10 11648 J Ajam 20221004 14
10 11648 J Ajam 20221004 14
10 11648 J Ajam 20221004 14
Email address:
m.elhatri@gmail.com (Mohamed El Hatri), infokhansa02@gmail.com (Rachid Ghenji)
∗
Corresponding author
Received: June 26, 2022; Accepted: July 26, 2022; Published: August 17, 2022
Abstract: In this work, we are interested in the mathematical study of the flow of a Newtonian Navier-Stokes fluid, coupled to
the energy equation, in a domain with axial symmetry. The study consists first of all in reducing this problem, which is posed in a
domain in dimension three (3-D), to a problem whose spatial domain is in dimension two, using the transformation of Cartesian
coordinates in cylindrical coordinates, assuming that the problem data does not depend on the angle of rotation. The problem
thus obtained is a so-called axially symmetric problem presenting a degeneracy on the axis of symmetry, hence the interest of
this study. The study of this problem is the subject of the first part of this article which deals with the existence and uniqueness of
the weak solution of the problem in a Sobolev space with appropriate weight. The results of this part have already been published
by the same authors that we recall here with some slight modifications in order to facilitate the reading and understanding of
the second part of the article. In this second part, we approach the existence and the unicity of the numerical solution of the
posed problem. It is obtained using the Lagrange finite element method whose polinomial space is of degree one. The study in
question highlights the necessary algebraic relations between the different physical parameters of the problem to which the flow
in question obeys.
Keywords: Newtonian Fluid, Navier-Stokes Equations, Axisymmetric Problem, Weak Solution, Numerical Solution,
Finite Element Method, Weighted Sobolev Space
1. Introduction and Formulation of the fluid, so called is then pumped into conventional exchangers
to produce superheated steam, called a thermodynamic fluid,
Problem which drives a turbine to generate electricity.
For the detailed description and operation of this device we
1.1. Introduction refer the reader, for example, to works [5, 9, 12, 17] and to the
abundant literature cited in these works.
For essentially economic reasons, one of the most In this work, we limit ourselves to the mathematical study
widely used technologies currently producing electricity (existence and uniqueness of the weak and the numerical
on the concentrated solar power principle (CSP) is solution) of stationary Navier-Stokes equations modeling the
the parabolic solar concentrator (PTC). The physico- flow process of a viscous incompressible Newtonian fluid in
mathematical modeling of the different processes involved in the PTC collector tube and subjected to temperature-dependent
this technical device has been the subject of intense studies by gravity force, when this is a solution of the heat equation.
the specialized scientific community for several decades. The force of gravity intervening in the Navier-Stokes system
Recall that a solar parabolic concentrator is a technology is approximated by the Boussinesq method.
used in a solar power plant to increase the thermal energy We will formulate the physical problem with respect to
of a heat transfer fluid charged with transporting heat and Cartesian coordinates (x, y, z). His mathematical study will be
circulating in a collector tube called solar collector. This
American Journal of Applied Mathematics 2022; 10(4): 141-159 142
carried out with respect to the cylindrical coordinates (r, θ, z) 1. the flow is completely developed turbulent;
in the case of an axisymetrical problem. 2. the regime is stationary;
3. the weight of the fluid is the only body force acting on
1.2. Formulation of the Physical Model the fluid.
Note that a turbulence is a flow property and not a fluid
property, which makes sense only in a three-dimensional space
and occurs at high Reynolds numbers.
t
Let u = (ux , uy , uz ) be the velocity field of the
through-flow in Q, where by wt we denote the transpose
of the vector (wx , wy , wz ) . Then, taking into account the
above assumptions, the flow of a viscous, Newtonian and
incompressible fluid is governed by the following equations:
1) Continuity equation:
Figure 1. Cylinder. ∂ux ∂uy ∂uz
∇u ≡ + + = 0, (1)
The model presented here is far from modeling all aspects ∂x ∂y ∂z
of the physical processes involved in a PTC-type device. We
2) Navier-Stokes equations in a vector form:
are interested here only in the velocity of the flow and the
temperature generated in the fluid from a given source. 1
Let Q ⊂ R3 be an open cylindrical tube of the PTC and −ν∆u + (u · ∇) u + ∇p = z, (2)
ρ0
placed horizontally with respect to a Cartesian coordinates
system (x, y, z) as shown in the Figure 1. We assume that Q is where, ∆ and ∇ are the usual differential operators of Laplace
t
occupied by a Newtonian fluid, visqueus and incompressible and divergence respectively and z = (0, 0, zz ) .
