Cette thèse de doctorat analyse les concepts impliqués dans la prise de décisions de groupes d'agents et applique ces concepts dans la création d'un cadre théorique et pratique qui permet la reconnaissance de comportements de groupes. Nous allons présenter une vue d'ensemble de la théorie de l'intention, étudiée dans le passé par quelques grands théoriciens comme Searle, Bratmann et Cohen, et nous allons montrer le lien avec des recherches plus récentes dans le domaine de la reconnaissance de comportements. Nous allons étudier les avantages et inconvénients des techniques les plus avancées dans ce domaine et nous allons créer un nouveau modèle qui représente et détecte les comportements de groupes. Ce nouveau modèle s'appelle Multiagent-Abstract Hidden Markov mEmory Model (M-AHMEM) et résulte de la fusion de modèles déjà existants, le but étant de créer une approche unifiée du problème. La plus grande partie de cette thèse est consacrée à la présentation détaillée du M-AHMEM et de l'algorithme responsable de la reconnaissance de comportements. Notre modèle sera testé sur deux applications différentes : l'analyse gesturale humaine et la fusion multimodale des données audio et vidéo. A travers ces deux applications, nous avançons l'argument qu'un ensemble de données constitué de plusieurs variables corrélées peut être analysé efficacement sous un cadre unifié de reconnaissance de comportements. Nous allons montrer que la corrélation entre les différentes variables peut être modélisée comme une coopération ayant lieu à l'intérieur d'une équipe et que la reconnaissance de comportements constitue une approche moderne de classification et de reconnaissance de patrons.

Safety is considered as one of the most important areas in future research and development within the automotive industry. New functionality, such as driver support and active/passive safety systems are examples where development mainly focuses on safety. At the same time, the trend is towards more complex systems, increased software dependence and an increasing amount of sensors and actuators, resulting in a higher risk associated with software and hardware failures. In the area of functional safety, standards such as ISO 26262 assess safety mainly focusing on qualitative assessment techniques, whereas usage of quantitative techniques is a growing area in academic research. This thesis considers the field functional safety, with the emphasis on how hardware and software failure probabilities can be used to quantitatively assess safety of a system/function. More specifically, this thesis presents a method for quantitative safety assessment using Bayesian networks for probabilistic modeling. Since the safety standard ISO 26262 is becoming common in the automotive industry, the developed method is adjusted to use information gathered when implementing this standard. Continuing the discussion about safety, a method for modeling faults and failures using Markov models is presented. These models connect to the previous developed Bayesian network and complete the quantitative safety assessment. Furthermore, the potential for implementing the discussed models in the Modelica language is investigated, aiming to find out if models such as these could be useful in practice to simplify design work, in order to meet future safety goals.

Stock Market is always being the most important role in modern capital market. And Stock Market is becoming one the most popular investment tools these days. Because of the Globalization of capital markets, the spreading of capital becomes faster and easier. The development of capital markets evoke the interesting of scholars and the field of stock market prediction attract scholars and researchers from different background. There are two approaches of predicting stock market - fundamental analysis and technical analysis. The purpose of my work was to predict three stock markets in the world, namely Taiwan Weighted Index (IDXWT), Hong Kong Hang Seng Index (HSI) and Dow Jones Industrial Average (DJI) using technical analysis and Dynamic Bayesian Network (DBN).This thesis is based on Wang¡¦s thesis [Wan05] ¡§Investment Decision Support with Dynamic Bayesian Networks¡¨. According to different characteristic of 3 stock markets, we divide 3 different markets into 3 experiments. For each market, we expect we can find the best indicators and trading signals. The first experiment involves Taiwan Weighted Index as our prediction target; the second one uses Hong Kong Hang Seng Index and the third experiment employs Dow Jones Industrial Average. As a result, Taiwan Stock market (both 15-day and 20-day Moving Average)can make higher returns than buy-and-hold, RSI_6 and KD. And we also have the same conclusion of Hang Seng Index and Dow Jones Industrial Average. The best return from 15-day MA and 20-day MA of Taiwan Stock market is 47.95% and 60.21%, respectively. Moreover, the best result of Hang Seng Index is 60.06% for 4 years and 25.83% for Dow Jones Industrial Average. All of the best results can make higher returns than each of their buy-and-hold, RSI_6 and KD. In the conclusion, we may say that this paper can provide a direction to investors while they are using these technical indicators to predict these particular stock markets.

