(IJACSA) International Journal of Advanced Computer Science and Applications,

Vol. 4, No.3, 2013

42 | P a g e
www.ijacsa.thesai.org

Resolution of Unsteady Navier-stokes Equations with
the

Boundary condition
Jaouad EL-Mekkaoui, Ahmed Elkhalfi
Mechanical engineering laboratory
Faculty of sciences and techniques-B.P.
2202 Route Imouzzer Fes

Abdeslam Elakkad
Department of mathematics
Regional Centre for Professions of Education and Training,
Fes, B.P: 243 Sefrou Morocco

Abstractin this work, we introduce the unsteady
incompressible Navier-Stokes equations with a new boundary
condition, generalizes the Dirichlet and the Neumann conditions.
Then we derive an adequate variational formulation of time-
dependent Navier- Stokes equations. We then prove the existence
theorem and a uniqueness result. A Mixed finite-element
discretization is used to generate the nonlinear system
corresponding to the Navier-Stokes equations. The solution of the
linearized system is carried out using the GMRES method. In
order to evaluate the performance of the method, the numerical
results are compared with others coming from commercial code
like Adina system.
KeywordsUnsteady Navier-Stokes Equations; Mixed Finite
Element Method;

boundary condition; Adina system.

I. INTRODUCTION
The two firsts PDEs given in section 2 constitute the basis
for computational modeling of the flow of an in compressible
Newtonian fluid. For the equations, we offer a choice of two-
dimensional domains on which the problem can be posed,
along with boundary conditions and other aspects of the
problem, and a choice of finite element discretizations on a
quadrilateral element mesh, whereas the discrete Navier-
Stokes equations require a method such as the generalized
minimum residual method (GMRES), which is designed for
non symmetric systems [9, 19]. The key for fast solution lies
in the choice of effective preconditioning strategies. The
package offers a range of options, including algebraic methods
such as incomplete LU factorizations, as well as more
sophisticated and state- of-the-art multigrid methods designed
to take advantage of the structure of the discrete linearized
Navier-Stokes equations. In addition, there is a choice of
iterative strategies, Picard iteration or Newton method, for
solving the nonlinear algebraic systems arising from the latter
problem.
This paper presents the unsteady Navier-Stocks equations
with a new boundary condition noted by

. This condition
generalizes the known conditions, especially the conditions of
Dirichlet, Neumann...
If are the real numbers strictly positive such that
, then

is the Neumann boundary condition and

if then

is the Dirichlet boundary condition. For

that is called the Dirichlet coefficient and is the Neumann
coefficient. So, we prove that the weak formulation of the
proposed modeling has a unique solution. To calculate this
latter, we use the discretization by mixed finite element
method. Moreover, to compare our solution with the some
previously ones, as ADINA system, some numerical results
are shown.
The paper is organized as follows. Section II presents the
model problem. In the section III we show the existence and
uniqueness of the solution of the standard weak formulation
and of semi-descretization.
In section IV we describe the approximation of the
standard weak formulation using mixed finite elements and
Picards nonlinear iteration and using GMRES algorithm to
solve it. Numerical experiments carried out within the
framework of this publication and their comparisons with
other results are shown in section V.
II. UNSTEADY INCOMPRESSIBLE NAVIER-STOKES
EQUATIONS
We consider the unsteady Navier-Stokes equations for the
flow with constant viscosity

O =
O = V
O = V + V + V
c
c
. in ) 0 (
, in 0 .
, in .
0
2
u u
u
f p u u u
t
u

v
(1)
Where 0 v a given constant is called the kinematic
viscosity. u

is the fluid velocity, p is the pressure field, V is

the gradient and . V is the divergence operator.
The boundary value problem that is considered is posed on
two or three-dimensional domain , is defined as:

( ) : :
,
O c = I = V + in t n pI u b u a C
b a


v (2)
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
43 | P a g e
www.ijacsa.thesai.org
Where n

that is the usual outward-pointing normal

boundary.
), (
2
1
I eH t

b a and are the function nonzero

continuous defined on I verify:
There are two strictly positive constants , and
1 1
| o such
that:

(3) all for
) (
) (

1 1
I e s s x
x b
x a
| o
This system is the basis for computational modeling of the
flow of an incompressible fluid such as air or water. The
presence of the nonlinear convection term u u

V means that
boundary value problems associated with the Navier-Stokes
equations can have more than one solution.
We define the spaces
(4) ) ( ; ; / : ) (
2 1
)
`

O e
c
c
c
c
O = O L
y
u
x
u
u IR u h



| | (5) ) ( ) (
2
1 1
O = O h H

{ } (6) on 0 . / ) ( ) (
1 1
I = O e = O n v H v H
N


{ } (7) on 0 . and . / ) ( I = V O e = n v v D v

{ } (8) on 0 . / ) ( ) (
1
I = V O e = O v H v V
N


{ } (9) on 0 . and . / ) (
2
I = V O e = n v v L v H



) ; , 0 ( X T L
p
is the space of strongly measurable functions
X T f ) , 0 ( : such that:
, ) (
2
1
0
2
) ; , 0 (

(

=
}

T
X X T L
dt t f f P

. ) ( ess sup
) , 0 (
) ; , 0 (
=
e

X
T t
X T L
t f f

For , u

and w

in ) (
1
O H we define, as usual, the
following forms:

: ) , (
0
v u
b
a
v u v u a

} }
O I
+ V V =v

. ) , (
}
O
V = u q q u b


. ) , (
}
O
= q p q p d

) ( ) , , (
1
}
O
V = v u z v u z a


) , , ( ) , ( ) , , (
1 0
v u z a v u a v u z c

+ =
. . . ) (
} }
I O
+ = v f v t
b
a
v l

For u

and w

in ) ; , 0 (
2
V T L we consider the function
) (
0
t u A t

defined a.e. on
] T , 0 [
by:
( ) . ), ( ), ( , ' ) (
0 0 0
V v v t u a v t u A V t u A e = e


It can be readily checked that ). ' ; , 0 ( ) (
2
0
V T L t u A t e


Next, consider the mapping )) ( ), ( (
1
t u t w A t

defined a.e
On
] T , 0 [
by:
( )
( ) ( )
.
), ( ), ( , ) ( ), (
, ' ) ( ), (
1 1
1

e =
e
V v v t u t w a v t u t w A
V t u t w A



. V v all for e

The standard weak formulation of (1), find ) (
1
O e
N
H u

and
) (
2
O eL p such that

, ) , ( ) , ; ( v t
b
a
v f q v b v u z c v
t
u

} } }
I O O
+ = + +
c
c
(10)
, 0 ) , ( = q u b

(11)

). ( ) (
2 1
O e O e L q and H v all for
N

The spaces
), (
1
O H

) (
1
O
N
H
and V equipped with
the norms
( )


:
2
1
2
, 0
2
, 1 , 1
2
1
,
O O O
O I
O
+ =
|
.
|

\
|
+ O V V =
} }
v v v
d v v
b
a
d v v v
J


v

The spaces H and ) (
2
O L equipped with the norm
( ) 2
1
, 0
, v v v

=
O

is new boundary condition, for that we need to show

the existence and uniqueness theorems for this modeling.
III. EXISTENCE AND UNIQUNESS OF THE SOLUTION
A. Weak formulation
THEOREM 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
O eH v all for

(12)

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 this result and the
result (3) give,

), (
1
1,
2
,
O e s
O O
H v all for v c v
J



( )

. with
2
1
2
1 2
v | + = c c On other hand, according to 5.55 in
[1], there exists a constant 0 such that


), ( ), (
1
2
, 0
2
, 0
2
, 0
O e + V s
I O O
H v v v v

using (3) and

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
44 | P a g e
www.ijacsa.thesai.org
,
2
,
2
, 0 O O
s V
J
v v

give
), ( c
1
J, , 1
1
O e s
O O
H v all for 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



) . ), ( (
, 1
1
O
O - H is a real space and ) (
1
O
N
H is closed
in ) (
1
O H and
.
, 1 O
and
.
,O J
are equivalent norms,
then
) . ), ( (
,
1
O
O
J
H
and
) . ), ( (
,
1
O
O
J
N
H
are a reals
Hilbert spaces.

