You are on page 1of 31

Chapter 3

Single phase flow hydrodynamics

In this chapter, we discuss single phase flow characteristics which are essential for the
description of multiphase flow. We will also have a brief look at:
• Dimensional analysis, which is a methodical way for construction of correlations in
fluid dynamics and physics in general.
• Similarity analysis, which is the theoretical basis for scaling of experiments from
“small scale” to “full scale”.

We start with the two latter items due to their importance also for single phase theory
presented in this chapter.

3.1. Dimensional analysis and similarity


Dimensional analysis is simply an application of the principle of “dimensional
homogeneity”, which means that the right-hand side (RHS) and left-hand side (LHS) of
any physical equation should be identical concerning units.

Example 1: Hydrostatic pressure


Derive the equation for hydrostatic pressure (P) in a fluid, based on fluid density (ρ),
gravitational acceleration (g) and depth in the fluid (h).
Solution: Assume there is a polynomial dependence of P on the other quantities. Thus
P=ρ a ⋅g b ⋅ hc . The trivial solution for the exponents a, b and c to ensure dimensional
homogeneity is a = b = c =1. Check this for yourself, based on the units.
Thus, the well-known law P = ρ g h may be derived on pure dimensional arguments.
This case is very simple and transparent, but even when many quantities enter a problem
it is still possible to apply dimensional homogeneity. This is the message of Buckingham’s
Π-theorem, which will be discussed in the following section.
3.1.1. The Buckingham π - Theorem
The Π-theorem can be expressed in the following apparently insignificant way:

“Assume a process or phenomenon is described by n quantities Q1, …, Qn, which may be


pressure, density, velocity, etc. These quantities are in turn assumed to contain m different
units U1, …, Um (in most cases only combinations of meter, seconds, kilograms in the
MKSA system). The process can instead be described by n-m dimensionless groups Π1,
…, Πn-m.”

However, this is in many cases a powerful tool to establish correlations in fluid mechanics.

Example 2. Prove the Buckingham Π-theorem.


Solution: Here we only give hints to solution. The most common way is by using a
derivation of the theorem which states that it is possible to express one of the
dimensionless quantities by the remaining n-m-1 quantities.
A particular form is the polynomial: Π1= Π2α,….., Πn-mω
A more general form is: Π1= f [Π2α,….., Πn,m]
The proof is straightforward when the polynomial is written in terms of units.
A particular proof is given by Rayleigh’s method. This method assumes a polynomial
combination to express one of the quantities (e.g. Q1) in terms of the remaining n-1
quantities as suggested in Example 1.

Dimensional homogeneity imposes a relation between the n-1 exponents that appear on
the RHS of the equation. This leads to a system of m linear equations in terms of the
exponents, one for each unit. However, the system is underdetermined if n > m. Thus,
only m exponents can be found, the other n-m-1 must be assigned some arbitrary
constant. From these the other m exponents may be expressed. The rest of the proof is
left to the reader.

Note that Buckingham’s Π-theorem can also be proved by making a product of all the
quantities as Q1a…Qnw and require the total product (Π) to be dimensionless, hence the
Π in the name of the theorem. This proof is nearly identical to the one given by Rayleigh’s
method.

Example 3. Friction factor in pipe flow: Flow friction is related to the shear stress, τw, at
the wall. Shear stress in turn is found to depend on average flow velocity (U), fluid viscosity

(μ), density (ρ) and pipe diameter (D). Determine the relation between τw and the other
quantities.
Solution: We may apply the technique based on the total π product in Buckingham’s

theorem to determine the relation between τw and the other quantities. There is n=5
quantities in this problem, with only m=3 fundamental units. We then expect n-m=2
dimensionless groups π1 and π2 from the analysis, thus there will be a relation:

Π1= f [Π2]

We form a total product of all the quantities and write: τwa Ub ρc μd De=Constant
where a, … , e are unknowns to be determines. Quantity kg m s exponent
We substitute units into the product and get 3 τ 1 -1 -2 a
equations relating the exponents, one for each U 0 1 -1 b
unit. The units for the quantities given in the ρ 1 -3 0 c
following table. μ 1 -1 -1 d
We then obtain the following underdetermined D 0 1 0 e
set of equations:
kg: a +c+d=0
m: -a + b – 3c – d + e = 0
s: -2a – b –d=0
Only 3 equations allow only three unknowns to be determined, while two must remain
undetermined. We may choose these two to be e.g. a = A, b = B and obtain for the other:
c= A+B
d=-2A-B
e= 2A+B
Dimensionless groups
Substituting the values for a,…,e into the original product, and collecting terms with
corresponding exponents, we get:

We see immediately that Π2 is the Reynolds number. However, Π1 is not easily


recognizable, although the solution is correct.

