Sie sind auf Seite 1von 36

21 227 MARINE HYDRODYNAMICS

DR P G SAYER

CONTENTS

Introduction Approach to mathematical modelling Continuity equation for incompressible flow Fundamental modelling tools - stream function Velocity potential and irrotational flow Circulation and vorticity Sources and sinks Doublets/dipoles and geometrical transformations Vortices and propeller flow Lift

1 3 4 8 10 12 14 17 19

P G Sayer NAME P G Sayer NAME

Boundary layers

NAME

22 24 26 26 28 29 31 32

Boundary layer velocity profiles

Displacement thickness, momentum thickness and skin friction Boundary layer thickness

NAME

Skin friction resistance in naval architecture Bibliography Tutorial exercises

MARINE HYDRODYNAMICS
Introduction This is a course involving applied mathematics!!....... Mathematics applied to marine hydrodynamics for naval architects. By marine hydrodynamics I mean the dynamic properties and characteristics of water, either on the large scale where we might wish to examine the flow of water along a ship hull and past the propeller, or on a smaller scale if we are looking at the flow of water (or some other fluid) through pipework or a piece of machinery. So there will be relevance to both naval architects and marine engineers. Naturally you may ask whether there is a need for yet another mathematics course. After all, surely naval architects and marine engineers simply need tools to do the job. The job in this case might be the prediction of how a ship performs in a given condition, how safe it is, how much power is needed, and similar factors. In a sense, maybe we just need to consider safety and economics. Whether you accept this view or not, some means is needed for quantifying the relevant criteria and data. Think for a moment about what most engineers take for granted when using Newtons laws of motion, or some similar equations or techniques fundamental to mechanics. Inherent in these equations are the principles of vectors (force and momentum), calculus (differentiation for acceleration and integration for positions) and all the underlying mathematical properties. Although we might perfectly happily treat the solutions of these equations via a black box , e.g. some software package, what happens if something goes wrong? - Invalid data, lack of convergence of some algorithm, etc. We need to be able to identify not only when something goes wrong, but also when something is likely to go wrong. This can only be achieved if we understand the fundamental mechanisms which we are using to try to solve the particular problem. At an earlier stage still, the greatest difficulty facing many people is how to formulate the physical problem in mathematical terms. In the case of ships or other marine vehicles, the laws of motion are complicated by the details of the excitation forces - the forces arising from the presence of the water (the hydrostatic and hydrodynamic effects) are complicated to compute because of the double difficulty of determining the pressure due to the waves and currents and the effect of the waters viscosity. Sometimes we can safely neglect the viscosity; sometimes we cannot, e.g. in the case of ship resistance. An engineer needs to know when various factors must be included and when they can be neglected; otherwise the governing equations become
2007 P.G. Sayer

Page 1

unmanageable in practice. It may be difficult to compute the solution and also, importantly, to enable the naval architect to get a feel for what is happening physically in the real situation. This course therefore introduces the fundamental characteristics of the flow of water and how forces are then exerted on a marine vehicle, by expressing the relevant physical properties in appropriate mathematical terms. Only then will it be possible to build on this knowledge and understanding to model ship motions. At each stage, we shall try to identify and explain any approximations and simplifications which can be made, or indeed which might be necessary. It is important to make clear at an early stage that engineers and scientists know the general equations that govern the flow of water: the so-called Navier-Stokes equations developed in the nineteenth century tell us about the way a lump of fluid (water) behaves when it is subjected to pressure forces (e.g. due to waves or current on a ship hull; or the action of a propeller or pump). The main practical difficulty, however, is that it has only become possible recently, with the advent of extremely powerful computers, to solve the equations with acceptable accuracy for many realistic situations in naval architecture and marine engineering. But such computational facilities are very expensive, not widely available and, importantly, we often do not need the level of detail or accuracy produced by this sophistication; it is a luxury that we do not need! Instead, we wish to gain more insight into the importance of various parameters and their effects (trends). Lots of factors complicate the real world, and often we have to decide what assumptions and simplifications are needed, and whether these are acceptable in a particular context. For example, most of us would perhaps accept that water is a non-viscous, incompressible fluid; in many situations this would be perfectly acceptable.......but not if we wish to study the resistance of a ship hull, for which a large contributing factor is the skin friction; this depends on the thin region of water adjacent to the hull where viscosity is important, and which gives rise to a boundary layer where our fluid lump is subjected to large shear forces caused by rapid changes in flow velocity across the boundary layer. Fortunately, the incompressibility of water is a very accurate assumption except in the very deepest parts of the ocean, where at a depth of 10,000m the ambient pressure will be around 1,000 atmospheres (108 Pa). [Interestingly, in aerodynamics, the compressibility of air usually only has to be allowed for at speeds greater than Mach 0.7]. Many of the fundamental concepts and properties of fluid flow are common to all branches of engineering and science. Bernoullis equation, for example, has very widespread use. But it has several forms, depending on the nature of the

2007 P.G. Sayer

Page 2

pressure forces giving rise to the flow of the fluid. We shall always attempt to emphasise how such principles relate to the marine context, and in doing so give equations and results in forms which tend to be most useful for naval architects. After these fundamentals of fluid flow, we proceed to introduce the theory of water waves. It turns out that the presence of a free surface (the so-called air/sea interface) is the source of many theoretical complications! To make a start, however, we shall adopt a widely-used first approximation and consider linear theory, in which the actual free surface is effectively replaced by the mean undisturbed water level, in other words by a horizontal surface. Nevertheless, this simplification turns out not only to be important in a systematic development of the theory but also leads to (surprisingly!) accurate results for many real-life situations. Only in the most extreme seas, and/or where largeamplitude motions of a ship may be expected, does the linear theory break down...but that, too, is beyond the scope of an introductory set of notes such as these. Approach to mathematical modelling We have already mentioned that water can be safely regarded as being incompressible. Now, how can this property be expressed mathematically? Quite often, this is the time when most students start to become alarmed!! As a general rule, we should try to clear in our minds the physical properties of the flow; in other words, either what is happening to our lump of fluid as it moves, or how the fluid moves with respect to a fixed point (or region) upon which we focus our attention. [The former gives rise to what is known as the Lagrangean description and Lagranges equations, while the latter leads to the Eulerian description and Eulers equations, in honour of these two eminent scientists]. The streamlines are the paths traced by the fluid particles; in other words, the fluid velocity vectors are aligned with the streamlines If we look at a region of incompressible fluid, then obviously however much fluid flows into this region must be matched by an equal amount flowing out. Otherwise, there would be a net build-up of fluid, which would contradict our assumption of the physics, namely the incompressibility criterion. To fix ideas, much of this course will concentrate on flow in two dimensions although an extension to the three-dimensional case is normally straightforward. [I say normally because, although the governing equations may become quite complicated, and correspondingly more difficult to solve, the underlying physics, and hopefully our understanding of the physics, is not affected. In three dimensions, it is often easier if we re-formulate the equations in terms of vectors and matrices]. We will need to set up our equations and formulae

