Thesis (M. Phil.)--University of Hong Kong, 1980.

Thesis (PH.D.) -- Syracuse University, 2004. / "Publication number AAT 3149061."

<p> Univariate cubic <i>L</i>¹ spline fits have been successful to preserve the shapes of 2D data with abrupt changes. The reason is that the minimization of <i>L</i>¹ norm of the data is considered, as opposite to <i>L</i>² norm. While univariate <i>L</i>¹ spline fits for 2D data are discussed by many, bivariate <i>L</i>¹ spline fits for 3D data are yet to be fully explored. This thesis aims to develop bicubic <i>L</i>¹ spline fits for 3D data approximation. This can be achieved by solving a bi-level optimization problem. One level is bivariate cubic spline interpolation and the other level is <i> L</i>¹ error minimization. In the first level, a bicubic interpolated spline surface will be constructed on a rectangular grid with necessary first and second order derivative values estimated by using a 5-point window algorithm for univariate <i>L</i>¹ interpolation. In the second level, the absolute error (i.e. <i> L</i>¹ norm) will be minimized using an iterative gradient search. This study may be extended to higher dimensional cubic <i>L</i>¹ spline fits research.</p><p>

<p> The development of expected-distance formulas for multi-dock-door, unit-load warehouse configurations is the focus of the dissertation. From formulations derived, the width-to-depth ratios minimizing expected distances are obtained for rectangle-shaped, unit-load warehouse configurations. Partitioning the storage region in the warehouse into three classes, the performance of a multi-dock-door, unit-load warehouse is studied when storage regions can be either rectangle-shaped or contour-line-shaped. </p><p> Our first contribution is the development of formulas for expected distance traveled in storing and retrieving unit loads in a rectangle-shaped warehouse having multiple dock doors along one warehouse wall and storage racks aligned perpendicular to that wall. Two formulations of the optimization problem of minimizing expected distance are considered: a discrete formulation and a continuous formulation with decision variables being the width and depth of the warehouse for single- and dual-command travel. Based on dock door configurations treated in the literature and used in practice, three scenarios are considered for the locations of dock doors: 1) uniformly distributed over the entire width of a wall; 2) centrally located on a wall with a fixed distance between adjacent dock doors; and 3) not centrally located on a wall, but with a specified distance between adjacent dock doors. </p><p> Our second contribution is the investigation of the effect on the optimal width-to-depth ratio (shape factor) of the number and locations of dock doors located along one wall or two adjacent walls of the warehouse. Inserting a middle-cross-aisle in the storage area, storage racks are aligned either perpendicular or parallel to warehouse walls containing dock doors. As with the warehouse having storage racks aligned perpendicular to the warehouse wall, discrete and continuous formulations of the optimization problem are developed for both single- and dual-command travel and three scenarios for dock-door locations are investigated. </p><p> Our final contribution is the analysis of the performance of a unit-load warehouse when a storage region or storage regions can be either rectangle-shaped or contour-line-shaped. Particularly, we consider two cases for the locations of dock doors: equally spaced over an entire wall of the warehouse and centrally located on a wall, but with a specified distance between adjacent dock doors. Minimizing expected distance, the best rectangle-shaped configuration is determined and its expected distance is compared with the expected distance in its counterpart contour-line-shaped configuration.</p><p>

<p> In deterministic computer experiments, a computer code can often be run at different levels of complexity/fidelity and a hierarchy of levels of code can be obtained. The higher the fidelity and hence the computational cost, the more accurate output data can be obtained. Methods based on the co-kriging methodology Cressie (2015) for predicting the output of a high-fidelity computer code by combining data generated to varying levels of fidelity have become popular over the last two decades. For instance, Kennedy and O&rsquo;Hagan (2000) first propose to build a metamodel for multi-level computer codes by using an auto-regressive model structure. Forrester et al. (2007) provide details on estimation of the model parameters and further investigate the use of co-kriging for multi-fidelity optimization based on the efficient global optimization algorithm Jones et al. (1998). Qian and Wu (2008) propose a Bayesian hierarchical modeling approach for combining low-accuracy and high-accuracy experiments. More recently, Gratiet and Cannamela (2015) propose sequential design strategies using fast cross-validation techniques for multi-fidelity computer codes. </p><p> This research intends to extend the co-kriging metamodeling methodology to study steady-state simulation experiments. First, the mathematical structure of co-kriging is extended to take into account heterogeneous simulation output variances. Next, efficient steady-state simulation experimental designs are investigated for co-kriging to achieve a high prediction accuracy for estimation of steady-state parameters. Specifically, designs consisting of replicated longer simulation runs at a few design points and replicated shorter simulation runs at a larger set of design points will be considered. Also, design with no replicated simulation runs at long simulation is studied, along with different methods for calculating the output variance in absence of replicated outputs. </p><p> Stochastic co-kriging (SCK) method is applied to an M/M/1, as well as an M/M/5 queueing system. In both examples, the prediction performance of the SCK model is promising. It is also shown that the SCK method provides better response surfaces compared to the SK method.</p><p>

