Doctor of Philosophy / Department of Mechanical and Nuclear Engineering / Mingjun Wei / In this study, a global Proper Orthogonal Decomposition (POD)-Galerkin based Reduced Order model (ROM) is proposed. It is extended from usual fixed-domain problems to more general fluid-solid systems with moving boundaries/interfaces. The idea of the extension is similar to the immersed boundary method in numerical simulations which uses embedded forcing terms to represent boundary motions and domain changes. This immersed boundary method allows a globally defined fixed domain including both fluid and solid, where POD-Galerkin projection can be directly applied. However, such a modified approach cannot get away with the unsteadiness of boundary terms which appear as time-dependent coefficients in the new Galerkin model. These coefficients need to be pre-computed for prescribed periodic motion, or worse, to be computed at each time step for non-prescribed (e.g. with fluid-structure interaction) or non-periodic situations. Though computational time for each unsteady coefficient is smaller than the coefficients in a typical Galerkin model, because the associated integration is only in the close neighborhood of moving boundaries. The time cost is still much higher than a typical Galerkin model with constant coefficients. This extra expense for moving-boundary treatment eventually undermines the value of using ROMs. An aggressive approach is to decompose the moving boundary/domain to orthogonal modes and derive another low-order model with fixed coefficients for boundary motion. With this domain decomposition, an approach including two coupled low-order models both with fixed coefficients is proposed. Therefore, the new global ROM with decomposed approach is more efficient. Though the model with the domain decomposition is less accurate at the boundary, it is a fair trade-off for the benefit on saving computational cost. The study further shows, however, that the most time-consuming integration in both approaches, which come from the unsteady motion, has almost negligible impact on the overall dynamics. Dropping these time-consuming terms reduces the computation cost by at least one order while having no obvious effect on model accuracy. Based on this global POD-Galerkin based ROM with forcing term, an improved ROM which can handle the parametric variation of body motions in a certain range is also presented. This study shows that these forcing terms not only represent the moving of the boundary, but also decouple the moving parameters from the computation of model coefficients. The decoupling of control parameters provides the convenience to adapt the model for the prediction on states under variation of control parameters. An improved ROM including a shit mode seems promising in model adaptation for typical problems in a fixed domain. However, the benefit from adding a shit mode to model diminishes when the method is applied to moving-boundary problems. Instead, a combined model, which integrates data from a different set of parameters to generate the POD modes, provides a stable and accurate ROM in a certain range of parametric space for moving-boundary problems. By introducing more data from a different set of parameters, the error of the new model can be further reduced. This shows that the combined model can be trained by introducing more and more information. With the idea of the combined model, the improved global ROM with forcing terms shows impressive capability to predict problems with different unknown moving parameters, and can be used in future parametric control and optimization problems.

abstract: The focus of this investigation includes three aspects. First, the development of nonlinear reduced order modeling techniques for the prediction of the response of complex structures exhibiting "large" deformations, i.e. a geometrically nonlinear behavior, and modeled within a commercial finite element code. The present investigation builds on a general methodology, successfully validated in recent years on simpler panel structures, by developing a novel identification strategy of the reduced order model parameters, that enables the consideration of the large number of modes needed for complex structures, and by extending an automatic strategy for the selection of the basis functions used to represent accurately the displacement field. These novel developments are successfully validated on the nonlinear static and dynamic responses of a 9-bay panel structure modeled within Nastran. In addition, a multi-scale approach based on Component Mode Synthesis methods is explored. Second, an assessment of the predictive capabilities of nonlinear reduced order models for the prediction of the large displacement and stress fields of panels that have a geometric discontinuity; a flat panel with a notch was used for this assessment. It is demonstrated that the reduced order models of both virgin and notched panels provide a close match of the displacement field obtained from full finite element analyses of the notched panel for moderately large static and dynamic responses. In regards to stresses, it is found that the notched panel reduced order model leads to a close prediction of the stress distribution obtained on the notched panel as computed by the finite element model. Two enrichment techniques, based on superposition of the notch effects on the virgin panel stress field, are proposed to permit a close prediction of the stress distribution of the notched panel from the reduced order model of the virgin one. A very good prediction of the full finite element results is achieved with both enrichments for static and dynamic responses. Finally, computational challenges associated with the solution of the reduced order model equations are discussed. Two alternatives to reduce the computational time for the solution of these problems are explored. / Dissertation/Thesis / Ph.D. Aerospace Engineering 2012

