Escolar Documentos
Profissional Documentos
Cultura Documentos
(2)
( ) . on I = V g A u u pI n
T T
v (3)
We also assume that O has a polygonal boundary
O c = I: , so n
) (
2
I eL g
=
) , ( ) , (
) , ( ) , (
) , (
y x b y x c
y x c y x a
y x A I e ) , ( y x all for (4)
- There are two strictly positive constants
, and
1 1
| o
such
that:
1 1
) , ( | o s s X y x A X
T
(5)
I e ) , ( y x all for and
{ }. 1 /
2
2
= e = e X IR X S X
Where are and , c b a
the function continuous defined on . I
III. THE WEAK FORMULATION
We define the following spaces:
)
`
O e
c
c
c
c
O = O ) ( ; ; / : ) (
2 1
L
y
u
x
u
u IR u h
(6)
| |
2
1 1
) ( ) ( O = O h H (7)
{ }, 0 / ) ( ) (
2 2
0
}
O
= O e = O q L q L
(8)
{ } , 0 . / ) ( ) (
1 1
0 ,
in n v H v H
n
= O e = O
(9)
{ } . in 0 . . / ) ( ) (
1
0 ,
1
0 ,
O = V O e = O v H v V
n n
(10)
The standard weak formulation of the Navier-Stokes flow
problem (1) - (2)-(3) is the following:
Find
) (
1
O eH u
and
) (
2
O eL p
such that
( )
= V -
+ =
V + V + V V -
}
} }
} } } }
O
I O
O O I O
, 0 .
,
. :
ds u q
v g v f
v p v A u v u u v u
T
v
(11)
( ) . ) ( ) ( ,
2
0
1
0 ,
O O e L H q v all for
n
Let the bilinear forms
IR L H B IR H H A
n n n
O O O O ) ( ) ( : ; ) ( ) ( :
2
0
1
0 ,
1
0 ,
1
0 ,
. ) ( ) ( :
2
0
2
0
IR L L d O O
: ) , ( v A u v u v u A
T
} }
O I
+ V V =v
(12)
. ) , (
}
O
V = u q q u B
(13)
. ) , (
}
O
= q p q p d
(14)
And the tri-linear forms
IR H H H D IR H H H C
n n n n n n
1
0 ,
1
0 ,
1
0 ,
1
0 ,
1
0 ,
1
0 ,
: ; :
) ( ) , , (
}
O
V = z v u z v u C
(15)
) , , ( ) , ( ) , , ( z v u C v u A z v u D
+ = (16)
Given the functional
IR L O) ( : L
2
0
} }
I O
+ = v f v g v L
. . ) (
(17)
The underlying weak formulation (11) may be restated as:
( ) such that ) ( ) ( ,
2
0
1
0 ,
O O e L H p u find
n
0 ) , (
) ( ) , ( ) , , ( ) , (
=
= + +
q u B
v L q v B v u u C v u A
(18)
( ) . ) ( ) ( ,
2
0
1
0 ,
O O e L H q v all for
n
In the sequel we can assume that . 0
= g
IV. THE EXISTENCE AND UNIQUENESS OF THE SOLUTION
In this section we will study the existence and uniqueness
of the solution of problem (18), for that we need the following
results.
Theorem 4.1. There are two strictly positive constants
1
c and
2
c such that:
O O O
s s
, 1
2
, , 1
1
c v c v v
J
) (
1
0 ,
O e
n
H v all for
(19)
with
( )
:
2
1
, } }
O I
O
+ V V = v A v v v v
T
J
v (20)
( )
2
1
2
, 0
2
, 1 , 1
O O O
+ = v v v
(21)
Proof. 1) The mapping ) ( ) ( :
2 1
0
I O L H is
continuous
(See [6] theorem 1, 2), then there exists : such that 0 c
). ( all for
1
, 1 , 0
O e s
O I
H v v c v
Using (5) gives,
,
2
, 0
1
2
, 0
1
I
I
I
s s
}
v v A v v
T
| o
(22)
then
), ( all for
1
1,
2
,
O e s
O O
H v v c v
J
( )
. with
2
1
2
1 2
v | + = c c
On the other hand. According to 5.55 in [1], there exists a
constant 0 such that ). (
2
, 0
2
, 0
2
, 0 I O O
+ V s v v v
Using (22), gives
), ( all for c
1
J, , 1
1
O e s
O O
H v v v
{ }. ; max and
1
with
1
2
1
1
1
v o
v vo
=
|
|
.
|
\
|
+ =
C
C
c
). ( all for c Finally,
1
, 1
2
, , 1
1
O e s s
O O O
H v v c v v
J
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
147 | P a g e
www.ijacsa.thesai.org
This result allows us to prove that ) . ), ( (
,
1
0 ,
O
O
J
n
H is a
Hilbert space which is obliged condition for to obtain the
existence and uniqueness of the solution.
Theorem 4.2.
space Hilbert real a is ) . ), ( (
,
1
0 ,
O
O
J
n
H
.
Proof.
) . ), ( (
, 1
1
O
O H
is a real space and
) (
1
0 ,
O
n
H
is
closed in ) (
1
O H and
.
, 1 O
and
.
,O J
are equivalent
norms, then
) . ), ( (
,
1
0 ,
O
O
J
n
H
is a real Hilbert space for two
norms.
Theorem 4.3
) 1
O O
s
, ,
) , (
J J
v u v u A
(23)
( ) ) ( ) ( ,
1
0 ,
1
0 ,
O O e
n n
H H v u all for
O
O
,
1
0 ,
. norm for the elliptic - ) ( is 2)
J
n
H A
and
) , (
2
J,O
= v v v A
(24)
). (
1
0 ,
O e
n
H v all for
.
