Beruflich Dokumente
Kultur Dokumente
ABSTRACT
In the paper, it is constructed with difference schemes which approximately mixed problem for heat equation and shown
their stability. There exist various methods to develop difference schemes which are mainly based on exchanging
derivatives with difference schemes.
KEYWORDS: Difference Schemes, Wave Equation, Approximation, Finite Difference, Stability, Boundary Conditions
Article History
Received: 10 Apr 2019 | Revised: 05 Oct 2019 | Accepted: 12 Nov 2019
INTRODUCTION
Preliminaries
Problem 1: We consider the following mixed problem for the equation below
u tt − u xx − u yy = 0 (1)
For y = 0, (t , x) ∈ R+2
ut − a 2 u x − b2 u y = 0 , (3)
In the domain R+
3
= { (t , x, y) t , x, y > 0}.
Here a1 , b1 , a2 , b2 ∈ R and r → 0
u t = o( r − 2 ), u x = o( r − 2 ), u y = o( r − 2 ), r = x 2 + y 2 .
1 1 1
www.iaset.us editor@iaset.us
38 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
• x = 0 in a1 > 0, b1 < 1 ;
• y = 0 in a2 > 0, b2 < 1 ;
π
e 2ξ utt − uθθ − uξξ = 0, for t > 0, 0 < θ < , ξ ∈ R1 (5)
2
π
eξ ut + a1uθ − b1uξ = 0 , for θ = , t > 0, ξ ∈ R 1 (6)
2
π
Then boundary conditions, t = 0, 0 < θ <
2
− 12ξ 2ξ 2ξ
In ξ → ∞ an initial conditions ut = o(e ), uθ = o(e− ), uξ = o(e −
1 1
) . Obtained problem (5)-(8) in the
domain t > 0, 0 < θ < π , ξ ∈ R1 is reduced to the following mixed problem consisting of system of symmetric t-hyperbolic
2
equations
ξ ∂ ∂ ∂
e A0 − B0 − C0 + Q0 V = 0 (9)
∂t ∂θ ∂ξ
Boundary conditions
π
θ= , t > 0, ξ ∈ R1 for ϑ1 + a1ϑ2 − b1ϑ3 = 0 , (10)
2
π ′
Initial condition, for t = 0, 0 < θ < , ξ ∈ R V = ( eξψɶ (θ , ξ ), ϕɶθ′ (θ , ξ ), ϕɶξ′ (θ , ξ ) ) and when
2
ξ → ∞ V = o (e ξ )
1
2
(12)
Here
k l m l k 0 m 0 k m 0 0
A0 = l k 0 , B0 = k l m , C0 = 0 − m l , Q0 = 0 0 0 ,
m 0 k 0 m −l k l m k 0 0
ϑ1 e ut
ξ
V = ϑ2 = uθ ,
ϑ u
3 ξ
µξ
according to [8], taking notation V = e Y , we get equality of dissipative energy integral
d
eξ ( A0Y , Y )t − ( B0Y , Y )θ − (C0Y , Y )ξ + ([Q0 + Q0* − 2 µC0 + B0 ]Y , Y ) = 0, (13)
dθ
d ξ
dt Π
∫∫ e ( A0Y , Y ) d ξ dθ − ∫
R
{( B Y ,Y )
0 θ = π2 }
− ( B0Y , Y ) θ = 0 d ξ +
d
+ ∫∫ Q0 + Q0* − C0 + B0 Y , Y d ξ dθ = 0
Π
dθ
H (θ ) O
• A0 (θ ) = T0* T0 ;
O H (θ )
O − H (θ )
• B0 (θ ) = T0* T0 ;
− H (θ ) O
− H (θ ) O
• C0 (θ ) = T0* T0 ;
O H (θ )
2m(θ ) 0 k (θ )
− H (θ ) L(θ )
• Q0 (θ ) + Q (θ ) = 0
*
0 0 = T0* T0 ;
0
k (θ ) 0 L (θ ) H (θ )
0
0 0 − l (θ ) m(θ )
O = , L(θ ) = ,
0 0 m(θ ) l (θ )
www.iaset.us editor@iaset.us
40 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
k (θ ) − m(θ ) − l (θ )
H (θ ) = ,
− l (θ ) k (θ ) + m(θ )
1 0 − 1
1 0 −1 0
T0 =
2 0 − 1 0
1 0 1
w1
W1 w2
W = = = T0V
W2 w3
w
4
d − H L − H O O − H ′
Q0 + Q0* − C0 + B0 Y , Y = e −ξ − + V ,V
dθ L
H O H − H ′ O
d
If H′ = H = L , that’s if ( k − m )′ = −l , ( k + m)′ = l , − l ′ = m , then square form is equal to 0.
