You are on page 1of 27

Applied Mathematical Modelling 37 (2013) 9455–9481

Contents lists available at SciVerse ScienceDirect

Applied Mathematical Modelling


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

A numerical assessment of parabolic partial differential


equations using Haar and Legendre wavelets
Siraj-ul-Islam a,⇑, Imran Aziz a,b, A.S. Al-Fhaid c, Ajmal Shah a,d
a
Department of Basic Sciences, University of Engineering and Technology, Peshawar, Pakistan
b
Department of Mathematics, University of Peshawar, Pakistan
c
Department of Mathematics, King Abdulaziz University, Jeddah, Saudi Arabia
d
Department of Mathematics, Edwardes College, Peshawar, Pakistan

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

Article history: In this paper we present two new numerically stable methods based on Haar and Legendre
Received 21 September 2012 wavelets for one- and two-dimensional parabolic partial differential equations (PPDEs).
Accepted 19 April 2013 This work is the extension of the earlier work [1–3] from one- and two-dimensional
Available online 16 May 2013
boundary-value problems to one- and two- dimensional PPDEs. Two generic numerical
algorithms are derived in two phases. In the first stage a numerical algorithm is derived
Keywords: by using Haar wavelets and then in the second stage Haar wavelets are replaced by Legen-
Haar wavelets
dre wavelets in quest for better accuracy. In the proposed methods the time derivative is
Legendre wavelets
Parabolic partial differential equations
approximated by first order forward difference operator and space derivatives are approx-
Dirichlet jump problem imated using Haar (Legendre) wavelets. Improved accuracy is obtained in the form of
Burgers’ equation wavelets decomposition. The solution in this process is first obtained on a coarse grid
Collocation methods and then refined towards higher accuracy in the high resolution space. Accuracy wise per-
formance of the Legendre wavelets collocation method (LWCM) is better than the Haar
wavelets collocation method (HWCM) for problems having smooth initial data or having
no shock phenomena in the solution space. If sharp transitions exists in the solution space
or if there is a discontinuity between initial and boundary conditions, LWCM loses its accu-
racy in such cases, whereas HWCM produces a stable solution in such cases as well. Con-
trary to the existing methods, the accuracy of both HWCM and LWCM do not degrade in
case of Neumann’s boundary conditions. A distinctive feature of the proposed methods is
its simple applicability for a variety of boundary conditions. Performances of both HWCM
and LWCM are compared with the most recent methods reported in the literature. Numer-
ical tests affirm better accuracy of the proposed methods for a range of benchmark
problems.
Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction

In recent years, wavelets have attracted much attention due to their extensive use in a variety of numerical approxima-
tions. A survey of some of the early work can be found in [4]. Specific applications of wavelets in the context of numerical
methods include numerical solutions of integral equations and numerical integration [5–9], ordinary differential equations
[10,2,3], partial differential equations [11,1] and fractional partial differential equations [12]. Various type of wavelets have
been used in such applications, for example, Daubechies [13], Battle-Lemarie [14], B-spline [10], Chebyshev [15], Haar wave-

⇑ Corresponding author.
E-mail addresses: siraj-ul-islam@nwfpuet.edu.pk, siraj.islam@gmail.com (Siraj-ul-Islam).

0307-904X/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.apm.2013.04.014
9456 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

lets [16,17,2], Legendre wavelets [18]. Haar wavelets have been used by many researchers because of their simplicity and
better convergence in the case of sharp transition of a function. Some of the relevant work can be found in Refs.
[17,19,20,16,21–23]. The weaker side of using the Haar basis functions for approximating smooth functions is that they
are lower in accuracy due their non-smooth character. To cover this aspect, smooth Legendre wavelets are considered to
get more accurate approximation in the absence of jumps or sharp transitions. Legendre wavelets use Legendre polynomials
as their basis functions. Because of their improved smoothness and good interpolating properties, accuracy of Legendre
wavelets is better than Haar wavelets in case of approximating smooth functions even on a coarse grid. Applications of
Legendre wavelets for numerical approximations can be found in Refs. [24,25].
In case of partial differential equations (PDEs), there are certain situations when an initial condition or boundary condi-
tions inherit some sort of discontinuity or shock phenomena. In such situations, accuracy of the numerical method cannot be
improved by using high order basis functions. Higher order basis functions suffer from the so-called Gibbs phenomena which
causes wild oscillations near the points of sharp transitions. Consequently the unwanted high frequency components of the
numerical solution caused due to Gibbs phenomena are removed through some post processing technique which introduces
artificial viscosity to retain stable solution. Artificial diffusion, although retaining stability, but at the same time adversely
affects the accuracy of the numerical method due to smearing or over smoothing. The use of Haar basis act as a natural filter
and can provide a better alternative in such situations. The solution thus obtained is both stable and accurate without resort-
ing to some post processing technique. This is made possible due to natural congruence between the structure of the Haar
basis functions and the shape of the functions having sharp transitions.
Mathematical models of basic flow equations which describe unsteady transport problems are governed by a single or a
system of nonlinear parabolic and hyperbolic PDEs. In general parabolic equations with or without reaction terms are used as
a fundamental tool to model a wide class of phenomena occurring in physical and biological sciences. These include heat
transfer, transport phenomena and reaction of chemical species, population dynamics, adsorption of pollutants in soil and
the diffusion of neutrons etc. (for details see [26–36] and the references therein).
In this paper we apply the HWCM and LWCM to one-dimensional diffusion problem [36], two-dimensional Dirichlet jump
problem [37,38], two-dimensional reaction diffusion equation [39,40] and two-dimensional Burgers’ equation [41]. Existing
literature contains different numerical algorithms designed for the numerical solution of different classes of PPDEs. These
include [42,43,29,44,32,33,35,34,45,36–40] and the references therein.
This paper proposes two new collocation methods HWCM and LWCM for numerical solution of PPDEs. The present work
is a continuation of our earlier work [1–3] where Haar and Legendre wavelets are being applied for the numerical solution of
boundary value problems (BVPs). This approach has the following advantages:

i. HWCM uses simple box functions and hence formulation of the numerical method based on Haar wavelets is
straightforward.
ii. The box functions are the natural choice when there is a shock or discontinuity in the initial and boundary conditions.
iii. HWCM and LWCM can be applied easily and accurately to different types of boundary conditions in the context of
PPDEs.
iv. LWCM provides a more accurate solution because of their best approximating properties compared to HWCM and
other well-established methods in situations where initial data is smooth.

2. Haar wavelets

A wavelet family ðwj;i ðxÞÞj2N; i2Z is an orthonormal subfamily of the Hilbert space L2 ðRÞ with the property that all function
in the wavelet family are generated from a fixed function w called mother wavelet through dilations and translations. The
wavelet family satisfies the following relation:

wj;i ðxÞ ¼ 2j=2 wð2j x  iÞ: ð1Þ

The Haar wavelet family defined on the interval ½a; bÞ consists of the following functions:
8
<1
> for x 2 ½a; bÞ;
Hi ðxÞ ¼ 1 for x 2 ½b; cÞ; i ¼ 2; 3; . . . ; ð2Þ
>
:
0 elsewhere;
where
k
a ¼ a þ ðb  aÞ ;
m
ðk þ 0:5Þ
b ¼ a þ ðb  aÞ ;
m ð3Þ
ðk þ 1Þ
c ¼ a þ ðb  aÞ ;
m
j
m ¼ 2 ; j ¼ 0; 1; . . . ; k ¼ 0; 1; . . . ; m  1:
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9457

The integer j indicates the level of the wavelet and k is the translation parameter. The relation between i; m and k is given
by i ¼ m þ k þ 1. Any function f ðxÞ in L2 ð½a; bÞ can be expressed as an infinite sum of Haar wavelets as follows:
X
1
f ðxÞ ¼ ai Hi ðxÞ;
i¼1

where ai are real constants. The function H1 ðxÞ is called the scaling function for the Haar wavelets family and is defined as
follows:

1 for x 2 ½a; bÞ;
H1 ðxÞ ¼ ð4Þ
0 elsewhere:
For approximation purpose we consider a maximum value J of the integer j that denotes level of the Haar wavelet in the
above definition. The integer J is then called maximum level of resolution. We also define the integer M ¼ 2J . With these
notations any function f ðxÞ in L2 ð½a; bÞ can be approximated as a finite sum of Haar wavelets as follows:
X
2M
f ðxÞ  ai Hi ðxÞ:
i¼1

The following notations are introduced:


Z x
ð1Þ
Hi ðxÞ ¼ Hi ðx0 Þ dx0 ;
a

Z x
ð2Þ ð1Þ
Hi ðxÞ ¼ Hi ðx0 Þ dx0 ;
a

Z b
ð1Þ
CH
i ¼ Hi ðxÞ dx:
a

These integrals can be evaluated using Eq. (2) and are given as follows:
8
< x  a for x 2 ½a; bÞ;
>
ð1Þ
Hi ðxÞ ¼ c  x for x 2 ½b; cÞ;
>
:
0 elsewhere;
8
>
>
1
ðx  aÞ2 for x 2 ½a; bÞ;
>
> 2
>
< ðbaÞ2  1 ðc  xÞ2
ð2Þ 4m2 2
for x 2 ½b; cÞ;
Hi ðxÞ ¼
> ðbaÞ2
>
>
> for x 2 ½c; 1Þ;
>
: 4m
2

0 elsewhere:

ðb  aÞ2
CH
i ¼ ;
4m2
where a, b, c and m are as defined in Eq. (3).

