Artikel Decatic B-Spline
Artikel Decatic B-Spline
DOI: 10.1002/num.22747
RESEARCH ARTICLE
KEYWORDS
1 INTRODUCTION
Several approaches are applied for computing the solution of different partial differential equations
such as exponential cubic B-spline scheme [12] and extended cubic B-spline with finite element
technique [21] for solving numerical solution generalized Burgers’ Fisher equation, extended cubic
B-spline for Kdv-Burgers’ equation [22], quartic B-spline scheme for solving MRLW equation [25,
52], MacCormack/Lax–Friedrichs approach to solve shock-capturing [3], optimized composite
Numer Methods Partial Differential Eq. 2021;1–19. wileyonlinelibrary.com/journal/num © 2021 Wiley Periodicals LLC 1
2 JENA AND GEBREMEDHIN
approaches for solving hyperbolic conservative laws [2], Legendre spectral element scheme for
generalized Benjamin–Bona–Mahony–Burgers equation [17], and so on [52].
This article deals with decatic B-spline basis function to approximate the following Burgers’
equation
zt + zzx = azxx , c ≤ x ≤ 𝑑, (1)
where a > 0 is the kinematic viscosity parameter. The initial and boundary conditions of Equation (1)
are shown in Equations (2) and (3), respectively.
z (x, 0) = g (x) (2)
⎫
z (c, t) = g0 (t)⎪
⎪
z (𝑑, t) = g1 (t)⎬ . (3)
⎪
zx (c, t) = zx (𝑑, t) = 0, 0 ≤ t ≤ T ⎪
⎭
The governing Equation (1) is widely known as Burgers’ equation [9] after the introduction by Bateman
[7]. Equation (1) is nonlinear partial differential equation which is applicable in many fields of sciences
such as in approximation theory involving wave propagation in viscous fluid [16], heat conduction and
continuous stochastic process [8], gas dynamics [34], longitudinal elastic waves in an isotropic solid
[42]. Various efforts are taken to obtain the approximate solution of Equation (1) using different numer-
ical schemes such as meshfree quasi-interpolation scheme [33], variational scheme [1], optimized
weighted essentially non-oscillatory third order scheme [5], quasi-linear technique [11], weighted
average differential quadrature scheme [29], sinc differential quadrature scheme [31], septic B-spline
scheme [45], cubic B-spline approaches [14, 51], non-polynomial spline approach [44], hybrid numer-
ical approach [27], quadratic and cubic B-splines finite element scheme [13, 46], finite difference
method [20, 32], finite element scheme [24, 41], automatic differentiation [6], differential quadrature
scheme [36, 37], efficient numerical scheme [39], linearized implicit scheme [40], B-spline Galerkin
methods [15], Haar wavelet quasilinearization scheme [26], multiquadratic quasi-interpolation tech-
nique [10], exponential twice continuously differentiable B-spline scheme [18], quartic B-spline
collocation method [30], implicit and fully implicit exponential finite difference schemes [23], quin-
tic B-spline collocation scheme [47], combinations of the wavelet and finite volume scheme [38],
semi-implicit finite difference approach [43], modified cubic B-spline scheme [35], improvised col-
location scheme with cubic B-spline as basis functions [19], radial basis functions with meshfree
algorithms [28], two meshfree approaches [48], cubic trigonometric B-spline approach [50], meshless
technique by applying Lie group integrator with multiquadric radial basis functions [49], MIEELDLD
technique [4], and so on.
Each method used to solve Equation (1), still involves certain drawbacks like high arithmetic
computations, complexity, lower accuracy, a difficulty for computer programming, and limitations of
certain special cases. For instance, in the finite difference and finite element schemes, a larger grid size
is required to find values at (or around) the point of interest. In the scheme like Galerkin, the selection
of trial functions is a tedious task itself. In each of the B-spline methods, we have observed that the
schemes involve less accuracy in terms of error.
In the present work, the decatic B-spline method is proposed to incorporate the setbacks of other
numerical studies and to obtain a better approximate solution than other methods of Equation (1).
The higher-order derivatives instead of taking the lower order derivatives of the decatic B-spline on
JENA AND GEBREMEDHIN 3
Equation (3) are used to construct additional constraints for obtaining a unique solution of the problem.
The summary of this paper is organized in the following manner: in Section 2, the construction of
decatic B-spline scheme is illustrated. The implementation of the method is explained in Section 3. The
stability analysis of the method is discussed in Section 4. Numerical tests and the results are discussed
in Section 5. In Section 6, conclusions are elaborated.
Let us take a uniform mesh c = x0 < x1 < · · · < xM = d, c, by the mesh points xj with 𝛥x = h = xj+1 −xj =
𝑑−c
M
, j = 0,1,2, … , M.
The numerical solution Z(x, t) to Equation (1) is given by:
∑
M+4
Z (x, t) = 𝜇j (t) 𝜌j (x) . (4)
j=−5
The boundary conditions and collocation form of Equation (1) are used to estimate the known
parameters 𝜇j (t).
And 𝜌j (x) determined as follows:
Let us consider the zero degree B-spline of the form
{ [ )
1, x ∈ xj , xj+1
𝜌j,0 = . (5)
0, otherwise
Equation (5) is Cox de Boor recursion formula for the jth B-Spline basis function of qth degree.
x − xj xj+q+1 − x
𝜌j,q (x) = 𝜌j,q−1 (x) + 𝜌j+1,q−1 (x) , q ≥ 1. (6)
xj+q − xj xj+q+1 − xj+1
Equation (6) starts from first-degree B-splines and builds the functions of successively higher degrees
of B-splines.
