Anda di halaman 1dari 34

BMED3300 Biotransport Lecture Notes C.

Ross Ethier 2004, 2013


Page 1 of 34
Lecture 11: Dimensions. Buckingham Pi theorem. Similitude.
Model design.
All (or nearly all) physically measurable quantities have units associated
with them. But nature does not use any system of units units are an
invention of humans. This means that natural processes should be
independent of any system of units. We will see that this can be
formalized by the idea of a H-group.
Dimensions: All systems of units are based on fundamental quantities
such as mass, length and time. We refer to such quantities as dimensions.
Note that dimensions are independent of any system of units, and
instead reflect what is being measured or expressed.
We will consider a hierarchy in which the fundamental dimensions are:
Mass, symbol M
Length, symbol L
Time, symbol T
Temperature, symbol
There are other fundamental dimensions, but we will not need them for
this class. The dimensions of all other quantities can be derived from
these fundamental dimensions. For example, volume is a product of
three lengths, and therefore has dimensions of L
3
. We write this as
[Volume] ~ L
3
, where the notation [ ] ~ means has dimensions of.
Example: What are the dimensions of force?
Answer: We can determine this by using F = ma to write:
[F] = [m][a] ~ M LT
-2

Note that we are not concerned with numerical coefficients when
working with dimensions.
Buckingham H-Theorem: We argued earlier that nature does not care
about systems of units. This implies that mathematical and empirical
relationships describing a system should be independent of units, i.e.
cannot depend on what are known as dimensional quantities. Instead,
such relationships in nature must be formulated in terms of quantities
that are independent of units, i.e. that are dimensionless. We call
quantities that are dimensionless H-groups, and say that nature talks in
H-groups.
Example: The Reynolds number for flow in a pipe is defined by
Re


BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 2 of 34
where is the diameter of the pipe and is the average fluid velocity in
the pipe. Show that the Reynolds number is a H-group.
Answer: To show that ReD is a H-group, we must show that it has no
dimensions. We can do this by listing out the dimensions of the quantities
appearing in ReD:
[]
1

[]
[]
3

[]
[]
[

/]

1
=
1

1

That means that the dimensions of

are
[

3

1

1
= 1
In other words,

has no dimensions and is therefore a H-group. Such


a combination of quantities is said to be dimensionless.
The Buckingham H-theorem provides a formal way to generate H-groups
from a set of dimensional quantities. It also tells us how many H-groups
we can create from such a set. It works are follows. Suppose that there
are n variables that are known to play a role in a given problem,

1
,
2
,

. Note that the identification of these n variables can be quite


tricky, and is extremely problem-dependent. These n variables will each
include some dimensions
1
,
2
,

, e.g. M, L and T. We create a pn


matrix of the form shown below

[




]

where entry (i,j) is the exponent of

in the quantity

. We denote the
rank of this matrix as . Note that s . (Usually = , but not always.)
Then the number of H-groups, , is given by
=
To form the H-groups: We choose r of the n quantities, and refer to these
r quantities as the core group. The r quantities must contain all p
dimensions, and cannot themselves form a H-group. Suppose for the
purposes of this example we choose
1
,
2
,

as the core group. Then


we formulate the first H-group by writing:

1
=
1

+1

BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 3 of 34
The exponents a, b, and c are determined by requiring that H1 be
dimensionless. We then repeat this process to get a second H-group

2
=
1

+2

where the exponents a, b, and c are different than for H1. We repeat this
process, each time replacing the last entry in the H-group until we have
formed = such groups.
Example: The pressure drop per unit length in turbulent pipe flow, /,
is known to depend on the pipe diameter, , the pipe wall roughness, c,
the fluid velocity, , and the fluid viscosity and density, and . What H-
groups can be formed for this problem?
Answer: It is useful to list all of the relevant variables and their dimensions.
Variable Dimensions
/ ML
-2
T
-2

ML
-3

ML
-1
T
-1

c L
L
LT
-1

The resulting matrix is:

[
1 1 1 0 0 0
2 3 1 1 1 1
2 0 1 0 0 1
]

This matrix has rank = 3. Therefore, there will be
= = 6 3 = 3 H-groups.
We arbitrarily choose , and as the core group. Then we form the
first H-group as:

1
=

(/)
We can compute the dimensions of H1 from:
[
1
]

(
1
)

(
3
)

2
=
0

0

where we set the exponents of M, L and T to zero since H1 must be
dimensionless. Equating exponents of M, L and T gives 3 equations for a,
b, and c:
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 4 of 34
M: c +1 = 0 therefore c = -1
L: a + b -3c -2 =0
T: -b -2 = 0 therefore b = -2 and a = 1
This means that the H-group is:

1
=

2

This H-group appears so often in pipe flow problems that it is given a
special name. Actually, twice H1 is called , the friction factor.
Now we can do the same procedure to get the next H-group, H2.

2
=


Once again substitute in dimensions:
[
2
]

(
1
)

(
3
)

=
0

0

Solving, we get a = -1; b = c = 0, so that H
2
= c/, the relative
roughness. Finally, doing the same thing for the last H-group we get:

3
=


Once again substitute in dimensions:
[
3
]

(
1
)

(
3
)

1
=
0

0

Solving, we get a = b = c = -1, so that
3
=

=
1
Re

.
Notes:
- Any product of H-groups is also a H-group.
- A H-group to any power is also a H-group.
- A H-group multiplied by a numerical factor is also a H-group.
From the above it should be clear that there are infinitely many possible
H-groups. However, only = of these are independent. Different
choices for the core group of variables will lead to different H-groups. To
keep things manageable, certain H-groups are standard and are always
used. In this problem they are , c/ and