Proof: it is easy.
Theorem 4.4
1)
( ) ,
2
,
, , 0 O O
s
J
v q q v B
v
(25)
( ) ) ( ) ( , all for
2
0
1
0 ,
O O e L H q v
n
2) The bilinear form b is satisfies the inf-sup: There exists a
constant
0 |
such that
) (
) , (
sup
2
0
, 0
,
) (
1
0 ,
O e >
O
O
O e
L q all for q
v
q v B
J
H v
n
|
(26)
Proof.
1) Let
( ) ), ( ) ( ,
2
0
1
0 ,
O O e L H q v
n
we have
( )
.
2
2
. ,
, , 0
, 0 , 0
, 0 , 0
O O
O O
O O
s
V s
V s
J
v q
v q
v q q v B
v
), ( Let 2)
2
0
O eL q we have
), [6] (see '
) , (
sup
, 0
, 1
) (
1
0
O
O
O e
> q
v
q v B
H v
|
{ }
. '
) , (
sup
) , (
sup
) , (
sup
) ( all for and
) ( 0 / ) ( ) (
since
, 0
, 1
) (
, 1
) (
, 1
) (
1
0
, 1 , 1
1
0 ,
1 1
0
1
0
1
0
1
0 ,
O
O
O e
O
O e
O
O e
O O
>
=
>
O e =
O c I = O e = O
q
v
q v B
v
q v B
v
q v B
H v v v
H in v H v H
H v
H v H v
n
n
|
Using (19) gives
.
c
'
with ,
) , (
sup
2
, 0
,
) (
1
0 ,
|
| | = >
O
O
O e
q
v
q v B
J
H v
n
Theorem 4.5
1) There exists a constant such that 0 m
( ) (27) , ,
, , , O O O
s
J J J
z v u m z v u C
( )
1
0 ,
1
0 ,
1
0 ,
, ,
n n n
H H H z v u all for e
.
( ) ( ) (28) , , , , ) 2 v z u C z v u C
=
( ) . , ,
1
0 ,
1
0 ,
1
0 , n n n
V V V z v u all for e
3) ( ) (29) 0 , , = u u v C
( ) . ,
1
0 ,
1
0 , n n
V V v u all for e
4)
( ) ( ) (30) , , ,
2
J,O
= = v v v A v v w D
imply ) (as ) ( in weakly ) 5
0 ,
O m V u u
n m
(31) ) , , ( ) , , ( that v u u D v u u D
m m
Proof
have we ), ( , , z Let 1)
1
0 ,
O e
n
H v u
( )
'
[6]) in (see ' , ,
, , , 3
1
, 1 , 1 , 1
O O O
O O O
s
s
J J J
z v u
c
m
z v u m z v u C
have we ), ( , , z Let 2)
1
0 ,
O e
n
V v u
( ) ( )
}
}
O
O
V =
V + V = +
) . ( .
) . . .( , , , ,
v u z
u v v u z u v z C v u z C
By Green formula, we have
( ) ( )
( ) ( ) . , , , , finally
, 0 and 0 . then ) ( Since
) . .( ) . )( . ( , , , ,
0 ,
u v z C v u z C
z div n z V z
v u z div v u n z u v z C v u z C
n
=
= = O e
= +
} }
O O c
3) Its easy, just take (28). in u v
=
4) It suffices to apply (29).
5) The same proof of V.Girault and P.A. Raviart in [6] page
115.
According the theorems 1.2 and 1.4, chapter IV in [6], the
results (18)-(30) ensure the existence at least one pair
( ) ) ( ) ( ,
2
0
1
0 ,
O O e L H p u
n
satisfies (18).
We define
( )
O O O
e
=
, , ,
, ,
, ,
sup
1
0 ,
J J J
H z v u
z v u
z v u C
N
n
(32)
.
sup
,
0 ,
O
O
e
-
}
=
J
V v
v
v f
f
n
(33)
Then a well-know (sufficient) condition for uniqueness is
that forcing function is small in the sense that
1
N
f s
-
(it
suffices to apply theorems 1.3 and 1.4 chapter IV in [6]).
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
148 | P a g e
www.ijacsa.thesai.org
Theorem 4.6. Assume that and
) (
2
O eL f
satisfy the
following condition
(34) ) ( all for .
1
0 ,
,
O e s
O
O
}
n
J
H v v
N
v f
o
[0,1[. number fixed some For e o
Then there exists an unique
( ) ) ( ) ( ,
2
0
1
0 ,
O O e L H p u
n
(35)
Proof. The some proof of theorem 2.4 chapter IV in [6].
V. MIXED FINITE ELEMENT APPROXIMATION
In this section we assume that c b a f and , ,
are the
polynomials.
Let
, 0 ; h T
h
be a family of rectangulations of O. For
any
,
h
T T e
T
e
is of rectangles sharing at least one edge
with element T,
T
e
~
is the set of rectangles sharing at least one
vertex with T. Also, for an element edge E,
E
e denotes the
union of rectangles sharing E, while
E
e
~
is the set of rectangles
sharing at least one vertex whit E.
Next, T c is the set of the four edges of T we denote by
) (T c
and
T
N the set of its edges and vertices, respectively.
We let ) (T
h
T T h
c c
e
= denotes the set of all edges
split into interior and boundary edges.
I O
=
, , h h h
c c c
Where
{ } O c e =
O
E E
h h
:
,
c c
{ } O c c e =
I
E E
h h
:
,
c c
We denote by
T
h the diameter of a simplex, by
T
h the
diameter of a face E of T, and we set { }
T T T
h h
h
e
= max
.
