In a real world pattern recognition application a user cannot assess the performance of a classifier on an unlabeled data set. Classifiers cannot give their best performance because they require user-controlled parameters. As a Solution, a Sequential Function Approximation (SFA) method has been' developed for classification that determines the values of the control parameters during learning. In this dissertation, experiments were carried out on real world data sets where SFA, using only the training subset, had comparable performance to a number of other popular classification schemes whose user-defined parameters were optimized utilizing the entire data set. By the statistical significance of the results it was concluded at 95% confidence that the performance of SFA will be equivalent or significantly better than those of the other popular classification tools. After establishing SFA as a proper classification tool in this dissertation, it is applied to a US Navy flight test problem. The current problem at hand is to predict pilot ratings from HH-60H Sea-Hawk helicopters based on 369 at sea take-off and landing DI tests. Least significant inputs with respect to classification were pointed out with the potential of accelerating through the DI test matrix. And finally an effort was made to give the DI test pilots an estimate of how many tests were necessary to be conducted before generating enough data for the SFA classification tool to satisfactorily learn.

Haptic interfaces are robotic systems designed to provide information to a user regarding a remote or virtual environment via the sense of touch; these robotic systems accomplish this feat through force feedback. Designers of haptic interfaces desire to produce an interaction using the haptic interface wherein the virtual or remote environment feels indistinguishable from the actual environment; however, it is difficult to define when a haptic interaction feels the same as the real interaction it is reproducing. This thesis discusses transparency as a measure that quantifies the performance of a haptic device by comparing the desired environment to be displayed to the actual environment displayed. In order to demonstrate the utility of transparency as a performance measure, haptic interactions are defined by the causality relationship between the user and the device. These interaction types are introduced as active and passive user interactions. In the active user interaction, the user is treated as an energy source and the environment is a dynamic system without a source of energy; the passive user interaction is the opposite case wherein the environment is an energy source and the user is a dynamic system without a source of energy. Methods of improving transparency, and hence performance, of a haptic device are compared against the definitions of haptic user interactions. These comparisons show that transparency for purely active user interactions is dependent on the user, and that for passive user interactions, transparency is dependent on the user and the haptic interface. In addition, other performance improvement methods often rely linear assumptions which are not general. This thesis proposes a method for improving transparency while maintaining stability without regard to assumptions of linearity.

The subject of this study is the vibration isolation effect of linear and nonlinear hysteretic single degree of freedom oscillators attached to a structure. The system of oscillators, attached to the main structure, reduce the amplitude of its structural response, over a wide frequency band excitation. This is achieved by distributing the natural frequencies of the attachments over an a priori specified frequency interval, which is related to the excitation power spectrum density. In order to introduce hysteretic restoring force to the oscillators and study its effect to the vibration reduction on the main structure the Bouc-Wen differential model of hysteresis is used. The introduction of hysteresis to the oscillators is observed to increase the vibration isolation efficiency of the attachments in certain cases. The study of the vibration isolation system behavior is conducted using the Monte Carlo technique. An explicit matrix formulated Newmark integration scheme is used for the linear attachment case. In the case of hysteretic attached oscillators, the system of equations of motion is integrated by an iterative decoupled Newmark integration technique both for the computation of the restoring force as well as the total response of the system. This scheme improves significantly the efficiency of the numerical integration of the equations of motion and accelerates the computational intensive Monte Carlo method. Further, statistical linearization of the hysteretic system is conducted. The agreement of the statistics of the nonlinear system response computed by the Monte Carlo method with those computed by the statistical linearization method enables the use of the latter as an efficient method for the computation of the nonlinear system response statistics in the frequency domain.