LEMMA 2.
) (
1
0
O eH v All

satisfy
. 2
2
1
, 0
2
1
, 1
4
1
, 4 , 0 O O O
s v v v

(13)

PROOF. See lemma 1.5 chapter V [2].
LEMMA 3.
) (
1
O eH v All

satisfy
. 2
2
1
, 0
2
1
, 1
4
1
, 4 , 0 O O O
s v v v

(14)
PROOF. is a bounded open of
2
IR with Lipschitz
continuous boundary, then there exists a sequence
( )
0 >
O
n n
of
opens, such that

. lim , 0 O = O O c O >
n n n
and n all For

Let
n
be a function defined as:

O O =
O =
. 1
, 1
n
n
on
on
n
n

(15)
Then for all
), (
1
O eH v

we get
), (
1
0 n
O eH v

for all
, 0 > n

According to lemma 2,
. 2
2
1
, 0
n
2
1
, 1
n
4
1
, 4 , 0
n
O O O
s v v v



By applying dominated convergence Theorem, we get

. 2
2
1
, 0
2
1
, 1
4
1
, 4 , 0 O O O
s v v v



LEMMA 4. when w

and u

belong both to
) ; , 0 ( ) ; , 0 (
2
H T L V T L

then
) ' ; , 0 ( ) , (
2
1
V T L u w A e

(16)
PROOF. , w

, u

, V v e

we have
( ) ( )
. ) . ( .
) . . .( , , , ,
1 1
}
}
O
O
V =
V + V = +
v u w
u v v u w u v w a v u w a




Use Greens theorem to show that
( ) ( )
0.
) . .( ) . )( . ( , , , ,

1 1
=
= +
} }
O O c
v u z div v u n z u v w a v u w a


Then,
( ) ( ), , , , ,
1 1
u v w a v u w a

=
and we have the upper
bound:

( )
O O O
s
, 1 , 4 , 0 , 4 , 0
1 1
, , v u w c v u w a


Therefore
. ) ( ) ( )) ( ), ( (
, 4 , 0 , 4 , 0
1 1
O O -
s t u t w c t u t w A

(17)
We make use of lemma 2. According to (14) and (17),
we have
O O O O -
s
, 0 , 0 , 1 , 1
1
2
1
) ( ) ( ) ( ) ( 2 )) ( ), ( ( t u t w t u t w c t u t w A

(18)
Therefore,


)) ( ), ( (
) ; , 0 ( ) ; , 0 ( ) ; , 0 ( ) ; , 0 (
0
2
1
2 2
V T L V T L H T L H T L
T
u w u w C
dt t u t w A



s
} -

This proves (16).
In the course of this proof we have shown that, if u

belongs both to
) ; , 0 ( ) ; , 0 (
2
H T L V T L

then

) , (
) ; , 0 ( ) ; , 0 ( ) ' ; , 0 (
1
2 2
V T L H T L V T L
u u C u u A

s
(19)
With these spaces and forms, consider the following
variational formulation of problem (1)-(2).
For a given function
)) ( ; , 0 (
1 2
O e

H T L f

and a given
element
0
u

of
, H
find
) ; , 0 ( ) ; , 0 (
2
H T L V T L u

e

Such that
( )

O = -
e
+ = + -
} }
I O
. in ) 0 (
[), , 0 (] ' in ,
, ) ), ( ); ( ( ), (
0
u u
T D V v
v t v f v t u t u c v t u
dt
d

(20)
THEOREM 5. Let
) ; , 0 ( ) ; , 0 (
2
H T L V T L u

e

be
solution of (20). Then
). ' ; , 0 (
2
V T L
dt
u d
e

PROOF. We have
v t u t u A t u A v t u t u c

)), ( ), ( ( ) ( ) ), ( ); ( (
1 0
+ =
. V v all for e

Therefore, by Lemma 1.1 chapter V in [2], each solution u

of
(20) satisfies in D(]0, T[):
v t u t u A t u A v t
b
a
v f v t
dt
u d

)), ( ), ( ( ) ( ), (
1 0
+ + =
} }
I O

. V v all for e

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
45 | P a g e
www.ijacsa.thesai.org
Now, by hypothesis
) ' ; , 0 ( :
2
V T L v t
b
a
v f v l e +
} }
I O

and we have mentioned previously that ) ' ; , 0 (
2
0
V T L u A e

when
). ; , 0 (
2
V T L u e

Furthermore ( ) ) ' ; , 0 ( ,
2
1
V T L u u A e


Hence,
). ' ; , 0 (
2
V T L
dt
u d
e

According to Theorem 5 and Theorem 1.1 in [2] chapter V,
). ]; , 0 ([
0
H T C u e