For instance the B-spline of degree one (first degree B-spline):
The first degree B-spline has been obtained by putting q = 1 into Equation (6) and using
Equation (5), which gives
x − xj xj+2 − x
𝜌j,1 (x) = 𝜌j,0 (x) + 𝜌j+1,0 (x)
xj+1 − xj xj+2 − xj+1
{ {
x − xj 1, xj ≤ x < xj+1 xj+2 − x 1, xj+1 ≤ x < xj+2
= +
xj+1 − xj 0, otherwise x j+2 − xj+1 0, otherwise
,
⎛x − xj , xj ≤ x < xj+1
1⎜
𝜌j,1 (x) = ⎜xj+2 − x, xj+1 ≤ x < xj+2 , (7)
h⎜
⎝0, otherwise
The decatic B-spline basis functions denoted by 𝜌j (x), j = − 5(1)M + 4 at the knots are also obtained
by putting q = 10 in Equation (5) and then using the nine degree B-spline, yields
⎛p1 , x ∈ Ij−5
⎜ ( )
⎜ 11
⎜p1 − 1 p2 , x ∈ Ij−4
⎜ ( ) ( )
⎜ 11 11
⎜p1 − p2 + p3 , x ∈ Ij−3
⎜ 1 2
⎜ ( ) ( ) ( )
⎜ 11 11 11
⎜p1 − p2 + p3 − p4 , x ∈ Ij−2
⎜ 1 2 3
( ) ( ) ( ) ( )
⎜ 11 11 11 11
⎜p1 − p2 + p3 − p4 + p5 , x ∈ Ij−1
⎜ 1 2 3 5
⎜ ( ) ( ) ( ) ( ) ( )
⎜ 11 11 11 11 11
1 ⎜p6 − p7 + p8 − p9 + p10 + p11 , x ∈ Ij
𝜌j (x) = 10 ⎜ 1 2 3 4 5 , (8)
h ⎜ ( ) ( ) ( ) ( )
⎜p − 11 11 11 11
⎜ 6 p7 + p8 − p9 + p10 , x ∈ Ij+1
1 2 3 4
⎜ ( ) ( ) ( )
⎜ 11 11 11
⎜p6 − p7 + p8 − p9 , x ∈ Ij+2
⎜ 1 2 3
⎜ ( ) ( )
⎜ 11 11
⎜p6 − p7 + p8 , x ∈ Ij+3
⎜ 1 2
( )
⎜ 11
⎜p6 − p7 , x ∈ Ij+4
⎜ 1
⎜
⎜p6 , x ∈ Ij+5
⎜
⎝0, otherwise
(
hZj′ = 10 − 𝜇j−5 − 501𝜇j−4 − 14106𝜇j−3 − 73626𝜇j−2 − 67956𝜇j−1 + 67956𝜇j
)
+ 73626𝜇j+1 + 14106𝜇j+2 + 501𝜇j+3 + 𝜇j+4 , (10)
(
h2 Zj′′ = 90 𝜇j−5 + 245𝜇j−4 + 3800𝜇j−3 + 7280𝜇j−2 − 11326𝜇j−1 − 11326𝜇j
)
+ 7280𝜇j+1 + 3800𝜇j+2 + 245𝜇j+3 + 𝜇j+4 , (11)
JENA AND GEBREMEDHIN 5
(
h3 Zj(3) = 720 − 𝜇j−5 − 117𝜇j−4 − 834𝜇j−3 + 798𝜇j−2 + 2604𝜇j−1 − 2604𝜇j
)
− 798𝜇j+1 + 834𝜇j+2 + 117𝜇j+3 + 𝜇j+4 , (12)
(
h4 Zj(4) = 5040 𝜇j−5 + 53𝜇j−4 + 80𝜇j−3 − 568𝜇j−2 + 434𝜇j−1 + 434𝜇j
)
− 568𝜇j+1 + 80𝜇j+2 + 53𝜇j+3 + 𝜇j+4 , (13)
(
h5 Zj(5) = 30240 − 𝜇j−5 − 21𝜇j−4 + 54𝜇j−3 + 54𝜇j−2 − 276𝜇j−1 + 276𝜇j
)
− 54𝜇j+1 − 54𝜇j+2 + 21𝜇j+3 + 𝜇j+4 , (14)
(
h6 Zj(6) = 151200 𝜇j−5 + 5𝜇j−4 − 40𝜇j−3 + 80𝜇j−2 − 46𝜇j−1 − 46𝜇j + 80𝜇j+1
)
− 40𝜇j+2 + 5𝜇j+3 + 𝜇j+4 , (15)
(
h7 Zj(7) = 604800 − 𝜇j−5 + 3𝜇j−4 + 6𝜇j−3 − 42𝜇j−2 + 84𝜇j−1 − 84𝜇j + 42𝜇j+1
)
− 6𝜇j+2 − 3𝜇j+3 + 𝜇j+4 , (16)
(
h8 Zj(8) = 1814400 𝜇j−5 − 7𝜇j−4 + 20𝜇j−3 − 28𝜇j−2 + 14𝜇j−1 + 14𝜇j − 28𝜇j+1
)
+ 20𝜇j+2 − 7𝜇j+3 + 𝜇j+4 , (17)
(
h9 Zj(9) = 3628800 − 𝜇j−5 + 9𝜇j−4 − 36𝜇j−3 + 84𝜇j−2 − 126𝜇j−1 + 126𝜇j − 84𝜇j+1
)
+ 36𝜇j+2 − 9𝜇j+3 + 𝜇j+4 . (18)
The Crank–Nicolson method and discretization of the time derivative using the finite difference
scheme are applied in Equation (1) for the implementation of the proposed scheme.
