Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
a r t i c l e
i n f o
a b s t r a c t
Article history:
The effects of inlet angle on the ow pattern and pressure drop in cyclones have been
numerically investigated using Large Eddy Simulations with the dynamic Smagorinsky-Lilly
subgrid-scale. Five cyclones with helical-roof inlets of different inlet angles and ve cyclones
2015
with tangential inlets of different inlet heights at the same other geometric dimensions are
considered. The results show that, increasing the inlet angle as well as the inlet height (inlet
area) decreases the absolute values of positive (close to the cyclone wall) and negative (in the
Keywords:
reduce the collection efciency. Also, increasing the inlet angle reduces the gas ow rates
central region) static pressure and tangential velocity in the cyclone body that will probably
Cyclone
along the cyclone axis in both downward (outer) and upward (inner) vortices and increases
Inlet angle
the maximum radial velocity under the vortex nder that can enhance the number of small
Flow pattern
particles entrained by the gas ow and transferred from that region into the vortex nder
Pressure drop
and negatively affect the overall collection efciency. The cyclone pressure drop is mainly
generated by the losses in the cyclone body (under the vortex nder) and in the vortex
nder. There is a signicant decrease in pressure drop with increase of inlet angle. Based on
the simulations an expression for the dimensionless pressure drop normalized by the inlet
velocity for the cyclone with helical-roof inlet of different inlet angles is derived. Cyclones
with helical-roof inlets have a higher aerodynamic efciency as compared to cyclones with
tangential inlets, and the highest aerodynamic efciency was reached with an inlet angle
of 20 .
2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
1.
Introduction
Corresponding author. Postal address: Division of Fluid and Experimental Mechanics, Department of Engineering Sciences and Mathematics, Lule University of Technology, SE-971 87 Lule, Sweden. Tel.: +46 0 920 492392; fax: +46 0 920 491074.
E-mail addresses: dzmitry.misiulia@ltu.se, dzmitry.misiulia@gmail.com (D. Misiulia), anders.g.andersson@ltu.se (A.G. Andersson),
staffan.lundstrom@ltu.se (T.S. Lundstrm).
1
Permanent address: Department of Machines and Apparatus for Chemical and Silicate Production, Belarusian State Technological
University, 13a Sverdlova str., 220006 Minsk, Belarus
http://dx.doi.org/10.1016/j.cherd.2015.06.036
0263-8762/ 2015 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
308
exemplify, Elsayed and Lacor (2011a) showed that, increasing the inlet dimensions decreases the maximum tangential
velocity in the cyclone and pressure drop and increases the
cyclone cut-off diameter. They found that the effect of changing the inlet width is more signicant than the inlet height
especially for the cut-off diameter and the optimum ratio
of inlet width b to inlet height a is somewhere between 0.5
and 0.7. Gimbun et al. (2005) showed that predicted pressure drop coefcients are proportional to the inlet area,
which has also been conrmed experimentally by Hsiao et al.
(2015). Hsiao et al. (2015) reported that reducing the inlet area
under a constant operation ow rate increased the pressure
drop and cut-off diameter while the pressure drop coefcient decreased. Their experimental results on examining the
effect of the inlet aspect ratio (a/b) did not show a clear trend
for the pressure drop coefcient, but the cut-off diameter
decreased gradually with increasing a/b. However Iozia and
Leith (1990) reported that the cyclone efciency is independent of a/b, and the pressure drop of a cyclone with large
a/b is lower than that of a cyclone with a small a/b. Lidn
and Gudmundsson (1997) argued that the cut-off size ought
to be independent of the inlet geometry. Kenny and Gussman
(2000) tested cyclones with different circular inlet model parts
and suggested that the inlet was most important with respect
to the effect of the cut-off diameter. Erdal and Shirazi (2006)
reported that, the gradually reduced inlet nozzle geometry is
the preferred geometry. Furthermore, the signicant effects
of the cyclone inlet dimensions on the cyclone performance
have been reported by Movafaghian et al. (2000) and Avci and
Karagoz (2001).
