At Domfer Metal Powders, the ball milling operation is critical since it determines many characteristics of the final product, such as density, green strength, compressibility and dimensional change. Highly variable input material properties and operational procedures accompanied by poor product feedback frequency generate large variations in powder properties. The goal of the research is to develop a model-based control system for the ball milling process of metal powder. The following research objectives are aimed at reducing process variations while maximizing throughput: include ball mill physics into design and computer models, develop control strategies and computer architectures for real-time control, and develop process monitoring and fault diagnosis techniques. / Ball mill size reduction theory is presented as a basis for process characterization. Next, process physics are described along with the measurability and controllability of the variables. Then, plant trials are performed to define system behavior and performance specifications of variables and sensors. After that, a sampler for metal powder is developed to automate the entire sieve analysis process. / A new ball mill model is created for open-circuit dry ball milling of metal powders. The process model is a combination of rules, equations and heuristics and is implemented using an agent based architecture that can deal with multiple data streams and a network of related sub-models of different sizes and operating time scales. / The model architecture is integrated using a programmable logic controller. Control and monitoring algorithms are developed in low-level PC language. A performance plant trial validated the control system and demonstrated that ball milling product specifications, namely size distribution and apparent density, are achievable and maintainable at a 99.7% confidence interval. This new technology will endow Domfer with a serious lead in metal powder manufacturing. / Key words. model-based control, ball milling, metal powder.

The dynamical stability, wave propagation, response, and mode localization in a train of flexibly interconnected rigid cylinders travelling in a confined cylindrical "tunnel" subjected to fluid dynamic forces are studied theoretically. Each cylinder, which is coupled and supported by springs and dampers, has two degrees of freedom of translational and rotational motions. The kinetic, dissipation, and potential energies of the system and the generalized forces associated with the fluid dynamic forces acting on the system, such as inviscid fluid dynamic forces, viscous frictional forces, and form drag, are obtained first. Then the equations of motion are derived by application of the Lagrange equation. The principal aim of this study is to investigate the effect of aerodynamic forces on the dynamics of a high-speed train running in a tunnel, or more generally of a train-like system travelling in a coaxial cylindrical tube. / The results of this study show that (a) the system loses stability by flutter; (b) viscous frictional drag has a considerable effect on stability; (c) when the aerodynamic forces act on the train, the frequency bands of the dispersion relation of wave propagation shift, and thus no classical normal modes (standing wave solutions) exist in the system; (d) the wavelength of the travelling sinusoidal force controls phase differences among cylinders in the train: and (e) the imperfections in the supporting springs have a great influence on mode localization and considerably alter the stability of the system.