such that: The vector equation (2) in a scalar form can be written as
∂ 2 ux ∂ 2 ux ∂ 2 ux
∂ux ∂ux ∂ux 1 ∂p
−ν + + + ux + uy + uz =− , (3)
∂x2 ∂y 2 ∂z 2 ∂x ∂y ∂z ρ0 ∂x
∂ 2 uy ∂ 2 uy ∂ 2 uy
∂uy ∂uy ∂uy 1 ∂p
−ν 2
+ 2
+ + ux + uy + uz =− , (4)
∂x ∂y ∂z 2 ∂x ∂y ∂z ρ0 ∂y
1.3. Formulation of the Axisymmetric Problem of the flow velocity, the pressure and the temperature will not
depend on the angle of rotation θ. In this case, one speaks
of an axisymetric flux without swirl. For more details about
the relation between the swirl flux and the no swirl flux, see
[4]. In Thus, the vector u of velocity of the fluid is reduced
to two components, one is radial ur, the other is axial uz , i.e.
− →
→ −
u = (ur , uz ) with respect the basis ir , k , where
→
− →
−
ir = (cos θ, sin θ, 0) , k = (0, 0, 1) ,
∂ 2 ur
1 ∂ ∂ur ur ∂ur ∂ur 1 ∂p
−ν r + 2
− 2 + ur + uz =− + β0 T, (9)
r ∂r ∂r ∂z r ∂r ∂z ρ0 ∂r
∂2T
1 ∂ ∂T ∂T ∂T
−λ r + + ur + uz = g. (11)
r ∂r ∂r ∂z 2 ∂r ∂z
t t
Let’s put r = x1 , z = x2 , ur = u1 , uz = u2 , T = u3 , u = (u1 , u2 ) and U = (u1 , u2 , u3 ) , where (x1 , x2 ) ∈ Ω,
Ω =]0, 1[×]0, 1[, see Figure 2, where the axis (Oz) represents that of the cylinder Q in Figure 1.
Therefore, we can write the equations (8) , (9) − (10) and (11) in the following vector form:
where
t
∆r U = (ν∆r u, λ∆r u3 )t , ∆r u = (∆r u1 , ∆r u2 ) ,
1 u1 ∂
∆r u1 = D1 (x1 D1 u1 ) + D22 u1 − 2 , Di = (i = 1, 2);
x1 x1 ∂xi
1
∆r uj =D1 (x1 D1 uj ) + D22 uj ( j = 2, 3) ;
x1
X2
(u · ∇) U = ui Di U, Di U = (Di u1 , Di u2 , Di u3 ) ;
i=1
R
Our goal is to establish the existence and uniqueness of the dµ (x), where µ (Ω) = Ω x1 dx is a density measure.
weak solution of the vectorial axisymetrical problem (12) − Let Lp (Ω, µ) be the completed of C0∞ Ω with respect to
(13) in an appropriate weighted sobolev space. k·kp,µ
Unlike the approach of decoupling the field of flow the norm kukp,µ , i.e. Lp (Ω, µ) = C0∞ Ω , see [8].
velocities from temperature to study the existence and We denote by W k,p (Ω, µ) , 1 ≤ p < ∞, the so called
uniqueness of the weak solution of the Navier-Stokes problem weighted Sobolev space of all scalar functions u = u (x)
coupled to the energy equation (3D), see [2, 6, 16], here which are defined a.e. on Ω and whose generalized derivatives
we treat the corresponding axisymmetric problem (2D) in Dα u (x) ∈ Lp (Ω, µ) for all α such that |α| ≤ k, where
vectorial form. Of course, here, the study takes place in an
appropriate weighted Sobolev space. ∂ |α|
Dα = D1α1 D2α2 = , |α| = α1 + α2 .
The mathematical study of the axisymmetric problem of the ∂ α1 x1 ∂ α2 x2
Navier-Stokes equations has been the subject of many works
The space Wk,p (Ω, µ) is a Banach space as completed
under different aspects, existence uniqueness and regularity of
space of C0∞ Ω with respect to the norm
the weak solution of the problem, see for example [1, 4, 15, 20]
and the references cited in these works.
In the case of an elliptical problem with axial symmetry, we 1/p
refer the reader to the article [14]. X Z p
kukk,p,µ = |Dα u (x)| dµ (x) ,
|α|≤k Ω
1/p
X Z p
kukk,p = |Dα u (x)| dx .
|α|≤k Ω
1,2 1
|·|k,p,µ
W0,Γ (Ω, µ) = C0,Γ Ω .
