Thesis (Ph.D.)--Computing, Georgia Institute of Technology, 2009. / Committee Chair: Abowd, Gregory; Committee Member: Edwards, Keith; Committee Member: Grinter, Rebecca; Committee Member: LaMarca, Anthony; Committee Member: Starner, Thad.

Thesis (Ph. D.)--State University of New York at Binghamton, Thomas J. Watson School of Engineering and Applied Science, Department of Electrical and Computer Engineering, 2008. / Includes bibliographical references.

Thesis (Ph. D.), Memorial University of Newfoundland, 1999. / Bibliography: p. 131-137.

Thesis (Ph. D.)--Lehigh University, 2006. / Includes vita. Includes bibliographical references (leaves 114-124).

Vibration-based fault diagnosis (FD) with a simple spectrum can be complex, especially when considering FD of rotating machinery with multiple bearings like a multi-stage turbine. Various studies have sought to better interpret fault spectra, but the process remains equivocal. Consequently, it has been accepted that the simple spectra requires support from additional techniques, such as orbit analysis. But even orbit analysis can be inconclusive. Though promising, attempts at developing viable methods that rival the failure coverage of spectrum analysis without gaining computational complexity remain protracted. Interestingly, few researchers have developed FD methods for transient machine operation, however, these have proven to be involved. Current practices limit vibration data to a single machine, which usually requires a large unique data history. However, if sharing of data between similar machines with different foundations was possible, the need for unique histories would be mitigated. From readily available works, this has not been encountered. Therefore, a simple but robust vibration-based approach is warranted. In light of this, a novel on-bearing vibration response integration approach for condition monitoring of shaft-related faults irrespective of speed and foundation type is proposed in the present study. Vibration data are acquired at different speeds for: a baseline, unbalance, bow, crack, looseness, misalignment, and rub conditions on three laboratory rigs with dynamically different foundations, namely: rigid, flexible support 1 (FS1) and flexible support 2 (FS2). Testing is done on the rigid rig set up first, then FS1, and afterwards FS2. Common vibration features are computed from the measured data to be input to the proposed approach for further processing. First, the proposed approach is developed through its application to a machine at a steady speed in a novel Single-speed FD technique which exploits a single vibration sensor per bearing and fusion of features from different bearings for FD. Initially, vibration features are supplemented with bearing temperature readings with improved classification compared to vibration features alone. However, it is observed that temperature readings are insensitive to faults on the FS1 and FS2 rigs, when compared to vibration features, which are standardised for consistent classification on the different rigs tested. Thus, temperature is not included as a final feature. The observed fault classifications on the different rigs at different speeds with the standardised vibration features are encouraging. Thereafter, a novel Unified Multi-speed FD technique that is based on the initial proposed approach and which works by fusion of vibration features from different bearings at different speeds in a single analysis step for FD is proposed. Experiments on the different rigs repeatedly show the novel Multi-speed technique to be suitable for transient machine operation. Then, a novel generic Multi-foundation Technique (also based on the proposed approach) that allows sharing of vibration data of a wide range of fault conditions between two similarly configured machines with similar speed operation but different foundations is implemented to further mitigate data requirements in the FD process. Observations made with the rigs during steady and transient speed tests show this technique is applicable in situations where data history is available on one machine but lacking on the other. Comparison of experimental results with results obtained from theoretical simulations indicates the approach is consistent. Thus, the proposed approach has the potential for practical considerations.

abstract: In today's global market, companies are facing unprecedented levels of uncertainties in supply, demand and in the economic environment. A critical issue for companies to survive increasing competition is to monitor the changing business environment and manage disturbances and changes in real time. In this dissertation, an integrated framework is proposed using simulation and online calibration methods to enable the adaptive management of large-scale complex supply chain systems. The design, implementation and verification of the integrated approach are studied in this dissertation. The research contributions are two-fold. First, this work enriches symbiotic simulation methodology by proposing a framework of simulation and advanced data fusion methods to improve simulation accuracy. Data fusion techniques optimally calibrate the simulation state/parameters by considering errors in both the simulation models and in measurements of the real-world system. Data fusion methods - Kalman Filtering, Extended Kalman Filtering, and Ensemble Kalman Filtering - are examined and discussed under varied conditions of system chaotic levels, data quality and data availability. Second, the proposed framework is developed, validated and demonstrated in `proof-of-concept' case studies on representative supply chain problems. In the case study of a simplified supply chain system, Kalman Filtering is applied to fuse simulation data and emulation data to effectively improve the accuracy of the detection of abnormalities. In the case study of the `beer game' supply chain model, the system's chaotic level is identified as a key factor to influence simulation performance and the choice of data fusion method. Ensemble Kalman Filtering is found more robust than Extended Kalman Filtering in a highly chaotic system. With appropriate tuning, the improvement of simulation accuracy is up to 80% in a chaotic system, and 60% in a stable system. In the last study, the integrated framework is applied to adaptive inventory control of a multi-echelon supply chain with non-stationary demand. It is worth pointing out that the framework proposed in this dissertation is not only useful in supply chain management, but also suitable to model other complex dynamic systems, such as healthcare delivery systems and energy consumption networks. / Dissertation/Thesis / Ph.D. Engineering 2010