3. Legendre wavelets

For any positive integer k, the Legendre wavelets family is defined [46] as follows:
( qffiffiffiffiffiffiffiffiffiffiffiffi
m þ 12 2k=2 Lm ð2k x  2n þ 1Þ; for 2n2
2k
6 x < 2n
2k
Lm;n ðxÞ ¼
0; otherwise
k1
where n ¼ 1; 2; . . . ; 2 and m ¼ 0; 1; 2; . . .. Here, Lm ðxÞ are the Legendre polynomials of order m. Legendre polynomials can
be calculated recursively by using the following relation:
L0 ðxÞ ¼ 1; L1 ðxÞ ¼ x;
   
2k þ 1 k
Lkþ1 ðtÞ ¼ xLk ðxÞ  Lk1 ðxÞ; k ¼ 1; 2; 3; . . . :
kþ1 kþ1
Equivalently, for any positive integer k, we can define the Legendre wavelets family as:
9458 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

( qffiffiffiffiffiffiffiffiffiffiffiffi
m þ 12 2k=2 Lm ð2k x  2n þ 1Þ; for 2n2
2k
6 x < 2n
2k
;
Li ðxÞ ¼ ð5Þ
0; otherwise;
where n ¼ 1; 2; . . . ; 2k1 ; m ¼ 0; 1; 2; . . . and i ¼ n þ 2k1 m. Any function f ðxÞ in L2 ð½0; 1Þ can be expanded by Legendre wave-
lets series [24] as given below:

X 2k1
1 X X
1
f ðxÞ ¼ am;n Lm;n ðxÞ ¼ ai Li ðxÞ: ð6Þ
m¼0 n¼1 i¼1

For approximations, the above series may be truncated and written as follows:
X
N
f ðxÞ ¼ ai Li ðxÞ ð7Þ
i¼1

where N ¼ 2k1 M and Legendre polynomials used in the approximation are of degree less than M. The following notations
are introduced:
Z x
ð1Þ
Li ðxÞ ¼ Li ðx0 Þdx0
0

Z x
ð2Þ ð1Þ
Li ðxÞ ¼ Li ðx0 Þdx0
0

Z 1
ð1Þ
CLi ¼ Li ðx0 Þdx0 :
0

4. Governing equations

In this module, we consider one- and two-dimensional PPDEs. A short description of the governing equations is also gi-
ven. In some cases, these equations pose challenge for numerical methods either due to discontinuity between boundary and
initial conditions or appearance of shock phenomena in the solution space at later time.

4.1. One-dimensional deterministic diffusion model

The optimal deterministic one-dimensional diffusion model in optic tomography for signal reconstruction is considered
in [36]. The diffusion variable uðx; tÞ denotes the photon density at any position x 2 ða; bÞ  R and at time t 2 ½t 0 ; t 0 þ t, where
t0 is the initial time and t is the time period. The diffusion parameter DðxÞ is defined by Addam et al. [36] in the following
manner,
1
DðxÞ ¼ ;
3ðla ðxÞ þ lr ðxÞÞ
where ls and la denote the scattering and absorption functions respectively, lr ¼ ð1  #Þls ðxÞ; # 2 ½1; 1 is the anisotropic
parameter. The time-domain diffusion problem is described in [36] such that for all ðx; tÞ 2 ða; bÞ  ½t0 ; t0 þ t, the equation of
photon density uðx; tÞ satisfies the following one-dimensional PPDE
 
1 @uðx; tÞ @ @uðx; tÞ
 DðxÞ þ la ðxÞuðx; tÞ ¼ f ðx; tÞ; ð8Þ
CðxÞ @t @x @x
Co
where CðxÞ ¼ VðxÞ , is the speed of light in the domain ½a; b; C o is the speed of light in vacuum, V is non-negative real valued
refractive index function defined on ½a; b. To deal simultaneously with the Dirichlet boundary conditions together with the
Robin boundary conditions, the following constants ba and bb were introduced in [36]:
( (
0 for fa ¼ 0 0 for fb ¼ 0
ba ¼ 1 bb ¼ 1
fa
for fa > 0; fb
for fb > 0:

The optic parameters V; D and la can either be chosen constants or variables. In the present work, we consider both the
constant optic parameters D ¼ Do ; V ¼ V o ; la ¼ lao and variable optic parameters VðxÞ ¼ V o ð1 þ 0:5ex Þ; la ðxÞ ¼ lao ex ;
DðxÞ ¼ D2o ðx þ 2Þ. The boundary conditions are given below:
ba @uða; tÞ
uða; tÞ  DðaÞ ¼ g a ðtÞ; ð9Þ
2 @x
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9459

bb @uðb; tÞ
uðb; tÞ þ DðbÞ ¼ g b ðtÞ: ð10Þ
2 @x

4.2. Two dimensional diffusion reaction equation

Consider the dimensionless form of two-dimensional diffusion reaction equation [40] defined on domain X with bound-
ary C
@uðx; tÞ
¼ gL½uðx; tÞ þ juðx; tÞ þ gðx; tÞ; ð11Þ
@t
with initial condition
uðx; t0 Þ ¼ u0 ; x2X[C
and Dirichlet or Neumann boundary conditions
B½uðx; tÞ ¼ f ðx; tÞ; t P t 0 ; x 2 C;
tr
where u; t; x ¼ ½x; y are the diffusion, time and space variables respectively, tr represents the matrix transpose, g and f are
the known functions of x and t, C ¼ CD þ CN , where CN and CD are boundaries that satisfy Neumann and Dirichlet boundary
conditions, respectively. The quantities g and j are real constants, L is a differential operator consisting of first- or second-
order derivatives of space variables and B is a first-order differential operator with respect to space variables in the case of
the Neumann boundary conditions and is identity operator in the case of the Dirichlet boundary conditions.

4.3. Dirichlet jump problem

Consider the two-dimensional diffusion equation having discontinuity between initial and boundary conditions [37] of
the form given below:
@uðx; tÞ
¼ g L½uðx; tÞ; ð12Þ
@t
with initial condition
uðx; 0Þ ¼ 1; x 2 X [ C
and Dirichlet jump boundary condition
uðx; tÞ ¼ 0; x 2 C; t P 0;
where g is real constant and L is differential operator as defined in the previous case.

4.4. Two-dimensional Burgers’ equation

Consider the dimensionless form of the nonlinear two-dimensional Burgers’ equation [33] defined on domain X and
boundary C
@uðx; tÞ
þ uðx; tÞ L½uðx; tÞ ¼ j52 uðx; tÞ; ð13Þ
@t
with initial condition
uðx; t0 Þ ¼ u0 ; x2X[C
and Dirichlet boundary conditions
uðx; tÞ ¼ f ðx; tÞ; t P t 0 ; x 2 C;
where j is real constant and L is differential operator consisting of first-order derivatives of space variables x; y and u; t; x
are defined as in previous case.

5. Time discretization

Let Dt be the time-step size, and t ¼ t0 þ Dt be the time discretization, where t 0 refers to the beginning time of every time
step, and t refers to the end of the time step. For a time period ½t0 ; t, the time derivative in given differential equations are
approximated by Euler’s formula
@uðx; tÞ uðx; tÞ  uðx; t0 Þ
 : ð14Þ
@t Dt
9460 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

Let h 2 ½0; 1. The parameter h is used in the time discretization of Eqs. (11)–(13) as given below:
uðx; t 0 þ h DtÞ  huðx; tÞ þ ð1  hÞuðx; t0 Þ;
gðx; t0 þ h DtÞ  hgðx; tÞ þ ð1  hÞgðx; t 0 Þ;
L½uðx; t0 þ h DtÞ  hL½uðx; tÞ þ ð1  hÞL½uðx; t0 Þ:
Then Eq. (11) can be discretized in time–space as
ð1  jDthÞuðx; tÞ  hDtðgL½uðx; tÞ þ gðx; tÞÞ ¼ uðx; t 0 Þ þ ð1  hÞDtðgLuðx; t 0 Þ þ juðx; t 0 Þ þ gðx; t 0 ÞÞ; x 2 X: ð15Þ
Similarly Eq. (12) can be discretized in time–space as
uðx; tÞ  Dt ghL½uðx; tÞ ¼ uðx; t 0 Þ þ ð1  hÞDtgL½uðx; t 0 Þ; x 2 X: ð16Þ
Also Eq. (13) can be discretized in time–space as
 
ð1 þ DthL½uðx; tÞÞuðx; tÞ  jDthr2 uðx; tÞ ¼ uðx; t 0 Þ þ Dtð1  hÞ jr2 uðx; t 0 Þ  uðx; t0 ÞLuðx; t 0 Þ ; x 2 X: ð17Þ

These are the commonly used two-level time stepping strategies which approximate uðx; tÞ in X from the values given at
the initial time t0 and further time t 0 þ Dt.
Eqs. (15)–(17) reduce to explicit, Crank–Nicolson and fully implicit schemes for h = 0, 0.5, 1, respectively. In this paper, we
take h = 1. Computational cost of the algorithm depends on the time-step size Dt. In the frame work of explicit time stepping
procedure, a small value of time step is required to ensure a stable solution. This in turn results in slow convergence. This
problem can be overcome by using Runge–Kutta method of order 4 for time integration. In the case of explicit method the
magnitude of time-step size in one-dimensional case is restricted by the following condition:
Dt 1
2
6 : ð18Þ
h 2
In two-dimensional problems the restriction is further squeezed and is given as
Dt 1
2
6 : ð19Þ
h 4

6. Solution procedure for space derivatives by Haar and Legendre wavelets

In this section, we use Haar and Legendre wavelets for spatial discretization of one- and two-dimensional PPDEs. We pro-
pose the new space discretization first for one-dimensional problem with Dirichlet together with Robin’s type boundary con-
ditions. In the next step, spatial discretization of two-dimensional PPDEs is derived for Dirichlet, Neumann, mixed type and
periodic boundary condition. The nodal points arrangement for the Haar wavelets algorithm is given as:
ðk  0:5Þ
xk ¼ a þ ðb  aÞ ; k ¼ 1; 2; . . . ; 2M; ð20Þ
2M

ðl  0:5Þ
yl ¼ c þ ðd  cÞ ; l ¼ 1; 2; . . . ; 2M:
2M
For Legendre wavelets algorithm, the collocation points are defined as:
k  0:5
xk ¼ a þ ðb  aÞ ; k ¼ 1; 2; . . . ; N; ð21Þ
N

l  0:5
yl ¼ c þ ðd  cÞ ; l ¼ 1; 2; . . . ; N;
N
where N ¼ 2k1 M.

