You are on page 1of 57

199

A NEW LOOK AT POROUS MEDIA FLUID MECHANICS - DARCY TO TURBULENT

A. Dybbs and R. V. Edwards

ABSTRACT 201
1. INTRODUCTION 202
2. BACKGROUND AND PERTINENT PREVIOUS WORK 204
3. THE FLOW REGIMES OF POROUS MEDIA 208
3.1 The Experiments 208
3.1.1. The Flow Visualization 210
3.1.2. The Velocity Measurements 211
3.2 Darcy Flow Regime 214
3.3 The Inertial Flow Regime - A Steady
Non-Linear Laminar Flow Regime 222
3.4 Unsteady Laminar Flow Regime 230
3.5 The Unsteady and Chaotic Flow Regime 246
4. FLUID MECHANIC IMPLICATIONS OF EXPERIMENTS 248
4.1 Modelling the Inertial Flow Regime 248
4.2 The Internal Heat Transfer Coefficient
in the Darcy and Inertial Flow Regime 250
5. CONCLUSIONS 251
ACKNOWLEDGEMENTS 251
REFERENCES 252
SYMBOLS 255

J. Bear et al. (eds.), Fundamentals of Transport Phenomena in Porous Media


© Martinus Nijhoff Publishers, Dordrecht 1984
201

A NEW LOOK AT POROUS ~ffiDIA FLUID MECHANICS - DARCY TO TURBULENT

A. Dybbs and R.V. Edwards

Department of Mechanical and Aerospace Engineering, Depart-


ment of Chemical Engineering, Case Western Reserve Universi-
ty, Cleveland, OR 44106

ABSTRACT

The purpose of this review paper is to present the results


of laser anemometry and flow visualization studies of the flow of
liquids in porous structures. Three dimensional velocity pro-
files and movies of dye streaklines will be shown. The porous
media consisted of plexiglas spheres in a hexagonal packing and
glass and plexiglas rods arranged in a complex, fixed three di-
mensional geometry. The liquids used were water, silicone oils,
Sohio MDI-57 oil and mineral seal oil. The Reynolds number based
on average pore size and average pore velocity ranged from 0.16
to 700.

The results indicate the existence of four flow regimes in a


porous medium: 1. The Darcy or creeping flow regime where the
flow is dominated by viscous forces and the exact nature of the
velocity distribution is determined by local geometry. This type
of flow occurs at Re < 1. At the Re ~ 1, boundary layers begin
to develop near the solid boundaries of the pores. 2. The iner-
tial flow regime. This initiates at Re between 1 and 10 where
the boundary layers become more pronounced and an "inertial core"
appears. The developing of these "core" flows outside the boun-
dary layers is the reason for the non-linear relationship between
pressure drop and flow rate. As the Re increases, the "core"
flows enlarge in size and their influence becomes more and more
significant on the overall flow picture. This steady non-linear
laminar flow regime persists to a Re - 150. 3. An unsteady lam-
inar flow regime in the Reynolds number range of 150 to 300. At
202

a Re - 150, the first evidence of unsteady flow is observed in


the form of laminar wake oscillations in the pores. These oscil-
lations take the form of traveling waves characterized by dis-
tinct periods, amplitudes and growth rates. In this flow regime,
these oscillations exhibit preferred frequencies that seem to
correspond to specific growth rateq. Vortices form at Re - 250
and persist to Re - 300. 4. A highly unsteady and chaotic flow
regime for Re > 300, qualitatively resembling turbulent flow ..

Based on these results, an analytical model is presented


that exhibits similar pressure drop behavior as porous media from
the Darcy to the inertial flow regime. In addition, the implica-
tions of these fluid mechanics results for other transport
phenomena are discussed.

1. INTRODUCTION

The understanding of the nature of flow in porous media is


essential for many science and engineering applications. Such
diversified fields as geohydrology, powder metallurgy, chemical,
mechanical, aerospace and petroleum engineering to name just a
few, need to understand transport phenomena in porous media.
Ever since the original work of Darcy [1] in 1856, much experi-
mental and analytical work has been devoted to the application of
Darcy's empirical relationship and its extensions. More recent-
ly, the need to understand heat and mass transfer phenomena in
porous media has gained importance because of such applications
as geothermal power production, secondary oil recovery, and tran-
spiration cooling. For a review of some of this research, see
Refs. 2-7.

Despite all of this work, little seems to be known about the


detailed nature of the interstitial velocities in porous media.
Analytically the problem of studying the fluid mechanics of a
cloud or cluster of arbitrary shaped closely packed particles has
proved intractable. Experimentally, velocity probes such as hot
film anemometers cannot be successfully used in porous media.
Thus, while the details of momentum transfer for a single sphere
are understood, those for a cluster of spheres are not. Conse-
quently little is understood about heat and mass transport for
clusters of spheres. As an example of this, consider the deter-
mination of the internal heat transfer coefficient for fluid
flowing in porous media. The inability of any theoretical
analysis to predict these coefficients has led to many experimen-
tal correlations (8-11). A comparison of these correlations
shows qualitatively a power law relationship between an appropri-
ately defined Nu and Re number in the form Nu = constant Re
where n ranges from 0.5 [9] to 2.0 [10]. The numerical value of
the heat transfer coefficient can differ as much as two orders of
203

magnitude (12) and this is true even for the same porous materi-
als tested under ostensibly similar conditions. The flow regimes
studied are from laminar to turbulent and these wide variations
occur at all Reynolds numbers. There is apparently an as yet
unknown parameter or set of parameters that governs interphase
heat transport in these systems. It is quite possible that the
details of the flow are different and thus the local heat
transfer coefficients are different. To adequately explain these
apparent discrepancies, there is a need to understand intersti-
tial velocities in porous media flows.

Fluid flow tracer experiments (residence time distribution


measurements) show that the flow in porous media is not slug flow
even in systems with large ratios of system diameter to particle
diameter. In reactors such as catalyst bed reactors, it is im-
portant and useful to control the flow patterns in order to effi-
ciently utilize the catalyst. A better understanding of the flow
patterns in porous media is also important for the design of
chromatographic columns and the understanding of adsorption and
multi-phase flow in porous media. Up until now, information
about the above areas has been inferred from measurements made
external to the media. Such areas as the following need clarifi-
cation and better understanding.

1) The effect of the packing geometry on the interstitial


and gross flow behavior.

2) The nature of the flow near a solid boundary.

3) The understanding of the transition from the Darcy flow


regime to the "inertial"' regime.

4) The understanding of the "inertial" flow regime.

5) The understanding of the transition from "inertial" to


the "turbulent" flow regime.

6) The nature of the mixing that takes place within an in-


terstitial space.

7) The nature and control of channeling phenomena.

The purpose of this paper is to review the results of laser


anemometry and flow visualization studies of the flow of liquids
in porous media. Three dimensional vel~city profiles and movies
of dye streaklines are presented. The porous media consisted of
plexiglas spheres in a hexagonal packing and glass and plexiglas
rods arranged .in a complex, fixed three dimensional geometry.
The liquids used were water, silicone oils, Sohio MDI-57 oil and
204
mineral seal oil. The Reynolds number based on average pore size
and average pore velocity ranged from 0.16 to 700.

In addition to clarifying some of the above areas, the


results indicate the existence of four flow regimes in a porous
medium: 1) The Darcy or creeping flow regime where the flow is
dominated by viscous forces and the exact nature of the velocity
distribution is determined by local geometry. This type of flow
occurs at Re < 1. At Re ~ 1, boundary layers begin to develop
near the solid boundaries of the pores. 2) The inertial flow
regime. This initiates at Re between 1 and 10 where the boundary
layers become more pronounced and an "inertial core" appears.
The developing of these "core" flows outside the boundary layers
is the reason for the non-linear relationship between pressure
drop and flow rate. As the Re increases, the "core" flows en-
large in size and their influence becomes more and more signifi-
cant on the overall flow picture. This steady non-linear laminar
flow regime persists to a Re - 150. 3) An unsteady laminar flow
regime in the Reynolds number range of 150 to 300. At a
Re - 150, the first evidence of unsteady flow is observed in the
form of laminar wake oscillations in the pores. These oscilla-
tions take the form of traveling waves characterized by distinct
periods, amplitudes and growth rates. In this flow regime, these
oscillations exhibi t preferred frequencies that seem. to
'correspond to specific growth rates. Vortices form at Re - 250
and persist to Re - 300. 4) A highly unsteady and chaotic flow
regime for Re > 300, qualitatively resembling turbulent flow.

2. BACKGROUND AND PERTINENT PREVIOUS WORK

Most of the experimental studies of flow in porous media


have been concerned with determining gross correlations of flow
properties. These studies treat the porous medium as a black box
through which fluid flows. This has been due, in part, to the
fact that instruments capable of measuring within porous media
were nonexistent. Such instruments as the hot wire, hot film,
and thermistor anemometers are difficult to use because of their
size, the problem of moving them to measure more than one point
in the flow, and the uncertainty as to how they would influence
the flow field.