2007 P.G. Sayer

Page 3

according to two sets of co-ordinates: Cartesian co-ordinates for the most part, but sometimes, especially for those flow descriptions which inherently involve circular-type motion (flow near a propeller, for example), it will be more convenient to adopt polar co-ordinates; once again, we choose the tools (in this case, the co-ordinate system) best suited to the job. Of course, co-ordinates other than Cartesians and polars could be used. Cartesian co-ordinates, however, are probably the best to start with. We shall use x and y, although in order to conform with the symbols and conventions widely used by naval architects, the two-dimensional study of water waves tends to use x and z (for horizontal and vertical directions respectively). You must not be put off by this!! Concentrate on understanding the approach to the modelling process; this should always take priority over the detailed mathematics. Forgive the digression, but this is an important point ... in a practical (work) situation, it will usually be possible to call on the services of someone with the appropriate expertise (a mathematician, for example) to solve some equations, but often a far greater and more important challenge is to determine which approach, model or set of equations is needed to formulate the real physical problem in mathematical terms. Continuity equation for incompressible flow Returning to our incompressible flow, we shall use u and v to denote the velocity components of the flow in the x and y directions, respectively. In mathematical terminology, we are saying that u is a function of x and y, namely u=u(x,y). Likewise for v. In Cartesian co-ordinates, it is easiest if we consider changes only along the co-ordinate directions, which in turn suggests that we should consider a rectangular region of fluid, to which we need to impose the equality of inflow and outflow from the physical statement of incompressibility. Note that we are at liberty to choose an arbitrary region of fluid to which we apply physical principles; the laws of physics must apply to all shapes, and so it is sensible to choose something that helps keep the mathematics as simple as possible. In standard mathematical style, the x-component of velocity at a horizontal coordinate x+x will be denoted by u+u; similarly the y-component at a vertical location y+y will be written as v+v.

2007 P.G. Sayer

Page 4

The inflow across the sides of the rectangular fluid region, of length x and y, is easily seen to be u y + v x. The outflow is (u + u) y + (v + v) x. Now from calculus, !u =
"u "v !x , !v = !y , which leads to the result "x "y !u !v "x "y + "y "x = 0 !x !y

The area of the rectangle is x y and this is non-zero. We conclude that


!u !v + =0 !x !y

This is referred to as the continuity equation. In practical terms, we shall usually employ Cartesian co-ordinates, certainly if we wish to consider the flow along a ships hull. Although you may correctly observe that such a flow is evidently three-dimensional in nature, we could easily extend our two-dimensional model to successive water-lines, and thereby account for the vertical (third) dimension. However, when looking at flow through or near a propeller, or near a bilge keel as a ship rolls, the resulting vorticity (see later) shows itself as a definite rotation of flow, for which polar co-ordinates are sometimes easier to work with. For r and , therefore, where x = r cos and y = r sin , the corresponding equations for the velocity components ur and u are:
ur =ucos! +vsin! , u! = " usin! +vcos!

2007 P.G. Sayer

Page 5

By similar reasoning as used for Cartesian co-ordinates, the incompressibility condition leads to the equation
! (rur ) + !u" = 0 !r !"

The speed (i.e. the magnitude of the velocity) is sometimes denoted by q; therefore q2 = u2 + v2 = ur2 + u2. The local angle that the streamline makes with the x-axis is tan-1(v/u) = + tan-1(u/ur).

So, we now know a mathematical constraint that the velocity components must satisfy for an incompressible fluid. We progress by considering what causes the fluid to move in the first place. Our lump of fluid will move only as a result of some force applied to it, either externally or internally. For the present, we consider the simple case of a fluid moving purely by the difference in pressure from one point in the fluid to another. We can apply Newtons second law directly: let us consider the case of motion along the x-axis. Denote by p the

2007 P.G. Sayer

Page 6

pressure acting on the side of the rectangle having horizontal co-ordinate x; following the same terminology as above, p+p denotes the pressure at x+x. Then, since pressure acts equally in all directions (it is a scalar quantity), the fluid lump experiences a (vector) force +py from left to right over the left-hand edge [since force equals (pressure x area), which in two-dimensions is effectively pressure x length], together with a force -(p+p)y over the righthand edge. Now, pressure is a function of both x and y; but since the only difference in co-ordinates between the left-hand and right-hand edges of our fluid region occurs in the horizontal (i.e. x) direction, we make use of calculus to express p as
!p x. !x

Since the acceleration of the fluid in the horizontal

direction (positive in the sense of positive x, i.e. left to right) is u or

equating the net force in the positive x-direction to the (mass x acceleration) of the fluid region we have:

!u then by !t

p( x )!y $ p( x +!x )!y = ( "!x!y )

#u #t

This equation can be integrated to give one form (the simplest!) of Bernoullis equation for fluid flow in the absence of external forces such as gravity (equivalently, flow in a horizontal plane). The integration is best done by noting that the acceleration can be re-written as u u/x, so that we can integrate both sides with respect to x. The above equation becomes -p/x = u u/x

2007 P.G. Sayer

Page 7

which integrates to give the Energy Equation, or Bernoullis Equation (more precisely, the simplest form of Bernoullis Equation):
p+ 1 2 !u = constant 2

The constant is the same for all fluid particles following the same streamline, and so we can generalise the above to give
p+ 1 2 !q = constant 2

