Applied Mathematics and Computation: Alper Korkmaz, Idris Da G

Applied Mathematics and Computation 274 (2016) 208–219
Contents lists available at ScienceDirect
Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc
Quartic and quintic B-spline methods for advection–diffusion

equation
Alper Korkmaz a,∗, Idris Dağ b
a
Mathematics Department, Çankırı Karatekin University, Çankırı, Turkey
b
Department of Mathematics and Computer Science, Eskişehir Osmangazi University, Eskişehir, Turkey
a r t i c l e i n f o a b s t r a c t
Keywords: Differential quadrature methods based on B-spline functions of degree four and five have been
B-spline introduced to solve advection–diffusion equation numerically. Two initial-boundary value
Differential quadrature method
problems modeling the transportation of a concentration and distribution of an initial pulse
Advection–diffusion equation
are simulated using both methods. The errors of the numerical results obtained by both meth-
ods have been computed. Stability analysis for both methods is also studied by the use of
matrix stability.
© 2015 Elsevier Inc. All rights reserved.
1. Introduction
In industrialized world, a wide variety of contaminants are released to the environment everyday from industrial sources and
plants. Modeling the contamination rate or transportation of contamination physically may probably the first step to solve an
environment problem. The advection-diffusion equation (ADE) is a mathematical model for transport, dispersion, diffusion or
intrusion in various media. Consider one-dimensional form of the ADE given by
∂U ∂U ∂ 2U
+α − β 2 = 0, 0 ≤ x ≤ L (1)
∂t ∂x ∂x
with the initial condition
U (x, 0) = U0 (x), 0≤x≤L (2)
and the boundary conditions
U (0, t ) = f (t ), U (L, t ) = g(t ) (3)
are parameters, ∂∂Ux and ∂∂ xU2 are advection and diffusion terms, respectively [1]. In many
2
in a finite domain [0, L] where α and β
environment problems, U(x, t) represents concentration of the pollutant or contaminant material at point x at the time t. Some-
times, the solutions refer to mass, heat, water or energy transportation in various media containing draining film or soil [2–4].
In some studies, the ADE models many engineering and chemistry problems covering dispersion in porous media, the intrusion
of fluids of different densities, the absorption of chemicals, dispersion of contaminants in rivers, lakes, embouchures and coasts,
flow of a solute material through a tube, the transportation of pollutants in atmosphere, cooling problems in generators, the ther-
mal pollution in water systems [5–13] etc. The ADE was solved numerically as a model in some financial forecasting problems
[14].
∗
Corresponding author. Tel.: +90 545 6497717.
E-mail address: [email protected] (A. Korkmaz).
http://dx.doi.org/10.1016/j.amc.2015.11.004
0096-3003/© 2015 Elsevier Inc. All rights reserved.
A. Korkmaz, I. Dağ / Applied Mathematics and Computation 274 (2016) 208–219 209
Differential Quadrature Method (DQM) is a derivative approximation technique constructed on the sum of the functional val-
ues at finite problem domain to solve partial differential equations numerically [15]. So far, many types of differential quadrature
methods derived from various basis functions covering Lagrange interpolation polynomials, Hermite polynomials, radial basis
functions, sinc functions and spline functions to solve many engineering and physics problems [16–37].
2. Numerical methods
Consider the uniform grid distribution a = x1 < x2 < · · · < xN = b with N grid points of the interval [a, b] and the functional
values of U(x, t) at each grid and time t are defined as U(xi , t), i = 1, . . . , N. The approximation to derivatives of U(x, t) with respect
to the variable x is
∂ U (r) (xi , t ) N
= w(i jr)U (x j , t ), i = 1, 2, . . . , N, (4)
∂ x(r) j=1
(r)
where wi j stands for the weighting coefficients of the approximation and r is the order of the derivative [15]. According to the
(r)
method, before approximate to derivative terms, the weighting coefficients wi j should be computed by substituting the basis
functions in Eq. (4) instead of U(xi , t). In this study, both quartic and quintic B-spline functions are considered to determine the
weighting coefficients.
2.1. Quartic B-spline differential quadrature method (QRTDQ)
Let ϕ i (x) be a quartic B-spline. Then, the quartic B-spline functions set {ϕ−1 (x), ϕ0 (x), . . . , ϕN+1 (x)} forms a basis set for
functions defined over [x1 = a, b = xN ]. The quartic B-spline functions ϕ i (x) are defined as:
⎧
⎪(x − xi−2 )4 , [xi−2 , xi−1 ]
⎪
⎨(x − xi−2 )4 − 5(x − xi−1 )4 , [xi−1 , xi ]
1
ϕi (x) = 4 (x − xi−2 )4 − 5(x − xi−1 )4 + 10(x − xi )4 , [xi , xi+1 ] (5)
h ⎪⎪
⎩(xi+3 − x)4 − 5(xi−2 − x) , [xi+1 , xi+2 ]
4 4
(xi+3 − x) , [xi+2 , xi+3 ]

