Professional Documents
Culture Documents
463
Abstract A horizontal turbulent boundary layer of air carry- U fluid free-stream velocity (m s~1)
ing heavy solid particles is investigated experimentally. Mean v@ vertical r.m.s. velocity of the fluid (m s~1)
f
and r.m.s. velocities of air and particles are measured by LDA, v@ vertical r.m.s. velocity of the particles (m s~1)
p
and particle mass flux distributions are obtained by means of V particle free fall velocity (m s~1)
g
a sampling method. The influence of the saltation mechanism x Cartesian co-ordinate in the streamwise direction
is revealed by the large particle r.m.s. velocity in the near-wall (m)
region, and by the velocity lag of the particles in the outer y Cartesian co-ordinate in vertical direction (distance
region of the boundary layer, which is shown to be closely from the bottom wall) (m)
related to their free fall velocity. The present original results are y` dimensionless distance from the wall (\yu /l)
q
discussed and compared with available experimental data z horizontal transverse Cartesian co-ordinate (dis-
concerning other kinds of horizontal flows. tance from the side wall) (m)
d boundary layer thickness (m)
List of symbols e mean volume fraction of the solid phase
d particle diameter (m) c density parameter [\(o [o )/o ]
p p p f f
dI dimensionless particle diameter [\d (gc /l2)1/3] l fluid kinematic viscosity (m2 s~1)
p p p
g gravitational acceleration (m s~2) o fluid density (kg m~3)
f
q volume flow rate (m3 s~1) o particle density (kg m~3)
V p
Re Reynolds number in the experiments of Kaftori et al. q Kolmogorov time scale (s)
h l
(1995) q time scale of large turbulent structures (s)
L
Re Reynolds number based on the momentum thick- q particle relaxation time (s)
h p
ness
S Stokes number q /q
l p l
S Stokes number q /q 1
L p L
u
f
mean fluid velocity (m s~1) Introduction
u mean particle velocity ( s~1) Although two-phase flows are very common in industrial and
p
u` dimensionless fluid velocity (\u /u ) environmental processes, their knowledge is still incomplete.
f q
u@ streamwise r.m.s. velocity of the fluid (m s~1) As an example, we can mention the problem of contaminant
f
u@ streamwise r.m.s. velocity of the particles (m s~1) dispersion in the atmosphere, which is directly related to the
p
u friction velocity (m s~1) understanding of take-off phenomena and particle behaviour
q
uJ dimensionless friction velocity [\u (gc l)~1/3] in a turbulent boundary layer. This paper presents an experi-
q q p
mental investigation related to the problem of solid particle
transport in a horizontal boundary layer.
The mechanism of particle transport in boundary layers has
Received: 17 June 1996 /Accepted: 3 April 1997 been investigated little as yet, except in the case of very fine
particles, whose behaviour is purely induced by turbulent
A. Tanière, B. Oesterlé diffusion (Poreh and Hsu 1971; Lee and Dukler 1981). With
Laboratoire Universitaire de Mécanique et d’Energétique de Nancy
increasing particle size, inertia effects and body forces may
(LUMEN) ESSTIN, Université Henri Poincaré-Nancy 1,
F-54500 Vandoeuvre-les-Nancy, France play a significant role on the dispersion process. An experi-
mental study of the dispersion of solid particles in a vertical
J. C. Monnier boundary layer was carried out by Rogers and Eaton (1990),
Institut de Mécanique des Fluides de Lille (IMFL) who showed that the mean fluid flow is not affected by the
5, boulevard Paul Painlevé, F-59000 Lille, France presence of the particulate phase at low loading, and that the
particle velocity fluctuations are strongly attenuated in the
Correspondence to: B. Oesterlé
transverse direction. Particular attention has been paid re-
This research was carried out at the Départment d’Energétique cently on the effect of organized turbulent structures: particle
Industrielle, Ecole des Mines de Douai, France, thanks to Prof. transport by the bursting events of the wall region has
B. Baudoin, whose assistance is gratefully acknowledged. been studied by Rashidi et al. (1990), whereas Eaton and
Fessler (1994) pointed out the occurrence of preferential
concentration due to coherent vortical structures. The role of
such coherent structures has been investigated experimentally
by Komori et al. (1995), as well as by Isdiyana (1993) or
Ushijima and Perkins (1995), who used a Lagrangian tech-
nique. The most comprehensive experimental work concern-
ing the particle behaviour in a horizontal turbulent boundary
layer is due to Kaftori et al. (1995), who studied the motion of
solid particles near the wall in a water channel by means of
visualization techniques and LDA, and measured the velocity,
flux and concentration distributions.
