Applied Mathematics and Computation 224 (2013) 166–177

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

Numerical solution of Burgers’ equation with modified cubic

B-spline differential quadrature method
Geeta Arora, Brajesh Kumar Singh ⇑
Department of Mathematics, School of Allied Sciences, Graphic Era Hill University, Dehradun 248002, Uttarakhand, India

a r t i c l e i n f o a b s t r a c t

Keywords: In this paper, a new numerical method, ‘‘modified cubic-B-spline differential quadrature
Burgers’ equation method (MCB-DQM)’’ is proposed to find the approximate solution of the Burgers’ equation.
Cubic B-spline The modified cubic-B-spline basis functions are used in differential quadrature to determine
Modified cubic B-spline the weighting coefficients. The MCB-DQM is used in space, and the optimal four-stage,
Differential quadrature method (DQM)
order three strong stability-preserving time-stepping Runge–Kutta (SSP-RK43) scheme is
Thomas algorithm
SSP-RK43 scheme
used in time for solving the resulting system of ordinary differential equations. To check
the efficiency and accuracy of the method, four examples of Burgers’ equation are included
with their numerical solutions, L2 and L1 errors and comparisons are done with the results
given in the literature. The proposed method produces better results as compared to the
results obtained by almost all the schemes available in the literature, and approaching to
the exact solutions. The presented method is seen to be easy, powerful, efficient and eco-
nomical to implement as compared to the existing techniques for finding the numerical
solutions for various kinds of linear/nonlinear physical models.
Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction

We consider the well known one dimensional nonlinear Burgers’ equation

@u @u @2u
þ au  m 2 ¼ 0; ðx; tÞ 2 X  ½0; T; ð1:1Þ
@t @x @x
where X ¼ ða; bÞ, with the initial condition

uðx; 0Þ ¼ f ðxÞ; x 2 ½a; b; ð1:2Þ

and the boundary conditions

uða; tÞ ¼ 0 and uðb; tÞ ¼ 0; t 2 ½0; T; ð1:3Þ

where m > 0 is a small parameter known as the coefficient of kinematic viscosity and a is some positive constant. Such type
of equations was first introduced by Bateman [5]. Also, he proposed the steady-state solution of the problem. Burgers [6,7]
has introduced this equation to capture some features of turbulent fluid in a channel caused by the interaction of the oppo-
site effects of convection and diffusion, and hence Eq. (1.1) is referred to as ‘‘Burgers’ equation’’. The structure of Burgers’
equation is similar to the one dimensional Navier–Stoke’s equation without the stress term. It is the simplest model of non-
linear partial differential equation for diffusive waves in fluid dynamics. This model arises in many physical problems

⇑ Corresponding author.
E-mail addresses: geetadma@gmail.com (G. Arora), bksingh0584@gmail.com (B.K. Singh).

0096-3003/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.amc.2013.08.071
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 167

including one-dimensional sound/shock waves in a viscous medium, waves in fluid filled viscous elastic tubes, magneto-
hydrodynamic waves in a medium with finite electrical conductivity, mathematical modeling of turbulent fluid, and in con-
tinuous stochastic processes.
In the last years, a great deal of effort to compute efficiently the numerical solutions of the Burgers’ equation for small and
large both values of the kinematic viscosity has been expanded. The Burgers’ equation is solved for both the infinite and the
finite domain [9]. The various numerical techniques to compute the numerical solutions of the Burgers’ equation are: auto-
matic differentiation method [1], finite elements method [32,34], Galerkin finite element method [13], cubic B-splines col-
location method [12], cubic B-spline quasi-interpolation [15], modified cubic B-spline collocation method [31], spectral
collocation method [27], sinc differential quadrature method [22], polynomial based differential quadrature method [23],
cubic B-spline differential quadrature method [20], quartic B-splines differential quadrature method [24], quartic B-splines
collocation method [38], quadratic B-splines finite difference element method [4,33], fourth-order finite difference method
[14], factorized diagonal Padé approximation [2], non-polynomial spline approach [37], a novel numerical scheme [45], ex-
plicit and exact-explicit finite difference methods [25], Hopf–Cole transformation [11,25], least-squares quadratic B-splines
finite element method [26], reproducing kernel function method [41], implicit fourth-order compact finite difference [29],
weighted average differential quadrature method [16], variational method [3], parameter-uniform implicit difference
scheme [17], one dimensional fourier expansion [30], etc.
The differential quadrature method (DQM), is an efficient technique to solve partial differential equations (PDEs), was first
introduced by Bellman et al. [8]. It was further improved by Quan and Chang [35,36] to solve the weighting coefficients. Var-
ious kinds of test functions such as spline functions, Lagrange interpolation polynomials, sinc function [35,36,22], etc. are used to
determine the weighting coefficients, for further details on DQM we refer to [42,43,40,36].
B-splines are a set of special spline functions that can be used to construct piece-wise polynomial by computing the
appropriate linear combination. These functions have their computational advantage from the fact that any B-spline basis
function of order m is nonzero over at most m adjacent intervals and zero otherwise. Due to smoothness and capability
to handle local phenomena, B-spline basis functions offer distinct advantages in comparison to other basis functions. Cubic
B-spline functions have already been used as basis functions to solve many physical models. Recently, Korkmaz and Dağ [21]
used cubic B-spline functions with DQM to solve advection–diffusion equation. Mittal and Jain [31] solved Burgers’ equation
by modified cubic B-spline collocation method.
In this paper, a new numerical method, ‘‘modified cubic-B-spline differential quadrature method (MCB-DQM)’’ is pro-
posed to find the approximate solution of the Burgers’ equation. In this method, the modified cubic-B-spline basis func-
tions are used in DQM to determine the weighting coefficients (i.e., spatial derivatives) which produces the system of
first order ordinary differential equations (ODEs). The resulting system of ODEs is solved by the optimal four-stage, order
three strong stability-preserving time-stepping Runge–Kutta (SSP-RK43) scheme [44]. The SSP-RK43 scheme needs less
storage space that causes less accumulation of numerical errors. This is why we preferred SSP-RK43 scheme. The
MCB-DQM solutions to the Burgers’ equation have been computed without transforming the equation and without using
the linearization. The comparison of the MCB-DQM numerical solutions with analytical solutions are presented to illus-
trate the efficiency and adaptability of the method. The L2 and L1 errors are also evaluated and compared with results
given in the literature.
This paper is organized as follows. In Section 2, the description of the modified cubic B-spline differential quadrature
method is given. In Sections 3, procedure for implementation of method is described. Numerical examples are given to estab-
lish the applicability and accuracy of the proposed method in Section 4. The conclusion is given in Section 5 that briefly sum-
marizes the numerical outcomes.