where h = xi − xi−1 is the equal sub interval size [38]. All the functional values of the quartic B-spline ϕ i (x) outside the sub
interval [xi−2 , xi+3 ] are zero. Substitution of each quartic B-spline function into the DQM equation (4) for a fixed xi and r gives
d(r) ϕm (xi ) (r)

m+2
(r )
= wi j ϕm (x j ), m = −1, 0, . . . , N + 1, (6)
dx j=m−1
The linear equation system (6) can be rewritten in the matrix notation as;
⎡ r) ⎤
w(i,−2
ϕ−1,−2 ϕ−1,−1 ϕ−1,0 ϕ−1,1 ⎢ (r) ⎥
ϕ0,−1 ϕ0,0 ϕ0,1 ϕ0,2 ⎢w ⎥= (7)
⎣ i,−1 ⎦
ϕN+1,N ϕN+1,N+1 ϕN+1,N+2 ϕN+1,N+3
w(i,N+3
r)
for any xi in the interval [a, b]. In the matrix form, ϕ i, j denotes ϕ i (xj ) and
T
d(r) ϕ−1 (xi ) d(r) ϕ0 (xi ) d(r) ϕN+1 (xi )
= ( )
, ( )
,...,
dx r dx r dx(r)
(r)
Even though the linear equation system (7) contains N + 3 equations, it has N + 6 unknowns, namely wi j , j = −2, −1, . . . , N + 3.
In order to reduce it to a solvable system, three more equations
d(r+1) ϕ−1 (xi ) 1

w(i,rj) ϕ−1 (x j )

=
dx(r+1) j=−2
d(r+1) ϕN (xi ) (r)

N+2
(r+1)
= wi, j ϕN (x j )
dx j=N−1
d(r+1) ϕN+1 (xi ) (r)

N+3
= wi, j ϕN+1 (x j )
dx(r+1) j=N
210 A. Korkmaz, I. Dağ / Applied Mathematics and Computation 274 (2016) 208–219
Fig. 1. Stability region when eigenvalues are complex.
are used. Finally, the equation and unknown number equalized system is of the form;
Mw = (8)
where ⎡ ⎤
ϕ−1,−2 ϕ−1,−1 ϕ−1,0 ϕ−1,1

⎢ϕ−1,−2 ϕ−1,−1 ϕ−1,0 ϕ−1,1 ⎥
⎢ ⎥
⎢ ϕ0,−1 ϕ0,0 ϕ0,1 ϕ0,2 ⎥
⎢ ϕN,N−1 ϕN,N ϕN,N+1 ϕN,N+2 ⎥
M=⎢ ⎥,
⎢
ϕN,N−1

ϕN,N

ϕN,N+1

ϕN,N+2 ⎥
⎢ ⎥
⎣ ϕN+1,N ϕN+1,N+1 ϕN+1,N+2 ϕN+1,N+3 ⎦

ϕN+1,N ϕN+1,N+1 ϕN+1,N+2 ϕN+1,N+3
T
(r) (r) (r)
w = wi,−2 , wi,−1 , . . . , wi,N+3 and
T
d(r+1) ϕ−1 (xi ) d(r) ϕ−1 (xi ) d(r) ϕ0 (xi ) d(r+1) ϕN (xi ) d(r) ϕN+1 (xi ) d(r+1) ϕN+1 (xi )
= , , ,..., , ,
dx(r+1) dx(r) dx(r) dx(r+1) dx(r) dx(r+1)
(r)
In order to obtain all weighting coefficients wi j , the system is solved for each grid xi covering the whole problem domain[26].
2.2. Quintic B-spline differential quadrature method (QNTDQ)
The quintic B-spline functions are defined as:

⎧
⎪η, [xi−3 , xi−2 ]
⎪
⎪
⎨η − 6χ ,
⎪ [xi−2 , xi−1 ]
1 η − 6χ + 15κ , [xi−1 , xi ]
i (x) = 5 (9)
h ⎪⎪η − 6χ + 15κ − 20(x − xi )5 , [xi , xi+1 ]
⎪
⎩η − 6χ + 15κ − 20(x − xi )5 + 15(x − xi+1 )5 ,
⎪ [xi+1 , xi+2 ]
5 5
η − 6χ + 15κ − 20(x − xi ) + 15(x − xi+1 ) − 6(x − xi+2 )5 , [xi+2 , xi+3 ]
where η = (x − xi−3 )5 , χ = (x − xi−2 )5 , κ = (x − xi−1 )5 , and the space step size h = xi − xi−1 , over the interval [x1 = a, b = xN ]
[38]. The N + 4 quintic B-splines set {−1 (x), 0 (x), . . . , N+2 (x)} forms a basis for the functions defined over [x1 = a, b = xN ].
Fig. 2. Eigenvalue distribution for α = 0.8, β = 0.005, N = 46.
Substitution of each quintic B-spline into the DQM approximation (4)for a fixed xi and r generates the equation system
d(r) m (xi ) (r)

m+2
(r )
= wi j m (x j ), m = −1, 0, . . . , N + 2, (10)
dx j=m−2
(r)
where r is only a notation that can be thought as a third index showing order of the derivative approximation in wi j . In order to
(r)
determine the weighting coefficient wi, j related to the grid point xi , the linear equation system
Mw = (11)
where
⎡ ⎤
−1,−3 −1,−2 −1,−1 −1,0 −1,1
⎢ 0,−2 0,−1 0,0 0,1 0,2 ⎥
M=⎣ ⎦,
N+1,N−1 N+1,N N+1,N+1 N+1,N+2 N+1,N+3
N+2,N N+2,N+1 N+2,N+2 N+2,N+3 N+2,N+4
T
w = w(i,−3
r)
, w(i,−1
r)
, . . . , w(i,N+4
r)
Fig. 3. Eigenvalue distribution for α = 0.5, β = 0, N = 26.
and
T
d(r) −1 (xi ) d(r) 0 (xi ) d(r) N+2 (xi )
= ( )
, ( )
,...,
dx r dx r dx(r)
is solved. Unfortunately, the linear equation system ( 11) contains N + 4 equations and N + 8 unknowns. Adding four more equa-
tions
d(r+1) −1 (xi ) 1

w(i,rj) −1 (x j )

= (12)
dx(r+1) j=−3
d(r+1) 0 (xi ) 2
w(i,rj) 0 (x j )

= (13)
dx(r+1) j=−2
d(r+1) N+1 (xi ) (r)

N+3
(r+1)
= wi, j N+1 (x j ) (14)
dx j=N−1
d(r+1) N+2 (xi ) (r)

N+4
(r+1)
= wi, j N+2 (x j ) (15)
dx j=N
Fig. 4. Transport simulation for various times.
Fig. 5. QRTDQ errors at time t = 9600 with the parameters h = 100, t = 100.
Fig. 6. QNTDQ errors at time t = 9600 with the parameters h = 100, t = 100.
Fig. 7. Fade out of initial pulse.
to the system (11), the number of unknowns and equations becomes equal. Thus, the system (11) becomes a solvable system. It
should be noted that, the weighting coefficients wi,−2 , wi,−1 , wi,0 , wi,N+1 , wi,N+2 , wi,N+3 , wi,N+4 are not originated from the grid
points in the interval, they are omitted in the DQM approximation.
Fig. 8. Error of QRTDQ at the terminating time t = 5.
3. Numerical discretization and stability analysis
Substituting the DQM approximations for the derivative terms with respect to x and boundary conditions reduces the ADE to
an ordinary differential equation system with respect to the independent variable t as,
∂ U (xi , t ) (1)
N−1 (2)
N−1
= −α wi j U (x j , t ) + β wi j U (x j , t ) + Si , i = 1, 2, . . . , N (16)
∂t j=2 j=2
(1) (1) (2) (2)

