This thesis presents a method for fault detection and precedent-free isolation for two types of channel flow systems, which were modeled with the finite element method. Unlike previous fault detection methods, this method requires no a priori knowledge or training pertaining to any particular fault. The basis for anomaly detection was the model of normal behavior obtained using the recently introduced Growing Structure Multiple Model System (GSMMS). Anomalous behavior is then detected as statistically significant departures of the current modeling residuals away from the modeling residuals corresponding to the normal system behavior. Distributed anomaly detection facilitated by multiple anomaly detectors monitoring various parts of the thermal-fluid system enabled localization of anomalous partitions of the system without the need to train classifiers to recognize an underlying fault. / text

Luge is an Olympic sport in which athletes ride feet-first on sleds down an ice-covered track. Competitors spring from the starting position and accelerate their sled by paddling with spiked gloves against the ice surface. Once the Luger leaves the starting section, their downhill motion is solely propelled by the effects of gravity. Athletes compete, one after the other, for the fastest time. Runs can differ by as little as a thousandth of a second, meaning that every minor sled adjustment, change of line choice, and shift of body position is critical. In the past, the sport of Luge has progressed through a series of steps involving trial and error, where changes to the sled and strategy rely more on intuition and race results, rather than in-depth, mathematical analysis. In an effort to try and improve track times for the US Olympic Luge team, a track and driver model is in development in order to simulate a sled going down the track. By doing this, the hope is to be able to pinpoint areas of possible improvement to the sled and see how adjustments can affect the optimum line down the track. A part of this model, which is the focus of the following paper, is the inclusion of an analysis to identify the frictional relationship between the ice surface and the steels of the sled. The model created of the ice-steel interaction was put in the form of a function file, which includes inputs of down force, ice temperature, sled velocity, and steel geometry. Creation of this model and completion of a set of parametric studies allowed for further understanding the interaction between the sled steels and ice surface, specifically applying to the sport of Luge. The model predicts for lower temperatures that at slower sled velocities the coefficient of friction is greater compared to faster sled velocities. This relationship inverts as the ice temperature moves closer to the melting temperature. A sharper steel edge radius was found to be beneficial in lowering the coefficient of friction at lower sled velocities. The sharp edge radius friction benefit decreases as the sled speed increases and is predicted to actually increase friction slightly compared to duller blades at greater velocities. A flat as possible rocker radius lowers friction at all sled velocities, as well as in banked turns where two contact patches are possible. On curves, the pressure on the steel is increased due to the effects of centripetal accelerations. A 1 g versus 5 g normal loading, experienced on the last turns of the track, increases the coefficient of friction on the blade, but also increases the allowable lateral force on the sled before side slip occurs. Understanding the relationships of these parameters, along with the information that may be gained from the driver model, may prove to be useful in choosing optimum sled characteristics and line choice.

<div>The objective of this dissertation is to develop new methods in the areas of multiscale topology optimization, thermomechanical topology optimization including heat convection, and thermal-fluid topology optimization. The dissertation mainly focuses on developing five innovative topology optimization algorithms with respect to structure and multistructure coupling fluid, thermal and mechanical analysis, in order to solve customary design requirements. Most of algorithms are coded as in-house code in MATLAB.</div><div><br></div><div><div>In Chapter One, a brief introduction of topology optimization, a brief literature review and the objective is presented. Five innovative algorithms are illustrated in Chapter Two</div><div>to Six. From Chapter Two to Four, the methods with respect to multiscale approach are presneted. and Chapter Five and Six aims to contribute further research associated with</div><div>topology optimization considering heat convection. In Chapter Two, a multiphse topology optimization of thermomechanical structures is presented, in which the optimized structure is composed of several phases of prescribed lattice unit cells. Chapter Three presents a</div><div>Multiscale, thermomechanical topology optimization of self-supporting cellular structures. Each lattice unit cell have a optimised porousity and diamond shape that benefit additive</div><div>manufacturing. In Chapter Four, the multiscale approach is extended to topology optimization involved with fluid mechanics problem to design optimized micropillar arrays in</div><div>microfludics devices. The optimised micropillars minimize the energy loss caused by local fluid drag force. In Chapter Five, a novel thermomechanical topology optimization is developed, in order to generate optimized multifunctional lattice heat transfer structure. The algorithm approximate convective heat transfer by design-dependent heat source and natural convection. In Chapter Six, an improved thermal-fluid topology optimization method is created to flexibly handle the changing of thermal-fluid parameters such as external heat source, Reynolds number, Prandtl number and thermal diffusivity. The results show the</div><div>changing of these parameters lead versatile optimized topologies. Finally, the summary and recommendations are presented in Chapter Seven.</div></div><div><br></div>