2. Description of the method

The differential quadrature method (DQM) is an approximation to derivatives of a function using the weighted
sum of the functional values at certain discrete points. Since the weighting coefficients are dependent on the spatial
grid spacing only, one can assume uniformly distributed N nodes/knots: a ¼ x1 < x2 ; . . . ; xN1 < xN ¼ b such that
xiþ1  xi ¼ h on the real axis. The first and second order spatial derivatives of the uðx; tÞ at any time on the knot
xi for i ¼ 1; . . . ; N are given by
X
N
ux ðxi ; tÞ ¼ aij uðxj ; tÞ; for j ¼ 1; . . . ; N; ð2:1Þ
j¼1

X
N
uxx ðxi ; tÞ ¼ bij uðxj ; tÞ; for j ¼ 1; . . . ; N; ð2:2Þ
j¼1

where aij and bij are weighting coefficients of the first and second order derivatives with respect to x, respectively [8].
The cubic B-spline basis functions at the knots are defined as follows
168 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177

8
>
> ðx  xj2 Þ3 x 2 ðxj2 ; xj1 Þ
>
>
>
> 3 3
> ðx  xj2 Þ  4ðx  xj1 Þ x 2 ðxj1 ; xj Þ
1<
uj ðxÞ ¼ 3 ðxjþ2  xÞ3  4ðxjþ1  xÞ3 x 2 ðxj ; xjþ1 Þ ð2:3Þ
>
h >
>
> ðx  xÞ3
>
> jþ2 x 2 ðxjþ1 ; xjþ2 Þ
>
:
0 otherwise;
where fu0 ; u1 ; . . . ; uN ; uNþ1 g forms a basis over the region ½a; b. The values of cubic B-splines and its derivatives at the nodal
points are tabulated in Table 0.1.
The cubic B-spline basis functions are modified in such way that the resulting matrix system of equations is diagonally
dominant. The modified cubic B-spline basis functions at the knots are defined as follows [31]
9
/1 ðxÞ ¼ u1 ðxÞ þ 2u0 ðxÞ >
>
>
>
/2 ðxÞ ¼ u2 ðxÞ  u0 ðxÞ >
>
=
/j ðxÞ ¼ uj ðxÞ for j ¼ 3; . . . ; N  2 ð2:4Þ
>
>
/N1 ðxÞ ¼ uN1 ðxÞ  uNþ1 ðxÞ >
>
>
>
;
/N ðxÞ ¼ uN ðxÞ þ 2uNþ1 ðxÞ
where f/1 ; /2 ; . . . ; /N g forms a basis over the region ½a; b.

2.1. To determine the weighting coefficients

The first order derivative approximation is given by

X
N
/0k ðxi Þ ¼ aij /k ðxj Þ; for i ¼ 1; . . . ; N; k ¼ 1; . . . ; N ð2:5Þ
j¼1

For the first knot point x1 , the approximation can be given as

X
N
/0k ðx1 Þ ¼ a1j /k ðxj Þ; for k ¼ 1; . . . ; N ð2:6Þ
j¼1

which gives a tridiagonal system of equation as

2 3
6 1 2 3
2 3 6=h
60 4 1 7 a11
6 7 6 7
6 76 7 6 6=h 7
6 1 4 1 76 a12 7 6 7
6 76 7 6 7
6 .. .. .. 76 .. 7 6 0 7
6 . . . 76 . 7 ¼ 6 . 7
6 76 7 6 .. 7
6 1 4 1 76 7 6 7
6 74 a1N1 5 6 7
6 7 4 0 5
4 1 4 0 5 a1N
0
1 6
We apply well known ‘‘Thomas algorithm’’ to solve the resulting tridiagonal system of equations whose solution provides
the coefficients a11 ; a12 ; . . . ; a1N . Similarly, for the second knot point x2 , the approximation can be given as

X
N
/0k ðx2 Þ ¼ a2j /k ðxj Þ; for k ¼ 1; . . . ; N ð2:7Þ
j¼1

which again gives a tridiagonal system of equations as

Table 0.1
The coefficients of cubic B-splines and its derivatives at knots xj .

xj2 xj1 xj xjþ1 xjþ2

uj ðxÞ 0 1 4 1 0
u0j ðxÞ 0 3=h 0 3=h 0
u00j ðxÞ 0 6=h
2
12=h
2
6=h
2 0
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 169

