JOURNAL OF LATEX CLASS FILES, VOL. , NO.

, 1

123
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 2

Application of The Boundary Element Method on

Two Dimensional Ideal Fluid Flow Problem
Salsabila Aurellia, Mahdhivan Syafwan, Arrival Rince Putri, and Anati bte Ali

Abstract—In this journal, two-dimensional ideal fluid flow Boundary element method (abbreviated BEM) is a method
problems will be discussed. The model derivation was completed that is often used in science and engineering applications.
numerically using the Boundary Element Method (BEM). In BEM has been widely used to solve problems involving the
this case, what is meant by ideal fluid flow is non-viscous and
incompressible fluid flow. The two-dimensional ideal fluid flow Schrödinger equation and the diffusion equation. The reason
problem is derived from the continuity equation which is based for choosing this method is because BEM with constant
on the law of conservation of mass. The BEM is constructed from elements provides a very good numerical solution. Further-
the idea that the solution to the partial differential equation more, the arrangement of the data is very simple, because the
is expressed in the limit integral equation which contains the discretization is only limited by the boundary conditions [7].
fundamental solution. In this method the domain boundary is
partitioned into a finite number of line segments which are then Another reason is the uniqueness of this method in reducing
used to evaluate the integral equation of the boundary. In this the dimensions of the problem, where a problem with a three-
final project BEM is implemented in solving two-dimensional dimensional domain can be reduced to a two-dimensional [2]
ideal fluid flow problems with mixed boundary conditions, problem. In addition, the advantages of this method are its
namely Dirichlet boundary conditions and Neumann boundary ability to handle open domain (infinite) problems, and it is
conditions..
effective in calculating the derivatives of the functions of flux,
Index Terms—Two Dimensional Ideal Fluid Flow Problem, stress, pressure and moment [7].
Boundary Element Method, Boundary Counditions.
Davies and Crann [5] suggest using a spreadsheet such as
Excel for the application used, because this application is well
I. I NTRODUCTION known to many people. However, spreadsheets have limita-
tions in matrix construction and manipulation. In addition,
A partial differential equation is an equation that contains
changes to parameter values (e.g. the number of bounded
the derivative of a function with two or more independent
elements resulting in the size of matrices and vectors) are
variables [4]. One of the problems described by the partial
not handled well by spreadsheets and are also inconvenient to
differential equation is the fluid flow problem. The fluid flow
work with. Although this difficulty can be solved using the
problem can be modeled from the continuity equation. The
Visual Basic Application (VBA) function provided by excel,
continuity equation is one of the three fundamental governing
it requires high-level programming [8].
equations of fluid flow, which is continuity equation, momen-
tum equation, and ennergy equation. The fluid flow problem Another application, MATLAB, offers all the functionality
itself is divided into various types such as; turbulent, laminar, for numerical methods such as BEM that involve matrix cal-
real, ideal, reversible, non-reversible, uniform, non-uniform, culations. In this case, the programming knowledge required
rotational and non-rotational [1]. in MATLAB is not too high compared to VBA. Additionally,
The fluid flow problem that will be discussed in this research MATLAB has an easy-to-use graphical viewer which makes
is a two-dimensional ideal fluid flow. An ideal fluid flow it very useful in the post-processing phase of BEM [8] .
is a fluid flow that has no viscosity and is incompressible Previously, Muhammad [10] had used the boundary element
(has zero pressure). Two-dimensional flow is a simplification method to solve fluid flow problems, one of which was for
that reduces the vector components in the three-dimensional ideal fluid flow. However, for this study using a flow through
momentum equation to two dimensions. Two-dimensional a two-dimensional plane. On the other hand, this research will
means the flow (and body shape) are identical in all parallel apply the boundary element method to the ideal fluid flow
planes [6]. problem that flows in a two-dimensional plane.
In general, fluid flow problems are difficult to be solved This research will discuss the implementation of BEM in
using analytical methods. To overcome this problem, a nu- solving fluid flow problems using the MATLAB application.
merical method is used to get an approximate solution. One The study in this research refers to [8] studies.
of the numerical methods commonly used to solve fluid flow
problems is the boundary element method [9]. II. P RELIMINARIES
S. Aurellia, M. Syafwan, and A. Rince Putri are with the Department A. The Gauss-Green Theorem
Mathematics and Data Science, Universitas Andalas, Padang, 25163 Indonesia
e-mail: (see [email protected]). The Gauss-Green theorem is a fundamental identity, which
Anati bte Ali is with Department of Mathematical Science, Universiti
Teknologi Malysia, UTM Johor Baru, 81310 Malaysia. relates the integral of the derivative of a function over a domain
Manuscript received ; revised . R to the integral of that function on its the boundary C.
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 3

