ARTICLE IN PRESS

Progress in Energy and Combustion Science 33 (2007) 409–452

www.elsevier.com/locate/pecs

Modeling the gas and particle flow inside cyclone separators

Cristóbal Cortés, Antonia Gil
Center of Research of Energy Resources and Consumptions (CIRCE), Department of Mechanical Engineering, University of Zaragoza,
Maria de Luna 3, 50018 Zaragoza, Spain
Received 17 January 2006; accepted 2 February 2007
Available online 19 April 2007

Abstract

This paper reviews the models developed for the flow field inside inverse-flow cyclone separators. In a first part,
traditional algebraic models and their foundations are summarized in a unified manner, including the formulae for
tangential velocity and pressure drop. The immediate application to the prediction of collection efficiency is also reviewed.
The approach is the classical, treating first the dilute limit (clean-gas correlations), and afterwards correcting for ‘‘mass
loading’’ effects. Although all these methods have had a remarkable success, more advanced ideas are needed to model
cyclones. This is put forward by exploring the work done on the so-called ‘‘natural’’ length of the cyclone, that has led to
the discovery of instability and secondary flows. The resort to computational fluid dynamics (CFD) in this case is difficult,
however, due to the very nature of the flow structure. A closing section on the subject reviews past and recent CFD
simulations of cyclones, both single- and two-phase, steady and unsteady, aiming at delineating the state-of-the-art,
present limitations and perspectives of this field of research.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: CFD; Cyclone; Gas-solid flow; Swirling flow

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
2. Basics of cyclone separators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
3. Flow field and pressure drop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
3.1. Velocity distribution inside cyclones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
3.2. Models of the velocity distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417

Abbreviations: CFBC, circulating fluidized bed combustion; CFD, computational fluid dynamics; DNS, direct numerical simulation;
FCC, fluid catalytic cracking; LDA, laser-Doppler anemometry; LES, large eddy simulation; LRR, Launder, Reece and Rodi
implementation of a differential RSTM, also known as the ‘‘basic’’ implementation. Variant: LRRG, after the modification by Gibson
and, Launder [91]; PFBC, pressurized fluidized bed combustion; PSD, particle size distribution; PSI-Cell, particle-source-in cell methods
of calculation of two-way coupled two-phase flow; PVC, precessing vortex core; RANS, Reynolds averaged Navier–Stokes equations;
RNG, Re-normalization group theory; RSTM, Reynolds stresses transport model; SGS, subgrid scale model; SSG, Speziale, Sarkar and
Gatski implementation of a differential RSTM [91]; TRANS, transient Reynolds averaged Navier–Stokes equations; equivalent to
URANS; URANS, unsteady Reynolds averaged Navier–Stokes equations; equivalent to TRANS.
Corresponding author. Tel.: +34 976 762034; fax: +34 976 732078.
E-mail address: [email protected] (C. Cortés).

0360-1285/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pecs.2007.02.001
 ARTICLE IN PRESS
410 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

3.3. Pressure field in cyclones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420

3.4. More on cyclone velocity patterns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
3.5. Pressure drop in cyclones. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
3.6. Calculating the pressure drop. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
4. Collection efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
4.1. Models of collection efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
4.2. Mass loading effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
5. Special phenomena associated with the flow field in cyclones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
5.1. Natural turning length. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
5.1.1. Relevant parameters influencing natural turning length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
5.1.2. Calculation of Ln . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
5.1.3. Influence of cyclone natural length on collection efficiency and pressure drop . . . . . . . . . . . . . 434
5.2. Precessing vortex core (PVC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
5.2.1. Influence of the PVC in the flow field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
5.2.2. Influence of operating conditions on PVC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
6. Computational fluid dynamics applied to cyclones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
6.1. CFD studies on single-phase cyclone flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
6.1.1. Non-intrusive velocity measurements in cyclone separators. . . . . . . . . . . . . . . . . . . . . . . . . . . 441
6.2. Numerical computation of two-phase flow in cyclones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
6.2.1. Eulerian–Lagrangian models of cyclone operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
6.2.2. Eulerian–Eulerian models of cyclone operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
7. Concluding remarks and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448

1. Introduction a fundamental step to understand pressure drop and

separation efficiency. At the same time, and based
Cyclone separators have been a decisive factor in on the data gathered, theoretical models have been
the development of coal combustion technologies. advanced to predict the basic features of the flow
Among diverse possibilities for hot gas cleaning, field, mostly on semi-empirical grounds. This
these devices have demonstrated the most favorable generic modeling strategy is still in use; in fact,
balance of separation efficiency and cost of invest- many formulae and methods derived in the past
ment, operation and maintenance. Able to handle remain very useful indeed for design purposes
any combination of gas pressure, temperature and nowadays. Nevertheless, as in many other fields of
very high solids loading, their performance is study, advances in experimental and computational
tolerable as compared with more efficient separa- methods have brought to light many additional
tion equipment (i.e., ceramic filters), being at once details and subtleties of the question.
much more simple, robust and reliable. Presently, In the case of cyclone separators, some of them
cyclones are a key component in most advanced have turned out to be of a rather fundamental
coal utilization concepts, such as pressurized and nature, and at the same time, of paramount
circulating fluidized bed combustion (PFBC and importance from the point of view of applications.
CFBC). In PFBC, cyclones are essential to maintain Unsteadiness and asymmetry are for example two
the integrity of the gas turbine, and thus the features not considered in classical cyclone theory
advantages of the concept itself [1,2]. In CFBC, that may affect the velocity distribution to a great
the scaling-up of the equipment to sizes compared extent, thus changing the model of the separation
to conventional coal firing is being developed partly mechanism. In close relationship, our picture of the
based on new designs of integrated, compact end of the separation vortex has been evolving until
cyclonic separators [3]. very recently. Latest research is revealing that in the
As a consequence, there is still a great need of a vicinity of this region, the flow can be unstable and
sound knowledge of the principles of operation. the vortex can attach to the solid wall, rotating on
Since cyclones were developed in the last decades of it. Consequently, the prediction of the total length
19th century, extensive experimental work has been of the vortex (the so-called ‘‘natural length’’ of the
done in order to explain their flow characteristics, as cyclone) is subjected to great uncertainty, as the
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 411

Nomenclature vi inlet gas velocity

vr radial gas velocity
a inlet section height vt tangential gas velocity
Ai inlet area vt max maximum tangential velocity [21]
AS inner cyclone friction surface area vte tangential gas velocity at the inner vortex
b inlet section width [13]
B discharge duct diameter vtw velocity in the vicinity of the wall
Ccr critical load vtw wall velocity just after inlet [15]
Csi inlet solids loading vz axial gas velocity
dp generic particle size vz0 mean axial velocity, vz0 ¼ V_ =pðr2c  r2t Þ
D diameter vze axial gas velocity at vortex finder
Dc cyclone diameter x particle size in cyclones
De vortex finder diameter x50 cut size of the cyclone
DLn diameter of the cyclone cone at the xm particle mass median diameter
vortex end position
f(x) particle size distribution Greek symbols
fg parameter of Alexander pressure drop
model, Eq. (39) w coefficient from Meissner and Löffler [15]
Frc inlet Froude number, Frc ¼ v2i =gDc L parameter [21]
Fre vortex finder Froude number, a coefficient in Barth theory [13]
Frc ¼ v2ze =gDe ap particle volume fraction
g acceleration of gravity estr porosity of the strands, Eq. (49)
H total cyclone height Z(x) fractional collection efficiency
h height of cylindrical section of cyclone Z total collection efficiency
h height of the separation surface in the l friction factor
models of Barth [13] mg gas dynamic viscosity
K empirical constant [21] rb bulk density
L generic interparticle spacing in two-phase rg gas density
flow rs solids density
Ln cyclone vortex length xc cyclone pressure coefficient,
m slope of the fractional efficiency curve, xc ¼ Dpc =12 rg v2i
Eq. (52) Fc angle of the cyclone cone
p pressure G constant, Eq. (30)
r radius O angular velocity, Eq. (29)
ra radius of maximum tangential velocity,
in particular as defined in [21] Subscripts
re radius of the vortex finder
rc cyclone radius b body of the cyclone
Rec cyclone Reynolds number, c cyclone
Rec ¼ Dc rg vi =mg e vortex finder
rt transition radius [21] g gas
S height of the vortex finder i inlet
St cyclone Strouhal number, Eq. (65) r radial component
Sw swirl number, Eqs. (63) and (64) s solids
T temperature t tangential component
V_ volumetric gas flow rate w wall
vb mean axial velocity at cyclone body, z axial component
vb ¼ 4V_ =pD2c
 ARTICLE IN PRESS
412 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

great discrepancy between different calculation state and transient. Predicted flow field, cyclone
methods found in the literature already attested. natural length, pressure drop and collection effi-
This parameter is essential for an optimal design, ciency are discussed and compared with experi-
specially in those units equipped with a dipleg,1 and mental data and between different models and
thus constitutes a current challenge for new cyclone calculations. The paper discusses all the peculiar
models. features mentioned above, analyzes the relative
The modern study of cyclone separators has performance of the models developed throughout
taken advantage of several sophisticated experi- the years, and attempts to outline general recom-
mental and numerical techniques. Amongst the mendations and future perspectives.
former, particle image velocimetry (PIV) and laser
Doppler anemometry (LDA) are prominent, even 2. Basics of cyclone separators
though the need of seeding the gas with small
particles, in turn too prone to separate, poses some The basic principle of cyclone separators is the
intrinsic difficulties. These are absent in intrusive forcing of the particle-laden gas into a vortex, where
methods (whose chief example is the miniature X- inertia and gravitational forces effect particle
hot-wire anemometer), that in exchange may suffer separation. Among existing cyclones and diverse
from inaccuracy when dealing with unsteady flow, cyclonic equipment, there is a basic model that at
and from probe deterioration and damage due to once has been used by industry exhaustively: the
the aggressive environment in two-phase flow. On inverse flow cyclone, Fig. 1.
the other hand, as in many other fields, computa- In this device, the fluid enters tangentially into the
tional fluid dynamics (CFD) currently emerges as an cylindrical chamber with a high rotational compo-
efficient alternative to traditional, approximate nent. The flow descends rotating near the wall, until
analytical models of the time-averaged flow. a certain axial location where the axial velocity
A validated numerical technique is of course component reverses itself, thus making the flow to
extremely powerful for analyzing geometry and ascend. This is referred to as the vortex end
operating conditions and pursuing an optimal position. The ascension proceeds near the cyclone
design. However, the flow inside cyclone separators axis and, since the flow rotation continues, a double
again entails special difficulties in this respect, so vortex structure is formed, as indicated in the figure.
that the use of CFD is not as widespread as it can be The inner vortex finally leads the flow to exit
expected. The high anisotropy of the turbulent field through a central duct, called the vortex finder. The
in such a confined, strongly swirling flow demands vortex finder protrudes within the cyclone body,
the adoption of modeling expedients that are quite which serves both to shield the inner vortex from the
expensive in terms of computing resources, such as high inlet velocity and to stabilize it. It is also worth
Reynolds stresses transport models (RSTM) and to mention that the inversion leading to this peculiar
higher order discretization. Furthermore, the very flow structure is apparently originated by the
unsteady nature of the flow implies that any valid pressure field inside the cyclone, and not directly
CFD scheme should be transient as well, which influenced by the conical shape or the geometrical
strictly leads to extremely costly techniques of length.
turbulence modeling, i.e., large eddy or direct The textbook explanation of the separation is
numerical simulation (LES or DNS). Simpler that the solid particles, denser than the gas, are
schemes, such as the traditional steady-state models, subjected to a high centrifugal force, which directs
or even time-dependent versions of them, are not them to the walls, where they collide, lose momen-
sound from a fundamental standpoint, and should tum and become disengaged from the flow. The
be scrutinized in depth. solids thus separated descend sliding on the conical
In this paper, we undertake a review of the most wall and are collected or extracted at the lower part.
relevant semi-empirical models proposed for the This mechanism obviously suffers from a number of
time-averaged flow in cyclones, as found in the imperfections: small particles that follow the gas,
literature. These are in turn connected with the particles that rebound and are re-entrained, and
results of the latest CFD simulations, both steady direct re-entrainment or by-pass at the lowest
conical section, near the inversion zone, at the
1
A prolonged tube used to accommodate the flow of solids in vortex finder lip, and all along the inner/outer
some applications, notably FCC and PFBC. vortex boundary. All these factors add up to the
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 413

forming dust strands that descend helicoidally along

Vortex finder
the walls. The centrifugal force only acts upon the
remaining dust, usually a small fraction of the total
[6,7]. As we will see later, this effect is explained
essentially by inertia and interaction between
particles.
There are several systems to insert the flow into
the cyclone with a high tangential velocity compo-
Downward Upward nent; some inlet designs are shown in Fig. 2. Most
outer region inner region frequent are the tangential and the scroll configura-
tions. The scroll inlet is usually designed to wrap
around up to 1801 of the cylindrical cross-section.
Higher arcs are not used, because they are
ineffective and unnecessarily increase pressure drop
[9], although shorter ones are not unusual. Depend-
ing on their use and particular properties of the
dusts, many different cyclone designs have been
developed throughout the years. As an example,
De Table 1 and Fig. 3 show sixteen different designs of
the same inlet area (0.01 m2) performing the same
b
duty at a given inlet velocity, as compiled in the
monograph by Hoffmann and Stein [10].
(3) (5) From an engineering point of view, cyclone
a
S
performance is measured by collection efficiency
h (the fraction of solids separated) and pressure drop.
(1)
These two parameters are the direct outcome of the
Dc flow developed inside the device, in turn described
by the velocity, solids concentration and pressure
H
fields. Given the wide range of solids loadings that
(2) cyclones are apt to handle, the flow is generically
biphasic; interaction between particles and two-way
coupling can only be neglected for low concentra-
tions of solids. In spite of this, traditional cyclone
models proceeded from clean-gas velocity measure-
ments, through explanation and correlation of the
(4) observed profiles, to arrive at relatively simple
formulations of measured efficiency and pressure
B drop. We will follow roughly this historical path,
signaling here and there the effects of a high solids
Fig. 1. (a) Qualitative drawing of the principle of operation and
concentration and how these are taken into account
flow patterns in cyclones. (b) Main parts and dimensions of an
inverse-flow cyclone: (1) cyclone body, (2) conical part, (3) inlet in traditional cyclone modeling. A final section on
duct, (4) exit duct, (5) vortex finder [4,5]. CFD calculations will give a more integrated view,
dealing with work done and perspectives on clean-
gas and two-phase flow simulations—mostly the
variables that affect the centrifugal force (mainly latter in the latter case.
geometry and inlet velocity), to make the collection
efficiency of cyclones highly variable. 3. Flow field and pressure drop
Actually, the centrifugal force (or any other
parameter of the gas dynamics) can only explain 3.1. Velocity distribution inside cyclones
the separation of solids in dilute flows. It is well
known that above a certain solids loading, particles The first studies of the features of cyclone inner
are collected as soon as they enter the device, flow were undertaken in 1930–1950 [11,12],
 ARTICLE IN PRESS
414 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Fig. 2. Main inlet arrangements: (a) tangential, (b) scroll, (c) helicoidal, (d) axial [8].

