Você está na página 1de 99

C URTIN U NIVERSITY OF T ECHNOLOGY

Fluid Mechanics 433


Lecture Notes

Department of Mechanical Engineering


Andrew King
October 2007 (Part Revised July 2014).
(non-Newtonian flow section based on Lecture
Notes originally compiled by Dr. David Devenish,
Curtin University)

Contents
Contents

ii

.
.
.
.
.

1
1
1
2
4
5

.
.
.
.
.
.
.
.

7
7
7
8
8
10
10
11
12

.
.
.
.

14
14
15
16
16

Free Settling - Spherical Particles


The Stokes Equation . . . . . . . . . . . . . . . . . . . . . . . .
Drag Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . .

18
18
19

Free Settling Non-Spherical Particles


Shape factor . . . . . . . . . . . . . . .
Generalized Curves . . . . . . . . . . .
Fall diameter . . . . . . . . . . . . . . .
Drag coefficient . . . . . . . . . . . . .
The effect of turbulence on fluid drag
Effect of concentration on drag . . . .

.
.
.
.
.
.

22
22
23
24
24
27
28

Classification of Particles
Free Settling . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hindered Settling . . . . . . . . . . . . . . . . . . . . . . . . . .

29
29
31

Background Concepts
Bernoulli Equation . . .
Viscosity . . . . . . . . .
Pressure Drop in pipes .
Turbulence . . . . . . . .
Pressure Drop in Fittings

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Pumps
Types of Pumps . . . . . . . . . . . . . . . .
Positive Displacement Pumps . . . .
Rotodynamic Pumps . . . . . . . . .
Centrifugal Pumps . . . . . . . . . . . . . .
Pump Curves . . . . . . . . . . . . . . . . .
Power consumption and Efficiency .
Net Positive Suction Head . . . . . .
Pump Similarity Laws . . . . . . . . . . . .
Flow in pipe networks
Simple systems of multiple pipes . . .
Other systems . . . . . . . . . . . . . .
Other flow devices . . . . . . . . . . .
1-Dimensional pipe network software

.
.
.
.

.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.

.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.

ii

Solid Liquid Separation


Dewatering processes . . .
Sedimentation Separators
Continuous Sedimentation
Thickener Sizing . . . . . .

.
.
.
.

32
32
33
33
35

Pressure drop through packed beds


Kozenys hydraulic model . . . . . . . . . . . . . . . . . . . . .
The Carman-Kozeny equation . . . . . . . . . . . . . . . . . . .
Specific surface of a packed bed . . . . . . . . . . . . . . . . . .

37
38
40
42

Fluidization

43

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

10 Non-Newtonian Fluids
Rheology . . . . . . . . . . . . . . . . . .
Newtonian Fluids . . . . . . . . . . . . .
Non-Newtonian Fluids . . . . . . . . . .
Time Independent Fluids . . . .
Time dependent fluid behaviour

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.
.

45
45
45
46
46
49

.
.
.
.

51
51
53
54
55

.
.
.
.
.
.
.
.
.
.
.

57
57
57
59
59
60
60
60
60
61
61
62

13 Rheometry
Tube type viscometers . . . . . . . . . . . . . . . . . . . . . . .
Coaxial cylinder viscometer . . . . . . . . . . . . . . . . . . . .

64
65
66

14 Turbulent Boundary Layers


Boundary Layer Velocity Profile . . . . . . . . . . . . . . . . . .
Application to pipe flow . . . . . . . . . . . . . . . . . . . . . .

69
69
71

15 Scale up methods for homogeneous flow


Laminar flow scale-up . . . . . . . . . . . . . . . . . . . . . . .
Turbulent flow scale-up . . . . . . . . . . . . . . . . . . . . . .
Transition to turbulence . . . . . . . . . . . . . . . . . . . . . .

73
74
74
75

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

11 A Generalized Method for Flow in Pipes


Solution for Laminar Flow . . . . . . . . . . . .
Generalized method for laminar flow . . . . .
Turbulent Flow and Dimensional Analysis . .
Generalized method applied to Turbulent Flow
12 Physics of Rheology
Solid Suspensions . . . . . . . . . . . . . .
Forces acting on particles . . . . . . . . . .
Rheograms of suspensions . . . . . . . . .
Shear thinning suspensions . . . .
Shear thickening suspensions . . .
Time dependent effects . . . . . .
Irreversible phenomena . . . . . .
Viscosity . . . . . . . . . . . . . . . . . . .
Dilute suspensions . . . . . . . . .
Concentrated suspensions . . . . .
Influence of particle-particle interactions .

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

iii

16 Pipe Flow of Heterogeneous Slurries


Distribution of Particles in Turbulent Flow . . . .
The Durand Method . . . . . . . . . . . . . . . . .
Pressure losses in the carrier fluid . . . . .
Pressure losses due to the particle phase . .
The Durand-Wasp Method . . . . . . . . . . . . . .
Application of the Durand-Wasp method . . . . .
Initial assumptions . . . . . . . . . . . . . .
Iteration . . . . . . . . . . . . . . . . . . . .
Pressure loss for the system . . . . . . . . .
Changes for multiple particle size fractions
Pressure loss versus flowrate . . . . . . . .
Deposition Velocity . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

77
77
78
78
79
79
80
80
81
82
82
83
84

17 Wear in Slurry Pipelines


Corrosion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Abrasion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85
85
87

18 Centrifugal Slurry Pumps


Effects of solids on pump performance . . . . . . . . . . . . . .

90
90

References

94

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

iv

1 Background Concepts
Fluid Mechanics is a very broad area of engineering, covering many aspects
of fluids, flowing or otherwise. This unit is designed to be relatively practical,
emphasising pipe flow (and pipe systems), pumps and non-Newtonian fluids (such as slurries). This chapter is an overview of the important concepts
which will be built upon in the rest of the unit.

Bernoulli Equation
The Bernoulli equation is one of the fundamental equations of fluid mechanics. It can be thought of as a simplified version of the energy equation, and is
given as
v2
P + + gz = k
(1.1)
2
where P is the static pressure, v is the flow velocity, z is the height (relative
to some reference point) and k is a constant. This equation is derived by
applying conservation of energy principals to a fluid particle as it moves in
a flow. The more common form is obtained by applying the above equation
between two points, i.e.
P1 +
or, in terms of head,

v2
v21
+ gz1 = P2 + 2 + gz2
2
2

P2 v2
P1 v21
+ + z1 = + 2 + z2

2g

2g

(1.2)

(1.3)

In each of these equations the first term, P, is the called the Pressure head,
the second term, v2 /2, is called the velocity head, and the last term , gz, is
called the static head. The Bernoulli equation can be applied to one dimensional flows, or between two points along a single stream line in two or three
dimensional flow. The limiting assumptions of equations 1.2 and 1.3 above,
are that the flow is steady and that the flow is frictionless. It is generally the
second assumption which causes the most difficulty for flow in pipes and
pipe networks.

Viscosity
While it is convenient to ignore frictional forces in fluids, in general (and
especially in pipe flows) fluid friction is too great to be ignored. The frictional
(shear) forces acting on a fluid are dependent on the viscosity of the fluid in
question. To illustrate the concept of viscosity consider Figure 1.1.
This figure shows the flow occurring between two parallel plates, of which
one is moving at a velocity , and the other remains stationary. In this case the
shear stress is constant across the gap and is proportional to the velocity gradient, i.e.
1

Figure 1.1: Flow between two parallel plates

u
y

(1.4)

or alternatively
=

du
dy

(1.5)

The constant of proportionality, , is called the viscosity (or more correctly the dynamic viscosity1 ). For most fluids the viscosity can be considered
constant with respect to shear rate. These fluids are called Newtonian fluids. Fluids for which this is not the case (for example slurries) are called
non-Newtonian. Many subclasses of non-Newtonian fluids exist, and some of
these will be examined later in the unit.

Pressure Drop in pipes


Pipes and pipe networks make up a very large segment of mechanical engineering, and for design purposes the chief consideration is usually the pressure drop along a pipe (or network). The pressure drop in combination with
the flowrate is what determines the selection of pumps, pipe sizes and other
requirements.
For laminar flow, an expression for the pressure drop through a segment
of pipe is relatively easy to develop, since a pipe is essentially a wrapped up
plate. This expression will be developed now, and will also be used later in
the unit when looking at non-Newtonian fluids.
Consider the force balance on a small volume of fluid in a pipe as shown
in Figure 1.2.

Figure 1.2: Flow in a section of pipe


1 As

compared with the kinematic viscosity, = /

For a steadily flowing fluid, the two forces indicated (the pressure force
and the wall shear force) must be equal, ie
w cdx = dp A

(1.6)

where c is the circumference. For circular pipes of course, c = D, and


A = D2 /4. Taking a section of pipe of length L, and simplifying gives
L
D
In addition, for Newtonian fluids, the shear stress is given by:
 
du
w =
dr w
P = 4w

(1.7)

(1.8)

which requires du/dr. This can be determined from the velocity profile
for the flow through the pipe. It can be shown that the velocity profile for
Newtonian laminar flow in a pipe is
"
 2 #
2r
(1.9)
u = 2V 1
D
differentiating and evaluating at the wall (r = D/2) gives,
 
8V
du
=
dr w
D

(1.10)

where V is the bulk velocity of the flow through the pipe. Substituting
this back into Equation 1.7 results in
V L
(1.11)
D2
Equation 1.11 allows the pressure drop to be determined for laminar flow
only. To obtain a more useful form this equation is non-dimensionalised by
dividing both sides by V 2 /2, giving
p = 32

32V L
p
=
2
V /2
V 2 D2 /2

 

L
= 64
V D
D
 
64 L
=
Re D

(1.12)
(1.13)
(1.14)

where Re is the Reynolds number. A new dimensionless quantity, the


friction factor, f , is introduced to separate the dependence on the Reynolds
number, and a more generalised form of the pressure drop equation is obtained:
L V 2
(1.15)
D 2
where f = 64/Re, is the friction factor for laminar Newtonian flow in a
pipe.
p = f

Turbulence
While the analysis in the previous section allows the pressure drop for laminar flow to be predicted, in reality most engineering flows are turbulent.
Turbulence occurs when shear forces in the fluid are too great, and packets
of fluid roll-up on each other, creating random unsteady fluctuations in the
flow. While turbulent flow is inherently unsteady, the randomness of the fluctuations allows time-averaging to be used, and in general properties like the
bulk velocity can still be defined and used with good accuracy.
For pipe and pipe system analyses the chief effect of turbulence is to increase the observed pressure drop. While increasing the pressure drop is generally undesirable, turbulent flow can sometimes be beneficial, for example
in heat transfer applications, or for slurry flows. In any case due to economic,
and practical limitations, turbulent flow is the rule rather than the exception.
For pipe flows, laminar flow is observed below Reynolds numbers of ~2100,
while above Reynolds numbers of ~4000 turbulent flow occurs. The in between range is known as the transition region.
For calculating the pressure drop in turbulent flow, Equation 1.15 is sufficient, providing that a suitable value for f can be found. For laminar flow f is
simply 64/Re as developed previously, but for turbulent flow f is dependent
on both the Reynolds number and the relative roughness of the pipe wall,
e/D. The values of f for turbulent flow can be found from the Colebrook
equation:


2.51
/D
1
= 2.0 log
+
(1.16)
3.7 Re f
f
Usually though, the Colebrook equation is presented graphically as the
Moody Diagram, shown in Figure 1.3.
Moody Diagram
0.1
0.09
0.08
0.05

0.07

0.04
0.06

0.03

0.05

0.02
0.015

0.04

0.03
f

0.004

/D

0.01
0.008
0.006

0.035

0.025
0.002
0.02

0.001
0.0008
0.0006
0.0004

0.015

0.0002
0.0001
0.00005

0.01

0.008

0.00001
103

104

105

106
Reynolds Number

107

108

109

Figure 1.3: Moody Diagram

For smooth pipes, a simplified version of the Colebrook equation can be


used
4

f = 0.184Re0.2

(1.17)

Pressure Drop in Fittings


The above equations are for determining pressure losses for sections of pipe
only. For long pipelines, where the pipe itself contributes most of the frictional losses, calculating the pressure drop using the above methods is usually sufficient. In most practical cases though losses due to fittings (such as
valves and reducers) and bends are not insignificant.
The most common method for allowing for the pressure losses in fittings
is by defining a loss coefficient. The loss coefficient relates the pressure drop
through the fitting to the dynamic pressure, V 2 /2, as in the following equation
V 2
(1.18)
2
Strictly speaking, KL is dependent on both the geometry of the component and the Reynolds number, though at typical flow speeds the Reynolds
number dependence can usually be neglected. Loss coefficients for components are usually presented in charts or tables, after being determined either
through experiments or, more recently, using computational fluid dynamics.
Figure 1.4 shows a cutaway of a gate valve and a typical loss coefficients at
various degrees of closure.
p = KL

Loss coefficient for typical gate valve


100

KL

10

0.1
0

20

40

60

80

100

% closed

Figure 1.4: Gate Valve, and loss coefficient

2 Pumps
Pumps are a type of fluid machine1 that add energy to a fluid. Energy needs
to be added to a fluid to overcome losses in pipework, changes in elevation,
or to increase the momentum of a fluid. Pumps are a very important piece of
mechanical plant, and it is estimated that approximately 20% of the worlds
electricity supply is consumed by motors that drive pumps [1]. For this reason correct pump selection is an important consideration, and is the chief
focus of this part of the unit.

Types of Pumps
When discussing pumps, there is generally two main classes either positive
displacement pumps or rotodynamic pumps. Other types of pump exist (such
as jet pumps), though most pumps used in industry are one of the above two
types and it is these that will be discussed in the following sections.

Positive Displacement Pumps


Positive displacement pumps use changes in volume in combination with
valves to move the fluid through the pump. Familiar examples of positive
displacement pumps include bicycle pumps, some car oil pumps and the
heart. Though in industry, it is more common to see pumps like that shown
in Figure 2.1.

Figure 2.1: Positive displacement pump[2]

Another common type of Positive displacement pump is a peristaltic pump,


as shown in Figure 2.2. Peristaltic pumps provide very accurate flowrates,
and for this reason are commonly encounted in process industries, where accurate control of reagents is required, and the medical industry.
1 which

includes compressors and turbines

Figure 2.2: Peristaltic pump[3]

Rotodynamic Pumps
Rotodynamic pumps differ from positive displacement pumps, the rotation
of a series of blades around an axis causes dynamic effects in the fluid that
add energy. Generally rotodynamic pumps are further separated into centrifugal pumps, mixed-flow pumps, or axial flow pumps. In reality there is
a smooth transition when moving from each of these sub groups to the next.
Figure 2.3 shows a typical centrifugal pump, identifying each of the major
components.

Figure 2.3: Centrifugal pumps, Schematic (left) and photo


(right)[4]

As centrifugal pumps are such an important device, a more in depth look


at their operation is in order. This is conducted in the next section.

Centrifugal Pumps
To start the analysis of a typical centrifugal pump, consider the simplified
impeller as shown in Figure 2.4
This figure considers the flow between two vanes of a pump impeller rotating at an angular velocity of . U is the absolute velocity of the impeller
at its outside edge, W is the velocity of the exiting fluid relative to the impeller and V is the absolute velocity of the fluid. Also, it is assumed that the
fluid enters the impeller through the eye with negligible tangential velocity
and that the flow exits the impeller at an angle to the outer circumference
of the impeller. Both these assumptions are reasonable for a pump running
at its design point.
8

Figure 2.4: Centrifugal pump impeller

Given these assumptions, the following expressions for velocities at the


exit can be written
U = R
V = U W cos
Vr = W sin

(2.1)
(2.2)
(2.3)

Taking the moment of momentum allows the shaft torque, Tw to be calculated


Tshaft = m(RV

(2.4)
)
or
Tshaft = Q(RV )

(2.5)

Power is then determined from


W shaft = Tshaft
W shaft = Q(RV )

(2.6)
(2.7)

also from Equation 2.1 changes the above to


W shaft = Q(UV )

(2.8)

If all of the work supplied to the shaft, then an ideal head rise can be
determined from
W shaft = QgH
(2.9)
Combining equations 2.8 and 2.9 gives
H=
From Figure 2.4
cot =

UV
g

(2.10)

U V
Vr

(2.11)

and Equation 2.10 can be rewritten as


H=

U 2 UVr cot

g
g

(2.12)
9

The flowrate, Q, can also be related to the radial exit velocity via
Q = 2RbVr

(2.13)

where b is the depth of the vane passage. This results in the ideal head rise of
a centrifugal pump being given by
H=

U 2 U cot

Q
g
2Rbg

(2.14)

a linear variation of H, with respect to the flowrate.


A real pump has additional losses which decrease the actual achieved
head. These are due to friction, leakages, eddies, amongst others. The resulting head curve for a centrifugal pump is shown in Figure 2.5.

Figure 2.5: Centrifugal pump curve

Pump Curves
As seen in the last section, in essence a pump curve is a plot of the actual
output head, H, of a pump versus the flowrate, Q. These curves are obtained
experimentally and are only for a specific pump running at a specific speed,
though it is possible to determine the characteristics for geometrically similar
pumps running at different speeds (covered in the next section).
For design purposes additional information is also included on the curve.
This additional information is usually curves for power consumption, efficiency, different impellers, and also the required Net Positive Suction Head
(NPSH). Figure 2.6 shows a typical pump curve as supplied by a pump manufacturer.

