Group Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Title: 11.5.3 - Characteristics of Fibre Suspensions in a Turbulent Pipe Flow
ALL VOLUMES CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00102023/00289
 Material Information
Title: 11.5.3 - Characteristics of Fibre Suspensions in a Turbulent Pipe Flow Particle-Laden Flows
Series Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Physical Description: Conference Papers
Creator: Dearing, S.
Visentini, R.
Soldati, A.
Publisher: International Conference on Multiphase Flow (ICMF)
Publication Date: June 4, 2010
 Subjects
Subject: fiber suspensions
turbulent flow
phase discrimination
PIV
 Notes
Abstract: Additive induced drag reduction in turbulent flows is a well known phenomenon. Since Toms discovered that a small concentration of linear polymers could reduce drag considerably, there has been a plethora of work on polymer induced drag reduction. Despite such a large amount of attention the drag reducing mechanism is still unclear. Less attention has been paid to fibre suspensions. Although drag reduction can be modest, they resist degradation unlike polymer additives, and studying the behaviour of fibre-laden flows may elucidate the mechanisms behind drag reduction. The overreaching motivation of this work is to characterise fibre suspensions in order to understand more about the effect of different types of fibres in turbulent pipe flows at high Reynolds number, such as those typically used in industrial applications. An important tool in the investigation of such fibre laden flows is the development of a robust phase-discrimination algorithms to discriminate fibres from seeding in Particle Image Velocimetry (PIV) images. The phase discrimination algorithm under development is based on an object based approach which sorts objects according to parametric combinations of fibre length and aspect ratio. The algorithm has been adapted to include the discrimination of ellipsoidal particles (fibres). The algorithm discriminates between seeding and fibres, subsequently PIV techniques are used to calculate the flow velocity fields. The number density and fibre orientation of fibres is calculated using an ad hoc algorithm is calculated and compared to current DNS data for similar fibre suspensions in turbulent flows (Marchioli et al. 2010). Velocity profiles and bulk measurements of fibre suspensions in a turbulent pipe flow are measured using Particle Imaging Velocimetry (PIV) and pressure readings. The effect of fibre concentration and Reynolds number on the shape of the velocity profile and pressure drop along the pipe has been investigated for three different type fibres: wood fibres, nylon fibres and a rigid rod (fibre-like) polymer, Xanthan Gum. Each fibre suspension has very different pressure drop and velocity profile characteristics. The length scale of fibres are also very different which has a role in the different fibre transport behaviours. In all fibre types it has been found that the behaviour does not degrade with the length of time that fibres have been subjected to high shear stress nor with length of time fibres are allowed to sit in the pipe circuit. Information about the distribution of the fibres within the pipe are extracted from number density plots calculated and it is found at the lower Reynolds numbers examined the fibres tend to deposit at the wall. Furthermore this Reynolds number effect is more pronounced at the higher concentrations.
General Note: The International Conference on Multiphase Flow (ICMF) first was held in Tsukuba, Japan in 1991 and the second ICMF took place in Kyoto, Japan in 1995. During this conference, it was decided to establish an International Governing Board which oversees the major aspects of the conference and makes decisions about future conference locations. Due to the great importance of the field, it was furthermore decided to hold the conference every three years successively in Asia including Australia, Europe including Africa, Russia and the Near East and America. Hence, ICMF 1998 was held in Lyon, France, ICMF 2001 in New Orleans, USA, ICMF 2004 in Yokohama, Japan, and ICMF 2007 in Leipzig, Germany. ICMF-2010 is devoted to all aspects of Multiphase Flow. Researchers from all over the world gathered in order to introduce their recent advances in the field and thereby promote the exchange of new ideas, results and techniques. The conference is a key event in Multiphase Flow and supports the advancement of science in this very important field. The major research topics relevant for the conference are as follows: Bio-Fluid Dynamics; Boiling; Bubbly Flows; Cavitation; Colloidal and Suspension Dynamics; Collision, Agglomeration and Breakup; Computational Techniques for Multiphase Flows; Droplet Flows; Environmental and Geophysical Flows; Experimental Methods for Multiphase Flows; Fluidized and Circulating Fluidized Beds; Fluid Structure Interactions; Granular Media; Industrial Applications; Instabilities; Interfacial Flows; Micro and Nano-Scale Multiphase Flows; Microgravity in Two-Phase Flow; Multiphase Flows with Heat and Mass Transfer; Non-Newtonian Multiphase Flows; Particle-Laden Flows; Particle, Bubble and Drop Dynamics; Reactive Multiphase Flows
 Record Information
Bibliographic ID: UF00102023
Volume ID: VID00289
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: 1153-Dearing-ICMF2010.pdf

Full Text



7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


Characteristics of Fibre Suspensions in a Turbulent Pipe Flow


S. Dearing* R. Visentinit and A. Soldati*

Dipartimento di Energetica e Macchine, Universith di Udine, Udine, UD 33100, Italy
t University of Toulouse, Institute of Fluid Mechanics Toulouse, All6e de Professeur Camille Soula, 31400 Toulouse, France
stella.dearing @uniud.it; alfredo.soldati@uniud.it
Keywords: Fiber suspensions, turbulent flow, phase discrimination, PIV