Note. Always verify the solution by checking that the groups are dimensionless.

We have found previously in Chapter 2 that where f is a function of the


Reynolds number. We can obtain the form by combining Π1 and Π2 into a new group:

Check this. This illustrates a fundamental and important property of solutions obtained:

“From one set of dimensionless groups Π1,….., ΠN, we may always form a new set
Π′1,...,Π′N by recombination of groups from the first set by the fundamental operations of
calculus (+, - , * , / ).”

3.1.2. Similarity
Similarity is a concept which in physics has a meaning roughly equivalent to its
mathematical origin. Two different physical systems A and B are similar provided they
have:
1. Geometrical similarity: Length ratios and angles in A are the same as in B.
2. Kinematical similarity: Velocities ratios in A are the same as in B
3. Dynamical similarity: Force ratios in A are the same as in B.
3.1.3. Similarity in flow based on Navier-Stokes equation
A physical system may be described by its equations of motion. Examples are Newtons
2nd law in mechanics and Navier-Stokes equations in fluid mechanics. We will study how
the ratio between forces in a fluid system appears naturally from the equations of motion
when we make them dimensionless. The ratios between forces appear in dimensionless
groups of the type we found in dimensional analysis, e.g. the Reynolds number. Consider
the Navier-Stokes equation:

The equation may be written more compact by introducing the material derivative, e.g. for
⃗:
𝑢

Thus, we may write

This equation is made non-dimensional by introducing dimensionless quantitites for


length, time, velocity and pressure as shown in the table below

Quantity Characteristic quantity Dimensionless quantity


Length L ⃗⃗⃗
𝑥′ = 𝑥 /𝐿
Velocity U ⃗⃗⃗
𝑢′ = 𝑢⃗ /𝑈
Time L/U 𝑡′ = 𝑡𝑈 /𝐿
Pressure ρu2 𝑝′ = 𝑝/𝜌𝑈 2

In most cases of pipe flow, the pipe diameter is used as characteristic length L, while U is
the mean flow velocity. The characteristic pressure is taken as the dynamic pressure
. Since the static pressure is of no relevance, at least in incompressible flow. By
introduction of the appropriate dimensionless quantities we obtain

or by reorganizing

Thus, we have obtained a dimensionless equation where the viscous term is divided by
the Reynolds number. The flow conditions are thereby introduced in this prefactor. The
higher the Reynolds number, the less important is the viscous term. Correspondingly, we
find the Froude number (or rather the inverse)

as a prefactor to the hydrostatic term. Then a high Fr makes the gravitational forces less
important. From this analysis, we see that the physics of a problem depends only on the
Re and Fr, since the equations are otherwise identical.

Exercise: Scale P using instead of ρU 2. Show that in this case the Reynolds number will
appear associated with the convective term. In addition, show that the gravitational term
now gets associated with a Stokes number instead of the Froude number.

Example 4. Slug flow in a full scale two phase flow pipe S is investigated by a small-scale
model pipe M. In both cases oil is used as liquid. The pipe diameter in S is D and in M it
is d. Determine whether it is possible to obtain equal value of Reynolds number and
Froude number in the two systems.
Solution: The Froude number and Reynolds number are seeked equal in the two systems.
The oil has density ρ and viscosity μ. Thus;

which requires that U D= u d or U=u d/D

which requires

Substituting U from 1 into 2, we obtain which implies d = D. Thus, with only one
fluid we cannot obtain dynamical similarity in the two systems.

Example 5. We reconsider the previous example and want to investigate slug flow in S by
the small-scale model pipe M. The full-scale system contains oil, while water is used for
S. In this case we require also the Weber number to be identical in the two systems. Is it
possible to have equal value of all three numbers? And, can water be used as test fluid?

