mt520
July 2003
FundamentaJs of Ship Resistance and
Propulsion
Technical University Deift
G. Kui.per'
January 29, 2003
1710-K
3
Vakcode: mt520
Vaknaam: Weerstand en Voortstuwing 2
.
Het betreft een college
TUD studiepunten: 1,5
ECTS studiepiinten: 2,75
Sùbfaculteit der Werktuigbouwkiinde en Maritiéme Techniek
Docent: Koning Gans, dr.ir.H.J. de Tel: 015/27 81852
Tefwoorden:
Stronaingsmodel, Vereenvoudigen van de stromiigsmode1ien, Potentia1theorie, Stationaire vrije vloeistofoppervlaktheorie, Omstromingsberekeningen rondom profiel, Lifting line theorie, Schroefóntwerp
Curstisjiar:
3Periode:
3/O/O/OColl.uren p/w:
3 Toetsvorm: schriftelijkTentamenperiode: 1,2
Voorkennis:Weerstand en Voortstuwing i, stromingsleer
TJitgebreide beschrijving van bet onderwerp:
Overzicht van programmatuur voor het berekenen van de omstroming van
romp en voortstuwer. Uitgangspunten en berekeningstechnieken.
Aan de orde komen het gebruik van bewegingsvergelijkingen,
potentiaalberekeningen, Navier Stokes oplossingen, linearisaties, invloed van viscositeit, grenslagen
C1'legemateriaal: collegedictaat
Referenties vanuit de literatuur: Principle of Naval Architecture Opmerkingen: tentaminering: Open boek tentamen
Leerdoelen:
- Het begrijpen van de Navier Stokes Vergelijkingen. Het kunnen afleiden
vaneenvoudige stromingsmodeilen4zoals Euler, grenslaag en
potentiaal-vergelijkingen) en weten welke fysica erachter steèkt. Het kmrien
toepassen van de j niste randvoorwa arde bij de bovengenoemde
stromingsmodellen, het kunnen toepassenvan elementaire
potentiaai-oplossingen. Het kunnen afleiden van de vrijvloeistofoppervlakte stromingen.
Begrij pen van diverse numerieke stromingsmodellenen weten warop deze
stromingsmodellen zijn gebaseerd en wanneer deze nmnerieke modellen mogen worden to egepast.
- Het begrijpen van de omstroming van pröfielen en bet kunnen uitrekenen van de locale snelheden en drukken op bet oppervlak van het profiel, met
behuip van profieltabellen. Kennismaken met de dragende lijn theorie.
Het kunnen reproduceren van géinduceerde sneiheden met behuip van deze
Contents
i
2
Equations of Motion
1.1 The Continuity Equation
1.2 The Equations of Motion
1.2.1 Rotation and Deformation
1.2.2 Relation between Stresses and Strain
1.2.3 Navier-Stokes Equations
...
1.3. A Simple Example
1.4 The Euler Equations . . .
1.5 The Bernoulli Equation
...
1.6 Summary
Potential flow
2.1 Singularities in Potential Flow
2.1.1 Uniform Flow . . 2.1.2 Source
213 Vottex
16 23 25 13 14 18 19 20 22 24 27 28 28 29 30 3 2.1.4 Dipole (3D)2.2 A Simple Example of Potential Flow
2.3 Forces on a Vortex
2.4 Panel Methods
2.4.1 The Lifting Problem
2.5 Summary
Boundary Layers
3.1 The non-dimensional Navier Stokes Equations
...
3.2 The Boundary Layer Equation
3.2.1 Scaling the Thickness of'the Boundary Layer
3.3 Solutions of the Boundary Layer Equations: B1aius
41 32 33 36 37 40 43 44 44 47 47
4
3.4 Turbulence
Flow Calculations without Waves
48
51
4.1 Potential Flow Calculations 54
4.1.1 Panel 'Methods withòut Free Surface . . . .,
4.1.2 Assessment of Various Btilb Designs
54
55
4.1.3 Knuckles and Bilge. Keels.. 58
4.1.4 Assessment of the Aftbody 59
4.2 Navier-Stokes Solutiòns 62
5 Fiów Calculations with
a Free Sürface
695.1 The Linearized Fee Surface Condition 69
5.2 Kelvin Sources 73
5.3 Applications of the Kelvin 'Sources 73
5.3.1 The Michell Theory 73
5.4 Kelvin Sources for Catamaran Hulls, an Example 76
5.5 Dawson's Method ' 78
5.5.1 Applications of Dawson's Method 79
56
General Considerations to Assess Programs. 82
'6
Profile Characteristics
896.1 The Pressure Distribution 89
6.2 The Loading Distribution 92
6.2. I The Lift Curve
' 93
6.3 The 'Zero Lift Angle ' ' 95
6.4 The Leading Edge Suction Peak 95
.6.4.1 The' Ideal Angle of Attack ,
6.42 Proffle Drag ' 96 97 6.5 Profile Series ' '99 6.5.1 Thickness Distributions 99 6.52 Caniber Distributions ' 100.
6.5.3 Derivation of the Local Pressure of a Proffle 103
6.5Á Considerations to Choose or Design a Profile 104
7 Lifting Line Propeller Design
' 107'7.1 Lifting Line Thèory . 110'
7.1.1 Two-dinîensionai Lifting Lines.,.
..
110 7.1.2 Lifting Lines, in Three Dimensions (finite span) . 111January 29, 2003, Preface 7
7.1.3 Lifting Line Theory for a Propeller 113
7.2 Optimum R.adiaJ Loading Distribution 115
7.3 induction Factors 116
7.4 Propeller Design using the Induction Factors 117
7.4.1 Determiñatjon of the Inflow 118
7.4.2 Determination of the Blade Sections 119
7.4.3 Strength of the Propeller 122
7.4.4 Stresses due to Loading 122
7.4.5 Stresses due to Centrifugal Forces 123
7.4.6 Approximate Methods 124
7.4.7 Lifting Surface Corrections 126
7.4.8 Viscous Forces 131
A TABLES
iB WOORDENLIJST
viiPrefâce.
This is an introductory course on ship resistance and propulsion for the
Mar-itime. Technology Department of the Deift University of Technology. The text is written fôr students who have only basic knowldge of mathematics and fluid' dynamics Vector and tensor notation is therefore avoided. The propeller inflow is averaged in time and' space to an average uniform inflow,
and the propeller loading is consequently assumed to be steady. Variable
conditions will be adressed in a more advanced course
The intention of the course is to describe the models whichare used. '1t
does not 'contain the complete diagrams, data and 'Fonnulas necessary fôr the actual application of the method. These will, nowadays, often be contained
in a computer program. The use of computer programs in routine
calcu-lations makes it even more necessary that the tiser undérstands the model which is used and the restrictions which are inherent to such a model.For an engineer it is risky 'to refer only to "a formula" without understanding the basic theory behind it. Relying only on computer programs, which may centin fudge factors or errors, can cause-problems.
it is essential for an 'engineer to be able to formulate rapidly a crude
approximation of a problem and to grasp the main rariables involved. Forthis and for proper use of complicated programs understanding the basic approach
is more important than the detailed' development of a theory. ]eaching the
basic theory 'is the aim of this course.
