The main aim of this thesis is the study of the two mechanisms underlying fluidelastic instability in cylinder arrays subjected to cross-flow: the negative fluid damping and the fluid stiffness-controlled mechanisms, especially with regard to the effect of fluidelastic coupling between cylinders and the effect of system parameters in each case. / An in-line square array with pitch-to-diameter ratio P/d = l.5 has been tested in both water- (with low mass ratio, $m/pd sp2$ = 6.5) and air-flow (with high mass ratio, $m/pd sp2$ = 860). The effect of fluidelastic coupling between cylinders on the critical flow velocity for fluidelastic instability was investigated via frequency detuning experiments, including an extreme case of one flexible cylinder in an otherwise rigid cylinder array. The water tunnel experimental results showed that fluidelastic coupling between cylinders in water-flow has little effect on the critical flow velocity for fluidelastic instability, whereas the wind tunnel experimental results demonstrated that fluidelastic coupling between cylinders in airflow has a significant effect on the critical flow velocity. Therefore, for the cases with low mass ratio parameter the negative fluid damping mechanism dominated the fluidelastic instability of the array; however, for cases with high mass ratio, fluidelastic instability of the array was controlled by fluid stiffness effects. / An in-line square array with P/d = l.5 was also tested in the wind tunnel, with cylinders with high and low levels of mechanical damping. The experimental results showed that fluidelastic coupling between cylinders with low mechanical damping has a more significant effect on the critical flow velocity for fluidelastic instability than with high mechanical damping in airflow (with high mass ratio, $m/pd sp2$ = 860); hence, fluid stiffness-controlled mechanism becomes more important for fluidelastic instability of the array with low mechanical damping in airflow. Next, an in-line square array with P/d = 3.0 was tested in the wind tunnel. The results showed that one flexible cylinder becomes unstable when positioned in row 2 of the array; however, it does not become unstable when positioned in rows 3 or 4. This suggests that P/d has a large effect on fluidelastic instability behaviour of in-line square arrays. / Finally, a new constrained-mode approach, with a one-cylinder-kernel, has been developed for the instability analysis of both in-line and staggered arrays. The approach was developed to reduce the computational effort when a fully flexible cylinder array has to be analyzed, in order to take into account fluidelastic coupling; between cylinders.

The fundamental properties of dust flame propagation have been studied experimentally and theoretically. Experiments were carried out in a Pyrex tube having a 5 cm inside diameter and a length of 1.2 meters. The dust dispersion system in the present experiment has been demonstrated to produce a laminar flow with uniform dust concentration. It is observed that the propagating flame exhibits a truly laminar character of propagation. Propagation and quenching of dust-air flames were measured for aluminum dust for a relatively wide range of dust concentration. It was observed that the process of the dust flame propagation in the tube can be divided into three different stages: laminar flame, oscillating flame and turbulent accelerating flame. During the first stage, the flame propagates with approximately constant speed and the flame is laminar. This stage is about 1/3 to 1/2 of the tube length and the shape of the flame front during this phase is close to the usual parabolic shape. In the second stage, flame starts to oscillate. Bright regular flashes alternate with stages when the flame is almost invisible. The third stage of the flame propagation can be observed only if the dust concentration is close to or higher than stoichiometric. The quenching distance and flame velocity of an aluminum dust flame under different initial oxygen concentrations and with different inert gases such as helium and nitrogen were also measured. When the amount of the initial oxygen concentration decreases, both the quenching distance and the lean limit increase, while the flame velocity decreases. On the other hand, after using helium as the inert gas of the mixture, it was observed that the whole value of the quenching distance, the lean limit, and the flame velocity increase. However, the objective is to understand the mechanism of the dust flame propagation. Theoretical models have also been developed to correlate the data to achieve a better understanding of the propagation mechan