( )
zn+1
j − znj (zzx )n+1
j + (zzx )nj (zxx )n+1
j + (zxx )nj
+ −a = 0. (19)
k 2 2
Taylor series expansion is applied for linearizing the nonlinear term in Equation (19) as follows:
[ ] ( )
(zzx )n+1
j = (zzx )nj + k (znt znx )j + (zn znxt )j + o k2
[ { }n+1 { }n ]
n zj (znx )j − (znx zn )j zj (zx )n+1
j {− (zn znx )}j ( )
= (zzx )j + k + + o k2
k k
( )
= zn+1
j (zx )nj + znj (zx )n+1
j − (zzx )nj + o k2 . (20)
Collecting the like terms (labels) after substituting Equation (20) in Equation (19), yields
( )
k k ka ka
1 + (zx )nj zn+1
j + znj (zx )n+1
j − (zxx )n+1
j = znj + (zxx )nj . (21)
2 2 2 2
Arranging the coefficients of 𝜇j−e
n+1
, e = 1 (1) 5 and 𝜇j+l
n+1
, l = 0 (1) 4, respectively with the help of
Equations (9)–(11) in Equation (21) to obtain Equation (22) as follows:
ka
b1 𝜇j−5
n+1
+b2 𝜇j−4
n+1
+b3 𝜇j−3
n+1
+b4 𝜇j−2
n+1
+b5 𝜇j−1
n+1
+b6 𝜇jn+1 +b7 𝜇j+1
n+1
+b8 𝜇j+2
n+1
+b9 𝜇j+3
n+1
+b10 𝜇j+4
n+1
= q1 + q3 ,
2
(22)
6 JENA AND GEBREMEDHIN
where j = 0(1)M
kq2 5kq1 45ka 1013kq2 2505kq1 11025ka
b1 = − − 2 + 1, b2 = − − + 1013
2 h h 2 h h2
70530kq1 171000ka
b3 = 23920kq2 − − + 47840,
h h2
368130kq1 327600ka
b4 = 227596kq2 − − + 455192
h h2
339780kq1 509670ka
b5 = 655177kq2 − + + 1310354,
h h2
339780kq1 509670ka
b6 = 655177kq2 + + + 1310354
h h2
368130kq1 327600ka
b7 = 227596kq2 + − + 455192,
h h2
70530kq1 171000ka
b8 = 23920kq2 + − + 47840
h h2
1013kq2 2505kq1 11025ka kq 5kq1 45ka
b9 = + − 2
+ 1013, b10 = 2 + − 2 + 1.
2 h h 2 h h
The values of q1 , q2 , and q3 are the same with Zj , Zj′ , and Zj′′ given in Equations (9)–(11) at level
n, respectively. The system in Equation (22) consists of (M + 1) equations with (M + 10) unknowns.
Nine additional constraints are required to find a unique solution of Equation (22). The nine additional
constraints are selected from the boundary conditions Equation (3) and using some of the possible
higher derivatives of the decatic B-spline on Equation (3). In general, the additional equations at c = x0
for j = 0 and d = xM , j = M can be represented as Equations (23)–(27) as follows:
⎛ 𝜇n+1 + 1013𝜇j−4
n+1 n+1
+ 47840𝜇j−3 n+1
+ 455192𝜇j−2 ⎞
( ) ⎜ j−5 ⎟ ( )
Z xj , t = ⎜ +1310354𝜇j−1n+1
+ 1310354𝜇jn+1 ⎟ = G xj , t , (23)
⎜ n+1 ⎟
⎝ +455192𝜇j+1 + 47840𝜇j+2 + 1013𝜇j+3 + 𝜇j+4 ⎠
n+1 n+1 n+1
( )
( ) 30240 − 𝜇j−5
n+1 n+1
− 21𝜇j−4 n+1
+ 54𝜇j−3 n+1
+ 54𝜇j−2 n+1
− 276𝜇j−1 + 276𝜇jn+1 − 54𝜇j+1
n+1
Z (5)
xj , t = =0
h5 n+1
− 54𝜇j+2 n+1
+ 21𝜇j+3 + 𝜇j+4
n+1
(24)
( )
( ) 151200 𝜇j−5
n+1 n+1
+ 5𝜇j−4 n+1
− 40𝜇j−3 n+1
+ 80𝜇j−2 n+1
− 46𝜇j−1 − 46𝜇jn+1 + 80𝜇j+1
n+1
Z (6)
xj , t = = 0, (25)
h6 n+1
− 40𝜇j+2 n+1
+ 5𝜇j+3 + 𝜇j+4
n+1
( )
( ) 604800
n+1
−𝜇j−5 n+1
+ 3𝜇j−4 n+1
+ 6𝜇j−3 n+1
− 42𝜇j−2 n+1
+ 84𝜇j−1 − 84𝜇jn+1 + 42𝜇j+1
n+1
Z (7)
xj , t = = 0, (26)
h7 n+1
− 6𝜇j+2 n+1
− 3𝜇j+3 + 𝜇j+4
n+1
( )
( ) 3628800
n+1
−𝜇j−5 n+1
+ 9𝜇j−4 n+1
− 36𝜇j−3 n+1
+ 84𝜇j−2 n+1
− 126𝜇j−1 + 126𝜇jn+1
Z (9)
xj , t = = 0. (27)
h9 n+1
− 84𝜇j+1 n+1
+ 36𝜇j+2 n+1
− 9𝜇j+3 + 𝜇j+4
n+1
Using Equations (23)–(27) for j = 0, Equation (22) for j = 0(1)M, and Equation (23), (25)–(27) for
j = M are considered respectively to generate a matrix equation as follows:
[ n+1 n+1 ]
n+1 T
where 𝜇n+1 = 𝜇−5 𝜇−4 … 𝜇M+3 n+1
𝜇M+4 , A(𝜇n ) and B(𝜇n ) are two square matrices of order
(M + 10) and G(tn ) is a column vector.
[ n ]T
Z(x, t) is determined with computation of the vector 𝜇n = 𝜇−5 𝜇−4
n
… 𝜇M+3n
𝜇M+4
n
by the repeated application of the recurrence relation Equation (28). The initial vectors 𝜇0 =
[ 0 ]T
𝜇−5 𝜇−40
… 𝜇M+3 0
𝜇M+4
0
can be computed from the initial state and boundary values of the
derivatives of the initial states as follows:
Z (4) (x0 , 0) = 0
Z (5) (x0 , 0) = 0
Z (6) (x0 , 0) = 0
Z (7) (x0 , 0) = 0
Z (9) (x0 , 0) = 0
( ) ( )
Z xj , 0 = g xj , 0 , j = 0 (1) M
Z (5)
(xM , 0) = 0
Z (6)
(xM , 0) = 0
Z (7)
(xM , 0) = 0
Z (9)
(xM , 0) = 0 (29)
D𝜇0 = 𝛾, (30)
[ ]T
where 𝜇0 = 𝜇−5
0
𝜇−4
0
… 𝜇M+3
0
𝜇M+4
0
and 𝛾 = [ 0 0 0 0 0 g(xj , 0) 0 0 0 0 ]T , j = 0, 1, … , M
⎡1 53 80 −568 434 434 −568 80 53 1 ⎤
⎢ ⎥
⎢−1 −21 54 54 −276 276 −54 −54 21 1 ⎥
⎢1 5 −40 80 −46 −46 80 −40 5 1 ⎥
⎢ ⎥
⎢−1 3 6 −42 84 −84 42 −6 −3 1 ⎥
⎢−1 9 −36 84 −126 126 −84 36 −9 1 ⎥
⎢ ⎥
⎢1 1013 47840 455192 1310354 1310354 455192 47840 47840 1 ⎥
⎢ … … … … … … … … … … ⎥
D=⎢ ⎥.