where Si = −α(wi1 f (t ) + wiN g(t )) + β(wi1 f (t ) + wiN g(t )). Let A be a coefficient vector of the system (16) whose entries are
(1) (2)
−α wi j + β wi j . Thus, the system can be written in matrix form as
du
= [A]u + s (17)
dt
where s is the vector resulted by boundary conditions and u = [U (x1 , t ), U (x2 , t ), . . . , U (xN , t )]. Since s is zero due to the homoge-
nous boundary conditions used in this study, the stability of Runge–Kutta methods for Eq. (17) can be discussed by applying the
method to the scalar equation
du
= λi u
dt
where λi , i = 1, 2, . . . , N are the eigenvalues of the matrix A. The Runge–Kutta methods are absolutely stable if
|Ei (λi t )| < 1, i = 1, 2, . . . N
where Ei (λi t) is an approximation to eλi t and real component of the eigenvalue Re(λi ) is negative [39].
Fig. 9. Error of QNTDQ at the terminating time t = 5.
The stability condition for the fourth order method is as t → ∞ requires that −2.78√< λi t < 0 if all eigenvalues are pure
real. For pure imaginary eigenvalues the condition becomes of the form 0 < |λi t | < 2 2. If the eigenvalues has both real and
complex components, the stability region for λi t is extended to the closed region, Fig. 1, [39].
One way to determine the eigenvalues of A is to follow the idea that the eigenvalues of similar matrices are equal. The trans-
formation of A to its similar Hessenberg matrix H by the QR method and determination of the eigenvalues are accomplished
by the use of the algorithm modified from the C routine given in [40]. The eigenvalue distribution of the matrix H for various
parameters are shown in Figs. 2–3.
In this study, the time integration of Eq. (16) is accomplished by both fourth order Runge–Kutta method (RK4) and sixth order
Runge–Kutta–Butcher method (RKB6).
The numerical approximation error is computed by use of the discretized maximum norm and L2 norm defined by

L∞ = max Uiexact − Uinumeric ,
i
12
N
2 (18)
L2 = h (
Uiexact − Uinumeric )
i=1
to check the accuracy of the proposed methods.
3.1. Test problem A
Consider pure advection problem obtained by choosing β = 0. The analytic solution of pure advection problem used in the
study [1] is
1

U (x, t ) = 10 exp − (x − x0 − αt )2 (19)
2ρ 2
Table 1
Comparison of methods at t = 9600 s.
Time integration method RK4 RKB6
Method h t L2 × 10 2
L∞ × 10 2
CPUT(s) L2 × 102 L∞ × 102 CPUT (s)
QRTDQ 400 100 1805.4 54.2 0.30 1805.4 54.1 0.67