For more details on the weighted sobolev spaces we refer the reader to the monograph [13].
To establish the variational formulation of problem (12) and (13), we will need
the following lemma:
1,2 1,2
Lemma 2.1. Let ui ∈ W0,Γ (Ω, µ) , i = 1, 2, 3 and v ∈ W0,Γ (Ω, µ) ∩ C 1 Ω . Then, there is a sequence x1,k ∈]0, 1[, k = 1,
2, . . . , lim x1,k = 0, such that
k→∞
Z 1
∂ui (x1,k , x2 )
lim x1,k v (x1,k , x2 ) dx2 = 0, i = 1, 2, 3. (14)
k→∞ 0 ∂x1
Proof For the proof of this lemma, see [11].
The following lemma play an essential role in the proof of unicity and
existance of the weak solution the problem (12) − (13) .
1 1,2
For the proof of this lemme, we use the density of the space C0,Γ Ω in the space W0,Γ (Ω, µ) .
1,2
Lemma 2.2. Let v ∈ W0,Γ (Ω, µ). Then, we have the following apriori estimate
1
kvk2,µ ≤ |v| . (15)
2 1,2,µ
In addition, for all v = (v1 , v2 ) ∈ V we have
2 !1/2
Z
v1 (x) √
dµ (x) ≤ 2 |v|1,2,µ . (16)
Ω x1
1
Proof For all v ∈ C0,Γ Ω , we have
Z 1 Z 1 21 Z 1 21
1 2
|v (x1 , x2 )| = − D1 v (ξ1 , x2 ) dξ1 ≤ dξ1 x1 |D1 v (x)| dx1 ,
x1 x1 ξ1 0
R1 1
Integrating the above inequality over Ω, we obtain (because − 0
x1 ln x1 dx1 = ):
4
2 1 2 1 2 2
kvk2,µ ≤ kD1 vk2,µ ≤ kD1 vk2,µ + kD2 vk2,µ ,
4 4
1
1,2
from where we deduce the estimate (15) by density of C0,Γ Ω in W0,Γ (Ω, µ).
For the proof of (16) , let v = (v1 , v2 ) ∈ V . Then, from the equality:
D1 (x1 v1 ) + x1 D1 v1 = v1 + x1 (D1 v1 + D2 v2 ) = 0.
We deduce
v1 1/2
1/2
= −x1 (D1 v1 + D2 v2 ) .
x1
American Journal of Applied Mathematics 2022; 10(4): 141-159 146
Therefore, we obtain
Z 2 Z
v1 2 2
2 2
2
x1 dx ≤ 2 (D1 v1 ) + (D2 v2 ) x1 dx ≤ 2 |v1 |1,2,µ + |v2 |1,2,µ = 2 |v|1,2,µ
Ω x1 Ω
Let’s rematk that, if p = 2 then the estimate (17) is true for all 1 < q < 6.
Remark 2.1. The inequality in (15) is a Poincaré-type inequality. It can be obtained for a more general weight function, see
([6], p. 5).
So, the inequalities (15) and (17) justify the existence of the following constants:
kvk2,µ kvk4,µ
P = sup , S= sup , (18)
v∈WΓ1,2 (Ω,µ) |v|1,2,µ v∈WΓ1,2 (Ω,µ) |v|1,2,µ
ν1 = ν2 = ν, ν3 = λ.
R
Integrate by parts the terms Ωk
Di (x1 Di uj ) vj dx,
i = 1, 2.
For i = 1, we obtain:
Z Z 1 Z
D1 (x1 D1 uj ) vj dx = x1,k D1 uj (x1,k , x2 ) × vj (x1,k , x2 ) dx2 − D1 uj (x1 , x2 ) D1 vj (x1 , x2 ) x1 dx.