Solution: The Weber number is defined by

where σ is the surface tension. We now require

equivalent to

Having chosen a model liquid, we have fixed ρM, μm and σM . Thus in 1, K is fixed.
Substituting for U from 1 into 2 and 3, we get

which implies that D3 = d 3K 2. This may together with the substitute


for U into 3 to give:

or equivalent:

This gives the final scaling relation:


which should be used for selecting the test fluid. For water and oil we have typical fluid
properties as in the table below:

Fluid Density (kg/m3) Viscosity (mPa-s) Surface tension (N/m)


Oil 800 2 30
Water 1000 1 70

The left side of the scaling relation then yields a value 0.997 which is a fairly useful
approximation although not perfect. However, this implies that K = 1.6 and consequently:

This means that the diameter is downscaled to 13% while the velocity is nearly 8 times
higher than in the full-scale case. As an alternative to selection of test fluid one may select
the test pipe diameter. In this case however, conditions will be put on density, viscosity
and surface tension. This might not be possible to fulfill in all cases. Thus fluid choice
might be more attractive in cases like this. Note also that it is not in all cases that the
scaling relation can be fulfilled.

3.1.4 Important dimensionless groups in fluid dynamics


A short survey of important dimensionless groups is presented here for reference.
Reynold's number

expresses the relation between inertia forces and viscous forces. Here D is a
characteristic length which is typical or important for the flow (pipe diameter, obstruction
size, liquid depth, etc.). In many cases there may be several dimensions describing the
problem. In that case also several Reynolds numbers will appear in the problem, or the
Reynolds number may be expressed in terms of several characteristic lengths. For flow
in annulus there will be two length scales, due to an outer and inner diameter. When the
Reynold's number is small, the dissipative (viscous) mechanisms are important, provided
the flow does not involve to strong gradients in space and time. At high Reynold's number
(typically > 4000) the flow becomes turbulent.

Froude number

gives the ratio between inertia forces and gravity forces (or buoyancy). The Froude
number is important in flows involving gravity waves. Fr may describe the degree of wave
resistance against a ship hull, as well as the degree of slugging in a horizontal pipe line.
It may be interpreted as the ratio of flow velocity U to the velocity of long gravity waves on
shallow water. When the Froude number is small the gravity forces keep the liquid
interface between two fluids plane. While at large Fr the dynamical pressure ρU 2 starts
to dominate and waves are easily generated

Weber number

is the ratio between dynamical forces and surface tension. A drop with diameter D falling
with a velocity U in air will be closely spherical if the Weber number is small. In horizontal
stratified wavy two phase flow the Weber number could be given by the flow velocity U
and D being either the pipe diameter, the liquid height or a typical wavelength or
amplitude. At large values of We the capillary effects are small, while they will dominate
at small We.

Strouhal number

is associated with the characteristic frequency that is generated when a fluid stream
passes a restricting bluff body. If D is the dimension of the restriction and the flow velocity
is U, the Str may be used to relate the frequency at various flow velocities, since it is found
that Str is practically constant over large domain of Reynolds numbers. This is applied in
the Vortex meter for flow measurement. It consists of a pipe segment with flanges to be
fit to the pipeline where measurement is to be carried out. Inside the meter a rod (“bluff
body”) is placed transverse to the flow. In turbulent flow where the effect of viscosity is
small, a pressure oscillation is set up between the two lateral sides of the rod. This is well
known as the “singing” around house corners and flag rods. The frequency may be
detected by a ΔP pressure sensor or a microphone.

The velocity is then calculated as:

Stokes number

expresses the friction forces relative to gravity forces.

3.2. Viscous flow


In this section, we will primarily study viscous flow. In the next section turbulence will be
studied. The Navier-Stokes equation

contains a “troublesome” term


most often referred to as the convective term. This apparently simple term actually is a
sum of 9 individual terms. Physically speaking it is responsible for viscous instability and
is the ”reason for existence” of turbulence. It mixes the local flow velocity components with
all the derivative of the velocity components. Expressed in another way, the flow velocity
in a given direction is coupled to the transversal velocity components and can be free
within the limits set by the continuity equation. Transverse instability is not appreciable at
low Reynolds numbers because it would set up excessive shear flow which is effectively

damped by the viscosity term .

