ISSN 1995-0802, Lobachevskii Journal of Mathematics, 2016, Vol. 37, No. 3, pp. 268273.


c Pleiades Publishing, Ltd., 2016.

Solution of Stokes Flow Problem Using Biharmonic Equation

Formulation and Multiquadrics Method
R. F. Mardanov* and S. K. Zaripov**

(Submitted by A. V. Lapin)
Kazan Federal University, Kremlevskaya ul. 18, Kazan, Tatarstan, 420008 Russia
Received August 14, 2015

AbstractThe biharmonic equation formulation of the Stokes ow problem for multiquadrics

method is developed. The main advantage of the approach is the iteration free method to nd the
solution. The numerical method is applied for the problem of steady incompressible uid ow past a
cylinder in the periodic cell of the Kuwabara model. The comparison with known analytical solution
and the analysis of absolute and relative errors show that proposed approach gives satisfactory
accuracy. The nonmonotonic dependence of the relative errors on the shape parameter typical for
multiquadrics method is observed.
DOI: 10.1134/S1995080216030161
Keywords and phrases: Stokes ow, Kuwabara model, multiquadrics method.

1. INTRODUCTION
Over the years there has been a high interest in meshless methods of solution of partial dierential
equations (PDE) describing various physical processes. One of the reasons is the simple formulation
and realization of the corresponding numerical problem. The meshless (grid free) scheme used for
representing surfacesthe multiquadrics method (MQ) was rstly proposed to solve the PDE by
Kansa [1, 2]. The radial basic functions (RBF) were used to interpolate the variable functions. The
method was applied to solve the Stokes ow problem by Young et al. [3], Chatsiriwan [4], Kutanaei
et al. [5]. The velocity-vorticity formulation and MQ with iterative scheme were used to solve the
cavity ow in the mentioned papers. The biharmonic problem for the stream function is reduced
to harmonic problems for the ow velocity components and vorticity. These harmonic problems are
solved iteratively using the relation of the vorticity with the derivatives of the velocity components as
the boundary conditions for the harmonic problem of vorticity. The extension of MQ that includes
the numerical scheme, the Method of Approximate Particular Solutions, for Stokes ow problems in
bounded domains was developed by Bustamante et al. [6]. The unsteady Stokes ow problem was solved
using MQ by Golbabai [7].
In this work the MQ method is applied directly to the solution of the biharmonic equation for
the stream function in the Stokes ow problem. This approach has the advantages over the method
described by Young et al. [3] as the only dependent variable is the stream function instead of two velocity
components and the vorticity and we can calculate it without iterations. The solution of the PDE for the
stream function of fourth order requires the determination of two conditions on the boundary. To provide
these conditions the additional outer points are introduced as in [7]. Direct solution of the biharmonic
equation by MQ (Kansas method) was also applied in the work [8]. But due to the two conditions on the
one boundary an over-determined algebraic system for the coecients in the linear combination of RBF
was obtained. The system was solved using least square method. The proposed approach of the MQ
method is applied to solve the problem of uid ow through a cylinder in the periodic cell of the Kuwabara
model [9].
*
E-mail: Renat.Mardanov@kpfu.ru
**
E-mail: Shamil.Zaripov@kpfu.ru

268
 SOLUTION OF STOKES FLOW PROBLEM 269

(a) (b)

Fig. 1.

2. THE PROBLEM FORMULATION

Consider the two-dimensional problem of steady state incompressible uid ow at small Reynolds
numbers in the region G limited by the boundary (g. 1, a). The problem, assuming the Stokes ow
approximation, can be reduced to the solution of the biharmonic equation [10]
4 4 4
L = 2 = + 2 + =0 (1)
x4 x2 y 2 y 4
for the stream function (x, y) with the boundary conditions
L1 = f1 (x, y), L2 = f2 (x, y), (2)
where L1 and L2 may be Dirichlet, Neumann, Robin or Laplace type boundary conditions, f1 (x, y) and
f2 (x, y) are the given functions.

3. SOLUTION
To apply the multiquadric method, the unknown stream function is presented as a sum in the form
given in equation (3)

N
(x, y) = j rj (x, y), (3)
j=1

where rj (x, y) = (x xj )2 + (y yj )2 + c2 are the RBF, {xj , yj } are the collocation points, c is the
free shape parameter. The partial derivatives of the radial basic functions rj (x, y) can be written as:
rj x xj rj y yj 2 rj (y yj )2 + c2 2 rj (x xj )2 + c2
= , = , = , = ,
x rj y rj x2 rj3 y 2 rj3
rj2 + c2 rj4 + 6rj2 c2 15c2
rj = , 2 rj = . (4)
rj3 rj7
The partial derivatives of the stream function in the form given in equation (4) are substituted into
equation (1) and the boundary conditions (2). By the collocation method we obtain the system of linear
algebraic equations (SLAE), given in equation (5), for the unknown quantities j

N
aij j = bi , i = 1, N, (5)
j=1

where the coecients aij and components bi of the free term vector are determined by the formulas
aij = Lrj (xci , yci ), bi = 0, i = 1, n, j = 1, N,
aij = L1 rj (xci , yci ), bi = f1 (xci , yci ), i = n + 1, n + m, j = 1, N,
aij = L2 rj (xci , yci ), bi = f2 (xci , yci ), i = n + m + 1, n + 2m, j = 1, N.