Abstract

Additive induced drag reduction in turbulent flows is a well known phenomenon. Since Toms discovered that a small
concentration of linear polymers could reduce drag considerably, there has been a plethora of work on polymer induced
drag reduction. Despite such a large amount of attention the drag reducing mechanism is still unclear. Less attention
has been paid to fibre suspensions. Although drag reduction can be modest, they resist degradation unlike polymer
additives, and studying the behaviour of fibre-laden flows may elucidate the mechanisms behind drag reduction. The
overreaching motivation of this work is to characterise fibre suspensions in order to understand more about the effect
of different types of fibres in turbulent pipe flows at high Reynolds number, such as those typically used in industrial
applications.
An important tool in the investigation of such fibre laden flows is the development of a robust phase-discrimination
algorithms to discriminate fibres from seeding in Particle Image Velocimetry (PIV) images. The phase discrimination
algorithm under development is based on an object based approach which sorts objects according to parametric
combinations of fibre length and aspect ratio. The algorithm has been adapted to include the discrimination of
ellipsoidal particles (fibres). The algorithm discriminates between seeding and fibres, subsequently PIV techniques
are used to calculate the flow velocity fields. The number density and fibre orientation of fibres is calculated using
an ad hoc algorithm is calculated and compared to current DNS data for similar fibre suspensions in turbulent flows
(Marchioli et al. 2010).
Velocity profiles and bulk measurements of fibre suspensions in a turbulent pipe flow are measured using Particle
Imaging Velocimetry (PIV) and pressure readings. The effect of fibre concentration and Reynolds number on the
shape of the velocity profile and pressure drop along the pipe has been investigated for three different type fibres:
wood fibres, nylon fibres and a rigid rod (fibre-like) polymer, Xanthan Gum. Each fibre suspension has very different
pressure drop and velocity profile characteristics. The length scale of fibres are also very different which has a role
in the different fibre transport behaviours. In all fibre types it has been found that the behaviour does not degrade
with the length of time that fibres have been subjected to high shear stress nor with length of time fibres are allowed
to sit in the pipe circuit. Information about the distribution of the fibres within the pipe are extracted from number
density plots calculated and it is found at the lower Reynolds numbers examined the fibres tend to deposit at the wall.
Furthermore this Reynolds number effect is more pronounced at the higher concentrations.


Nomenclature AP Pressure drop (Pa)
U Mean velocity (m/s)
Roman symbols N Number density (-)
a,b,c,d,e,f numerical constants (-) < N > Normalised mean number density (-)
x,y cartesian coordinates S Fibre to fluid density ratio (-)
d, Fibre image length (m) a fibre semiminor axis (m)
I Intensity (W/m2) b fibre semimajor axis (m)
z Depth (m) Greek symbols
AZo Laser sheet thickness m 0 Orientation (0)











r] mass loading (-)
A Fibre aspect ratio (-)
Tp Fibre relaxation time (-)
Subscripts
o Fibre centre of mass
fibre Fibre suspension property
water Reference fluid (water) property
f Focus point location
x With respect to the streamwise direction
Superscipts
In fibre reference frame
+ Wall units


Introduction

Turbulent drag reduction has been observed experimen-
tally in suspensions involving a wide array of materials;
these include paper and cloth pulp (Radin et al. 1975),
asbestos (McComb, W. & Chan, K. 1985), and chopped
nylon (Lee et al. 1974; Kale et al. 1976). Dilute solutions
of "fibre-like" polymers, Xanthan gum and schizophyl-
lum polysaccharide, whose molecules having lengths
comparable to their persistence lengths have also been
observed to reduce drag Sasaki (1991). Although drag
reduction can be modest, they resist degradation unlike
polymer additives, and studying the behaviour of fibre-
laden flows may elucidate the mechanisms behind drag
reduction. The overreaching motivation of this work is
to characterise fibre suspensions in order to understand
more about the effect of different types of fibres in turbu-
lent pipe flows at high Reynolds number, such as those
typically used in commercial applications.
Fibres have a complex behaviour since they can ro-
tate and this leads to a preferential alignment in a shear
flow, such as that found in a turbulent layer (Marchi-
oli et al. 2010; Gillissen et al. 2007; Zhang et al. 2001).
Furthermore the hydrodynamic drag, which controls the
translational motion of the fibre is strongly coupled to
the fibre orientation. For example, a fibre depositing
parallel close to wall has a higher hydrodynamic than
one depositing perpendicular to the flow. The preferen-
tial alignment of the fibres generates large normal and
shear stresses. Paschkewitz et al. (2005) suggest that
these stresses are responsible for drag reduction mech-
anism since they oppose Newtonian accelerations in the
spanwise and wall normal directions thus weaken near-
wall structure.
The way in which fibres distribute/accumulate close
to wall and orientate plays an important role in the
turbulent-fibre interaction, hence the drag reducing
mechanism. For example it is believed that discrepan-
cies between numerical and experiments results lay in