Some investigations have been done to derive the effect of
inlet angle on cyclone performance with somewhat contradictory results. Funk et al. (2001) experimentally investigated
cyclones with square and rectangular inlets at three inlet
angles (10 , 0 , 10 ). They found that square and angled
inlet modications lowered cyclone performance. Introducing dustladen air into a cyclone at an upward angle of 10
such that it strikes the top of the cyclone reduced collection
efciency by 0.25%, from 0.9931 to 0.9906, while introducing
the airow downward did not reduce efciency substantially
from the efciency realized with a horizontal inlet duct, the
difference was 0.025%. Bernardo et al. (2005) numerically
investigated a tangential cyclone with inlet angle of 45 and
compared their results to experimental measurements of the
cyclone with a conventional 0 angled inlet. They reported that
increasing the inlet angle reduced the collection efciency
from 92% to 90.5% and increased pressure drop from 579 to
620 Pa. However later, Bernardo et al. (2006) computationally
investigated an industrial-sized cyclone with a normal scroll
0 angled inlet and the three scroll inlet angles 30 , 45 and
60 . Their numerical results showed that increasing inlet section angle decreased the total pressure drop, while the overall
collection efciency increased for the cyclone with inlet section angle 30 and 45 and decreased for the cyclone with 60
inlet angle. Under the same operating conditions the predicted
overall collection efciency for the cyclone with = 45 was
77.2%, while that for the normal inlet duct was 54.4%. Qian
and Zhang (2007) numerically investigated the gas ow eld
of a tangential cyclone with three inlet section angles (0 , 30 ,
45 ) at the same inlet velocity (15 m/s) perpendicular to the
inlet plane. They found that the inlet section angle increased
the pressure drop of the cyclone whereas the pressure drop
coefcients decreased. For calculating the pressure drop coefcients, they used the velocity in the inlet direction rather
2.
Numerical set-up
2.1.
Selection of the numerical model and governing
equations
The rst numerical derivations of the ow eld in a cyclone
separator were probably performed by Boysan et al. (1982).
From that time, CFD has been a successful method for performance estimations of cyclone separators. CFD is based
on Navier-Stokes equations and turbulence is often modeled.
However, the most accurate approach is direct numerical simulation (DNS) where the whole range of spatial and temporal
scales of turbulence are resolved. The number of grid points
required for fully resolved DNS is, however, enormously large,
especially for high Reynolds number (Re) ows, and hence
DNS is restricted to relatively low Re and limited geometries.
An alternative approach is the large eddy simulation (LES)
technique that is based on a separation between large and
small turbulent scales. Scales that are of a characteristic size
greater than the grid size are calculated directly and called
large or resolved scales, and others are called small or subgridscales (SGS) and are modelled. LES is accepted as a promising
numerical tool for solving the large-scale unsteady behavior of
complex turbulent ows. The LES methodology has been used
in many studies on highly swirling ow in cyclone separators
(Derksen and Van den Akker, 2000; Slack et al., 2000; Shalaby
et al., 2005; Lipowsky and Sommerfeld, 2007; Martignoni et al.,
2007; Shalaby, 2007; Derksen et al., 2008; Shalaby et al., 2008;
Elsayed, 2011; Elsayed and Lacor, 2011b, 2013; Misiulia et al.,
2015).
(1)
u i
u
1 p
+
+ u j i =
xj
I xi
xj
t
u j
u i
+
xj
xi
sgs
ij
xi
sgs
= uiuj ui uj .
(2)
(3)
2.3.
sgs
ij
sgs
s ij
1
2
sgs
3 kk
sgs
= 2 sgs s ij ,
u j
u i
+
xj
xi
(4)
,
sgs
2.2.
309
The governing equations were solved using the commercial solver ANSYS CFX 15.0, wherein a unique hybrid
nite-element/nite-volume approach is used to discretize
NavierStokes equations. As a nite volume method, it satises strict global conservation by enforcing local conservation
over control volumes that are constructed around each
mesh vertex or node. The nite element methodology is
used to describe the solution variation (needed for various
surface uxes and source terms) within each control volume.
The choice of the discretization schemes has a tremendous
inuence on the simulation results. The unbounded central
difference advection scheme and the more accurate implicit
(5)
De D
.
4ab
(6)
310
Fig. 1 The cyclone dimensions and congurations: is the inlet angle; a is the inlet height; b is the inlet width; D is the
barrel (body) diameter; De is the vortex nder diameter; d is the cone-tip (dust outlet) diameter; H is the cyclone height; h is
the height of the cylindrical section; hc is the height of the conical section; l is the length of the inlet section; s is the vortex
nder length. (2-column tting image).
