Sie sind auf Seite 1von 8

Acta Mech. Sin.

(2011) 27(4):494501
DOI 10.1007/s10409-011-0471-7

RESEARCH PAPER

Measurement and simulation of the two-phase velocity correlation


in sudden-expansion gasparticle flows
Li-Xing Zhou Yang Liu Yi Xu

Received: 8 March 2010 / Revised: 22 July 2010 / Accepted: 9 August 2010


The Chinese Society of Theoretical and Applied Mechanics and Springer-Verlag Berlin Heidelberg 2011

Abstract In this paper the present authors measured the


gasparticle two-phase velocity correlation in sudden expansion gasparticle flows with a phase Doppler particle
anemometer (PDPA) and simulated the system behavior
by using both a Reynolds-averaged NavierStokes (RANS)
model and a large-eddy simulation (LES). The results of the
measurements yield the axial and radial time-averaged velocities as well as the fluctuation velocities of gas and three
particle-size groups (30 m, 50 m, and 95 m) and the gas
particle velocity correlation for 30 m and 50 m particles.
From the measurements, theoretical analysis, and simulation, it is found that the two-phase velocity correlation of
sudden-expansion flows, like that of jet flows, is less than the
gas and particle Reynolds stresses. What distinguishes the
two-phase velocity correlations of sudden-expansion flow
from those of jet and channel flows is the absence of a clear
relationship between the two-phase velocity correlation and
particle size in sudden-expansion flows. The measurements,
theoretical analysis, and numerical simulation all lead to the
above-stated conclusions. Quantitatively, the results of the
LES are better than those of the RANS model.
Keywords PDPA measurement Simulation Two-phase
velocity correlation Sudden expansion gasparticle flows
The project was supported by the National Natural Science Foundation of China (50606026 and 50736006).
L.-X. Zhou ( ) Y. Liu Y. Xu
Department of Engineering Mechanics,
Tsinghua University, 100084 Beijing, China
e-mail: zhoulx@mail.tsinghua.edu.cn
Y. Liu
Marine Engineering College,
Dalian Maritime University,
116026 Dalian, China

1 Introduction

The gasparticle velocity correlation term hvgi vpi i (where the


subscripts g and p refer to gas and particle, respectively) is
important in two-fluid models, such as in the unified secondorder moment (USM) model or the two-phase Reynolds
stress model proposed by Zhou et al. [1,2]. The role of this
correlation term is akin to that of the pressurestrain term
in the single-phase Reynolds stress equation. In the USM
model, hvgi vp j i is closed by either dimensional analysis or by
a transport equation. The rationality of these closure models
needs to be validated by both experiments or direct numerical simulation (DNS) and large-eddy simulation (LES).
The relationship between the two-phase velocity correlation hvgi vp j i and the gas Reynolds stress hvgi vg j i, particle Reynolds stress hvpi vp j i, and particle size or relaxation
time needs to be studied both theoretically and experimentally. The phase Doppler particle anemometer (PDPA) can
measure not only the gas velocity, particle size and velocity, and Reynolds stresses for gas and particles, but also the
two-phase velocity correlation. Prevost et al. [3] measured
hvgi vp j i using a PDPA in the far field of an axisymmetric jet,
and Wang et al. [4] made the measurement in a gasdroplet
mixing layer. Their results indicate that the two-phase velocity correlation is less than the Reynolds stresses for both
gas and particles, and that it decreases with the increase of
particle size or particle Stokes number. Recently, Longmire
and Khalitov [5] measured the gasgas, gasparticle, and
particleparticle correlations in channel flows using simultaneous two-phase particle image velocimetry. They found
that the spatial distributions of all these correlations were
similar to each other. The gasgas correlation was found
to be greater than the gasparticle correlation, which in turn
was found to be greater than the particleparticle correlation.
In addition, the gasparticle velocity correlation was found
to decrease with the increase of particle size.