.
The Buckingham H-theorem is useful in two different ways.
Efficiency in Data Correlation and Relationships: Think about our
example of the pressure drop in pipe flow. Without H-groups, the most
that we can say about the problem is:
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 5 of 34
A/ =
1
(, , , c, )
which means that the unknown function
1
depends on 5 variables.
Discovering this function is very hard from experimental measurements,
and even once it is determined, presenting it in charts or tables is almost
impossible.
It is much simpler to use H-groups. Knowing that the fundamental
relationships about pressure drop in pipe flow must involve only H-
groups, we can write:
=
2
(
c

)
Notice that f is the H-group that includes the pressure drop, and depends
on only two other quantities. The form of
2
is much easier to determine
from experiments than the form of
1
, and it is also easier to express. In
fact, it all fits on one page (the Moody chart). Notice that H-group
analysis does not tell us anything about the form of
2
. To determine this
we still need to do experiments or analysis, but the H-groups make this
easier.
Example: The pulse wave speed in an artery, , is a function of the fluid
density, , and the distensibility of the artery, |, defined by:
=
1


where is the cross-sectional area of the artery. Find an expression for
in terms of and |.
Answer: We first determine the dimensions of the various quantities
involved in the problem, as follows:
Variable Dimensions
LT
-1

ML
-3

| LT
2
M
-1

Notice from the definition of |, it has the same units as 1/p. We can then
form our matrix of coefficients as:

[
0 1 1
1 3 1
1 0 2
]

This matrix is degenerate, and has rank = 2. Therefore, we can form
= 3 2 = 1 H-group. It is easy to compute this H-group as
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 6 of 34
H
1
=
2
|
This does not seem that useful, but if we think about it a bit more we can
actually get something very useful from it. In general, the behaviour of a
physical system must be governed by a relationship of the form H1 =
function(H2, H3, ). However, for the case of this example there is only
one H-group, so that the function in the above expression takes no
arguments. The only such function is a constant, so that we immediately
have that H1 = const. Therefore we can write the wave speed as:
=


Dimensional analysis does not tell us what the constant is in the above
expression, but it does tell us about the form of this expression. This is
quite handy.
Similitude: The second way in which H-groups are useful is in similitude,
which is the scaling of transport phenomena from one situation to
another. This usually involves trying to extrapolate the performance of a
model to that of a full-scale device. To understand how this works we
define 3 different types of similitude.
1. Geometric Similitude: Two objects are said to be geometrically
similar if one is a simple enlargement of the other, i.e. all lengths are
scaled up by some constant factor. This scaling can be quite tricky,
including, for example, surface roughness, etc.
2. Kinematic similitude: The velocity fields around two geometrically
similar objects are kinematically similar if all velocities around one
object are identical to the velocities around the second object to
within a scaling factor.

In other words, vA = C vA for all pairs of points AA. Note that if model
2 is k times larger than model 1, then the distance from A to the
origin must be k times larger than the distance from A to the origin.
In other words, we must compare velocities at scaled locations.
Model 1
Model 2
A
A
vA vA
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 7 of 34
3. Dynamic similitude: Each of the terms in the Navier-Stokes equations
can be interpreted as a force per unit volume (or inertia per unit
volume). As is the case for all equations, each term must have the
same dimensions as the other terms. We can therefore form H-
groups by taking ratios of the terms in the Navier-Stokes equations.
There are three common H-groups that result when this is done.
The Reynolds Number: Re

=
inertia
viscous forces
The Froude Number: =

=
inertia
gravitational forces
The Euler Number: =

2
=
pressure
inertia

where , and are suitable length, velocity and pressure scales. If
these H-groups match between two flows, then the flows are said to
exhibit dynamic similitude. In such cases, all force-related quantities
in the flows can be scaled using similitude.
Aside: Show that the Reynolds number is the ratio of inertial to
viscous effects. We can do this by considering the relevant terms in
the governing equations:
Inertia scales as



Viscous forces scale as
2

2


The ratio is

2
/
/
2
=

= Re


In addition to the H-groups arising from the governing equation,
there can also be H-groups arising from the boundary conditions.
This is very problem-dependent. Note that in real problems, one can
often argue on physical grounds that one or more of the above H-
groups do not play a role in the problem, e.g. the Froude number is
usually only important when there are free-surface flows. In that case
we do not need to include the unimportant H-groups in the analysis.
Knowing what groups are/are not important can be very TRICKY!
The important point from all of this is that if both geometric similitude
and kinematic similitude exist for flow over two different objects, then
dynamic similitude also exists for these flows. Alternatively, if geometric
similitude and dynamic similitude exist, then kinematic similitude does
also. In practice, we usually try to enforce geometric and dynamic
similitude in experimental testing of fluid mechanical performance using
scaled models. More specifically, when we make a model, we must
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 8 of 34
ensure that it has geometric similitude to the actual device. Then we
determine what H-groups are important based on the details of the
problem, and enforce dynamic similitude using those H-groups. Note
that if we have H-groups in a problem, we must only match 1 of
them, since we know that H
i
= (H
1
, H
2
, H
i1
). In other words,
matching 1 of the H-groups is sufficient to fully constrain the system
and ensure matching of all the H-groups.
Example: A blood oxygenator is being tested to ensure that the pressure
drop on the blood side is not too large. A 3 scale model is made, and a
blood analogue fluid with viscosity 19 cP and density 1.8 g/cm
3
is pumped
through it. If the design flow rate is 5 L/min, what should be the blood
analogue fluid flow rate? If a 32 mmHg pressure drop is measured across
the 3 scale model, what will the corresponding pressure drop be for the
real system?
Answer: It is convenient to summarize everything in a table:
Quantity Configuration 1 (Blood) Configuration 2 (3 larger)
3.5 cP 19 cP
1.05 g/cm
3
1.8 g/cm
3

Q 5 L/min ?
p ? 32 mmHg
Treating the oxygenator as a series of pipes (tubes), we expect that the
pressure drop can be computed from f = fcn(c/D, ReD). To ensure
similitude we need c
1
/
1
= c
2
/
2
, which is part of the requirement for
geometric similitude. We also need Re1 = Re2, which we can write as:

4
1

1
=
4
2

2
(*)
where we have rewritten the Reynolds number in terms of the flow rate
for a pipe of diameter D. By manipulating equation (*) we obtain a
constraint on the flow rate in Configuration 2, namely

2
=
1
(

1
)(

1
) (

2
)
= (5

)(3) (
19
3.5
) (
1.05
1.8
)
= 47.5

min

This is the flow rate which will ensure that Reynolds numbers match
between the two configurations.
In this example there are only three relevant H-groups, so that if we
specify two of them, then the third must be determined uniquely. This
implies that by matching Reynolds number and relative roughness
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 9 of 34
between the two configurations, the friction factor must match between
the two configurations. We can therefore write:

1
=
2

1
2

1
=

2
2

2
(**)
From the geometry of the model, we have that
2
= 3
1
and
2
= 3
1
,
so that the / terms cancel out. We can therefore write:

1
=
2
(

2
) (

1
2

2
2
)
It is convenient to eliminate the velocities in favour of flow rates ,
which we can do by noting that =
4

2
, so that equation (**) becomes:

1
=
2
(

2
) (

2
)
2
(

1
)
5
(
L
1
L
2
)
= 32 (
1.05
1.8
) (
5
47.5
)
2
(3)
4

= 16.8
Lecture 12: What is a boundary layer? Laminar boundary layer
equations & Blasius solution.
Consider what happens when we immerse an object, say a sphere, into a
stream of moving fluid. We will suppose that the Reynolds number is
large, which means that viscous effects are not very important. Yet, at
the surface of the sphere we must still satisfy the no-slip condition, which
comes about due to viscosity. This presents an apparent contradiction: it
seems that viscosity is both important and not important.
The resolution to this apparent paradox is the boundary layer, or more
specifically, the momentum boundary layer. This is a thin region next to
a solid surface where viscous effects are important. Note that it only
makes sense to talk about a boundary layer in the context of a high
Reynolds number flow, since by definition the boundary layer is a small
region of the flow. As the Reynolds number increases, we expect that the
thickness of the boundary layer will decrease.
To analyze the boundary layer in more detail, it is best to consider the
simpler case of a thin flat plate in a uniform flow. This is shown
schematically below. Outside the boundary layer, viscous effects are
unimportant. For the case of the flat plate, the streamlines outside the
boundary layer are essentially parallel, and the velocity (the free-stream
velocity) is constant and equal to

.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 10 of 34
Inside the boundary layer, viscous effects are important and there is an
appreciable velocity gradient. We arbitrarily define the edge of the
boundary later to be the location where

is 99% of its free-stream value.


This defines a boundary layer thickness, o(), which depends on axial
position .

The boundary layer has an initial laminar region, followed by a turbulent
region. The transition between the laminar and the turbulent parts
depends on the Reynolds number based on x:


The transitional Reynolds number is somewhere between 2 10
5
and
3 10
6
, depending on the roughness of the plate and the level of free-
stream turbulence.
Let us now carry out a more quantitative analysis of the boundary layer.
In the y-direction, all changes take place over the length scale o. In the x-
direction, the length scale for changes to take place is x. Notice that o <
< . This means that / will scale with 1/, while / will scale with
1/.
We will carry out this analysis by considering the magnitudes of the
various terms that appear in the governing equations. We will see that
some terms will be small and can be neglected. The remaining terms must
balance, and from this balance we can learn about the sizes of some of
the physical quantities of interest. Let us first consider the continuity
equation:

= 0
y
x
V
Edge of
boundary
layer
laminar
turbulent
o(x)
vx(y)
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 11 of 34
Based on the above argument, we expect that the first term will be
proportional to V/x, since vx scales with V. Since the second term must
be of the same order as the first one, we can then write


The symbol ~ means here is proportional to or scales with. The
above expression tells us the scale (magnitude) of the y-component of
the velocity.
Now let us consider the y-component of the Navier-Stokes equations:

1
=

2
+
2

3

We can estimate the scaling (magnitude) of each term as follows:
Term 1:

= [

] ~ [

+
1

)
2
] ~

2

Term 2:

. Note here that we have assumed that changes


in pressure scale with

2
, as predicted by Bernoullis equation.
Term 3:
2

= [

2
+

2
] ~
[

Small compared
to next term
since x
+

2
]

2
~


Now that we have estimates for the size of each term, let us compare
their relative magnitudes.
Term 1
Term 2
~
/
2
/
~(

)
2
1
Term 3
Term 2
~

2
/
~

~
1
Re

1
The last ratio is small because the boundary layer concept is only valid
when the Reynolds number, Rex, is large.
The above analysis shows that terms 1 and 3 are very small compared to
term 2. That means that we can ignore terms 1 and 3 in the y-component
of the Navier-Stokes equations, so that this equation reduces to

0.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 12 of 34
This means that the pressure does not change across the boundary layer,
i.e. that it is transmitted unchanged from the free stream flow across
the boundary layer. But in the outer flow for flow over a flat plate, =
constant and so = constant by Bernoulli. (If the object is some other
shape that a flat plate, then the outer streamlines curve and the outer
pressure is no longer constant.) When the outer pressure is constant,

0 therefore implies that the pressure is everywhere constant in the


flow, including in the boundary layer.
Now let us consider the x-component of the Navier-Stokes equations:

1
=

2
+
2

3

We can again consider the magnitude of each of the terms:
Term 1:

= [

] ~ [

+ (

] ~


Term 2: this is zero since the pressure is uniform in the boundary layer.
Term 3:
2

= [

2
+

2
] ~
[

Small compared
to next term
since x
+

2
]