2.4.
The hexahedral computational grids were generated by dividing the whole cyclone geometry into a number of blocks and
then meshing each block separately using Ansys Meshing 15.0.
A grid independence study was performed for cyclone 2 h
(see Table 1) with four levels of grid, 0.85, 1.27, 1.92 and 2.63
million nodes, to be sure that the obtained results are grid
independent. The detailed results of this are presented in
Misiulia et al. (2015) and it can be concluded that even the
coarsest grid gave rather good predictions. However, to reduce
the uncertainty, for all ten cyclones meshes with approximately 1.31.5 million nodes have been created where the total
number of grid points was not that critical with respect to the
computational resources. This is presented in Table 2 which
shows that all important measures of mesh quality (mesh
orthogonality, aspect ratio and expansion factor) are within
correct ranges indicating that all meshes are of rather high
quality. Resolving the boundary layers with LES leads to high
resolution requirements and computing times as it should be
2.5.
Boundary conditions
At the inlet, velocity prole boundary condition is applied corresponding to the mass ow rate Qin of 0.19075 kg/s and the
mean (supercial) axial velocity in the cyclone body of 3.5 m/s
being the optimal velocity for that type of cyclones (Lazarev,
2006). The velocity proles for fully developed ow in rectangular ducts (Fig. 3) corresponding to cyclone inlet dimensions
(height and width) were obtained using the Reynolds stress
Table 1 The geometrical dimensions of the cyclone considered (normalized with the cyclone internal body diameter
D = 0.24 m).
Cyclonea
ab /D
Cyclone 1 h
Cyclone 1t
Cyclone 2hd
Cyclone 2t
Cyclone 3hd
Cyclone 3t
Cyclone 4 h
Cyclone 4t
Cyclone 5 h
Cyclone 5t
7
0
11
0
15
0
20
0
25
0
a
b
c
d
sb /D
hb /D
0.31
1.09
0.48
0.66
0.91
bb /D
0.26
De b, c /D
0.59
1.17
db /D
0.35
l/D
0.6
hc /D
H/D
Sg
1.91
4.21
5.75
1.26
2.08
4.38
3.71
1.44
2.26
4.56
2.7
1.69
2.51
4.81
1.96
1.95
2.77
5.07
1.52
Mesh characteristic
Cyclone no.
Minimum
Average
Maximum
Average
390.6
RMS
Maximum
4.292
0.27
1h
1t
2h
2t
3h
3t
4h
4t
5h
1.298
1.252
56.8
1.292
1.247
34
1.319
1.273
61.3
1.351
1.303
33.9
1.375
1.326
39.8
1.407
1.357
34
1.436
1.384
53.2
1.478
1.424
34
1.508
1.455
53.8
5t
1.533
1.475
33.9
0.381
0.991
0.806
0.068
429.7
0.904
3.065
4.373
0.276
0.36
0.99
0.775
0.069
293.4
0.968
3.4
4.445
0.28
0.565
0.991
0.705
0.068
330.8
0.721
2.817
4.572
0.288
0.359
0.99
0.775
0.069
239.3
0.807
2.744
4.607
0.291
0.504
0.991
0.732
0.069
271.4
0.607
2.704
4.78
0.302
0.359
0.988
0.775
0.069
193.1
0.687
2.671
4.83
0.305
0.582
0.99
0.743
0.071
220.3
0.499
1.915
5.069
0.32
0.358
0.989
0.775
0.069
173.4
0.576
2.118
5.065
0.339
0.555
0.981
0.762
0.084
191.9
0.448
1.961
5.37
0.32
0.358
0.989
0.775
0.069
0.495
2.083
311
312
2.6.
Selection of the time step, convergence criteria and
other settings
The time dependency is specied by the time duration and
the time steps. The selection of an appropriate time step size
is essential in order to obtain a good convergence. At each
time step in a transient simulation, several coefcient iterations or loops are carried out. A value of 3 iterations per time
Fig. 3 Velocity prole inlet boundary condition [m/s] for cyclones (a) 1, (b) 2, (c) 3, (d) 4 and (e) 5. (2-column tting image).
3.