The domain may be two or three-dimensional. For simplic- If g is another function of x and y, then equation (6) and (7)
ity of presentation, this relationship is derived for the two- result in
dimensional case. Consider the plane domain R bounded by
Z Z Z
∂f ∂g
the curve We shall work first with the derivative of a function g dR = − f dR + f gnx ds, (8)
R ∂x R ∂x C
f = f (x, y) with respect to x. The integral over R may be
written as a double integral, for which the integration is carried and Z Z Z
∂f ∂g
out first with respect to x and then with respect to y. Thus, g dR = − f dR + f gny ds. (9)
R ∂y R ∂y C
we can write
Z Z y2 Equation (8) and (9) state the integration by parts in two
∂f dimensions and are known as the Gauss-Green theorem [3].
dR = {f (x2 , y) − f (x1 , y)} dy, (1)
R ∂x y1

where B. Green’s Second Identity

x1 = x1 (y) and x2 = x2 (y). (2) Consider the functions u = u(x, y) and v = v(x, y) which
are twice continuously differentiable in R and once on C.
Applying equation (8) for g = v, f = ∂u ∂x and equation (9) for
g = v, f = ∂u∂y and adding the resulting equations, we arrive
at the following
 2  R  ∂u ∂v 
∂ u ∂2u ∂u ∂v
R
R
v ∂x 2 + ∂y 2 dR = − R ∂x ∂x
+ ∂y ∂y dR
Z  
∂u ∂u
+ v nx + ny ds. (10)
C ∂x ∂y
∂v
Similarly, applying equation (8) for g = u, f = partialx and
∂v
equation (9) for g = u, f = ∂y and adding the resulting
equations, we obtain
 2  R  ∂v 
∂ v ∂2v
dR = − R ∂u ∂u ∂v
R
R
u ∂x 2 + ∂y 2 ∂x ∂x + ∂y ∂y dR
Figure II-A: Integration over a plane domain R bounded by Z  
a curve C [7]. ∂v ∂v
+ u nx + ny ds. (11)
C ∂x ∂y
From the detail of figure II-A, we have
Subtracting equation (11) from equation (10) yields
dy Z Z 
∂u ∂v

= cos α = nx ⇒ dy = nx ds, (3) 2 2
(u∇ u − u∇ v)dR = v −u ds. (12)
ds ∂n ∂n
R C
dx Where ∇2 is known as the Laplace operator or harmonic
− = cos α = ny ⇒ dx = −ny ds. (4)
ds operator and it is defined as
where nx and ny are the components of the unit vector n, 
∂ ∂
 
∂ ∂

∂2 ∂2
2
which is normal to the boundary C. The negative sign in ∇ ≡ ∇·∇ = i +j · i +j = + ,
∂x ∂y ∂x ∂y ∂x2 ∂y 2
equation (4) is due to the fact that the dx and the sin α have (13)
opposite signs when the angle α is measured in the counter- while
clockwise sense with respect to the positive x-direction. Con- ∂