2
~

2

Let us look at the ratio of the two remaining terms:
Term 1
Term 3
~

2
/
V

/
2
~(

)
2

~(

)
2
Re


Since these are the only two terms remaining in the governing equation,
they must balance each other, and therefore their ratio must be of order
one. (The only other possibility is that one term is much greater than the
other, so that the smaller term can be neglected and the larger term is
approximately zero. But this would imply that there is no velocity
gradient in the boundary layer, which we know cannot be true.) The fact
that term 1 and term 3 are of the same order also agrees with our idea
that the boundary layer is the part of the flow where viscous and inertial
forces balance one another.
Since this ratio is of order one, we can write:
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 13 of 34
(

)
2
~
1
Re

~ Re

1/2

This analysis predicts that the laminar boundary layer will grow as
Re

1/2
, or as

. A more rigorous analysis that involves solving a


reduced form of the Navier-Stokes equations shows that

= 5.0 Re

1/2

This is known as Blasius (laminar) boundary layer theory.
Lecture 13: Skin friction coefficient. Turbulent boundary layers.
Examples.
It is of interest to compute the drag on the flat plate due to the presence
of the boundary layer. The drag arises from shear stresses acting on the
surface of the plate, given by:

|
=0
~


From last lecture we have an expression for o, which we substitute into
the above equation to obtain:

Re

1/2
(*)
Note that this predicts that the shear stress will be infinite at the leading
edge of the plate ( = 0). This is a singular point in the flow where the
boundary layer formulation breaks down. The shear stress then
decreases as
1/2
.
It is convenient to express the local value of the shear stress in terms of
a dimensionless coefficient which we call the skin friction coefficient,

.
This is defined by

1
2

2

Substitution of equation (*) into the above definition gives

Re

1/2

2
~ Re

1/2

The Blasius analysis gives
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 14 of 34

,
= 0.664 Re

1/2

Laminar boundary layer
Notice that this is the local skin friction coefficient, i.e. it gives the shear
stress at a particular location . For this reason I have added a subscript
to the above expression. To get the total drag on the plate we have to
integrate the shear stress over the entire surface of the plate. Denoting
the drag force by

we have for a plate of flow-side length and width


:

()

0
(**)

We can define an average skin friction coefficient for the entire plate as:

,
=

/
1
2

2

Substituting equation (**) into this definition gives:

,
=
1

()
1
2

2
=

0
1

,
()

0

For the laminar boundary layer,
,
is proportional to

, so that we
obtain:

,
= 2
,
( = )
Laminar boundary layer
Turbulent Boundary Layer: Flows can be characterized as laminar,
turbulent, or transitional. Laminar flow is the easiest to understand and
analyze: it is characterized by smooth, well-organized streamlines. The
flow is deterministic in the sense that if the same flow is measured twice,
the same results will be obtained. Turbulent flow is fundamentally
different. Fluid layers no longer slide over one another in smooth paths.
Instead, the flow is inherently unsteady and chaotic in appearance, so
that it is impossible to precisely replicate the same flow in two
subsequent measurements. Transitional flow is a flow that jumps back
and forth between turbulent and laminar flow.
All of the above development for boundary layers assumed that the flow
was steady. Since one of the characteristics of turbulence is that it is
inherently unsteady, we cannot use the above equations to analyze
turbulent boundary layers. Unfortunately, the analysis of turbulent flow
is much more complex than for laminar flow, and we will therefore rely
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 15 of 34
on qualitative arguments and experimental results to treat turbulent
boundary layers.
For purposes of this course, the key point about turbulence is that it
drastically enhances mixing. This can be best understood by considering
a flow in which there is a velocity gradient, such as shown below. For
laminar flow, there is exchange of high and low x-momentum fluid
between layers (different y-locations), but this is only due to molecular-
level fluctuations. The net effect of all of these fluctuations is to create
what we call viscosity. When turbulence is present, there is again x-
momentum exchange between layers, but now large packets of fluid
are moving instead of single molecules. The net effect is that the fluid is
better mixed, so that x-momentum tends to be more uniform, i.e.
velocity gradients are diminished. This can be interpreted as being due to
a large increase in the effective viscosity of the fluid,

. Unfortunately,

is a complex function of flow characteristics, so that analytic


treatment of the governing equations is usually not possible.

Applying the above arguments to the turbulent boundary layer, we
expect that the velocity profile will be flatter over much of the boundary
layer when compared to the laminar case. However, we still have to
satisfy the no-slip condition, so that there is a very steep velocity gradient
right next to the wall (see figure). This means that the wall shear stresses
are higher in turbulent flows than in laminar flows.
y
vx(y)
For laminar flow, the circles represent fluid
molecules. For turbulent flow, they are packets of
fluid being vertically transported by random
fluctuations in the velocity field.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 16 of 34

We use semi-empirical results to quantify boundary layer thickness and
skin friction coefficient. The results for smooth flat plates are as follows.

= 0.376Re

1/5

,
= 0.0576Re

1/5

,
=
5
4

,
( = )
}

valid for 5 10
5
Re

1 10
7

= 0.16Re

1/7

,
= 0.027Re

1/7

,
=
7
6

,
( = )
}

valid for 10
5
Re

10
9

The first set of equations hold over a smaller range of Reynolds number,
but are more accurate over this range. Note that the equations for
,

given above only hold if the entire plate is covered by a turbulent
boundary layer.
Example: A flat plate of length has a fluid flowing over it, creating a
boundary layer. Suppose that the boundary layer over the first half of the
plate (0 /2) is laminar, while over the second half the boundary
layer is turbulent. Using the first set of expressions above for the
turbulent portion, derive an expression for the average skin friction
coefficient over the entire plate,
,
.
Answer: We can use the Blasius (laminar) solution for the first half of the
plate and the above equations for the second (turbulent) half. Note that
this will require integration. The situation is as shown below.
y/o
vx(y)/V for turbulent
boundary layer

vx(y)/V for laminar
boundary layer

Edge of
boundary layer

BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 17 of 34

It is easiest to consider a plate that is a depth into the page. Then the
surface area of the plate is and the area of a small element of the
plate is . From the definition of
,
, we can calculate the drag on
the plate as
=
1
2
V