dθ
Finding the solution of these differential equations, we have
We rewrite Boundary
π
condition (10) as W1 = SW2 when θ = , where
2
2a1 1 − b1
−
S = 1 + b1 1 + b1 .
1 0
([
− (B0Y , Y ) θ = π = e −ξ S * H + HS W2 ,W2 ] )
θ=
π
2 2
2a 2 1 − b2
− −
R = 1 + b2 1 + b2 .
1 0
For n + 1 , solutions are obtained by a formula explicitly while the solutions for n of the difference schemes are
known, this scheme is called explicit. Even though samples of some difference schemes to be considered look like the
samples of explicit schemes, they are not actually.
Explicit Right Difference Scheme: To solve problem-1 numerically, we employ explicit right difference scheme which
approximates differential problem. For this, we rewrite the system (9) in the following form (7):
∂Y ∂[B0Y ] ∂Y d
e ξ A0 − − C0 + Q0 − µC 0 + B0 Y = 0 , (14)
∂t ∂θ ∂ξ dθ
∂Y ∂Y ∂Y
eξ A0 − B0 − C0 + [Q0 − µC0 ]Y = 0 (15)
∂t ∂θ ∂ξ
We multiply the systems (14) and (15) by D = diag ( y1 , y2 , y3 ) from left side. Adding obtained systems, we find
∂Y ∂Y ∂[ B0Y ] ∂Y ∂Y
eξ DA0 + eξ DA0 −D − DB0 − DC0 − (16)
∂t ∂t ∂θ ∂θ ∂ξ
∂Y d
− DC0 + D[Q0 − µC0 + B ]Y + D[Q0 − µC0 ]Y = 0.
∂ξ dθ
∆t = ∆ t , ∆θ = ∆ θ , ∆ξ = ∆ ξ .
Yijn = Y (n ∆t , i ∆θ , j ∆ξ ) = ( y1 (n ∆t , i ∆θ , j ∆ξ ), y2 (n ∆ t , i ∆θ , j ∆ξ ), y3 (n ∆ t , i ∆θ , j ∆ξ ))′,
i = 0, I , n, j = 0,1,…;
I −1 ∞
ξj
= ∆θ ∆ ξ ∑ ∑e
2
Yn ( A0Yijn , Yijn ) ,
A0
i = 0 j = −∞
www.iaset.us editor@iaset.us
42 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
1
L = (1, 1, 1)′ , µ = .
2
Using the above notations, we develop difference scheme approximating (16):
for i = 0 , j = 0,1,2,…
( y1 ) 0n j − a2 ( y 2 ) 0n j − b2 ( y3 ) 0n j = 0 , (18)
for i = I , j = 0,1,2,…
( y1 ) nI j + a1 ( y2 ) nI j − b1 ( y3 ) nI j = 0 , (19)
′
1ξ
( 1ξ 1ξ
Yij0 = e 2 jψɶ (ξ j , θ i ), e 2 j ϕɶθ′ (ξ j , θ i ), e 2 j ϕɶξ′ (ξ j , θ i ) . ) (20)
Sample of this difference scheme as shown in figure-1.1, consists of the systems of equations which are not linear.