The motion control of mechanical systems with flexible links is investigated. Issues addressed are the modelling and simulation of such systems, the design of a feedback control scheme and its implementation on an actual system. The modelling method used is a combination of a spline-based spatial discretization technique, which allows a definition of the state-variable vector as the set of curvature values at the supporting points of the spline and their time-rates of change, and the natural orthogonal complement of the kinematic constraints that eliminates the constraint forces and moments from the mathematical model. The control algorithm consists of two parts, namely, the decoupling of nonlinear equations of motion and the filtering of non-working constraint forces. The former is achieved using the unconstrained equations of motion, expressed in terms of extended generalized coordinates with which the motion of each separate link is defined. The latter is accomplished using the fact that the constraint forces thus introduced indeed lie in the nullspace of the transpose of the NOC. This control scheme is implemented on a prototype four-bar flexible mechanism. Strain gauges are used to measure link curvature at the supporting points of the spline, while the time-rate of change of curvature is estimated with a Kalman filter. Moreover, the angle of rotation of the input link and its time-rate of change are measured and then used to infer the rest of the rigid-body motions. Results show that the proposed control scheme provides successful trajectory tracking while suppressing the vibration triggered by a doublet-type of disturbance. This disturbance is induced by the singularities of the mechanism coupled with the rapid inertia changes. While the main motivation of this study is the control of robotic manipulators with long and slender links, typically found in space applications, the results presented are applicable to the control of a much broader class of mechan

A general class of manipulators with hybrid kinematic chains is introduced. These manipulators are modular and can readily lead to kinematically redundant structures. Moreover, they may contain serial subchains as well as subchains of parallel modules, acting in parallel and connecting a base to a common end-effector. / At the onset, a novel formalism is introduced to study modular kinematic structures. This includes some new concepts and definitions. The proposed formalism provides a systematic way to represent a complex hybrid manipulator through a hierarchy of its constituting modules. Next, a formulation is devised to study the instantaneous kinematics of these manipulators. The proposed formulation is general and can be applied to most of the existing parallel and hybrid manipulators. To realize the concept of general hybrid manipulator in practice, a few prototype designs are also introduced. The proposed prototypes are superior to their conventional counterparts in terms of dexterity and the volume of their workspace. / The study is then extended to the theory of hyper-redundant manipulators that comprise, as a subclass, the variable-geometry trusses. A spline-based solution method is proposed for the inverse kinematics of hyper-redundant manipulators. The method is applicable to both extensible and nonextensible cases and includes planar as well as spatial manipulators. Also, a new variable-geometry truss is introduced whose modules themselves are kinematically redundant. / Another practical aspect of redundancy is the use of redundant-sensor data to simplify inherent nonlinear direct kinematics of parallel manipulators. For a six-degree-of-freedom general parallel manipulator, we introduce a formulation of the direct kinematics whereby the positioning and orientation problems are decoupled by introducing two auxiliary parameters in the forms of either two angles or two lengths. This is in accordance with the type of redundant sensors, i.e., rotary or translational, to be used. Moreover, a real-time implementation of extra-sensor data, with a unique direct kinematics solution, is proposed by resorting to an eigenvalue problem. The parallelism in the proposed formulation enables the user to benefit from a parallel-computing environment. Hence, we introduce a parallel-computing algorithm that highly increases the robustness of the computational algorithm. / The concept of kinematic isotropy has been used as a criterion in the design of serial and parallel robotic manipulators. However, all notions adopted to express isotropy in parallel manipulators have been based on the structure of the Jacobian matrices of serial manipulators. Here, we introduce a definition of kinematic isotropy that is well-suited to parallel manipulators. This is done by identifying the special structure of the Jacobian matrices involved in the differential kinematics of such manipulators. This leads to a partitioning of the Jacobian matrices into submatrices with dimensionally homogeneous entries. Moreover, based on the proposed definitions of isotropy and kinematic optimality, a set of conditions is derived that provides a systematic way for the optimum kinematic design of parallel manipulators, with or without structural constraints.