2 3
6 1 2 3
2 3 3=h
60 4 1 7 a21
6 7 6 7
6 76 7 6 0 7
6 1 4 1 76 a22 7 6 7
6 76 7 6 7
6 .. .. .. 76 .. 7 6 3=h 7
6 . . . 76 . 7 ¼ 6 7
6 76 7 6 0 7
6 1 4 1 76 7 6 7
6 74 a2N1 5 6 .. 7
6 7 4 . 5
4 1 4 0 5 a2N
0
1 6
whose solution provides the coefficients a21 ; a22 ; . . . ; a2N . Proceeding in the similar manner, up to the second last knot point
xN1 the coefficients ak1 ; ak2 ; . . . ; akN for k ¼ 3 . . . N  1 are determined. At the knot point xN1 , the tridiagonal system of equa-
tions is given as
2 3
6 1 2 3
2 3 0
60 4 1 7 aN11
6 7 6 .. 7
6 76 7 6 7
6 1 4 1 76 aN12 7 6 . 7
6 76 7 6 7
6 .. .. .. 76 .. 7 6 0 7
6 . . . 76 . 7¼6 7
6 76 7 6 3=h 7
6 1 4 1 76 7 6 7
6 74 aN1N1 5 6 7
6 7 4 0 5
4 1 4 0 5 aN1N
3=h
1 6
Now, the matrix system of equations corresponding to the last knot point xN is given as
2 3
6 1
2 3 2 3
60 4 1 7 aN1 0
6 7
6 76 7 6 . 7
6 1 4 1 76 aN2 7 6 .. 7
6 76 7
6 .. .. .. 76 .. 7 6 7
6 . . . 76 . 7 ¼ 6 6 0 7;
7
6 76 7 6 7
6 1 4 1 76 7 4 6=h 5
6 74 aNN1 5
6 7
4 1 4 0 5 aNN 6=h
1 6
which provides the coefficients aN1 ; aN2 ; . . . ; aNN . Thus, we have evaluated the weighting coefficient aij for
i ¼ 1; 2; . . . ; N; j ¼ 1; 2; . . . ; N.
Using these coefficients, the weighting coefficient bij for i ¼ 1; 2; . . . ; N; j ¼ 1; 2; . . . ; N is evaluated as follows [40]
  X
N
1
bij ¼ 2aij aij  ; for i – j; and bii ¼  bij :
xi  xj i¼1;i – j

3. Implementation of method

On substituting the first and second order approximation of the spatial derivatives, obtained by using MCB-DQM, the Bur-
gers’ Eq. (1.1) can be rewritten as
@ui XN XN
¼ m bij uðxj Þ  auðxi Þ aij uðxj Þ; i ¼ 1; . . . ; N: ð3:1Þ
@t j¼1 j¼1

Thus, Eq. (3.1) is reduced into a set of ordinary differential equations in time, that is, for i ¼ 1; . . . ; N, we have
dui
¼ Lðui Þ; ð3:2Þ
dt
where L denotes a spatial nonlinear differential operator. There are various methods to solve this system of ODE. We pre-
ferred the optimal four-stage, order three strong stability-preserving time-stepping Runge–Kutta (SSP-RK43) scheme [44]
to solve the system of ODE. In this scheme the Eq. (3.2) is integrated from time t 0 to t0 þ Mt through the following operations
Mt
uð1Þ ¼ um þ Lðum Þ
2
Mt
uð2Þ ¼ uð1Þ þ Lðuð1Þ Þ
2
ð3Þ 2 m uð2Þ Mt
u ¼ u þ þ Lðuð2Þ Þ
3 3 6
mþ1 ð3Þ Mt ð3Þ
u ¼ u þ Lðu Þ;
2
170 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177

and consequently the solution uðx; tÞ at a particular time level is completely known.

4. Numerical experiments and discussion

In this section, the numerical solutions by the proposed method (MCB-DQM) are evaluated for some examples of Burgers’
equation. Existence of analytical solutions help to measure the accuracy of numerical methods. In the present study, the
accuracy and the efficiency of this method is measured for various numerical examples by evaluating the discrete L2 and
L1 error norms which are defined as follows
!1=2
N h
X i2 
N 


L2 ¼ h uexact
j  uj ; and L1 ¼ maxuexact
j  uj ;
j¼1
j¼1

where uj represent the numerical solution at node j.

Example 1. The Burgers’ Eq. (1.1) with a ¼ 1 is solved over the region ½0; 1:2 and the initial and boundary conditions are as
given in [1,4,23]
x
uðx; 1Þ ¼ 1 ; and uð0; tÞ ¼ 0; uð1:2; tÞ ¼ 0; for t > 1:
1 þ exp 4m
ðx2  14Þ
In this problem the initial condition is taken at t ¼ 1. The exact solution of the problem is
x  
t 1
uðx; tÞ ¼  1=2  x2  ; for t P 1; where t 0 ¼ exp :
t 8m
1þ t0
exp 4mt

In this example, the parameter values are taken as m ¼ 0:005; h ¼ 0:01; Mt ¼ 0:01. The comparison of the numerical solutions
obtained by MCB-DQM, at different time levels are presented with the solutions obtained by Mittal and Jain [31], Shu et al.
[41] and the exact solutions in Table 1.1. In Table 1.2, L2 and L1 errors at different time levels t 6 3:50 are compared with the
errors obtained by several earlier schemes. Further, the L2 and L1 errors at t ¼ 3:60 are compared with errors obtained by the
three methods recently proposed by Korkmaz–Dağ [20] and are reported in Table 1.3. It is found that our results are much
better than all the three methods. It is evident from Table 1.1, 1.2 and 1.3 that our method produces better approximate nu-
meric solutions than almost all the earlier schemes and approaching towards the exact solutions.
The absolute errors at t ¼ 3:5 are plotted in Fig. 1. It is evident from Fig. 1 that the absolute errors are very small as
compared to that given by Mittal and Jain [[31] Fig. 10]. The absolute errors for different time levels are also plotted in Fig. 2.
It is found that the errors are decreasing with increasing time and the maximum error is shifting towards the boundary
x ¼ 1:2 only. The physical behavior of the numerical solutions at m ¼ 0:005 for different time levels t 6 3:5 is depicted in
Fig. 3.