Rising global energy demands have led to an increase in demand for clean, sustainable energy. A leading technology for reducing greenhouse gas (GHG) emission for existing coal-power infrastructure is gasification, which has sparked an interest in reactor modelling for design and performance analysis. Reduced order models (ROMs) have seen an increase in popularity for entrained flow gasifiers, as they offer a low-computational alternative to conventional computational fluid dynamic (CFD) modelling while maintaining the integrity of important operational parameters, such as carbon conversion and syngas yield. However, ROMs require more physical parameter inputs than are normally required for CFD modelling, such as the geometry of the gas-solid jet (specifically the jet half-angle). Experiments were conducted to understand the relation between the required input parameters for ROMs, such as fuel flow rate, transport gas flow rate, and jet half-angle, and develop useful correlations for ROM systems. A new configuration for pneumatic conveying was developed and tested at the pilot-scale system at NRCan CanmetENERGY. It was used to study the pneumatic conveying of pulverized fuels, specifically the influence of operating parameters such as pressure drop and gas flow rates on the fuel flow rate, and the geometry of the gas-solid fuel jet (notably the jet half-angle) injected into the gasifier. The mean fuel flow rate of pulverized fuels was shown to increase with increasing pressure drop and with decreasing gas flow rates in the fuel transfer line. The jet half-angle was shown to increase as the solid loading ratio in the jet core was decreased. Finally, the relative fuel flow variability was observed to be significantly influenced by the design of the pneumatic conveying system, with the fluctuations increasing with increasing pressure drop and with decreasing gas flow rate, similar to the mean flow rate.

This thesis deals with the formulation and application of reduced order models based on extraction of dominant structures from a system utilizing the method of proper orthogonal decomposition. Time evolution of computed modes is described by a system of ordinary differential equations, which is gained by means of Galerkin projection of these modes onto the Navier-Stokes equations. This methodology was applied on two test cases Kármán vortex street and vortex rope. In both cases, a CFD simulation of one refference point was carried out and by utilizing gained modes, the corresponding reduced order models were formulated. Their results were assessed by comparing to the refference simulation.

<div>Vacuum is needed in many applications and, there are many types of pumps that can provide the vacuum level needed. One widely used pump is the liquid-ring vacuum pump, which does not involve any solid-solid contacts at interfaces where moving and stationary parts meet. Though liquid-ring vacuum pumps are efficient and robust, manufacturers have aggressive goals on improving efficiency, performance, and range of operations.</div><div> </div><div> In this research, time-accurate computational fluid dynamic (CFD) analyses were performed to study the flow mechanisms in a liquid-ring vacuum pump to understand how it works and how the design can be improved. Based on the understanding gained, a physics based reduced order model was developed for preliminary design of the liquid ring vacuum pumps.</div><div> </div><div> In the CFD analyses, the liquid (water) was modeled as incompressible, the gas (air) as an ideal gas, and turbulence by the shear-stress transport model. The gas-liquid interface was resolved by using the volume-of-fluid method, and rotation of the impeller was enabled by using a sliding mesh. Parameters examined include the suction pressure (75, 300, and 600 Torr) and the impeller's rotational speed (1150, 1450 and 1750 rpm) with the temperature of the gas at the inlet of the suction chamber kept at 300 K and the pressure at the outlet of the exhaust chamber kept at one atmosphere. The CFD solutions generated were verified via a grid sensitivity study and validated by comparing with experimental data. When compared with experiments, results obtained for the flow rate of the gas ingested by the pump had relative errors less than 6\% and results obtained for the power consumed by the pump had relative errors less than 13\%.</div><div> </div><div> Results obtained show the shape of the liquid ring to play a dominant role in creating the expansion ratio or the vacuum needed to draw air into the pump through the suction port and the compression ratio needed to expel the air through the discharge ports. Results were generated to show how centrifugal force from rotation and how acceleration/deceleration from the difference in pressure at the pump's inlet and outlet along with the eccentricity of the impeller relative to the pump's housing affect the shape of the liquid ring. Results were also generated to show how the rotational speed of the impeller and the pressure at the suction port affect the nature of the gas and liquid flow in the pump and the pump’s effectiveness in creating a vacuum. </div><div> </div><div> With the knowledge gained from the CFD study, a physics-based reduced-order model was developed to predict air ingested and power consumed by the pump as well as the liquid ring shape and pressure of the gas and liquid in the pump as a function of design and operating parameters. This model was developed by recognising and demonstrating that the amount of air ingested and power consumed by the pump is strongly dependent on the shape and location of the liquid ring surface. The flow rates of the gas ingested by the pump and the power consumed by the pump predicted by the model were compared with experimental data and relative errors were less than 12\% and 17\% respectively.</div>

