0% found this document useful (0 votes)
90 views8 pages

Numerical Solutions of Integral Equations

This document discusses numerical solutions of integral equations using the finite volume method. It begins by introducing integral and partial differential equations and conservation laws, describing how they can be formulated in both integral and differential form. It then discusses the finite volume method, how it applies conservation laws to finite control volumes. Finally, it covers initial and boundary conditions that must be specified to solve the conservation equations.

Uploaded by

garkin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
90 views8 pages

Numerical Solutions of Integral Equations

This document discusses numerical solutions of integral equations using the finite volume method. It begins by introducing integral and partial differential equations and conservation laws, describing how they can be formulated in both integral and differential form. It then discusses the finite volume method, how it applies conservation laws to finite control volumes. Finally, it covers initial and boundary conditions that must be specified to solve the conservation equations.

Uploaded by

garkin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

CHAPTER 3

NUMERICAL SOLUTIONS OF INTEGRAL EQUATIONS


• Integral and partial differential equations
• Conservation laws
• Integral form
• Differential form
• Conservation of mass
• Heat conduction equation
• Initial and boundary conditions
• Finite Volume method.
• Convective and diffusive fluxes
• Explicit Euler time integration

İ.H. TUNCER & Y. ÖZYÖRÜK 61 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Integral and Partial Differential Equations

• The conservation laws, conservation of mass, momentum and energy, are formulated in
integral equations.
• Each equation employs a certain physical quantity as its dependent variable being mass,
momentum, energy per unit volume, and implies that there must be a balance among the
factors that influence the distribution of the variable in space and time.
• Dependent variables are, in general, a function of three space coordinates, and time. Thus

φ = φ(t, x, y , z)

where x, y , z and t are the independent variables.


• The dependent variables are expressed as; ρ ≡ mass per unit volume,
~ ≡ momentum per unit volume and ρE ≡ total energy per unit volume.
ρV

İ.H. TUNCER & Y. ÖZYÖRÜK 62 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Integral and Partial Differential Eqns (cont’)

• Integral equations are formed by applying a • Differential equations are formed by


conservation law in a finite control-volume applying the conservation laws in a
bounded by its control surface. infinitesimally small control-volume.

İ.H. TUNCER & Y. ÖZYÖRÜK 63 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Conservation Laws

• Let’s consider a scalar quantity per unit volume, q, in a control volume (open system) Ω,
non-deforming and fixed in space, bounded by a closed surface S.
• The general form of a conservation law states that
• the rate of change of the quantity q within the control volume Ω
Z

q dΩ
∂t Ω

~ , is the amount of q passing


• PLUS the net contribution from the outgoing q flux , (Flux, F
through a unit surface area per unit time) through the boundary surface, S,s
I
~ · d~
F S
S

where the differential surface vector d ~


S points outward normal to the control surface.

• is equal to the contributions from the sources of the quantity q in the control volume Ω.

İ.H. TUNCER & Y. ÖZYÖRÜK 64 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Conservation Laws (cont’d)

• In the absence of sources, the integral conservation equation , which is the basis for the
Finite Volume (FV) method, becomes:
Z I
∂ ~ · d ~S = 0
q dΩ + F
∂t Ω S

• Note that with the application of the divergence theorem, the surface integral can be
converted into a volume integral:
Z Z
∂ ~ ·F~ dΩ = 0
q dΩ + ∇
∂t Ω Ω

• For a fixed control volume the terms can be combined as


Z
∂q ~ ~
( + ∇ · F ) dΩ = 0
Ω ∂t

İ.H. TUNCER & Y. ÖZYÖRÜK 65 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Conservation Laws (cont’d)


• For an infinitesimally small control volume, the fact that as the volume goes to zero, the
integrand should goto zero faster leads to the differential form of the equation :

∂q ~ =0
~ ·F
+ ∇
∂t
which may be solved numericall by the Finite Difference method.
• 2-D CONSERVATION OF MASS
The differential equation for the conservation of mass is given by
∂ρ ~ ~) = 0
+ ∇ · (ρV
∂t
∂ρ ~ ~
+∇·F = 0
∂t
where the flux vector is the conserved quantity multiplied by the flow velocity:
~ = ρV
F ~ = (ρu)~i + (ρv )~j