Structure of the
course
The prediction of resistance, propeller and propulsion characteristics are
January 29, 2003, Preface 9
extrapolation frm modl test results.
systematic or statistical data
flow ca1cuiations
This sequence is natural since model tests form the baths formany
sys-tematic data sets The development of computatiónaj fluids dynamics (CFD) has been rapid in the last. decade, so these methods have becóme a
consider-able help in the prediction of the behavior of ships at fullscale.
Model tests and computations are often cOmplementary, both having their
advantages and disadvantages. Model tests have the disadvantage ofpossible
scale effects, but have the advantage that complex flOw jihenomena
canbe
simulated.
Calculations. have the advantage that the flow can be calculated in detail and that variations can be made rather easily. However, drastic simplifications
such as inviscid flow are used: in the calcuintión& An important aim of this,
course is to explain the complementary role of cakulations and model tests.
Textbooks
This course is not intended to provide a full inventory of practical rhethods
of ship design or for the prediction of resistance and propulsion, fOr this the
Principles of Naval Architecture [3t] is more suitable. The basis of the math-ematical description of marine hydrodynamics can be founìd in Newman's
book with the samè title [40] Reiated'specialized books are Lighthill's book
on waves [32]and the books of Knapp [26] and Young [53] on cavitation. An introduction in basi aerodynamics with numerical solutions of potential flow problems can be found in Katz and Plòtkin [24]
The emphasis in this course is on the practical application of first principles
to the prediction of the behavior of ships and propellers Insightm these first
principles increases understanding of the complex phenomena and forms a basis for intelligent problem solving.
10 C.Kuiper Resistance and Propulsion, January 29, 2003
Additional data
The text includes some additional çhita, such as 'common formuis, wich is not
nessesarily relevent to the main idea. This information is. on]y given for the
convenience of the reader using the material for his own purposes It is not
a part of' the text and does not add to the understanding of the problems.
important Fórrnulàs.
or statements
Some conclusions, fórmulas' and' definitions are important throughout the
text. A 'box 'has been placed around this information to smilifr recognition
and retrieval.
References
Since this course is, intended to provide an understanding of the basic ap-proach of a topic, 'a limited use of references has been made. In most cases users of this course will not have studied literature of the subjects in depth. No efforts have been made to refer to the most recent literature. The. refer-erces in the. Proceedings of the. International 'Towing Tank Conference [21]
or textbooks can be used to locate this information. Whennames are linked
to formulations or theories, thesê names are given with the year, but without the reference. E.g. the Betz 'condition for optimum efficiency is mentioned to date-baek-te-19-29 j-but-is-not-mentioned in the references. Full references are
only made when full sets of diagrams, dataor formulas have been used. This
makes' the list of references less dependent on the most recent publications.
'Acknowledgements
Many students and colleagues from Marin have given comments, corrections and material fór this 'course. The help of Mr. Hoekstra, van' Gent, van
Wijngaarden, Holtrop, de Koriing-G'ans. and Mrs.. Raven is gratefully
ac-'knowl'edged. Special: help was. received 'from' SM.M.Bernaert. She has done the redactional work and has improved many topics The text will be
January 29, 2003, Preface 11 pages and on the title page Any comments can be helpful to improve it and will be very welcome.
Chapter 1
Equations of Motion
Objective:
To show the origin of the equations of motion The derivation of the formula 's is not impOrtant. The purpose is to show that the equations of motion are the equivalent of Newtons second law, -to show the assumptions used to arrive at the equations (Hooke 's law, Newtonian fluidj. Also the con-cept of rotation is introduced.The -equations of rnotiòn describe the relation between forces and motiòns
in the fluid. The equations are rewritten forms of Newtons'second 1aw:
F(t). = mü (1.1)
where F=force in Newtons, mmass in kg and x=position in m.
The equations of motion therefore describe a relation, not a situation!
The situation, local velôcities and pressures (forces), can -only be found from
a combination of the equations of motion and boundary conditions. The,
boundary conditions are situations inpressureor velocity at a certain position
at a certain- time. When the boundary conditionsare suficient, the velo:cities
-and pressures at other times and: positions can then be cakulated using the
equations of motion.
Example A simple and well known example is: the applicationof Newtons'
law directly to a bod' with -mass m. 'The equation of motion is eq. 1.1 and the boundary conditions are e;g.
14 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
y(0) = O x(0) = O
F(t)=f(t)
This makes it possible to calculate the velocity of thebody at an arbitrary time t from:
From the boundary conditions it is found that C = O. In this case the
equation of motion is integrated.
When the velocity ±(t) is given as a boundary condition instead of f(t) the forces can be derived directly from the équation of motion by using the time-derivative of the given velocity.
This example is meant to ilhistrate the character of the equations of
motion in a fluid. A very useful notation for three-dimensional flow fields
is vector notation. In the follöwing , however, the equations öf motion will
be derived without using this notatiön because it is not essential. To avoid
excessive long formulat ions. the derivation will be inone direction only and for
incompressible flow. The purpose of this derivation is to give basicinsight
in the equations of motion and to show the concepts of stress-strain and
of-rotation=deförmation--Abasic-insight--iii-the-equations of mttiorris-also necessary to understand simplifications which will be made later on, in order to make calculations feasible.
Before deriving these equations of motion another basic equation has to be formulated.
1.1
The Continuity Equation
Nomenclature:
In flow calculations the velocity of particulars has .a direction and a mag-nitude. The components are mentioned u, y and w. These are the velocities
pt
January 29, 2003, Equations of Motion 15
in respectively in x-, y- and z-direction. These components are local
veloci-ties. There are also global velbcity components, they are expressed with the
capitals: U, V and W (respectively inx-, y-, and z-direction).
Sometimes pertuated velocities are considered. When this is the case, it
will be mentioned. Then, the pertuation components are respectively u, y
and w.
The continuity equation expresses that no mass is lost when there is no flow through a control surface. When a cube of fluid is observed at a cer-tain time its shape may change over, time, its deni'ty may change, but the amountS of mass is unchanged regardless the deformations or compression.
For an incompressible fluid this means that not only the mass, but also the
volume of such a cube remains unchanged. We will assume incompressible
fluid here, because the line of reasoning remains the same as in compressible flow. Water can nearly always be considered as incompressible.
The formulation of the continuity equation is an example of the use of an arbitrary control volume which is fixed in time and position.
Figure 1.1: Control Volume Fixed in Time and Position
Consider a control volume as shown in Fig. 1.1 with sides dx, dy and dz. The velocity of the fluid entering the control volume from the left side
ADFG is now called u. (Note that this is the average velocity over the plane,
not a local velocity. This is useful because afterwards the size of the control
16 G.Kuiper, Resistance and Prvpulsion 2, January 29, 2003
the local velocity in point G.). The velocity leaving he cóntrol volume at the
right side BCEH is then u+ dx. Partial derivatives are used here although
only the derivative in x-direction is relevant. Sirnilary, the velocity of the
fluid entering the control volume from the back is called y and that of the
fluid entering from the bottom is called w.