7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


the presence of an inhomogeneous concentration distri-
bution in the experiment. Numerical experiments1 sug-
gest a monotonic increase in drag reduction with in-
creasing concentration(Paschkewitz et al. 2005) whereas
experimentally (Paschkewitz et al. 2005) drag reduction
stabilises around 1' for increasing concentration of
schizophyllum polysaccharide. The reason for this is un-
clear and the authors highlight a current lack of data for
comparison. It is suggested that the fibres may be inho-
mogeneously distributed within the turbulent boundary
layer, whereas the numerical model imposes a homoge-
neous fibre concentration distribution. There is evidence
to ,II-i.c-sI il.ii this is true in the case of fibres: Marchioli
et al. (2010); Zhang et al. (2001) find they preferentially
accumulate and segregate in a similar fashion to spher-
ical particles. However these results are obtained using
models are based on one-way coupling approach, where
the particles "feel" the fluid but fluid does not see the
fibres. There is still a need for validating these results
with "real world" situations.
To this aim Particle Image Velocimetry (PIV) im-
ages has been adapted to include "phase discrimina-
tion"(for more details see for example Khalitov & Long-
mire (2002)) which sorts tracer particles from non-tracer
objects. The phase discrimination algorithm based on an
object based approach which sorts objects according to
parametric combinations of fibre-length and aspect ratio.
Fibres are identified using an ellipse fitting algorithm
from which the number density and fibre orientation of
fibres is calculated using an ad hoc program developed
using matlab. The main challenge for this type of anal-
ysis, which will be discussed subsequently, is correct-
ing for the fibres that are perpendicularly orientated in
the visualisation plane, which therefore can be confused
with seeding particles.

Experimental facility

Experiments of dilute fibre suspension were carried out
on a large closed water loop (total working length is
31m) with a 0.lm pipe diameter. Pressure drop is calcu-
lated at regular intervals along the circuit, from pressure
measurements taken along the water loop using high ac-
curacy pressure transducers2. The Reynolds numbers
examined range from 70,000-200,000. The mean ve-
locities range from 0.7 m/s to 2.1 m/s. The velocity is
kept stable within I'. The smallest scales of the flow,
the kolmogorov's scale, calculated to range from 66 mi-
crons to 28 microns.
Visualisations, for PIV and fibre orientation data, are
carried out in a test section constructed from Plexiglas
'where the presence of fibres are modelled using a Fokker-Planck
type equations which are to describe fibres in a two-fluid model
2IMPRESS SENSORS; 0-400mBar gauge; 1% FS accuracy







7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


to allow visual access. This section comprises 2mm
thickness pipe surrounded by a box filled with water.
The pipe wall thickness inside the test section is thinner
than the rest of the pipe circuit. The latter reduces image
distortion due to refraction resulting from the curved
geometry of pipe which acts like a lens, whereas the
box acts to minimize image distortion due to refraction
resulting from change in material. The suspension is
visualized in the streamwise-wall normal plane at the
pipe centre line using a pulsed laser sheet and a CCD
camera. The system comprises a pulsed Nd:Yag laser,
optics for the light sheet formation, a 1280 x1024 PCO
Sensicam camera, a frame grabber and PC for image
acquisition and analysis. The camera field of view
was 17mm x 21mm with the longest direction in the
streamwise direction with a spatial resolution of 17
micron per pixel. Seeding particles used are around 20
microns. A total of 800 files are analysed at for each
Reynolds number and each concentration.



Phase discrimination technique

Phase discrimination is used to separate the fibres from
seeding/impurities in the captured images.
Raw images of laser illuminated fibres must be pre-
processed to remove the background noise. Sources of
background noise comprise: light reflecting off the cir-
cular pipe and fibres/particles behind the light sheet that
reflect light scattered by the particles/fibres within the
laser sheet. High frequency noise is eliminated using a
median filter: the median of neighboring pixels' val-
ues is calculated and repeated for each pixel in the im-
age. The image intensity is adjusted such that I'. of
data is saturated at low and high intensities of original
image. This increases the contrast of the output image
and makes the fibres appear brighter. Despite this pro-
cessing there was still a considerable amount of back-
ground noise. To magnify the intensity of the fibre the
image was dilated. This increases the brightness of ob-
jects by taking the neighborhood maximum when pass-
ing the structuring element over the image as well as
increasing the size of objects and fills holes and broken
areas. This worked well for fibres that were not illumi-
nated uniformly and could not be converted into binary
image as whole fibre even though to the naked eye they
could be identified as whole fibres. Once dilated, the
background noise was eliminated by thresholding, that
is any pixel with a value less than a certain limit (0.55)
was set to zero. After which images were eroded back,
such that fibres returned to their original size. Once
completed images are converted into binary images. The
pre-processing steps are summarised in Fig. 1
The image is then split into regions of 64 pixels by


Figure 1: Summary of main processing steps: far left-
raw image after contrast equalisation; centre-
image dilation; far right-image erosion.