∂ ∂

∂ ∂
sequently, equation (1) becomes ≡ n · ∇ = (nx i + ny j) · i +j = nx + ny ,
R y2 ∂n ∂x ∂y ∂x ∂y
R
{f (x2 , y) − f (x1 , y)} dy = s2 f (x2 , y)nx ds− (14)
y1
Z
f (x1 , y)nx ds. (5) C. Laplace Equation
s1 One of the easiest ways to explain BEM is to study the
In the previous expression the integration on s1 is performed solution of the two-dimensional Laplace equation
in the negative direction (clockwise) when y varies from y1 ∂2u ∂2u
to y2 . Using uniform direction for the integration over s, both ∇2 u = + 2 = 0, for (x, y) ∈ R, (15)
∂x2 ∂y
terms in equation (5) can be combined in a single expression
Z Z with boundary conditions
∂f
dR = f nx ds. (6) u = f (x, y) for (x, y) ∈ Cα , (16)
R ∂x C
∂u
By substituting x with y in equation (6), we obtain and q= = g(x, y) for (x, y) ∈ Cβ , (17)
∂n
Z Z
∂f
dR = f ny ds. (7)
R ∂y C
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 4

Furthermore, it can be shown that

(
∂u∗
Z  
∗ ∂u 0 if (ξ, η) ∈/ R∪C
u −u ds = 1
Cε ∂n ∂n 2 u(ξ, η) if (ξ, η) lies in section smooth C
(22)
Can be written:
∂u∗
Z  
∂u
λu(ξ, η) = u − u∗ ds (23)
C ∂n ∂n
where

Fig. 1. Two-dimensional Laplace’s equation in the region R [8]. if (ξ, η) ∈
/ R∪C
0

1
λ= if (ξ, η) lies in section smooth C (24)
2
1 if (ξ, η) ∈ R

where R is the two-dimensional area in the plane xy
bounded by the closed curve C = Cα ∪ Cβ , as shown in Equation (23) is the limit integral equation for Laplace’s
Figure 1. equation. In this case the problem in the form of a partial dif-
The fundamental solution of the two-dimensional Laplace’s ferential equation with boundary conditions is converted into
equation is given by an integral equation problem which contains the fundamental
solution of Laplace’s equation. If the value on the right side
1
u∗ = ln (x − ξ)2 + (y − η)2 .


(18) of the equation (23) is known, then the value of u(ξ, η) can
4π be determined for each point (ξ, η) in R [8].
Equation (18) satisfies Laplace’s equation at every point in
the region R except at point P (ξ, η). However, if the region D. Fluid Flow Problem
R − Cε , where Cε is a small loop containing and centered at
An ideal fluid is one that has no viscosity and is incom-
P (ξ, η), then u∗ is defined and ∇2 u∗ = 0 in this region.
pressible (has zero pressure). One approach that is quite good
in solving the problem of non-viscous fluid flow is to solve
continuity equation, namely
∂ρ
+ ∇ · (ρv) = 0 (25)
∂t
where ρ is the density of the fluid, v(vx , vy ) is the velocity
of the fluid at the point (x, y). For an ideal fluid it is known
that the fluid is incompressible so that the density is constant,
then Equation (25) becomes
∇ · v = 0. (26)
Fig. 2. The two-dimensional Laplace equation in the region R with a domain
that does not contain a circle Cε centered on P (ξ, η) [8].

By using Green’s second identity for the region Ω = R−Cε

and the C + Cε curve, we get

∂u∗
Z Z  
2 ∗ ∗ 2 ∗ ∂u


u∇ u − u ∇ dΩ = u −u ds.
Ω C+Cε ∂n ∂n
(19)
It can be seen that the left-hand side of the above equation is
zero, so
∂u∗
Z  
∗ ∂u
u −u ds = 0, (20)
C+Cε ∂n ∂n Fig. 3. Two-dimensional fluid flow [7].

which can be rewritten as

Observe the two-dimensional fluid flow in Figure 3. If the
∂u∗ ∂u∗
Z   Z  
∗ ∂u ∗ ∂u flow is irrotational, then the vorticity is zero, obtaining
u −u ds = − u −u ds.
C ∂n ∂n Cε ∂n ∂n 
∂vy ∂vx