Fundamental modelling tools - stream function Let us pause and consider the information that we now have: the velocity components u and v satisfy the continuity equation; and the pressure and the two velocity components satisfy Bernoullis equation; so we have two equations for three unknowns, which does not enable us to solve for p, u and v separately. At this stage it is appropriate to introduce two new concepts: stream function (which is a physical measure of fluid flow, somewhat like a meter which measures the flow-rate of the water) and velocity potential (which is a rather abstract mathematical quantity, but has some direct similarities with potential energy in mechanics). Both turn out to be very important in the sense that they provide fundamental information about the flow, from which the physically meaningful quantities such as pressure and velocity can be obtained (and, of course, once the pressure is known, it can be integrated/summed around the surface of a body to get the total force acting on the body). Which of these two functions we use again depends on which one is better suited to the constraints of the real situation we are trying to model. For example, although stream function relates more closely with the physics of the problem, and so may well be the first choice for most people, it can only be defined in two dimensions. This causes no difficulties just now (nor indeed in the twodimensional study of water waves later) but when we wish to extend to a full three-dimensional model then other ways have to be sought. On the other hand, velocity potential can be defined for both two and three dimensions, but (and this is a crucial point) it is defined only for so-called potential flow (hence its name!) which means that there is no rotation of the flow. Rotation, and the related circulation will be precisely defined later; for now, it is sufficient to know that these terms relate to the twisting or spinning of the flow particles. Returning to the physics of what is happening, such rotation stems from the action of viscosity within the fluid...so, although boundary layers in water are

2007 P.G. Sayer

Page 8

thin, they are nonetheless responsible for generating circulation within the flow, which in turn modifies the pressure and velocity distribution compared to what would exist in the absence of boundary layers. If all this seems too abstract and perhaps unnecessary, let us not lose track of the purpose being modelling the hydrodynamic flow characteristics. We ultimately wish to calculate the forces acting on marine vehicles and structures. In a laboratory experiment, we could mount the model on a sensitive balance and measure the forces or moments directly. Alternatively, particularly if we also wish to learn more of the characteristics of the flow as it passes the model, we could measure the fluid pressure at a number of points around the surface of the model. In practice, either at full-scale or in a laboratory experiment, we could measure fluid pressure using a Pitot tube (or pressure transducer) and the fluid velocity using a current meter for water or a hot-wire anemometer for air speed in a wind tunnel (or, in a more sophisticated way, by means of laser anemometry). So, for both theoretical model and laboratory experiment, we need to measure flow pressure and velocity. Regard this new concept of stream function, therefore, as a device to measure flow velocity in our mathematical model. The S.I. units of the stream function are m2s-1 (equivalent to volume flow per metre width of fluid). Consideration of the flow in the diagram below enables us to define the velocity components u and v in terms of x- or y-derivatives of the stream function. Based purely on a dimensional argument, we can obtain the units of velocity (ms-1) by dividing those of the stream function by length; which is equivalent to a spatial derivative with respect to x or y. The symbol is usually used to denote stream function; whether it is positive or negative depends solely on how we define the positive direction of flow; so the definition is arbitrary but once we have chosen a particular sign convention then we must stick to it! (There is no standard convention in the literature and there is no need to worry about this ambiguity in sign). If we can determine then we can calculate u and v by differentiation, then p by Bernoulli, and so the hydrodynamic force.

If we consider the flux of flow between two neighbouring stramlines, then


2007 P.G. Sayer

Page 9

continuity (i.e. equating inflow and outflow) gives


!" = u!y # v!x

Noting from differential calculus that


!# = "# "# !x + !y "x "y

we can equate the coefficients of the x and y terms to obtain


u=

"# "y

and

v=!

"# "x

By considering the flow in a wedge of fluid, comprising arcs of a circle, we can derive the following corresponding results in polar co-ordinates:
!" = rur !# % u# !r
ur = 1 $" $" and u# = % r $# $r

As a simple example of a stream function, we consider for a uniform stream flowing at speed U in the negative x-direction. So u = -U and so trivially it follows that = -Uy. Why is there no integration constant? Think about this; we shall return to it in a moment. Velocity potential and irrotational flow As mentioned above, the concept of velocity potential is somewhat more difficult to grasp. Its applicability to three dimensions, however, together with the fact that there are many situations where we can neglect viscosity, means that it does play a vital role in hydrodynamic modelling. It is usually denoted by the symbol (and once again whether it is defined as a positive or negative quantity is quite arbitrary, though having selected a sign convention fixes everything thereafter). Its units are those of (velocity x distance) and so it is clear that its spatial derivative will generate a velocity. An important point to note is that the definition and value of at any particular point in the flow is completely independent of the path taken by the flow in reaching that point. Path-independence turns out to be a key property in defining the potential energy of a body subjected to the force of gravity (where
2007 P.G. Sayer

Page 10

we implicitly neglect other forces, such as air resistance, in comparison). The only mathematical difference between potential energy and velocity potential is that potential energy has the units of (force x distance) whereas velocity potential has units of (velocity x distance). [Note that for distance we could equally well write length]. The relationships between u and v and are given in the following diagram.

The change in the velocity potential is seen to be u x + v y. We then adopt the same reasoning as for the stream function by using calculus to express as (/x) x + (/y) y and then equating coefficients:
u=

!" !" and v = !x !y

Using the same technique but without going through the details, the corresponding formulae in polar co-ordinates are:
ur =

!" 1 !" and u# = !r r !#

Before we continue to see how and are used in practice, we need to remember that a body in general motion (and this includes the motion of a fluid particle) can be split into translational components as well as rotational components - you can perhaps relate these to a ball hit through the air during its flight, the path followed by its centre of gravity represents the translational motion, whilst any spin about its centre of gravity corresponds to rotation, in every sense of the word! In the absence of external forces, we have already noted that only the fluid viscosity can generate shear forces, which in turn cause the fluid to spin. So for a non-viscous, or inviscid, or frictionless flow, there is no rotation and so a velocity potential can be defined. In this case we use the term irrotational flow. How we actually calculate may be somewhat
2007 P.G. Sayer