Table 1
Relevant geometric dimensions of several standard cyclone designs [10]

Name Dc De S H h A b B

Muschelknautz E 680 170 311 934 173 173 58 228

Muschelknautz D 357 119 318 863 262 187 54 195
Storch 4 260 117 176 1616 909 260 38 91
Storch 3 192 107 200 821 462 167 60 92
Storch 2 225 108 239 1097 464 188 53 84
Storch 1 365 123 142 1943 548 100 100 64
Tengbergen C 337 112 145 930 187 100 100 112
Tengbergen B 210 112 224 604 324 179 56 112
Tengbergen A 277 112 157 647 180 135 74 202
TSN-11 348 136 242 959 219 184 54 154
TSN-15 266 158 350 1124 589 166 60 119
Stairmand high efficiency 316 158 158 1265 474 158 63 119
Stairmand high flow 190 141 165 755 283 141 71 71
VanTongeren AC 325 100 325 1231 436 149 67 130
Vibco 286 111 124 720 228 111 90 66
Lapple GP 283 141 177 1131 566 141 71 71

promptly revealing their extraordinary complexity where C is a constant, r the radius and the
that initiated an enormous wealth of experimental exponent n depends on r. This variation is
and theoretical work on the subject. Figs. 4 and 5 sketched in Fig. 6. In the inner region near the
show classical measurements and explanations, cyclone axis, n is close to 1 (forced vortex),
attempting to describe the three velocity compo- whereas n approaches 1 (free vortex) near the
nents inside a cyclone separator. wall. The forced vortex only encompasses a region
The time-averaged flow is made up mainly of a fairly close to the centerline, so that the point
vortex, thus dominated by tangential velocity and at which the velocity attains a maximum is well
strong shear in the radial direction. As a first and inside the radius covered by the vortex finder, as
simple approximation, the profile can be descri- seen in Fig. 5. Actually, the inner rotation is
bed as a Rankine vortex, a combined free and somewhat different from that of a solid, and
forced vortex. The tangential velocity distribution vt shear in the outer region is too high for the influence
in the radial direction is assumed to obey a law of of viscosity to vanish. In practice, a ‘‘forced
the form vortex’’ exponent close to 1 is observed for the
inner part of the profile, but the outer part
vt rn ¼ C, (1) (excluding the sudden decrease very close to the
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 415

Fig. 3. Scale drawings of the cyclone designs of Table 1 [10].

Fig. 4. Measured profile of velocity components in a reverse-flow cyclone: (a) tangential, (b) radial, (c) axial [12].

wall) is better correlated by a exponent n in the space. As we will see, many algebraic models have
range 0.4–0.8. relied on the adequate correlation of n, but this
A further common assumption is no axial approach cannot take into account two important
variation of vt, acknowledging the fact that it is parameters, namely, the wall friction and, for dense
indeed fairly low, at least within the main separation flows, the concentration of solids. Both directly
 ARTICLE IN PRESS
416 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Fig. 5. Sketch of cyclone velocity profiles: 1—radial, 2—axial,

3—tangential [14]. Fig. 6. Sketch of the tangential velocity profile in cyclones [4].

influence the strength of the vortex, and thus the maximum roughly at the radial position of the
exponent n changes with them, in a manner that vortex finder; sometimes the drop in momentum is
cannot be ascertained but by blind empiricism. so severe as to cause backflow [10]. This curious
Clearly, more elaborated models of the velocity field behavior is frequently observed in experimental
are needed. measurements and CFD simulations; we will return
Also of importance are the other two velocity later to its explanation and effects. An implication is
components. The axial velocity is responsible, more that the locus of flow inversion cannot be simply a
than gravity, for the transport of particles to the cylinder, for obvious reasons of continuity.
collection device [12]. A simple quantitative model is The radial velocity is important in some models
suggested by the double vortex structure: radially- of particle collection; evidently, it always will be a
constant values for the outer vortex/downward flow factor when analyzing by-pass and losses of
and the inner vortex/upward flow zones. Both efficiency. Frequently, it is assumed of much lesser
values should be zero at the axial position of the magnitude than the other components, but this is
vortex end. At the vortex finder, the inner value is only true concerning the outer vortex. The radial
given by the volumetric flow rate and the cross- velocity grows steeply towards the vortex core,
sectional area; its axial variation can be then aimed inwards, specially in the vicinity of the vortex
adjusted by a simple (linear) function, and the outer finder [9,14]. An average, perhaps characteristic,
value deduced from continuity. This requires how- value can be derived [6,13,15] by assuming that the
ever some rather drastic assumptions about the gas flows evenly through the imaginary cylinder CS
entire flow structure, Fig. 7: a vortex end position seen in Fig. 7, which needs the approximations
dictated by geometry or pre-established in some previously mentioned.
other, simple form, and a hypothetical cylinder Finally, in cyclones equipped with diplegs, several
having the vortex finder diameter as the locus of flow regimes may develop within them, depending
flow inversion. on the procedures for extracting the solids. In PFBC
Obviously, actual profiles are not flat, but exhibit applications, a deep penetration of the swirl inside
maxima. Descending flow always has a maximum the dipleg has been experimentally measured
close to the wall, as can be expected, but axial [5,16,17], caused by the (small) fraction of gas used
velocity at the inner vortex is either reported as an for particle transport. This is mainly absent in FCC
inverted V or W-shaped profile, i.e., with a cyclones, where the gravity-assisted, intermittent
maximum or a dip at the symmetry axis, as can be flow of solids gives way to zones of dense-phase
observed in Figs. 4 and 5. The W-pattern exhibits a transport [18].
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 417

friction, having a significant effect on the flow field,

is not easily handled in this manner.
Barth [13], Eqs. (5)–(8): This is a simple and still
useful model, by which friction was first introduced
in cyclone modeling. As we shall see later, the
velocity profiles of Barth were immediately applied
as a first construct to predict collection efficiency.
Similarly to Alexander’s, this model considers as
a geometric constant the ratio a between average
angular momentum of the gas at the inlet and that
of the gas rotating inside the cylindrical body of the
cyclone, close to the wall. This constant is obtained
for several entrance geometries, being unity for
scroll, 3601 inlets and less than unity for tangential
inlets, where a considerable acceleration obtains.
The tangential velocity at the wall vtw is then
related to the tangential velocity at the control
surface CS of Fig. 7, which approximately repre-
Fig. 7. Imaginary cylinder of Barth theories [13]. sents the swirl intensity of the inner vortex. The
method is an angular momentum balance that
assumes
p ffiffiffiffiffiffiffiffiffiffiffi another imaginary surface of diameter
3.2. Models of the velocity distribution Dc De and height h* where all frictional losses are
concentrated; the analysis leads to Eq. (6). Losses
Table 2 summarizes the most relevant, alge- are represented by a lumped wall friction coefficient
braic models of cyclone flow, as taken from the lg, empirically adjusted. The height h can be
literature. A scheme of the key hypotheses and main naturally identified as the length of the vortex;
formulae is given; the complete details can be Barth made a purely geometrical interpretation of
looked up at the original references. As we will this parameter, via Eqs. (8).
see, cyclone models started from crude considera- Muschelknautz [6]: Muschelknautz and co-work-
tions on vortex flow, and evolved to incorporate ers have worked upon Barth’s ideas to develop
more classes of phenomena and more sophisticated empirical models that combine in admirable mea-
principia. Most models deal only with the tangential sures both simplicity and realism. Concerning the
component; axial and radial velocities are usually velocity distribution, the table summarizes one of
handled through the simplified expedients men- their earliest models. The essential concepts of the
tioned above. coefficient a and the friction surface are maintained
Alexander [11], Eqs. (2)–(4): This is a purely and perfected. Data on a is correlated by means of
empirical model that addresses two separate ques- analytical formulae, Eqs. (9) and (10), to replace the
tions. Firstly, it correlates the ratio of the tangential original graphs. In addition, the value of the gas
velocity in the vicinity of the wall vtw to the (given) friction coefficient lg is readjusted.
mean inlet velocity vi, considering it as a purely But beyond that, the model is modified to give a
geometrical parameter, which is reasonable for the quantitative prediction of the effect of the concen-
high Reynolds numbers usually found in practice. tration of solids in the regime of dense flow. By
Secondly, a correlation is given for the exponent n means of streaks of particles directly separated at
that characterizes the radial profile of the tangential the entrance, this is to augment wall friction and
velocity in the outer vortex. This is made to depend thus weaken the vortex intensity. Muschelknautz’s
on the cyclone diameter and also on absolute original expressions for the increase of lg with the
temperature, since the experimental census com- inlet solids loading Csi are given in the table as Eqs.
prised cyclones treating hot gases, whose tangential (11). As we will see later, this discovery opened the
velocity additionally changes due to the change of way to modeling mass loading effects, both in
viscosity. Compared to more modern models and pressure drop and separation efficiency.
measurements, the value of n is normally under- Meissner and Loffler [14,15]: Similar to Barth,
predicted; on the other hand, variation of the wall they derived an empirical expression for the
 ARTICLE IN PRESS
418 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Table 2
Algebraic models of the velocity field in cyclones

Model Equation Remarks

r n
Alexander [11] c Experiments with scroll and
vt ¼ vtw (2)
r tangential inlets
Ai n
 
vtw Air and combustion gases, up to
¼ 2:15 (3)
vi Dc De 1100 1C

T 0:3
 
T in K and Dc in m
n ¼ 1  1  0:67D0:14
 
c (4)
283

Barth [13] 2vi r̄i Scroll inlets aE1

a¼ (5)
vtw Dc

v r =r
vte ¼  tw c e (6)
1 þ h rc plvtw =V_


r̄i ¼ ðDc þ bÞ=2 (7) For tangential inlets, values of a

are given by graphs
h ¼ H  S; De pB (8a)

ðH  hÞðDc  De Þ l ¼ lgE0.02
h ¼ þ h  S; De 4B (8b)
Dc  B

Muschelknautz [6] a ¼ 1  1:2b=Dc for tangential inlets (9) Eq. (9) is valid for
pffiffiffi pffiffiffiffiffi 0:9oab=pr̄2i o1:8
a ¼ 1 þ 3plr̄i = ab for scroll inlets (10)a
 pffiffiffiffiffiffiffi
l ¼ lg 1 þ 2 C si ; C si o1 (11a) Csi in kg solids/kg gas

 pffiffiffiffiffiffiffi
l ¼ lg 1 þ 3 C si ; C si 41 (11b) lgE0.005 for high Reynolds
numbers
1
Meissner and Loffler vtw 0:204b
¼ þ 0:889 (12)
[15] vi rc
" 0:5 #
vtw 1 lg hz vtw
¼ 0:25 þ  0:5 (13)
vb lg hz vb

_ 2
vb ¼ 4V=pDc (14)
 
a 2p  a cos b=rc  1 h w accounts for the angular
hz ¼ 1 þ (15)
rc 2p rc momentum exchange between
wall and gas
vtw lg, lx, lkE0.0065–0.0075 for
vt ¼   (16)
r=rc 1 þ w 1  r=rc smooth walls at room
temperatures
vtw lk
w¼ lx þ (17)
vb sinðFc Þ

V_
Radial velocity : vr ðrc Þ ¼ 0; vr ðre Þ ¼ (18a,b)
2pre ðH  SÞ

V_ ðH  zÞ
Axial velocity : vz0 ¼  2  (19)
p rc  r2e ðH  SÞ
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 419

Table 2 (continued )

Model Equation Remarks

0:72
Reydon and Gauvin Freevortex : vt ¼ 14:79vi r (20) r in cm
[19]
Forced vortex : vt ¼ 1:35vi r exp 0:153vi0:17 r (21)

Ogawa [20,21] Free vortex : vt rn ¼ K 0 vi ðrc  bÞn (22) K0, n ¼ f (Re, geometry)

Forced vortex : vt ¼ Krð1  LrÞ (23)

2vt max
K¼ (24a)
ra

1þn
Lrt ¼ (24b)
2þn

ra L ¼ 0:5 (24c)

K ¼ 66:5De2:65 V_ (24d)

V_
Mean axial velocity : vz0 ¼  2  (25)
p rc  r2t

Radial and
 axial velocities
 in the vortex
vr rc r rc vz z
¼  ; ¼1  (26a,b)
vz0 2h rc r vz0 h

Radial and axial velocities in the forced-vortex

vr r2  r2 vz r2  r2  z 
¼  c 2 t r; ¼ c 2 t  1 (27a,b)
vz0 2rt h vz0 rt h

a
This equation does not always give consistent values [10].