Thus the black box approach is characterized by measuring


the pressure drop generated by various flow rates through a
porous medium. The data generally display linear behavior in the
viscous-dominated Darcy flow regime changing smoothly to a
higher-order relationship in the inertia-dominated flow regime.
Although it is generally accepted that this first transition to
non~linear flow is both laminar in nature and due to the increas-
ing effects of inertia in the internal microscopic flow field,
205

disagreement still exists as to when a second transition from


non-linear laminar to fully turbulent flow occurs and what the
physical mechanisms governing that transition are. For these
reasons, examination of the miscroscopic flow field of a
representative porous medium model offers an opportunity to gain
an understanding of the physics of flow through porous media at
non-Darcy Reynolds numbers.

As noted by Scheidegger (2) and Bear (6) in their comprehen-


sive reviews of porous media research up to 1974, two main ap-
proaches are used in the reduction and characterization of flow
rate pressure drop data. The first, introduced by Forcheimer
(2) is to fit a second- or third-order equation to the data in
the following relationship:
2 3
Vp = aq + bq + cq (2.1)

where Vp is the pressure gradient across the porous bed, q is the


average macroscopic fluid velocity defined as the total system
flow rate divided by the cross-sectional area of the vessel con-
taining the porous structure, and a, b, and c are constants of
fit. Note that the average macroscopic velocity q is also re-
ferred to as the superficial or filter velocity in the litera-
ture.

A second method of data reduction uses similarity theory to


reduce the black box data to a functional relationship involving
appropriately defined dimensionless parameters (6). The primary
parameters resulting from dimensional analysis are the Euler
number:

Eu=~2 (2.2)
pq
and the Reynolds number:

Re =~ (2.3)
II
where p is the fluid density, II is the fluid viscosity, and d is
a characteristic length scale typical of the internal structure
of the porous medium. In addition, a set of parameters associat-
ed with the actual porous structure include the porosity and a
set of length scale ratios which characterize constituent parti-
cle or pore size, internal surface roughness and bed packing
geometry.

Ergun (13) in 1952 defined a modified friction factor:


r =~ . i . _E_3_ (2.4)
2 L 1 - E
pq
and a modified Reynolds number:
206
Re' = .e.g;!. 1 (2.5)
J1 ~
Also, he found that an equation of the form:

f'=~+B
ReI (2.6)

fits the data well over the entire range of Reynolds numbers.
Specifically, he found the values of the constants to be A=150,
B=1.75. The Ergun equation has become a more commonly used rela-
tionship characterizing porous media for practical engineering
applications, although a recent paper by Macdonald, et al. (14)
suggests that modifying the above coefficients to be A=lBO,
B=1.B-4.0 provides a better correlation to a large body of exist-
ing data. Macdonald suggests that the variation in B is due to
internal surface roughness differences at high flow rates, analo-
gous to turbulent flow in pipes, although roughness data verify-
ing this assertion was not available for the media involved.

Introduction of the following simple relationship, known as


Dupuit's approximation (2):

v=3. (2.7)
t:
into the modified friction factor and Reynolds number of Ergun
yield the following microscopic or "bed" quantities:
d t: tP t:
tP* = ~2 1 - t: bed
(2.B)
L 1 - t:
pv
Re* = pvd t:
= Re.. (2.9)
J1 bed 1 - t:
where v is the average microscopic or "pore" velocity. In this
case, a velocity is used which is consistent with a pore diameter
d and thus more suitable for an investigation that emphasizes mi-
croscopic flow. The porosity factor t:/(1-t:) assumes new physical
meaning as a void-to-solid fraction ratio characterizing the
medium. For these reasons, the quantities defined in Eqs. 2.8
and 2.9 above will be used in this review.

Flow field characteristics have been mainly inferred from


local averaged measurements made external to the porous media.
For example, measurements of the flow field were made downstream
of a packed bed by Schwartz and Smith (15), Rhoades (16), Murphy
(17) and Musser (18,19) in order to determine the effect of chan-
neling. These studies are not conclusive and in some cases are
contradictory (19). In particular, each study observed average
peak velocities near the wall, and each developed a different
criteria for the position of this peak velocity. The relation-
ship between this peak, the geometry of the packing and the gross
flow rate was not clearly established.
207

Flow visualization studies (20,21) and some transport stu-


dies (22,23) of a single particle in a packed bed have inferred
some interesting flow patterns which deserve further study.
Visualization of the internal flow in a porous medium has been
accomplished by Hanratty and co-workers (21,23). Jolls and Han-
ratty (23) examined the pathlines of flow through a packed bed of
plexiglas spheres using dye injection into a fluid with a refrac-
tive index closely matching that of the plexiglas. They found
that transition to unsteady flow occurred between Reynolds
numbers of 110 and 150, where the Reynolds number was based on
macroscopic filter velocity and sphere diameter. Using the
porosity of 0.41 cited, this equates to bed Reynolds numbers of
268 and 366 for this study. In addition, the gradual transition
to unsteady flow observed did not yield any discontinuity in mass
transfer, consistent with the smooth behavior of the traditional
friction factor-Reynolds number plot.

In a second study by Wegner, Karabelas, and Hanratty (21)


detailed visualization of the flow on a sphere in a geometry
identical to the above was performed. At a bed Reynolds Number
200, characterized by steady flow, nine distinct regions of re-
versed flow were observed on the surface of a test sphere, which
the authors referred to as "separation bubbles". At a bed Rey-
nolds number of 488, the flow was unsteady as evidenced by path-
line movement in time, and the same regions of reversed flow were
found, but were larger. The authors referred to the unsteady
flow at this Reynolds number as "turbulent".

In addition to these findings, several authors (6) have


placed the transition to "turbulent flow" between macroscopic
Reynolds numbers of 150 and 300, again based on filter velocity.
However, the use of the term turbulent appears inappropriate in-
sofar as some authors, according to Bear (6), simply interpret
this as unsteady flow, which could still be laminar in nature;
whereas others consider it to mean fully turbulent flow in the
classical sense. As a result, there is little consensus as to
what the distinct flow regimes and associated critical Reynolds
numbers are for flow beyond the Darcy range. Moreover, several
candidate mechanisms have been proposed to govern these transi-
tions. These include non-uniformity of pore size distribution,
internal surface roughness, mixing of intersecting stream tubes,
and separation of the microscopic flow field from the internal
local geometry.

Analytical approaches to flow in porous media take essen-


tially three different approaches. One approach is the method of
geometric modelling in which one postulates a geometry resembling
the structure of the porous media, yet is simple enough to allow
the governing differential equations to be solved. Such a model,
the "straight- capillary" model, has been used by Kozeny (24) and
208

modified by Carmen (3) to obtain expressions for the permeability


of porous media. A minor variation of the model is the skewed
capillary model. A review of the above "capillaric" models has
been given in Refs. 2 and 7.

Brenner (25) has proposed a model made up of an infinite ar-


ray of unit cells having boundaries of arbitrary shape and con-
taining one or more particles of arbitrary shape. To make the
flow amenable to analysis, it is required that the boundary of
the unit cell be a fluid surface. Since a porous medium is gen-
erally not a well ordered structure, the idea of using statisti-
cal models is appealing, and a number of attempts have been made
in this area (26,27,28). Both Beran (27) and Scheidegger (28)
feel that this approach will ultimately prove the most fruitful.

A third approach is the development of the correct averaged


forms of the governing differential equations (6,7,29-32). These
resulting equations should be valid for any porous media geometry
and any statistical model should be in accord with these equa-
tions.

All of the three methods lead to unspecified parameters


which must be determined experimentally.

3. THE FLOW REGIMES OF POROUS MEDIA

Four different flow regimes of fluids flowing in porous


media have been identified by interstitial velocity measurements
and flow visualization experiments. The regimes are the Darcy or
creeping flow regime, the inertial flow regime, the unsteady lam-
inar flow regime, thE' highly unsteady and the chaotic flow re-
gime. The porous media studied consisted of spheres in a hexago-
nal close packing and rods in a complex three dimensional
geometry. The Reynolds number based on average pore size or par-
ticle size and average pore velocity was varied from 0.16 to 700.

3.1 The Experiments

Flow visualization was performed throughout the four flow


regimes and three dimensional interstitial velocity profiles were
measured to determine the details of each flow regime. The velo-
city measurements were made using laser anemometry and were per-
formed on porous media consisting of spheres in a close hexagonal
packing and cylindrical rods in a complex geometrical structure.