Page 11

complicated! Nevertheless, many textbooks are available which give or for particular situations. We shall see that wherever both and are valid, then one can be easily derived from the other. So, if both exist, it makes sense to know which is the easier to derive and work with. Naturally some worked examples will be given in class but, more importantly in some ways, I want you to remember how we can derive or for simple case and then build up more complicated fluid flow models/pictures by using the simpler examples in a kind of building-block approach. No-one should expect to be able to spot the form of or for complicated flows; moreover, it is not really helpful just to be able to memorise a formula from a book. The real advances are possible only when the student understands how to tackle the formulation of a problem. Circulation and vorticity If a velocity potential can be defined (we sometimes say exists) then the characteristics of the flow are path-independent. Therefore, if we consider two points A and B in the fluid, then the value of at B is not only independent of the path taken by the flow in moving from A, but also will return to the same value at A no matter how we travel from B back to A. [This agrees with the definition of potential energy: in raising a body between two levels, the change in potential energy is independent of the path taken between the two levels; furthermore, there is net zero change in potential energy if the body returns to its original level, regardless of the path taken to get there]. But, for a general type of fluid motion, there may be a difference in the velocity and pressure which do depend on the paths taken. We therefore define the circulation to be the integral of velocity around any closed path (there is another criterion which is important mathematically, namely that the path must not enclose a flow singularity, i.e. a singular point in the mathematical model of the flow).

2007 P.G. Sayer

Page 12

If we divide the expression for the circulation by the area enclosed by the path taken, the terms disappear and we are left with what is termed the vorticity of the flow. [In lay terms, most of you will be aware of flow down a bath plughole, or the flow in a tornado, as being like a vortex, or as having vorticity; so this vorticity measure the spin of the flow and clearly will be important when we try to model the flow past a propeller, impeller pump or similar device].
!= "v "u # "x "y

So, if the flow is irrotational, then the vorticity is zero. You may be tempted to assume, therefore, that if the vorticity is zero then the flow must also be irrotational. This turns out to be true in three dimensions, but not for the twodimensional problem. The explanation lies in the mathematical result, known as Stokess theorem, which relates the strength of the vorticity (or vortex tubes) to the cross-sectional area of the region within the closed path used to define circulation. Path-independence turns out to require that the irrotationality result must hold for any closed path, regardless of its size or shape; in particular, as the length of the path and the area within it shrink to zero. This can be done for all paths in 3-d, but not in 2-d (e.g. consider a path surrounding a 2-d circular
2007 P.G. Sayer

Page 13

cylinder). In practice, though, we are attempting to model a fully 3-d situation, so the 2-d special case tends to be mainly of theoretical interest. Do not, however, incorrectly suppose that movement of fluid along a circulartype path must imply vorticity is present (regardless of whether we are looking at a 2-d or 3-d model). Vorticity is a local property of the motion of a lump of fluid and measures how the lump of fluid spins about its centre of gravity. The centre of gravity of the fluid can still follow a circular path while not spinning this is irrotational motion (and so can be described by a velocity potential). A practical example of this will be illustrated later by the flow past (or through) a propeller. However, before we demonstrate how to bring together the relevant mathematical ideas and concepts, it is instructive, and also allows us to follow a more systematic route, if we introduce some other modelling tools. Sources and sinks A source is a mathematical device which causes flow (again mathematically) to be generated at a singularity (where the source is located) and projected outwards uniformly in the radial directions. A sink has simply the opposite effect; it is a negative source. The fact that such singularities do not exist in reality is not important. Provided the mathematical model generates the correct physical behaviour of the flow, in the region of the flow where it is needed, then we need not worry about these mathematically singular points. We have seen that a singularity exists at the centre of our mathematical model of a vortex; the fact that a singularity (where something tends to infinity) does not occur at the centre of a real vortex (viscosity prevents infinite values, even though large finite values can still occur) means that we must limit the range of validity of our model. Therefore, we have to realise that the mathematical model must break down before we reach the origin of the singularity (or the eye of the vortex). We cannot expect to model everything using relatively simple tools!

2007 P.G. Sayer

Page 14

The stream function for a source can be derived by considering the uniform radial flow that it generates outwards from the origin. Denoting its strength by m means that ur = m/(2r) whereas u = 0. Using the formulae for velocity components in terms of derivatives of leads to = m/(2) Should there be a constant of integration, or some other function? The answer is that we can set constants of integration arbitrarily to zero, without loss of generality, because the physically relevant quantities, namely velocity, acceleration, pressure, etc. depend only on the derivatives of . The next question, then, is probably: What use is a mathematical source? The answer lies in the velocity that it imparts to the fluid. Since velocity is a vector, it has both magnitude and direction. A flow model can be thought of as a means of determining the velocity components at every point in the fluid. So in any locality, we might compute the fluid velocity components and represent them as arrows, the length of the arrow corresponding to the magnitude of the velocity, with the arrow aligned with the local direction of the flow. In such a vector diagram, we can make use of standard vector algebra: two or more arrows can be combined to find the resultant vector. In this way, it is possible to imagine the vectors cancelling each other, or reinforcing each other. In the former case, the resultant velocity can become exactly zero; this occurs at a stagnation point. From Bernoullis equation, if the velocity is zero, then the pressure is a (local) maximum. Extending these ideas, a source (or line of sources) can deflect an incident flow to such an extent that we could envisage the flow lines tracing out the lines of an actual ship hull. If this can be done, then our model lets us take a next step along the route of modelling the velocity (and hence pressure) distribution along a hull, and so to the net force acting on the hull. [Remember, though, that this non-viscous model can only partly account for the force acting on the hull; more precisely, this model leads to what is known as the form drag; there are other important components, for example skin frictions which depend intrinsically on the viscosity of the boundary-layer flow and the hull roughness]. Again, examples in class will be calculated to show how these ideas are implemented in practice. One of the simplest examples is that of a source place in a uniform incident flow. This results in the incident flow being deflected laterally, before continuing downstream parallel to its original direction. We can extend this in a simple way by considering a source and a sink of equal strengths, giving flow around an oval-shaped body. The lateral deflection of the flow, or equivalently
2007 P.G. Sayer

Page 15

the maximum breadth (beam) of the body, can be calculated directly as a function of the source strength and incident flow speed. Conversely, given the breadth of the body, we can calculate the required source strength to model the correct flow behaviour.