64. The software scans the region for any bright pixels
and is then able identify connected particles. Fibres are
then selected using two parameters to distinguish them:
they had to be over a certain length and over a certain
aspect ratio (aspect ratio> 1.1). The centre coordinates
of pixels that make up the fibre are used as input for
ellipse fitting algorithm described below. Once fibres
have been identified a nonlinear least squares method is
used to fit the data points, x,z, onto an ellipse. Since
an ellipse is a special case of a general conic, it can be
described by a second order polynomial:

F(x, z) = ax2 + bz + cz2 +dx +ez + f = 0 (1)

with ellipse specific constraints:


b2 4ac < 0


where a, b, c, d, e and f are constants to be estimated,
x and z are coordinate measurement of the centre of
the pixels assigned as fibres. The polynommial F(x,z),
known as the algebraic distance of point x,z to a given
conic, can be rewritten as a vector:


where


F(aX) ,= X a 0


a = [a, b, c, d, e, f]


S[x2, xy, y, 1] (5)
The fitting of the general conic to the set of points,
Xi, zi, i=l,...., N, where xi, zi represent coordinates of
the centre of a pixel, can be approximated by minimiz-
ing the sum of squared algebraic distances of the point
to the conic represented by coefficient d (Fitzgibbon et
al. 1996):


N
mnin iF(xi, Zi)2
i 1


N
in i


The orientation is calculated from a non zero value of the
constant from equation 1, b, since this represents the tilt
of an ellipse in the x-z plane. The orientation of the fibre,
0x, is defined as the angle deviant from the streamwise







7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


axis parallel to the pipe, shown in Fig. 3 (a). Fig. 3 (b)
shows an ellipse fitted to a tilted fibre (in red) compared
to a reference fibre which is parallel to the mean flow.
Fig. 2 shows final result of the phase discrimination for
fibres.


(a) (b)


Figure 2: Raw image (a), results of phase discrimina-
tion algorithm (b). Pipewall is at the bottom of
the image.Insets show close up of fibres and
identification method.


Validation methodology

Only fibres that have been fit using the least squares
techniques are used to calculate statistics. This means
that fibres which are aligned or nearly aligned to the
spanwise plane and have small deviations from the
streamwise direction will not be included in this process-
ing step. In the limit case the fibre would be perpendic-
ular to streamwise direction and look like a dot. Fur-
thermore fibres that are partially illuminated by the laser
sheet could be misinterpreted as seeding/impurities. An
example of this can be found from Fig. 4 which shows
two artificial images of fibres illuminated by laser light
projected onto a 2D plane, one with random orienta-
tion and the other with no rotation about the wall-normal
axis. Another intrinsic limitation of using a 2D approach
is to capture 3D orientation is that the resulting angle ori-
entation is a projection of the actual fibre on a 2D plane.
In order to circumvent both limitations it is necessary to
use a Monte Carlo simulation. This provides pseudo-


Figure 3: Schematic of angle used for calculation of di-
rection cosine, 0, (a), An example of ellipse
fitted fibres(fit in red)and orientation of fibre
compared to reference ellipse, which is paral-
lel to the wall, in blue (b).


experiment information on the fibre position and orien-
tation within a laser sheet. In this way correction factors
fibre-orientation and fibre-number density have been de-
veloped. This approach uses a large number of data to
ensure statistical reliability. This type of approach has
also been used for the calculation of bubble intercept for
gas hold up calculations (Busciglio et al. 2010).
Artificial images are created by drawing 3D ellipsoids
with random orientation, projecting the fibre image onto
a plane and defining intensity distribution across the fi-
bre. An extension of the assumptions for spherical par-
ticles has been used; these are described as Gaussian in-
tensity profiles (Raffel et al. 2002). Fibres are assumed
to have a Gaussian intensity profiles along their minor
axes and a constant intensity along their major axes, up
to the focus points of the ellipse, as described:



I(x')=Ioexp [ ( )2] {x1fl < X1 < Xf2
(7)


Then intensity becomes a function of both axes:





I(x',y')=Io,ep [f I -Xf > { 1 >' > .