Measurement and simulation of the two-phase velocity correlation in sudden-expansion gasparticle flows

However, the behavior of two-phase velocity correlation in highly sheared separating flows, such as suddenexpansion flows, is unclear. Laser Doppler velocimeter measurements of sudden-expansion gasparticle flows
were made by Shahnam and Morris [6] and Hishida and
Maeda [7], but no information about gasparticle velocity
correlation was reported. In a numerical simulation, Simonin
et al. [8] applied DNS and LES to validate their two-fluid
models, including the closure models of gasparticle correlation, but they did this only for isotropic homogeneous turbulent gasparticle flows. Thus, in this paper, the present
authors study gasparticle velocity correlations for suddenexpansion gasparticle flows by using phase Doppler particle
anemometer (PDPA) measurements and a large-eddy simulation (LES) with a two-phase subgrid scale stress model
proposed by the present authors. Also, a Reynolds-averaged
NavierStokes (RANS) modeling using a USM two-phase
turbulence model [1] was applied. The measurement results
are analyzed using theoretical expressions, and the numerical simulation results are validated by measurement results.
2 Experimental setup and measurement technique
The experimental setup for measuring sudden-expansion
gasparticle flows is shown in Fig. 1. It includes the test
section, a section for supplying air, and the measurement
instrumentation. The upper part of the test section is a
610 mm long tube made of plexiglass with an inner diameter of 120 mm. The diameter of the inlet on the top of the
test section is 50 mm. A piece of optical glass is mounted
on one side of the upper part of the test section in a slot.

495

The lower part consists of a tube of 1 500 mm long, which


has a larger sudden-expansion chamber is connected to its
exit to protect the test section disturbances from downstream
flows. The air-supply system consists of a blower, valves,
and a D-shaped flow meter. The exit of the test section is
connected to a blower, which closes the gas circuit. The benefit of the cyclic gas circuit is that the powder feeder may be
omitted. Therefore, the apparatus does not suffer from disturbances due to mechanical powder feeding, which makes
the particle flow more stable and uniform and reduces the
number of particles used. Initially, the particles were fed into
the system via the inlet. The measurement instrumentation
consisted of a three-dimensional (3D) phase Doppler particle anemometer (PDPA, model 58N50, Dantec Company,
Skovlunde, Denmark), which was used in the backward scattering configuration.
To measure the gasparticle velocity correlation, the
particle arrival time must be much shorter than the gasfluctuation time scale. In addition, to ensure that enough particles arrive within a short time interval and to avoid a large
mass-loading ratio, the two-dimensional measurement technique was adopted, keeping the data efficiency above 60%.
Spherical glass beads ranging in sizes from 5 m to 130 m
were used in the measurements, and glass beads smaller than
10 m were used as gas tracers so as to avoid the difficulties
associated with (i) the difference in refractive index between
seeding particles and dispersed-phase particles and (ii) the
interference between particles. The broad size distribution
(up to 130 m) made it possible to observe different particle behavior for different size particles, but it extended the
arrival-time interval for particles of a given group.

Fig. 1 Experimental setup

Measurements were taken at cross sections 39 mm,


119 mm, 248 mm, and 495 mm downstream from the inlet.
The ratios of these distances to the step height of 35 mm are
1.1, 3.4, 7.1, and 14, respectively. The time limit was set
to 3 minutes for measurements at each point, and 100 000
samples were acquired. Three size groups (30 m, 50 m,
and 95 m) with widths of 10 m were chosen to express
the particle phases; particles with the size range of 5 m to

10 m were used as gas tracers. The inlet volumetric flow


rate was 212.4 m3 /h, and the particle volumetric loading ratio
was about 0.005%. Therefore, particleparticle interactions
were negligible.
3 Measurement results and discussion
The axial gas velocity, radial gas velocity, particle-fluc-

496

L.-X. Zhou, et al.

tuation velocity, and the two-phase velocity correlation are


shown in Figs. 2 and 3. In the recirculation zone, the maximum of the two-phase fluctuations corresponds to the maximum mean-velocity gradient. In the downstream region,
the radial fluctuations are intensified near the wall, because
of particle rebound. The 30 m and 50 m particles exhibit
axial-fluctuation velocities different from that of the gas, but
their radial fluctuations decrease significantly with the increase of particle size. The 95 m particles exhibit axial fluctuations different from those of the gas and the smaller particles. The radial fluctuations of the 95 m particles are less
than those of the gas and the smaller particles, except near
the wall in the downstream region, where these fluctuations
are similar.

Fig. 3 Radial Reynolds stresses for gasparticles and two-phase


velocity correlation (m2 /s2 )

Fig. 2 Axial Reynolds stresses for gasparticles and two-phase


velocity correlation (m2 /s2 )

The gas and particle fluctuations were noticeably


anisotropic in all regions. The axial fluctuations were much
larger than the radial fluctuations, and the anisotropy increases with the increase of particle size. To measure the correlation between the gas and particle velocity fluctuations,
the velocity correlation is analyzed as a function of time (see
Fig. 4). The gasparticle velocity correlation function from
t = 0 to t = is ugi upi = ugi (t)upi (t + ), which is determined
from experimental results. The two-phase velocity correlation at = 0 is obtained by interpolating the data using an
exponential function that is fit to the data.