⎢ ⎥
⎢ ⎥
⎢ … … … … … … … … … … ⎥
⎢ 1 1013 47840 455192 1310354 1310354 455192 47840 1013 1⎥
⎢ ⎥
⎢ 1 −21 54 54 −276 276 −54 −54 21 1⎥
⎢ 1 5 −40 80 −46 −46 80 −40 5 1⎥
⎢ ⎥
⎢ −1 3 6 −42 84 −84 42 −6 −3 1⎥
⎢ −1 9 −36 84 −126 126 −84 36 −9 1⎥⎦
⎣
Therefore, the value of the initial vector 𝜇0 is easily obtained from Equation (30).
4 STABILITY ANALYSIS
To investigate the linear stability of the present scheme using Von Neumann approach, the nonlinear
term zzx of the Burgers’ equation is linearized by taking z as a local constant 𝜔 in Equation (1) to
transform
zt + 𝜔zx − azxx = 0. (31)
8
TABLE 1 Comparison of the computed approximate solution with exact and result of others at t = 0.001 with b = 2, h = 0.025, k = 0.0001
a=1 a = 0.5
x Exact Present [35] [6] [33] Exact Present [35] [6] [33]
0.1 0.65354448682 0.65354448773 0.653547 0.653589 0.653563 0.32786955239 0.32786955250 0.327870 0.327874 0.327870
0.2 1.30553352975 1.30553353240 1.305540 1.305611 1.305519 0.65506922222 0.65506922242 0.655071 0.655078 0.655028
0.3 1.94936356540 1.94936357030 1.949376 1.949485 1.949321 0.97841249918 0.97841249953 0.978416 0.978427 0.978449
0.4 2.56592491417 2.56592492245 2.565949 2.566103 2.565977 1.28846349693 1.28846349751 1.288469 1.288485 1.288417
0.5 3.11073888461 3.11073889691 3.110778 3.110992 3.110769 1.56306385243 1.56306385328 1.563074 1.563096 1.563014
0.6 3.49286571491 3.49286572765 3.492910 3.493222 3.492902 1.75664210905 1.75664210995 1.756654 1.756691 1.756653
0.7 3.54959512912 3.54959512594 3.549585 3.550079 3.549538 1.78720639755 1.78720639746 1.787204 1.787281 1.787184
0.8 3.05013447864 3.05013443170 3.049957 3.050702 3.050089 1.53769439067 1.53769438774 1.537649 1.537794 1.537658
0.9 1.81666037028 1.81666029947 1.816379 1.817077 1.816492 0.91685979884 0.91685979401 0.916786 0.916941 0.916795
L∞ … 7.31e−8 2.85e−4 … … … 1.95e−8 7.44e−5 … …
L2 … 2.72e−8 1.07e−4 … … … 3.59e−9 2.79e−5 … …
JENA AND GEBREMEDHIN
JENA AND GEBREMEDHIN 9
TABLE 2 Comparison of the current approximate solution with analytical solution and result of others at t = 0.001 with
b = 2, h = 0.025, k = 0.0001
a = 0.2 a = 0.1
x Exact Present [35] [6] Exact Present [35] [6]
0.1 0.13141154744 0.13141154746 0.131412 0.131412 0.06574975908 0.06574975909 0.065750 0.065750
0.2 0.26258120931 0.26258120932 0.262581 0.262582 0.13138293549 0.13138293549 0.131383 0.131383
0.3 0.39226230949 0.39226230952 0.392263 0.392263 0.19628086783 0.19628086784 0.196281 0.196281
0.4 0.51670947721 0.51670947725 0.516710 0.516711 0.25857573781 0.25857573782 0.258576 0.258576
0.5 0.62707950292 0.62707950297 0.627081 0.627082 0.31384935555 0.31384935556 0.313850 0.313850
0.6 0.70512028177 0.70512028182 0.705122 0.705124 0.35297182087 0.35297182089 0.352972 0.352972
0.7 0.71788221709 0.71788221712 0.717882 0.717890 0.35944285955 0.35944285957 0.359443 0.359444
0.8 0.61813645848 0.61813645842 0.618129 0.618148 0.30958038487 0.30958038487 0.309579 0.309583
0.9 0.36881372310 0.36881372291 0.368802 0.368824 0.18475374278 0.18475374273 0.184751 0.184756
L∞ … 5.19e−9 1.22e−5 … … 1.56e−9 3.08e−6 …
L2 … 8.48e−10 4.57e−6 … … 2.62e−10 1.15e−6 …
TABLE 3 Computed error norms by the current method for b = 2, h = 0.025, k = 0.0001 and different values of t and a
FIGURE 1 Numerical and exact solutions at different values of a and t = 0.001 [Color figure can be viewed at
wileyonlinelibrary.com]
Applying the Crank–Nicolson method and discretizing the time derivative using a finite difference
scheme in Equation (31) yields
𝜔k ak 𝜔k ak
zn+1
j + (zx )n+1
j − (zxx )n+1
j = znj − {(zx )}nj + (zxx )nj . (32)
2 2 2 2
The growth factor of a typical Fourier mode is defined as follows:
√
𝜇jn = 𝜑n eij𝜆h , i = −1, (33)
TABLE 4 Comparison of the computed error norm L∞ with others at different values of k and t for b = 2, h = 0.01
TABLE 5 Comparisons of the computed results L2 and L∞ with others at t = 1 with a = 0.01, b = 100,
k = 0.01, h = M1
Using Equations (9)–(11), the Fourier mode Equation (33), and implementation of Euler’s formula
(e±i𝜆h = cos(𝜆h) ± isin(𝜆h)) in Equation (32) for further simplification, gives
( ) ( )
q1 + iq2 𝜑n+1 = q3 + iq4 𝜑n , (34)
where
q1 = (1 + e1 ) cos (5𝜆h) + (1014 + e2 ) cos (4𝜆h) + (48853 + e3 ) cos (3𝜆h) + (503032 + e4 ) cos (2𝜆h)
+ (1765546 + e5 ) cos (𝜆h) + (1310354 + e7 ) ,
q2 = (𝑑1 − 1) sin (5𝜆h) + (𝑑2 − 1012) sin (4𝜆h) + (𝑑3 − 46827) sin (3𝜆h) + (𝑑4 − 407352) sin (2𝜆h)
+ (𝑑5 − 855162) sin (𝜆h) ,