In this case, it is perfectly allowable to prescribe the value of
u

at t=0. Moreover, it stems from the above proof that

problem (20) can be equivalently stated as follows:
For a given function
)) ( ; , 0 (
1 2
O e

H T L f

and a given
element
0
u

of
, H
find

O = -
= + + -
e
e -

. in ) 0 (
, ) , (
) ' ; , 0 (
) ; , 0 ( ) ; , 0 (
0
1 0
2
2
u u
L u u A u A
dt
u d
that such V T L
dt
u d
with
H T L V T L u

(21)
We consider the following problem:
For a given function
)) ( ; , 0 (
1 2
O e

H T L f

and a given
element
0
u

of
, H
find
) , ( p u

such that

O = -
O = V + V + V
c
c
-
O e
e -

. in ) 0 (
, in .
[) , 0 ] ( ' ) ' ; , 0 (
) ; , 0 ( ) ; , 0 (
0
2
2
2
u u
f p u u u
t
u
T D and V T L
dt
u d
with
H T L V T L u

v
(22)
THEOREM 6. Problem (21) and (22) are equivalent.
To show this Theorem we need the following lemmas
LEMMA7. Let ) (
2
1
I eH g

satisfies
}
I
= . 0 d n g

Then
there exists a function ) (
1
O eH u

such that
. 0 I = = on g u and u div


PROOF. See Temam [20].
LEMMA8. The divergence operator is an isomorphism from

V onto ), (
2
0
O L verifies: There exists 0 o such that
.
2
,
2
, 0

O O
e > V v v v div
J

o

PROOF. Let ) (
1
O e
N
H v

By Greens formula:
} }
O I
= = . 0 . d n v dx v div


Let q a function of ). (
2
0
O L As O is bounded, there exists
some function ) (
2
O eL u such that
. O = A in q u
We set .
1
O = in grad v u

Then
;
1
O = A = in q v div u

moreover, by Greens formula

} } }
I O O
= = = . 0 .
1 1
dx q dx v div d n v

Also ). (
2
1
1 0
I eH v

Therefore, we can apply Lemma 7:

there exists ) (
1
1
O eH w

such that 0
1
= w div

and
.
1 0 1 0
w v

= then
1 1
w v v

= is the required function since
) (
1
0
O eH v

and . q v div =

Finally, it follows from the open mapping Theorem (cf.Yosida
[21]) that the inverse of div is continuous from ) (
2
0
O L onto
,

V therefore div is an isomorphism.

LEMMA 9. Let
) (
1
O e

H l

and satisfy

e = V v v l

, 0 ,

Then there exists exactly one function in
) (
2
0
O L
such that:
}
O
= = . , , v grad dx v div v l

(23)
). (
1
O e
N
H v

PROOF. Consider the following problem:
Find

eV u

satisfying
. , , ,

e = V v v l v div u div

(24)
By the Lemma 8 and the Lax-Milgrams Theorem, (24) has a
unique solution u

in .

V Then,
). ( , , ,
1
O e =
N
H v v l v div u div

We set ) (
2
0
O e = L u div

and we find (23).
It remains to prove that is unique in ). (
2
0
O L But clearly, if
) (
2
0
O eL and 0 , = v div

), (
1
O e
N
H v

then 0 =
since div maps ), (
1
O
N
H onto ). (
2
0
O L
PROOF OF THEOREM 6. Clearly, if ) , ( p u

is a solution
of (22), thenu

satisfies (21).
Conversely, let
) ; , 0 ( ) ; , 0 (
2
H T L V T L u

e

A solution of (21) and consider the mapping defined on
), (
1
O
N
H by:
{ } ) , ( ) ), ( ( ) ), ( ); ( ( ), (
) , (
0
0
v u v t u ds v t u t u c v s f
t v L
t