At a critical Reynolds number, the inertial energy in the flow becomes strong enough to
sustain instabilities and make them “live” long enough to generate new instabilities.
Normally one refers to a transition region of Re for transition to turbulence from 2000 to
4000.
Turbulence transition 2000 < Re < 4000

As Re increase beyond 4000 the instabilities become more and more pronounced. We
need Reynolds numbers higher than 105 - 106 to obtain what is sometimes referred to as
isotropic homogeneous turbulence. This is an idealized picture of turbulence which will be
discussed in more detail in the section about turbulent flow.

In steady state, laminar one-dimensional flow (straight pipe) we may neglect the
convective term. However, in expanding or contracting pipe sections this term is
important, and it will occur in connection with Bernoulli’s equation.

Example 6. Derive the friction factor for steady-state laminar flow from the Navier-Stokes
equation.
Solution: From Navier-Stokes equation we may in this case neglect both the explicit time
derivative term as well as the convective term. Thus, the whole left-hand side vanishes.
Since we intend to study only the frictional forces the gravitational force may be neglected
as well, and we have
or

This is a vector equation, but we will consider only one-dimensional flow in the z-direction
(cylinder coordinates) along the pipe. Since the flow is assumed uni-directional the other
components are zero (ur=uƟ=0). In a circular pipe, the steady-state solution for flow
velocity u (r) z is only dependent on r (distance from pipe center). We consider then only
the component of the above simplified Navier-Stokes equation in z-direction which
becomes

This equation is solved to obtain the velocity profile uZ(r) in terms of pressure gradient.
Next we integrate the velocity profile to find the average pipe flow velocity U in terms of
the pressure gradient. This relation is then simply reordered to express pressure gradient
in terms of U.

• Rearranging the equation yields: which may be integrated to

give . This equation may be used to find uZ(r), renamed simply to


u(r) by rewriting it in differential form:

By integration from pipe center (r’ = 0) to pipe wall (r’ = r), we obtain:

From this expression u(0) is determined by the condition that u(r=R) = 0. Thus
which is replaced in the previous equation to give

which shows that the velocity profile is parabolic with the maximum at the pipe center. In
one-dimensional flow, the partial derivative may be replaced by ordinary derivative, thus

Integrating the velocity profile over the pipe cross section (area integration) we have

Since A = πR2 we get

By reordering and simple manipulation, we get

When we compare this result with the general expression from Chapter 1 and from the
dimensional analysis,

we obtain the friction factor in laminar flow as an exact analytical result.

3.2.1 Some other important relations for laminar flow

Average velocity related to the maximum of the velocity profile


Cross sectional integration (averaging) of the velocity profile give, from the previous
section
This applies in steady state laminar pipe flow. Note however that the laminar profile may
be disturbed e.g. by bends or restrictions (valves). Thus, this result and also the laminar
flow profile are restricted to stabilized situations.

Pump effect
Pumping with constant velocity U gives rise to a steady state pressure gradient dp/dz
along the pipe (note that in most of this book we use x as the flow direction). The pumping
effect PE may be expressed as: PE1 = work done per unit time = pump force times flow
velocity. Thus

Here we have used the previous result that the pressure gradient is proportional to U, thus
K now contains all other quantities than U. The first part of this equation applies to all
steady state situations, even turbulent flow. However, the latter part PE=K.U2, is the
laminar relation. It is found in the next sections that the pressure gradient is proportional
to U1.8 in strong turbulence, thus the appropriate expression for effect would be PE=K.U2.8.

The symbol used for effect is normally P. However, P is here used for pressure, thus PE
is used here to avoid confusion.

3.3. Turbulent flow


From a fluid engineering point of view the most dramatic effect of turbulence is the strong
increase of friction. This may be seen e.g. in the plot of friction factor in Chapter 2. This
increase may be understood when the velocity profile is considered. In developed
turbulent flow the common way to express the velocity profile is based on the power law
form:

where n ranges from 1/5 in weak turbulence, to 1/7 in strong turbulence. Compared to the
parabolic laminar flow profile the turbulent profile is flat with a relatively higher velocity
closer to the pipe wall. Then the velocity has to fall to zero at the wall in a much shorter
distance. Thus, the shear becomes stronger, and with it the friction.

Another important feature is the relation between cross sectional average velocity and
pipe center line velocity (umax). It is shown later that:

for the turbulent flow profile.