q3 = (1 − e1 ) cos (5𝜆h) + (1014 − e2 ) cos (4𝜆h) + (48853 − e3 ) cos (3𝜆h) + (503032 − e4 ) cos (2𝜆h)
+ (1765546 − e5 ) cos (𝜆h) + (1310354 − e6 ) ,
TABLE 6 Comparative studies for the error norms at t = 1 with a = 0.005, b = 100,
k = 0.01, h = M1
M 10 20 40 80
−9 −11 −12
Present L∞ 1.259e 6.399e 2.469e 3.037e−13
method L2 5.523e −10
2.146e −11
7.530e −13
2.144e−13
[35] L∞ 1.215e−7 3.062e−8 7.644e−9 1.917e−9
L2 8.631e−8 2.153e−8 5.378e−9 1.345e−9
[43] L∞ 1.246e−7 3.394e−8 1.125e−8 5.549e−9
L2 8.819e−8 2.403e−8 7.942e−9 3.918e−9
[27] L∞ 2.062e−9 1.951e−10 1.461e−10 3.182e−10
L2 1.198e−10 2.160e−11 1.184e−11 2.665e−12
[26] L∞ 2.076e−9 2.732e−9 2.191e−10 3.272e−10
L2 1.278e−10 2.183e−11 2.735e−11 5.852e−11
[29] L∞ 4.708e−8 1.091e−8 1.980e−9 7.182e−9
L2 6.459e−8 4.465e−9 2.786e−10 2.665e−10
TABLE 7 Results of the error norms L2 and L∞ by current method at different value t with a = 0. 1, 0.2, 1, 2, b = 200,
k = 0.001, h = M1
FIGURE 2 Numerical and exact solutions at t = 1, 2, 3, 4 with a = 0.01, k = h = 0.01 [Color figure can be viewed at
wileyonlinelibrary.com]
FIGURE 3 Numerical and exact solutions for t = 1, 2, 3, 4 with a = 0.005, k = h = 0.01 [Color figure can be viewed at
wileyonlinelibrary.com]
FIGURE 4 Physical behavior of numerical and exact solutions for t = 1, h = 0.025, k = 0.0001 with different values a [Color
figure can be viewed at wileyonlinelibrary.com]
TABLE 8 Comparisons of the computed error norms with others at different values of t with
k = 0.0002, h = 0.02, a = 0.002
In this part, the approximate results using the present approach are computed for certain problems
of Equation (1). The efficiency and accuracy of the proposed scheme are tackled for different test
problems by computing the error norms L2 and L∞ which are mentioned as follows:
√
√ M
√ ∑ |( )2 |
L2 = √ h | Zj − zj |,
| |
j=0
L∞ = max ||Zj − zj || ,
j
TABLE 9 Comparisons of the current error norms with others at different values of t with
k = 0.0002, h = 0.02, a = 0.000666666
TABLE 10 Results of error norms computed by the present method at different values of t with k = 0.0002, h = 0.02,
a = 0.1, 0.2, 1, 2
FIGURE 5 Comparison of the numerical and exact solutions at t = 5 with h = 0.025, k = 0.0002, a = 0.002 [Color figure can
be viewed at wileyonlinelibrary.com]
Numerical experiment 1: We solve the model problem Equation (1) with the initial value
2a𝜋 sin (𝜋x)
z (x, 0) = , b>1
(b + cos (𝜋x))
and boundary conditions z(0, t) = z(1, t) = 0, which are obtained from the exact solution
2a𝜋e(−a𝜋 t) sin (𝜋x)
2
z (x, t) =
b + e(−a𝜋 t) cos (𝜋x)
2
as given by [6, 35, 43]. There is a brief comparison between our computed results with the analytical
solution and solutions of others. The numerical results and the error norms L∞ and L2 obtained by
the present method and other methods in [6, 33, 35] for the problem corresponding to t = 0.001 are
reported in Tables 1 and 2 and the error norms L∞ and L2 computed by the current method for t = 0.1,
0.2, 0.5, 1 are also described in Table 3 with the common values b = 2, a = 1, 0.5, 0.2, 0.1, h = 0.025,
TABLE 11 Comparisons of analytical and approximate solutions of Numerical experiment 3 at different values of a and h
0.25 0.4 0.013572156 0.013572130 0.01357 0.01363 0.01354 0.30889423 0.30889360 0.30910 0.30895 0.30887 0.30892
0.6 0.001888835 0.001888830 0.00189 0.00190 0.00188 0.24073902 0.24073869 0.24093 0.24079 0.24070 0.24077
0.8 0.000262448 0.000262447 0.00026 0.00026 0.00026 0.19567557 0.19567540 0.19586 0.19572 0.19566 0.19572
JENA AND GEBREMEDHIN
1 0.000036458 0.000036458 0.00004 0.00003 0.00004 0.16256486 0.16256478 0.16274 0.16261 0.16255 0.16261
3 9.8e−14 9.7e−14 0.00000 0.00000 0.00000 0.02720232 0.02720229 0.02720 0.02722 0.02721 0.02718
0.5 0.4 0.019235462 0.019235424 0.01923 0.01932 0.01920 0.56963245 0.56963209 0.56973 0.56978 0.56956 0.56970
0.6 0.002672020 0.002672012 0.00267 0.00269 0.00266 0.44720552 0.44720529 0.44736 0.44735 0.44715 0.44729
0.8 0.000371173 0.000371172 0.00037 0.00037 0.00037 0.35923606 0.35923594 0.35943 0.35937 0.35920 0.35930
1 0.000051560 0.000051560 0.00005 0.00005 0.00005 0.29191596 0.29191589 0.29213 0.29204 0.29188 0.29195
3 1.38e−13 1.38e−13 0.00000 0.00000 0.00000 0.04020492 0.04020488 0.04032 0.04024 0.04022 0.04016
0.75 0.4 0.013631023 0.013630995 0.01363 0.01369 0.01360 0.62543790 0.62543923 0.62573 0.62572 0.62540 0.62520