Currently at Bombardier Aerospace, aeroelastic analyses are performed using the Doublet Lattice Method (DLM) incorporated in the NASTRAN solver. This method proves to be very reliable and fast in preliminary design stages where wind tunnel experimental results are often not available. Unfortunately, the geometric simplifications and limitations of the DLM, based on the lifting surfaces theory, reduce the ability of this method to give reliable results for all flow conditions, particularly in transonic flow. Therefore, a new method has been developed involving aerodynamic data from high-fidelity CFD codes which solve the Euler or Navier-Stokes equations. These new aerodynamic loads are transmitted to the NASTRAN aeroelastic module through improved aerodynamic influence coefficients (AIC). A cantilevered wing model is created from the Global Express structural model and a set of natural modes is calculated for a baseline configuration of the structure. The baseline mode shapes are then combined with an interpolation scheme to deform the 3-D CFD mesh necessary for Euler and Navier-Stokes analyses. An uncoupled approach is preferred to allow aerodynamic information from different CFD codes. Following the steady state CFD analyses, pressure differences (Cp), calculated between the deformed models and the original geometry, lead to aerodynamic loads which are transferred to the DLM model. A modal-based AIC method is applied to the aerodynamic matrices of NASTRAN based on a least-square approximation to evaluate aerodynamic loads of a different wing configuration which displays similar types of mode shapes. The methodology developed in this research creates weighting factors based on steady CFD analyses which have an equivalent reduced frequency of zero. These factors are applied to both the real and imaginary part of the aerodynamic matrices as well as all reduced frequencies used in the PK-Method which solves flutter problems. The modal-based AIC method's evaluation, performed with CFD data calculated by the DLM, is essential to find the natural modes which are most influential on the flutter solutions of the different configurations. Finally, Euler and Navier-Stokes results are used to obtain improved flutter solutions for a subsonic case at Mach 0.7 and dispositions are made to accomplish the same exercise for transonic speeds. / En ce moment chez Bombardier Aéronautique, les analyses aéroélastiques sont effectuées à partir de la méthode Doublet-Lattice (DLM) inclue dans le logiciel NASTRAN. Cette méthode est plutôt fiable et rapide lors des étapes préliminaires de design lorsque les données expérimentales sont inexistantes. Malheureusement, les simplifications géométriques nécessaires et les limites mathématiques de la méthode DLM, qui est basée sur la théorie des surfaces portantes, réduit la capacité de celle-ci d'obtenir des résultats satisfaisants pour toutes les conditions de vol, spécialement pour le domaine transsonique. Ainsi, une nouvelle méthode a été développée à l'aide de données aérodynamiques provenant de codes CFD (Computational Fluid Dynamics) qui solutionnent des équations d'Euler et Navier-Stokes. Ces nouvelles données aérodynamiques sont ensuite transmises au module aéroélastique de NASTRAN à partir de coefficients d'influence aérodynamiques (AIC) améliorés. Une aile provenant du modèle structurel du Global Express est mise en porte-à-faux et une série de modes naturels sont calculés pour une configuration de base de la structure. Ces modes naturels sont ensuite combinés à une méthode d'interpolation qui déforme le maillage volumique nécessaire aux analyses par codes Euler et Navier-Stokes. Une méthode "découplée" est privilégiée pour permettre l'utilisation de données aérodynamiques provenant de différents logiciels. Suite aux analyses CFD statiques, les différences de pressions (Cp), calculées entre les modèles déformés et la géométrie originale, produisent les forces aérodynamiques qui seront transférées au modèle DLM. Une méthode d'AIC modales est appliquée aux matrices aérodynamiques de NASTRAN à l'aide d'une approximation de moindres carrés pour évaluer les forces aérodynamiques d'une configuration différente de l'aile qui détient des déformations modales de types similaires. La méthodologie développée dans cette recherche permet de créer des facteurs de pondération basés sur les analyses CFD statiques qui ont une fréquence réduite équivalente à zéro. Ces facteurs sont appliqués simultanément sur les parties réelles et imaginaires des matrices aérodynamiques ainsi que pour toutes les fréquences réduites utilisées dans la "PK-Method" qui résout les problèmes de flottement aéroélastique. L'évaluation de la méthode des AIC modales, effectuée à partir des données obtenues de la méthode DLM, est essentielle afin de trouver les modes naturels qui influencent le plus les solutions deflottement des différentes configurations. Finalement, les résultats des analyses Euler et Navier-Stokes sont utilisés afin d'obtenir des solutions de flottement améliorées pour une vitesse subsonique de Mach 0.7 et des dispositions sont mises en place dans le but d'accomplir le même exercice pour des vitesses transsoniques.

ABSTRACT Space marching methods are widely used to calculate steady supersonic internal and external flows in view of their superior efficiency as compared to time marching approaches. A very general space-marching procedure was proposed by Nakahashi and Saitoh in 1996. The method can be used with any explicit time-marching procedure, allows for embedded subsonic regions, and is well-suited for unstructured grids, enabling a maximum of geometric flexibility. The flow field is only updated in the so-called active domain. Once the residual has fallen below a preset tolerance, the active domain is shifted. The present work is aimed at further refining of the Nakahashi and Saitoh's method. This is achieved via the following new approaches: (a) A non-reflecting boundary condition is implemented at the exit of the active domain, thus eliminating non-physical reflections propagating upstream; (b) A general, user-independent procedure for partitioning of the computational domain into a set of active domains is proposed; it allows to minimize the size of the active domain and the residual monitor region, thus contributing to the method's efficiency; (c) Local grid adaptation is performed in the course of space marching. These ideas are verified and tuned, at first, on two simple test problems (supersonic flows over compression and expansion corners). Finally, the results of the application of the improved method to supersonic flows in air-breathing engine inlets and ram accelerators are presented. / ABRÉGÉ Les méthodes de la marche spatiale sont couramment utilisés pour calculer les flux supersoniques internes et externes, grâce à l'efficacité de ces méthodes plus élevées comparé aux approches de la marche temporelle. Une procédure trés générale de la marche spatiale a été proposée en 1996 par Nakahashi et Saitoh. Cette méthode peut étre utilisée avec n'importe quelle procedure explicite de la marche temporelle, va bien pour des regions subsoniques encastrées et pour des grilles non structurées, permettant un maximum de flexibilité géométrique. L'espace du flux est seulement mis à jour dans ce qui s'appelle le domaine actif. Lorsque le résidu tombe en-dessous du niveau de tolérance, le domaine actif est changé. Le travail présent vise à étudier plus profondement la méthode de Nakahashi et Saitoh. Cet objectif est atteint par ces nouvelles approches : (a) Une condition non-réfléchissante de la frontiére est mise en place à la sortie du domaine actif, ce qui élimine les réflexions non-physiques qui se propagent dans le sens opposé du flux; (b) Une procédure générale, non dépendante de l'utilisateur est proposée pour partitionner le domaine calculatoire en plusieurs domaines actifs, cela permet de réduire la taille du domaine actif et la région du moniteur en résidu, contribuant ainsi à l'efficacité de la méthode; (c) Une adaptation locale de la grille est réalisée au cours de la marche spatiale. Ces idées sont vérifiées et réglées sur deux tests simples pour commencer (des flux supersoniques sur des angles en compression et en expansion). Finalement, les résultats de l'application de la méthode améliorée concernant les flux supersoniques dans les moteurs fonctionnant avec de l'air et les ram accélérateurs sont presentés.