In the case of an actual ships waterline, its shape is rather more complicated than the example just given! A large number of sources are needed, their strengths being determined from the condition that the deflected flow must follow the waterline itself. This leads to a set of simultaneous equations for the source strengths. So, for practical purposes, the ships beam and the other waterline offsets form the input, while the solution of the simultaneous equations generate the source strengths as the output. A major weakness of the non-viscous model is that it would appear that a body that has fore and aft symmetry will have zero net force acting on it. The situation is even worse!! This model predicts zero force on a body of any shape when subjected to a steady flow: this is known as dAlemberts Paradox. Its resolution lies in realising that the model neglects viscosity, which means that no separation of flow occurs after the flow encounters the body and therefore that no wake is generated downstream of the body. In reality the separation of flow results in an imbalance between the fluid pressures fore and aft, and hence to a non-zero net force in the direction of the flow (put simply, this is called drag!). [We have also neglected any wave generation that may occur - this also leads to a component of drag, independent of viscous effects]. So, once more, we need to appreciate the limitations of our models, while at the same time building on what we can already model accurately. We must do things gradually. We must walk before we run!

2007 P.G. Sayer

Page 16

Doublets/dipoles and geometrical transformations Intuitively, the effects of a source and a sink should cancel out. But think more carefully about the vector arrows, and what happens if the source and sink are brought close to each other in such a special way that the product of their strength and the separation distance remains constant, i.e. the strengths increase in inverse proportion to the distance between their centres. The result is a doublet or dipole. What do the streamlines look like?

The expressions for a doublet or dipole are as follows:


! = sin# 2"r cos# 2"r 2 sin# 2"r 2

ur =

u# =

If we consider the streamlines for a source and a sink, then there is much in common with the lines of force of a bar magnet. But as the source and sink coalesce, without annihilating each other, then so the streamlines bend more and more to form circles. If we take things one step further by putting the doublet in a uniform stream then the zero streamline (across which flow cannot pass; and so physically can be viewed as an impermeable barrier) is a circle. We simply add the stream function for a uniform stream to the stream function for a doublet: = -U (1 - a2 / r2 ) r sin , where we have written a2 = /(2U)

2007 P.G. Sayer

Page 17

For offshore engineers, the flow past circular sections is commonly encountered, e.g. platform legs. But mathematical modelling turns out to have far greater scope. By using conformal mapping techniques (a special kind of geometrical transformation) many flows around complicated shapes can be reduced to flows around much simpler shapes, including circles.

In the field of ship hull generation, the 1930s saw the introduction of Lewis transformation techniques, in which several typical hull cross-sections could be mapped into semi-circles, and hence for which flow characteristics could be determined analytically (just as well, in those pre-computer days!!). Earlier still, in the 1900s, Joukowski showed how flow around foil sections could also be reduced to flow around circles or ellipses. This led to pioneering developments in the design of aircraft wings for several decades afterwards. Much more recently (but still before the arrival of personal computers) the same ideas were extended to hydrofoils and then to sails, the latter being somewhat complicated examples of thin, cambered and flexible foils. Little wonder, then, why much advance in the design of yachts has had to be done empirically and via extensive physical model testing. Although I have mentioned previously that the governing equations of fluid motion, including the effects of viscosity, have been known for over a century, a detailed study of viscous flow past a single circular cylinder was only feasible computationally in the 1980s. To put matters into perspective, it is still impossible to predict mathematically the time history of droplets of water as they drip from a tap! [Nor, incidentally, can we predict the eventual configuration of, say, a piece of string which is held vertically by one end and then dropped to the ground......and there are many other examples in many other fields].

2007 P.G. Sayer

Page 18

Vortices and propeller flow The streamlines of a vortex are circular. To derive the stream function (or the velocity potential) we use elementary mechanics and consider the motion of our fluid lump following a circle path. Polar co-ordinates are appropriate and we consider the resultant of the pressure acting over the four edges of the lump and the acceleration towards the centre of rotation u2 / r.

Newtons second law leads to a fairly simple equation:


dp !u" 2 = dr r

which also shows that pressure increases as we travel outwards across the vortex. In irrotational flow, the energy is constant along a streamline, and so we can make use of Bernoullis Equation in conjunction with the above vortex pressure equation to obtain
u! = c p # p1 c2 $1 1 & and 2 = 2 # r " 2 r1 r2 2 %

Such a motion is what we term a free vortex.

2007 P.G. Sayer

Page 19

In contrast, for a forced vortex the fluid rotates rather like a solid body; so for angular velocity we have u = r and, of course, ur = 0. The pressure satisfies
p2 ! p1 # 2 2 = (r2 ! r12 ) " 2

Exercise: Show that the stream function for a free vortex is given by
! =#
$ r ln 2" a

where is the circulation and is defined to be zero on the circle r = a. This

2007 P.G. Sayer

Page 20

last condition is equivalent to saying that the fluid is bounded internally by a circle of radius a; no fluid can cross the surface of the circle, so in practice we could use this to model the flow around a circular cylinder of radius a. A more obvious marine application, however, is a ship propeller. The flow between the propeller shaft and the blade tips is acted upon by the blades, and so an external force (provided, of course, by the engine) alters the energy of that region of fluid. The resultant flow is rotational (and so only the stream function, not the velocity potential, can be used). In contrast, just beyond the blade tips, the fluid flow is irrotational so, although the centre of gravity of the fluid lump traces out a circular path, the fluid lump does not spin about its centre of gravity. The fluid as a whole still of course follows as circular-type path, and continues to do so for quite some distance away from the blade tips (the flow clearly cannot just switch off beyond the blade tips), but the vorticity in this region exterior to the propeller itself has no circulation; its total energy remains constant, and this means that Bernoullis equation can be used (in contrast to the inner region encompassed by the blades where Bernoullis equation is not valid). These physical differences between the flow interior and exterior to the propeller blades can be modelled by forced and free vortices, respectively. Since we cannot use Bernoullis equation in the forced vortex case we need to employ other results of mechanics (analogous to rigid-body rotation) to make further progress. Beyond the blade tips, however, the free-vortex motion can be determined using the Bernoulli relation between pressure and velocity. Without going into the mathematical details, we can still make some initial observations which turn out to serve as boundary conditions; these enable us to determine, for example, constants of integration when we solve the equations, as well as checking that the pressure and velocity changes tend to zero as we move away from the propeller. It is always sensible to carry out such checks. There is no point at all in striving to solve complicated equations if the equations themselves do not satisfy fundamental conditions. As a specific example, the equations for the pressure and velocity in the inner and outer regions of a propeller must generate equal values of the pressure and equal values of the velocity at the boundary between the inner and outer regions; i.e. the two models must agree at the circle described by the tips of the propeller blades. Lift As a final illustration of our combined Lego-brick modelling we examine the flow past a spinning circular cylinder placed at right angles to a uniform stream of speed U. Mathematically, this is achieved by adding together the flow due to a uniform incident stream, a doublet and a vortex.