The numerical model was validated by comparing the predicted tangential and axial velocity proles at three cut planes,
total pressure drop and cut-off diameter with experimental
data for the cyclone with = 11 and the results were published
by Misiulia et al. (2015). Considering the complexity of the turbulent swirling ow in the cyclone, the agreement between
the predicted and measured velocity proles was considered
to be acceptable. The total pressure drop and cut-off diameter obtained with LES agreed very well with the experimental
data, having only 3% maximum discrepancy. The comparison showed that the LES with dynamic Smagorinsky-Lilly SGS
model can be used to analyze the gas ow eld and performance of the cyclone separator with helical-roof inlet.
4.
4.1.
Flow pattern
The ow pattern in a cyclone can be described from the distribution of the static pressure and velocity elds within it. The
velocity of the uid ow in the cyclone can be resolved into
three components: tangential, axial and radial. Fig. 4 shows
the contours plots for the time-averaged static pressure, tangential, axial and radial velocity components at a vertical YZ
plane located at X = 0 for all 10 cyclones investigated. Also,
comparison of the tangential and axial velocity proles predicted in cyclones with different inlet angle and inlet heights
has been made at the three horizontal cut planes given in
Fig. 1. The radial proles of the predicted tangential and axial
velocities are presented in Fig. 5, and are scaled with the mean
axial velocity in the cyclone body according to
wz =
4.1.1.
4Qin
.
D2
(7)
Static pressure
4.1.2.
Tangential velocity
313
4.1.3.
Axial velocity
314
Fig. 4 The contour plots for the time averaged (a) static pressure [Pa], (b) tangential velocity (anticlockwise is positive and
clockwise is negative) [m/s], (c) axial velocity (upward is positive and downward is negative) [m/s] and (d) radial velocity
(positive in the direction of increasing radial coordinate and negative in the direction of decreasing) [m/s] at YZ plane (X = 0).
(2-column tting image).
wall, and an upward ow directed to the outlet at smaller
radii. Near the centerline, there is a region of zero axial velocity
or even ow reversal. The shape of the axial velocity proles
is also affected by the inlet angle, however this effect is not
315
Fig. 5 Comparison between the radial prole for the time-averaged tangential (absolute value) and axial velocities in the
cyclones at different sections on the YZ plane (X = 0): (a) section II, (b) section IIII, (c) section IIIIII. (2-column tting image).
r2
Q =
2rwax dr,
(8)
r1
Where, r is a radius.
The calculated time-averaged downward and upward
dimensionless ow rates (normalized by the inlet ow rate)
along the cyclone axis are presented in Fig. 7.
The inlet angle has the following effects on the ow rates
along the axis in the cyclone with helical-roof inlet:
(1) in the angular space between the cyclone body and the vortex nder the negative mass ow rate is slightly larger than
at the inlet that can be explained by involving a small part
of ow in the recirculation zones (so-called secondary
ows created by pressure gradients caused by the swirling
316
Fig. 6 Dimensionless time-averaged maximum tangential velocity in the cyclone body (a) and in the vortex nder (b).
(2-column tting image).
317
Fig. 8 Time dependent axial velocity [m/s] at the outlet for the cyclone 2 h at different simulation time (a) 0, (b) 0.25 and (c)
0.5 s. (2-column tting image).
Fig. 8 just helps to explain that for a correct calculation
of the mass ow rate at the outlet time dependent axial
velocity must be used rather than time-averaged.
Comparing cyclones with a helical-roof and tangential
inlet, the outer and inner vortices are more intense in the latter since the ow rate in the outer vortex in the direction to the
dust hopper is larger as well as at the gas outlet. That is agreed
with the predicted maximum radial velocity, which is lower
in the cyclone with tangential inlet. The effect of increasing
the inlet height in the cyclone with tangential inlet has the
similar tendency as the effect of inlet angle in a cyclone with
helical-roof inlet.
4.1.4.
Radial velocity
The radial velocity affects the particle bypass and is important for the analysis of the particle collection and losses of
efciency. Fig. 4(d) shows that in general, the radial velocities
are much smaller than the tangential velocities and are not
uniform with height. In the inner region the distribution of
the radial velocity is positive on one side and negative at the
other side, that can be explained by the vortex core precession. There is a region right under the vortex nder, where gas
directly ows into the vortex nder instead of spinning down
to the conical part and then owing upward, thereby creates
short-circuiting ow. It is often referred to as lip leakage
which deteriorates cyclone separation capability. Moreover,
the radial velocity in that region has the maximum absolute
value and comparable with the maximum tangential velocity
in the cyclone. The effect of changing inlet dimensions (inlet
angle and inlet height) on the maximum time-averaged radial
velocity in the cyclone with helical-roof and tangential inlet is
presented in Fig. 9.
