La diminution en pétrole brut léger nécessite de convertir les fractions lourdes en produits valorisables (essences, gazoles, huiles, etc.). Dans ce contexte, l'hydrocraquage (HCK) fournit des produits de très haute qualité à partir de distillats sous vide (DSV) du pétrole brut. La qualité des coupes obtenues est caractérisée par des propriétés physico-chimiques qui sont soumises à des spécifications. L'optimisation du procédé nécessite des expérimentations longues et coûteuses. IFPEN a donc de plus en plus recours à des tests sur unité d'expérimentation haut débit (EHD). Ces derniers posent cependant un problème d'accessibilité aux coupes d'intérêt. Par ailleurs, pour comprendre et prédire l'impact des conditions opératoires sur la qualité des produits, des simulateurs sont développés. Certaines propriétés de produits sont cependant complexes et difficiles à modéliser voire mal comprises. Ce travail de thèse a porté sur l'amélioration de la compréhension moléculaire des propriétés produits pour une meilleure prédiction. Dans cette étude, nous nous sommes focalisés sur le point de trouble (PT) de la coupe gazole et l'indice de viscosité (VI) de l'huile obtenue lors de l'hydrocraquage de DSV. Deux techniques d'analyse moléculaire ont été utilisées : la chromatographie en phase gazeuse bidimensionnelle (GC×GC) qui permet de déterminer la composition par famille chimique des différentes coupes et la résonance magnétique nucléaire (RMN) du 13C qui fournit des informations sur la structure chimique des hydrocarbures présents dans ces mélanges. Nous présentons les résultats obtenus par une régression multivariée parcimonieuse (sparse Partial Least Squares) appliquée aux données GC×GC et 13C RMN. Il s'agit d'une variante de la PLS classique qui permet de réduire le nombre de facteurs tout en privilégiant ceux qui sont les plus corrélés à une propriété d'intérêt donnée. Globalement, cette étude a notamment permis de mieux comprendre l'impact des différents hydrocarbures (n-paraffines, isoparaffines, aromatiques,…) et de leur structure moléculaire (longueur de chaînes, degrés de branchements,…) sur le PT des gazoles et le VI des huiles. La bonne qualité des modèles obtenus par sparse PLS montre par ailleurs la possibilité d'accéder à la qualité des produits lors de l'utilisation d'EHD. Des modèles de prédiction par krigeage ont également été développés. Cette méthode d'interpolation permet de prédire une propriété en un point donné en effectuant une moyenne pondérée des observations au voisinage de ce point. Les modèles de krigeage sont des modèles locaux adaptés aux structures de données complexes. Ce sont des approches probabilistes qui permettent d'estimer les incertitudes de prédiction. Aussi bien dans le cas du PT de la coupe gazole que dans celui du VI de la coupe huile, les résultats montrent une amélioration des performances. Cette approche est tout à fait novatrice dans le domaine des produits pétroliers. Lors de l'utilisation d'unités EHD, elle permet d'accéder au VI des huiles de base plus aisément que via des données chromatographiques ou spectroscopiques, qui sont de plus non accessibles en raffinerie / The rapid decline in light crude oils requires to convert heavy petroleum fractions into more valuable products (naphtha, diesel, lubricants, etc.). In this context, hydrocracking process (HCK) consists on upgrading vaccum gas oil (VGO) into high quality products. The quality of petroleum products is based on some chemical and physical properties that should fulfill prerequisite specifications. The hydrocracking process optimization requires to set up time consuming and costly experiments for developing catalysts and setting operating conditions. High throughput experimentation (HTE) units are then increasingly used at IFPEN. However, these units do not enable to obtain end products. Otherwise, predictive models were developed in order to understand and predict the impact of operating conditions about products quality. However, some complex properties are very difficult to model and require a better understanding. This work is mainly concerned with the understanding of diesel cloud point (CP) and viscosity index (VI) of base oils. Two analytical techniques were used: the two-dimensional gas chromatography (GC×GC) that enables to identify hydrocarbons compounds in petroleum products and the 13C nuclear magnetic resonance (NMR) spectroscopy which provides structural characteristics of these compounds. A sparse multivariate regression (sparse Partial Least Squares) was performed using chromatographic and spectroscopic data. The sparse PLS is derived from classical PLS. It allows to reduce the number of factors by performing a variable selection. The selected factors are the most correlated to the property to model. Globally, this approach enabled to better understand how hydrocarbon compounds (nparaffins, isoparaffins, aromatics,…) and their molecular characteristics (carbon number, degree of branching,…) affect the diesel CP and the VI of base oil. Furthermore, the good performances of developed sparse PLS models show that it is possible to access to the products quality when using HTE units. Kriging models were also developed. Kriging is an interpolation method that predicts the value of a function at a given point by computing a weighted average of the known values of the function in the neighborhood of the point. Kriging models have local aspect which is well adapted to complex data. Its probabilistic approach enables to provide an estimate of predicted value uncertainty. Results show that kriging improves predictive performances for both diesel CP and VI of base oil. This approach is quite innovative in modelling of petroleum products properties. When using HTE units, it allows to estimate the VI of base oil more easily than from chromatographic or spectroscopic data which are not available for the refiners