Because the characteristic time of the two-phase velocity fluctuations is determined by both the particle relaxation
time and the time scale of gas turbulence, which is a variable
throughout the flow field, choosing the time interval for the
two-phase velocity correlation function is very important and
requires careful thought. The Eulerian time scales, which are
estimated from the correlation functions of time, are given
in Fig. 5, which shows that the Eulerian time scale changes
dramatically from the recirculation region to the downstream
region. Because the characteristic time of the two-phase fluctuation velocity decreases with particle relaxation time, the
characteristic time for the two-phase velocity correlation for
95 m particles is so small that it is difficult to be obtained.
The two-phase velocity correlation for 30 m and
50 m particles is shown in Figs. 2 and 3. The profiles of
hugi upi i are similar to those of hugi ugi i and hupi upi i. The functions hugi upi i, hugi ugi i, and hupi upi i have peaks in the shear
layer. In most regions, the two-phase fluctuation velocity
correlation for 50 m particles is slightly less than that for
30 m particles. In addition, the axial component of the gas
particle velocity correlation is much smaller than the normal
Reynolds stresses for gas and particles, and the same may be
said for the radial component, with the caveat that the difference is much smaller. Similar to normal Reynolds stresses
for gas and particles, the two-phase velocity correlation is
also anisotropic (its radial component is smaller than its axial component).

Measurement and simulation of the two-phase velocity correlation in sudden-expansion gasparticle flows

497

Fig. 4 Two-phase velocity autocorrelation and cross-correlation function

These results are qualitatively similar to those obtained


by Prevost et al. [3] for a gasparticle jet. However, because
of the strong shear layer in sudden-expansion flows, many
new features appear in the present results. In gasparticle
jets, hvg vp i is very small (only about 0.2 m2 /s2 ), so Prevost
et al. [3] suggested that hvg vp i can be neglected in the transport equation for hvp vp i. In the present study, hvg vp i is about
3 m2 /s2 which, although smaller than hug up i (1018 m2 /s2 ),
but is much larger than that in the jet and so can not be neglected. According to Prevost, hvgi vpi i should decrease with
the increase of particle size in the range of 10 m to 45 m,
but this tendency is not clear in the present study.

Looking at transport equations, it is easy to understand


that hup up i will peak in regions where the velocity gradient is large, such as in the shear layer. However, in the
gasparticle jet, the peak value of hup up i occurs on the axis,
which seems unreasonable. Because the streamwise, transverse and shear stresses maintain the same profiles in the
mixing-layer flow [4], the conclusion that the two-phase velocity correlation function expresses turbulent kinetic energy
and shear stresses is not a universal conclusion. In complex
sudden-expansion flows, the axial and radial two-phase velocity correlations have the same profiles in each respective
direction as those of the Reynolds stresses of gas and parti-

498

L.-X. Zhou, et al.

cles. The dependence of the two-phase velocity correlation


on the two-phase turbulent kinetic energy can not represent
the anisotropy of turbulence.

4 LES and RANS modeling


4.1 LES governing equations
For a two-fluid LES and neglecting all forces acting on the
particles other than the drag force, the filtered continuity and
momentum equations for gas and particle phases can be obtained as

s = g, p,
(1)
(s s ) +
(s s usi ) = 0,
t
x j

(g g ugi ) +
(g g ugi ug j )
t
x j
=

pg g,i j gs,i j g g
+
+
+
(ugi upi ),
x j
x j
x j
r

(2)

(p p upi ) +
(p p upi pup j )
t
x j
=

p,i j ps,i j p p
+
+
(ugi upi ),
x j
x j
r

(3)

where is the volume fraction, is the material density, u is


the velocity, p is the pressure, and r is the particle relaxation
time. The filtered gas and particle viscous forces are
 u
ug j  2 ug j
gi
g,i j = gl
+
gl
i j ,
x j
xi
3
x j
 u
up j  2 up j
pi
p,i j = p
+
p
i j .
x j
xi
3 x j

Fig. 5 Eulerian timescale (ms)

The gas and particle subgrid scale (SGS) stresses are defined
as
Xu and Zhou [9] proposed the following expression for
the two-phase fluctuation velocity correlation
hvgi vp j i = hvgi vg j i