geometric relationship between the tangential velo- radial, Eq. (18), and axial velocity at the outer
city at the cyclone wall vtw and the inlet velocity vi, vortex, Eq. (19). Note in the formulae that they
Eq. (12), and a momentum balance to take wall refer to geometries more restrictive than those
friction into account, Eq. (13). Only that two values considered by Barth.
of the tangential velocity are considered: vtw just at The second angular momentum balance implies
the inlet slot where acceleration occurs and the the use of two additional friction factors, different
‘‘developed’’ value vtw at the cyclone body. As from lg: those corresponding to the upper and lower
indicated by the formulae, the momentum balance metal surfaces of the cyclone that bound the control
is different from Barth’s; it only refers to the volume. Although in latter studies the same value
reduction from vtw to vtw as a result of flow along was given to the three coefficients, it is by no means
the cyclone wall. Only slot or tangential inlets are clear that their physical significance be equivalent.
handled by this model; scrolls are explicitly ex- For this reason, the model is only strictly valid for
cluded. dilute flows; the inclusion of mass loading effects is
To calculate the tangential velocity at CS, an more difficult than with Barth or Muschelknautz.
angular momentum balance is applied to a hollow Reydon and Gauvin [19]: Both theoretical and
cylinder of differential thickness between r and experimental flow studies were carried out at
r+dr, which leads to Eq. (16) for the tangential different operating conditions and for different
velocity at any radius vt(r). The velocity at CS is geometric parameters, in an effort to obtain more
simply found by substituting r ¼ re. Ideas similar to general expressions for vt. The results are divided in
those discussed at the beginning, are adopted for the two regions corresponding to the outer and inner
 ARTICLE IN PRESS
420 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

vortex, Eqs. (20) and (21), and the coefficients were

adjusted with experimental data by linear regres-
sion.
Ogawa [20,21]: This author developed the most
complex algebraic construct to date, based on
theoretical considerations. The wall tangential
velocity vtw, that intermediate variable used in
previous work, is absent here. A classical outer
‘‘free vortex’’ region is deduced, Eq. (22), with
constants K0 and n that depend on Reynolds
number and geometry. A law for the ‘‘forced
vortex’’ region is also deduced, Eq. (23), introducing
a maximum tangential velocity vt max at a radial
position ra from which solid body rotation prevails
in the inner vortex. These results are based on
considerations on axial swirl stability, as can be
consulted in [22]. Axial and radial velocities at the
two flow regions are also derived, Eqs. (25)–(27).
Finally, the strategy proposed in [23] can serve to
integrate recent models [15,19,20] in a supposedly
coherent manner:

1. Use Ogawa’s Eq. (23) to determine the radial

evolution of vt in the inner vortex.
2. The radius of maximum tangential velocity ra,
needed to obtain the constant L, is obtained by
intersection of Eqs. (16) and (21), i.e., the ‘‘forced
vortex’’ of Reydon and Gauvin and the ‘‘free
vortex’’ of Meissner and Löffler. Fig. 8. Comparison of tangential velocity profiles predicted with
3. The ‘‘free vortex’’ region is described by Eq. (16), different flow models. Adapted from [23].
and the transition radius rt between this region
and the ‘‘forced vortex’’ can be obtained as the tional, swirl-free flows. These have caused more
intersection with the Ogawa curve, Eq. (23). than one confusion in cyclone literature.
Let us begin with the equilibrium between
centrifugal force and radial pressure gradient
Fig. 8 taken from [23], shows the results and a
comparison of different models and measurements. v2t dp
r ¼ . (28)
Interestingly, the framework of the most recent and r dr
complex models is apparently inviscid, although It is worth to remember that this equation is exact
friction is of course introduced by the adoption of for idealized, axially symmetric one-dimensional
correlating exponents and functions for the inner flows, being directly derived from the momentum
and outer vortexes. However, this hardly can take equation in the radial direction. In other words, it is
into account the effect of a high concentration of equally valid for ideal or viscous, for laminar or
solids [24]; in this sense, older approaches, such as turbulent, steady flows.
Muschelknautz’s, are perhaps more practical. If we assume that the inner zone is a pure, forced
vortex vt ðrÞ ¼ Or, being O ¼ const. the angular
3.3. Pressure field in cyclones velocity, Eq. (28) is easily integrated to obtain
pðrÞ ¼ p0 þ 12rO2 r2 , (29)
Some discussion of pressure distribution in swirl
flows is in order here, since some fluid mechanics where p0 ¼ p(0) is the pressure at the cyclone axis.
effects are very special, leading to ideas that In a similar fashion, for an outer zone that obeys a
contradict the usual intuition drawn from unidirec- pure, free vortex law vt ðrÞ ¼ G=r with G ¼ const.,
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 421

the integral results in A pattern recently verified by numerical calcula-

1

1 1
 tions, e.g. in [4,25], is the so-called ‘‘lip leakage’’: a
pðrÞ ¼ pa þ rG2 2  2 , (30) strong radial component of velocity, inwardly
2 ra r
directed just under the rim of the vortex finder.
where ra is an arbitrary radius and pa ¼ p(ra). There is possibly much more to say about secondary
Simply matching both velocity and pressure laws flow structures; for instance, Ref. [26] documents
at r ¼ ra, we get (also numerically) four different classes of them
along the cyclone. However, the question is possibly
G ¼ Or2a ,
much harder to rationalize.
pa ¼ po þ 12 rO2 r2a ð31Þ Other feature that can be explained now is the
and using Eq. (31) on Eq. (30), the outer pressure is existence of W-shaped profiles of axial velocity.
They result from the attenuation of swirl by the
r2
 
1 walls of the vortex finder, which flattens a pressure
pðrÞ ¼ po þ rO2 r2a 2  a2 . (32)
2 r distribution like that in Fig. 9. In this manner, the
Fig. 9 shows an example of the radial pressure pressure gradient is positive downstream and near
distribution given by Eq. (29) for rora and Eq. (32) the centerline, so that a dip in the profile of axial
for r4ra, and normalized to a unitary maximum momentum obtains [27,28]. The effect is more
pressure difference. The implication is clear: due to pronounced with larger vortex finder diameters.
the centrifugal force, the effect of a swirl is to Inside and nearby the vortex finder area, the
decrease pressure towards the axis of rotation. In velocity normally becomes negative, i.e., a region
real cyclones, it has been estimated that the of backflow develops; the phenomenon is concep-
difference may be as high as to rise pressure 30% tually the same as the recirculation ‘‘bubble’’ of
above the cross-sectional average [10]. swirl tubes [29] and burners [30], an can be classified
as a vortex breakdown of type 0 [31]. Some results
3.4. More on cyclone velocity patterns. that further document the flow in this region of a
cyclone are the numerical calculations of [4,32] and
The consequences of this fact are diverse. Firstly, the visualization described in [33].
it somewhat serves to complete the picture of inner In many cases, the dip in velocity persists inside
cyclone velocity patterns. Due to the cross-sectional the inner separation space, well below the vortex
pressure gradient, any rotated flow develops sec- finder. This can be seen for instance in the data of
ondary components that are forced to evolve in the [25,28,34,35] and in older and newer numerical
axial and radial directions. As we have already simulations [32,36–38]. According to Hoffmann
mentioned, these secondary flows can be of rele- et al. [27], this reflects the high sensitivity of the
vance, since they directly contribute to by-pass or swirling flow to the conditions in the vortex finder:
leakage of particles to the inner vortex, thus leading If swirl attenuation is also provided at the other end
to losses of separation efficiency. A generic sketch of (the dust extraction end) by means of a prolonged
secondary flow patterns in cyclones is given in [10]. tube, inner W-shaped profiles are suppressed due to
the same influence coming from the opposite
direction. Numerical results of [4] for a cyclone
equipped with a dipleg only exhibit V-shaped
profiles, thus contrasting with those of [32] with
direct dust discharge, and confirming the explana-
tion of Ref. [27]. Although a W-shaped profile with
backflow may appear beneficial for the separation
of particles, it is not; separation is actually more
efficient with V-shaped profiles due to the fact that
the ascending flow region is narrower and thus
subjected to higher swirl.
Finally, the radial pressure gradient makes the
vortex flow inherently unstable. Separation of solids
Fig. 9. Radial distribution of static pressure in a Rankine vortex can be upset as a result if the flow near the wall
with ra/rc ¼ 0.6 [4]. encounters otherwise normal deviations from an
 ARTICLE IN PRESS
422 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

ideal surface: weld seams, measuring probes, riveted second effect is always the largest by far, so that,
unions and the like. This makes cyclone perfor- surprisingly (or not so by now), increased wall
mance difficult to predict, even by sophisticated friction decreases pressure drop. This has been
calculations. In fact, as we will see later, instability documented by many studies; see a good example
of the double-vortex structure influences all the in [24]. Exactly the same argument explains the
cyclone flow features, being the key to completely effect of solid loading and body length. For
explain its operation. instance, the experimental studies of [39,40] verified
that lower tangential velocities do result when dust
3.5. Pressure drop in cyclones concentration is increased, which was attributed to
increased wall friction due to separated solids
Logically, vortex motion and its associated radial covering the wall.
pressure gradient cannot be ignored when consider- A second subtlety concerns the very meaning of
ing pressure drop in cyclones. As if the flow were pressure drop measurements. As we have defined it,
unidirectional, pressure drop for clean, smooth- cyclone total pressure is equivalent to stagnation
wall cyclones can be split in three contributions: pressure, or, in other words, the pressure that would
(1) losses at the inlet; (2) fluid friction in the be measured by reversibly stopping the fluid that
double vortex within the separation space; and swirls at a velocity vt in an ideal flow rectifier.
(3) losses in the vortex finder and exit duct. Among However, this is irrelevant here. Pressure drop in
them, the first is usually of minor importance and cyclones is measured by the usual means of static
the last is the largest. However, the underlying pressure taps on the wall of inlet and outlet ducts.
mechanism is not the usual one, as the following Assuming good instrumentation practices, such a
experimental trend clearly points out: cyclone pressure probe senses perpendicular force on the
pressure drop decreases with increased wall friction wall, be it under a swirl component of velocity or
coefficient, concentration of solids or length of the not. Therefore, the magnitude detected is the static
apparatus. pressure in any case. The fact that there is a
The explanation of these perplexing effects is centrifugal force has nothing to do with stagnation
common and lies in the fact that it is not merely the whatsoever; otherwise, the equilibrium expressed by
normal or ‘‘static’’ fluid pressure p what is being Eq. (28) will be counted twice, so to speak.
lost, decreased or dissipated. As a correctly written The question is then that measurements at
mechanical energy balance may put forward, cyclone outlet (and thus pressure drop data) are in
viscous dissipation results in a decrease of the principle difficult to interpret. For normal cyclone
quantity p þ 12 rv2 , which we usually call ‘‘total’’ arrangements, it happens that a tangential velocity
pressure. Here v is the modulus of the velocity, but component persists at the exit duct connected to the
the axial component is of like magnitude at cyclone vortex finder. Therefore, according to our argu-
inlet and outlet, as a result of continuity and similar ments, what is really measured is a static pressure
cross-sectional areas. Accordingly, one can think of at the wall that is higher than the cross-sectional
a total pressure inside the cyclone made up of the average that should make up an overall balance.
static pressure plus the kinetic energy of the vortex, On the other hand, the ‘‘dynamic’’ component
p þ 12 rv2t . 1 2
2 rvt of the ‘‘total’’ pressure is neither taken into
Pressure losses in cyclones are dominated by the account.
viscous dissipation of this quantity in the vortex If (and only if) we define ‘‘pressure drop’’ or
finder. Such a dissipation is roughly proportional to ‘‘pressure losses’’ as the viscous dissipation of total
the absolute magnitude of v2t , so that any influence pressure, which is of course what makes more sense,
that tends to increase the strength of the vortex both effects are opposite: detecting only static
increases the losses and vice versa. For instance, an pressure tends to increase apparent losses, whereas
increase of the wall friction coefficient results in the fact that the measurement is higher than the
increased losses in the separation space, as it would average tends to decrease them. Strictly speaking,
be in the absence of rotation. However, at the same there is no reason why these two tendencies should
time it also brings about a decrease in the magnitude exactly compensate, so that the question remains
of vt, which in turn leads to decreased losses in the open and we can still expect a high uncertainty, say
vortex finder. Since vt is higher in the inner vortex, perhaps of 720%, when interpreting pressure drop
and the relevant variable is actually vt squared, the reports.
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 423