La nutrition est un facteur de risque modifiable pour le cancer puisqu'environ un tiers des cas pourraient être évités en adoptant une meilleure alimentation. La relation entre nutrition et cancer est complexe, et son étude est enrichie par de nouveaux défis apportés par les récentes avancées technologiques dans le domaine des « -omiques ». Cette thèse a pour but de développer de nouvelles approches biostatistiques afin d'étudier la relation entre nutrition et cancer au sein de la cohorte EPIC. Pour ce faire, l'applicabilité de nouvelles méthodologies multivariées dans le domaine de l'épidémiologie nutritionnelle a été étudiée.Une nouvelle méthode multivariée pour la réduction de la dimensionnalité, le Treelet Transform (TT), a été examinée afin d'extraire des patterns de nutriments issus de questionnaires. Les patterns ainsi obtenus par le TT étaient plus facilement interprétables que par les méthodes classiques. Ensuite, un cadre analytique pour implémenter le concept du « meeting-in-the-middle » (MITM) a été développé et appliqué dans 2 études cas-témoin nichées sur le cancer hépatocellulaire avec des données métabolomiques. Le MITM cherche à identifier des biomarqueurs qui soient à la fois des marqueurs de certaines expositions passées et des prédicteurs de maladies. L'implémentation s'est focalisée sur l'application de la PLS et de l'analyse de médiation.Enfin, nous avons examinés la relation entre les niveaux plasmatiques de 60 acides gras issus de biomarqueurs et le risque de cancer du sein dans une étude cas-témoin nichée dans EPIC.Cette thèse servira de base pour des applications épidémiologiques futures examinant la relation nutrition-cancer / Diet is a modifiable risk factor for many cancers. It has been estimated that about a third of cancer cases can be prevented by complying with a healthy diet and adhering to the recommendations in terms of nutrition. The nutrition-cancer relationship is a complex one, and its study is currently at a turning point with the opportunity and challenges brought by the recent technological advances in the fields of « -omics ».This thesis aims to develop new biostatistical approaches to investigate the nutrition-cancer relation within the European Prospective Investigation into Cancer and nutrition (EPIC) study. To do so, the applicability of new methodologies in the field of nutritional epidemiology has been examined.First, a new multivariate dimension reduction method, the Treelet Transform (TT) was applied to extract nutrient patterns relying on questionnaire data. The extracted patterns were more easily interpretable than those obtained with more classical methods.Then, an analytical framework was conceived for the « meeting-in-the-middle » (MITM) principle and applied to two nested case-control studies on hepatocellular carcinoma, with targeted and untargeted metabolomics data. The MITM aims to identify overlap biomarkers of certain exposures that are at the same time predictive of disease outcomes. The implementation focused on the application of partial least squares and mediation analyses. Last, the association between 60 plasma fatty acids levels assessed from biomarkers and breast cancer risk was examined in a nested case-control study in EPIC. This thesis will serve as a basis for future epidemiological applications looking into the nutrition-cancer relation

A travers le cas des fournisseurs de Marques de Distributeurs (MDD) en agro-alimentaire, cette thèse vise à montrer que la satisfaction du fournisseur peut être expliquée par la proximité entre ce dernier et son distributeur. En adoptant une posture épistémologique positiviste et une démarche hypothético-déductive, nous avons réalisé une revue de littérature qui a permis d’élaborer un modèle conceptuel. Celui-ci postule que la proximité, déclinée en quatre dimensions-proximité relationnelle, proximité identitaire, proximité de processus et proximité géographique, est un antécédent de la confiance et in fine de la satisfaction de cette relation B to B. Afin de tester les hypothèses formulées, nous avons ensuite mené une étude empirique auprès de 93 fournisseurs de MDD de la Grande Distribution Alimentaire. Une première phase de collecte des données s’est déroulée au salon MDDEXPO en avril 2014 et a permis de recueillir 51 questionnaires. La deuxième phase de collecte a été réalisée via le réseau social numérique professionnel VIADEO et a permis de recueillir 42 réponses. Puis, nous avons soumis notre modèle au test avec la méthode PLS sous le logiciel SMARTPLS3. Les résultats de cette recherche montrent que deux dimensions de la proximité permettent d’accroitre directement la satisfaction du fournisseur de MDD de la relation, sans « passer par la confiance du fournisseur de MDD envers son distributeur ». / Through the case of private label supplier in food sector, this thesis aims to show that the satisfaction of private label supplier can be explained by the proximity with his distributor. By adopting a positivist epistemological position and hypothetical-deductive approach, we conducted a literature review that allowed us to develop an integrative model made up of four dimensions: relational proximity, identity proximity, process proximity and geographical proximity. Then realized an empirical study alongside 93 suppliers of private label. A first data collection stage took place in MDDEXPO on April 2014 and has enabled to collect 51 questionnaires. The second phase of collection was done via VIADEO and which came up with 42 responses. Further, we submitted our model to test using PLS method under SMARTPLS3 software. The results of this research show that two dimensions of proximity have direct relationship with increasing private label supplier satisfaction. They are, in the order of importance of structural relations, level of cooperation (relational proximity) and adaptations (process proximity). They also indicate that three dimensions of proximity are used to increase the satisfaction of private label supplier. They are , in order of importance of structural relationships, identity proximity (shared values), social bonds (sub-scale relational proximity) and geographic proximity. Moreover, they show that two sub-dimensions of proximity do not significantly influence the confidence nor the satisfaction of private label supplier to its retailer. It is a question of information exchange (relational proximity) and idiosyncratic assets (process proximity).