In this dissertation, comprehensive thermal models are developed to account for the influence of grinding conditions on the temperatures generated. Various aspects of the heat transfer problem are analyzed and experimentally investigated including the fluid flow through the grinding zone, energy partition to the workpiece, heat flux distribution at the workpiece surface, cooling by the grinding fluid, transient temperatures during a grinding pass, and burn-out heat flux limits for creep-feed grinding. The work begins with an analysis of fluid flow through the grinding zone. The model predicts tangential and radial fluid velocities relative to the wheel, the depth of fluid penetration into the wheel pores, and the useful flow rate through the grinding zone. The useful flow rates predicted using the model are combined with experimental measurements for both conventional and creep-feed wheels to estimate the effective wheel porosity, which provides a measure of the ability of a wheel to pump fluid through the grinding zone. Thermal models are then developed to predict the partition of the total grinding heat to the workpiece, the fluid, and the grinding wheel. The thermal properties of the composite depend on the effective wheel porosity as described above. Results are obtained for the local partition of the total energy to the workpiece and to the composite along the grinding zone for different grinding conditions. For creep-feed grinding in the absence of fluid burn-out, the energy partition to the workpiece is typically only a few percent of the total. Inverse heat transfer methods are developed to determine the heat flux distribution to the workpiece from measured temperature distributions for regular grinding of steels with both conventional aluminum oxide wheels and CBN wheels. These results are also used to estimate the energy partition to the workpiece, as well as the wheel-workpiece contact length and convective heat transfer coefficient on the workpiece surface. A transient thermal analysis was developed. For creep-feed grinding, the temperature rise is found to be much bigger than the quasi-steady state value at the end of the cut. The transient temperature model is subsequently applied to predicting burn-out heat flux limits for creep-feed grinding. (Abstract shortened by UMI.)

Several design methodologies for automatically designing injection molds were developed to reduce warpage of injection molded parts. The part warpage was defined from warpage simulation, so as to represent various deformation behavior of the molded part. A concept for deliberately varying part wall thicknesses, within feasible dimensional tolerances to reduce the warpage, was introduced. The wall thicknesses, that minimized the effect of process fluctuation inherent to molding process, were obtained using the Taguchi method. The results of the example indicated that one can considerably reduce the warpage, and may also reduce material costs, by varying the wall thicknesses. A modified complex method was developed to optimize various domains in injection mold design. The modified complex method significantly reduced the part warpage with a moderate number of independent analysis evaluations for the proposed optimization of: part wall thicknesses; location and sizes of cooling channels; and process conditions. The runner system of multiple cavity molds was successfully balanced by adjusting the sizes of runners and gates using the iterative redesign method integrated with packing simulation in two case studies. A design technique for automatically determining the "best" gate location(s) of injection molds was presented in association with the three important parameters in gating design--the part warpage, the weld and meld line location at the critical areas, and the Izod impact strength at a specific region of the part. The difficulty in predicting accurate values of engineering property like Izod impact strength is that they vary throughout a part with respect to the thermomechanical history. Upon evaluating each gating design, the trained neural network computation predicts, regardless of part geometry, Izod impact strength by a non-parameteric modeling of the complex relation with thermomechanical processing histories. As the results, when the polymer resin and the part geometry were predetermined, the part warpage could be considerably reduced by the developed design tools for the part wall thicknesses, the mold systems including gating, runner, and cooling system, as well as the process conditions.