2007 P.G. Sayer

Page 21

% a2 ( r ! = $ U 1 $ 2 ' r sin" $ ln r & 2# a % a2 1 ! ur = = $ U 1 $ 2 ' cos" r " r & % a2 ! )a 2 u" = $ = U 1 + 2 ' sin" + r r & r

On the surface of the cylinder r = a, we have ur = 0 and u = 2U sin + a, giving a stagnation point at
# &a ! = !stag = sin "1 " % 2U $

for which the pressure coefficient is


cp p ! p0 'a " = 1 ! 4 sin& + $ 1 2U # 2 %U 2
2

The following diagram shows the variation of cp with when = 0. What can we say when 0?

2007 P.G. Sayer

Page 22

From a practical viewpoint, the resulting transverse (lift) force which is generated (known as the Magnus Effect) gives a means of controlling the depth and orientation of an underwater vehicle or towed body, more efficiently than using flaps and thrusters, as well as increasing the turning effect of a rudder (in the latter case, however, there are more efficient ways of achieving the same end via flaps on the rudder). Whilst the expression for may appear complicated, you should recall that it comprises simply a summation of simple flow models. After going through the mathematical details, we can integrate the pressure distribution around the circle to obtain the net force. Of course, we must remember that force is a vector and so take into account its direction when doing the pressure integration; hence the sin and cos factors in the integrals. The final results for the in-line (drag) and transverse (lift) forces are interesting: the model still predicts zero drag (because viscosity and hence separation are not modelled) whereas the lift force is computed very accurately: its magnitude is U, quite a remarkably simple expression for a complex piece of flow modelling. In fact, the result is the same for any lifting surface: different shapes of bodies are accounted for by different values of . Symmetrical foil sections are widely used in marine applications. Examples include the NACA 4-digit series for rudders and the NACA 60 series for yacht keels. Some important characteristics of lift and drag are given in the following figure.

2007 P.G. Sayer

Page 23

Boundary Layers We have already remarked that all fluids have some viscosity. Although the viscosity of water is often neglected, we cannot do this in those narrow regions of fluid immediately next to a body, i.e. in the boundary layer. The resultant shear stresses give rise to a drag force exerted on the body. There are two basic types of flow, namely laminar and turbulent. Especially within the boundary layer it is important to determine the flow characteristics accurately; otherwise, the drag predictions can be seriously in error. For laminar flows we can make use of a simple relationship, known as Newtons Law of Viscosity. This relates the shear stress (i.e. force per unit area, or force per unit length in two dimensions) in the boundary layer to the viscosity and the rate of change of (tangential) velocity u/y as we cross the boundary layer at right angles to the surface of the body: = u/y The extent of the boundary layer occupies that region of fluid adjacent to the body in which the tangential velocity component decreases from its mainstream (undisturbed) value at the edge of the boundary layer, to zero (relative velocity) at the surface. This implies that the fluid satisfies a no-slip condition at the body surface, again an essential criterion if basic laws of physics are considered. The thickness of the boundary layer is usually denoted by . For convenience two more non-dimensional quantities are also introduced: the ratio of the boundary layer velocity component u to the mainstream velocity U, i.e. u/U, and a normalised boundary layer thickness parameter = y/. Therefore both parameters u/U and y/ vary from 0 at the surface of the body to 1 at the edge of the boundary layer. Note that in practice we need to place a definite, and sensible, limit on the edge of the boundary layer (which mathematically extends indefinitely); it is usual to define the edge of the boundary layer where u/U=0.99.

2007 P.G. Sayer

Page 24

The study of flow over a flat plate gives a very good insight into general boundary layer characteristics as well as having direct practical applications in naval architecture; for example, one important component of ship resistance, namely skin friction, is computed from extensive tests and empirical formulae based on flat plates. The boundary layer characteristics obviously depend on whether the incident flow itself is laminar or turbulent, but regardless of this, there will be a (short) region extending from the leading edge of the plate where the boundary layer flow must be laminar. Here, the effects of viscosity are so pronounced that the flow is forced to exhibit this laminar behaviour. In general, flows for which viscous effects are strong, or which have sufficiently low speeds (in fact, for sufficiently low values of the Reynolds number Rn ) have streamlines corresponding to layers of fluid slipping past each other, in a fairly well-structured manner. In contrast, turbulence characterises itself by flow instabilities and a subsequent mixing of the layers of fluid; energy and momentum are interchanged between fluid layers, across the whole boundary layer, and it is no longer possible to specify, or determine computationally, the exact, deterministic properties of the flow, such as velocity. We have to be content to use mean values, to which are added random perturbations. Further downstream, along the plate, inherent perturbations in the flow will trip the flow, changing it from laminar to turbulent. This occurs in the socalled transition zone. Even further downstream, where the boundary layer can be regarded as turbulent, there will nevertheless still be a thin laminar sub-layer next to the body; as before, the effects of viscosity very close to the body are sufficient to retain the laminar characteristics, and of course at the body surface itself there will be the no-slip condition. If x measures distances downstream from the leading edge of the plate, then the transition zone is located where the value of the local Reynolds number Rx = Ux/ lies between about 3x105 and 106. [For flow inside a pipe of circular cross-section, of diameter D, the corresponding transition value of Reynolds number is given by RD = UD/ 2300].