The steam generator plays a primary role in the safety and performance of a pressurized water reactor nuclear power plant. The cost to utilities is in the order of millions of Rands a year as a direct result of damage to steam generators. The damage results in lower efficiency or even plant shutdown. It is necessary for the utility and for academia to have models of nuclear components by which research and analysis may be performed. It must be possible to analyse steam generator performance for both day-to-day operational analysis as well as in the case of extreme accident scenarios. The homogeneous model for two-phase flow is simpler in its implementation than the two-fluid model, and therefore suffers in accuracy. Its advantage lies in its quick turnover time for development of models and subsequent analysis. It is often beneficial for a modeller to be able to quickly set up and analyse a model of a system, and a trade-off between accuracy and time-management is thus required. Searches through available literature failed to provide answers to how the homogeneous model compares with the two-fluid model for operational and safety analysis. It is expected to see variations between the models, from the analysis of the mathematics, but it remains to be shown what these differences are. The purpose of this study was to determine how the homogeneous model for two-phase flow compares with the two-fluid model when applied to a u-tube steam generator of a typical pressurized water reactor. The steam generator was modelled in both RELAP5 and in Flownex. A custom script was written for Flownex in order to implement the Chen correlation for boiling heat transfer. This was significantly less detailed than RELAP5’s solution of a matrix of flow regimes and heat transfer correlations. The geometry of the models were based on technical drawings from Koeberg Nuclear Power Plant, and were simplified to a one-dimensional model. Plant data obtained from Koeberg was used to validate the models at 100%, 80% and 60% power output. It was found that the overall heat transfer rate predicted with the RELAP5 two-fluid model was within 1.5% of the measured data from the Koeberg plant. The results generated by the homogeneous model for the overall heat transfer were within 4.5% of the measured values. However, the differences in the detailed temperature distributions and heat transfer coefficient values were quite significant at the inlet and outlet ends of the tube bundle, at the bottom tube sheet of the steam generator. In this area the water-level was not accurately modelled by the homogeneous model, and therefore there was an under-prediction in heat transfer in that region. Large differences arose between the Flownex and RELAP5 solutions due to difference in the heat transfer correlations used. The Flownex model exclusively implemented the Chen correlation, while RELAP5 implements a flow regime map correlated to a table of heat transfer correlations. It was concluded that the results from the homogeneous model for two-phase flow do not differ significantly when compared with the two-fluid model when applied to the u-tube steam generator at the normal operating conditions. Significant differences do, however, occur in lower regions of the boiler where the quality is lower. We conclude that the homogeneous model offers significant advantage in simplicity over the two-fluid model for normal operational analysis. This may not be the case for detailed accident analysis, which was beyond the scope of this study. / Thesis (MIng (Nuclear Engineering))--North-West University, Potchefstroom Campus, 2013.

The steam generator plays a primary role in the safety and performance of a pressurized water reactor nuclear power plant. The cost to utilities is in the order of millions of Rands a year as a direct result of damage to steam generators. The damage results in lower efficiency or even plant shutdown. It is necessary for the utility and for academia to have models of nuclear components by which research and analysis may be performed. It must be possible to analyse steam generator performance for both day-to-day operational analysis as well as in the case of extreme accident scenarios. The homogeneous model for two-phase flow is simpler in its implementation than the two-fluid model, and therefore suffers in accuracy. Its advantage lies in its quick turnover time for development of models and subsequent analysis. It is often beneficial for a modeller to be able to quickly set up and analyse a model of a system, and a trade-off between accuracy and time-management is thus required. Searches through available literature failed to provide answers to how the homogeneous model compares with the two-fluid model for operational and safety analysis. It is expected to see variations between the models, from the analysis of the mathematics, but it remains to be shown what these differences are. The purpose of this study was to determine how the homogeneous model for two-phase flow compares with the two-fluid model when applied to a u-tube steam generator of a typical pressurized water reactor. The steam generator was modelled in both RELAP5 and in Flownex. A custom script was written for Flownex in order to implement the Chen correlation for boiling heat transfer. This was significantly less detailed than RELAP5’s solution of a matrix of flow regimes and heat transfer correlations. The geometry of the models were based on technical drawings from Koeberg Nuclear Power Plant, and were simplified to a one-dimensional model. Plant data obtained from Koeberg was used to validate the models at 100%, 80% and 60% power output. It was found that the overall heat transfer rate predicted with the RELAP5 two-fluid model was within 1.5% of the measured data from the Koeberg plant. The results generated by the homogeneous model for the overall heat transfer were within 4.5% of the measured values. However, the differences in the detailed temperature distributions and heat transfer coefficient values were quite significant at the inlet and outlet ends of the tube bundle, at the bottom tube sheet of the steam generator. In this area the water-level was not accurately modelled by the homogeneous model, and therefore there was an under-prediction in heat transfer in that region. Large differences arose between the Flownex and RELAP5 solutions due to difference in the heat transfer correlations used. The Flownex model exclusively implemented the Chen correlation, while RELAP5 implements a flow regime map correlated to a table of heat transfer correlations. It was concluded that the results from the homogeneous model for two-phase flow do not differ significantly when compared with the two-fluid model when applied to the u-tube steam generator at the normal operating conditions. Significant differences do, however, occur in lower regions of the boiler where the quality is lower. We conclude that the homogeneous model offers significant advantage in simplicity over the two-fluid model for normal operational analysis. This may not be the case for detailed accident analysis, which was beyond the scope of this study. / Thesis (MIng (Nuclear Engineering))--North-West University, Potchefstroom Campus, 2013.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第19681号 / 工博第4136号 / 新制||工||1638(附属図書館) / 32717 / 京都大学大学院工学研究科機械理工学専攻 / (主査)教授 西脇 眞二, 教授 稲室 隆二, 教授 松原 厚 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