+ =
}

For each t, L is a linear functional on ), (
1
O
N
H that
vanishes on V. Hence, according to Lemma 9, for each t there
exists exactly one function ), ( ) (
2
0
O eL t P such that
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
46 | P a g e
www.ijacsa.thesai.org
v t P grad t v L

), ( ) , ( = ). (
1
O e
N
H v

In other words,
{ } ) , ( ) ), ( ( ) ), ( ); ( ( ), (
) ), ( (
0
0
v u v t u ds v s u s u c v s f
v t P
t

+ =
}

). (
1
O e
N
H v

(25)
By using Lemma 4, it can be checked that
( ) ) ( ]; , 0 [
2
0
0
O e L T C P
Next, by differentiating (25), we get:

{ } ) , ( ) ,
) (
( ) ), ( ); ( ( ), (
) ,
) (
(
0
v u v
dt
t u d
v s u s u c v s f
v
dt
t dP

+ =

). (
1
O e
N
H v

(26)
Thus, if we set
dt
dP
p =
in
[), , 0 ] ( ' T D O
we find the
second equation of (22).

THEOREM 10. Problem (21) has a unique solution in
) ; , 0 ( ) ; , 0 (
2
H T L V T L u

e

PROOF. The same steps of proof of Theorem 1.5 chapter VI
in [2] but the spaces V, H,
)... (
1
O
N
H
are note the same. .
B. Semi-discritisation
In this paragraph, we propose to analyze a very simple one-
step method in order to illustrate the type of argument that is
often used when dealing with semi-discretization. Consider
again the problem (20).

find ) ; , 0 ( ) ; , 0 (
2
H T L V T L u

e

Such that
( )

O = -
e
= + -
. in ) 0 (
[), , 0 (] ' in ,
), ( ) ), ( ); ( ( ), (
0
u u
T D V v
v l v t u t u c v t u
dt
d

(27)

Let
N
T
k =
and
n
t the subdivisions of [0, T]: ; nk t
n
=
. 0 N n s s
Now, suppose that an approximation, , V u
n
e

of ) (
n
t u

is
available and consider the following problem:

( )

e + =
+ -
e -
} }
I
+
O
+
+ +
+
. , ) ( ) (
) , ; ( ,
1
1 1
1 1
1
V v v t t v t f
v u u c v u u
k
that such V u find
n n
n n n n
n

(28)

,
n
u

) ( and ) (
1 1 + + n n
t t t f

are given respectively in V and

V, it follows that (28) can be expressed in the from:
( )

e +
(

+ =
+ -
e -
} }
I
+
O
+
+ +
+
. ) , ( , ) ( ) (
) , ; ( ,
1 1
1 1
1
V v v u v t t v t f k
v u u kc v u
that such V u find
n n n
n n n
n

(29)

Thus, we are asked to solve a linear boundary value
problem associated with the bilinear form:
( ) ( ) ). , ; ( , , v u u kc v u v u
n

+
This form is continuous in V V and V-elliptic since
( )
2
,
2
, 0
) , ; ( ,
O O
+ = +
J
n
v k v v v u kc v v


Therefore, by Lax-Mailgrams theorem [2], problem (29)
has a unique solution
1 + n
u

in V.
IV. DISCRETIZATION BY MIXED FINITE
ELEMENTS
Our goal here is to consider the unsteady Navier-Stokes
equations with
b a
C
,
boundary conditions in a two dimensional
domain and to approximate them by a mixed finite element
method.
Mixed finite element discretization of the weak
formulation of the Navier- Stokes equations gives rise to a
nonlinear system of algebraic equations. Two classical
iterative procedures for solving this system are Newton
iteration and Picard iteration
Let , 0 ; h T
h
be a family of rectangulations of O. For
any .
h
T T e
We denote by
T
h the diameter of a simplex, by
E
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 1
h
O c
N
H and ) ( M
2
0
h
O c L
The discrete version of (10)-(11) is:
: such that and
1 h
h h h
M p X u find e e