(21) ∂x − ∂y = 0.
It can be shown for (ξ, η) ∈ R,
Thus v is the slope of the velocity potential φ, namely
∗
limε→0+ Cε u ∂u
R
ds = u(ξ, η) and ∂φ ∂φ
R ∂n ∗ ∂u v = ∇φ = i+ j, (27)
limε→0+ Cε u ∂n ds = 0. ∂x ∂y
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 5

and its velocity component is E. Boundary Element Method

∂φ ∂φ The boundary element method can be obtained to find an

vx = dan vy = . (28) approximation of the solution to the boundary value problem,
∂x ∂y
one of which is the fluid flow problem. The initial step in
Next, substitute Equation (27) into Equation (26) we get constructing the BEM is to approximate the closed curve C
by a polygon with N straight line segments, which is
∇2 φ = 0. (29)
C ' C1 ∪ C2 ∪ C3 ∪ · · · ∪ CN ,
The solution of the differential equation (29) with boundary as illustrated in Figure 4.
conditions, which will be explained, determines the potential
function φ. Furthermore, the velocity components are obtained
from Equation (28).
The boundary conditions on the potential velocity may be
derived from physical considerations. For this purpose, review
immobilized and non-viscous flow in the pipe, as shown in
Figure 3.
In the absence of penetration through the walls AB and DC
of the pipe, the normal vn component of the velocity is zero.

∂φ
vn = = 0. (30) Fig. 4. Closed curve C that is approximated by boundary elements
∂n C1 , C2 , C3 , . . . , CN [8].
Along the intersection plane AD the distribution of vn is
written as,
These segments are constructed by selecting N points on
∂φ the boundary, say (x1 , y1 ), (x2 , y2 ), . . . , (xN , yN ) and defin-
vn = = vn , (31)
∂n ing (xN +1 , yN +1 ) = (x1 , y1 ). Then, Ck is a straight line
while in the BC intersection the conditions are: connecting point (xk , yk ) to point (xk+1 , yk+1 ). The straight
line segments Ck for k = 1, 2, . . . , N are boundary elements.
1) The velocity components of vn can be formulated, but For simplification of the problem, it is assumed that the
the distribution of Equation (30) and (31) must obey the values of u and ∂u/∂n are constant with respect to each
law of conservation of mass in the given domain, which boundary element. More specifically, it takes the value at the
is midpoint of each element to be the value across elements.
Z Z Z Z
vn ds = − ds + vn ds = − f dΩ. (32) ∂u
u ≈ uk and ≈ q k for (x, y) ∈ Ck , k = 1, 2, · · · , N,
Γ vn BC Ω ∂n
(36)
2) Steady flow conditions (fully developed laminar flow) ∂u
where uk and q k are the values of u and ∂n at midpoint
are applicable, provided the cross section BC is located (xk , yk ) of boundary element Ck .
far enough from the region of strong variation in the Based on the explanation above, the integral on the right
velocity field. This condition is stated as side of Equation (23) can be approximated as follows:
N
∂vn ∂2φ X
= = 0. (33) λu(ξ, η) = (uk Gk (ξ, η) − q k Fk (ξ, η)) , (37)
∂n ∂n2
k=1
If the cross section BC is placed away from the region of where
strong velocity variation, as was the case in case 2, and shaped Z
such that the velocity is normal at every point, the tangential Fk (ξ, η) = u∗ (ξ, η)ds, (38)
component of the velocity disappears. Therefore, Ck
∂u∗ (ξ, η)
Z
∂φ Gk (ξ, η) = ds. (39)
vn = = 0, (34) Ck ∂n
∂s
If the values of uk and q k are known, then Equation (37)
which result in can be used to determine the value of u(ξ, η) for (ξ, η) ∈ R.
φ = C, (35) However, for a more realistic boundary value problem only
one of uk and q k is known at any given boundary.
where C is an arbitrary constant number. This condition allows Equation (37) can be used to approximate unknown bound-
the formation of velocity potential φ for arbitrary constant ary values, before using it again to approximate values at
approximations. However, it does not affect the velocity field, interior points (points that are in the R domain). To do
which is determined by Equation (28) [7]. so, (ξ, η) is made the midpoint of the elements Ck for
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 6

