The prime objective of this work is to prove the feasibility of the polyhedral cell methodology in practical situations where turbulent channel or river flow is encountered. To get to this point, several other objectives must be met first. A first step is the design of algorithms for generating a polyhedral mesh and its subsequent software implementation. This is followed by the derivation of the generic discretised equations of flow and turbulence and their implementation in a numerical code, which is to be properly validated against a number of measurements in different flow situations. In the following thesis chapters all required mathematical derivations, as well as the results of validation and application runs are discussed, while the implementation is done in a software model called RSim-3D. This name is short for River Simulation in 3D and it consists of a pre- and postprocessor written in the Java programming language, hence allowing for a platform-independent usage, and a solver module, coded in GNU compliant C because of speed considerations. Due to all of these objectives, the work employs knowledge in the scientific fields of mathematics, geometry, informatics and hydraulic research alike. The work is arranged into five core chapters, each representing a distinct step in model development. First of all, chapter 2 reviews a number of commercial and non-commerical 3D models for computational fluid dynamics, listing their numerical capabilities along with usual fields of application and past project references relevant for hydraulic engineering and research. At the end of this chapter, the RSim-3D model is positioned within the framework of these models to allow for a comparison. In chapter 3, the design and application of polyhedral computation grids is discussed. Algorithms for point distribution and grid generation are the core of this chapter, but is also discusses issues like grid refinement in practical situations and equations for obtaining cell volumes and surface areas in a geometrically complex grid configuration. For such a general grid requires a very general treatment of the governing equations of flow and turbulence, chapter 4 derives the discretised equations in an appropraite way. Furthermore this chapter outlines the boundary conditions of all flow properties required to obtain a solution, before theoretical and practical considerations about numerical issues like stability and convergence conclude that section. The verification and validation of the model is subject to discussion in chapter 5. Validation is done by applying the model to four different flow cases: a wind-channel duct curved by 90 degrees is computed first, followed by a rectangular laboratory flume with an 180 degree bend, and subsequently a channel exhibiting a 270° bend. In the latter two cases, the used grid type is varied to assess its influence on the results obtained. Finally, an S-shaped trapezoidal channel is investigated to make a first step towards the modelling of realistic real-world flow situations. The validation work of chapter 5 is followed by an exemplary application of the model to a reach of the river Danube in chapter 6. Finally, a summary and the discussion of possible future perspectives conclude the work.
