Civil Engineering Reference
In-Depth Information
is a so-called reduction operator , i.e., a linear mapping defined on the finite element
space for the shear terms which does not affect the elements in h .
If possible, the finite element calculations are performed using the displace-
ment model (6.13), since it leads to systems of equations with positive definite
matrices and with fewer unknowns. On the other hand, for the convergence anal-
ysis, it is still best to use the mixed formulations. However, there is a problem: In
general, the functions in h cannot be represented in the form
γ h =
grad r h +
curl p h ,
where r h and p h again belong to finite element spaces. Thus, except for a special
case treated by Arnold and Falk [1989], a modification of the Helmholtz decom-
position is necessary.
The notation for the following finite element spaces is suggested by the vari-
ables in (6.11).
6.3 The Axioms of Brezzi, Bathe, and Fortin [1989]. Suppose the spaces
W h H 0 (),
h H 0 () 2 ,Q h L 2 ()/ R , h H 0 ( rot ,)
and the mapping R h defined in (6.14) have the following properties:
(P 1 ) W h h , i.e., the discrete shear term γ h :
= t 2 ( w h R h θ h ) lies in h .
(P 2 ) rot h Q h - this requirement is consistent with γ h H( rot ,) , and thus
with rot γ h L 2 () .
(P 3 ) The pair ( h ,Q h ) satisfies the inf-sup condition
( rot ψ h ,q h )
ψ h 1 q h 0 =
inf
q h Q h
sup
ψ h h
: β> 0 ,
where β is independent of h .
The spaces are thus suitable for the Stokes
problem.
(P 4 ) Let P h be the L 2 -projector onto Q h . Then
for all η H 0 () 2 ,
rot R h η = P h rot η
i.e., the following diagram is commutative:
rot
−−→ L 2 ()
H 0 () 2
R h 4
4
P h
rot
−−→ Q h .
h
Search WWH ::




Custom Search