No description available.

Large, ``rigid'' structures made of interconnecting beams and tendons represent a type of construction that has many engineering benefits. Lightweight, strong, and generally constructed of standard-shaped metal framework, they also incur significant challenges to control vibration. The construction that makes them strong, light, and ``rigid'' also makes them have very small inherent structural damping, and highly-complex modal structure. A myriad of control techniques have been developed to work on this problem with success usually in very small bandwidths related to a specific frequency ``hot-spot'' around a few specific modes. This work describes the design, analysis, and implementation of a novel controller configuration applicable to broadband vibration suppression on a large, uncertain resonant structure. The measurement, identification, characterization, and modeling of a large, flexible, lightly-damped test structure with in excess of 1000 modes in 50-5000Hz range was used as the basis for choosing a control configuration. This choice leverages the relative benefits of different control types to obtain one with a combination of the best features of all of them. High-order and low-order feedback, and feed-forward controller configurations were all used in different frequency ranges. Real design tradeoffs such as computation complexity, model accuracy, and available actuator technologies were fundamental to the design choices. Measured individual modal reduction was as much as 15dB for feedback control, 20dB for feed-forward control, and 4dB broadband over the range of 50-5000Hz. / Ph. D.

Thesis (Ph. D.)--University of Washington, 2004. / Vita. Includes bibliographical references (leaves 82-92).

No description available.

In practice, machine learners often care about two key issues: one is how to obtain a more accurate answer with limited data, and the other is how to handle large-scale data (often referred to as “Big Data” in industry) for efficient inference and optimization. One solution to the first issue might be aggregating learned predictions from diverse local models. For the second issue, integrating the information from subsets of the large-scale data is a proven way of achieving computation reduction. In this thesis, we have developed some novel frameworks and schemes to handle several scenarios in each of the two salient issues. For aggregating diverse models – in particular, aggregating probabilistic predictions from different models – we introduce a spectrum of compositional methods, Rényi divergence aggregators, which are maximum entropy distributions subject to biases from individual models, with the Rényi divergence parameter dependent on the bias. Experiments are implemented on various simulated and real-world datasets to verify the findings. We also show the theoretical connections between Rényi divergence aggregators and machine learning markets with isoelastic utilities. The second issue involves inference and optimization with large-scale data. We consider two important scenarios: one is optimizing large-scale Convex-Concave Saddle Point problem with a Separable structure, referred as Sep-CCSP; and the other is large-scale Bayesian posterior sampling. Two different settings of Sep-CCSP problem are considered, Sep-CCSP with strongly convex functions and non-strongly convex functions. We develop efficient stochastic coordinate descent methods for both of the two cases, which allow fast parallel processing for large-scale data. Both theoretically and empirically, it is demonstrated that the developed methods perform comparably, or more often, better than state-of-the-art methods. To handle the scalability issue in Bayesian posterior sampling, the stochastic approximation technique is employed, i.e., only touching a small mini batch of data items to approximate the full likelihood or its gradient. In order to deal with subsampling error introduced by stochastic approximation, we propose a covariance-controlled adaptive Langevin thermostat that can effectively dissipate parameter-dependent noise while maintaining a desired target distribution. This method achieves a substantial speedup over popular alternative schemes for large-scale machine learning applications.

La production des états charmonia (par exemple le J/ψ et le ψ(2S)) est l’une des sondes étudiées pour comprendre les propriétés du Plasma de Quarks et de Gluons (PQG) formé dans les collisions d’ions lourds à très haute énergie. En effet, la présence d’un milieu déconfiné est susceptible de modifier le taux de production des charmonia par effet d’écrantage de couleur de la paire de quarks charme-anti-charme. Une telle suppression fut déjà observée dans les collisions d’ions lourds aux énergies du SPS et du RHIC. Dans les collisions Pb-Pb à √sNN = 2.76 TeV au LHC, une suppression relative du taux de production du J/ψ par rapport à celui mesuré dans les collisions pp normalisé aux nombres de collisions binaires proton-proton (facteur de modification nucléaire RAA) fut aussi mesurée. Cette suppression est cependant moins importante que ce qui fut observé à plus petites énergies dans le centre de masse. Ceci peut s’expliquer par la présence d’un nouveau mécanisme de production, la recombinaison de quarks charme et anticharme déconfinés au moment de l’expansion hydrodynamique du PQG ou durant la phase d’hadronisation. Dans cette thèse, nous présentons les résultats de la production de J/ψ dans les collisions pp et Pb-Pb, mesurée par le détecteur ALICE à √sNN = 5.02 TeV à rapidité avant dans le canal de décroissance dimuonique. La section efficace de production mesurée dans les collisions pp est comparée à différents modèles théoriques et est utilisée pour calculer le RAA en collisions Pb-Pb, lui-même comparé à différentes prédictions théoriques ainsi qu’aux précédentes mesures. / The production of charmonium states (for instance J/ψ and ψ(2S)) is one of the probes studied to investigate the properties of the Quark-Gluon Plasma (QGP) formed in high-energy heavy-ion collisions. Indeed, the presence of a deconfined medium should modify the charmonium production yield, due to the color screening of the charm quark anti-quark potential. Such a suppression was already observed in heavy-ion collisions at SPS and RHIC energies. In Pb-Pb collisions at √sNN = 2.76 TeV at the LHC, a clear suppression of the J/ψ yield with respect to the one measured in binary-scaled pp collisions (nuclear modification factor RAA) was observed, but much smaller than that at lower collision energies. This observation can be explained by the presence of a new production mechanism, the (re)combination of deconfined charm and anti-charm quarks during the hydrodynamical expansion of the QGP or at the hadronization stage. In this thesis, we report on the results of the J/ψ production in pp and Pb-Pb collisions measured with the ALICE detector √sNN = 5.02 TeV at forward rapidity in the dimuon channel. The cross-section in pp collisions is compared to various model calculations, and is used to compute the RAA in Pb-Pb collisions, which is also compared to theoretical predictions as well as to previous measurements.

No description available.

No description available.

No description available.

System availability is the probability of the system being operable at instant t. Markov chains are a model used for system availability analysis. The exact analytical solution in terms of component failure rates and repair rates for steady-state system availability is complex to find solving the large numbers of simultaneous linear equations that result from the model. Although exact analytical solutions have been developed for series and parallel systems and for some other small size systems, they have not been developed for large scale general systems with n distinct components. Some methods for approximate analytical solutions have been suggested, but limitations on network types, over simplified states merge conditions and lack of predictions of approximation errors make these methods difficult to use. Markov state transition graphs can be classified as symmetric or asymmetric. A symmetric Markov graph has two-way transitions between each pair of communicating nodes. An asymmetric Markov graph has some pair(s) of communicating nodes with only one-way transitions. In this research, failure rates and repair rates are assumed to be component dependent only. Exact analytical solutions are developed for systems with symmetric Markov graphs. Pure series systems, pure parallel systems and general k out of n systems are examples of systems with symmetric Markov graphs. Instead of solving a large number of linear equations from the Markov model to find the steady-state system availability, it is shown that only algebraic operations on component failure rates and repair rates are necessary. In fact, for the above class of systems, the exact analytical solutions are relatively easy to obtain. Approximate analytical solutions for systems with asymmetric Markov graphs are also developed based on the exact solutions for the corresponding symmetric Markov graphs. The approximate solutions are shown to be close to the exact solutions for large scale and complex systems. Also, they are shown to be lower bounds for the exact solutions. Design principles to improve systems availability are derived from the analytical solutions for systems availability. Important components can be found easily with the iteration procedure and computer programs provided in this research.