Power consumption and Efficiency


It is possible to determine the additional energy that is added by a pump to
a fluid. This is given by
Wfluid = QP
(2.15)
or (in terms of head)
Wfluid = QgH

(2.16)
10

Figure 2.6: Typical manufacturers pump curve, showing efficiency and NPSHR

where P (or H) is the pressure rise across the pump. It is necessary to supply
more than this amount of work to the shaft due to mechanical and hydraulic
losses in the pump. These losses are mostly dependent on the geometry of
the pump and thus need to be determined experimentally. As the consumed
power of a pump is an important piece of information to have when specifying a pump this information is included as a series of curves indicating the
efficiency of the pump at each flowrate, where
=

Wfluid
gHQ
=
Wshaft
Wshaft

(2.17)

This can be used to determine the shaft work at the operating point.
In addition to curves indicating the efficiency pump curves also indicate
the Best Efficiency Point (BEP). It is desirable to keep the operating point as
close to this point as possible.

Net Positive Suction Head


The Net Positive Suction Head (NPSH) is the positive pressure at the suction
side of the pump. Each pump also has a required NPSH, or NPSHR , which
is the suction side pressure required to ensure that cavitation does not occur
within the pump. Cavitation occurs when the absolute pressure in the pump
drops below the vapour pressure of the fluid being pumped2 . This causes
bubbles of vapour to form in the fluid, which rapidly collapse when the pressure increases above the vapour pressure again. The collapse of these bubbles
near to the pump casing or impeller causes undesirable wear, and can lead to
2 While

the pressure at the inlet of the pump may be high, there is only a certain amount
of energy that is available in the fluid. From Bernoullis equation, P + V 2 /2 = constant (neglecting elevation), meaning that anywhere that the velocity of the fluid in the pump is high,
the absolute pressure can be much lower than the pressure at the suction side.

11

early failure of the pump. Also, the inclusion of vapour in the flow decreases
the volumetric efficiency of the pump. For these reasons cavitation should always be avoided. Keeping the pressure at the suction side above the NPSHR
guarantees this to be the case.

Pump Similarity Laws


A useful set of expressions for pumps have been derived from dimensional
analysis. These expressions are commonly referred to as the pump similarity
laws. For dimensional analysis, the key dependent variables for a pump are:
the head rise, H, the shaft power , Wshaft , and the efficiency, . [5] These values
are expected to depend on the geometry of the pump (D, l and ), flowrate
(Q), pump speed () and the fluid properties ( and ). For dimensional
analysis this means that each of H, Wshaft and can be expressed as a function
of the independent variables,
H,Wshaft , = f (D, l, , Q, , , )

(2.18)

After non-dimensionalising each of the dependent variables, the following expressions are obtained for CH (dimensionless head increase), CP (dimensionless shaft power) and (efficiency)3


l Q D2
gH
, ,
(2.19)
,
CH = 2 2 = 1
D
D D D3



Wshaft
l Q D2
CP =
= 2
, ,
(2.20)
,
D D D3

3 D5


gQH
l Q D2
, ,
(2.21)
=
= 3
,
Wshaft
D D D3

To be usefull these laws are generally simplified further. Firstly, the influence of the viscous forces is small if the Reynolds number is high enough
(where D2 /) is a type of Reynolds number) and therefore dependence
on this term can be ignored. Secondly, the relative roughness /D is also neglected, as the irregular shape of the pump itself has a lot larger influence on
the performance than the surface roughness. Finally, if geometrically similar
pumps are compared, then l/D is constant for the analysis. This means that
the above equations simplify to


Q
gH
= 1
(2.22)
2 D2
D3


Wshaft
Q
= 2
(2.23)
D3
3 D5


Q
= 3
(2.24)
D3
Equations 2.22, 2.23 and 2.24 are now only dependent on the flow coefficient, CQ = Q/D3 . If two pumps from the same family are operated at the
3 These

are the commonly accepted forms of the terms. Other non-dimensional forms
can also be derived but are not in common usage

12

same flow coefficient, then





Q
D3 1


gH
2 D2 1


Wshaft
3 D5 1
1


=
=
=
=


Q
D3 2


gH
2 D2 2


Wshaft
3 D5 2
2

(2.25)
(2.26)
(2.27)
(2.28)

These scaling laws can be used to predict pump performance at different flowrates, pump speeds, or for different sized but geometrically similar
pumps.

13

3 Flow in pipe networks


Section 1 revised viscous flow in pipes and fittings. In reality single line pipe
are rare and usually pipes and pumps exist in a network. This section looks
at the methods used to determine flowrates and pressures in a network of
pipes, pumps and tanks.

Simple systems of multiple pipes


All pipe systems are governed by the same equations as those introduced
in Section 1. To start the analysis the simplest systems of multiple pipes are
examined, a group of pipes either in series or parallel. These are shown in
Figure 3.1.

(a) Series network


1

(b) Parallel network

Figure 3.1: Simple multiple pipe systems

For a fluid circuit there a balance is obtained between the pressure drop,

P, the flowrate, Q, and the flow resistance. R,


P = Q2 R

(3.1)

This is analogous to electrical systems where the voltage (pressure), e, is


related to the current (flow), i, and the circuit resistance, R by the expression e = iR. This similarity means that the techniques to solve multiple pipe
systems are similar to those used to solve electrical circuits, though nonlinearities due to Q2 mean not all the methods used for electrical circuits can
be used for fluid circuits.
The simplest multiple pipe system is that of pipes connected in series,
as shown in Figure 3.1(a). These systems have been discussed in Section 1.
When considering series pipes in a network the governing equations are
Q1 = Q2 = Q3

(3.2)

PLAB = PL1 + PL2 + PL3

(3.3)

The other of the simple multiple pipe systems is that where pipes are connected in parallel, as shown in Figure 3.1(b). In this case the governing equations become
Q = Q1 + Q2 + Q3
(3.4)
14

(3.5)

PLAB = PL1 = PL2 = PL3

Other systems
The rules described in the previous section form the basis for analysing more
complicated networks of pipes. A simple example of such a system is the
three-reservoir problem as shown in Figure 3.2.

Figure 3.2: Three reservoirs connected at a common node

This system consists of three reservoirs connected by pipes which meet at


a common junction. In this system we can also write additional governing
equations. Firstly, from continuity,
Q1 + Q2 + Q3 = 0

(3.6)

if we take all flows as positive towards the junction. Note that this implies
one or two of the flows are away from the junction as expected. Also, as for
the parallel pipes above, the pressure drop along each of the three pipes must
be such that the resulting static pressure at the junction is the same in each
case. In equation form
L1 V12
= gz1 PJ
D1 2
L2 V22
= f
= gz2 PJ
D2 2
L3 V32
= f
= gz3 PJ
D3 2

P1 = f

(3.7)

P2

(3.8)

P3

(3.9)

This system is fully defined, it has 4 equations and 4 unknowns, though


due to the dependence of f on the pipe velocities iteration is necessary to
obtain a solution.
The next step is to extend the rules in Section to arbitrary networks, such
as that shown in Figure 3.3. For a system such as this, for example a water
distribution network, the rules for determining the flowrates through pipes
and pressures at the junctions can be summarised as
The net flow into any junction must be zero; and
The net head loss around any closed loop must be equal to zero
15

Figure 3.3: Large network of pipes

Applying these rules to each junction and loop in the network a system of
equations is obtained and can be solved by iteration to find the flowrates and
pressures in the system.

Other flow devices


So far the analysis has only considered pipes. A more general approach is
needs to also consider flow devices; such as pumps, valves and fittings to
name a few. This can be achieved by thinking of about the frictional losses
in a pipe. The friction factor and pipe loss equation essentially describes a
relationship between pressure and flowrate for a pipe, ie.
P = f (Q)

(3.10)

By allowing this relationship to be arbitrary the rules described in Section can be applied to any flow device for which this relationship is known.
These components can either take energy out of the system (loss through a
valve) or add energy to the system (a pump). In the first case the pressure
decreases in the direction of the flow, while for the second the pressure increases. In any case, a system of simultaneous equations is obtained, which
can be solved.

1-Dimensional pipe network software


The system of equations that results from analysing a pipe network is highly
non-linear, and generally must be solved using iterative techniques. For this
reason computers are usually the only practical way of solving large flow networks. Many different software packages are available both commercial and
free that can solve these networks of pipes. In the software a schematic of the
system is constructed and appropriate properties assigned to the items that
join the nodes; pumps, valves, pipes, et cetera. Most commercially available
16

packages have an in-built catalogue of components, which can be added to


as needed. More advanced packages account for unsteady flow, and in some
cases also include flow control devices, allowing a full system to be analysed
in real-time.

17

4 Free Settling - Spherical Particles


This is a free motion when the particle is falling under the force of gravity and
enjoys complete freedom in its motion as it passes through the fluid. Motion
is said to be hindered when a number of particles interfere with each other,
thereby reducing the settling rate.

The Stokes Equation


Consider a solid particle falling from rest in a stationary fluid under the action
of gravity. As frictional force increases with velocity it will eventually reach
a condition where the drag force is equal to the lift force and the particle will
fall at a constant velocity. This velocity attained at the end of the acceleration
period is called the terminal settling velocity.
Figure 4.1 shows a solid particle of density p falling in a stationary fluid
of density .

Figure 4.1: Particle falling freely in an infinite fluid

For a particle of diameter D p the volume and mass of the particle are
D3p
Volume =
6
D3p
Mass =
6

(4.1)
(4.2)

If the particle has reached its terminal velocity then the forces are in equilibrium and the drag force, D, can be determined from
Drag Force = Weight Force Buoyancy Force
D3p p g D3p g
D =

6
6
3
D p
D =
( p ) g
6

(4.3)
(4.4)
(4.5)

If the difference in density between the fluid and the particle is large, (such
as sand settling in a gas) the buoyancy force is small and the drag force is
approximately equal to the weight force.
18

An analysis of the drag force suggests that it is a function of the particle


size, D p and velocity, Vs ; and the fluid density, , and viscosity, .
D = f (D p ,Vs , , )

(4.6)

Dimensional analysis of these variables results in two groups, and the dimensionless drag becomes a function of the Reynolds number.
D
= f (Re)
Vs2 D2p

(4.7)

In this case, the Reynolds number is based on the particle diameter and the
settling velocity
Vs D p
(4.8)
Re =

As in most fluid problems, the Reynolds number can be used to indicate


the flow regime, in this case the pattern of motion of the particle. For spherical particles it is accepted that laminar motion occurs up to a Re value of 0.2.
A good starting point to determine the drag in this region is to assume that it
is inversely proportional to Re, as is the case for pipe flow. This means that,
D
K
=
2
2
Vs D p
Re
D = KVs2 D2p

(4.9)

Vs D p

= KVs D p

(4.10)
(4.11)

The constant K has been determined for different shaped particles from experimental results and for a sphere has been found to be equal to 3. The
drag force is then
D = 3 Vs D p
(4.12)
Equation 4.12 is known as the Stokes Equation (as is Equation 4.14 below).
Substituting this into Equation 4.5 gives
3 Vs D p =

D3p
( p ) g
6

(4.13)

Rearranging allows the terminal settling velocity to be determined,


D2p ( p ) g
Vs =
18

(4.14)

Drag Coefficient
The above analysis allows the settling velocity to be determined for particles
falling in the laminar regime. However, for particles falling in the transitional
or turbulent regimes, as well as for non-spherical particles, the drag coefficient
provides a more useful means for determining the settling velocity. The drag
coefficient is defined as
D
CD =
(4.15)
(Vs2 /2) A
19

where A is the projected area of the body (in this case the particle). For a
sphere
D2p
(4.16)
A=
4
and therefore the drag force is given by
D2p Vs2
D = CD
8

(4.17)

Substituting this into Equation 4.5 again, gives


D3p

( p ) g
CD D2p Vs2 =
8
6

(4.18)

from which

4
D p ( p ) g = CDVs2
(4.19)
3
Therefore the terminal settling velocity for an arbitrary particle in terms of
the drag coefficient can be determined from
s
4 D p ( p ) g
(4.20)
Vs =
3CD

In a similar manner to determining the drag force, it is possible to correlate


the drag coefficient with the Reynolds number,
CD = f (Re)

(4.21)

Figure 4.2 shows the variation of drag coefficient in terms of the Reynolds
number. As with most flows three regions can be identified from Figure 4.2,

Figure 4.2: Drag coefficient of a settling spherical particle in


terms of the Reynolds number

namely
20

A Laminar (or viscous) Zone, for Re < 0.2


A Transition Zone, for 0.2 < Re < 500; and
A Turbulent Zone, for Re > 500
Alternatively these are known as the Stokes, Allen or Newton zones, respectively. For the laminar zone, the drag coefficient is inversely proportional to
the Reynolds number, where
CD =

24
24
=
Re D pVs

(4.22)

If this expression is substituted into Equation 4.20, then the settling velocity
becomes
s
s
4 D pVs D p ( p ) g
(4.23)
Vs =
3 24

simplifying this gives


Vs =

D2p ( p ) g
18

(4.24)

which is the same result as Equation 4.14.


For the turbulent region, CD is approximately 0.44, for Reynolds numbers
between 500 and 105 . Substituting this into Equation 4.20 gives
s
r
D p ( p ) g
4 1
(4.25)
Vs =
3 0.44

s
D p ( p ) g
= 1.74
(4.26)

Finally, for the transition region the drag coefficient can be described by
CD =


24 
1 + 0.15Re0.687
Re

(4.27)

21

5 Free Settling Non-Spherical Particles


For spherical particles, the evaluation of drag coefficient is relatively straightforward, requiring only a knowledge of the terminal fall velocity of the sphere.
However, the analysis in the previous section is of use only for a small range
of engineering design problems, as uniform spherical particles are rarely encountered in practice. In fact, mineral processing invariably deals with solids
of irregular shapes. The drag force on any particle is, is dependent on the
boundary layer condition at the surface of the particle which itself is dependent on the particle shape.

Shape factor
In order to generalise to practical cases, it is necessary to examine how particle shape affects the drag coefficient. A great deal of work has been done
on particles of regular shapes other than spheres, for example the studies of
McNown et al. [6, 7]. McNown et al. studied ellipsoids and non-ellipsoids
that were symmetrical with respect to each of three mutually perpendicular
axes as well as discs. Their approach was to assume that the drag force on the
particle could be related to that of a sphere in the stokes law regime according
to
D = K (3D pV )
(5.1)
This equation is simply the Stokes equation multiplied by a dimensionless
correction factor K, which McNown et al. suggested should be labeled the
Stokes number. Different values of K can be ascribed to different shapes.
For ellipsoids, a theoretical solution to the equation of motion is available
similar to the solution presented by Stokes for spheres, allowing values of
K to be determined for ellipsoids. McNown et al. claimed that from these
theoretical results, the drag coefficients for a wide range of shapes can be
estimated within 10 percent.
Drag coefficients associated with particles of irregular shape are of greater
interest to the pipeline designer. Albertson [8] studied the effect of shape on
gravel particles. He concluded that while it was unlikely that particle shape
could ever be accounted for by a single parameter, a shape factor (S.F) of the
form
c
(5.2)
SF =
ab
appeared to provide a satisfactory representation of particle shape. In Equation 5.2, a, b and c are the size of the particle along three mutually perpendicular axes, where a is the longest dimension and c is the shortest. This form of
shape factor had been employed previously by numerous people, including
McNown et al. The shape factor as defined in Equation 5.2 is roughly equivalent to sphericity and, as first used, axis a was taken to be the axis parallel
to the direction of motion. In the Stokes regime this can be important as at
low Reynolds numbers any orientation is stable and shape factors in excess
of unity can be obtained. However, outside the Stokes Law region, particles
22

will fall such that the maximum projected area is normal to the flow direction. Consequently, c will always be the shortest of the three dimensions and
the shape factor will be less than one in all cases.
The main conclusions of Albertsons studies are shown in Figure 5.1, which
consists of a plot of drag coefficient versus particle Reynolds number for particles of various shape factors. Particle Reynolds number is defined using the
nominal diameter of the particle as the length scale, ie
Re =

Vs Dn

(5.3)

where Dn is the diameter of a sphere having the same volume of the particle
itself. Albertsons studies showed that the shape factor defined by Equation 5.2 was able to adequately characterize the both naturally-worn gravel
and crushed gravel. However he did find that the lines of constant shape
factor for the naturally-worn particles and the crushed particles did not coincide, which is explained further in the next section.

Figure 5.1: Drag coefficient versus Reynolds number for different shape factors [8]

Generalized Curves
The end result of the generalization from spheres to irregular shaped particles
is the plot of CD versus Reynolds number shown in Figure 5.2. This plot
employs a shape factor as defined by Equation 5.2 and is a more extensive
version of Figure 5.1, referring to naturally worn particles.
Superimposed on the CD versus Re curves are the additional parameters
Cw and Cs , where
3 CD g( p )
Cw =
=
(5.4)
4 Re
2Vs3
and
Cs =

( p )gD3n
CD Re2 =
8
6
2

(5.5)

Note that Cw can be evaluated without any knowledge of the particle diameter, while Cs can be evaluated without any knowledge of the settling velocity. Thus, Figure 5.2 is a rather economical composite of a great deal of
23

information and can consequently be employed in several ways. The use of


primary concern is for the calculation of drag coefficients. The importance of
knowing the particle shape factor is that it permits the identification of the
correct CD -Re curve for the particles under study.
In most slurry pipeline applications, the solids are subjected to some sort
of size reduction and as Albertson discovered, drag coefficients of crushed
particles of a given shape factor do not coincide with the values for naturallyworn particles of the same shape factor. This does not preclude the use of
Figure 5.2 in practical applications; however, in using Figure 5.2 the engineer
should recognize exactly what it is he or she is evaluating. A knowledge of
the nominal diameter and the terminal fall velocity for a crushed particle allows both CD and Re to be calculated. This allows the correct shape factor
curve to be identified for these particles. The line will be for a naturallyworn particle and will not reveal the actual shape factor of the particle. However, what it does do is identify the shape factor for a naturally worn particle
that is hydraulically identical to the crushed particle. Using this shape factor
and Figure 5.2 allows the settling characteristics of the crushed particle to be
determined.