The total volume change dV in a time-step dt is now:
dV
= (dx)ddzdt + (dy)dxdzdt + (dz)dxdydt
In incompresi'ble 'flow this volume change must be zero for any volume
and for any time-step, so that
Ou 0v 0w
(1.3)
Ox ôy ôz
This is the continuity equation for incompressible flów. This equation can be used in the formulation of the equations of motion.
1.2
The Equati'ons of Motion
For the formulation of the equations of motion we consider the forces on a cube of fluid. These forces äre decomposed' in three directions, so that at each side of the cube there is one pressure force (indicated a) and two friction forces (indicatecLi-)
In Fig. 1.2 a cross section of a cube is shOwn with
at the bottom the
pressure force a and the fiction forceT fl x'- direction. The related forces at
the 'opposIte side of the cube are o- ± dz and y + Edz. These forces are
always in opposite directions and the resulting forces on the fluid particle
can therefore be written as dz and Edz only.
in Fig. 1.3 the resulting forces acting on a cube of fluid are indicated.
Note that in this case this is not a cöntrol volume, but a material cube of
fluid at a certain time. The first index indicates the plane at which the force
is acting, the second index is the direction of the force, So is a friction
force on a plane perpendicular to the x-axis in the direction of the z-axis.
We can now förmulate the resulting, force in each directión on the fluid particle. For sake of simplicity this will only be done in the x-direction
Figure i2; Forces on a Cross Section of a Cúbe in a Fluid
Figure L3: Resulting Forces in x-direction
The resulting force in xdirection is
18 C.Kuiper, Resistance and Propulsion 2, January 29, 2003
s
F
(&TXd)dd + (dy)dxdz ± (dz)dxdy
This force can now simply be used. ineq. 1.1, resulting in
F = pdxdydz (1.4)
where dxdydz is the volume of the element. The result of thissubstitution
du
ôa
a-t a 3y ôz (1.5)
In eq. 1.5 the relation between the fluid motion u. and the stresses in the fluid is given The stresses m the fluid are in turn related to the deformations and compressions in the fluid. For a further formulation of the equations of
motion the relation between the deformations (strain) in the fluid and the
stresses on a fluid element is required.
1.2..i
Fotation and Deformation
The stresses in the fluid will now be related to the deformations of the fluid
particle. The cube, considered before, will be deformed .by the forces on
its sides. For simplicity only two dimensions will be considered, so a cross
section of the fluid particle in the x-z plane as shown in Fig. 1.4.
The originally rectangular shape of the fluid particle under consideration
will he deformed after a time At to an arbitrary cross section as given in
Fig. 1.4. After the time-step:
B(ix, O) is in B'(Lx + x)
D(O, Ay) is in D'(Ly, Lìy +
C(x,
y) is in C'(tXx +-I- Ay, Ly +
x + y)C
January 29, 2003, Equations of Motion 19
A B A' X
Figure 1.4 Deformation of a Fluid Particle
the rotation of the diagonal AC, which is
-
). The value-is called the rotation of the element.
the deformation of the element, which is writteñ as
+
). This iscalled the dilatation of the element.
The dilatation of the element causes a strain in the element.
1.2.2
Relation between Stresses and Sffn
The relation between the dilatation of the element and the frictional stresses
is written as
0v Ou
=
+
(1.6)Ou 0w
20 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
Note that the frictional stress in the x-direction contains also the velocity
gradients in the y-direétion. The parameterji is the dynamic viscosity of the
fluid.
Also when no dilatation occurs a deformation is possible, as shown in
Fig 1.5.ln this case the pressure gradient in z-directioncauses a defórmation.
In such a case the follOwing relation yields:
8cY ô ôu
=
Here p is the mean static pressure. This relation is not evident. It is. valid: for an isotropic fluid and the stress-strain relation uses the same dy-namic viscosity as the relation between viscous stresses and deformations.
y
D
(1.7)
A B A B'
Figure_L Deformation_of_a_Fluid_Particle_due_to the Pressure_Gradient
Lbwgvgle}
The fluids for which the viscosity is a constant independent of the
flow-field are called Newtonian fluids Water is such a fluid. An example of a
non-newtonian fluid is blood. In blood the viscosity increases with increas-ing velocity gradient. This is also the case with many polymer solutions.
Using the foregoing the equations of motoncan be formulated.
1.2.3
Navier-Stokes Equations
Eq. 1.5 can be rewritten using the foregoing as:
y
C,
January 29, 2003, Equatiöns of Motion 21
The total derivative of the velocity u can be written in its partial
deriva-tives as
=
+ u + v ± w so that the formulation of the equationof motion in the x-direction yields:
Su Su Su Su
Sx Sy Sz (1.9)
Sp S2u S2u a2u
a Su
0v 5w+ /L-(-+-±-)
öxôx
Oy Sz
This equation is also called the Navier-Stokes equation, because the
French mathematician Navier and the English mathematician Stokes
deivel-oped this equation independently in the middlle of the nineteenthcentury.
In incompressible flow this equation simplifies because eq. 1.3 can be used, resulting in
au au au
au
p{- +u ±v-- +w} =
52u S2u 52u
52w
ôzS)
5v 5w ay ôz (1.8) (1.10)The equations in the other two directions çan simply be found by changing the x,y and z and u,v,w variab1es and indices.
du o a
a
a
3v ôu=
ô ôu 5w
ap 52u 52v 52u 82u
=
ap a2u a2u a2u
ô Su
22 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
1.3
A Simple Example
The Navier-Stokes equations are too difficult to solve in general. As a simple
illustration the flow between two parallel wallscan be considered (Fig. 1.6).
In this case there is only motion in one direction, and only the equation of motion in that direction has to be solved. Moreovèr, the flow is steady and does not change in time
Figure L6: The Flow between Two Parallel Wails
When the flow is in x-direction the Navier-Stokes equation in x direction
has to be considered. Since the flow is steady the time-derivatives are zero.
Also the x-derivatives of the velocity arezero since the flow proffle remains
the saxne-aallst-at-ions--T-here-is-onIy-a-pressure gradientg in-x direction.
The pressure gradient is negative since thepressure decreases in x-direction.
Because the flow is two-dimensional there are no derivatives in z-direction and the y- and w-components of the flow are zero. This simplifies the N.S. equation to:
ap 02u
The solution of equation 1.11 is now straightforward:
au =
yap
+ cl
(1.12)January 29, 2003, Equations of Motion 23
u =
+C1y+C2
2t ax
This is the integrated equation of motion. The constants can be deter-mined using boundary conditions. These conditions are:
s v=O
u(b) = u(b) =O
where b and b are the locations of the wall in y-direction. With these
boundary conditions it has been assured that the flow at the wall is parallel
to the wall and the condition of viscous flow has also been imposed: the
velocity of the fluid close to the wall is zero.
The solution is C1 = O and C2
=
. This leads to the solution:lap
2 2u=--(b y)
The velocity distribution between the walls is thus arabo1ic.