200 100 74.4 2.4 1.34 75.2 2.5 2.59
100 100 3.2 0.1 4.89 4.3 0.1 10.14
400 50 1805.4 54.3 0.56 1807.3 54.1 1.31
200 50 75.0 2.5 2.46 75.4 2.5 5.14
100 50 4.4 0.1 9.73 4.4 0.1 20.20
50 50 2.2 0.1 40.35 2.4 0.1 80.43
QNTDQ 400 100 964.0 27.3 0.15 964.0 27.1 0.40
200 100 11.0 0.3 0.54 10.3 0.3 1.82
100 100 2.3 0.1 2.09 1.1 0.1 7.21
400 50 964.4 27.2 0.32 964.0 27.1 0.93
200 50 10.0 0.3 1.09 10.2 0.3 3.73
100 50 2.2 0.1 4.14 2.2 0.1 14.46
50 50 2.2 0.2 16.57 2.3 0.2 57.78
25 50 2.1 0.2 74.28 2.6 0.2 230.56
Femlsf [1] 400 50 10571.2 345.1
200 50 3287.4 135.0
100 50 1059.6 49.4
50 50 798.4 38.0
25 50 788.1 37.7
Femqsf [1] 400 50 8173.6 224.7
200 50 1255.5 51.8
100 50 795.1 37.6
50 50 790.8 37.3
25 50 790.8 37.9
Table 2
Comparison of the error norms.
RK4 RKB6
Method N L2 L∞ CPUT L2 × 102 L∞ × 102 CPUT
−2 −1 −2 −1
QRTDQ 21 9.84 × 10 1.42 × 10 0.46 9.84 × 10 1.42 × 10 1.62
31 9.71 × 10−2 1.42 × 10−1 1.04 9.71 × 10−2 1.42 × 10−1 3.56
46 9.21 × 10−2 1.26 × 10−1 2.25 9.21 × 10−2 1.26 × 10−1 7.71
91 2.52 × 10−3 5.08 × 10−3 8.67 2.52 × 10−3 5.09 × 10−3 30.25
181 4.59 × 10−5 7.68 × 10−5 35.01 4.63 × 10−5 7.75 × 10−5 121.00
QNTDQ 21 1.02 × 10−1 1.40 × 10−1 0.50 1.03 × 10−1 1.40 × 10−1 1.70
31 1.00 × 10−1 1.53 × 10−1 1.03 1.00 × 10−1 1.53 × 10−1 3.56
46 1.05 × 10−1 1.24 × 10−1 2.18 1.05 × 10−1 1.24 × 10−1 7.87
91 7.70 × 10−3 1.17 × 10−2 8.48 7.70 × 10−3 1.17 × 10−2 30.56
181 4.35 × 10−6 6.52 × 10−6 38.39 4.23 × 10−6 6.28 × 10−6 123.60
where ρ is a real parameter. This solution is a model for transportation of an initial concentration of height 10 units whose
peak is at x0 initially along a channel preserving its shape and size as time goes. The numerical simulation is accomplished with
the transportation speed α = 0.5 m/s, initial peak location x0 = 2 kms and ρ = 264 by the terminating time t = 9600 s with
homogenous boundary conditions, Fig. 4. Both RK4 and RKB6 algorithms are run for various time steps (t) and space steps (h).
The errors for both methods at the terminating time t = 9600 are graphed in Figs. 5–6.
The comparison with least square finite element method [1] is reported in Table 1. Total simulation process times (CPUT) in
a standard home computer are also given in Table 1. Both quartic and quintic B-spline methods generate sufficiently accurate
results even larger space steps are selected when compared with least square finite element method. When QRTDQ and QNTDQ
are compared, it is seen that QNTDQ generate more accurate results. Even though the order of the RK4 is lower than RKB6, the
accuracy of both methods are the same, but RK4 is more economical owing to total simulation process times.
3.2. Test problem B
The analytical solution of the initial boundary value problem for the ADE