A discrete weak formulation is defined using finite
dimensional spaces ) ( X
1
0 ,
1
h
O c
n
H and
) ( M
2
0
h
O c L
The discrete version of (15) is:
( ) ( ) ( )
( )
0 ,
) ( , , , ,
: such that and
1
=
= + +
e e
h h
h h h h h h h h
h
h h h
p v B
v L p v B v u u C v u A
M p X u find
(36)
For all . and
1 h
h h h
M q X u e e
We define the appropriate bases for the finite element
spaces, leading to non linear system of algebraic equations.
Linearization of this system using Newton iteration gives the
finite dimensional System:
: such that and
1 h
h h h
M p X u find e e o o
( )
( ) ( )
( )
(37)
) ( ,
) ( , , ,
, , :
=
= + +
+ + V V
} }
O O c
h k h h
h k h h h h h
h h h h
T
h h h
q r q u B
v R p v B v u u C
v u u C v A u v u
o
o o
o o o v
For all . and
1 h
h h h
M q X u e e
Here, ) (
h k
v R
and ) (
h k
q r are the non linear residuals
associated with the discrete formulations (36). To define the
corresponding linear algebra problem, we use a set of vector-
valued basis functions { }
u
n i i ,..., 1 =
So that
= =
A = =
u u
n
j
j j h
n
j
j j h
u u u u
1 1
. ; o
(38)
We introduce a set of pressure basis functions { }
p
n k k ,..., 1 =
and set
. ;
1 1
= =
A = =
p p
n
k
k k h
n
k
k k h
p p p p o
(39)
Where
u
n and
p
n are the numbers of velocity and pressure
basis functions, respectively.
We find that the discrete formulation (37) can be expressed as
a system of linear equations
.
0 0
0
0 0
|
|
.
|
\
|
=
|
|
.
|
\
|
A
A
|
|
.
|
\
|
+ + f
P
U
B
B W N A
T
(40)
The system is referred to as the discrete Newton problem.
The matrix
0
A is the vector Laplacian matrix and
0
B is the
divergence matrix
(41) : ; ] [
, , 0
v
A a a A
T
i j i j i j i
} }
O I
+ V V = =
. ; ] [
, , 0
}
O
V = =
j k j k j k
b b B
(42)
The vector-convection matrix N and the Newton derivative
matrix W are given by
) . ( ; ] [
, ,
}
O
V = =
j i h j i j i
u n n N
(43)
) . ( ; ] [
, , }
O
V = =
j h i j i j i
u w w W
(44)
For
. ,..., 1 and ,..., 1 ;
p u
n k n j i = =
The right-hand side vectors in (40) are
, . ; ] [
i i i i
g f f f f
} }
O I
+ = =
(45)
for
, ,..., 1
u
n i =
For Picard iteration, we give the discrete problem
.
0 0
0
0 0
|
|
.
|
\
|
=
|
|
.
|
\
|
A
A
|
|
.
|
\
|
+ f
P
U
B
B N A
T
(46)
VI. A POSTERIORI ERROR ESTIMATOR
In this section we propose two types of a posteriori error
indicator, a residual error estimator and local Poisson problem
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
149 | P a g e
www.ijacsa.thesai.org
estimator, which are shown to give global error estimates that
are equivalent to the true error.
A. A Residual Error Estimator
The bubble functions on the reference element
) 1 , 0 ( ) 1 , 0 (
~
= T
are defined as follows:
) 1 ( ) 1 ( 2
4
~ y y x x b
T
=
) 1 )( 1 ( 2
2
~
,
~
1
y x x b
T E
=
x y y b
T E
) 1 ( 2
2
~
,
~
2
=
x x y b
T E
) 1 ( 2
2
~
,
~
3
=
) 1 )( 1 ( 2
2
~
,
~
4
x y y b
T E
=
Here
T
b~
is the reference element bubble function, and
T E
i
b ~
,
~ , 4 : 1 = i are reference edge bubble functions. For any
,
h
T T e
the element bubble functions is
T
T
T
F b b ~ = and
the element edge bubble function is , ~
,
~
, T
T E
T E
F b b
i
i
= where
T
F
the affine map form T. to
~
T
For an interior edge
O
e
, h
E c ,
E
b is defined piecewise, so
that
2 : 1 ,
,
= = i b b
i i
T E T E
where
2 1
T T E =
.
For a boundary edge
T E E h
b b
, ,
, E = e
I
c , where T is the
rectangle such that . E T c e
With these bubble functions, ceruse et al ([3], lemma 4.1]
established the following lemma.
Lemma 6.1. Let T be an arbitrary rectangle in and
h
T
For any
), ( and ) (
0 0
E P v T P v
k E k T
e e
the following
inequalities hold.
, 0
, 0
2
1
, 0 T
T k
T
T T
T
T k
v C b v v c
s s
(47)
, 0
1
, 1 T
T T k
T
T T
v h C b v
s
(48)
E
E k
E
E E
E
E k
v C b v v c
, 0
, 0
2
1
, 0
s s
(49)
, 0 , 0
2
1
E
E E k
T
E E
v h C b v
s
(50)
, 0 , 1
2
1
E
E E k
T
E E
v h C b v
s
(51)
Where
k
c and
k
C
are tow constants which only depend
on the element aspect ratio and the polynomial degrees
0
k
and
1
k .
Here,
0
k and
1
k are fixed and
k
c and
k
C can be
associated with generic constants c and C In addition,
E
v
which is only defined on the edge E also denotes its natural
extension to the element T.
From the inequalities (50) and (51), we established the
following lemma:
Lemma 6.2. Let T be a rectangle and
.
