A scalable domain decomposition method for simulation of 3D complex flows

Numerical simulation of 3D flows around complex objects at high Reynolds number is challenging. In order to obtain accurate simulations, very fine meshes are necessary, and such simulations are increasingly important for modern engineering practices, such as the aerodynamic design of high speed trains, cars and wind turbines, which can reduce the need for building physical models and wind-tunnel experiments. The computation for this kind of simulations is very challenging because of the complex geometry, the moving boundary, and the high Reynolds number; thus the use of massively parallel computers and scalable parallel algorithms has become indispensable. In this talk, we introduce a domain decomposition method based fully implicit algorithm for the 3D unsteady Navier-Stokes equations discretized with FEM/FVM on very fine unstructured meshes. We test the algorithm for some 3D complex unsteady flows, including flows passing a high speed train, a car, a wind turbine, and so on with realistic geometries, realistic Reynolds numbers, and realistic wind speed, on a supercomputer with thousands of processors.