LOBACHEVSKII JOURNAL OF MATHEMATICS Vol. 37 No. 3 2016

270 MARDANOV, ZARIPOV

In order to formulate the Stokes problem for the stream function, it is necessary to apply the two
boundary conditions given by equation (2) at each boundary point. This requires the addition of extra
nodes located outside the domain G [7]. The number of additional nodes should be equal to the number
of the boundary nodes. This approach allows to avoid the solution of over-determined algebraic system
for the coecients in the linear combination of RBF as in [8].
The node points {xj , yj }, j = 1, n + 2m (g. 1, b) consists of the interior points of G {xj , yj }, j = 1, n,
(circle symbols), the boundary points {xj , yj }, j = n + 1, n + m (square symbols), the additional outer
points {xj , yj }, j = n + m + 1, n + 2m (triangular points). We denote by {xcj , ycj } the collocation
points where equation (1) and boundary conditions (2) must be satised: {xcj , ycj } = {xj , yj }, j =
1, n + m; {xcj , ycj } = {xjm , yjm }, j = n + m + 1, n + 2m.
The solution of (5) gives the unknown quantities j . After that the stream function (x, y) is
determined from the formula (3). The velocity components and vorticity can be found from formula
 rj (x, y)  rj (x, y)
N N

u(x, y) = = j , v(x, y) = = j ,
y y x x
j=1 j=1


N
(x, y) = = j rj (x, y). (6)
j=1

The proposed approach of applying the multiquadrics method to the solution of the Stokes problem
provides two main advantages compared with the MQ method with iterative scheme. It enables a
reduction in the number of unknown functions and nd the solution of the Stokes problem without
iteration. We have considered a partial dierential equation of fourth order and it creates no additional
calculation diculties for MQ.

4. RESULTS
The proposed approach of the MQ method is used to solve the uid ow past a cylinder in the periodic
cell of the Kuwabara model [9]. The uid domain is the half annulus region formed by the outer half-circle
of radius h > 1 and the inner half-circle of unity radius (g. 2, a). On the cylinder surface BC the slip
conditions u = v = 0 are given in the form (Dirichlet and Neumann type boundary conditions)

= 0, = 0, (7)
n
where n is the exterior normal to the boundary. On the other boundary AD the conditions from Kuwabara
model are given (Dirichlet and Laplace type boundary conditions)

ur = = U cos , = = 0, (8)
s
where ur is the radial component of velocity vector, s is the tangent to the boundary, U is the given
outer ow velocity, is the polar angle. On the boundaries AB and CD we taken the zero ow rate and
symmetry condition u/y = 0 (Dirichlet and Laplace type boundary conditions)
2
= 0, = 0. (9)
y 2
The analytical solution of the boundary value problem (1), (7), (8) and (9) was obtained by
Kuwabara [9]. The expressions for stream function (x, y) and u (x, y), v (x, y), (x, y) are written
in cylindrical coordinates (r, ):
 
A
(r, ) = + Br + Cr ln r + Dr 3 sin , (10)
r
where
U (1 /2) U (1 ) U U
A= , B= , C= , D= ,
2 k 2 k k 4 k
LOBACHEVSKII JOURNAL OF MATHEMATICS Vol. 37 No. 3 2016
 SOLUTION OF STOKES FLOW PROBLEM 271

(a) (b)
U

r=h

r=1

A B C D

Fig. 2.

(a) 0 (b) 0
0.4 0.4
0.8 0.8
1.2 1.2
1.6 1.6
2.0 2.0

Fig. 3.

1 1 3
= h2 , k = 2 ln .
4 2 4
The velocity components and vorticity are written in the form
 