2

,
()

0

We split this integral into 2 parts: one for the laminar front half and one
for the turbulent second half.
=
1
2
V

,
()

2
0
+
,
()

2
}


=
1
2
V

0.664 Re
x
1/2

2
0
+ 0.0576 Re

1/5

2
}


=
1
2
V

0.664(

)
1/2

1/2

2
0
+0.0576 (

)
1/5

1/5

2
}


The integration is a bit ugly, but eventually gives
=
1
2
V

2
{1.328
2
1/2
+ 0.0720 (

1/5

1
2

/2
1/5
)}
Finally, the overall skin friction coefficient is:

,
=

1
2
V

= 1.328
2
1/2
+ 0.0720 (

1/5

1
2

/2
1/5
)
y
x
V
Edge of
boundary
layer
laminar
turbulent
o(x)
vx(y)
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 18 of 34
Boundary Layers on Objects Other Than Flat Plates
For objects with a more complex shape than a flat plate, the above
analysis is qualitatively correct. However, there is a significant
complication: for such objects, the outer streamlines are no longer
straight, so that the pressure outside the boundary layer is no longer
uniform. This outer pressure is still transmitted into the boundary layer,
so that the pressure in the boundary layer is no longer uniform. This turns
out to be very important in understanding form drag, as we will see
shortly. However, we will not quantitatively analyze such situations in this
course.
Lecture 14: Drag, lift, drag coefficient and lift coefficient. Example:
peregrine falcon.
We will now talk about forces exerted on objects that are moving relative
to a stream of fluid. This does not seem like it has much to do with H-
groups, but we will see that dimensional analysis crops up here too.
When an object is placed into a flowing fluid, there is a force exerted on
the object by the fluid. This force can be resolved into two components:
- Drag, parallel to the flowing fluid
- Lift, perpendicular to the flowing fluid.

To characterize the drag on the object, it is convenient to define a
dimensionless drag coefficient,

, by:

1
2

2

where

is the projected cross-sectional area of the object. Here the


projection is usually done onto the plane normal to the oncoming velocity,
. There is a special case where this does not work, where the object is
very thin and is aligned with . Then the projected area is zero (or very
close to it) and we write:

1
2

2


V
FL, lift force
FD,
drag force
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 19 of 34
where is the surface area of the object and the coefficient

is the skin
friction coefficient that we are familiar with from the section on boundary
layers.
Example: A peregrine falcon of mass 8 kg begins a dive at 50 km/h. If it
falls freely under gravity, how far must it fall before it is travelling at 150
km/h? Assume a frontal area of 0.05 m
2
and a constant drag coefficient
of 0.30.
Answer: Draw a free body diagram for the falcon. The forces are the
weight, mg, and the drag. We can express the drag force as

1
2

2
=
2

where the constant =

2
.

Summing forces in the downward () direction, we have



2
=

2
(
2
)



We can rearrange this to give
(
2
)

+
2

2
= 2
which has solution

2
=

+
1

2/

where
1
is a constant. We apply the initial condition =
0
at = 0 to
write:

1
=
0
2


Weight, mg
Drag, FD
x
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 20 of 34
Therefore, the velocity is given by

2
=

+ (
0
2

)
2/
(*)
Now let us plug in some numbers. Using = 1.2 kg/m
3
, we obtain:
=

2
=
0.3 0.05 1.2
2

3
= 0.009


Then / = 9720 m
2
/s
2
. Further, the starting velocity is
0
2
= 192.9
m
2
/s
2
, and the final velocity is
2
= 1736.1 m
2
/s
2
. Plugging this all into
equation (*), we solve to get = 976 m.
Drag Coefficient: In the above example, we treated the drag coefficient
as a constant. However, in general, this is not the case, and we should ask
how and why the drag coefficient changes. Since both

and

are H-
groups, they can only be a function of other H-groups and the geometry
of the body. For incompressible flow of a fully-submerged object, the only
other relevant H-group is the Reynolds number, . For example, the
drag on a cylinder for incompressible flow is shown in the graph below.
Notice that as expected, the drag coefficient is a function of Reynolds
number only, here defined in terms of the cylinder diameter.

Figure 1: Drag coefficient for circular cylinders as a function of Reynolds
number. Shaded regions indicate areas influenced by shear stress. [From W
3
R,
Figure 12.2]
1
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 21 of 34
The shape of the drag coefficient curve is quite complex, and it is useful
to examine why it has this shape. To do so, we can divide the graph up
into several different regimes, labelled I to IV. We will consider each
regime in turn.
Regime I: The Reynolds number is very small, less than one. That means
that the flow is everywhere dominated by viscous effects, as shown by
the shaded region. This is sometimes called the creeping flow or
Stokes flow regime. Because fluid inertia is relatively small, the
streamlines can follow the contour of the cylinder. All the drag is due to
viscous effects on the surface of the cylinder. This is a very important
regime for bioengineering, since we are often concerned with small
particles (proteins, cells) moving in liquids, which have small Reynolds
number.
For the case of a sphere of diameter D, there is an exact solution for the
Navier-Stokes equation in the limit of zero Reynolds numbers. From this
solution we can compute the drag as