Floating Production, Storage, and Offloading (FPSO) Systems are quite often subjected to stochastic sea wave loadings. In this thesis, a methodology is developed for estimating the fatigue life of topside structures. Proper Response Amplitude Operator (RAO) of the FPSO system, and the well-known Ochi-Hubble sea wave elevation spectrum are combined to provide the design spectrum at the deck level on topside FPSO. For ordinary Single-Degree-of-System (S-D-O-F) piece of equipment, the dynamic response is simulated by a time series model. A non-recursive Rainflow cycle counting method is applied to the equipment stress time history to identify significant cycles that produce fatigue damage in the time domain. The results of the Rainflow cycle counting method are supplemented by results from a power spectrum based, exclusively, approach. Further, pipe systems with/without limit stops on topside FPSO are modeled as Bernoulli-Euler beam. A Galerkin method is therefore employed in conjunction with the beam random vibration theory. Specifically, the statistical linearization technique is adopted to derive the equivalent linear system for the pipe example with nonlinear constraints. The applicability of the proposed approach is demonstrated by the analysis of both a simple S-D-O-F piece of equipment and by an illustrative example of pipeline conveying fluid with/without nonlinear constraints subject to sea wave loading. The proposed in this thesis integrated approach can be used for the stochastic fatigue analysis of structures on topsides FPSO during preliminary design when piping system responses must be estimated analytically. Further work can consider the coupling between the transverse and the longitudinal response of the pipelines on topside of FPSO and related issues.

The literature in engineering education research suggests that a cohesive series of laboratory exercises improves the learning and retention of the material presented in a laboratory course. The literature presents various devices that serve as the focus of a cohesive laboratory series. These devices and their accompanying laboratory series are designed to engage the students and generate enthusiasm in the course. One such device is the Haptic Paddle: a low cost single degree-of-freedom force-reflecting joystick that has enjoyed successful implementations in laboratory courses at other academic institutions. This thesis presents an adaptation of the Haptic Paddle and its accompanying laboratory series to meet specific goals for implementation in the laboratory component of the MECH 343: Modeling Dynamic Systems course offered at Rice University. Additionally, the device was developed to be an attractive platform for dissemination to colleagues who are interested in collaborating in engineering education research.

An experimental protocol for measuring the energy release rate in a non-linear reversible electromechanical body is proposed and summarized. The potential results are capable of shedding light on the true physical nature of the conditions prevailing at the crack surface and in the space within the crack. The experimental procedure is simulated numerically for a linear piezoelectric specimen in a four point bending configuration subjected to electrical loading perpendicular to the crack. Two efficient finite element formulations are presented for nonlinear crack face boundary conditions. Methods for the numerical determination of the crack tip energy release rate and the simulation of the experimental method for obtaining the total energy release rate are developed. It is shown that the crack tip energy release rate calculated under energetically consistent boundary conditions is equal to the total energy release, when the exact boundary conditions are used, there is no such agreement.

This dissertation introduces a new multiscale stochastic finite element method (MSFEM) for determining the mechanical properties of polymer nanocomposites (PNC) consisting of polymers reinforced with single-wall carbon nanotubes (SWCNT). Obviously, reliable characterization of the various properties of nanomaterials such as PNC is indispensable in engineering applications. In this context, it is noted that the results reported in the literature often overestimate the actual mechanical properties of PNC reflecting uncertainty in the assumptions and approximations made. The method proposed herein uses actual experimental characterization information at the nano and micro scales to model the spatial randomness induced by the non-uniform dispersion of SWCNT in polymers, and to determine the mechanical properties of PNC. First, the proposed method defines a material region and identifies randomness at the nanoscale. Second, it develops a random field model that quantifies the spatial randomness in PNC. Then, the method formulates a Monte Carlo finite element (FE) scheme used to solve a specific elasticity problem. This FE scheme incorporates the effects of the local mechanical properties of both phases in PNC and the size, shape, orientation, agglomeration, and dispersion of SWCNT in polymers. The developed MSFEM is used in three applications in the dissertation. In the first, tensile test results of two PNC presented in the literature are used to derive estimates of the Young's modulus (YM) and Poisson ratio (PR). The results demonstrate the success of the proposed method in quantifying the effect of the spatial randomness on the mechanical properties of PNC. The second application uses experimental information about nanoindentation (NI) testing to numerically generate NI data which are subsequently used to compute estimates of the overall YM of PNC. The third application addresses the elastic stability of PNC structures. The computed results show the effect of incorporating SWCNT in polymers, as well as of the material randomness on the buckling loads and modes of the PNC structures. Overall, the proposed MSFEM succeeds in modeling the effect of the spatial randomness on the mechanical properties of PNC by using actual experimental findings, and by efficiently combining information obtained at different length scales.