In this thesis, I introduce a multivariate discretization algorithm to discretize the continuous variables of clinical pathways of Hemodialysis and use the clustering algorithm to shift time stamps to reduce the number of nodes of Bayesian networks. The generalized sequential patterns algorithm is used to find the possible patterns, which have far-reaching effect on the next nodes of the Bayesian networks of Hemodialysis. Bayesian network is a graphical model that encodes probabilistic relationships among variables of interest, and easily incorporates with new instances to maintain rules up to date. Bayesian networks are used to represent knowledge of frequent state transitions in medical logs. Bayesian networks and sequential patterns algorithms can only handle discrete or categorical data. Therefore, we have to discretize the continuous variables with suitable technique to generalize the node, and shift the time stamps of nodes to reduce the variations in time. With these generalizations, we improve the problem of over-fitting of the Bayesian networks of Hemodialysis. We expect the discovered patterns can give more information to medical professionals and help them to build the reciprocal cycle of knowledge management of Hemodialysis.

Design problems in engineering are typically complex, and are therefore decomposed into a hierarchy of smaller, simpler design problems by the design management. It is often the case in a hierarchical design problem that an upstream design team’s achievable performance space becomes the design space for a downstream design team. A Bayesian network classifier is proposed in this research to map and classify a design team’s attainable performance space. The classifier will allow for enhanced collaboration between design teams, letting an upstream design team efficiently identify and share their attainable performance space with a downstream design team. The goal is that design teams can work concurrently, rather than sequentially, thereby reducing lead time and design costs. In converging to a design solution, intelligently narrowing the design space allows for resources to be focused in the most beneficial regions. However, the process of narrowing the design space is non-trivial, as each design team must make performance trade-offs that may unknowingly affect other design teams. The performance space mapping provided by the Bayesian network classifier allows designers to better understand the consequences of narrowing the design space. This knowledge allows design decisions to be made at the system-level, and be propagated down to the subsystem-level, leading to higher quality designs. The proposed methods of mapping the performance space are then applied to a hierarchical, multi-level metamaterial design problem. The design problem explores the possibility of designing and fabricating composite materials that have desirable macro-scale mechanical properties as a result of embedded micro-scale inclusions. The designed metamaterial is found to have stiffness and loss properties that surpass those of conventional composite materials. / text

The increasing complexity of today’s vehicles gives drivers help with everything from adaptive cruisecontrol to warning lights for low fuel level. But the increasing functionality also increases the risk offailures in the system. To prevent system failures, different safety analytic methods can be used, e.g.,fault trees and/or FMEA-tables. These methods are generally performed manually, and due to thegrowing system size the time spent on safety analysis is growing with increased risk of human errors. If the safety analysis can be automated, lots of time can be saved. This thesis investigates the possibility to generate fault trees from safety requirements as wellas which additional information, if any, that is needed for the generation. Safety requirements are requirements on the systems functionality that has to be fulfilled for the safety of the system to be guaranteed. This means that the safety of the truck, the driver, and the surroundings, depend on thefulfillment of those requirements. The requirements describing the system are structured in a graphusing contract theory. Contract theory defines the dependencies between requirements and connectsthem in a contract structure. To be able to automatically generate the fault tree for a system, information about the systems failure propagation is needed. For this a Bayesian network is used. The network is built from the contract structure and stores the propagation information in all the nodes of the network. This will result in a failure propagation network, which the fault tree generation will be generated from. The failure propagation network is used to see which combinations of faults in the system can violate thesafety goal, i.e., causing one or several hazards. The result of this will be the base of the fault tree. The automatic generation was tested on two different Scania systems, the fuel level displayand the dual circuit steering. Validation was done by comparing the automatically generated trees withmanually generated trees for the two systems showing that the proposed method works as intended. The case studies show that the automated fault tree generation works if the failure propagationinformation exists and can save a lot of time and also minimize the errors made by manuallygenerating the fault trees. The generated fault trees can also be used to validate written requirementsto by analyzing the fault trees created from them.