= 3

(Stokes solution)
From this we can compute the drag coefficient as

2
=
3

2
4
2

2
= 24


or

=
24
Re


Regime II: Fluid inertia is not so small that it can be ignored any longer.
However, there is still a significant region of the flow where viscous
effects are important (shaded zones). A fluid instability grows in
magnitude until vortices start to be alternately shed from the top and
bottom of the cylinder. This creates the von Karman vortex street, which
strengthens as the Reynolds number increases. Also, as the Reynolds
number increases, the streamlines can no longer follow the contours of
the cylinder, and flow separation results.
Regime III: Viscous effects are now confined to a thin region of fluid
behind the cylinder known as the wake. The wake is unsteady, even if the
oncoming flow is steady. The front part of the cylinder is covered by a
laminar boundary layer. The edges of the wake are defined by fluid
streamlines that separate from the cylinder at about 80 from the leading
stagnation point. Why does the fluid separate from the surface of the
cylinder? This has to do with the interaction between the pressure in the
outer flow and the boundary layer. In some regions of the flow field the
pressure gradient is favourable, i.e. pushes the fluid in the flow-wise
direction, while in other regions it is unfavourable. The fluid in the free-
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 22 of 34
stream has enough momentum that it can overcome an unfavourable
pressure gradient (think of exchanging V
2
for p in Bernoullis equation).
However, this adverse pressure gradient is transmitted into the boundary
layer, where the fluid is moving more slowly. Eventually the slow-moving
fluid near the surface of the cylinder will be stopped and then even
turned backwards by this adverse pressure gradient. The only way that
this can happen is if the boundary layer detaches from the surface of the
cylinder. This is known as separation, and it can be clearly seen in the
images below.
In Regime III, most of the drag force is no longer due to viscous shear
stress on the cylinder surface. Instead, most of the drag is now form drag
or pressure drag. This comes about because there are significant losses
in the wake, which is a region of very high mixing. The pressure on the
front of the cylinder is large (the stagnation pressure), but the pressure
in the wake is low, due to the losses. This means that there is an
unbalanced pressure distribution on the cylinder, which tends to push the
cylinder in the direction of flow. This is pressure drag, and it can be
minimized by reducing the size of the wake.
Regime IV: Notice that the wake is smaller in region IV than in III. This
occurs because the boundary layer on the cylinder has changed from
laminar to turbulent. Remember that mixing in a turbulent boundary
layer is more efficient than in a laminar boundary layer. This means that
in the turbulent boundary layer, low-momentum fluid that is close to
separating can efficiently receive some momentum from adjacent,
faster-moving fluid. Therefore, the fluid in the turbulent boundary layer
can continue against an adverse pressure gradient for longer than in the
laminar case, so that separation is delayed. In the case of the cylinder
with a turbulent boundary layer, separation occurs at approximately 90-
100 from the leading stagnation point, versus 80 for the laminar
boundary layer. This makes the wake smaller, so that pressure drag is
smaller, which explains the fairly drastic drop in drag coefficient at a
Reynolds number of about 5 10
5
.
Let us return to the example with the falcon. Assuming that a relevant
length L for the falcon is 50 cm, and the kinematic viscosity of air is
1.6 10
-5
m
2
/s, then we can compute a Reynolds number based on the
starting velocity (13.89 m/s) of 4.34 10
5
. This is heading into the steeply
changing part of the drag coefficient curve, so our assumption of constant

is probably not a good one.



BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 23 of 34

BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 24 of 34
Lecture 15: Laminar and turbulent flow in conduits. Relevant pi-
groups. Moody chart.
For steady, incompressible flow in an infinitely long pipe of circular cross-
section and diameter D, we previously derived the Hagen-Poiseuille law
=

4
128


No pipe is infinitely long, but in practice we can use the above formula if
we are sufficiently far from the entrance of the pipe. In such cases the
flow is said to be fully developed; the part of the flow that is close to the
entrance (or other source of flow disturbance) is said to be developing.
It is convenient to rewrite the above expression in terms of the average
velocity = 4/
2
to get
=

2
32

(*)
Now recall from our example in the dimensional analysis section that
there are three relevant H-groups when considering fully developed,
steady, incompressible flow in a pipe, namely c/,

and . Recall
further that the friction factor, , is given by
=

1
2


Darcy-Weisbach equation
To be more explicit, is known as the Darcy friction factor; shortly we
will see that there is another (closely related) friction factor that is in
common use. Because there are three H-groups, we can write
= (/,

)
For Poiseuille flow we can explicitly compute the form of the above
function by direct combination of equation (*) and the definition of the
Darcy friction factor, to obtain:
= 64

=
64


Laminar flow
Notes:
- The above formula shows that in laminar flow, friction factor is
independent of pipe roughness and varies inversely with Reynolds
number.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 25 of 34
- The transition between laminar and turbulent pipe flow is complex.
It primarily depends on the value of

. A good rule of thumb is that


the flow is laminar for

2300. For 2300

4000 the
flow is transitional, and for

4000, it is turbulent.
- Unfortunately, there are two different friction factors that are in use,
the Darcy friction factor (already introduced), and the Fanning
friction factor, which we will denote by

. These two friction factors


are related to one another by = 4

. The Darcy friction factor is


used more commonly. Unfortunately, the Fanning friction factor is
used in many chemical engineering texts, including the Welty et al.
book. We will use the Darcy friction factor, so beware that the Welty
book has a different friction factor than we use here!
Non-Circular Conduits: The above material is only true for flow in pipes of
circular cross-section. What do we do is the conduit has a non-circular
cross-section? For laminar flow there are exact results for common cross-
sectional cases, but this is not true for turbulent flow. As an
approximation in laminar flow (and as a necessity in turbulent flow), we
can use a concept known as hydraulic diameter,

is defined in
terms of the fluid-filled cross-sectional area of the pipe and the wetted
perimeter, which is the perimeter of the pipe wall touched (wetted) by
the flowing fluid, as shown below.

Example: What is

for a fully-filled circular pipe of diameter ?