La télédétection contribue à la cartographie et la modélisation des paramètres forestiers. Ce sont les systèmes optiques et radars qui sont le plus généralement utilisés pour extraire des informations utiles à la caractérisation de ces paramètres. Ces systèmes ont montré des bons résultats pour estimer la biomasse dans certains biomes. Cependant, ils présentent des limitations importantes pour des forêts ayant un niveau de biomasse élevé. En revanche, la télédétection LiDAR s’est avérée être une bonne technique pour l'estimation des paramètres forestiers tels que la hauteur de la canopée et la biomasse. Alors que les LiDAR aéroportés acquièrent en général des données avec une forte densité de points mais sur des petites zones en raison du coût de leurs acquisitions, les données LiDAR satellitaires acquises par le système spatial (GLAS) ont une densité d'acquisition faible mais avec une couverture géographique mondiale. Il est donc utile d'analyser la pertinence de l'intégration des hauteurs estimées à partir des capteurs LiDAR et des données auxiliaires afin de proposer une carte de la hauteur des arbres avec une bonne précision et une résolution spatiale élevée. En outre, l'estimation de la hauteur des arbres à partir du GLAS est difficile compte tenu de l'interaction complexe entre les formes d'onde LiDAR, le terrain et la végétation, en particulier dans les forêts denses. Par conséquent, la recherche menée dans cette thèse vise à: 1) Estimer et valider la hauteur des arbres en utilisant des données acquises par le LiDAR aéroportés et GLAS. 2) évaluer le potentiel de la fusion des données LiDAR (avec les données aéroportées ou satellitaires) et des données auxiliaires pour l'estimation de la hauteur des arbres à une échelle régionale (Guyane française). L'estimation de la hauteur avec le LiDAR aéroporté a montré une EQM sur les estimations de 1,6 m. Ensuite, le potentiel de GLAS pour l'estimation de la hauteur a été évalué en utilisant des modèles de régression linéaire (ML) ou Random Forest (RF) avec des métriques provenant de la forme d'onde et de l'ACP. Les résultats ont montré que les modèles d’estimation des hauteurs avaient des précisions semblables en utilisant soit les métriques de GLAS ou les composantes principales (PC) obtenues à partir des formes d’onde GLAS (EQM ~ 3,6 m). Toutefois, un modèle de régression (ML ou RF) basé sur les PCs est une alternative pour l'estimation de la hauteur, car il ne nécessite pas l'extraction de certaines métriques de GLAS qui sont en général difficiles à dériver dans les forêts denses.Finalement, la hauteur extraite à la fois des données LiDAR aéroporté et GLAS a servi tout d'abord à spatialiser la hauteur en utilisant les données environnementales cartographiées. En utilisant le RF, la spatialisation de la hauteur des arbres a montré une EQM sur les estimations de la hauteur de 6,5 m à partir de GLAS et de 5,8 m à partir du LiDAR aéroporté. Ensuite, afin d'améliorer la précision de la spatialisation de la hauteur, la technique régression-krigeage (krigeage des résidus de la régression du RF) a été utilisée. Les résultats de la régression-krigeage indiquent une diminution de l'erreur quadratique moyenne de 6,5 à 4,2 m pour les cartes de la hauteur de la canopée à partir de GLAS, et de 5,8 à 1,8 m pour les cartes de la hauteur de la canopée à partir des données LiDAR aéroporté. Enfin, afin d'étudier l'impact de l'échantillonnage spatial des futures missions LiDAR sur la précision des estimations de la hauteur de la canopée, six sous-ensembles ont été extraits de de la base LiDAR aéroporté. Ces six sous-ensembles de données LiDAR ont respectivement un espacement des lignes de vol de 5, 10, 20, 30, 40 et 50 km. Finalement, en utilisant la technique régression-krigeage, l’EQM sur la carte des hauteurs était de 1,8 m pour le sous-ensemble ayant des lignes de vol espacés de 5 km, et a augmentée jusqu’à 4,8 m pour le sous-ensemble ayant des lignes de vol espacés de 50 km. / Remote sensing has facilitated the techniques used for the mapping, modelling and understanding of forest parameters. Remote sensing applications usually use information from either passive optical systems or active radar sensors. These systems have shown satisfactory results for estimating, for example, aboveground biomass in some biomes. However, they presented significant limitations for ecological applications, as the sensitivity from these sensors has been shown to be limited in forests with medium levels of aboveground biomass. On the other hand, LiDAR remote sensing has been shown to be a good technique for the estimation of forest parameters such as canopy heights and above ground biomass. Whilst airborne LiDAR data are in general very dense but only available over small areas due to the cost of their acquisition, spaceborne LiDAR data acquired from the Geoscience Laser Altimeter System (GLAS) have low acquisition density with global geographical cover. It is therefore valuable to analyze the integration relevance of canopy heights estimated from LiDAR sensors with ancillary data (geological, meteorological, slope, vegetation indices etc.) in order to propose a forest canopy height map with good precision and high spatial resolution. In addition, estimating forest canopy heights from large-footprint satellite LiDAR waveforms, is challenging given the complex interaction between LiDAR waveforms, terrain, and vegetation, especially in dense tropical and equatorial forests. Therefore, the research carried out in this thesis aimed at: 1) estimate, and validate canopy heights using raw data from airborne LiDAR and then evaluate the potential of spaceborne LiDAR GLAS data at estimating forest canopy heights. 2) evaluate the fusion potential of LiDAR (using either sapceborne and airborne data) and ancillary data for forest canopy height estimation at very large scales. This research work was carried out over the French Guiana.The estimation of the canopy heights using the airborne showed an RMSE on the canopy height estimates of 1.6 m. Next, the potential of GLAS for the estimation of canopy heights was assessed using multiple linear (ML) and Random Forest (RF) regressions using waveform metrics and principal component analssis (PCA). Results showed canopy height estimations with similar precisions using either LiDAR metrics or the principal components (PCs) (RMSE ~ 3.6 m). However, a regression model (ML or RF) based on the PCA of waveform samples is an interesting alternative for canopy height estimation as it does not require the extraction of some metrics from LiDAR waveforms that are in general difficult to derive in dense forests, such as those in French Guiana. Next, canopy heights extracted from both airborne and spaceborne LiDAR were first used to map canopy heights from available mapped environmental data (geological, meteorological, slope, vegetation indices etc.). Results showed an RMSE on the canopy height estimates of 6.5 m from the GLAS dataset and of 5.8 m from the airborne LiDAR dataset. Then, in order to improve the precision of the canopy height estimates, regression-kriging (kriging of random forest regression residuals) was used. Results indicated a decrease in the RMSE from 6.5 to 4.2 m for the regression-kriging maps from the GLAS dataset, and from 5.8 to 1.8 m for the regression-kriging map from the airborne LiDAR dataset. Finally, in order to study the impact of the spatial sampling of future LiDAR missions on the precision of canopy height estimates, six subsets were derived from the airborne LiDAR dataset with flight line spacing of 5, 10, 20, 30, 40 and 50 km (corresponding to 0.29, 0.11, 0.08, 0.05, 0.04, and 0.03 points/km², respectively). Results indicated that using the regression-kriging approach, the precision on the canopy height map was 1.8 m with flight line spacing of 5 km and decreased to an RMSE of 4.8 m for the configuration for the 50 km flight line spacing.