The planning and scheduling of appropriate resources is essential in engineering design for delivering quality products on time, within cost and at acceptable risk. There is an inherent complexity in deciding what resources should perform which tasks taking into account their effectiveness towards completing the task, whilst adjusting to their availabilities. The right resources must be applied to the right tasks in the correct order. In this context, process modelling and simulation could aid in resource management decision making. However, most approaches define resources as elements needed to perform the activities without defining their characteristics, or use a single classification such as human designers. Other resources such as computational and testing resources, amongst others have been overlooked during process planning stages. In order to achieve this, literature and empirical investigations were conducted. Firstly, literature investigations focused on what elements have been considered design resources by current modelling approaches. Secondly, empirical studies characterised key design resources, which included designers, computational, testing and prototyping resources. The findings advocated for an approach that allows allocation flexibility to balance different resource instances within the process. In addition, capabilities to diagnose the impact of attaining specific performance to search for a preferred resource allocation were also required. Therefore, the thesis presents a new method to model different resource types with their attributes and studies the impact of using different instances of those resources by simulating the model and analysing the results. The method, which extends a task network model, Applied Signposting Model (ASM), with Bayesian Networks (BN), allows testing the influence of using different resources combinations on process performance. The model uses BN within each task to model different instances of resources that carries out the design activities (computational, designers and testing) along with its configurable attributes (time, risk, learning curve etc.), and tasks requirements. The model was embedded in an approach and was evaluated by applying it to two aerospace case studies. The results identified insights to improve process performance such as the best performing resource combinations, resource utilisation, resource sensitive activities, the impact of different variables, and the probability of reaching set performance targets by the different resource instances.

There is an emerging trend of manufacturing companies offering combined products and services to customers as integrated solutions. Availability contracts are an apt instance of such offerings, where product use is guaranteed to customer and is enforced by incentive-penalty schemes. Uncertainties in such an industry setting, where all stakeholders are striving to achieve their respective performance goals and at the same time collaborating intensively, is increased. Understanding through-life uncertainties and their impact on cost is critical to ensure sustainability and profitability of the industries offering such solutions. In an effort to address this challenge, the aim of this research study is to provide an approach for the analysis of uncertainties in Product Service System (PSS) delivered in business-to-business application by specifying a procedure to identify, characterise and model uncertainties with an emphasis to provide decision support and prioritisation of key uncertainties affecting the performance outcomes. The thesis presents a literature review in research areas which are at the interface of topics such as uncertainty, PSS and availability contracts. From this seven requirements that are vital to enhance the understanding and quantification of uncertainties in Product Service System are drawn. These requirements are synthesised into a conceptual uncertainty framework. The framework prescribes four elements, which include identifying a set of uncertainties, discerning the relationships between uncertainties, tools and techniques to treat uncertainties and finally, results that could ease uncertainty management and analysis efforts. The conceptual uncertainty framework was applied to an industry case study in availability contracts, where each of the four elements was realised. This application phase of the research included the identification of uncertainties in PSS, development of a multi-layer uncertainty classification, deriving the structure of Bayesian Network and finally, evaluation and validation of the Bayesian Network. The findings suggest that understanding uncertainties from a system perspective is essential to capture the network aspect of PSS. This network comprises of several stakeholders, where there is increased flux of information and material flows and this could be effectively represented using Bayesian Networks.

