Académique Documents
Professionnel Documents
Culture Documents
A.V.Smirnov
Contents
Contents Preface Nomenclature 1 Properties and Variables 1.1 Kinematic variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.2 Substantial derivative . . . . . . . . . . . . . . . . . . . . . 1.1.3 Acceleration 1.1.4 . . . . . . . . . . . . . . . . . . . . . . . . . . Strain rate and vorticity . . . . . . . . . . . . . . . . . . . . i v vii 1 1 1 3 4 4 5 5 6 8 8 9 12 13 15 15
1.2 Thermodynamic Variables . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 Equations of state . . . . . . . . . . . . . . . . . . . . . . . 1.2.2 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Fluid Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3.1 Thermodynamic properties . . . . . . . . . . . . . . . . . . 1.3.2 Transport Properties . . . . . . . . . . . . . . . . . . . . . . 1.3.3 Other properties . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Fundamental Laws 2.1 Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . . . . i
ii
CONTENTS
2.1.1 General formulation . . . . . . . . . . . . . . . . . . . . . . 2.1.2 Constant density ow . . . . . . . . . . . . . . . . . . . . . 2.1.3 Stream function . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Conservation of Momentum . . . . . . . . . . . . . . . . . . . . . . 2.2.1 General formulation . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Constant density ow . . . . . . . . . . . . . . . . . . . . . 2.2.3 Vorticity formulation . . . . . . . . . . . . . . . . . . . . . . 2.2.4 2.2.5 Potential ow . . . . . . . . . . . . . . . . . . . . . . . . . . 2D limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 16 17 18 18 21 21 23 24 27 28 29 32 32 32 33 33 34 34 37 39 39 40 40 43 44 46 50 53
2.2.6 Viscous limit . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.7 Inviscid limit . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.8 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . 2.3 Pressure Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 General formulation . . . . . . . . . . . . . . . . . . . . . . 2.3.2 Constant density ow . . . . . . . . . . . . . . . . . . . . . 2.3.3 Viscous limit . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.4 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . 2.4 Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 General formulation . . . . . . . . . . . . . . . . . . . . . . 2.4.2 Constant density ow . . . . . . . . . . . . . . . . . . . . . 2.4.3 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . 2.5 Curvilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . 2.5.1 Invariant forms . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.2 Non-inertial coordinate systems . . . . . . . . . . . . . . . 2.6 The Law of Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6.1 PI-Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6.2 Non-dimensional formulations . . . . . . . . . . . . . . . . . 2.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Laminar ows
CONTENTS
3.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Conned ows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 Flow between parallel plates . . . . . . . . . . . . . . . . . 3.2.2 Axially moving concentric cylinders . . . . . . . . . . . . . . 3.2.3 Rotating concentric cylinders . . . . . . . . . . . . . . . . . 3.2.4 3.2.5 3.2.6 Poiseuille ow through ducts . . . . . . . . . . . . . . . . . Combined Couette-Poiseuille ows . . . . . . . . . . . . . Non-circular ducts . . . . . . . . . . . . . . . . . . . . . . .
iii 53 54 54 56 57 59 63 64 65 66 68 71 72 76 76 81 81 85 87 92 94 98
3.3 Unsteady ows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Fluid oscillation above innite plate . . . . . . . . . . . . . . 3.3.2 Unsteady ow between innite plates . . . . . . . . . . . . 3.4 Creeping ows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Stokes ow around a sphere . . . . . . . . . . . . . . . . . 3.4.2 2D Creeping ows . . . . . . . . . . . . . . . . . . . . . . . 3.4.3 Lubrication theory . . . . . . . . . . . . . . . . . . . . . . . 3.5 Boundary layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.1 3.5.2 Flat plate integral analysis . . . . . . . . . . . . . . . . . . Laminar boundary layer equations . . . . . . . . . . . . . .
3.5.3 Blasius solution . . . . . . . . . . . . . . . . . . . . . . . . . 3.5.4 Reynolds analogy . . . . . . . . . . . . . . . . . . . . . . . 3.5.5 Free shear ows . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Integral methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 4
4.1 Transition to turbulence . . . . . . . . . . . . . . . . . . . . . . . . 105 Turbulence Modeling . . . . . . . . . . . . . . . . . . . . . . . . . 105 4.2.1 LES models . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.2.2 RANS models . . . . . . . . . . . . . . . . . . . . . . . . . 107
CONTENTS
113 115
Coordinates and Tensors . . . . . . . . . . . . . . . . . . . . . . . 116 Cartesian Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 A.2.1 Tensor Notation . . . . . . . . . . . . . . . . . . . . . . . . 120 A.2.2 Tensor Derivatives . . . . . . . . . . . . . . . . . . . . . . . 126
A.3
Curvilinear coordinates . . . . . . . . . . . . . . . . . . . . . . . . 128 A.3.1 A.3.2 Tensor invariance . . . . . . . . . . . . . . . . . . . . . . . 128 Covariant differentiation . . . . . . . . . . . . . . . . . . . . 132
A.3.3 Orthogonal coordinates . . . . . . . . . . . . . . . . . . . . 134 A.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 143 147 193 197 . . . . . . . . . . . . . . . . . . . . . . . . . . 198
B Curvilinear coordinate systems C D E Solutions to problems Midterm Exam Topics: Laminar Flow Solutions Final Exam Topics E.2 Analytical Solutions
E.1 Fundamental Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 E.3 Boundary Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 E.4 Turbulence Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . 202 Index 203
Preface
The idea behind this book is to provide a formal but concise introduction to theoretical uid mechanics. The book covers the traditional topics of uid mechanics, such as the fundamental equations of motion, compressible and incompressible forms, and invariant formulations, special analytical solutions, laminar ows, elements of boundary layer theory, main aspects of turbulence modeling and numerical methods. The emphasis is on viscous ow phenomena. The author tried to put more emphasis on mathematical rigor rather than on lengthy narrative. Tensor notation is used extensively throughout the book. However, the knowledge of tensor calculus is not required of the reader, since enough introductory material is provided in the appendix.
vi
Nomenclature
is dened as , or is equivalent to . Note: partial derivative over time:
vii
E 9 9 7 )DC@BA@( 86 (
partial derivative over : control volume time -th component of a coordinate ( =0,1,2), or uid velocity: strain tensor any variable of coordinates and time stress tensor viscosity : kinematic viscosity Right-hand-side Left-hand-side Navier-Stokes equation Partial differential equation Continued list of items
&(
a a
d b ca
(1.1)
The denition above can be inverted to dene a particle trajectory. Denition 1.1.2 Particle trajectory For a given vector eld of uid velocities, a solution to the following problem:
, particle trajectory,
From these denitions it follows that the vector of velocity is tangential to particle trajectory at each point in the uid. Particle trajectory is also called a streamline. In what follows we shall drop the super-index , and also adopt the following dot-notation for particle velocity:
(1.2)
meaning that a time derivative of uid particle position is taken along the particles trajectory at a space point . This notation should not create a confusion since space coordinates do not depend on time, while the uid-particle coordinates do. Hence when using time derivatives of coordinates, as in (1.2), we will mean uid particle coordinates. We shall use the same dot notation for partial time derivatives of other uid variables at a xed point in space. For example, for any variable,
(1.3)
SR4 I (
SR4 9 QPH I
SR4 9 Q I A
, is
(1.4)
where we used a Taylor expansion up to the rst order. We can rewrite it in a more compact tensor notation (Sec.A):
(1.5)
Considering that the displacement follows the uid element, it should be a product of velocity and time: . Then we can rewrite the expression above in terms of variable change, , as follows:
(1.6)
(1.7)
( ( ( I 4 RS4 9 4 ( IPH b SR4 9 ( PH b SR4 9 ( H SR4 b 4 9 b ( PH I I ( SR4 b 4 9 b ( I SR4 9 ( H I H 4 ( A ( ( I 4 SR4 9 ( P& H b SR4 9 ( I H b SR4 9 ( PH SR4 b 4 9 b ( PH I ( I A H b & H 4 H 4 H yifg"fed h 4 4 4 A SR4 9 ( I H b 4 SR4 9 ( P& H I SR4 9 ( H SR4 b 4 9 b ( PH H I ( I
, and taking the limit of
we have:
this expression represents a substantial derivative of a uid variable, which describes the change of that variable in the coordinate system moving with the uid element. The last term , in (1.7) is also called a convective derivative.
1.1.3 Acceleration
Applying (1.7) to velocity itself, we have the relation for ow acceleration:
(1.8)
1.1.4
where these parts becomes the newly introduced strain rate ( ( ) tensors. Denition 1.1.4 Strain rate tensor The strain rate tensor,
, is dened as:
(1.9)
is dened as
(1.10)
G F
n U ob G F A R m A f A I dl b R m b f A I dl k f A
f A
R m A f A I dl kU n
A R m b f A I dl kG F
j AjA f b & A 4 A
G F
R H A I G n
) and vorticity
G n
(1.11)
From this denition, and the denition of vorticity tensor (1.10), we have:
(1.12)
Using the denition (A.23) of the permutation tensor components of (1.11) explicitly as:
(1.13)
j Gqp
j R A j m A I j GSp dl 6 n
p Rr A p R A R qr A
n r n p n
, is dened as
where is the gas constant. Depending on the number of parameters in the system, there can be several equations of state so as to keep the number of independent variables, to comply to the Gibbs rule [1]:
where is the number of components and is the number of phases in the system. For example, for the water-vapor mixture we have and , then . Thus the state of water-vapor mixture can be completely described by the pressure or temperature only.
1.2.2 Energy
The rst law of thermodynamics expresses the principle of energy conservation related to the thermodynamic variables: internal energy1 , , heat, , and work, : (1.15)
that is, the change of energy of the uid element is equal to the heat inow plus the work done on that element. Rewriting this in terms of specic values, i.e. values related to the unit of mass ( ), we have: (1.16)
where is the specic volume: , and the minus sign signies that the work done on the system is positive when it is compressed and negative when it is inated. From the denition of entropy [1] we have:
z R Dr }I d |{u
s te Y
yxSw!u v
d yxSwu v
}|{!u z
2~ l u
(1.17)
is a function of
and
(1.19)
Relation (1.19) constitutes the equation of state as dictated by the energy conservation law. as: Another form of energy identied in thermodynamics is enthalpy. It is dened
(1.20)
(1.21)
Expressing the entropy, , from relations (1.17) and (1.21) we obtain the so-called relations, well known in thermodynamics:
F Y b F s b F s
Y b b
F Y r Y b F s
F s
only:
F s b F s
R Y9 FI
F s
F s
1.3.1
Thermodynamic properties
Specic heat at constant volume determines the amount of thermal energy that is needed to be transfered to the substance to heat it up by one degree, while keeping it at a constant volume:
(1.22)
For ideal gases the internal energy, depends only on temperature, and the relation above can be rewritten as:
(1.23)
Specic heat at constant pressure, , is dened as the amount of energy needed to be supplied to the substance to heat it up by one degree at constant pressure:
(1.24)
where the enthalpy, , is used instead of internal energy, , since it accounts for the work done by the pressure forces to extend/compress the substance. For an ideal gas this translates to:
(1.25)
v s 2v s s s 2v
2v
(1.26)
where is the kinematic or thermodynamic property, which represents a dependent variable of the problem, ux is the amount of property passed through a unit area per unit time and gradient is the direction of maximum change in property. Both ux and gradient are a vectors. The coefcient of proportionality, represents the transport property controlling the transport process (1.26). The coefcient of viscosity The coefcient of viscosity is introduced to quantify the process of momentum transport. Newtonian uid: mation: In elasticity theory the resistance force is proportional to defor-
Similarly in uid mechanics the resistance to uid motion is proportional to the velocity change in the direction normal to the uid motion (strain).
Since generally the stresses and strains are tensors (Sec.1.1.4, 2.2.1), the relation above is usually written as:
A uid with the linear relationship between stresses and strains, like (1.27) is called a Newtonian uid.
( R I3 t) R I3 wA6 a 5 4 t a t 4 F F 4 5 F U F G T !
10
Denition 1.3.2 Coefcient of viscosity The coefcient of viscosity is introduced as a proportionality constant between the shear stress and strain. Considering one component of stress tensor at: i=1, j=2, we have:
(1.28)
Viscosity usually decreases with temperature for liquids and increases for rareed gases [2]. Non-newtonian uid: For non-Newtonian uids the relation between the stress and the strain is non-linear, for example
Thermal Conductivity Following the gradient approximation (1.26), we presume the heat transport to obey the relation:
(1.29)
where is the wall heat ux per unit of time ( ). This is the expression of the Fouriers law, where is a heat conduction coefcient, or heat
(9E 9 7 )B9 @A A
R F r I X &
( R A 4 ` 4 I t) SR4 I Aq
W p p r A V r T D0 A V T Gf F G T
s t 6 &
&
E 9 9 )DCqq@( 7
11
then in SI units:
(1.30)
Mass Diffusivity The gradient approximation (1.26) for the mass transport can be written as:
(1.31) where
is in direction (
j Y X j Y
we have:
j j j Y j Y t R j Y I j Y j &
ustd 5 5 s
4 a a R 4 I 5 4 F F r F F
a F ( R 4 5 F I `) R F I wA6
j Y j &
F X r f
j &
where
X R F I R a a A 4 4 t 4 r R 4 I V
, will be determined
I X k 5 4
).
12
(1.32)
and relating the mass ux to the uid velocity across the boundary: we have:
There are two non-dimensional numbers relating the momentum-to-mass and mass-to-heat transport processes:
Schmidt number:
Lewis number:
where
3 j Y j &
R j I fd j Y j &
,
R j I fd j 3 r Y }p Y W
Dv X
1.4. PROBLEMS
Bulk modulus
13
The bulk modulus expresses the change of density with increasing pressure at a constant temperature:
(1.33)
1.4 Problems
Problem 1.4.1 Mass diffusivity in terms of concentration Show how to obtain (1.31) from (1.32). Problem 1.4.2 Lubrication
A plate of mass with an area slides down a long incline at angle , on which there is a lm of oil of thickness , with viscosity . Assuming the plate does not deform the oil lm estimate (1) the terminal sliding velocity , and (2) the time required for the plate to accelerate from rest to of the terminal velocity.
yds
q3
Y Y ' r Y s Y Y d
s R F I X V { l
14
The total change of mass inside the control volume will consist of changes of mass inside the volume because of density changes that may occur at each point of the ow, and the inux or out-ux of mass through the boundary. This can be expressed as:
where is the unit normal vector to the boundary, is the surface area element, and the last integral spans all the boundary of the control volume. 15
a A Y b 3 & Y 4
4 F a 3 Y 4 a
By Gauss theorem we can convert the last integral to the volume integral:
and nally:
Considering the arbitrary nature of the control volume selection, we conclude: (2.2)
This is a general relation of mass conservation valid for both compressible and incompressible ows. Differentiating the second term by parts, and using the relation of substantial differentiation (1.7) the latter can be rewritten as: (2.3)
which is also called the continuity equation or incompressibility condition. Vector eld satisfying (2.4) is also called solenoidal or divergence-free. Another form of this relation can be obtained by combining (2.3) and (2.4): (2.5)
3 R R A Y I b & Y I 3 R A Y I b 3 & Y 4
3 R A Y 6 A Y I
6 R A Y I b & Y
q f A Y b 4 Y
a 6 q f A 4Y
as:
&Y
17
or, using the nabla operator (A.32): (2.7) which analogously to (1.13) is
(2.8)
i.e. a vector normal to the plane, and therefore (2.8) are reduced to:
(2.9)
The following relation between the stream function and the mass ow rate can be shown for a two dimensional case:
(2.10)
where is the element of the surface normal to the velocity also be proved more rigorously for a 3D space.
A & &3 Y p p p A p A Y R r ( b ( r A I Y r ( r b (
p R pr R R qr
j j GSq A Y p
j p j GSq A Y E 7 B9 9 6 p qpr I r II
. This relation can
p p r A Y r AY
AY pr A Y AY
18
Remark 2.1.1 Existence of stream-function It should be noted that sometimes, instead of denition (2.6), the streamfunction is dened from a simpler relation:
(2.11)
where the density is omitted. In this case the stream function can only be used to describe incompressible ow. This can be shown by computing the divergence of velocity vector :
which is true due to the symmetric identity (A.27) and the symmetry of with respect to the order of differentiation. This becomes especially obvious in a 2D case:
Thus, in terms of denition (2.11) the stream function can only exist for incompressible ows.
which will apply to a particle of both constant and a variable mass. Applying this to a uid particle of density and velocity in a small control volume, we have
(2.12)
j U
p p pp r r r qr b A q f A A
3 R A YI 4
qy j USq f A j p A R I 4
j p j USq A
u A
19
The forces acting on a uid element come from the possible external forces, like gravity, electromagnetic elds, etc. (body forces), and forces caused by the interaction of this uid element with neighboring uid elements or boundaries (surface forces). Body forces relate to the unit of volume and surface forces relate to the unit of area.
(2.13)
where is the volumetric density of the body force . It corresponds to a force eld like electromagnetic, gravity, etc. Generally it can serve as a source term connecting this equation to other equations. Denition 2.2.1 Stress tensor The surface force term
Using this denition and applying Gauss theorem to the last term in (2.13), we have:
(2.14)
Using the fact the control volume was chosen arbitrarily, the integral sign could be dropped, and we have:
(2.15)
(2.16)
j j T b 3 b 6 3 j j T R A Y I 4 b jA j j T j R A Y I b R A Y I 4 3 j j T b 3 b j j T R A Y I 4 G T
20
Using the hypothesis of Newtonian uid (1.27), and general considerations of symmetry for the case of isotropic and homogeneous uid, a general relation between the stress tensor and the strain tensor can be written as [2, p.66]: (2.17)
where is the pressure, is the coefcient of bulk viscosity, which is only important for compressible ows. Sometimes it is convenient to separate the stress tensor (2.17) into the pressure-related and viscous parts: (2.18) where (2.19)
Parameter is the coefcient of bulk viscosity, which can only be important for variable density ows [2]. Thus, for incompressible ows (2.19) becomes: (2.20)
where we used the denition of strain rate tensor (1.9). Using the denition of the viscous stress (2.19) we can write (2.16) as: (2.21) When (2.21) as: (2.22)
Equations (2.2), (2.22) and (2.19) represent a general case of compressible unsteady ow. The dependent variables are two scalars: density, and pressure, , and a vector - velocity, . Considering that equation (2.21) is a vector equation, this leaves one more equation to close the system. The appropriate candidate for this is the equation of state (1.13).
jA ` Y b j j T j R A Y I b R A Y I 4
Y r j j b b j R A Y I b R A Y I 4 T jA
G F
U T
G T
21
(2.23)
where we used the continuity relation: Substituting this into (2.22), we have:
(2.24)
which is an incompressible form of momentum equation, also referred to as the Navier-Stokes equation (NS).
(2.25)
(2.26)
j j A jBj A j A Bjj A j Bjj f A I V j b jA b j fR j uR AR I V I j j @f V l I V j j R T j AV b F Aj A b Y Bjj f A W j f b & A R A R A SRR A j p j 2nj GS n j GSp A I j 2pG j p dl A I 2j p j GSp dl A I j Ip j Gmp dl 2nj GSp j
R A A I dl f A A ii n j GSp j A f A
Now lets consider the convective term in (2.24). Using the constant density assumption ( ) and now considering the cross product of the type , we can repeate the steps as in (2.26) and obtain1 :
j A n j Gmp
f A A } f A
j p 2nj GS} f A
q R m A I U m A
G m A f A j GSp jn
j j Sp j GSp
CHAPTER 2. FUNDAMENTAL LAWS
23
(2.31)
And substituting (2.29) and (2.31) into the momentum equation (2.24) we have:
(2.32)
(2.33)
(2.34)
2.2.4
Potential ow
Lets consider the irrotational ow where the vorticity vector is zero ( ). This ow is also called potential ow, since the velocity vector can be replaced by a gradient of a scalar function, , also called a velocity potential function:
This is possible, because the gradient of a scalar function also satises the condition of zero vorticity, which follows from the denition of the vorticity vector (1.12) and the symmetry of the second derivative of with respect to order of differentiation, :
Note that we achieved only a partial success in our objective to replacing the velocity vector with the vorticity vector, but thats the best we can do.