It is easily seen that, this scheme approximates differential equation with first order. The boundary conditions are precisely
approximated. As we mentioned above to compute schemes approximation error exact solution is calculated by the scheme
(17) and we denote this error δ fh as the norm of the vector E (t n , θ i , ξ j ) . Approximation of difference scheme for the
sample equation is shown in the figures 2-4 and table-1. Here t10 = 0.3, t20 = 0.6, t30 = 0.9, θ 2 = 0.524, ξ6 = 3.2 ,
www.iaset.us editor@iaset.us
44 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
Theorem 1: If for (18)–(19) it holds Shapiro–Lopatinski condition, that’s if a1 > 0, b1 < 1 any a2 > 0, b2 < 1 , then difference
Proof. We multiply (17) by vector L and for convenience, we perform this operation for each variable separately:
∆t
n +1
) ( )
=
1
∆t
{(( A ) (Y
0 i
n +1
ij ) ( ( A ) (Y
− Yijn ) , Yijn + 0 i
n +1
ij − Yijn ) , Yijn +1 )} =
=
1
∆t
{(( A ) Y
0 i
n +1
ij ) ( ) ( ) (
, Yijn − ( A0 )i Yijn , Yijn + ( A0 )i Yijn+1 , Yijn+1 − ( A0 )i Yijn , Yijn+1 = )}
=
1
∆t
(( A ) Y0 i
n +1
ij )
, Yijn+1 −
1
∆t
(( A ) Y
0 i
n
ij )
, Yijn =
1
∆t
( A0Y , Y )ijn+1 − 1 ( A0Y , Y )ijn ;
∆t
+
1
∆θ
(
[B0Y ]in+1 j , Yi +n1 j − Yijn = 1
∆θ
) (((B ) Yn
0 i +1 i +1 j ) ( ) (
, Yijn − (B0 )i Yijn , Yijn + (B0 )i +1 Yi +n1 j , Yi +n1 j − )
(
− (B0 )i +1 Yijn , Yi +n1 j = )) 1
∆θ
(B0Y , Y )in+1 j − 1 (B0Y , Y )ijn ;
∆θ
n Y n − Yijn n Y n − Yijn
Dij (C 0 ) ij +1
∆ξ
, L + Dij +1 (C 0 ) ij +1
∆ξ
, L =
1
((
(C 0 )i Yijn+1 , Dijn L − (C 0 )i Yijn , Dijn L + ) ( )
∆ξ
i i
•
( ) (
+ (C 0 )i Yijn+1 , Dijn+1 L − (C 0 )i Yijn , Dijn+1 L =
1
∆ξ
))
(C 0Y , Y )ijn +1 − 1 (C 0Y , Y )ijn ;
∆ξ
n d d
Dij 2Q0 − C0 + B0 Yijn , L = 2Q0 − C0 + B0 Yijn , Dijn L =
dθ i dθ i
•
d d
= 2Q0 − C0 + B0 Yijn , Yijn = Q0 + Q0* − C0 + B0 Yijn , Yijn
dθ i dθ
here it is employed (Q Y n
0 ij ) ((Q Y
, Yijn = 1
2
n
0 ij ) (
, Yijn + Q0*Yijn , Yijn . ))
Summing up above equalities, we get discreet analogue of differential representation of energy integral:
e
ξj 1
∆t
{( A Y , Y )
0
n +1
ij
− ( A0Y , Y )ij −
n
} 1
∆θ
{( B Y , Y )
0
n
i +1 j
− ( B0Y , Y )ij −
n
}
(21)
{(C Y , Y ) }
n
1 d
− − ( C0Y , Y )ij + Q0 + Q0* − C0 + B0 Y , Y = 0
n n
∆ξ ij +1
dθ
0
ij
(
Y n = Yijn , Yijn ) 1
2
→ 0 when ξ → ∞ , we have
2 2
Y n +1 − Yn =
A0 A0
∑ {( B Y , Y ) }
+∞ I −1 ∞ n
d
= ∆θ ⋅ ∆ξ ⋅ − ( B0Y , Y )0 j − ∑ ∑ Q0 + Q0* − C0 +
n n
B0 Y , Y
dθ
0
i = 0 j =−∞
I j
j =−∞ ij
H (θi ) O *
O − H (θ i )
A0 (θi ) = T0* T0 , B0 (θ i ) = T0 T0 ,
O H (θi ) − H (θ i ) O
− H (θ i ) O , − l (θ i )
0 0 , L(θ ) =
m(θ i ) k (θ ) − m(θ i ) − l (θ i )
C0 (θ i ) = T0* T0 O = m(θ ) , H (θi ) = i ,
k (θ i ) + m(θ i )
l (θi ) − l (θ i )
i
O H (θ i ) 0 0 i
2m(θi ) 0 k (θi )
− H (θi ) L(θi )
Q0 (θi ) + Q0* (θi ) = 0 0 0 = T0* T0 ,
k (θ ) 0 L(θi ) H (θi )
i 0
1 0 − 1
1 0 −1 0
T0 = , θ i = i ∆θ , i = 0,1,..., I .