A novel approach to better model nanomanipulation of a nanosphere lying on a stage via a pushing scheme is presented. Besides its amenability to nonlinear analysis and simulation, the proposed model is also effective in reproducing experimental behavior commonly observed during AFM-type nanomanipulation. The proposed nanomanipulation model consists of integrated subsystems that consistently define the dynamics of the nanomanipulator tip and nanosphere, friction between the nanosphere and the stage, and the contact deformation between the nanomanipulator tip and the nanosphere. The main feature of the proposed nanomanipulation model is the Lund-Grenoble (LuGre) dynamic friction model that reliably represents the stick-slip behavior of atomic friction experienced by the nanosphere. The LuGre friction model introduces a new friction state and has desirable mathematical properties making it a well-posed dynamical model that characterizes friction with fidelity. The proposed nanomanipulation model facilitates further improvement and extension of each subsystem to accommodate other physical phenomena that characterize the physics and mechanics of nanomanipulation. Finally, the proposed model is simulated and compared to existing modes in the literature to demonstrate its versatility and effectiveness.

A ballistic limit equation for hypervelocity impact on thin plates is derived analytically. This equation applies to cases of impulsive impact on a plate that is protected by a multi-shock shield, and is valid in the range of velocity above 6 km/s. Experimental tests were conducted at the NASA Johnson Space Center on square aluminum plates. Comparing the center deflections of these plates with the theoretical deflections of a rigid-plastic plate subjected to a blast load, one determines the dynamic yield strength of the plate material. The analysis is based on a theory for the expansion of the fragmented projectile and on a simple failure criterion. Curves are presented for the critical projectile radius versus the projectile velocity, and for the critical plate thickness versus the velocity. These curves are in good agreement with curves that have been generated empirically.

Combined radiation, convection, and conduction heat transfer are studied in a dispersed two phase flow of gray, laminar, axisymmetric media, where both phases are radiatively participating. The radiative transfer equation in curvilinear coordinates is coupled with an interpenetrating continua energy and momentum formulation for the two phase flow. The radiative transfer equation is set in a novel coordinate system which allows full variation of the radiation intensity in the plane of symmetry and expression of symmetric boundary conditions, without resort to the frequent assumption of azimuthal symmetry. The radiative transfer equation is solved using the discrete ordinates method, chosen for its relative accuracy and ability to numerically complement a differential energy formulation. An appropriate discrete ordinates quadrature is derived for the novel coordinate system. The heat transfer model is used to study a proposed arrangement in which heated particles are seeded into heat exchange tubes running through a furnace in order to enhance heat transfer to a gas flowing in the tubes. It is found that with particles heated to temperatures approaching the furnace temperature, and mass loading ratios (seeded particle specific mass to carrying gas specific mass) up to the order of 10, significant enhancement in heat transfer to the gas is achieved. Such enhancement has the effect of reducing the required heat exchange tube length, thereby reducing the furnace size. Interphase heat transfer between the dispersed particles and the semi-continuous gas is studied in detail by formally modeling combined radiation, conduction, and convection heat transfer between a particle and a semi-infinite medium. Results of this study demonstrate that in the seeding particle case, simple correlations for combined mode heat transfer are accurate. It is also found that the critical particle spacing at which interphase heat transfer is interfered with by neighboring particles is smaller for radiation dominated cases than for conduction dominated cases.