1.4
The Euler Equations
Because-the Navier-Stokes-equations are very difficult-to solve, both analyt-ically and numeranalyt-ically, simplifications have been applied. The first
simpli-fication is the neglect of viscosity. This removes all terms with from the
Navier Stokes equations. The result is called the Euler-equation. From the
Euler equations some specific results can be derived . Although the Euler
Equations are 3D, now only the 2D-equations are considered. The stationary Euler-equation in x-direction is:
au
auu + v =
ax aypax
This equation can be rewritten in terms of rotation as
24 G.Kuiper, Resistance and Propulsion , January 9, OO8 i ôp p 0x l&u2 5u
=
lau2
auia
2 2 =(u+v)
This form can also be integrated in x-direction:
p 2 2 u V
-- + C = (u + y ) + j v( - )dx
(1.16)p 2
J
ôy ôxEq. 1.16 is a form of the Euler equation . These equations of motion
describe the flow in an inviscid, rotational flow. The Euler equations can
be used in flow regions where vortices, generated elsewhere, are present, but where no new vorticity is generated. The path and behavior of the vorticity can be calculated by Euler solvers.
1.5
The Bernoulli Eqfuatiôn
The pressure can be expressed explicitly from the Euler equations.Thereare
two possibilities to do this:
along a streamline, where f {v(
-
) - u(.-
)} ds = Oin irrotational flow where Oy
-
Ox = OIn both cases the integration leads to the Bernoulli-equation:
p+pV2 =c
(1.17)When gravity is the only external force the constantcan be written as pgh+C, so that the Bernoulli equation takes the well known form
p+ pV +pgh= C
ôv Ox axau
0v +V(ay ax (1.15) (1.18)January 29, 2003, Equations of Motion 25 Note that in an inviscid and irrotational flow the Bernouffi equation is valid anywhere in the flow (it has the same constant everywhere in the flow)
In an inviscid but rotational flow the Bernoulli equation has the same constant along a streamline. Every streamline has its own constant, however.
1.6
Summary.
The equations of motion in the flow are a translation of the second law of
Newton:F orce = mass x acceleration When this law is applied to a particle
in the flow the Navier stokes equations evolve. In deriving the Navier Stokes equations use has been made of the law of Hooke, which relates the
deforma-tion of the fluid with the required stress. Also the fluid is characterized by a
single viscosity in all directions and for all deformations. The class of fluids with these characteristics are called Newtonian fluids.
The Navier Stokes equations can be simplified by neglecting the viscosity.
This results in the Euler equations. A further simplification can be made by
neglecting the rotation of the flow. In that case the equations of motion can be integrated and the Bermioulli equation evolves:
Chapter 2
Potential flow
Objective: To show the use of singularities in potential flow calculations.
The equations of motion relate the pressures in the fluid with the flow velocities.In an irrotational, inviscid flow the equation of motion can be writ-ten as the Bernouffi equation,, eq. 1. i& The continuity equation is given in
eq. 1.3. These two equations are the constitutive equations for an
irrota-tional, inviscid flow. That means that they are vaiid anywhere in the flow.
A potential flow is now defined as a flow field which can be described by a scalar function, generally indicated as (x, y, z), which is the potential. The-velocities in the flow in any-direetion ar-e-the par-thu derivatives-in that direction of the potential:
u =
-Ox
V =
-ay
28 G.Kuiper, Resistance and Propulsion .2, Jamuar' 29, 2003
The rotation of a potential flow e.g. in the x
- y
plane can be wrjtten interms of the potential
@u avaa
aa
ay ôx Oyäx ôxOy
and this is always zero for a function which is twice differentiable. So
a potential flöw is always irrotational
In such a flow the relation between the pressures and the velocities can be described by the Bernoulli equation.
'When the continuity equation is written in terms of the potential this results in
a2 52 32
a ay2 8z2
This equation is called the Laplace equation In order to describe a flow
field it is therefore necessary to find a solution of the Laplace equation.
The strength of the potential flow description is that it is possible to
define a number of elementary solutions of the Laplace equation. Because
the solutions are superimposable, an arbitrary flow field can be expressed in
terms of superpositions of these elementary solutions.
(2.1)
2.1
Singularities in Potential Flow
2.1.1
Uniform Flow
The simplest solution of the Laplace equation is
=Ux
Its second derivative in x-direction is zero and the derivatives in y and z directions are also zero So this potential satisfies the Laplace equation. The meaning of this potential is a uniform steady flow in x-direction, since the
January 29, 2003, Potential Flow 29
2.1.2
Source
Another solution is the potential of a source. In two dimensional flow its potential is
=--1n(r)
(2.2)r is the distance from the source with strength Q. It requires some alge-bra to show that this potential satisfies the Laplace equation.
The velocities in radial direction can be found from the derivative of the potential function to r, so
Q
Vr -2irr
There is only a velocity in radial direction because vo = O.
/\
Figure 2.1: Velocity Field of a Potential Source
This is called a source because the amount of fluid passing through a
circle at radius r is always Q. The flow field of a single source is shown in Fig. 2.1. VVhen the sign of the source is negative it is called a sink.
In three dimensions the potential of a source is
C.
4irr (2.3)
The velocities in radial direction are found from
a
30 C.Kuiper, Resistance and Propulsion 2, January 29, 2003
Here the amount of fluid passing through a sphere around the source per unit time is always Q.
2.1.3
Vortex
The potential of a two-dimensional vortex is
2ir The velocities are now in tangential direction
10cI F
Ve = -- =
rôO 2irr
There is no velocity in radial direction since the potential ftmction is
independent of r. Here F is the vortex strength. The integral of the velocity around a contour in the flow is called the circulation of the vortex, defined as
F =
j'vods
The flow field of a two-dimensional vortex is shown in Fig. 2.2.
The two-dimensional vortex can be considered as a vortex element per pendicular to the plane.
In three dimensions the same formulation 6f a vortex can be used. Only a vortex element is considered, and the velocity induced by such a vortex element is in a plane perpendicular to the vortex element, as shown in Fig. 2.3.
The velocity induced by the vortex element dl with strength F in an arbitrary point P can be written as
F sin O
dV= ds
4irr2
For an infinitely long straight vortex in 3 D the induced velocity is
r
2ira (2.5)
January 29, 2003, Potential Flow 31
Figure 2.2: Velocity Field of a Potential Vortex
Figure 2.3: Velocities Induced by a Vortex Line Element
where a is the distance perpendicular to the vortex line. This is a well
32 C.Kuiper, Resistance and PrOpulsion 2, January 29, 2003
An isolated vortex element cannot exist in a potential flow, field, because in such a flow field no vorticity is generated. So a vortex element has to be part of a continuous vortex or it splits into two vortex elements with different
directions, but with the same total strength. A vortex can only end on a solid
wall which is the boundary ofa potential flòw 'fie1d
2.1.4
Dipole (3D)
An elementary solution of the Laplace 'equation which is also often used is a combination of a source and a sink ('a negative source). The source and sink are infinitely close together, but the product of source/sinkstrength and distance remains finite This product is the, dipole or doublet strength ji.