k = 1, 2, · · · , N , so we get N equations. Since (ξ, η) is in an incompressible and non-viscous fluid through a tube. The
Ck , substituting the value λ = 1/2 into Equation (37), we get velocity potential in the tube will be determined. The shape
N of the curvilinear segment is determined by the following
1
(uk Gk (xm , y m ) − q k Fk (xm , y m )) for m = 1, 2, · · ·equations [7]:
X
um = , N,
2
k=1
Segment BC: y = 2x3 − 3x2 − 1,
0 0
(40) Segment B C : y = −2x3 + 3x2 + 1.
∗ 2 2
where u = (1/4π)ln((x − xm ) + (y − y m ) ) and (xm , y m ) The assumed boundary conditions are:
∂φ
is the midpoint of the element Cm . at the inlet cross-section AA0 : ∂n = vn = −1,
To determine the unknown boundary value, Equation (40) at the outlet cross-section DD : φ = 0. 0
∂φ
can be changed to Az = b. The unknown boundary values are at the rigid walls ABCD and A0 B 0 C 0 D0 : ∂n = vn = 0.
u or q and are elements in z with corresponding coefficients with a rectangle with the size of 2 × 1 between C 0 CD0 D seg-
∂φ
in the matrix A. Meanwhile, the b matrix consists of known ment. This rectangle has boundary condition of ∂n = vn = 0
boundary values. What needs to be done is to construct the for the vertical line and φ = 0 for the horizontal line.
matrix A and the matrix column vector b according to the type
of boundary value in each element z. In other words, for every
m = 1, 2, . . . , N , it needs to be constructed
N
X N
X
Amk Zk = bmk (41)
k=1 k=1

The matrices A and b are constructed systematically by

considering two possibilities, namely (a) when the value of
u is known, and (b) when the value of q = ∂u/∂n is known
at the limit of Ck .
For each m (i.e. each row in the matrix), it is necessary to
determine the entry Am k in the matrix A and the entry bm k
in the vector b in each boundary element of Ck . If, in the Fig. 5. The tube segment of the example [7].
boundary element Ck the value of u is known, then the value
of q is unknown and the element in the A matrix is Am k = Based on the explanation above and Figure 5 it can be
−Fk (xm , y m ). The corresponding entries in the matrix b are concluded that the flow discussed is an ideal fluid flow that
−uk Gk (xm , y m ) if k 6= m or −uk Gk (xm , y m ) + 1/2um if flows in a tubular two-dimensional plane. So that this fluid flow
k = m. problem can be described using Equation (29) to determine the
Conversely, if the value of q is known in the boundary value of the potential velocity in the tube.
element Ck , then u becomes a variable, so the entries in b In the previous sub-chapter, it has been explained that
are q k Fk (xm , y m ). Elements in A are Gk (xm , y m ) if k 6= m, using the MATLAB program consists of three processes.
or Gk (xm , y m ) − 1/2 if k = m. The first process, namely Pre-processing, is to write data
It can be concluded that, the entries for matrix A and vector about the boundary elements. This data is written in four
b are constructed as follows: columns namely, the coordinates x and y, then the type of
boundary conditions, and the last column contains the value

Fk (xm , y m )
 if u given over Ck
of the boundary conditions for each element. For this type
Amk = Gk (xm , y m ) if q given over Ck and k 6= m
of boundary conditions there are two, namely the Dirichlet
Gk (xm , y m ) − 12 if q given over Ck and k = m boundary condition (the value of u is known) is declared with


 (42) a value of 0, while the Neumann boundary condition (the value
 k
 q F k (xm , y m ) if u given over C k of q is known) is declared with a value of 1. The node points of
bm k = −uk Gk (xm , y m ) if q given over Ck and k 6= m the boundary elements are determined with the same distance
−uk Gk (xm , y m ) + 21 if q given over Ck and k = m.along 0.25 so as to produce 64 element node points.



(43) Based on the input data, a numerical solution will be found
When A and vector b have been formed, then z can be using the boundary elements using the Processing process.
solved, remembering that zk = q k if u is assigned to Ck , and Next, the surface plot will be obtained using the process
zk = uk if q is assigned to Ck . When all boundary values are Post-processing. The color at a point in the interior R in the
known, then Equation (37) can be used to find the value at image represents the value of u at that point, where the color
each point in the domain R [8]. degradation is between -1 and 0.4.