(8)


I is the intensity distribution, the centre of particle is
located at ,x, yo, x', y' refer to the coordinates on the
major and minor axis of the fibre, d, is the fibre im-
age length, xf, yf refer to focus point coordinates on the
major and minor axis of the fibre and fl, x f are the




















Figure 4: Artificial image -fibres only. Left hand side
shows the fibres generated with random ori-
entation. Right hand side shows the same fi-
bres without rotation about the wall normal
axis (as indicated by arrows).


focus point coordinates on the major axis. Io is the peak
intensity of the laser sheet defined as:


I(z)= q exp I
8p o


Az, is thickness light sheet, q is the efficiency with
which the particle scatters light. It is assumed that a fi-
bre, once projected into a 2D plane is at a fixed Z. This
simplification means that a fibre projections lie wholly
on one plane, hence the image does not possess depth.
Spherical particles for seeding/impurities are created us-
ing a relation for I(x,y) similar to equation 7 (Raffel et
al. 2002).
Once an artificial image has created, an instanta-
neous flow field for a turbulent channel is imposed on
the "artificial" seeding particles using a slice of DNS
data(Marchioli et al. 2010) similarly fibres are moved
according to particle flow field generated by the same
turbulent flow field(Marchioli et al. 2010). It is found
that with a small enough timestep between laser pulses
the fibres do not rotate. This means that the fluid flow
field around the fibre, once the fibre has been identified
and removed from the image, can be evaluated using a
cross correlation (PIV) algorithm.

Fibre characteristics

Nylon fibres are nominally 0.3mm in length and 14 mi-
crons in diameter with a specific gravity of 1.14. The
fibres tend to be uniform in length as confirmed from
images taken with microscope at a Ii.,gil.lik.ili', n of 10
x, as shown in Fig. 5 (a). Wood fibres tend to have a
large range of lengths, ranging anywhere from 10 mi-
crons to 3mm in size, as can be seen from Fig. 5 (b).
The wood fibre diameters estimated from these images
is approximately 20 microns. The specific gravity of the
wood calculated as 0.763.
Xanthan gum is a polysaccharide with a molecular
weight of 1.4 x 106. The persistence length of the


7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


polymer, a measure of chain flexibility, has been es-
timated to be between 330nm-760nm Paschkewitz et
al. (2005). Higher values indicate stiffer chains, hence
Xanthan gum is considered to be a rigid rod like. The
contour length, the maximum possible extension of a
polymer chain, is approximately 1000nm Meyer at al.
(1993). Using the contour length as the fibre length and
the diameter of polysaccharide molecule of 2.6nm, the
aspect ratio is 385. This considerably higher than aspect
ratio for nylon fibres (A z20), and larger than aspect
ratio for the longest wood fibre (A z50). Thus it can
be shown that ratio of the kolmogorov scale to the fibre
length ranges from 0.09 0.038 (becoming smaller with
increasing Reynolds) for the nylon fibres, 0.6 0.28 for
the shortest wood fibres, 0.03-0.02 for the longest wood
fibres and 65-45 for Xanthan gum. This indicates that
the length scale of Xanthan gum is smaller than the all
the scales of the turbulent flow. This is not true of the
macroscopic fibres.


Figure 5: High ill.ignilk .l i'il image of the wood fibres
(a) and nylon fibres (b) under study

The concentration of the fibres is defined using the
volume fraction p or the concentration parameter,
defined as nL3, where n is the number density of the
fibres and L is half length of the fibre-major axis. When
nL3 < 1 the solution is considered dilute, whereas when
1< nL3 < A the solution is considered semi-dilute.
Table 1 describes the concentration of the different fibre








7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


% change in drag


Table 1: Fibre concentrations used in
measurements


fibre type
nylon







wood






Xanthan Gum


mass fraction
0.01
0.02
0.05
0.1
0.5
1
0.01
0.05
0.1
0.5
1
0.05
0.1
0.15


pressure drop


nL3
0.018
0.035
0.089
0.177
0.890
1.78
0.073
0.363
0.725
3.633
7.298
0.002
0.004
0.007


I86 08


solutions studied.


% char in dng


Results


15


4 4


Fibre suspension characteristics

The effect of fibre concentration and Reynolds number
on the shape of the velocity profile and pressure drop
along the pipe has been investigated for three different
type fibres: nylon fibres, wood fibres and a rigid rod
("fibre-like") polymer, Xanthan Gum3. This is sum-
marised in Fig. 6, which shows drag reduction, defined
as Af.i t. for the different fibre suspensions.
It is found that the drag reduction calculated using phase
discriminated mean velocity profiles Usb ater is
within 2'. of the drag reduction found using pressure
drop profiles.
It is clear that the fibres act in different ways. Fig.
6(b) shows that Xanthan gum has a drag reducing ef-
fect at all Reynolds numbers and concentrations under
investigation. Maximum drag reduction is about 35 %.
This is considerably different to existing values found in
literature: Paschkewitz et al. (2005) found 12 % drag
reduction at Reynolds numbers based momentum thick-
ness of 490 (turbulent boundary layer) and Sasaki (1991)
found 6.5 % at Reynolds number based on pipe diame-
ter of 6800. This difference may be a Reynolds num-
ber effect and low Reynolds experiments are currently

3Wood fibres were donated by FANTONI industry, nylon fibres
bought from SWISS FLOCK, who, however also donated some
free fibres and Xanthan Gum was bought from CONTIPRO
GROUP


0 0 -


16. 1.8 2 2,2
X10


(c)

Figure 6: Pressure drop characterisation for nylon fibres
(a), xanthan gum (b) and wood fibres (c).