2
n
bj A p R A A I dl n j USk u A A
(A.34), we have:
j j
24
In the case of steady state incompressible potential ow the continuity condition (2.4) translates into the Laplace equation for the velocity potential:
(2.35)
Thus, the solution of the problem in this case is reduced to nding a single scalar function from equation (2.35). Another important relation in this case can be obtained from equation (2.34), which, after eliminating time derivatives and vorticity terms reduces to:
(2.36)
2.2.5
2D limit
(2.37)
Where we denoted the term in parentheses by . The equality above is a rst rank tensor equality with terms of type . Lets now form a cross product between this equality and of the type (see also (A.32)):
(2.38)
bj A R jnj GSp W n j GSp "I " 6 !" " p b & A " p " f s " 6 s " " p p s ! j bj A 2nj Gmp W n j GSp q ! b & A
j R j j I j GSp dl R A j m A I j Gmp dl q n 4 F a Y b ( R A A I dl
(A.32), we get:
q q f A
j GSp
" p
R "" n " " n I W iiy" n " mp j GSp W j " n " A " n " b A b & n
can be simplied as:
G
E 9 r qA9 p A A mn9 9 n E 7 7 q f A A " n " Ab " " n A " n " AA " n " b " " n A n " " b " n " A j " j R j n " A I b " R " n A I 6" R n A I j j m" b " R n A "I j t " R n A I R j y" " j I 6" R n A "I Sp j Gmp " m@" p 6" p p b & n " u A " n p " w" p B9 b j j "y n j b n A j j " p USp W b j " R A I " p p USp " R 2n j Gmp W n j USp I " " !w" p b " f & A " p
.
(2.40)
where we used abbreviations for the last two terms on the RHS. Since is a symmetric tensor then by (A.2.22), we have .
Noticing that by the denition of vorticity vector: the equation (2.39) to the form:
and In a 2D limit we have the 3-rd component of the equation above, i.e.
Applying the same manipulations to the Using the permutation property (A.24):
n EB9 9 7 Sn9 9 ky n E r BA9 p A A E r q(9 p ( ( E 7 7 7 } f j b G m R j Bjj f } f G m I dl b j j j j o dl j R j j I dl R I dl R j G j p j 2pG j Ip dl p 2 R j p j Ip j GSp dl 6 n
, and , and we have:
j p j USq A
n W R rr n R rr n
n W n A & n n pp p p R rr ob n I W pp n }I W n n I W
,
(2.44)
Substituting (2.44) into (1.12) we can nd relation between the vorticity vector and the stream-function:
In the 2D limit: , , the rst of the last two terms in (2.45) will vanish: Stream-function formulation After substituting
27
Thus, the ow is now completely dened by a scalar eld , which is obtained as a solution of (2.47). Note that since (2.47) contains 4th order derivatives of , it involves more complex boundary conditions, and poses higher differentiability requirements on . In the case of irrotational ow ( Laplace equation for the stream-function: (2.48)
with the boundary conditions derived from the relation between and the velocity eld (2.9). Solving the equation for stream function is usually preferred over solving an equation for the vorticity, since the velocity eld can be obtained from the stream function by a simple differentiation of type (2.6) or (2.9), whereas obtaining the velocity eld from the vorticity as in (1.12) would require a more laborious integration.
(2.49)
Using the expression of vorticity vector (1.12), we can express the above equation in terms of vorticity vector only: (2.50)
which is an incompressible viscous limit of the NS equation in vorticity formulation (see Problem.2.7.3).
SR4 9 ( I
SR4 9 ( I
28
In the case of steady state ow this equation simplies to a Laplace equation for the vorticity vector:
(2.51)
(2.52)
This is Euler equation for inviscid incompressible ow. It can also be rewritten in terms of substantial derivative (1.8):
(2.53)
Conservation of vorticity It can be shown that the inviscid ow preserves vorticity. For this purpose lets form a vector product between the nabla operator and equation (2.53):
where the last equality is due to the symmetry of transform the term according to: (2.54)
R y A f A I m j m A G j p b f A m j k f A m j p l p p
f A y j p G kG m j u A m j p 4 q A 4 m j p p
jA j R A Y I b R A Y I 4
Y j b A Aj A f &
Y 4 A
Bjj m n j T
and identity (A.27). We can
29
which means that the vorticity is conserved ( ). In particular, this means that if the ow was irrotational ( ), it will remain so4 . In this case the problem of inviscid ow can be solved using velocity potential function (Sec.2.2.4). The momentum ux Since
(2.56)
(2.57)
4 F a 6 n
j j R A A Y I b R A Y I 4
(2.55)
j Gqp
j j b R A A Y I j
j j R A Y I 4
n 4
q n
j j A
30 Inlet
At the inlet boundary the value of the velocity is usually specied . This boundary condition is known as Dirichlet boundary condition. Note, that this is not always the case, since a pressure can be prescribed as the inlet condition instead, when the Poisson equation for pressure (2.60) is used. Outlet Depending on the character of the equation system the boundary conditions may or may not need be specied at the outlet boundary. The most common outlet boundary condition is the condition of the zero boundary-normal velocity derivative (Neuman boundary). In more complex ow situations there may not be a clear distinction between the inlet and the outlet, since the ow may reverse. These types of situations are hard to solve in a consistent manner and should be avoided by repositioning the inlet/outlet of the domain so as to comply to either Dirichlet or Neuman boundary conditions. Fluid-solid interface (Wall) In the case of a uid-solid boundary the ow velocity is set equal to the velocity of the wall, which covers the cases of both stationary and moving boundaries. This boundary condition is called a no-slip boundary condition. In some cases a nite velocity jump may be imposed at the boundary, in which case this is called a slip boundary condition. The specication of the velocity alone at the boundary may not be enough, since the momentum equation (2.21) contains second order velocity derivatives in the viscous term ( ), which means that the rst order derivatives should be given at the boundary. However, with the no-slip condition at the wall, the velocity derivatives at the wall can be considered zero. This follows from the continuity relationship (2.4). For example, if we consider velocity components and as being parallel to the wall, and normal to the wall, then from the no-slip condition we have:
and consequently:
rA
pA
p r A A
A
j j T
31
In cases when the forces on the wall need to be estimated they can be related to the boundary normal forces due to pressures, and shear forces that can be related to the stress tensor via (2.14). It should be noted that if the boundary is moving with acceleration additional non-inertial terms should be introduced into the boundary conditions (Sec.2.5.2).
Fluid-uid interface (Free surface) The gas-liquid or liquid-liquid boundary conditions are also called free-boundary conditions. They consist of the requirements that the pressure, velocities and the uxes of mass and momentum be continuous functions across the interface. This means that these quantities should have the same values on both sides of the interface. The position of the interface surface will then be determined as a solution to the ow equations subjected to the free surface boundary conditions. In cases where surface tension effects are important, they should enter into the pressure boundary condition, namely the extra boundary pressure should be added on both sides of the interface. This pressure should be inversely proportional to the local surface curvature. Since the surface curvature generally depends on the direction selected on the surface to measure the curvature, one form of its estimate may be to set it proportional to the sum of inverse curvature radii in two orthogonal directions:
(2.58)
where is the coefcient of surface tension. Note, that the forces resulting from the pressure terms should always act normal to the surface. Shear forces at the boundary are usually considered to be zero.
pp r qr A A A e 0e R d b d I yx
pp r qr A A
32
(2.59)
After substituting the rst term on the LHS from (2.2), considering that a constant, and rearranging terms, we have:
(2.60)
which is a Poisson equation for pressure. It has to be solved together with the momentum equation (2.21) and the relation of the state law (1.13).
u A
T j j b f Y j j T b R
is
( X
j R j j T j R A Y I A I R A Y I SR4 I q
b R j R A Y I A I j Y I b g R j R A
b R A YI 4 j Y I A I b R A Y I 4
33
Using the incompressible form of the viscous stress tensor , (2.20), and the continuity relation, , it can be shown that the last term will be zero:
(2.61)
This is the incompressible form of the pressure equation, also called the Poisson equation for pressure. It should be considered together with the momentum equation (2.24).
(2.62)
jA j j b j A TT j u Y j A j j b 2Aj A j A j A f Y R f I Y R j j T j R A Y qII 6
34
comes out as the solution which satises this condition. However, because the Poisson equation is a second order equation, the Neuman boundary condition alone will result in an indeterminate solution, when adding any constant to the pressure will still satisfy the equation. Fixing the value for the pressure in at least one point will remove this uncertainty. Thus, other conditions at the inlet/outlet boundaries are usually applied. At the inlet/outlet pressure values are given and the boundary normal derivatives are usually set to zero. In the case when the velocity is also specied at the inlet/outlet, both pressure and velocity specications should be consistent so as not to create a overdened problem. Specifying either pressure or velocity alone will be enough in many cases. However, this will depend on the character of the ow and the discretization scheme used to solve the equations [4].
(2.63)
where the LHS is the rate of energy change inside the control volume, and the two terms on the RHS represent total heat inow into the control volume and work done on it. This relation is also known as the second law of thermodynamics. Note that both heat and work are transported into the control volume through its boundary, so they can be represented by ux-vectors. Energy:
(2.64)
3 4 A l A b A 4 A Y Y b 3 ( ` A A dl
b 4 4 4
b qRR ( I Y I 4 Y 4 4
35
where the minus sign in front of the gravity term means the potential energy increases as we move against the gravity force. Heat: The total heat change inside the volume can be related to the heat ux through the boundary of the volume:
(2.65)
where is the heat ux, which is the rate of heat inow through a unit area per unit of time, and the vector-element of the boundary introduced in (2.1). Minus sign occurs because of the convention of surface normal vectors to point outside of the volume, meaning that is actually the out-ux of heat whereas was dened as an incoming heat by the virtue of (2.63). Applying the Gauss theorem to (2.65) we have:
(2.66)
where the minus sign signies that the heat ows from higher to lower temperatures. This relation is known as the Fouriers law. Substituting it into (2.66), we have:
(2.67)
Work: In analogy to (2.65) we introduce the ux of external work through the uid element:
3 u & 6 & 4
s g s t 6 &
3 R s 4 I
& 4 & 4
&
&
36
We saw in (2.2.1) that the surface forces are described by the stress tensor . The time derivative of the displacement of uid element is given by velocity (1.1). Thus5 :
(2.69)
(2.70)
(2.71)
(2.72)
Note that the positive signs in (2.68) and (2.69) follow the convention that the work done on the system is positive when the direction of external force coincides with the direction of displacement.
3 m A G T b Y R A Y I 4 A 4
a 4 F 5 4 & 3 R A G T 4 I A G T q &
G T
37
which after rearranging terms and dropping the integration sign due to the arbitrariness of our choice of the control volume becomes:
(2.73)
which is the equation for the rate of change of energy density valid for both compressible and incompressible uid. In an ideal gas approximation we can express the LHS in terms of temperature using the thermodynamic relation (1.23): (2.74)
It is useful to express (2.73) in terms of enthalpy (1.20). For this purpose we can add to both sides of (2.73) and obtain: (2.75)
which is the equation for the rate of change of enthalpy valid for both compressible and incompressible uid.
j F b A R j A G G V l r G t I m b R s I b 4 4F a Y
m A G T b Y A R A Y I 4 6 b R s I 4 Y A l A b 4 A D A Y b A Y R Y I 4 A
And after canceling the same terms on both sides, we have:
4 l b A b s b 4 4 b Y A A m G T R I R Y I R
3 y A G T b 3 )4 Y A l A
4 l b A b s s 4 2v Y A A y G T R I R Y I
4 l b A b s 4 Y A A m G T R I R Y I
Y A b 4 A
A R A Y I 4 b R A Y b qRR ( ` I Y I
s I 4
4 Y j T
YI 4
4X
38
Using the denition of viscous stress (2.19), we can rewrite this as:
which after applying the continuity relation (2.4) this reduces to:
(2.77)
Equation (2.77) can be used in combination with the momentum equation (2.21) to derive a strong form of the Bernoullis equation6 :
(2.78)
We can also rewrite (2.77) in terms of temperature change, and velocity. Using the denition of specic heat (1.25), substituting the velocity from the relation of viscous stress (2.20), and assuming , we have:
(2.79)
In the cases with small pressure and velocity gradients, or when uid viscosity is small compared to the heat conductivity, which corresponds to small Prandtl number (1.30), equation (2.79) simplies to:
(2.80)
6
A R y b u A I m A V b s b 4 4 s Y
A A G T m b R s I b f 4 4 Y A G T m b R s I b 4 4 Y A R m b f A I V G V l kU T F 4F a b 4 F a C b A A dl s 4 s Y
39
where we also presumed that the heat conduction coefcient, , is a constant. This is a limit case of (2.79) for the case of heat dominated ow. Note that since the substantial derivative is used for T, we still have the convective terms present:
(2.81)
A b s R s & s I Y a s
40
(2.83)
(2.84)
(2.85)
where the covariant and contravariant velocities and stress tensors are linked by the conjugate tensor relations (A.42), (A.43):
where and is the metric tensor (A.40) and its conjugate (A.38). A more detailed description of curvilinear coordinate systems including an important case of an orthogonal coordinate system can be found in (App.A.3.3). A derivation of Laplace operator in cylindrical coordinates using chain differentiation is given in (Sec. B).
2.5.2
Generally, if we consider space and time as a single continuum represented by a single set of coordinates, we have no distinction between the curvilinear and
j fj b j R A Y I j b R A Y I 4 T A
G
G`
41
non-inertial coordinate systems. This treatment is used in relativistic uid dynamics [7]. This approach would also make sense in the treatment of general moving and deforming coordinate systems. However, in a variety of applications it makes sense to treat space and time as separate variables. In this case one has to distinguish between inertial and non-inertial coordinate systems. For any moving coordinate system one has to formulate an explicit dependence of coordinates on time. Generally, this time dependence can reveal itself in motions and deformations. In this section we shall only consider the case of a non-deforming and moving coordinate system. And in particular, we shall focus attention on an important case of rotating coordinate systems. Rotating coordinate systems Consider an inertial coordinate system and a non-inertial coordinate system , which is moving with respect to . Let the coordinates of a particular point in be described by a vector7 :
(2.87)
where is the time derivative with respect to the inertial frame of reference . We shall consider the motion of as composed of displacement determined by and rotation, given by angular velocity vector . Then can be represented as
is the velocity of the point as measured in the non-inertial coordi. Since the rotation does not affect the motion of the origin, , we
Here we are using vector notation, since only the vector quantities are involved
is the position vector the origin of system is the coordinate vector of the point in system
&
n b &
b Q
E (q9 p 7 r q(9 ( Q
SR4 I SR4 I
"
&
SR4 I
with respect to
42 have
(2.89)
(2.90) where
(2.92)
(2.93)
The extra acceleration terms involving arise due to rotation and are interpreted as being the result of the Coriolis force. An important special case of a non-inertial coordinate systems is a coordinate system undergoing a pure rotation with a constant angular velocity. Assume that the moving coordinate system is undergoing a rotation with and . Then we can align the origins of the two coordinate system to
4F a n
n n b b b R n n b n I n I R b n b 4 F a SR4 I n n
R n b I 4 &4 &
4 4 r r &
l b 4 b 4 n b 4 b 4 Q
4 4 b &Q
" &
" "
43
(2.94)
Where and are the additional acceleration vectors (Problem 2.7.7). The corresponding Coriolis forces are introduced into the equations of motion in a rotating coordinate system:
with being the displaced mass. In computations of continuum media dynamics is replaced with mass element :
where is the density of the uid and is the face-normal velocity across the face of area of a control volume (see Problem 2.7.8).
The number of parameters8 can be reduced by considering the dimensions of physical units in which they are measured:
A parameter can be looked at as just another independent variable, like space coordinate or time. However, we treat them separately, since parameters are specic for each physical law, whereas are not.
8
('&%$" #
E f G(uss p ( Q 7
w b b
bb r 4 r Q
9 R W !9 9 Q I A 6 A f f 4 Y W
9 W !9
R mn9 9 I n
p r q(9 ( Y
eliminate plane
altogether. Lets also assume that the axis of rotation is normal to the , that is . With these assumptions (2.93) becomes
44
The only non-dimensional combination of these parameters is provided by the Reynolds number:
We can non-dimensionalize other variables by scaling them with the appropriate length and velocity scales: (2.96)
Since units dimensions should be preserved in an expression of a physical law, a law formulated in dimensionless variables can only contain dimensionless parameters. Hence the new dimensionless variables (2.96) should enter into a only, since its the only dimensionless parameter derived from relation with the properties of the system. Following the convention that the velocity is the dependent variable (2.95), we can write this relation as:
Thus, using simple considerations of physical dimensions, we reduced the number of parameters from three, ( ) to one, ( ). Similar considerations allow to reduce the number of parameters in a more general case, which is proved in a so-called PI-theorem.
2.6.1 PI-Theorem
(2.97)
Suppose that the law requires that each variable can be expressed in units of length, time an mass, which we call the primary dimensions: (2.98)
6 7Guss p
a p 5 4 4 G q W r
2 5e
A q A
) p 5 4 4 aG 6 1 6 p R 7vyss I 8
9 W !9 2 R 4De9 Q I q A W 3e 2
) ( 6 (
0) 4 aG 6 2 3e
variables,
45
where represents the dimensionality, and are the powers of the 9 respective units . For example, the dimensionality of the gravity acceleration, , in SI units is expressed as . We can write a non-dimensional expression of the law by scaling variables with respect to any primary dimension. Lets rst non-dimensionalize the variables with respect to length through the transformation:
(2.100)
(2.101)
should be
(2.102)
9
9 9 H QP H V QP H T QP I
6 p 8 SRR I 7Guss R I I W
C S B9 R I k A
9ARC1 6
F r
This is an extra relation imposed in addition to our physical law (2.97) by virtue of scale invariance or homogeneity of our law with respect to length-scaling [9]. In a similar manner we can arrive at two more relations imposed because of homogeneity with respect to other two primary dimensions: time and mass: (2.104) (2.105)
Thus we have three more relations in addition to our physical law (2.97), . which means that the number of variables can be reduced from to If we use a more complex law that involves an additional primary dimension, such as temperature, then we can reduce the number of variables of the problem by 4. Generally, if we have primary dimensions and independent variables in the problem, then the independent variables can be reduced to nondimensional parameters. These parameters can be different depending on the choice of scaling factors used in transformation (2.100). Generally, normalization (2.100) does not have to be done by primary dimensions, but can be used with respect to any group of variables that do not form a so called PI-group, i.e. their products of the type (2.98) can not be reduced to a non-dimensional number, no matter what powers are used [10, 11]. This constitutes the essence of the PItheorem [11]. It lays a more rigorous foundation for the law of similarity [7, 8], which means that the same solution can be reused by rescaling the variables.
a a
6 R 7vyss p I 8 ) YP H
and get:
47
Space and time derivatives should also be scaled. Thus, if we consider space and time as independent variables: , and as dependent variables velocity, density, and pressure we can introduce the following non-dimensional variables:
After the dimensional variables are replaced withe the dimensionless ones by means of (2.106), one should look into the physics of the problem and see if some extra relations between the scales can be applied. For example, in some problems the characteristic velocity scale can be related to length and time scales as: . This can be the case in the problem of a steady ow around a xed object. However, a steady ow around a rotating object will have an independent time scale related to the period of rotation. After all possible eliminations of scales were done, one should try to construct dimensionless combinations of scales, or non-dimensional parameters. There can be several different ways in which these parameters can be selected. This process can be formalized somewhat [10], but there is still a room for subjective judgment on which dimensionless combinations of scales are most appropriate as parameters for the problem at hand. No matter how these parameters are selected the PI-theorem states that their minimum number can be as low as , where is the number of dimensional scales and is the number of primary dimensions of the problem. If all the dimensionless parameters have been correctly identied, it should be possible to replace all the dimensional scales with these parameters, thereby rendering the physical law in a dimensionless formulation with the minimum set of independent parameters. Lets consider several cases of non-dimensional formulations and of application of the PI-theorem. Mass conservation law Lets write a non-dimensional formulation of the mass conservation law (2.2): (2.107)
(2.106)
49 (
48
where we used the nabla operator (A.32) to simplify further analysis. There are three primary dimensions in this case ( ): [length], [time] and [mass]. Using the scaling transformations (2.106), the non-dimensional form of (2.107) is:
tion.