,I
c e
h
T E c
For any ), (
0
E P v
k E
e
=
E E
b v in the other three edges of rectangle
T, it can be extended to the whole of by setting 0
=
E E
b v
in
, T O
then
and
, 1 , 1 O
=
E E
T
E E
b v b v
O
=
, , J
E E
T J
E E
b v b v
Using the inequalities (19), (50) and (51) gives
O
=
, , J
E E
T J
E E
b v b v
O
s
, 1
2 E E
b v c
T
E E
b v c
, 1
2
=
( )
2
1
2
, 1
2
, 0
2
T
E E
T
E E
b v b v c
+ =
( )
E
E E E k
v h h C c
, 0
2
1
1
2
+ s
( )
E
E E k
v h D C c
, 0
2
1
2
1
2
2
1
+ s
E
E E
v Ch
, 0
2
1
s
With D is the diameter of and
( ) . 1
2
1
2
2
+ = D C c C
k
We recall some quasi-interpolation estimates in the
following lemma.
Lemma 6.3. Clement interpolation estimate: Given
), ( v
1
O eH
let
1
h
X e
h
v
be the quasi-interpolant of v
defined
by averaging as in [4].
For any
,
h
T T e
,
~
, 1 , 0
T
v Ch v v
T
T
h
e
s
(53)
and for all , T E c e
~
, 1
2
1
, 0
E
v Ch v v
E
E
h
e
s
(54)
We let ( ) p u,
denote the solution of (18) and let denote
( )
h h
p u ,
the solution of (36) with an approximation on a
rectangular subdivision .
h
T
Our aim is to estimate the velocity and the pressure errors
). ( and ) (
2
0
1
0 ,
O e = O e = L p p H u u e
h n h
c
The element contribution ,
,T R
q of the residual error estimator
R
q is given by
2
, 0
2
, 0
2
, 0
2 2
,
E
E
T E
E
T
T
T
T T T R
R h R R h
c e
+ + = q
(55)
and the components in (55) are given by
{ } T p u u u f R
h h h h T
/
2
V V V + =
v
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
150 | P a g e
www.ijacsa.thesai.org
{ } T u R
h T
/ .
V =
| |
( ) ( )
;
;
2
1
, ,
,
e V +
e V
=
I
O
h T E h h
T
h
h h h
E
E n I p u A u g
E I p u
R
c v
c v
With the key contribution coming from the stress jump
associated with an edge E adjoining elements T and S:
| | ( ) ( ) S I p u T I p u I p u
h h h h h h
/ / V V = V
v v v
The global residual error estimator is given by:
e
=
h
T T
T R R
2
,
q q
Our aim is bound
h
X
h
p p u u and
with respect to
the norm
J
. for velocity
J, X
v v
=
and the quotient
norm for the pressure
.
0,
p p
M
=
For any
,
h
T T e
and , T E c e we define the following two
functions:
;
T T T
b R w
= ,
E E E
b R w
=
T, on 0 c = -
T
w
. on 0 then if
, E E h
w T E e c c = c e -
O
. T rectangle of edges
other in the 0 then if
,
= c e -
I E h
w T E c
. if T - in 0
. if - in 0
T - in 0
: setting by
of whole to extended be can and
,
,
I
O
c e O =
c e O =
O =
O
-
h E
h E E
T
T E
T E w
T E w
w
w w
c
c e
With these two functions we have the following lemmas:
Lemma 6.4. For any
h
T T e we have:
( ) ( ) . . : .
} } }
V + V V =
T T
T T
T
T
w u u w pI u w f
v
(56)
Proof
Using (1) gives
( )
} }
V + V + V =
T
T
T
T
w p u u u w f
2
. . .
v
By applying the Green formula and T, on 0 c =
T
w we
obtain
( )
( ) ( )
( ) ( ) . . . :
. . :
. .
} }
} }
} }
V + V V =
V + V V +
V =
c
T
T
T
T
T
T
T
T
T
T
T
T
w u u w pI u
w u u w pI u
w n pI u w f
v
v
v
Lemma 6.5
: have we if i)
,O
c e
h
T E c
( ) ( ) . . : .
} } }
V + V V =
E E E
E E E
w u u w pI u w f
e e e
v
(57)
: have We , if )
,I
c e
h
T E ii c
( )
( ) ( ) . .
: .
} }
} }
V + +
V V =
c T
T
T
E
T
T
E
T
E
w u u w g A u
w pI u w f
v
(58)
Proof.
i) The same proof of (56).
: have We , If )
,I
c e
h
T E ii c
( )
( ) ( )
( )
}
} }
} }
V +
V V V =
V + V + V =
c
T
T
T
E
T
E
T
T
T
T
w u u
w n pI u w pI u
w p u u u w f
. .
. :
. . .
2
v v
v
We have
( ) 0 and
= I c = V
E
T T
w E on g A u u pI n v
in the other tree edges of rectangle T, then
( )
( ) ( ) . . .
: .
} }
} }
V + +
V V =
c T
T
T
E
T
T
E
T
E
w u u w g A u
w pI u w f
v
We define the bilinear form
( ) ) , ( ) , ( ) , ( ) , ( ); , ( p v B q u B v u A q v p u G
+ + =
We define also the following functional
( ) ( ) ( )
). ( , , and all for
. . . . , , ,
1
0 ,
O e O c
V V =
} }
n
K K
H v y x K
v y y v x x v y x K
u
Lemma 6.6. There exists 0 and 0
0
h C such that
( )
. and ) ( , all for
, , , ,
1
0 , 0
, ,
O c O e s
s
K H v h h
v e C v u u K
n
K J K J
h
u
(59)
Proof. Using (27), (32) and (35), we have that for
) (
1
0 ,
O e
n
H v
( ) ( ) ( )
( ) ( )
(60) ) 2 (
) . 2 (
) . (
. . . .