In this report, the results of a theoretical and experimenta. l investigation of the propagation of shocks and blast waves in a detonating gas are presented. In particular, cylindrical diverging detonations are studied in detail: In the theoretical phase of the investigation, analytical solutions have been obtained for two classes of problems: The first class involves a steady-state reaction front where self-similar solutions exists, and the second class of problems involve a non-steady reaction front where self-similar solutions are not possible. [...]

Surge chambers are sometimes included in the hydraulic circuits of hydroelectric power plants as a means of absorbing pressure waves formed by the opening/closing of a turbine. Surge chambers, however, result in additional power loss and therefore reduce the efficiency of the plant. This work aims to 1) investigate the physical phenomena and flow within a surge chamber under normal operation (i.e. no opening/closing of a turbine), and 2) obtain experimental data for the validation of numerical simulations of this complex flow.Experiments and numerical simulations have been conducted for a simplified model of a surge chamber operated under multiple configurations at a constant input flow rate. This 3-D, unsteady, incompressible, swirling, two-phase flow has been experimentally characterized by global values, such as head losses, and local values, such as free-surface profiles, free-surface oscillations, reduced pressure profiles and velocity fields. The same quantities were also obtained numerically using the "rasInterFoam" solver of the open source code "OpenFOAM-1.5," for incompressible two-phase flows. This solver implements a one-fluid, volume-of-fluid (VOF) method with an interface-capturing scheme. Overall agreement between the experimental and numerical quantities is good, although there are local discrepancies. The periodic oscillations of the flow observed in the experiments and the numerical simulations of the simplified model (operated under constant input flow rate) were associated with the phenomena of i) oscillating mass, and ii) self-induced sloshing. / Des chambres d'équilibre sont parfois intégrées aux circuits hydrauliques des centrales hydroélectriques afin d'absorber les ondes de pression se formant lors de l'ouverture/fermeture d'une turbine. Celles-ci affectent l'efficacité des centrales, en augmentant les pertes d'énergie. Ce projet de maîtrise vise à 1) étudier les phénomènes physiques ainsi que l'écoulement à l'intérieur d'une chambre d'équilibre sous opération normale (i.e. aucune ouverture/fermeture de turbine), et à 2) obtenir des données expérimentales visant à valider les simulations numériques de cet écoulement complexe.Les mesures expérimentales et les simulations numériques ont été effectuées sur un modèle simplifié d'une chambre d'équilibre. Ce dernier a été opéré sous de multiples configurations à débit d'entrée constant. L'écoulement tridimensionnel, instationnaire, incompressible, tourbillonnant et biphasique a été caractérisé expérimentalement par des quantités globales, telles que des pertes de charges, ainsi que par des quantités locales, telles que des profils et des périodes d'oscillations de surface libre, des profils de pression réduite et des champs de vitesses. Les mêmes quantités ont aussi été obtenues par calculs numériques en utilisant l'exécutable "rasInterFoam" du code à source ouverte "OpenFOAM-1.5", limité aux écoulements incompressibles et biphasiques. Ce dernier traite l'écoulement comme étant un mélange localement homogène composé de deux phases en utilisant une méthode "volume-of-fluid" (VOF) et un schéma de capture d'interface.Globalement, les résultats numériques concordent avec les mesures expérimentales, malgré quelques variations locales. Les oscillations périodiques de l'écoulement survenant à un débit d'entrée constant, qui ont été observées tant sur le banc d'essai que dans les simulations numériques, sont associées aux phénomènes i) d'oscillation de masse et de ii) ballottement auto-induit.

The Very Large Optical Telescope (VLOT) is a concept, proposed by the National Research Council - Herzberg Institute of Astrophysics, to address future demands of the international astrophysics community. To develop and evaluate the VLOT designs, an Integrated Modeling approach is employed. The Integrated Model (IM) is a single simulation tool which accounts for the interactions between different telescope modules to produce the final image quality generated by the telescope. The work in this thesis relates directly to two of the modules of the IM: (1) the structural dynamics and (2) the primary mirror control module. / In order to obtain accurate optical performance predictions from the IM, a high fidelity structural model of the telescope is required. However, the finite-element model of the telescope, which is its nascent structural model, is prohibitively large in size for practical simulation. This problem can be addressed by applying a suitable Model Order Reduction technique. In the first half of this thesis, we compare five Model Order Reduction techniques from the literature for the VLOT application. These methods are evaluated and compared in frequency and time domains. Based on these results, we conclude that the Singular Perturbation Approximation method is suitable for the VLOT application. / The primary mirror, which is the main light-gathering surface of the telescope, must be maintained in a particular profile in order to provide the desired image quality. Currently, this control objective is achieved with an enhanced integral controller based on a SISO control approach. The primary mirror, however, is a MIMO system and hence the second objective of this thesis is to evaluate a MIMO control design approach, in particular, the Linear Quadratic Gaussian (LQG) controller. The LQG design is evaluated in two steps: initially with a single mirror segment model and then with the VLOT structural model. Based on our results, we conclude that even with currently available computational resources, it is not feasible to employ the LQG control approach in a centralized manner to a large scale system such as VLOT.

