Detached Eddy Simulations (DES) is a hybrid URANS-LES technique that was proposed to obtain computationally feasible solutions of high Reynolds number flows undergoing massive separation with reliable accuracy. Since its inception, DES has been applied to a wide variety of flow fields, but mostly limited to unbounded external aerodynamic flows. This is the first study to apply and validate DES to predict the internal flow and heat transfer in non-canonical flows of industrial relevance. The prediction capabilities of DES in capturing the effects of Coriolis forces, which are induced by rotation, and centrifugal buoyancy forces, which are induced by thermal gradients, are also authenticated. The accurate prediction of turbulent flows is sensitive to the level of turbulence predicted by the turbulence scheme. By treating the regions of interest in LES mode, DES allows the unsteadiness in these regions to develop and hence predicts the turbulence levels accurately. Additionally, this permits DES to capture the effects of system rotation and buoyancy. Computations on a rotating system (a sudden expansion duct) and a system subjected to thermal gradients (cavity with a heated wall) validate the prediction capability of DES. The application of DES is further extended to a non-canonical, internal flow which is of relevance in internal cooling of gas turbine blades. Computations of the fully developed flow and heat transfer shows that DES surpasses several shortcomings of the RANS model on which it is based. DES accurately predicts the primary and secondary flow features, the turbulence characteristics and the heat transfer in stationary ducts and in rotating ducts, where the effects of Coriolis forces and centrifugal buoyancy forces are dominant. DES computations are carried out at a computational cost that is almost an order of magnitude less than the LES with little compromise on the accuracy. However, the capabilities of DES in predicting the transition to turbulence are inadequate, as highlighted by the flow features and the heat transfer in the developing region of the duct. But once the flow becomes fully turbulent, DES predicts the flow physics and shows good quantitative agreement with the experiments and LES. / Ph. D.

Increasing the turbine inlet temperature of gas turbine engines significantly increases their power output and efficiency, but it also increases the likelihood of thermal failure. Internal passages with tiny ribs are typically cast into turbine blades to cool them, and the ability to accurately predict the flow and heat transfer within these channels leads to higher design reliability and prevention of blade failure resulting from local thermal loading. Prediction of the flow through these channels is challenging, however, because the flow is highly turbulent and anisotropic, and the presence of rotational body forces further complicates the flow. Large Eddy Simulations are used to study these flows because of their ability to predict the unsteady flow effects and anisotropic turbulence more reliably than traditional RANS closure models. Calculations in a stationary duct are validated with experiments in the developing flow, fully developed, and 180° bend regions to establish the accuracy and prediction capability of the LES calculations and to aid in understanding the major flow structures encountered in a ribbed duct. It is found that most flow and heat transfer calculations come to within 10-15% of the measurements, typically showing excellent agreement in all comparisons. In the developing flow region, Coriolis effects are found to destabilize turbulence and increase heat transfer along the trailing wall (pressure side), while decreasing leading wall heat transfer by stabilizing turbulence. Coriolis forces improve flow turning in the 180° bend by shifting the shape of the separated recirculation zone at the tip of the dividing wall and increasing the mainstream flow area. In addition, turbulence is attenuated near the leading wall throughout the bend, while Coriolis forces have little effect on trailing wall turbulence in the bend. Introducing and increasing centrifugal buoyancy in the developing flow region increases trailing wall heat transfer monotonically. Along the leading wall, buoyancy increases the size of the recirculation zones, shifting the peak heat transfer to a region upstream of the rib, which decreases heat transfer at low buoyancy parameters but increases it as the buoyancy parameter is increased beyond a value of 0.3. Centrifugal buoyancy in the 180° bend initially decreases the size of the recirculation zone at the tip of the dividing wall, increasing flow area and decreasing flow impingement. At high buoyancy, however, the recirculation zone shifts to the middle of the bend, increasing flow resistance and causing strong flow impingement on the back wall. The Boussinesq approximation is used in the buoyancy calculations, but the accuracy of the approximation comes into question in the presence of large temperature differences. A variable property algorithm is developed to calculate unsteady low speed flows with large density variations resulting from large temperature differences. The algorithm is validated against two test cases: Rayleigh-Bénard convection and Poiseuille-Bénard flow. Finally, design issues in rotating ribbed ducts are considered. The fully developed assumption is discussed with regard to the developing flow region, and controlling the recirculation zone in the 180° bend is considered as a way to determine the blade tip heat transfer and pressure drop across the bend. / Ph. D.