Vortex Lattice Method for Hydrofoils

Gerrie Thiart, Mechanical Engineering,

University of Stellenbosch
Problem Group:
Eyaya Eneyew, AIMS, Cape Town
Alistair Fitt, University of Southampton, UK
Neville Fowkes, UWA, Perth, Australia
David Mason, WITS, Johannesburg
Eric Newby, WITS, Johannesburg
Prosper Ngabonziza AIMS, Cape Town
15th January 2010
Basic Problem
How to deal with difculties in implementing vortex lattice
method

Validity of model

What happens as the hydrofoil moves near to the water

surface

How to deal with some nearly-divergent integrals that occur

How to deal with some highly oscillatory integrals that may

occur
NOTE: we do not know the details of how Gerrie currently deals
with all these difculties - so if you already know all this - sorry!
Description of Hydrofoil

Hydrofoil is essentially a wing placed underneath a boat

The hydrofoil lifts the boat out of the water

The drag is reduced (no hull drag), allowing the boat to

travel faster
Vortex Lattice Method

Historically, one of the rst CFD methods

Pioneered as long ago as late 1940s at ARC (Aeronautical

Research Council - founded 1909, UK)

A completely INVISCID IRROTATIONAL

INCOMPRESSIBLE method

- but in spite of this does allow the prediction of lift and drag

Based on Prandtl lifting line theory

Often used in the early rough stages of aircraft design

Vortex Lattice Method (VLM) - General Theory

Assume irrotational ow of an incompressible inviscid uid

so that uid velocity q = where is velocity potential

We need to solve the Laplace equation

2
= 0 with
suitable BCs
VLM essentially relies on the following observation:
For ow past a body of length L with free stream velocity U

e
x
,
we have
= Ux +
_
L
0
f (t ) log
_
(x t )
2
+y
2
_
dt +
_
L
0
g(t ) arctan
_
y t
x
_
dt
where
f (t ): distribution of sources providing displacement but not lift
g(t ): distribution of vortices providing lift but not displacement.
VLM - Basic
Basic idea of VLM is to use vortices only (zero-width hydrofoil) -
discretize wing into a sequence of panels, each panel
associated with a vortex whose strength
ij
is to be determined
VLM - general 3-D wings with dihedral and camber
Coordinates for 3-D hydrofoil with dihedral and camber
VLM - Theory and Boundary Conditions
When vortex strengths
ij
are known, lift and drag follow easily
Trailing vortex contributions cancel since F = Ux
To set up equations to determine
ij
:

Must impose boundary condition of no ow normal to

hydrofoil

Here we have a free surface (water surface) z = (x, y, t )

Need to satisfy boundary conditions

z
=
t
+
x

x
+
y

y
at z = (x, y, t )
1
2
(
2
x
+
2
y
+
2
z
) + g = 0 at z = (x, y, t )
0 as z
VLM - Linear Theory

This theory is nonlinear

z
=
t
+
x

x
+
y

y
at z = (x, y, t )
1
2
(
2
x
+
2
y
+
2
z
) + g = 0 at z = (x, y, t )
0 as z
No hope of a nice simple solution for the fully nonlinear
problem.
The only hope is to linearize by observing that if the hydrofoil
only creates small waves on the free surface, then
Ux +O()
so the boundary conditions become

z
=
t
+U
x
at z = 0
1
2
U
2
+ g = 0 at z = 0
0 as z

The linear theory assumes that the waves created on the

surface are small

The boundary condition can be linearized

The equations can be solved analytically for simple cases

and numerically for more complicated cases

As we get closer to the surface, more waves are generated

on the surface. So the linearized theory fails.
Three types of problem integral
Though in principle the method is straightforward, it has been
observed that there can be problems in actually carrying out
the required computations.
We will identify three general circumstances where the
integrations have to be carried out carefully and may cause
problems.
Where do all the troublesome integrals come from?
- to set up the equations to determine the vortex strengths
ij
,
we need to know the velocity components induced by each
vortex on the system. After much simplication, this boils down
to calculating and then summing a whole lot of integrals like:
Integrals required for
ij
equations
u

ij
2
_
/2
/2
Im[f (, , , )[J(,
r
,
r
,
r
) J(,
l
,
l
,
l
)] cos d
v

ij
2
_
/2
/2
Im[f (, , , )[J(,
r
,
r
,
r
) J(,
l
,
l
,
l
)] sind
w

=

ij
2
_
/2
/2
Re[f (, , , )[J(,
r
,
r
,
r
) J(,
l
,
l
,
l
)]d
where
J(, , , ) = 2iH()

exp(

(z + ) +i

)
+
1

_
1
(z + ) +i )

_

0
e
t
dt
t +

((z + ) +i ))
_
f (, , , ) =
sec + tan
cos + sin +

=
g
U
2
sec
2
, = (x ) cos + (y ) sin
Problem integral 1:
The rst difculties that we shall discuss come from the function
f (, , , ) =
sec + tan
cos + sin +
.
Assume that for simplicity = 0 (at hydrofoil). Then (for
example)
u