A cross section of the packed bed of spheres is shown in


Fig. 3.1. This porous medium consisted of 1.27 cm (1/2 inch)
plexiglas spheres in a hexagonal close packing. Each layer of
spheres was identical to the one twice removed from it. In all,
209
PLEXIGL ASS SPHERES

PL EXIGL ASS STRIPS

Fig. 3.1 Cross sec:ion of porous medium packed bed of spheres


and locat10n of axial velocity measurements.

four hundred and eighty-five spheres were used, eight layers of


thirty-seven spheres and seven layers of twenty-seven spheres.
Plexiglas strips along the wall of the pipe test section held the
spheres in a hexagonal close packing arrangement without the use
of glue, which was optically desirable. Additional experimental
details can be found in Refs. 33-35. The void fraction of ths
porous medium was 0.394 and the permeability, k, as calculated
from Darcy's law was 8.6± 0.9 x 10 4 Darcys.

A model of the rod porous medium is shown in Fig. 3.2. It


was constructed of pyrex glass rods 1.27± 0.04 cm in diameter
and consisted of vertical rods fixed symmetrically in a hexagonal
shape; the space between these rods being filled by horizontal
rods, some perpendicular to the side walls of the test section
and the bulk flow direction and som~ at ± 30° angles to the side
walls of the test section. A total of eighty-eight rods was
used, thirty-three were fixed vertically by two 1.27 cm thick
aluminum plates and the remaining fifty-five filled the void
spaces as shown in Fig. 3.2. Wall effects were eliminated by
cutting the horizontal rods and fitting them along the walls.
Such an arrangement of rods needs no glue. This porous structure
has an inhomogeniety in porosity in the bulk flow direction that
varies from 0.33 to 0.79. The friction factor - Reynolds number
210

Fig. 3.2 Photograph of half-scale model of complex rod structure


porous medium.

plot for this porous medium is shown in Fig. 3.3, along with the
Ergun equation, which correlates to the experimental data to
within the experimental error. Note that the pore length scale
used is the diameter of the constituent rods. See Refs. 36 and
37 for details.

3.1.1 The flow visualization. The flow visualization was done in


the complex three dimensional rod bundle porous medium shown in
Fig. 3.2. The test section was placed in an open loop flow sys-
tem. A dye solution consisting of 5% potassium permanganate by
weight was injected by a variable-flow infusion pump from a 50ml
syringe through microbore tubing to a set of 22 gauge injection
needles. The infusion pump was capable of flow rates between
0.04 and 33 ml/min. The injection needles in turn were inserted
into the top of the test section through gasketed fittings locat-
ed above selected rods. At these locations, glass rods were re-
placed by plexiglas rods with grooves cut down their lengths.
The injection needles were then inserted into these grooves, al-
lowing a minimum disturbance of the surrounding local flow field
during dye injection. By sliding the needles up or down in these
211

k, ,, PA,Rtl.MEtERS
POROSITY' £ . 0 496
~
SfM
o
o
FLUIO MUlTURE
OIL
w.o.T(R
l E'IIC,TH ROllO 0084 o
,
'HOTE.R +- GLvC.EROL

'!'!;' "-
~ I02 "- ,.,
~... , ,
"'
g
~
....
~ '0'
Q

:3'"
,
0

MODIFIED REYNOLOS NUMBER Re- ' ( p~d)( I ~E")

Fig. 3.3 Modified friction factor - Reynolds number plot;


correlation of the data of Rosenstein [36,37] to
equation of Ergun.

grooves, dye illumination was possible throughout a field of in-


terest that avoided entrance and exit regions of the bed. This
field, delineated by a dotted line, is shown in Fig. 3.4, which
is a side view of the rod bundle in the photographic plane. Due
to the relatively small differences in the refractive indices of
water and glass/plexiglas, reasonable resolution of dye streak-
lines was limited to three or four rod diameters away from the
near test section wall. Care was taken to approximately match the
dye injection flow rate to the overall water flow rate so as to
minimize disturbing the internal flow field being visualized.
The ratio of dye-to-water flow rate averaged 10- 4 for most cases.

Streakline patterns were recorded by means of both a 35 mm


Nikon F2 camera and a Canon 1014 Electronic Super 8 move camera.
Further experimental details can be found in Refs. 38 and 39.

3.1.2 The velocity measurements. The laser anemometer used to


make the measurements consisted of a two component three beam po-
larized TSI Inc. system. The application of the laser anemometer
to this study of flow in porous media requires minimal distur-
bance of the incident and scattered light as it transverses in-
side the plexiglas spheres and glass rods and as it crosses the
fluid-solid interfaces. This was accomplished by closely
212

TOP VIEW

SIDE VIEW

, IElD OF INTEREST
CELLS STuDIED IN DETAI L

Fig. 3.4 Diagram of Rod bundle: partial top view shows grooved
injection rods (shaded), side view defines general
field of interest (dashed line), cells studied in
detail (solid line).

matching the index of refraction of the plexiglas spheres and


pyrex glass rods with a mixture of silicone fluids.

The TSI laser anemometer included all the optics necessary


to make two overlapping velocity sample volumes in the intersti-
tial spaces of the porous medium,. Each sample volume had a cross
stream length of 0.165 cm and a width of 0.012 cm (36). In addi-
tion, the system is capable of sensing the direction of the flow
by means of a Bragg cell frequency shifter that is driven by a 40
MHz oscillator. The receiving electronics are designed to deter-
mine if the received signal is above or below 40 MHz, indicating
if the flow direction is plus or minus. The outputs of the two
receivers (one for each velocity component) are sent to a fre-
quency tracker which was specially designed for the two component
system (see Ref. 40).

Each channel of the tracker has a local oscillator which


continually tracks the frequency of the incoming signal, averages
213

HORIZONTAL ROD
BULK ~O NORMAL TO THE WALLS
FLOW "-
DIRECTION ,
/
l'

Z • .&2
HORIZONTAL ROD
30' TO THE WALLS Z' .50
y

Fig. 3.5 Structure of the cell in complex rod bundle porous


medium.

Fig. 3.6 Cross section of representative measurement cell in


plane Z = 3.00 in. (from bottom of test section)
--- • ---, represents cell boundary; --, represents
cylinders in plane;-- •• --, represents cylinders
in the plane below, and -- --, represents cylinders
in the plane above.
214

the mean frequency of each component of velocity for 1, 10, or


100 seconds and converts the resulting frequency to binary digi-
tal form and displays the result.

The test section of either plexiglas spheres or glass rods


and the mixture of silicone oil was of uniform refractive index.
In addition, the pyrex glass windows in the complex rod structure
were parallel to each other and together perpendicular to the
optical axis of the laser anemometer. The test sections were
oriented so that one of the velocity components measured was
aligned with the direction of the bulk flow and the other in the
vertical plane perpendicular to the bulk flow direction.

The working fluid that matched the index of refraction of


the plexiglas spheres and test section components consisted of a
mixture of 73% by weight Dow Corning 550 fluid (n = 1.4974, v
125 cs at 22°C) and 27% by weight Union Carbide L42 fluid (n
1.4711, v = 400 cs at 22°C). The resulting mixture had the fol-
lowing properties at the operating temperature, 22°C: n = 1.4905,
v = 188 cs, and p = 1.04 g/cm3 •

The working fluid, that matched the index of refraction of


the pyrex rods was a mixture of silicone oils composed of 48.5%
by weight Dow Corning 556 (n = 1.4600, v = 22.5 cs, p = 0.98 g/cm
at 25°C) and 51.5% by weight Dow Corning 550 (n = 1.495, v = 125
cs, p = 1.07 g/cm at 25°C). The working fluid was used at 28.1°C
where the mixture had the same index of refraction of pyrex glass
n = 1.4731 and the following properties: v = 43.3 cs and p
1.03 g/cm3 . The details of the index matching procedure can be
found in Refs. 36 and 37.

The representative cell of the complex porous structure


(Fig. 3.2) in which the velocity measurements were made is shown
in Fig. 3.5 and its location in the porous medium is shown in
Fig. 3.6. The cell sliced in the Z direction is shown in Fig.
3.7. Since the cell is symmetric, only one-half, from Z = 0.5 to
Z 10, is shown. The bulk flow direction is always in the Y
direction.

The two measured velocity components were collected and


analyized using a real time interactive computer system (41,42).
The third component of velocity was calculated using an
analytic-numeric scheme discussed in Ref. 43.

3.2 Darcy Flow Regime

The Darcy or creeping flow regime is dominated by viscous


forces and the exact nature of the velocity profiles is deter-
215

IlOO
THE wAllS

t-MIZONTAl ROO
_ NORMAL TO THE WALL
DfGRf ASfS IN SlZf

Z '0 ee

lOwER EOGE fF HORIZONTAL


ROO NORMAL TO WALL

Z '0 75
UPP(R EDGE Of HORIZONTAL
ROO :50. TO THE W.c.LLS

111 HORIZONTAL RODS


'~- )0. TO THE WALLS
FULL SIU

Z 'O~
I --
. I
x

Fig. 3.7 Sliced cell.

mined by the local geometry. This flow regime occurs at Re < I


and is illustrated by velocity profiles in both the packed bed of
spheres and the complex rod bundle structure.

Velocity profiles for the packed bed of spheres (Fig. 3.1)


have been made at Re = 0.154. The bulk or axial component was
measured:

1. Across several representative geometric sections of the


porous medium, normal to the gross flow direction.