Dans cette thèse, nous introduisons des modèles de régression ”Gini-PLS”. Les algorithmes proposés combinent les propriétés des estimateurs relatifs aux régressions Gini et PLS. Les quatre modèles construits dans cette thèse permettent de résoudre simultanément les problèmes : de valeurs extrêmes (”outliers”), de multi-colinéarité, de faible taille de l’échantillon, de données manquantes, d’erreurs de mesure et d’endogénéité. En présence des problèmes cités, les modèles uni-variés (Gini-PLS1) sont robustes pour estimer une variable dépendante en fonction d’une ou plusieurs variables explicatives ; tandis que les modèles multi-variés (Gini-PLS2) servent à estimer une matrice de variables dépendantes en fonction d’une matrice de variables explicatives.Notre application dans le cadre de la thèse concerne l’estimation de contributions des variables technico-économiques aux inégalités des rémunérations pour les pays européens adhérents à la Politique Agricole Commune.Nous proposons deux approches de régressions basées sur les modèles Gini-PLS (RISD-Gini-PLS) pour estimer les contributions des variables technico-économiques (sources de revenus, superficies, main d’œuvre, etc.) aux inégalités des revenus agricoles pour les pays de l’union européenne avant et après les réformes de Mac Sharry et de l’accord de Luxembourg. / In this thesis we propose ”Gini-PLS” regressions. The proposed algorithms combine the properties of the estimators related to the Gini and PLS regressions. The four models built in this thesis solve simultaneously the problems of : extreme values (outliers), multicollinearity, small sample, missing data, measurement errors,and endogeneity. In presence of these problems, the univariate models (Gini-PLS1) are robust to estimate a dependent variable with one or more explanatory variables. While, the multivariate models (Gini-PLS2) are used to estimate a matrix of dependent variables with a matrix of explanatory variables.Our application in this thesis is the estimation of the contributions of technico-economic variables to the whole inequality of farm’s income for European countries acceding to the Common Agricultural Policy. We also propose Gini-PLS regressions approaches based on income source decomposition (RISD-Gini-PLS) to estimate the contributions of techno-economic variables (income sources, areas, labor, etc.) to the incomei nequalies of productions (total output crops and output livestock) for european countries.

L’objectif de la recherche est d’étudier, dans une optique partenariale de la firme, l’efficacité des stratégies d’alliance dans le secteur financier français en termes d’amélioration de niveau d’orientation client des entreprises partenaires. Nous nous fondons, d’une part, sur l’approche par les connaissances et en particulier celle de l’apprentissage organisationnel, et d’autre part, sur la littérature portant sur le management des relations interentreprises. D’un point de vue théorique, des différentes approches traitant du management des alliances, notre recherche instrumentalise celle concevant ces stratégies comme un processus d'apprentissage organisationnel. L’analyse causale, étudie l’influence des modes de coordination des stratégies d’alliance sur le niveau d’orientation client des entreprises partenaires : Il s’agit des deux variables : confiance et contrôle. Cette influence s’exerce par le biais d’une variable centrale à savoir la qualité d’apprentissage organisationnel. Cette analyse explicative est réalisée grâce aux méthodes d’équations structurelles, de type PLS. Trois principaux résultats sont obtenus : quand les rapports entre les partenaires sont basés sur la confiance, la qualité d’apprentissage organisationnel dans l’alliance se retrouve meilleure ; l’exercice du contrôle dans l’alliance n’affecte pas la qualité d’apprentissage organisationnel ; et l’alliance, par l’apprentissage organisationnel qu’elle permet, améliore le niveau d’orientation client des entreprises partenaires. / Based on the perspective of the partnership of the firm we aim to study, alliance strategies in the French financial sector, in terms of the improvement of the level of customer orientation business partners. We are based, firstly, on the approach by the knowledge and especially that of organizational learning, and secondly, on the linkages’ management literature. From a theoretical point of view, in terms of the various approaches dealing with the management of alliances, our research exploits those which designing these strategies as a process of organizational learning. Causal analysis, studies the influence of the modes of coordination of alliance strategies on the level of customer orientation of companies: These are two variables: trust and control. This influence is exerted through a central variable namely the quality of organizational learning. Such explanatory analysis is performed using the methods of structural equation of PLS. Three main results are obtained: when the relationship between the partners is based on trust, quality of the organizational alliance learning is found best ; the exercise of control within the alliance does not affect the quality of learning organizational and The alliance allows organizational learning and improves the level of customer orientation business partners.