Ωk 0 Ωk
Let’s now pass to the limit in the above equality when k → ∞, by using the Lemma 2.1. We derive
Z Z
− D1 (x1 D1 uj ) vj dx = D1 uj (x1 , x2 ) D1 vj (x1 , x2 ) x1 dx. (21)
Ω Ω
147 Rachid Ghenji and Mohamed El Hatri: Axisymmetric Problem of Stationnary Navier-Stokes Equations
Coupled with the Heat Equation
For i = 2, we obtain
Z Z 1 Z
1
D2 (x1 D2 uj ) vj dx = x1 D2 uj (x1 , x2 ) vj (x1 , x2 )|x2 =0 dx1 − D2 uj (x1 , x2 ) D2 vj (x1 , x2 ) x1 dx
Ωk x1,k Ωk
Z
=− D2 uj × (x1 , x2 ) D2 vj (x1 , x2 ) x1 dx,
Ωk
because v (x1 , 0) = v (x1 , 1) = 0, ∀x1 ∈ (0, 1) . Then, after passing to the limite when k → ∞, we derive
Z Z
− D2 (x1 D2 uj ) vj dx = D2 uj (x1 , x2 ) D2 vj (x1 , x2 ) x1 dx. (22)
Ω Ω
On the other hand, because of the equality
Z Z
∂(x1 v1 ) ∂(x1 v2 ) 1 ∂(x1 v1 ) ∂v2
p + dx = p + x1 dx = 0,
Ω ∂x1 ∂x2 Ω x1 ∂x1 ∂x2
Therefore, by combining the equalities (21), (22) and (23) with the equality (20), we get
Z X 3 Z
2 X
b(u, U, V) = (u · ∇) U · Vdµ (x) = ui (Di uj ) vj dµ (x) , (26)
Ω i=1 j=1 Ω
Z Z
(f (U) , V) = β0 u3 v1 dµ (x) + gv3 dµ (x) . (27)
Ω Ω
Definition 3.1. We say that a function U ∈ W is a weak solution of the problem (12) − (13), if U verify the integral equation
(24) for all V ∈ W.
Let’s remark that b(·,·, ·) is a trilinear form on the space W.
Proposition 3.1. For all U = (u, u3 ) ∈ W and V, W ∈ W we have the following estimates:
2 2
ν0 |U|1,2,µ ≤ ar (U, U) ≤ 3λ0 |U|1,2,µ , (28)
√
|b (u, V, W)| ≤ 2S 2 |u|1,2,µ |V|1,2,µ |W|1,2,µ (29)
|(f (U) , V)| ≤ β0 P 2 |U|1,2,µ + γ |V|1,2,µ , (30)
where
R
gvdµ (x)
Ω
ν0 = min (ν, λ) , λ0 = max (ν, λ) , γ = sup , g ∈ L2 (Ω, µ) . (31)
06=v∈W 1,2 (Ω,µ) |v|1,2,µ
0
Z 2 X 3 2 XZ 3
X 2 u 1 2
X 2 2
ar (U, U) = νj (Di uj ) + 2 dµ (x) ≥ ν0 (Di uj ) dµ (x) = ν0 |uj |1,2,µ = ν0 |U|1,2,µ .
Ω i=1 j=1
x 1 i,j Ω j=1
American Journal of Applied Mathematics 2022; 10(4): 141-159 148
Therefore, by combining the above inequality with the previous one, we obtain (28) .
Proof of (29). Let’s now apply the generalized Hölder’s inequality to the second side of the following equality:
1/4 1/2 1/4
x1 ui Di vj wj = x1 ui x1 Di vj x1 wj .
Therefore, according to the Cauchy-Schwarz inequality, we deduce from the inequality above
1/2 1/2
2
X X3 3
X
2 2
|b (u, V, W)| ≤ S 2 |ui |1,2,µ |vj | 1,2,µ
× |wj |1,2,µ
i=1 j=1 j=1
P2 √
= S2 i=1 |ui |1,2,µ |V|1,2,µ |W|1,2,µ ≤ 2S 2 |u|1,2,µ |V|1,2,µ |W|1,2,µ .
Proof of (30) . Let U = (u1 , u2 , u3 ) ∈ W and V = (v1 , v2 , v3 ) ∈ W. According to definition (27) and the inequality of
Hölder taking and the first inequality in (18), we obtain
Z R
Z
g (x) v3 dµ (x)
Ω
|(f (U) , V)| ≤ β0 |u3 v1 | dµ + g (x) v3 dµ ≤ β0 ku3 k2,µ kv1 k2,µ + sup
|v3 |1,2,µ
Ω Ω v3 6=0 |v3 |1,2,µ
≤ β0 P 2 |u3 |1,2,µ |v1 |1,2,µ + γ |v3 |1,2,µ ≤ β0 P 2 |U|1,2,µ + γ |V|1,2,µ .