2. Along a wall region of the bed, axially and radially.

3. In an interstitial space of a unit cell of the geometric


packing.
216

Axial velocity measurements were made in the representative


geometric section of the porous medium in the plane of the third
bead from the bed entrance normal to the flow direction. This
region is shown in Fig. 3.1. An unsmoothed three dimensional
plot of this axial velocity is shown in the plots of Figs. 3.8
and 3.9. Each figure shows a different perspective of the pro-
files. The height of the graph represents the velocity at the X,
Y position. The wall effect is clearly demonstrated. The average
velocity in the region near the wall is twice that at the center
of the porous medium. In addition the fact that the wall effect
is felt more than one sphere diameter from the wall may indicate
cross flows. Further evidence of the wall effect can be seen in
Fig. 3.10 where the axial velocity at the center of the wall re-
gion along the entire length of the porous medium is shown. As
shown in Figs. 3.8 and 3.9 the average axial velocity is about
twice that observed in the enter of the bed. This average value
of the axial velocity persists the entire length of the bed and
changes in response to the local geometry of the spheres. Radjal
measurements of the axial velocity were made in the region shown
in Fig 3.11. These measurements were made along five coplanar
lines spaced 1 mm apart. The resulting velocity scans are shown
in Figs. 3.12-3-.14. The data does suggest a ·'jet like" velocity
in the wall region.

Schwartz and Smith (15) observed similar velocity distribu-


tions in this flow regime by measuring velocity profiles above
porous media packed with particles of various sizes. They found
that for ratios of dt/dp (where dtis the porous medium diameter
and dp is the particle diameter) of less than 30, a peak velocity
near the wall ranged from 30 to 100 percent greater than the
velocity at the center of the porous medium. In particular for
the case of thi& study (Fig. 3.1) adt/dp ratio of 8, they found
that velocity gradients are large and that the peak velocity was
about 100% above that found in the center of the porous medium.
Musser (17,18) using hot wire anemometry also measured velocity
profiles above packed beds of spheres. He found similar peak
velocities near walls, but found them to exist at much higher
dt/d ratios than the 30 value given by Schwartz and Smith (15).
Refe¥ring to Figs. 3.8-3.9, the distance from the wall to the
point where the velocity has reached its minimum value averages
approximately 1.5 sphere diameters. This data would imply that
the velocity "defect" at the wall propogates no more that two
particle diameters in from the wall of the porous medium. This
result is consistent with that of Schwartz and Smith (15), who
found that the velocity remains relatively constant to within two
particle diameters from either wall, where it begins to increase
to its peak velocity.
217

Fig. 3.8 Perspective view of axial velocity profile of porous


medium shown in Fig. 3.1.

Fig. 3.9 Perspective view of axial velocity profile of porous


medium shown in Fig. 3.1, also see Fig. 3.8.
218

VELOCITY (em".e)

0 . 30

j I! f
0.20

0. 10

o L-~0~~ln~~2n~3~n~4~.0~5~
.0~6~n~~~0~B.~0-9~.0~I*Q~
0 ~1~1.0~12~D~
DISTANCE FROM BED ENTRANCE (em)

Fig. 3.10 Axial velocity for porous medium shown in Fig. 3.1
at the center of the wall region along the entire
length of porous medium.

Fig. 3.11 Region in which radial measurements of axial velocity


were made. Scans are lines along which velocity
measurements were made.
219

VELOCITY (cmlSe< I SCAN 2


0.30

f
1 fI
0 .20
!
0.10

____-L____ ~ ____ ~ ____ ~ __- - J


OL---~~--~

BED WALL 1.0 2.0 3,0 4 .0 5.0 6 .0


RADIAL DISTANCE FROM WAL L (m m)

Fig. 3.12 Axial veloci ty scan #2, made radial ly in wall region .
See Fig. 3.11.

VELOC I TY Icrn/secl SCAN 3


030

0 .20

0 , 10

SEaWALL 1,0 2.0 3.0 4 .0 50 6 .0


RADIAL DISTANCE FROM WALL 1m m)

Fig. 3.13 Axial veloci ty scan #3, made radial ly in wall region .
See Fig. 3.1l.
22Q

VELOCITY (em/seel SCAN 4


0.30

0.20

0.10

o '--_~~_-----'L. I o
BED WALL 1.0 2.0 3 .0 4 .0 5 .0 6 .0
RADIAL DISTANCE FROM WALL (mm)

Fig. 3.14 Axial velocity scan #4, made radially in wall region.
See Fig. 3.11.

VELOCITY (em/sec)
0.30

~t::.
t::. ~
'1

0.20 &
~ '"
0

J ~ <0
0 ~
0.10
OVELOCITY MEASUREMENT IN PLANE
OF THIRD ROW OF SPHERES

[). VELOCITY MEASUREMENT IN PLANE OF


ELEVENTH ROW OF SPHERES
'--BED WALL BED WALL-"""" I

o ~~~--~~----~~--~~----~~--~~--__-L____~L-__-LI-L__
0.0 1.0 2.0 3.0 4.0 5.0 6 .0 7.0 B.O
RADI AL DIS TANCE (em)
Fig. 3.15 Comparison of axial velocity profiles in the plane
of third row of spheres from the entrance with those
in the eleventh row of spheres from the entrance.
221

In this flow regime the velocity profiles, once developed,


remain constant throughout the porous medium. Fig. 3.15 shows a
comparison of axial velocity profiles in the plane of the third
row of spheres from the entrance with those from the plane of the
eleventh row of spheres from the entrance in the geometrically
representative region (Fig. 3.1). To within experimental error
the two sets of measurements are identical. Thus the axial velo-
city profile remains constant from the third row of spheres on.
These results also delineate an upper bound on the entrance re-
gion of a porous medium of this type. The entrance region being
defined as the distance from the beginning of the porous medium
to that place at which the flow becomes a constant flow. For
this case it appears to be three particle diameters.

Velocity measurements made with a unit cell of a cluster of


spheres halfway down the porous medium indicate approximately the
same interstitial velocity at all points in the cell. This was
anticipated from the profiles shown in Figs. 3.8 and 3.9.

Similar results were found for the complex rod bundle struc-
ture shown in Fig. 3.2.

Figure 3.16 shows the velocity component, Uy , in the bulk


flow direction for Re; 0.8. Each of the plots in Fig. 3.16
corresponds to a horizontal "slice" of the representative cell at
constant Z-coordinate. (Also see Fig. 3.7.)

Even at this Reynolds number a comparison of the velocity


profiles shows the presence of boundary layers in the most narrow
"pass" of the cell. In between these boundary layers, the velo-
city profile is almost flat (see Fig. 3.16a). With the increas-
ing influence of a new horizontal rod (see Fig. 3.16c), the velo-
city profile changes completely and gets redeveloped according to
the geometry at Z ; 0.8 and Z ; 0.64. The flow at this Re number
is very sensitive to the continuously changing boundary condi-
tions along the Y coordinate. No evidence of separation can be
seen.

Figure 3.17 shows the vertical component of velocity, UZ ' in


the Z direction. The positive and negative areas of the velocity
profiles can be distinguished. An interesting feature at
Re = 0.8 can be seen by comparing some of the plots from Figures
3.16 and 3.17.

In Fig. 3.17, the vertical component of velocity of the op-


posite sign does not correspond to a change of sign in the hor-
izontal flow direction. Such a flow corresponds to mixing taking
place inside the porous medium.
222
Re • 08 R. ' 0 .8

~
~
I
Z ' LO
I
<<I ) :~Z'1.0
I I
I I
I I
I I
I I : I I I
I ; I

:~
I
:
I: ~ I
I Z'0 .8
:~
I
\:
I
,
I
IZ'
I
,
0 .8

I I (11) I ,
I I' I I
I I I I
I I I I
I I I
I I ' ,

~:
I

:~I\
I I
I
I Z . 064
I
.
'~"O"
I (t:)

3.16 3.17

Fig. 3.16 Velocity component in bulk flow direction UY' Re = 0.8

Fig. 3.17 Velocity component in vertical direction UZ ' Re - 0.8

For Re = 0.8, the flow is an almost viscous mixing of two


flows; one of them is directed upwards, the other downwards, but
both having positive velocity in the bulk flow direction.

The velocity distribution at Re = 0.8 in the bulk flow


direction (see Fig. 3.16) displays some flatness in the region of
the most contraction. This distribution is different from the
one at higher Reynolds numbers where viscous forces are not
strong enough to resist the formation of the "jet like" flow.

3.3 The Inertial Flow Regime - A Steady Non-Linear Laminar Flow


Regime

This flow regime initiates at a Reynolds number between 1


and 10. At a Re - 1, boundary layers begin to develop near the
solid boundaries of the pores. As the Reynolds number is further
increased the boundary layers become more pronounced and an
223

"inertial core" appears. This core flow enlarges in size and its
influence becomes more and more significant in the overall flow
picture. The developing of these core flows outside the boundary
layers is the reason for the non-linear relationship between
pressure drop and flow rate.