1 A 2
ur (r, ) = = + B + C ln r + Dr cos ,
r r2
 
A
u (r, ) = = 2 + B + C(ln r + 1) + 3Dr 2 sin ,
r r
 
   
1 1 2 C
(r, ) = r + 2 = 2 + 4Dr sin . (11)
r r r r 2 r
In this work the nodes of a structured radial mesh were taken as a centers and collocation points
(g. 2, b). The number of collocation points is larger close to the cylinder wall where the higher gradients
are expected. Taking U = 1, h = 5 the comparison of the distributions of stream function (streamlines)
and vorticity from the numerical (g. 3, a) and analytical (g. 3, b) models shows good agreement.
To estimate the accuracy of the method the absolute (E (x, y), Eu (x, y), Ev (x, y), E (x, y)) and
relative errors ( , u , v , ) were calculated for the functions (x, y), u(x, y), v(x, y), (x, y):
E (x, y) = (x, y) (x, y), Eu (x, y) = u (x, y) u(x, y),
Ev (x, y) = v (x, y) v(x, y), E (x, y) = (x, y) (x, y),
max |E (x, y)| max |Eu (x, y)|
=
, u =  ,
max | (x, y)| max u2 (x, y) + v 2 (x, y)
max |Ev (x, y)| max |E (x, y)|
v =  , = .
2
max u (x, y) + v (x, y)2 max | (x, y)|
The distributions of E (x, y), Eu (x, y), Ev (x, y), E (x, y) for N = 642 at c = 0.37 are shown in (g. 4).
It is seen that the lowest values of relative error are obtained for the stream function (x, y), the errors for
u(x, y) and v(x, y) are higher. The greatest errors are observed for (x, y) that is obtained as a second

LOBACHEVSKII JOURNAL OF MATHEMATICS Vol. 37 No. 3 2016

272 MARDANOV, ZARIPOV

p U
0.002 0.02
0.006 0.01
0.010 0
0.014 0.01
0.018 0.02
0.022 0.03
0.026 0.04
0.030 0.05

V
0.035 0.08
0.025 0.04
0.015 0
0.005 0.04
0.005 0.08
0.015 0.12
0.025 0.16
0.035 0.20

Fig. 4.

100 100
u
101 101

102 102

103 103
0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0
c c
100 100

101 101

102 102

103 103
0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0
c c

Fig. 5.

derivative of the stream function. The values of relative error are consisted = 0.0065, u = 0.035,
v = 0.024, = 0.098.
The relative errors calculated as a shape parameter c function at the various values of number of
collocation points are shown in (g. 5), (solid, dashed, dotted curves correspond to N = 332, 642, 1319).
Typical for MQ the nonmonotonic dependence of relative errors on the shape parameter is observed.
The functions (c), u (c), v (c), (c) have one minimum for the various values of c. The value of the

LOBACHEVSKII JOURNAL OF MATHEMATICS Vol. 37 No. 3 2016

 SOLUTION OF STOKES FLOW PROBLEM 273

shape parameter at which this minimum occurs decreases with an increase in N . The conclusion can be
made that the optimal value of c, c , is proportional to the average distance between adjacent collocation
points. The increase in the number of collocation points gives no higher accuracy due to round-o errors.
As an optimal parameter c for the entire task the average value between the minimums of ( , u , v ,
) can be taken. In the problem considered c = 0.5, 037, 0.27 for N = 332, 642, 1319.

5. CONCLUSIONS
The multiquadrics method for the biharmonic equation formulation of the Stokes ow problem is
developed. Two main advantages of the approach are the reduction of unknown functions and the
iteration free method to nd the solution. The approach is applied to the problem of steady incompressible
uid ow past a cylinder in the periodic cell model of Kuwabara. It has been shown that proposed
approach gives satisfactory accuracy.The nonmonotonic dependence of the relative errors on the shape
parameter typical for multiquadrics method is observed.

ACKNOWLEDGMENTS
This work was supported by RFBR (grant 15-01-06135).

REFERENCES
1. E. Kansa, Computers & Mathematics with Applications 19 (8-9), 127145 (1990).
2. E. Kansa, Computers & Mathematics with Applications 19 (8-9), 147161 (1990).
3. D. Young, S. Jane, C. Lin, C. Chiu, and K. Chen, Engineering Analysis with Boundary Elements 28 (10),
12331243 (2004).
4. S. Chantasiriwan, International Communications in Heat and Mass Transfer 34 (3), 313320 (2007).
5. S. S. Kutanaei, N. Roshan, A. Vosoughi, S. Sagha, A. Barari, and S. Soleimani, Engineering Analysis with
Boundary Elements 36, 5, 633638 (2012).
6. C. Bustamante, H. Power, Y. Sua, and W. Florez, Applied Mathematical Modelling 37 (6), 45384547
(2013).
7. A. Golbabai and H. Rabiei, Engineering Analysis with Boundary Elements 36 (11), 15551559 (2012).
8. G. Yao, C.H. Tsai, and W. Chen, Engineering Analysis with Boundary Elements 34, 625631 (2010).
9. S. Kuwabara, Journal of Physical Society of Japan 14 (4), 527532 (1959).
10. J. Happel and H. Brenner, Low Reynolds Number Hydrodynamics with Special Application to Particu-
late Media (Prentice-Hall, Englewood Clis, 1965).

LOBACHEVSKII JOURNAL OF MATHEMATICS Vol. 37 No. 3 2016