As can be seen, all the dimensional parameters canceled out from the equa-
Momentum equation Consider the steady-state limit of the incompressible momentum equation given by the Navier-Stokes equation (2.24):
(2.108)
Since this is a constant density formulation, density becomes a parameter of the problem: . Considering this, and transforming to the non-dimensional variables according to (2.106) we obtain:
(2.109)
To simplify things, lets select for the pressure scale the dynamic pressure: . By doing this we state that pressure scale, , is not an independent parameter of our problem, but is related to the density and velocity scales. Now, lets make each term of (2.109) dimensionless, by multiplying the whole equation by :
The equation includes four dimensional parameters: . Since this is a steady-state formulation, time is no longer a dimension of the problem, and there are only two primary dimensions left ( ): [length], and [mass]. Thus,
9 W 9 q(9 A
( A ( j A b Y %j j r A W j
A b ( A j A A ` r ( j j W j A
A ` b Y j j W q j A A j
l
R A Y I b & Y
( A A r
Y Y
A ( r X A r Y
49
(2.110)
(2.111)
(2.112)
with the four non-dimensional variables: parameters: (see also Problem 2.7.9). Boundary conditions
and
Some non-dimensional parameters arise from the boundary conditions. For example, non-dimensionalizing the boundary condition of the energy equation (2.82) leads to:
(2.114)
where is the wall heat ux ( ), the heat conduction coefcient, (1.29), and difference between the wall and the uid.
s R F r I X v s u
where
(2.113)
l l a
. If we
9 A 9 ( x b 2 A 3e j A d j j d j A
2 ( W A 3e
( A x r
A u 6 a s
x 3e 92
50
Boundary conditions at the free surface give rise to additional parameters, such as Froude number, relating inertia forces to gravity, (2.111), Weber number, relating inertia to surface tension:
Other non-dimensional parameters may appear as new phenomena are added into the physical law [2].
2.7 Problems
Problem 2.7.1 Derivation of the vorticity equation Obtain the result outlined in (2.30). Problem 2.7.2 2D vorticity limit Perform the missing steps in (2.42). Problem 2.7.3 Incompressible viscous limit Derive (2.50) from (2.49). Problem 2.7.4 Conservation of circulation The velocity circulation is dened as
(2.116)
where the integration is over any closed loop inside the uid. Show that for irrotational ow ( Problem 2.7.5 Bernoullis equation
4 F a b q n
):
( A d b c
r A Y
.
2.7. PROBLEMS
51
Problem 2.7.6 Volume change inside a moving boundary Suppose that a region of space is enclosed by a moving boundary. The velocity of motion of the boundary, , is given at each point on the boundary. Show that the rate of change of the volume, , of that region will be equal to:
(2.117)
where is the unit normal vector to the boundary and and nd the coefcient . Problem 2.7.7 Rotating coordinates
Obtain explicit relations for the components of acceleration vectors in in (2.94) in terms of . Problem 2.7.8 Rotation with separated coordinate origins
in this case.
Consider a simple rotation with origin of the rotating coordinate system origin of :
as in (2.94), but now let the rotate with the same around the
4F a j j b b T
surface area element,
A G T m b R s I b 4 4 Y E 7 mn9 9 n a A 43 n 4 3 b C b A A dl A
jA j R A Y I b R A Y I 4
9 9 r 9 p 9 B9 r B9 p qn
Q
52
Problem 2.7.9 Nondimesionalizing energy equation Write a non-dimensional form of the heat convection equation (2.79):
selecting for the pressure scale. Determine the minimum number of dimensionless parameters. Write the equation using the Eckert number (3.114) as one of the parameters:
(2.118)
AY r A R y b u A I m A V b s b 4 4 s Y z s A r
(3.4)
Lets make an assumption of laminar ow which states that the time scale of changes in the ow can not be lower than the time-scale of the motion of the 53
j T
Bjj f A W j j T p Y
( ` Y b
where
A A b G T y b s R s A & A s I Y j p Y j j T p Y j f b & A q f A
, has the form
54
boundary or any external sources. In other words, if there is any repeatability in the motion of the boundary or in the external forces then the frequencies associated with either factors can not be lower than the frequencies of the ow motion. It means that neither the boundaries not external forces can induce any additional frequencies in the ow. In the limit case of non-moving boundaries and non-changing forces the ow should not depend on time, which means that all the dependent variables should become functions of spatial coordinates only. The conditions of laminar ow dened by (3.1.1) are realized when the contribution of the non-linear term in the momentum equation (3.2) is small, or when the contribution of the viscous term dominates. This is usually the case when non-dimensional Reynolds number (2.110):
(3.6)
Navier-Stokes equation, (3.2) is known to have very few analytical solutions. This is mainly due to the non-linear convective term , which is the main cause for the rich dynamical features of uid ow. For this reason, most of the cases that provide analytical solution do not include the convective term. Below we shall consider several such cases.
3.2
Conned ows
Probably the simplest of conned ows are the ows between moving surfaces, which belong to the category of Couette ows [2].
f A A
2 5e
corresponds to
d h 2 i3e
W 3e 2
Bjj f A W
E r q(9 p ( 7 Q
55
Figure 3.1: Flow between parallel plates: the lower plate is at rest, the upper plate is moving with velocity . . Since the plates are considered to be innite no variable should change in direction either. Thus, the only independent variable of the problem becomes the vertical coordinate , which we shall denote as . Likewise, from the symmetry of the problem the only non-zero component of velocity is , which we shall denote by . In addition to this we can also assume the pressure to be constant. This can be explained by the absence of normal stresses in this ow. With these assumptions the momentum and energy equations (3.2), (3.3) reduce to:
(3.7) (3.8)
Equations (3.7) and (3.8) can be solved with the boundary conditions u(0) = 0 and u(H) = U, and and , with the solution (See Problem 3.7.2): (3.9) and the shear stress: (3.10)
pA
2 3e Y Y k l Vl rTl
ps s R I s R I s A V b r s r r rA V
V A V T
R I A
r(
p( E r BA9 p A A 7
(3.12)
Solution to the temperature equation (3.8) produces a quadratic dependence on (See Problem 3.7.2):
The dimensionless Brinkman number is introduced as a relative measure of viscous forces to thermal uxes:
(3.13)
In the momentum equation (3.7) the effect gravity was neglected under the assumption that the gravity force acts normal to the direction of the ow. Generally it may not be the case, but the solution procedure remains essentially the same (see Problem 3.7.3).
(3.14)
(3.15)
Er 9p 9x 7 sqAqBAA q A
b l b b l s s V s ps r V R I
r 9 r R I sA R BC%D9 I sA
p r r b R I fd R I qA
qr A d
p R s sI x r V
2 e { l 3ik
Er qBA9 9 7
57
(3.16)
from which it follows that . Thus the problem of pulling an innite rod does not have a steady-state solution.
-momentum:
(3.17)
p p R I qA qA
e pe p p R e I ud r 9 R X p I fd
p p R e X e I ud p tb R e Xp e I ud r sA R e X I fd R X e I ud R I
p
.
r r R 0uI qA 9 R e I qA
p r r R e I qA 9 R e I qA p A r Y
p kp r r b R &uI ud
E qBA9 qBAqBAA 6 A 9p 7 Er 9p 9x 7
p r D9
are the constants, which can be determined from the boundary con-
(3.18)
Boundary conditions:
The equation that determines the velocity satisfy this equation has the form:
(3.19)
Substituting it into the boundary conditions, we nd the constants and the nal solution becomes:
(3.20)
which is a solid body rotation. Thus the steady-state solution for the ow inside a rotating cylinder is a solid body rotation. Remark 3.2.3 Rotating an innite rod If we solve the problem above with the boundary conditions:
p r
p e X e e X p e p n p b p e X e eX
p w pq A x sA v V b R s I r w v xsp A i b qrp A R I p ss
p sA p p p R eI s 9 n R eI s 9 n
p b p sA r
p p n qA e p pe e e e p eX X X p X e n e sA
e p e sA p e RR e II sA p
59
(3.22)
Substituting it into (3.17) and integrating it, we can obtain the pressure distribution:
where the constant is found from the boundary condition: the pressure distribution is: (3.23)
It is useful to compute the rotational momentum (torque) that arises in a system of two rotating cylinders (Problem 3.7.6).
3.2.4
Lets consider a case of a straight duct with a constant cross-sectional area, . Since the area does not change its form, the length-scale of the problem will be the characteristic size of the duct1 : .
1
R eI
p r D9
R r X r e d I r e r n Y dl b
p e p u qA 9 n qeqA
b r e ln Y R I r e n p A r Y r Y }p r
cn sA e p r e cn D9 p r p r u b u cn r e p e r b e
. Thus,
60
Compared to the case of innite plates, a duct has an entrance and it has a nite cross-sectional area. The existence of an entrance causes entrance effects, such that the ow is three dimensional over some distance from the entrance, that is all three components of velocity are non-zero: , and each component also depends on all three spatial coordinates: . However, we assume that this transition region will end at some distance after the entrance and be replaced by a fully developed ow region, where axial velocity does no longer depend on the axial coordinate, i.e.
(3.24)
It should be noted that the existence of a fully developed laminar regime is only an assumption, but it happens so, that there is a solution satisfying this assumption. However, it also happens that there are other solutions, which do not satisfy this assumption, i.e. unsteady turbulent regime. Which solution is realized in reality depends on the magnitude of the Reynolds number. At the low Reynolds numbers the fully developed regime occurs in circular ducts at distances between 30 and 100 duct diameters from the entrance. In a fully developed ow in addition to (3.24), the velocity components normal to the axis should be zero: . This is because the appearance of any velocity component normal to the axis can not be sustained for a long period of time since there is no pressure gradient imposed in that direction. On the other hand, any short time appearance of such components will violate the assumption of a steady-state laminar nature of the ow in a fully developed region. Thus, we can use only the axial component, which we shall denote as . With these assumptions, in a fully developed ow the axial momentum equation (3.2) can be simplied to:
(3.25)
Since the second term on the RHS of (3.25) does not depend on , and then it follows from (3.25) that should not depend on either. But
The other two momentum equations simplify to only on : , and we can write:
p(
r p A R b rr A I V b
E 9 9 R q(9 r ( I p A t A 7
p p( p( p
p p R S(9 r ( I A A
p p R (I (
R q(9 r ( I A A
p A A
depends
p(
61
since depends only on , it follows that is a constant. Thus, dividing equation (3.25) by , and introducing dimensionless variables:
where is the appropriate measure of the duct cross-sectional size, we obtain a the following boundary value problem for the dimensionless velocity :
(3.26)
where subscript stands for wall and index spans only the variables that depends on: . This is a Poisson equation in a conned 2D domain with the no-slip velocity at the boundary (Dirichlet boundary conditions). Remark 3.2.4 Reynolds number independence Note that the Reynolds number does not enter the equation (3.26), and therefore, should not affect the solution. This is because we excluded non-steady solutions and turbulence from our consideration. The circular pipe Consider a fully developed ow region in a circular pipe of radius . The natural coordinate system for this case is cylindrical: , with being the axial coordinate. Following the discussion of the previous section, only the axial velocity will be non zero, which we shall denote for simplicity as . As it component, was shown, it can only depend on and . However, because of the symmetry of the duct, the dependence on will inevitably lead to time-dependent solution and violate the assumption of fully developed ow. Thus, we should have , and using the Laplacian operator in cylindrical coordinates, (3.26) reduces to:
(3.27)
R I A A
p( X A A 9 9 ( V r q ( 6 p p(X p( (X C DC%D9 9 R d I A d A d A r 5
A y d A
q(9 r (
sr A
62
(3.28)
(3.29)
(3.30)
Skin-friction coefcient
(3.31)
where the Reynolds number is based on duct diameter: Poiseuille number (3.32)
2 W X A e
0 A l ( V w | d e A r ( V e A t l A
A r T Y x e ( l A V e A V T
e ( V A r r d R I 23e A Bd r T lY
.
d r d R I A d 3ek 2 {
63
3.2.5
Couette ows are driven by shear boundary motion, and Poiseuille ows - by the axial pressure gradient. The equations describing both types of ows do not have a nonlinear convective terms, which makes these equations linear, and thus enables linear superpositions of solutions. Consider the Couette ow between parallel plates (Sec.3.2.1) but with a constant pressure gradient applied in the axial direction. The momentum equation in this case will be a combination of (3.7) and (3.25) in the form:
(3.33)
where we are using for the axial and for the vertical coordinates. Since the LHS does not depend on and the pressure is a function of only: , we conclude that must be a constant. Introducing dimensionless coordinates: , we have:
(3.34)
(3.35)
we have: form:
(3.36)
R (I
r
b p b A r r
d b R I A
( V r A r r
( rA V d R d I A R I A
p r d B9
A `A 9 ( X X X
64
(3.37)
From (3.36) we can see, that when the velocity changes sign at the lower wall ( ). This is called ow separation point, and according to (3.37) it corresponds to the pressure gradient:
(3.38)
A pressure gradient greater than this will cause the ow at the lower wall to reverse2 .
3.2.6
Non-circular ducts
Because the problem of the fully developed duct ow was reduced to a well posed and well studied boundary value problem based on the Poisson equation there are variety of analytical solutions obtained for ducts of various shapes [2]. There are several convenient measures that are introduced for ducts of arbitrary shapes. The cross-sectional length-scale of the duct is called the Hydraulic diameter, which is introduced as a generalization of relation for a diameter of a circle. For a circle of diameter the relation between its area and a perimeter is: . Thus, for any non-circular duct of perimeter and area the hydraulic diameter is dened as:
(3.39)
(3.40)
2
T c d T
d ( Vl r ( r V l
X
65
In a fully developed ow each element of the uid between cross-sectional planes at and moves with a constant velocity. Thus the sum of forces on that element should be zero. This means that the friction force at the wall should exactly balance the axial force pushing the uid element due to the pressure drop in the duct. Thus, we have the following relation between the mean wall shear stress and the pressure gradient:
and using the denition of the wall mean shear stress (3.40), we have:
Some unsteady ows in the ducts can still be solved analytically. To introduce unsteadiness we formally use the same assumptions that lead to (3.26), but now we shall reinstall the time derivative from (3.2), which with these assumptions becomes: (3.41)
where as in Sec.3.2.4, is the axial component of velocity in the duct, which is the only non-zero velocity component. Following the same reasoning as in Sec.3.2.4, we conclude that the last term can not depend on any spatial coordinate. Since we consider unsteady ows, this term can still depend on time. However, in this problem we can combine velocity and pressure into a single joint variable:
(3.42)
A W & A
p b 4 p Y A A
R S(9 r ( I A A p p Y A W A &
( T c ( T
( ( b
66
This equation has a parabolic character, which means that one independent variable - time - is asymmetric with respect to direction. Specically, at any point in time the solution will depend only on the previous points on the time axis, but not on the subsequent points. This difference in the directions of time: the future and the past, is the result of the rst order time derivative in the equation (3.42). In contrast, the second order Laplacian space derivative, makes all space directions equivalent. It should be noted that since , the Laplacian operator in (3.42) includes only two components: . Equation (3.42) is still simple enough to provide analytical solutions in several cases. The important cases include: 1. Starting ow in a duct. 2. Pipe ow due to oscillating pressure gradient. 3. Fluid oscillating above an innite plate. 4. Unsteady ow between innite plates.
In this case the solution for the uid velocity will depends on only one direction : . Then equation (3.42) can be rewritten as:
(3.44)
We can note that in this case the motion in time is periodic, while the changes in space are aperiodic. Correspondingly, we may look for a solution form which is a periodic function in time and a decaying function of space. One form of the solution that will satisfy this equation is: (3.45)
A b rr A q A R (q9 r ( I A A A r(
d SR4 n I eG 2 " w
r A W 4A
R I A A
r (
67
where is an unknown function of . If we substitute the latter into (3.44), well get the equation for :
where
Constant and can be found from the initial conditions. In case of a moving plate and a stagnant ow-eld at innity (3.43), we have:
(3.47)
p 9 R ! 4 n I F n g SR4 I A
Thus
b R ! 4 n I F n SR4 I A p 9
where
b p p" 9 n P n 6f H SR4 I A
and for
we obtain:
W g r n 5 g q
r }p R W l X n I
R I
X g
9 SR4 I A
R I
, we can write:
68
In case of a stationary plate and a ow-eld oscillating as we can obtain the solution by subtracting the equation above from
(3.48)
It can be checked by a direct substitution that the above equation satises (3.44) and the boundary conditions: , and .
(3.49)
, we have:
d SR4 n I eG SR4 I
(t):
A A ) 4 4 t) s
Ar sA r tX r 4 W Ar r W sA 4 r
69
where (3.50) describes the xed lower plate, (3.51) describes the moving upper plate, (3.52) is the initial and (3.52) the nal velocity distributions. Note, that the nal velocity distribution was obtained before as the steady state solution for this case (Sec.3.2.1). As can be seen, boundary condition (3.51) is non-zero. To simplify our search for the right solution, it would be nice if we could look for a function which is zero at the boundaries. To make the boundary conditions both zero lets look for a solution that is represented by the difference between the steady-state solution , (3.53), and , since this function will be zero at both plates:
Now, if we replace with the new unknown function: following boundary value problem:
where we have all the boundary values zero. We can look for a solution in form of separated variables:
(3.58)
and :
R s 9 wI
r
R 9 u 9 I A A R 9 u 9 I d d R R s d I I A d R R s d I I 9 9 R s 9 I A R s 9 }I s
R s 9 I A R s 9 wI
R s I R I R s 9 I
R s 9 I A
r g gg
R 9 u 9 IA A d RR s d wI I A 9 R s 9 I A
R u 9 wI A
R s 9 I A
, we obtain the
70
where as in (3.46) we introduced a constant . This time, however, is not known in advance, and should be determined from the boundary conditions. The relation above is identically satised by:
(3.59) (3.60)
with , and unknown constants in addition to . Out of these the constant can be absorbed into and , since and enter as a product in (3.58). So, without loss of generality we can set . The boundary conditions on (3.55) - (3.58) can be translated to and as:
where the last equality (3.64) is satised identically by virtue of (3.60). From (3.61) and (3.59) if follows that . Similarly, from (3.62) it follows that , where is an integer number. The only way to reconcile boundary condition (3.63): , with boundary conditions (3.61) and (3.62) and the analytical form of given by (3.59), is to express as a Fourier series:
Thus for
we have:
Coefcients
, we obtain:
R s 9 wI
a
R I R R wI v RdI v R I v
f f R a I fe d w R I
R sI d R I (u 9 d R sI R I eG b R I fe d R wI R wI
0uI R wI R u 9 wI R }I R I R 9 9 I R s I R d I R s d I R s I R I R s 9 }I
R wI
R wI R wI a
71
(3.65)
The LHS of this equation represents the inertial forces. The assumption of creeping ow or Stokes ow states that the inertial forces are negligible. With this assumption the last equation becomes:
tain:
f a d R a I ue {f R d I w l R wI } f A V q q j GSp
72
Swapping indexes and in this equation and subtracting one from another we can express it in terms of vorticity vector (1.12):
(3.68)
Thus both the vorticity and pressure satisfy Laplace equation in a creeping ow. Important cases of creeping ow include: 1. Fully developed duct ow. Re-number independent. 2. Flow about immersed bodies (Stokes solution or the sphere). 3. Flow in narrow but variable passages. (Lubrication theory). 4. Flow through porous media.