0.6 0.001889972 0.001889966 0.00189 0.00190 0.00188 0.48721498 0.48721544 0.48760 0.48752 0.48716 0.48694
0.8 0.000262470 0.000262469 0.00026 0.00026 0.00026 0.37392175 0.37392177 0.37434 0.37419 0.37389 0.37365
1 0.0000364587 0.0000364585 0.00004 0.00003 0.00026 0.28747441 0.28747424 0.28788 0.28770 0.28743 0.28724
3 9.8e−14 9.7e−14 0.00000 0.00000 0.00000 0.02977213 0.02977208 0.29881 0.02980 0.02978 0.02974
TABLE 12 Comparisons of current results with others for numerical experiment 3 at t = 0.1, a = 1 and different values of h
TABLE 13 Values of error norms obtained by the current method for numerical experiment 3 at different
values of t and a with k = 0.002, 0.001, and h = 0.025
a t 0.1 0.5 1 2 3 5
k = 0.002 0.2 L∞ 1.84e−6 4.66e−7 3.62e−7 9.08e−8 1.90e−8 6.13e−10
L2 1.26e−6 3.46e−7 2.41e−7 6.42e−8 1.34e−8 4.33e−10
1 L∞ 1.19e−5 1.15e−6 1.65e−8 1.71e−12 1.31e−16 2.00e−18
L2 8.47e−6 8.12e−7 1.17e−8 1.21e−12 9.25e−17 1.15e−18
−5 −8 −12 −19 −19
2 L∞ 3.56e 6.62e 6.85e 5.87e 6.07e 5.94e−19
−5 −8 −12 −19 −19
L2 2.52e 4.68e 4.84e 1.70e 1.49e 1.75e−19
−7 −7 −8 −8 −9
k = 0.001 0.2 L∞ 4.59e 1.16e 9.05e 2.27e 4.75e 1.53e−10
−7 −8 −8 −8 −9
L2 3.16e 8.64e 6.02e 1.61e 3.36e 1.08e−10
1 L∞ 2.98e−6 2.87e−7 4.13e−9 4.27e−13 2.57e−17 2.85e−17
L2 2.12e−6 2.03e−7 2.92e−9 3.02e−13 1.52e−17 1.96e−17
2 L∞ 8.91e−6 1.66e−8 1.71e−12 4.36e−18 4.43e−18 4.38e−18
−6 −8 −12 −18 −18
L2 6.30e 1.17e 1.21e 2.41e 2.50e 2.49e−18
k = 0.0001. In Figure 1, the approximate and exact solutions corresponding to t = 0.001 with different
values of a be depicted and it is similar with the figure reported in [6, 35].
In the second part of numerical experiment 1, computed error norm L∞ using the present method
and comparison with the methods in [27, 37] at different values of k and t with the values b = 2,
h = 0.01, a = 10−3 , 10−5 are reported in Table 4.
In the third part of numerical experiment 1, we have computed the error norms L∞ and L2 for t = 1
with the values b = 100, h = M1 , k = 0.01 and comparison of the error norms of the present method
with the others in [26, 35, 43] at a = 0.01 and with the methods in [26, 27, 29, 35, 43] at a = 0.005
are depicted in Tables 5 and 6, respectively. The computed results of error norms L∞ and L2 using
the present method at t = 0.1, 0.5, 2,5 with a = 0.1, 0.2, 1,2, b = 200, k = 0.001, h = M1 are also
represented in Table 7. The physical behavior of exact and numerical solutions for t ≤ 4 with values
b = 100, h = 0.01, k = 0.01 and at a = 0.01, 0.005 is depicted in Figures 2 and 3, respectively and
the two figures are similar with the figures reported in [27, 35, 43]. Figure 4 describes the physical
behavior of the numerical and exact solutions in three-dimension at t = 1 with h = 0.025, k = 0.0001,
a = 0.5, 0.1 and 0.01 and it is shown that the solutions of the governing equation rise up from xt-plane
as the value of the parameter “a” decreases.
JENA AND GEBREMEDHIN 17
Numerical experiment 2: The analytical solution of Equation (1) is given in [35, 46] as:
( ( ))
a x
z (x, t) = x + tan
1 + at 2 + 2at
with initial value ( ( ))
x
z (x, 0) = a x + tan
2
and boundary conditions
( ( )) ( ( ))
a 0.5 a 1.5
z (0.5, t) = 0.5 + tan , z (1.5, t) = 1.5 + tan .
1 + at 2 + 2at 1 + at 2 + 2at
The physical measure of L∞ and L2 for different values of t with parameters h = 0.02, k = 0.0002
using the present method and other methods at a = 0.002 and 0.0006666 are reported in Tables 8 and
9, respectively. The results of L∞ and L2 by the present scheme is very small as compared to results
obtained in [35, 46]. The error norms computed by the present method at different values of t with
k = 0.0002, h = 0.02, a = 0.1, 0.2, 1,2 are also reported in Table 10. The physical behavior of numerical
and exact solutions for t = 5 with h = 0.025, k = 0.0002, a = 0.002 is illustrated in Figure 5 and it is
observed that the numerical solution is in a good agreement with exact solutions.
Numerical experiment 3: Let us take the Burgers’ equation (1) with the initial and homogeneous bound-
ary values given as z(x, 0) = sin(𝜋x) and z(0, t) = z(1, t) = 0, respectively. The analytical solution of
Equation (1) with given initial-boundary values was given by Cole [11] as follows:
∑∞ 2
2𝜋a p=1 bp e−(𝜋p) at p sin (p𝜋x)
z (x, t) = ∑∞ 2 ,
b0 + p=1 bp e−(𝜋p) at cos (p𝜋x)
where b0 and bp (p = 1, 2, 3, … ) are Fourier coefficients described by:
1 (−1+cos(𝜋x))
b0 = e 2𝜋a dx
∫0
1 (−1+cos(𝜋x))
bp = e 2𝜋a cos (p𝜋x) dx, p ≥ 1.