Initiation of detonation in the cylindrical geometry is important in validating a general theory of initiation, and has particular significance to the interaction of high velocity projectiles with combustible gas. In the present study, an experimental investigation of the direct initiation of gaseous detonations by a line source of condensed explosives (detonating cord) is presented. / Due to the high velocity of detonation in the condensed explosive cord (6--7 km/s), the rapid chemical energy release along its length generates a quasi-cylindrical blast wave in the surrounding combustible mixture. For a sufficiently strong detonating cord, the generated blast wave is observed to decay to a stable Chapman-Jouguet (CJ) conical detonation. Below a certain critical energy, the blast wave decays to low velocities without initiating detonation. In the critical regime of initiation, measurements of the combustion front velocity and direct flow visualization indicate that the blast wave decays to sub-CJ velocities before the onset of detonation. The onset of detonation is associated with discrete explosion centers. This phenomenon is similar to what has been previously observed for spherical initiation, suggesting a universal mechanism of detonation initiation. / The critical shock radius at which the onset of detonation occurs is correlated to the detonation cell size (lambda) of the mixture, consistently yielding a radius of 4 to 8lambda. Using the invariance of the explosion length R0 in the critical regime of initiation for spherical and cylindrical detonations, it is shown how the critical radii and critical energies scale between the two geometries, clarifying the correct length scales that should be used in the theory of direct initiation of detonation.

Metal and composite plates can be bonded together to form a joint known as the "single-lap" joint. The single-lap joint is studied under two different loading conditions: (i) out-of-plane load (bending) and (ii) in-plane load (tension). The different joint configurations are studied analytically and experimentally in order to achieve the optimum design. In configurations such as lap joints, the presence of stress singularities eliminates the possibility of using any stress-based failure criteria. A strain energy method is used to predict the strength of adhesively bonded joints because of its convergence with mesh refinement and it is found to be in good agreement with experimental results. Failure of single-lap joints is governed by the load case under consideration and the way in which the stress distribution varies at the joint ends. Failure varies with the taper angle (inner and outer), with or without additional epoxy beads. It is observed that designing the joint for one kind of load will not always be satisfactory because, for other load cases, different parameters will govern the design. It is shown that the optimum design for the single lap joint under bending loads will not be the optimum design for the tension case. Therefore, the optimum design can be chosen in a way that satisfies both loading conditions.

Milling of thin-walled aerospace structures is a critical process due to the high flexibility of the workpiece. Available models for the prediction of the effect of the fixture on the dynamic response of the workpiece are computationally demanding and fail to represent practical cases for milling of thin-walled structures. Based on the analysis of typical structural components encountered in the aerospace industry, a generalized unit-element, with the shape of an asymmetric pocket, was identified to represent the dynamic response of these components. Accordingly, two computationally efficient dynamic models were developed to predict the dynamic response of typical thin-walled aerospace structures. These models were formulated using Rayleigh's energy and the Rayleigh-Ritz methods. In the first model, the dynamics of multi-pocket thin-walled structures is represented by a plate with torsional and translational springs. A methodology was proposed and implemented for an off-line calibration of the stiffness of the springs using Genetic Algorithms. In the second model, the dynamics of a 3D pocket is represented by an equivalent 2D multi-span plate. Through a careful examination of the milling of thin-walled structures, a new formulation was developed to represent the continuous change of thickness of the workpiece due to the material removal action. Two formulations, based on holonomic constraints and springs with finite stiffness, were also developed and implemented to take into account the effect of perfectly rigid and deformable fixture supports. All the developed models and formulations were validated numerically and experimentally for different workpiece geometries and / Le fraisage des structures aérospatiales à parois minces est un processus critique dû à la flexibilité élevée de la pièce. Les modèles disponibles pour la prévision de l'effet du système de fixation sur la réponse dynamique de la pièce sont basés sur des méthodes numériques très lentes et n'arrivent pas à représenter les cas pratiques du fraisage des structures à parois minces. Basé sur une analyse des composants structurels typiques produits dans l'industrie aérospatiale, un élément généralisé de base avec la forme d'une poche asymétrique, a été identifié pour représenter la réponse dynamique de ces composants. En conséquence, deux modèles dynamiques efficaces ont été développés pour prévoir la réponse dynamique des structures aérospatiales types à parois minces. Ces modèles ont été formulés en utilisant les méthodes de Rayleigh et Rayleigh-Ritz. Dans le premier modèle, les réponses dynamiques des structures de poches multiples à parois minces sont représentées par des plaques avec des ressorts de torsion et de translation. Une méthodologie a été proposée et mise en application pour calibrer la rigidité des ressorts en utilisant les algorithmes génétiques. Dans le deuxième modèle, la réponse dynamique d'une poche en 3D est représentée par une plaque équivalente de multi-travées en 2D. À travers une étude approfondie du fraisage des structures à parois minces, une nouvelle formulation a été développée pour représenter le changement continu de l'épaisseur de la pièce durant l'usinage. Deux formulations, basées sur des contraintes holonomes et des ressorts avec des rigidités finies, ont été$