(3.69)
In spherical coordinates the relation between the velocity and stream-function (2.6) will become:
(3.70) (3.71)
n
fe d x fp e d r
}G
} n
p qA x A
73
(3.72)
where we neglected the azimuthal direction angle because of the symmetry of the problem. The boundary conditions for this problem are:
(3.73)
which can be checked by direct substitution3 into (3.72). The velocity components can be found from (3.70) and (3.71) to be:
(3.74) (3.75)
The important quantity is the uid drag on the sphere. It consists of the contribution of the shear stress (tangential friction at the surface), and the pressure forces normal to the surface (Fig.3.2). Pressure distribution on the surface of the sphere is computed from the momentum equation (3.66), and results in the following expression:
(3.76)
3
e e r b e r ue d r d R D9 I rl
e le
d er }r e l V
b d q qA b d p e fe l d eG A b d x e
where
is any constant.
b R Ir 4 r r d r r b d r fe r R eI p dl R &uI R eI x
74
(3.77)
The total force on the sphere can be obtained by integrating both shear (surface friction) and the pressure components over the surface of the sphere:
(3.78)
where the integral should be evaluated at the surface of the sphere: , and is the outward normal unit vector and is a tangential unit vector at the surface (Fig.3.2). Equation above gives the total vector force acting on the sphere. We shall consider projections of this force on two important direction: parallel to the ow, and normal to it. We shall refer to the former as the drag force, , and the . From the symmetry of the problem we can immediately latter as the lift force, conclude that . For the drag force we should take the projection of the force (3.78) onto the direction of the x-axis, which is given by a scalar product between and unit vector in x-direction, , as shown in Fig.3.2:
I~ a
d 0 ~a ~a R I eG R `~ a I ~ a I'g
, ,
wx e
0 d d R I fe px T b R I eG R ~ I wx ~a
p A x fe l V w qp A x qb p A v V px T d e
~ a px T b `~ a ~
~a
~ a
"
"
0 R I ue d R s ~ a I ~a
75
(3.79)
where the negative sign in the second term is due to the fact the direction of increase of is opposite to the selected direction of . Substituting from (3.77) and from (3.76), and setting , we obtain:
where we omitted the term involving the constant pressure component , since its contribution will become zero after the integration. Computing the integrals:
where the plus sign signies that the force is directed toward the ow velocity . It is interesting to note, that the contribution of the viscous wall friction due to the shear stress is twice as big as that one from the normal pressure term. A useful engineering formula for a drag coefcient is obtained by relating the drag net pressure, to ow inertia (kinetic energy), :
d 2 5e
R ue e I e l d
, we
becomes:
px T
2 W tX }e l e
lXr Y
er d ue d z g V d b e r fe z e ! ~a d d d R r ue qpx T z yG fe z I r e l e r rl Y
}e V R l b I t V wx e d d z y ( eG ( R ( I r er R ( I ue d z l z y R I d d l ( d ( ( d R ( I yG{ R ` I eG R I fe z
2 4e l
R r e I X
wx
wx
px T
76
(3.82)
where is the known free-stream velocity. With these assumptions the drag coefcient becomes:
(3.83)
In addition to this there are various engineering approximations obtained for the drag coefcients of a sphere and a cylinder for various Reynolds numbers [2]. A reasonably good curve-t approximation for the sphere is:
2 e s
ssb@4e d b 3e 2 s 2 d l
2 d l 3e 2 b 2 3e d 4e s b m b l
f A V b q t Y A
77
Pressure inside a non-uniform gap Lets consider a ow between two plates one of which is at an angle to the other (Fig.3.3). We shall use the coordinates in horizontal, and in the vertical directions, and denote the corresponding velocity components as and . Without loss of generality we may presume the lower plate at to be horizontal and the upper plate is given by a known prole of its height at each -position: . Also without loss of generality we can assume that the lower plate is moving at a constant velocity . Lets consider a ow in a limited section stretching from coordinate to . From Sec.3.2.1 we know that the steady ow between parallel plates has a linear prole given by (3.9). In the case of tilted plates we can not expect this prole to hold, because it would to the loss of mass-ow conservation, i.e. more ow will enter the section at than exit at . Indeed, since the ow velocity at the lower moving plate should always be and at the upper xed plate it should always be zero, there will always be more ow entering at , where the plates separation is wider than at where it is narrower. This is because a linear velocity prole takes shape of a triangle, and the mass ow rate is proportional to the area of this triangle. The area of the triangle at will . From this we conclude that the linear always be larger than the one at velocity prole can not be a solution to our problem. In this situation a pressure distribution arises in the ow that leads to nonlinear ow proles at the inlet and the outlet, such that the mass conservation is satised. Thus, the nature of this ow will be that of combined Couette-Poiseuille ow discussed in Sec.3.2.5. If we assume that the solution is of combined Couette-Poiseuille type, we
R ( I
( (
( (
(
(
( (
78
can nd the equation for the pressure distribution in the section between the converging plates, that will lead to mass conservation. For this purpose, lets use the incompressibility condition (2.4), and apply it to our case:
(3.85)
. The solution where we used the no-slip condition at the walls: for the combined Couette-Poiseuille ow should be obtained from equation (3.34):
(3.87)
R }I R I 0 A t 0 b 0 0 R }I R I A A t R b A I b b p b A r r ( V r A r r ( V r l d r p b d r R d I A d R }I A
0 A pp b A r qr b A q f A
79
where the rst two equations follow from the boundary conditions, and the last equality follows from the substitution of (3.87) into (3.86). If we dene a constant as:
(3.89)
Since is now a function of , is also a function of , given by (3.88). Transferring (3.85) to dimensionless variables, and substituting (3.89), we obtain:
(3.90)
Substituting
This is the equation for pressure distribution inside the section between the two converging plates. It should be solved with a known prole of the gap width as a function of : , and with the given pressure distribution at the inlet ( ) and the outlet ( ) of the section. Constant pressure boundary conditions can be used in a simplied case: . Remark 3.4.1 Contracting vs expanding gap
When we solve equation (3.91) with the constant pressure boundary condi, for a linear prole of we should obtain a parabolic tions: solution with an extremal point, , (minimum or maximum) between and . The condition for this point is . Opening the parentheses on the LHS of (3.91), we can obtain for the point the following relation:
(
R I R }I
R ( I
(3.91)
0 b I 0 b R o p R d I d b b
( ( db b R d I r A
0 b dl R d r p
0 0 V
2 s( 0 2 R q( I s( 2
( V l r ( R ( ( I d b 0 I A
R I g R }I g
d 0 p A
(
(
80
from which we see that if the slope of the upper wall is negative: (contracting gap), then , and the extremal point is a maximum. This means that the pressure will be everywhere higher than the ambient inside the contracting gap. The opposite conclusion follows for the expanding gap. In reality the case of the expanding gap may lead to higher ow instabilities and cavitation. This is the result of the counter-ow pressure gradients arising in the case of expanding gap that may lead to the possibility of ow reversal and separation as determined by relation (3.38). Validity of the pressure equation In arriving at solution (3.91) we assumed that we can use the solution of combined Couette-Poiseuille ow given by (3.36). But that solution was obtained under the assumption that ow velocity has only component which depends only on one coordinate - . This assumption made the convective term in the NavierStokes equation (3.2) equal to zero. However, this assumption is not strictly valid in our case since we have a non-zero vertical velocity component, , and both horizontal and vertical velocity components are functions of and . In this situation we can still justify dropping the convective term if we use the assumption of Stokes ow, that is, consider the inertial forces to be negligible as compared to viscous forces. Mathematically this means that:
From the form of these equations we can deduce more specic relations between the parameters of our problem ( , , , ). In particular, if we impose condition of a narrow gap: , then it would lead to and . With these conditions we can neglect the equation for and simplify the -equation by neglecting a smaller order term on the RHS:
A X ( X
W S b D00 W mR A D00 I R b A I W
0 sb qA 0A Ab A A (
or expressed in terms of , , ,
j f qA Aj
} f A W f A A
81
Both terms on the LHS are of the same order and can be approximated by the order of the rst one, which is . Approximating the term on the RHS as , we nally have:
(3.92)
This condition together with justify the Stokes ow assumption and constitute the validity limits for equation (3.91).
3.5
3.5.1
Boundary layers
Flat plate integral analysis
We shall consider a two-dimensional ow above a semi-innite plate (Fig.3.4). Our objective is to introduce a quantitative measure of the thickness of the boundary layer, and to estimate its growth with the distance from the edge of the plate, .
Xr m A W b 0 A A A d W r
r X W (
(3.94)
Momentum thickness
Applied to the case of the boundary layer (Fig.3.4) it will provide the expression for a drag force, , on the plate:
expressing
w A dv A r Y r A r Y A R A Y I R Y I
R A dI R A I b R I R I A b r R A d I R I
R I A
83
(3.95)
Using this denition we can also dene the skin-friction coefcient as:
(3.96)
where is the distance from the edge of the plate in the downstream direction, and consequently, , we can rewrite (3.96) as:
(3.98)
(3.99)
Remark 3.5.1 Displacement vs Momentum thickness Displacement thickness is a more universal notion than a momentum thickness, since the former is obtained from the mass conservation, which is a always true in incompressible ows, whereas the momentum conservation may be violated due to the dissipative effects. Relation (3.97) will only hold for the at-plate boundary layers.
( 0 0 ( l R ( I ( r l Y 0 r Y l
( ( k R I l R I d
w A d v A
X l r l Y r T l Y k ( T 0
0 T (
Imposing the boundary conditions appropriate for the boundary layer ow:
(3.100)
(3.103)
R R t I A ) t I A ) R I A
, and obtain: (1.28):
b p b A r r R I R Il R I A A r V l Y k A V T Y k V d l
85
Integrating, we obtain:
(3.104)
(3.105)
which means that we have a thinning of the boundary layer with the increase or a Reynolds number according to
(3.106)
3.5.2
The ow incompressible equations (2.4), (2.24), (2.77) can be used as a basis for the boundary layer formulation in the form:
A A b R m b f A I m A W b s Y R s & s I b Y u A W f A b & A A q f A
2 V`X ( 5e Y V r
( Y Vd 2 `3e ( }p ' (
, we have:
( d k
2 3e (m
86
where we shall assume a 2D approximation for the boundary layer: , , that is, we consider that there is no variation in the span-wise direction. Consider also, that the layer is thin:
(3.107)
Therefore,
Using this assumption we can introduce the following scales for dimensionless variables.
(3.108) (3.109)
where is the operator (A.32). The factors were introduced to make all the dimensionless variables of the same order of magnitude as we take the limit . After transferring to dimensionless units, equations (3.107), (3.107), (3.107) will become (see Problem 3.7.13):
p R r q(9 ( I @ ( ( 4A
z p b R r A I xr r p p b q f A
p r ( ( ) p A r A }p2 e ( r d r }p2 e p A rd rA
x A p p A b s b s & s rr s r r p p p p p A p rr A r A r b A b & A A
A p ) p A A
( ( p ) r ( r }p2 e r ( ) ( (
p R r BA9 A I q A
2 4e
where is the Froude number (2.111), and we introduced the dimensionless Eckert number:
(3.114)
[2].
Remark 3.5.2 Parabolic character Note, that all the second derivatives over have disappeared from the momentum equations. This means the that equations of motion are parabolic with respect to direction . This in turn enables to use simpler solution procedures. Flow separation ), where If we apply the momentum equation (3.111) at the wall ( , and consider the gravity forces acting normal to the wall ( ), we obtain:
from which we see that the curvature of the axial velocity prole along the vertical axis ( ) will have the sign of the pressure gradient. At some point a strong positive curvature may cause the velocity prole to bend to the extend that velocity values may become negative close the wall. This means that ow separation at the wall may become possible.
(3.115)
p(
p p p A 0 A r A p A A r ( p ( (
t p p R 9 (I A
p(
S A W 0A b 0 A A b A
z s A
p rr A
p S p rr A A r A A
E p` 9 7 r R 9 (I r A
87
r(
88
(3.116)
Considering this, lets look for a similarity solution . That is, we assume that the proles at different are obtained by scaling the solution by a factor . This means that the solution can be expressed in terms of a single dimensionless variable , such as:
(3.117)
where the coefcient was introduced for convenience. Since we have a 2D case it is convenient to look for a solution in terms of a stream-function, , such that:
(3.118)
9 R B@( I
9 R ` X I A R B@( I A
( W l o R 0uI g R ( I g R ( I
R I R (I R B9@( I (
as:
9 R u @( I A
lm
R (I
d R 0uI g ) R }I g R }I g g 6b g g g
b g g g g g g g g g g g
g g g ( l m A W g g R r g k I ( l A A g g g ( l r 0 A gg ( W lr o gg g A g g 4( l ( l g g g 0 g g g 0 A d ( ( ( R g k I l W o R g d m l d m l I W l m 0 b ( 0 R g q( m P d m l I W l m gg A ( 9 R I 6 W l m R B@( I
( W l m R (I
R (I R (I
d R &uI g R &uI g
Since set
(3.122)
and nally:
(3.121)
(3.120)
(3.119)
(3.123)
is a constant, we can incorporate it into without loss of generality, and obtain for
90
where the last condition also implies that . This is a Blasius equation. It has no analytical solutions in a general case, but can be solved numerically. Knowing function the boundary layer.
Thus:
(3.124)
(3.125)
y s R I g g g g g R g d I g g g R g d I R g d I g
Rg
( Wl
s o d l d R I ugfei d ( W l o dIg ( d Wl o R A I
Rg d Ig
(l gg W o R I Y A V T
R 0uI g g
( W l o R }I g g
( Wl o
R I
91
(3.127)
Wedge ows
A similarity solution for the wedge ows was found as an extension of the Blasius solution. This solution, called after the authors . This solution is obtained by eliminating -velocity from the continuity equation:
m A W b 0 A 0 A 0 A A
R I k l ( R ( I k d
( g g Y k W l o R }I r T l ( W l b d R (I R I o 9 R I g R ( I R B@( I A 0 A ( k
R (I
92
where is the rst derivative of the so-called Blasius stream function, . With these assumptions one can obtain the Falkner-Skan equation in terms of :
where , and the boundary conditions are the same as (3.123). The Falkner-Skan formulation is consistent with the free-stream velocity distribution of the type:
I happens so that this equation provides similarity solutions that represent wedge ows, with the stream function of the type:
Wall suction or blowing The case of wall suction or blowing can be modeled in the Blasius solution by a non-zero vertical velocity at the plate surface. It can be seen from (3.120) that this be accomplished by specifying a non-zero value of the Blasius stream function at the wall:
An interesting feature of the solution is the effect of blow-off of the boundary layer, where the velocity becomes identically zero. This occurs at the blow-off limit of :
R I
d p R b I fe d qh | R D9 I
R r g d I b g g 6b g g g ( l o R }I R @( I 9 W | ( R ( I V g
d l m Rs A R }I
b X R d I l
R Ig
R I
93
where is the specic heat at constant pressure, (1.24), is the coefcient of viscosity, dened as a proportionality constant between the shear stress and velocity gradient (1.28):
and is the heat conduction coefcient dened through the relation between the heat ux and temperature gradient (1.29):
The Nusselt number was introduced to quantify the heat conduction at the walls with the temperature difference (2.114):
where
Reynolds analogy is the statement of proportionality between the heat ux and shear stress at the wall:
(3.128)
such that
where and are empirical constants depending on geometry. The Reynolds analogy is approximately valid for shear layers, boundary layers, and pipe ows.
x Wik ~
r T l Y k
(3.11):
s t 6 &
x 2 S3e ~ u
D0 A V T
T &
&
).
94
which in terms of explicit variables , and to the Blasius equations for the boundary layer (3.115):
Consider a shear layer created by two streams with uniform velocities and initially separated by a at plate. After passing the plate the streams mix forming a shear layer. The treatment of this case is done similarly to the Blasius approach, by selecting the non-dimensional variables as in (3.117), and looking for a solution in the form (3.119), but applied separately for each of the streams:
(3.130)
r
9 9 R A I A R q@( I ( p 9 A A A l d 9 9 5 rr W u l d 5 q f A p
s A
aP H aP H b
m A W
( ) W p l o 6
A b 0 A A b0A
becomes identical
95
We can sew both solutions at a horizontal plane at , which is the imaginary continuation of the plate. For this purpose we need to consider boundary conditions on that plane5 . One condition is derived from continuity: , :
where . It is an interesting fact, that in the case of the same uid in both jets ( ) and zero free stream velocity in the one of the jets ( ), the vertical component of the second jet at innity is a nite constant, determined by , which is the same as the blow-off limit in the case of the plate with suction or blowing considered in Sec.3.5.3.
(3.131)
The similarity solution for this case as obtained by Schlichting [2] is:
This plane does not represent any physical interface between the two streams. The latter should be dened as a surface formed by ow streamlines.
5
p R }I r A R }I A !r
r p R I s r }p R }I s p p R }I qr A r V R I A V p p } r R I ) R }I r R }I
p p s r R u Ir A R u Ir
4F a A r Y
p ) d R 0uI p A R 0uI s
d l m iDs R u I r p pd R V YI Xr Vr Y
p R }I r R I
, which
96
and the boundary conditions (axial symmetry), and (quiescent ambient uid), which translate into:
with the constant determined from the conservation of the total jets momentum. Thus, substituting (3.132) into (3.131) we obtain:
Wakes are ow patterns generated by bodies moving in a stagnant uid. We need to consider these ow patterns far enough from the body for similarity solutions
9 R u @( I A
( ' ( 0r w | ( }p ` 9 }p R @( I A R ( I A rl
R &uI g R }I g g R I 9 9 R @( I R @( I A r g b g g b g g g R I f7 l R I }p V Y m d
` ( R g l I }p r W ( }p ` A r R Ig R I }p ( }p W r ( r }p W r
9 R u @( I
97
to be valid. Since it is more convenient to select a coordinate system moving with the body, we will be considering a non-moving body immersed in a uid with the free stream velocity . It is also convenient to describe the wake in terms of deviation of velocity from the free-stream velocity:
(3.134)
and at
we obtain:
(3.135)
since usually constitutes a small value compared to (3.129), and considering and we obtain:
9 9 R @( I R u @( I
9 R B@( I A
m 0 W W
9 R B@( I
. Replacing
with
in
98
The constant is evaluated from the second Newton law, that the drag force should equal to the momentum decit in the wake:
and
(3.136)
When substituted into (3.134) it shows that the wake defect is proportional to body drag coefcient.
where is the vector area element. Summing up over all the four faces, and considering that the ow is uniform in the transverse direction (2D ow), we can expand the equation above as:
9 b 9 b 9 9 @( t( b @( t( @( @(
A Y t W o Yl Y Y
W o
r Y dl
6 A c
where
t
99
where is the normal velocity at the wall, which is non-zero in the case of wall suction or effusion. Rewriting the second integral as:
(3.137)
where the summation is done over all faces, and we are considering projection of forces and uxes on the horizontal direction. Writing the terms explicitly, we obtain:
b ( Y A Y h b Y A Y A b Y g A Y A Y h ( R A I wA6 w 6 w bt A ( (
100
where is the pressure. Canceling terms and neglecting higher order terms, we obtain:
(3.139)
One can relate the pressure gradient to the free-stream velocity gradient:
It is also convenient to express (3.139) in terms of displacement thickness, (3.94) and momentum thickness (3.95):
A ( A ( b ( r Y Y Y t T
R t I r DRt I A
A g A b b r Y r Y Y A Y ( r T ` r b I R I R l X I b b b R (X A ( ( b ( r Y r Y T 0 W Y ( Y (
r A r A
3.7. PROBLEMS
101
(3.140)
3.7
Problems
Show how to obtain equation (3.7) and (3.8).
Problem 3.7.2 Couette plates solutions Solve equations (3.7) and (3.8) with the boundary conditions u(0) = 0 and u(H) = U, and and .
Consider a wide uid lm of constant thickness, , owing steadily due to the gravity down the inclined plate at angle . Find an analytical expression of a uid velocity distribution as a function of a distance from the plate surface: . Assuming the viscosity and the density of the uid are , and respectively, nd the volumetric ow rate, , per 1m of the plate. Atmospheric pressure can be considered constant.
s tRd V
{ s d
0 b 2R l b I 0 DR t I ( DR t I 0 l BR I r r d BR g I ( ( r A r d A d b 0
0 0 DR b Il b r T Y 0 DR b Il b 0 kl ps R }I s
0 DRt I b 0 0 l ( d b0 0 r TY F `X AR I X R I s R }I s
102
Problem 3.7.4 Couette solution for non-Newtonian uids How will the solution (3.9) change for a non-Newtonian uid? Problem 3.7.5 Momentum equation for Couette ow between concentric cylinders Using the assumptions on the Couette velocity prole between the rotating concentric cylinders (Sec.3.2.3) and the expression for the momentum equation and the Laplacian operator in cylindrical coordinates:
In the system of two rotating cylinders (Sec.3.2.3) consider the torque applied to the inner rotating cylinder when the outer cylinder is xed ( ). What is the power required to rotate the inner cylinder? Problem 3.7.7 Flow between parallel plates under pressure
A viscous uid with viscosity ( ) is driven between two parallel plates apart by an imposed pressure gradient of . The upper plate is moving with velocity . Find the volume ow rate per 1m of the plates width. What pressure gradient will cause the ow to reverse? Problem 3.7.8 Verifying the Stokes solution Verify that the solution (3.73) satises the equation (3.72). Problem 3.7.9 Stokes velocity and shear stress Using Stokes solution for the stream function (3.73), obtain velocity components (3.74), (3.75) and the shear stress component (3.77).
FX d (X F s R I X d V
p n
`X s l
3.7. PROBLEMS
103
A sphere of density is dropped into oil of density and viscosity . Estimate the terminal velocity of the sphere if its diameter is (a) , (b) and (c) . Problem 3.7.11 Boundary layer analysis for cubic velocity prole le: Repeat the boundary layer analysis of Sec.3.5.1 with assumed velocity pro-
Compute ( .
), (
), (
), (
), (
), where
Hint: Assume
A thin equilateral triangle plate is immersed parallel to a stream of air with the at temperature and pressure . Assuming velocity laminar ow, estimate the drag on this plate. Problem 3.7.13 Boundary layer equations Derive Prandtl boundary layer equations (3.110) - (3.113):
r Y e s { Y 2 3e
4 d
2 4e m 4e m k 4e 2 2 wv l d
d
{ld s
m R }I A 2 2 m R ( X I 3e m R ( X I 3e m R ( X I w v l A
d yds s F uds r Y % Y
FX l d
V
W ( X
104
The transition to turbulence for the case of rotating cylinders is usually measured in terms of Taylor number and occurs at the critical value of:
4.2
Turbulence Modeling
Lets consider the incompressible forms of the mass and momentum equations, (2.4), (2.24): 105
2 l 5e
t d
W p e p e e 6s w R n n I R I
t d
W 3e 2
106
(4.1) (4.2)
where we introduced an abbreviation for pressure-to-density ratio: also use the conservative expression for the pressure term: which is true due to continuity (4.1).