planned. It is found that there is a threshold Reynolds
which demarks regions of low drag reduction and high
reduction. Futhermore the drag reduction levels off at
the higher Reynolds to a constant gradient. This be-
haviour is reminiscent of polymer additive induced drag
reduction, where by, beyond a certain Reynolds number
drag reduction reaches a maximum, referred to as the
maximum drag reduction asymptote.
Nylon and wood behave very differently despite the
fact they are both macroscopic fibres. For wood fibres
it is found the behaviour of the fibres, shown in Fig.
6 (c), tends to that of water or is drag augmenting at
all concentrations at the Reynolds numbers under in-
vestigation with the exception of the lowest concentra-
tion. On the otherhand, Fig. 6(a) shows that the pressure
drop of nylon fibres is relatively insensitive to both in-


a 101


* 00cool,prova
0005, proval
A 01, proval
* 05 prova 1
C1,proval


Y *


1~6 08


12 14 1 6
Reynolds number


% chnge Inag


1.2 14 16 18l


2 22
x,10


4 4


. f


18 2 22
x105


06 0,8 1 1.2 1.4
Re













Table 2: Summary of relevant fibre parameters for com-
parison with numerical simulations.


Re
71,938
111,568
145,804
178,218
226,531


kolmogorov scale/b Re,
0.108 1737
0.077 2507
0.063 3131
0.0523 3513
0.044 4514


7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


all fibres, Np,i found in the region i and dividing by Np,i:


< cos(0) >= lEOkcos(x) (10)


A direction cosine close to 1 indicates alignment to
the streamwise direction. The normalised number den-
sity profile is defined as the sum fibres at equidistant
points along z+ divided by the total number of fibres all
the images. 800 images are used for the mean statistics.


creases in concentration and in Reynolds numbers. Ny-
lon fibres exhibit very small changes in pressure drop
which changes from drag reducing regime to drag reduc-
ing regime for a threshold Reynolds of about 140,000.
For similar Reynolds numbers and concentration Lee et
al. (1974) found drag reductions up to 19 %. It should
be noted that in the experiments described by Lee et al.
(1974); Radin et al. (1975); Kale et al. (1976) a surfac-
tant was used (Aerosol OT) to ensure good dispersion
of fibres, which was not used in our case. We are cur-
rently exploring the effect of this surfactant on the flow
properties.
The synergistic effect of fibres and polymers mixtures
is investigated. Lee et al. (1974) reports that using a
combined suspension of both fibres and polymers results
in a higher drag reduction that individual suspensions.
Here a mixture of wood fibres and Xanthan gum was
studied; no change in the measured drag reduction
compared to the pure Xanthan Gum case was found.
In both fibre types it has been found that the behaviour
does not degrade with the length of time that fibres have
been subjected to high shear stress nor with length of
time fibres are allowed to sit in the pipe circuit.

Fibre distribution and orientation statistics

The way in which fibres distribute/accumulate close
to wall and orientate plays an important role in the
turbulent-fibre interaction. For example: correctly
quantifying fibre deposition and accumulation at the
wall will permit us to evaluate local concentration,
which in turn will allows us to evaluate the effect of
localised fibre-fibre interaction and changes in viscosity.
Mean distribution and fibre orientation statistics are
presented for dilute suspensions of nylon fibres. The
reason for using nylon fibres, is that they most resemble
the fibres found in the DNS studies. Table 2 summarises
the relevant fibre parameters for comparison with
numerical data.
The statistics presented are found in the following
way. The mean direction cosines of orientations at
equidistant points along z+ are found by summing over


< N >= 1 = N
NP,, I
hpi ~


x10 3 C 0 01% normalised mean number density


5

24
E

E 3


0 Re=71,000
Re=111,568
1 Re=145,804
Re=178,218
Re=226,531
0 200 400


600 800
z m


1000 1200


(a)
103 C 0 02% normalised mean number density


S........... *-~.~++ +


o Re=71,000
Re=111,568
1 + Re=145,804
Re=178,218
+ Re=226,531
0 200 400
0 200 400


600 800
z+ m


1000 1200


(b)

Figure 7: Wall-normal fibre concentration profiles for
a fibre mass fraction of 0.01%(a) and 0.02
1)i).


Fig. 7 compares the wall normal fibre concentration
for two different mass fractions. It can be seen that there
are a greater number of fibres relatively close to the wall
at lower Reynolds numbers for a comparable wall nor-
mal distance (z+ = 100). This difference is more marked
for higher concentrations. Further away from the wall


' o * -*--ISB---' i








7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


the concentrations profiles collapse onto a single line,
in a similar fashion to Marchioli et al. (2010), albeit at
much higher z+.


x 10
6

5

S4
E
3
E
1 2
E
z1


3 Companons of both concentrations
0 C001, Re=71,000
o C002,Re=71,000

0 o Qj < <


x 10
6

-5
5

J4
E

E
2
E

z 1


Comparnons of both concentrations
S 001, Re=226,531
o C002, Re=226,531
S\- -- 4


0 100 200 300 400 500 600 700 800


0 100 200 300 400 500 600 700 800
z



S10 3 Companons of both concentrations
6
SC001, Re=111,568
SC002, Re=111,568


0 100 200 300 400 500 600 700 800
z



x 10 3 Companons of both concentrations
6
C001, Re=178,218
SC002, Re=178,218
5 -- -


0 100 200 300 400 500 600 700 800
z


Figure 8: Wall-normal fibre concentration profiles: in-
fluences of fibre mass fraction on each indi-
vidual Reynolds investigated.