Consider a source of strength Q and a sink of strength -Q at a distance i from each other. The potential in an arbitrary point P is
47r \r1 r2
(2.6)
The factor between brackets can be written as
r2 r1
r1 T2
When i - O the product r2r1 -* r2 and the distance r2 - r1 - i.cosû, where
O is the. angle between the axis of the dipole and the arbitrary point P. When
Ql -*
ji
the potential of the dipole can be writtenasjicos(0)
3D= 2
4irr
Analogue as in 3D, a dipole in 2D can be derived, the result is:
jicos (9')
2D =
2irr
The velocity field belonging to a dipole with its axis in the x-direction is shown in Fig. 2.4.
The velocity field 'indicates that dipoles can be written as vortices and
vice-versa. In fact it can be shown that a vortex can be written as a deriva-tive of the dipole. So a velocity field containing dipoles is equivalent with a
(2.7)
January 29, 2003, Potential Flow 33
Figure 2.4: Velocity Field of a Doublet or Dipole velocity field with vortices.
Note that a dipole has a direction. This direction is the axis of the position
of source and ink. The potential ofa dipole is in fact the derivative of the potential of a source, so
(dzpole) a_(source) (2.9)
This relation will be used in panel methods (see below).
The velocity of a source or vortex is infinite at the location of the source
or vortex itself. The mentióned potentials are therefore called singularities.
Note that a single singularity describes the fow field everywhere in the flow.
2.2 A Simple Example of Potential Flow
The use of singu]arities to describe a flow field is now illustrated by the
combination of a uniform flow with a single dipole (in 2D). A uniform flow in
x-direction has a potential function 1 Ux. When a dipole with a strength
N3, in x direction is put in a unifòrm flow the potentials can be added, so:
(stream) = (parale1) ± 1(dipole) = Ux N3, cosO
where r2 = x2 + y2 and x r cos O. The velocity in radial direction is then
found from
34 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
N
vr==UcosO(1+
2irUr2
For r -' oo this reduces to
Vr = LT cos O, which is the undisturbed uniform flow in x-direction.There is another special location, where
2_
Nr
2irU
At that location the radial velocity is exactly zero, so there is only tan-gential velocity. When this location is considered as the boundary between
inner and outer flow it turns out that the combination of a dipole and a
uniform flow represents the flow around a cylinder!.
Figure 2.5: Flow around a Dipole or Doublet in Uniform Flow
Note that the tangential velocity along the cylinder is not zero, as it
would be in real fluid. This is because the solution is in potential flow, which is inviscid.
The tangential velocity along the cylinder is also be derived from the
potential function:
(2.11)
vo =
= U sin 0(1
January 29, 2003, Potential Flow 35
At the location of the cylinder R2
= -, since there is only a tangential
velocity, so
vo = 2UsinO
So there is a sinusoidal distribution of the velocity on the cilinder. The maximam velocity at O = ir/2 is twice the incoming velocity U. Because this is in a potential flow (irrotational) Bernoulli's law can be applied to find the pressure from the velocities. The inviscid pressure distribution can thus be found. At the top and bottom position (O = ir/2) the velocity is maximum
and the pressure is consequently minimum, the minimum pressure is found
from: Po + 1/2pU2= prnjn + 1/2p{2U]2 Or PminPO
-3
1/2pU2
-where Po iS the pressure in the undisturbed flow. The value-3 is called
the pressure coefficient.
This example shows that a single dipole can describe the flow around a cylinder. Combinations of singularities in the flow can describe almost any flow patterns. When the flow around an arbitrary body has to be calculated the potential flow theory is used to locate singularities in the flow and to
determine the strength of these singularities in such a way that the body
contour is exactly a streamline.
Because the pressure distribution in our example of the cylinder is sym-metrical in x-direction there is no resulting dragforce on the cylinder. This is a general feature of bodies in a potential flow (Paradox of d'Alembert).
36 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
The resistance of a body in
a parallel potential flow is zero.
This is due to the absence of the boundary layer, which causes not only the absence of a friction force along the surface, but also the absence of
pres-sure drag.
2.3
Forces on a Vortex
It is easy to add a circnlation around the cylinder by adding a vortex with strength F in the center of the cylinder. The potential of the flow is similar to eq. 2.10 with the potential of the vortex added. The radial velocity in eq. 2.11
does not change, because the radial derivative of the vortex potential iszero.
The tangential velocity component changes from eq. 2.12 into:
N F
vo=UsinO(1
)+-2'ïrUR2 2irr (2.13)
The velocity distribution now becomes asymmetrical.
At the top of the
cylinder with 6 = the velocity is
N
F= U(1
- 27rUR2) + tebottoinoLthecyiinder the velocity isN F
= +U(i
- 2irUR2)+
The pressure distribution, which can again be found by application of Bernoulli's law, also is asymmetrical. Integration of the pressure over the cylinder gives after some algebra the resulting force pUF in the direction perpendicular to the flow.
So a vortex does give a force, perpendicular to the flow. The relation
between the side force L on a vortex and the incoming flow U is called the law of Kutta-Joukowsky:
L = pUF
January 29, 2003, Potential Flow 37
Vortices are so important because they can represent a lift force in the fluid. This will be used for the description of airfoil wings or propeller blades. An example of the application of lift on a single vortex is the Flettner rotor
The rotating cylinders generate a circulation around the cylinder due to frictional forces in the boundary layer close to the wall. In wind the result is a side force perpendicular to the wind.
2.4
Panel Methods
Without proof it is stated here that any potential flow field may be writtenas
a surface integral over the boundary S of a source distribution and a dipole distribution on S. (See e.g [24]). The dipole distribution has its direction normal to S. Note that the boundary surface S is the complete boundary of the flow, as shown in Fig. 2.6.
38 .G.Kuiper1, Resistance and' Propulsion , January 9, 003
Figure 2.7 Panelling of an Axbitrary Body
When the outer surface goes to 'iifinity the surface S can be considered.
as the surface of the body and the cut from the body to hifinity. When this cut has no thiciness only dipoles are present there.. When. there is no lift on
the body a source distributibn over the surface of the body remains.
For a short indication Of panel methods it is sufficient to consider 'a source
distribution on a body. This situation is different fromthe examplé of a
'cilin-der, because in that example the dipolewas in the. center_oLthecyJinder and
not on the surface.
The boundary condition On the. .body is that the normal velocity on the body is zero, so
On (2.14)
This condition is valid everywhereon, the surface of the body.
The panel method is a :d:i,scretization of this boundary problem.' The
surface' of the body is approximated by flat elements. At each element a
January 29, 2003, Potential Flow 39
So the souice distribution on the body is also discretised. The boundary
condition eq., 2.14 is now applied on one point of the element. This point is the control point, where the boundary condition is fulfilled.
Consider the velocity, induced by an arbitrary source panel in one contröt point P1, as: sketched in Fig. 2.7. The potential in P1 ofan arbitrary panel is
fUpa,e
dS J panel 4irrwhere dS is a part of the paneL This integral is the influence function of a panel on a control point The nain variable is the distance r the control pomt and an arbitrary point on the panel S The potential m P is the sum
of the influence functions of ali panels and the normal velocity is the normal derivative of this sum.