The solution to this equation system for high Reynolds numbers will result in a turbulent ow eld. This eld is highly unsteady with a broad spectrum of eddies, which makes it difcult, if not impossible to resolve it on even the most powerful computers. However, for some limited range of Reynolds numbers the solution can be obtained numerically. The technique that uses this direct approach of computing turbulence is called direct numerical simulation (DNS). The difculty with this approach is that in order to reproduce all the turbulent eddies from the largest to the smallest, the simulation has to resolve the smallest space and time scales of turbulence. This in turn may require a very ne grid and time-resolution.
where is the molecular viscosity, and is the eddy viscosity, associated with the cumulative action of unresolved eddies on the resolved large eddies. This approach to turbulence modeling is called Large eddy simulation. In the rst proposed LES model, Smagorinsky model, the turbulent viscosity is set proportional to the strain-rate tensor (1.9), and the computational grid size:: (4.4)
where is the grid cell size, and the coefcient of proportionality, Smagorinsky constant.
R A AY X I u A A W
} f A W b k R A A I b & A q f A }p r R G F U F I r R I W W W b W W
. We ,
, called the
In the Smagorinsky model the constant is considered xed, and is selected by matching the experimental data or those produced by DNS. In more sophisticated LES models is no longer considered a constant, and its value is determined in a more complex way, for example by comparing some integral measures produced from solutions obtained from space-averaged equations using different averaging lters [13]. The common feature of LES models is that they are based on solving for a time-dependent ow eld using space-averaged form of the Navier-Stokes equation.
(4.5)
(4.6) (4.7)
(4.8)
The averaging time is usually much greater than the time scale of turbulent uctuations. Since the time averaging interval does not stretch to innity, the resultant average quantities can still be slowly varying functions of time. But the time scale of these variations will be larger than largest turbulence time scales. Under this assumptions decomposition (4.8) is called the Reynolds decomposition.
It is also common to denote the uctuating velocity as
Lets use the Reynolds assumption that any turbulent quantity, decomposed it into the time average and uctuating components:
9 9 SR4 @( I g b R ( I SR4 @( I
107
, can be
and after and time averaging the latter and the continuity relation (4.1), and using (4.8), we obtain:
(4.10)
As can be seen, both the mean and uctuating elds satisfy the continuity. The last equation is called the Reynolds averaged Navier-Stokes equation (RANS), thus the name of the approach. As can be seen the equation contains and extra unknown term composed of derivatives of the so-called the Reynolds stress . tensor: To close the new equation system one needs to formulate extra equations for the components of Reynolds stress tensor (closure). One of the simplest and rst closure was suggested by Boussinesq, and is called the Boussinesq approximation. In this approximation the Reynolds stress tensor is considered proportional to the mean velocity gradient:
(4.11) (4.12)
(4.13)
where we introduced the mean strain rate tensor, analogous to (1.9). Quantity dened by (4.12) is called the Turbulent kinetic energy, and is called the eddy viscosity,. Both and represent two new unknown variables in the model. Empirical algebraic relations can be devised for and connecting them to and length-scales of the problem, as is done in the mixing-length theory [2], or in a more sophisticated RANS models discussed below. On the other end of the turbulence closure spectrum is the Reynolds stress model (RSM). In this model each component of the Reynolds stress tensor is
R b I W b R b I k R b I R b I b & b &
f W b R I b R I b & q f q f U R m b f I dl kG A A dl G W l U l k W
@" T f
109
obtained from a separate equation. Since there are six independent components, there should be six equations. In the full Reynolds stress model these equations are represented by the PDEs of the transport type. In an approximate version of RSM - algebraic Reynolds stress model these equations are given by algebraic relations. To obtain the equations for Reynolds stress tensor, lets rst subtract (4.10) from (4.9):
(4.14)
b A j A A b j A W j A k R j j I b b j j & A
5 jA jA jA j j j u A W b mA k R I b f b u b & A
j qA u W b k R I b f b f b &
: of this equation, we get:
This is the equation for the components of the Reynolds stress tensor. Since the Reynolds stress tensor is symmetric, it has only six independent components, and the number of equations is 6. As can be seen the equation contains 3-rd order correlations: . One can write an equation for as well, but it will depend on 4-th order correlations, and so on. In practice, an empirical relationship is proposed, that links the third order tensor to the second order tensor . This relationship is called closure. In fact all the terms on the RHS
j A A A
j A A A
A A
j A A A
R }j j
b j b j j Sj b b } ff I j W b R R j I I j & b
changing the order of time differentiation and time averaging, and canceling terms we obtain:
b j b j I bj bj R j I b R b I j R R b I I b f R j I b bj
111
of (4.16) require some kind of closure to make the problem complete. If such closures are established, the obtained equation system will constitute a turbulence model. In this particular case, when the equation system provides PDEs for Reynolds stress tensor components, the corresponding turbulence model is called Reynolds stress model (RSM). Considering that the Reynolds stress tensor is symmetric, this model may include as many as 12 PDEs: 3 - velocity, 1 - pressure, 6 - Reynolds stress tensor. The 12-th equation is the one for the turbulent dissipation rate, which will be discussed in the next section. Two equation turbulence models The RSM model described above may be prohibitively expensive in terms of the number of equations and the complexity of implementation. In this case simpler models can be used, which are based on smaller number of equations. The rst step to reduce the number of equations is to relate the components of the Reynolds tensor to the mean velocity gradients following the Boussinesq approximation (4.11). Then one can formulate a separate transport equation for and an algebraic relation for as a function of the latter and the mean velocity gradients. This approach will constitute a one equation turbulence model. A more popular approach is to formulate two transport equations: one for turbulent kinetic energy, and another for its dissipation rate . The eddy viscosity, is then related to and as
(4.17)
which can be shown from dimensional reasoning. Indeed, if represents the rate of change of : then its physical dimensions should be , Considering that , and , we can obtain (4.17). This approach is called the turbulence model (KE) [14, 15], which is the most popular turbulence model for engineering computations today. The equation for is formulated as a transport equation of the form:
(4.18)
where is the effective eddy viscosity of . This equation is solved with the boundary conditions of zero at the walls. Usually a non-zero is set at the
F F tX r X
R m b u I u W b R P W I 4
FX r c W
W
F r X r c
112
open boundaries, where its value is related to level of turbulent uctuations expected in each particular ow case. The equation for the turbulence dissipation rate, , is written in analogy to the one for . Namely, we multiply the -equation by , and introduce the effective eddy viscosity of : to obtain:
(4.19)
where we introduced two new constants . Boundary conditions on can be obtained from those on , by relating to through equation (4.18) where the non-steady term should be set to zero at the boundary: . (diffusivNow equations (4.18) and (4.19) have three effective viscosities: ity of momentum), (diffusivity of turbulent kinetic energy, ), and (diffusivity of turbulence dissipation rate, ). It should be noted that unlike the molecular viscosity, , which is a property of the uid and is usually independent on coordinates or time2 , all turbulent viscosities are functions of space, and are therefore represent the new dependent variables of the problem along with and . Another assumption of the model is that with different proportionality constants: and are both proportional to
The system of equations (4.18), (4.19), (4.20), (4.21), (4.17) is closed and constitutes the turbulence model. It has 5 empirical constants:
the values of which are determined by comparison of computations with experimental data. KE model belongs to the class of two-equation turbulence models. Another model [16], where instead important two equation turbulence model is the of the turbulence dissipation rate a turbulence frequency scale, , is used as an extra variable.
2
W W 4X p r D9 r R m tb f I f 9
n
d ) d ) l d r ) d p ) s s s s s
W W
W W W W
p W b R W I 4 W
Bibliography
[1] Y.A. Cengel and M.A. Boles. Thermodynamics. McGraw-Hill, Inc., 2002. [2] Frank White. Viscous Fluid Flow. Second Edition. WCB/McGraw-Hill, 1991. [3] K.Jr. Wark. Advanced Thermodynamics for Engineers. McGraw-Hill, Inc., 1995. [4] C.A.J Fletcher. Computational techniques for uid dynamics. SpringerVerlag, 1991. [5] C.R. Chester. Techniques in partial differential equations. McGraw-Hill, NY, 1971. [6] J.H. Ferziger and M. Peric. Computational Methods for Fluid Dynamics. Springer Verlag, 1997. [7] L.D. Landau and E.M. Lifshitz. Fluid Mechanics. Course of Theoretical Physics, volume 6. Butterworth-Heinemann; 2nd edition, 1987. [8] O. Reynolds. An experimental investigation of the circumstances which determine whether the motion of water shall be direct or sinuous, and of the law of resistance in parallel channels. Royal Society, Phil. Trans., 1883. [9] E. Buckingham. Model experiments and the form of empirical equations. Trans. ASME, 37, 1915. [10] F. White. Fluid Mechanics. Fifth Edition. WCB/McGraw-Hill, 2002. [11] E. Buckingham. On physically similar systems: Illustrations of the use of dimensional equations. Phys. Rev., 4(4):345376, 1914. [12] C.W. Pseen. Ueber die stokessche formel und ueber eine verwandte aufgabe in der hydrodynamik. Ark. f. Math. Astron. och Fys., 6(29), 1910. 113
114
BIBLIOGRAPHY
[13] M. Germano. Turbulence: the ltering approach. Journal of Fluid Mechanics, 238:325336, 1992. [14] B.E. Launder and D.B. Spalding. The numerical computation of turbulent ows. Computer Methods in Applied Mech. and Eng., 3:269289, 1974. [15] W.P. Jones and B.E. Launder. The prediction of laminarization with a twoequation model of turbulence. Int. J. Heat Mass Transfer, 15:301314, 1972. [16] D.C. Wilcox. Turbulence modeling for CFD. DCW Industries, Inc., 1993. [17] Barry Spain. Tensor Calculus. Oliver and Boyd, 1965. [18] J.L. Synge and A. Schild. Tensor Calculus. Dover Publications, 1969. [19] P. Morse and H. Feshbach. Methods of Theoretical Physics. McGraw-Hill, New York, 1953.
115
116
There are two aspects of tensors that are of practical and fundamental importance: tensor notation and tensor invariance. Tensor notation is of great practical importance, since it simplies handling of complex equation systems. The idea of tensor invariance is of both practical and fundamental importance, since it provides a powerful apparatus to describe non-Euclidean spaces in general and curvilinear coordinate systems in particular. A denition of a tensor is given in Section A.1. Section A.2 deals with an important class of Cartesian tensors, and describes the rules of tensor notation. Section A.3 provides a brief introduction to general curvilinear coordinates, invariant forms and the rules of covariant differentiation.
A.1
, and a single real time, Consider a space of real numbers of dimension , . Continuum properties in this space can be described by arrays of different dimensions, , such as scalars ( ), vectors ( ), matrices ( ), and general multi-dimensional arrays. In this space we shall introduce a coordinate system, , as a way of assigning real numbers1 for every point of space There can be a variety of possible coordinate systems. A general transformation rule between the coordinate systems is (A.1)
Consider a small displacement . Then it can be transformed from coordinate system to a new coordinate system using the partial differentiation rules applied to (A.1):
(A.2)
This transformation rule2 can be generalized to a set of vectors that we shall call contravariant vectors:
(A.3)
1
Super-indexes denote components of a vector ( ) and not the power exponent, for the reason explained later (Denition A.1.1) 2 The repeated indexes imply summation (See. Proposition A.21)
l
d fe a
( ( ( ( ( ( R f G(yss p ( I ( (
( (
R!
f p E ( 7 (
117
That is, a contravariant vector is dened as a vector which transforms to a new coordinate system according to (A.3). We can also introduce the transformation matrix as:
(A.4)
(A.5)
Transformation rule (A.3) will not apply to all the vectors in our space. For example, a partial derivative will transform as:
(A.6)
that is, the transformation coefcients are the other way up compared to (A.2). Now we can generalize this transformation rule, so that each vector that transforms according to (A.6) will be called a Covariant vector:
(A.7)
This provides the reason for using lower and upper indexes in a general tensor notation. Denition A.1.1 Tensor Tensor of order is a set of numbers identied by integer indexes. and an -order tenFor example, a 3rd order tensor can be denoted as sor can be denoted as . Each index of a tensor changes between 1 and n. For example, in a 3-dimensional space (n=3) a second order tensor will be represented by components. Each index of a tensor should comply to one of the two transformation rules: (A.3) or (A.7). An index that complies to the rule (A.7) is called a covariant index and is denoted as a sub-index, and an index complying to the transformation rule (A.3) is called a contravariant index and is denoted as a super-index.
j G
( ( ( ( ( ( (
( (
|a ( X
( (
r
118
Each index of a tensor can be covariant or a contravariant, thus tensor is a 2-covariant, 1-contravariant tensor of third order. Tensors are usually functions of space and time:
Remark A.1.2 Tensor character of coordinate vectors are not tensors, since generally, they are not Note, that the coordinates transformed as (A.5). Transformation law for the coordinates is actually given by (A.1). Nevertheless, we shall use the upper (contravariant) indexes for the coordinates.
is dened as
(A.8)
From the denition (A.1.3) and the summation convention (A.21), follows
(A.10)
Gj f
( SR4 9 f G(yss p ( T T I
d UU vv 55
k2G
( G (
G
119
Assume that there exists the transformation inverse to (A.5), which we call
(A.11)
(A.12)
From this relation and the independence of coordinates (A.9) it follows that , namely:
(A.13)
A.2
Cartesian Tensors
Cartesian tensors are a sub-set of general tensors for which the transformation matrix (A.4) satises the following relation:
(A.14)
(A.15)
(see Problem A.4.3), which means that both (A.5) and (A.6) are transformed with the same matrix . This in turn means that the difference between the covariant and contravariant indexes vanishes for the Cartesian tensors. Considering this we shall only use the sub-indexes whenever we deal with Cartesian tensors.
je
( ( G j ( j ( j j
j j ( ( j ( ( j ( ( ( ( j e
( ( j j ( (
( ( e
( (
( e (
j j e
120
A.2.1
Tensor Notation
Tensor notation simplies writing complex equations involving multi-dimensional objects. This notation is based on a set of tensor rules. The rules introduced in this section represent a complete set of rules for Cartesian tensors and will be extended in the case of general tensors (Sec.A.3). The importance of tensor rules is given by the following general remark: Remark A.2.1 Tensor rules Tensor rules guarantee that if an expression follows these rules it represents a tensor according to Denition A.1.1. Thus, following tensor rules, one can build tensor expressions that will preserve tensor properties of coordinate transformations (Denition A.1.1) and coordinate invariance (Section A.3). Tensor rules are based on the following denitions and propositions. Denition A.2.2 Tensor terms A tensor term is a product of tensors.
(A.16)
(A.17)
Generally the terms in the expression may come with plus or minus sign. Proposition A.2.4 Allowed operations The only allowed algebraic operations in tensor expressions are the addition, subtraction and multiplication. Divisions are only allowed for constants, like . If a tensor index appears in a denominator, such term should be redened, so as not to have tensor indexes in a denominator. For example, should be redened as: .
t X d
j j G b
j j G
For example:
X d
Xd
121
(A.18)
Denition A.2.6 Free indexes A free index is any index that occurs only once in a tensor term. For example, index is a free index in the term (A.16). Proposition A.2.7 Free index restriction Every term in a tensor equality should have the same set of free indexes. For example, if index is a free index in any term of tensor equality, such as (A.18), it should be the free index in all other terms. For example
is not a valid tensor equality since index is a free index in the term on the RHS but not in the LHS. Denition A.2.8 Rank of a term A rank of a tensor term is equal to the number of its free indexes. For example, the rank of the term
It follows from (A.2.7) that ranks of all the terms in a valid tensor expression should be the same. Note, that the difference between the order and the rank is that the order is equal to the number of indexes of a tensor, and the rank is equal to the number of free indexes in a tensor term. Proposition A.2.9 Renaming of free indexes Any free index in a tensor expression can be named by any symbol as long as this symbol does not already occur in the tensor expression.
j f j b j j kG
is equal to 1.
kG
j j G
122
(A.19) is equivalent to
(A.20)
Here we replaced the free index with . Denition A.2.10 Dummy indexes
A dummy index is any index that occurs twice in a tensor term. For example, indexes
Proposition A.2.11 Summation rule Any dummy index implies summation, i.e.
(A.21)
Proposition A.2.12 Summation rule exception If there should be no summation over the repeated indices, it can be indicated by enclosing such indices in parentheses. For example, expression:
R % I
5 j j f w 6w 5 G k QP H QP H
is expressed in tensor
U q9 9 9
123
In case if an index occurs more than twice in a term this term should be redened so as not to contain more than two occurrences of the same index. For example, term should be rewritten as , where is dened as with no summation over in the last term. Proposition A.2.16 Renaming of dummy indexes Any dummy index in a tensor term can be renamed to any symbol as long as this symbol does not already occur in this term. For example, term . is equivalent to , and so are terms
Remark A.2.17 Renaming rules Note that while the dummy index renaming rule (A.2.16) is applied to each tensor term separately, the free index naming rule (A.2.9) should apply to the whole tensor expression. For example, the equality (A.19) above
The components of a third order permutation tensor are dened to be equal to 0 when any index is equal to any other index; equal to 1 when the set of
j Gqp
j j G
j
j j t
j j
kG
j j j j j j P H P H
and
124
indexes can be obtained by cyclic permutation of 123; and -1 when the indexes can be obtained by cyclic permutation from 132. In a mathematical language it can be expressed as:
(A.23) where
is a permutation group of a triple of indexes abc, i.e. . For example, the permutation group of 123 will consist of three combinations: 123, 231 and 312, and the permutation group of 123 consists of 132, 321 and 213. Corollary A.2.19 Permutation of the permutation tensor indexes From the denition of the permutation tensor it follows that the permutation of any of its two indexes changes its sign:
(A.24)
A tensor with this property is called skew-symmetric. Corollary A.2.20 Vector product A vector product (cross-product) of two vectors in vector notation is expressed as
(A.25)
Tensor expression (A.26) is more accurate than its vector counterpart (A.25), since it explicitly shows how to compute each component of a vector product.
R 2e I W
E ie 9 D 9 2fe2e 7 R 2e I W
125
we have:
(A.28)
from which we conclude that (A.27) is true. Theorem A.2.23 Tensor identity The following tensor identity is true:
(A.29) Proof
This identity can be proved by examining the components of equality (A.29) component-by-component.
j j 2 Sp j Uqp
j j GSp j D j Sp
j j GSp j j GSp
into
and
j j Gm j j GSp p
into
j j GSp G
in the RHS of this expression, according
G
126
Corollary A.2.24 Vector identity Using the tensor identity (A.29) it is possible to prove the following important vector identity:
(A.31)
that is, the index of the spatial component that the derivation is done over is delimited by a comma (,) from other indexes. For example, is a derivative of a second order tensor . Denition A.2.27 Nabla Nabla operator acting on a tensor
is dened as
(A.32)
j yGt
U
127
Even though the notation in (A.31) is sufcient to dene the derivative, In some instances it is convenient to introduce the nabla operator as dened above. Remark A.2.28 Tensor derivative In a more general context of non-Cartesian tensors the coordinate independent derivative will have a different form from (A.31). See the chapter on covariant differentiation in [17]. Remark A.2.29 Rank of a tensor derivative The derivative of a zero order tensor (scalar) as given by (A.31) forms a rst order tensor (vector). Generally, the derivative of an -order tensor forms an order tensor. However, if the derivation index is a dummy index, then the rank of the derivative will be lower than that of the original tensor. For example, the rank of the derivative is one, since there is only one free index in this term. Remark A.2.30 Gradient Expression (A.31) represents a gradient, which in a vector notation is
d b r qr b p p ( 6 f ( d bd b ( (
f 6R ~ I
G k u (
mG
d b
128
Remark A.2.33 Laplace operator : The Laplace operator in vector notation is represented in tensor notation as
Examples (A.2.30), (A.2.32) and (A.2.33) clearly show that tensor notation is more concise and accurate than vector notation, since it explicitly shows how each component should be computed. It is also more general since it covers cases that dont have representation in vector notation, for example: .
A.3
Curvilinear coordinates
In this section 3 we introduce the idea of tensor invariance and introduce the rules for constructing invariant forms.