Answer: For a fully-filled pipe, the area = t
2
/4 and the wetted
perimeter is t. Therefore, the hydraulic diameter is

=
4

=
Note that

can be defined for a pipe of any cross-sectional shape. Since


it reduces to the diameter for a circular pipe, it suggests that we can
rewrite the Darcy-Weisbach equation using

instead of :
Area, A
Wetted
perimeter, P

=
4


I
n
d
e
p
e
n
d
e
n
t

s
t
u
d
e
n
t

r
e
a
d
i
n
g

BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 26 of 34
=

1
2


Darcy-Weisbach equation for non-circular conduits
(Note that in the above formula, = /, just as before.) Hagen-
Poiseuille becomes:
=
64


Example: Oil ( = 900 kg/m
3
; v = 1.4 10
-4
m
2
/s) flows in a square channel
1 cm 1 cm. If the flow rate is 3 litres/min. what pressure drop will result
over 22 m of channel?
Answer: The first step is to compute the hydraulic diameter.

Next we compute the velocity in the duct:
= 3

min
10
3

min
60
= 5 10
5

3
/
=

=
5 10
5

3
/
1 10
4

2
= 0.5 /
Re

=
0.5 0.01
1.4 10
4
= 35.71

This means that the flow is laminar. We can therefore compute the Darcy
friction factor as:
=
64

= 1.79
Finally, the pressure drop is obtained from the Darcy-Weisbach equation
as:
=
1
2


= (1.79) (
22
0.01
)
1
2
(900)(0.5)
2

2

= 4.435 10
5

= 64.33
0.01 m
P = 0.04 m
A = 1 10
-4
m
2

=
4

=
4 10
4

2
4 10
2

= 0.01
I
n
d
e
p
e
n
d
e
n
t

s
t
u
d
e
n
t

r
e
a
d
i
n
g

BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 27 of 34
Turbulent flows: Most engineering flows are turbulent, and therefore the
Hagen-Poiseuille law no longer holds. Unfortunately, there are no
analytical results for turbulent flow as is the case for laminar flow.
However, there is a large body of empirical and semi-empirical data from
which working relationships can be derived. Without going through the
details, the bottom line is that in turbulent flow, the Darcy friction factor
can be expressed as
1

= 2.0log
10
{
/
3.7
+
2.51
Re

}
Colebrook formula (Darcy friction factor for turbulent pipe flow)
Notes:
- The above expression is implicit for the friction factor, and there
messy to use in practice. A plot of the Colebrook formula is called the
Moody chart and is frequently used in practice (see next page).
- Notice that the friction factor depends on both c/ and

for
turbulent flow. c is a measure of the roughness of the pipe wall, and
depends on the material that the pipe is made from. Values of c are
tabulated in standard texts (see below).

[From White, 4
th
Edition].
- If

is large enough, the second term in the Colebrook equation


vanishes and the friction factor depends only on c/, not on

.
This is called the fully turbulent regime and corresponds to the flat
part of the curves on the Moody chart.
- If c/ is small enough, the first term in the Colebrook equation goes
away and the friction factor is independent of c/. In this case the
pipe is said to be hydraulically smooth.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 28 of 34

Moody chart for estimating the Darcy friction factor as a function of Reynolds number and pipe relative
roughness. From http://blogs.mathworks.com/pick/2011/12/02/will-my-flow-be-turbulent/
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 29 of 34
Example: A 800-foot long, 6-inch diameter commercial steel pipe carries
water ( = 62.4 lbm/ft
3
; v = 1.2 10
-5
ft
2
/s). The pressure drop along the
pipe is known to be 2 psi. What is the corresponding flow rate?
Answer: We can look up c = 0.00015 feet for commercial steel in the table
below. Then the relative roughness is

=
0.00015
0.5
= 0.0003
At this point we would like to use the Darcy-Weisbach equation
=
1
2


but we dont know the friction factor or the velocity V, so the best we can
do is to get a relationship between the two of them. Substituting
numerical values into Darcy-Weisbach gives:

2
=
2


=
2 [(2

2
) (
144
2

2
) (
32.2

2
)] (0.5 )
(62.4

3
) (800 )

= 0.1858

2
(*)
At this point the best way to proceed is to guess a value of , plug it into
the above equation to get , the compute

and use the Moody chart


to get a new value of . A useful trick in this process is to start the
guessing for in the fully turbulent regime (the flat part of the curve on
the Moody chart). In this problem, that means a starting guess of f = 0.015.
Friction factor, from equation (*) [ft/s]


0.015 3.52 1.466 10
5

0.019 3.13 1.30 10
5

0.0195 3.09 1.29 10
5

The iteration has converged with = 3.09 /. We can then compute
the flow rate as:
=

2
=

4
(3.09

) (0.25
2
) = 0.607


Lecture 16: Generalized Bernoulli equation (Mechanical Energy
equation). Examples: flow in pipe systems.
Generalized Bernoulli equation (Mechanical Energy equation): To date,
we have only been able to consider simple pipe systems, consisting only
of a single pipe of constant diameter with no elevation differences
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 30 of 34
between the two ends. Now we will show how more general problems
can be tackled. The starting point is the Bernoulli equation:

+
1
+

1
2
2
=

+
2
+

2
2
2


It is convenient to define the total head, =

+ +

2
2
, in which case
we can rewrite Bernoullis equation as h1 = h2. As previously noted, the
total head is a measure of the useful energy available in the fluid to do
work.
The main difficulty with the Bernoulli equation for most practical
situations is that it does not include losses (friction). Friction will act to
decrease the total potential for driving fluid flow, i.e. total head. If there
are losses between and we need to modify Bernoullis equation. It
seems reasonable to write:

1
=
2
+

(*)
where

represents the head lost between and due to friction. This


is known as the Mechanical Energy equation, or Generalized Bernoulli
equation. Since head is proportional to

, we see that the head loss term


will be equal to

for a simple pipe system (pipe of constant diameter


with no elevation change between and ). This means that we can
generalize the Darcy-Weisbach equation to:

2
2

In this equation, we should interpret

as the head lost due to friction


between the flowing fluid and the pipe walls. We can use this in
conjunction with equation (*) to solve more complex problems involving
flow in conduits.