Lemma 3.1. The bilinear form ar (·, ·) : W × W → R is continuous, i.e. there is a constant C = 3λ0 such that
Proof Let U ∈ W and V ∈ W. So, according to the inequalities of Hölder and Cauchy-Schwarz and the estimate (16), we
obtain
3 2 Z Z
X X u v
1 1
|ar (U, V)| =
Di uj Di vj dµ (x) + ν 2 dµ (x) ≤
j=1 i=1 Ω Ω x1
X3 X 2
≤ λ0 kDi uj k2,µ kDi vj k2,µ + 2 |u|1,2,µ |v|1,2,µ ≤
j=1 i=1
P3 P 1/2 P 1/2
2 2 2 2
≤ λ0 j=1 i=1 kDi uj k2,µ i=1 kDi vj k2,µ + +2 |u|1,2,µ |v|1,2,µ =
X3
= λ0 |uj | 1,2,µ |v j | 1,2,µ + 2 |u| 1,2,µ |v|1,2,µ
≤ λ0 |U| |V| + 2 |u|
1,2 1,2 1,2,µ |v| 1,2,µ ≤ 3λ0 |U|1,2 |V|1,2 .
j=1
149 Rachid Ghenji and Mohamed El Hatri: Axisymmetric Problem of Stationnary Navier-Stokes Equations
Coupled with the Heat Equation
b (u, V, V) = 0, (33)
Z 1 Z Z 1 Z
1 1 1 1
= x1 u1 vj2 x Ddx2 −
1 (x 1 u 1 ) v 2
j dx + x 1 u2 v 2
j (x 1 , x2 ) dx1 − D2 (x 1 u2 ) v 2
j dx
2 1 =0 2 x2 =0
0 Ω 0 Ω
Z
1 1
=− x1 D1 (x1 u1 ) + D2 u2 vj2 dx = 0.
2 Ω x1
In the last equality above, we used the fact that (x1 u1 ) (0, x2 ) = 0 (since u1 ∈ C 1 Ω ) and u2 (x1 , 0) = 0 and the condition
1
D1 (x1 u1 ) + D2 u2 = 0. So, we get (33) by density of V (resp. W) in V (resp. in W).
x1
From the property (33) we deduce (34). Indeed, according to the trilinearity of b (·, ·, ·), we obtain
According to the Hölder’s inequality and the first equality in (18), we have
Z
u3 u2 dµ (x) ≤ ku3 k ku2 k ≤ P 2 |u3 |
2,µ 2,µ 1,2,µ |u2 |1,2,µ . (41)
Ω
So, according to the estimates (28) , (29) and (30), we derive from (40) taking into account the estimates (16) , (29) and
(41) :
2 √
ν0 U1,2,µ ≤ ar U, U ≤ S 2 6 |u|1,2,µ U1,2,µ |U∗ |1,2,µ + β0 P 2 |u3 |1,2,µ |u2 |1,2,µ
√ √ !
√ 2 2 2 γS 2
2 2 2 γS 2
2 2
≤ 2S U1,2,µ |U∗ |1,2,µ + β0 P 2 U1,2,µ ≤ + β0 P 2 U1,2,µ = + β0 P 2 U1,2
U
ν0 − P 2 β0 1,2,µ ν 0 − P 2 β0
V ∈ Wm , ∀U = (u, u3 ) ∈ Wm , ∀V ∈ Wm .
Let us mention that the existence of such element Pm (U ) ∈ Wm follows from Riesz’s theorem, because the left side of (46)
is actually a scalar product in Wm and the right hand side in (46) is a bounded linear functional of V ∈ Wm .
∞
Lemma 5.2. The operator Pm defined by (46) is continuous on Wm , i.e. for all sequence {Un }n=1 in Wm convergent to
U0 ∈ Wm :
lim |Un −U0 |1,2,µ = 0,
n→∞
Therefore, by combining the estimates (49) − (51) with the equality (48), we derive
|ar (P (Un ) − P (U0 ) , V)| ≤ |a (Un −U0 , V)| + |b (un −u0 , Un , V)|
i.e.
lim ar (P (Un ) − P (U0 ) , V) = 0, ∀V ∈ Wm .
n→∞
Because the space Wm is finite, this is enough to establish the convergence lim P (Un ) = P (U0 ).
n→∞
Now we will try to use Lemma 5.1. First, we are looking the condition (44) . According to the apriori estimates (28) and (30),
we obtain
2
[Pm (U) , U] = ar (Pm (U) , U) = ar (U, U) − (f (U) , U) ≥ ν0 |U|1,2,µ − β0 P 2 |U|1,2,µ + γ |U|1,2,µ
2
= ν0 − β0 P 2 |U|1,2,µ − γ |U|1,2,µ = ν0 − β0 P 2 |U|1,2,µ − γ |U|1,2,µ
i.e.
ν0 − β0 P 2 |U|1,2,µ − γ |U|1,2,µ .
ar (Pm (U) , U) ≥
The above inequality implies that the condition (44) is fulfilled for |U|1,2,µ = r, if
γ
r> ,
ν0 − β0 P 2
which is true for large enough r.