Fortunately, things are this once much more 3.6. Calculating the pressure drop
friendly to the researcher or engineer. For instance,
if we assume solid body rotation inside the vortex Dimensional analysis helps to identify relevant
finder, Eq. (28) applies for 0prpre. Using it, it is variables and organize empirical or theory-based
easy to show that both static pressure at the wall formulae for pressure drop in cyclones. Taking for
p(re) and the cross-sectional average of the total instance the approach of [16], we get, as a rather
pressure pðrÞ þ 12 rO2 r2 ¼ p0 þ rO2 r2 attain a com- complete list of variables
mon value, p0 þ 12 rO2 r2e . These are of course very
idealized velocity and pressure profiles, but it has Dpc
xc  1 2
¼ f ðgeometry; Frc ; C si ; Rec ; rs =rg Þ. (33)
been shown that the same sort of compensation 2 rg vi
takes place approximately with real vortex finder
flows [24]. Dimensionless pressure drop xc (also called the
Summarizing, for simple exit arrangements, Euler number, Eu) is customarily defined with
the usual (static) pressure measurement gives reference to the inlet gas velocity, although there
an adequate account of cyclone pressure losses are other possibilities. As in most equipment
if the latter are understood as total, dissipative operating under turbulent flow, the dependence on
viscous losses, not just simply the input/output Reynolds number is only relevant up to a certain
difference in static values. Cyclone designers and value, and usually negligible for the values found in
users should be aware however than dissipation of practice [41]. Other two parameters, the Froude
the swirl will certainly continue downstream of the number Frc ¼ v2i =gDc and the solid to gas density
exit duct, up to the point that almost all the ratio rs/rg, can be also eliminated, on the basis that
‘‘dynamic’’ pressure is lost, with no reversible their variation for a certain class of cyclone designs
recovery of any (static) pressure. Therefore, total and operating conditions is usually small; it can also
system loss is higher than the figure reported by be added that their influence is nil in the dilute flow
cyclone studies. Conversely, if the cyclone dis- limit.
charges to the atmosphere, and no exit pressure This leaves us with the sole influences of cyclone
measurement is done, the excess pressure at inlet geometry and solids loading; this is a possible
directly indicates total losses, since the sudden explanation of the structure of most pressure drop
dissipation of swirl at the outlet is added up by this models for cyclones. In many of them, the two
procedure. But then the pressure drop figure is not effects are introduced as independent, multiplicative
comparable to that obtained by using pressure taps factors to compose an overall Euler number
at the exit duct. xc ¼ xg xs , (34)
On the other hand, the use of flow rectifiers really
diminishes pressure losses by means of a (partial) where xg represent the limit of dilute flow, i.e., the
recovery of the dynamic component as a static loss that would occur in the absence of particles and
pressure, and certainly makes the measurement thus it only can be a function of geometry. xs is a
downstream unequivocal. The idea is very old, as correction factor that accounts for the presence of a
it is the dispute about the pressure recovery high concentration of solids. Of course, the latter is
attainable and the side effects on vortex dynamics an ad hoc correction, so that the factorization
and separation efficiency. The recent work reported expressed by Eq. (34) is devoid of any fundamental
in [33] reviews the question and contributes to its significance.
enlightenment. According to it, a rectifier located The most widely used correlations for the ‘‘clean’’
downstream of the vortex breakdown inside the pressure loss coefficient xg are summarized in
vortex finder has little effect on cyclone perfor- Table 3, Eqs. (35)–(44). Some of them are empirical,
mance but a limited potential, since the swirl has such as the simple formulae of [42,44], but also the
been largely dissipated already. To attain reduc- more complex formulation of Barth [13]. Actually,
tions of 30–50%, as reported in early work, the the formula for the loss in the cyclone body, Eq.
device should be located somewhat protruding from (39), is theoretical, based on the friction surface
the vortex finder, where the tangential velocity is concept, but, according to the data, it gives figures
still high. But then, logically, there is also a considerably lower than the main contribution: the
significant, deleterious effect on vortex stability loss in the vortex finder calculated by Eq. (40),
and separation. which is purely empirical. Muschelknautz and
 ARTICLE IN PRESS
424 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Table 3
Algebraic models of the ‘‘clean’’ pressure drop xg in cyclones

Model Equation Remarks

Shepherd and 16ab Tangential inlet; ambient

xg ¼ (35)
Lapple [42] D2e air conditions

"  2p !  2n #

Alexander [11] Experiments with scroll
 
ab Dc 1n Dc
xg ¼ 4:62 1 þfg (36) and tangential inlets
Dc De De n De

4  22n
     
1 1n  1n Air and combustion gases,
þ 0:2 22n  1 þ 1:5 22n
  
f g ¼ 0:8 
nð1  nÞ 3 n n up to 1100 1C
(37)

T 0:8
 
n ¼ 1  0:67Dc0:14
 
(4)
283
!2
Barth [13]   ab
x l ¼ lg ¼ ðxb þ xe Þ (38)
pD2e =4

Loss in the cyclone body

 2 !
De 1 vte
xb ¼ 2  (39)
Dc vze =vte  ððH  SÞ=ð0:5De ÞÞl vze


Loss in the vortex finder

 4=3  2
vte vte 3.41oKo4.4
xe ¼ K þ (40)
vze vze
!
Muschelknautz and   ab Tangential and scroll inlets
Kambrock [43] x l ¼ lg ¼ ðxb þ xe Þ (41)
pD2e =4

AS rg Flow field based on Barth0 s

xb ¼ l ðvtw vte Þ1:5 (42)
0:9V 2 model [13]
 4=3  2
vte vte Ambient P,T conditions
xb ¼ 2 þ 3 þ (43) l ¼ lgE0.006
vze vze
AS is the total inner area of
cyclone contributing to
friction
!2
Casal et al. [44] ab Comparative study of six
xg ¼ 11:3 þ 2:33 (44) correlations
D2e

Kambrock method [43] patterns the general scheme The models of Barth [13] and Muschelknautz and
of Barth. Kambrock [43] use the velocities vtw and vte given by
According to our analysis, some methods just Eqs. (5) and (6), which amounts to introduce a
assume that xg is only a function of geometry, which explicit friction coefficient lg for cyclone walls. This
in fact constitutes an elementary scaling rule, for is included in ‘‘geometry’’ or already assumed of
smooth-walled cyclones operating at high Reynolds typical value in the dimensionless Eq. (33), but in
numbers and low solid loadings. However, Alex- this way, variations of wall roughness can be
ander [11] already took into account variations of handled.
gas viscosity with temperature, by using in his Table 4 compares data from the experimental rig
formulae, Eq. (36) and (37), the exponent n, Eq. (4). of [45] with the predictions of formulae in Table 3.
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 425

The experimental value of xg is an actual ‘‘clean’’ calculated as above, taking into account vortex
pressure drop, i.e., measured without solids loading. penetration in dipleg. These results offer again a
All the correlations perform rather modestly, good agreement with measured data, although the
although Muschelnautz’s is clearly the best. How- recommendations of Baskakov et al. perform
ever, the experimental cyclone of [45] was a model remarkably well, if not better.
of a PFBC unit equipped with a long dipleg. The studies that led to the correlations of Table 5
Allowing the vortex end to penetrate the latter and Fig. 10 also revealed another significant fact.
(which is actually an observed fact of the experi- For low-to-medium inlet solid loadings, the pressure
ment), we were able to predict xg with total accuracy coefficient xs deceases monotonically with solids
by Muschelknautz method. The comparison is fair concentration, in accordance with the physical
for the rest of formulae because such a modification explanation of the effect. Nevertheless, for very
is not possible with them, whereas tangential high solid loadings (such as in CFBC) the opposite
velocities of Eqs. (42) and (43), and thus xg, depend has been detected. Baskakov et al. [48] found a
on vortex end position through Barth’s vortex minimum located in the xs–Csi curve at 200 g/kg.
length, Eqs. (8), Fig. 7. This most probably represents the very limits of the
The factor xsp1 estimates the reduction in effect, i.e., the point at which no more friction
pressure drop due to a high solids loading. There due to particle strands can decrease further the
is considerable uncertainty in cyclone literature as tangential velocities, so that the ‘‘normal’’ effect
to what should be understood by ‘‘high’’ in this of an increased friction begins to dominate. Chen
context. The right answer lies possibly within the et al. [49] obtained similar results for various
range of 25–50 g/kg, considering only effects in cyclone geometries, although the minimum was
observed pressure drop. Eqs. (45)–(49) in Table 5 located at higher inlet concentrations, in the range
summarize the most cited models for xs. Most of the 400–700 g/kg. Similar trends have been found for
studies have taken the simplified approach of CFBC cyclones in other experimental studies
assuming xs only a function of the inlet solids [50–52].
loading Csi—the alternate variable Csirg (kg/m3)
being also very popular, if not dimensionally 4. Collection efficiency
coherent. This amounts to our reasoning that
Froude and density ratio numbers do not vary All the ideas on the mechanics of cyclone flow we
typically much, plus the additional assumption that have discussed thus far are only a part of the way to
the multiplicative factor xs is universal, not depen- explain how the basic purpose of this kind of
dent on geometry, or that the formula is restricted equipment is effected: separation of solids from
to geometrically similar apparatuses of certain dust-laden gas streams. Let us give here a brief
design. account of the question. Collection of particles
On the other hand, if we recall Eqs. (11) in inside a cyclone is naturally a result of the forces
Table 2, it is clear that Muschelknautz method [6] acting on them, whose resultant drives them to
can go a step beyond, being able to account for the cyclone walls. Literature always lists centrifugal,
effect through an augmented friction coefficient, drag and gravitational forces, but in addition, there
which is much more sound from a physical might be others, not entirely understood and often
standpoint. The table shows the way of introducing neglected, such as particle-particle and particle-wall
this within the framework of Eq. (34). The method interaction, that surely influence the collection
also considers geometry and Froude and den- process.
sity ratio numbers, possibly being the most com- The collection ability of a cyclone is measured by
plete available (and not proprietary of cyclone its collection efficiency Z, defined as the fraction of
manufacturers). the inlet flow rate of solids separated in the cyclone.
Fig. 10 evaluates the different methods by Since a cyclone usually collects particles possessing
comparing their predictions with data from the a wide range of sizes, it is common to work also
PFBC cyclone of [45]. The value of xs is the truly with different efficiencies, each defined for a
correction factor defined in Eq. (34), calculated particular and narrow interval of particle sizes.
from measured pressure drops with and without Imagining indefinitely small intervals, we get a
solids loading. Calculations following Muschel- continuous function Z(x) that can be thought of as
knautz need the clean pressure drop, which is the fractional or grade-efficiency of the cyclone for
 ARTICLE IN PRESS
426 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

particles of size x. Ref. [10] gives a mathematical always laminar and ordered, there will be an abrupt
definition of Z(x) in relationship with Z that is very cut at some intermediate x at which particles would
enlightening. Here we limit ourselves to the reverse cease to be separated to escape or vice versa. (Some
relationship, most obvious: if f(x) is the particle size slight dispersion will appear however due to
distribution (PSD) at cyclone inlet differences in the position of the particle at the inlet
Z 1 section.) This is not the case; for many reasons, a
Z¼ f ðxÞZðxÞ dx. (50) fixed particle size is separated with a probability
0 greater than cero and less than unity. As a
consequence, Z(x) has the generic shape of a Sigma
The grade-efficiency curve can be conceived too function between the limits Z(0) ¼ 0 and Z(N)-1.
as the true measure of the cyclone effect, since by its Then, a simple characterization is made up of the
own definition, it depends only on cyclone char- so-called cut size x50, the size which is separated half
acteristics, but not on inlet PSD. On the contrary, the time, Z(x50) ¼ 0.5, and the slope of the grade-
we can see in Eq. (50) that the total efficiency efficiency curve at that point. Fig. 11 is a scheme of
depends on both, so that it is not only a the relevant definitions concerning cyclone collec-
characteristic of the apparatus. tion efficiency.
It is also pretty obvious that very large particles
will be always separated, whereas very fine material
will always escape. If fluid and particulate flow were 4.1. Models of collection efficiency

As with pressure drop, cyclone efficiency was first

Table 4 modeled in the dilute limit, and afterwards corrected
Comparison between measurement of xg and predictions [45]. for high concentrations of dust, the so-called ‘‘mass
xg Discrepancy loading effects.’’ For the first class of models, it has
(%) been tradition to build theoretical constructs from
very idealized arrangements of particle forces and
PFBC cyclone, measured [45] 11.6 — velocities, far removed from the chaos associated
Shepherd and Lapple [42], Eq. (35) 15.7 35.3
Alexander [11], Eq. (36) 13.6 17.2
with turbulent fluctuations and dispersion, and the
Muschelknautz [6,43], Eq. (41) 14.1 21.6 natural instability of the vortex structures. In spite
Muschelknautz [6,43], Eq. (41), 11.7 1 of this, some models have been surprisingly
including vortex penetration in successful.
dipleg [45] Most popular hypothesis are: (1) gravitational
Casal et al. [44], Eq. (44) 14.2 22.4
field negligible compared to centrifugal forces, (2)

Table 5
Formulae for the coefficient of solid loading effects xs [45]

Reference Equation Remarks

Briggs [46] 1
xs ¼ (45) Air at ambient conditions
1 þ 0:0086ðC si rg Þ0:5 Multivane cyclone of 0.23 m diameter
Solids: rock dust 0–44 mm, Csi range: 0.17–192 g/kg air
Smolik [47] xe ¼ 1  0:02ðC si rg Þ0:6 (46) Experimental data from several sources

Baskakov et al. [48] x ¼ 1

s þ 0:67C si (47) Hot gases at ambient pressure and 150–250 1C
1 þ 3:1C 0:7
si Helical inlet cyclone of 0.204 m diameter
Solids: alumina xm ¼ 40 mm, Csi up to 0.7 kg/kg gas
Muschelknautz [6] xðl ¼ ls Þ
xs ¼ (48)
xg

5=8 See Table 3 for xg

rg
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Dc
ls ¼ lg þ þ0:25 ZC si Fr0:5 e (49)
rs ð1  str Þ ðDc  bÞ
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 427

1
0.9

Cyclone pressure coeficient, ξs

0.8
0.7
0.6
0.5
0.4 Briggs [46]
Smolik [47]
0.3 Baskakov et al. [48]
Muschelknautz [6]
0.2 Measured data [45]
0.1
0
0 50 100 150 200 250
Csi, g solids / kg gas

Fig. 10. Cyclone pressure coefficient vs. solids concentration at inlet. Comparison between correlations and measured data [45].

