An Optimised Recoupling Strategy for the Parallel Computation of Turbomachinery Rows with Domain Decomposition

摘要

The parallel simulation of two relevant classes of turbomachinery flow is presented. The parallel algorithm adopts a simple domain decomposition, which is particularly tailored to flow in turbomachines. The loss in implicitness brought by the decomposition is compensated by a sub-iterative procedure, which has been optimised to reduce the number of data exchanges and the time spent by MPI calls. The code has been applied to the simulation of an axial turbine stator and a centrifugal impeller. With 16 processors, speed-up factors of up to 14.7 for the stator and 13.2 for the impeller have been achieved at fixed residual level.