These features can be seen from the velocity measurements in


the complex rod bundle structure porous medium. Profiles for
the three velocity components in the unit cell for Reynolds
number 7 are shown in Figs. 3.18-3.20. Even though the velocity
profiles are somewhat similar for Re = 0.8 and Re = 7.0 (see
Figs. 3.16 and 3.18; 3.17 and 3.19), the behavior of the flow
field is different.

At- Re = 7 the boundary layers are steeper and there is evi-


dence of a core like flow. While the Reynolds numbers (Figs.
3.16 and 3.18) differ by a factor of nine, the local velocities
differ by a greater factor. Thus more flow takes place in the
bulk flow direction at Re = 7 than at Re = 0.8.

Re J 7
Re ; 7

~
:
Z'0. 88
. ( )
:~Z'0.88
I I I I
I I I
I I I
I I I I I
I I
I
I
IZ ' 0 .76 I Z. 0 .76
I. I
I (b)
I
I I
I
I
I I
I I
I I I I I

i~: ~
I ~
I I
iZ ' 0 .67 I I IZ ' 0 .67
I I I I
, (c) I

3.18 3.19

Fig. 3.18 Velocity component in bulk flow direction Uy ' Re = 7.0

Fig. 3.19 Velocity component in vertical direction UZ ' Re = 7.0


224

Roe • 7 Re " 28

:~l'096
I I
I
I I I
I I I

I~I:
I
I
I I I
I z, 0 76
I . Iz· 0 81
I I (b) I I
I
I I I

I I
I I
I
I I I I I
I I
:I:

~
I I I
I I I
I I :Z'O 67
~I
,~,.,,,
I (C)

3.20 3.21

Fig. 3.20 Calculated horizontal component of velocity UX' Re 7

Fig. 3.21 "Bulk" flow Uy component at Re = 28

The other two components Uz and IX are useful in order to


complete the general picture of this three-dimensional flow. The
"bulk" component of velocity Uy (Fig. 3.18) is the only component
that is positive everywhere.

The vertical velocity UZ is formed in the same way as the


vertical component of the flow around horizontal cylinders: One
would expect the flow to be of the opposite sign in the lower and
upper parts. This also can be seen in Fig. 3.20 which represents
the calculated transversed horizontal velocity. As in the case
of a single cylinder, the flow displays better memory of the
events (such as change of boundary conditions, mixing, etc.) for
higher Re.

Figure 3.21 shows the velocity component U in the bulk flow


direction for Re = 28. The velocity distribution between boun-
dary layers is no longer flat and corresponds to the increasing
importance of inertial forces. In this "core" area, inertia is
the dominant force whereas in the boundary layers, both the iner-
tia and viscous forces are important.
225

Figure 3.22 shows the vertical component of velocity U in


the Z direction. The velocity patterns in Fig. 3.22 are slmilar
to those of Figs. 3.17 and 3.19.

The third component, UX' was numerically calculated using


the continuity equation as described in Ref. 43. Figure 3.23
shows the calculated Ux component of the velocity for Re = 28.
The flow in the direction normal to the bulk flow is very sensi-
tive to the change in the boundary conditions. The Ux component
does not display any boundary layer or "core"-type flow. The po-
sitive or negative sign (Fig. 3.23) corresponds to a three-
dimensional flow mixing, positive or negative in the X-direction.

Figure 3.24 shows the velocity component Uy in the bulk flow


direction for three different Re numbers at Z = 0.93. The change
in the Re from 0.8 to 7 produces steeper boundary layers and less
flatness in the velocity of the "core". The Re = 28 displays
even steeper gradients of the velocity outside the boundary
layers. This velocity profile persists until the flow reaches
the horizontal rod located downstream in the Y direction.

Re I 2a

Fig. 3.22 Flow pattern in vertical direction UZ ' Re 28


226

Re '28 Z ' 08 1 Re ' 28 Z ' 0.7 5 Re '28 Z · 0 .5 6

u,
y I
x

o o o

114.2c.m Isec

Fig. 3.23 Calculated component of vector-velocity UX' Re 28

Re • 0 . 8 Z ' 095 Re' 7 Z' 0 92

o o o

1.4 em I sec I 3.6 em I sec I 14 .2 em/sec

Fig. 3.24 Developing of boundary layers and "core" flow for


Re = 0.8, 7 and 28. Bulk flow velocity component Uy
227

Figure 3.25 shows the vertical component of fluid velocity


for the same Re number as shown in Fig. 3.24. Comparison with
Fig. 3.17 shows that for Re = 0.8 the flow is an almost viscous
mlxlng of two flows--one of them is directed upwards, the other
downwards, but both have positive velocity in the bulk flow
direction. At Re = 28 the mixing is more jet-like in nature with
each of the jets more inertial than viscous.

This steady non-linear laminar flow persists to a Reynolds


number of 150. The flow visualization of this regime can be seen
in Figs. 3.26-3.30, Figs. 3.26-3.28 show an overview of the
porous medium. The flow is always from right to left and I is
the dye injection flow rate in ml/min. Figure 3.26 taken at a
Reynolds number of 45, shows two injection needles in two of the
grooved vertical rods. The slight negative buoyancy of the dye
is apparent in this photograph. At higher Reynolds numbers this
buoyancy effect is not seen as evidenced in Fig. 3.27 which is
for a Reynolds number of 86. Features to note in these two fig-
ures are the well defined streaklines indicative of laminar flow,
and the dispersion of the dye downstream, indicative of the com-
plicated tortuous path an individual fluid particle follows as it
is transported through the porous medium. Figures 3.29 and 3.30
show a sequence of photographs reproduced from Super 8 movie
frames taken of the microscopic celli. (See Fig. 3.4). At Rey-
nolds numbers of 86 and 130 in Figs. 3.29 and 3.30 respectively,
the dye streakline boundaries are smooth and the flow is still
able to conform to local geometry.

Re ' 0 .8 Z' 095 Re·7 Z' 0 . 92 Ro '28Z '0. 91

l_ 4em/sec. 13.6 emf sec 114-2 em Isec

Fig. 3.25 Vertical component of velocity, Uz for Re 0.8, 7 and


28
N
N
00

Flow {3.26

+--

3.27}

Fig. 3.26 Re=4.5±6, 1= 2 x 0.30 cc/min. Fig. 3.27 Re=86 ±16, I = 2 x 0.30 cc/min.
Fig. 3.28 Re= 14&t6, I = 2 x 1.1 cc/min.
Fig. 3.29 Re=86 ±lO, I 30 cc/min.

{3.28

Flow

+--

3.29}
{3.30
Flow

+--
3.3l}

Fig. 3.30 Re,.,130 ±10, I 0.30 cc/rnin.


( "'- _ _1
'[ I
Fig. 3.31 Re=190±12, I 2 x 1.1 cc/rnin.
Fig. 3.32 Re=215±18, I 2 xLI cc/rnin.
Fig. 3.33 Re=225tl5, I 2 x 2.2 cc/rnin.

{3.32
now

---
3.33}

N
N

'"
230

The non-linear laminar flow regime varies from a Reynolds


number between 1 and 10 to a Reynolds number about 150. The
transition from the Darcy flow regime to this inertial flow re-
gime is due to the development of boundary layers near Reynolds
number one and the subsequent appearance of an "inertial core",
which assumes more significance as the Reynolds number increases.

The fluid flow in this flow regime can be represented as


flow inside a stream tube bounded by solid boundaries. At the
locations where two such stream tubes (or more) merge, complicat-
ed mixing takes place and the flow starts adjusting itself to the
local geometry. The geometry changes continuously and, in this
sense, the flow never becomes fully developed (even for very low
Reynolds numbe rs).

The nature of the complicated geometry of a porous medium


and the influence of the inertia terms appears to have been
neglected in all previous theoretical models. The fluid flow has
been analyzed as being fully developed. The presence of the mix-
ing process creates a new velocity profile different from a fully
developed one and the adjustment of the flow must show up in any
model of a porous medium until the next mixing takes place.

An important and interesting feature of the flow through the


porous media studied is that there is no change in the flow
structure at a given Reynolds number from pore space to pore
space as one goes downstream. But in any given pore the velocity
profiles are developing (for Re > 1). Hence at any position in
the porous medium the inertial effects are important since the
boundary effects have not penetrated the core flow. This pattern
of flow development repeats itself time and time again in the
pores as one proceeds downstream.