Fall diameter
Another concept which is sometimes used to characterize the effect of shape
on particle fall velocities is that of fall diameter. Fall diameter is defined as
the diameter of a sphere of the same density as the particle that has the same
settling velocity as the particle under consideration. Figure 5.3 shows the
relationship between particle nominal diameter and fall diameter for various
shape factors. In reality the fall diameter is merely a substitute for the more
basic concept of fall velocity.
The ratio of nominal diameter to fall diameter has been used as a hydraulic shape factor. Again, on this basis, spheres would have a shape factor
of unity, but increasing resistance is indicated by values greater than one.
This differs from the shape factor given in Equation 5.2, where increasing resistance is indicated by values less than one. Of the two, the shape factor
given in Equation 5.2 is preferred.

Drag coefficient
The recommended procedure for finding the drag coefficient for a particle is
as follows:
1. Establish the terminal settling velocity for the particles of interest by
experiment. Since most samples contain particles of wide granulometry,
this involves separation of the solids in to their respective sieve sizes
and the evaluation of the settling velocity for several different sizes at
the coarse end of the size distribution.
2. Calculate the nominal particle diameter Dn for each different sieve size.
This is usually considered to be the geometric mean of the apertures of
the sieves between which the particles are retained.
24

25

Figure 5.2: Drag coefficient versus Reynolds number for irregular shaped particles [9]

Figure 5.3: Relationship between particle nominal diameter


and fall diameter

3. Obtain the shape factor from Figure 5.2. This may vary from one size
fraction to another.
4. From the curves located by the various shape factors, obtain the drag
coefficient as needed.
For work in situations where the accurate knowledge of shape factors is
critical, it may be necessary to evaluate the settling velocities in a fluid of the
same viscosity as the slurry itself.
For preliminary designs, Figures 5.4 and 5.5 show typical experimental
results for coal, and sand and gravel particles.

Figure 5.4: Particle CD for coals

CD can be correlated against the Archimedes Number, Ar, which is defined


as
Ar = CD Re2 =

4 ( p ) gD3p
3 2

(5.6)

26

Figure 5.5: Particle CD for Sand and Gravel

Particle
Coal

Sand

Range
Ar < 24
24 < Ar < 4600
4600 < Ar
Ar < 24
24 < Ar < 2760
2760 < Ar < 46100
46100 < Ar

a
576
128
2.89
576
80.9
8.61
1.09

b
-1
-0.482
-0.0334
-1
-0.475
-0.193
0

Table 5.1: Coefficients for CD correlation

This removes the necessity for iteration in calculating particle settling velocities. For the two sets of experimental data given, the drag coefficient can
be calculated from
CD = aArb
(5.7)
where a and b are found from a piecewise fit of the experimental data. The
values of a and b are given in Table 5.1
The coal particles were fractured by mining and cleaning processes and
are indicative of crushed or ground particles. The sand and gravel particles
were rounded to some extent and are indicative of naturally worn particles.

The effect of turbulence on fluid drag


The preceding sections have indicated how we may characterize the drag
forces on a single particle falling at its terminal velocity through a still fluid.
The environment that exists in a pipeline is obviously different from the still
fluid case, being complicated by factors such as particle rotation, the presence of more than one particle. Also, the fluid will almost invariably be in a
turbulent flow condition.
The effect of both scale and intensity of turbulence on fall velocities would
appear to be an obvious area of weakness in translating the fall velocity information to pipe flows. However, as a review by Graf [10] shows, the full
impact of turbulence is not well understood, mainly due to the difficulties in
27

design and execution of experiments in this area. One relatively well established effect is that the particle Reynolds number at which transition from the
Newton regime to the turbulent regime occurs can be reduced from the order
of 105 to about 103 .

Effect of concentration on drag


The discussion so far was applicable to single particles settling in infinite fluids. When there are a number of particles dispersed in a fluid, the settling
velocity of the particles will differ from that of a single particle due to mutual
interference. When a group of randomly oriented particles settle in a fluid,
the velocity of the cluster is found to be larger than that of individual particles. This phenomenon is different from agglomeration since in the cluster
the particles are not in contact, the groups are stabilized only by the fluid dynamics of the systems. As the suspension concentration increases the particle
acceleration due to cluster formation becomes less marked as the increased
drag from the returning fluid begins to slow them down. At still higher concentrations solids settle as a mass with an interface between the solid phase
and the fluid. This type of settling behaviour is known as hindered settling.
In the hindered settling regime the settling velocity decreases with an increase in solids concentration. For particles settling in the Stokes regime, the
effect of concentration is given by the following equation
Vs,h
= 4.5
Vs

(5.8)

where, Vs,h is the hindered settling velocity, Vs is the settling velocity of a


single particle in an infinite fluid and is the volume fraction of the voids
(equal to 1 CV , where CV is the concentration by volume of the particles).
In a suspension made up of a mixture of particles sizes the coarse particles
settle in a matrix of smaller particles. For a binary mixture, Davies and Kaye
[11] have shown that the mixture settles without segregation when the distance between larger particles is such as to trap the smaller particles. Thus,
particle segregation would occur up to a critical concentration beyond which
the smaller particles will remain trapped between larger particles.

28

6 Classification of Particles
Free Settling
Consider in Figure 6.1, a particle rising on a stream of fluid.
Let U be the upward velocity of the fluid, and Vp be the absolute velocity
of the particle, then using the terminal settling velocity w as a reference, the
following limits can be set up:
U = 0,Vp = w particle falling in a stationary fluid at its terminal settling
velocity
U = w,Vp = 0 particle suspended in the fluid
U < w,Vp > 0 particle falling at an absolute velocity of Vp = w U
U > w,Vp < 0 particle rising at an absolute velocity of Vp = U w

Figure 6.1: Particle falling in a upward flowing fluid

The above limits explain the principle involved in the separation of particles of different sizes into fractions of desired characteristics. The process is
known as classification.
Particles subjected to classification may vary in density as well as size. The
separation of mixtures of this kind is called sorting, as distinct from sizing,
which is the name given to classification of particles of the same density.
The limits set up for a single particle above, or more generally for particles at infinite dilution cannot be applied to suspensions without some refinement. For the sake of simplicity however, further discussion will refer to
very dilute suspensions, when is approaching unity and particles may be
assumed to follow the pattern adopted in free settling. On the basis of this
simplifying assumption, the process of separating particles into fractions of
desired characteristics may be explained with the aid of a sorting diagram, as
follows.
Consider a very dilute suspension containing a mixture of two different
materials, A and B, with material A the lighter of the two. Let D1 and D2 be
the diameters of the smallest and largest particles of the mixture. Plotting
the terminal settling velocity against the particle diameter within the whole
range of sizes available will result in two curves, as shown in Figure 6.2.
Next it is assumed that classification of the two materials is carried out
by a rising stream of fluid. This allows two pure fractions of material to be
29

Figure 6.2: Sorting Diagram

obtained, by maintaining the upward velocity of the fluid at a predetermined


level. If this velocity is made equal to the settling velocity of the smallest
particle of B (U = wB1 ) a pure fraction of A will separate from the mixture
as an overhead product. This fraction will contain particles ranging in size
between D1 and D3 . Similarly a pure fraction of B ranging in size from D4
to D2 , can be obtained as a bottom product (sediment), by maintaining the
velocity of the classifying fluid equal to the settling velocity of the largest
sized particle of A.
For operation at either wA2 or wB1 , the ratio of the smallest and largest
diameters is given by Equation 6.4, which is derived as follows. Equation 4.20
gave the settling velocity for a particle in an infinite fluid, repeated here
s
r
D p ( p ) g
4
(6.1)
w=
3CD

taking wA = wB , where the subscript A refers to the largest particle A for


the pure bottom fraction, or vice versa for the pure overhead fraction,
w2A =

4DA (A ) g
3CD,A

(6.2)

w2B =

4DB (B ) g
3CD,B

(6.3)

and

Equating the above two equations, for wA = wB , simplifying and rearranging gives
DA B CD,A
=
(6.4)
DB A CD,B
The ratio DA /DB is often known as the settling ratio. In the turbulent
regime, CD,A /CD,B equals 1 and Equation 6.4 reduces to
DA B
=
DB A

(6.5)

In the laminar range,


CD =

24
24
=
Re wD p

(6.6)
30

since wA = wB

CD,A DB
=
CD,B DA

Substituting this back into Equation 6.4 and rearranging gives


r
DA
B
=
DB
A

(6.7)

(6.8)

Inspection of this equation and Equation 6.5 shows, at least for the two
ranges considered, and other things being equal, the sorting range can be
widened by increasing the density of the classifying fluid, . This is occasionally done by adding fine particles or some reasonably soluble and cheap
salts, such as calcium chloride, to the classifying medium.

Hindered Settling
As the proportion of solids in the pulp increases, the effect of particle crowding becomes more apparent and the falling rate of the particles begins to decrease. The system begins to behave as a heavy liquid whose density is that
of the pulp, rather than that of the carrier fluid: this condition is referred to
as hindered settling. Because of the high density and viscosity of the slurry
through which a particle must fall in separation by hindered settling, the resistance to falling is mainly due to the turbulence created and a modified
form of Newtons law can be used to determine the approximate falling rate
of the particles:
q
w=k

D p (s p )

(6.9)

where p is the pulp density and s is the solids density.


The lower the density of the particle the more marked is the effect of reduction of the effective density, s p , and the greater the reduction in the
settling velocity. Similarly, the larger the particle the greater the reduction in
settling rate as the pulp density increases.
This is important in classifier design, in effect hindered settling reduces
the effect of size, while increasing the effect of density on classification.
The hindered settling ratio is given as
DA B p
=
DB A p

(6.10)

The hindered settling ratio is always greater than the free settling ratio
and the denser the pulp the greater the hindered settling ratio. Hindered
settling classifiers are used to increase the effect of density on the separation,
whereas free settling classifiers use relatively dilute suspensions to increase
the effect of size on the separation.

31

7 Solid Liquid Separation


Mineral processing opeartions are generally carried out in the aqueous phase.
At some stage during the concentrating or processing operation a liquid-solid
mixture is produced which must be partially or completed dewatered. This
is generally because
Pyrometallurgical processes such as smelters or refineries cannot handle the exess amounts of liquid
Tranportation of liquid with the concentrate is expensive (except for
pipelines)
Recycling of process water is desirable
Solids concentration may need to be adjusted prior to a further processing stage
Disposal of clean water is necessary due to environmental concerns
amongst other reasons

Dewatering processes
Dewatering can be broadly classified into four processes. These are:
Sedimentation
Filtration
Mechanical Dewatering
Thermal Drying
The choice of process used for dewatering depends on the particle size
and solids concentration of the pulp. For instance, in the case of coarse particle sizes the dewatering process need simply be dewatering screens, such as
a vibrating screen or sieve bend where the screen aperture is chosen (down
to about 250 microns) such that it retains the solid particles but allows the
liquid medium to pass. As the particle size in the slurry decreases the screen
aperture required to retain the particles also decreases until a stage is reached
where the pressure drop across the screening surface is too large to allow free
drainage of the liquid through the screen. At this stage additional pressure
(or vacuum) is required to maintain a reasonable separation rate. The separation is then referred to as filtration.
When the solids concentration of the slurry is low, the primary dewatering
step is some sort of sedimentation settling.
The most difficult dewatering situation occurs when the percentage of
solids is low and the particle size of the solids is very low (sub-micron). This

32

presents a problem because very fine particles form a filter cake of low porosity and high pressure drop and require lengthy times for filtration separation.
When the solids concentration of the slurry is low, the primary dewatering
step is a sedimentation process such as thickening or cycloning followed by
filtration and possibly thermal drying, depending on the desired condition of
the product.

Sedimentation Separators
Sedimentation separators employ the principles of inertial or gravitational
forces acting on solid particles in a fluid to bring about settling. The velocity
at which a particle will settle out of a fluid is proportional to its density and
the square of its diameter and inversely proportional to the viscosity of the
fluid. For spherical particles
D2p
(7.1)
V

and for non-spherical particles


D2p
V k

(7.2)

where k is a constant.
This relationship is valid provided the particle size is not so small that
its motion is affected by Brownian motion below this size the terminal velocity is so reduced that Brownian motion is sufficient to keep the particles
suspended indefinitely in a static fluid.
Settling equipment utilising gravitational forces to bring about sedimentation are termed clarifiers or thickeners.
Settling rates may be increased artificially by introducing a centrifugal
motion to the fluid, increasing particle velocities relative to the bulk of the
fluid. Equipment using this method are called cyclones or centrifuges.
Settling rates may also be increased by increasing the effective particle
diameter (since settling rates are proportional to D2p ) by aspects of colloidial
science referred to as coagulation and flocculation.

Continuous Sedimentation
Sedimentation may be defined as the process of separation of solid particles
from a slurry into a substantially clear liquid and a slurry of a higher concentration of solids, called the sludge. The force causing this separation may be
gravitational, centrifugal or of some other form, but the term sedimentation
generally implies settling due to gravity.
If the slurry treated in the sedimentation process contains large amounts
of solids, the operation is called thickening, otherwise it is known as clarification. Alternatively, the operation may be given either name based according
to whether the thickened slurry of the overflowing liquid is the desired product.
33

Figure 7.1: Clarifier[12]

Figure 7.2: Typical settling zones in a thickener

The mechanism of sedimentation is that of free and hindered settling. It is


usually conveniently described in terms of the characteristic zones which appear as the process is in progress. The work of Coe and Clevenger, published
as early as 1916, is still recognized as an important contribution in this field.
Their method of correlating the data from small-scale batch settling tests is
often utilized in scaling up operations. Later work may be regarded as useful extensions of this method or merely its refinement. The method will not
be described here, and the discussion will be confined to a brief survey of
the settling zones in a continuous thickener and to the basic principle of its
design.
Figure 7.1 shows a typical thickener design, while Figure 7.2 is a diagrammatic representation of a conventional thickener for handling large volumes
of slurries. The settling zones which appear as the process is in operation
have the following characteristics:
Zone A Clarification zone
The clarification zone presents an ideal condition for free settling. Any particle
which may have entered this zone has sufficient time to resettle before the
liquid has reached the overflow level.

34

Zone B Constant concentration zone


The constant-concentration zone, or feed zone, may also be expected to favour
free settling. The concentration of solids in this zone is substantially of the
same order as that in the slurry feed.
Zone C Transition zone
The transition zone lies between the feed zone and the compression zone. The
concentration of solids in this zone passes through all values between the
concentration of the feed and the ultimate concentration at the top of the
compression zone. The presence of this zone is a result of the settling velocity
of the particles decreasing as the concentration increases.
Zone D Compression zone
In the compression zone, also known as the thickening zone, consolidation of
the settled particles takes place. The particles have a limited downward motion, particularly in the upper portion of the zone, as their consolidation is in
progress. This motion produces a slow upward flow of the liquid displaced
by the moving particles.

Thickener Sizing
The usual requirement in design practice is the specification of the crosssectional area and the depth of a thickener. A brief outline of a method leading to this specification is given below.
There may be limiting capacities in both the clarifying and thickening
zones, thus both pssibilities must be considered. There is always a clarifying zone capacity limitation, but there is not necessarily a thickening zone
capacity limit.
The cross-sectional area of a thickener must be large enough to enable
solids to settle at a rate equal to the feed rate. If this area is less, the solids
will accumulate at the transition zone with the effect that this zone will move
upwards until it reaches the overflow level. If this happens, the thickener will
not produce clear overflow.
The successful operation of a thickener also depends on its clarification capacity which, for a given throughput, is a function of its cross-sectional area.
As this area is closely related to the settling velocity of the slowest moving
particle, the problem confines itself to the evaluation of this critical velocity.
This is normally achieved by a proper interpretation of batch settling curves.
Once the cross-sectional area of the thickener is established, the depth of
the compression zone (zone D) can be determined from the retention time of
particles in this zone. This time is controlled by the rate of discharge of the
sludge, which in turn depends on the desired concentration of the product.
Having fixed this concentration, the volume of the compression zone may be
estimated from consideration of the time each layer of the sediment has been
in compression. The estimate is usually made with the aid of test data. The
full procedure is outside the scope of this course, however. A rough estimate
35

of the total depth of a thickener can be approximated by making allowances


for the clarification and settling zones, about 0.6 m for each, and for the extra
storage capacity required to cover possible operational irregularities.