6.1. One dimensional space discretization based on Haar and Legendre wavelets

To construct approximation for Eq. (8) by Haar and Legendre wavelets, we assume that

@ 2 uðx; tÞ X2M

2
¼ ai ðtÞ Hi ðxÞ: ð22Þ
@x i¼1

Similarly for Legendre wavelets, we assume


Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9461

@ 2 uðx; tÞ X N

2
¼ ai ðtÞ Li ðxÞ; ð23Þ
@x i¼1

where N ¼ 2k1 M.

6.2. Two-dimensional space discretization based on Haar and Legendre wavelets

In this part we derive Haar and Legendre wavelets based algorithms for two-dimensional PPDEs.
To further this idea, we assume the following Haar approximations for second order derivatives w.r.t x; y

@ 2 uðx; y; tÞ X2M X2M

2
¼ aij ðtÞ Hi ðxÞ Hj ðyÞ ð24Þ
@x j¼1 i¼1

and

@ 2 uðx; y; tÞ X2M X2M

2
¼ bij ðtÞ Hi ðxÞ Hj ðyÞ: ð25Þ
@y j¼1 i¼1

Similarly in case of Legendre wavelets, the approximation for second order derivatives w.r.t x; y are given by,

@ 2 uðx; y; tÞ X N X N

2
¼ aij ðtÞ Li ðxÞ Lj ðyÞ ð26Þ
@x j¼1 i¼1

and

@ 2 uðx; y; tÞ X N X N

2
¼ bij ðtÞ Li ðxÞ Lj ðyÞ: ð27Þ
@y j¼1 i¼1

7. Boundary conditions for one-dimensional PPDE

In this section, we consider Dirichlet boundary conditions together with Robin type boundary conditions for one-dimen-
sional deterministic diffusion model (8).

7.1. Boundary conditions using Haar wavelets

Consider the boundary conditions given by Eqs. (9) and (10). Integrating Eq. (22) from a to b, we get
@uðb; tÞ @uða; tÞ
 ¼ a1 ðtÞðb  aÞ: ð28Þ
@x @x
Similarly integrating Eq. (22) from a to x and then from a to b, we get

@uða; tÞ X2M
uðb; tÞ  uða; tÞ  ðb  aÞ ¼ ai ðtÞCH
i : ð29Þ
@x i¼1

Again integrating Eq. (22) twice from a to x, we get

@uða; tÞ X2M
uðx; tÞ ¼ uða; tÞ þ ðx  aÞ þ ai ðtÞH2i ðxÞ: ð30Þ
@x i¼1

7.2. Boundary conditions using Legendre wavelets

In the case of Legendre wavelets, we obtain the following equations by adopting a similar procedure as above,
k1
@uðb; tÞ @uða; tÞ ðb  aÞ X2
 ¼ pffiffiffik1 ai ðtÞ; ð31Þ
@x @x
2 i¼1

@uða; tÞ X N
uðb; tÞ  uða; tÞ  ðb  aÞ ¼ ai ðtÞCLi ; ð32Þ
@x i¼1
9462 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

@uða; tÞ X N
ð2Þ
uðx; tÞ ¼ uða; tÞ þ ðx  aÞ þ ai ðtÞLi ðxÞ: ð33Þ
@x i¼1

8. Boundary conditions for two-dimensional PPDEs

In this section we consider Dirichlet, Neumann, mixed type and periodic boundary conditions for both Haar and Legendre
wavelets.

8.1. Dirichlet conditions using Haar wavelets

In general, Dirichlet boundary conditions can be written as;


uða; y; tÞ ¼ gða; y; tÞ; uðb; y; tÞ ¼ gðb; y; tÞ; y 2 ½c; dÞ; ð34Þ

uðx; c; tÞ ¼ gðx; c; tÞ; uðx; d; tÞ ¼ gðx; d; tÞ; x 2 ½a; bÞ; ð35Þ


where g is some known function. Integrating Eq. (24), we obtain
 
@uðx; y; tÞ gðb; y; tÞ  gða; y; tÞ X2M X2M
ð1Þ CH
¼ þ aij ðtÞ Hi ðxÞ  i Hj ðyÞ; ð36Þ
@x ba j¼1 i¼1
ba

which on integration and further simplification gives the following form,


X
2M X2M  
gðb; y; tÞ  gða; y; tÞ ð2Þ CH
uðx; y; tÞ ¼ gða; y; tÞ þ ðx  aÞ þ aij ðtÞ Hi ðxÞ  i ðx  aÞ Hj ðyÞ: ð37Þ
ba j¼1 i¼1
ba

Similarly, integrating Eq. (25), we obtain


!
@uðx; y; tÞ gðx; d; tÞ  gðx; c; tÞ X2M X2M
ð1Þ CH
j
¼ þ bij ðtÞ Hi ðxÞ Hj ðyÞ  : ð38Þ
@y dc j¼1 i¼1
dc

Integration and further simplification gives


!
gðx; d; tÞ  gðx; c; tÞ X
2M X2M
ð2Þ CH
j
uðx; y; tÞ ¼ gðx; c; tÞ þ ðy  cÞ þ bij ðtÞ Hi ðxÞ Hj ðyÞ  ðy  cÞ : ð39Þ
dc j¼1 i¼1
dc

Equating Eqs. (37) and (39), and substituting the collocation points, we obtain system of 4M 2 equations. Another system
of 4M 2 equations is obtained by substituting the expressions of uðx; yÞ and its partial derivatives into the differential equation
of the concerned problem. These two sets of equations are solved simultaneously for the unknown Haar coefficients aij ’s and
bij ’s. The solution can be easily obtained by substituting these coefficients either in Eq. (37) or in Eq. (39).

8.2. Dirichlet conditions using Legendre wavelets

Like previous case, integrating Eq. (26), we obtain


 
@uðx; y; tÞ gðb; y; tÞ  gða; y; tÞ X N X N
ð1Þ CL
¼ þ aij ðtÞ Li ðxÞ  i Lj ðyÞ ð40Þ
@x ba j¼1 i¼1
ba

and
 
gðb; y; tÞ  gða; y; tÞ X N X N
ð2Þ CL
uðx; y; tÞ ¼ gða; y; tÞ þ ðx  aÞ þ aij ðtÞ Li ðxÞ  ðx  aÞ i Lj ðyÞ: ð41Þ
ba j¼1 i¼1
ba

Similarly, integrating Eq. (27), we get


!
@uðx; y; tÞ gðx; d; tÞ  gðx; c; tÞ X N X N
ð1Þ CLj
¼ þ bij ðtÞ Li ðxÞ Lj ðyÞ  ð42Þ
@y dc j¼1 i¼1
dc

and
!
gðx; d; tÞ  gðx; c; tÞ X N X N
ð2Þ CLj
uðx; y; tÞ ¼ gðx; c; tÞ þ ðy  cÞ þ bij ðtÞ Li ðxÞ Lj ðyÞ  ðy  cÞ : ð43Þ
dc j¼1 i¼1
dc
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9463

The numerical solution can be obtained in a similar fashion, as we did for Haar wavelets.

8.3. Neumann boundary conditions using Haar wavelets

The generic form of Neumann boundary conditions can be written as


@uða; y; tÞ @uðb; y; tÞ
¼ g x ða; y; tÞ; ¼ g x ðb; y; tÞ; x 2 ½a; bÞ; ð44Þ
@x @x

@uðx; c; tÞ @uðx; d; tÞ
¼ g y ðx; c; tÞ; ¼ g y ðx; d; tÞ; y 2 ½c; dÞ: ð45Þ
@y @y
Integration of the above equations yield
X
2M
g x ðb; y; tÞ  g x ða; y; tÞ ¼ ðb  aÞ a1j ðtÞHj ðyÞ: ð46Þ
j¼1

X
2M
g y ðx; d; tÞ  g y ðx; c; tÞ ¼ ðd  cÞ bi1 ðtÞHi ðxÞ: ð47Þ
i¼1

Substituting collocation points yl and xk in Eqs. (46) and (47) respectively, we obtain two systems of equations which can
be solved for the unknown Haar coefficients a1j ; j ¼ 1; 2; . . . ; 2M and bi1 ; i ¼ 1; 2; . . . ; 2M . Integrating Eqs. (26) and (27), we
get

@uðx; y; tÞ X
2M X2M
ð1Þ
¼ g x ða; y; tÞ þ aij ðtÞ Hi ðxÞ Hj ðyÞ; ð48Þ
@x j¼1 i¼1

@uðx; y; tÞ X
2M X2M
ð1Þ
¼ g y ðx; c; tÞ þ bij ðtÞ Hi ðxÞ Hj ðyÞ; ð49Þ
@y j¼1 i¼1

which yield the following equations after integrating Eq. (48) form a to x and Eq. (49) form c to y respectively,
X
2M X
2M
ð2Þ
uðx; y; tÞ ¼ uða; y; tÞ þ ðx  aÞg x ða; y; tÞ þ aij ðtÞ Hi ðxÞ Hj ðyÞ ð50Þ
j¼1 i¼1

and
X
2M X
2M
ð2Þ
uðx; y; tÞ ¼ uðx; c; tÞ þ ðy  cÞg y ðx; c; tÞ þ aij ðtÞ Hi ðxÞ Hj ðyÞ: ð51Þ
j¼1 i¼1

Substituting these expressions in given differential equation and equating two expressions of uðx; yÞ, we get a system of
8M 2 equations having 8M 2 unknowns aij ; i ¼ 2; 3; . . . ; 2M; j ¼ 1; 2; . . . ; 2M; bij ; i ¼ 1; 2; . . . ; 2M; j ¼ 2; 3; . . . ; 2M; uða; yl Þ; l ¼
1; 2; . . . 2M and uðxk ; cÞ; k ¼ 1; 2; . . . 2M. Solving this system, we obtain the unknowns which can then be substituted in Eq.
(50) or in Eq. (51), to find the approximate solution.