gas density negligible vs. particle density, (3) for discrete sizes xi and fractional efficiencies Zi. The
particles are spherical, of low size and the relative exponent m is the slope of the curve, to be adjusted
velocity is small enough for Stokes law to apply, to the data.
and (4) relative velocity is purely radial. Under these Algebraic efficiency models pertain to either of
hypotheses, the following equation of motion of a two broad classes. In principle, the idea could be to
spherical particle rotating at an arbitrary radial perform the integral of Eq. (51) from the inlet, and
position applies: see whether the particle in question has time to
reach the wall. A model that does exactly this is
px3 dur px3 v2t called a ‘‘time-of-flight’’ model. The idea needs
r ¼ r  3pmxur , (51) however a complete battery of simplifications,
6 s dt 6 s r
especially if a more or less closed, algebraic method
of prediction is pursued. For this reason, such
where ur is radial velocity of the particle relative to drastic hypotheses have been used that the concept
the gas, x is particle diameter, rs is particle density of a Lagrangian track of particles is almost lost. The
and vt is tangential velocity of both particles and most successful has been the so-called notion of
gas. The first right-side term is the centrifugal force ‘‘equilibrium orbit’’ firstly put forth by Barth.
and the second the drag. For the particle sizes and Most popular formulae for cyclone efficiency
Reynolds numbers encountered in industrial cy- calculation are summarized in Table 6.
clones, the Basset and displaced-mass terms are Lapple [53]: It is similar to the method for sizing
clearly negligible. However, turbulent diffusion is sedimentation chambers, the simplest version of the
flagrantly ignored. time-of-flight ideas. A radial initial position is taken
Classical models for calculating cyclone efficiency at the middle of the entrance duct (b/2), a uniform
result from integrating Eq. (51) under a manifold of particle distribution across the inlet section is
flow situations and hypotheses. Afterwards, some assumed and a constant tangential velocity is used,
notion (also simplified) must be imposed to decide if all of this in order to ease the integration of Eq. (51).
a given particle is collected or not. The calculated This serves to calculate the time for a particle of
parameter is often x50; the complete curve can be given size to reach the wall. The particle whose time
then adjusted by some other means, frequently of equals the gas residence time is considered to
empirical nature. For instance, a very popular be separated with 50% efficiency, and thus its size
curve-fit has the form is x50.
Barth [13]: Barth introduced the notion, reason-
1 able but arbitrary, that a particle subjected to radial
Zi ¼ (52)
1 þ ðx50 =xi Þm force equilibrium just at the CS surface (Fig. 7) will
 ARTICLE IN PRESS
428 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

100
90
80

Fractional efficiency, %
70
60
50 X50
40
30
20
10
0
1 10 100 1000
Particle size, μm

Fig. 11. Typical fractional efficiency curve.

be separated half the time, since under ideal geometric by Barth. In Section 5.1 we will return to
conditions it would remain there forever. Therefore, the importance of this fact.
equating Eq. (51) to zero is all it is needed to Although very popular, the model of Leith and
calculate x50. The result is included in Table 6 as Licht was incorrect in its derivation, as put forward
Eq. (54). Its resemblance with the time-of-flight in [55], due to a basic flaw in the definition of
formula given by Eq. (53) is at least surprising. average residence time for a continuous flow system.
However, differences are significant. One of them is Correcting the mistake led to a simpler formula, but
the fact that Barth used his own tangential velocity not to a proper model of the fractional efficiency
at the surface CS. This makes the method sensible to curve.
an important parameter: the diameter of the vortex More sophisticated models can be found in the
finder De, a geometrical feature that greatly literature, such as Dietz’s [56], where the cyclone is
influences cyclone efficiency. In contrast, only the separated in three regions, assuming perfect mixing
inlet velocity is used in most time-of-flight models, of solids in the radial direction. This model is
such as Lapple’s; they lack accordingly this im- hybrid, using ideas both from time-of-flight and
portant capability. equilibrium-orbit paradigms. It had the same basic
In this respect, it has been shown recently [25], flaw as Leith and Licht’s, but could be corrected
both by measurement and numerical calculation, satisfactorily, with a lesser shift of the predicted
that the reasonable ‘‘Barthian’’ cylinder CS does not grade efficiency curve [55]. As can be imagined,
really exist, in the sense that the width of the Muschelknautz work [6] served also to improve
ascending flow tube seems to be not influenced by Barth’s efficiency, by including the effects of wall
De, but only by the cyclone diameter Dc, at least for friction and solids loading through the tangential
some kind of apparatuses. The effect of De on velocity calculation (more on this in Section 4.2).
collection efficiency would result from the change of Other models available are even more complex in
local flow patterns just under the vortex finder. nature, although in essence, most reduce to mod-
Leith and Licht [54]: This is a more elaborated ifications of Barth’s equilibrium orbit, a seminal
time-of-flight model that permits to determine the idea. For instance, the extension to cyclone zones
entire grade–efficiency curve. To this end, the different from the main separation space is a
authors considered a continuous flux of dust instead recurrent idea, as in [57]. Finally, Mothes and
of single particles. Eq. (51) is thus not used, but the Löffler [14] is a hybrid model that added a finite
continuous advection of material balanced by particle dispersion coefficient in an effort to
centrifugal forces and inertia. Other important improve the prediction.
assumptions are negligible gas radial velocity (as
in most time-of-flight models) and a constant 4.2. Mass loading effects
concentration of particles along the radius. Assum-
ing this being a result of turbulent diffusion served All these models of collection efficiency usually
to overcome the limitation we commented above. perform reasonably well for low solids loadings, less
Results are expressed as a function of the cyclone than 5–10 g/m3 at cyclone inlet. However, many
‘‘natural length’’, a parameter that was assumed industrial units handle higher loads, that can ever
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 429

Table 6
Formulae for the cyclone collection efficiency

Reference Equation
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Lapple [53] 9mg b h þ ðH  hÞ=2
x50 ¼ ; Nc ¼ (53a,b)
2prs vi N c a

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Barth [13] 9mg De vre
x50 ¼ (54)
rs v2tw

" #
Leith and Licht. [54], corrected by Clift et al. [55] rs 2xvi 2
 
Z ¼ 1  exp  tres (55)
9mg Dc

V
tres ¼ (56)
V_

reach 10 kg/kg, e.g. in FCC, PFBC or CFBC dense (Csi4Ccr) two-phase flow
applications. Under these conditions, the gas cannot
be treated as if the particles were absent, and the Z ¼ Z0 ; C si pC cr , (57a)
latter cannot be modeled as a single particle,  
roughly following the gas and without interactions C cr C cr
Z¼ 1 þ Z; C si 4C cr . (57b)
with other particles. In other words, the simplicity C si C si 0
of a dilute flow should be abandoned and more
In Eq. (57b), the first term accounts for the
involved ideas should be applied.
efficiency of the solids separated by ‘‘sedimenta-
This has been accomplished only to a partial
tion’’ at cyclone inlet, whereas the second stands for
success. Experimental facts are, again, perhaps
the collection efficiency Z0 of the remaining load in
perplexing: collection efficiency Z increases as the
the separation space. In principle, Z0 is equal to the
concentration of solids at inlet Csi is increased. A
collection efficiency of the cyclone at low solids
part of the perplexity is explained because the
loadings, and this is assumed in Eq. (57a). However,
increase in Z proceeds at a lower pace than that of
as we have seen, the solids separated at the entrance
Csi, so that the absolute amount of solids emitted—
have a definite influence on cyclone flow, decreasing
lost—always increases with Csi. Another part can be
tangential velocity through increased wall friction.
accepted if we consider that the physics of a
This obviously influences collection efficiency and
concentrated, two-phase flow actually favors parti-
thus can make the Z0 in the second formula different
cle separation
from the Z0 in the first one. As we have seen too,
This is the point adopted by the main approach to
some models of flow and efficiency in cyclones can
model mass loading effects: the concept of a
handle this difference.
‘‘critical load’’, firstly proposed by Muschelknautz
Other than this, the question reduces itself to
in his early studies [6]. Tracing an analogous with
estimate the value of the critical load Ccr. The
sedimentation effects in pneumatic transport,
original formula of Muschelknautz [6] was derived
Muschelknautz reasoned that a form of sedimenta-
from reasoning on his velocity distributions
tion of particles occurs under the centrifugal force,
pffiffiffiffiffiffiffiffiffiffiffi
just at the cyclone entrance. The turbulent energy of l Dc De m
the gas stream only supports a limited weight of C cr ¼   pffiffiffiffiffiffiffiffiffiffiffi , (58)
2 1  De =Dc rs x2m vte vtw
solids, a critical load Ccr, settling the rest unclassi-
fied. Therefore, if solid inlet concentration Csi is where xm is the particle mass-median diameter
increased above Ccr, cyclone efficiency increases and the other terms are derived from flow field
regardless of particle size. calculations.
In this manner, collection efficiency is calculated However, more recent experimentation [7] has led
differently for the regimes of dilute (CsipCcr) and to different expressions, not directly based on
 ARTICLE IN PRESS
430 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

sedimentation studies efficiency is calculated taking into account an

x50 extended vortex length, as we did with the pressure
C cr ¼ 0:025 ð10C si Þ0:4 ; C si o0:1, (59a) drop.
xm

x50 5. Special phenomena associated with the flow field in

C cr ¼ 0:025 ð10C si Þ0:15 ; C si 40:1. (59b) cyclones
xm
Ref. [7], based on the work of Muschelknautz [6] 5.1. Natural turning length
and Barth [13], also developed a method to correct
the dilute efficiency Z0 for the effect of a decreased As we have seen, in a reverse-flow cyclone, the
vortex intensity, that can reach up to 60% for values outer vortex weakens and changes its direction at a
of Csi ¼ 1 [58]. But the main novelty of this work certain axial distance Ln from the vortex finder.2
was the hypothesis that some classification also This magnitude is usually called the ‘‘turning
occurred at the inlet, changing the PSD of the length’’, ‘‘natural length’’ or ‘‘vortex length’’ of
particles finally suspended in the vortex. By the cyclone, and the axial position is referred to as
considering a mean centrifugal acceleration, a ‘‘the end of the vortex’’. This point can be imagined
settling velocity was calculated at the entrance duct, as an effective end of the apparatus, since almost all
from which an inlet cut-size was determined. the gas has leaked entirely out to the inner vortex at
This has been signaled as inconsistent with an this position. Therefore, both pressure losses and
observed uniform increase of the grade-efficiency particle separation are mainly determined by events
curve that would not warrant any kind of sharp occurring above; what is left below is an induced,
classification due to the cyclone effect. In any case, secondary vortex that cannot contribute much, only
fractional efficiency and its variations with load to re-entrainment, in poorly-designed extraction
usually exhibit lifts and hooks that are not easily systems.
reconciled with physical evidence. An example from As a consequence, cyclone designs with a natural
[45] is shown in Fig. 12. The fact that efficiency length some measure greater than the physical
seems to be higher for smaller particles leads us to a length are advisable [60,62], since a lack or
second mass-loading effect that has been far less performance or an oversized unit results otherwise.
explored. The question is then what determines this para-
It consists in the possibility that the temporary meter. In his influential theories, Barth assumed
adhesion of small particles to larger ones, or the that h*, the effective cyclone length, Fig. 7, was a
swept of the former by the latter inside the vortex, mere function of geometry, i.e., that it changed
could be another significant factor for an increased exclusively with cyclone design. Although many
efficiency at high solids loadings [14,59,60]. In fact, useful results were derived in part from this
this is almost the only way a minimum in grade hypothesis (as we have seen), it is not generally
efficiency can be explained, once particle attrition is true. We have learned from experience that the
accounted for. The agglomeration effect seems to natural length of cyclones is influenced by dynamic
have been detected for hot gas cleaning applications factors. Moreover, the vortex end can be itself a
in the Grimethorpe PFBC cyclones [61]. However, dynamic and complicated phenomenon.
up to now, particle interaction effects have not been It is thus not strange that its true nature wasn’t
neither confirmed nor modeled in this context, fully explained until very recently. Initially it was
neither as a modification of the Z(x) curve, nor to assumed to be an axisymmetric flow structure
the total Z value. caused by the axial and radial pressure gradients,
Finally, to give an idea of the performance of perhaps related to the phenomenon of vortex
present efficiency models when used in the high breakdown as observed in once-through swirling
loading regime, Fig. 13 compares data from [45] flows [31]. However, this is possibly true only at low
with the calculations according to Trefz and velocities and high wall friction coefficients, i.e., at
Muschelknautz [7]. Agreement is much better than low swirl [63]. Under realistic conditions, the double
that obtained with simpler, older efficiency models, vortex bends and attaches itself to the lateral wall,
limited to low concentrations, which is well ex- and, superposed to the vortex swirl, the bend itself
plained by the separation of the critical load at inlet.
2
Moreover, agreement is even better if the dilute Customarily measured from its lower rim, as in Fig. 7.
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 431

105

100
Fractional efficiency, %

95
71 g /kg

90
114 g /kg

85 169 g /kg

80
1 10 100
Particle size, μm

Fig. 12. Experimental collection efficiency as a function of solids inlet concentration [45].