36

8 Pressure drop through packed beds


The first scientific approach to the study of flow through beds of solid particles is attributed to Darcy. Apart from his major contribution to hydraulics
in the fields of fluid friction in pipes, Darcy is also noted for his work on the
flow of water through beds of sand. The result of his investigations was published in 1856 in a treatise which is generally regarded as the first work on
the subject. The leading conclusion drawn by Darcy from the results of his
investigations was that the drop in pressure through a bed of solid particles
was proportional to the first power of the flowrate rather than, as it was then
popularly supposed, to its square root.
A number of attempts have been made since the time of Darcy to express the drop in pressure in terms of the flowrate and physical properties
of fluid-solid systems. Among these was a correlation of experimental data,
presented in 1913 by Chilton and Colburn, which attracted a great deal of
attention. Based on the Darcy equation for flow in pipes, the correlation assumes the form
2
0 U h
Fw Fp
(8.1)
p = 4 f
2 Dp
where
p = pressure drop,
U = superficial velocity, based on empty cross-section,
D p = nominal size of packing particle,
h = depth of packing,
Fw = correction factor for wall-effects,
Fp = correction factor for hollow particles
The friction factor, f 0 , which appears in the above equations may be obtained from a diagram (as shown in Figure 8.1), which plots this factor versus
a Reynolds number. In this case the Reynolds number is defined by
Re =

UD p

(8.2)

Figure 8.1 is the outcome of an excellent survey of data available on geometrically regular shapes, as well as on irregularly shaped particles.
The original plot of this data shows a curve drawn through widely scattered points, particularly in the region of high Reynolds numbers. For the
sake of simplicity however, the curve has been approximated by two straight
lines intersecting at a point corresponding to Re = 40, which is accepted as
the transition point between the laminar and turbulent flow regimes.
The two lines are approximated by the following equations: For the laminar region, Re < 40
850
f0 =
(8.3)
Re
and for the turbulent region, Re > 40
f0 =

38
Re0.15

(8.4)
37

1000

100

10

10

100

1000

Reynolds Number

Figure 8.1: Friction factor for Chilton-Colburn equation

(a) Berl Saddles

(b) Raschig Rings

Figure 8.2: Packed bed packings [13]

The wall effect correction factor, Fw , which appears in Equation 8.1, has
been determined experimentally by Furnas for solid particles only. This factor is presented as a plot against the ratio of particle diameter to column diameter. For hollow shapes this factor is taken as unity and an alternative
correction factor, Fp , is used instead. From the correlation of data presented
by White, this factor may be evaluated in terms of the nominal particle diameter, D p using
C
Fp = 0.5
(8.5)
Dp
The constant C in the above equation has a value of 0.24 for Raschig rings or
Lessing rings, and 0.13 for Berl saddles, as shown in Figure 8.2.

Kozenys hydraulic model


The flow through a packed bed of solid particles takes place in the irregularly
shaped channels provided by the voids of the bed. It seems logical therefore
that any equation relating to this kind of flow should refer to these channels,
rather than to the size of the bed forming particles, as is suggested by Equation 8.1. The Kozeny hydraulic model assumes that the channels are formed
by capillary tubes running parallel to the direction of flow, as shown in Figure 8.3. The equation that determines the pressure drop in this model is the
Carman-Kozeny equation.
38

Figure 8.3: The Kozeny hydraulic model

In Figure 8.3, the fluid flows through a bed of solid particles contained in
a cylindrical vessel of diameter D and length h, which is also the depth of the
bed. If there is n channels formed in the bed in total, then from the definition
of voidage fraction,
=

volume of voids nhd 2 /4


=
volume of bed
hD2 /4

(8.6)

d2
(8.7)
D2
If v is the average velocity through the parallel channels, then to satisfy
continuity
d 2
D2
= nv
(8.8)
U
4
4
from which
 
U D 2
n=
(8.9)
v d
=n

and

U
(8.10)

Similarly, the hydraulic mean radius may also be expressed in terms of


the voidage fraction.
If Vp is the volume of a single particle and A p is its surface area, then in a
uniformly sized bed of N particles,
NVp = total volume of particles
NA p = total surface area of particles

and NVp
= total volume of voids
1
The hydraulic mean radius is generally defined as the ratio of the crosssectional area and the wetted perimeter. For a bed of solid particles, the parameter is conveniently defined as
v =

rH =

Total volume of voids


Total surface area of particles

(8.11)

39

It follows that
rH =

NVp /(1 )
Vp
=
NA p
A p (1 )

(8.12)

If the specific area of a single particle is defined as


Ap
Vp

(8.13)

S p (1 )

(8.14)

Sp =
then
rH =

The Reynolds numbers for this Kozeny hydraulic model is defined as


vr
H

(8.15)

[/S p (1 )](U/)

(8.16)

U
S p (1 )

(8.17)

Re =
substituting rH and v into this gives
Re =
Simplifying gives
Re =

The Carman-Kozeny equation


The drawbacks inherent in Equation 8.1, which is simply a form of the Darcy
equation for flow in pipes, can be removed by adapting it to the theory advanced by Kozeny as outlined above.
Ignoring the factors Fp and Fw in Equation 8.1 and replacing the terms U
and D p with their equivalents, v and rH , leads to
p =

4 f 00 h v2
2rH

(8.18)

where f 00 is a new friction factor.


Further substituting the expressions for v and rH into this equation, and
letting fc = 2 f 00 , then
fc h (U/)2
p =
(8.19)
/[S p (1 )]
Rearranging the terms in this equation gives


2 1
hS p
p = fcU
3

(8.20)

Equation 8.20 is known as the Carman-Kozeny equation.


From the available data, Carman made a logarithmic plot of the factor
fc , now called the Carman friction factor, against the Reynolds number as
defined by Equation 8.17 The resulting line indicated a linear relationship up
40

100

Solid packings
Hollow Packings

fc

10

0.1

0.1

10
Reynolds Number

100

1000

Figure 8.4: Carman friction factor

to approximately Re = 4. From the slope of the straight line portion of the


curve Carman deduced the following relationship between the two variables
5
, Re 4
(8.21)
Re
Similar to most flow situations, the range up to Re = 4 to which Equation 8.21
applies is called laminar or viscous flow. For the same reason, the range
above Re = 4 is referred to as turbulent. Within this range, the line on the
Carman friction factor plot curves moderately, and can be approximately described by
5
0.4
fc =
+ 0.1 , Re > 4
(8.22)
Re Re
Further experimental evidence has confirmed the validity of Equation 8.22
for a wide range of particles, with the exception of hollow particles, for which
the following equation is used
fc =

fc =

1
5
+ 0.1 , Re > 4
Re Re

(8.23)

A plot of the Carman friction factor is shown in Figure 8.4


For spherical particles, the specific surface is given from Equation 8.13.
Taking A p = D2p , and Vp = D3p /6, gives
Sp =

6
Dp

(8.24)

where D p is the particle diameter.


Substituting this into the Reynolds number equation gives
Re =

D pU
6(1 )

(8.25)

In the laminar range, the friction factor expression becomes


fc =

30(1 )
D pU

Finally, substituting this into Equation 8.20 gives


"
#
Uh (1 )2
p = 180
, Re 4
D2p
3

(8.26)

(8.27)
41

which gives the pressure drop for a packed bed of spheres in the laminar
range.

Specific surface of a packed bed


The surface area of individual particles in a unit volume of a packed bed
is known as the specific surface of the bed. The usual symbol given to this
quantity is a , and its relation to the specific surface of the individual particle
may be deduced as follows.
If Nc is the number of individual particles contained in a unit volume of
the packed bed, then by definition
a =

surface area of all particles


unit volume of bed

(8.28)

Nc A p

1
NcVp 1

(8.29)

a =

Taking S p = A p /Vp and simplifying


a = S p (1 )

(8.30)

The Reynolds number and Carman-Kozeny equation expressed in terms


of a assume the forms
U
(8.31)
Re =
a
and
p = fc

a 

hU 2

(8.32)

respectively.

42

9 Fluidization
A fluidised bed is a bed of particles held in suspension by a fluid stream, as
for example a bed of coal particles in a stream of air (fluidized bed combustion), or a bed of catalyst particles in a stream of petroleum product (catalytic
cracker).
If a fluid is passed through a bed of solid particles, the pressure increases
with the velocity until a point is reached when the bed starts to expand. At
this point, the particles lose permanent contact with each other and are free
to move throughout the whole bed. The bed then resembles a boiling liquid,
and is said to be in a fluidized state.
A typical result for a gas fluidized bed is shown in Figure 9.1, which is a
plot of superficial gas velocity versus the pressure drop, and bed height.
The curve ABC in Figure 9.1 is for slowly increasing flow through a bed
which has not previously been fluidized, the point C being defied as the point
of incipient fluidization. The pressure drop at this point is just sufficient to
support the weight of the particles. If p is the pressure drop, then on the
basis of a unit cross section
p = hg ( p ) (1 )

(9.1)

where is the voidage fraction (or porosity) of the bed.


At point B, the pressure drop is slightly more than enough to support the
weight of the particles. The difference between the two values for pressure
drop is due to the friction between the individual particles before they are
able to move freely. If the velocity is slowly decreased from point C, the
particles are then more loosely packed than they were prior to fluidization,
and the pressure drop is less for the same gas velocity. Additionally, the bed
height is greater than for the previously unfluidized bed.
Below the point of incipient fluidization, the behaviour of the bed is the
same for both gas and liquid fluidised beds. Above this point they behave
differently though.
Increasing the gas velocity above point C results in the formation of bubbles through the bed which burst on reaching the bed surface. This kind of
behaviour is called aggregative fluidization. The bubbles agitate the bed, and
its height fluctuates. With still higher velocity, the bubbles grow and appear
more frequently until their frontal diameters nearly equal the diameter of the
containing vessel. The bed is then said to be slugging. A further increase in
velocity causes the particles to be carried out of the vessel. The porosity of
the bed is then very high, and the phenomenon is that of pneumatic transport,
with the gas velocity greater than the freefall velocity of the particles.
For a liquid fluidized bed above the incipient point, the height increases
with velocity but there is usually no marked fluctuation of the bed surface
and no bubble formation occurs. This kind of behaviour is known as particulate fluidization.
Although aggregative and particulate fluidization are typical for gas and
liquid fluidized beds, there are exceptions to this generalization.

43

Figure 9.1: Pressure drop and bed height for a fluidized bed

44

10 Non-Newtonian Fluids
Rheology
Rheology is defined as the science of deformation and flow of matter [14]. The
word rheology comes from the Greek words rheo, meaning to flow, and logi,
meaning science. The study of rheology deals with the stress, strain and time
relationships of any matter, although usually it deals with cohesive matter
which may be solid, liquid or gaseous. For the purposes of this study the
rheological behaviour of fluid only will be examined.

Newtonian Fluids
The simplest rheological behaviour that a fluid may exhibit is Newtonian behaviour.
To illustrate Newtonian rheological behaviour many authors use the example of a fluid being contained between two large parallel plates, as depicted in Figure 10.1 [1517]. In this situation the plates are of area A, and are
separated by a small distance, dy. One plate is set in motion by the application of a force, F, which is applied parallel to the plate. The resulting velocity
of the plate is u. If the fluid displays Newtonian rheological behaviour then
after a sufficient time, t, the steady state condition shown in Figure 10.1 will
be reached. In this situation, provided the flow is laminar, the velocity of
the fluid will increase in direct proportion to the distance from the stationary
plate. The relationship between the force required to maintain the motion of
the upper plate and the velocity of this plate is given by the Newtonian law
of viscosity as
du
F
=
(10.1)
A
dy
or
du
=
(10.2)
dy
or
=
(10.3)

Figure 10.1: Flow of a Newtonian fluid between two plates

45

where is the shear stress in the direction parallel to the plate and is the rate
of shear in this direction. The constant of proportionality, , is the Newtonian
viscosity and is independent of any of the other terms of equations 10.1, 10.2
or 10.3. Factors such as temperature and pressure may however affect the
value of viscosity for a Newtonian fluid.

Non-Newtonian Fluids
Many fluids flow under the action of an applied force, F, but do not exhibit
Newtonian behaviour. Consequently the rheological behaviour of these fluids cannot be adequately defined by a single constant viscosity. Such fluids
are broadly classified as non-Newtonian. Fluids that exhibit non-Newtonian
rheological behaviour may be further divided into three distinct groups [15,
16]; time independent fluids, time dependent fluids and viscoelastic fluids.

Time Independent Fluids


For time independent fluids the relationship between shear stress and shear
rate is still of the form

= f ()
(10.4)
but these two quantities cannot be related by the one constant of viscosity as
for Newtonian fluids. Instead a minimum of two constants are required to
adequately define the relationship.
Time independent fluids are generally grouped into four categories [9, 16]
1. Bingham fluids
2. Pseudoplatic fluids
3. Dilatant fluids
4. Yield pseudoplastic fluids
The general shear stress versus shear rate characteristics of each of these
groups of fluids are indicated in Figure 10.2.
Whilst it can easily be seen that a single viscosity will not define the rheological behaviour of these fluids, since the shear stress versus shear rate
relationship is no longer linear or the shear stress is no longer zero as the
shear rate approaches zero or both, it is still useful to have some method for
comparing and correlating data between these different fluids. One of the
more commonly used parameters for this purpose is the apparent viscosity, a ,
which is defined as
shear stress
(10.5)
a =
shear rate
The concept of apparent viscosity is indicated in Figure 10.3.
The value of apparent viscosity will vary with shear rate. Consequently, a
value of apparent viscosity for a non-Newtonian fluid is meaningless unless
the shear rate at which it applies is also given. In the case of Bingham and
pseudoplastic fluids, the value of apparent viscosity decreases with increasing shear rate, whilst for dilatant fluids the apparent viscosity will increase
46

Figure 10.2: Shear stress versus strain for time independent


non-Newtonian fluids

Figure 10.3: Apparent viscosity for a non-Newtonian fluid

47

with increasing shear rate. As a result Bingham and pseudoplastic fluids are
termed shear thinning whilst dilatant fluids are shear thickening [9].
As the name suggests yield stress fluids require some minimum shear
stress to be applied before flow will occur; the minimum shear stress being
termed the yield stress. Fluids which exhibit a yield stress and a linear relationship between shear stress and shear rate after the yield stress is has been
exceeded are known as Bingham fluids (or Bingham Plastics). These fluids are
described by the equation
= y +
(10.6)
where y is the yield stress and is known as plastic viscosity or the coefficient of rigidity [18]. Whilst few fluids exhibit exact Bingham plastic behaviour many can be approximated by this model, especially at higher shear
rates. Common examples are drilling muds, oil paints, toothpaste and sewage
sludges [15]. Wasp et al. [9] explains the cause of Bingham behaviour as being the result of a three dimensional structure which is sufficiently rigid to
resist any stress less than the yield stress. When the yield stress is exceeded
this structure disintegrates and the fluid behaves as a Newtonian fluid under
a shear stress of ( y ). The structure reforms when the applied shear stress
again falls below y . This reformation can often be a function of time which
could be virtually infinite.
Psuedoplastic fluids exhibit no yield stress but do exhibit a non-linear
relationship between shear stress and shear rate. A popular mathematical
model used to represent pseudoplastic fluid behaviour is the Ostwald-de Waele
or Power law model [19] and is expressed as
n , where n < 1
= k ()

(10.7)

In this model k and n are constant for any particular pseudoplastic fluid.
The term k is called the consistency. The term n is called the flow behaviour
index, and is an indication of the degree of non-Newtonian behaviour. The
smaller the value of n the greater the non-Newtonian behaviour. If n = 1, then
the equation represents a Newtonian fluid with a viscosity of k. Bhattacharya
[20] points out a deficiency of the power law model in the initial slope and
hence apparent viscosity at a shear rate of zero. There have been numerous
other pseudoplastic models outlined in the literature to try and overcome the
deficiencies of the power law model. Govier and Aziz [16] summarize many
of these, and the more common ones are
Prandtl

2)
= C1 sin1 (/C

(10.8)

1 +C2 sin (/C3 )


= /C

(10.9)

= C1 +C2 sinh1 (C3 )

(10.10)

(C2 + )
+
= C1 /

(10.11)

+C2 sinh1 (k2 )


+ ...
= C1 sinh1 (k1 )

(10.12)

Eyring
Powell-Eyring
Williamson
Ree-Eyring

48

Whilst these models may offer an improvement over the power law model
for empirically defining the rheological behaviour of pseudoplastic fluids,
especially at very high or very low rates of shear, they are considerably more
difficult to use [15]. For engineering problems involving the flow of fluids in
pipelines the power law model gives excellent results with the advantage of
being very simple to use [16].
Another useful model for pseudoplastic rheological flow behaviour was
proposed by Cross [21] and is described by Govier and Aziz [16] as being
superior to any other models of comparable simplicity. This model uses the
relationship
i
h

a = + (0 ) / 0.66

(10.13)

where a and are the apparent viscosities at shear rates of zero and infinity
and is a coefficient related to the shear stability of the structure.
Dilatant fluids are the opposite to pseduoplastic fluids in rheological behaviour. The power law model can be fitted to dilatant fluid behaviour with
the flow behaviour index, n, being greater than 1. In a similar manner most
of the other models for pseudoplastic fluids can be fitted to dilatant fluid behaviour [20].
Certain fluids also display a combination of yield stress and pseudoplastic
or dilatant behaviour. One of the simplest models to accommodate this form
of rheological behaviour is the Herschel-Bulkley model which is an adaption
of the Bingham model to incorporate a power law term [20]. The model is
given by
= y + kn
(10.14)
Another well used and comparatively simple model for yield pseudoplastic behaviour was proposed by Casson [22] and is described as giving as good
a fit to experimental data as many more complex models. The model is described by
1

2 = y2 + k 2

(10.15)

Time dependent fluid behaviour


The second major category into which non-Newtonian fluids are divided are
those which exhibit time dependent behaviour. For these fluids the shear
stress versus shear rate relationship is also a function of time and, depending
on the rheological response of the fluid with time, are further classified as
either thixotropic or rheopectic fluids.
A thixotropic fluid has a decreasing shear stress with time at a constant
shear rate. After a considerable review of the literature Bauer and Collins
[23] gave the generally agreed definition of thixotropic behaviour as
A thixotropic system exhibits a time dependent, reversible and
isothermal decrease of viscosity with shear in flow.
Flow curves for a thixotropic fluid are given in Figure 10.4
Figure 10.4(a) demonstrates the decreasing shear stress with time that is
exhibited by thixotropic fluids that are sheared at a constant rate. For increasing rates of constant shear this decrease in in shear stress with time becomes
49

(a)

(b)

Figure 10.4: Behaviour of thixotropic fluids [15]

far more rapid. Figure 10.4(b) illustrates the shear stress against shear rate
relationship of a thixotropic fluid that has been left to recover after initial
shearing. The longer the thixotropic fluid is left to recover the greater the
shear stress which will be exhibited on subsequent shearing.
Wasp et al. [9] explain thixotropic behaviour as being caused by the breakdown of a three dimensional structure within the material being sheared. The
structure rebuilds once the shearing has stopped. Examples of fluids which
exhibit thixotropic behaviour are iron oxide slurries, alumina hydrate slurries, starch pastes, clay suspensions, and many crude oils [16, 23].
The opposite time dependent behaviour to thixotropic is when the shear
stress exhibited by a fluid, at a constant shear rate, increases with time. Fluids
that exhibit this type of behaviour are often called rheopectic fluids [9, 15, 16].
Dilatancy and antithixotropy are also often used to describe the behaviour
of these fluids [23]. Bauer and Collins [23] distinguish between dilatancy
and rheopectic behaviour by defining a dilatant system as one which recovers its original state once shearing has stopped almost instantaneously,
whilst a rheopectic systems requires a finite time to recover. Fluids that exhibit rheopectic behaviour are relatively rare, although tailings from mineral
sand processing and some clay suspensions have been found to behave in
this way [16].
Whilst there are numerous models to describe the flow behaviour of fluids, some of which have been briefly outlined, they are of little value without
a means to develop data to apply these models. Later sections investigate
techniques by which the flow behaviour of fluids can be determined.