8.4. Neumann boundary conditions using Legendre wavelets

For this case integrating Eqs. (26) and (27), we obtain the following equations.
k1
ðb  aÞ X N X2
g x ðb; y; tÞ  g x ða; y; tÞ ¼ pffiffiffik1 aij ðtÞ Lj ðyÞ; ð52Þ
2 j¼1 i¼1

k1
ðd  cÞ X2 XN
g y ðx; d; tÞ  g y ðx; c; tÞ ¼ pffiffiffik1 bij ðtÞ Li ðxÞ; ð53Þ
2 j¼1 i¼1

Integration of the above equations give

@uðx; y; tÞ XN X N
ð1Þ
¼ g x ða; y; tÞ þ aij ðtÞ Li ðxÞ Lj ðyÞ; ð54Þ
@x j¼1 i¼1
9464 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

@uðx; y; tÞ XN X N
ð1Þ
¼ g y ðx; c; tÞ þ bij ðtÞ Li ðxÞ Lj ðyÞ ð55Þ
@y j¼1 i¼1

and
X
N X
N
ð2Þ
uðx; y; tÞ ¼ uða; y; tÞ þ ðx  aÞ g x ða; y; tÞ þ aij ðtÞ Li ðxÞ Lj ðyÞ; ð56Þ
j¼1 i¼1

X
N X
N
ð2Þ
uðx; y; tÞ ¼ uðx; c; tÞ þ ðy  cÞ g y ðx; c; tÞ þ bij ðtÞ Li ðxÞ Lj ðyÞ: ð57Þ
j¼1 i¼1

The numerical solution can be obtained in a similar fashion, as we did in the case of Haar wavelets.

8.5. Mixed boundary conditions using Haar wavelets

We consider general form of mixed boundary conditions,


@uða; y; tÞ @uðb; y; tÞ
¼ g x ða; y; tÞ; ¼ g x ðb; y; tÞ; x 2 ½a; bÞ; ð58Þ
@x @x

uðx; c; tÞ ¼ gðx; c; tÞ; uðx; d; tÞ ¼ gðx; d; tÞ; y 2 ½c; dÞ: ð59Þ


Integrating Eq. (24) and using the conditions in (58), we get Eq. (46). Substituting collocation points yl ; l ¼ 1; 2; . . . ; 2M in
Eq. (46), we obtain a system which gives the values of a1j ; i ¼ 1; 2; . . . ; 2M. Equating Eqs. (50) and (39), we get another sys-
tem which can be solved simultaneously with the system obtained by substituting uðx; yÞ and its partial derivatives in the
differential equation of the concerned problem. The values thus obtained can be used either in Eq. (50) or Eq. (39) for the
numerical solution.

8.6. Mixed boundary conditions using Legendre wavelets

Integrating Eq. (26) and using boundary conditions (58), we get Eq. (52). Substituting collocation points
yl ; l ¼ 1; 2; . . . ; 2M to obtain first system of equations which can be solved for a1j ; i ¼ 1; 2; . . . ; 2M. Equating Eqs. (56) and
(43), we get another system which can be solved simultaneously with the system obtained by substituting uðx; yÞ and its
partial derivatives in the differential equation of the concerned problem. The values thus obtained can be used either in
Eq. (56) or Eq. (43) for the numerical solution.

8.7. Periodic boundary conditions using Haar wavelets

The general form of periodic boundary conditions is given by,


@uða; y; tÞ @uðb; y; tÞ
uða; y; tÞ ¼ uðb; y; tÞ; ¼ ; y 2 ½a; bÞ; ð60Þ
@x @x

@uðx; c; tÞ @uðx; d; tÞ
uðx; c; tÞ ¼ uðx; d; tÞ; ¼ ; x 2 ½c; dÞ: ð61Þ
@y @y
In order to obtain numerical solution by Haar wavelets, we integrate Eqs. (24) and (25) and use boundary conditions
above to obtain the following expressions:
X
2M
0 ¼ ðb  aÞ a1j ðtÞHj ðyÞ; ð62Þ
j¼1

X
2M
0 ¼ ðd  cÞ bi1 ðtÞHi ðxÞ: ð63Þ
i¼1

Substituting collocation points yl and xk ; l; k ¼ 1; 2; . . . ; 2M in Eqs. (62) and (62) respectively, we obtain two systems
which give the values of unknown Haar coefficients a1j ; j ¼ 1; 2; . . . ; 2M and bi1 ; i ¼ 1; 2; . . . ; 2M. Now integrating Eqs.
(24) and (25) and using boundary conditions, we have
 
@uðx; y; tÞ X2M X2M
ð1Þ CH
¼ aij ðtÞ Hi ðxÞ  i Hj ðyÞ; ð64Þ
@x j¼1 i¼1
ba
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9465

X
2M X2M  
ð2Þ CH
uðx; y; tÞ ¼ uða; y; tÞ þ aij ðtÞ Hi ðxÞ  i ðx  aÞ Hj ðyÞ ð65Þ
j¼1 i¼1
ba

and
!
@uðx; y; tÞ X2M X2M
ð1Þ CH
j
¼ bij ðtÞHi ðxÞ Hj ðyÞ  ; ð66Þ
@y j¼1 i¼1
dc

!
X
2M X
2M
ð2Þ CH
j
uðx; y; tÞ ¼ uðx; c; tÞ þ bij ðtÞHi ðxÞ Hj ðyÞ  ðy  cÞ : ð67Þ
j¼1 i¼1
dc

Proceeding as in the case of Neumann conditions, we obtain the unknown coefficients by solving system of simultaneous
equations. These coefficients then yield the approximate solution.

8.8. Periodic boundary conditions using Legendre wavelets

Like the previous case, we integrate Eqs. (26) and (27) and use boundary conditions given in Eqs. (60) and (61), we have
k1
ðb  aÞ X N X2
0 ¼ pffiffiffik1 aij ðtÞ Lj ðyÞ; ð68Þ
2 j¼1 i¼1

k1
ðd  cÞ X2 XN
0 ¼ pffiffiffik1 bij ðtÞ Li ðxÞ: ð69Þ
2 j¼1 i¼1

Substituting the collocation points yl and xk , we obtain two systems which can be solved for the unknowns aij and bij . Dif-
ferent expressions in terms of Legendre wavelets and their integrals are given by
 
@uðx; y; tÞ X N X N
ð1Þ CL
¼ aij ðtÞ Li ðxÞ  i Lj ðyÞ; ð70Þ
@x j¼1 i¼1
ba

XN X N  
ð2Þ CL
uðx; y; tÞ ¼ uða; y; tÞ þ aij ðtÞ Li ðxÞ  ðx  aÞ i Lj ðyÞ ð71Þ
j¼1 i¼1
ba

and
!
@uðx; y; tÞ X N X N
ð1Þ CLj
¼ bij ðtÞ Li ðxÞ Lj ðyÞ  ; ð72Þ
@y j¼1 i¼1
dc

!
X
N X
N
ð2Þ CLj
uðx; y; tÞ ¼ uðx; c; tÞ þ bij ðtÞ Li ðxÞ Lj ðyÞ  ðy  cÞ : ð73Þ
j¼1 i¼1
dc

Repeating the same procedure as we did for Haar wavelets, we obtain the unknown coefficients. These coefficients then
yield the approximate solution.

9. A hybrid FDM and wavelets based numerical scheme

Coupling of finite forward difference approximation for time derivative along with Haar and Legendre based approxima-
tions for solution variable as well as its derivatives produce a variety of discretization schemes for different cases which are
given as follows.

9.1. One-dimensional deterministic diffusion equation

9.1.1. Constant optic parameters


Hybrid scheme for one-dimensional deterministic diffusion equation (8) in the case of Haar wavelets for constant optic
parameters is given by,
!
 X
2M
1 þ C Dtla0 uðx; tÞ ¼ uðx; t0 Þ þ Dt C D0 ai Hi ðxÞ þ f ðx; tÞ : ð74Þ
i¼1
9466 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

Substituting the value of uðx; tÞ given in Eq. (30) in above equation and discretizing at the collocation points xk , for
k ¼ 1; 2; . . . ; 2M, we obtain a system of 2M equations. Combining these 2M equations with Eqs. (9), (10), (28) and (29),
and solving them simultanuously, we get the values of the unknowns uða; tÞ; uðb; tÞ; @uða;tÞ @x
; @uðb;tÞ
@x
and ai for
i ¼ 1; 2; . . . ; 2M. Replacing these values in Eq. (30), we get the desired numerical solution at time t.
Similarly in the case of Legendre wavelets we obtain,
!
 X
N
1 þ C Dtla0 uðx; tÞ ¼ uðx; t 0 Þ þ Dt C Do ai Li ðxÞ þ f ðx; tÞ : ð75Þ
i¼1

In this case we substitute value of uðx; tÞ given in Eq. (33) in above equation. The solution is obtained in a similar way as in
case of Haar wavelets described above.

9.1.2. Variable optic parameters


Hybrid scheme for one-dimensional deterministic diffusion equation (8) in the case of Haar wavelets for variable optic
parameters is given by,
!
 @uðx; tÞ X
2M
1 þ DtCðxÞla ðxÞ uðx; tÞ  DtCðxÞD0 ðxÞ ¼ uðx; t 0 Þ þ Dt CðxÞ DðxÞ ai Hi ðxÞ þ f ðx; tÞ : ð76Þ
@x i¼1

Similarly in the case of Legendre wavelets we obtain,


!
 @uðx; tÞ X
N
1 þ DtCðxÞla ðxÞ uðx; tÞ  DtCðxÞD0 ðxÞ ¼ uðx; t 0 Þ þ Dt CðxÞ DðxÞ ai Li ðxÞ þ f ðx; tÞ : ð77Þ
@x i¼1

The solution procedure is similar as in the case constant optic parameters.