Submitted by Nadir Basilio (nadirsb@uninove.br) on 2017-08-04T15:51:49Z No. of bitstreams: 1 Marcelo Luiz do Amaral Goncalves.pdf: 4473007 bytes, checksum: 5fa0b469e762031a2eafd0475ed7574c (MD5) / Made available in DSpace on 2017-08-04T15:51:49Z (GMT). No. of bitstreams: 1 Marcelo Luiz do Amaral Goncalves.pdf: 4473007 bytes, checksum: 5fa0b469e762031a2eafd0475ed7574c (MD5) Previous issue date: 2017-02-15 / In a wide range of activities, companies have been using the Business Model to represent the organization's strategy in delivering value and meeting the needs of its clients and segments with organizational efficiency and competitive differential. The business model presents itself as a very useful conceptual tool for capturing, sharing and creating a common view of the organization model. The projects are prominent in the strategic issues, because through them it is possible to materialize the planned strategic objectives and described in the Business Model of the company. The evaluation of project success can be accomplished by considering short and long-term objectives, uncertainties associated with the market, and technologies that may affect expectations about project success. This research analyzed the influence of the configuration of the business model in the project success in national and international companies from different fields of activity. In this study, each of the dimensions that constitute the Business Model was identified, as well as the dimensions of the Project Success. Additionally, was verified the contribution of the Business Model configuration in the overall configuration of the Business Model and the contribution of the Project Success dimensions in the overall success of the Project. This research is of an applied nature, characterized as confirmatory-descriptive and the data were obtained through a survey, using as a research instrument a structured questionnaire for data collection. The questionnaire was sent to a universe composed of professionals from various fields of activity, with leadership positions and who have already participated in projects, providing a number of 181 valid answers. The quantitative analysis was used to study the data collected in the research, using the Structural Equation Modeling (SEM) as the method and the Partial Least Square (PLS-SEM) as a technique for analyzing the data. The results showed that the configuration of the business model has a positive influence to the project success, explaining 46% of the effects on the success of projects developed in organizations of several branches of activities in several countries. As a contribution to professional practice, was proposed the Business Model´s Components Evaluation Model. This model aims to assess the maturity of each components that constitutes a Business Model, helping managers to diagnose which components need action to improve the maturity of Business Models in their organizations. / Nos mais diversos ramos de atividades, as empresas vêm utilizando o Modelo de Negócios para representar a estratégia da organização na entrega de valor e atender as necessidades de seus clientes e segmentos com eficiência organizacional e diferencial competitivo. O modelo de negócios apresenta-se como uma ferramenta conceitual muito útil para capturar, compartilhar e criar uma visão comum do modelo da organização. Os projetos são destaques nas questões estratégicas, pois através deles é possível materializar os objetivos estratégicos planejados e descritos no Modelo de Negócios da empresa. A avaliação do sucesso do projeto pode ser realizada considerando os objetivos de curtos e longos prazos, as incertezas associadas ao mercado e as tecnologias que podem afetar as expectativas em torno do sucesso do projeto. Esta pesquisa analisou a influência do modelo de negócios no sucesso do projeto em empresas nacionais e internacionais de diversos setores de atividades. Neste estudo, foi identificado cada uma das dimensões que constitui o Modelo de Negócios, bem como as dimensões do Sucesso do Projeto. Adicionalmente, foi verificada a contribuição da configuração do Modelo de Negócios no sucesso global do projeto e em cada uma das dimensões que compõe o sucesso do projeto. Esta pesquisa é de natureza aplicada, caracterizada como confirmatória-descritiva e os dados foram obtidas por meio de levantamento (Survey), utilizando como instrumento de pesquisa um questionário estruturado para a coleta dos dados. O questionário foi encaminhado a um universo composto por profissionais de vários ramos de atividade, com cargos de liderança e que já participaram de projetos, proporcionando um número de 181 respostas válidas. Foi utilizada a análise quantitativa para estudar os dados coletados na pesquisa, tendo a Modelagem de Equações Estruturais (MEE) como método e o Partial Least Square - Structural Equation Modeling (PLS-SEM) como técnica para analisar os dados. Os resultados demonstraram que a configuração do modelo de negócios influencia positivamente o sucesso do projeto, explicando 46% dos efeitos no sucesso do projeto desenvolvidos em organizações de diversos ramos de atividades em diversos países. Como contribuição para a prática profissional foi proposto o Modelo de Avaliação dos Componentes do Modelo de Negócios. Este modelo tem como objetivo avaliar a maturidade de cada um dos componentes que constituem um Modelo de Negócios, auxiliando os gestores a diagnosticar quais componentes necessitam de ações para melhorar o nível de maturidade dos Modelos de Negócios em suas organizações.

