An immersed boundary parallel RANS Solver for turbomachinery flows

This paper describes some recent progress in developing a parallel immersed boundary method for compressible viscous flows in complex turbomachinery configurations. Using the immersed boundary technique, the grid generation process is greatly simplified and simple Cartesian grids can be generated around highly complex bodies in a very short time. The accuracy and efficiency of the method are enhanced by using wall functions as well as a flexible local grid refinement technique, which coarsens a uniformly fine mesh far from solid walls and other high-gradient-flow regions. The numerical method is combined with a domain-decomposition strategy which provides partitioned grid-blocks balancing the load among the processors, so as to minimize the computational cost on parallel-architecture computers. In the present paper, the domaindecomposition process is validated and the numerical method is employed to compute two- and three-dimensional turbinecascade flows. In particular, as a very severe test case, the geometry of a cooled turbine blade is considered, and the proposed method is used to compute the flow field around the blade as well as the cooling-air flow inside of it. The method is shown to be accurate and very flexible. On the other hand, the data-structure of the code is to be modified in order to minimize both the memory requirement and data exchange between processors so as to fully exploit a parallel architecture, thus leading to a very efficient methodology.