Fig. 9 shows that in the cyclone with helical-roof inlet
the maximum radial velocity signicantly increases with the
changing inlet angle from 7 to 11 and to a smaller extent for
inlet angles in-between 11 and 25 . From that point of view,
increasing the inlet angle can enhance the number of small
particles entrained by the gas ow and transferred from that
region into the vortex nder and negatively affect the overall
collection efciency. In the cyclone with tangential inlet the
maximum radial velocity is smaller compared to the cyclone
with helical-roof inlet at the same geometric dimensions (inlet
aspect ratio and geometric swirl number).
4.2.
(9)
It can be seen from the Fig. 10(b) that the inlet angle
greatly affects the pressure drop in a cyclone with helicalroof inlet. Increasing the inlet angle from 7 to 11 , 15 , 20
and 25 reduced the pressure drop from 3385 to 1770, 1122,
706 and 511 Pa respectively. Also, using a helical-roof inlet
instead of tangential one signicantly reduces the pressure
drop. Increasing the inlet angle from 0 to 7 and 11 reduces
Pressure drop
Friction at the walls and in the vortex core leads to dissipation of mechanical energy which gives rise to the permanent
318
Fig. 10 Time dependent (a) and time-averaged (b) pressure drops of the cyclones. (2-column or two single column tting
images).
the cyclone pressure drop by 22.1 and 28.2% respectively, while
cyclones with helical-roof inlet with the inlet angle in-between
15 and 25 have approximately 30% less pressure drop compared to the cyclone with tangential inlet. The maximum
pressure drop reduction of 31.2% was obtained for the inlet
angle of 20 .
The pressure drop over a cyclone pc can be subdivided in
four contributions:
(1) losses in the inlet section,
(2) losses in the annular space between a cyclone body and a
vortex nder,
(3) losses in the main cyclone body (under the vortex nder),
(4) losses in the vortex nder.
The calculated dimensionless pressure losses (scaled with
the total pressure drop) are illustrated in Fig. 11.
The pressure losses in the inlet section are the smallest and
do not exceed 3% of the total pressure drop of a cyclone with
Euin =
2pc
w2z
2pc
w2in
(10)
(11)
(12)
4.3.
Aerodynamic efciency
2pc
2
(wmax
)
t
319
(13)
5.
Conclusions
320
Acknowledgments
The authors gratefully acknowledge nancial support provided by the Swedish Institute through the Visby Programme
scholarship for postdoctoral research to the rst author.
References
Andersson, A.G., Hellstrm, J.G.I., Andreasson, P., Lundstrm, T.S.,
2014. Effect of spatial resolution of a rough surface on a
numerically computed ow eld with application to hydraulic
engineering. Eng. Appl. Comput. Fluid Mech. 8 (3), 373381.
ANSYS CFX-Solver Modeling Guide, 2013. Release 15.0, ANSYS,
Inc, Canonsburg, Pennsylvania.
Avci, A., Karagoz, I., 2001. Theoretical investigation of pressure
losses in cyclone separators. Int. Commun. Heat Mass Transf.
28 (1), 107117,
http://dx.doi.org/10.1016/S0735-1933(01)00218-4.
Bernardo, S., Peres, A.P., Mori, M., 2005. Computational study of
cyclone ow uid dynamics using a different inlet section
angle. Engenharia Trmica (Therm. Eng.) 4 (1), 1823.
Bernardo, S., Mori, M., Peres, A.P., Dionsio, R.P., 2006. 3-D
computational uid dynamics for gas and gas-particle ows
in a cyclone with different inlet section angles. Powder
Technol. 162 (3), 190200,
http://dx.doi.org/10.1016/j.powtec.2005.11.007.
Boysan, F., Ayer, W.H., Swithenbank, J., 1982. A fundamental
mathematical modelling approach to cyclone design. Trans.
Inst. Chem. Eng. 60 (4), 222230.
Burstrm, P.E.C., Antos, D., Lundstrm, T.S., Marjavaara, B.D.,
2015. A CFD-based evaluation of selective non-catalytic
321