9.2. Linear diffusion reaction equation

Hybrid scheme for Eq. (11) in the case of Haar wavelets is given by,
!
X
2M X
2M X
2M X
2M
ð1  0:1Dt Þuðx; y; tÞ ¼ uðx; y; t 0 Þ þ 0:2Dt aij Hi ðxÞHj ðyÞ þ bij Hi ðxÞHj ðyÞ : ð78Þ
j¼1 i¼1 j¼1 i¼1

Similarly in the case of Legendre wavelets we obtain,


! !
X
N X
N X
N X
N
ð1  0:1Dt Þuðx; y; tÞ ¼ uðx; y; t 0 Þ þ 0:2Dt aij Li ðxÞ Lj ðyÞ þ 0:2Dt bij Li ðxÞ Lj ðyÞ : ð79Þ
j¼1 i¼1 j¼1 i¼1

9.3. Dirichlet jump problem

Hybrid scheme for Eq. (12) in the case of Haar wavelets is given by,
" #
X2M X
2M X
2M X
2M
uðx; y; tÞ ¼ uðx; y; t 0 Þ þ Dt aij Hi ðxÞHj ðyÞ þ bij Hi ðxÞHj ðyÞ : ð80Þ
j¼1 i¼1 j¼1 i¼1

Similarly in the case of Legendre wavelets we obtain,


" #
XN X
N X
N X
N
uðx; y; tÞ ¼ uðx; y; t 0 Þ þ Dt aij Li ðxÞ Lj ðyÞ þ bij Li ðxÞ Lj ðyÞ : ð81Þ
j¼1 i¼1 j¼1 i¼1

9.4. Burgers’ equation

After time discretization, Eq. (13) takes the following form


!  
Dt @ 2 u @ 2 u @u @u
uðx; y; tÞ ¼ uðx; y; t 0 Þ þ þ  D tu þ : ð82Þ
Re @x2 @y2 @x @y

The non-linear terms are linearized by using the following formulae:


@u @uðx; y; tÞ @uðx; y; t 0 Þ @uðx; y; t 0 Þ
u ¼ uðx; y; t0 Þ þ uðx; y; tÞ  uðx; y; t0 Þ :
@x @x @x @x
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9467

@u @uðx; y; tÞ @uðx; y; t 0 Þ @uðx; y; t 0 Þ


u ¼ uðx; y; t 0 Þ þ uðx; y; tÞ  uðx; y; t0 Þ :
@y @y @y @y

60 20
Exact sol. Exact sol.
Numerical sol. 18 Numerical sol.
50
16

14
40
12

30 10

8
20
6

4
10
2

0 0
−5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5

Fig. 1. Comparison of exact and numerical solutions, Test Problem 1, HWCM at t ¼ 1, Dt ¼ 102 ; N ¼ 34; x 2 ½5; 5.

−9 −10 −11
x 10 x 10 x 10
4.5 4.5 4.5

4 4 4

3.5 3.5 3.5

3 3 3

2.5 2.5 2.5

2 2 2

1.5 1.5 1.5

1 1 1

0.5 0.5 0.5

0 0 0
−5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4

Fig. 2. Absolute errors for variable optic parameters, Test Problem 1, HWCM at t ¼ 1, Dt ¼ ð102 ; 103 ; 104 Þ; N ¼ 34, from left to right respectively.

−3 −4 −4
x 10 x 10 x 10
2.5 6 1.4

5 1.2
2
1
4
1.5
0.8
3
0.6
1
2
0.4
0.5
1
0.2

0 0 0
−5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5

Fig. 3. Absolute errors for constant optic parameters, Test Problem 1, HWCM at t ¼ 1; Dt ¼ 102 ; N ¼ 130; 258; 514, (left to right respectively).
9468 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

0Þ 0Þ
The terms @uðx;y;t
@x
; @uðx;y;t
@y
are obtained at initial time t 0 by differentiating Eq. (83), and the terms @uðx;y;tÞ
@x
; @uðx;y;tÞ
@y
are given
by Eqs. (36) and (38) for Haar wavelets and by Eqs. (40) and (42) for Legendre wavelets. These expressions can be used in Eq.
(82) to obtain the hybrid schemes for Haar as well as Legendre wavelets.

−3
10
HWCM
−4 LWCM
10

−5
10

−6
10

−7
10

−8
10

−9
10

−10
10
0 5 10 15 20 25 30 35

Fig. 4. Comparison of HWCM and LWCM for constant optic parameters, Test Problem 1, t ¼ 1; N ¼ 4; 8; 16; 32; x 2 ½0; 1.

50 50

45 45

40 40

35 35

30 30

25 25

20 20

15 15
10 10
5 5
0 5 0 5
−1 −1
0 0 0 0
1 1
2 −5 2 −5

Fig. 5. Comparison of exact solution and HWCM for variable optic parameters, Test Problem 1, N ¼ 64; t 2 ½1; 2.

150 150

100 100

50 50

0 0
0 5 5
0
0.2 0.2
0.4 0 0.4 0
0.6 0.6
0.8 0.8
1 −5 1 −5

Fig. 6. Comparison of exact solution and HWCM for constant optic parameters, Test Problem 1, N ¼ 64; t 2 ½0; 1.
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9469

Exact

−1

−2
10
8
5 6
4
2
0 0

2 2 2

1 1 1

0 0 0

−1
−1 −1

−2
10 −2 −2
10 10
8
8 8
5 6
5 6 5 6
4
2 4 4
0 0 2 2
0 0 0 0

2 2 2

1 1 1

0 0 0

−1 −1 −1

−2 −2 −2
10 10 10
8 8 8
5 6 5 6 5 6
4 4 4
2 2 2
0 0 0 0 0 0

Fig. 7. Comparison of exact and HWCM (top), LWCM (bottom), Test Problem 2, t ¼ 0:1; Dt ¼ 103 ; N ¼ 8  8.

Table 1
Comparison of HWCM for Test Problem 2, with GRBFCNCM and GRBFICM at t ¼ 0:1; Dt ¼ 104 in X ¼ ½0; 2p2 .

N HWCM LWCM N GRBFCNCM [40] GRBFICM [40]


L1 L1 L1 L1
44 2:2616  103 2:2730  103 11  11 8:34  103 8:34  103
88 9:0577  104
1:0886  10 4 21  21 6:57  10 2
6:58  102
16  16 2:4716  104 2:7439  105 41  41 8:17  102 8:18  102
32  32 6:3066  105
9:7540  10 8 55  55 7:92  10 2
7:95  102
9470 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

10. Numerical results and discussion

In this section, we examine the accuracy of the new methods on several benchmarks and compare the numerical results
with the existing methods in literature. These include spline based methods for one-dimensional diffusion problem [36],
ELMFS [47], Global and Local Meshless Methods [45,40], Explicit Local Meshless Methods [37]. These problems are consid-
ered in order to demonstrate an improved and more stable approximating properties of Haar and Legendre wavelets. The
accuracy of the schemes is measured in terms of absolute and relative errors which are defined as:

Labs ¼ ju  u ^ j;
Labs
Lrel ¼ ;
u
where u and u ^ are the actual and numerical solutions respectively. We use the following nomenclature for the proposed
methods along with different sets of boundary conditions: Haar wavelets collocation method with Dirichlet boundary con-
ditions as (DHWCM), Legendre wavelets collocation method with Dirichlet boundary conditions as (DLWCM), Haar wavelets
collocation method with mixed boundary conditions as (MHWCM), Legendre wavelets collocation method with mixed
boundary conditions as (MLWCM), Haar wavelets collocation method with Neumann boundary conditions as (NHWCM),

Table 2
Comparison of HWCM for Test Problem 2, with LRBFECM [40] and LRBFICM [40] at (t ¼ 103 ; 102 ; 101 ), Dt ¼ 103 ; X ¼ ½0; 2p2 .

N HWCM LWCM N LRBFECM LRBFICM


E1 E1 E1 E1

(t ¼ 103 )
22 3:7885  105 3:7885  105 11  11 2:43  102 3:10  102
44 2:2831  10 5
2:2831  10 5 21  21 1:60  10 3
2:13  103
88 9:1069  106 5:4985  107 41  41 2:68  104 3:13  104
16  16 2:4872  10 6
9:4894  10 9 101  101 2:70  10 4
1:78  103
2
(t ¼ 10 )
22 3:7857  104 3:7857  104 11  11 2:43  102 3:06  102
44 2:2810  104 2:2821  104 21  21 2:68  103 2:19  103
88 9:1022  105
5:4708  10 6 41  41 2:68  10 3
3:42  104
16  16 2:4851  105 9:4836  108 101  101 2:70  103 4:73  103

(t ¼ 101 )
22 3:7582  103 3:7857  104 11  11 2:64  102 3:43  102
44 2:2608  103
2:2821  10 4 21  21 2:67  10 2
2:68  103
88 9:0488  104 5:2077  105 41  41 2:67  102 5:60  104
16  16 2:4626  10 4
9:4289  10 7 101  101 2:69  10 2
2:69  102

Table 3
Comparison of DHWCM, MHWCM, and NHWCM, Test Problem 2, at t ¼ 1; ðDt ¼ 102 ; 103 ; 104 Þ; X ¼ ½0; 12 .

