Structured grid generation for gas turbine combustion systems
Commercial pressures to reduce time-scales encourage innovation in the design and analysis cycle of gas turbine combustion systems. The migration of Computational Fluid Dynamics (CFD) from the purview of the specialist into a routine analysis tool is crucial to achieve these reductions and forms the focus of this research. Two significant challenges were identified: reducing the time-scale for creating and solving a CFD prediction and reducing the level of expertise required to perform a prediction. The commercial pressure for the rapid production of CFD predictions, coupled with the desire to reduce the risk associated with adopting a new technology led, following a review of available techniques, to the identification of structured grids as the current optimum methodology. It was decided that the task of geometry definition would be entirely performed within commercial Computer Aided Design (CAD) systems. A critical success factor for this research was the adoption of solid models for the geometry representation. Solids ensure consistency, and accuracy, whilst eliminating the need for the designer to undertake difficult, and time consuming, geometry repair operations. The versatility of parametric CAD systems were investigated on the complex geometry of a combustion system and found to be useful in reducing the overhead in altering the geometry for a CFD prediction. Accurate and robust transfer between CAD and CFD systems was achieved by the use of direct translators. Restricting the geometry definition to solid models allowed a novel two stage grid generator to be developed. In stage one an initial algebraic grid is created. This reduces user interaction to a minimum, by the employment of a series of logical rules based on the solid model to fill in any missing grid boundary condition data. In stage two the quality of the grid is improved by redistributing nodes using elliptical partial differential equations. A unique approach of improving grid quality by simultaneously smoothing both internal and surface grids was implemented. The smoothing operation was responsible for quality, and therefore reduced the level of grid generation expertise required. The successful validation of this research was demonstrated using several test cases including a CFD prediction of a complete combustion system.