50

11 A Generalized Method for Flow in


Pipes
This section starts with laminar flow and the analytical methods that are
available, before moving to the more complex methods available for turbulent flow. In a similar progression from simple to complex, the treatment
starts with simple rheological models before moving onto a more generalized method which makes few assumptions about the rheological behaviour
of the fluid.
The engineer is primarily concerned with the frictional pressure drop which
will result from a given flowrate through a given pipe. A simple force balance, as shown in Figure 11.1, indicates that the pressure loss due to friction
is related to the wall shear stress by
DP
4L

w =

(11.1)

Solution for Laminar Flow


The flowrate through a pipe, of any arbitrary velocity distribution, is given
by
Z R

Q=

(11.2)

2rvdr
0

where v is the local velocity at radial position r from the central axis of the
pipe.
Integration by parts gives:

R
Z
2
2
Q = r v r dv
(11.3)
0

At the wall the velocity is zero, that is v = 0 at r = R, and


Q =

Z R

r2

dv
dr
dr

(11.4)

It is now assumed that the flow is laminar and consequently


=

dv
dr

(11.5)

Figure 11.1: Pipe flow

51

and from the law of linear distribution of shear stress


r
= w
R

(11.6)

where is the shear stress at radius r and w is the shear stress at the wall.
Therefore
Z
R3 w 2
Q= 3
d
(11.7)
w 0
Equation 11.7 is a general relationship valid for any time independent
fluid. If can be expressed as a function of shear stress, , then the equation
may be integrated. In the following the average flow velocity is defined as V ,
and the pipe diameter, D, is used instead of the pipe radius.
For a Newtonian fluid in laminar flow
=

(11.8)

substituting this into Equation 11.7 and integrating gives


w =

8V
D

(11.9)

This is the equivalent solution to the flow between flat plates (w = V /y), but
for a pipe.
For a power law fluid in laminar flow,
= kn

(11.10)

Again, substituting and integrating gives



 1
4n
w n
8V
=
D
3n + 1
k
or

3n + 1
w = k
4n

n 

8V
D

(11.11)

n
(11.12)

For a Bingham plastic in laminar flow,


= y +
Again, substituting and integrating gives
"
  #
w
4 y 1 y 4
8V
=
1
+
D

3 w 3 w

(11.13)

(11.14)

which is known as the Buckingham equation. Equation 11.14 can be solved


numerically, but often the last term is considered negligible and


8V
w
4 y
=
1
(11.15)
D

3 w
This is equation is referred to as the truncated Buckingham equation, and is
much simpler to solve. Rearranging for the wall shear stress gives
 
4
8V
w = y +
(11.16)
3
D
52

In summary for laminar flow the following expressions for the wall shear
stress apply

Newtonian Fluid

w =


3n + 1
w = k
4n

Power Law Fluid

8V
D
n 

4
w = y +
3

Bingham Plastic

8V
D

8V
D

n

The importance of the above relationships is in the ability to now extract


the required fluid properties (y , , k, n) from easily obtained pseudo-shear diagrams plots of the wall-shear, w , versus the average strain rate, 8V
D . Laminar
flow pipe design can also be performed using the above relationships if the
fluid type is known.

Generalized method for laminar flow


Similarities between the above relationships suggest the use of a general form
of equation that would include all of the above fluid types.
Metzner and Reed [24] suggested a generalized model for pipeline flow,
w = k

8V
D

n0
(11.17)

where the generalized pipe index, n0 is given by


n0 =

d (ln w )
d (ln(8V /D))

(11.18)

or, the tangent slope of ln(w ) versus ln(8V /D).


Unlike the n parameter in the power law model n0 is allowed to vary with
shear rate. Thus the model is quite general and designed to include all types
of rheological behaviour even though, for instance, it does not appear to fit
the Bingham plastic relationship very well. This is a point which will be
addressed in an example presented shortly.
Using Equation 11.17 in pipe design requires n0 and k0 to be first determined. In practice this can be done quite simply if a rheogram is available,
although the mathematics involved in a rigorous proof of the method is quite
difficult.
At the heart of the method is the idea that every fluid can be approximated
as a power law fluid. Then comparing equations 11.10 and 11.12 it follows
that


8V 3n0 + 1
w
(11.19)
D
4n0

53

For shear rates close to w , n0 may be treated as a constant and then the
average shear rate, 8V /D, and the actual shear rate at the wall, w , differ only
by a constant of proportionality. Thus n0 may be obtained from a rheogram
that provides w plotted against w .
d (ln w )
(11.20)
d (ln w )
What w should be used? The wall shear rate that is actually experienced
in the pipe flow. Since this is not known prior to the solution being obtained
some iteration may be necessary.
The generalized pipe consistency, k0 , may be obtained by comparison between equations 11.12 and 11.17,
0
 0
3n + 1 n
0
(11.21)
k =k
4n0
n0

The procedure may be illustrated by an example.

Turbulent Flow and Dimensional Analysis


For turbulent flow no analytical solutions to the pipe flow problem are available and recourse is made to experimental results which are presented in
non-dimensional form.
For a Newtonian fluid and a smooth walled pipe, the physics presents
itself as,
F(p, ,V, D, L, ) = 0
(11.22)
Dimensional analysis of this parameter set gives rise to the following


P
V D
L
(11.23)
= G
V 2 D

where G is an unknown function of Reynolds number which must be determined experimentally. In terms of the Fanning friction factor, C f , Equation 11.23 takes the form of the familiar Darcy Equation,
L V 2
(11.24)
D 2
and the friction factor is presented graphically on the Moody diagram as a
function of Reynolds number.
For a power law fluid the parameter set is slightly different, since is no
longer well defined, and becomes
P = 4C f

F(p, ,V, D, L, n, k) = 0

(11.25)

Applying a similar type of dimensional analysis to the above produces


the dimensionless grouping


P
L
V 2n Dn
= G n,
(11.26)
V 2 D
k
This suggests that a diagram similar to the Moody diagram could be constructed from experimental data where now the friction factor would be a
V 2n Dn
function of both the flow index, n, and the Reynolds type number,
.
k
54

Generalized method applied to Turbulent Flow


The Reynolds type number suggested in Equation 11.26 is not unique, and
other forms of the number have been proposed. Dodge and Metzner [25]
proposed the use of a generalized Reynolds number, Re0 , which is essentially
a Reynolds number based on the effective viscosity of the fluid. Thus,
0

Dn V 2n
Re =
0
k0 8n 1
0

(11.27)

Note that if n0 = 1, then Equation 11.27 reduces to the Newtonian definition of Reynolds number. Using this generalized type of Reynolds number
in Equation 11.26 can only be strictly valid if in fact the fluid is a true power
law fluid. For other fluids, its use will be approximate.
As for the generalized model, the generalized index, n0 , and consistency,
k0 , should be evaluated at the actual wall shear stress that is experienced by
the flow. Again, as this is not known before the calculation, iteration will
generally be necessary.
The form of this generalized Reynolds number is especially convenient
because it is consistent with laminar, Newtonian behaviour. Namely,
Cf =

w
V 2 /2

(11.28)

by definition.
For laminar flow, and using the generalized model, substituting Equation 11.17 into Equation 11.28 above, yields
Cf =

16
Re0

(11.29)

This is the same results that is obtained from laminar, Newtonian flow
theory.
For the generalized model, Dodge and Metzner [25] developed an equation for smooth walled pipes based on the logarithmic resistance of vonKarman (Newtonian) fluids.



n0
4
0.4
1
0 1 2
= 0 0.75 log10 Re f
0 1.2
(11.30)
(n )
(n )
f
If n0 = 1 von Karmans equation results. Equation 11.30 has been plotted
in the chart form in Figure 11.2.
The consistency between the generalized model and Newtonian behaviour
is helpful in determining transition between laminar and turbulent flow. For
Newtonian fluids the transition from laminar to turbulent flow is given by
the lower critical Reynolds number for the breakdown of laminar flow, at
Re 2100, to the commencement of fully turbulent flow at Re 3000.
Data of Dodge and Metzner [25] show critical generalized Reynolds numbers corresponding to the onset of turbulence increases with decreasing values of the flow index, n0 . Re0 critical increases from 2100 at n0 = 1 to 3100 at
n0 = 0.38, but no exact criterion has been established for Re0 critical.

55

Figure 11.2: Fanning friction factor for Newtonian and nonNewtonian fluids

56

12 Physics of Rheology
Solid Suspensions
Slurries can be divided arbitrarily into three categories; thin, thick and maximally packed. Thin slurries are such that the particles can move freely and
independently without any physical constraints from neighbouring particles,
or where the interparticle distance is greater than the mean particle diameter. The extreme case of this is the suspending fluid alone, with zero solids
concentration. Thick slurries on the other hand are where the interparticle
distance is less than the mean particle diameter and the particles experience
physical barriers to flow. Maximally packed slurries represent the other extreme, where the solid particles have no space in which to move and the
slurry is essentially a solid. Figure 12.1 shows these slurry categories.
Movement will only occur when the slurry is less than maximally packed
and the concept of maximal packing fraction, designated M , is an important rheological parameter. At concentrations less than M the material can
and will flow, but the ease of flow depends on the ability of a particle to
move into a vacant space. Thus in scientific analysis the relevant measure
of the amount of material suspended in the liquid is that fraction of space of
the total suspension that is occupied by the suspended material. This is the
volume-per-volume fraction, otherwise known as the phase volume or volume fraction, and not the weight-by-weight fraction that is commonly used
in defining concentration. The reason why volume fraction, designated by ,
is so important is that the rheology to a large extent depends on the space in
which a particle has to move, and on the forces which act on the surfaces of
particles, and not so much on the particle density.

Forces acting on particles


So far it has been assumed that no particle-fluid or particle-particle interactions have been present of course this is not really the case. Generally three
types of forces coexist in flowing suspensions to various degrees. Firstly there
are those of colloidal origin that arise from direct particle-particle interaction
and are controlled by the properties of the fluid, such as polarizability, but
not viscosity. They can result in repulsive forces generated by electrostatic

Figure 12.1: Categories of slurries

57

Figure 12.2: House of cards structure of clay

charges on the particles or from the presence of polymeric or surfactant material on the particle surface. Attractive forces arise from the ever present
London-van der Waals attraction, or from electrostatic attraction between unlike charges on different parts of the particle surface, an example being the
edge-face attraction between clay platelets resulting in the familiar house of
cards structure, as shown in Figure 12.2.
If the net result of these charges is attraction, then the particles tend to
flocculate into larger aggregates, whereas overall repulsion keeps the particles separate and dispersed. Each colloidal force has a different rate of decrease as the distance from the particle increases, and the overall result of the
combination of these forces can be quite complicated. These forces are more
important as the particle size decreases and the effect is almost negligible for
particles much greater than 1 micron in size.
Secondly, there is the ever present Brownian or thermal randomizing forces
which randomize the position and spatial orientation of the particles. Again,
these forces are only really important for sub-micron particles.
Thirdly, there are the fluid-particle forces arising from the viscosity of the
fluid. These viscous forces are proportional to the local velocity difference
between the particle and the surrounding fluid, and affect the overall suspension viscosity accordingly.
The composition of the fluid can also influence particle-particle interactions. Anions or cations dissolved in the fluid can preferentially adsorb on
the surface of the particle and alter the surface charge. This is particularly
true of polyionic species and long chain polymers.
Clearly, the macroscopic rheology is strongly dependent on all these microstructural considerations, and although most mineral slurries have average particle sizes well in excess of colloidal dimensions, there is usually a sufficient number of sub-micron particles for these interactions to be significant.
The presence of isolated particles means deviation of the fluid flow lines and
hence an increased viscosity. At higher concentrations there is more resistance as the particles have to force their way past each other. When particles
flocculate there is even more resistance because not only are the flocculated
particles much bigger, but the flocs enclose and hence immobilize some of the
continuous phase, thus increasing the apparent volume fraction and again
giving a higher than expected viscosity.

58

Figure 12.3: Shear thinning fluid rheogram

Figure 12.4: Deflocculation under shear

Rheograms of suspensions
Shear thinning suspensions
Most mineral slurries encountered are either Newtonian of shear thinning,
the latter behaviour predominating in concentrated suspensions. A typical
rheogram of a shear thinning suspension is shown in Figure 12.3.
The first region (a) is a Newtonian region at low shear rate and can be
interpreted as a solid deformation region of high viscosity. As the shear rate
increases, the viscosity decreases and then becomes constant again at high
shear rates, represented by the second Newtonian region (b). In certain situations the viscosity in the first Newtonian region can be so high as to be inaccessible to measurement. In such cases the low shear rate is often described
as an apparent yield stress.
The microscopic interpretation of this behaviour is often visualized in
terms of particle-particle interaction. In the low shear rate region the particles interact to form flocs which occlude a portion of the continuous phase,
thus increasing the apparent volume fraction. At this stage it is a thick slurry,
almost maximally packed and the viscosity is correspondingly very high. The
slurry behaves as an elastic solid and the response is Newtonian deformation.
As shearing increases, the flocs break under the mechanical shearing force, releasing some of the occluded fluid and thus decreasing the apparent volume
fraction. The viscosity consequently decreases until a point is reached when
no more floc structure exists and the particles move freely in a thin slurry.
This process is indicated in Figure 12.4.

59

It is this formation of structure on resting that gives rise to an apparent


yield stress, and the breaking down of this structure that leads to shear thinning behaviour. Both of these effects arise from particle-particle interaction.
If the flocs are chemically destroyed using dispersants, the slurry is not shear
thinning and behaves as a Newtonian fluid (c). On the other hand, a Newtonian suspension can be made shear thinning by the use of flocculants.

Shear thickening suspensions


Shear thickening is characterized by an increase in apparent viscosity with
increasing rate of deformation. It is shown in particular by thick suspensions
of non-interacting particles where the volume fraction approaches the maximum packing fraction, and there is only sufficient fluid available to fill the
voids between the particles. As the shear rate is increased, this dense packing
must be broken down to permit the particles to flow past one another. The
resulting expansion leaves insufficient fluid to fill the voids and is opposed
by surface tension forces. This explains why wet sand apparently becomes
dry and firm when walked upon. Coarse suspensions of pure silica particles
show a degree of shear thickening at high concentrations.

Time dependent effects


In the above consideration it was assumed that the flocculation-deflocculation
processes were instantaneous and depended only on the rate of shear. Many
mineral slurries do behave in this way, but some, particularly finely divided
clays and materials which have been artificially flocculated with long-chain
polymers, exhibit a property known as thixotropy, where the restructuring
process is not instantaneous and is time dependent. Thus the slurry will flow
easily after shearing and will remain thinned after the shearing force has been
removed. Rethickening only takes place after some time has elapsed (which
may be in the order of weeks). Thixotropy is particularly important in the
paint industry, as it is desirable that the paint should flow only when brushed
onto the appropriate surface (high rate of shear); and must stay fluid long
enough for the brush marks to level out before the paint re-gels.

Irreversible phenomena
Shearing sometimes leads to an irreversible breaking of linkages between the
structural elements of a material (rheodestruction), which results in a permanent shear thinning. This is particularly common with suspensions which
have been treated with high molecular weight polymers.