1 (x − x0 − αt )2
U (x, t ) = √ exp − (20)
4t + 1 β(4t + 1)
models fade out of an initial bell-shaped concentration of height 1 [41]. The initial pulse, the peak of which is located at x0
initially, moves to the right along the horizontal axis as time goes.
The numerical solution for this problem is obtained by use of the initial condition derived from the analytical solution Eq. (20)
for t = 0 and homogenous boundary conditions U (0, t ) = U (9, t ) = 0. The simulations are accomplished with the propagation
speed 0.8 m/s and the diffusion coefficients 0.005 m2 /s for t ∈ [0, 5], Fig. 7. The errors obtained both methods at the end of
simulation time t = 5 for N = 181 are depicted in Figs. 8–9.
The QNTDQ generates better results than the QRTDQ when the same time and space step sizes are used, Table 2. The time
integration method RK4 seems more suitable for this test problem owing to its accuracy and its computational cost.
4. Conclusion
In the present paper, the quartic and the quintic B-spline functions are used to develop differential quadrature methods for
the numerical solutions for the ADE. The weighting coefficients for the derivative approximations are computed by solving a set
of linear equation systems resulted by the use of B-spline functions in the differential quadrature derivative approximation. The
resultant ordinary differential equation systems are integrated in time by the use of RK4 and RK6. A matrix stability analysis has
also been performed by computing eigenvalues of coefficient matrices for both methods.
The accuracy and validity of both approaches are investigated by solving two initial boundary value problems obtained by
combining initial and boundary conditions to the ADE.
The first problem is chosen as pure advection problem. Both methods have generated acceptable solutions and simulated
the transportation of concentration successfully. The numerical results obtained by both proposed methods are compared with
analytical solution by measuring the error between them. When compared with the least square finite elements, the results
obtained via both QRTDQ and QNTDQ seem more accurate.
The numerical solution for initial boundary value problem modeling fade out of an initial concentration has also been studied.
The numerical simulations show that both methods have generated acceptably accurate results.
Acknowledgment
A part of this study was presented orally in Karatekin Mathematics Days 2014, Turkey.
References
[1] I. Dağ, D. Irk, M. Tombul, Least-squares finite element method for the advection-diffusion equation, Appl. Math. Comput. 173 (2006) 554–565.
[2] B.J. Noye, Numerical solution of partial differential equations, Lecture Notes, 1990.
[3] J. Isenberg, C. Gutfinger, Heat transfer to a draining film, Int. J. Heat Transf. 16 (1972) 505–512.
[4] J.Y. Parlarge, Water transport in soils, Ann. Rev. Fluids Mech. 2 (1980) 77–102.
[5] Q.N. Fattah, J.A. Hoopes, Dispersion in anisotropic homogeneous porous media, J. Hydraul. Eng. 111 (1985) 810–827.
[6] P.C. Chatwin, C.M. Allen, Mathematical models of dispersion in rivers and estuaries, Ann.Rev. Fluid Mech. 17 (1985) 119–149.
[7] F.M. Holly, J.M. Usseglio-Polatera, Dispersion simulation in two-dimensional tidal flow, J. Hydraul Eng. 111 (1984) 905–926.
[8] J.R. Salmon, J.A. Liggett, R.H. Gallager, Dispersion analysis in homogeneous lakes, Int. J. Numer. Meth. Eng. 15 (1980) 1627–1642.
[9] L. Lapidus, N.R. Amundston, Mathematics of absorption in beds, J. Phys. Chem. 56 (8) (1952) 984–988.
[10] Z. Zlatev, R. Berkowicz, L.P. Prahm, Implementation of a variable stepsize variable formula in the time-integration part of a code for treatment of long-range
transport of air pollutants, J. Comput. Phys. 55 (1984) 278–301.
[11] C.R. Gane, P.L. Stephenson, An explicit numerical method for solving transient combined heat conduction and convection problems, Int. J. Numer. Meth. Eng.
14 (1979) 1141–1163.
[12] M.H. Chaudhry, D.E. Cass, J.E. Edinger, Modelling of unsteady-flow water temperatures, J. Hydraul. Eng. 109 (5) (1983) 657–669.
[13] N. Kumar, Unsteady flow against dispersion in finite porous media, J. Hydrol. 63 (1988) 345–358.
[14] K.W. Morton, Numerical Solution of Convection–Diffusion Equation, Chapman & Hall, London, 1996.
[15] R. Bellman, B.G. Kashef, J. Casti, Differential quadrature: a tecnique for the rapid solution of nonlinear differential equations, J. Comput. Phys. 10 (1972)
40–52.