A singularity occurs in the influence function when P1 is located in the panel
which is considered, because in P1 the radius i is zero. This leads to a
sin-gular integrali over the panel with P1. Fortunately this integral has a finite
sölùtion and can be analytically formulated.
When r is relatively large (say afe times the panelsize), the integration
can be omitted because the panel distribution can be consideredas a single
source with strength cr8. The majority fthe panels can thus be treated: as
discrete sources,
The normal velocity induced by all panels in the contröl point P1 can thus be formulated as the sum of the influence functions of ail panels Application of the boundary condition of zero normal velocity results in a set of N linear equations, where N is the number of panels. A matrix of N x N results and this matrix can be solved.
There are different methods to formulate the problem, but the basic
assump-tions are the same. Refinements are possible by using curved panels instead
of flat ones and by using linear or quadratic formulations for the source
dis-tribution on the panels. For a short review see [1:5]. An introduction in the formulation of the problem is given eg in [24].
40 C.Kuiper, Resistance and Propulsion 2, January 29, 2003
2.4.1
The Lifting Pröblem
In the. foregoing description it was assumed that there was no lift on the
body. When lift is present two important additions have to be made.
The first one is that there ae not 'only singularities on thebody, but also in the wake, which is expressed by the cut S in Fig. 2.6. Consequently the wake has to be panelled too. This can be a problem because the position of the wake is not always known. In case of a flat wing it can be approximated by positioning it in the plane of the wing, which is only true when the dipole strength or vortex strength is smalL This wake model therefore introduces
again sorne linearization. On a propeller the problem is more complicated
because the wake has: a complicated shape. This shape can be determined
only by the condition that the wake is in the direction of the (induced) flow.
The induced flow is again dependent on the dipole strength, and this dilemma
is characteristic, for the propeller problem. Simplified wakemodels will be
treated in the sectiOns on propeller design and analysis.
On the body and in the wake a dipole distribution is added to the
prob-lem. Every panel has not only a (uniform) source strength o, but also a
(umform) dipole strength ¡i of unknown magmtude The influence functions
of the panels become more complicatedas a result. The general shape of the
influence functhin is 'given in eq. 2.15.
I
Ii
1ia1\J-a+--jdS
(2.15)panel n
P1 =
In this equation the factor has been absorbed in the strength of the
sin-gularities.
The boundary condition remains the same, but an additional boundary
condition is necessary because there are additional unknown dipole strengths
in the influence functions. A solution can be found by prescribing as an
adcli-tiönal condition that the floW at the aft stagnatiòn point remains finite it is
easier to consider a wing, where the flow velocity at the trailingedge has to
remain flnite This condition is càJled the Kuttci conditIOn and it determines the strength of the dipoles or vörtiçes in' the wake. It adds one equation to the matrix of influence functions.
LST1NG STIP OF PANEU OUND VOTICITY N-LINES TRAILING EbGE
January 29, 2003, Potential Flow 41 A common problem of a panel method with lift is the flow around an airplane, as shown in Fig. 2.8.
-4-tRAIUNG EDGE
SEGMENT
Figure 2.8: Panels at an Airplane Sürface (From Hess)
The lift on the wings is represented by dipoles over the wings and the wake. The fuselage is covered with sources only.
Although the-for
often not so easy. When the panels are chosen wrongiy or inconsistently, the matrix of iufluence-,fanctions becomes ilbconditioned and the solutIon is no longer reliable.
2.5
Summary
The velocities in an incompressible, irrótational flow can be described by the Laplace equation and the Bernoulli equation. The latter equàtion relates the pressures with the velocities:
42 G.Kuiper, Resistance and Propulsion 2, January 29, 2008
These flows are called potential flow fields.
Potentizl flow fields can be described with singularities. These are poten-tials which are sirigular(infinite) in the core. Typical singularities are vortices and sources. A body in the flow can be described by a closed streamline, which is generated by a singularity distribution in the flow. In a potential flOw no force is exerted on a body without circulation. (Paradox of d'Alembert). When circulation is present the force is perpendicular to the flow and its magnitude
is
F=pVF
where F is the circulation around the body contour.
Single vortIces have the property that they cann'ot begin or end in the flow. They are either closed or are connected to a solid vial 1.
The velocity induced by a vortex in the flow can be found from its potential. For a straight line vortex element this velocity is
given by the law of Biot and Savart
Fsinü
dv=
ds4irr2
The flow around an aTbitrary three dimensional body can be described with a surface distributïon of singularities. When no circulation, is present a surface distribution of sources (and sinks,) is enough. When circulation is present also vortices or dipoles have to be distributed over the surface and
Chapter 3
Boundary Layers
'Objective: A basic approach of boundary layer theory and the definition of a turbulence model
Consider the flow over a two-dirnensïonaJ body in the x-y plane. The mean flow is in the x- direction.The flow can be considered inviscid, except in a region close to the body, where a strong velocity gradient is present.
In this boundary layer region the equations of motions e the Navier-Stokes
equations When the flow remains attached the viscous region is thm relative
to the chord f the profile and this thin layer is called 'the boundary layer.
A two-dimensional body is. considered, so the two-dimensional Navier-Stokes equations apply. These equations are:
ôu au
u F'
V-ôx ay 0v 0vu +v-
Ox Oy Ou 0v Ox ayIn these equations the dynamic viscosity a has 'been replaced by the kine-matic viscosfty u, which 'is found from' ii =
lap
82u ô2ulop
02v 02v-
± V('+
Oect.
(3.1)
44 G.Kuiper Resistance and Propulsion 2, January 29, 2003
3.1
The non-dimensional Navier Stokes
Equa-tions
For further use these equations are first non-dimensionalized by replacing
/ u by u =
xbyx' =
ybyy'
pbyp' =
where U is the incoming undisturbed flow velocity and I is a relevant length of the body. For one term this is döne as follows:
=
(3.2),Du' U2
=
T)
T11s can be done for all terms in eq. 3d, which leads to the non-dimensional equivalent of the Navier-Stokes equation:
Du Du i 87) u A2, A2,
U-
-4-'ii-DxDy =
p3xUlDxDy2)
(3.3) Dv Dv= ---+{----+--
lap
y (D2v
D2v Dx Dy p Dy Ut \ Dz2 Dy2 Du Dv±-
Dx DyIn eq. 3.3 all primes are deleted, but all velocities and distances are now non- dimensional.The quantity u/UI is the inverse of the Reynolds number.
3.2
The Boundary Layer Equation
Terms which are small when the boundary layer is thin will now be neglected. The boundary layer with thickness 6 is thin when q «i. Because the profile
=0
January 29, 2003, Boundary Layers 45 is assumed to be thin also the vertical velocity y will be small. This leads to the following orders of magnitude of the non-dimensional quantities:
u=ø(1)
x=ø(1)
vø(6)
.y=ø(S)
This makes it possible to identify terms of order i5. In x-direction eq. 3.3
is
Ou Ou lOp i
ôu
02uThe terms of this equation have the following order. of magnitude:
1
i
lOpi
I iF1- +8=
-
±(- ±
The last term between the brackets can only be of order i when R
-which is the case for iarge Reynolds numbers. Neglecting the terms of order 8 the boundary layer equation is found:
Ou Ou
löp__i Ou
(3.4)
It is important to remember that the Reynolds number in this case has to be of the order 1/82. So when the boundary layer thickness is of the order of 1 percent of the length of the body, the Reynolds number based on the
length of the body has to be of the order of iü. Still these equations are only valid for laminar boundary layers.