Sweden’s Facebook users produce large amounts of data on a daily basis by sharing their personal data with the platform despite being unaware of what or who the information is used for or by. The reason for this behavior is barely touched in previous research, therefore this quantitative study’s aim is to investigate factors influencing the attitude Facebook users in Sweden have towards sharing data. Beyond this the study also aims to investigate if there’s any difference in factors affecting the attitude between a generation which has grown up surrounded by technology and a generation without the same presence of technology in their upbringing. The study uses a modified version of Technology Acceptance Model (TAM) created by inspiration from previous research by related research. The constructs Previous privacy invasion, Awareness, Perceived risk and Social influence were added to the original TAM and used as a theoretical model for the study. 189 responses were gathered through an online-survey. The answers were analyzed with the multivariate analysis method Partial Least Squares Structural Equation Modeling (PLS-SEM). The method was used to evaluate the hypothesis in the modified model and in turn to answer the research questions posed by the study. After the results shows that the attitude towards sharing data on Facebook's services is affected by perceived risk, social influences from others as well as the useability and perceived ease of use. Furthermore the results show that there are different factors affecting the attitude towards sharing data on Facebook's services among people that have grown up with technology and those that haven't. None of the significant factors in the model were the same for the compared generation. The older generation’s attitude towards sharing data was affected by social influences and perceived ease of use and the younger generation was affected by useability. / Facebooks användare i Sverige genererar stora mängder data varje dag genom att dela med sig av information till plattformen. Användarna delar med sig av denna information trots att många av dem inte vet vad den används till eller vilka som kan ta del av den. Om det finns en eller flera specifika anledningar till varför användarna trots detta fortsätter att använda och således dela information på Facebooks tjänster är okänt. Denna kvantitativa studie har som syfte att undersöka vilka faktorer som påverkar Facebook-användares attityd till att dela information. Studien syftar även till att undersöka om det finns någon skillnad i vad som påverkar attityden hos en generation uppvuxen med teknologi och en äldre generation som inte vuxit upp med samma närvaro av teknologi. Studien använder en version av Technology Acceptance Model (TAM) som modifierats med inspiration hämtad från tidigare forskning inom ämnesområdet för attityd och beteende relaterat till informationssytem. Konstrukten Tidigare integritetsintrång, Kännedom, Upplevd oro och Socialt inflytande har adderats till TAM:s teoretiska grund. i studien används en undersökningsstrategi där 189 svar samlades in från respondenter via en enkätundersökning. Svaren analyserades med multivariatanalysmetoden Partial Least Squares Structural Equation Modeling (PLS-SEM) för att undersöka hypoteserna från den modifierade TAM samt svara på forskningsfrågorna. När samtliga svar från studien analyserats visade resultatet att attityd till datainsamling påverkas av upplevd oro, socialt inflytande från andra samt det sociala mediets nytta och användbarhet. Dessutom visade resultatet att det finns olika faktorer som påverkar attityden hos den yngre och den äldre gruppen. I studien var ingen av de faktorer som var signifikanta för den äldre gruppen signifikanta för den yngre gruppen och vice versa. För den äldre gruppen var socialt inflytande och nytta faktorer som påverkade deras attityd till datainsamling på Facebooks tjänster och för den yngre var den upplevda användbarheten något som påverkade deras attityd till datainsamling.

Soil spectroscopy has experienced a tremendous increase in soil property characterisation, and can be used not only in the laboratory but also from the space (imaging spectroscopy). Partial least squares (PLS) regression is one of the most common approaches for the calibration of soil properties using soil spectra. Besides functioning as a calibration method, PLS can also be used as a dimension reduction tool, which has scarcely been studied in soil spectroscopy. PLS components retained from high-dimensional spectral data can further be explored with the gradient-boosted decision tree (GBDT) method. Three soil sample categories were extracted from the Land Use/Land Cover Area Frame Survey (LUCAS) soil library according to the type of land cover (woodland, grassland, and cropland). First, PLS regression and GBDT were separately applied to build the spectroscopic models for soil organic carbon (OC), total nitrogen content (N), and clay for each soil category. Then, PLS-derived components were used as input variables for the GBDT model. The results demonstrate that the combined PLS-GBDT approach has better performance than PLS or GBDT alone. The relative important variables for soil property estimation revealed by the proposed method demonstrated that the PLS method is a useful dimension reduction tool for soil spectra to retain target-related information.