Thus, according to Lemma 5.1, we deduce that there exists a solution Um of the form (42) verifying the system (43) .
American Journal of Applied Mathematics 2022; 10(4): 141-159 152
Let’s now multiply (43) by ξk,m and sum the obtained equalities for k = 1, . . . , m. We obtain
and
lim kUm0 − Uk2,µ = 0. (53)
m0 →∞
∞
Lemma 5.3. If {Un = (u1,n , u2,n , u3,n )}n=1 be a sequence weakly converges to a function U ∈ W, strongly convergent in
3
(L2 (Ω, µ)) and uniformely bounded in W, then
P R R
=− i,j Ω
(un,i − ui ) un,j Di vj dµ+ + Ω
ui (un,j − uj ) Di vj dµ + b (u, U, V) . (56)
As V is smooth, Un is uniformely bounded, then according to the Hölder inequality and the strongly convergence of (Un ) in
3
(L2 (Ω, µ)) , we deduce
Z
(un,i − ui ) un,j Di vj dµ ≤ kun,i − ui k kun,i k kDi vj (x)k
2,µ 2,µ ∞,0 → 0. (57)
n→∞
Ω
Similarly, we have
Z
ui (un,j − uj ) Di vj dµ ≤ kun,j − uj k kun,i k kDi vj (x)k → 0. (58)
2,µ 2,µ ∞,0 n→∞
Ω
Therefore, by combining the inequalities (57) and (47) with the equality (56), we derive (54) .
According to the Hölder’s, we obtain
Z
|(f (Un ) , V) − (f (U) , V)| ≤ β0 |un,3 − u3 | |v2 | dµ ≤ β0 P 2 |un,3 − u3 |1,2,µ |v3 |1,2,µ → 0, n → ∞,
Ω
6. Numerical Solution Note that given the geometric simplicity of our domain Ω,
we have
In this paragraph, we will establish the existence and [
uniqueness of a numerical solution, obtained by the finite Ω= K, K = K ∪ ∂K.
element method, approaching the weak solution of the K∈Th
variational problem (24) .
6.2. Definition of the Finite Element Approximation Space
6.1. Finite Element Method
At the triangulation Th we associate the space Vh of
Let Th be a regular triangulation of Ω formed by triangles 1
functions vh ∈ C0,Γ Ω whose restriction vK = vh |K ∈ PK ,
Ki ⊂ Ω, i = 1, . . . , N of diameter hKi , see [7], and let put such that
h = max hK → 0,
Vh = vh ∈ C0,Γ1
Ω : vhK = vh |K ∈ PK .
K∈Th
where hK = max |x − y| . Note that the space Vh is a subspace of the weighted Sobolev
x,y∈K 1,2 1,2
Let Uh be the set of vertices of all the triangles K ∈ Th . space W0,Γ (Ω, µ) : Vh ⊂ W0,Γ (Ω, µ). In the following, we
◦ consider the space Vh of functions vh ∈ Vh × Vh such that
We denote by U h,Γ the set of vertices belonging to the interior Z
of the domain ΩΓ = ΩΓ. Here, the nodes located on the
divr vh dx = 0 ∀K ∈ Th (62)
boundary Γ0 = {(0, x2 ) : 0 < x2 < 1} of the boundary ∂Ω K
◦
belong to the set U h,Γ . and we put Wh = Vh × Vh which is a subspace
◦ 3
1,2
Here, any triangle K ∈ Th has nonempty interior K and of W0,Γ (Ω, µ) , which is called a finite element
Lipschit boundary ∂K, moreover, any two triangles K and K 0 approximation space. The space Wh is a Hilbert space with
◦ ◦
satisfies the condition K∩K 0 6= ∅ and no vertex of one belongs respect of the induced scalar product:
to a side of the other. 2 X
3 Z
Let P1 (K) be the space of polynomials of first degree
X
((U, V)) = Di uj Di vj dµ.
defined on K. It is shown that there exist unique linearly i=1 j=1 Ω
independent polynomials λK,i (x) ∈ P1 (K) , i = 1, 2, 3, such
that Let τB ⊂ Th be the set of all the elements of Th having
3 ◦
X a common vertex B ∈ U h,Γ . Consider a continuous function
λK,i (aj ) = δij , λK,i (x) = 1. (60)
whB on τB such that:
i=1
(i) its restriction whB |K = whB,K on each element K ∈
called nodal functions or barycentric functions, where δij is τB is a first degree polynomial whB,K ∈ PK ,
the Kronecker symbol, see ◦
We call finite element the triple (K, PK , ΣK ) , where ΣK = (ii) whB (B) = 1 and whB (M ) = 0 ∀M ∈ U h,Γ , M 6= B.