2M ¼ N Dt ¼ 104 Dt ¼ 103 Dt ¼ 102


E1 E1 E1
DHWCM DLWCM DHWCM DLWCM DHWCM DLWCM
22 3:5070  104 3:5070  104 3:4953  104 3:4953  104 3:3782  104 3:3782  104
44 2:0939  104 1:6472  104 2:0736  104 1:6639  104 1:8707  104 1:8304  104
88 5:8647  105 5:8647  105 5:6559  105 5:6559  105 3:5717  105 3:5717  105
16  16 1:4736  10 5
2:3069  10 7
1:2654  10 5
2:3065  10 6
8:1277  10 6
2:3032  105
MHWCM MLWCM MHWCM MLWCM MHWCM MLWCM
22 1:1614  103 6:5158  104 1:6165  103 6:5393  104 1:6373  103 6:7140  104
44 7:1113  104 3:4670  104 7:1362  104 3:4948  104 7:3842  104 3:7741  104
88 2:2727  104 9:8304  105 2:2997  104 9:4802  105 2:5686  104 6:0158  105
16  16 6:3887  105 3:8377  107 6:6651  105 3:8362  106 9:4385  105 3:8202  105
NHWCM NLWCM NHWCM NLWCM NHWCM NLWCM
22 8:3147  103 3:1069  103 8:3183  103 3:1135  103 8:3543  103 3:1785  103
44 2:9477  103 7:6924  104 2:9520  103 7:7602  104 2:9952  103 8:4375  104
88 2:2727  10 4
8:5919  10 4
8:6385  10 4
8:6385  10 4
9:1050  10 4
9:1050  104
16  16 6:3887  105 7:6222  107 2:3608  104 7:6221  106 2:8369  104 7:6210  105
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9471

Legendre wavelets collocation method with Neumann boundary conditions as (NLWCM). In nutshell, HWCM has subclasses
DHWCM, MHWCM and NHWCM and LWCM has subclasses DLWCM, MLWCM and NLWCM for two-dimensional PPDEs. We
also add L; R; T and B to indicate the respective conditions defined at Left, Right, Top and Bottom boundaries.

10.1. Test Problem 1

For constant optic parameters the exact solution of one-dimensional diffusion equation (8) is given by

uðx; tÞ ¼ exp½ðx þ t 2 Þ;


with the source term
 
2V o t
f ðx; tÞ ¼  Do þ lao exp½ðx þ t 2 Þ:
Co
The boundary functions g a ; g b are defined as
 
ba
g a ðtÞ ¼ 1þ Do exp½ða þ t 2 Þ;
2

−5
x 10

0.5

0
1

0.8

0.6
1
0.4 0.8
0.6
0.2 0.4
0.2
0 0

Fig. 8. Absolute error graph, Test Problem 2, MHWCM at t ¼ 0:1; Dt ¼ 103 ; N ¼ 34  34.

−2
10
DHWCM
DLWCM
MHWCM
−3
10 MLWCM
NHWCM
NLWCM

−4
10

−5
10

−6
10

−7
10
0 50 100 150 200 250 300

Fig. 9. Comparison of HWCM and LWCM, Test Problem 2, for different types of boundary conditions and different N ¼ N x  N y , at
t ¼ 0:1; Dt ¼ 102 ; X ¼ ½0; 12 .
9472 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

−2 −4
10 10

−4 −6
10 10

−6 −8
10 10

MHWCML
MLWCML
−8 MHWCMR −10
10 10 MLWCMR
NHWCML
NLWCML
NHWCMR
NLWCMR
NHWCMT
NLWCMT
−10 NHWCMB −12
10 10 NLWCMB

−12 −14
10 10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fig. 10. Absolute error profile for HWCM and LWCM, Test Problem 2, for Neumann boundaries using mixed and Neumann boundary conditions at
t ¼ 0:1; Dt ¼ 103 ; N ¼ 10  10.

−9
−4 x 10
x 10
5
1

4
0.8

3
0.6

2
0.4

0.2 1

0 0
1 1
1 1
0.8 0.8
0.5 0.5 0.6
0.6
0.4 0.4
0.2 0.2
0 0 0 0

Fig. 11. Absolute error for HWCM, Test Problem 3, at (t ¼ 0:5; 1, left to right), Dt ¼ 103 ; N ¼ 18  18.

−3
10
HWCM
−4 LWCM
10

−5
10

−6
10

−7
10

−8
10

−9
10

−10
10

−11
10
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3

Fig. 12. Comparison of HWCM and LWCM, Test Problem 3, for different time steps. t ¼ 0:4; 0:6; 0:8; 1; 1:2; Dt ¼ 103 ; N ¼ 16  16.
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9473

−4 −4
x 10 x 10

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0.5 0.5
0.5 0.5
0 0
0 0
−0.5−0.5 −0.5−0.5

Fig. 13. Absolute error for HWCM and LWCM, Test Problem 3, at time t ¼ 0:5; Dt ¼ 102 and N ¼ 18  18.

−5 −5
x 10 x 10
9 9

8 8

7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
−0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5

Fig. 14. Absolute error for HWCM and LWCM, Test Problem 3, x ¼ 0:0 in y-direction at t ¼ 0:5; Dt ¼ 103 ; N ¼ 34  34.

 
b
g b ðtÞ ¼ 1  b Do exp½ðb þ t 2 Þ:
2
The exact solution in case of variable optic parameters is given by

uðx; tÞ ¼ ðx þ 2Þ2 expðt2 Þ;


with the source term
!
2VðxÞðx þ 2Þ2 t
f ðx; tÞ ¼  4DðxÞ þ la ðxÞðx þ 2Þ2 expðt 2 Þ:
Co

The boundary functions g a ; g b are defined as


 
ba
g a ðtÞ ¼ 1 Do ða þ 2Þ2 expðt 2 Þ;
2

 
b
g b ðtÞ ¼ 1 þ b Do ðb þ 2Þ2 expðt2 Þ:
2
In Fig. 1, the plot of the exact and the numerical solutions using HWCM for constant as well as variable optic parameters is
given at time t ¼ 1; Dt ¼ 102 ; N ¼ 34 and x 2 ½5; 5. The agreement of the numerical solution to the exact solution is obvi-
ous from the figure. The absolute errors are graphically shown in Fig. 2. For the case of variable optic parameters, we choose
time t ¼ 1; N ¼ 34 and different time steps Dt ¼ 102 ; Dt ¼ 103 and Dt ¼ 104 , respectively. For constant optic parameters
case, as shown in Fig. 3, we consider t ¼ 1; Dt ¼ 102 , and different nodal points N ¼ 130; 258, and 514, respectively. The
two extra points are taken for solution near the boundaries. A comparison of the HWCM and LWCM in terms of L1 is shown
9474 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

a b
−4 −4
x 10 x 10

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
1 1
1 1
0.8 0.8
0.5 0.6 0.5 0.6
0.4 0.4
0.2
0 0 0 0

c d
−5 −5
x 10 x 10

8 2

6 1.5

4 1

2 0.5

0 0
1 1
1 1
0.8 0.8
0.5 0.6 0.5 0.6
0.4 0.4
0.2 0.2
0 0 0 0

Fig. 15. Error profiles for HWCM, Test Problem 4, at ((a) t ¼ 0:10; (b) t ¼ 1:00; (c) t ¼ 3:00; (d) t ¼ 9:00), Dt ¼ 0:01; N ¼ 8  8; Re ¼ 1.

in Fig. 4. Better performance of the LWCM over HWCM is obvious from this figure. Exact profiles of the solutions for variable
and constant optic parameters in the case of HWCM are shown in Figs. 5 and 6. From this comparison, it is seen that numer-
ical solution obtained through HWCM resembles exact profiles for N ¼ 64.

10.2. Test Problem 2

The differential operator and other parameters given in Eq. (11) are defined as follows: L ¼ r2 ; g ¼ 0:2 , j ¼ 0:1 and
g ¼ 0. Accordingly the exact solution of Eq. (11) is given by

uðx; y; tÞ ¼ expð0:1t Þ½cosðxÞ þ sinðyÞ:


The exact and the numerical solutions for different types of boundary conditions are graphically depicted in Fig. 7 using
HWCM and LWCM. The resemblance between exact and numerical solutions is clear from these figures. The numerical re-
sults in terms of L1 error norm of the proposed methods for different number of collocation points are compared with global
meshless collocation methods (GRBFCNCM and GRBFICM) [40] in Table 1 and local meshless collocation methods (LRBFECM
and LRBFICM) [40] in Table 2. Improved performance of the proposed methods is clear from these tables. As shown in Table 1,
accuracy of global meshless methods (GRBFCNCM and GRBFICM) [40] is of order Oð102 Þ for 55  55 nodal arrangement
whereas accuracy of HWCM is of order Oð104 Þ and LWCM is of order Oð105 Þ corresponding to Dt ¼ 104 , for 16  16 col-
location points. Similarly in Table 2, accuracy of local meshless method LRBFECM [40] is of order Oð104 Þ; Oð103 Þ; Oð102 Þ
corresponding to Dt ¼ 103 ; Dt ¼ 102 ; Dt ¼ 101 for 101  101 collocation points. Accuracy of local meshless method LRB-
FICM [40] is of order Oð103 Þ; Oð103 Þ; Oð102 Þ corresponding to same time and spatial steps. Accuracy of HWCM is of order
Oð105 Þ; Oð104 Þ; Oð103 Þ and that of LWCM is Oð105 Þ; Oð104 Þ for 2  2 collocation points. The accuracy of HWCM in-
creases further to Oð106 Þ and that of LWCM to Oð109 Þ for 16  16 collocation points and Dt ¼ 103 . Better performance
of the HWCM and LWCM is maintained for Dt ¼ 102 and Dt ¼ 101 as well. Neumann boundary conditions which are more
problematic for the classical numerical methods such as finite difference, finite volume and meshless methods are also dealt
with accurately. The error introduced at the boundary due to the Neumann boundary conditions diffuses to the interior
nodes which badly effect the accuracy of the classical methods. In the proposed methods the Neumann conditions are inte-
grated in the solution procedure and is no more a challenge for the present methods. The numerical results in terms of L1
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9475

a b
−6 −6
x 10 x 10

5 1.2

4 1

0.8
3
0.6
2
0.4
1 0.2

0 0
1 1
1 1
0.8 0.8
0.5 0.6 0.5 0.6
0.4 0.4
0.2 0.2
0 0 0 0

c d
−5 −6
x 10 x 10

1 2

0.8 1.5

0.6
1
0.4
0.5
0.2

0 0
1 1
1 1
0.8 0.8
0.5 0.6 0.5 0.6
0.4 0.4
0.2 0.2
0 0 0 0

Fig. 16. Error profiles for LWCM, Test Problem 4, at ((a) t ¼ 0:10; (b) t ¼ 1:00; (c) t ¼ 3:00; (d) t ¼ 9:00), Dt ¼ 0:01; N ¼ 8  8; Re ¼ 1.

error norm for different types of boundary conditions and at different number of collocation points are calculated at t ¼ 1
and Dt ¼ 102 ; Dt ¼ 103 ; Dt ¼ 104 are given in Table 3. From this table, it is clear that HWCM and LWCM are showing
stable behavior for Dirichlet, Neumann and mixed type of boundary conditions. The absolute errors at t ¼ 0:1; Dt ¼ 103
and 34  34 nodal points are shown in Fig. 8 using MHWCM. L1 error norm for HWCM and LWCM for different types of
boundary conditions is shown in Fig. 9. Marginal improvement in terms of accuracy is seen in the case of LWCM. The abso-
lute error profiles for Neumann and mixed type of boundary conditions are given in Fig. 10. For Haar wavelets the results of
NHWCMT, NHWCMB and for Legendre wavelets the results of MLWCML, MLWCMR and NLWCM, NLWCMB are almost sim-
ilar and their graphs overlap each other.