( )
( )
h h h h
h h h h h h h h
h
v t
b
a
v f p v b
v u u a v u
b
a
v u v
t
u

} }
} } }
I O
O O c O
+ = +
+ + V V +
c
c
-
,
, , . :
1
v
(30)
( ) 0 , = -
h h
q u b

(31)
. and
1 h
h h h
M q X v all for e e

We define the appropriate bases for the finite element
spaces, leading to a non linear system of algebraic equations.
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
47 | P a g e
www.ijacsa.thesai.org
To define the corresponding linear algebra problem, we use a
set of vector-valued basis functions { }
u
n i i ,..., 1 =

So that

=
=
u
n
j
j j h
y x t u y x t u
1
) , ( ) ( ) , , (

(32)
We introduce a set of pressure basis functions { }
p
n k k ,..., 1 =

and set

=
=
p
n
k
k k h
y x t p y x t p
1
) , ( ) ( ) , , (
(33)
Where
u
n and
p
n are the numbers of velocity and pressure
basis functions, respectively.
| | ) ( ) ( ) ( ] )) ( ( ) ( t L t BP t U M t U N t
dt
dU
D = + + +
(34)
0 ) ( = t U B
T
(35)
With
( )
T
n
t u t u t u t U
u
) ( ),..., ( ), ( ) (
2 1
=

( )
T
n
t p t p t p t P
p
) ( ),..., ( ), ( ) (
2 1
=

}
O
= , . ); (
, , j i j i j i
d d D


( )

}
=
O
V = =
u
n
k
i k j k j i j i
t u c c t U N
1
, ,
, . ) ( ), ( )) ( (


, : ), (
, , i j i j j i j i
b
a
m m M v

} }
O O
+ V V = =

, : ), (
, , i j i j j i j i
b
a
m m M v

} }
O O
+ V V = =

. ; ] [
, ,
}
O
V = =
j k j k j k
b b B


( ) ) ( ). ( ) ( ; ) ( ) (
i i i i
t t
b
a
t f t l t l t L

} }
O I
+ = =

for
, ,..., 1 ,
u
n j i =
and
. ,..., 1
p
n k =

Using the backward Euler method for the time derivative
and substituting into (34)-(35), one obtains the following
system of nonlinear equations in tensor notation:

( ) , ) 0 ( ),..., 0 ( ), 0 ( ) 0 (
2 1
T
n
u
u u u U =
(36)
| | ), ( ] ) (
1
1 1 1
1
+
+ + +
+
= + + +

n
n n n
n n
t L BP U M U N
k
U U
D
(37)
. 0
1
=
+ n T
U B (38)

With ) 0 ( ),..., 0 ( ), 0 (
2 1
u
n
u u u are the coordinates in the
basis { }
u
n i i ,..., 1 =

of the approximation of
0
u

in .
h
X Solution
of the nonlinear system of equations, Eq. (34)-(35), can be
carried out efficiently using Picards method, where we start
with an initial guess
p u
n n n n
IR P U
+
e ) , (
0 , 0 ,
and construct a
sequence of iterates
p u
n n
m n m n
IR P U
+
e ) , (
, ,
it converges to
the solution of (34)-(35). In this approach we approximate the
nonlinear convection term as follows:

| | | |
1 , 1 , 1 1 1
] ) ( ] ) (
+ + + + +
+ ~ +
m n m n n n
U M U N U M U N

For the finite-element basis functions, we chose to work
with stable rectangular elements (Q2-Q1), where we use
biquadratic approximation for the velocity components,
bilinear approximation for the pressure, and stable triangular
elements (P2-P1), where we use quadratic approximation for
the velocity components and linear approximation for the
pressure.
The linear system we need to solve within each iteration of
Picards method has the following generic form:
.
0 0

0
0 0
|
|
.
|

\
|
=
|
|
.
|

\
|
|
|
.
|

\
|
+ f
P
U
B
B N A
T
(39)
We use the generalized minimum residual method
(GMRES) for solving the nonsymmetric systems.
Preconditioning is a technique used to enhance the
convergence of an iterative method to solve a large linear
system iteratively. Instead of solving a system
x = b, one solves a system ,
1 1
b P x P