Example 2. In this example, we take the particular solution of the Burgers’ Eq. (1.1), for a ¼ 1, over the region ½0; 2 as con-
sidered in [32]:

Table 1.1
Comparison of the MCB-DQM numerical solutions of Example 1 with exact solutions, for m ¼ 0:005.
x t Shu et al. [41] with h ¼ 104 ; Mt ¼ :01 Mittal & Jain [31] MCB-DQM Exact value

b¼1 b ¼ 0:5 h ¼ 0:005; Mt ¼ 10 3 h ¼ 0:01; Mt ¼ 0:01

0.20 1.7 0.1176565 0.1174841 0.1176452 0.1176450 0.1176452

2.5 0.0800527 0.0798389 0.0799990 0.0799989 0.0799990
3.0 0.0667147 0.0665176 0.0666658 0.0666658 0.0666658
3.5 0.0571820 0.0570060 0.0571422 0.0571422 0.0571422
0.40 1.7 0.2332111 0.2348504 0.2351690 0.2351680 0.2351677
2.5 0.1591735 0.1596608 0.1599771 0.1599770 0.1599769
3.0 0.1328314 0.1330273 0.1333211 0.1333210 0.1333209
3.5 0.1139606 0.1140077 0.1142780 0.1142780 0.1142779
0.6 1.7 0.2940048 0.2961269 0.2958570 0.2959160 0.2959097
2.5 0.2347876 0.2376699 0.2381299 0.2381200 0.2381207
3.0 0.1973222 0.1990478 0.1994839 0.1994800 0.1994805
3.5 0.1697753 0.1708231 0.1712257 0.1712240 0.1712242
0.8 1.7 0.0008917 0.0006640 0.0006381 0.0006464 0.0006465
2.5 0.1103866 0.1036067 0.1021325 0.1020930 0.1020957
3.0 0.2088346 0.2093735 0.2088032 0.2088380 0.2088359
3.5 0.2119293 0.2143409 0.2145938 0.2145870 0.2145869
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 171

Table 1.2
Comparison of L2 and L1 errors in the MCB-DQM solutions of Example 1 for m ¼ 0:005 at different time levels t 6 3:50 with the errors obtained by earlier
schemes.

Methods N Mt t ¼ 1:7 t ¼ 2:4 t ¼ 3:1 t ¼ 3:25

3 3 3 3 3 3
L2  10 L1  10 L2  10 L1  10 L2  10 L1  10 L2  103 L1  103
MCB-DQM 121 0.01 0.00191 0.00777 0.00086 0.00308 0.00065 0.00331 0.001341 0.00918
QRTDQ [24] 101 0.001 0.109 0.434 0.100 0.339 0.091 0.266
BS.FEM [10] 50 0.100 0.857 2.576 0.423 1.242 0.230 0.680
C.S.C.[39] 50 0.100 0.857 2.576 0.423 1.242 0.235 0.688
Galerkin [46] 200 0.010 0.857 2.576 0.423 1.242 0.235 0.688
QBCM1[28] 200 0.010 0.017 0.061 0.012 0.058 0.601 4.434
QBCM2 [28] 200 0.010 0.358 1.211 0.251 0.807 0.630 4.790
PDQ [18] 200 0.010 0.015 0.056 0.011 0.064 0.584 4.301
CBCDQ [19] 101 0.001 0.210 0.680 0.190 0.530
t ¼ 2:5
QBCM [12] 200 0.010 0.072 0.311 0.051 0.189 1.129 8.983
CBCM [12] 200 0.010 2.466 27.577 2.111 25.15 1.925 21.084
QRKM [12] 200 0.010 0.026 0.091 0.031 0.115 1.111 8.000
t ¼ 3:00 t ¼ 3:50
MCB-DQM 121 0.010 0.00191 0.00777 0.00778 0.00275 0.00056 0.0017 0.006177 0.04335
MCB-CM [31] 241 0.010 0.0252 0.0994 0.0151 0.0549 0.0118 0.0414 0.0117 0.0486
[41] (b ¼ 0:5) 12001 0.010 0.38421 1.34728 0.49135 1.55470 0.51508 1.5529 0.525855 1.52196
[41](b ¼ 1) 12001 0.010 3.08966 10.4040 2.72048 8.29747 2.39922 6.9880 2.12110 5.94321
MCB-DQM 121 0.010 0.00191 0.00777 0.00778 0.00275 0.00056 0.0017 0.006177 0.04335

Table 1.3
Comparison of L2 and L1 errors in the MCB-DQM solutions of Example 1 for m ¼ 0:005 at t ¼ 3:6 with the errors obtained in [20].

MCB-DQM Korkmaz & Dağ [20]

Method I Method II Method III

L2  10 3 0:01 0:18 0:16 0:14

L1  103 0:07 0:46 0:52 0:54

−5
x 10
4.5

3.5

3
Absolute Error

2.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2
x

Fig. 1. Absolute errors in the MCB-DQM numeric solutions of Example 1 for m ¼ 0:005 at t ¼ 3:5 with h ¼ 0:01; Mt ¼ 0:01.

sinðpxÞexpðp2 m2 tÞ þ 4 sinð2pxÞexpð4p2 m2 tÞ
uðx; tÞ ¼ 2pm ; for x 2 ð0; 2Þ and t P 0; ð4:1Þ
4 þ cosðpxÞexpðp2 m2 tÞ þ 2 cosð2pxÞexpð4p2 m2 tÞ
where the initial condition is evaluated from (4.1), and the boundary conditions are taken to be uð0; tÞ ¼ 0 and uð2; tÞ ¼ 0.
In this example, we have computed L1 and L2 errors at t ¼ 0:1; 1:0 with the parameter h ¼ 0:01; Mt ¼ 0:01, and at the
different values of m. The comparison of the computed errors with the errors obtained by Mittal and Jain [[31] Table 5.1] are
reported in Table 2.1. It is evident that as the value of m decreases the absolute error decreases rapidly. Also, for a given value
of m, the computed errors are less than that obtained by Mittal and Jain [31], and hence, the numerical solutions produced by
172 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177

−5
x 10
4.5
t=1.7
4 t=2.5
t=3.0
3.5 t=3.5

Absolute Error
2.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2
x

Fig. 2. Absolute errors in the MCB-DQM numeric solutions of Example 1 for m ¼ 0:005 at t 6 3:5 with h ¼ 0:01; Mt ¼ 0:01.

0.45

0.4 t=0.0
t=1.7
0.35 t=2.5
t=3.0
0.3 t=3.5

0.25
u(x,t)

0.2

0.15

0.1

0.05

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
x

Fig. 3. Physical behavior of the MCB-DQM numeric solutions of Example 1 for m ¼ 0:005 at t 6 3:5 with h ¼ 0:01; Mt ¼ 0:01.

