Appendix C

Edge-based finite element formulation

The edge-based data structure has been used in many recent finite element formulations for flow problems. As mentioned in Sec. 6.8, Chapter 6, this formulation has many advantages such as smaller storage, etc. To explain the formulation we shall consider the Euler equations and a few assembled linear triangular elements on a two-dimensional finite element mesh as shown in Fig. C. 1. From Eq. (1.24) we rewrite the following Euler equations

where @ are the conservative variables. If the element-based formulation for the above equation omits the stabilization terms, the weak form can be written as A@ In a fully explicit form of solution procedure, the left-hand side becomes M(A@/At) and here M is the consistent mass matrix (see Chapter 3). We can write the RHS of the above equation for an interior node I (Fig. C.l(a)) by interpolating F; in each element and after applying Green's theorem as

where AE is the area and I, J and K are the three nodes of the element (triangle) E. This is an acceptable added approximation which is frequently used in the TaylorGalerkin method (see Chapter 2). In another form, the above RHS can be written as (Fig. C.l(a))
-~

aN~ (F, 3 ax;

A2 dN1 A3 aNI + F, + F ~ +) -(F, + F~ + F,) + (F, + F, + F ~ ) (c.4) 3 ax; 3 axj

~

where A I , A2 and A3 are the areas of elements 1, 2 and 3 respectively. For integration over the boundary on the RHS, we can write the following in the element formulation
BI

jr N ' ( N k F f ) d r n B
B

=
BE1

[%(2F: + Fi).]

Appendix C

299

Fig. C.l Typical patch of linear triangular elements: (a) inside node; (b) boundary node.

where n is the boundary normal. The above equation can be rewritten for the node I in Fig. C. 1(b) as

(2Fj + F?)n, + r B 2 (2Fj + Ff)n2 (C.6) 6 6 where r B ] and r B 2 are appropriate edge lengths. The above equations (C.3) and (C.5) can be reformulated for an edge-based data structure. In such a procedure, Eq. (C.3) can be rewritten as (for an interior node I)
rBI
~ ~

C J n Ed " +NI'Fm=
EI

2{C [4 + }
A E 'lvl

(F!

Fis)

(C.7)

S=l

EEII,

where mSis the number of edges in the mesh which are directly connected to the node I and the summation C E E I l , extends over those elements that contain the edges I [ T . The user can readily verify that the above equation is identically equal to the standard element formulation of Eq. (C.4) if we consider the node I in Fig. C.l(a). The inclusion of boundary sides is direct from Eqs (C.5) and (C.6).