RECENT PROGRESS IN INCOMPRESSIBLE REYNOLDSAVERAGED NAVIER-STOKES SOLVERS

ＲＥＣＥＮＴＰＲＯＧＲＥＳＳＩＮＩＮＣＯＭＰＲＥＳＳＩＢＬＥＲＥＹＮＯＬＤＳＡＶＥＲＡＧＥＤＮＡＶＩＥＲ－ＳＴＯＫＥＳＳＯＬＶＥＲＳ￥Ｃ．Ｈ．Ｓｕｎｇ；Ｔ．Ｔ．Ｈｕａｎｇ（ＤａｖｉｄＴａｙｌｏｒＭｏｄｅｌＢａｓｉｎ，ＣＤ／ＮＳＷＣＢｅｔｈｅｓｄａ，Ｍ...

RECENT PROGRESS IN INCOMPRESSIBLE REYNOLDSAVERAGED NAVIER-STOKES SOLVERS

A new numerical approach based on a multiblock, multigrid, local refinement method has been developed. The multiblock structure makes grid generation for complex geometries easier,multigrid techniques significantly accelerate the rate of convergence and the local refinement method provides high spatial resolution of boundary lay6r and separated vortical flows with much reduced computer memory and CPU time. Sample three-dimensional flow computations for complex geometries are presented to illustrate the advantages in term of memory savings and reductions of CPU time in obtaining high spatial resolution solutions. High spatial resolution in the inner viscous layer is achieved by using the values of y for the first grid centers rom the wall in the order of 1 to 7. The required solution resolution in the dominantly viscous flow region (boundary layer, vortex core, or three-dimensional separated zone) dictates the acceptable spacing of grid cells in that region. It is very desirable to be able to reduce grid spacing locally in the critical region without changing the overall grid. The local refinement technique doubles grid numbers in all three directions for every level of refinement. Multiple levels of local grid refinement with moderate grid stretching can be used to provide fine grid spacing in rapidly changing flow regions such as near the wall, vortex core and separated shear layer. Thus, the method provides the required fine spatial resolution in the dominantly viscous region which is relatively small compared to the entire computational domain. This also avoids the used of excessively large aspect ratios of the grid cells near the rapidly changing now regions thus reduces truncation error and improves robustness. Furthermore, the results of numerical experimentation indicate that the present multigrid local refinement technique provides very effective numerical communications across the fine and coarse grid interfaces because the finer mesh is embedded entirely in the coarse mesh, and the occurrence of numerical instability in the interface is reduced.Other recent progress in the application of the preconditioning method to improve the rate of convergence and the modification of the Baldwin-Lomax turbulence model for accurate prediction of the cross-now separation problems is also discussed, Numerical examples are presented to illustrate that these advances have significantly improved our computational capability. Some future research trends are also indicated.