Vp j

hvgi vg j i
,
p
x j

T gi j
p

T gi j + p

p2

T gi j
p

T gi j + p

gs,i j = g Rgs,i j = g (ugi ug j ugi ug j ),


ps,i j = p Rps.i j = p (upi up j upi up j ),
where R expresses the subgrid scale stress per unit mass, and
the subscript s denotes the SGS value.
4.2 The two-phase energy equation SGS stress (SGS-k-kp)
model

where T gi j is the gas Lagrangian integral time scale seen by


the particle. The crossing-trajectory effect, continuity effect,
inertia effect, and anisotropy effect are incorporated into the
time scales seen by the particle. It is apparent that the twophase velocity correlation depends on gas turbulence, particle velocity gradient, and particle inertia. The particle velocity gradient increases the transfer of turbulent kinetic energy
between two phases, which clarifies why the radial Reynolds
stress of the particle decreases with increasing particle size;
namely, the radial Reynolds stress mainly depends on the
two-phase velocity correlation. Because the gas and the
30 m and 50 m particles have very similar velocity gradient distributions, the gas Reynolds stress, particle Reynolds
stress, and two-phase velocity correlation have similar distributions, as shown in Figs. 2 and 3.

The subgrid scale two-phase stresses can be closed by


1
gs,i j = g Kgs,i j = 2g kgs,i j = 2g (vgi vg j vgi vg j ),
2
1
ps,i j = p Kps,i j = 2p kps,i j = 2p (vpi vp j vpi vp j ),
2
Kgs = 2kgs .
The two-phase SGS energies (SGS-k and SGS-kp) are closed
by

(g Kgs,i j ) +
(g vg j Kgs,i j )
t
x j
 e Kgs,i j 
sgs
sgs
=
+ Ggk + Gpg,i j g g ,
x j k x j

(4)

Measurement and simulation of the two-phase velocity correlation in sudden-expansion gasparticle flows

(p Kps,i j ) +
(p vp j Kps,i j )
t
x j
 p Kps,i j 
sgs
=
+ Gpk + p p .
x j p x j

(s sm u0si u0s j )
t

(s sm usk u0si u0s j )


xk

= Ds,i j + Ps,i j + s,i j s,i j + Gs,gp,i j .

(5)

The terms on the right-hand side of Eqs. (4) and (5) are similar in meaning to those in the two-phase turbulent kinetic
energy equations in RANS modeling. For example
mp
sgs
sgs
[np (Rgp,si 2kg )],
Gpg,i j =
r

499

(8)

The boundary conditions for solving Eqs. (7) and (8) can be
found in Ref. [10].

5 Simulation results and their validation by measurement results

Rgp,si = vgi vpi v gi v pi .


The SGS two-phase velocity correlation Rgp,si is determined
by

Rgp,si + (vk + v pk )
Rgp,si
t
xk



(e + p )
Rgp,si
=
xk
xk
1
[g Rp,si + g Rg,si (g + p )Rgp,si ]
+
g rp

vpi
vgk  g
Rgp,sik
+ Rgp,ski
sgs Rgp,si i j .
xk
xk
kg

(6)

4.3 RANS modeling using the USM two-phase turbulence


model
The RANS modeling using a unified second-order moment
(USM) two-phase turbulence model [1, 2] was also made
by the present authors. In USM-RANS modeling, transport
equations are adopted for Reynolds stresses of gas and particles and for gasparticle velocity correlations. The equations
for two-phase Reynolds stresses are
(g gm u0gi u0g j )

The geometrical configuration of the sudden-expansion


chamber is shown in Fig. 6. The heavy particles are glass
beads with an average size of 30 m and a material density
of 2 500 kg/m3 . The length of the chamber is 1 000 mm. The
inlet gas volumetric flow rate is 212.4 m3 /h, and the particle mass loading is 0.005. For LES, the grid size is taken
as 1 mm, the grid number in a half two-dimensional computation domain is 60 000. The time step is 106 s. Figure 7
shows the LES with an SGS-k-kp model predicted two-phase
axial time-averaged velocities and compares them with the
experimental results. It is clear that the predicted values are
in good agreement with the experimental results. Figure 8
shows the predicted two-phase axial RMS fluctuation velocities using both LES-k-kp and RANS-USM and compares
them with the experimental results. The results of both models are in good agreement with the experimental results, but
the LES results are more accurate than the RANS results.

(g gm ugk u0gi u0g j )

+
t
xk
= Dg,i j + Pg,i j + g,i j g,i j + Gg,gp,i j ,

(7)

Fig. 6 Schematic of sudden-expansion chamber

Fig. 7 Two-phase time-averaged axial velocities. a For gas; b For particles

500

L.-X. Zhou, et al.