A.3.1
Tensor invariance
The distance between the material points in a Cartesian coordinate system is . The metric tensor, is introduced to generalize the computed as notion of distance (A.39) to curvilinear coordinates. Denition A.3.1 Metric Tensor The distance element in curvilinear coordinate system is computed as:
(A.35) where
Thus, if we know the metric tensor in a given curvilinear coordinate system then the distance element is computed by (A.35). The metric tensor is dened as a tensor since we need to preserve the invariance of distance in different coordinate
3
In this section we reinstall the difference between covariant and contravariant indexes.
j j g
G`
( ( G r
( ( r
G`
129
systems, that is, the distance should be independent of the coordinate system, thus: (A.36)
The metric tensor is symmetric, which can be shown by rewriting (A.35) as follows:
, we get:
The metric tensor is also called the fundamental tensor. The inverse of the metric tensor is also called the conjugate metric tensor, , which satises the relation: (A.38)
Let be a Cartesian coordinate system, and - the new curvilinear coordinate system. Both systems are related by transformation rules (A.5) and (A.11). Then from (A.36) we get:
(A.39)
When we transform from a Cartesian to curvilinear coordinates the metric tensor in curvilinear coordinate system, can be determined by comparing relations (A.39) and (A.35):
(A.40)
where we rst swapped places of and , and then renamed index and into . We can rewrite the equality above as:
Gt ( ( ( j ( ( ( ( ( j ( ( j ( ( j ( ( r
( ( ( ( G G r
into
130
(A.41)
Using these expression one can compute nate systems (see Problem A.4.6). Denition A.3.2 Conjugate tensors
For each index of a tensor we introduce the conjugate tensor where this index is transfered to its counterpart (covariant/contravariant) using the relations:
(A.42) (A.43)
Conjugate tensor is also called the associate tensor. Relations (A.42), (A.43) are also called as operations of raising/lowering of indexes. Remark A.3.3 Tensor invariance Since the transformation rules dened by (A.1.1) have a simple multiplicative character, any tensor expression should retain its original form under transformation into a new coordinate system. Thus if an expression is given in a tensor form it will be invariant under coordinate transformations. Not all the expressions constructed from tensor terms in curvilinear coordinates will be tensors themselves. For example, if vectors and are tensors, 4 then is not generally a tensor . However, if we consider the same operation on a contravariant tensor and a covariant tenso then the product will form an invariant:
(A.44)
Thus in curvilinear coordinates we have to rene the denition of the scalar product (Corollary A.2.13) or the index contraction operation to make it invariant (Problem A.4.12).
4
For Cartesian tensors any product of tensors will always be a tensor, but this is not so for general tensors
U Ut j( ( ( j ( U
and
6
U` G
131
The invariant form of the scalar product between two covariant vectors and is . Similarly, the invariant form of a scalar product between two contravariant vectors and is , where is the metric tensor (A.40) and is its conjugate (A.38). Corollary A.3.5 Two forms of a scalar product According to (A.42), (A.43) the scalar product can be represented by two invariant forms: and . It can be easily shown that these two forms have the same values (see Problem A.4.12). Corollary A.3.6 Rules of invariant expressions To build invariant tensor expressions we add two more rules to Cartesian tensor rules outlined in Section A.2.1: 1. Each free index should keep its vertical position in every term, i.e. if the index is covariant in one term it should be covariant in every other term, and vise versa. 2. Every pair of dummy indexes should be complementary, that is one should be covariant, and another contravariant. For example, a Cartesian formulation of a momentum equation for an incompressible viscous uid is
(A.45)
, and
j T
A j j A
U`
j Aj A j T W b Y j u b & A
A fjj T W b Y j u A j b & A
DG`
G w U
G T j
132
A.3.2
Covariant differentiation
A simple scalar value, , is invariant under coordinate transformations. A partial derivative of an invariant is a rst order covariant tensor (vector):
However, a partial derivative of a tensor of the order one and greater is not generally an invariant under coordinate transformations of type (A.7) and (A.3). In curvilinear coordinate system we should use more complex differentiation rules to preserve the invariance of the derivative. These rules are called the rules of covariant differentiation and they guarantee that the derivative itself is a tensor. According to these rules the derivatives for covariant and contravariant indices will be slightly different. They are expressed as follows:
and is also known in tensor calculus as Christoffels symbol of the second kind [17]. Tensor represents the inverse of the metric tensor (A.38). As can be seen differentiation of a single component of a vector will involve all other components of this vector. In differentiating higher order tensors each index should be treated independently. Thus differentiating a second order tensor, , should be performed as:
and as can be seen also involves all the components of this tensor. Likewise for
G`
G
is dened as
( | E j | 7 | E j | 7 Gj j mG
( j y j b ( j yGj f
( q
G
133
(A.48)
(A.49)
Despite their seeming complexity, the relations of covariant differentiation can be easily implemented algorithmically and used in numerical solutions on arbitrary curved computational grids (Problem A.4.8). Remark A.3.7 Rules of invariant expressions As was pointed out in Corollary A.3.6, the rules to build invariant expressions involve raising or lowering indexes (A.42), (A.43). However, since we did not introduce the notation for contravariant derivative, the only way to raise the index . of a covariant derivative, say , it to use the relation (A.42) directly, that is: For example, we can re-formulate the momentum equation (A.45) in terms of contravariant free index as:
(A.50)
where the index of the pressure term was raised by means of (A.42). Using the invariance of the scalar product one can construct two important differential operators in curvilinear coordinates: divergence of a vector (A.51) and Laplacian, (A.55).
5
U
T TT E 1 b b T E T b T TT E f 7 7 b iiii b T E Tf 7 7 b j T ( T 0 T &1 T a j j E | b E | b j ( j | 7 | 7 G G
j b Y j A b A A j T W j j j & j ! j ! 5
G
we have:
134
(A.52)
(A.53) where
(A.55)
The denitions (A.3.8), (A.3.9) of differential operators are invariant under coordinate transformations. They can be programmed using a symbolic manipulation packages and used to derive expressions in different curvilinear coordinate systems (Problem A.4.9).
t G` (m m d (m m d
.
( j E j b 7 j ! j m G
is dened as a divergence of its
b (
135
Consider three unit vectors, , each directed along one of the coordinate axis (tangential unit vectors), that is: (A.56) (A.57) (A.58)
The condition of orthogonality means that the scalar product between any two of these unit vectors should be zero. According to the denition of a scalar product (Denition A.3.4) it should be written in form (A.44), that is, a scalar product between vectors and can be written as: or . Lets use the rst form for deniteness. Then, applying the operation of rising indexes (A.42), we can express the scalar product in contravariant components only:
(A.59)
where we used the symmetry of , (A.37). Since vectors and were chosen to be non-zero, we have: . Applying the same reasoning for scalar products of other vectors, we conclude that the metric tensor has only diagonal components non-zero5 : (A.60) Lets introduce stretching factors, ponents of : (A.61)
Now, consider the scalar product of each of the unit vectors (A.56)-(A.58) with itself. Since all vectors are unit, the scalar product of each with itself should be one:
5
pp p ) r }p R I ) r }p R r` I kr ) r }p R I r
e e
QP H DG kG
E D9 9 7 E 9 fe 9 7 e E 9 r 9 p 7
B9 fe9
G
136
Now, consider the rst term above and substitute the components of (A.56). The only non-zero term will be:
and consequently:
(A.62)
where the negative solution identies a vector directed into the opposite direction, and we can neglect it for deniteness. Applying the same reasoning for each of the tree unit vectors , we can rewrite (A.56), (A.57) and (A.58) as:
which means that the components of unit vectors in a curved space should be scaled with coefcients . It follows from this that the expression for the element of length in curvilinear coordinates, (A.35), can be written as: (A.66)
d 6 e e 6
from
D9 ge9
137
Consider a direction in space determined by a unit vector . Then the physical component of a vector in the direction is given by a scalar product between and (Denition A.3.4), namely:
(A.69)
Suppose the unit vector is directed along one of the axis: From (A.63) it follows that:
where is dened by (A.61). Thus according to (A.69) the physical component of vector in direction 1 in orthogonal coordinate system is equal to:
or, repeating the argument for other components, we have for the physical components of a covariant vector:
(A.70)
E p 9 D9` 7
G QP H QP H G
X 9 r X 9 p X p r
R I
p r Br 9 p
G R I
p X p R d I
d QP H X QP H p Xd p p
. , we have:
138
General rules of covariant differentiation introduced in (Sec.A.3.2) simplify considerably in orthogonal coordinate systems. In particular, we can dene the nabla operator by the physical components of a covariant vector composed of partial differentials:
(A.71)
where the parentheses indicate that theres no summation with respect to index . In orthogonal coordinate system the general expressions for divergence (A.53) and Laplacian (A.55)) operators can be expressed in terms of stretching factors only [19]:
(A.72)
Important examples of orthogonal coordinate systems are spherical and cylindrical coordinate systems. Consider the example of a cylindrical coordinate system: and :
According to (A.40) only few components of the metric tensor will survive (Problem A.4.5). Then we can compute nabla, divergence and Laplacian operators according to (A.71), (A.52) and (A.55), or using simplied relations (A.72)(A.73):
p f ( QP H ( d H ( QP d
E q(9 r S(9 p ( 6 ( 7
139
Note, that instead of using the contravariant components as implied by the general denition of the divergence operator (A.51) we are using the covariant components as dictated by relation (A.70). The expression of the Laplacian becomes:
The advantages of the tensor approach are that it can be used for any type of curvilinear coordinate transformations, not necessarily analytically dened, like cylindrical (C.64) or spherical. Another advantage is that the equations above can be easily produced automatically using symbolic manipulation packages, such as Mathematica (wolfram.com) (Problems A.4.6,A.4.7,A.4.9). For further reading see [17, 18].
b p ( d p ( d
x d b r C b p d b p p ( ( ( p ( d b b rr d C t 9 d 9 C rR I b b r Rr ( I r
rR I b R I rd p r p b rr R r ( I ( b r R ( I r r d r r
x b p p ( 5
140
A.4
Problems
Check if the following Cartesian tensor expressions violate tensor rules:
Construct a valid Cartesian tensor expression, consisting of three terms, each including some of the four tensors: . Term 1 should include tensors only, term 2 tensor and term 3 tensors . The expression should have 2 free indexes, which should always come rst among the indexes of a tensor. The free indexes should be at A and B in the rst term, at B and C in the second term and C and D in the last term. How many different tensor expressions can be constructed? Problem A.4.3 Cartesian identity Prove identity (A.15) Problem A.4.4 Vector identity Using tensor identity (A.29):
Cylindrical coordinate system (C.64) is given by the following transformation rules to a Cartesian coordinate system, :
9 9
E 9 9 7 %BCqB@( (
R ~ ~ I ~ R ~ ~ I ~ R ~ ~ I ~
U 9 D9 G `j Gt 9
B9 9
A.4. PROBLEMS
141
Obtain the components of the metric tensor (A.40) (A.38) in cylindrical coordinates. Problem A.4.6 Metric tensor in curvilinear coordinates
Using Mathematica Compute the metric tensor, , (A.40) and its conjugate, , (A.38) in spherical coordinate system ( ):
(A.73)
Problem A.4.7 Christoffels symbols with Mathematica Using the Mathematica package, write the routines for computing Christoffels symbols. Problem A.4.8 Covariant differentiation with Mathematica Using the Mathematica package, and the routines developed in Problem A.4.7 write the routines for covariant differentiation of tensors up to second order. Problem A.4.9 Divergence of a vector in curvilinear coordinates Using the Mathematica package and the solution of Problem A.4.8, write the routines for computing divergence of a vector in curvilinear coordinates. Problem A.4.10 Laplacian in curvilinear coordinates Using the Mathematica package and the solution of Problem A.4.8, write the routines for computing the Laplacian in curvilinear coordinates. Problem A.4.11 Invariant expressions
G
Ut
D29 9
eG C d d d efd er ffe e d (
C d efd eG (
142
not:
(A.74)
(A.75)
(A.76)
Laplace equation in different coordinate systems. When solving boundary value problems in more than one dimension it is often necessary to use other coordinate systems than the Cartesian. It is then important to be able to express the Laplacian operator in these coordinate systems. We rst consider POLAR COORDINATES
PROOF; Let us for now apply the convention that subscript implies partial differentiation e.g.
b b D9 A R D9 r I A r r d R D9 I A d R D9 r I A r R I r d yfd Ge (
143
x A A
d er Y (
which is the desired result! CYLINDRICAL COORDINATES It is easy to generalize the result for polar coordinates to cylindrical coordinates
b b m b D00 A A Ar R D9 r I A r r d R B9 I A d R D9 r I A r m p qA b p b x x b x ( l A ( pA ( l psA ( xA r A r r r pqA ( A pA ( psA xA D00 A bx b x p b x l r l r rr ( 0 D00 ( l ( R l 0 I d b 0 r R r I r R I d r ( d D00 r r ( r b ( r ( 0 ( p f b ( r r ! 0 pqb 0 0 xsA 0 0 qA pA p p r 0 R 0 BI pb 0 xA 0 0 R A I xA x x 0D0 Dpqb 0 0 qA r bR D00 I b 0 0 R A I D00 A A p xA x 0R BqIp b 0 A 0 A R I A x
APPENDIX B. CURVILINEAR COORDINATE SYSTEMS
Finally we give without proof the result for Laplaces equation in spherical coordinates:
For proof see e.g. Chapter 8 of Riley et al. or Chapter 2 of Arfken and Weber.
ue d 9 R D%Dr 9 I A r r d R R D% I A r 9
b R D%D 9 9 R I I r d
d A ue d I d fe b 9 R DC%B9 I A r
SPHERICAL COORDINATES
146
A plate of mass with dimensions slides down a long incline at angle , on which there is a lm of oil of thickness , with viscosity . Assuming the plate does not deform the oil lm, and considering the weight of the plate to be much greater than the weight of the oil lm covering the same area, nd expressions for (1) the terminal sliding velocity , and (2) the time required for the plate to accelerate from rest to of the terminal velocity ( ). Also compute the numerical values for both and . Solution: To nd a terminal velocity consider the balance of forces on a plate moving with the terminal velocity :
147
q3 q3 X 3
s
RF l
Y `X j Y j
j fd j Y j j j Y j Y R j Y I j Y
j Y
I { X d
V
148
from which
To nd the time required to accelerate to the velocity equation of motion of the accelerated plate:
dividing by
, we get:
after rearranging:
Dene
as:
. Then:
q3 3
4 V $ p n~ | 3 fe V 3 V 4 3 d
4 d 3 V fe 3
d 4 V 3 fe 3
fe V 3 d
3 V fe d
, consider the
, we obtain:
Matlab solution: % Lubrication g=9.8 m=2 tet=10/180*pi mu=5e-3 h=3e-4 alp=0.99 A=0.3*0.4
% % % % %
4 V
where
. Thus
Perform the missing steps in (2.42). Solution: Using the tensor identity (A.29), we have:
(C.1)
(C.3) Solution:
(C.4)
(C.5)
R "" n n j " " I W j R m" n j m" y" n " j I W "y jn R j "m " j I W 6m" jn " Sp j GSp W j R A j m A I j Gmp dl 6 n p Y Bjj f A W & A Bjj m n W k & n
151 where the pressure term on the RHS became zero by symmetric identity (A.27). Now we can form another equation similar to (C.5): (C.6)
(C.7)
where the integration is over any closed loop inside the uid. Show that for irrotational ow ( time, we have:
Using (1.2), we can rewrite the last term on the RHS as:
the last equality stems from the fact that the integral of a total differential over a closed loop is zero. Thus we have for the rate of change of circulation:
Substituting the velocity derivative from the Euler equation in form (2.53), we have:
l A c A A c 4 ( A c
152
According to the Stokes theorem the contour integral on the RHS can be rewritten as the integral over the surface encircled by the contour:
where the last equality is due to the symmetry of and the symmetric identity (A.27). Thus, the time derivative of circulation is zero:
This is the law of circulation or Kelvins theorem, which states that in an ideal uid the velocity circulation round a closed contour is constant in time. Problem 2.7.5: Bernoullis equation. Using the energy equation (2.77):
(C.8)
(C.9)
Assuming steady state inviscid uid at constant temperature, we have according to (C.8):
j 6 j j Gmp d Y 6 j j GSp dY 4 b j j b b j R A Y I b R A Y I 4 T jA A G T m b R s I b 4 4 Y b 4 F a C b A A dl 4b
( c dY 4 b
153
(C.10)
On the other hand with the same assumptions (C.9) can be rewritten as
(C.12)
Collecting terms:
"
R ( I 4 Y b 4 6 A ` Y b A 6 A 4 A Y
R A A I 4 dl Y R ( ` I 4 Y 4
R A A I 4 dl R ( ` I 4 4
b ( ` A A dl 4
A ` Y b 6 A 4 Y
, we have
A b r& 4
4 dY 4
4 4 Y
, we have:
&
(C.13)
This is the strong form of a Bernoullis equation valid for steady-state inviscid ow. When gravity is directed opposite to z-axis the last term on the RHS will be equal to .
Method 2
where is the integration constant. Substituting it into (C.11) above and dividing by we obtain:
Expansing the expression for substantial derivative (1.7), rearranging terms, and using the steady state assumption , we have:
j R 2(j I
A b Aj q`u A j f A A
b 4 F a 6 ( A A dl
: :
j f A b Aj q & A ` Y b q A 4
b Y
C 6 ( Y
155 From which we obtain (C.13). Problem 2.7.6: Volume change inside a moving boundary Suppose that a region of space is enclosed by a moving boundary. The velocity of motion of the boundary, , is given at each point on the boundary. Show that the rate of change of the volume, , of that region will be equal to: Solution Consider any volume of a moving uid
where the integration is done over an arbitrary control volume inside the uid. Using identity (A.34), we can rewrite the latter as:
where is a coordinate vector at the boundary and is an element of the boundary surface area (2.1). Using the denition of velocity (1.1), we can then obtain the relation for a volume change:
(C.14)
Obtain explicit relations for the components of acceleration vectors in (2.94) in terms of . Solution Writing out (2.93):
A d 6 4 ( d 6 ( 4 d 4 3
R n n b n l b 4 b 4 Q I
( d 3 f ( d 3
3 f ( d 3
3 3
9 9 r 9 p 9 B9 r B9 p qn
156
in this case.
(C.17)
(C.18)
Consider a simple rotation with origin of the rotating coordinate system origin of :
A R y b u A I m A V b s b 4 4 s Y
qRR b n n b n l b 4 Q I I
R n n 4 n 4 I r r
R b b & Q I n
E 7 Sn9 9 n
n 4
pr e e
(C.15)
e pr n r n r p n l r n l
r p
as in (2.94), but now let the rotate with the same around the
Q
157 Select for the pressure scale. Determine the minimum number of dimensionless parameters. Write the equation using the Eckert number as one of the parameters:
Solution
(C.19)
Rearranging:
considering that the velocity scale should relate to length and time scales as , we have:
(C.20)
This equation contains 7 dimensional parameters: . According to the PI-theorem, the number of dimensionless parameters can be as low as three. A conventional choice of these parameters is:
V 9 9 D9 Y 9 qs9 qA9 ( ( A s s ( A R m b f A I m A A V Y b s Y b 4 A 4 s r
( s s ( A R y b f A I m A 4 A Y V r b s Y r b 4 A 4 s 4 r r ( ( A R m b f A I m A r A V b s r s b 4 Y 4 A 4 s s 4 Y r r Y Y A s A ( r Y ) s s ) A 6 A ) ( 6 ( ) 4 4 4
( VA Y s A
z s A
, and obtain:
r r
2 3e z
A r Y
4X ( A
2 3e S3e z x 2 A R m b f A I m A b s d b 4 4 s z
x @ V
APPENDIX C. SOLUTIONS TO PROBLEMS
and (C.20) becomes: 158
159 Chapter 3 Problem 3.7.1: Couette ow equations Show how to obtain equation (3.7) and (3.8). Solution Consider (3.2) and (3.3):
Using the denition (2.17), and since the only non-zero velocity components are and the only independent variable is , and considering continuity (2.4) and constancy of pressure , we arrive at (3.7) and (3.8). Problem 3.7.2: Couette plates solutions Solve equations (3.7) and (3.8) with the boundary conditions u(0) = 0 and u(H) = U, and and . Solution
(C.24) (C.25)
Aligning the coordinate origin with the lower plate (y=0) and solving the equation (C.24) with the boundary conditions: u(0) = 0 and u(H) = U, we have
E p ( g ( 7
R I A p p X Av R I A R I b g p R I A
6
A V b rs r r rA r
ps R }I s
A b R s A A & s I Y j j f b & A
A G T y b s p Y Bjj f A W q f A
s R }I s
E r A A 7
160
(C.26)
Consider a wide uid lm of constant thickness, , owing steadily due to the gravity down the inclined plate at angle . Find an analytical expression of a uid velocity distribution as a function of a distance from the plate surface: . Assuming the viscosity and the density of the uid are , and respectively, nd the maximum ow velocity and the volumetric ow rate, , per 1m of the plate. Atmospheric pressure can be considered constant.
