%0 Journal Article
%T A Parallel Adaptive Newton-Krylov-Schwarz Method for 3D Compressible Inviscid Flow Simulations
%A Marzio Sala
%A P¨¦n¨¦lope Leyland
%A Angelo Casagrande
%J Modelling and Simulation in Engineering
%D 2013
%I Hindawi Publishing Corporation
%R 10.1155/2013/694354
%X A parallel adaptive pseudo transient Newton-Krylov-Schwarz ( NKS) method for the solution of compressible flows is presented. Multidimensional upwind residual distribution schemes are used for space discretisation, while an implicit time-marching scheme is employed for the discretisation of the (pseudo)time derivative. The linear system arising from the Newton method applied to the resulting nonlinear system is solved by the means of Krylov iterations with Schwarz-type preconditioners. A scalable and efficient data structure for the NKS procedure is presented. The main computational kernels are considered, and an extensive analysis is reported to compare the Krylov accelerators, the preconditioning techniques. Results, obtained on a distributed memory computer, are presented for 2D and 3D problems of aeronautical interest on unstructured grids. 1. Introduction The aim of this paper is to provide an overview of the methods required for an efficient parallel solution of the compressible Euler equations (CEE) on unstructured 2D and 3D grids. The ingredients include space and time discretisation schemes for the underlying partial differential equations (PDEs), a nonlinear solver based on the Newton¡¯s method, and a parallel Krylov accelerator with domain decomposition preconditioners. Nowadays, unstructured grids are of particular interest for industrial applications. With respect to structured grids, it is often easier to produce unstructured grids of good quality in domains of complex shape, especially if the unstructured grid generator can be coupled to a CAD system. This can provide¡ªat least in principle¡ªa fast process to solve the problem at hand, with minimal intervention of the user, once the geometry of the domain and the boundary conditions have been specified. However, such techniques imply an almost perfect geometric representation (CAD level) and automatic grid generation remains a challenge in computational engineering, especially for moving boundaries. The framework presented here can be successfully applied to the solution of sets of PDEs problems discretised on unstructured grids. The focus in this paper is restricted to the solution of the CEE. In particular, the space discretisation technique considered is based on the so-called multidimensional upwind residual distribution (MURD) schemes; see for instance [1¨C5]. These schemes are supposed to render higher accuracy and less spurious numerical anomalies. This is due to the fact that they take into account the multidirectionality of the wave structure, rather than being based upon a summation
%U http://www.hindawi.com/journals/mse/2013/694354/