It is called a convective flux


İ.H. TUNCER & Y. ÖZYÖRÜK 66 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Conservation Laws (cont’d)
• 2-D HEAT CONDUCTION EQUATION
The differential heat conduction equation, which is a simple form of the conservation of
energy equation and also known as the two-dimensional diffusion equation, is given with a
constant diffusivity α by

∂T ∂2T ∂2T
− α[ 2 + ] = 0
∂t ∂x ∂y 2
∂T ∂ ∂T ∂ ∂T
− [ (α )+ (α )] = 0
∂t ∂x ∂x ∂y ∂y
∂T ~ · (α ∂T ~i + α ∂T ~j) = 0
−∇
∂t ∂x ∂y
where the flux vector is

~ = −(α ∂T ~i + α ∂T ~j) = −α∇T


F ~
∂x ∂y

which is a function of the gradient of the conserved quantity, and called a diffusive flux

İ.H. TUNCER & Y. ÖZYÖRÜK 67 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Initial and Boundary Conditions


• The solution of the conservation equations requires a set of conditions to be specified at
the boundaries of the control volume. They are classified as initial conditions and
boundary conditions.
• A boundary condition is the requirement that the magnitude of dependent variable and/or
its flux must satisfy on the boundaries. Various types of BCs may be given:
• Dirichlet BC: The dependent variable is prescribed along the boundary (surface, s).

T s = Ts

• Neumann BC: The normal gradient of the dependent variable, which is related to the flux,
is prescribed along the boundary.


~ = −k ∂T = qs00
−k~n · ∇T s ∂n s
• Robin BC: A linear combination of Dirichlet and Neumann BC is specified.


~ = −k ∂T = h(T − Ta )
−k~n · ∇T s ∂n s
s

• Mixed BC: A certain portion of the boundary is Dirichlet type, another portion is of
Neumann type.
İ.H. TUNCER & Y. ÖZYÖRÜK 68 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Finite Volume (FV) Method


• In Finite Volume method, an integral conservation law
Z I n2
∂ ~ · d ~S = 0 B
q dΩ + F C
∂t Ωe Se n
n3 n1
D

is applied to each finite control volume (cell or element (e)) A


n4
Ωe and its boundaries Se , ABCD in the discretized solution
domain.
e : elem.
s : elem. surface
• A simple discrete form of the conservation equation n : elem. neighbor

applied to a 2D triangular cell e is then given by B


n2

3 n1
d(qe Ωe ) X ~ ~
s2
e s1
+ (F · S)e,s = 0 C s3
dt s=1
A
n3

where qe is the average q over element e, and the flux


terms are summed over its three sides AB, BC , CA.
Here Ωe is the area of the triangle ABC .

İ.H. TUNCER & Y. ÖZYÖRÜK 69 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Finite Volume (FV) Method (cont’d)
• If the flux and the surface area vectors are defined as B
C
~
F = f ~i + g~j
~S f dy
= ∆y ~i − ∆x ~j
= ∆yAB ~i − ∆xAB ~j g dx
S
= (yB − yA ) ~i − (xB − xA ) ~j A dx
• Here ~S points outward normal to the cell surface. Note that |~S| = S = |AB|.
~ ~S is
~
obtained by rotating the AB vector 90 degrees clockwise. ∆x and ∆y are evaluated as
the boundary edges are traversed in the counter-clockwise direction.
• Then, the total flux for a triangular element e may be expressed as follows:
X 3
X
~ · ~S)e,s =
(F (fe,s ∆ye,s − ge,s ∆xe,s )
sides,s s=1

• Flux evaluations along the surfaces will be discussed later.