464 Studies related to particle take-off phenomena in a turbulent
boundary layer have been initiated by Bagnold (1943), who Fig. 1. Various behaviours illustrated by Foucaut’s take-off curve
showed that particles take off in the same manner whatever (q is defined by q \v@ /g)
g g f
their diameters are. After the take-off, various behaviours
are observed, since particles are more or less influenced
by the turbulence. It has been proposed recently to define
A B
gc 1/3
four specific behaviours (Foucaut et al. 1996), which can be dI q\dp p (4)
characterized by criteria based on some time scale ratios which v2
are similar to Stokes numbers. The particle relaxation time
where u is the friction velocity, d is the particle diameter, l is
q is compared to the time scale q and q of the smallest q p
p l L the fluid kinematic viscosity, g is the gravitational acceleration,
(Kolmogorov) and biggest turbulent structures, respectively.
and c \(o /o )[1 (o particle material density, o fluid
For very small particles, which satisfy p p f p f
density). For a particle with given d and o , the minimum
p p
q take-off friction velocity can be deduced from the take-off
Sl\ p@1 (1) curve in Fig. 1.
ql
The aim of the present study is to identify the particle
the velocity fluctuations are practically equal to those of the behaviour in the flow after take-off, by means of the experi-
fluid. This behaviour is called ‘‘pure suspension’’. On the mental investigation of a horizontal air—solid turbulent
contrary, ‘‘pure saltation’’ is observed for particles whose boundary layer on a flat plate. The particle concentration is
behaviour is not influenced by turbulence effects, but is sufficiently low for the disturbance of the flow and the effect of
governed by inertia, gravity, and particle-wall collision phe- collisions between particles being negligible. After describing
nomena, keeping in mind that we are dealing with a horizontal the experimental method, the flow conditions and the particle
boundary layer. Such particles, which exhibit ballistic trajecto- characteristics, original results concerning the profiles of mean
ries, are characterized by the following inequality: and r.m.s. velocities of both phases, as well as the mass flux
distributions, are presented and discussed with reference to the
q work of Kaftori et al. (1995).
SL\ p A1 (2)
qL
Between such asymptotic behaviours, i.e. when the particle 2
relaxation time is of the same order as some time scales of the Experimental apparatus and techniques
fluid turbulent fluctuations, one may expect some evolution of
the mean relative velocity of the particles with respect to the 2.1
fluid, as well as some changes in the effect of particle—wall Wind tunnel
interactions. This kind of mixture is called herein ‘‘interactive The wind tunnel used in this study, sketched in Fig. 2, has
medium’’. In such a medium, two possible behaviours can be a total length of 13.2 m. The co-ordinate system referred to in
distinguished, according to the value of the ratio V /v@ , where the following sections is shown in the figure. The working
g f section is 5.7 m long, and has a 0.5 m wide, 0.47 m high
V is the particle free fall velocity, and v@ is the vertical r.m.s.
g f rectangular cross-section, with smooth walls. The boundary
velocity of the fluid. The parameter V /v@ , which is often called
g f layer thickness is about 10 cm at the end of the test section. The
the crossing-trajectory parameter, was introduced by Yudine
(1959). If this quantity is smaller than 1, we define a ‘‘modified maximum velocity is 25 m/s. A rough paper tape is set at the
suspension state’’. Should it exceed 1, the ‘‘modified saltation wall between the end of the contraction and the inlet of the test
state’’ is defined. Hence, a continuous progress between the section, in order to activate the boundary layer transition. Solid
two asymptotic behaviours is obtained, as illustrated in Fig. 1 particles are introduced in section T1 (see Fig. 2), at x\3.17 m.
by Foucaut’s take-off curve (Foucaut et al. 1996), which is Velocity and mass flux measurements are carried out in section
a plot of the dimensionless take-off threshold friction velocity T2, which extends from x\5.0 m to x\5.6 m, by Laser
uJ as a function of the dimensionless particle diameter dI , these Doppler anemometry and by sampling, respectively. The
q p Dantec Laser equipment is made up of a one-component argon
two quantities being defined by
Laser (power 1.6 W, wave length 514.5 nm) and a Burst
uq Spectrum Analyser (BSA), intended to perform the photomul-
uJ q\ (3)
(gcpv)1/3 tiplicator signal analysis.
Fig. 2. Experimental facility
465
A B
ld 1/2 d
ql+ , qL+ (5) by the wind tunnel can be considered to be of satisfactory
(0.1 U)3 0.1 U quality.