[ES] El cacao es un producto de alto valor, no únicamente por sus características sensoriales, sino porque también presenta un alto contenido en antioxidantes y alcaloides estimulantes con efectos saludables. Debido a la alta demanda, la industria del cacao en polvo tiene el desafío de asegurar la calidad de grandes volúmenes de producción de una manera rápida y precisa, evitando la presencia de contaminantes o adulterantes en la materia prima, ofreciendo productos donde se preserven las propiedades saludables. La espectroscopia del infrarrojo cercano (NIR) es una tecnología rápida y no destructiva útil en el análisis de productos alimentarios. La presente tesis doctoral se centra en evaluar el potencial uso del NIR como una herramienta de control de calidad con el fin de poder resolver problemas que se presentan en la industria del cacao en polvo. Los problemas a resolver incluyen la detección de materiales no deseados o adulterantes en el cacao en polvo, y la monitorización rápida y precisa del contenido de flavanoles y metilxantinas del cacao en polvo durante el proceso de alcalinización. El primer capítulo evalúa la viabilidad del NIR, en combinación con análisis quimiométricos, en la detección de la presencia de materiales no deseados o adulterantes como son cascarilla de cacao o harina de algarroba. Para ello, diferentes muestras de cacao en polvo natural y con diferentes niveles de alcalinización (suave, medio y fuerte) fueron mezcladas con distintas proporciones de cascarilla de cacao (con cacao natural) o harina de algarroba (con cacao natural y alcalinizado). Los resultados obtenidos indican que el NIR, combinado con modelos estadísticos tales como el análisis discriminante por mínimos cuadrados parciales (PLS-DA) y la regresión parcial de mínimos cuadrados (PLS), es un método rápido y eficaz para identificar cualitativa y cuantitativamente materiales no deseados o adulterantes como la cascarilla y la algarroba en cacao en polvo, independientemente del grado de alcalinización o el nivel de tostado de la harina de algarroba. En el segundo capítulo, el análisis composicional del cacao en polvo se orientó al control de los cambios producidos en el contenido de flavanoles y metilxantinas debidos al proceso de alcalinización al que se somete el caco en polvo. Se determinó el contenido de catequina, epicatequina, cafeína y teobromina mediante cromatografía líquida de alta resolución (HPLC), correlacionándose los contenidos obtenidos para cada uno de estos compuestos con las determinaciones NIR. Se obtuvieron buenos modelos para la predicción de los compuestos mediante regresión PLS con valores superiores a 3 para la relación entre el rendimiento y la desviación (RDP), lo cual demuestra que los modelos obtenidos pueden ser utilizados para la rápida y fiable predicción del contenido de flavanoles y metilxantinas en cacaos naturales y con diferentes niveles de alcalinización. / [CAT] El cacau és un producte d'alt valor, no sols per les seues característiques sensorials, sinó perquè també presenta un elevat contingut en antioxidants i alcaloids estimulants amb efectes saludables. A conseqüència a l'alta demanda, l'industria del cacau en pols té el desafiament d'assegurar la qualitat de grans volums de producció d'una manera ràpida i precisa, evitant la presència de contaminants o adulterants en la matèria cosina, oferint productes a on se preserven les propietats saludables. L'espectroscòpia de l'infrarroig proper (NIR) és una tecnologia ràpida i no destructiva útil en l'anàlisi de productes alimentaris. La present tesis doctoral se centra en avaluar el potencial ús del NIR com una eina de control de qualitat amb l'objectiu de poder resoldre problemes que es presenten en l'industria del cacau en pols. Els problemes a resoldre inclouen la detecció de materials no desitjats o adulterants en el cacau en pols, i la monitorització ràpida i precisa del contingut de flavanols i metilxantines del cacau en pols durant el procés d'alcalinització. El primer capítol avalua la viabilitat del NIR, en combinació amb anàlisis quimiométrics, en la detecció de la presència de materials no desitjats o adulterants com són pellofa de cacau o farina de garrofa. Per a això, diferents mostres de cacau en pols natural i amb diferents nivells d'alcalinització (suau, mig i fort) foren barrejades en distintes proporcions de pellofa de cacau (en cacau natural) o farina de garrofa (en cacau natural i alcalinisat). Els resultats obtinguts per a NIR, combinats amb models estadístics com l'anàlisi discriminant per mínims quadrats parcials (PLS-DA) i la regressió parcial de mínims quadrats (PLS), és un mètode ràpid i eficaç per identificar materials no desitjats o adulterants com la pellofa de cacau o la farina de garrofa, amb independència del grau d'alcalinització del cacau o de torrat de la farina de garrofa. En el segon capítol, l'anàlisi composicional del cacau en pols s'orientà al control dels canvis produïts en el contingut de flavanols i metilxantines a causa del procés d'alcalinització al que se sotmet el cacau en pols. Es va determinar el contingut de catequina, epicatequina, cafeïna i teobromina mitjançant cromatografia líquida d'alta resolució (HPLC), i es van correlacionar els continguts obtinguts per a cadascun d'estos composts amb les determinacions NIR. Es van obtindré bons models per a la predicció dels composts mitjançant regressió PLS amb valors superiors a 3 per a la relació entre el rendiment i la desviació (RDP), la qual cosa demostra que els models obtinguts poden ser emprats per a la ràpida i fiable predicció del