Table 2.1
Comparison of L1 and L2 in the MCB-DQM numeric solutions of Example 2 with the errors obtained in [31].

m t ¼ 0:1 t ¼ 1:0
Mittal & Jain [31] MCB-DQM Mittal & Jain [31] MCB-DQM
h ¼ 0:025; Mt ¼ 103 h ¼ 0:1; Mt ¼ 0:01 h ¼ 0:025; Mt ¼ 103 h ¼ 0:1; Mt ¼ 0:01

L1 L2 L1 L2 L1 L2 L1 L2

102 4:41E  03 3:55E  03 3:89E  03 3:41E  03 3:13E  02 2:66E  02 2:92E  02 2:63E  02

103 4:60E  05 3:72E  05 4:09E  05 3:55E  05 4:45E  04 3:59E  04 3:93E  04 3:45E  04
104 4:62E  07 3:74E  07 4:11E  07 3:56E  07 4:61E  06 3:72E  06 4:09E  06 3:55E  06
105 4:62E  09 3:74E  09 4:11E  09 3:56E  09 4:62E  08 3:74E  08 4:11E  08 3:56E  08
106 4:62E  11 3:74E  11 4:11E  11 3:56E  11 4:62E  10 3:74E  10 4:11E  10 3:56E  10

our method are accurate than [31], for the current problem. Also, the absolute errors at t ¼ 1 taking v ¼ 104 ; 105 ; 106
have been shown in Fig. 4. The physical behavior of the numerical solutions at m ¼ 0:01 for different time levels are depicted
in Fig. 5.

Example 3. The initial and the boundary conditions of the Burgers’ Eq. (1.1) over the region ½0; 1 with a ¼ 1, are considered
as in [26,34,16]:
uðx; 0Þ ¼ 4xð1  xÞ and uð0; tÞ ¼ uð1; tÞ ¼ 0:
The description of the numerical solutions of this example for m ¼ 0:1 and 0:01 is given below:
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 173

−4
x 10
8
v=1.0 E−06
6 v=1.0 E−05
v=1.0 E−04
4

Errors
0

−2

−4

−6

−8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x

Fig. 4. Errors in the MCB-DQM numeric solutions of Example 2 at t ¼ 1 for m ¼ 104 ; 105 and 106 taking h ¼ 0:01; Mt ¼ 0:01.

Table 3.1
Comparison of the computed results form Example 3 to Jiwari et al. [16], Kutluay et al. [26] for m ¼ 0:1, and the exact solutions.
x T [26] [16] MCB-DQM Exact
Mt ¼ 0:0001 Mt ¼ 0:0001 Mt ¼ 0:001
h ¼ 0:0125 h ¼ 0:04 h ¼ 0:025
0.25 0.4 0.32091 0.31744 0.317526 0.31752
0.8 0.20211 0.19952 0.199558 0.19956
1.0 0.16782 0.16557 0.165601 0.16560
3.0 0.02828 0.02775 0.027761 0.02775
0.50 0.4 0.58788 0.58443 0.584541 0.58454
0.8 0.37111 0.36733 0.367406 0.36740
1.0 0.30183 0.29830 0.298352 0.29834
3.0 0.04185 0.04106 0.041069 0.04106
0.75 0.4 0.65054 0.64556 0.645641 0.64562
0.8 0.39068 0.38526 0.385369 0.38534
1.0 0.30057 0.29582 0.295885 0.29586
3.0 0.03106 0.03043 0.030443 0.03044

Table 3.2
Comparison of the MCB-DQM numerical solutions of Example 3 with the numeric solutions due to Mittal & Jain [31] for m ¼ 0:01.

x t [31] MCB-DQM Exact

h ¼ 0:025 h ¼ 0:025
Mt ¼ 0:001 Mt ¼ 0:001
0.25 0.4 0.36225 0.36226 0.36226
0.6 0.28202 0.28204 0.28204
0.8 0.23044 0.23045 0.23045
1.0 0.19468 0.19469 0.19469
3.0 0.07613 0.07613 0.07613
0.50 0.4 0.68368 0.68368 0.68368
0.6 0.54832 0.54832 0.54832
0.8 0.45371 0.45371 0.45371
1.0 0.38567 0.38568 0.38568
3.0 0.15218 0.15218 0.15218
0.75 0.4 0.92052 0.92049 0.92050
0.6 0.78300 0.78297 0.78299
0.8 0.66272 0.66271 0.66272
1.0 0.56932 0.56932 0.56932
3.0 0.22782 0.22775 0.22774

(a) In Table 3.1, we have computed the numerical solutions with parameter values m ¼ 0:1; h ¼ 0:025; Mt ¼ 0:001 at dif-
ferent time levels. The comparison of our results with the results obtained in [16,26] are repored in Table 3.1. Also, the
numerical solutions obtained by Jiwari et al. [[16]Table 3] are better than the solutions obtained in [26,34]. Thus, we
found that our solutions are more accurate than the solutions obtained in [16,26,34] and approaching to exact
solutions.
174 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177

1 1

0.8 0.8

0.6

u(x,t)
0.6
u(x,t)

0.4 0.4

0.2
0.2 0 0
0.2 0.2
0
0 0.4 0 0.4
0 0.6 0.2 0.6
0.2 0.4
0.4 0.8 0.6 0.8
0.6
0.8 0.8
1 1 t 1 1 t
x x

Fig. 6. Physical behavior of the MCB-DQM numeric solutions of Example 3 at m ¼ 0:01 (left) and at m ¼ 0:1 (right) for t 6 1 with h ¼ 0:025; Mt ¼ 0:001.