3 According to the properties (60), we deduce that for all
{λK,i }i=1 .
Uh = (uh , uh,3 ) ∈ Wh we have
In the following instead of P1 (K) we simply write PK . In
this way any polynomial p ∈ PK can be written in the form
3
X
p (x) = λK,i (x) p (ai,K ) . (61)
i=1
X X X
Uh = Uh (B) whB = uh (B) whB , uh,3 (B) whB (63)
◦ ◦ ◦
B∈U h,Γ B∈U h,Γ B∈U h,Γ
Note that the approximation space Wh is not a subspace of the energy space W, Vh W, since for everything vh (x) ∈ Vh
the divergence divr vh (x) is not necessarily zero on Ω.
Proposition 6.1. For all vh ∈ Vh , there is a constant P > 0 no depending of h such that
where ν1 = cos (x1 , ν) is the exterior normal to ∂K. The scalar function xα 2
1 vK is continuous on Ω, and therefore, on each side
of the element K ∈ Th , moreover, it cancels out on Γ and on the boundary Γ0 = {(0, x2 ) : 0 < x2 < 1} . In this way, if γ is a
side common to two adjacent elements K and K 0 of Th , then we will have
2 2
x 1 vK ν1 γ∈K = − x1 vK ν1 γ∈K 0 .
Therefore,
X Z
2
x 1 vK ν1 dx = 0.
K∈Th ∂K
Z 2 3+ε (1+ε)/2 Z 1
(1−ε)/2
1/4 1+ε 2
× x1 |vh | dx ≤ x12(1−ε) |vh | dx
Ω Ω
Z 2 3+ε (1+ε)/2p
1/4 1+ε p (1+ε)/2p0
× x1 |vh | dx × (mes (Ω)) ,
Ω
0
where 1/p + 1/p = 1. Therefore, taking p = 2 (1 + ε) / (3 + ε) in the above inequality, we get
4 1−ε 3+ε
kvh k0,4,µ ≤ kvh k0,2,µ kvh k0,4,µ , mes (Ω) = 1,
155 Rachid Ghenji and Mohamed El Hatri: Axisymmetric Problem of Stationnary Navier-Stokes Equations
Coupled with the Heat Equation
Therefore
kvh k0,4,µ kvh k0,4,µ
≤ sup =S≤P
|vh |1,2,µ vh 6=0 |vh |1,2,µ
where
2 X
3 Z
X X 1
b0 (uh , Vh , Wh ) = uK,i (Di vK,j ) wK,j dµ,
2 K
K∈Th i=1 j=1
2 X
3 Z
00
X X 1
b (uh , Vh , Wh ) = − uK,i vK,j (Di wK,j ) dµ.
2 K
K∈Th i=1 j=1
Discreet problem: we search Uh = (uh , uh,3 ) ∈ Wh of the form ( 63) such that
Proposition 6.3. Let Uh ∈ Wh be a solution of (68) and let that the condition (35) holds. Then
γ
|Uh |1,2,µ ≤ , (69)
ν0 − β0 P 2
where R
gvdµ (x)
Ω
γ = sup , g ∈ L2 (Ω, µ)
06=v∈Vh |v|1,2,µ
Proof Let Uh = (uh,1 , uh,2 , uh,3 ) ∈ Wh verify (65). Then, by property (70), we deduce that
Z Z
ar (Uh , Uh ) = (f (uh,3 ) , Uh ) = β0 uh,3 uh,2 dµ + guh,3 dµ.
Ω Ω
Hence, by Hölder’s inequality, we get
Z
2 2
ν0 |Uh |1,2,µ ≤ ar (Uh , Uh ) = (f (uh,3 ) , Uh ) ≤ β0 kUh k0,2,µ + guh,3 dµ
Ω
R
2
gvh,3 dµ (x) 2
≤ β0 kUh k0,2,µ + sup Ω
|uh,3 |1,2,µ ≤ β0 P 2 |Uh |1,2,µ + γ |Uh |1,2,µ .
06=vh,3 ∈Vh |vh,3 |1,2,µ
From where, we deduce
ν0 − β0 P 2 |Uh |1,2,µ ≤ γ.