2 0 −1 0
1 0 1
=e
(1− 2 µ ) ξ j
{( H (θ )W ,W ) 1
n
1 ij + ( H (θ )W2 ,W2 )ijn },
e
ξj
(( A ) Y n
0 i ij )
, Yijn > 0 , (22)
If
www.iaset.us editor@iaset.us
46 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
then
d − H (θi ) L(θi )
Q0 + Q0 − C0 + B0 Yijn , Yijn = −
*
′
Hence we find (k (θ ) − m(θ ) ) = −l (θ ), (k (θ ) + m(θ ))′ = l (θ ), − l ′(θ ) = m(θ ) or
− (B0Y , Y )I j = − (B0 )I e
n
( − 12 ξ j
VInj , e
− 12 ξ j
)
VInj = −e
−ξ j
(B0V ,V )nI j =
n
= −e
−ξ j * O
T0
− H (π )
− H ( π2 )
O
T0V , V = e j (H ( π2 )W2 , W1 )nI j + (H ( π2 )W1 , W2 )nI j
−ξ
{ }
2 I j
− ( B0Y , Y ) I j = e ( S H ( )
−ξ j n
) + H ( π2 ) S W2 ,W2
n * π
2 .
Ij
According to lemma D.L. Tkachev, M.V. Gomolina ([7]) if for boundary conditions a1 > 0 and b1 < 1 ,
[S H (
* π
2
) + H ( π2 ) S ]
i= I
>0
− ( B0Y , Y ) I j ≥ 0.
n
(24)
Analogously we rewrite (B0Y , Y )0n j taking into account that boundary condition (18) is (W1 )0n j = R(W2 )0n j as
2a2 1 − b2
where −
R = 1 + b2
− . If a 2 > 0 and b2 < 1 for boundary conditions, then according to Lyuapunov
1 + b2
1 0
[ ]
theorem R * H (0) + H (0) R i = 0 < 0 . Consequently, it yields
( B0Y , Y )0 j ≥ 0.
n
(25)
2 2
Y n +1 ≤ Yn .
A0 A0
− µξ j
From Yijn = e Vijn it is easy to understand that
(( A ) e )
I −1 ∞ I −1 ∞
ξ ξj − 12 ξ j − 12 ξ j
∆θ ∆ξ ∑ ∑ e j ( A0Y , Y )ijn+1 = ∆θ ∆ξ ∑ ∑ e 0 i Vijn +1 , e Vijn +1 =
i =0 j =−∞ i =0 j =−∞
I −1 ∞ I −1 ∞
= ∆θ ∆ξ ∑ ∑ ( ( A0 )i Vijn+1 ,Vijn+1 ) =∆θ ∆ξ ∑ ∑ ( A0V ,V )ijn +1 .
i = 0 j =−∞ i =0 j =−∞
Therefore, it holds
I −1 ∞ I −1 ∞
J n = ∆θ ∆ξ ∑ ∑ ( A V ,V ) 0
n
ij ≤ ∆θ ∆ξ ∑ ∑ ( A V ,V ) 0
0
ij =J 0 (26)
i = 0 j = −∞ i = 0 j = −∞
Theorem is proved.
To solve the first problem, we make use the following difference scheme:
U ijn +1 − U ijn
− (U t )ij = 0.
n
(27)
∆t
The above difference scheme is obtained using U t − U t ≡ 0 . Now we multiply (27) by the vector 2U ijn+1 :
(
2 (U ijn +1 ,U ijn +1 ) − 2 (U ijn , U ijn +1 ) − 2∆t (U t )ij , U ijn +1 = 0
n
)
and from this equality, we have
(U n +1
ij ( n n
)
, U ijn +1 ) − (U ijn , U ijn ) − ∆ t (U t )ij , (U t )ij − ∆t (U ijn +1 , U ijn+1 ) ≤ 0 .