The influence of the particles on the fluid flow can be
Two kinds of spherical particles have been used in our estimated in measuring the fluid velocity profile in the
experimental investigations: 60 lm diameter glass beads presence of solid particles. Measurements have been made
(density 2500 kg/m3) and 130 lm diameter PVC beads (density possible by using incense smoke as tracer particles for the fluid
1430 kg/m3). Diameter standard deviations, reported in flow, and in adjusting the apparatus in order to exclude the
Table 2, show that the PVC particles have a rather wide size signals coming from the glass or PVC beads, so that only
distribution. The corresponding dimensionless diameters dI , signals diffused by tracer particles are detected. Such a distinc-
p
according to Eq. (4), are 2.7 and 4.9, respectively. From Fig. 1, tion was performed via the Burst Spectrum Analyser, using an
the values of the take-off friction velocities can be deduced for amplitude discrimination technique. As emphasized by Kulick
each kind of particles: one obtains 0.20 m/s for d \60 lm, and et al. (1993), inter-phase crosstalk may be caused by large
p
0.17 m/s for d \130 lm. It can be concluded that the present particles passing through the edge of the measurement volume,
p
friction velocity of 0.40 m/s is high enough for both kinds of thus having scattering intensities which are comparable with
particles to take off and to be transported by the boundary those of incense tracer particles. In order to reduce this source
layer flow. In Table 2, q denotes the volumetric solid flow of uncertainty, the number of fringes has been chosen as large
V
rate, and e is an estimate of the mean volume fraction of the as possible (namely 32), and particular attention has been paid
solid phase. The particle relaxation time q and free fall velocity to the scattered light amplitude reduction, as recommended by
p
V , which are related by V \g q , were calculated using the Fessler and Eaton (1995).
g g p
drag coefficient correlation proposed by Morsi and Alexander The mean fluid velocity profiles obtained at x\5.06 m for
(1972), inasmuch as the particle Reynolds numbers based on V the pure gas flow and for the particle laden flow (with 60 lm
g
lie in the range 1—10, thus precluding the use of Stokes’ law. particles) are presented in Fig. 5, which is a logarithmic
467
Fig. 6. Influence of the solid phase on the r.m.s. fluid velocity profiles
(x\5.06 m); d Without particles; n with particles (d \60 lm) Fig. 8. Particle and fluid mean velocity profiles at x\5.51 m: d air;
p n 60 lm particles; K 130 lm particles (the solid line represents the
law of the wall u`\2.43 ln y`]5.6)
presentation of the dimensionless velocity u`\u /u as boundary layer, due to the very low concentration. In Fig. 7,
f q
a function of the dimensionless wall distance y`\yu /l (u the dimensionless mean velocity profiles of the solid phase at
q f
being the local mean fluid velocity). Comparison between both x\5.06 m are displayed and compared with the mean fluid
profiles shows that the flow is not significantly altered by the velocity profile. The same comparison at x\5.51 m is illus-
particles. A slight reduction of the turbulence level near the trated in Fig. 8. The fluid and particle distributions are seen to
wall can be observed in examining the streamwise r.m.s. fluid have similar shapes. However, it can be noticed that the mean
velocity profiles (Fig. 6). However, such a reduction was not particle velocity is lower than that of the fluid, except very close
detected for the 130 lm particles. Taking into account the very to the wall. Such velocity distributions are similar to the
low concentration of the solid phase, whose local volume profiles observed in gas-solid pipe flows, e.g. by Tsuji and
fraction does not exceed 10~4, even near the wall, it is believed Morikawa (1982). A linear plot of the reduced velocities u /U
p
that the apparent turbulence level reduction is rather due to the and u /U as a function of y/d is displayed in Fig. 9, in order to
f
noisy character of the data. examine the particle behaviour near the wall. It is seen that the
particle velocity noticeably deviates from the fluid velocity, this
3.2 effect being more pronounced for larger particles. For both
Mean velocity profiles diameters, the particle mean velocity reaches a nonzero value
Particle mean velocities have been measured in averaging over at the wall. The presence of such a relative velocity between the
no less than 2000 particles, except near the outer edge of the two phases confirms that the particle behaviour belongs to the
468
469
However, it is a little surprising that the streamwise particle
turbulence intensities are larger than those of the fluid. This
phenomenon, which is related to the measurement technique,
may be explained in keeping in mind that we are dealing here
Fig. 13. Vertical r.m.s. velocity profiles in the near-wall region
with a flow where particles undergo the effect of inertia and
(x\5.06 m) r v@ /u , n v@ /u (60 lm), K v@ /u (130 lm)
gravity. So, among the particles which are located within the f q p q p q
probe volume of the LDA, some are moving upwards and some
are moving downwards, as illustrated by Fig. 12. On the
average, the upwards moving particles have a smaller horizon- smaller than the value of v@ outside the near-wall region. These
p
tal velocity, due to the wall collision they have undergone, and observations clearly show that there are two different behav-
the downwards moving particles have a larger horizontal iours in the interactive medium. The difference lies essentially
velocity, since they have been accelerated by the fluid for in the characteristics of the transverse (or vertical) velocity
a longer time. This argument is consistent with the experi- fluctuations in the near-wall region.