0.1

0.05
u(x,t)

−0.05

1
−0.1 0.8
0 0.2 0.6
0.4 0.6 0.4
0.8 1 1.2 1.4 0.2
1.6 1.8 0
2
t
x

Fig. 5. Physical behavior of the MCB-DQM numeric solutions of Example 2 for m ¼ 0:01 at t 6 1 with m ¼ 0:01; h ¼ 0:01; Mt ¼ 0:01.

Table 4.1
Comparison of the MCB-DQM numeric solutions of Example 4 for m ¼ 1:0 with the numercal solutions obtained by Dağ et al. [12], Korkmaz [22], Mittal & Jain
[31], and exact solutions.

x t [12] [31] [22] MCB-DQM Exact

h ¼ 0:0125 h ¼ 0:025 h ¼ 0:025 h ¼ 0:025
Mt ¼ 104 Mt ¼ 0:00025 Mt ¼ 0:000125 Mt ¼ 0:00025

0.25 0.4 0.01357 0.01354 0.01363 0.0135710 0.01357

0.6 0.00189 0.00188 0.00190 0.0018888 0.00189
0.8 0.00026 0.00026 0.00026 0.0002624 0.00026
1.0 0.00004 0.00004 0.00003 0.0000365 0.00004
3.0 0.00000 0.00000 0.00000 0.0000000 0.00000
0.50 0.4 0.01923 0.01920 0.01932 0.0192336 0.01923
0.6 0.00267 0.00266 0.00269 0.0026719 0.00267
0.8 0.00037 0.00037 0.00037 0.0003712 0.00037
1.0 0.00005 0.00005 0.00005 0.0000516 0.00005
3.0 0.00000 0.00000 0.00000 0.0000000 0.00000
0.75 0.4 0.01362 0.01360 0.01369 0.0136298 0.01363
0.6 0.00189 0.00188 0.00190 0.0018899 0.00189
0.8 0.00026 0.00026 0.00026 0.0002625 0.00026
1.0 0.00004 0.00004 0.00003 0.0000365 0.00004
3.0 0.00000 0.00000 0.00000 0.0000000 0.00000

(b) In Table 3.2, we have computed the numerical solutions with parameter values m ¼ 0:01; h ¼ 0:025; Mt ¼ 0:001. Since
the numerical solutions obtained by Mittal and Jain [[31] Table 5.1] are better than the solutions obtained in [1,2,45],
hence the comparison of the obtained solutions with the exact solutions as well as with the solutions obtained in [31]
are reported in Table 3.2. It is observed that our solutions are accurate than the results obtained in [1,2,31,45].
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 175

Table 4.2
Comparison of the MCB-DQM numeric solutions of Example 4 for m ¼ 1:0 with the solutions obtained by Dağ et al. [12].
x [12] MCB-DQM [12] MCB-DQM Exact
h ¼ 0:0125 h ¼ 0:025 h ¼ 0:00625 h ¼ 0:0125
Mt ¼ 105 Mt ¼ 104 Mt ¼ 105 Mt ¼ 104
0.1 0.10952 0.109530 0.10953 0.109526 0.10954
0.2 0.20975 0.209771 0.20977 0.209766 0.20979
0.3 0.29184 0.291860 0.29186 0.291855 0.29190
0.4 0.34785 0.347874 0.34788 0.347869 0.34792
0.5 0.37149 0.371517 0.37153 0.371512 0.37158
0.6 0.35896 0.358981 0.35900 0.358975 0.35905
0.7 0.30983 0.309845 0.30986 0.309839 0.30991
0.8 0.22776 0.227773 0.22778 0.227766 0.22782
0.9 0.12065 0.120666 0.12067 0.120659 0.12069

Table 4.3
Comparison of the MCB-DQM numeric solutions of Example 4 for m ¼ 0:1 with the numercal solutions obtained by Dağ et al. [12],
Korkmaz [22], Mittal & Jain [31], and exact solutions.

x t [12] [31] [22] MCB-DQM Exact

h ¼ 0:0125 h ¼ 0:025 h ¼ 0:025 h ¼ 0:025
Mt ¼ 104 Mt ¼ 0:0025 Mt ¼ 0:00125 Mt ¼ 0:004

0.25 0.4 0.30890 0.30892 0.30910 0.3089280 0.30889

0.6 0.24075 0.24077 0.24093 0.2407550 0.24074
0.8 0.19569 0.19572 0.19586 0.1956840 0.19568
1.0 0.16258 0.16261 0.16274 0.1625700 0.16256
3.0 0.02720 0.02718 0.02720 0.0272047 0.02720
0.50 0.4 0.56965 0.56970 0.56973 0.5696530 0.56963
0.6 0.44723 0.44729 0.44736 0.4472170 0.44721
0.8 0.35925 0.35930 0.35943 0.3592450 0.35924
1.0 0.29192 0.29195 0.29213 0.2919250 0.29192
3.0 0.04019 0.04016 0.04032 0.0402085 0.04021
0.75 0.4 0.62538 0.62520 0.62573 0.6253490 0.62544
0.6 0.48715 0.48694 0.48760 0.4872040 0.48721
0.8 0.37385 0.37365 0.37434 0.3739350 0.37392
1.0 0.28741 0.28724 0.28788 0.2874930 0.28747
3.0 0.02976 0.02974 0.29881 0.0297753 0.02977

1
1
0.8
0.8
0.6
u(x,t)

0.6
u(x,t)

0.4
0.4
0.2
0.2 0
0 0
0.2 0
0 0.2
0.4 0.2
0 0.4
0.2 0.4
0.6
0.4 0.6 0.6
0.6 0.8 0.8
0.8
0.8
1 1 t x 1 1 t
x

Fig. 7. Physical behavior of MCB-DQM numeric solutions at m ¼ 0:1 (left) and at m ¼ 1:0 (right) of Example 4 for t 6 1 with h ¼ 0:025; Mt ¼ 0:001.

The physical behavior of the current problem for different time levels t 6 1, is depicted in Fig. 6. The similar figures are also
depicted in [1,31].

