|
Organizers |
Parallel Computation of Turbomachninery Cascade Flows
by
Ismail H. Tuncer
Department of Aeronautical Engineering, METU
In the overset grid system, grids around each blade are overlaid onto a rectangular background grid, which covers the multi-passage flow domain. Blade grids are free to move with respect to the background grid as an out-of-phase unsteady motion is imposed on the blade. At the inter-grid boundaries, the flow variables are interpolated by first localizing an inter-grid boundary point in a triangular stencil on the donor grid. The main advantage of this approach lies in its versatility to resolve the multi-passage computational domain with high quality, orthogonal subgridsi, and it lends itself readily for a coarse domain decomposition.
Computations are performed in parallel by a coarse domain decomposition. First, the background grid is equally partitioned for each blade. Each blade and the partitioned background grid may then be assigned to a separate processor in a dual processor PC cluster operating on linux. pvm library routines are employed for parallel communication. The unsteady flow in each subgrid is computed using an approximately factored, implicit Euler/Navier-Stokes solver. Osher's third-order-accurate flux upwind scheme is implemented in the discretization of convection terms. Turbulence is modeled by the Baldwin-Lomax model only in blade grids.
Unsteady flows are computed as the blades undergo 180 and 90 degree out-of-phase plunging oscillations. In additions, influence wave modes for a single plunging blade is computed in five and nine passage domains. The numerical predictions are compared with the available experimental data and parallel efficiency of the computations are addressed.
Date received: March 12, 2001
Copyright © 2001 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cagk-88.