You are on page 1of 6

Lecture Notes CHE 312 Fluid Mechanics (Winter 2012)

8. Fluid-Particle Systems & Porous Media


8b. Multiple particles & Porous media
Partly based on Chapter 6 and 11 of the De Nevers textbook.
LN8a: 6.13
LN8b: 11.1, 11.4, 11.5

Fluid friction in porous media, fixed beds
A porous medium is a solid with pores (i.e. void spaces). In impermeable porous media these
voids are isolated, i.e. not connected. A polystyrene cup is made of impermeable porous
material. It has many air pockets in it; since these are not connected, the material does not
leak. In permeable porous media the voids are connected and allow for fluid flow through
them; the interconnected voids act as narrow, tortuous channels. Permeable porous media, and
flow through them occur in many natural and engineered systems (soil and groundwater, oil
and gas fields, filters, packed beds etc). Flow through porous media therefore has important
engineering implications.
As a model porous medium we consider a packing of solid spheres, all having the same
diameter d
p
, see Figure 1. The main issue we would like to address relates to fluid friction in
this (model) porous medium. If we know the fluid friction
fr
e we can (with help of
Bernoullis equation) answer questions such as what pressure drop is required to maintain a
certain flow rate through the porous medium.


Figure 1

The model situation is given in Figure 1: a bed with cross sectional area A, length L,
consisting of spheres with diameter d
p
. A fluid flow through it is driven by a pressure drop
p .
First we introduce some terminology. Suppose we are pushing a volumetric fluid flow
rate of
V
through the bed. Then we define the superficial velocity
sup
v as
sup
V
v
A

. The
porosity of the bed (symbol ) is that part of the volume of the bed that is void of solid
particles. The channel in Figure 1 has a volume V AL = . If the channel contains N spherical
solid particles then the solids volume fraction is
3
6
p
N d
AL

= . Where there is no solid, there


is void; the porosity then is 1 = .
The fluid velocity in the bed is higher than
sup
v , since the fluid has to squeeze itself
through the channels of the porous medium; the fluid has less than the cross sectional area
A to flow through. The cross sectional area available for fluid flow is A . The average
interstitial velocity (the velocity of the fluid in between the spheres) in the bed then is
sup
V
ins
v
v
A


= = which is higher than
sup
v since 0 1 < < . From a practical point of view the
superficial velocity is more important than the interstitial velocity since the former is a direct
measure for the flow rate. From a theoretical point if view the interstitial velocity is more
important. It determines the flow structures in the bed (e.g. it determines if the flow in the
voids is laminar or turbulent).
In order to describe friction in the particle bed, we recall the concept of the hydraulic
diameter. In a straight channel it was defined as four times the cross sectional area available
for flow, divided by the wetted perimeter. In each cross section of the packed bed the
configuration of spheres is different, and the hydraulic diameter will fluctuate. If we average
over the length L of the bed we can redefine the hydraulic diameter D
h
as four times the
void volume divided by the total surface area. If we neglect the area of the channel walls, the
surface area is due to the N sphere surfaces. The total wetted area in the channel then is
2
w p
A N d = . The void volume we write as
( )
3
6 1
p
N d
AL

. This is a bit of a weird


construction. It says that since the volume fraction occupied by the spheres is
3
1
6
p
N d
AL

= = , we can write
( )
3
6 1
p
N d
AL

. Now the hydraulic diameter is


4 2
3 1
h p
w
AL
D d
A

= =

.
This whole exercise we are still undertaking to determine friction in a particle bed. The
general equation for friction in channels with length L and (hydraulic) diameter D
h
was
2
1
4
2
fr
h
L
e f U
D
= . In terms of the particle bed, the velocity U here must be the interstitial
velocity
sup
ins
v
v

= . If we substitute the expressions for D


h
and
ins
v we get for the friction
( )
2
sup 3
1
1
3
fr
p
e fL v
d

= . Now (unfortunately) the friction factor in porous media is defined as


3f: 3
PM
f f so that finally


( )
2
sup 3
1
1
fr PM
p
e f L v
d

= (1)