= A where P is
the preconditioned. A good preconditioned should lead to fast
convergence of the Krylov method. Furthermore, systems of
the form Pz = r should be easy to solve. For the Navier-Stokes
equations, the objective is to design a preconditioned that
increases the convergence of an iterative method independent
of the Reynolds number and number of grid points. We use a
least-squares commutator preconditioning [10, 11,12].
V. NUMERICAL SIMULATIONS
In this section, some numerical results of calculations with
mixed finite element method and ADINA system will be
presented. Using our solver, we run two traditional test
problems (driven cavity flow [9, 14, 15, 16, 17], Backward-
facing step problem [10, 13]) and the flow over an obstacle [9]
with a number of different model parameters.
EXAMPLE 1. Driven cavity flow. It is a model of the flow
in a square cavity with the lid moving from left to right. Let
the computational model: { }, 1 / 1 1 ; 1
4
x u x y
x
= s s =

a regularized cavity. The streamlines are computed from the
velocity solution by solving the Poisson equation numerically
subject to a zero Dirichlet boundary condition.


Fig.1. Vorticity (left plot), and velocity magnitude solution (right plot) using
P2 P1 approximation a 64 64 square grid and Reynolds number Re =100.
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
48 | P a g e
www.ijacsa.thesai.org

Fig.3. Vorticity (left plot), and velocity magnitude solution (right plot) using
P2 P1 approximation a 64 64 square grid and Reynolds number Re =400.



Fig.2. Vorticity (left plot), and velocity magnitude solution (right plot) using
P2 P1 approximation, a 64 64 square grid and Reynolds number
Re=1000.

Fig.4. The velocity component u at vertical center line (left plot), and the
velocity component v horizontal center line (right plot) with a 129 129 grid
and Re=100.


Fig.5. The velocity component u at vertical centerline (left plot), and the
velocity component v at horizontal center line (right plot) with a 129 129
grid and Re=1000.
Figures 4 and 5 shows the velocity profiles for lines
passing through the geometric center of the cavity. These
features clearly demonstrate the high accuracy achieved by the
proposed mixed finite element method for solving the
unsteady Navier-Stokes equations in the lid-driven squared
cavity.
EXAMPLE 2. L-shaped domain , parabolic inflow
boundary condition, natural outflow boundary condition.
This example represents flow in a rectangular duct with a
sudden expansion; a Poiseuille flow profile is imposed on the
inflow boundary
{ }, 1 0 ; 1 s s = y x
and a no-flow (zero
velocity) condition is imposed on the walls.
The Neumann condition is applied at the outflow boundary
(x=5; 1 < y < 1) and automatically sets the mean outflow
pressure to zero.

Fig.6. Equally spaced streamline plot associated with a 32 96 square
grid,
0 1
P Q approximation and
100
1
= v
( t=100 seconds).
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 4, No.3, 2013
49 | P a g e
www.ijacsa.thesai.org