Statistical aspects of turbulent flow


The velocity components fluctuate in turbulent flow. The fluctuations are largely random,
i.e., there is no way to predict the time variation. One may only describe average values.
For steady-state situations it is useful to describe the velocity at a given position 𝑟 at a
given instant t, as a sum of an average velocity and a fluctuation. For space component
i (i = x, y, z), we may write

The average velocity is calculated as

Time averaging u follows the same simple rules as averages 〈𝑢〉 discussed in Chapter 1.

3.3.1. The structure of turbulence


In developed turbulence, the wall shear rates dy/du and transverse eddy momentum is so
high that instabilities cannot be damped by viscous dissipation. Turbulence is therefore
primarily a wall phenomenon, although the turbulent eddies spread into the central part of
the pipe, and there initiates new eddies. However, experiments also show that although
the turbulence is created near the wall, the zone very the wall is laminar. This is called the
viscous or laminar sublayer.

In this layer, the flow velocity is very small (zero at the wall) and thus easily damped by
the viscosity. This leads to a more complicated total velocity profile than discussed initially
in this section. As a preliminary summary, we have:

• The fully turbulent zone. This zone - which covers most of the pipe cross section –
is adequately described by the power law profile. Note however, that the flow profile is a
time averaged picture. At a certain instant, the profile may differ very much from the
idealized mathematical expression.
• In a transition zone close to the wall, typically a millimeter away, the power-law
relation must cease. One may easily show that the power law profile leads to infinite shear
forces at the wall, and thus is mathematically unfavorable. Here we find a transition zone
before we enter the laminar zone.
• The laminar or viscous sub- layer immediately outside the wall. Here the shear
stress may be assumed to be constant. Thus, the velocity profile must be a linear function
of distance from the wall.

The turbulent zone


For fully turbulent flow the shear stress can be modelled as

Here l is a length which is connected to the size of the turbulent eddies and is often
referred to as Prandtl’s mixing length. This expression allows us to calculate l as a function
of r. In steady-state turbulent flow this is easily calculated by force balance considerations.

Example 7. Derive Prandtl’s mixing length in steady-state turbulent flow as a function of


distance r from the pipe center.
Solution: We first show that τ is a function of r. Since the velocity profile u(r) is known, it is
straightforward to calculate dy/du and then l follows immediately. To find τ consider a
section of the pipe having length L and cross section area A. The pipe radius is R and the
perimeter associated with R is O. In steady-state flow the friction force balances the
pressure force, thus

where we use 0 = 2πR and A =πR2


and τw = τ(R) is the shear stress at the wall. In steady-state flow this also applies to any
fluid cylinder segment inside the pipe. Assuming a fluid cylinder with radius r we get (using
0= 2π r)

We then get that the shear stress is proportional to the distance from the center of the
pipe

where y = R-r is the distance from the pipe wall. Combining this with

we get the relation

The quantity is called the friction velocity or sometimes the wall-friction


velocity, since it is defined by the friction and having dimension of velocity. If the velocity
profile is known, it is straightforward to calculate dy/du. The friction velocity can be
calculated from the pressure gradient and the density ρ. The mixing length is of technical
importance in many situations ranging from chemical reaction speed in chemical plants,
burning velocity in combustion, friction factor in pipe flow as well as mixing and dispersion
of gas bubbles in two-phase flow.

The viscous sub-layer


In the viscous sub-layer, the flow is laminar, thus the relation between shear-stress and
shear rate is as follows

The velocity at the wall is zero, and we obtain the velocity profile

where

is the kinematic viscosity

We may introduce the dimensionless velocity u+ and distance y+, which gives the
dimensionless profile

Thus, Laminar layer velocity profile: u+=y+

The viscous layer is very thin, typically having a dimensionless thickness y+W=11.6
Other researches state that the value is y+W=5.
This profile is often referred to as “law of the wall”.