3.4 Unsteady Laminar Flow Regime

This flow regime begins at a Reynolds number about 150 and


persists to a Reynolds number of 300. At Reynolds number 150 ~n~­
tial unsteadiness is seen in the form of slight oscillations of
dye streamers in cells 2 and 3 of the complex rod bundle porous
medium (see Figs. 3.2 and 3.4). As Figs. 3.31 and 3.32 show, the
flows remains laminar throughout the field of interest after the
onset of unsteadiness. By a Reynolds number of 225 (Fig. 3.33),
however, dye streamers are beginning to break up as they exit
cell 1, and the dye exiting the bed is dispersed. Super 8 movies
of this sequence of Reynolds numbers showed that as oscillations
set in, they first appeared in the "pore" or open spaces between
rods, typified by cells 2 and 3, and that at certain Reynolds
numbers between 150 and 250, the oscillations had regular periods
which were measureable. Moreover, all the oscillations in the
231

bed at a particular Reynolds number had the same period. Beyond


a Reynolds number of 300, though, the unsteadiness ceased to be
regular and took on a random nature instead.

Details of the microscopic flow through cell lover this


Reynolds number range are shown in Figs 3.34-3.36. At a Reynolds
number of 199 in Fig. 3.34 the streakline edges are kinked or
wavy and dye is being flung into the upper yawed rod surface as
opposed to flowing smoothly around it as in Fig. 3.29. This pro-
vides qualitative evidence of the increasing effects of inertia
in the microscopic flow field, as would be expected when the Rey-
nolds number is increased. Figures 3.35 and 3.36 also show a
laminar flow which was unsteady, as seen from the movie se-
quences. Another feature of the flow through this cell is the
crossing of streaklines originating on either side of the
upstream rod and exiting the pore on the opposite side of the
downstream rod. This "mixing" of apparent "stream tubes", as
mentioned previously, has been proposed as a mechanism governing
the transition to turbulence (2), although the flow in Figs.
3.34-3.36 is not turbulent.

Detail movie sequences were made of individual streakline os-


cillations in the unsteady transitional range of Reynolds numbers
between 150 and 250. Two such sets are shown in Figs. 3.37 and
3.38. First, a sequence of prints reproduced from movie frames
separated by 0.112 sec are shown. Taken at a Reynolds number of
161, this sequence corresponds to the onset of initial unsteadi-
ness. A traveling wave of very small amplitude, on the order of
1/20 of a rod diameter, can be seen in the lower pore streakline
(cell 3). A more apparent traveling wave can be seen in the Fig.
3.38 sequence by following a trough in the lower dye streakline
of cell 3. This wave, at a Reynolds number of 173, appears to
have an amplitude of about 1/10 a rod diameter and a period of
0.40± 0.05 sec, which places one complete cycle between Figs.
3.38-4 and 3.38-5. A more accurate method of determining the
period (counting the number of oscillations over a given period
of time while viewing the actual movie film), was used to obtain
a value of 0.41± 0.06 sec.

A second oscillation of the same period is shown in Fig.


3.39. This sequence of four exposures was taken at a Reynolds
number 212 with the 35 mm camera connected to a motor drive fir-
ing five times a second. As this sequence clearly shows, the
period of the oscillation is exactly 0.40 sec within the accuracy
of the motor drive unit which was calibrated to ±.01 sec.
Another striking feature of this wave is the oscillation about
either end of the downstream rod. This implies a traveling wave
with an amplitude that has apparently grown from near zero at the
injection point near the forward boundary of cell 2 to a length
on the order of a full rod diameter as it impinges on the down-
N
W
N

Fig. 3.34 Re= 199±15, I = 0.59 cc/min. Fig. 3.35 Re=225±15, I 1.1 cc/min.

Fig. 3.36 Re=242±13, I 2.2 cc/min.


233

Fig. 3.37 Cell 3 streakline oscillation; Reynolds


number = 161±10; Injection rate = 0.59
cc/min.; 3.37-1. T = 0 sec; 3.37-2.
T = 0.11 sec; 3.37-3. T = 0.22 sec;
3.37-4. T = 0.34 sec; 3.37-5. T = 0.45 sec.

3.37-1

3.37-2

3.37-3
234

3.37-4

3.37-5

Fig. 3.38 Cell 3 streakline oscillation; Reynolds number


173±9; Injection rate = 2.2 cc/min.
3.38-1. T 0 sec.; 3.38-2. T = 0.11 sec.;
3.38-3. T 0.22 sec.; 3.38-4 T = 0.34 sec.;
3.38-5. T 0.45 sec.

3.38-1
235

3.38-2

3.38-3

3.38-4

3.38-5
N
I.;.J
0\

Fig. 3.39 Cell 2 wake oscillation; Reynolds number = 212 ± 12;


Injection rate = 1.1 cc/min. 3.39-1. T = 0 sec.;
3.39-2. T 0.20 sec.; 3.39-3. T = 0.40 sec.;
.3...3..9-4. T = 0.60 sec.

3.39-1
237
238
239
240
Fig. 3.40 Cell 2 streakline oscillation; Reynolds number =
22~15; Injection rate = 2.2 cc/min. 3.40-1.
T = 0 sec.; 3.40-2. T = 0.17 sec.; 3.40-3. T
0.34 sec.; 3.40-4. T = 0;50 sec.; 3.40-5. T
0.67 sec.

3.40-1

3.40-2

3.40-3
241

3.40-4

3.40-5

Fig. 3.41 Cell 2 vortex shedding; Reynolds number =232 ±8;


Injection rate = 4.4 cc/min. 3.41-1. T = 0 sec.;
3.41-2. T 0.11 sec.; 3.41-3. T 0.22 sec.;
3.41-4. T = 0.34 sec.; 3.41-5. T = 0.45 sec.

3.41-1
242

3.41-2

3.41-3

3.41-4

3.41-5
N
.p.
w
Fig. 3.42 CellI vortex shedding. 3.42-1 Re 237 ±1l,
I = 0.59 cc/lilin.
244

~
a
......
(.)
(.)

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

.H

N
I
N

"'"
C'"l
245

stream rod. Based on the increase in growth rate with increase


in Reynolds number seen in photo sequences of Figs. 3.37-1
3.37-5, 3.38-1 - 3.38-5, and 3.39-1 - 3.39-4, one might expect to
ultimately see actual vortex formation in the pore spaces between
horizontal rods at sufficiently high Reynolds numbers comparable
to the flow about an isolated circular cylinder.

Two final sequences of streakline oscillations, again repro-


duced from movie frames, are shown in Figs. 3.40 and 3.41. Fig-
ures 3.40-1 - 3.40.5 show a traveling wave on the densest dye
streakline in cell 2. One cycle can be seen over approximately
four of the five frames, which are separated by 0.166 sec,
corresponding to a period of 0.55 ± 0.08 sec. Verification by
direct movie viewing yields a period of 0.53 ± 0.06 sec. Note
also that the amplitude at the downstream rod, and thus the
growth rate through this pore, does not seem to be as large as
that of the preceding wave at a Reynolds number of 212.

The second sequence of Figs. 3.41-1 - 3.41-5 at a Reynolds


number of 232 shows a sharp change in the character of the flow.
Instead of a traveling wave, the dye appears to be illuminating
actual vortex formation in the pore space of cell 2. Individual
vortices are evidenced by the sharp reversals of the lower dye
streamer as it enters the pore space. These reversals subse-
quently expand as they are transported downstream. A regular fre-
quency of this shedding was not apparent, either from the photo
sequence or the movie.

Further evidence of the existence of distinct vortices in


the pore spaces of the representative cells is seen in Fig. 3.42.
These two single instant 35 mm exposures, at Reynolds numbers of
237 and 305, clearly show individual vortices being shed into the
pores of cells 2 and 3. Again, characteristic frequencies proved
too difficult to obtain with the visualization techniques used.

The existence of vortex formation in the pore spaces strong-


ly suggests that a laminar, wake instability may be responsible
for the transition from laminar, steady flow to highly unsteady,
chaotic flow in this particular rod bundle geometry. This is typ-
ical of flows about bluff bodies. Interestingly, the frequency
of the traveling waves was not a monotonically increasing func-
tion of Reynolds number. This is seen in Figure 3.43 which plots
a dimensionless frequency of laminar oscillation versus bed Rey-
nolds number. Data points were obtained by both viewing movie
sequences and actually counting oscillations in real time obser-
vation of the bed in connection with 35 mm photography. Note the
apparent sharp breaks in the data at Reynolds numbers of about
210 and 255. This would seem to indicate that the flow
throughout the bed as a whole prefers certain critical frequen-
cies. This speculation is reinforced by the fact that oscilla-
246

SYMBOL SOURCE

q
16 0 35mm STILL

~+
0 SUPER B MOVIE
15
~I>
1.4

,.. 1. 3
u
;z
w 1. 2
:>
0
w
a:
..... 1, 1 ,

4= ~
V>
V>
w 1.0
..J

.-f-+~
;Z

+
0 09
u;
;z
w
::;: 0 ,8
is
07
200 210 220 230 240 250 260 270 280 290 300 3 10
pVd
REYNOLDS NUMBER Re' -".
-

Fig. 3.43 Dimensionless frequency of wake oscillation versus


Reynolds number in unsteady flow regime; parameters
based on average pore velocity and rod diameter.

tions at different points in the bed were observed to have the


same frequency at a given Reynolds number. These characteristic
critical frequencies, in turn, may correspond to laminar insta-
bilities having the largest growth rates which allow a small per-
turbation to amplify in preference to disturbances at other fre-
quencies. In order to verify this, a determination of the param-
eters of the cylinder wakes, including wake width and velocity,
will be necessary.