Figure C.2: Flow of a liquid lm. Solution Selecting the coordinate parallel to the plate and in the direction of the steepest decent, and normal to the plate, we can conclude that the steady
s Dd V
{ s d
s l b b l s b V s ps r V R I l b p V s ps s R I s bg p b l R I s r V
`X (
F R I X d R I A
161 state solution should satisfy the following constraints: , . Then the momentum equation (3.2) reduces to:
and
The second condition is the negligible shear stress at the free surface:
thus:
Substituting and into (C.27) we obtain the nal solution for the axial ow velocity distribution with :
The volumetric ow rate is obtained by integrating the velocity along the thickness and the width of the lm:
R I A A
, and
~ D0 V b R Il R I fe d l Y y A V R I T
V A ~ |p R I a5 F Y
v R }I A b b r R I fe V d l Y A
0 A Vl w| r R I ue d Y R Il R I fV e ld Y A
R I fe d Y r A V
dV R I fe Y
X ( X
162
A cubic dependence of the ow-rate on means that the draining of the lm is strongly dependent on the lm thickness.
h=1.5e-3 % m: film thickness theta=pi/6 % RAD: inclination of the plate mu=1.6e-3 % kg/(m s): viscosity rho=8e2 % kg/m3: fluid density g=9.8 % m/s2: gravity acceleration umax=(rho*g*h2*sin(theta))/(2*mu) %=2.756 m/s Q=(rho*g*h3*sin(theta))/(3*mu) %=0.002756 m3/s Problem 3.7.4: Couette solution for non-Newtonian uids How will the solution (3.9) change for a non-Newtonian uid? Solution: For non-Newtonian uids the relation between stress and strain is nonlinear, thus, instead of (3.5) we have:
which after the assumptions of Couette ow between parallel plates (Sec.3.2.1) leads to a modied equation (3.7):
which for any non-zero and has the same solution as (3.7). So, the solution for a non-Newtonian uid will be the same. Problem 3.7.5: Momentum equation for Couette ow between concentric cylinders Using the assumptions on the Couette velocity prole between the rotating concentric cylinders (Sec.3.2.3) and the expression for the momentum equation and the Laplacian operator in cylindrical coordinates:
f R Bjj f A I W j j T p Y
f rA V r
0 A w|
The values of
and
Solution
, and using the assumptions of (Sec.3.2.3) Renaming for simplicity: that , the equation (C.28) simplies to:
(C.28)
In the system of two rotating cylinders (Sec.3.2.3) consider the torque applied to the inner rotating cylinder when the outer cylinder is xed ( ). What is the power required to rotate the inner cylinder? Solution: In this case we need to multiply the force applied at the surface of the cylinder by its radius:
(C.29)
where is the length of the cylinder. Since we are using a curvilinear coordinate system, the expression for the shear stress tensor used in the momentum equation is (A.45), which in cylindrical coordinate system has a form:
Considering the specic form of velocity dependence: and evaluating the expression above at , we have
p n
E 9 R I qBA9 q A p 7
e l T e T e e
x b w A v xx A b r A x A xx A r A x R x A I xw r A x A A v e x T p p x T
(Sec.3.2.3),
p sA A w qp A v b srp A r
T
R I A A
164
) is driven between two A viscous uid with viscosity ( parallel plates apart by an imposed pressure gradient of . The upper plate is moving with velocity . Find the volume ow rate per 1m of the plates width. What pressure gradient will cause the ow to reverse? Adapting the general expression for the volumetric ow rate (3.29) to the case of the ow in a square duct of unit width and height , we have:
Solution
which is a volumetric ux per 1m width of the duct. Substituting the solution (3.36), (3.37) into the above, and performing the integration, we obtain:
F X d (X F s R I X d V pe e r p e r e n V n e n & r r r pe e r p e r e n V r r
d p p ee XX e
p pr er
e e e b r ep n p p e r X p e n e b p e p e e n p e x y qp A T e X e d r e r e r r e qA
l x $ x w sp A v x p qA b x w qp A v T p X R I
R I A t
`X s l
165 where
The pressure gradient causing the ow reversal at the lower plate can be computed from (3.38). Below is the Octave (Matlab) solution to this problem. % Q=Integrate[u(y),y] % where u(y) is given by (CFM:Sec.3.2.5): % u = U/H*y*(C*y/H+1-C) % where C=H2/(2*mu*U)*dpdx % Integrating, we obtain: % Q=U*H*(3-C)/6 % Express all values in SI units: U=0.15 % m/s H=0.005 % m mu=1.4e-4 % kg/(m s) dpdx=-2.55 % kg/(m2 s2) C=H2/(2*mu*U)*dpdx Q=U*H*(3-C)/6 %=0.0005647 m2/s =(m3/s)/m % The pressure gradient that will case the % reverse flow: dpdx=2*mu*U/H2 %=1.68 kg/(m2 s2) Problem 3.7.8 Verifying the Stokes solution Verify that the solution (3.73):
(C.30)
(C.31) Solution:
R Ir 4 r r d b r r
e e r b e r ue d r d R D9 I rl
0 Vl r
166
Using the symbolic manipulation library GiNaC (www.ginac.org), we can easily check the consistency of the solution. Below is the example of a C++ code using the GiNaC library. /******************************************** This function computes the Laplacian of the Stokes solution in spherical coordinates. It should be compiled as c++ diff.cc -o diff -lcln -lginac Using GiNaC system (www.ginac.org) Running the executable produces the following output: Laplace(F)=3/2*sin(t)2*a*U*r(-1) Laplace2(F)=0 ********************************************/ #include <iostream> #include <ginac/ginac.h> using namespace std; using namespace GiNaC; ex Laplace ( const ex & F, const symbol & r, const symbol & t ) { return normal ( F.diff(r,2) + F.diff(t,2)/pow(r,2) - cos(t)*F.diff(t)/(sin(t)*pow(r,2)) ); } int main() {
167 symbol r("r"), t("t"), U("U"), a("a"); ex F=normal ( U/4*pow(a*sin(t),2)*(a/r-3*r/a+2*pow(r/a,2)) ), LF=Laplace(F,r,t), LF2=Laplace(LF,r,t); cout << "Laplace(F)=" << LF << endl << "Laplace2(F)=" << LF2 << endl; return 0; } Problem 3.7.9 Stokes velocity and shear stress Using Stokes solution for the stream function (3.73):
(C.32) (C.33)
(C.34) Solution:
Using the symbolic manipulation library GiNaC (www.ginac.org), we can easily obtain the desired expressions. Below is the example of a C++ code using the GiNaC library. /********************************************
p A x fe q l V w qp A x qb p A v px T d e
e e r b e r ue d r d R D9 I rl
(3.77):
e le
px T b d q qA b d p e fe l d eG A b d x e
168
This function computes velocity components and viscous stress of the Stokes solution. It should be compiles as: c++ stokes.cc -o stokes -lcln -lginac Using GiNaC system (www.ginac.org) Running the executable produces the following output: Ur=1/2*(2*r3+a3-3*a*r2)*cos(t)*U*r(-3) Ut=3/4*sin(t)*a*U*r(-1)-sin(t)*U+1/4*sin(t)*a3*U*r(-3) tau=-3/2*mu*sin(t)*a3*U*r(-4) ********************************************/ #include <iostream> #include <ginac/ginac.h> using namespace std; using namespace GiNaC; int main() { symbol r("r"), t("t"), U("U"), a("a"), mu("mu"); ex F=normal ( U/4*pow(a*sin(t),2) *(a/r-3*r/a+2*pow(r/a,2)) ), Ur= F.diff(t)/(pow(r,2)*sin(t)), Ut=-F.diff(r)/(r*sin(t)), dU= Ur.diff(t)/r + Ut.diff(r) - Ut/r; cout << "Ur=" << Ur << endl << "Ut=" << expand(Ut) << endl << "tau=" << mu*expand(dU) << endl; return 0; }
A sphere of density
r Y e s { Y
s r Y % Y
169
Solution is:
The balance of forces on the sphere when it reached the terminal velocity
where cient
are densities of the oil and sphere respectively, and the drag coefcan be rst assumed for a laminar case:
Or, alternatively the expression for Stokes drag force can be used:
e d 3e b 2 s b R I b l
l r3 Y p r R Y YI e
d5e 2 p d R Y Y I V 3 r
3r V
3V Y e X l
e
where
is a drag force:
d
d yds F uds V
. Estimate the terminal velocity of the sphere if its , (b) and (c) .
p Y 9 Y
170
% (a): d=0.1e-3 V=d2*(den1 - den0)*g/(18*mu) Re=den0*V*d/mu % (b): d=1e-3 V=d2*(den1 - den0)*g/(18*mu) Re=den0*V*d/mu % (c): d=1e-2 V=d2*(den1 - den0)*g/(18*mu) V1=0.0
% =2.5117
% Its a turbulent case so we loop until convergence: while (abs(V1-V)>0.1*abs(V)) V1=V Re=den0*V*d/mu Cd=24/Re + 6/(1+sqrt(Re)) + 0.4 Fb=pi*d3/6*(den1 - den0)*g Fd=Fb V=sqrt(Fd/(pi/4*d2*Cd*den0/2)) end
% m/s
171 Problem 3.7.11: Boundary layer analysis for cubic velocity prole le: Repeat the boundary layer analysis of Sec.3.5.1 with assumed velocity pro-
Compute ( .
), (
), (
), (
), (
), where
Using the denition of momentum thickness and introducing dimensionless coordinate , we obtain:
(C.35)
(C.36)
(C.37)
(C.38)
2 3e
2 4e m 4e m k 4e 2 2 wv l d
y l w A
b s s R s d d I R s d I d v A
X R X A d I
s X V td A V T
( ty l l Y Y Xr V r T l
m R }I A 2 2 m R ( X I 3e m R ( X I 3e m R ( X I w v l A ( l k X
W ( X
172 and
(C.39)
Substituting
A thin equilateral triangle plate (Fig.C.3) with the edge length , is immersed parallel to a stream of air with the velocity at temperature and pressure . Assuming laminar ow, estimate the drag on this plate. Solution Using (3.125), we have:
The area element is , where , and . For two sides of the triangle, we have for the total drag force:
R X I ue d
l
FX l d
R D9 }I ( R X ( d I ( T l ( ( }p V l s l W o R }I g g Y T r Y
s 2 d 3e Rl s 3e k 2
s 1td Rs
id ( VY d
2 ( Rs 3e 2 3e 2 3e
4 d
R (I
{ld s lX m
l }p V I l s }p V r R Y r R Y I }p m ts r l }p V }p m r R Y I R l I r }p }p }p r( X X l l }p ( ~ r d ( }p r ( ~ ( }p ( R X ( r r r
( }p ( ( d }p V s r R X I ~ r R Y I R l }I l T l
Figure C.3: Triangular plate pulled in a viscous uid
s R l I
dI ~
173
p r t b Y r A r r W b r A p r p p p b pY A r r W b A
p p
W r A r W p Ar
A r b r A A p p r b A
p pA b r & A pA p b & A
Since scaling transformations (C.44), (C.45) are anisotropic (different scales are used for different components), we have to rewrite the momentum equations (C.44) for each component separately:
q f A
( 4A
A p ) p A ( A ( p p ) r ( r }p2 e r ( ) ( (
A b W b s Y R s & s I W f A b & A A
x A p p A b s b s & s rr s r r p p A p p p A p rr A r A r b A b & A
z p b R r A I r r px b p q f A
r
, we have:
2 3e x }p2 e 2 3e 2 3e p p 3e 3e 2 2 p p 2e b `r r d r r A r r d b r A r d r A r r A d b r A A d b r & A d
x b `r d r }p2 }p2 e r r A r r A r d b }p2 e r }p2 e p p 2 e e r A r r r d b r A r d }p2 e p p }p2 e r A A r d b r &A r d
:
In the limit
and dividing by
Similarly, introduce the dimensionless variables into the momentum equation for , (C.47):
(C.49)
(C.48)
(C.50)
rA
p A 3e b p A 3b p A p A ( s b 2 2e r R qr I d r R r I qr r l A W 2 e pp A ( A p p A b R rr s 4b s I Y r s r b s & s
and using the denition of the Reynolds number
A sX ( p A 4e b p A 3b p A p A ( b 2 2e r R qr I d r R r I qr r l r A W 2 e pp ( A p p rA b ( R rr s 3b s I r s Y R r s r b s & s I s A
q u A
p A b pA r R qr I pp r R r I s R rr b s I Y
bp p qr A r A l b r R f A I W l b A p pA b R r s r b s & s I
X W
A A b R m b f A I m A W b s Y R s & s I p(
.
where is zero by continuity (2.4). Transferring, to dimensionless variables (C.44) and (C.45), we obtain: Writing it out explicitly, we have: This means that the pressure only changes in the direction of Now consider the energy equation (C.44): , and rearranging the RHS, we obtain:
using the denition of Prandtl (1.30) and Eckert numbers (3.114), we obtain:
177
178 Chapter A
Problem A.4.1: Check tensor expressions Check if the following Cartesian tensor expressions violate tensor rules:
Problem A.4.2: Construct tensor expression Construct a valid Cartesian tensor expression, consisting of three terms, each including some of the four tensors: . Term 1 should include tensors only, term 2 tensor and term 3 tensors . The expression should have 2 free indexes, which should always come rst among the indexes of a tensor. The free indexes should be at A and B in the rst term, at B and C in the second term and C and D in the last term. How many different tensor expressions can be constructed? Solution One possibility is:
Since there are four locations for dummy indexes in each term, there could be three different combinations of dummies in each term. Thus, the total number of different expression is
9 9
9 !B9 U 9 D9 G `j Gt 9
2 2 b b j j
j j G wG j j G j j j 2 b 2 b j j 2G
l
B9 9
179 Problem A.4.3: Cartesian identity Prove identity (A.15). Proof Integrating (A.5) in the case of constant transformation marix coefcients, we have:
(C.56)
(C.57)
(C.58)
( j j ( ( ( ( e b ( e b (j e b ( j ( j j
, we have:
into :
( ( G j ( j ( j j
e b ( j ( j
( ( ( ( ( j ( j
. Then we get:
D 2k
is the same as the latter can be
j j j j R j j I j R Sp j GS Ip
R ~ ~ I ~ R ~ ~ I ~ R ~ ~ I ~ j j Sp j GSp
(C.60)
(C.62)
(C.61)
Proof
180
(C.63) Using (A.10), and since rewritten as: (C.59) Using the tensor identity:
Now rename the dummy indexes: sion looks like one in (A.29): From (A.24) it follows that
. Then we have:
Problem A.4.5: Metric tensor in cylindrical coordinates. (C.64) is given by the following Cylindrical coordinate system transformation rules to a Cartesian coordinate system, :
Obtain the components of the metric tensor (A.40) (A.38) in cylindrical coordinates. First compute the derivatives of
with respect to
(C.64)
E%B9 7 9 (
d C ( C C ( d p eG y r (( p r ( fe d q( ( pr ( p ( x fe d ` p ( d x eG ( ( r (( p
E%DC9B9@( 7 (
Solution:
G
E 9 9 %DCB@(
Ut
7
R ~ ~ I ~ R ~ ~ I ~ E)D9 7 ( 9 C d efd eG (
:
182
Problem A.4.6: Metric tensor in curvilinear coordinates Using Mathematica, write a procedure to compute metric tensor in curvilinear coordinate system, and use it to obtain the components of metric tensor, , (A.40) and its conjugate, , (A.38) in spherical coordinate system ( ):
(C.65)
Solution with Mathematica NX = 3 (* Curvilinear Y = Array[,NX] Y[[1]] = r; Y[[2]] = th; Y[[3]] = phi;
cooridnate system *) (* Spherical coordinate system *) (* radius *) (* angle theta *) (* angle phi *)
(* Cartesian coordinate system *) X = Array[,NX] X[[1]] = r Sin[th] Cos[phi]; X[[2]] = r Sin[th] Sin[phi]; X[[3]] = r Cos[th]; (* Compute the Jacobian: dXi/dYj *)
D29 9
pp x x
183 J = Array[,{NX,NX}] Do[ J [[i,j]] = D[X[[i]],Y[[j]]], {j,1,NX},{i,1,NX} ] (* Covariant Metric tensor *) g = Array[,{NX,NX}] (* covariant *) Do[ g [[i,j]] = Sum[J[[k,i]] J[[k,j]],{k,NX}], {j,1,NX},{i,1,NX} ]; g=Simplify[g] (* Contravariant metric tensor *) g1 =Array[,{NX,NX}] g1=Inverse[g]
Problem A.4.7: Christoffels symbols with Mathematica Using the Mathematica package, write the routines to compute Christoffels symbols Solution
(************* File g.m ************* The metric tensor and Christoffel symbols *************************************)
9 9 77 EE r ar 9 9 7 E 9 9 7 E 9 d P d 9 EE r R I d 9 9 E r 9 9 E 9 d 7 7 9 9 9 r R I fe r 7 r 7
184 DIM = 3 (*
The metric tensor *) g = Array[,{DIM,DIM}] (* covariant *) g1 =Array[,{DIM,DIM}] (* contravariant *) Do[ g [[i,j]] = 0; g1[[i,j]] = 0 , {j,1,DIM},{i,1,DIM} ] (* Cylindrical coordinates *) Z=Array[,DIM] Z[[1]] = r Z[[2]] = th Z[[3]] = z g [[1,1]] = 1 g [[2,2]] = r2 g [[3,3]] = 1 g1[[1,1]] = 1 g1[[2,2]] = 1/r2 g1[[3,3]] = 1 (* Christoffel symbols of the first and second type *) Cr1 = Array[,{DIM,DIM,DIM}] Cr2 = Array[,{DIM,DIM,DIM}] Do[ Cr1[[i,j,k]] = 1/2 ( D[ g [[i,k]], Z[[j]] ] + D[ g [[j,k]], Z[[i]] ] - D[ g [[i,j]], Z[[k]] ] ), {k,DIM},{j,DIM},{i,DIM} ] Do[ Cr2[[l,i,j]] =
Problem A.4.8: Covariant differentiation with Mathematica Using the Mathematica package, write the routines for covariant differentiation of tensors up to second order. solution
(************** File D.m ******************* Rules of covariant differentiation ********************************************) (* B.Spain Tensor Calculus, 1965 Eq.(22.2) *) D1[N_,A_,k_,X_,j_]:= (* Computes covariant derivative of a mixed tensor of second order with index k - covariant (upper) *) Module[ {i,s}, s = Sum[Cr2[[k,i,j]] A[[i]],{i,N}]; D[A[[k]],X[[j]]] + s ] Dl1[N_,A_,l_,X_,t_]:= (* Computes covariant derivative of a mixed tensor of second order
186
*) Module[ {s,r}, s =Sum[Cr2[[r,l,t]] A[[r]],{r,N}]; D[A[[l]],X[[t]]] - s ] D1l1[N_,A_,m_,l_,X_,t_]:= (* Computes covariant derivative of a mixed tensor of second order with index m - contravariant (upper) and index l - covariant (lower) *) Module[ {s1,s2,r}, s1 =Sum[Cr2[[m,r,t]] A[[r,l]],{r,N}]; s2 =Sum[Cr2[[r,l,t]] A[[m,r]],{r,N}]; D[A[[m,l]],X[[t]]] + s1 - s2 ] D2[N_,A_,i_,j_,X_,n_]:= (* Computes covariant derivative of second order tensor with both m and l contravariant (upper) indexes B.Spain Tensor Calculus, 1965 Eq.(23.3) *) Module[ {s1,s2,k}, s1 =Sum[Cr2[[i,k,n]] A[[k,j]],{k,N}]; s2 =Sum[Cr2[[j,k,n]] A[[i,k]],{k,N}]; D[A[[i,j]],X[[n]]] + s1 + s2 ] D2l1[N_,A_,i_,j_,k_,X_,n_]:= (* Computes covariant derivative of third order tensor with i and j contravariant (upper)
187 and k contravariant (lower) indexes B.Spain Tensor Calculus, 1965 Eq.(23.3) *) Module[ {s1,s2,s3,m}, s1 =Sum[Cr2[[i,m,n]] A[[m,j,k]],{m,N}]; s2 =Sum[Cr2[[j,m,n]] A[[i,m,k]],{m,N}]; s3 =Sum[Cr2[[m,k,n]] A[[i,j,m]],{m,N}]; D[A[[i,j,k]],X[[n]]] + s1 + s2 - s3 ] D4l1[N_,A_,i1_,i2_,i3_,i4_,i5,X_,i6_]:= (* Computes covariant derivative of 5 order tensor with 4 first indexes contravariant (upper) and the last one contravariant (lower) B.Spain Tensor Calculus, 1965 Eq.(23.3) *) Module[ {k,s1,s2,s3,s4,s5}, s1= Sum[Cr2[[i1,k,n]] A[[k,i2,i3,i4,i5]],{k,N}]; s2= Sum[Cr2[[i2,k,n]] A[[i1,k,i3,i4,i5]],{k,N}]; s3= Sum[Cr2[[i3,k,n]] A[[i1,i2,k,i4,i5]],{k,N}]; s4= Sum[Cr2[[i4,k,n]] A[[i1,i2,i3,k,i5]],{k,N}]; s5=-Sum[Cr2[[k,i5,n]] A[[i1,i2,i3,i4,k]],{k,N}]; D[A[[i1,i2,i3,i4,i5]],X[[i6]]]+s1+s2+s3+s4+s5 ]
Problem A.4.9: Divergence of a vector in curvilinear coordinates Using the Mathematica package and the solution of Problem A.4.8, write the routines for computing divergence of a vector in curvilinear coordinates. Solution
188
Using the algorithms of covariant differentiation developed in Problem A.4.8 we have: <<"./g.m" (* The g-tensor and Christoffel symbols *) <<"./D.m" (* Rules of covariant differentiation *) (* The original coordinates: *) NX = DIM X = Array[,NX] (* Variables: *) NV = DIM U = Array[,NV] (* New coordinate system *) Y = Array[,NX] Y[[1]] = r; Y[[2]] = th; Y[[3]] = z; X[[1]] = r Cos[th]; X[[2]] = r Sin[th]; X[[3]] = z; (* Compute the Jacobian *) J = Array[,{DIM,DIM}] Do[ J [[i,j]] = D[X[[i]],Y[[j]]], {j,1,DIM},{i,1,DIM} ] J1=Simplify[Inverse[J]] (* Derivatives of a vector *) V0 = Array[,NX] V0[[1]] = Vr[r,th,z]; V0[[2]] = Vt[r,th,z]; V0[[3]] = Vz[r,th,z]; (* Rescaling for physical (dimensionally correct) coordinates
189 (\cite[5.102-5.110]{SyScTC69}) *) V = Array[,NX] Do[ V[[i]] = PowerExpand[V0[[i]]/g[[i,i]](1/2)], {i,1,NX} ] (* Transform vectors as first order contravariant tensors *) U = Array[,NX] SetAttributes[RV1,HoldAll] RV1[NX,V,U] (* Compute first covariant derivatives of vectors *) DV = Array[,{NX,NX}]; Do[ DV[[i,j]] = D1[NX,V,i,Y,j], {j,1,NX},{i,1,NX} ] (* Divergence *) div=0 Do[ div=div+DV[[i,i]], {i,NX} ] div0 = div/.th->0
Problem A.4.10: Laplacian in curvilinear coordinates Using the Mathematica package, write the routines for computing Laplacian in curvilinear coordinates. solution Using the algorithms of covariant differentiation developed in Problem A.4.8
190 we have:
<<"./g.m" (* The g-tensor and Christoffel symbols *) <<"./D.m" (* Rules of covariant differentiation *) (* The original coordinates: *) NX = DIM X = Array[,NX] (* Variables: *) NV = DIM U = Array[,NV] (* New coordinate system *) Y = Array[,NX] Y[[1]] = r; Y[[2]] = th; Y[[3]] = z; X[[1]] = r Cos[th]; X[[2]] = r Sin[th]; X[[3]] = z; (* Compute the Jacobian *) J = Array[,{DIM,DIM}] Do[ J [[i,j]] = D[X[[i]],Y[[j]]], {j,1,DIM},{i,1,DIM} ] J1=Simplify[Inverse[J]] (* Derivative of a scalar *) DP = Array[,NX]; Do[ DP[[i]] = D[p[r,th,z],Y[[i]]], {i,1,NX} ] DDP = Array[,{NX,NX}]; Do[ DDP[[i,j]] = Dl1[NX,DP,i,Y,j], {i,1,NX},{j,1,NX}
191 ] DDQ = Array[,{NX,NX}]; Do[ DDQ[[i,j]] = Sum[DDP[[k,l]] J1[[k,i]] J1[[l,j]],{k,NX},{l,NX}], {i,1,NX},{j,1,NX} ] (* Laplacian *) (*** lap=lap+Sum[g[[i,j]]*Dl1[NX,DS,j,Y,i],{i,1,NX},{j,1,NX}],*) lap=Sum[DDQ[[i,i]],{i,NX}] lap0=lap/.th->0
Problem A.4.11: Invariant expressions not: Check if any of these tensor expressions are invariant, and correct them if
(C.66)
(C.67)
Answers:
Since there are two combinations for an invariant combination of dummy indexes (Corollary A.3.5), there can be several different invariant expressions.
