| The focus of the current work is to develop turbulence models that can solve complex flows in a physically realistic and economical manner, while maintaining a reasonable level of computational robustness. Specifically, this thesis presents physics-based models that improve the predictive capability of RANS-based CFD for unsteady or non-stationary, turbulent flows. This is achieved by the development and implementation of a semi-deterministic stress model (SDSM), which captures effects of unsteadiness (particularly Kelvin-Helmholtz or roller vortices) on the time-averaged flow without performing costly, time-dependent simulations. This is an important item to consider in a design environment, where an unsteady simulation is generally impractical. The novel approach is applied to a wide variety of test cases, and significant improvement is achieved compared to results from currently available or "off-the-shelf" turbulence models. For the first time in the literature, it is shown that the effects of this type of unsteadiness can be captured in a steady framework through a model.; To develop the SDSM, it was necessary to obtain a database of time-dependent flow data from a number of test cases. This was accomplished by performing a series of unsteady RANS (URANS) simulations on a variety of fundamental turbulent flows, which exhibited Kelvin-Helmholtz vortex shedding. However, the implementation of currently available turbulence models did not allow for the resolution this type of unsteadiness. To overcome this deficiency, it was necessary to develop a new turbulence model that could actually be used in URANS calculations. Implementation of the new model led to realistic, time-dependent data for this type of unsteadiness. This is the first time that this has been achieved using a URANS-based approach.; For flows that exhibit roller vortices, it is shown that the SDSM approach reduces CPU time by at least 95% for 3-D flows when compared to the URANS approach, while producing essentially the same results. This is due to the much finer grid and small time step size required for URANS. Total CPU costs for the SDSM approach are equivalent to using "off-the-shelf" turbulence models in steady simulations. Thus, a significant amount of physics can be captured in the SDSM approach at a small price. |