[16] R. Bellman, K. Bayesteh, E.S. Lee, R. Vasudevan, Differential quadrature and splines, in: Computers and Mathematics with Applications, Pergamon, Oxford,
1976, pp. 371–376.
[17] J.R. Quan, C.T. Chang, New sightings in involving distributed system equations by the quadrature methods-i, Comput. Chem. Eng. 13 (1989) 779–788.
[18] J.R. Quan, C.T. Chang, New sightings in involving distributed system equations by the quadrature methods-ii, Comput. Chem. Eng. 13 (1989) 71017–71024.
[19] C. Shu, B.E. Richards, Application of generalized differential quadrature to solve two dimensional incompressible Navier stokes equations, Int.J. Num. Meth-
ods Fluids 15 (1992) 791–798.
[20] J. Cheng, B. Wang, S. Du, A theoretical analysis of piezoelectric/composite laminate with larger-amplitude deflection effect, part ii: hermite differential
quadrature method and application, Int. J. Solids Struct. 42 (2005) 6181–6201.
[21] H. Zhong, Spline-based differential quadrature for fourth order equations and its application to kirchhoff plates, Appl. Math. Model. 28 (2004) 353–366.
[22] Q. Guo, H. Zhong, Non-linear vibration analysis of beams by a spline-based differential quadrature method, J. Sound Vib. 269 (2004) 413–420.
[23] H. Zhong, M. Lan, Solution of nonlinear initial-value problems by the spline-based differential quadrature method, J. Sound Vib 296 (2006) 908–918.
[24] I. Bonzani, Solution of non-linear evolution problems by parallelized collocation–interpolation methods, Comput. Math. Appl. 34 (12) (1997) 71–79.
[25] C. Shu, Y.L. Wu, Development of rbf-dq method for derivative approximation and its application to simulate natural convection in concentric annuli, Comput.
Mech. 29 (2002) 477–485.
[26] A. Korkmaz, A.M. Aksoy, I. Dağ, Quartic b-spline differential quadrature method, Int. J. Nonlinear Sci. 11 (4) (2011) 403–411.
[27] T.S. Lee, G.S. Hu, C. Shu, Application of gdq method for study of mixed convection in horizontal eccentric annuli, Int. J. Comput. Fluid Dyn. 18 (1) (2004)
71–79.
[28] A. Korkmaz, I. Dağ, A differential quadrature algorithm for simulations of nonlinear schrödinger equation, Comput. Math. Appl. 56 (9) (2008) 2222–2234.
[29] B. Saka, I. Dağ, Y. Dereli, A. Korkmaz, Three different methods for numerical solutions of the ew equation, Eng. Anal. Boundary Elements 32 (2008) 556–566.
[30] A. Korkmaz, I. Dağ, A differential quadrature algorithm for nonlinear schrödinger equation, Nonlinear Dyn. Volume 56 (2009) 1–2.
[31] A. Korkmaz, I. Dağ, Crank-nicolson - differential quadrature algoritms for the kawahara equation, Chaos&Solitons&Fractals 42 (1) (2009) 65–73.
[32] A. Korkmaz, I. Dağ, Solitary wave simulations of complex modified korteweg-de vries equation using differential quadrature method, Comput. Phys. Com-
mun. 180 (9) (2009) 1516–1523.
[33] R. Jiwari, J. Yuan, A computational modeling of two dimensional reaction–diffusion brusselator system arising in chemical processes, J. Math. Chem. 52 (6)
(2014) 1535–1551.
[34] 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 (12) (2013) 6680–6691.
[35] R. Jiwari, S. Pandit, R.C. Mittal, Numerical simulation of two-dimensional sine-gordon solitons by differential quadrature method, Comput. Phys. Commun.
183 (3) (2012) 600–616.
[36] R. Jiwari, S. Pandit, R.C. Mittal, A differential quadrature algorithm to solve the two dimensional linear hyperbolic telegraph equation with dirichlet and
neumann boundary conditions, Appl. Math. Comput. 218 (13) (2012) 7279–7294.
[37] R.C. Mittal, R. Jiwari, Numerical study of two-dimensional reaction–diffusion brusselator system by differential quadrature method, Int. J. Comput. Methods
Eng. Sci. Mech. 12 (1) (2011) 14–25.
[38] P.M. Prenter, Splines and Variational Methods, A Wiley-Interscience Publication, John Wiley & Sons, New York-London-Sydney-Toronto, 1975.
[39] M.K. Jain, Numerical Solution of Differential Equations, John Wiley & Sons, 1984.
[40] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in C, Cambridge University Press 1988, 1992.
[41] S. Sankaranarayanan, N.J. Shankar, H.F. Cheong, Three-dimensional finite difference model for transport of conservative pollutants, Ocean Eng. 25 (6) (1998)
425–442.

Applied Mathematics and Computation: Alper Korkmaz, Idris Da G

Uploaded by

Applied Mathematics and Computation: Alper Korkmaz, Idris Da G

Uploaded by

Applied Mathematics and Computation 274 (2016) 208–219

Contents lists available at ScienceDirect