Viscosity
An important effect of the addition of solid particles to a fluid is its influence
on the system viscosity. The presence of the particles invariably increases
the suspension viscosity to a value greater than that of the fluid itself and in
many cases results in a suspension which is non-Newtonian in character.
60

Here, the case of the viscosity of suspensions of uniform spherical particles will be considered, which are usually Newtonian in character. This type
of system is in practice relatively uncommon, but it is of importance since its
viscosity represents a minimum value for the viscosity of non-Newtonian
suspensions.

Dilute suspensions
The viscosity of dilute suspensions of solids was the subject of a theoretical
study by Einstein in 1905, who concluded that the viscosity could be represented as follows
= 0 (1 + 2.5 )
(12.1)
Here represents the concentration by volume, Cv expressed as a fraction. Equation 12.1 applies to laminar flow of suspensions of rigid spheres in
which particles are large compared to molecular dimensions, but small with
respect to the characteristic length scale of the measuring instrument. It is
also assumed that the suspensions are extremely dilute ( < 0.01); that is,
that there is no particle interactions.

Concentrated suspensions
The viscosity of more concentrated suspensions must take into account the
various types of particle interactions that may occur. Thomas [26] reviewed
the existing theoretical and experimental expressions for suspension viscosities. Many take the form of a Taylor-series expansion,

= 0 1 + k1 + k2 2 + k3 3 + ...
(12.2)
The value of k1 is generally assumed to be that determined by Einstein,
i.e. 2.5. Various values have been assigned to k2 by accounting for certain
interactions: Guth and Simha [27] obtained a value of k2 = 14.1. The higher
order coefficients are not readily calculable and according to Thomas, this
leads to errors in excess of 10% for suspension concentrations greater than
= 0.20. Thomas [26] suggests an equation which is closed in form, (and
plotted in Figure 12.5)

= 1 + 2.5 + 10.05 2 + 0.00273e16.6


0

(12.3)

Equation 12.3 is a useful average of results from a number of experimental


studies using deflocculated monosize spheres.
Another form of equation which recognizes the importance of the maximum packing fraction, M is



1
= 1
(12.4)
0
M
This expression holds for uniform spheres. For irregularly shaped particles entropy considerations require the vacancy fraction term to be raised to
the power of -3, although in practice it may be best to determine the exponent
experimentally.
61

Figure 12.5: Effect of volume fraction on viscosity

Influence of particle-particle interactions


Attractive electrostatic interaction between particles causes deviation from
Newtonian behaviour, arising from the formation of flocs and subsequent increase in apparent volume fraction. This interaction can arise naturally if the
slurry is composed of at least two different solid components that have opposite surface charges. The surface charge on a particle can be measured and
is referred to as the zeta potential, . Common examples of this effect occur in
certain gold ores which are generally composed of mixtures of silica and iron
oxide, among other things. Around neutral pH conditions, silica particles
carry a negative surface charge, whereas iron oxide particles carry a positive charge, and the resulting interactions cause deviation from Newtonian
behaviour and the appearance of a yield stress. Interactions can also be induced artificially by the addition of flocculants. These usually consist of ions
carrying multiple charges which attract particles carrying opposite charges.
Quite often these are very high molecular weight polymeric ions which physically string the particles together. These invariably increase the yield stress
of the slurry because the binding of the solid particles enclose and immobilize part of the suspending fluid and effectively increase the apparent volume
fraction. A practical problem arising from this is irreversible shear thinning
owing to the mechanical breakdown of the flocculant molecules. These longchain polymers are fragile molecules, and are particularly vulnerable when
bound to slid particles which are under shearing stresses. These stresses are
sufficiently large to break down the molecule into smaller units which diminishes or destroys its flocculating ability.
The reverse of flocculation can be accomplished using dispersants. These
are small molecules carrying multiple charges, or surfactant molecules with
hydrophobic and hydrophilic regions. The mechanism by which they act is
to bind tightly to the solid particle and thus alter the surface charge or zeta
62

Figure 12.6: Polyphosphate anion

potential, such that all the particles in the suspension have similar charges.
This will ensure that only repulsive charges exist between the particles and
they effectively repel one another. Such compounds are used as viscosity
modifiers in many industrial applications as they effectively reduce the yield
stress and cause a return to Newtonian behaviour. An example of this is the
shear thinning of iron oxide-silica slurries using sodium tripolyphosphate
(Figure 12.6), a highly charged polyanion.
Even simple ions can act as flocculants or dispersants to some extent. A
common example of this is the hydroxide ion, which although singly charged,
is able to alter the surface charges on particles if the concentration is sufficiently high. Thus an aqueous slurry of silica and calcium carbonate will have
a decreased yield stress as the pH of the suspension is increased. This can be
confirmed by zeta potential measurements which show a decrease in with
increasing pH, changing sign from positive to negative at about pH = 10.7.
Above this pH the slurry shows minimum yield stress and is almost Newtonian.

63

13 Rheometry
The most basic measurement of a fluids rheology is a rheogram. Methods
for obtaining a rheogram come under the heading of rheometry. Since the
concept of viscosity applies only to Newtonian fluids, a viscometer should
perhaps more generally be referred to as a rheometer. A mathematical relation used to describe an experimentally determined rheogram is known as a
rheological model. But still more terminology is required.
Since a rheogram is a plot of shear stress against shear rate, it may also be
called a shear diagram, as shown in Figure 13.1(a). Thus if experimental data
is obtained in the form of shear stress at the wall, w , versus an average shear
rate, 8V /D, then a plot of this data is called a pseudo-shear diagram, as shown
in Figure 13.1(b).
A pseudo-shear diagram can sometimes be converted to a rheogram and
vice versa. An example of this is for a Newtonian fluid, for which the pseudoshear diagram and the rheogram are identical. Another example is for a pseudoplastic, for which it is apparent that
 

8V
3n + 1
w =
(13.1)
4n
D
These two examples are special cases of the Mooney-Rabinowitsch transformation (see Mooney [28]), which states that for laminar flow, the shear
stress at the wall w is proportional to the average shear rate 8V /D according
to
 0
 
3n + 1
8V
(13.2)
w =
0
4n
D
where

d(ln w )

n0 = 
8V
d ln
D

(13.3)

In this transformation, the index n0 is not required to be a constant as in


the case of a power law fluid. In applying the Mooney-Rabinowitsch tech-

(a) Rheogram

(b) Pseudo-shear diagram

Figure 13.1: Rheological diagrams

64

Figure 13.2: Primary output of a tube type viscometer

nique, the first step is to plot the pseudo-shear diagram on logarithmic coordinates. The slope of the plot in a given region of shear rate gives the value
of the index n0 to use in applying the transformation equation above. It is of
course possible to argue in reverse and extract a pseudo-shear diagram from
a rheogram. Note however that the Mooney-Rabinowitsch transformation is
based on a truncated Taylor series approximation of the rheogram and the
implications of this are not well understood.
In summary then, rheological models are based on the rheogram, this being the more fundamental of the two shear diagrams. If however, a pseduoshear diagram is the primary data available, then the parameter values for a
rheological model may be obtained either by
1. transforming the diagram to a rheogram as described above and curve
fitting, or
2. by using the integrated relationships, ie Equations 11.12 and 11.16 from
Section 11.

Tube type viscometers


Tube type viscometers are based on Poiseuille flow. These viscometers are of
particular value because of their geometric similarity with the pipeline which
is the final object of the study. Thus tube viscometer data can be used directly
in scale-up methods without the intermediary of a rheological model. This
is attractive since rheological models only approximate the rheogram, and
many fluids do not easily fit two parameter models, such as the power law
or Bingham plastic models.
The primary output of the tube viscometer is a pseudo-shear diagram. A
typical example is shown in Figure 13.2. Notice that the transition to turbulence is well defined and that below transition, the results should be independent of tube diameter. From a rheological point of view, the results in the turbulent region are meaningless. From a pipe scale-up point of view however,
the turbulent results are very useful and the provision of this information is
one of the primary advantages of the tube type viscometer.
One experimental difficulty with tube viscometers is the presence of effective slip. In the region near the wall, the higher velocity gradients produce
a Magnus type effect which drives particles towards the centre of the pipe.
65

Figure 13.3: Force balance in a coaxial cylinder viscometer

This depletion of particles near the wall gives a lower apparent viscosity in
this region and effective slip. To see whether effective slip is significant or
not, it is necessary to experiment with various tube diameters and check for
agreement in the laminar flow region.
While the tube viscometer should be favoured for pipeline design applications because of its geometric similarity, several difficulties with its use
should be noted. The first difficulty is that testing is slow several hours
at each shear rate and tedious. Secondly, it is difficult to maintain constant
test conditions over the long test period resulting in a low level of reputability.
This is particularly evident when testing a fluid at an elevated temperature
for instance. Finally, a relatively large sample of slurry is required compared
with that required for a coaxial cylinder viscometer, which is discussed next.

Coaxial cylinder viscometer


An advantage of the coaxial cylinder viscometer is that a wide range of shear
rates, including very low values, can be tested. When combined with the
considerations that the instrument may easily be automated, that temperature control is readily achieved and that only small samples are required, it is
apparent that these viscometers are very convenient.
The primary output of a coaxial cylinder viscometer is a plot of applied
torque, T , against angular speed of the bob, , which must be reduced to the
form of a rheogram. The actual method of data reduction will vary depending on the gap size that is used. The importance of the gap is evident from
a force balance. With reference to Figure 13.3, and neglecting end effects, the
torque T is resisted by the shear stress, at some location r such that
=

T
2hr2

(13.4)

This indicates that the shear stress, and hence shear rate, varies across the
width of the gap.
For small gaps, the shear stress across the gap is relatively constant and
the flow approaches that of Couette flow that of flow between two parallel
66

Figure 13.4: Results from a coaxial viscometer test on a Bingham plastic

sliding plates. For this case, no prior assumption need be made about the
rheology of the fluid being tested and the analysis reduces to
= app = app

ravg
ro ri

(13.5)

When the gap between the bob and the cup is significant, the shear stress,
and hence shear rate, can no longer be considered constant. The end result is
that Equation 13.5 must be replaced by a rheological model and an integration performed. The classical results for various rheological models are given
as
For Newtonian fluids, the Margules equation



1
1
T
(13.6)

=
4h
ri2 ro2
For Bingham plastic fluids, the Reiner-Riwlin equation



1
T
1
1
o ro
=
ln

4h
ri
ri2 ro2
and for power law fluids, the following

1 "
 2 #
n
n
T
ri n
=
1

2 2ri2 hk
ro

(13.7)

(13.8)

These equations illustrate the importance of gap size and the limitations
of the coaxial viscometer. Namely that for a large gap, either the Couette
flow assumption will breakdown or an a priori assumption must be made regarding the fluids rheology. Note also that special care must be taken when
testing Bingham plastics. In testing a Bingham plastic at low shear rates the
results shown in Figure 13.4 may be obtained. In this instance, the curvature of the rheogram at low shear rates indicates that the fluid is not shearing
across the whole gap. The Reinler-Riwlin equation could be used in this instance to check at what angular speed shearing across the whole gap would
begin. Data obtained below this speed could then be disregarded.
There are a few practical problems with the coaxial cylinder viscometer.
Particle segregation in the radial direction may occur. Calibrations are required to take into account the end effects of the cup and the bob. Wall slip as
67

observed in tube type viscometers may also occur to some extent in coaxial
cylinder viscometers. It is also worth reinforcing that the coaxial cylinder viscometer can only provide meaningful results while the flow within the gap
remains laminar.

68

14 Turbulent Boundary Layers


Most slurry pipelines operate in the turbulent regime. Boundary layer theory
becomes important for several reasons, including
much work is based on the smooth pipe assumption which must be
understood within the context of a turbulent boundary layer.
scale-up methods for turbulent flow are based on similarity of the boundary layers at the pipe wall position.
the equations presented graphically in the Moody diagram are derived
from boundary layer theory.
For these reasons boundary layer theory is again presented here, although
similar treatments can be found in almost any elementary fluid mechanics
text.

Boundary Layer Velocity Profile


The presence of fine-scale turbulence gives rise to momentum convection
across the boundary of an infinitesimal control volume within a boundary
layer. Expressed mathematically within the momentum equation for the control volume, these momentum transfers have the dimensions of stress and occur alongside the Newtonian viscous stresses. Consequently they are called
turbulent stresses or Reynolds stresses.
Figure 14.1 shows the distribution of lam and turb from typical measurements across a turbulent boundary layer. Laminar shear is dominant near
the wall (the wall layer or the viscous sub layer, and turbulent shear dominates
in the outer layer. There is an intermediate region, called the overlap layer,
where both laminar and turbulent shear are important.
In the outer layer turb is two or three orders of magnitude greater than
lam , and vice versa in the wall layer. These experimental facts enable a crude
but very effective model for the velocity distribution u across a turbulent
boundary layer to be used.
For the wall layer, Prandtl deduced in 1930 that u must be independent of
the boundary layer thickness, , and
u = f (, w , , y)

(14.1)

Figure 14.1: Typical velocity and shear distributions in a turbulent boundary layer

69

Using non-dimensional terms, this is equivalent to


u+ = f (y+ )
u
u+ =
U
r
w

U =

yU
and y+ =

where

(14.2)

is the dimensionless flow velocity


is the friction velocity
is the dimensionless distance from the wall

Equation 14.2 is called the law of the wall, and the quantity U is termed
the friction velocity because it has the dimensions of velocity, although it is not
actually a flow velocity. A specific form of the law of the wall for the laminar
sub-layer is
u+ = y+
(14.3)
Subsequently, von Karman in 1933 deduced that u in the outer layer is
independent of molecular viscosity, but its deviation from the stream velocity U must depend on the boundary layer thickness, , and the remaining
properties in Equation 14.1
(U u)outer = f ( , w , , y)

(14.4)

Again, in non-dimensional form


y
U u
=
f
U

(14.5)

Equation 14.5 is called the velocity defect law for the outer layer.
Both the law of the wall and the defect law are found to be accurate for a
wide variety of experimental turbulent duct and boundary layer flows. They
are different in form, yet they must overlap smoothly in the intermediate
layer. In 1937, Millikan, a student of von Karman, showed that this can be
true only if the overlap region varies logarithmically with y


1
yU
u
= ln
+B
(14.6)
U

Equation 14.6 is called the logarithmic overlap layer. A good summary of


the dimensional analysis arguments used to derive this is given by Massey
[29]. Over the full range of turbulent wall flows, the dimensionless constants
and B are found to have the approximate values of = 0.4, and B = 5.5.
Thus by dimensional analysis and physical insight it is inferred that a plot
of u versus ln y in a turbulent shear layer will show a curved wall region, a
curved outer region, and a straight line logarithmic overlap region, as shown
in Figure 14.2. Although the inner wall law, Equation 14.3, appears only to
apply to approximately, y+ = 5 , for many practical purposes, it is sufficient
to assume that the viscous sub-layer extends all the way to the point where
Equation 14.3 and Equation 14.6 intercept, at y+ = 11.6.
The logarithmic scale used in Figure 14.2 belies the importance of the overlap region. The logarithmic law, instead of being a short overlapping link,
actually approximates nearly the entire velocity profile, except for the outer
70

Figure 14.2: Experimental verification of turbulent velocity profiles

law when the pressure is increasing strongly downstream, as in a diffuser


flow. The inner wall law typically extends over less than 2% of the profile
and can usually be neglected. Thus, Equation 14.6 can be used as an excellent approximation of the whole boundary layer in many practical boundary
layer problems.

Application to pipe flow


For turbulent pipe flow, it is reasonable to assume that the flow approximates
that of a turbulent boundary layer with a thickness, , equal to the pipe radius, r. Starting with von Karmans velocity profile for a turbulent boundary
layer


yU
u
= 2.5 ln
+ 5.5
(14.7)
U

it is relatively simple to integrate over the pipe area to show that the average
velocity for turbulent flow in a pipe is given by


V
DU
= 2.5 ln
(14.8)
U

which is known as von Karmans average velocity result.


Rearranging this in terms of friction factor give yet another form of von
Karmans equation,
 p 
1
= 4 log10 Re f 0.6
(14.9)
f
This will be recognised as very similar to the Colebrook Equation, upon
which the Moody diagram is based. The agreement between von Karmans
Equation, Equation 14.9, and the experimental results presented in the Moody
diagram is not unexpected. The form of von Karmans equation was derived
theoretically from dimensional analysis, but the constants involved were obtained from experiments which included pipe flow experiments.
Another application of the preceding theory with respect to pipe flow is
that of surface roughness. It was previously stated that the laminar sublayer
may be assumed to exist up to the position, y+ = 11.6. If it is known that
roughness effects are of minor importance in laminar flow, then this thickness
of the laminar sublayer allows an estimate of the value of surface roughness
71

below which the pipe may be considered hydraulically smooth. Specifically,


when the surface roughness of the pipe wall increases to the same order of
magnitude as the thickness of the laminar sublayer, the pipe may no longer
be considered smooth. This is an important point since it can be appreciated
that most of the preceding material has assumed that the pipe is hydraulically
smooth.

72

15 Scale up methods for homogeneous


flow
As opposed to heterogeneous flow, which may be deposition controlled, homogeneous or non-settling flow is very often transition controlled. That is
to say, it is desired to transport the slurry close to the transitional point or
just within the turbulent flow regime to guarantee that the solids have no
tendency to settle. With reference to Figure 15.1, the slope of the friction loss
versus velocity curve is typically very flat in the laminar region. Operation
too far into the turbulent region will generally incur an uneconomical pressure drop and hence the transition point is an important design parameter in
homogeneous slurry flow.
So far, various methods for evaluating both laminar and turbulent pressure losses have been presented, as have methods for determining the transition point between the two. It is well to note some of the limitations inherent
in the methods covered so far. These limitations provide some of the justification for the scale-up methods that are presented in this section. The scaleup method is recommended by Wilson et al. [30] as being the most accurate
method. Note however that the method is relatively capital intensive since,
despite the scale-up process, the test equipment is generally large.