American Journal of Applied Mathematics 2022; 10(4): 141-159 156
∗ |hlh (Uh ) , Vh i| ν0 γ
klh (Uh )k = sup ≤ . (71)
Vh 6=0 |Vh |1,2,µ ν0 − β0 P 2
Proof We have
|hlh (Uh ) , Vh i| = |(f (uh,3 ) , Vh )| ≤ β0 kUh k0,2,µ kVh k0,2,µ + γ |vh,3 |1,2,µ ≤ β0 P 2 |Uh |1,2,µ |Vh |1,2,µ + γ |Vh |1,2,µ .
Then
∗ ||(f (uh,3 ) , Vh )|| γβ0 P 2 ν0 γ
klh (Uh )k = sup ≤ β0 P 2 |Uh |1,2,µ + γ ≤ 2
+γ =
Vh 6=0 |Vh |1,2,µ ν0 − β 0 P ν0 − β0 P 2
2 3
2 1,2 3 1,2
Proposition 6.5. For all uh ∈ (Vh ) ⊂ W0,Γ (Ω, µ) and Vh , Wh ∈ (Vh ) ⊂ W0,Γ (Ω, µ) , we have
√
S2 2
|bh (uh , Vh , Wh )| ≤ |uh |1,2,µ |Vh |1,2,µ |Wh |1,2,µ , (72)
2
where S is the Sobolev constant, see (18) .
Proof The estimate of the term b0 (uh , Vh , Wh ). Let us apply Hölder’s generalized inequality to the right side of
the following equality:
Z Z
1/4 1/2 1/4
uK,i (Di VK,j ) WK,j dµ = x1 uK,i x1 Di vK,j × x1 wK,j .
K K
We find Z Z 1/4
uK,i (Di vK,j ) wK,j dµ ≤ 4
|uK,i | dµ (x) ×
K K
Z 1/2 Z 1/4
2 4
× |Di vK,j | dµ (x) × |vK,j | dµ (x) ≤ kuK,i k0,4,µ |vK,j |1,2,µ kwK,j k0,4,µ .
K K
where Z 1/2
2 2
|ϕK |1,2,µ = |D1 ϕK | + |D2 ϕK | dµ .
K
The estimate of the term b00 (uh , Vh , Wh ). We proceed in the same way as in the previous case. We have
Z Z
uK,i vK,j (Di wK,j ) dµ = 1/4 1/4 1/2
x1 uK,i x1 vK,j x1 Di wK,j
K K
Z
uh,3 uh,2 dµ (x) ≤ kuh,3 k kuh,2 k ≤ P 2 |uh,3 |
2,µ 2,µ 1,2,µ |uh,2 |1,2,µ . (75)
Ω
[10] M. El Hatri, Estimation d’Erreur Optimale et de type [15] J. Neustupa, Axysimetrical flow of Navier-Stokes fluid
Superconvergence de la Méthode des Eléments Finis pour in the whole space with non-zero angular velocity
un Problème aux Limites Dégénérés, M2AN, Tome 21, component, Math. Bohemica, N◦ 2, 2001, pp. 469-481.
n◦ EndExpansion1 (1987), p. 27 − 61.
[16] C. Pérez, J-M. Thomas, S. Blancher, R. Creff, The
[11] M. El Hatri, R. Ghenji1 and N. Popivanov, coupled steady Navier-Stokes/Energy equations problem
Axisymmetric problem of non-stationary Navier- with temperature-dependent viscosity in open channel
Stokes equations coupled with the heat equation . flows, Part 1: Physical model and Analysis of the
AIP Conference Proceedings 2333, 120005 (2021); continuous problem, 2014.
https://doi.org/10.1063/5.0041937
[17] Q. Sylvain, Les centrales Solaires à Concentration,
[12] A.A. Hachicha, Numerical modelling of a parabolic Faculté des Scienses Appliquées, Université de Liège,
trough solar collector, Thesis, Universitat Politècnica de 2007.
Catalunya, 2013, pp.99. [18] R. Temam, Navier-Stokes equations, Theory and
numerical analysis, Elsevier-North Holland.1979.
[13] A. Kufner, O. John, S. Fučik, Function Spaces,
Academia, Prague, 1977. [19] F.M White, Viscous Fluid Flow, Third Edition, Ed.
McGraw Hill, 2006.
[14] B. Mercier and G. Raugel, Résolution d’un problème aux
limites dans un ouvert axisymétrique par éléments finis en [20] Z. Zhang, A pointwise regularity criterion for axysymetric
r, z et série de Fourier en θ. RAIRO, Numerical Analysis Navier-Stokes system, J. Math Anal. and Appl. 461
(VoL. 16, n◦ 4, p. 405-4061). (2018) 1-6.