Example 4. In this example, the initial condition for the Burgers’ Eq. (1.1), for a ¼ 1, over the region ½0; 1 is considered as
given in [22]:
176 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177

uðx; 0Þ ¼ sinðpxÞ; ð4:2Þ

and the boundary conditions
uð0; tÞ ¼ uð1; tÞ ¼ 0: ð4:3Þ
The analytical solution of this problem is given by Cole [11] in terms of an infinite series as
P 2 2
4pm 1 j¼1 jIj ð2pmÞ sinðjpxÞexpðj p mtÞ
1
uðx; tÞ ¼ P 2 2
; ð4:4Þ
I0 ð2p1 mÞ þ 2 1 j¼1 I j ð2pmÞ cosðjpxÞexpðj p mtÞ
1

where Ij are the modified Bessel’s functions.

The numerical solutions of this example for different values of m are given below:

(a) In Table 4.1, we have computed the numerical solutions of this example at different time levels with parameter values
m ¼ 1:0; h ¼ 0:025 and Mt ¼ 0:00025. The comparison of our results with the exact solutions as well as the solutions
obtained in [12,31,22] are reported in Table 4.1. It is found that our method produces comparable results as obtained
in [12,31,22].
Further, the numerical solutions are computed at t ¼ 0:1 with parameter values h ¼ 0:025; 0:0125
m ¼ 1:0; Mt ¼ 0:0001, and compared with the solutions obtained in Dağ et al. [[12] Table 1]. The results are reported
in Table 4.2. It is found that we require half of the grid points to produces the results similar to [12].
(b) Also, the numerical solutions of this example are computed at different time levels with parameter values
m ¼ 0:1; h ¼ 0:025 and Mt ¼ 0:004. The comparison of our solutions with the exact solutions as well as the solutions
obtained in [12,31,22] are reported in Table 4.3. It is evident that the MCB-DQM numerical solutions are better as com-
pared to the results obtained in [12,31,22].

The physical behavior of this example for m ¼ 0:1 and m ¼ 1:0 are depicted in Fig. 7.

5. Conclusion

In this paper, we have developed a method (MCB-DQM) to solve nonlinear partial differential equations. In this method,
the modified cubic B-splines are used in differential quadrature method as basis function to evaluate the weighting coeffi-
cients, and hence the derivatives. In this way, we find a system of ordinary differential equations (ODEs) which is solved by
SSP-RK43 scheme. To check the efficiency and accuracy of the method, four examples of Burgers’ equation are included with
their numerical solutions, L2 and L1 errors and done the comparisons with the results given in the literature. It is evident that
our method produces better results as compared to the results obtained by almost all the schemes available in the literature,
and approaching to the exact solutions.
The cubic B-spline basis functions are modified in such a way that it reduces matrix size and complexity when applied
with differential quadrature method. In this method we require less number of grid points as compared to the earlier given
methods. This method is hence easy to implement and economical in terms of data complexity, which results in less errors
and so, the easiness of the implementation of MCB-DQM, and low memory storage can be counted as advantages of this
method. Also, this method can be easily implemented to solve two-dimensional nonlinear partial differential equations.

Acknowledgement

The authors thank the anonymous referees for their time, effort, and extensive comments which improve the quality of
the presentation of the paper.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/
j.amc.2013.08.071.

References

[1] A. Asaithambi, Numerical solution of the Burgers’ equation by automatic differentiation, Appl. Math. Comput. 216 (2010) 2700–2708.
[2] K. Altparmak, Numerical solution of Burgers’ equation with factorized diagonal Padè approximation, Int. J. Numer. Methods Heat Fluid Flow 21 (3)
(2011) 310–319.
[3] E.N. Aksan, A. Ozdes, A numerical solution of Burgers’ equation, Appl. Math. Comput. 156 (2004) 395–402.
[4] E.N. Aksan, Quadratic B-spline finite element method for numerical solution of the Burgers’ equation, Appl. Math. Comput. 174 (2006) 884–896.
[5] H. Bateman, Some recent researches on the motion of fluids, Mon. Weather Rev. 43 (1915) 163–170.
[6] J.M. Burgers, Mathematical example illustrating relations occurring in the theory of turbulent fluid motion, Trans. Roy. Neth. Acad. Sci. Amsterdam 17
(1939) 1–53.
[7] J.M. Burgers, A mathematical model illustrating the theory of turbulence, Adv. Appl. Mech., vol. I, Academic Press, New York, 1948, pp. 171–199.
 G. Arora, B.K. Singh / Applied Mathematics and Computation 224 (2013) 166–177 177