contingut de flavanols i metilxantines en cacaus naturals o amb diferents nivells d'alcalinització. / [EN] Cocoa is a product of high value, not only because of its sensory characteristics, but also because it has a high content of antioxidants and stimulating alkaloids with health effects. Due to the high demand, the cocoa powder industry has the challenge of ensuring the quality of large volumes of production in a fast and accurate way, avoiding the presence of contaminants or adulterants in the raw material, offering products where the healthy properties are preserved. The near infrared spectroscopy (NIR) is a rapid and non-destructive technology useful in the analysis of food products. The present doctoral thesis focuses on evaluating the potential use of NIR as a quality control tool in order to solve problems that arise in the cocoa industry powdered. The problems to solve include the detection of unwanted materials or adulterants in the cocoa powder, and the rapid and accurate monitorization of the flavanols and methylxanthines content in the cocoa powder during the alkalization process. The first chapter evaluates the viability of the NIR, in combination with chemometric analysis, in the detection of presence of unwanted materials or adulterants such as cocoa shell or carob flour. For this, different samples of natural cocoa powder and with different levels of alkalization (light, medium and strong) were mixed with different proportions of cocoa shell (with natural cocoa) or carob flour (with natural and alkalized cocoa). The results obtained indicate that the NIR combined with statistical models such as the partial least squares discriminant analysis (PLS-DA) and the partial least squares regression (PLS), is a fast and efficient method to identify qualitative and quantitative unwanted materials or adulterants such as shell and carob in cocoa powder, regardless of the degree of alkalization or level of roasting of carob flour. In the second chapter, the compositional analysis of cocoa powder was oriented to the control of the changes produced in the content of flavanols and methylxanthines due to the process of alkalization to which the cocoa powder is subjected. The content of catechin, epicatechin, caffeine and theobromine were determined by high performance liquid chromatography (HPLC), correlating the contents obtained for each of these compounds with the NIR determinations. Good models were obtained for the prediction of compounds by regression PLS with values above 3 for the ratio of performance to deviation (RDP), which shows that the obtained models can be used for the quick and reliable prediction of flavanol content and methylxanthines in natural cocoas and with different alkalization levels. / This Doctoral Thesis has been carried out thanks to a doctoral studies scholarship granted by the Ministry of Higher Education, Science, Technology and Innovation (SENESCYT) of the Republic of Ecuador / Quelal Vásconez, MA. (2019). Application of Infrared Spectroscopy and Chemometrics to the Cocoa Industry for Fast Composition Analysis and Fraud Detection [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/135258 / TESIS

Massiver wirtschaftlicher Schaden und der Verlust des Verbrauchervertrauens hervorgerufen durch die BSE - Krise führten zu einer radikalen Änderung in der Futtermittel- und Lebensmittelgesetzgebung. Neben diversen anderen Maßnahmen wurden u. a. das Gehirn und Rückenmark von über 12 Monate alten Wiederkäuern vom Gesetzgeber als so genanntes spezifiziertes Risikomaterialien (SRM) definiert. Die spezielle Entsorgung und das Verarbeitungsverbot dieser Materialien in Lebens- und Futtermitteln hatte zum Ziel die Infektkette zu unterbrechen und dadurch das Risiko einer Infektion mit dem BSE-Erreger für den Verbraucher zu senken. Um die Einhaltung des Verarbeitungsverbotes von SRM in Lebensmitteln zu überprüfen, wurden diverse ZNS-Nachweisverfahren entwickelt. Jedoch ist keines, der hierfür entwickelten molekularbiologischen und immunochemischen Verfahren in der Lage sowohl Spezies, als auch Tieralter des nachgewiesenen ZNS zu bestimmen und damit eine potentielle Einordnung zum SRM zu ermöglichen. Darüber hinaus hat sich für fast alle ZNS-Marker der Einfluss von hohen Prozesstemperaturen als nachteilig für die Nachweisbarkeit erwiesen. NIEDERER und BOLLHALDER (2001) entwickelten ein auf Fettsäureanalytik basierendes ZNS-Nachweisverfahren, welches am Institut für Lebensmittelhygiene der Universität Leipzig weiterentwickelt wurde. Vorteile dieses Verfahrens sind die hitzestabilen Marker und die Möglichkeit mit Hilfe von spezifischen Fettsäureverhältnissen Spezies und Tieralters des nachgewiesenen ZNS zu ermitteln. Somit ist es mit diesem Verfahren erstmals möglich das nachgewiesene ZNS der Gruppe der spezifizierten Risikomaterialien zuzuordnen. Ziel dieser Arbeit war es, das am Institut für Lebensmittelhygiene weiterentwickelte Verfahren auf Praxistauglichkeit anhand eines externen Blindversuchs zu testen. Hierbei sollten mögliche Schwachstellen identifiziert und im weiterführenden Verlauf dieser Arbeit Lösungen erarbeitet werden. Für den Blindversuch wurden vom Max-Rubner-Institut, Standort Kulmbach, insgesamt 72 Brühwurstproben zur Verfügung gestellt, welche mittels des nach LÜCKER et al. (2005) beschriebenen Verfahrens untersucht wurden. Die Ergebnisse des Blindversuchs deuten auf eine grundsätzliche Eignung des Verfahrens für den Nachweis von spezifiziertem Risikomaterial in Fleischerzeugnissen hin. Jedoch konnte bei der Tierartdifferenzierung und vor allem bei der Altersbestimmung