In y-direction a similar procedure can be followed. The Navier-Stokes equation in that direction is
46 CKuiper, Resistance and Propulsion 2, January 29, 2003
which has terms with the following order of magnitude
6 6.
ia
26
6All terms except the pressure term are now of order ô or smaller, so the
equation of motion in y-direction reduces to
ay
This means that the pressure in the boundary layer is constant over the
boundary layer thickness. This is an important conclusion because now the
pressure. outside the boundary layer can be identified with the wall pressure.
Outside the boundar layer the flow is modeled kike a imdisturbed
poteri-tial fiw and the velocitycomponents are respectively U and y. The stream-line at the edge of the boundary layer is along the profile of the boundary
layer.
So there is a relation between the paralll and
cross components,namely that the direction of the streamline: is equal to the tangent of the
proffle of the boundary layer.
- = - or y = U
u
Ox Ox
Because the boundary layer is small, the velocity component y is small.
using BernoiiilWs equation, the square of y is very smaJrd can be negld.
Thsi way, the pressure is only dependent of the velocity component in
x-direction:
au
Opox ox
This can be. integrated in x-direction, which leads to
p+,pU2
= C
So outside the boundary layer the BernouJli equation is valid again. The
January 29, 2003, Boundary Layers 47
3.2.1
Scaling the Thickness of the Boundary 'Layer
in the foregoing, derivation of the bóundary layer equations the assumption
was made that R (X 1/i2. Inversely this means that ' oc (The prime
means the nondimensiona1 boundary layer thickness f) When a model is
now towed at scale ratio 25 at the full scale 'Froude number -, the ratio
between R, and Rnm is 125,, where the index m stands for ml and s for
ship. This means that
=Vi
So the boundary layer at the model is relatively more than ten. times thicker than at the ship. Relatively, because the boundary layer is taken
relative to the size of the ship This causes the scale effects on the wake distribution and on the resistance.
3.3
Solutions of the Boundary Layer
Equa-tions: Biasius
The 'boundary layer equation eq. 3.4 can be. integrated when it is assumed that the velocity profile at every x-position has a similar shape
J (3.5)
Such a solution' is called a similarity solution. The specific friction at the wall is found from
Ou
Tw
0Y
y0
Blasius assumed' a simple similarity solution by taking the velocity distribu-tion in the boundary layer as
U -
"5This made it possible to integrate the boundary layer equation in combination with the no slip condition at the wall. The resistance D of a flat plate with
48 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
length i and width b canbe found from the integration of the specific friction
coefficient 'r iver the length of the plate at both sides: D
=
2bf
r(s)dse
The solution of Blasius for a flat plate is
1.328 cf
=
where D=
:pU23.4
Turbulence
The foregoing boundary layer equations are valid för laminar flow. The vis-cosity used in the equations i the kinematic visvis-cosity, a property of the fluid.. Simultaneously a high Reynolds number is assumed, so the Blasius solution
is only valid for R iO4 and until the boundary layer becomes turbu1nt at
R7., io
Thrbulence is the occurrence of random motions additional to the mean velocity. So the instantaneous velocity of a fluid particle can be described by the mean velocity and a turbulent component:
u
=
u' with Oy = + y' with = O
p + p' with O
To illustrate the effect of turbulence the N:avierStokes equation in x-direction (eq. 3.1) is considered:
Ou Ou Op 02u 02u.
p(u- + v-h)
-- ± i(- +
Substituting the instantaneous velocity in this equation and taking the time average results in a new formulation, the term p can be written as
i (2 + 2iu' + u'2)
-
2 axi ôz i au'2 an
i a'
= --+ =+---
2ax
2ax
ax20x
The turbulent equation of motion in x-direction becomes in this way:
a
a
p(u +v)
-ax ay
January 29, 2003, Boundary Layers 49 The latter is true because the time-average of the turbulent pressure com-ponent is zero. So this pressure term is not changed by turbu1ence This is true for all right hand terms, but not for the left hand terms. E.g. the term
u can be written as au 13u2
2a
a2i
a2u iai
aJI.
-
-ä--
ay (3.7) (3.8)The extra terms require further attention, because they describe the éffect
of tiirbulence The term 2 will be small relative to the derivative
in y direction, because the variations of the velocities normal to the. wall (in y-direction) are much larger than those in flow direction. (See also the consideration of a thin boundary) To assess the effects of turbulence only the
second term is sufficient. Similarly the term will be small relative
These term can therefore alsobeneg1ected.
When it is assumed that the term can be written as
a
AT
ay it possible to. rewrite eq. 3.8 as:
/ ô
au\
ai
a2u 52up(+ I
\ ax
ayj
=----+,i--+A--
ax ay2ay2
(3.9)
(3.10) When written in this form is is clear that the turbulence can be taken as an additional viscosity, also called the eddy viscosity . This is of course due
50 GKuiper, Resistance and Propulsion , January 9, 003
to the assumption of eq. 3.9. This assurnption is called the turbulence model,
which in this case relates the turbulent viscosity in a !jfl fashion rith thê
velocity gradient in. the bomdary layer. Other turbulent mode1 lave been formulated. No final turbulence model that .can describe the phenomena in
rai flow has been formulated'yet.
The eddy. viscosity is generally much greater than the dyflamic viscosity, so AT is greater than p. The determination of A is therefore important. A well-known model is that of Prandtl, who assumed that
pl
where i is the average distance of the turbulent motion in the, boundary
layer. The model. is therefore called the mixing length model. The problem
is of courseto determine this length. In practice there is not such a single
length scalè, but this model' gives a conceptual physical background to the turbulence model of eq 310
Chapter 4
Flow Calculations without.
Wave
s
Objective:
Review of the use Qf calculations for resistance and flow No extensive mathematical formulations are used, the purpose is to be able to use the available programs intelligently.A complete description of the flow around an arbitrary body is given by the Nàvier-Stokes equations, provided that at high Reynolds numbers a so-lution can be found for the formulation of the eddy viscosity, the ficticious viscosity which is caused by the turbulence in the flow. The flow around a ship hull can be found in priciple by applying the no slip condition at the 'wall and by applying the appropriate boundary conditions at the free sur-face. The free surface conditions are twofold. The first is the kinematical free surface condition : the fluid. particles at the free surface have to remain
at that surface The second is the dynamical free surface condition : the pressure at the free surface is always equal to the atmospheric pressure. In principle this problem is well fórmulated' and using a volume discretization of the fluid it can be solved.
The solution of this problem is still not really feasible. One reason is that the viscous phenomena require a very fine grid because the sçale at which energy is dissipated is very small. On the other hand the scale of the waves at the free surface. is large. In principle a large flow region with a very fine
52 GKuiper, Resistance and Propulsion 2, January 29, 2003
volume grid has to be used for the solution and this is still beyond the present computing capacity. Another reason is that the free surface is a part of the solution. It is not known beforehand where the, free surface boundary condi-tion has to be applied.
