A local block processing strategy for multiblock flow computations

Publisher Summary The numerical solution of complex flow fields on high-performance parallel architectures is widely performed by using multiblock structured methods. The success of this technique is because of the considerably simplified grid-generation task and because of the intrinsic parallel nature of the multiblock approach that enables satisfactory speedup and efficiency. It is interesting to note that, by this extra-cell states computation, during the multistage time stepping, the flow states are frozen at the initial stage and the time integration is modified with respect to a single block computation. This local modification could be avoided by recomputing extra-cell states at each stage of the Runge–Kutta scheme obtaining a standard subdomain calculation. The method can significantly reduce the computational work necessary to perform steady-state calculations of inviscid sub/transonic flows when an explicit multistage time-stepping integration is used. The number of relaxations is significantly reduced especially for the blocks far from the airfoil. The proposed test cases showed that no loss in accuracy is obtained, included the case of shock crossing the interface.