Laminar flow
As an application of the material covered in Section 11, once the fundamental
properties of the slurry are known in terms of a rheological model, it is a
trivial matter to use the appropriate integrated equation to evaluate the wall
shear stress and hence the pressure drop in the pipe1 . While this method is
reasonably simple, some shortcomings of the method exist. These are:
The Buckingham equation is only easy to use in its truncated form
which results in some loss of accuracy. To overcome this, some graphical representations of the full Buckingham equation are available, for
example in Wasp et al. [9].
The method implies that a rheological model applies in the first place,
whereas the simple two parameter rheological models do not always
accurately represent the slurry behaviour. The use of the generalized
method for these situations can introduce errors.

Turbulent flow
The generalized method has the attraction of being able to use the data from
a rheogram, easily available from a cylinder viscometer, without the need for
tedious tube testing. The price of this convenience however may be accuracy,
as:
1 such

as the Buckingham equation for laminar flow of a Bingham plastic

73

this method does not allow for hydraulically rough pipes,


the data does not have extensive experimental verification, and
it is questionable whether the laminar flow information from a concentric cylinder viscometer will accurately translate to the boundary layer
of a turbulent pipe flow.

Laminar flow scale-up


The tube viscometer is geometrically identical to a pipeline. This makes it
possible to directly scale the test results (ie. the pseudo shear diagram) from
the model to the prototype. This method avoids the intermediary of a rheological model. To make scale-up simple the average shear rate, 8V /D, is
maintained as a constant between the model and the prototype, thus allowing eff (which will now be constant due to the unique relationship between
eff and 8V /D) to be treated as a simple fluid property from a dimensional
analysis point of view. By equating the groups for the prototype or alternatively, from the definition of effective viscosity
w = eff

8V
D

(15.1)

it follows that the wall shear stress in the test model and in the full size pipe
must be equal. Thus the full size pipe pressure drop can be determined for
the particular 8V /D considered.
Note that it is also possible to represent this scale up information on a
Moody type diagram, making the Moody type diagram an indirect scale-up
method. For the Moody type diagram, Re may be evaluated on the basis of
= eff , which must be found from a test at the particular shear rate being
considered. The generalized method of Metzner and Reed [24] provides an
example of such a diagram (see also Section 11).

Turbulent flow scale-up


At this stage an important result of turbulent boundary layer theory is revisited, namely that viscous stresses are only significant within the viscous
sublayer and to a diminishing extent in the logarithmic overlap layer. If the
bold simplifying assumption is made that viscous stresses are only significant
in the viscous sublayer, then the scale-up problem for turbulent flow becomes
tractable. The wall shear stress is given by
w = app w

(15.2)

and provided that the correct w is applied, then the slurry may resemble a
Newtonian fluid with a viscosity of = app . This has encouraged the use of
various Moody type diagrams where = or = eff . Wasp et al. [9] gives
other examples of this approach besides the generalized method.
As for laminar flow before, the scale-up method avoids the intermediary of a rheological model. The shear rate at the wall, w is maintained as
74

a constant between the model and the prototype thus allowing app (which
will now be constant) to be treated as a simple fluid property. Invoking von
Karmans average velocity result across a hydraulically smooth pipe


V
DU
= 2.5 ln
(15.3)
U

allows the scaling law to be obtained,


V2 = V1 + 2.5U ln

D2
D1

(15.4)

A final comment on the generalized method of Dodge and Metzner [25]


can now be made in the light of the scale-up method presented. Although
the parameters k0 and n0 are allowed to vary with the shear rate in the generalized method, it is questioned what shear rate applied in evaluating these?
Applying the logic used in scale-up suggests that the shear rate prevailing in
the viscous sublayer would be the most appropriate, and this is indeed what
was prescribed in the earlier sections.

Transition to turbulence
Similarities between Newtonian and non-Newtonian slurries in turbulent
flow suggest that a type of Reynolds number criteria where = app would be
appropriate for describing the transition to turbulence for a non-Newtonian
fluid. In practice however, it is more common to use a Reynolds number criteria based on = eff . A common example of this is the criteria of Thomas
[26] for a Bingham plastic


y D
V D
= 2100 1 +
(15.5)

6V
The generalized method of Dodge and Metzner [25] is another example
more suited to a power law fluid (see Section 11).
The transition point can be predicted from small scale tests by conducting
tests in both the laminar and turbulent regions. Upon scale-up, the two sets
of results will intersect, or can be extrapolated to intersect, at a point which
is the predicted transition point for the scaled-up flow. An illustration of this
technique is given in Figure 15.1. Here the turbulent test data has been scaled
to a 200 mm pipe size and the transition point between laminar and turbulent
flow for this pipe size has been estimated by extrapolation.

75

Figure 15.1: Illustration of test data scaled to full pipe size

76

16 Pipe Flow of Heterogeneous Slurries


The flow of heterogeneous slurries is a very complex subject; not easily modelled by mathematics, nor is the physics of the flow completely understood.
The treatment presented here is based on the simple and widely used, if not
accurate, method of Durand [31] together with enhancements subsequently
introduced by Wasp et al. [9]. Attention will be limited to the case of a horizontal pipe and the carrier fluid will be assumed Newtonian. Indeed in Durands original work, the carrier fluid was water. With these limitations, this
section can only be considered as an introduction to the topic as a whole.

Distribution of Particles in Turbulent Flow


Suspension mechanisms for particles are many and varied. For small particles, the predominant mechanism is their entrainment within the turbulent
eddies of the flow. For particle sizes less than 75 to 100 m, the settling velocity is so small that slurries comprised of these particles may be classified as
non-settling or homogeneous.
For larger particles, the mechanisms of lift and collision become significant. Particles may experience lift due to the Magnus effect when they acquire rotation from collision with the pipe wall or other particles. Particleparticle and particle-wall collisions also provide a net upward force serving
to suspend particles.
At an extreme of heterogeneous flow is a sliding or stationary bed of particles deposited on the floor of the pipe. In this situation inter-particle contact
loads provide the dominant support mechanism. Bernoulli effects also come
into play in lifting particles from the surface of the bed of solids in saltation
flow.
Regardless of the mechanisms present, if the distribution of solids across
a pipe cross-section is steady, then it may be assumed that at all points in the
flow the rate of upward transport of solids is equal to the rate of downward
transport of solids. Observing that generally there is a characteristic concentration gradient across the pipe cross section suggests that the upward
transport can be described as a diffusion process.
Mathematically the equation will have the form
ES
where

Es
C
y
and Vs

dC
+VsC = 0
dy

(16.1)

is the mass transfer coefficient for the particles


is the concentration by volume at some height y
is the height from the bottom of the pipe
is the settling velocity of the particles in a still fluid.

Integration of this equation should provide the basic form of an equation


which describes the distribution of solids across the cross-section of the pipe.
It is still unknown however what actual mass transfer coefficient, Es , should
77

be used. A semi-emprical approach provided by Wasp et al. [9] reasons that


Es should be similar to the momentum transfer diffusion coefficient, turb ,
which is closely related to the friction velocity, u . While the process of obtaining the final distribution is fully described by Wasp et al. [9], the final
result is simply
1.8Vs
C
=
(16.2)
log10
CA
u
where

and

CA
C

is the von Karman constant (from section 14)


is an empirical constant the ratio between the mass and momentum transfer coefficients
is the concentration at the centre of the pipe
is the concentration at a distance of 0.08 diameters from the
top of the pipe
(These locations correspond to the probe locations used in the
original testing for this relationship).

It is known that the presence of particles in the fluid suppresses the von
Karman constant from its clear water value of 0.4. There is also considerable doubt over the correct value of the constant . Values of 0.4 and unity,
respectively, give the generally conservative result
log10

Vs
C
= 4.5
CA
u

(16.3)

This result provides a convenient way of quantifying the nature of the


slurry flow. Values of C/CA greater than 0.8 may be regarded as an indication
of strong homogeneity, while flows with C/CA below 0.1 may be regarded as
heterogeneous. The result also has use later in providing a means for proportioning some particles to the carrier fluid and some to the particle phase.

The Durand Method


The method of Durand [31] models the heterogeneous slurry flow as a twophase flow comprising of the phases
1. Carrier fluid
2. Solids carried by the carrier fluid
Durands model determines the pressure losses of the carrier fluid independently of any solids and then adds an over-pressure, or additional pressure
loss, to account for the effect of the solids phase.

Pressure losses in the carrier fluid


For Durand, the carrier fluid was water and evaluation of the carrier fluid
losses was a simple matter of applying the Darcy equation with a friction factor supplied from the Moody diagram or similar. In practice however, the
carrier fluid is not always best represented by clear water. The addition of
fines may present a situation where the carrier fluid is best modelled as an
78

equivalent fluid, with the density and viscosity of this fluid determined by the
amount of fines which are in homogeneous type suspension. These modifications, proposed by Wasp et al. [9] are discussed below. In the unmodified
Durand model the carrier fluid losses are determined as discussed in Section 1.

Pressure losses due to the particle phase


To determine the pressure losses in the particle phase an additional overpressure is calculated. Durand presents the over pressure in terms of the
function
= 82 1.5
(16.4)
and

fslurry fcarrier
fcarrierCv

(16.5)

p
V2
carrier
CD
gD solid carrier

(16.6)

=
where is calculated as follows
=

where V is the flow velocity, D is the diameter of the pipe, CD is the drag
coefficient of the particles and carrier and solid are the densities of the carrier
fluid and solids respectively.
Alternatively, Equation 16.5 can be rearranged to give the total friction
factor of the slurry, i.e.
h
i
1.5
fslurry = fcarrier 1 + 82Cv
(16.7)
If more than one size of solid is present in the slurry, then this may be
modified to
h
i
fslurry = fcarrier 1 + 82Cv,1 11.5 + 82Cv,2 21.5 + ...
(16.8)
where Cv,1 to Cv,n are the volume fractions of each size, and
Cv,total = Cv,1 +Cv,2 + ... +Cv,n

(16.9)

This friction factor is then applied in the Darcy equation to give the pressure loss, or
L slurryV 2
P = fslurry
(16.10)
D
2

The Durand-Wasp Method


The work of Wasp et al. [9] in arriving at Equation 16.3 enabled them to propose a modification to Durands basic method. They proposed that for each
size fraction of particles in a slurry, a certain proportion of these particles, as
indicated by the C/CA ratio for that size fraction, could be considered as being in homogeneous suspension and therefore part of the carrier fluid. The remainder of the particles are treated as being in the solid phase and are treated
as per the Durands method above.
79

The proportioning between the carrier and solid phases is done using the
C/CA ratio as is. Simply, if a size fraction had a C/CA ratio of 0.65, then 65%
of the particles in this size fraction would be proportioned to the carrier fluid
and 35% would be included in the solid phase.
The method is potentially confusing from the respect that there are now
three phases that must be considered:
1. clear water
2. the carrier fluid; and
3. the slurry
The viscosity of the carrier fluid is determined using the equivalent fluid
model, Equation 12.3, and is repeated here for convenience
i
h
16.6Cv,carrier
2
(16.11)
carrier = water 1 + 2.5Cv,carrier + 10.05Cv,carrier + 0.00273e
and the density becomes
carrier = water +Cv,carrier (solid water )

(16.12)

In the above equations, Cv,carrier , is the total amount of solids that is proportioned to the carrier, i.e. the sum across all of the particle size fractions.
The over pressure terms for the remaining particles in each size fraction
are calculated using Equation 16.5, and a total friction factor for the slurry
is determined from Equation 16.8. The carrier friction factor used in Equation 16.8 is calculated using the properties of the carrier fluid determined
from equations 16.11 and 16.12.

Application of the Durand-Wasp method


Implementation of the method as described above is necessarily iterative, due
to the C/CA ratio being dependent on the friction velocity, and in turn the
friction factor. An outline of the process, step-by-step, is given below.

Initial assumptions
Initially none of the solids are proportioned to the carrier phase, and the carrier properties are those of clear water. Also, it is assumed that the flowrate
and pipe size are known.
These assumptions allow the pipe velocity, V , to be determined and an
initial Reynolds number to be calculated
Rec =

cV D
c

(16.13)

where, c and c are initially the values for water, and the subscript c denotes
the carrier fluid.

80

Next, the friction factor for the carrier, fc , can be determined. This can be
found using the methods described in Section 1, though for convenience here
the expression for smooth pipes will be used
fc = 0.184Rec0.2

(16.14)

The last step prior to iteration is to determine an initial value for the friction velocity
r
r
w
fc

u =
=V
(16.15)
c
8
It is now possible to start the iterative part of the process.

Iteration
1. Calculate the drag coefficient for the particles, CD . This can be done using any of the methods described in Section 5. The archimedes number
correlations are usually the easiest to implement.
2. Calculate the settling velocity for the particles
s
( p c )
4
Vs =
D pg
3CD
c
where p is the density of the solids.
3. calculate the C/CA ratio

C
= 104.5(Vs /u )
CA

4. Determine the solids concentration in the carrier


Cv,c =

C
Cv
CA

5. Calculate the new carrier properties, c and c




2
c = water 1 + 2.5Cv,c + 10.05Cv,c
+ 0.00273e16.6Cv,c
c = water +Cv,c ( p water )
6. Calculate the new Reynolds number
Rec =

cV D
c

7. Calculate the new carrier friction factor, using the Moody diagram, or
for smooth pipes
fc = 0.184Re0.2
c

81

8. Calculate the new friction velocity


r
u = V

fc
8

9. Calculate the Durand coefficient for the particles in the solid phase
=

V 2 c p
CD
gD p c

10. Calculate the slurry friction factor


h
i
1.5
fslurry = fc 1 + 82(Cv Cv,c )
The above steps are iterated until the values converge, usually after 5 or so
iterations.

Pressure loss for the system


Finally the pressure loss for the system can be determined. The Durand-Wasp
method is developed based on the carrier fluid properties, so the density of
the carrier must be used, i.e.
P = fslurry

L slurryV 2
D
2

(16.16)

Changes for multiple particle size fractions


The steps outlined above are for a uniform size of particle. The method is
easily modified to allow for varying particle sizes, as follows.
Firstly, for steps 1 to 4 are performed for each size fraction. Each size
fraction now has a proportion allocated to the carrier phase and a proportion
in the solid phase.
For the carrier phase, the proportions from each size fraction are added
together to give the total proportion of solids in the carrier phase
Cv,c(total) = Cv,c,1 +Cv,c,2 + ... +Cv,c,n

(16.17)

This total amount is used to calculate the new carrier fluid properties in step
5.
For the remaining solids in each size fraction the Durand coefficient is calculated for each separately (step 9), and the slurry friction factor is then calculated from Equation 16.8, suitably modified for the Durand-Wasp method.
h
i
fslurry = fcarrier 1 + 82(Cv,1 Cv,c,1 )11.5 + 82(Cv,2 Cv,c,2 )21.5 + ... (16.18)
This friction factor is then used to calculate the slurry pressure drop, as
before
L slurryV 2
(16.19)
P = fslurry
D
2
82

Figure 16.1: Pressure loss versus flow velocity for DurandWasp method

Pressure loss versus flowrate


For a uniform particles, calculating the pressure loss at various flowrates results in a curve such as that shown in Figure 16.1. At low velocities the pressure loss predicted by the Durand-Wasp method tends to infinity. This represents the situation where the solids in the pipe have settled and eventually
plugged the pipeline. At higher velocities the overpressure due to the solids
phase reduces and the slurry tends to behave as an equivalent fluid, with a
similar slope on the pressure-velocity plot as a pure fluid (indicated for water
in Figure 16.1). This characteristic of slurries was not taken into account be
the original Durand model, but is captured in the Durand-Wasp model.
Between these two extremes other characteristics are observed. As the
velocity is decreased from an initially high velocity, the uneven distribution
of solids in the pipeline becomes more pronounced until, at the curve minimum, a stationary or sliding bed of particles starts to deposit in the pipe. The
velocity at which this occurs is called the deposition velocity. For systems consisting of uniformly sized particles, the deposition velocity coincides with the
minimum of the pressure loss-velocity curve. If the velocity is reduced below
this minimum, the bed of solids may build up in the pipe and the friction loss
increases due to the restriction in available flow area. This region of flow can
be relatively unstable so that the exact location on the pressure loss-velocity
curve may be time dependent.
Unlike the transition velocity of a homogeneous suspension, which corresponds to the break between laminar and turbulent flow, the deposition
velocity is strictly a turbulent phenomenon. At the deposition velocity, the
tendency of the particles to settle under gravitational forces just exceeds the
turbulence forces acting to maintain the particles in suspension. The basic
design information is therefore not only the pressure loss-velocity response
of a given system, but also the deposition velocity to be expected. In commercial terms, operation below the deposition is impractical. Apart from the
obvious danger of plugging the pipeline, it also produces excessive erosion
in the lower part of the pipe. Additionally, for a system containing particles
of a wide granulometric range it can result in the larger particles being deposited in the line while the smaller particles remain in suspension. In effect
the pipeline behaves as a classifier. Thus, the long-term stability of a heterogeneous flow depends on a reliable prediction of the deposition velocity.