. . . . , , ,
, , ,
,
2
, , , ,
, , , , , ,
K J K J K J
K J K J K J K J K J
K J K J
h
K J K J K J K J
K
h
K
K
h h
K
h
v e e N
v e v e u N
v u e v e u N
v u e v e u
v u u v u u v u u K
+ s
+ s
+ s
V + V =
V V =
} }
} }
o
u
We have
) ( in lim
1
0 ,
0
O =
n h
h
H u u
, then there exists 0
0
h such that
. all for 1
0
, ,
h h u u e
K J
h
K J
s s =
Using this result and (60), we obtain
( )
. 2 C with ), ( and all for
, , ,
1
0 , 0
, ,
N H v h h
v e C v u u K
n
K J K J
h
+ = O e s
s
ov
u
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
151 | P a g e
www.ijacsa.thesai.org
We have
( )
( ) ( )
( )
). , (
) , ( ) , ( ) , , ( - ) (
) , ( ); , ( ) , , ( - ) (
) , ( ); , ( - ) , ( ); , (
) , ( ) , ( ) , ( ) , ( ); , (
q u B
p v B v u A v u u C v L
q v p u G v u u C v L
q v p u G q v p u G
p p v B q u u B v u u A q v e G
h
h h
h h
h h
h h h
=
=
=
+ + = c
( ) ( )
). ( ) ( ) , ( all for
) , ( ) , ( ) , (
) ( ) , , ( , , , ) , ( ); , (
2
0
1
0 ,
O O e
+ = O +
L H q v
q u B p v B v u A
v L v u u C v u u q v e G
Then
n
h h h
h h h
u c
(61)
) ( and ) ( errors that the find We
2
0
1
0 ,
O e O e L H e
n
c
Satisfy the non-linear equation
( ) ( )
). ( ) ( ) , ( all for
) , ( ) , ( ) , (
, , , ) , ( ); , (
2
0
1
0 ,
O O e
)
`
+ =
O +
e c c
L H q v
q R v R v R
v u u q v e G
n
T T T E
T T E E T T
h
h
u c
(62)
) ( ) ( ) , ( pair a define We
2
0
1
0 ,
O O e L H
n
to be the Ritz projection of the modified residuals
( )
( ) ( )
). ( ) ( ) , (
, , , ) , ( ); , (
, , , ) , ( ) , ( ) , ( ) , ( ) , (
2
0
1
0 ,
O O e
O + =
O + + + = +
L H q v all for
v u u q v e G
v u u v B q e B v e A q d v A
n
h
h
u c
u c
(63)
Lemma 6.7
1
all for 0 ) , (
h h h
X v v A e =
(64)
Proof: we set
h
v v q
= = and 0 in (62), we obtain
( )
. 0 ) ( ) (
) , , ( ) , , (
) , ( ) , ( ) , ( ) , (
, , , ) , ( ) , ( ) , (
= =
+
+ =
O + + =
h h
h h h h
h h h h h h
h h h h h
v L v L
v u u C v u u C
p v B p v B v u A v u A
v u u v B v e A v A
u c
Next, we establish the equivalence between the norms of
) ( ) ( ) , (
2
0
1
0 ,
O O e L H e
n
c
of (63).
Theorem 6.8. Let the conditions of theorem 4.6 hold.
There exist two positive constants , and
2 1
k k independent of
h , such that
{ } { } { }
2
, 0
2
,
2
2
, 0
2
,
2
, 0
2
,
1
O O O O O O
+ s + s + c
J J J
k e k
Proof. The same proof of theorem 3 in [9].
Theorem 6.9.
have we ), ( ) ( ) , ( all For
2
0
1
0 ,
O O e L H s w
n
). (
2
1
) , ( ) , (
sup
, 0 ,
, 0 ,
) , (
2
0
1
0 ,
O O
O O
e
+ >
+
+
s w
q v
q s d v w A
J
J
L H q v
n
(65)
Proof.
have we ), ( ) ( ) , ( Let
2
0
1
0 ,
O O e L H s w
n
(66)
) , (
0
) 0 , ( ) , (
) , ( ) , (
sup
,
,
, 0 , , 0 ,
) , (
2
0
1
0 ,
O
O
O O O O
e
= =
+
+
>
+
+
J
J
J J
L H q v
w
w
w w A
v
s d w w A
q v
q s d v w A
n
(67) .
) , (
0
) , ( ) 0 , (
) , ( ) , (
sup
have also We
, 0
, 0
, 0 , , 0 ,
) , (
2
0
1
0 ,
O
O
O O O O
e
= =
+
+
>
+
+
s
s
s s d
s
s s d w A
q v
q s d v w A
J J
L H q v
n
We gather (66) and (67) to get
). (
2
1
) , ( ) , (
sup
, 0 ,
, 0 ,
) , (
2
0
1
0 ,
O O
O O
e
+ >
+
+
s w
q v
q s d v w A
J
J
L H q v
n
Theorem 6.10. For any mixed finite element
approximation (not necessarily inf-sup stable) defined on
rectangular grids
h
T , the residual estimator
R
q satisfies:
,
, 0 ,
R
J
C e q c s +
O O
and
{ } .