This research is an assessment of the machinability of a nickel-based alloy Inconel 718 using a finish turning operation with carbide coated and uncoated tools having fine and medium grain size substrates, and the same tool geometry (a round tool shape geometry with a 0° rake angle and 7° clearance angle). In addition, orthogonal machining tests were carried out to generate experimental data for future identification of the constitutive laws of Inconel 718 (its behavior under high stress, strains, strain rates, and temperatures). The finish turning of this alloy was evaluated in terms of tool life, tool wear mechanisms, cutting forces, surface roughness, and chip formation process. The cutting speed range evaluated was from 50 to 100 m/min, and the feed range was from 0.075 to 0.125 mm/rev. Throughout the machining tests, the depth of cut was kept constant and equal to 0.25 mm. Tungsten carbide uncoated tool having fine grain size (<1 μm), strong and sharp cutting edges were found to be the most suitable tool material and tool geometry for the finish turning of Inconel 718 at a cutting speed of 50 m/min, and, moreover, yielded the best tool life of 53 minutes. When finish turning at a cutting speed of 75 m/min, a single layer PVD (TiAlN=3μm) carbide tool yielded the best tool life of 12 minutes. At a cutting speed of 100 m/min, triple layer CVD coated tools (TiCN/Al2O3/TiN=15 μm) produced the best performance, yielding an 8 minutes of tool life. During the machining tests, the common wear mechanisms of the three different tools were adhesion, abrasion, and diffusion. Flank wear was the dominant tool failure mode of all three tools. The analysis of chip morphology showed that no phase transformation existed even though the material was deformed. The increase of micro-hardness was confirmed by micro-hardness measurements on chips, as compared to the initial micro-hardness of the material (Inconel 718). Cutting forces for all the different tools were relatively low (< 300 N) when the cutting edge was new. For the finish turning operation, surface roughness of (<1 mum) was achieved in all the cutting tests.

Computational and experimental studies of fluid flow and heat transfer phenomena in interrupted-surface passages are presented. The computational investigation was focused on developing and fully-developed (spatially and temporally) laminar flow and heat transfer in essentially two-dimensional regions of the following passages: (1) a straight rectangular duct with spatially periodic in-line plate inserts; and (2) staggered-plate arrays. A second-order finite-volume method was developed, validated, and used to solve these problems. Time-mean modular friction factors and Colburn factors were obtained from the domain inlet to the spatially-periodic fully-developed region for a Reynolds number range of 100 to 600, thus exploring the laminar steady and unsteady regimes. Additionally, various cyclic domains were investigated. In the experimental investigation, single hot-wire measurements were used to obtain ensemble-averaged power spectrums and Strouhal numbers in the fully-developed region of the rectangular duct with spatially periodic in-line plate inserts, for Reynolds numbers ranging from 2000 to 30000.

The algorithms used for solving the forward dynamics problem of a complex multibody system are essential for an efficient simulation of robotic systems. The efficiency is measured by the CPU time and the number of operations per number of bodies in the system. In this work, we discuss the solution of the forward dynamics problem for real-time simulations of space robotic systems using the so-called Articulated Body Method. Traditional methods use the calculation of the inverse of the generalized mass matrix. This makes the required number of operations proportional to the cube of the number of bodies in the multibody system. For real-time simulations of complex systems, O(n) algorithms seem to be the best choice because the number of operations is linearly proportional to the number of bodies. They become more efficient than O( n3) algorithms as soon as the number of bodies in the system exceeds 12--14. In this thesis, we will present a detailed discussion of O(n) algorithms derived based on the Articulated Body Method (ABM). This algorithm can be presented using Lagrangian and Hamiltonian variables. Such algorithms can be used for the real-time simulation of robotic systems by taking into account both joint-flexibility and approximations for the gear-ratio effect. A unified derivation of the ABM algorithms using both Lagrangian and Hamiltonian variables will be discussed. The intended, primary application of the algorithms is to develop real-time simulation engines for the complete robotic system of the International Space Station. The implementation and use of these algorithms will be analyzed in detail.