10.3. Test Problem 3

If the differential operator L ¼ r2 and g ¼ 1, then the exact solution of the PPDE in Eq. (12) is given by

16
uðx; y; tÞ ¼ uðx; tÞuðy; tÞ;
p2
with n ¼ x; y
h i
i 2
1 ð1Þ exp ð2i þ 1Þ p2 t cos½ð2i þ 1Þpn
X
uðn; tÞ ¼ :
i¼0
ð2i þ 1Þ

This problem is challenging in the sense that there is jump discontinuity between initial and boundary conditions. Sudden
transition can cause unstable numerical solution in the beginning of the simulation time. Labs error norm of HWCM for dif-
ferent values of t is shown in Fig. 11. It is clear from this figure that the error is maximum at the boundaries due to Dirichlet
jump phenomenon. The error drastically drops at the interior nodes as simulation time progresses from smaller to higher
value. L1 error norm of both the proposed methods HWCM and LWCM for different values of t is shown in Fig. 12. This graph
verifies fast convergence as the solution marches in time. Absolute error profile using HWCM and LWCM for
t ¼ 0:5; Dt ¼ 102 and N ¼ 18  18 is given in Fig. 13. The graph for x ¼ 0:0 in the y-direction is depicted in Fig. 14. From
9476 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

a b

−3
x 10 x 10
−3

1.5 1.5

1 1

0.5 0.5

0 0 1
1 0 0.8
0
0.8 0.2
0.2 0.6
0.6 0.4
0.4 0.4
0.4 0.6
0.6 0.8 0.2
0.8 0.2
1 0
1 0

c d
−3
x 10 −3
x 10
1.5
1.5

1
1

0.5 0.5

0 0
0 1 0 1
0.2 0.8 0.2 0.8
0.4 0.6 0.4 0.6
0.6 0.4 0.6 0.4
0.8 0.2 0.8 0.2
1 0 1 0

Fig. 17. Absolute error profiles, HWCM, Test Problem 4 at ((a) t ¼ 0:5; (b) t ¼ 1; (c) t ¼ 0:75; (d) t ¼ 1:25), Dt ¼ 0:001; N ¼ 1024; Re ¼ 20.

the numerical results shown in the figures above it is clear that the Dirichlet jump problem is solved in a stable manner by
the proposed methods HWCM and LWCM. It is obvious from the figure that both HWCM and LWCM show almost similar and
stable performance throughout the domain. Unlike the previous case, accuracy of the numerical method LWCM does not im-
prove by using high order Legendre wavelets. Degradation in the accuracy of LWCM is due to Dirichlet jump, which is a
source of nuisance for the numerical methods specially at the initial phase of simulation time.

10.4. Test Problem 4

If the differential operator L ¼ r2 and j ¼ Re


1
then the exact solution of the PPDE in Eq. (13) discussed in [41] is given by,

1
uðx; y; tÞ ¼ : ð83Þ
1 þ expðReðx þ y  tÞ=2Þ
Numerical results corresponding test Problem 4 are shown in Figs. 15–25. The absolute errors for HWCM are depicted in
Fig. 15 and for LWCM in Fig. 16 at time levels 0:1; 1; 3 and 9 respectively for Reynolds number Re ¼ 1 and Dt ¼ 0:01. The
diffusion process varies smoothly, because the convective term is not large in comparison with the diffusion term. The abso-
lute errors are quite small, which shows that the proposed methods can handle the Burgers’ equation at low Reynolds num-
ber. The absolute error profiles for Re ¼ 20; N ¼ 1024; Dt ¼ 103 and different times are shown in Fig. 17. It is clear that
error moves due to movement of shock as the time progress. Velocity profiles for HWCM are given in Fig. 18 where
Re ¼ 20; Dt ¼ 0:001 and different time levels 0:5; 0:75; 1:0 and 1:25 are taken. Numerical solution of the problem with
the global meshless collocation method [45] is shown in Figs. 19,20 for different values of Re ¼ 120; 150; 200; 300. It is clear
from these figures that an inaccurate numerical solution is obtained through meshless method for Re ¼ 150; 200; 300. For
Re ¼ 300, the solution is highly distorted. Instabilities in the form of wiggles and spikes can be seen clearly in the solution
profile. Similarly solutions produced by LWCM are also unstable for Re ¼ 120; 150; 200; 300 which are shown in Figs. 21 and
22. Legendre wavelets basis behaves like meshless methods (which uses MQ RBFs) and are unable to capture sharp transi-
tions accurately. The solution obtained through LWCM is also highly unstable for Re ¼ 200; 300. The solutions obtained
through LWCM for Re ¼ 120; 150; 200; 300 are shown in Figs. 23 and 24. From these figures it clear that HWCM produces
spikes free, stable and accurate solution for Re ¼ 120; 150; 200. This is because of the similarity between Haar basis and
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9477

a b
1
1

0.8 0.8

0.6 0.6

0.4
0.4

0.2
0.2
0
0
0
0
0.5
0.5
0.8 1
0.6 0.8 1 0.6
1 0.2 0.4 1 0.2 0.4
0 0

c d
1
1

0.8
0.8

0.6
0.6

0.4
0.4

0.2
0.2

0
0
0
0

0.5 0.5
1
0.6 0.8 1
0.4 0.6 0.8
1 0 0.2 1 0.2 0.4
0

Fig. 18. Velocity profiles for HWCM, Test Problem 4, at ((a) t ¼ 0:50; (b) t ¼ 0:75; (c) t ¼ 1:00; (d) t ¼ 1:25), Dt ¼ 0:001; N ¼ 32  32; Re ¼ 20.

1.2
1
0.8
0.6
0.4
0.2
0
−0.2
0 0.2 1
0.6 0.8 1 0.4 0.6 0.4 0.6 0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1 0 0.2

Fig. 19. Numerical solution Meshless methods, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 120 (left), Re ¼ 150 (right), N ¼ 26  26.

the shape of the sharp jump present in the solution. For high values of Reynolds number the performance shown by HWCM
is much better than the meshless method and LWCM. For Reynolds Re ¼ 300 the spike seen in the solution is due to time
truncation error which can be controlled by using higher order time integration solver. Exact profiles of the solution for dif-
ferent values of Re are shown in Fig. 25. The solution obtained from HWCM resemble closely with the exact profiles for
Re ¼ 120; 150. No post processing technique is needed in the case of HWCM. In the case of LWCM and meshless methods
stable solution can only be obtained if some filtering technique (like rank order filter (ROF), digital total variation (DTV),
or gegenbauer polynomial) etc. are used for removal of unwanted spikes appearing in the solution.
9478 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

14000
1.4 12000

1.2 10000
8000
1
6000
0.8
4000
0.6 2000
0.4 0
−2000
0.2
−4000
0
−6000
0 1
−0.2 0.2 0.8 0
0.4 0.6 0.4 0.2
1 0.8 0.6 0.4 0.6
0.6 0.8 0.2 0.8
0.4 0.2 1 0 1
0

Fig. 20. Numerical solution Meshless methods, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 200 (left), Re ¼ 300 (right), N ¼ 26  26.

1.4 1.4

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.2
0 1 0 1
0.6 0.8 0.6 0.8
0.5 0.2 0.4 0.5 0.4
1 0 1 0 0.2

Fig. 21. Numerical solution LWCM, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 120 (left), Re ¼ 150 (right), N ¼ 16  16.

1.4 2

1.2
1.5
1
0.8
1
0.6
0.4 0.5
0.2
0
0
−0.2
−0.5
0
0.2 1 0
0.4 0.8 1
0.6 0.8
0.6 0.4 0.5 0.6
0.8 0.2 0.4
1 0 0.2
1 0

Fig. 22. Numerical solution LWCM, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 200 (left), Re ¼ 300 (right), N ¼ 32  32.
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9479

1.2 1.2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2
−0.2
0 0
1 0.5 0.8 1
0.5 0.8 0.6
0.4 0.6 0.4
1 0.2 1 0 0.2
0

Fig. 23. Numerical solution HWCM, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 120 (left), Re ¼ 150 (right), N ¼ 32  32.

1.4
1.2
1.2
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0

−0.2 0
0
−0.2
1 0
0.5 0.8 1
0.6 0.5 0.8
0.4 0.4 0.6
0.2 1 0 0.2
1 0

Fig. 24. Numerical solution HWCM, Test Problem 4, t ¼ 1; Dt ¼ 103 ; Re ¼ 200 (left), Re ¼ 300 (right), N ¼ 32  32.