O crescente volume de informações disponíveis na rede mundial de computadores, gera a necessidade do uso de ferramentas que sejam capazes de localizá-las e ordenálas, de forma cada vez mais precisa e que demandem cada vez menos recursos computacionais. Esta necessidade tem motivado pesquisadores a estudar e desenvolver modelos e técnicas que atendam esta demanda. Estudos recentes têm sinalizado que utilizar vários ordenamentos (rankings) previamente montados possibilita o retorno e ordenação de objetos de qualquer natureza com mais eficiência, principalmente pelo fato de haver uma redução no custo da busca pela informação. Este processo, conhecido como fusão de rankings, permite que se obtenha um ordenamento com base na opinião de diversos juízes (critérios), o que possibilita considerar um grande número de fontes, tanto geradas automaticamente como por especialistas. Entretanto os modelos propostos até então tem apresentado várias limitações na sua aplicação: desde a quantidade de rankings envolvidos até, principalmente, a utilização de rankings parciais. A proposta desta dissertação é apresentar um modelo de fusão de rankings que busca estabelecer um consenso entre as opiniões (rankings) dos diferentes juízes envolvidos, considerando distintos graus de relevância ou importância entre eles. A base desta proposta está na Análise de Preferência, um conjunto de técnicas que permite o tratamento da multidimensionalidade dos dados envolvidos. Ao ser testado em uma aplicação real, o modelo mostrou conseguir suprir algumas limitações apresentadas em outras abordagens, bem como apresentou resultados similares aos das aplicações originais. Esta pesquisa, ainda contribui, com a especificação de um sistema Web baseado em tecnologias open source, o qual permite que qualquer pessoa possa realizar a fusão de rankings. / The growing volume of available information on the web creates the need to use tools that are capable of retrieve and ordering this information, ever more precise and using less computer resources. This need has motivated researchers to study and develop models and techniques that solve this problem. Recent studies have indicated that use multiple rankings previously mounted makes possible the return and sorting of the objects of any kind with more efficiency, mainly because there is a reduction in the cost of searching for information. This process, called ranking fusion, provide a ranking based on the opinion of several judges (criteria), considering a large number of sources, both generated automatically and also by specialists. However the proposed models have shown severe limitations in its application: from the amount involved rankings to the use of partial rankings. The proposal of this dissertation is to show a model of ranking fusion that seeks to establish a consensus between the judgement (rankings) of the various judges involved, considering different degrees of relevance or importance among them. The baseline of this proposal is the Preference Analysis, a set of techniques that allows the treatment of multidimensional data handling. During tests in a real application, the model supplied some limitations presented by other approaches, and presented results similar to the original applications. Additionally, this research contributes with the specification of a web system based on open-sources technologies, enabling the realization of fusion rankings by anyone.