It seems reasonable that streamlining a body will help prevent separation of the
2007 P.G. Sayer

Page 25

flow and therefore this might be expected to reduce the drag or resistance. Of course, there are several components of drag, and excessive streamlining will actually tend to increase skin friction, at the expense of reducing form drag or separation drag. So, in practice, a compromise has to be sought. One of the most startling results for drag from classical fluid mechanics relates to the drag on a two-dimensional foil section compared to that for a circular cylinder. The separation drag for a circular cylinder of diameter D is at least an order of magnitude (a factor of 10) larger than that for the foil. Boundary layer velocity profiles (b.l.v.p.) Given our non-dimensional parameters defined above, we often make use of standard expressions for the velocity profile of the flow within the boundary layer; these are based on practical observations/measurements, although a few originate from theoretical arguments. The most frequently used expressions for laminar boundary layers are: u/U = 2 - 2 u/U = (3 - 3)/2 u/U = 2 - 23 + 4 u/U = sin (/2) For a turbulent boundary layer, the most widely used expression is u/U = 1/7 The thickness of the boundary layer, not surprisingly, affects the magnitude of the drag of a body. However, before we can derive expressions for this and other important results, we need to introduce two additional concepts, namely displacement thickness * and momentum thickness . Displacement thickness, momentum thickness and skin friction If we track a fluid particle as it enters the boundary layer near the leading edge of the plate, we observe that the streamline is deflected away from the body. Physically, the reason is quite simple: the fluid near to the plate is retarded by viscosity and so, to maintain continuity, all streamlines have to be shifted laterally outwards. For flow incident along a thin flat plate, we can view the
2007 P.G. Sayer

Page 26

displacement thickness as the difference between the original streamline spacing and the boundary layer thickness where the streamline penetrates the boundary layer.

By continuity, the volume flow (i.e. flux of fluid) between any pair of streamlines remains constant. So
!

U y = u dy
0

Then, noting * = - y and = y/, we have


!* = !
1

(1 # u / U ) d"

So, for example, the laminar velocity profile u/U = 2 - 2 leads to */ = 1/3, whereas the turbulent profile u/U = 1/7 gives */ = 1/8. To calculate the drag arising from skin friction (in principle this is the only component of drag for a thin flat plate) we use Newtons second law and consider the initial momentum per second of a thin strip of fluid before it enters the boundary layer. Its value is ( u dy) U. In contrast, within the boundary layer, the corresponding velocity component is u rather than U, and so the final momentum per second of the strip is ( u dy) u, from which the total rate of change of momentum for the whole boundary layer is obtained by integrating the difference between these two expressions across the boundary layer:
"

( ! u dy ) ( u # U )
0

By Newtons third law, the skin friction R is minus this value. Alternatively, we may write R = U2 where the momentum thickness is

2007 P.G. Sayer

Page 27

given by
&

!=

u # u 1 " % dy U U$ 0

Equivalently, in non-dimensional form, R = U2 where


1

!=

u $ u 1 # & d" U U% 0

But we still do not have an explicit expression for the actual boundary layer thickness ! This we now obtain. Boundary layer thickness For laminar flow we can simply equate the shear stress to the gradient (spatial derivative) of the skin friction, via Newtons law of viscosity.
!= u d$ dR = " U2# = y dx dx
! 1 = 5.48 Rn " 2 x

When we use u/U = 2 - 2 then integration gives

The important point here is that the laminar boundary layer thickness is proportional to Rn-1/2. A more general form of the above result is
! $" 2 2 & = x #Rn %
1

where

!=

d # u % d" U $"=0

Turbulent flow is more difficult, partly because we cannot use Newtons law of viscosity, since the actual value of, say, the velocity component u comprises both a mean value and a turbulent variation (perturbation). [It is beyond the present scope of work to introduce further concepts such as eddy viscosity].

2007 P.G. Sayer

Page 28

Instead we take an empirical result, based on the turbulent flow of a fluid through a circular cylindrical pipe and use this as a means of extracting an expression for . It turns out that the resultant expression agrees remarkably well with actual measurements (and so that may be regarded as our justification for using it!). We assume that a sufficiently accurate expression for the shear stress is given by
4 %# ! = 0.0225 " U ' U$ & 1

The next step is to equate this to ! U 2 "

d# dR and then integrate. [In other = dx dx

words we assume that the empirical expression gives a result for an equivalent flat plate]. For the usual turbulent profile u/U = 1/7 this gives
! 1 = 0.37 Rn " 5 x

So a fundamental difference is that the boundary layer thickness is proportional to Rn-1/5 for turbulent flow, but proportional to Rn-1/2 in laminar flow. This has important practical consequences. For example, when Rn = 106 the skin friction in turbulent flow is more than three times its value in laminar flow; so there are obvious benefits in trying to retain laminar flow over as long a length of the plate as possible. [Without wishing to complicate matters, the reader should be aware that turbulence can be put to beneficial use, for example in controlling the width of the wake, or in retaining flow attachment to the body; but these effects are properties of the shape of the body, and are not relevant to a simple thin, flat plate]. Skin friction resistance in naval architecture Over the first half of the twentieth century, several famous theoreticians and experimentalists in fluid mechanics gradually refined the boundary-layer formulae, according to greater accuracy in experiments and advances in theoretical knowledge. Not surprisingly, a mean skin friction coefficient cf is often used to compare various expressions and approaches, where cf = 2 L / L , and L is the maximum boundary layer thickness for a flat plate of length L. Our present approach gives the result cf = 0.072 [Rn(L)]-1/5
2007 P.G. Sayer

Page 29

for turbulent flow, whereas experimental values now suggest a slightly different value cf = 0.074 [Rn(L)]-1/5 In fact, as Reynolds number increases, the turbulent velocity profile u/U = 1/7 becomes less accurate. For Rn up to about 108 we can use u/U = 1/N , increasing the value of N. However, a logarithmic profile is better, for example
! % y u #+ C = 2.5 ln U $ &"

where the constant C lies typically between 6 and 8.5. Schlichting obtained the empirical result cf = 0.455 [log10 Rn(L)]-2.58 using a logarithmic velocity profile. Prandtl then made a correction to allow for the laminar region near the plates leading edge: thus we have the PrandtlSchlichting skin friction formula cf = 0.455 [log10 Rn(L)]-2.58 - A [Rn(L)]-1 where A = Rcrit ( cfT - cfl )