The friction factor
PM
f depends on the flow regime in the particle bed, and thus on the
Reynolds number. The Reynolds number we choose here is based on the hydraulic diameter
and the interstitial velocity:
( )
sup
2 2
Re Re
3 1 3
f h ins f p
PM
D v d v

= =

, so that
( )
sup
Re
1

f p
PM
d v
.
Figure 2 shows the dependency of
PM
f on Re
PM
(note the slight differences in notation in the
Figure compared to the notation in these lecture notes).
Figure 2 shows two important regimes: laminar for low Reynolds numbers, turbulent
for high Reynolds numbers (turbulent in quotation marks since it is hard to develop real
turbulence in a porous medium due to the limited space). As we have seen before (pipe flow,
drag on a single sphere), in the laminar regime the friction factor is inversely proportional to
the Reynolds number. Here the correlation is
150
Re
PM
PM
f = so that in this regime


( )
2
sup
2 3
1
150
fr
p
v
e L
d

= (2)

This is called the Kozeny-Carman (or sometimes Blake-Kozeny, see Figure 2) equation. In
the high Re
PM
regime,
PM
f tends to a constant value (just as was the case for rough pipes):
1.75
PM
f = , so that


( )
2
sup 3
1
1
1.75
fr
p
e L v
d

= (3)

known as the Burke-Plummer equation. We can smoothly transit from the low Reynolds to
the high Reynolds regime by proposing
150
1.75
Re
PM
PM
f = + , and as a result

( ) ( )
2
sup 2
sup 3 2 3
1 1
1
1.75 150
fr
p p
v
e L v L
d d



= + (4)

which is the (famous) Ergun equation. It is the drawn curve in Figure 2 and pretty well
matches experimental results.


Figure 2 (reprinted from the De Nevers book, pg 404)

At this stage, Example 11.1 from the De Nevers book is very illustrative; we will discuss it in
class.

Permeability
In the Kozeny-Carman limit (Eq. 2),
fr
e is proportional to the velocity (since superficial and
interstitial velocity are linearly related it does not really matter which velocity). In the
Kozeny-Carman limit, velocities are usually very low. When applying Bernoullis equation
we therefore neglect inertia (the
2
1
2
v term):
p vL
gz
k


| |
+ =
|
\
. Here we have lumped part
of the constants of Eq. 2 in a new constant k. For practical reasons the velocity v here is the
superficial velocity. The constant k is called permeability, it has units m
2
(check for yourself).
Applying this equation to a thin slice if porous material dx, the equation can be written as:


( ) d p gz
v
dx k


+
= (5)

This is the Darcy equation. Considering two specific situations may help in understanding Eq.
5 better. In the first case we do not consider a pressure gradient, constant, and align x with
z . Then Eq. 5 reads g v
k

= which describes a fluid trickling through a porous medium


under the influence of gravity (e.g. water through coffee powder). In the second case we apply
a pressure gradient in the horizontal (x) direction which creates a flow in horizontal direction,
and we do not need to consider gravity:
dp
v
dx k

= .

Filters

Figure 3

An example of applying Darcys equation is filtration. In Figure 3 we show a horizontally
placed filter. The resistance to flow (i.e. friction) is due to the filter itself, and the filter cake.
If we have a superficial velocity v going through the system, the pressure drop over the cake
is
c
c
c
D
p v
k
= , over the filter it is
f
f
f
D
p v
k
= , so that the total pressure drop is


f
c
c f
D
D
p v
k k

| |
= + |
|
\
(6)