rotates at frequencies in the order of several throughout the years. For instance, the erosion ring
tens of Hz. Thus, the cyclone vortex does not at the lower part of the conical section, frequently
end inside the fluid, but on its boundary, and it is encountered in cyclones operating with hard pow-
not, definitely, an axisymmetric and stationary ders [68–70], is simply a witness of the circular
structure. movement of the vortex end on the wall. So is the
The history of how this was discovered deserves ring made of residual dust frequently observed in
some attention; a swirling flow that descends, bends experimental units with transparent walls, e.g., in
on a wall, reverses its stream direction and [17]—see also the interpretation of [66]. Also, the
continuously changes the axial plane where every- decrease in cyclone efficiency with a shortened
thing happens is certainly not easy to detect. In fact, vortex that cannot be completely explained by the
modern studies, both experimental and numerical, mere reduction of Ln [71]. And finally the drop in
normally miss the detail if not looking specifically at static wall pressure and tangential velocity measure-
it [32,64], and papers continue to be published still ments observed by many, e.g., [4,17].
seemingly unaware of the true nature of the vortex In the numerical arena there is clear evidence too.
end [65]. Pioneers are Montavon et al. [72], who reported
Observations of a bended vortex attached to the ‘‘interaction with the wall’’ apparently unaware of
wall were already made by Muschelknautz in the the experimental work done on the question. In Ref.
1970s [43]; by the end of the 1990s, the fact seemed [4], CFD data was used to ‘‘visualize’’, in striking
almost to pertain to the traditional knowledge on resemblance to the stroboscopic experiments, the
cyclones, see for instance the ‘‘regimes’’ explained in attachment of the vortex to the wall. (We will return
[63]. A definite evidence was blatantly absent to this later.)
however. Hoffmann and Stein [10] firstly reported This picture cannot be complete without explain-
unpublished visualization experiments with strobo- ing why the cyclone vortex behaves in this way. The
scopic lighting that fully supported the fact. Finally, reason is that the whole vortex is inherently an
the same research group was able to duplicate and unstable structure, due to the radial pressure
document the experiment, publishing solid evidence gradient developed by itself, as already mentioned
[66]. They further characterized the phenomenon in in Section 3.4. As a consequence, it oscillates
terms of frequencies and parametric dependencies; radially at any axial location, which is known as
more recently, they have made consistent observa- the phenomenon of the precessing vortex core
tions by visualization with a neutrally buoyant (PVC), common to any swirl flow of practical
tracer and time-resolved wall pressure measure- interest. In cyclone separators, the vortex bends,
ments [67]. attaches to the wall and rotates on it because this is
The explanation agreed very reasonably with a simply the way in which the PVC ends in that case.
vast ensemble of indirect evidence obtained The implication is that unsteadiness of cyclone flows
 ARTICLE IN PRESS
432 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

100
PFBC cyclone
Predicted
Cold model overall efficiency %
99

98 inner separation

97

96 separation at inlet

95

94
0 50 100 150 200 250
Inlet solid concentration, g solids/kg gas

100
PFBC cyclone
Cold model overall efficiency %

99 Predicted

98 inner separation

97

96 separation at inlet

95

94
0 50 100 150 200 250
Inlet solid concentration, g solids/kg gas

Fig. 13. Comparison of total collection efficiency obtained experimentally and calculated: (a) model of Trefz and Muschelknautz [7], (b)
same model, accounting for the length of the vortex [45].

is not limited to the position of the vortex end, but is inlet Reynolds number. In another exhaustive
rather a feature of the whole pattern [73]. review [74], two distinct types of cyclone operation
were identified, as depicted in Fig. 14:
5.1.1. Relevant parameters influencing natural
turning length  Operation with LnXH, Fig. 14a. In this case, the
Predicting natural length of cyclones is a complex natural vortex length is equal or longer than
endeavor, and there is no complete agreement about the physical length, and the vortex end reaches
the influential parameters. A review of the possibi- the bottom of the cyclone, or even the dust
lities most explored in the literature is given below. collection or extraction hardware. This is the
Inlet Reynolds number: There was no documented most advisable mode of operation, since it leads
evidence of the influence of this variable until the to a high collection efficiency. It corresponds to
last decade, what amounts to say that the standard, cyclones with relatively high inlet Reynolds
‘‘Barthian’’ picture of a stable flow only determined numbers (approximately ranging from 2  103
by geometry has been the dominant view until to 2  104) and H/Dc ratios within 2–10.
recent times. A review of cyclone research [63]  Operation with LnoH, Fig. 14b. It entails poor
found, after analyzing an exhaustive database, a collection performance, because the vortex end
strong dependency of the natural length Lm on the attaches to the cyclone wall, disturbing the solids
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 433

strands that are already separated and decreasing

efficiency through instability and re-entrainment.
This mode of operation corresponds to low inlet
Reynolds numbers (o103) or very tall cyclones
(H/Dc415).

These ‘‘regimes’’ have been recently observed by

flow visualization [67]. Reynolds numbers in prac-
tice are usually high enough to allow high values of
cyclone height, so that the majority of cyclones
should normally operate in the range of good
performance (LnXH). There is however not total
agreement on the quantitative effect. For instance, Fig. 14. (a) Cyclone operation with Ln4H. (b) Operation with
LnoH and the vortex-end attached to the lateral wall. Adapted
the influence detected in [75] is significant but not as from [63].
strong as that found in [74].
Geometric parameters: Aside from the ratio H/Dc
itself, the most relevant geometric parameters are In relation with this, also the use of gas
the following: extraction flows for solids evacuation is known
to increase the natural length of the cyclone and
exert a strong stabilization effect on the vortex
 Inlet area: Alexander found [11] that Ln de-
[45].
creased proportionally to the inlet area Ai, but
the opposite trend has been also reported [76].
Wall roughness and solids loading: In laboratory
More exhaustive research seems to support
tests with gas cyclones operating with smoke, it was
classical Alexander’s findings [74].
found that the increase in wall roughness caused by
 Length and diameter of the vortex finder. There is
condensate covering the wall decreases Ln making
experimental evidence [71,75] that the length of
the vortex shorter [71]; the same effect was
the vortex finder S does not influence the position
consistently found with higher solids loadings in
of the end of the vortex.3 On the contrary, the
standard operation. The same observation has been
vortex finder diameter De has been identified as a
reported also in experiences with PFBC cyclones
definite influence. The trend most frequently
[17,45]. Recalling the influence of the Reynolds
observed is an increased natural length with an
number, the general implication seems to be that
increased De [10,11,75]. However, again, the
any parameter that attenuates the swirl intensity
opposite has been reported too [76].
leads to a shortening of the natural length.
 Discharge zone design. In industrial cyclones,
Instability and apparent randomness of the vortex
vortex stabilization devices are frequently used to
end position: The deleterious effect of a short vortex
attach the vortex at the lower end of the cyclone,
is not only due to the reduced value of Ln and the
thus making the natural length equal to the
movement of the vortex end structure. Actually,
geometrical length and stabilizing long cyclones.
reduction of the natural length is accompanied by
These are conical-shaped pieces, with an upward-
instability, in the sense that a short vortex losses
oriented apex and an axis aligned with the
space- and time-coherence to a great extent. For
cyclone axis.
instance, the investigation in [71] found that the
In the opposite situation of a short cyclone, the
vortex-end position (as visualized by smoke) was
vortex end is located in the dust hopper and
much more variable for larger vortex finder
induces there an unusual swirl, which may
diameters, i.e., for shorter natural lengths. Also in
jeopardize the collection efficiency through re-
[71], and in [17] among others, the vortex end
entrainment [67]. A swirl tube or dipleg can be
position (visualized by a ring of dust) was reported
then desirable, to accommodate the vortex end
to move up and vary continuously and apparently
and avoid this effect.
at random under some operating conditions. The
3
Which would invalidate the traditional origin for measuring ensuing highly variable flow patterns obviously
Ln, making more meaningful the roof of the cyclone for this entail a reduced efficiency and an increased pressure
role [71]. drop, and a high uncertainty in any measurement or
 ARTICLE IN PRESS
434 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

prediction. The reverse trend is also true, i.e., a 5.1.3. Influence of cyclone natural length on
long vortex is more coherent and stable, which collection efficiency and pressure drop
favors the separation effect and reduces losses and The influence of the natural length is clearly
uncertainty. shown when this parameter is included in some
efficiency models of advanced capabilities, as we
have demonstrated in Fig. 13 [17,45] using the
5.1.2. Calculation of Ln
calculation methods of Muschelknautz, [6,7]. How-
Up to now, adequate calculation methods of the
ever, there the vortex end was stable and located
cyclone natural length have not been devised. None
well inside a dipleg, so that the sole effect of a larger
of the available proposals (e.g., [11,76–78]) takes
separation space manifests, as it should be for a
into account all the observations made in relation-
properly operated cyclone.
ship with this phenomenon. Moreover, most of
In other circumstances, if a short vortex is
them still consider Ln as a geometric parameter,
present, we get a regime similar to Fig. 14b and
and, when compared among themselves, they
efficiency falls. As an example, the study in [75]
deliver diverging results and also trends, as it can
demonstrated that efficiency increases with geo-
be easily verified with the following expressions:
metric length: a 20% increase was detected when
Ln
 1=3
De D2c using different models of cyclones with H/Dc equal
Alexander ½11 : ¼ 2:3 , (60) to 4.65 and 5.65 instead of 2.65 and 3.15,
Dc Dc ab
respectively. However, when H/Dc was higher than
 1=2 5.65, collection efficiency slightly decreased, which
Ai was interpreted as the effect of the vortex end
Zenz ½77; Bryant ½78 : Ln ¼ 2Dc , (61)
Ae entering the separation space and disturbing the
flow there.
 2:25  2 0:361
Ln De Dc Only two approaches in the classical cyclone
Zhongli ½76 : ¼ 2:4 . literature account for the effect of the vertical
Dc Dc Ai
dimension of the cyclone on the total pressure drop.
(62)
Barth [13] included the length H–S in his calcula-
Considering the characteristics of the vortex end tion, Eq. (39), and in tangential velocity models,
we have just mentioned, it is not strange that these Eq. (6). In [61], tangential velocity and pressure
formulae perform very badly. (Alexander’s geome- drop are related to the length of the separation zone
tries and conditions were also too different from by considering the total inner friction surface of
modern designs.) Although no correlation effort has cyclone walls AS.
been reported incorporating dynamic parameters, As in the case of efficiency, if the measured vortex
there has been some success in predicting Ln by length is included in classical models, for instance,
numerical calculation. Velilla [4] used CFD results those of Muschelknautz, Table 3, the predicted
to calculate a natural length, defining it as the locus values fairly agree with experimental data. This has
where isobaric surfaces bended to the wall. The been already shown in Table 4; Fig. 15 further
figure thus obtained was in very reasonable agree- shows the question under variable solids loading.
ment with experimental observations of the dust The explanation lies in the identification of Ln as an
ring. Table 7 shows those results; the (completely effective length. Then, as explained in Section 3.5, a
deviant) predictions of the above formulae are also longer vortex increases friction, causing lower
shown for illustration. tangential velocities and thus a reduced pressure
Nevertheless, these experiences were for clean gas drop.
and the vortex end stabilized inside the dipleg. More Again, all this refers to stable vortex ends; it can
general conditions, i.e., a short and unstable vortex be expected that the higher dissipation resulting
provoked or not by a high solid loading, have not from a highly variable flow will increase pressure
been considered to date by this kind of methods. drop.
There are however favorable indications in the
literature; for instance the numerical simulations of 5.2. Precessing vortex core (PVC)
[72], and somewhat also those of [65], apparently
reproduced very realistically a highly chaotic PVC The PVC is a flow phenomenon originally
for non-interacting two-phase flows. observed in once-through swirling flows, such as
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 435

Table 7 The peculiarity of a reverse-flow disposition is that

Calculations of the natural vortex length [4] the PVC extends in both directions: down to the
Ln/Dc Remarks
separation space encompassing the double-vortex
and the vortex end structure, and also downstream
Experiment 9.51 Stable vortex end on the dipleg wall of the vortex finder, inside the connecting tube. As
CFD 9.85 +3.5% off from experiment an example, Fig. 16 illustrates flow visualization at
Alexander [11] 1.56 Predicts vortex end on cyclone cone
the exit cross-section of a laboratory hydrocyclone
Bryant et al. [78] 2.24 Ibid
Zhongli et al. [76] 14.62 Vortex end out of the dipleg model [81]; the displacement of the vortex core from
the axis is clearly noticed. Cyclone separators
operating at practical values of swirl and Reynolds
those in swirl tubes, burners and combustors. In numbers always exhibit a PVC as a quasi-periodic
general, the swirl intensity is characterized by the feature of the flow. Its properties depend however
swirl number Sw [79] on flow parameters. The coherence is high or low
under the conditions for a stable or unstable vortex
2G t
Sw ¼ , (63) end, as explained in the preceding section, respec-
Gx D
tively. Frequency of the movement is also variable,
where Gt is the axial flux of angular momentum, Gx as explained in Section 5.2.2.
is the axial flux of axial momentum, and D is a In swirl burners and combustors, where mixing is
characteristic diameter. In most occasions, the exact desired, a PVC is indeed highly beneficial [79]. On
calculation of Sw is too involved. But using reason- the contrary, given the objective of cyclone separa-
able approximations, in the constant-density case, tors, the phenomenon can be detrimental in them.
Eq. (63) can be shown to reduce to a geometric With independence of this, the cyclone PVC has
factor. The appropriate expression for a cyclone been found to influence aerodynamic noise [82].
separator is, for instance [80] Under some conditions, it may lead to deterioration
of the equipment through coupling with the natural
p De Dc
Sw ¼ . (64) frequency of large apparatuses, and even to damage
4 Ai
of some elements [83]. In any case, the PVC
Swirling flow patterns change dramatically with phenomenon has associated a significant energy
Sw. For swirl numbers about 0.6 and increasing expense, which means that it increases pressure
Reynolds numbers, a recirculation ‘‘bubble’’ is losses [84].
formed. It is initially axisymmetric, but if the swirl Other peculiarity of the PVC in cyclones is that
is further increased, it develops a fundamental the backflow zone, being confined by the vortex
instability and starts to rotate axially. At the same finder, adopts the spiraling shape and movement of
time, the zone close to the centerline of the flow also the PVC [4]. A common feature with once-through
becomes unstable downstream, loses symmetry and PVCs is the formation of coherent eddies in the
starts to precesse (locally) at a frequency depending axial-radial plane downstream of the backflow zone,
on geometry [64]. The backflow zone also oscillates that detach in an alternating pattern (synchronized
axially; it is finally shaped by the downstream with the PVC movement), and migrate following a
oscillation itself. In this way, a coherent, quasi- helicoidal trajectory to the exit section of the
periodic flow structure is formed, which we known cyclone, being quickly dissipated some diameters
as the ‘‘PVC’’. The term precession is somewhat downstream [4,82].
misleading; the whole structure resembles more a
spiraling or helical movement. 5.2.1. Influence of the PVC in the flow field
In once-trough swirling flows, all these features By studying tangential velocity profiles in a cross-
are also characteristic of what is called a vortex sectional plane at cyclone exit, the motion of the
breakdown of type 0. For an unconfined flow, the vortex core has been exhaustively described, rotat-
recirculation is caused by the downstream attenua- ing around the geometric center of the apparatus
tion of the swirl in the open volume; the dip in the [85]. Fig. 17 represents a sketch of the flow at this
profile of radial pressure accounts for the instability position. The precession movement displaces the
[31]. As we have seen, cyclone separators develop vortex core from the geometrical center of the
similar structures for similar reasons, with the device. This produces a ‘‘negative’’ tangential
backflow ‘‘bubble’’ located inside the vortex finder. velocity near the axis and a maximum of ‘‘positive’’
 ARTICLE IN PRESS
436 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