_
/2
/2
J
r
J
l
sin( )
d
where
tan =

.
This exists only as a Cauchy Principal Value (CPV) integral and
MUST be evaluated by suitable methods.
Problem integral 1:
Standard numerical integration methods (e.g. Gaussian
quadrature) will not, in general, cope with CPV integrals.
To give some concrete illustrative examples, lets take the case
where

We consider a at non-tapered Hydrofoil

We use a one by one grid to calculate the integrals

The term involving the rst (non-integral) part of J looks like
- showing a typical Cauchy-type singularity behaviour where
cos +sin = 0 - i.e. at = /4 (note also the inuence of
the Heaviside step function - which causes no problems at all).
Numerical evaluation of CPV integrals
Normal tactic is to approximate the integrand using a spectral
method (e.g. splines) and use exact integration (approximate
the function UNDER the integral sign and do all integrals in
closed form) to ensure that the collocation points do not bump
into the singularity as the mesh is rened.
EXAMPLE: use trapezium rule to evaluate

_
1
0
t
4
t
1
3
dt =
49
108
+
1
81
log 2 0.4622610763
using a regular mesh of N intervals that does not bump into the
point t =
1
3
.
Problem integral 1:
Numerical results - integral value vs N - the middle points are
the right answer (spectral method) the upper and lower points
are N = 1 mod 3 and N = 2 mod 3!
Problem integral 2:
Another severe difculty arises from the integral term appearing
in the function J. The third term in J (, , , )) is of the form

_

0
e
t
t +

((z + ) +i )
dt
which in the simple case of a 1 1 grid on a at, non-tapered
1 1 hydrofoil becomes

0
sec
2

_

0
e
t
t +
0
sec
2

_
2z +
1
2
i (cos +sin)
_dt
So there will be a term in u

of the form
_
2

2
sec
2

cos +sin
Im
_
_

0
e
t
t +
0
sec
2

_
2z +
1
2
i (cos +sin)
_dt
_
d.
(Note that the sec
2
term in the outer integral is saved by the
one in the inner integral as /2).
Problem integral 2:
multiplying by the c.c., we get
_
2

2
sec
2

cos +sin
_

0
e
t
_
1
2

0
sec
2
(cos +sin)
_
(t +2z
0
sec
2
)
2
+
1
4

2
0
sec
4
(cos +sin)
2
dt d
Though the troublesome cos +sin terms cancel, the
denominator in the t -integral is 0 exactly when =

4
and
t = 4z
0
. Since z < 0 this will inevitably happen, giving a
hypersingular integral and severe numerical problems.
Suggestion: Do not use the Giesing & Smith simplication of J,
but instead use
J = i

exp(

(z+)+i

)+
1

_

0

exp((z+)+)d
- this avoids hypersingular integrals, replacing them with a CPV
integral (together with the one discussed previously - which are
NOT simultaneously singular).
Problem integral 3:
Finally, the rst term in J (, , , ) is of the form
2iH()

exp(

(z + ) +i

)
The exponential part of this term becomes highly oscillatory
under the right conditions, which correspond to the hydrofoil
being moved close to the surface of the water. As an illustrative
example, for a at hydrofoil, the exponential can be written as
E = exp
_

0
sec
2
(2z +i )
_
.
Since z < 0, when is within a distance of /2, we have
E exp
_

0

2
+O(
4
)
(2|z| +i )
_
The complex term in the exponential causes high frequency
oscillations as 0, but the real term damps them out.
For the oscillation to be sufciently damped we require (by
numerical experiment - more could be done with more time)
2|z|

10
Allowable distance from uid surface:
Now consider a hydrofoil (span 2l , chord ) with the centre of
the leading edge at (x, y) = (0, 0).

Divide the hydrofoil into a 2M N grid

Task: determine the largest possible value of

= (x ) cos + (y ) sin

this will allow us to determine a bound for |z|

- thereby telling us how close to the surface we can bring

the hydrofoil.
Obervation: max() ocurs when calculating the value induced
at the top corner control point by the vortex segment starting at
the origin.
Allowable distance from uid surface:
This maximal value of was found to be given by
=
_


2N
_
cos +
_
l
l
2m
_
sin
The maximum value of is

max
=

_


2N
_
2
+
_
l
l
2m
_
2
So we have
|z|
1
20

_
c
c
2N
_
2
+
_
l
l
2M
_
2
Placing the hydrofoil any closer to the surface then this causes
the integrand to become highly oscillatory corresponding to the
breakdown of the linearised boundary condition.
Effect of placing the hydrofoil close to the surface
Final Conclusions

You MUST use a suitable method to calculate the CPV

integrals - spectral methods are highly recommended

Do not use the Giesing & Smith simplication of J, but

instead use the less convenient-looking original version.
This has CPV integrals, but no hypersingular integrals

Use the criterion that we have found to determine how

close to the surface the hydrofoil may be placed

Dont complain that things go wrong when z is small and

negative - at this point the theory cannot be linearised and
the model is no longer valid.