For Rcrit 5 x 105 we find A 1650. Using a different approach, Schoenherr proposed the formula

c f = 0.242 / log 10 c f Rn ( L)

Finally, in the 1950s the I.T.T.C. put forward a formula which is still in widespread use today:
cf = 0.075

(log

10 Rn ! 2 )

+ 0.0004

The factor 0.0004 makes an allowance for average effects of hull roughness.
2007 P.G. Sayer

Page 30

Courses on resistance and propulsion of ships will expand on these formulae and how they are used in practice. Bibliography
Fundamentals of fluid mechanics - B R Munson, D F Young and T H Okiishi, Wiley, 1998. Fluid physics for oceanographers and physicists - J Williams and S A Elder, Pergamon Press, 1989. Mechanics of fluids B.S. Massey and J Ward-Smith, Chapman and Hall, 1998. Introduction to fluid mechanics - J A Fay, MIT Press, 1994. Mechanics of fluids - I H Shames, McGraw-Hill, 1982. Solving problems in fluid mechanics - J F Douglas, Longman Scientific and Technical, 1986. Fluid mechanics with engineering applications - R L Daugherty, J B Franzini and E J Finnemore, McGraw-Hill, 1985. Fluid flow - a first course in fluid mechanics -R H Sabersky, A J Acosta and E G Hauptmann, Macmillan, 1971. Fluid mechanics - VL Streeter and E B Wylie, McGraw-Hill, 1985. The fluid mechanics and dynamics problem solver - The Research and Education Association, New York, 1987. Introduction to Naval Architecture - Eric Tupper, Butterworth Heinemann, 1996.

2007 P.G. Sayer

Page 31

Tutorial Exercises 1

1. (i) (ii) (iii) (iv) (v) (vi) (vii)

Verify that =2xy represents a stream function that is valid for irrotational, incompressible flow. (Hint: consider continuity and vorticity). Sketch the streamlines for x>0, y>0. Calculate the magnitude and direction of the fluid velocity at the points (0.5,1.0), (2.0,0.25) and (4.0,0.125), indicating these vectors on your sketch. What type of real flow could represent? Determine the corresponding velocity potential Sketch the equipotentials for x>0, y>0. What is the angle between corresponding streamlines and equipotentials? Does =xy represent a valid incompressible flow field? (=real number).

2. (i)

What type of flow can be represented by =2y(1+x) ? Sketch the streamlines. Derive the corresponding velocity potential. Repeat this for =2y(1+x+y-1) and comment on your answer.

(ii)

3. (i) (ii) (iii) (iv)

Determine which of the following velocity fields satisfy continuity for an incompressible flow: v=6xi-6yj-7tk (t=time) v=10i+(x2+y2)j-2xyk v=10i+(x2+y2)j-2tk v=(ur,u), where ur= -U cos + a2 r -2 U cos , u=U sin + a2 r -2 U sin , (U=constant).

4. (i) (ii) (iii) (iv) (v) (vi) Verify that =ax2-y2, where a=constant, is a valid stream function. Calculate the velocity at a general point (x,y). Determine the asymptotes. Sketch the streamlines when a=tan , =30o, indicating the direction of the fluid flow. What general type of flow could be modelled by ? Determine whether this flow is irrotational.

2007 P.G. Sayer

Page 32

Tutorial Exercises 2

1.

For an irrotational, incompressible fluid show that both and satisfy Laplace's equation

! 2" ! 2" ! 2# ! 2# + = + =0 !x2 !y2 !x2 !y2


2. Sketch the streamlines for a doublet, indicating the direction of flow. Calculate the velocity vector at (x,y)=(1,1) for a doublet of unit strength. A source of unit strength is placed in a uniform stream of velocity 1 ms-1 (i) (ii) (iii) Sketch the streamlines. Calculate the position of the stagnation point. Calculate the transverse width of the zero streamline far downstream.

3.

4.

Show that the potential flow generated by a doublet in a uniform stream has velocity components ur = - U (1-a2 r - 2 ) cos u = U (1+a2 r - 2 ) sin in the usual notation. The pressure coefficient cp is defined by the relationship cp = (p-p)/(0.5U2) Show that cp=1-4 sin2 for the combined doublet and stream flow. Hence determine where the fluid pressure is least.

5.

A simple pump comprises an impeller of diameter 1.9 m inside a casing of diameter 2 m. Sketch the pressure distribution for an impeller speed of 1000 rpm. and hence calculate the maximum difference in radial pressure across the pump.

6.

A circular cylinder is spinning steadily about its longitudinal axis, normal to a uniform incident stream. Determine the maximum variation in pressure on the surface of the cylinder in terms of the cylinder and flow parameters. Calculate the magnitude and direction of the resultant force.

2007 P.G. Sayer

Page 33

Tutorial Exercises 3

1.

You are given the following velocity profiles: Laminar Turbulent Calculate the ratio of (a) (b) displacement thickness momentum thickness u/U=2-23+4 u/U=1/7

between laminar and turbulent boundary layers when R n=105

2.

For the above laminar boundary layer velocity profile, calculate the boundary layer thickness at the trailing edge of a flat plate of length 0.6m immersed in water having a uniform velocity of 2ms-1, =1.2x10-6 m2s-1, =1025 kg m-3. Calculate also the power required to overcome the resistance of both sides of the plate which is 1m wide. For the velocity profile u/U=2-2 show that the mean skin friction coefficient is approximately equal to 1.46 Rn-1/2. A thin flat plate of length 10m is placed in a uniform flow of velocity 1ms-1. Compare the maximum thickness of the boundary layer with that given by the velocity profile u/U = sin (/2). What flow speed in air would be needed to generate the same boundary layer characteristics?

3.

4.

For the two velocity profiles in Question 3, calculate the ratio of the corresponding skin-friction coefficients cf and comment on this value. Show that when Rn=106 the plate has a maximum boundary layer thickness in turbulent flow more than four times that generated by laminar flow. (You may assume u/U=1/7 for turbulent flow). How important is the skin friction component when considering the overall drag of a marine vehicle?

2007 P.G. Sayer

Page 34

Das könnte Ihnen auch gefallen