∫0
The comparisons of present numerical solutions for the parameters a = 1, 0.1, 0.01, h = 0.05, 0.025,
0.0125, 0.01, k = 0.00025, 0.0025, 0.002, 0.001, 0.00125, 0.0001, 0.000125, 0.00001 at t = 0.4, 0.6,
0.8, 1,3 and t = 0.1 with analytical solutions and the result of other numerical studies are reported in
Tables 11 and 12, respectively. Finally, the computed error norms by the proposed method at different
values t and a with k = 0.002, 0.001 and h = 0.025 are recorded in Table 13.
6 CONCLUSION
In this paper, we have constructed a decatic B-spline collocation approach to compute the approximate
solution of Burgers’ equation and the nonlinear term of the Burgers’ equations is linearized using the
Taylor series technique. The present scheme is an efficient method to obtain a better approximate solu-
tion of the nonlinear Burgers’ equation by avoiding the difficulties for a very small value of kinematic
viscosity “a” which is a highly advantageous and remarkable point of our work over others in the liter-
ature. The stability analysis using Von Neumann approach for the present method is investigated and
it is found to be unconditionally stable. The approximate results of the numerical test problems com-
puted using the proposed approach are in good agreement with analytical solutions as well as much
better than the results of other numerical studies in the literature.
18 JENA AND GEBREMEDHIN
ORCID
Saumya R. Jena https://orcid.org/0000-0001-6247-6109
REFERENCES
[1] E. N. Aksan and A. Özdeş, A numerical solution of Burgers’ equation, Appl. Math. Comput. 156 (2004), 395–402.
[2] A. R. Appadu, Optimised composite numerical schemes in 2-d for hyperbolic conservation laws, Int. J. Numer.
Methods Fluids 69 (2012), 1522–1549.
[3] A. R. Appadu, M. Z. Dauhoo, and S. D. V. Rughooputh, Control of numerical effects of dispersion and dissipation
in numerical schemes for efficient shock-capturing through an optimal courant number, Comput. Fluids 37 (2008),
767–783.
[4] A. R. Appadu and S. N. N. Nguetchue, The technique of MIEELDLD as a measure of the shock-capturing property
of numerical methods for hyperbolic conservation laws, Prog. Comput. Fluid Dyn. 15 (2015), 247–264.
[5] A. R. Appadu and A. A. I. Peer, Optimized weighted essentially nonoscillatory third-order schemes for hyperbolic
conservation laws, J. Appl. Math. 2013 (2013), 1–12.
[6] A. Asaithambi, Numerical solution of the Burgers’ equation by automatic differentiation, Appl. Math. Comput.
216 (2010), 2700–2708.
[7] H. Bateman, Some recent researches on the motion of fluids, Mon. Weather Rev. 43 (1915), 163–170.
[8] M. P. Bonkile, A. Awasthi, C. Lakshmi, V. Mukundan, and V. S. Aswin, A systematic literature review of Burgers’
equation with recent advances, Pramana 90 (2018), 69.
[9] J. M. Burgers, A mathematical model illustrating the theory of turbulence, in Advances in Applied Mechanics,
Elsevier, Vol. 1, 1948, 171–199.
[10] R. Chen and Z. Wu, Applying multiquadric quasi-interpolation to solve Burgers’ equation, Appl. Math. Comput.
172 (2006), 472–484.
[11] J. D. Cole, On a quasi-linear parabolic equation occurring in aerodynamics, Q. Appl. Math. 9 (1951), 225–236.
[12] I. Dag and O. Ersoy, Numerical solution of generalized Burgers–Fisher equation by exponential cubic B-spline
collocation method, in AIP Conference Proceedings, Vol. 1648, AIP Publishing LLC, Rhodes, Greece, 2015,
370008.
[13] I. Dag, D. Irk, and A. Sahin, B-spline collocation methods for numerical solutions of the Burgers’ equation, Math.
Probl. Eng. 2005 (2005), 521–538.
[14] I. Dağ, D. Irk, and B. Saka, A numerical solution of the Burgers’ equation using cubic B-splines, Appl. Math.
Comput. 163 (2005), 199–211.
[15] İ. Dağ, B. Saka, and A. Boz, B-spline Galerkin methods for numerical solutions of the Burgers’ equation, Appl.
Math. Comput. 166 (2005), 506–522.
[16] T. E. Danaf, Efficient and accurate numerical methods for the Burgers and related partial differential equations,
Ph.D. Thesis, Menoufia University, 1998.
[17] M. Dehghan, N. Shafieeabyaneh, and M. Abbaszadeh, Numerical and theoretical discussions for solving nonlinear
generalized Benjamin–Bona–Mahony–Burgers equation based on the Legendre spectral element method, Numer.
Methods Partial Differ. Eq. 37 (2021), 360–382.
[18] O. Ersoy, I. Dag, and N. Adar, Exponential twice continuously differentiable B-spline algorithm for Burgers’
equation, Ukr. Math. J. 70 (2018), 906–921.
[19] S. Gupta and V. K. Kukreja, An improvised collocation algorithm with specific end conditions for solving modified
Burgers equation, Numer. Methods Partial Differ. Eq. 37 (2021), 874–896.
[20] I. A. Hassanien, H. A. Salama, and A. A. a nd Hosham, Fourth-order finite difference method for solving Burgers’
equation, Appl. Math. Comput. 170 (2005), 781–800.
[21] O. E. Hepson, An extended cubic B-spline finite element method for solving generalized Burgers–Fisher equation,
in AIP Conference Proceedings, Vol 1978, AIP Publishing LLC, Thessaloniki, Greece, 2018, 470100.