Cette thèse concerne la perception de l’environnement pour le guidage automatique d’un robot mobile. Lorsque l’on souhaite réaliser un système de navigation autonome, plusieurs éléments doivent être abordés. Parmi ceux-ci nous traiterons de la franchissabilité de l’environnement sur la trajectoire du véhicule. Cette franchissabilité dépend notamment de la géométrie et du type de sol mais également de la position du robot par rapport à son environnement (dans un repère local) ainsi que l’objectif qu’il doit atteindre (dans un repère global). Les travaux de cette thèse traitent donc de la perception de l’environnement d’un robot au sens large du terme en adressant la cartographie de l’environnement et la localisation du véhicule. Pour cela un système de fusion de données est proposé afin d’estimer ces informations. Ce système de fusion est alimenté par plusieurs capteurs dont une caméra, un télémètre laser et un GPS. L’originalité de ces travaux porte sur la façon de combiner ces informations capteurs. A la base du processus de fusion, nous utilisons un algorithme d’odométrie visuelle basé sur les images de la caméra. Pour accroitre la précision et la robustesse l’initialisation de la position des points sélectionnés se fait grâce à un télémètre laser qui fournit les informations de profondeur. De plus, le positionnement dans un repère global est effectué en combinant cette odométrie visuelle avec les informations GPS. Pour cela un procédé a été mis en place pour assurer l’intégrité de localisation du véhicule avant de fusionner sa position avec les données GPS. La cartographie de l’environnement est toute aussi importante puisqu’elle va permettre de calculer le chemin qui assurera au véhicule une évolution sans risque de collision ou de renversement. Dans cette optique, le télémètre laser déjà présent dans le processus de localisation est utilisé pour compléter la liste courante de points 3D qui matérialisent le terrain à l’avant du véhicule. En combinant la localisation précise du véhicule avec les informations denses du télémètre il est possible d’obtenir une cartographie précise, dense et géo-localisée de l’environnement. Tout ces travaux ont été expérimentés sur un simulateur robotique développé pour l’occasion puis sur un véhicule tout-terrain réel évoluant dans un monde naturel. Les résultats de cette approche ont montré la pertinence de ces travaux pour le guidage autonome de robots mobiles. / This thesis addresses the perception of the environment for the automatic guidance of a mobile robot. When one wishes to achieve autonomous navigation, several elements must be addressed. Among them we will discuss the traversability of the environment on the vehicle path. This traversability depends on the ground geometry and type and also the position of the robot in its environment (in a local coordinate system) taking into acount the objective that must be achieved (in a global coordinate system).The works of this thesis deal with the environment perception of a robot inthe broad sense by addressing the mapping of the environment and the location of the vehicle. To do this, a data fusion system is proposed to estimate these informations. The fusion system is supplied by several low cost sensors including a camera, a rangefinder and a GPS receiver. The originality of this work focuses on how to combine these sensors informations. The base of the fusion process is a visual odometry algorithm based on camera images. To increase the accuracy and the robustness, the initialization of the selected points position is done with a rangefinder that provides the depth information.In addition, the localization in a global reference is made by combining the visual odometry with GPS information. For this, a process has been established to ensure the integrity of localization of the vehicle before merging its position with the GPS data. The mapping of the environment is also important as it will allow to compute the path that will ensure an evolution of the vehicle without risk of collision or overturn. From this perspective, the rangefinder already present in the localization process is used to complete the current list of 3D points that represent the field infront of the vehicle. By combining an accurate localization of the vehicle with informations of the rangefinder it is possible to obtain an accurate, dense and geo-located map environment. All these works have been tested on a robotic simulator developed for this purpose and on a real all-terrain vehicle moving in a natural world. The results of this approach have shown the relevance of this work for autonomous guidance of mobile robots.