İ.H. TUNCER & Y. ÖZYÖRÜK 70 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Finite Volume (FV) Method (cont’d)


• The FV equation for element (cell) e then becomes
3
dqe X
Ωe + [fe,s ∆ye,s − ge,s ∆xe,s ] = 0
dt s=1

• The time derivative is approximated by the first order TSE in time;


dqe dqe qe (t + ∆t) − qe (t)
qe (t + ∆t) = qe (t) + ∆t + O(∆t 2 ) −→ =
dt dt ∆t
k
• If the time level (t) is denoted by the superscript where qe (t) ≡ qek is known:
3
qek+1 − qek X k k
Ωe + [fe,s ∆ye,s − ge,s ∆xe,s ] = 0
∆t s=1

• Such a formulation leads to the explicit Euler time integration :


3
∆t X k
qek+1 = qek − k
[f ∆ye,s − ge,s ∆xe,s ]
Ωe s=1 e,s
İ.H. TUNCER & Y. ÖZYÖRÜK 71 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Finite Volume (FV) Method (cont’d)

• The conserved quantity/variable qe is taken as an average value of the variable over the
cell volume. It is usually attached to the cell centroid.
• Finite volume method can, in general, be applied to any cell structure. Unstructured grids
with triangular/tetrahedral cells offer greater flexibility for discretizing complex domains.
• Cells should cover the whole solution domain, Ω.
• Cells should not overlap.
• The coordinates of the cell nodes, A,B,C, are needed to determine the cell volume (area in
2D) and the surface areas(lengths in 2D).
• The total flux through a cell edge contributes to the neighboring cell with the opposite
sign.
• The flux vector on a cell edge should be evaluated exactly the same for both cells.
• Neumann type boundary conditions are applied naturally. For example solid walls..

İ.H. TUNCER & Y. ÖZYÖRÜK 72 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Evaluation of Volumes/Areas
• For a triangle ABC, the area, Ω, can be evaluated from the e : elem.
s : elem. surface
vector products of two edge vectors: n : elem. neighbor

B
1 ~ ~ n2
Ω = AB × AC n1
2 s2
e s1
1 C s3
= [(xB − xA )(yC − yA ) − (yB − yA )(xC − xA )] A
2 n3

1
= (∆xAB ∆yAC − ∆yAB ∆xAC )
2

• For a general quadrilateral ABCD, the area, Ω, can be


n2
evaluated from the vector products of the the diagonals: C
B


1 ~ ~ n3
n
n1
Ω = AC × BD D
2
A
1 n4
= [(xC − xA )(yD − yB ) − (yC − yA )(xD − xB )]
2
1
= (∆xAC ∆yBD − ∆yAC ∆xBD )
2
İ.H. TUNCER & Y. ÖZYÖRÜK 73 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Flux Evaluation Methods

There are 2 fundamental approaches to compute fluxes:

1 Averaging of cell-based quantities – straightforward


• Averaging of fluxes of the cells sharing the interface −→
simplest !
• Averaging the cell variables sharing the interface, and
then computing the fluxes −→ simple !

2 Averaging of node-based quantities – some overhead


• Averaging node-based fluxes −→ requires transferring
solution to nodes, and computing fluxes there first –
expensive, but better connectivity
• Averaging node-based solutions, and then computing
fluxes −→ requires transferring solutions to nodes first –
expensive, better connectivity

İ.H. TUNCER & Y. ÖZYÖRÜK 74 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Cell-based Flux Evaluation


• The simplest cell-based flux evaluation approach for surface AB takes and averages the
cell-based flux values of the cells sharing this surface (face).

• As shown in the Figure, the interface AB is shared


by discrete elements e3 and e7. Hence,

1 1
fAB = [f (qe7 ) + f (qe3 )] ≡ (fe7 + fe3 ),
2 2
1 1
gAB = [g (qe7 ) + g (qe3 )] ≡ (ge7 + ge3 )
2 2

• Similarly,
1 1
fBC = (fe7 + fe9 ), fCA = (fe7 + fe1 )
2 2
1 1
gBC = (ge7 + ge9 ), gCA = (ge7 + ge1 )
2 2
İ.H. TUNCER & Y. ÖZYÖRÜK 75 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Explicit Time Integration with Cell-Based Flux Evaluation

The Euler explicit time integration now becomes:


3
∆t X k
qek+1 = qek − [f ∆ye,s − ge,sk
∆xe,s ]
Ωe s=1 e,s e : elem.
 s : elem. surface
k ∆t 1 k k 1 n : elem. neighbor
= qe − (f + fn1 )∆ye,1 − (gek + gn1
k
)∆xe,1
Ωe 2 e 2 n2
B

1 1 n1
+ (fek + fn2k
)∆ye,2 − (gek + gn2
k
)∆xe,2 s2
e s1
2 2  s3
C A
1 k k 1 k k n3
+ (fe + fn3 )∆ye,3 − (ge + gn3 )∆xe,3
2 2
where
∆xe,1 = xB − xA , ∆xe,2 = xC − xB , ∆xe,3 = xA − xC ,
∆ye,1 = yB − yA , ∆ye,2 = yC − yB , ∆ye,3 = yA − yC ,
1 ~ 
~ | = 1 (xB − xA )(yC − yA ) − (xC − xA )(yB − yA )]
Ωe = |AB × AC
2 2

İ.H. TUNCER & Y. ÖZYÖRÜK 76 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Evaluation of Gradient Vector


• In the case of diffusive fluxes, where the flux vector involves the gradient vector of
variables, namely, first-order space derivatives of the variables, it becomes necessary
to Rdefine average values for the space derivatives of variables for a cell, that is;
1 ∂T
Ω Ω ∂x dΩ.
• A general procedure can be derived by the application of the divergence theorem. For a
control volume Ω and its closed surface S the divergence theorem states that:
Z I
~
∇TdΩ = Td ~S
Ω S

where in 2D d ~S = dy ~i − dx ~j
and Z I
∂T ~ ∂T ~
( i+ j)dΩ = Tdy ~i − Tdx ~j
Ω ∂x ∂y S

• The average derivatives may then be defined as


Z I Z I
∂T 1 ∂T 1 ∂T 1 ∂T 1
= dΩ = Tdy , = dΩ = − Tdx
∂x Ω Ω ∂x Ω S ∂y Ω Ω ∂y Ω S
İ.H. TUNCER & Y. ÖZYÖRÜK 77 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Evaluation of Gradient Vector (cont’d)


• For a triangular Finite Volume cell:
3
∂T 1 X
= Te,s ∆ye,s
∂x e Ωe s=1
e : elem.
3 s : elem. surface
1 X1 n : elem. neighbor
= (Te + Tns ) ∆ye,s
Ωe s=1 2
B
n2
Similarly
n1
3 s2
∂T 1 X e s1
= − Te,s ∆xe,s
∂y e Ωe s=1 C s3
A
3 n3
1 X1
= − (Te + Tns ) ∆xe,s
Ωe s=1 2

İ.H. TUNCER & Y. ÖZYÖRÜK 78 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Evaluation of Gradient Vector (cont’d)

~ for the marked cell


Example: Compute ∇φ
Element id : e = 6, no. of vertices = 3 (triangular)

A/v5 B/v7 C/v2


Cell vertices : 5, 7, 2 ---> [ (4.0,1.0), (3.0,2.0), (1.0,1.0) ]
x y x y x y

s1 s2 s3 n1 n2 n3
Cell Faces (sides) : 5-->7, 7-->2, 2-->5; and neighbor elements : 9, 10, 8

Cell area(6) = 0.5*( (x(B)-x(A))*(y(C)-y(A)) - (y(B)-y(A))*(x(C)-x(A)) )


= 0.5*( (3.0-4.0)*(1.0-1.0) - (2.0-1.0)*(1.0-4.0) )
= 0.5*( 0.0 + 3.0 ) = 1.5

dphi/dx(6) = (1/area(6))*( ( 0.5*( phi(6) + phi(9) ) * ( y(7) - y(5) ) +


0.5*( phi(6) + phi(10)) * ( y(2) - y(7) ) +
0.5*( phi(6) + phi(8) ) * ( y(5) - y(2) ) )

= (1/1.5) *( 0.5*( 20 + 40 ) * ( 2.0 - 1.0 )