ein Optimierungsbedarf ermittelt werden. Somit ergaben sich für diese Arbeit folgende weiterführende Aufgabenstellungen: 1. Die Optimierung der Tierartbestimmung auch im Hinblick auf Erweiterung des Speziesspektrums und 2. die Optimierung der Tieraltersbestimmung von Rinder- und Schaf-ZNS. Insgesamt 257 ZNS-Proben der Spezies Rind, Schaf, Schwein, Ziege und Geflügel wurden einer Fettsäureanalyse unterzogen. Aus allen analysierten Fettsäuren wurden 67 Fettsäureverhältnisse gebildet. Zur Identifikation für die Speziesdifferenzierung geeigneter Fettsäureverhältnisse wurde das statistische Verfahren der Partial Least Square – Discriminant Analysis (PLS-DA) eingesetzt. Hierbei ergab sich, dass von den 67 untersuchten Fettsäureverhältnissen insgesamt 14 für die Differenzierung zwischen den Tierarten Rind, Schwein, Schaf, Ziege und Geflügel geeignet sind. Für die Optimierung der Tieraltersbestimmung wurden 37 ZNS-Proben vom Rind und elf ZNS-Proben vom Schaf untersucht. Die analysierten Fettsäuren wurden auf ihre Korrelation mit dem Alter untersucht. Stark korrelierende Fettsäuren und deren Verhältnisse wurden mit Hilfe der Regressionsanalyse auf ihre Vorhersagefähigkeit geprüft. Hierbei gelang die Identifikation von vier neuen Fettsäureverhältnissen. Für die Altersschätzung bei der Tierart Rind scheinen sich die FS-Verhältnisse 2OH-C24:0/2OH-C25:0 und 2OH-C24:1(n-7)/2OH-C25:0 am besten zu eignen. Für die Tierart Schaf sollten die Verhältnisse 2OH-C25:0/2OH-C26:0 und 2OH-C25:0/2OH-C26:1(n-7) bevorzugt eingesetzt werden. Im Vergleich zur bisherigen Verfahrensweise ist mit Hilfe dieser Fettsäureverhältnisse und deren Regressionsformeln eine deutlich präzisere Altersschätzung für die Tierarten Rind und Schaf möglich. Der neue Ansatz bietet die Möglichkeit flexibel auf zukünftige Änderungen der Altersgrenze von SRM zu reagieren. / The immense economical damage and the loss of consumer trust caused by the bovine spongiform encephalopathy (BSE) - crisis resulted in a radical alteration of the feed- and foodlegislation. Besides several other measures, the legislator defined the brain and spinal cord of ruminants older than 12 month as specified risk material (SRM). The special disposal and the processing prohibition of these materials in food and feedstuffs aimed to interrupt the infection chain and to reduce the risk of an infection caused by the BSE pathogen for the consumer. In order to control the processing prohibition of SRM in food, several CNS detection methods were developed. But none of the designed molecular biological or immuno chemical methods has the ability to detect species and age of the CNS. Therefore, a classification of the detected CNS as SRM is not possible. Furthermore, high process temperatures influence nega-tively the detection of almost all CNS markers. NIEDERER and BOLLHALDER (2001) developed a CNS detection procedure based on the analysis of fatty acids, which was improved at the Institute of Food Hygiene, University Leipzig. Advantages of this procedure are the heat stability of the markers and the possibility to identify species and age of the detected CNS. Therefore, this procedure is the first, which facilitates the potential of identifying SRM. The aim of this work was to test the practicability of the improved CNS detection procedure in an external blind trial. During this, possible weak points should be identified and solutions for their elimination presented. Furthermore, the sample preparation of the method should be optimized with regard to cost and time reduction. For the external blind trial the Max-Rubner-Institute, Kulmbach, produced 72 emulsion type sausages. These sausages were analysed according to the procedure described by LÜCKER et al. 2005. The results of the blind trial show the suitability of the procedure for the detection of SRM in meat products in principle. However, some results revealed that the species and age identification required further enhancement. Therefore, the following additional topics of this work were: 1. Optimization of species identification especially with regard to other species 2. Optimization of animal age prediction for cattle and sheep CNS. Selected fatty acids of total 257 CNS samples from cattle, pig, sheep, goat and several poultry species were analysed. These fatty acids were combined to 67 fatty acid ratios. Afterwards, the application of these ratios for the species detection was tested by using the statistical method of Partial Least Square – Discriminant Analysis (PLS-DA). In result, 14 out of 67 fatty acid ratios are suitable for differentiation between the species cattle, pig, sheep, goat and poultry. For the optimization of animal age prediction 37 CNS samples from cattle and eleven CNS samples from sheep were examined. All analyzed fatty acids were checked for their correlation with age. Afterwards strong correlating fatty acids and their fatty acid ratios were examined for their predictive ability by the application of regression analysis. In result, four new fatty acid ratios for age prediction could be identified. For the age prediction of cattle CNS the fatty acid ratios 2OH-C24:0/2OH-C25:0 and 2OH-C24:1(n-7)/2OH-C25:0 are the best choice. In sheep CNS the ratios 2OH-C25:0/2OH-C26:0 and 2OH-C25:0/2OH-C26:1(n-7) should be preferred for prediction of age. In comparison to the previous age prediction method, the application of these fatty acid ratios and their regression formula led to more accurate results. Furthermore, it offers the possibility to adopt to possible variations of the age limits within the SRM definition in future.