The purpose of this study was to develop a robust fluid-structure-interaction (FSI) technology that can accurately model non-linear flutter responses for sub- and transonic fluid flow. The Euler equation set governs the fluid domain, which was spatially discretised by a vertex-centred edge-based finite volume method. A dual-timestepping method was employed for the purpose of temporal discretisation. Three upwind schemes were compared in terms of accuracy, efficiency and robustness, viz. Roe, HLLC (Harten-Lax-Van Leer with contact) and AUSM+-up Advection Up-stream Splitting Method). For this purpose, a second order unstructured MUSCL (Monotonic Upstream-centred Scheme for Conservation Laws) scheme, with van Albada limiter, was employed. The non-linear solid domain was resolved by a quadratic modal reduced order model (ROM), which was compared to a semi-analytical and linear modal ROM. The ROM equations were solved by a fourth order Runge-Kutta method. The fluid and solid were strongly coupled in a partitioned fashion with the information being passed at solver sub-iteration level. The developed FSI technology was verified and validated by applying it to test cases found in literature. It was demonstrated that accurate results may be obtained, with the HLLC upwind scheme offering the best balance between accuracy and robustness. Further, the quadratic ROM offered significantly improved accuracy when compared to the linear method. / Dissertation (MEng)--University of Pretoria, 2011. / Mechanical and Aeronautical Engineering / unrestricted

We investigate the implementation of nontraditional open-loop and closed-loop control techniques to systems at the micro and macro scales. At the macro level, we consider a quay-side container crane. It is known that the United States relies on ocean transportation for 95% of cargo tonnage that moves in and out of the country. Each year over six million loaded marine containers enter U.S. ports. Current growth predictions indicate that container cargo will quadruple in the next twenty years. To cope with this rapid growth, we develop a novel open-loop input-shaping control technique to mitigate payload oscillations on quay-side container cranes. The proposed approach is suitable for automated crane operations, does not require any alterations to the existing crane structure, uses the maximum crane capabilities, and is based on an accurate two-dimensional four-bar-mechanism model of a container crane. The shaped commands are based on a nonlinear approximation of the two-dimensional model frequency and, unlike traditional input-shaping techniques, our approach can account for large hoisting operations. For operator-in-the-loop crane operations, we develop a closed-loop nonlinear delayed-position feedback controller. Key features of this controller are that it: does not require major modifications to the existing crane structure, accounts for motion inversion delays, rejects external disturbances, and is superimposed on the crane operator commands. To validate the controllers, we construct a 1:10 scale model of a 65-ton quay-side container crane. The facility consists of a 7-meter track, 3.5-meter hoisting cables, a trolley, a traverse motor, two hoisting motors, and a 50-pound payload. Using this setup, we demonstrated the effectiveness of the controllers in mitigating payload oscillations in both of the open-loop and closed-loop modes of operation. At the micro level, we consider a micro optical device known as the torsional micromirror. This device has a tremendous number of industrial and consumer market applications including optical switching, light scanning, digital displays, etc. To analyze this device, we develop a comprehensive model of an electrically actuated torsional mirror. Using a Galerkin expansion, we develop a reduced-order model of the mirror and verify it against experimental data. We investigate the accuracy of representing the mirror using a two-degrees-of-freedom lumped-mass model. We conclude that, under normal operating conditions, the statics and dynamics of the mirror can be accurately represented by the simplified lumped-mass system. We utilize the lumped-mass model to study and analyze the nonlinear dynamics of torsional micromirrors subjected to combined DC and resonant AC excitations. The analysis is aimed at enhancing the performance of micromirrors used for scanning applications by providing better insight into the effects of system parameters on the microscanner's optimal design and performance. Examining the characteristics of the mirror response, we found that, for a certain DC voltage range, a two-to-one internal resonance might be activated between the first two modes. Due to this internal resonance, the mirror exhibits complex dynamic behavior. This behavior results in undesirable vibrations that can be detrimental to the scanner performance. Torsional micromirrors are currently being implemented to provide all-optical switching in fiber optic networks. Traditional switching techniques are based on converting the optical signal into electrical signal and back into optical signal before it can be switched into another fiber. This reduces the rate of data transfer substantially. To realize fast all-optical switching, we enhance the transient dynamic characteristics and performance of torsional micromirrors by developing a novel technique for preshaping the voltage commands applied to activate the mirror. This new approach is the first to effectively account for inherent nonlinearities, damping effects, and the energy of the significant higher modes. Using this technique, we are able to realize very fast switching operations with minimal settling time and almost zero overshoot. / Ph. D.

No description available.

No description available.

To improve the performance of the three-way catalytic (TWC) converter, advanced control strategies and onboard diagnostics (OBD) systems are needed. Both rely on a relatively accurate but computationally efficient TWC converter model. This thesis aims to develop a control-oriented model that can be employed to develop the control strategies and OBD systems of the TWC converter. The thesis consists of two parts, i.e., the high-fidelity model development and the model reduction. Firstly, a high-fidelity model is built using the energy and mass conservation principles. In this model, a constant inlet simulation is used to validate the warming-up characteristics, and a driving cycle simulation is used to calibrate the reaction rate parameters. The results of the simulation show that the high-fidelity model has adequate accuracy. Secondly, a reduced-order model is developed based on phase and reaction simplifications of the high-fidelity model. The aim of the development of the reduced-order model is to propose a computationally efficient model for further development of control strategies and state estimators for OBD systems. The accuracy of the reduced-order model is then validated by means of simulations. / Thesis / Master of Science (MSc)