal education degrees. From the multivariate data analysis,the results support the thesis and that gender has influence in the SRL level. The metaconclusion of this thesis confirms the referenced studies, estating that learning can be dominated and controlled by individuals through the adoption of individual strategies of regulation and motivation. An important contribution of this study is to offer empirical conclusions that might help teachers, students themselves, researchers, educational institutions and graduate programs to understand more systematically the aspects of self-regulated learning and self-determined learning that characterize the Accounting graduate students. The major limitations of the present study can be seen as opportunities for future researches: the sample involves a particular audience, research can provide common methods bias and the low participation of professional master\'s degree students in the sample. Future studies can take further methodological strategies and/or involve more diversified samples or consider longitudinal approaches

Ionospheric scintillation is a rapid variation in the amplitude and/or phase of radio signals traveling through the ionosphere. This spatial and time-varying phenomenon is of interest because its occurrence may affect the reception quality of satellite signals. Specialized receivers at strategic regions can track multiple variables related to the phenomenon, generating a database of historical observations on the regional behavior of ionospheric scintillation. The analysis of such data is very challenging, since it consists of time-varying measurements of many variables which are heterogeneous in nature and with possibly many missing values, recorded over extensive time periods. There is a need to introduce alternative intuitive strategies that contribute to experts acquiring further knowledge from the ionospheric scintillation data. Such challenges motivated a study on the applicability of visualization techniques to support tasks of identification of relevant attributes in the study of the behavior of phenomena described by multiple time-varying variables, of which the ionospheric scintillation is a good example. In particular, this thesis introduces a visual analytics framework, named TV-MV Analytics, that supports exploratory tasks on time-varying multivariate data and was developed following the requirements of experts on ionospheric scintillation from the Faculty of Science and Technology of UNESP at Presidente Prudente, Brazil. TV-MV Analytics provides an interactive visual exploration loop to analysts inspecting the behavior of multiple variables at different temporal scales, through temporal representations associated with clustering and multidimensional projection techniques. Analysts can also assess how different feature sub-spaces contribute to characterizing a certain behavior, where they may direct the analysis process and include their domain knowledge in the exploratory analysis. We also illustrate the application of TV-MV Analytics on multivariate time-varying data sets from three alternative application domains. Experimental results indicate the proposed solutions show good potential on assisting time-varying multivariate data mining tasks, since it reduces the effort required from experts to gain deeper insight into the historical behavior of the variables describing a phenomenon or domain. / A cintilação ionosférica é uma variação rápida na amplitude e/ou na fase dos sinais de rádio que viajam através da ionosfera. Este fenômeno espacial e variante no tempo é de grande interesse, pois pode afetar a qualidade de recepção dos sinais de satélite. Receptores especializados em regiões estratégicas podem rastrear múltiplas variáveis relacionadas ao fenômeno, gerando um banco de dados de observações históricas sobre o comportamento regional da cintilação. O estudo do comportamento da cintilação é desafiador, uma vez que requer a análise extensiva de dados multivariados e variantes no tempo, coletados por longos períodos. Medições são registradas continuamente, e são de natureza heterogênea, compreendendo múltiplas variáveis de diferentes categorias e possivelmente com muitos valores faltantes. Portanto, existe a necessidade de introduzir estratégias alternativas, eficientes e intuitivas, que contribuam para a adquisição de conhecimento, a partir dos dados, por especialistas que estudam a cintilação ionosférica. Tais desafios motivaram o estudo da aplicabilidade de técnicas de visualização para apoiar tarefas de identificação de atributos relevantes no estudo do comportamento de fenômenos ou domínios que envolvem múltiplas variáveis, como a cintilação. Em particular, esta tese introduz um arcabouço visual, o qual foi denominado TV-MV Analytics, que apoia tarefas de análise exploratória sobre dados multivariados e variáveis no tempo, inspirado em requisitos de especialistas no estudo da cintilação, vinculados à Faculdade de Ciências e Tecnologia da UNESP de Presidente Prudente, Brasil. O TV-MV Analytics fornece aos analistas um ciclo de interativo de exploração que apoia a inspeção do comportamento temporal de múltiplas variáveis, em diferentes escalas temporais, por meio de representações visuais temporais associadas a técnicas de agrupamento e de projeção multidimensional. Também permite avaliar como diferentes sub-espaços de atributos caracterizam um determinado comportamento, podendo direcionar o processo de análise e inserir seu conhecimento do domínio no processo de análise exploratória. As funcionalidades do TV-MV Analytics também são ilustradas em dados variantes no tempo oriundos de outros três domínios de aplicação. Os resultados experimentais indicaram que as soluções propostas têm bom potencial em tarefas de mineração de dados multivariados e variantes no tempo, uma vez que reduz o esforço e contribui para os especialistas obterem informações detalhadas sobre o comportamento histórico das variáveis que descrevem um determinado fenômeno ou domínio.

L’objectif de cette thèse est de contribuer à l’amélioration du processus de dépouillement d’essais de sûreté visant étudier le comportement d'un combustible nucléaire en contexte d’accident d’injection de réactivité (RIA), via la technique de contrôle par émission acoustique. Il s’agit notamment d’identifier clairement les mécanismes physiques pouvant intervenir au cours des essais à travers leur signature acoustique. Dans un premier temps, au travers de calculs analytiques et des simulation numériques conduites au moyen d’une méthode d’éléments finis spectraux, l’impact du dispositif d’essais sur la propagation des ondes est étudié. Une fréquence de résonance du dispositif est identifiée. On établit également que les mécanismes basses fréquences ne sont pas impactés par le dispositif d'essais. En second lieu, diverses techniques de traitement du signal (soustraction spectrale, analyse spectrale singulière, ondelettes. . . ) sont expérimentées, afin de proposer des outils permettant de traiter différent types de bruit survenant lors des essais RIA. La soustraction spectrale s’avère être la méthode la plus robuste aux changements de nature du bruit, avec un fort potentiel d’amélioration du rapport signal-à-bruit. Enfin, des méthodes d’analyse de données multivariées et d’analyse de données fonctionnelles ont été appliquées, afin de proposer un algorithme de classification statistique permettant de mieux comprendre la phénoménologie des accidents de type RIA et d’identifier les mécanismes physiques. Selon l’approche (multivariée ou fonctionnelle), les algorithmes obtenus permettent de reconnaître le mécanisme associé à une salve dans plus de 80% des cas. / The objective of the thesis is to contribute to the improvement of the monitoring process of nuclear safety experiments dedicated to study the behavior of the nuclear fuel in a reactivity initiated accident (RIA) context, by using the acoustic emission technique. In particular, we want to identify the physical mechanisms occurring during the experiments through their acoustic signatures. Firstly, analytical derivations and numerical simulations using the spectral finite element method have been performed in order to evaluate the impact of the wave travelpath in the test device on the recorded signals. A resonant frequency has been identified and it has been shown that the geometry and the configuration of the test device may not influence the wave propagation in the low frequency range. Secondly, signal processing methods (spectral subtraction, singular spectrum analysis, wavelets,…) have been explored in order to propose different denoising strategies according to the type of noise observed during the experiments. If we consider only the global SNR improvement ratio, the spectral subtraction method is the most robust to changes in the stochastic behavior of noise. Finally, classical multivariate and functional data analysis tools are used in order to create a machine learning algorithm dedicated to contribute to a better understanding of the phenomenology of RIA accidents. According to the method (multivariate or functional), the obtained algorithms allow to identify the mechanisms in more than 80 % of cases.