Valid for lossless, steady,
incompressible flow
along a streamline.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 31 of 34
Example: In the diagram below, what is if is known to be 1 m
3
/s?

Answer: From the given data we can directly compute:

=
0.02
20
= 0.001
=

=
1
(0.1)
2

3

2
= 31.83 /
Re

=
(31.83 /)(0.2 )
10
6

2
/
= 6.37 10
6

Now we use the Moody chart to directly look up f = 0.0195, and the use
the Darcy-Weisbach equation to get

2
2
= 0.0195
100
0.2
(31.83)
2
2(9.81)

= 503.5
This is the head loss due to friction in the pipe, but what does this say
about ? To answer this question we need to use the Mechanical Energy
equation. The first step is to define points and as shown in the figure.
Then we notice that
1
=
2
= 0,
1
~ 0, and
1

2
= , so that we can
write the head difference as
1

2
=

2
2
2
. Then the Mechanical
Energy equation reduces to
=

2
2
2
= 503.5 +
(31.83 /)
2
2(9.81 /)
= 566.7
Notice that the height is being used to overcome friction in the pipe
and also to give the exiting fluid kinetic energy.
Minor Losses: There are other mechanisms to lose head besides pipe wall
friction. One of the most common sources of head loss in piping systems
is the presence of valves, fittings, bends and other devices that locally
disturb flow. Ultimately the head lost due to these items is due to pipe
wall friction, but the losses tend to be quite intense and quite localized,
and it is therefore convenient to express these losses in a different way
that pipe wall friction for straight pipes. We write such losses as

2
2

Water
= 10
3
kg/m
3

v = 1 10
-6
m
2
/s

D = 20 cm
L = 100 m
c = 0.02 cm
H
Q


BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 32 of 34
where is known as the minor loss coefficient. Values of can be
looked up in reference books. A partial listing is shown in the table below.

Resistance coefficients =

/ [

] for open valves, elbows and tees. [Table


6.5 from White, 5
th
edition.]
Pumps and Turbines: Head can be added to the fluid by a pump. We
denote the total amount of head added by a pump as

.
Aside: Related to this is the issue of how
much power is being added to the fluid by a
pump. This is given by =


Head can also be removed from the fluid by a turbine. The amount
removed is denoted by

.
Considering all of the above factors, we can write a more complete form
of the generalized Bernoulli equation as:

+
1
+

1
2
2
=

+
2
+

2
2
2
+


Mechanical energy equation
where we have included the possibility of having multiple pipes, minor
losses, turbines and pumps by explicitly showing summation signs. Note
that the head added by the pump must be subtracted from the right hand
side, since it is acting to increase total head, which is opposite to all the
loss terms.
Example: A pump delivers fluid between two reservoirs as shown. If the
pump adds 55 meters of head, compute the resulting flow rate.
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 33 of 34

There are four sites of minor loss:
- Entrance to the pipe from the supply reservoir, = 0.5
- Expansion after the pump, = 0.6 (based on the velocity in the
small pipe)
- Valve, = 7.5
- Expansion into the receiving reservoir, = 1.0
Answer: Label and as shown. Apply the generalized Bernoulli
equation between these points, noting that:
-
1
=
2
=0
-
1

2
= 25
-
1
~ 0
-
2
~ 0
The head difference is therefore
1

2
= 25 . The generalized
Bernoulli equation reduces to:
25 =

2
2
+

2
2
+ 0.5

2
2
+ 0.6

2
2
+ 7.5

2
2
+1.0

2
2
55
where subscript A refers to values in the smaller pipe and subscript B
refers to values in the larger pipe.
By using continuity we can write:

2
=

(0.2)
2
=

(0.4)
2

= 4


Therefore the Mechanical Energy equation becomes
Water
= 10
3
kg/m
3

v = 1 10
-6
m
2
/s

D = 20 cm
L = 300 m
c = 0.05 mm


P
D = 40 cm
L = 600 m
c = 0.3 mm
25 m
BMED3300 Biotransport Lecture Notes C. Ross Ethier 2004, 2013
Page 34 of 34
30 =

2
2
{

16 +

+16 1.1 +8.5}


=

2
2
{2.4 10
4

+1.5 10
3

+26.1}
To use this equation we need to be able to get friction factors from the
Moody chart, which requires us to compute relative roughnesses for both
pipe A and pipe B.
Pipe A:

= 0.00025
Pipe B:

= 0.00075

At this point we must iterate. We guess fA and fB (starting in the fully
turbulent regime), and then use the Mechanical Energy equation to
compute

. Then we get

from the continuity equation get the


Reynolds numbers in pipes A and B, and then use the Moody chart to
update

and

. It is convenient to summarize this in a table as shown


below.
fA fB VB (m/s) ReD,A ReD,B
0.0145 0.0185 1.21 9.72 10
5
4.84 10
5

0.0150 0.0190 1.18
This has converged, so that we can now easily compute Q as
=

2
= 0.149
3
/
As a check, it is interesting to compute the head loss across each
component in the piping system. This is shown in the table below. It will
be noticed that the sum of all the head losses equals the net head
available (within numerical round-off). This table clearly shows that
essentially all of the head loss occurs in the first pipe A, and therefore if
it is desired to increase the flow rate, the first thing to do would be to
replace pipe A.
Component Head loss
Pipe entrance 0.5

2
/2 = 0.568
Expansion after pump 0.6

2
/2 = 0.681
Valve 7.5

2
/2 = 0.532
Expansion 1.0

2
/2 = 0.071
Pipe A

2
/2 = 25.89
Pipe B

2
/2 = 2.04
Total 29.78 m

Anda mungkin juga menyukai