Une des conséquences les plus marquantes de l'évolution actuelle de l’industrie ferroviaire est l'augmentation des contraintes exercées aussi bien sur les voies que sur les matériels roulants ; tant en termes de sollicitations, de charges, de fréquences, qu'en termes d'exigences de disponibilité et de sécurité. De ce fait, la recherche de politiques de maintenance optimales répondant aux objectifs de disponibilité, de coûts, de sécurité est devenue un sujet particulièrement d'actualité. Pour répondre à cette demande d’ajustement des stratégies de maintenance, le formalisme des réseaux bayésiens est une approche de plus en plus utilisée pour développer des outils d'aide à la décision. Afin de s’affranchir de l’hypothèse markovienne restrictive imposée par l’utilisation « standard » des réseaux bayésiens, une structure originale a été proposée pour modéliser finement un processus de dégradation dans le cadre discret à partir de distributions de temps de séjour quelconques. Cette approche, dénommée Modèles Graphiques de Durée, autorise une finesse de modélisation du processus de dégradation qui permet de reproduire le comportement de systèmes multi-composants et multi-états, tout en tenant compte de variables exogènes. Cette modélisation semi-markovienne de la dégradation a, jusqu'à présent, été utilisée surtout pour évaluer ou comparer des stratégies de maintenance pouvant mêler des approches correctives, systématiques ou conditionnelles. Cette thèse vise à étendre les travaux précédents aux actions de maintenance prévisionnelle. Cette approche, qualifiée également de pronostic, offre en effet l’avantage d’une prédiction de l’instant optimal d’intervention maximisant la durée de fonctionnement du système avant intervention, tout en satisfaisant les contraintes d’exploitation et d’entretien. Les systèmes considérés sont à espaces d’états discrets et finis, périodiquement observables, situation fréquente pour de nombreuses applications industrielles, notamment dans le domaine des transports. Ces travaux de thèse proposent, à partir du formalisme des réseaux bayésiens dynamiques et des modèles graphiques de durée, des outils de pronostic dans le but de permettre la modélisation de politiques de maintenance préventives prévisionnelle. Pour répondre à cet objectif, un algorithme de pronostic basé sur des distributions de temps de séjour a tout d’abord été introduit, dans le but de calculer une estimation de la durée de vie résiduelle (RUL) d'un système et de la mettre à jour à chaque fois qu’un nouveau diagnostic est disponible. Pour améliorer la précision des calculs de pronostic, un nouveau modèle de dégradation a ensuite été proposé pour tenir compte de l'existence éventuelle de plusieurs dynamiques de dégradation coexistantes. Son principe consiste à identifier à chaque instant un mode de dégradation actif, puis à répercuter cette information sur les temps de séjour considérés dans les états suivants par l'utilisation de lois de temps de séjour conditionnelles. Enfin, des solutions pour diminuer la complexité des calculs d'inférence exacte sont proposées / One of the most important consequences due to current developments in the rail industry is the increase of stresses on tracks and rolling stock; in terms of loads, frequencies, and both in terms of availability and security requirements. Therefore, looking for optimal maintenance policies to meet the availability, cost and security objectives has become a particularly topical subject. To address this need of maintenance strategy adjustment, approaches using bayesian networks have increasingly been used for the development of decision support tools. To overcome the restrictive Markovian assumption induced by the use of standard bayesian networks, a specific structure has been proposed to accurately model a degradation process in discrete case using any kind of sojourn time distributions. This approach called "Graphical duration model" make possible to describe multicomponent and multi state system behaviours by taking into account many exogenous variables. This semi-markovian modelling of the degradation has mainly been used to evaluate and compare different maintenance strategies based on corrective, systematic and conditional approaches. This PhD thesis aims to extend previous works to predictive maintenance policies. This approach, based on prognosis computations, has the advantage to predict the optimal intervention time maximizing the remaining useful life of the system and both satisfying operating and maintaining constraints. Considered systems have finite discrete state spaces and are periodically observable as many existing ones in the industry and particularly in the field of transport systems. The presented works, based on the dynamic bayesian network formalism and the graphical duration model, propose prognostic tools in order to model the set of predictive maintenance policies. A prognosis algorithm is first introduced to compute the remaining useful life (RUL) of the system and update this estimation each time a new diagnosis is available. To improve the prognosis estimation accuracy, a new degradation model is proposed to take into account the possible existence of many coexisting degradation modes. The principle is to identify at each time the active degradation mode and then to use this information to choose sojourn times considered in next states using conditional sojourn times distributions. At last, some solutions to reduce the complexity of inference computations are proposed

The focus of this study is to gain a better understanding of the hazards affecting the transportation of avocados from farm to packhouse by developing an effective risk assessment tool farmers can use. The transport related factors considered in this study encompass all hazards which may affect the avocado, from the point the fruit is picked to the point the avocado is packed at the packhouse. The study has been undertaken in five stages, namely:  A literature study split into four main stages, including an investigation into avocado specific hazards, transportation related hazards, market influencers and investigating analysis tools.  Data collection (including environmental indicators, accelerations and GPS measurements) stemming from field tests conducted with a smart avocado device (smAvo);  Data analysis of the smAvos, including assessing the kinetic energy the avocado experiences;  Risk analysis and Bayesian Network Development including those hazards identified in the literature study as well as from the smAvo, and  Bayesian Network analysis, using Delphi Fuzzy methodology and smAvo data to determine the influence of the combination of risk factors identified. The risk assessment tool was developed through the use of Bayesian Networks. This tool eliminates the guesswork of what causes the largest reduction in shelf life/waste and therefore profit. The Network considers the joint probability of these hazards, and posterior probabilities of any subset of variables when evidence is introduced. The Bayesian Network is analysed and optimised by means of finding factors that will cause the greatest improvement of shelf life and decreased damage. A converse analysis is done by determining the effect of, for example poor road conditions or truck type. The result of this analysis provides the farmer with a decision-making tool which will optimise processes, increase profits (by reducing waste) and eliminate any guesswork. The Network can be used by the farmer and updated as new evidence is discovered. The analysis concludes with the most damaging areas within the network is at harvest, followed by truck transportation effects, packhouse conditions and lastly farm transportation effects. In order to optimise the network, emphasis is put on the plant condition, followed by any delay in transportation and the picking technique used during harvest. A “what-if” analysis was done which concluded poor road conditions can increase overall damage by 0.44 per cent, whereas poor harvest conditions can increase this to 12.57 per cent. / Dissertation (MEng (Transportation Engineering))--University of Pretoria, 2020. / Civil Engineering / MEng (Transportation Engineering) / Unrestricted