The objective of this work was to model a transferred arc metal evaporator including the behavior of a plasma arc and the molten anode being continuously evaporated. The model predicted the system temperature and velocity distributions as well as the rate of metal evaporation. The effects of arc current, arc length, plasma gas flow rate, crucible temperature, surrounding temperature, and hydrogen addition to an argon arc have been investigated and some optimization of operating conditions was done. / The appropriate governing equations were solved numerically. The numerical results of the present investigation revealed that the geometry of the crucible did not significantly affect the temperature distribution of anode surface or the metal vaporization rate. Increasing plasma torch flow rate also did not lead to an increase of the temperature of the anode surface and the metal vaporization rate. However, the crucible temperatures significantly affected the vaporization rate; the crucible should be maintained at a high temperature, which is compatible to the physico-chemical properties of the crucible material. The arc current strongly affected the vaporization. Addition of hydrogen to the plasma gas caused a significant increase of vaporization rate. The arc length greatly affected the surface temperature distribution, the metal vaporization rate, the arc voltage, and the total power. Shorter arcs were found to be more effective. The modeling results were in reasonable agreement with the experiment results.

We examined the design, development and implementation of a modular system of multiple, wheeled mobile manipulators that can team up to cooperatively transport a large common object. Each individually autonomous mobile manipulator consists of a differentially driven wheeled mobile robot (WMR) with a passive, two-degree-of-freedom, planar revolute jointed arm mounted in the plane parallel to the base of the WMR. / The composite multi-degree-of-freedom vehicle, formed by placing a common object on the end-effector of two such mobile manipulator systems, possesses the ability to change its relative configuration as well as accommodate relative positioning errors of the mobile bases. However, closed kinematic loops are also formed constraining the relative motions of the overall system and requiring a careful treatment. / Two variants of the control schemes developed for mobile manipulators are adapted for the control of the overall collaborating system of two mobile manipulators carrying a common object along a desired trajectory. (Abstract shortened by UMI.)

The goals of the research in this thesis are twofold. First, I designed and tested a stair-traversing controller, which allows RHex to ascend and descend a wide variety of human sized stairs. I tested the stair-ascending controller on nine different flights of stairs, and the stair-descending controller on four different flights of stairs. Rhex was run ten times on each flight, and met with only a single failure out of the one-hundred-and-thirty attempts. Second, we built and tested three competing leg designs. The second and third leg designs proved adequate for use on RHex, enabling dynamic and off road behaviors. The stair controller and leg design processes both drew on lessons from biology, in a process in we call "functional biomimesis". This framework guided us while we mimicked the functionally important features of animal morphology and behavior, while ignoring those features that are irrelevant to locomotion. We compared these advances with previous work in the field, including work on wheeled and tracked vehicles, as well as stair-traversing legged robots. Finally, RHex is a new robot, so we present an overview of RHex's basic mechanical and electrical designs.