Professional Documents
Culture Documents
An Taya 2012
An Taya 2012
FEDSM2012
July 8-12, 2012, Rio Grande, Puerto Rico
FEDSM2012-72379
4.0 DISCUSSION
In this section the results of the computed test cases are
evaluated and compared with the experimental values
determined by Hashemi et al [7] and Matousek [8]. As stated in
Section 1.0, the purpose of the present work is to assess
different dispersed phase turbulence models and non-drag
interphase forces. First, the local particle velocity and
concentration profiles obtained from numerical simulations are
compared with experimental data. Specifically, the Case 1 (see
Table 2) results reported by Hashemi et al [7] are compared
with the simulations. The numerical results from various
dispersed phase turbulence models and non-drag interphase a)
forces are compared to determine which model most accurately
predicts the experimental results. The model providing most
favourable comparison with the experimental data was then
used to compute the other flow conditions (Cases 2-6 in Table
2) and the results are discussed herein. Although neither of the
experimental studies provides information on turbulent intensity
or modulation, the accurate prediction of some integral
parameters such as frictional pressure drop and heat transfer
characteristics are highly dependent on the knowledge of the
effect that the dispersed phase has on the continuous phase.
Therefore, a discussion of the effect of the presence of the
particles on the fluid turbulence intensity obtained from the
numerical study conducted as part of this investigation is given. b)
This discussion provides qualitative insight into the fluid flow Figure 2: a) Normalized sand velocity and b) concentration
structure and also highlights the type of measurements needed distributions. See Table 1 for the meanings of the acronyms
for experimental verification. used in the legend.
It can be seen that treatment of the dispersed phase as
Dispersed Phase Turbulence Models: Figures 2a and 2b
a laminar flow does not provide good predictions of the
show the normalized mean particle velocity, Us (= us/uc) and
dispersed phase velocity distribution. In fact, the velocity
concentration (Is) profiles over the pipe cross section,
profile is more parabolic in shape which is typical of laminar
respectively. These figures show the results obtained from the
flow and, consequently different from other profiles, including,
simulations using various dispersed phase boundary conditions,
most critically, the experimental results. This is an interesting
and both dispersed and continuous phase turbulence models.
observation since this approach has been employed in the
The reader is referred to the legend of Figure 2 and the
previous work by Ekambara et al [4], although they did not
information of Table 1 for specifics. In addition, the
report the particle velocity distribution. On the other hand, Fig.
experimental data of Hashemi et al [7] are shown on the figure.
2b shows that in the near-wall region, the assumption of laminar
The results for k-e-DisNS1 and k-e-DisFS1 on Fig. 2a
flow for the disperse phase results in improved prediction of the
demonstrate the effect of using no slip and free slip as wall
concentration distribution; however, no difference was observed
boundary condition for the sand, with k-ε turbulence model for
among the results in the core region. The other results of Fig. 2
water. It appears that the free-slip boundary condition is better
show that increasing the Prandtl number (see Table 1) does not
able to predict the velocity profile. The runs implementing the
necessarily have a significant effect on the predicted results
no-slip boundary condition overestimated and underestimated
The results from SST-DisNS1 simulations were not
the velocity values in the core region and near-wall region,
shown in Fig. 2 because they are almost identical to SST-Homo
respectively. However, it can be seen from Fig. 2b, which shows
case (which is shown). It is clear from the results presented in
the local concentration profile, that the no-slip result is similar
Fig. 2 that the SST model (i.e. the SST-Homo case) provided
to the experimental result, especially in the near-wall region at
much better predictions of the velocity and concentration
the top of the pipe. It should be said that the concentration
distributions than the k-H model, particularly in the near-wall
a)
b)
Figure 5: a) Normalized sand velocity and b) concentration
distributions for multifluid with data from Matousek [8] (Cases
5-6 in Table 2)
Turbulence Kinetic Energy: In this section the turbulent
kinetic energy of slurry flows are compared to that expected for
single phase flow under identical conditions. The comparison is
shown as so-called turbulent modulation, TM ( k s / kl -1),
which is expressed as a per cent of the value expected for
single-phase flows. In the regions y/D > 0.5 and y/D < 0.12, the
presence of the dispersed phase suppresses the turbulent kinetic
energy irrespective of the bulk velocity, concentration, and pipe
Figure 7: Turbulence Modulation
diameter, as shown in Fig. 7. Outside these regions, there is
turbulence enhancement for C = 0.2 for both uc = 2 and 3 m/s,
CONCLUSIONS
and C = 0.43 for uc = 6 m/s. For the other flow conditions, there
For this study, several test cases exploring the effects
is suppression in the entire pipe. For the dense slurry flows that
of turbulence models, wall boundary conditions, interstitial
are considered here, particle-particle collisions are known to be
forces and bimodal mixture were performed to determine their
a source of particle kinetic energy. It should be noted that at
effect on numerical predictions for sand velocity and
higher bulk concentrations, the flows are relatively
concentration distributions, and frictional pressure drop. The
homogeneous as a result of the competing forces, including
predicted results were compared with the experimental
particle-particle collisions, turbulent dispersion, drag and
measurements of Hashemi et al [7] and Matousek [8].
viscous forces (as mentioned earlier).
MATHEMATICAL MODEL
¦U ¦D
1
(D i U i u i ) 0 and i 1 (4)
i i i
&
where α is the concentration or volume fraction, index i is for either solid (s) or liquid (l), u is the velocity vector for respective phase,
ρ is the density, g is acceleration of gravity, p is the thermodynamic pressure which is the sum of dynamic, static and turbulent
fluctuation components, M is the sum of interphase forces, P is effective viscosity which is the sum of the dynamic viscosity and eddy
&
or turbulent viscosity, and W s is the solid stress tensor. It is worth commenting that, in two-fluid approach, it is assumed that both
fluids/phases share the same pressure field. In the subsequent sections, model equations for eddy viscosity, solid stress tensor, and
interphase forces are discussed.
A.2 Closure Models
A.2.1 Turbulence Models
For the present study, the continuous phase is assumed to be turbulent and therefore, effective viscosity is computed using
two-equation turbulence models, the standard linear-eddy-viscosity k-H model and shear stress transport which hereafter referred to as
k-H and SST, respectively. The general expression for effective viscosity, P is given as:
Pl Plm Plt (5)
where m and t denote, respectively, dynamic and eddy components of the viscosity. The k-H assumes that, the turbulence viscosity is
linked to the turbulence kinetic energy, k and dissipation, H via the relation:
k2
Plt CP U (6)
H l
where CP is a constant of value 0.09 and all other variables have their usual meanings. The values of k and H are computed directly
from differential transport equations for the turbulence kinetic energy and turbulence dissipation rate which are given in Eqns. (7) and
(8).
&
§ wk wul k · w §¨ §¨ P t · wk ·¸
D l U l ¨¨ ¸ Dl
¸ P l ¸ D l Pk U l H (7)
© wt wxi ¹ wxi ¨© ¨© V k ¸¹ wxi ¸¹
&
§ wH w ul H · w §¨ §¨ P lt ·¸ wH ·¸ H
¨
D l Ul ¨ ¸ Dl
¸ P D l cH 1 Pk cH 2 U l H (8)
¨ ¨ ¸
V H ¹ wxi ¹ ¸
© wt wxi ¹ wxi © © k
where cH1, cH2, Vk and VH are constants with values 1.44, 1.92, 1.0 and 1.3, respectively, Pk is the production term which is responsible
for the generation of the turbulent kinetic energy and is a function of the mean velocity gradient.