Finite Volume Method on Twodimensional

Anh Ha LE March 20, 2012

Introduction

We will use the nite volume method to discretize the following equation with = (0, 1) (0, 1) and f L2 (): u = f (x, y) u(x, y) = g(x, y) in on . (1.1)

2
2.1

Rectangular mesh
Rectangular mesh
2 2

On interval [0, 1], we make two partition (xi+ 1 )i0,N1 , (yj+ 1 )j0,N2 such that 0 = x 1 < x 3 < < xN1 1 < xN1 + 1 = 1,
2 2 2 2 2 2

0 = y 1 < y 3 < < yN2 1 < xN2 + 1 = 1.

2 2

Choosing the sequences (xi )i0,N1 +1 and (yj )j0,N2 +1 such that x0 x 1 , x i =
2

xi 1 + xi+ 1
2 2

y0 y 1 , y j =
2

2 yj 1 + yj+ 1
2 2

, xN1 +1 xN1 + 1 ,
2

, jN2 +1 yN2 + 1 .
2

Let T = (Kij )i1,N1 ,j1,N2 be a admissible mesh of (0, 1) (0, 1) such that Kij = [xi 1 , xi+ 1 ] [yj 1 , yj+ 1 ]
2 2 2 2

and the point (xi , yj ) is the control point of control volume Ki,j for all i 1, N1 , j 1, N2 (see Fig. 1). Let hi = |xi+ 1 xi 1 |,
2 2

kj = |yj+ 1 yj 1 |
2 2

for all i 1, N1 and j 1, N2 for all i 0, N1 and j 0, N2 .

and hi+ 1 = |xi+1 xi |,

2

kj+ 1 = |yj+1 yj |
2

Then, the area of control volume |Kij | = hi kj . Let h = max{hi , kj } be the mesh size. 1

yN2 + 1 yN2 +1
2

yN2 yN 1
2

y4 y7
2

y3 y5
2

y2 y3
2

(x4 , y2 ) K42

y1 y0 y 1 2 x0 x 1
2

x1 x 3 x2 x 5 x3 x 7 x4xN 1xN
2 2 2 2

xN1 + 1 xN1 +1
2

Figure 1: Rectangular mesh

2.2

The scheme

As in the 1D case, the nite volume scheme is found by integrating the rst equation of (1.1) over each control volume Ki,j , which yields 1 |Kij | 1 |Kij |
yj+ 1
2

yj 1
2 2

1 ux (xi+ 1 , y)dy + 2 |Kij | uy (x, yj+ 1 )dx +

2

yj+ 1
2

ux (xi 1 , y)dy
2

yj 1

2 2

xi+ 1 xi 1

1 |Kij |

xi+ 1 xi 1
2

uy (x, yj 1 )dy =
2

1 |Kij |

f (x, y)dxdy
Kij

(2.1)

These terms is approximated following: 1 |Kij | 1 |Kij | 1 |Kij | 1 |Kij |

yj+ 1
2

ux (xi+ 1 , y)dy =
2

kj ux (xi+ 1 , yj )
1

yj 1

|Kij | kj ux (xi 1 , yj )
1

ui+1,j ui,j , hi hi+ 1

2

2 2

yj+ 1

ux (xi 1 , y)dy =
2

yj 1

|Kij | hi uy (xi , y |Kij | hi uy (xi , yj 1 )

2

ui,j ui1,j , hi hi 1
2

2 2

xi+ 1

uy (x, yj+ 1 )dy =

2

j+ 1 2

) =

xi 1
2 2

ui,j+1 ui,j , kj kj+ 1

2

xi+ 1

uy (x, yj 1 )dy =
2

xi 1
2

|Kij |

ui,j ui,j1 . kj kj 1
2

The approximate equation (2.1) becomes Fi+ 1 ,j Fi 1 ,j + Fi,j+ 1 Fi,j 1 = fij ,

1 2 2 2

(2.2)

where Fi+ 1 ,j =
2

ui+1,j ui,j hi hi+ 1

2

for all i 0, N1 , j 1, N2 , for all i 1, N1 , j 0, N2 ,

Fi,j+ 1 =
2

ui,j+1 ui,j kj kj+ 1

2

and ui,j 1 = kj 1 = hi

j+ 1 2

g(xi , y)dy
yj 1
2 1 i+ 2

for i {0, N1 + 1}, j 1, N2 ,

ui,j

g(x, yj )dx for i 1, N1 , j {0, N2 + 1}.

xi 1

References
[1] R. Eymard, T. Gallout, and R. Herbin, Finite Voulume Methods. In: Ciarlet, Ph.G. and e Lions, J.-L., Eds. Handbook of Numerical Analysis, Vol. VII, Amsterdam: North-holland, pp. 1731020 (2000).