Fig.7. The solution computed with ADINA system. The plot show the
streamlines associated with a 32 96 square grid and
100
1
= v
(t=100
seconds).
The two solutions are therefore essentially identical. This
is very good indication that my solver is implemented
correctly.
VI. CONCLUSION
In this work, we were interested in the numerical solution
of the partial differential equations by simulating the flow of
an incompressible fluid. We introduced the unsteady Navier-
Stokes equations with a new boundary condition noted Ca,b.
We have shown the existence and uniqueness of the solution
of the weak formulation and the solution of the semi-
discretization. We used the discretization by mixed finite
element method.
Numerical experiments were carried out and compared
with satisfaction with other numerical results, either resulting
from the literature, or resulting from calculation with
commercial software like Adina system.
ACKNOWLEDGMENTS
The authors would like to express their sincere thanks for
the referee for his/her helpful suggestions.
REFERENCES
[1] Alexandre Ern, Aide-memoire Elements Finis, Dunod, Paris, 2005.
[2] P.A. Raviart, J. Thomas, Introduction lanalyse numerique des quations
aux derives partielles, Masson, Paris, 1983.
[3] R.E. Bank, B. Welfert, A posteriori error estimates for the Stokes
problem, SIAM J. Numer. Anal 28, pp. 591-623, 1991.
[4] C. Carstensen, S.A. Funken. A posteriori error control in low-order finite
element discretizations of incompressible stationary flow problems.
Math. Comp., 70, pp. 1353-1381, 2001.
[5] R. Verfurth, A Review of A Posteriori Error Estimation and Adaptive
Mesh-Refinement Techniques, Wiley-Teubner, Chichester, 1996.
[6] D.H. Wu, I.G. Currie. Analysis of a posteriori error indicator in viscous
flows. Int.J.Num. Meth.Heat Fluid Flow., 12, pp. 306-327, 2001.
[7] E. Creuse, G. Kunert, S. Nicaise, A posteriori error estimation for the
Stokes problem: Anisotropic and isotropic discretizations, M3AS 14, pp.
1297-1341, 2001.
[8] P. Clement, Approximation by finite element functions using local
regularization, RAIRO. Anal. Numer. 2, pp. 77-84, 1975.
[9] H. Elman, D. Silvester, A. Wathen, Finite Elements and Fast Iterative
Solvers: with Applications in Incompressible Fluid Dynamics, Oxford
University Press, Oxford, 2005.
[10] M. ur Rehman, C. Vuik and G. Segal, A comparison of preconditioners
for incompressible Navier-Stokes solvers, Int. J. Numer. Meth. Fluids,
57, pp.1731-1751,2007.
[11] H. Elman, V.E. Howle, J. Shadid, D. Silvester, and R. Tuminaro, Least
squares preconditioners for stabilized discretizations of the Navier-
Stokes equations. SIAM J. Sci. Comput., 30, 290-311, 2007.
[12] A. Gauthier, F. Saleri, and A. Veneziani, A fast preconditioner for the
incompressible Navier-Stokes equations, Comput. Visual. Sci, 6, pp.
105 -112, 2004.
[13] P. Gresho, D. Gartling, J. Torczynski, K. Cliffe, K. Winters, T. Garratt,
A. Spence, and J.Goodrich, Is the steady viscous incompressible 2d flow
over a backward facing step at Re =800 stable?, Int. J. Numer. Methods
Fluids, 17, pp. 501-541, 1993.
[14] S. Turek, A comparative study of some time-stepping techniques for the
incompressible Navier-Stokes equations: From fully implicit nonlinear
schemes to semi-implicit projection methods, Int. J. Numer. Meth.
Fluids, 22, pp.987-1011, 1996.
[15] E. Erturk, T.C. Corke and C. Gokcol, Numerical solutions of 2-D steady
incompressible driven cavity flow at high Reynolds numbers, Int. J.
Numer. Meth. Fluids, 48, pp.747-774, 2005.
[16] S. Garcia, The Lid-Driven Square Cavity Flow: from Stationary to Time
Periodic and Chaotic, Commun. Comput. Phys, 2, pp.900-932, 2007.
[17] E. Barragy and G.F. Carey, Stream function vorticity driven cavity
solution using p finite elements, Comput. Fluids, 26, pp.453-468, 1997.
[18] F. Brezzi and M. Fortin, Mixed and Hybrid Finite Element Method,
Springer Verlag; New York, 1991.
[19] A. Nejat, and C. Ollivier-Gooch, Effect of discretization order on
preconditioning and convergence of a high-order unstructured Newton
GMRES solver for the Euler equations, J. Comput. Phys. 227, pp.2366
2386.2008.
[20] R.Temam Navier-Stokes Equations. North Holland, Amsterdam,1977.
[21] K.Yosida, Functional Analysis. Spriger-Verlag, Berlin,1965.
[22] Elman, H. and Ramage, A. and Silvester, D.J. Algorithm 866: IFISS: a
Matlab toolbox for modelling incompressible flow. ACM Transactions
on Mathematical Software, 33, pp.1-18, 2007.
[23] S.Alami1, A.Elakkad, and A. Elkhalfi, A posteriori error estimation for
incompressible flow problem, IJMER, 2, pp. 533-538, 2012.
[24] Mohamed S. Ebeida, Roger L. Davis and Roland W. Freund, Unsteady
Incompressible Flow Simulation Using Galerkin Finite Elements with
Spatial/Temporal Adaptation, American Institute of Aeronautics and
Astronautics, pp.1-11, 2007.