mental trajectories reported by Ciccone et al. (1990), who
developed an original technique to track saltating sand 3.4
particles in a turbulent boundary layer. Such velocity differ- Mass flux measurements
ences are taken into account by the BSA in performing the It is well known that isokinetic sampling is theoretically
averaging operation, and that is the reason why the particle required for the concentration measurements to be accurate.
r.m.s. velocities plotted in Fig. 11 must be interpreted as Isokinetic conditions, which are completed when the velocity
agitation velocities, not only due to fluid turbulence but also of air through the entry of the probe matches that of the
due to inertia and gravity effects. The present results indicate approaching stream, are obtained by adjusting the suction flow
that the latter effect may prevail over the turbulence effect in rate. However, Azzi (1986) showed that there is a range of
the case of saltation flows. Another cause of particle velocity sampling flow rates where the sampled mass is only weakly
fluctuations may be the size distribution, since the particles dependent upon the suction flow rate. This fact is confirmed by
used here cannot be regarded as monodisperse. Soo et al. the results displayed in Fig. 14, which is a plot of the mass of
(1960), who experimentally studied a horizontal gas—solid pipe particles collected during 60 s as a function of the suction flow
flow, with 100 lm and 200 lm glass particles, observed the rate, for three different locations of the probe. It is seen that,
same fact but they did not give any explanations. In the same for a suction rate lying between 2.5 and 3.5 m3/h, the measured
kind of flow, Tsuji and Morikawa (1982) also obtained large mass flux does not depend on the suction velocity (arrows
fluctuating velocities for 200 lm plastic particles. These indicate the theoretical flow rate for isokinetic sampling).
authors pointed out that the main mechanism of particle Additional tests were carried out by performing, for both kinds
velocity fluctuations is different from gas turbulence. Particle of particles, a mass balance using the measured values of
velocity fluctuations exceeding the fluid turbulent fluctuations the mass fluxes in terms of the height y: the overall mass
have also been observed in a liquid-solid channel flow by conservation was satisfied within 8%. Such preliminary
Kaftori et al. (1995), whose results will be discussed and measurements allow to ascertain the reliability and the
compared with the present results in the next section. accuracy of the mass flux distributions presented hereafter.
Contrary to streamwise fluctuations, results concerning the In Fig. 15, profiles of particle mass flux at x\5.06 m have
transverse velocity fluctuations near the wall show a significant been presented in plotting the percentage of the sampled mass
difference between both kinds of particles, as can also be against the total collected mass versus the dimensionless wall
seen in Fig. 13, which is a zoom of the near-wall region. In distance y`. Samplings were performed each 3.5 mm, which
particular, a peak appears on the 130 lm particle curve, corresponds to the height of the inlet section of the probe.
revealing phenomena which take place at the wall, connected As could be expected, the particle flux is decreasing with
with particle rebounds. In carefully examining the values of increasing distance from the wall for both kinds of particles,
v@ /u near the wall, and taking into account V /u \0.63 for the mass flux profiles exhibiting an exponential shape. It
p q g q
d \60 lm and V /u \1.3 for d \130 lm, it is found that the can be observed that the 60 lm particles, which are lighter,
p g q p
maximum value of v@ is close to the value of the free fall reach higher elevations than the 130 lm, and that their flux
p
velocity. The absence of such a peak for the 60 lm particles distribution is a little more homogeneous. Near the wall, the
would therefore be due to the fact that their free fall velocity is 130 lm particle flux is larger, due to the gravity force, which
470
Fig. 14. Influence of the suction flow rate on the mass collected by the
Fig. 16. A comparison between the present mass flux profiles and the
sampling probe
results of Kaftori et al. (1995): K Present results, d \130 lm, S \58.5,
p l
S \0.867; n Present result, d \60 lm, S \27.9, S \0.41; ] Kaftori
L p l L
et al. (1995), d \900 lm, S \1.50, S \0.045; s Kaftori et al. (1995),
p l L
d \275 lm, S \0.138, S \0.0041; j Kaftori et al. (1995),
p l L
d \100 lm, S \0.018, S \0.0005
p l L