∆θ ∆ξ and is summed up with respect i from 0 till I − 1 , with respect to j from −∞ till +∞ :
www.iaset.us editor@iaset.us
48 M. U. Khudayberganov, S. U. Dadabayev, H. D. Yuldashev & O. M. Raxmonov
I −1 ∞ I −1 ∞
∆θ ∆ξ ∑ ∑ (U
i =0 j =−∞
n +1
ij , U ijn +1 ) ≤ ∆θ ∆ξ ∑ ∑ (U
i = 0 j =−∞
n
ij , U ijn ) +
∑ {( (U ) , (U ) ) + (U }
I −1 ∞
+ ∆ t ∆θ ∆ξ ∑ n +1
,U ijn +1 ) .
n n
t ij t ij ij
i = 0 j =−∞
+ ∆ t (U t )
n +1
U n +1 ≤ Un + ∆ t U n +1 . (28)
A0 A0 A0 A0
Writing (26) as
I −1 +∞
n +1
C1 ∆θ ∆ξ ∑ ∑ (U t ,U t )ij + (U x , U x )ij + (U y , U y ) ≤ J n ≤
n +1 n +1
ij
i = 0 j =−∞
I −1 +∞
n
≤ C2 ∆θ ∆ξ ∑ ∑ (U t , U t )ij + (U x , U x )ij + (U y , U y )ij
n n
i = 0 j =−∞
2 2
U n +1 ≤ const U n
A1 A1
REFERENCES
1. Aloev R.D. The method of dissipative energy integrals for difference schemes. Novosib. un-t Novosibirsk,
1993.68s.
2. Aloev R.D., Blokhin A.M., Druzhinin I.Yu. Stability of explicit difference schemes for symmetric t-hyperbolic systems.
// Proceedings of the seminar of academician S.L. Soboleva.IM SB RAS, Novosibirsk. 1986. No. 1.-p. 26–39
3. Blokhin A.M. Energy integrals and their applications to gas dynamics problems. Novosibirsk: Science. 1986.-240 p.
4. Volkov Yu.A., Markov M.B. Numerical solution of vector wave equations. KIAM Preprint No. 26, Moscow, 1996
5. Godunov S.K., Gordienko V.M. The mixed problem for the wave equation. In the book: Partial differential
equations. Seminar proceedings. S.L. Sobolev. 2. Novosibirsk, ed. Institute of Mathematics SB AS, 1977.- p. 5–32.
6. Marchuk N.G. On the existence of solutions to the mixed problem for the vector wave equation. DAN USSR, 1980.
Vol. 252, No. 3 — pp. 546–550.
7. Khudoyberganov M.U. Stability of difference schemes for the wave equation. // Proceedings of the international
scientific conference. Modern problems of mathematical physics and information technology.-Tashkent, 2005.
Volume 2.s. 237-241.
8. Blokhin A.M., Tkachev D.L. Mixed problems for the wave equation in coordinate domains. Nova science
publishers, INC.1998.133p.
9. Falk R.S., Richter G. R. Explicit finite element methods for symmetric hyperbolic equations. SIAM J. Numer. Anal.
Vol. 36, No. 3, pp. 935–952
10. A.M Blokhin, R.D Alaev, M.U Hudayberganov. One Class of Stable Difference Schemes for Hyperbolic System.
American Journal of Numerical Analysis 2 (3), 85–89. 2014.
11. R.D Aloev, S.O Davlatov, Z.K Eshkuvatov, N.M.A.N.Long. Uniqueness solution of the finite elements scheme for
symmetric hyperbolic systems with variable coefficients. Malaysian journal of mathematical sciences 10, 49–60.
2016.
12. A.Rakhmatillo, K.Mirzoali, B.Alexander. Construction and research of adequate computational models for
quasilinear hyperbolic systems. Numerical Algebra, Control & Optimization 8 (3), 287–299. 2018.
www.iaset.us editor@iaset.us