83

Figure 16.2: Variation of FL with particle diameter

Deposition Velocity
The characteristic curve shown in Figure 16.1 suggests that the deposition
velocity loosely corresponds to the location of the curve minimum. Indeed
an approach similar to this was used by Durand to determine a correlation
for the deposition velocity, Vd The correlation is given by
r
p c
(16.20)
Vd = FL 2gD
c
where the factor FL is determined from Figure 16.2
Many other methods for determining the deposition velocity are available, and Wilson et al. [30] provide an overview of these. However, Wasp
et al. [9] recommend the use of the Durand correlation as it accurately predicts the deposition velocity for sand, and predictions for other suspensions
are conservative.

84

17 Wear in Slurry Pipelines


Progressive loss of pipe metal (or wear) due to corrosion and abrasion is an
important consideration in pipeline systems affecting capital cost and the life
of the system. Prevention of this wear depends on a proper understanding
of the corrosion and abrasion processes which will now be treated separately,
although in practice they tend to synergize.

Corrosion
Corrosion of metal in an aqueous environment is a complex process as Chilton
[32] explains:
Wet corrosion may take many forms: there may be uniform destruction of the material (as is normally encountered in oxidation);
the corrosion may be highly localized, as in pitting or stress corrosion; or attack may be concentrated at areas adjacent to a second
more noble metal or at points where the oxygen supply is limited.
In contrast to dry corrosion, wet corrosion is an electrochemical phenomenon.
As such, it requires that certain areas of the metal form an anode where the
metal is oxidized and other areas of the metal form a cathode where the oxidant is reduced. A positive current flows through the electrolyte from the
anodic areas to the cathodic areas. Since the majority of slurry pipelines are
made of steel, the corrosion of iron is of particular interest. The reactions at
the anode and cathode being as follows:
Anode:
Fe = Fe2+ + 2e
(17.1)
Cathode:
O2 + 4e + 2H2 O = 4OH
+

2H + 2e = H2

neutral solution

(17.2a)

acid solution

(17.2b)

Hydrogen reduction is the main cathodic reaction in acid solutions where


the metal dissolves with the evolution of hydrogen gas. Oxygen reduction
is the main cathodic reaction in naturally aerated waters. Corrosion may
also occur in caustic solutions, but the reactions are more complex than those
listed above.
The charged iron ions travel through the solution towards the cathode
and they may further react with the cathodic reaction products to form rust.
Formation of rust is actually multistage comprising the formation of ferrous
hydroxide and then the further oxidation of this with oxygen to form a hydrated
ferric oxide, Fe2 03 H2 O, although the structure is better represented by the
formula FeO(OH). These corrosion products are in contrast to the oxide films
(magnetite and haematite) which are produced in dry corrosion. Oxide films
may still occur on a metal immersed in a solution through direct oxidation
with oxygen when present, eg.
4Fe + 3O2 = 2Fe2 O3

(17.3)
85

however, this will not usually be the primary corrosion mechanism.


While the oxide film may not be a primary corrosion mechanism in itself,
its effect on the wet electrochemical corrosion can be profound. An oxide film
may have a passivating effect, whereby the corrosion film forms a barrier to
the diffusion of oxygen to the cathodic region and thus reduces the corrosion
rate. Steel and other metals subject to passivation are sometime vulnerable to
pitting. Pitting describes a localized attack and may occur in regions where
the oxide film is defective. The defect area immediately becomes an anode
due to the easy release of metal ions into solution at these sites. The susceptibility of steel to pitting increases with higher pH. This is due to the increase
in passivity of steel at high pH.
The prevention of corrosion depends on preventing the anodic process
taking place although prevention of the cathodic reaction will have the same
effect. The standard methods of corrosion prevention include:
1. Elimination of oxygen (for neutral solutions)
protective coatings. ie. greasing, painting, epoxy lining, etc.
de-aerating, gas-stripping
cathodic inhibitors
oxygen scavenging additives
2. Prevention of the anodic reaction using an anodic inhibiter
3. Making the steel cathodic (cathodic protection)
use of sacrificial anodes ie. zinc or magnesium
supplying a source of electrons (eg. from a battery or generator) to
the steel.
4. Material selection
plastics, alloys etc.
For slurry pipelines in particular, some specific comments can now be
made. Firstly, it is noted that higher corrosion rates might be expected in
the first few kilometres of a slurry pipeline since the dissolved oxygen level
should reduce with the time of travel (providing there is no fresh oxygen
supply accessing the slurry). Secondly, the most common type of attack is
erosion-corrosion where the rate of corrosion is accelerated by scouring of the
surface and removal of protective oxide or scale films by the solid particles.
With this in mind, corrosion control in long pipelines is often attempted with
oxygen inhibitors which combine with the oxygen to prevent it from reacting
with the pipe. This type of corrosion control generally works best in slurries
which do not have strong ionising potentials, such as coal or limestone slurries. The use of anodic/cathodic inhibitors are also commonly reported in the
literature. Finally, it is worthwhile to remember that corrosion is a complex
problem.

86

Figure 17.1: Abrasion mechanisms

Abrasion
Abrasive wear is a particular type of erosion which results from the impingement of moving particles on a surface. Depending on the flow conditions, the
abrasion can be one of the following three types.
1. Deformation (impact) wear,
2. Cutting (sliding) wear, or
3. Fatigue
These mechanisms are depicted in Figure 17.1. Deformation wear is caused
by the normal impact of solids particles. Some of these have sufficient kinetic
energy to cause local stresses higher than the yield strength of the pipe. An
accumulation of these stresses and the resultant strain lead to surface breakdown of the pipe.
Cutting wear is caused by the oblique impact of solid particles, some of
which have sufficient energy to shear the surface of the wearing material and
gouge a fragment loose.
The abrasion mechanism that dominates may dictate the type of material
that is best suited for the wearing surface. A soft and ductile material is very
useful at larger impact angles where it is able to absorb the kinetic energy
of the particle without plastic deformation. A hard and brittle material is
useful when the impact angle is small. In most slurry pipelines, the angle of
impact is expected to be small, while in pumps and fittings the impact angle
is expected to be greater.
For abrasion to take place, the solids must generally be harder than the
pipe surface. Thus the abrasive wear can be eliminated by making the pipe
wall harder than the solid particles. However, with reference to Figure 17.2, it
is noted that the most common abrasive component is silica which is much
harder than steel.
Abrasion is also governed by the size distribution of the solids, slurry concentration, and flow velocity. In a slurry pipeline these are interdependent to
87

Figure 17.2: Approximate hardness of various ore minerals and


metals

some extent. For example, use of large size solids requires an increase in minimum transportation velocity. It has been found that abrasion increases as the
size of solid particles increases. Thus, by reducing the size of solids, abrasion
can be substantially reduced due to the combination of lower required velocity and reduction in wear due to a smaller particle size. The effect of slurry
concentration on the abrasion is more complicated.
The additional cost of grinding and dewatering of finer solids should be
compared with the savings in pumping cost and abrasive wear before selecting the size to which the material should be ground, though sometimes the
size can be a process requirement. Wasp et al. [9] reports that
From experience it has been found that the metal loss due to
abrasion is quite insignificant if the flow velocity is less than about
3 m/s. For long-distance slurry pipelines, velocity in the range of
1.5 - 2 m/s give economic design. Thus, a particle size should be
selected such that the slurry is nearly homogeneously suspended
at velocities in the range of 1.5 - 2 m/s.
It has been found from experience that when the C/CA of the
maximum size particle in the slurry is greater than about 0.5, then
abrasion is negligible provided the velocity of flow is less than
3 m/s.
When the value of C/CA for the solids is very much lower than
0.5, the abrasivity of the slurry has been found to increase with an
increase in slurry concentration. At low values of C/CA , a bed of
solids slides along the bottom of the pipe and gives rise to abrasion. The value of C/CA is also found to increase with slurry concentration. Thus, the abrasivity of a slurry may reach a maximum,
after which the abrasivity starts decreasing with an increase in
88

slurry concentration. It should be recognized that these criteria


apply only when the velocities of flow are above the critical deposition velocity.

89

18 Centrifugal Slurry Pumps


Effects of solids on pump performance
The presence of solid particles in the flow tends to produce adverse effects
on pump performance, and detailed information on these effects is needed
to achieve reliable and energy efficient operation. Head-discharge curves for
centrifugal pumps are often rather flat. Also, the pipeline characteristics for
slurry flow usually displays a minimum followed by a slow rise in head with
increasing discharge. As a result, the two curve often intersect at a rather
shallow angle. It is important to note that even a small reduction in pump
head can produce a disproportionally large drop in flowrate. Proper pump
selection is critical therefore, but also difficult. The result is that many systems are over designed causing these large and expensive systems to run
inefficiently.
Since published pump ratings are generally based on pumping clear water, it is desirable to use the pump similarity laws to convert the ratings for
operation on some given slurry. Two difficulties arise in this application of
the similarity laws, however. Firstly, the standard laws do not take into account viscous effects, which can be very important when handling viscous
slurries1 . Secondly, slurries, especially those comprising of larger particles,
do not behave as an equivalent fluid.
Even so, application of the similarity laws gives results as illustrated in
Figure 18.1(a). If only the fluid density is changed, then for a given discharge,
for the same pump and rotational speed, the kinematics of the flow are unchanged and hence the efficiency remains unaltered. Similarly, the head delivered is the same although the pressure and power change in accordance
with the change in density.
1 one of the assumptions made when deriving the similarity laws in Section 2 was that the
flow was at high Reynolds numbers, D2 / , which implies negligible viscous forces.

(a) Density

(b) Viscosity

Figure 18.1: Effects on a pump characteristic curve due to a


change in fluid properties

90

Figure 18.2: Effect of solids on pump performance

As mentioned above, the effects of a change in fluid viscosity are also well
known. If the density of the fluid being pumped is kept constant while its
viscosity is increased, then the head and efficiency reduce, while the power
needed to maintain the flowrate increases. This is illustrated in Figure 18.1(b).
Note that the effects of viscosity are more pronounced at high flowrates, and
as a result the flowrate at best efficiency tends to become smaller as the viscosity is increased. This shift may not always be significant, but another influence, which cannot be represented on the curves for a given pump, is of
greater importance. This is the scale effect an increase in viscosity has a
greater influence on small pumps than large ones.
Having noted this effect of increased viscosity on pump performance, no
quantitative method of determining the effect will be determined here. For
the second difficulty, that of the equivalent fluid assumption, a simple derating method based on the work of McElvain [33] will be presented. This
is one method amongst others that are available, for example the methods of
Wilson et al. [30] or Burgess and Reizes [34].
Although the presence of solid particles introduces more complicated effects than those accompanying a viscosity increase in a simple fluid, there is
some rough qualitative similarity between slurry flow and the flow of a fluid
having values of both density and viscosity greater than those for water. The
effects on pump performance are shown schematically in Figure 18.2. This
figure illustrates the reduction in head and efficiency of a centrifugal pump
operating at the same speed, but handling a solid-water mixture. In this figure, and the discussion that follows, m represents the pump efficiency in
slurry service, and w the clear water equivalent. Likewise, Pm and Pw are the
power requirements for slurry service and water service respectively. The
head, Hw represents the head developed in water service, in height of water,
while Hm represents the head developed in slurry service in height of slurry.
The head ratio, HR, and the efficiency ratio, ER are defined as Hm /Hw and
m /w respectively.
McElvain [33] proposed that for centrifugal slurry pumps the head reduction and the efficiency reduction are proportional to the delivered concentration, Cvd , giving
Cvd
(18.1)
HR = ER = 1 K
0.20
where 0.20 is simply a convenient reference value of concentration. McElvain
91

Figure 18.3: Solid-effect reduction curve [33]

related the factor K to a relative density of solids and to average particle size
by means of reduction curves, which are shown in Figure 18.3. For example,
if a coarse sand with Ss = 2.65 and d5 0 = 1.5mm is pumped at Cvd = 0.20 the
reduction of head and efficiency is indicated to be 25 - 30%.
The curves and relationships proposed by McElvain were compared to the
data of other investigators by Sellgren and Addie [35], who included their
own data for large pumps tested at the GIW Hydraulic Laboratory. They
concluded that the general tendencies expressed by the reduction curves in
Figure 18.3 hold for many types of centrifugal slurry pumps and industrial
slurries. However, specific applications of the curves are effectively limited
to products with a narrow particle size distribution, to solids concentrations
below 20-25% by volume, and to small pumps with impeller diameters less
than about 0.5 m.
The experimental results of Sellgren and Addie [35] also confirmed that
the solids effect could be considered to be independent of rotary speed. A
slight dependence on flowrate was found, but it was concluded that this
could normally be neglected in practical applications.
An extension of the curves of McElvain is presented in Figure 18.4.

92

Figure 18.4: Expanded solid-effect reduction curve

93

References
[1] Trey Walters. Cutting costs in pump and pipe sizing. Chemical Processing,
2002.
[2] Feluwa pump catalogue, 2004. FELUWA Pumpen GmbH.
[3] Confessions of a chemical feed pump manufacturer, nd.
http://www.asiawaterbusiness.com/news_show.php?language=english&n_id=195.

[4] Warman centrifugal slurry pumps, 2006. Weir Minerals.


[5] B R Munson, D F Young, and T H Okiishi. Fundamentals of Fluid Mechanics. John Wiley & Sons, 3rd edition, 1998.
[6] J S McNown, J Malaika, and H Pramanik. Particle shape and settling
velocity. In Proc. Int. Assoc. Hyd. Res, 4th Meeting, Bombay, India, 1951.
[7] J S McNown and J Malaika. Effects of particle shape on settling velocity
at low reynolds numbers. Trans. Amer. Geophys. Union, 31, 1950.
[8] M L Albertson. Effects of shape on the fall velocity of gravel particles.
In Proc. 5th Iowa Hyd. Conf., Iowa, 1953.
[9] Edward J Wasp, John P Kenny, and Ramesh L Gandhi. Solid Liquid Flow:
Slurry Pipeline Transportation. Trans Tech Publications, 1977.
[10] W H Graf. Hydraulics of Sediment Transport, chapter 4, pages 58 62.
McGraw-Hill, 1971.
[11] R Davies and B H Kaye. Experimental investigation into the settling
behaviour of suspensions. In Proc. Int. Power Technology & Bulk Granular
Solids Conference, 1971.
[12] Aquatec maxcon website, nd. Accessed Online, 4th Sept, 2007.
[13] Vereinigte fllkrper-fabriken, nd.
http://www.vff.com/ .
[14] BS 5168:1975. British Standards Institute.
[15] W L Wilkinson. Non-Newtonian Fluids. Pergamon Press, 1960.
[16] G W Govier and K Aziz. The Flow of Complex Mixtures in Pipes. Litton
Educational Publishing Inc., 1972.
[17] R B Bird, W E Stewart, and E N Lightfoot. Transport Phenomena. J. Wiley
and Sons, 1953.
[18] E C Bingham. Fluidity and Plasticity. McGraw Hill, 1922.
[19] J R van Wazer, W J Lyons, K Y Kim, and R E Colwell. Viscosity and Flow
Measurement: A Laboratory Handbook of Rheology. Interscience, 1963.
94

[20] S N Bhattacharya. Introduction to fluid rheometry. In Proceedings of


Flow and Rheological Measurements of Complex Fluids, Royal Melbourne
Institute of Technology, July 1987.
[21] M Cross. Rheology of non-newtonian fluids, a new flow equation for
pseudoplastic systems. Journal of Colloid Science, 20:417437, 1965.
[22] N Casson. A flow equation for pigment-oil suspensions of the printing
ink type. In C C Mill, editor, Rheology of Dispersed Systems, pages 84104.
Pergamon Press, 1959.
[23] W H Bauer and E A Collins. Thixotropy and dilatancy. In R R Eilrich,
editor, Rheology: Theory and Applications, volume 3. 1960.
Tcorrelation

[24] A B Metzner and J C Reed. Flow of non-newtonian fluidsA


of the laminar, transition, and turbulent-flow regions. AIChE Journal, 1,
1955.
[25] D W Dodge and A B Metzner. Turbulent flow of non-newtonian systems.
AIChE Journal, 5, 1959.
[26] D G Thomas. Transport characteristics of suspensions:VIII. Journal of
Colloid Science, 20:267277, 1965.
[27] E Guth and R Simha. Kolloid-Z. 74:266, 1936.
[28] M Mooney. Explicit formulas for slip and fluidity. Journal of Rheology, 2
(2):210222, 1931.
[29] B S Massey. Mechanics of Fluids. van Nostrand Reinhold, 4th edition,
1979.
[30] K C Wilson, G R Addie, and R Clift. Slurry transport using centrifugal
pumps. Elsevier Applied Science, London, 1992.
[31] R Durand. The hydraulic transportation of coal and other materials in
pipes. In Colloquium of the National Coal Board, London, Nov 1952.
[32] J P Chilton. Principles of Metallic Corrosion.
Burlington House, London, 1973.

The Chemical Society,

[33] R McElvain. High pressure pumping. Skilings Mining Review, 63(4):114,


1974.
[34] K Burgess and J Reizes. The effect of sizing, specific gravity and concentration on the performance of centrifugal slurry pumps. Porc. IMechE,
190(36):391399, 1976.
[35] A Sellgren and G Addie. Effects of solids on large centrifugal pump
head and efficiency, 1989. Paper presented at CEDA Dredging Day, Amsterdam, The Netherlands.

95

Você também pode gostar