This was the rst di erence scheme to be based on the

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: me to be based on the solution of a Riemann problem. This early work and a subsequent work of Glimm 17] and Chorin 9] stimulated a great deal of interest in using Riemann problems to construct numerical ux functions. A summary of a large number of choices appears in Cockburn and Shu 12]. 10.3 Multidimensional Discontinuous Galerkin Methods Let us extend the discontinuous Galerkin method to multidimensional conservation laws of the form ut + r f (u)x = b(x y z t u) (x y z) 2 t>0 (10.3.1a) where f (u) = f (u) g(u) h(u)] (10.3.1b) 10.3. Multidimensional Discontinuous Galerkin Methods and r f (u) = f (u)x + g(u)y + h(u)z : 33 (10.3.1c) The solution u(x y z t) componenets of the ux vector f (u), g(u), and h(u) and the loading b(x y z t u) are m-vectors and is a bounde region of <3. Boundary conditions must be prescribed on @ along characteristics that enter the region. We'll see what this means by example. Initial condtions prescribe (x y z) 2 u(x y z 0) = 0 @: (10.3.1d) Following our analysis of Section 10.2, we partition into a set of nite elements j , j = 1 2 : : : N , and construct a weak form of the problem on an element. This is done, as usual, by multiplying (10.3.1a) by a test function v 2 L2( j ), integrating over j , and applying the divergence theorem to the ux to obtain (v ut)j + < v f n >j ;(rv f )j = (v b)j 8v 2 L2( j ) (10.3.2a) where Z (v u)j = vT udxdydz (10.3.2b) j (rv f )j = Z T T vx f (u) + vy g(u) + vzT h(u)]dxdydz (10.3.2c) j f n = f n = f (u)n1 + g(u)n2 + h(u)n3 and < v f n >j = Z @ vT f ndS: (10.3.2d) (10.3.2e) j The vector n = n1 n2 n3 ]T is the unit outward normal vector to @ and dS is a surface in nitessimal on @ j . Only the normal component of the ux is involved in (10.3.2) hence, its approximation on @ j is the same as the one-dimensional problems of Section 10.2. Thus, the numerical normal ux function can be taken as a one-dimensional numerical ux using solution values on each side of @ j . In order to specify this more precisely,...
View Full Document

This document was uploaded on 03/16/2014 for the course CSCI 6860 at Rensselaer Polytechnic Institute.

Ask a homework question - tutors are online