The objective of this investigation is to evaluate the ability of analytical and computational models to describe the momentum and heat transfer between the gas and particles in a falling-bed heat exchanger. Experimental data are presented for a test falling-bed heat exchanger. Measured temperatures, pressures, and overall heat transfer rates are compared to predicted values from analytical and computational models, and the capabilities and deficiencies of these modeling methods are discussed. In addition, the effect of the addition of a particle distributor on the performance of the falling-bed heat exchanger is measured. In the falling-bed heat exchanger, solid particles fall through a vertical column against a counterflowing gas stream flowing upward with a velocity less than the terminal velocity of the particle. Heat is exchanged between the falling particles and rising gas. This arrangement has been proposed for heat recovery and regeneration in power plants and other process applications. The ability to model and predict the heat transfer rate between the gas and particles is critical to the design of the falling-bed heat exchanger. The heat transfer between the gas and solid particles in these devices has typically been modeled by assuming steady-state and ideal, uniform, one-dimensional flow of the continuous fluid and the particle or droplets. This model, termed the uniform mixing model in this study, has been used in many instances to estimate the effective heat transfer coefficient and Nusselt number of the falling droplets and particles as a function of effective Reynolds number from experimental data. The addition of a particle distributor has been shown to increase the heat transfer effectiveness of the falling-bed heat exchanger in experiments. It has been determined that the uniform mixing model generally does not provide an accurate representation of the falling-bed heat exchanger, as it cannot account for gas and particle maldistributions such as those created by a particular particle distributor design. Computational fluid dynamics, which can permit the modeling of these spatial maldistributions, has been used to model the falling-bed heat exchanger. The predictions of the overall heat transfer rate from computational fluid dynamics are in better agreement with the measured values. However, discrepancies between the predicted and measured pressures and local temperatures indicate that the modeling of the turbulent mixing of momentum and energy is inadequate.

A numerical simulation of the injection-compression molding process is developed with the capability of predicting product quality attributes including residual stress, birefringence, and warpage. A hybrid finite element/finite difference method is employed to calculate the temperature and pressure fields of the process with a non-isothermal compressible flow model. The process simulation is coupled with viscoelastic constitutive models to predict the flow and thermally induced residual stresses. A structural finite element analysis is formulated to predict the warpage of the disc due to asymmetric thermal stress and gravity after demolding. The flow and thermally induced birefringence of injection-compression molded optical media is predicted by applying a stress-optical rule to the flow and thermally induced stresses. The resulting model considers the contributions of flow and cooling induced molecular orientation, and the transient effect of thermal stress and pressure on the birefringence. The simulation is validated by compact-disc-recordable moldings with an optical grade polycarbonate under different processing conditions.

Distractions outside the vehicle are a growing cause of concern. Effects of external distractions are not just limited to younger novice drivers, as is the case with in-vehicle distractions. Experienced drivers' ability to maintain their attention on the forward roadway and their ability to anticipate hazards is also compromised in the presence of external distractions. This dissertation seeks to both understand the extent to which external-to-vehicle distractions affect driver safety and study the potential remedies for this problem. Driver behaviors, specifically the ability to anticipate and respond to hazards in the presence of external-to-vehicle distractions, were evaluated for different age groups in a driving simulator. Drivers' performance was evaluated on three categories of hazard anticipation scenarios; (a) top-down explicitly cued scenarios (drivers were cued before they encountered the external-to-vehicle task), (b) top-down implicitly cued scenarios (drivers were cued before they encountered the external-to-vehicle task) and (c) bottom-up cued scenarios (drivers were cued after they encountered the external-to-vehicle task). Eye movement information and vehicle parameters were collected and analyzed. Drivers of all age groups were significantly affected by the presence of external-to-vehicle distractions regardless of the type of cue: drivers took long glances at the external-to-vehicle task, did not scan the appropriate locations for risk relevant elements, and responded significantly slower to hazards in the presence of external-to-vehicle distractions (Experiments 1 and 2). Training was considered as an intervention that could mitigate the effects of external distractions. As a start, the effect of a training program (FOCAL) that reduced the duration of in-vehicle glances on the duration of external-to-vehicle glances was evaluated in a simulator. The training was not effective in reducing the long duration of glances at the external task (Experiment 3). This finding lead to the development of a PC-based attention maintenance training program, FOCAL-Ex, that specifically addressed the issue of external-to-vehicle distractions. The evaluation was done on a PC. The percentage of external-to-vehicle tasks in which trained drivers took especially long glances was significantly less than this percentage for placebo trained drivers (Experiment 4).