3.5 The Unsteady and Chaotic Flow Regime

A highly unsteady and chaotic flow regime exists for Rey-


nolds numbers greater than 300. The flow visualization results
for the complex rod bundle porous medium clearly show this re-
gime. As seen in Fig. 3.44, at a Reynolds number of 305, the dye
streaklines are breaking up and dye is dispersed even in cell 1.
This photo, along with Fig. 3.45, portrays a highly unsteady,
chaotic flow that is not laminar, but rather is reminiscent of
the turbulent mixing of dye first seen by Reynolds in his classic
pipe transition experiment and visualized by many other research-
ers over the years.

The microscopic flow through cell 1 indicates a similar type


of flow. Figures 3.46 and 3.47 at Reynolds numbers of 449 and
622, show a highly unsteady, chaotic flow that exhibits the dye
0.44

3.45}

Fig. 3.44 Re=305±13, I 2 x 4.4 cc/min. Fig. 3.45 Re=514±22, I 2 x 8.6 cc/min.

Fig. 3.46 Re=44~±24, I 4.4 cc/min. Fig. 3.47 Re=622±30, I 8.6 cc/min.

{3.46

3.47}
N
-I'-
-..J
248
dispersion characteristic of classic turbulent flow. At best,
these photos provide qualitative evidence of the existence of
true turbulent flow in a porous medium, although quantitative,
time-dependent measurements will be required to verify this.

4. FLUID MECHANIC IMPLICATIONS OF EXPERIMENTS

The above results can be used to aid in the modelling of


porous media flows and to improve the understanding of heat and
mass transport in porous media. As an illustration of this, con-
sider the following: 1. The modelling of the inertial flow re-
gime and the tansition from the Dracy to the inertial flow re-
gime, and 2. The difference in the internal heat transfer coef-
ficient of a porous medium in the Darcy and the inertial flow re-
gimes.

4.1 Modelling the Inertial Flow Regime

The behavior of the flow in the Reynolds number range 1-28,


appeared to be a "developing" laminar flow: The momentum defect
due to the solid boundaries did not penetrate to the center of
the flow pores.. A calculation of the pressure drop through a
typical pore structure would thus proceed as a calculation for an
entrance flow where the intertial terms are important - even in
the laminar flow region (for a pipe). The pressure drop per unit
length would be higher than for a fully developed flow. What is
needed at this point is an idea of how long the "entrance length"
is compared to the total length. Two additional experimental ob-
servations help to clarify this length scale.

1. The friction factor behavior of the porous medium is


identical to that of other porous media.

2. For the Reynolds number range studied the flow pattern


after the first three "unit cells" from the entrance, are identi-
cal within experimental error. Thus, on a macroscopic scale, the
flow is a developing one in that the wall effects do not
penetrate to the core of the pores.

The porous medium can be modelled in the Re number range of


the study as a series of geometric sections each with the identi-
cal entrance flow and within each the flow is developing. In the
porous medium of this study "length of a section" is well de-
fined, being the downstream length of a unit cell. Denote this
length 2. Since the porous medium behaves on a macroscopic scale
the same as other porous media, it is reasonable to assume that
the concept of a section length, 2, would be valid for any porous
medium. Further, the friction factor should be a function of the
Reynolds number (based on a pore diameter d) and the ratio 2/d.
249

,('
f = f (Re, d)' (4.1)

In most porous media ~/d would be of order one.

To get a feel for how a system like this should behave quan-
titatively, we performed a calculation of the friction factor as
a function of Reynolds number for pipe sections of ~/d - 1. The
result of the calculation was

f =~
Re
+ B, (4.2)

where A and B are function of ~ /d. The pressure drop in the


porous media is then n times the pressure drop for one section if
the bed is n sections long. The shape of the friction factor
curve should then mimic the curve of one section. The computed
friction factor curve matched the measured data for a ~/d value
of 1.1, as is shown in Fig. 4.1.

/ADJUSTED VALUE OF 1jJ : oV' 64 FOR Re "

_ .- ERGUN EOUATION { 19521


- EXPERIMENTAL DATA

d Il
~----2 . 0

:;:~::3:::::::===
":: 1. 2
1.1
~~:::::=1.0
.9

DARCY' 5 LAW .5

10 ' 10 2
pVd
REYNOLDS NUMBER Re ' T

Fig. 4.1 Adjusted friction factor - Reynolds number plot;


Rosenstein results [36]; experimental, analytical,
correlation with Ergun equation.
250

--
I I I
S'~OERS i.'J5 ;'./

10
-== - . rrrt=!
:~
>~ -l IGANSON

, ;"'-:--E~ ~
~ . ,~
,
>1 I,
:1 I .;~r
'I
I .e°1" -
1
c--- '--'--
.r-r-
'0 '" '-" -~

r - KAR AND f-~


I~ ,
f-- DYBBS ~ ~ 7) _ " V
. -~--rH ~ ,~ I~
H-4
I II '}
'.." III
I . ~ ;r. I f- LU
I - - r- -t '" :!.!Lf:tt -- ~ - l-
I -~-~ I
1'71 I
1------'- - 1-
-
I? I II I I I I!I I
";'1 II II I I II

. 10 2
10 10

md p
Re
"
Fig. 4.2 Internal heat coefficients for various porous media
after Kunii [44].

4.2 The Internal Heat Transfer Coefficient in the Darcy and


Inertial Flow Regime.

The experiments indicate a difference between the fluid


mechanics in the Darcy and intertial flow regimes. One would ex-
pect this difference reflected in the heat and mass transport in
porous media. Figure 4 . 2 shows a compilation of the internal
heat transfer coefficient data for many different porous media.
At a Reynolds number between 1 and 10 the Nusselt vs Reynolds
number correlation changes. In the Darcy regime the power of the
Reynolds numbers is - 1.36, in the inertial flow regime it is
- 1 . 0. It is interesting to note that between Reynolds numbers
150 and 400 there appears to be another change in the correla-
tion , perhaps corresponding to the unsteady laminar flow regime.
251

The above two examples are illustrative of the manner in


which the fluid mechanics results reported in this review paper
can be used.

5. CONCLUSIONS

This paper has presented the results of experiments that in-


dicate some new and interesting phenomena about flow in porous
media. In particular four flow regimes have been indentified.
These are: 1) The Darcy or creeping flow regime where the flow
is dominated by viscous forces and the exact nature of the velo-
city distribution is determined by local geometry. This type of
flow occurs at Re < 1. At Re 1, boundary layers begin to
develop near the solid boundaries of the pores. 2) The inertial
flow regime. This initiates at Re between 1 and 10 where the
boundary layers become more pronounced and an "inertial core" ap-
pears. The developing of these "core" flows outside the boundary
layers is the reason for the non-linear relationship between
pressure drop and flow rate. As the Re increases, the core
flows enlarge in size and their influence becomes more and more
significant on the overall flow picture. This steady non-linear
laminar flow regime persists to a Re - 150. 3) An unsteady lam-
inar flow regime in the Reynolds number range of 150 to 300. At
a Re - 150, the first evidence of unsteady flow is observed in
the form of laminar wake oscillations in the pores. These oscil-
lations take the form of traveling waves characterized by dis-
tinct periods, amplitudes and growth rates. In this flow regime,
these oscillations exhibit preferred frequencies that seem to
correspond to specific growth rates. Vortices form at Re -250
and persist to Re - 300. 4) A highly unsteady and chaotic flow
regime for Re > 350, qualitatively resembling turbulent flow.

These results indicate some areas of future study. The de-


tailed nature of the unsteady flow regime needs to be measured,
using laser anemometry, in particular the oscillation frequency
as a function of Reynolds number. Pressure distributions in each
of the flow regimes should be computed along streamlines using
the detailed velocity measurements. Finally the relationship
between the velocity profiles and transport behavior has to be
elucidated.