1
0.95

Cyclone pressure coeficient, ξ s

0.9
0.85
0.8
0.75
0.7
0.65
0.6 Measured data [45]
Muschelknautz [6]
0.55 Muschelknautz [6] , including vortex penetration in dipleg
0.5
0 50 100 150 200 250
C si , g solids / kg gas

Fig. 15. Comparison of experimental and predicted values of the pressure drop coefficient, as a function of inlet solids loading. Lines are
power fits to the data [45].

velocity near the wall; both can be explained by an

angular momentum balance through the plane
formed by the precession and the geometric centers.
The flow is ‘‘pressed’’ to the wall by the displace-
ment of the vortex core, forcing it to pass through a
restricted area, and causing an increase of the
tangential velocity [82].
The vortex nucleus describes a closely circular
path around the geometrical center of the device. A
stationary observer located at a point on the xx-line
detects tangential velocity fluctuations, which are
roughly sinusoidal. Thus, a sinusoidal wave can be
measured, such as that reported by [81] in a cold-
flow model of a swirl combustor. The statistical
distribution of the tangential velocity is obviously Fig. 16. Visualization of the precessing vortex core (PVC) at the
bimodal at the points situated within the regions of exit duct of a hydrocyclone [81].
change of the flow direction. This behavior has been
used to detect the existence of the PVC phenomenon now, but numerical simulations show it clearly
[86,87]. [4,88].
Some numerical simulations and experimental
campaigns have dwelt on the effect of the PVC on 5.2.2. Influence of operating conditions on PVC
velocity fluctuations [88]. The outcome is, briefly, The frequencies of oscillation of the PVC in
that a maximum of the RMS value of axial and cyclones do not differ much from those of the swirl
tangential velocity fluctuations is induced by the of the gas itself. Customarily, for a frequency f (Hz),
PVC in the central part of the cyclone. This a cyclone Strouhal number is defined as
influence roughly accounts for 80% of the total
fDc
fluctuation, the rest being due to ‘‘unstructured’’ St ¼ . (65)
turbulence, whose contribution is of the same vi
magnitude as in elementary flows such as boundary Several studies consistently report a value of St of
layers. about 0.5–0.6 for PVC and vortex end structures in
As for pressure patterns, the PVC appearance cyclones [66,88,89], almost independent of the
implies that an irregular and twisting zone of low Reynolds number. Other references agree in report-
pressures (the displaced inner vortex zone) must ing an approximately linear increase of frequency
be detected. No measurements confirm this up to with Rec [81,90]. A decrease has been found with
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 437

increased swirl number [82], which can be inter-

preted, in agreement with previous observations, as
a stabilization effect. Consistently, a slight decrease
of frequency along the length of the vortex has been
reported [66].
On the other hand, many efforts have been
devoted to control PVC by cyclone design as a
means to favor cyclone efficiency. Under a fixed
inlet Reynolds number, the behavior of the PVC at
the lower part of the cyclone is a mere question
of geometry. But things can be changed at the
other end [85]. When a 901-bend is attached to the
vortex finder, it has been found that the PVC is
distorted and damped. By adding a straight section
6.5 diameters long, the intensity of the PVC is
further decreased, reducing tangential velocity Fig. 17. Sketch of tangential velocity profiles as affected by the
and centering the vortex. Finally, if a 601-bend is phenomenon of the PVC. Adapted from [85]
added, the precession movement completely dis-
appears. Research has also focused at mitigating
PVC amplitude and frequency in the vortex finder that at least second-order closure, i.e., a Reynolds
by installing centering devices and guide vanes stresses transport model (RSTM) is needed to
[64,82]. capture anisotropy and achieve a realistic simula-
tion of cyclone flow [91,92]. Both algebraic and
6. Computational fluid dynamics applied to cyclones differential RSTMs have been employed. The
algebraic RSTMs made a great difference with
Up to this point, we have seen what traditional, respect to simpler models, but it became clear very
semi-empirical and algebraic, models of complex early that a differential RSTM was always prefer-
flow phenomena can offer to the practical task of able—in case the extra cost of the calculation was
calculating cyclones. Let us see what the modern affordable [93–95]. With the computer capabilities
computational artillery can bring in. of nowadays, even the difference between a basic
Flow in the inner space of a cyclone separator and an advanced differential RSTM can be of
possesses several characteristics that makes its relevance in this case [91].
numerical simulation difficult. Obviously, intense There are also purely numerical issues in the
swirl and shear and confined and unstable flow hydrodynamic simulation of cyclones. Specifically,
structures will not be easy to duplicate. But, to be the discretization of advection terms cannot be the
more concrete, the main issue here is rather the usual, first-order scheme, since this brings along
necessity of a model of turbulence that simulta- excessive numerical diffusion (for any kind of
neously accounts for discretization geometry of reasonable size) and
prevents an adequate simulation of flows possessing
1. High curvature of the average streamlines. the features mentioned above. Along with an
2. High swirl intensity and radial shear. adequate turbulence model, higher accuracy dis-
3. Adverse pressure gradients and recirculation cretization schemes, at least second order, are
zones. needed.
At a first sight, the hypothesis of an axisymmetric
These three features make turbulence inside swirling flow can be advantageous. It is, of course,
cyclones highly anisotropic, which means that most in terms of computational time and cost. However,
models based on first order turbulence closure, such a computation cannot duplicate the features
including two-equation models (i.e., the ke model derived from the entrance, inherently asymmetric,
and its variations), are very weak at capturing the and, most importantly, fully 3-D effects, such as
real flow. secondary eddies and flows, and the PVC. This is
As we will see, several attempts were made to especially important at the lower part of the cyclone
overcome this limitation. The outcome has been and in the vortex finder.
 ARTICLE IN PRESS
438 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

These phenomena arise because of a transient structures in external flows, but its computational
behavior: cyclone flow turns asymmetric mostly cost grows as the wall is approached, since the
because it turns unsteady. Accordingly, although ‘‘large’’ scales gets progressively smaller there. LES
much work has been done on steady CFD simula- for full-size equipment is presently possible, but still
tion of cyclones, one must face the fact that a costly. URANS methods are comparatively far less
realistic flow field can only be duplicated by solving expensive; in fact, most commercial CFD solvers
a time-dependent, transient problem. In other usually handle a transient, taking as convergence
words, it is almost a certitude that early CFD criterion the cancellation of time derivatives. Only
modelization of cyclones produced steady-state the extra memory and the size and number of time
solutions only because the limited spatial resolution steps needed to attain fully developed periodicities
available at the time, which resulted in artificial make the difference between solving RANS or
damping of instabilities. In modern times, it has URANS.
been recognized that the problem really demands an In an attempt to take advantage of the LES
adequately space- and time- resolved numerical concept at a lower cost, there are also methods that
calculation. can be qualified as a mixture of LES and RANS.
Computational cost grows even further; the fact This is somewhat in exchange for a minor loss of
that the unsteadiness is quasi-periodic is hardly of accuracy, although some RANS abilities, i.e., the
any help. To accomplish the task, there are however treatment of the boundary layer, can be very
several degrees of complexity: valuable in a LES context. Very large eddy
simulation (VLES) consists in an advanced URANS
1. Simulation of the unsteady or transient Rey- scheme for the SGS, allowing less space resolution
nolds-averaged Navier–Stokes equations for the resolved scales. Detached eddy simulation
(TRANS or URANS). (DES) is essentially a RANS solver that changes to
2. Large eddy simulation (LES). subgrid mode when the computed turbulent length
3. Direct numerical simulation (DNS). scale exceeds the grid side; or put in other way, it
is a LES that switches to RANS near the wall.
The first consists in simply retaining the transient The economy of all these methods is still under
term in the RANS equations, thus solving a time- test; on the other hand, and contrary to pure
dependent problem instead of a stationary one. URANS or pure LES, they are not standard, easily-
Turbulence closure can be accomplished by first- or implemented methods.
second-order methods; roughly speaking, the ob- As for the realism of the simulation, the ladder
servations above still hold, so that a RSTM is goes naturally the other way. There is a basic
preferable for swirling flows, although this can be problem with URANS: it assumes two distant
arguable (see later). In the second family of ranges of time scales, those pertaining to the
methods, filtered, although exact, conservation turbulent fluctuations and to the unsteady struc-
equations are solved, so that large scales (‘‘eddies’’) tures of the main flow. The question of exactly
are exactly duplicated. Filtered-out details are which flows meet such a condition is as diffuse as
represented by means of a standard turbulence the condition itself, so that it has always been a
model; this is called a subgrid scale model (SGS). source of confusion; for instance, literature on the
Most commonly, a isotropic eddy viscosity method simulation of combustion inside the cylinders of IC
suffices as a SGS. Finally, DNS is the numerical reciprocating engines is a good example. In the case
solution of all the spatial and temporal scales of the of cyclones, experimental observations and numer-
time-dependent Navier–Stokes equations, thus at- ical calculations readily showed that there is not a
tempting to duplicate all the detail of the turbulent clear-cut of frequencies in this kind of flow. There-
flow. fore, URANS methods lack a sound theoretical
This ladder of complexity obviously implies a basis [96], and the preferred method—DNS
gradation of realism and computational cost. Direct discarded—should be LES [97].
simulation of turbulence is presently feasible only at However, although the superiority of LES has
limited sizes or for model problems, and low been established against URANS, the specific
Reynolds numbers. Flows in industrial equipment models used in the later were limited to two-equa-
are, and will continue to be, out of the scope of tion models. The simulations were able in any case
DNS. LES is relatively affordable to duplicate great to show up periodicities and coherent structures
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 439

but, as can be expected, these were somewhat RNG ke model showed great improvement over
schematic, not completely duplicating intermediate this, but as expected, the Reynolds stresses model
details of the flow. In contrast, when powerful, exhibited the best behavior. Fig. 18 pictures the
differential RSTMs are brought into play, differ- results for tangential velocity. The same tendency
ences between LES and URANS diminish, and it was found when comparing the same models when
seems that the second is outperforming its very simulating an exit device called postcyclone [105].
theoretical basis [98]. Among all RSTMs available, it is still not clear
Presently, URANS with advanced RSTM ap- which is the most suitable for a specific application.
pears as an attractive option, offering realism at a For instance, another comparative study was made
much lower cost than LES [99]. This seems to be [106] including two differential RSTM implementa-
true not only for turbulent flows with large coherent tions, those known as LRR and SSG (see nomen-
structures [100], but also for more intricate cases clature), as well as the standard ke and a ke
such as flow in cyclone separators. This is, however, modified to account for the streamline curvature
still a point for research and discussion; for (ke cc). Although the pressure distribution showed
instance, there have been URANS simulations of acceptable agreement for all models, the flow field
swirling flows in which a k– e model was deemed was very sensitive, and the typical Rankine profile
more adequate than second-order closure at captur- could only be obtained by means of RSTMs. Fig. 19
ing the instabilities [101]. gives a clear illustration. Only small differences are
In any case, it should not be forgotten that the observed between LRR and SSG models, especially
fast growing capabilities of modern computers will at the lower part of the cyclone. At this position, the
surely favor complexity, which in this case is almost SSG model produces more acceptable results,
a pledge of LES or intermediate methods. whereas the LRR method underpredicts tangential
velocity near the cyclone center.
6.1. CFD studies on single-phase cyclone flow The high computational costs associated to
Reynolds stresses models justified great interest in
The first CFD simulation of cyclone flow was alternate methods that account for the turbulence
performed about 20 years ago [36–38], using the anisotropy without excessive computational effort.
finite-element method. It was the first to detect that Aside from the RNG ke and the ke cc, other
the standard ke turbulence model was not able to hybrid models based on two-equation ideas, such as
accurately simulate this kind of flows. Using a that of [107] have been used to obtain more
mixed algebraic-differential, stationary RSTM, the acceptable results
authors found instead good agreement between In later years, the need of using unsteady solvers
experimental data and simulation of overall pres- and second-order discretization was established;
sure drop (both in absolute value and tendencies), some key papers on this issue are Refs.
and they could correctly duplicate the radial [72,108,109]. For instance, in [72] a URANS scheme
velocity profile. At the same time, advanced ke was used to detect that the PVC structure develops
schemes, that took partial account of the turbulence thorough the cyclone length, is destroyed at a
anisotropy, were employed for simulating hydro- certain height, and finally stabilizes itself in the form
cyclone flow [102–104], arriving at acceptable of a quasi-periodic phenomenon. The study re-
results. ported in [109] used a stationary RSTM simulation
In one of the first comparative studies of several as an initial condition to trigger a LES, to finally
turbulence models, the relative performance of the capture the PVC phenomenon with great realism. It
ke model, the differential RSTM known as the was also established that the flow was dominated by
LRRG model [91], and a variation of the ke model the greater scales, since the influence of the specific
based on ReNormalization Group (RNG) theory SGS used in the LES solver was minor.
was evaluated [28]. Simulations were also compared To provide some illustrations to the simulation of
with velocity measurements carried out by means of PVC phenomena, lets quote recent research on a
LDA. Tests were performed with three different pilot PFBC cyclone, equipped with a long dipleg [4].
vortex finder diameters, which produced three Single-phase gas flow was simulated by three
different swirl numbers. For all runs, the ke model methods: ke, RNG ke and the SSG implementa-
predicted solid-body rotation, a logical result given tion of a RSTM. Predictions were compared with
its foundations, albeit contrary to experiment. The hot-film anemometry and pressure measurements
 ARTICLE IN PRESS
440 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Fig. 18. Comparison of tangential velocity profiles [28].

