This preview shows page 1. Sign up to view the full content.
Unformatted text preview: ent e. We have divided the vector
ce of parameters into its contributions ce j , j = 1 2 ::: n, from the shape functions of
element e. Thus, we may write c = c 1 c 2 ::: c ]:
T
e T
e T
e T
e np (5.6.4b) In this form, we may write U0 as U0 = N c = c N
T e T
e (5.6.4c) 5.5. Assembly of Vector Systems 39 where N is the npm m matrix N = N1 I N2I ::: N I]:
T (5.6.4d) np and the identity matrices have the dimension m of the partial di erential system. The
simple linear shape functions will illustrate the formulation.
Example 5.6.2. Consider the solution of a system of m = 2 equations using a
piecewiselinear nite element basis on triangles. Suppose, for convenience, that the
node numbers of element e are 1, 2, and 3. In order to simplify the notation, we suppress
the subscript 0 on U0 and V0 and the subscript e on ce. The linear approximation on
element e then takes the form U1 = c11 N (
U2
c21 1
where N1 ( )=1; ) + c12 N2(
c22 N2( ; ) + c13 N3(
c23 N3 ( )= ) )= : The rst subscript on cij denotes its index in c and the second subscript identi es the
vertex of element e. The expression (5.6.4c) takes the form
2 U1 = N1 0 N2 0 N3 0
U2
0 N1 0 N2 0 N3 3 c11
6 c21 7
6
7
6 c12 7
6
7
6 c22 7 :
6
7
4 c13 5
c23 Substituting (5.6.4c) and a similar expression for V0 into (5.6.3a,e) yields Ae(V U) = dT (Ke + Me)ce
e
where K= ZZ e (V f )e = dT fe
e (5.6.5) N G1 N + N G2 N + N G2 N + N G3eN ] det(J ) d
e T e T e T T e d (5.6.6a) 0 M= ZZ e NQN det(J )d d f =
T e ZZ e 0 e 0 Problems Nf det(J )d d : (5.6.6b) 40 Mesh Generation and Assembly
1. It is, of course, possible to use di erent shape functions for di erent solution components. This is often done with incompressible ows where the pressure is approximated by a basis having one degree less than that used for velocity. Variational
formulations with di erent elds are called mixed variational principles. The resulting nite element formulations are called mixed methods. As an example, consider
a vector problem having two components. Suppose that a piecewiselinear basis is
used for the rst variable and piecewise quadratics are used for the second. Using...
View Full
Document
 Spring '14
 JosephE.Flaherty

Click to edit the document details