(maybe you notice the analogy with Ohms law with p in the role of voltage, v in the role of
current, and
f
c
c f
D
D
k k

| |
+ |
|
\
as two resistances in series).
In the course of time, the filter cake grows thicker. If the slurry pushed through the filter
has a solids volume fraction , and if all the solids are captured by the filter (and the cake) the
cake grows at a rate
c
dD
v
dt
= (this can be appreciated by setting up a solids mass balance per
unit of cross sectional area over the filter cake). If we rewrite this as
1
c
dD
v
dt
= and substitute
it in Eq. 6 we get
f
c c
c f
D
dD D
p
dt k k

| |
= + |
|
\
. In quite many practical cases p is constant so
that the only time-dependency in this equation is due to the filter cake growing in time. If we
separate variables we get
f
c
c
c f
D
D
p dt dD
k k

| |
= + |
|
\
. This we integrate from t=0 at which we
have no cake at all, to some moment in time t:
2
2 2
f c f
c c
c
c f c f
D D D
D D
p t D
k k k k

| |
= + = + |
|
\
. At
some stage in the process
2
f
c
c f
D
D
k k
. At that stage
2
2
c
c
D
p t
k

= so that then the thickness of


the cake grows with the square root of time.

Fluidization

Figure 4 (reprinted from Fluid Mechanics for Chemical Engineers by J.O. Wilkes)

In the situation of Figure 4a we have a vertically placed packed bed of solid particles. If we
supply fluid through a porous plate from the bottom with a certain volumetric flow rate
V
uA = (with u the superficial fluid velocity, and A the cross sectional area of the bed) a
pressure drop over the bed (
1 2
p p p = ) will develop. In the case of laminar flow,
fr
e
depends linearly on the superficial velocity (see Eq. 2) so we expect a linear relationship
between u and p (see Figure 4b). At some u, the drag on the solid particles forming the bed
gets so high that is comparable to the gravity (minus buoyancy) force on the particle. At that
stage the particles start losing their mutual contact; the beds starts expanding and gets
fluidized (starts behaving like a fluid), and the height H increases. If we then further increase
the flow rate (i.e. u) the pressure drop does not increase much more: the higher flow rate
expands the bed more which makes the space between the particles bigger and (relatively)
reduces friction losses.
The concept of (at least the initial stage) fluidization can be quantified as follows.
Check out the forces on the particle bed as a whole:
An upward force due to the pressure drop: pA .
A gravity force down: ( ) 1
s f
HA g ( +

with H the bed height.
If we assume laminar flow, Eq. 2 can be used for
fr
e , and Bernoullis equation between top
and bottom of the bed reads
( )
2
2 3
1
150
fr
f f p
u
p
gH e H
d

= = so that
( )
2
2 3
1
150
f
p
u
p H gH
d

= + . The force balance over the solids phase then reads


( )
( )
2
2 3
1
1 150
s f f
p
u
g g
d

( + = +

or ( )( )
( )
2
2 3
1
1 150
s f
p
u
g
d

= , which can
be rewritten as an equation in u:


( )
( )
2 3
150 1
p
s f
d
u g



=

(7)

With this equation we can estimate at what u the transition from a packed to a fluid bed takes
place since Eq. 7 expresses a balance of gravity and (fluid) friction. A typical case of a gas
fluidized bed would be a bed with in packed state a porosity of =0.4, a particle size of
p
d = 100 m,
s
=2000 kg/m
3
, =1.8 10
-5
Pa s, and
f
=1.2 kg/m
3
. Then fluidization will
start at (roughly) u=7.7 mm/s. This we can compare with the settling velocity of the spheres in
air based on Stokes law
( ) ( )
2
3
1
150
18 18
s f p
s
gd
u u



= = =0.6 m/s. (applying Stokes law is
not fully appropriate here since Re
f s p
u d

= =2.4>1, but it is a fair estimate; if you would do


the full iteration you would end up at
s
u =0.55 m/s). The superficial velocity required to
fluidize a bed of particles is roughly 100 times less than the settling velocity. This is the
results of:
The interstitial velocity between the particles being higher than the superficial velocity.
The drag on a particle in a close packing being higher than the drag on a particle in
unbounded fluid.

You might also like