[8] R. Bellman, B.G. Kashef, J. Casti, Differential quadrature: a technique for the rapid solution of nonlinear differential equations, J. Comput. Phy. 10 (1972)
40–52.
[9] E. Benton, G.W. Platzman, A table of solutions of the one dimensional Burgers’ equations, Q. Appl. Math. 30 (1972) 195–212.
[10] J. Chung, E. Kim, Y. Kim, Asymptotic agreement of moments and higher order contraction in the Burgers equation, J. Diff. Equ. 248 (10) (2010) 2417–
2434.
[11] J.D. Cole, On a quasi-linear parabolic equations occurring in aerodynamics, Quart. Appl. Math. 9 (1951) 225–236.
_ Dağ, D. Irk, A. Sahin, B-Spline collocation methods for numerical solutions of the Burgers’ equation, Math. Probl. Eng. 5 (2005) 521–538.
[12] I.
[13] A. Dogan, A. Galerkin, Finite element approach to Burgers’ equation, Appl. Math. Comput. 157 (2004) 331–346.
[14] I.A. Hassanien, A.A. Salama, H.A. Hosham, Fourth-order finite difference method for solving Burgers’ equation, Appl. Math. Comput. 170 (2005) 781–
800.
[15] Z. Jiang, R. Wang, An improved numerical solution of Burgers’ equation by cubic B-spline Quasi-interpolation, J. Inf. Comput. Sci. 7 (5) (2010) 1013–
1021.
[16] R. Jiwari, R.C. Mittal, 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.
[17] M.K. Kadalbajoo, K.K. Sharma, A. Awasthi, A parameter-uniform implicit difference scheme for solving time-dependent Burgers’ equation, Appl. Math.
Comput. 170 (2005) 1365–1393.
[18] A. Korkmaz, Numerical solutions of some nonlinear partial differential equations using differential quadrature method, Thesis of Master Degree,
Eskisßehir Osmangazi University, 2006.
[19] A. Korkmaz, Numerical solutions of some one dimensional partial differential equations using B-spline differential quadrature method, Doctoral
Dissertation , Eskisßehir Osmangazi University, 2010.
[20] A. Korkmaz, I. _ Dağ, Cubic B-spline differential quadrature methods and stability for Burgers’ equation, Eng. Comput. Int. J. Comput. Aided Eng. Software
30 (3) (2013) 320–344.
[21] A. Korkmaz, I. _ Dağ, Cubic B-spline differential quadrature methods for the advection-diffusion equation, Int. J. Numer. Methods Heat Fluid Flow 22 (8)
(2012) 1021–1036.
[22] A. Korkmaz, I. _ Dağ, Shock wave simulations using sinc differential quadrature method, Eng. Comput. Int. J. Comput. Aided Eng. Software 28 (6) (2011)
654–674.
[23] A. Korkmaz, I. _ Dağ, Polynomial based differential quadrature method for numerical solution of nonlinear Burgers’ equation, J. Franklin Inst. 348 (10)
(2011) 2863–2875.
[24] A. Korkmaz, A.M. Aksoy, I. _ Dağ, Quartic B-spline differential quadrature method, Int. J. Nonlinear Sci. 11 (4) (2011) 403–411.
[25] S. Kutluay, A.R. Bahadir, A. Ozdes, Numerical solution of one-dimensional Burgers’ equation: explicit and exact-explicit finite difference methods, J.
Comput. Appl. Math. 103 (1999) 251–261.
[26] S. Kutulay, A. Esen, I._ Dağ, Numerical solutions of the Burgers’ equation by the least-squares quadratic B-spline finite element method, J. Comput. Appl.
Math. 167 (2004) 21–33.
[27] A.K. Khalifa, K.I. Noor, M.A. Noor, Some numerical methods for solving Burgers’ equation, Int. J. Phys. Sci. 6 (7) (2011) 1702–1710.
[28] S. Lin, C. Wang, Z. Dai, New exact traveling and non-traveling wave solutions for ð2 þ 1Þ dimensional Burgers equation, Appl. Math. Comput. 216 (10)
(2010) 3105–3110.
[29] W. Liao, An implicit fourth-order compact finite difference scheme for one-dimensional Burgers’ equation, Appl. Math. Comput. 206 (2008) 755–764.
[30] R.C. Mittal, P. Singhal, Numerical solution of Burgers’ equation, Commun. Numer. Methods Eng. 9 (1993) 397–406.
[31] R.C. Mittal, R.K. Jain, Numerical solutions of nonlinear Burgers’ equation with modified cubic B-splines collocation method, Appl. Math. Comput. 218
(2012) 7839–7855.
[32] M.M. Cecchi, R. Nociforo, P.P. Grego, Space-time finite elements numerical solution of Burgers problems, Le Matematiche LI (Fasc. I) (1996) 43–57.
[33] T. Özis, A. Esen, S. Kutluay, Numerical solution of Burgers’ equation by quadratic B-spline finite elements, Appl. Math. Comput. 165 (2005) 237–249.
[34] T. Ozis, E.N. Aksan, A. Ozdes, A finite element approach for solution of Burgers’ equation, Appl. Math. Comput. 139 (2003) 417–428.
[35] J.R. Quan, C.T. Chang, New insights in solving distributed system equations by the quadrature methods-I, Comput. Chem. Eng. 13 (1989) 779–788.
[36] J.R. Quan, C.T. Chang, New insights in solving distributed system equations by the quadrature methods-II, Comput. Chem. Eng. 13 (1989) 1017–1024.
[37] M.A. Ramadan, T.S. El-Danaf, F.E.I. Abd Alaal, Application of the non-polynomial spline approach to the solution of the Burgers’ equation, Open Appl.
Math. J. 1 (2007) 15–20.
_ Dağ, Quartic B-spline collocation method to the numerical solutions of the Burgers’ equation, Chaos Solitons Fractals 32 (2007) 1125–1137.
[38] B. Saka, I.
[39] A.H. Salas, Symbolic computation of solutions for a forced Burgers equation, Appl. Math. Comput. 216 (1) (2010) 18–26.
[40] C. Shu, B.E. Richards, Application of generalized differential quadrature to solve two dimensional incompressible Navier–Stokes equations, Int. J.
Numer. Meth. Fluids 15 (1992) 791–798.
[41] Shu-Sen Xie, S. Heo, Seokchan Kim, Gyungsoo Woo, Sucheol Yi, Numerical solution of one-dimensional Burgers’ equation using reproducing kernel
function, J. Comput. Appl. Math. 214 (2008) 417–434.
[42] C. Shu, Y.T. Chew, Fourier expansion-based differential quadrature and its application to Helmholtz eigenvalue problems, Commun. Numer. Methods
Eng. 13 (8) (1997) 643–653.
[43] C. Shu, H. Xue, Explicit computation of weighting coefficients in the harmonic differential quadrature, J. Sound Vib. 204 (3) (1997) 549–555.
[44] J.R. Spiteri, S.J. Ruuth, A new class of optimal high-order strongstability-preserving time-stepping schemes, SIAM J. Numer. Anal. 40 (2) (2002) 469–
491.
[45] Min Xu, Ren-Hong Wang, Ji-Hong Zhang, Qin Fang, A novel numerical scheme for solving Burgers’ equation, Appl. Math. Comput. 217 (2011) 4473–
4482.
[46] L. Zhang, J. Ouyang, X. Wang, X. Zhang, Variational multiscale element-free Galerkin method for 2D Burgers equation, J. Comput. Phy. 229 (19) (2010)
7147–7161.