京都大学 / 新制・課程博士 / 博士(工学) / 甲第24226号 / 工博第5054号 / 新制||工||1789(附属図書館) / 京都大学大学院工学研究科機械理工学専攻 / (主査)教授 平山 朋子, 教授 岩井 裕, 教授 松原 厚 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

abstract: First, in a large-scale structure, a 3-D CFD model was built to simulate flow and temperature distributions. The flow patterns and temperature distributions are characterized and validated through spot measurements. The detailed understanding of them then allows for optimization of the HVAC configuration because identification of the problematic flow patterns and temperature mis-distributions leads to some corrective measures. Second, an appropriate form of the viscous dissipation term in the integral form of the conservation equation was considered, and the effects of momentum terms on the computed drop size in pressure-atomized sprays were examined. The Sauter mean diameter (SMD) calculated in this manner agrees well with experimental data of the drop velocities and sizes. Using the suggested equation with the revised treatment of liquid momentum setup, injection parameters can be directly input to the system of equations. Thus, this approach is capable of incorporating the effects of injection parameters for further considerations of the drop and velocity distributions under a wide range of spray geometry and injection conditions. Lastly, groundwater level estimation was investigated using compressed sensing (CS). To satisfy a general property of CS, a random measurement matrix was used, the groundwater network was constructed, and finally the l-1 optimization was run. Through several validation tests, correct estimation of groundwater level by CS was shown. Using this setup, decreasing trends in groundwater level in the southwestern US was shown. The suggested method is effective in that the total measurements of registered wells can be reduced down by approximately 42 %, sparse data can be visualized and a possible approach for groundwater management during extreme weather changes, e.g. in California, was demonstrated. / Dissertation/Thesis / Doctoral Dissertation Mechanical Engineering 2015

This study solves the problem of unsteady free convection from an inclined heated tube both numerically and analytically. The tube is taken to have an elliptic cross-section having a constant heat flux applied to its surface. The surrounding fluid is viscous and incompressible and infinite in extent. The Boussinesq approximation is used to describe the buoyancy force driving the flow. The underlying assumptions made in this work are that the flow remains laminar and two-dimensional for all time. This enables the Navier-Stokes and energy equations to be formulated in terms of the streamfunction, and vorticity. We assume that initially an impulsive heat flux is applied to the surface and that both the tube and surrounding fluid have the same initial temperature. The problem is solved subject to the no-slip and constant heat flux conditions on the surface together with quiescent far-field and initial conditions. An approximate analytical-numerical solution was derived for small times, t and large Grashof numbers, Gr. This was done by expanding the flow variables in a double series in terms of two small parameters and reduces to solving a set of differential equations. The first few terms were solved exactly while the higher-order terms were determined numerically. Flow characteristics presented include average surface temperature plots as well as surface vorticity and surface temperature distributions. The results demonstrate that the approximate analytical-numerical solution is in good agreement with the fully numerical solution for small t and large Gr.

This study solves the problem of unsteady free convection from an inclined heated tube both numerically and analytically. The tube is taken to have an elliptic cross-section having a constant heat flux applied to its surface. The surrounding fluid is viscous and incompressible and infinite in extent. The Boussinesq approximation is used to describe the buoyancy force driving the flow. The underlying assumptions made in this work are that the flow remains laminar and two-dimensional for all time. This enables the Navier-Stokes and energy equations to be formulated in terms of the streamfunction, and vorticity. We assume that initially an impulsive heat flux is applied to the surface and that both the tube and surrounding fluid have the same initial temperature. The problem is solved subject to the no-slip and constant heat flux conditions on the surface together with quiescent far-field and initial conditions. An approximate analytical-numerical solution was derived for small times, t and large Grashof numbers, Gr. This was done by expanding the flow variables in a double series in terms of two small parameters and reduces to solving a set of differential equations. The first few terms were solved exactly while the higher-order terms were determined numerically. Flow characteristics presented include average surface temperature plots as well as surface vorticity and surface temperature distributions. The results demonstrate that the approximate analytical-numerical solution is in good agreement with the fully numerical solution for small t and large Gr.