The turbulent layer


The turbulent velocity profile may be determined from the previous relation
If we restrict the analysis to regions close to the wall (typically y/R < 0.01) we may write

Prandtl hypothesized that the mixing length increased with distance from the wall in this
region, l = k.y. Prandtl’s student Nikuradse in his measurements (1930) found k = 0.4. By
inserting l, and rearranging the equation we obtain the differential equation

By indefinite integration we obtain

The constant C is found by using the condition u = umax at y = R (pipe center), thus

By taking the difference between the two expressions, we obtain

Inserting the value 0.4 for k and using Briggs’ logarithm this transforms to

Note that umax is developed for the near wall region and therefore not really comparable
to the umax for the full pipe profile. It will be removed from the equations by requiring that
the velocity profile of the laminar and turbulent regions is continuous:
Connecting the laminar and turbulent regions
The two velocity profiles must equal at the edge y=yw of the laminar region, where

Combining with the turbulent solution at y = yw, we get

thus we find
Also, we transform the argument of the logarithm, and get

or in dimensionless form Turbulent layer velocity profile: u+=5.5 +5.76 log(y+)

This profile is often called “the universal velocity profile”.

A unified near wall profile


The two velocity profiles are continuous at y = yw. However, the derivatives are not
continuous. In the range 5 ≤ y+ ≤ 30, which is referred to as the buffer layer, the velocity
profile is something between linear and logarithmic. Spalding (1961) suggested a single
formula which covers the whole wall region up to y+ > 100.

Note that this equation is transcendent and has to be solved numerically if u+ is to be


calculated as a function of y+. For plotting purposes, it may be sufficient to have a relation
between u+ and y+, in which case one may express y+ as a function of u+ as above.

The connection between velocity profile and friction factor


Since friction is related to the shear rate dy/du it is obvious that also the velocity profile
and friction factor must be connected. For power-law velocity profiles of the turbulent type
it may be shown that it is connected to the Blasius type friction factor f = C ⋅Re−m by the

relation . The n = 1/7 velocity profile thus is connected to the m = 0.25 friction
factor. It is also possible to show that the wall velocity profile is directly related to the
Nikuradse-Colebrook friction factor:

where fM means that the friction factor is of Moody type.

3.4. Turbulence models


It is possible to illustrate the consequence of the Prandtl mixing length in a conceptually
simple way. Assuming turbulence in a shear field with velocity profile u(y), there is a shear
rate dy/du. Consider two nearby points (1) and (2) along a line in y-direction away from
the pipe wall. The distance between the points is l and is taken as representative for the
mixing length, typically the diameter of the turbulent eddies. Then there will be a difference
between the velocities at the two points given by

There is then an obvious relation appearing from the velocity profile

If homogeneous isotropic turbulence may be assumed, there are fluctuations in all


directions with the same amplitude, so that v′ = u′ . We the have the expression for
turbulent shear stress

Comparing this expression with the relation used in laminar flow


we may write in a similar way a turbulent shear stress

where we have introduced the eddy viscosity

We note that the eddy viscosity depends on shear rate and thus in some respect
resembles a non-Newtonian viscosity. However, the analogy with non-Newtonian fluids is
superficial. For ordinary non-Newtonian fluids (polymers) the mechanism is on a
molecular level, while the turbulent eddy viscosity in fact is a consequence of (complete)
removal from molecular viscosity.

3.4.1. Turbulence from time averaging of the Navier-Stokes equation


We will show in a little more rigorous way how the turbulent shear stress τR arises naturally
by time averaging of the Navier-Stokes equation. It is a vector equation, but for simplicity
(with no loss of generality) we consider only one spatial component (i) of the Navier-
Stokes equation. We introduce as defined in the beginning of section 3, the flow variables
(pressure and velocity essentially) as a sum of a mean value and a fluctuation. We use
the notation that the mean value is written with large letters, while time dependent values
are written with small letters. The fluctuations are primed (‘).

On macro length scale and limited time resolution it is practically impossible to keep track
of detailed fluctuations. For engineering purposes, it is of little interest to keep track of
individual fluctuations. However, we shall see that they have large scale consequences,
giving rise to new terms in the Navier-Stokes equation. We consider first the continuity
equation under the assumption that the flow is incompressible.

The continuity equation

can be time averaged to give:

The last left side term vanishes because the time average of fluctuations are zero.

Furthermore and thus we obtain

Inserting this in the starting equation gives

We introduce in similar way average + fluctuation for the Navier-Stokes equation

By time averaging and by using the previous result

we obtain the following equation which is form equivalent to the original Navier-Stokes
equation:

We now consider a steady-state two dimensional system with i=x and j=y and get
The terms named 1 and 2 disappear. Both U and is independent of x.