The influence of concentration is studied through a
comparison of the concentration profiles for fibre mass
fraction of 0.0 1'. and 0.02 % at fixed Reynolds number.
Fig. 8 shows that at higher mass fractions there is a drop
in concentration close to the wall, with an equal or near
equal increase in concentration closer to the pipe centre-
line (bear in mind that the field of view is 1/3 of the pipe
radius). At the highest Reynolds numbers the difference
from the 0.01 % concentration becomes localised closer
to the wall than the centreline, such that main deviation
is a decrease in concentration close to the wall. This flux
may indicate that the transport mechanism has changed
when the mass fraction increased. There is no increase
in the normalised number density value. It would ap-
pear that a higher mass fraction there is a larger degree
of transport from near wall region.
Fig. 9 compares the absolute values of mean direc-
tion cosines for two different mass fractions. Close to
the wall, fibres are more strongly aligned than further
away from the wall. This trend can also be seen from
the numerical data, Fig. 10. Away from the wall the di-
rection cosines tend to the values of approximately 0.75
for mass fraction of 0.01 % and 0.8 for mass fraction
0.02 %. Marchioli et al. (2010) find a direction co-
sine approaching a value of 0.5 near the centreline. The
source of the different comes from the interpretation of
the mean direction cosines. At the centreline the tur-
bulence has a strong randomising effect, hence we ex-
pect all fibres orientations to be equally likely. When the
mean cosine direction fibre is calculated based on orien-
tations that are defined in 3D, as in the case of the DNS
calculations, it can be shown through calculus that the
mean direction cosine for an uniform distribution tends
to 0.5. Since in the imaging method the fibre is projected
into a 2D plane, it can be shown that the mean direction
cosine tends to 2/r.
The influence of concentration is studied through a











C 0 01% Icos(O6)l


0956
09
085U


Re=71,000
Re= 111,568
Re=145,804
Re=178,218
Re=226,531
200 400


800 1000 1200


(a)
C 0 02% Icos(Oe)1


095

085 '
n ^*


o Re=71,000
SRe=111,568
* Re=145,804
Re=178,218
+ Re=226,531
200 400


800 1000 1200


Figure 9: Absolute values of mean direction cosines for
a fibre mass fraction of 0.01' (a) and 0.02 %
(b)



comparison of the fibre orientation statistics for fibre
mass fraction of 0.01'. and 0.02 % at fixed Reynolds
number. Fig. 9 show that at higher mass fractions
there is an increase in alignments at all values from
the wall. This difference becomes more marked at
higher Reynolds numbers, lower T7. This trend is
in agreement with Mortensen et al. 12" is., who find
that faster particles become more aligned with the
wall. The reason for maintaining such a high degree of
alignment at such high values of z+, however, needs to
be investigated.



Conclusions and Future Developments

The main objective of this work was to characterise fi-
bre suspensions for high Reynolds numbers pipe flow.
To this aim we have presented a method using imaging
to identify fibres close to the wall in a turbulent flow and


7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010







0. ine ( 5 0 rs r

0,4 ,(o)- 30,4 (u) 10,()
S 50. (Marchioli40 60 1 et al. (40 20 10)) 1 120 1

Figure 10: Absolute values of mean direction cosines,
cos(O,)|; Panels:e a represent fibres with
low inertias, b represent fibres with higher
inertias (T =5 and 30 respect ively) where

e f 2(+)2sitting algo m ) Symbols: (o)
A 1.001, (o) A 3, (m) A 10, (o)
A 50. (Marchioli et al. (2010))





calculate the fibre orientation. The phase discrimination
algorithm based on an object based approach which sorts
objects according to parametric combinations of fibre-
length and aspect ratio. Fibres are identified using an
ellipse fitting algorithm from which the number density
and fibre orientation of fibres is calculated using an ad
hoc program developed using matlab. The intrinsic limi-
tations of this methods have identified and the validation
technique is presented.

The effect of fibre concentration and Reynolds num-
ber on the shape of the velocity profile and pressure drop
along the pipe has been investigated for three different
type fibres, wood fibres, nylon fibres and a rigid rod
(fibre-like) polymer, Xanthan Gum. It was found that the
fibre-turbulent interaction for each of these fibres types
is different and we are exploring the mechanisms be-
hind these differences using the methods presented. In
particular we examine fibre distribution and orientation
close to the wall. Mean distribution and fibre orientation
statistics are presented for dilute suspensions of nylon
fibres. Since nylon fibres are most like the fibres found
in the DNS studies, it has been possible to compare the
trends identified DNS data with those found with exper-
imental data. In the near future we will carry out ex-
periments with Reynolds numbers more similar to those
found in numerical studies.