+0.5*( 20 + 30 ) * ( 1.0 - 2.0 )
+0.5*( 20 + 10 ) * ( 1.0 - 1.0 ) ) = 3.33

dphi/dy(6) = -(1/area(6))* *( 0.5*( phi(6) + phi(9) ) * ( x(7) - x(5) ) +


+0.5*( phi(6) + phi(10)) * ( x(2) - x(7) ) +
+0.5*( phi(6) + phi(8) ) * ( x(5) - x(2) ) )

= -(1/1.5) *( 0.5*( 20 + 40 ) * ( 3.0 - 4.0 )


+0.5*( 20 + 30 ) * ( 1.0 - 3.0 )
+0.5*( 20 + 10 ) * ( 4.0 - 1.0 ) ) = 23.33

İ.H. TUNCER & Y. ÖZYÖRÜK 79 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Solution of Potential Flow Equation w/ FV Method


• The 2-D potential flow equation is derived for incompressible, irrotational flow as follows
∇ ~
~ ·V = 0 ← conservation of mass
~ ×V
∇ ~ = 0 ← irrotationality condition
since
~ × ∇φ
∇ ~ ≡0 ∴ if ~ = u~i + v~j = ∇φ
V ~
then irrotationality condition is satisfied trivially and the conservation of mass eqn.
becomes the Laplace’s equation:
~ · ∇φ
∇ ~ ≡ ∇2 φ = 0

• With the introduction of a pseudo time derivative and a diffusion constant Laplace’s
equation may be cast as a diffusion equation:
∂φ ~ · ∇φ
~ = 0 where F ~ = −ν ∇φ
~
− ν∇
∂t
• The integral form of the equation then becomes
Z I
∂ ~ · d ~S = 0, ~ = f ~i + g~j = −(ν ∂φ~i + ν ∂φ~j)
φ dΩ + F where F
∂t Ω S ∂x ∂y
İ.H. TUNCER & Y. ÖZYÖRÜK 80 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Solution of Potential Flow Equation w/ FV Method (cont’d)

k
• The discrete form of the equation for a triangular cell e at time level becomes :
3
∂φ k X ~ k ~
Ωn + (F · Se,s ) = 0
∂t e s=1 e,s
3
φk+1
e − φke 1 X ~k ~
= − (F · Se,s )
∆t Ωe s=1 e,s
3
∆t X ~ k ~
φk+1
e = φke − (F · Se,s )
Ωe s=1 e,s
3
∆t X 1 k 1 
φk+1
e = φke − k
(fe + fe,ns ) ∆ye,s − (gek + ge,ns
k
) ∆xe,s
Ωe s=1 2 2

İ.H. TUNCER & Y. ÖZYÖRÜK 81 AE305 - Numerical Methods in Aerospace Engineering @ 20201
Boundary Conditions for Potential Flow Equation
Boundary conditions are specified at farfield and wall boundaries.

• Farfield BC: Velocity is set to free stream value, V∞

~
∇φ ~∞
= V
∂φ
= u∞
∂x
∂φ
= v∞
∂y

• Wall BC: No flow through the wall dictates that the normal component of the velocity at
the wall boundary is zero;

~ · ~S
V = 0
~ · ~S
∇φ = 0 which means ~ w = 0.
F

• Note that Neumann type boundary conditions are specified.

İ.H. TUNCER & Y. ÖZYÖRÜK 82 AE305 - Numerical Methods in Aerospace Engineering @ 20201

Grid Data for Finite Volume Solutions

A FV grid should provide the following information :


List of node coordinates :
node index x y
1 0.25000 1.49050
2 1.88482 1.42583
... ... ...

List of node and neighbor connectivity for cells :


Cell index node1 node2 node3 neigh1 neigh1 neigh1
1 11 1 2 -1 9 6
2 2 9 6 7 8 6
3 3 ... ... ... ... ...
... ... ... ... ...

where -1 indicates Boundary 1

İ.H. TUNCER & Y. ÖZYÖRÜK 83 AE305 - Numerical Methods in Aerospace Engineering @ 20201

You might also like