The two last terms in the remaining equation then give

The term is called the Reynolds stress. The negative sign may indicate that the
Reynolds number decreases the viscosity and thereby the friction. However: this is not
the case. One usually finds that u’ and v’ are anti-correlated; thus if u’ is negative v’ is

positive and vice versa. In that case the time average and we may write

Total shear stress may then be written

which is a sum of viscous and turbulent shear, as discussed earlier.

3.5. Other important flow phenomena


We include some other important contributions to dynamics in two-phase flow. The most
important basic theory is Bernoulli’s equation which will be considered briefly. We also
discuss the Marangoni effect which depends on surface tension. Bernoulli’s equation
describes pressure in flow as a function of velocity changes due to acceleration. It explains
a lot of phenomena like slug flow, aeroplane wing lift, and Venturi flowmeters. The Magnus
effect describes the movement of a rotating ball (e.g. footballs and tennis balls).
Transverse forces arise which causes the ball to move sidewise.
3.5.1. Bernoulli’s equation

It is derived straightforward from the Navier-Stokes equation. It expresses simply that the
sum of pressure energy and kinetic energy is constant. Bernoulli’s equation is valid only
if the flow friction is negligible. Paradoxically, this most often happen in strong turbulence,
where the friction is very strong. Another criterion for validity is that the flow must be

steady-state (no explicit time dependence: ). The hydrostatic pressure must also
be neglected. It is possible to validate Bernoulli’s equation:

1. Locally: along a stream line

2. Globally: provided the flow is irrotational (defined by )

This is shown in the following examples.

Example 8. Show that Bernoulli’s equation is valid along a streamline.


Solution: Typical for a streamline is that all particles close to the line flows parallel to the
line. We let u be the velocity along the streamline, and v be the velocity normal to the line.
The spatial directions are l along the streamline and r is normal to the line. We simplify
the Navier-Stokes equation by demanding steady-state and remove viscous force and
hydrostatic force. Then we have only .In the proposed coordinate
system it becomes

We consider only the component along the streamline

Now
The simplified Navier-Stokes equation then becomes

which is equivalent to

Bernoulli´s equation follows immediately from this expression.

The following example illustrates global validity of Bernoulli´s equation.

Example: Show that Bernoulli’s equation is valid generally for a flow if is fulfilled.
Solution: To prove this result we recall the simplified Navier-Stokes equation
𝜌(𝑢 ⃗ ) = −∇𝑝
⃗ ∙ ∇𝑢

Furthermore, we need the vector identity:


1 1
(𝑢 ⃗ ) = ∇(𝑢
⃗ ∙ ∇𝑢 ⃗)−𝑢
⃗ ∙𝑢 ⃗ ) → ∇𝑢2
⃗ × (∇ × 𝑢
2 2

The rest of the proof is left to the reader.

3.5.2. Koanda effect


The Koanda effect is a consequence of Bernoulli’s equation. An example of the Koanda
effect is shown in Figure 1. A jet coming from a nozzle to the left on the figure will be
drawn towards the nearest wall. This is caused by a back-flow close to the the wall. The
back flow is set up due to continuity to conserve volume when the fast jet flows forward.

Figure 1. Koanda effect


The fluid is caught up by the jet at the exit and accelerates to the right. This requires
energy which is taken from the pressure. A lowered pressure between jet and wall thus
forces the jet to the wall.

Similar phenomena are seen between two close jets, as illustrated on the figure to the
right. The pressure is lowered between the two jets due to the re-circulation zone, and the
two jets may merge into one.

3.5.3. Magnus force


The Magnus force is a consequence of Bernoulli pressure variation for rotating bodies
moving in fluids. The transverse Magnus force is given by

where is the velocity relative to the fluid, is the particle spin vector, ρ is the density,
and C is a geometry factor of magnitude close to 1.

There are two common flow regimes of the Magnus effect referred to as the
• “positive” or “conventional”, and
• “negative” or “anomalous”

Magnus effect. These are shown in the next figures. The conventional effect gives the
force in accordance with the above figure. In this case the stagnation point is shifted
counter clockwise, i.e. in the rotation direction, and the “wind speed” as seen from the ball
becomes highest on the “right side” in this Figure 3. The ball moves from right to left in
both figures.