j j b j w
" j " j
(C.68)
2 j U
192
which proves that both forms have the same values. If we now consider the rst form then:
which can not be reduced further and, therefore is not invariant, since it has a different form from the LHS.
j j j j j G j 2 U 6
( j j j j (
j j ( ( ( ( 6
( ( 6
w
is not.
194
11. Obtain equation of motion for axially moving concentric cylinders: [(3.14)] on the basis of general NS equation in cylindrical coordinates. 12. Obtain the solution [(3.16)] to equation (3.14). 13. Show that the problem of pulling an innite rod does not have a steady-state solution [Remark 3.2.1]. 14. Obtain equations of motion for axially moving concentric cylinders: [(3.17)], [(3.18)] on the basis of general NS equation in cylindrical coordinates (Problem 3.7.5). 15. Obtain the solution for the ow between axially moving concentric cylinders, [(3.20)], on the basis of equation (3.18) and the appropriate boundary conditions. 16. Obtain solutions for the ow inside and outside of a rotating cylinder: Remarks [3.2.2] and [3.2.3]. 17. Obtain the pressure distribution in a ow outside a rotating cylinder: [(3.23)] on the basis of the solution (3.22) and equation and the appropriate momentum equation [(3.17)]. 18. Solve problem 3.7.6. 19. Derive the equation for axial velocity in a fully developed ow region [(3.25)] using the appropriate assumptions and equation (3.2). Write a non-dimensional formulation of the boundary value problem [(3.26)]. 20. Formulate a boundary value problem for a Poiseuille ow in a circular duct [(3.27)] on the basis of a NS equation in cylindrical coordinates, and obtain the Poiseuille solution [(3.28)]. 21. For a Poiseuille ow through a duct (3.28) compute the volumetric ow rate [(3.28)], the wall shear stress [(3.30)], the skin friction coefcient [(3.31)] and Poiseuille number [(3.32)]. 22. For combined Couette-Poiseuille ows formulate the equation of motion [(3.33)], obtain the solution [(3.36)], and formulate the criterion of separation [(3.38)]. 23. Solve problem 3.7.7. 24. Give a denition of a hydraulic diameter [(3.39)].
195 25. Formulate the equation for the uid oscillating above an innite plate [(3.44)], and obtain the solution for the case of oscillating plate [(3.47)] and an oscillating uid [(3.48)]. Hint: Use the identity: , where . 26. Obtain the solution for an unsteady ow between two innite plates [(3.65)]. . Hint: Use integral relation: 27. Formulate the assumptions of creeping ow, and show how to obtain the Laplace equations for pressure [(3.67)] and vorticity [(3.68)]. 28. Using the expressions for pressure (3.76) and shear stress (3.77) of a Stokes ow around a sphere, compute the total drag force [(3.80)], and the drag coefcient as a function of the Reynolds number [(3.81)]. Hint: Use the integral relation: . 29. Using the assumptions of the lubrication theory, derive the equation for pressure distribution in a ow between moving plates with a non-uniform gap [(3.91)]. 30. Derive criteria of validity of equation (3.91): [
], [(3.92)].
d m 5
d d( R ( I yG4( R ( I ue d ( R ( I fe 4x m l nX R d b I5 r }p 5
d ( d d l d X SRR ( I eG{ R ` I er I ( R ( I fe xx
196
198
8. Formulate the energy equation in terms of temperature [(2.74)] and enthalpy [(2.75)]. Explain the meaning of each term. 9. Problem CFM.2.7.5: Using the energy equation (2.77):
(E.1)
(E.2)
derive the strong formulation of the Bernoullis equation. 10. Derive the expression for Coriolis forces [(2.93)]. 11. Problem CFM.A.4.4: Using tensor identity:
j j b b j R A Y I b R A Y I 4 T jA j j 2 Sp j GSp
w6 R ~ ~ I ~ R ~ ~ I ~ R ~ ~ I ~
A G T m b R s I b 4 4 Y
, and both are invariant, while
199
15. Problem CFM.3.7.3: Consider a wide uid lm of constant thickness, , owing steadily due to the gravity down the inclined plate at angle . Find the velocity distribution, , and the volumetric ow rate, . Atmospheric pressure can be considered constant. 16. Obtain equation of motion for axially moving concentric cylinders: [(3.14)] on the basis of general NS equation in cylindrical coordinates. 17. Obtain the solution to equation (3.14) for the case of axially moving concentric cylinders [(3.16)]. Show that the problem of pulling an innite rod does not have a steady-state solution [Remark 3.2.1]. 18. Problem CFM.3.7.5: Using the assumptions on the Couette velocity prole between the rotating concentric cylinders (Sec.3.2.3) and the expression for the momentum equation and the Laplacian operator in cylindrical coordinates:
Derive equation for , [(3.18)], and obtain the solution for the ow between rotating concentric cylinders, [(3.20)], on the basis of the appropriate boundary conditions. 19. Problem CFM.3.7.6: In the system of two rotating cylinders (Sec.3.2.3) consider the torque applied to the inner rotating cylinder when the outer cylinder is xed ( ). What is the power required to rotate the inner cylinder? 20. Derive the equation for axial velocity in a fully developed ow region [(3.25)] using the appropriate assumptions and the momentum equation (3.2):
Write a non-dimensional formulation of the boundary value problem [(3.26)]. 21. Formulate a boundary value problem for a Poiseuille ow in a circular duct [(3.27)] on the basis of a NS equation in cylindrical coordinates, and obtain the Poiseuille solution [(3.28)]. Compute the volumetric ow rate [(3.28)], the wall shear stress [(3.30)], the skin friction coefcient [(3.31)] and Poiseuille number [(3.32)].
Aj A p Y j j T p Y j f b & A
R I A
p n
200
22. For combined Couette-Poiseuille ows formulate the equation of motion [(3.33)], obtain the solution [(3.36)], and formulate the criterion of separation [(3.38)]. Using the obtained solution, consider a viscous uid with viscosity ( ) driven between two parallel plates apart by an imposed pressure gradient of (Problem 3.7.7). The upper plate is moving with velocity . Find the volume ow rate per 1m of the plates width. What pressure gradient will cause the ow to reverse? 23. Formulate the equation for the uid oscillating above an innite plate [(3.44)], and obtain the solution for the case of oscillating plate [(3.47)] and an oscil, where . lating uid [(3.48)]. Hint: Use the identity: 24. Obtain the solution for an unsteady ow between two innite plates [(3.65)]. Hint: Use integral relation: . 25. Formulate the assumptions of creeping ow, and show how to obtain the Laplace equations for pressure [(3.67)] and vorticity [(3.68)]. 26. Using the expressions for pressure (3.76) and shear stress (3.77) of a Stokes ow around a sphere, compute the total drag force [(3.80)], and the drag coefcient as a function of the Reynolds number [(3.81)]. Hint: Use the integral relation: . 27. Using the assumptions of the lubrication theory, derive the equation for pressure distribution in a ow between moving plates with a non-uniform gap [(3.91)], and derive criteria of its validity: [ ], [(3.92)].
[(3.104)], [(3.105)].
d m 5
d d( R ( I eG4( R ( I fe d ( R ( I fe 3x l m X R d b I5 r }p 5
FX d X s l ( X
bg p b A r r R I
d ( d d l d X qRR ( I er R I er I ( R ( I fe x
F s R I X d V
201
derive Prandtl boundary layer equations [(3.110)] - [(3.113)]. Hint: Use the Froude [(2.111)] and Eckert [(3.114)] numbers as dimensionless parameters. 31. Formulate the boundary value problem for the boundary layer [(3.115), (3.116)], and reduce it to the Blasius problem [(3.122), (3.123)]. 32. Formulate the Blasius problem [(3.122), (3.123)], and obtain relations for momentum thickness [(3.124)], wall shear stress [(3.125)], skin friction coefcient [(3.126)], and the total drag on the plate: [(3.127)]. 33. Formulate the equations for the free shear ows [(3.129)], and reduce them to the case of a wake [(3.133)]. Assume solution of a heat-conduction type:
Using the appropriate boundary conditions and the integral relation between momentum and drag determine the constants and as functions of , , , , and the characteristic size of the object, : [(3.135), (3.136)]. 34. Performing integral analysis of the boundary layer to obtain the von Karman integral momentum relation [(3.140)].
( 4A
A R m b f A I m A b Y u A q f A
}p ( 9 ( |jh r R q@( I k r
A b W b s Y R s & s I W f A b & A A e r A p ( A p
) p A A ( ( p ) r ( r }p2 e r ( ) ( ( Y
202
35. Problem CFM.3.7.12: A thin equilateral triangle plate with the edge length of is immersed parallel to a 12m/s stream of air at and 1atm, as in Fig.C.3. Assuming laminar ow estimate drag of this plate (in N). First give an answer in symbolic form in terms of , , compute it to a number. , and . And then
{l
Y V
l
Index
Acoustic problems, 13 Algebraic Reynolds stress model, 109 Associate tensor, 130 Bernoullis Equation, 24 Bernoullis equation, 38, 51, 152, 154, 198 Blasius equation, 90 Blasius stream function, 92, 96 Boundary conditions, 29 Boundary layer, 85, 87 Boundary layer blow-off, 92 Boussinesq approximation, 108 Brinkman number, 56 Bulk modulus, 13 Cartesian Tensors, 119 Christoffels symbol, 132 Coefcient of bulk viscosity, 20 Coefcient of thermal expansion, 13 Coefcient of viscosity, 9, 10, 93 Conjugate metric tensor, 129 Conjugate tensor, 40, 130, 134 Continuity equation, 16, 32 Contraction of indexes, 123 Contraction operation, 130 Contravariant index, 117 Contravariant tensor, 133 Contravariant vectors, 116 Convective derivative, 4 Coordinate system, 116 Coriolis force, 42, 43 Couette ows, 54 Covariant differentiation, 132 203 Covariant index, 117 Covariant vectors, 117 Creeping ow, 71, 76 Cross product, 124 Cylindrical coordinates, 138 Darcy friction factor, 62 Dependent variable, 9 Dependent variables, 1 Direct numerical simulation, 106 Dirichlet boundary, 30, 61 Displacement thickness, 82, 90, 100 Divergence operator, 127, 133, 138, 139 Divergence-free vector eld, 16 Dot-notation, 2 Drag coefcient, 83, 98 Drag force, 74 Dummy index restriction, 123 Dummy indexes, 122 Eckert number, 52, 87, 157, 177 Eddy viscosity, 106, 108 Enthalpy, 7 Enthapy, 8 Entrance effect, 60 Entropy, 6 Equation of state, 5, 7, 32 Euler equation, 28, 29 Eulerian description, 1 Falkner-Skan equation, 92 Falkner-Skan wedge ow, 91 First law of thermodynamics, 6 Flow separation, 64
204 Fluid element, 3 Fluid particle, 1 Fluid particles, 1 Fluid properties, 8 Fluid velocity, 2 Flux, 9 Fourier series, 70 Fouriers law, 10, 35 Free boundary, 31 Free indexes, 121 free jet, 95 Friction coefcient, 55, 90, 93 Froude number, 49, 50, 87, 175 Fully developed ow, 60, 61, 65 Fundamental tensor, 129 Gas constant, 6 Gauss theorem, 16, 19, 35, 36 Gibbs rule, 6 Gradient, 9, 127 Gradient approximation, 911 Heat conduction coefcient, 10, 39, 49, 93 Heat conduction equation, 38 Heat conductivity, 10 Heat convection equation, 39 Heat dominated ow, 39 Hydraulic diameter, 64 Hydrostatic pressure, 53 Ideal uid, 28, 152 Ideal gas law, 6 Incompressibility condition, 16 Independent variables, 1, 6, 8 Inlet boundary, 30, 34 Integral momentum relation, 101 Internal energy, 6, 8 Invariance, 39, 116, 128 Invariant, 40, 130, 132 invariant expressions, 39 Invariant forms, 40, 128, 131 Inviscid ow, 28 Irrotational ow, 23, 27, 29 KE turbulence model, 111 Kelvins theorem, 152 Kinematic variables, 1 Kinematic viscosity, 10, 43 Kronecker delta tensor, 118
INDEX
Lagrangian description, 1 Laminar ow, 56 laminar ow, 53 Laplace equation, 24, 27, 28, 71 Laplace equation for pressure, 33 Laplacian, 61, 66, 72, 73, 102, 128, 134, 138, 139, 162, 199 Large eddy simulation, 106 Law of circulation, 152 Law of similarity, 43, 46 Lewis number, 12 Lift force, 74 Lowering indices, 130 Lubrication theory, 76 Mean wall shear stress, 64 Metric tensor, 40, 128, 131 Momentum equation, 131 Momentum ux, 29 Momentum thickness, 83, 90, 100 Momentum transport, 9 Moving boundaries, 30 Nabla, 17, 28, 47, 48, 86, 126, 138 Natural convection, 13 Neuman boundary, 30, 33 No-slip boundary, 30, 61 Non-dimensional parameters, 46, 47, 49 Non-dimensional variables, 48, 49 Non-inertial coordinate systems, 41 Non-Newtonian, 10 Nusselt number, 49, 93
INDEX
One equation turbulence model, 111 Order of a tensor, 117, 121 Orthogonal coordinate system, 137 Orthogonal coordinates, 40, 134 Oseen approximation, 76 Outlet boundary, 30, 34 Parabolic equation, 66 Particle trajectory, 2 Permutation tensor, 123 Physical component, 137 Physical components of tensors, 39 PI-group, 46 PI-theorem, 46, 47, 157 Poiseuille ow, 62 Poiseuille number, 56, 62 Poisson equation for pressure, 33 Potential ow, 23 Prandtl number, 11, 38, 92, 177 Primary dimensions, 44, 46, 48 Properties of the uid, 1 Quasi-equilibrium approximation, 5 Raising indices, 130 Rank of a tensor derivative, 127 Rank of a term, 121 RANS models, 108 Renaming indexes, 123 Renaming of dummy indexes, 123 Reynolds analogy, 92, 93 Reynolds averaged Navier-Stokes equation, 108 Reynolds averaging, 108 Reynolds decomposition, 107 Reynolds number, 44, 49, 54, 55, 60, 61, 75, 85, 94, 105, 175 Reynolds stress model, 108, 111 Reynolds stress tensor, 108 Scalar product, 40, 122, 123, 130, 131 Schmidt number, 12
205 Second law of thermodynamics, 34 Shear layer, 94 Similarity solution, 88 Skew-symmetric tensor, 124 Skin-friction coefcient, 62, 83 Slip boundary, 30 Smagorinsky constant, 106 Smagorinsky model, 106 Solenoidal vector eld, 16 Solid body rotation, 58 Spatial derivative of a tensor, 126 Specic heat, 8, 11, 38, 93 Speed of sound, 12 Stanton number, 93 Stokes ow, 71, 80 Stokes paradox, 76 Stokes theorem, 152 Strain rate tensor, 4, 20 Stream function, 17, 72 Streamline, 2 Stress tensor, 19 Stretching factors, 135, 138 Substantial derivative, 3, 40 Surface area vector, 16 Surface tension coefcient, 31, 50 T-s relations, 7 Taylor number, 105 Tensor, 117 Tensor derivative, 127 Tensor equality, 121 Tensor expression, 120 Tensor identity, 125 Tensor notation, 116, 120 Tensor rules, 120 Tensor terms, 120 Thermodynamic properties, 1 Thermodynamic variables, 1 Time derivative of a tensor, 126 Transformation matrix, 117, 119 Transformation rule, 116
206 Transport properties, 1 Transport property, 9 Turbulence dissipation rate, 112 Turbulence model, 111 Turbulent closure, 110 Turbulent kinetic energy, 108 Two equation turbulence models, 111 Two-equation turbulence models, 112 Vector product, 124 Velocity circulation, 50, 151 Velocity potential function, 23, 29 Viscous ow, 27 Viscous stress tensor, 20, 33 Volumetric ow rate, 62 von Karman, 101 Vorticity tensor, 4 Vorticity vector, 5 Wake, 96 Wall shear stress, 62, 90 Weber number, 50 Wedge ows, 91
INDEX