ACKNOWLEDGMENTS

The authors acknowledge the useful discussions on transition


flows with Prof. Eli Reshotko. In addition, the partial finan-
cial support of this work by the National Science Foundation
Fluid Dynamics Program Grant HCME79-13389 is gratefully appreci-
ated.
252

REFERENCES

.1. Darcy, H.. Les Fontaines Publiques de la Ville de Dijon,


Paris, Victor Dallmont, (1856).
2. Scheidegger, A.E., The Physics of Flow Through Porous Media.
New York, MacMillan Co., (1960).
3. Carmen, P.C., Flow of Gases Through Porous Media. New
York, Academic Press, (1956).
4. Collins, R.E., Flow of Fluids Through Porous Materials.
New York, Reinhold Publishing Co., (1961).
5. Irmay, S. Ed., Physical Principles of Water Percolation and
Seepage. Arid Zone Research XXIX, UNESCO (1968).
6. Bear, J., Dynamics of Fluids in Porous Media. New York,
American Elsevier Co., (1972).
7. Dullien, F.A.L., Porous Media Fluid Transport and Pore
Structure. Academic Press, New York, N.Y., (1979).
8. Jakob, M., Heat Transfer Vol. 2, John Wiley Sons, New York,
New York, 394, (1957).
9. Barker, J.J., Heat Transfer in Packed Beds. Ind. and Eng.
Chemistry, 57:43, (1965).
10. Kunii, D., and Suzuki, Particle-to-Fluid Heat and Mass
Transfer in Packed Beds of Fin~ Particles., Int. J. Heat
Mass Transfer, 10;845, (1967).
11. Nelson, P.A. and Galloway, T.R., Particle-to-Fluid Heat and
Mass Transfer in Dense Systems of Fine Particles.,Chem.
Engr. Sci., 30:1, (1975).
12. Curry, D.M., and Cox, J.E., The Effect of the Porous Materi-
al Characteristics on the Internal Heat and Mass Transfer.
ASME Paper 73-HT-49.
13. Ergun, S., Fluid Flow Through Packed Columns. Chem. Eng.
Progress, 89.48, (1952).
14. Macdonald, I.F., EI-Sayed, M.S., Dullien, F.A.L., Flow
Through Porous Media - The Ergun Equation Revisited. Ind.
Eng. Chem., Fundamentals, 18.3. (1979).
15. Schwartz, C.E., and Smith, J.M., Flow Distribution in Packed
Beds. Ind. and Eng. Chem., 75:1209, (1953).
16. Rhoades, R.G., Flow of Air in Beds of Spheres; A Statistical
and Theoretical Approach. Ph.D. Thesis Rensselaer Polytech-
nic Institute, (1963).
17. Murphy, D., Experimental Study of Low Speed Momentum
Transfer in Packed Beds. Ph.D. Thesis, Case Western Reserve
University, (1967).
18. Musser, W.N., The Design of Inlet J)istributors for Large-
Scale Chromatographic Columns. M.S. Thesis, Case Western
Reserve University, (1968).
19. Musser, W.N., Low Dispersion Columns for Large Scale Chroma-
tographic Separations. Ph.D. Thesis, Case Western Reserve
University, (1971).
253

20. Irmay, S .. The Mechanism of Filtration of Non-Colloidal


Fines in Porous Media. Symposium on the Fundamentals of
Transport Phenomena in Porous Media, Haifa, Israel, (1969).
21. Wegner, T.H., Karabelas, A.J., and Hanratty, T.J., Visual
Studies of Flow in a Regular Array of Spheres. Chem· Engr.
Science. 2659, (1971).
22. Jolls, K.R., and Hanratty, T.J., Use of Electro-Chemical
Techniques to Study Mass Transfer Rates and Local Skin Fric-
tion to a Sphere in a Dumped Bed. AIChE J., 15.199, (1969).
23. Jolls, K.R., and Hanratty, T.J., Transition to Turbulence
for Flow Through a Dumped Bed of Spheres. Chem. Engr. Sci-
ence, 21;1185, (1966).
24. Kozeny, J., Flow in Porous Media. S.B. Akad, Wiss. Wien,
Abt., IIa.126, (1927).
25. Brenner, H., Rheology of Two Phase Systems. in Vol. 2 Annu-
al Review of Fluid Mechanics ed. Van Dyke et aI, (1970).
26. Aranon, R.H., Statistical Approach to Flow Through Porous
Media. Phys. Fluids, 9;1721, (1966).
27. Beran, M.J., Statistical Continuum Theory, New York, Inter-
science, (1968).
28. Scheidegger, A.E., Statistical Theory of Flow Through Porous
Media Trans. Soc. Rheol., 9;313, (1965).
29. Slattery, J.C., Momentum, Energy. and Mass Transfer Con-
tinua, New York, McGraw Hill Co" (1972).
30. Whitaker, S., Advances in the Theory of Fluid Motion in
Porous Media. Ind. and Eng. Chem., 61.14, (1969).
31. Dybbs, A., and Schweitzer, S" Forced Convection in Saturat-
ed Porous Media. Heat Transfer 1970, U. Grigull and E.
Hane, eds. Elsevier Pub. Co., Amsterdam (1970).
32. Gray, W.G., A Derivation of the Equations for Multi-Phase
Transport. Chem. Engr. Sci. 30, (1975).
33. Johnston, W., Dybbs, A., and Edwards, R.V., Measurements of
Fluid Velocity Inside Porous Media with a Laser Anemometer
Department of Fluid Thermal and Aerospace Sciences, Case
Western Reserve University, Report FTAS/TR-74-98 (1974).
34. Johnston, W., Dybbs, A., and Edwards, R.V., Measurement of
Fluid Velocity Inside Porous media with a Laser Anemometer
Phys. Fluids Case Western Reserve University, 18,7, (1975).
35. Stephenson, D.B., Laser Anemometry Measurements of Fluid
Velocities Inside Porous Media. M.S. Thesis, Case Western
Reserve University, (1976).
36. Rosenstein, N.D., Non-Linear Laminar Flow in a Porous Medi-
um. Ph.D. Dissertation, Department of Mechanical and
Aerospace Engineering, Case Western Reserve University,
(1980).
37. Rosestein, N.D., Dybbs, A., and Edwards, R.V., Porous Medium
Veloity Profiles - Darcy to Inertial. in press.
38. Bilardo, Jr., V.L., The Flow Regimes of Flow in a Porous
Medium. M.S. Thesis, Department of Mechanical and Aerospace
Engineering, Case Western Reserve University, (1982).
254

39. Bilardo, Jr. V.L., Dybbs, A., and Edwards, R.V., Visualiza-
tion of the Flow Regimes in a Porous Medium. in press.
40. Edwards, R.V., Lading, L., and Coffield, F .. Design of a Fre-
quency Tracker for Laser Anemometer Measurement. University
of Missouri, Rolla, (1977).
41. Dybbs, A. and Bradshaw, F., A User Oriented Mini-Computer
System for a Fluid Mechanics Laboratory. Computers and Edu-
cation, Vol. 1, 167-175 (1977).
42. Dybbs, A., and Bradshaw, F., A Microprocessor Based Ring
Network for Experimentation. Advances in Computer
Technology--1980, Vol. 2, ed. A. Seireg, ASME Press,
(1980).
43. Rosenstein, N.D., Dybbs, A., and Edwards, R.V., Data Ac-
quisition and Processing from A Laser Anemometer. in Comput-
ers in Flow Predictions and Fluid Dynamic Experiments, eds.
K.N. Ghis, T.J. Mueller and B.R. Patel, ASME Publications,
New York, New York, (1981).
44. Kunii, D. and Smith, J. M., Heat Transfer Characteristics of
Porous Rocks: II. Thermal Conductivity of Unconsolidated
Particles with Flowing Fluids. AIChE Journal, Vol. 7, 29
(1961).
45. Saunders, O. A. and Smoleniak, S., General Discussion on
Heat Transfer. Section V, Inst. Mech. Eng., 443 (1951).
46. Gamson, B. W.) Thodos, G., and Hougen, O. A., Heat and Mass
and Momentum Transfer in the Flow of Gases Through Granular
Solids. Presented at the AIChE Meeting at Cincinnati, OH,
Nov. 16-17 (1942).
47. Kar, K. K., Internal Heat Transfer Coefficients of Porous
Metals. to be presented at ASME Winter Annual Meeting,
Phoenix, Arizona, Nov. 1982.
255

SYMBOLS

a constant
A slope of Ergun equation
b constant
B constant in Ergun equation
c constant
d characteristic pore dimension (m)
d particle diameter (m)
P
dt tube diameter (m)
2
Eu Euler number = IIp/(pq )
3
f ~~_E:_
I modified friction number
2 L l-E:
pq
k permeability (Darcy)
1 passage length (m)
L characteristic length of porous medium
n exponent
N frequency of oscillation in Hz
P pressure (kP a )
lip Pressure drop across rod bundle
Pe Peclet number
q macroscopic fluid velocity (m/sec)
Q volumetric flow rate
Re Reynolds number pqd/)..l
modified Reynolds number =~_l_
Re'
)..l l-E:
Re* modified Reynolds number
S Strouhal number
t time in seconds
T temperature in °c
v average pore velocity (m/sec)
v velocity (m/sec)
x co-ordinate directions (m)
256

y co-ordinate directions (m)


z co-ordinate directions (m)

Greek symbols

E porosity = void volume/total volume of rod bundle


n index of refraction
~ absolute fluid viscosity (kg/m-sec)
v kinematic viscosity (m 2 /sec)
p fluid density (kg/m3 )
~ friction factor
~* modified friction factor
T dimensionless time or time constant
t passage length or length of section

You might also like