Good agreement of trends have been found between
the DNS and numerical data. It has been found that wall
concentration profiles are sensitive to fibre mass frac-
tion: at a higher mass fraction there is a flux of fibres
away from the wall, toward the centre of the pipe. The
mean orientations, described using direction cosines,
show that there is strong alignment close to the wall.











Acknowledgements

Financial support from CIPE Comitato Interministeri-
ale per la Programmazione Economica under Grant on
"Caratterizzazione ed abbattimento di inquinanti e anal-
isi del rischio nei process di lavorazione del legno", and
from the Regional Authority of Friuli Venezia Giulia un-
der Grant "Nuove metodologie per la riduzione e la ges-
tione di emissioni di COV e particolato per lindustria di
pannelli di particelle e fibra di legno" are gratefully ac-
knowledged.
We would like to gratefully acknowledge Fantoni
group & Swiss Flock for fibre samples.
Thank you to Prof. Dr. Dieter Bothe and to Alessan-
dro Busciglio for the stimulating discussions during 12th
Workshop on two phase flow predictions, 22 25 March
2010, Halle-Wittenburg, Germany.

References

D. A, Khalitov. and E. K. Longmire, Simultaneous two
phase PIV by two parameter discrimination, Exp. Fluids,
Vol. 32, pp. 252-268, 2002

Marchioli C., Fantoni M. & Soldati A., Orientation, dis-
tribution, and deposition of elongated, inertial fibres in
turbulent channel flow, Phys. Fluids, 22, 2010

Sasaki, S., Drag reduction effect of rod-like polymer so-
lutions. i. Influences of polymer concentration and rigid-
ity of skeletal back bone, J. Phys. Soc. Japan 60, 868878,
1991.

McComb, W. & Chan, K., Laser-Doppler anemometer
measurements of turbulent structure in drag-reducing fi-
bre suspensions, J. Fluid Mech., 152, 455478, 1985.

Lee, W., Vaseleski, R. & Metzner, A., Turbulent drag
reduction in polymeric solutions containing suspended
fibres, AIChE J. 20, 128133, 1985.

Kale, D. & Metzner, A., Turbulent drag reduction in
dilute fibre suspensions: mechanistic considerations,
AIChE J., 22, 669674, 1976.

Paschkewitz J.S., Dmitropoulos C.D., Hou Y.X., So-
mandepalli V.S.R., Mungal M.G., Shaqfeh E.S.G. &
Moin P., An experimental and numerical investigation
of drag reduction in a turbulent boundary layer using a
rigid rodlike polymer, Phys. Fluids, 17, 2005.

Radin, I., Zakin, J. & Patterson, G., Drag reduction in
solidfluid systems, AIChE J., 21, 358371, 1975.

Zhang H., Ahmadi G., Fan F.G., and McLaughlin J.B.,
Ellipsoidal particles transport and deposition in turbu-
lent channel flows, Int. J. Multiphase Flow,27, 971-
1009, 2001.


7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 -June 4, 2010


Krochack P.J., Olson J.A., Martinez D.M., The stress
generated by non-Brownian fibres in turbulent channel
flow simulations, Phys. Fluids, 19, 115107, 2007.

Paschkewitz J.S., Dmitropoulos C.D., Hou Y.X., So-
mandepalli V.S.R., Mungal M.G., Shaqfeh E.S.G., and
Moin P, The dynamic mechanism for turbulent drag re-
duction using rigid fibres based on Lagrangian condi-
tional statistics, Phys. Fluids, 17, 063102, 2005.

Meyer, E. L., Fuller, G. G., Clark, R. C, Kulick, W.
M, Investigation of Xanthan Gum solutions behaviour
under shear stress using Rheoptical techniques, Macro-
molecules, 26, 504-511,1993.

Fitzgibbon, A. W., Pilu, M and Fischer, R. B., Direct
least squares fitting of ellipses, In Proc. of the 13th In-
ternational Conference on Pattern Recognition Vienna,
September 1996, pages 253257, 1996

A. Busciglio, F.Grisafi,F.Scargiali,A.Brucato, On the
measurement of local gashold-up and interfacial area in
gasliquid contractors via light sheet and image analysis,
Chemical Engineering Science, in press.

Raffel, M., Willert, C. & Kompenhans, J. Particle
Imaging Velocimetry: a practical guide Springer; 2nd
edition, 2002.

Mortensen P.H., Andersson H.I., Gillissen J.J.J., and
Boersma B.J., Dynamics of prolate ellipsoidal particles
in a turbulent channel flow, Phys. Fluids, 20, 093302,
2008a.




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.7 - mvs