2
1
'
2
' , 0
2
' ,
,
|
|
.
|
\
|
+ s
e
T
T
T T J
T R
e C
e
c q
Note that the constant C in the local lower bound is
independent of the domain, and
. . :
2
, } }
c
+ V V =
T T
T
T J
e A e e e e
v
Proof. To establish the upper bound we let
1 2
0
1
0 ,
and ) ( ) ( ) , (
h h n
X v L H q v e O O e
be the clement
interpolation of . v
\
|
|
|
.
|
\
|
+
|
|
.
|
\
|
|
|
.
|
\
|
s
+
+ s
)
`
+ =
+ = +
e c e e c e
e e
c e e
e c c
h h
h h
h
h
T T T E
E
h
E T T T E
T
E E
T T
T
h
T
T T
T
T T
E
T
T
T T
E
h
T
E
T T
T
h
T
T
T T T E
T T E h E T h T
h
v v
h
R h
v v
h
R h
R q
v v R v v R
q R v v R v v R
q d v v A q d v A
Using (52) and (53), then gives
{ }
{ } .
) , ( ) , (
2
1
2
, 0
2
, 0
2
, 0
2
2
1
2
, 0
2
,
|
|
.
|
\
|
+ +
|
|
.
|
\
|
+ s +
e
c e
e
h
h
T T
E
E
T E
E
T
T
T
T T
T T
T T J
R h R R h
q v C q d v A
Finally, using (65) gives:
{ }
e
c e
+ + s +
h
T T
E
E
T E
E
T
T
T
T T
T T J
R h R R h C
2
, 0
2
, 0
2
, 0
2
, 0 ,
According to theorem 6.8, we have
{ }
e
c e
O
+ + s +
h
T T
E
E
T E
E
T
T
T
T T
T T J
R h R R h C e
2
, 0
2
, 0
2
, 0
2
, 0 ,
c
This establishes the upper bound.
Turning to the local lower bound. First, for the element
residual part, we have:
} }
} }
} }
V V +
V V =
V V V + =
c T
T h h
T
T h h
T T
T h h T
T T
T h h h h T T
w u u w n I p u
w I p u w f
w p u u u f w R
). ( . ) (
: ) ( .
). . ( .
2
v
v
v
See that T in 0 c =
T
w , using (56) and (57) gives:
( )
) e C( .
gives (48), Using
, then T, in 0 Since
' ' ) e ( C'
, , , : ) ( .
, 0
1
2
1
2
, 0
2
,
, 1 ,
, , , 1 , 0 , 1
T
T T
T T J
T
T T
T
T
T J
T
T J T J T
T
T T
T h
T T
T T T
R h w R
w w w
w e C w
w u u T w I e w R
+ s
= c =
+ + s
+ V V =
}
} }
c
v
c
u c v
In addition, from the inverse inequality (47)
, .
2
T 0,
2
, 0
2
1
T k
T
T T
T
T T
R c b R w R
> =
}
Thus, ) e (
2
, 0
2
,
2
T 0,
2
T T J
T T
C R h c + s
(68)
Next comes the divergence part,
(69)
2 2
2
) ( .
.
, ,
, 1
, 0
, 0 , 0
T J T J
h
T
h
T
h
T
h
T
T
e u u
u u
u u
u R
v v
= s
s
V =
V =
Finally, we need to estimate the jump term. For an edge
have We , if
,O
c e
h
T E c
}
}
=
c
V =
E
i
T
E h h E E
i
w n I p u w R
2 : 1
. ) ( . 2
} }
=
V V + V V =
2 : 1
2
. ). ( : ) (
i
T
E h h E h h
i E
w p u w I p u
v v
e
: gives , in 0 and (57) Using
E
e c =
E
w
( )
E E
i
i
i
E E E
i
i
E
J
E
J
i
T
E
T
T E
i
T
E h E E T
E
E
E E
w e C
w R w e
w u u w R
w I e w R
e e
e e e
e
c v
e u
c v
, ,
2 : 1
, 0
, 0
, 1 , 0 , 1
2 : 1
) (
, , ,
: ) ( . 2
+
+ + s
+
V V =
}
} }
=
=
E E
, 1 ,
E
then , in 0 Since
e e
v e
T
J
E
w w w
= c =
Using (50) and (51), to get
E
E E
i
T
T
E
E E
E
E E
R h R
R h e C w R
i
i
E E
, 0
2
1
2 : 1
, 0
, 0
2
1
2
1
2
, 0
2
, 1
) ( ' . 2
}
=
+
+ s
e e
c
Using (68), gives
) ( . 2
, 0
2
1
2
1
2
, 0
2
,
E
E E
J
E
E E
R h e C w R
E E
+ s
} e e
c (70)
Using the inverse inequality (49) and thus using (70) gives
) e C(
2
, 0
2
,
2
E 0, E E
J
E E
R h
e e
c + s
(71)
Also need to show that (70) holds for boundary edges.
have We , an For
,I
c e
h
T E c
| |
. ) ( .
} }
c
V + =
E T
E h h
T
h E E
w g n I p u A u w R
v
( ) ( )
( ) ( )
( ) . ) (
: ) ( .
. ) ( .
2
}
} }
} }
V V +
V V + =
V + =
c
c c
T
E h h
T T
E h h E
T
h
T T
E h h E
T
h
w p u
w I p u w g A u
w n I p u w g A u
v
v
v
Using (58), (59) and (4), gives
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
153 | P a g e
www.ijacsa.thesai.org
( )
( )
) (
'
) (
, , , .
. : ) ( .
, 0
, 0
, , 0 ,
, 0
, 0
, ,
, 0 , 0
1
, 1 , 0 , 1
T
E
T
T
T J
E
T T J
T
E
T
T
T J
E
T J
T
E
T T
E
T T
T
E h E T
T
E
T
E T
E E E
w R w e C
w R w e C
w e w e
w u u T w R
w A e w I e w R
+ + s
+ +
+ + s
+
V V =
c c
c
}
} } }
c
| c v
u
c v
Using (50) and (52) gives
E
E E
T
T
E
E E
T T J
E
E E
R h R
R h e C w R
, 0
2
1
, 0
, 0
2
1
2
1
2
, 0
2
,
1
) ( .