Fig. 8 Two-phase RMS axial fluctuation velocities. a For gas; b For particles

Figure 9 gives the predicted axial and radial components of the gasparticle velocity correlation using LES-k-kp
and RANS-USM and compares them with the experimental
results. The results of both models give the same tendency
as the experimental results, and the values predicted by the
LES model are greater than those predicted by the RANS

model. In most regions, the LES results are closer to the experimental results than the RANS results. Comparing Fig. 9
with Fig. 8, it is seen that the gasparticle velocity correlation distribution is similar to the gas and particle root-mean
square (RMS) fluctuation velocities but smaller than the gas
and particle RMS fluctuation velocities.

Fig. 9 Gasparticle velocity correlation. a The axial direction; b Radial

Qualitatively, these results are similar to those reported


by Prevost et al. [3], but are different from those reported by
Longmire et al. [5], who reported that the gasparticle velocity correlation is larger than the particle RMS fluctuation
velocity. This highlights the difference between the flow features of sudden-expansion flows and those of jet and channel
flows and implies that different shear rates in various flow
configurations lead to different gasparticle turbulence interaction behavior.
6 Conclusions
The gasparticle velocity correlations for sudden-expansion
gasparticle flows are studied in this paper by using both
PDPA measurements and LES with a two-phase subgridscale stress model and Reynolds-averaged NavierStokes
(RANS) modeling with a USM two-phase turbulence model.

It is found that the spatial distribution of the two-phase velocity correlation in sudden-expansion flows, like that in jet
and channel flows, is similar to that of Reynolds stresses
for gas and particles (i.e., it peaks in the shear region).
In addition, the two-phase velocity correlation in suddenexpansion flows is smaller than the Reynolds stresses of gas
and particles, as for jet flows, but it is different from that
in channel flows where the two-phase velocity correlation is
greater than the Reynolds stresses for particles. For suddenexpansion gasparticle flows, the relationship between the
two-phase velocity correlation and particle size is unclear.
However, in both jet and channel flows the two-phase velocity correlation decreases clearly with the increasing particle
size or particle Stokes number.
Measurements and theoretical analysis indicate that the
two-phase velocity correlation is not a simple function of
particle turbulent kinetic energy, and that it can not be neglected in the transport equation of the radial component of

Measurement and simulation of the two-phase velocity correlation in sudden-expansion gasparticle flows

Reynolds stresses for normal particles. Finally, the results of


the LES-k-kp and USM-RANS models agree with the measurement results and theoretical analysis, although quantitatively the LES results are more accurate than those of the
RANS modeling.

References
1 Zhou, L.X., Liao C.M., Chen, T.: A unified second-ordermoment two-phase turbulence model for simulating gas
particle flows. ASME-FED 185, 307331 (1994)
2 Zhou, L.X., Chen, T.: Simulation of swirling gasparticle flows
using USM and k--kp two-phase turbulence models. Powder
Technology 114(1-3), 111 (2001)
3 Prevost, F., Boree, J., Nuglisch, H.J., et al.: Measurements of
fluid/particle correlated motion in the far field of an axisymmetric jet. International Journal of Multiphase Flow 22(4), 685
701(1996)
4 Wang, M.R., Huang, D.Y.: Measurements of u0gl u0pl in mixinglayer flow with droplet loading. Atomization and Sprays 5,

501

305328 (1995)
5 Longmire, E.K., Khalitov, D.A.: Simultaneous gas and particle
velocity measurements in turbulent channel flows. In: Matsumoto et al. eds. Proceedings of the 5th International Conference on Multiphase Flow, Yokohama, Japan (2004)
6 Shaham, M., Morris, G.J.: Gas-solid flow in an axisymmetric
sudden expansion. ASME-FED 78, 8399(1989)
7 Hishida, K., Maeda, M.: Turbulence characteristics of particleladen flow behind a reward facing step. ASME-FED 121, 207
212 (1991)
8 Simonin, O., Fede, P., Patino, G., et al.: Mathematical models and closure laws for gasparticle turbulent flows. In: Matsumoto et al. eds. Proceedings of the 5th International Conference on Multiphase Flow, Yokohama, Japan (2004)
9 Xu, Y., Zhou, L.X., Fan, L.S.: Simulation of swirling gas
particle flows using an improved second-order moment twophase turbulence model. Powder Technology 116, 178189
(2001)
10 Zhou, L.X., Zhang X.: Simulation of sudden-expansion and
swirling gasparticle flows using a two-fluid particle-wall collision model with consideration of the wall roughness. Acta
Mechanica Sinica 20, 447454 (2004)

Das könnte Ihnen auch gefallen