[22] O. E. Hepson, A. Korkmaz, and I. Dag, Extended B-spline collocation method for Kdv–Burgers equation, TWMS
J. Appl. Eng. Math. 9 (2019), 267–278.
[23] B. Inan and A. R. Bahadir, Numerical solution of the one-dimensional Burgers’ equation: Implicit and fully implicit
exponential finite difference methods, Pramana 81 (2013), 547–556.
[24] L. Iskandar and A. Mohsen, Some numerical experiments on the splitting of Burgers’ equation, Numer. Methods
Partial Differ. Eq. 8 (1992), 267–276.
[25] S. R. Jena, A. Senapati, and G. S. Gebremedhin, Approximate solution of MRLW equation in B-spline environment,
Math. Sci. 14 (2020), 345–357.
[26] R. Jiwari, A Haar wavelet quasilinearization approach for numerical simulation of Burgers’ equation, Comput.
Phys. Commun. 183 (2012), 2413–2423.
[27] R. Jiwari, A hybrid numerical scheme for the numerical solution of the Burgers’ equation, Comput. Phys. Commun.
188 (2015), 59–67.
JENA AND GEBREMEDHIN 19
[28] R. Jiwari, S. Kumar, and R. C. Mittal, Meshfree algorithms based on radial basis functions for numerical
simulation and to capture shocks behavior of Burgers’ type problems, Eng. Comput. 36 (2019), 1142–1168.
[29] R. Jiwari, R. C. Mittal, and K. K. Sharma, A numerical scheme based on weighted average differential quadrature
method for the numerical solution of Burgers’ equation, Appl. Math. Comput. 219 (2013), 6680–6691.
[30] A. Korkmaz, A. M. Aksoy, and I. Dag, Quartic B-spline differential quadrature method, Int. J. Nonlinear Sci. 11
(2011), 403–411.
[31] A. Korkmaz and İ. Dağ, Shock wave simulations using sinc differential quadrature method, Eng. Comput. 28
(2011), 654–674.
[32] S. Kutluay, A. Bahadir, and A. Özdeş, Numerical solution of one-dimensional Burgers equation: Explicit and
exact-explicit finite difference methods, J. Comput. Appl. Math. 103 (1999), 251–261.
[33] M. Li, L. Chen, and Q. Ma, A meshfree quasi-interpolation method for solving Burgers’ equation, Math. Probl.
Eng. 2014 (2014), 1–8.
[34] M. J. Lighthill, Viscosity effects in sound waves of finite amplitude, Vol 250351, Batchelor and Davies (Eds),
Surveys in Mechanics, Cambridge University press, London, 1956.
[35] R. C. Mittal and R. K. Jain, Numerical solutions of nonlinear Burgers’ equation with modified cubic B-splines
collocation method, Appl. Math. Comput. 218 (2012), 7839–7855.
[36] R. C. Mittal and R. Jiwari, A differential quadrature method for numerical solutions of Burgers’ type equations,
Int. J. Numer. Methods Heat Fluid Flow 22 (2012), 880–895.
[37] R. C. Mittal, R. Jiwari, and K. K. Sharma, A numerical scheme based on differential quadrature method to solve
time dependent Burgers’ equation, Eng. Comput. 30 (2012), 117–131.
[38] K. P. Mredula, V. D. Pathak, and B. M. Shah, Numerical solution of nonlinear Burger equation using wavelet
approximation combined with finite volume formulation, Math. Today 34 (2018), 174–187.
[39] V. Mukundan and A. Awasthi, Efficient numerical techniques for Burgers’ equation, Appl. Math. Comput. 262
(2015), 282–297.
[40] V. Mukundan and A. Awasthi, Linearized implicit numerical method for Burgers’ equation, Nonlinear Eng. 5
(2016), 219–234.
[41] T. Öziş, E. N. Aksan, and A. Özdeş, A finite element approach for solution of Burgers’ equation, Appl. Math.
Comput. 139 (2003), 417–428.
[42] L. A. Pospelov, Propagation of finite amplitude elastic waves (longitudinal elastic wave of finite amplitude
propagation in isotropic solid), Soviet Phys. Acoust. 11 (1966), 302–304.
[43] K. Rahman, N. Helil, and R. Yimin, Some new semi-implicit finite difference schemes for numerical solution
of Burgers equation, 2010 International Conference on Computer Application and System Modeling (ICCASM
2010), vol. 14, IEEE, 2010, pp. 451–455.
[44] M. A. Ramadan, T. S. El-Danaf, and F. E. Abd Alaal, Application of the non-polynomial spline approach to the
solution of the Burgers’ equation, Open Appl. Math. J. 1 (2007), 15–20.
[45] M. A. Ramadan, T. S. El-Danaf, and F. E. A. Alaal, A numerical solution of the Burgers’ equation using septic
B-splines, Chaos Solitons Fractals 26 (2005), 1249–1258.
[46] K. R. Raslan, A collocation solution for Burgers equation using quadratic B-spline finite elements, Int. J. Comput.
Math. 80 (2003), 931–938.
[47] B. Saka and İ. Dağ, A numerical study of the Burgers’ equation, J. Franklin Inst. 345 (2008), 328–348.
[48] M. Sarboland and A. Aminataei, On the numerical solution of one-dimensional nonlinear nonhomogeneous
Burgers’ equation, J. Appl. Math. 2014 (2014), 1–15.
[49] M. Seydaoğlu, A meshless method for Burgers’ equation using multiquadric radial basis functions with a Lie-group
integrator, Mathematics 7 (2019), 1–11.
[50] M. Yaseen and M. Abbas, An efficient computational technique based on cubic trigonometric B-splines for time
fractional Burgers’ equation, Int. J. Comput. Math. 97 (2020), 725–738.
[51] C.-G. Zhu and R.-H. Wang, Numerical solution of Burgers’ equation by cubic B-spline quasi-interpolation, Appl.
Math. Comput. 208 (2009), 260–272.
[52] S. R. Jena, A. Senapati and G. S. Gebremedhin, Numerical study of solitions in BFRK Scheme, International Journal
of Mechanics and Control. 21 (2020), 163–175.
How to cite this article: Jena SR, Gebremedhin GS. Decatic B-spline collocation scheme for
approximate solution of Burgers’ equation. Numer Methods Partial Differential Eq.
2021;1–19. https://doi.org/10.1002/num.22747