+
+ s
}
c
Using (69) gives
) ( .
, 0
2
1
2
1
2
, 0
2
,
E
E E
T T J
E
E E
R h e C w R
+ s
}
c (72)
Using (49), then
gives (71), using and
, .
2
, 0
2
, 0
2
1
E
E
E
E E
E
E E
R c b R w R
> =
}
) e C(
2
, 0
2
,
2
E 0,
T T J
E E
R h c + s
(73)
Finally, combining (68), (69), (71) and (73) establishes the
local lower bound.
B. The Local Poisson Problem Estimator.
The local Poisson problem estimator defined as:
e
=
h
T T
T P P
2
,
q q
2
, 0
,
2
,
,
2
,
T
T P
T J
T P T P
e c q + =
(74)
Let
{ } O c c = e = - T on 0 . : ) (
1
n v T H v V
T
} }
c
+ V V = -
T T
T
T P T P T P T
v A e v e v e A . : ) , (
, , ,
v
T T P
V e e
,
e
=
) (
,
) , ( ) , ( ) , (
T E
E E T T T P T
v R v R v e A
c
(75)
. any for
T
V v e
And . / .
,
T u
h T P
V = c (76)
Theorem 6.11.
T P,
q is equivalent to
T R,
q estimator:
.
, , , T P T R T P
C c q q q s s
Proof. For the upper bound, we first let
T T T
b R w
= (
T
b is
an element interior bubble).
From (75),
T
T
T
T P
T T T P T T T
w e
w e w R
, 1 , 1
,
,
) , ( ) , (
v
v
s
V V =
Using (48) we get
( ) ) , (
2
1
2
, 0
,
2
,
,
, 0
1
T
T P
T J
T P
T
T T T T T
e R Ch w R c + s
(77)
In addition, from the inverse inequalities (47),
T T T T
w R R ) , C.(
2
T 0,
s
And using (77), to get
( ).
2
, 0
,
2
,
,
2
, 0
2
T
T P
T J
T P
T
T T
e C R h c + s
(78)
Next, we let
E E E
b R w
= (
E
b is an edge bubble function).
If
,
,I
c e
h
T E c
using (75), (78), (50) and (51) give
( )
) , ( ) , ( ) , (
2
1
2
, 0
,
2
,
,
, 0
, 0
, 0
, ,
,
,
2
1
T
T P
T J
T P
E
E E
T
E
T
T
T J
E
T J
T P
T E T T E T P T E E E
e R Ch
w R w e
w R w e A w R
c + s
+ s
+ =
, if
,O
c e
h
T E c
See that the matrix A defined just in , I then we can posed
. in 0 I O = A
Using (75), (50), (51) and (78) give
) , ( ) , ( ) , (
, 0
, 0
, 1 , 1
,
,
T
E
T
T
T
E
T
T P
T E T T E T P E E E
w R w e
w R w e w R
+ s
+ V V =
v
v
( )
2
1
2
, 0
,
2
,
,
, 0
2
1
T
T P
T J
T P
E
E E
e R Ch c + s
( )
) , (
have we , any and T any for Finally,
2
1
2
, 0
,
2
,
,
, 0
2
1
T
T P
T J
T P
E
E E E E E
h
e R Ch w R
T E T
c + s
c e e
From this result and the inverse inequalities (49), give
( ) .
2
, 0
,
2
,
,
2
, 0 T
T P
T J
T P
T
T E
e C R h c + s
(79)
We have also
, .
, 0
,
, 0 , 0 T
T P
T
h
T
T
u R c == V =
then
( ).
2
, 0
,
2
,
,
2
. 0 T
T P
T J
T P
T
T
e R c + s
(80)
Combining (78), (79) and (80), establishes the upper bound in
the equivalence relation.
For the lower, we need to use (65):
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
154 | P a g e
www.ijacsa.thesai.org
( )
( )
) , ( ) , (
sup 2
, 0 ,
, ,
) ( ) , (
T 0,
,
,
,
2
1
2
, 0
,
2
,
, ,
2
0
T T J
T P T P T
T L V q v
T P
T J
T P
T
T P
T J
T P T P
q v
q d v e A
e
e
T
+
+
s
+ s
+ =
e
c
c
c q
Using (75) and (76) give
,
) , ( ) , ( ) , (
sup 2
, 0 ,
) ( ) , (
,
2
0
T T J
T E
T T E E T T
T L V q v
T P
q v
q R v R v R
T
+
+
s
c c
e
q
then
sup 2
, 0 ,
, 0 , 0 , 0
, 0
, 0
, 0
) ( ) , (
,
2
0
T T J
T E
T T
T
E
E
E
T
T
T
T L V q v
T P
q v
q R v R v R
T
+
+ +
s
c c
e
q
(82)
Now, since v
satisfies
T
E
E
v Ch v
, 1
2
1
, 0
s
(83)
T
T
T
v Ch v
, 1 , 0
s (84)
Combining these two inequalities with (82) immediately
gives the lower bound in the equivalence relation.
Consequence 6.12. For any mixed finite element
approximation (not necessarily inf-sup stable) defined on
rectangular grids
h
T , the residual estimator
P
q
satisfies:
P
J
C e q c s +
O O , 0 ,
Note that the constant C in the local lower bound independent
of the domain.
VII. NUMERICAL SIMULATION
In this section some numerical results of calculations with
mixed finite element Method and ADINA system will be
presented. Using our solver, we run the flow over an obstacle
[15] with a number of different model parameters.