O crescente volume de informações disponíveis na rede mundial de computadores, gera a necessidade do uso de ferramentas que sejam capazes de localizá-las e ordenálas, de forma cada vez mais precisa e que demandem cada vez menos recursos computacionais. Esta necessidade tem motivado pesquisadores a estudar e desenvolver modelos e técnicas que atendam esta demanda. Estudos recentes têm sinalizado que utilizar vários ordenamentos (rankings) previamente montados possibilita o retorno e ordenação de objetos de qualquer natureza com mais eficiência, principalmente pelo fato de haver uma redução no custo da busca pela informação. Este processo, conhecido como fusão de rankings, permite que se obtenha um ordenamento com base na opinião de diversos juízes (critérios), o que possibilita considerar um grande número de fontes, tanto geradas automaticamente como por especialistas. Entretanto os modelos propostos até então tem apresentado várias limitações na sua aplicação: desde a quantidade de rankings envolvidos até, principalmente, a utilização de rankings parciais. A proposta desta dissertação é apresentar um modelo de fusão de rankings que busca estabelecer um consenso entre as opiniões (rankings) dos diferentes juízes envolvidos, considerando distintos graus de relevância ou importância entre eles. A base desta proposta está na Análise de Preferência, um conjunto de técnicas que permite o tratamento da multidimensionalidade dos dados envolvidos. Ao ser testado em uma aplicação real, o modelo mostrou conseguir suprir algumas limitações apresentadas em outras abordagens, bem como apresentou resultados similares aos das aplicações originais. Esta pesquisa, ainda contribui, com a especificação de um sistema Web baseado em tecnologias open source, o qual permite que qualquer pessoa possa realizar a fusão de rankings. / The growing volume of available information on the web creates the need to use tools that are capable of retrieve and ordering this information, ever more precise and using less computer resources. This need has motivated researchers to study and develop models and techniques that solve this problem. Recent studies have indicated that use multiple rankings previously mounted makes possible the return and sorting of the objects of any kind with more efficiency, mainly because there is a reduction in the cost of searching for information. This process, called ranking fusion, provide a ranking based on the opinion of several judges (criteria), considering a large number of sources, both generated automatically and also by specialists. However the proposed models have shown severe limitations in its application: from the amount involved rankings to the use of partial rankings. The proposal of this dissertation is to show a model of ranking fusion that seeks to establish a consensus between the judgement (rankings) of the various judges involved, considering different degrees of relevance or importance among them. The baseline of this proposal is the Preference Analysis, a set of techniques that allows the treatment of multidimensional data handling. During tests in a real application, the model supplied some limitations presented by other approaches, and presented results similar to the original applications. Additionally, this research contributes with the specification of a web system based on open-sources technologies, enabling the realization of fusion rankings by anyone.