The negative Magnus force is observed at low Reynolds numbers. In this case the “wind
velocity” at the ball boundary on the upper side can be lower than the value required for
turbulence, while the lower side develops a turbulent boundary layer due to higher
apparent wind velocity and stronger shear. A turbulent boundary layer dissipates more
energy and thus the pressure is reduced more on the lower side. This causes a
downwards deflection.

Figure 2. a) Negative or anomalous Magnus force (Re<Recr)


b) Positive or conventional Magnus force (Re>Recr)

In dispersed flow with bubbles and droplets the rotation is normally set up by the velocity
profile. As shown in the figure, a droplet with upward rise velocity will rotate counter
clockwise and will experience a force directed towards the wall. This may lead to a higher
concentration of bubbles close to the wall.

In real flows a variety of different concentration profiles are found. There is a complexity
of flow effects which set up the observed profiles, and even CFD simulation will have
problems with accurate prediction of dispersed multiphase flow.

Nevertheless, this is an important issue, since various engineering problems are


associated with the concentration profile. It concerns water droplets and wetting of walls,
and thus is important for corrosion problems. In the same way, it also is important for
erosion caused by particle movement. In gas liquid flow a higher concentration of gas
bubbles close to the wall may reduce the heat conduction between wall and liquid, and
thus influence heat exchangers, and sub-sea pipeline flow temperature.
Figure 3. Concentration profiles

3.5.4. Flow driven by surface tension


Marangoni convection is flow which is driven by variations in surface tension. Examples
are:
• “Tear flow” - which can be observed on the wall above the wine in wine glasses.
• Surface movement in Benard convection. This flow is often called Marangoni-
Benard convection.

The “tear flow” in wine glasses is due to a varying alcohol-water concentration in the vapor
over the wine surface. The surface tension is a function of the difference in concentration
in the vapor and in the flow tears.

The surface tension effect in Marangoni-Benard convection is similar to what happens if


a small amount of washing detergent is dropped onto water. If some dust is present on
the water surface, the flow pattern is easily seen.
The Figure 4 below illustrate the two types of flow.

Figure 4. a) Marangoni convection b) Marangoni-Benard convection.


The hot spot at the top of the liquid has a higher
temperature than the rest of the surface. The surface
tension is lower here than for colder liquid and the
flow from the hot spot increases compared to if the
effect of surface tension reduction can be neglected.

Exercises
1. Assume that viscous friction F against sphere with diameter d, which moves with
constant velocity v through a liquid with density ρ and viscosity μ.
a) Show that F can be written as

where φ (x) means “function of x”


b) Show that Stokes result for very low velocities: F = 3πμvd is in accordance with the
general formula above.

2. Show that flow and wave resistance R against a surface vessel can be written as

where density is ρ, and ν is kinematic viscosity of the liquid. The vessel has a characteristic
dimension l, U is the velocity and g is the acceleration of gravity.

3. Write the convective term in Navier-Stokes equation in full for:


• The x-component in Cartesian coordinates
• The r-component in cylinder coordinates

3. Calculate the shear stress at the wall based on the equations and

in viscous flow. Can these equations be used in turbulent flow?

4. Assume in turbulent flow that we have a “power-law” velocity profile given by:

Show for this case that:


a) The average velocity U in the pipe will be

b) There will be a relation between the velocity profile and the friction factor
f = C Re−m so that n = m/(2-m).

Hint: This may be done by assuming that u(r) is a function of r/R only. Further,
express the shear stress at the wall, W by u(r) and r. Finally, assume that τw is
explicitly independent of R, so that the exponent for R i the resulting expression
for τw must be zero.

5. Calculate the thickness yw of the laminar zone in turbulent flow, for the case U=1m/s,
pipe diameter D=10 cm. Density and viscosity is as for water.

6. Explain why Bernoulli’s equation (“non-viscous”) can be assumed valid in strong


turbulence even though friction is large.

Answer: The condition of no friction for validity of Navier-Stokes equation should

be refined. The essential condition is that . As long as the viscous term


μ∇2u􀁇 is large compared to the convective term , this is not obeyed. However,
as turbulent friction becomes larger, the term becomes more important,
and equivalently the viscous friction becomes negligible.

You might also like