Example: Flow over an obstacle. This is another classical
problem. The domain is O and is associated with modelling
flow in a rectangular channel with a square cylindrical
obstruction. A Poiseuille profile is imposed on the Inflow
boundary ), 1 1 ; 0 ( s s = y x and noflow (zero velocity)
condition is imposed on the obstruction and the top and
bottom walls. A Neumann condition is applied at the outflow
boundary which automatically sets the mean outflow pressure
to zero. O a disconnected rectangular region
1) (-1, 8) (0,
generated by deleting the square 1/4). (-1/4, 9/4) (7/4,
Fig.1. Equally distributed streamline plot associated with a 3280 square
grid
0 1
P Q
approximation and
.
500
1
= v
Fig.2. uniform streamline plot computed with ADINA System, associated
with a 32 80 square grid and
.
500
1
= v
Fig.3. Velocity vectors solution by MFE with a 32 80 square grid and
.
500
1
= v
Fig.4. The solution computed with ADINA system. The plots show the
velocity vectors solution with a 32 80 square grid and
.
500
1
= v
The two solutions are therefore essentially identical. This
is very good indication that my solver is implemented
correctly.
Fig.5. Pressure plot for the flow with a 32 80 square grid.
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
155 | P a g e
www.ijacsa.thesai.org
Fig.6. Estimated error
T R,
q
associated with 32 80 square grid and
0 1
P Q
approximation
TABLE I.
TABLE1. The local Poisson problem error estimator Flow over
an obstacle with Reynolds number Re = 1000.
O
V h u .
estimated velocity divergence error.
Grid
O
V
, 0
. h u
P
q
8 20 5.892389e-001 3.210243e+001
16 40 1.101191e-001 6.039434e+000
32 80 3.707139e-002 2.802914e+000
64 160 1.160002e-002 1.484983e+000
TABLE II. A residual error estimator for Flow over an obstacle with
Reynolds number Re = 1000.
Grid
R
q
8 20 9 ,309704e+00
16 40 1,727278e+000
32 80 8,156479e
-001
64 160 4.261901e-001
VIII. CONCLUSION
We were interested in this work in the numeric solution for
two dimensional partial differential equations modelling (or
arising from) model steady incompressible fluid flow. It
includes algorithms for discretization by mixed finite element
methods and a posteriori error estimation of the computed
solutions. Our results agree with Adina system.
Numerical results are presented to see the performance of
the method, and seem to be interesting by comparing them
with other recent results.
ACKNOWLEDGMENT
The authors would like to express their sincere thanks for
the referee for his/her helpful suggestions.
References
[1] Alexandre Ern, Aide-mmoire Elments Finis, Dunod, Paris, 2005.
[2] P.A. Raviart, J. Thomas, Introduction lanalyse numrique des quations
aux drives partielles, Masson, Paris, 1983.
[3] E. Creuse, G. Kunert, S. Nicaise, A posteriori error estimation for the
Stokes problem: Anisotropic and isotropic discretizations, M3AS,
Vol.14, 2004, pp. 1297-1341.
[4] P. Clement, Approximation by finite element functions using local
regularization, RAIRO. Anal. Numer, Vol.2, 1975, pp. 77-84.
[5] T. J. Oden,W.Wu, and M. Ainsworth. An a posteriori error estimate for
finite element approximations of the Navier-Stokes equations, Comput.
Methods Appl. Mech. Engrg, Vol.111, 1994, pp. 185-202.
[6] V. Girault and P.A. Raviart, Finite Element Approximation of the
Navier-Stokes Equations, Springer-Verlag, Berlin Heiderlberg New
York, 1981.
[7] A. Elakkad, A. Elkhalfi, N. Guessous. A mixed finite element method for
Navier-Stokes equations. J. Appl. Math. & Informatics, Vol.28, No.5-6,
2010, pp. 1331-1345.
[8] R. Verfurth, A Review of A Posteriori Error Estimation and Adaptive
Mesh-Refinement Techniques, Wiley-Teubner, Chichester, 1996.
[9] M. Ainsworth and J. Oden. A Posteriori Error Estimation in Finite
Element Analysis. Wiley, New York, [264, 266, 330, 334, 335], 2000.
[10] M. Ainsworth, J. Oden, A posteriori error estimates for Stokes and
Oseens equations, SIAM J. Numer. Anal, Vol. 34, 1997, pp. 228245.
[11] R. E. Bank, B. Welfert, A posteriori error estimates for the Stokes
problem, SIAM J. Numer. Anal, Vol.28, 1991, pp. 591623.
[12] C. Carstensen,S.A. Funken. A posteriori error control in low-order finite
element discretizations of incompressible stationary flow problems.
Math. Comp., Vol.70, 2001, pp. 1353-1381.
[13] D. Kay, D. Silvester, A posteriori error estimation for stabilized mixed
approximations of the Stokes equations, SIAM J. Sci. Comput, Vol.21,
1999, pp. 13211336.
[14] R. Verfurth, A posteriori error estimators for the Stokes equations,
Numer. Math, Vol.55, 1989, pp. 309325.
[15] H. Elman, D. Silvester, A. Wathen, Finite Elements and Fast Iterative
Solvers: with Applications in Incompressible Fluid Dynamics, Oxford
University Press, Oxford, 2005
[16] V. John. Residual a posteriori error estimates for two-level finite element
methods for the Navier-Stokes equations, App. Numer. Math., Vol.37,
2001, PP. 503-518.
[17] D. Acheson, Elementary Fluid Dynamics, Oxford University Press,
Oxford, 1990.