Programs using either volume discretizations or surface panelling there-fore contain essential simplifications. Only when the implications of these simplificaticns are understood a proper use can be made Of Computational Fluid Dynamics (CFD).
The first and basic simplification is that the regions of viscous flow and
the regions of potential flow with a free surface are separated, This is an
assumption comparable with the Froude hypothesis, which separates viscous and residuary resistance.
By far the most important 'calculation 'retbods are those which calcu- " '
late the inviscid outer flow. The presence of a thin boundary layer makes it possible to regard the flow outside the boundary layer as a potential flòw
and the pressure at the wall i equal to the pressure at the outside of the thin boundary layer. In that case the flow around a ship hull can be
re-gardad as a potential flow, because the viscosity can be neglected and as a
result no rotation is generated. The uthforrn inflow in front of the ship is
albo irrcitational, so the fiöw around: the 'ship can be described by the Laplace equation. The boundary conditions are the free' surface conditions 'and tan-gential flow along the outside of the boundary 'layer. Since the boundary layer is considered thin this can be approximated by tangential flow along the hull.
The viscous region can in principle 'be calculated using 'boundary layer equations, because the boundary is considered to be thin. Efforts have been made to use two-dimensional boundary layer equations along streamlines The boundary layer is' considered two-dimensional when no cross flow per-pendicular to the stre rnkne occurs. The streamlines and the pressure distri-bution along 'the streamlines have to be found from the potential calculations of the outer flow. The results of such calculations give a distribution of lo-cal friction coefficients along the streamline integration of the longitu'di:nal component over the hull gives the resistance.
aver-G.Kuiper January 29, 2OO, Flow calculations 53
UMIT1NG STREAMLINES
Figure t L Longitudinal, separation at the bow
t
t i
F
age ship hull the effects Of cross-flow and separation are too 'importarït to be ignored, and the presence of a thick boundary layer in the stern region has too much influence on the rsistance Cross-flöw is too important to be ignored. Three-dimensional boundary layer calculations are complicated and they blow up in regions of separation, where vortices leave the hull. In those regions the basic assumptions of boundary layer flow are violated. Until now no reliable calculatiOn methods are available to calctilä.tê the resistance. Ef-forts to db this are madè using the full Naviér-Stokes solutions, as will be discused later in this chapter.
The separation of the viscous and the inviscid regions neglcts the interL action between the viscous flow region and the potential flow region. The
54 G.Kuiper, Resistance and Propulsion 2, January 29, 2003
main region where problems occur is the wake, where the flow is 'highly ro-tational and viscosity cannot be neglected. Still it is a part of the potential
flow region. But also elsewhere on the hull where any type of separation
occurs, such as at the bilges in the forebody (see Fig. 4.1) the assumptions are viohited. Also when the boundary layer becomes very thick, as in some regions in the aftbody, the solution of the potential theory will be inaccurate. Keeping in mind these restrictions the potential flOw calculations can be used intelligently to optimize hull. forms or to calculate flow patterns which are difficult to measure at modél scale.
4.1
Potential Flow Calculations
4.1.1
Panel Methods without
ee Surface
The simplest category of calculations of the flow arouiid the ships hull are the potential flow calculations without a free surface. The undisturbed free surface is assumed to be unchanged by the flow around the hull. in practice this is realised by mirroring the ship hull at the undisturbed: free surface. This socalled: double hull is used for flOw calculations in an unbounded fluid. Because the undisturbed waterline is a plane of symmetry, the velocities nor-mal to that plane will always be zero. A drawback is, of course, :that twice the amount of panels have to be used to cover the wholé double body.
When no lift is present the singularity distribution used is a source
dis-tribution. The sources are distributed on the discretized hull surface. The
elements of the hull surface are called panels. As boundary conditions the condition of tangential flow at the ships hull, is used. This condition is ap-plied at the center of the panels, the so-calléd control points.
The simplest panel shape is a quadri1ateral flat plane , on which a uniform
distribution of sources with constant strength is placed Tins implies that
the distribution of the sources is a step-function instead of a continuous dis-tribution. This approach, first applied by Hess and Smith [14], approximated the shape of a ships double hull with a large number (N) of such flat panels. On each panel a umform source/srnk distribution is placed with an unknown
source strength Q. The potential function (x, y, z') can then be written
as the sum of the potential functions of all elements. The velocity at an
C.Kuiper January 29, 2003, Flow Calculations 55 The influence function of the own panel is singular,, but the singularity 'can be integrated over'the.panel.So after some'mathematics the velocity in an ar-bitrary control point can be expressed as a linear function of the N unknown source strengths Qi..
Next the boundary condition of tangential flow is applied at eaçh control point. This results in N boundary conditions 'for the N panels. This system can be solved, resulting in the strength Q at every panel. The velocities at every position at the ships hull and around the ship' can then be found 'from
the derivative of the potential function 1(x, y, z) in the desired direction
Note that the panel size limits the accuracy of the derivative, because the' derivative has to be determined from the difference between two panels. The pressures can be derived from the velOcities using Bernoulli's law.
It is. important to realize the consequences of the simplifications made. First it is potential' flow, so the fluid is inviscid'. That means that necèssar-uy the resistänce is zero (Paradox of d'Alembert) and .that flow separation
does not occur. Just as in the case of a cylinder .the pressure recovery in
the aftbody 5 complete. The water surface is also undisturbed, so the wave resistance is also zero.
How can' the results of such calculations be used? They can be used to assess qualitatively the relative merit of various alternatives and therefore to optimize the hull before model tests or further calculations are carried out. They can also' be used to indicate', improvements, because the calculations provide data such as pressure 'and velocity distributions, which are not mea-sured Examples are giveu below.
4.1.2
Assessment of Various Bulb Designs
in Fig. 4.2 three forebodies are shown of a full bulk carrier. [23] For these
designs panel calculations were carried' out, The results are shown in Fig. 4.3 The dotted lines in 'this Figure are lines Of equal pressure coefficient. Since the pressure coefficient 'Cp is dèflned' as
i2, a
low (highly negative) value of C means a low pressure on the hull.A bulbous bow is 'used to minimize wave resistance, but excessive flow separation has. to be avoided as well, since this increases the residuary resis-tance' again. 'The risk of flow separation is largest whèn the.pressure gradients
56 G. Kuiper, Resistance and Propulsion 2, January 29, 2003
SHIP Pi
SHIP P2
Figure 4 2 TwoBulb Configurations for a Bulk_Carrier
are large. In this case, the pressure gradients are largest when the minimum pressure is low. From the calculations this mmi mum pressure occurs at frame
19 near the bottom. The location does not vary significantly between the
alternatives. The fórebody with the smallest risk of flow separation is the one with the smallest perturbation of the flow, so the :011e with the highest minimum pressure coefficient. Ship Pl has a minimum pressure coefficient of -0.35, which is lower than that of ship P2. This indicates that ship. Pl has a potentially higher residuary resistance.