1 1
0.9
0.8 0.8

0.7
0.6 0.6

0.5
0.4 0.4

0.3
0.2 0.2

0.1
0
0 0
0
1 0.5 1
0.5 0.8 0.6 0.8
0.6 0.4
0.4 1 0.2
1 0.2 0
0

Fig. 25. Exact solution, Test Problem 4, at t ¼ 1; Dt ¼ 103 ; Re ¼ 120 (left), Re ¼ 150 (right), N ¼ 32  32.
9480 Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481

11. Conclusions

In this paper we proposed two new numerical methods for PPDEs. The new methods are based on Legendre and Haar
(box) functions. Suitability of Haar functions to capture sharp transitions is tested on the convection dominated flows.
HWCM which uses constant box functions provide a stable solution free of wild oscillations whereas meshless method
and LWCM which use higher order basis (RBF and Legendre wavelets) fails to have a stable solution for convection domi-
nated models. The performance of the Legendre wavelets is superior when the initial data is smooth. The methods work well
for Neumann, mixed, periodic and Dirichlet type of boundary conditions. Stable solution with improved accuracy is obtained
as a result of the application of the new methods.

Acknowledgments

This paper was funded by the Deanship of Scientific Research (DSR), King Abdulaziz University, Jeddah, Saudi Arabia. The
authors, therefore acknowledge with thanks DSR technical and financial support. The second author also acknowledges par-
tial funding from Higher Education Commission Islamabad in terms of Indigenous Scholarship program.

References

[1] Imran Aziz, Siraj-ul-Islam, Bozidar Sarler, Wavelets collocation method for numerical solution of elliptic problems, Appl. Math. Model. 37 (2013) 676–
694.
[2] Siraj-ul-Islam, Imran Aziz, B. Šarler, The numerical solution of second-order boundary-value problems by collocation method with the Haar wavelets,
Math. Comput. Model. 50 (2010) 1577–1590.
[3] Siraj-ul-Islam, B. Šarler, Imran Aziz, Fazal Haq, Haar wavelet collocation method for the numerical solution of boundary layer fluid flow problems, Int.
J. Therm. Sci. 50 (2011) 686–697.
[4] Wolfgang Dahmen, Andrew Kurdila, Peter Oswald, Multiscale Wavelet Methods for Partial Differential Equations, Academic Press, 1997.
[5] Imran Aziz, Siraj-ul-Islam, New algorithms for numerical solution of nonlinear Fredholm and volterra integral equations using Haar wavelets, J. Appl.
Math. Comput. 239 (2013) 333–345.
[6] Siraj-ul-Islam, I. Aziz, M. Fayyaz, A new approach for the numerical solution of integro-differential equations via Haar wavelets, Int. J. Comput. Math.
(2013).
[7] Siraj-ul-Islam, Imran Aziz, Fazal Haq, A comparative study of numerical integration based on Haar wavelets and hybrid functions, Comput. Math. Appl.
59 (2010) 2026–2036.
[8] Imran Aziz, Siraj-ul-Islam, Wajid Khan, Quadrature rules for numerical integration based on haar wavelets and hybrid functions, Comput. Math. Appl.
61 (9) (2011) 2770–2781.
[9] Siraj-ul-Islam, Imran Aziz, Wajid Khan, Numerical integration of multi-dimensional highly oscillatory, gentle oscillatory and non-oscillatory
integrands based on wavelets and radial basis functions, Eng. Anal. Bound. Elem. 36 (2012) 1284–1295.
[10] M. Dehghan, M. Lakestani, Numerical solution of nonlinear system of second-order boundary value problems using cubic B-spline scaling functions,
Int. J. Comput. Math. 85 (2008) 1455–1461.
[11] V. Comincioli, G. Naldi, T. Scapolla, A wavelet-based method for numerical solution of nonlinear evolution equations, Appl. Numer. Math. 33 (2000)
291–297.
[12] J.L. Wu, A wavelet operational method for solving fractional partial differential equations numerically, Appl. Math. Comput. 214 (2009) 31–40.
[13] L.A. Diaz, M.T. Martin, V. Vampa, Daubechies wavelet beam and plate finite elements, Finite Elem. Anal. Des. 45 (2009) 200–209.
[14] X. Zhu, G. Lei, G. Pan, On application of fast and adaptive Battle–Lemarie wavelets to modelling of multiple lossy transmission lines, J. Comput. Phys.
132 (1997) 299–311.
[15] E. Babolian, F. Fattahzdeh, Numerical solution of differential equations by using Chebyshev wavelet operational matrix of integration, Appl. Math.
Comput. 188 (2007) 417–426.
[16] Ü. Lepik, Numerical solution of evolution equations by the Haar wavelet method, Appl. Math. Comput. 185 (2007) 695–704.
[17] C.F. Chen, C.H. Hsiao, Haar wavelet method for solving lumped and distributed-parameter systems, IEE Proc. Control Theory Appl. 144 (1997) 87–94.
[18] Xing Tao Wang, Numerical solution of time-varying systems with a stretch by general Legendre wavelets, Appl. Math. Comput. 198 (2008) 613–620.
[19] C.H. Hsiao, Haar wavelet approach to linear stiff systems, Math. Comput. Simul. 64 (2004) 561–567.
[20] Chun hui Hsiao, Wen-June Wang, Haar wavelet approach to nonlinear stiff systems, Math. Comput. Simul. 57 (2001) 347–353.
[21] Ü. Lepik, Haar wavelet method for nonlinear integro- differential equations, Appl. Math. Comput. 176 (2006) 324–333.
[22] K. Maleknejad, F. Mirzaee, Using ratiionalized Haar wavelet for solving linear integral equations, Appl. Math. Comput. 160 (2005) 579–587.
[23] E. Babolian, A. Shshsawaran, Numerical solution of nonlinear Fredholm integral equations of the second kind using Haar wavelets, J. Comput. Appl.
Math. 225 (2009) 87–95.
[24] K. Maleknejad, S. Shorabi, Numerical solution of fredholm integral equation of the first kind by using Legendre wavelets, Appl. Math. Comput. 186
(2007) 579–587.
[25] Ibrahim Sadek, Taher Abualrub, Marwan Abukhaled, A computational method for solving optimal control of a system of parallel beams using Legendre
wavelets, Math. Comput. Model. 45 (2007) 1253–1264.
[26] J.D. Murray, Mathematical Biology I, Springer-Verlag, Berlin, 2002.
[27] J. Bear, Hydraulics of Groundwater, Dover, Minneola, 2007.
[28] D.K. Hetrick, Dynamics of Nuclear Reactors, University of Chicago, Chicago, 1971.
[29] S.N. Atluri, The Meshless Method MLPG for Domain and BIE Discretizaion, Tech. Sci. Press USA, 2004.
[30] J.M. Burger, A mathematical model illustrating the theory of turbulence, Adv. Appl. Mech. 1 (1948) 171–199.
[31] J. Nee, J. Duan, Limit set of trajectories of the coupled viscous Burgers’ equations, Appl. Math. Lett. 11 (1998) 57–61.
[32] J. Perko, B. Šarler, Weight function shape parameter optimization in meshless methods for non-uniform grids, CMES, Comput. Model. Eng. Sci. 19
(2007) 55–68.
[33] D.L. Younga, C.M. Fana, S.P. Hua, S.N. Atluri, The Eulerian–Lagrangian method of fundamental solutions for two-dimensional unsteady Burgers’
equations, Eng. Anal. Bound. Elem. 32 (2009) 395–412.
[34] W. Zhang, C.H. Zhang, G. Xi, An explicit Chebyshev pseudospectral multigrid method for incompressible Navier–Stokes equations, Comput. Fluids 39
(2010) 178–188.
[35] X.H. Zhang, J. Ouyang, L. Zhang, Element-free characteristic Galerkin method for Burgers’ equation, Eng. Anal. Bound. Elem. 33 (2009) 356–362.
[36] M. Addam, A. Bouhamidi, K. Jbilou, A numerical method for one-dimensional diffusion problem using fourier transform and the B-spline Galerkin
method, Appl. Math. Comput. 215 (2010) 4067–4079.
Siraj-ul-Islam et al. / Applied Mathematical Modelling 37 (2013) 9455–9481 9481

[37] G. Yao, C.S. Chen, B. Sarler, A comparison of three explicit local meshless methods using radial basis functions, Eng. Anal. Bound. Elem. 35 (2011) 600–
609.
[38] R. Vertnik, B. Šarler, Meshless local radial basis function collocation method for convective–diffusive solid–liquid phase change problems, Inter. J.
Numer. Methods Heat Fluid Flow 16 (2006) 617–640.
[39] Yong-Tao Zhang, Qing Nie, Frederic Y.M. Wana, Xin-Feng Liu, Compact integration factor methods in high spatial dimensions, J. Comput. Phys. 227
(2008) 5238–5255.
[40] G. Yao, Siraj-ul-Islam, Bozidar Sarler, A comparative study of global and local meshless method for diffusion-reaction, CMES 59 (2010) 127–154.
[41] Y. Li Duan, Ru Xun Liu, Lattice Boltzmann model for two-dimensional unsteady Burgers’ equation, J. Comput. Appl. Math 206 (2007) 432–439.
[42] C.A.J. Fletcher, A comparison of finite element and finite difference solution of the one and two dimensional Burgers’ equations, J. Comput. Phys. 51
(1983) 159–188.
[43] S.F. Radwan, On the fourth-order accurate compact ADI scheme for solving the unsteady nonlinear coupled Burgers’ equations, J. Nonlinear Math. Phys.
6 (1999) 13–34.
[44] A.R. Bahadir, A fully implicit finite difference scheme for two dimensional Burgers’ equations, Appl. Math. Comput. 137 (2003) 131–137.
[45] Arshed Ali, Siraj-ul-Islam, Siraj ul Haq, A computational meshfree technique for the numerical solution of the two-dimensional coupled Burgers’
equations, Int. J. Comput. Methods Eng. Sci. Mech. 10 (2009) 406–412.
[46] M. Razzaghi, S. Yousefi, Legendre wavelets direct method for variational problems, Math. Comput. Simul. 53 (2000) 185–192.
[47] D.L. Young, C.M. Fan, S.P. Hu, S.N. Atluri, The Eulerian–Lagrangian method of fundamental solutions for two-dimensional unsteady Burgers’ equations,
Eng. Anal. Bound. Elem. 32 (2008) 395–412.

You might also like