III. R ESULT AND D ISCUSSION

In this sub-chapter an example will be presented to illustrate
the implementation of the boundary element method in fluid
flow problems using the MATLAB application. Consider the
steady, laminar flow (streams moving in the same path) of
JOURNAL OF LATEX CLASS FILES, VOL. , NO. , 7

From Figure 6 above it can be seen that the potential

velocity of the fluid flow from the example is in the range
-0.1 to 0.4. The fluid flow in the AA0 segment has a velocity
potential of -1, the closer the fluid is to the BC and B 0 C 0
segments, the more the velocity potential value approaches 0.
And in the BC and B 0 C 0 segments to the DD0 segment, the
fluid velocity potential has a value of around 0. The AA0 BB 0
segment has a smaller value compared to the BB 0 C 0 D0 DC
segment because the area of AA0 BB 0 is smaller than the
BB 0 C 0 D0 DC area.

IV. C ONCLUSION
This research has discussed the implementation of the
boundary element method (BEM) in fluid flow problems.
Fig. 6. Surface plot of φ velocity potential.
The fluid flow problem studied in this final project is a
two-dimensional ideal fluid flow, that is, a fluid that has no
viscosity (thickness) and is incompressible (has zero pressure),
which is modeled as
∇2 φ = 0 (44)
BEM is a method of solving partial differential equations
by dividing the domain of the equation into a number of finite
line segments, then these segments are used to evaluate the
limit integral equation.
In this final project, the boundary element method has been
implemented to determine the potential velocity of fluid flow
which has been described previously using the MATLAB
application. And obtained numerical results from the potential
velocity.

ACKNOWLEDGMENT
The authors would like to thank Universitas Andalas for
Fig. 7. Surface plot of velocity vx . supporting this research and Universiti Teknologi Malaysia for
the hospitality and support while Salsabila Aurellia do her
credit earning program there.

R EFERENCES
[1] Anderson, J.D. 1992. Computational Fluid Dynamics. Springer, Berlin,
Heidelberg.
[2] Azis, M.I. 2019. Metode Elemen Batas: Untuk Media Anisotropik Ho-
mogen. Brilian Internasional, Sidoarjo.
[3] Becker, A.A. 1992. The Boundary Element Method in Engineering: A
complete course. McGraw-Hill Companies, Nottingham.
[4] Boyce, W.E. dan R.C. DiPrima. 2009. Elementary Differential Equations
and Boundary Value Problems. Ninth Edition. John Wiley and Sons, Inc,
New York.
[5] Davies, A. dan D. Crann. 1998. The boundary element method on a
spreadsheet. Int. J. Math. Edu. Sci. and Technol. 29(6): 851-865.
[6] Flandro, G.A, H.M.McMahon dan R.L.Roach. 2012. Basic Aerodynamics.
Cambridge University.
[7] Katsikadelis, J.T. 2002. Boundary Element: Theory and Applications.
Elsevier Science, Oxford.
[8] Keng C.A. 2008. Introducing the boundary element method with MAT-
LAB. Int. J. Math. Edu. Sci. and Technol. 39(4): 505-519.
Fig. 8. Surface plot of velocity vy . [9] Manaqib, M. 2017. Boundary element method untuk menyelesaikan
masalah syarat batas persamaan Laplace dimensi dua. Jurnal Logika. 7(2):
122-136.
Based on Figures 7 and 8 it can be seen that the velocity vx [10] Muhammad, G. 2011. Boundary Element Methods for Incompressible
Fluid Flow Problems. Tesis S-3, tidak diterbitkan. Departement of Math-
of the inlet cross section AA0 decreases towards the right. And ematics, University of Engineering and Technology, Pakistan.
in Figure 8 the speed vy in the AA0 BB 0 region of the A0 B 0
segment is getting smaller and smaller for the BCDD0 C 0 B 0
region, almost all regions have the same speed .