Fig. 19. Comparison of tangential velocity profiles [106].

 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 441

taken at the dipleg. The study discovered that, helicoidal movement than just a simple wavering of
under normal operating conditions, the vortex the double vortex structure.
enters well within the dipleg, so that the flow The latest CFD studies [25,110] continue to show
possess their many characteristics of the flow in a the capacity of CFD models for predicting interest-
cyclone body [16,17]. In order to simulate it, second- ing aspects of centrifugal separators. Further work
order discretization and a differential RSTM were using RSTMs is reported in [26,65]. LES simulation
needed. When the latter was run, however, it did not of swirling and cyclone flows is presently becoming
converge to a stationary solution. With an adequate a new standard, based as already noted on increased
time step and retaining an adequate number of time computer capabilities. For example, Derksen and
iterations, quasi-periodicities were discovered and Akker at Delft University [88] were among the first
the simulation transformed itself in a URANS to simulate the PVC phenomenon by this method,
implementation. but only attempting to model small scale cyclones at
Results showed a reasonable agreement in pres- a moderate inlet Reynolds number of 14 000.
sure drop and tangential velocity up to the lower Subsequent investigations with the same numerical
part of the dipleg. Velocity data suffered however method have been able to cope with industrial
from the fact that experimental measurements were equipment at Rec ¼ 280 000 [32]. The capabilities
taken as pure averages, i.e., without acknowledg- and sensibilities of the particular technique (LES
ing the fact that large oscillations will show up with standard SGS and a lattice-Boltzmann dis-
at fixed positions as a consequence of the PVC. cretization) in simulating confined turbulent vortex
Fig. 20 shows the simulated, phase-averaged, flows have been further reported in [29].
tangential velocity at the cylindrical part of the It is not easy to forecast the future of this line of
cyclone (it does not vary much axially, even at the research. On the one hand, LES methods seem to
conical part), compared with several algebraic offer a completely realistic simulation barring some
models. The scheme resulting from the combi- minor details. However, the relative competence of
nation Meissner–Reydon–Ogawa shows the best URANS with an advanced RSTM has not been
agreement. established yet, and it should be, given the very
Special phenomena associated with cyclone flow relevant cost factor involved. On the other hand,
were also detected for the PFBC cyclone dipleg, only relatively stable and ordered flows have been
both experimentally and numerically. The position simulated. LES and URANS methods must be
of the vortex end was detected by pressure tested for conditions of a highly incoherent and
measurements (see also Ref. [17]) and by the variable PVC, to ascertain their usefulness under
numerical simulation. The adopted URANS scheme such conditions. Obviously, this might demand also
seemed to capture the PVC in a similar fashion to empirical input. In any case, a full perspective must
other unsteady calculations [25,72,109,110], some of take into account two-phase flow; we address the
them being also URANS models but including LES question in Section 6.2 below.
as well. Fig. 21 shows the evolution of the vortex
core in a cross-sectional plane of the dipleg. It 6.1.1. Non-intrusive velocity measurements in
describes a nearly circular path around the geome- cyclone separators
trical center, rotating in the same direction as the Finally, it is worth noting that modern CFD
flow. To complete the picture, Fig. 22 shows the 3-D studies have relied largely on advanced, non-
surface of zero axial velocity inside the cyclone and intrusive measurements of velocity patterns inside
dipleg, which separates the upward and downward the cyclone, as a source of detailed validation data.
vortex flows. The surface is colored in proportion to Although we cannot intend here a full account of
the velocity modulus. In very apparent agreement the question, a brief review is in order. Early
with the stroboscopic experiments of Hoffmann measurements of velocity profiles inside cyclones
[66], the vortex end, identified as a 90o twist of the were undertaken by means of Pitot tubes—see for
structure, just separating the zones of high and low instance [42], which quickly began to compete with
velocity, is seen to adhere to the wall of the dipleg hot-wire (or hot-film) sensors. Both kinds of
and rotate on it. Observations of a ring of dust in instruments have been in use until recent times,
the same experimental rig, coherent with pressure the second having the advantages of a smaller size
measurements, confirm this [4,17]. On the other and the possibility of very good time resolution. In
hand, the ‘‘precession’’ is seen in this way more as a both, however, determination of the direction of
 ARTICLE IN PRESS
442 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Fig. 20. Comparison between predictions of tangential velocity by algebraic models and CFD computation [4].

velocity can be an issue and there is of course the limited [28] or even no measurement is possible at
essential problem of an intrusive measurement. all [25].
The sensor head and its support may indeed Recent developments in PIV have qualified this
disturb the flow and introduce a lingering and technique as a competent alternative to LDA, with
mostly unknown uncertainty in the magnitude and the added capability of a quantitative appreciation
profile of the measured velocity. This fact has been of global aspects of the flow pattern, instead of
long recognized by cyclone experimenters, as single-point velocity values. (The problem of tracer
attested by reports of induced asymmetry [73], segregation is however pretty much the same.)
attenuation of the swirl [25] and inconsistent read- Actually, the reports of its use in swirling flows
ings near the centerline [4]. In the context of a date also from the end of the 1990s. In particular,
highly unstable and easily disturbed flow, the the group at Cardiff University has been using both
necessity of non-intrusive measurements soon be- LDA and PIV in cyclonic combustors and separa-
came evident. tors to measure velocity and detect coherent
As we have already mentioned, these have been structures [64,67,84]. Their method for resolving
accomplished mostly by laser Doppler anemometry, the PVC consists in phase-locking the PIV measur-
starting with the classical investigations at Delft [28] ing system with the aid of fast pressure sensors.
and continuing with many other works, most More recent references related to this technique in
notably [25,29,106]. LDA techniques are used in cyclones are [35,111], having been the first studies to
cyclones with apparent success, not only for the introduce stereoscopic PIV in this context.
determination of time-averaged profiles, but also A perennial field in cyclone experimentation is by
resolving turbulent quantities and detecting periodic far and large flow visualization. This has been
fluctuations (PVC). The only drawback is that the accomplished by diverse methods, ranging from the
tracer particles may be segregated inside the simple observation of the separated particle streaks
volume, leaving a void in the immediate vicinity of on Perspex model walls to the use of streamers and
the centerline where the time resolution can be tracers. A couple of recent and representative
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 443

Fig. 21. Unsteady field of axial velocity obtained by a URANS simulation of cyclone flow [4], signaling the existence of a PVC.

examples are the already-mentioned works on PVC results concerning cut sizes and grade-efficiency, are
characterization by stroboscopic lighting [66] and relatively scarce in the literature about cyclones. Let
by neutrally-buoyant tracers combined with pres- us give a brief account of the state-of-the-art.
sure measurements of high time-resolution [34]. Dispersed two-phase flows can be classified
according to the importance of the interaction
6.2. Numerical computation of two-phase flow in mechanisms [112,113]. Fig. 23 shows an elementary
cyclones and practical sketch. Generally, two different
regimes are distinguished, depending on the exis-
Overwhelmed by the complexity of the flow of gas tence of mutual, significant interaction between
in cyclones, we have perhaps forgotten that it is the particles: dilute and dense two-phase flow. The
combined flow of gas and particles what really approximate borderline is a volume fraction
matters in the operation of such an apparatus. apE10–3, which translates generically as an inter-
Compared to single-phase, gas flow simulations, particle spacing L/dpE8. Within the dilute regime,
simulations of two-phase flows, and its immediate all the influence of particles on the gas can be
 ARTICLE IN PRESS
444 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

Fig. 22. Instantaneous locus of the surface of zero axial velocity inside the ensemble of a cyclone+dipleg, colored according to the
modulus of velocity [4].

neglected for apo10–6 (L/dp480), which is known collisions can lead to coalescence and break-up,
as ‘‘one-way coupling’’; for higher volume fractions, which must be considered too. This regime is called
it needs to be accounted for (‘‘two way coupling’’). frequently ‘‘four-way coupling’’.
Aside from obvious issues of volume and continu- Within this framework, numerical models have
ity, the generic flow feature most affected is developed in two parallel paths according to the
turbulence. In our context, the effect is normally manner in which the dispersed phase is treated,
an attenuation, but an enhancement may also result either by a Lagrangian or by an equivalent Eulerian
under some circumstances. In the dense regime formulation. In the Lagrangian model, particle
(ap410–3, L/dpo8), interparticle interactions be- trajectories are obtained by integrating the particle
come of importance, both physical collisions and equation of motion. In the Eulerian approach,
indirect influence through the nearby flow field. The conservation equations are written for the mass,
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 445

Interparticle spacing L/dp Eq. (51) neglects however a basic fact: only
100 10 1 particles of large size will behave as exclusively
influenced by the time-averaged gas flow. Very small
particles will tend to fluctuate following turbulent
Dilute-dispersed Dense-dispersed fluctuations of the gas velocity, and there will be a
two-phase flow two-phase flow
complete range of intermediate behaviors between
these two extremes. This phenomenon is known as
turbulent diffusion. There are several approaches to
Two-way Four-way
One-way
coupling coupling simulate it by including additional terms in a time-
coupling
averaged equation such as Eq. (51). The most
popular nowadays amount to assume a prefixed
probability distribution of velocity, solve the
equation of motion for discrete velocities (and
10-8 10-6 10-4 0.01 0.1 particle sizes) and make an average of the forces.
Volume fraction αp These are referred to as ‘‘stochastic’’ models of
particle turbulent diffusion [117–119].
Fig. 23. Regimes of dispersed two-phase flow as a function of the Lagrangian calculations following these princi-
particle volume fraction/interparticle spacing. Adapted from ples have been profuse in cyclone literature
[113].
[26,28,36,120–122]. Computational simulations fre-
quently obtained larger cut-sizes than measured for
momentum and energy of the particle cloud as a different flow conditions, which is clearly a signal of
continuum, and integrated to predict volume- the complexity of the flow and the approximate
averaged properties thorough the field. nature of the models. To the intricacy of the single-
The suitable choice is often problem-specific. phase gaseous flow, it is indeed very easy to add-up
Generally speaking, Lagrangian methods are more difficulties arising from the dispersed phase: For
suited to dilute flows, free of numerical diffusion, instance, with independence of inlet solids loading,
less influenced by other errors and more stable for the accumulation of particles in specific regions
large particle velocity gradients. Also, the treatment (i.e., cyclone walls as a result of the collection
of realistic poly-dispersed particle systems is process) leads to high local values of particle
straightforward. However, for highly loaded concentration, which violate the basic assumption
flows, limitations related to computer storage, of a simple Lagrangian tracking, and are not easily
calculation times and convergence arise [114]. In modeled, even considering their effect as a bound-
those cases, the Eulerian approach becomes more ary condition.
adequate. It seems also the right framework for In recent years, as we have seen, LES is
modeling particles’ mutual interaction [115], increasingly being applied to simulate time-resolved
although Lagrangian methods have been also gas flow in cyclones. Of course, the resulting
applied [113,116]. velocity field has also been used to calculate particle
flow by means of a Lagrangian tracking [32,123]. As
6.2.1. Eulerian– Lagrangian models of cyclone an example of such a modern calculation, Fig. 24
operation. shows snapshots of particle concentration at five
For dilute, one-way coupled flows, a Lagrangian instants of time inside a high-efficiency, lightly-
tracking of particles is adequate, which consists in loaded cyclone [32]. The leftmost drawing repre-
integrating the equation of motion of a particle sents the time at which the feeding of particles
given by the general form of Eq. (51). In the context is stopped; subsequent evolution is shown at
of cyclone CFD, gas velocities are not longer the equally-spaced intervals to the right. As expected,
result of an algebraic model but the computed 3-D particles tend to accumulate at the bottom bin (note
velocity field; the calculation is equivalent to a very also the spiral streaks so familiar in Perspex
refined time-of-flight model. A Lagrangian tracking experiments), but there is also a persistent, high
is normally decoupled from the gas flow calculation; local concentration at the roof of the cyclone. The
it can be undertaken as a post-process computation fate of these particles is still undecided, which
by a CFD solver, i.e., as a calculation using previous means that the collection process can be quite a
gas velocity results from a clean, gas-only run. lengthy one.
 ARTICLE IN PRESS
446 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

senting large groups with the same properties) are

tracked through the flow field, typically a minimum
of 10 000–20 000. Source terms are thus obtained for
a second Eulerian calculation of the gas flow. The
procedure is repeated iteratively until convergence;
to avoid strong fluctuations, under-relaxation is
frequently used [125]. Special models of turbulence
accounting for the effect of particles are needed, at
least for steady RANS—see a good review in [126].
The main drawback of this type of methods is
obviously the high computational effort, if large
numbers of particles are needed to represent the
dispersed phase.
This seems to be the case of cyclone separators; in
contrast with simple Lagrangian trackings, the use
of PSI-Cells methods is still at its very infancy.
Actually, we can only signal a relevant reference at
the time of writing (December 2006): [127], which is
a logical continuation of the work of the same
authors we have just reviewed.
Curiously enough, the intermediate step of a
stationary two-phase calculation seems to be absent,
and the work deals directly with LES-resolved,
time-dependent flow. In any case, the contribution
Fig. 24. Snapshots of particle concentration (scale not given) at is worth a good look. In it, the previously-reported
five instants of time inside a Stairmand cyclone separator, as LES+Lagrangian tracking [32] is modified to
calculated by LES+Lagrangian tracking [32]. assemble a PSI-Cell calculation, attempting to
simulate mass-loading effects for moderate, but still
high, values of Csi ¼ 0.05–0.1, that fall into the
This fact leads to an excessively costly method to regime of dilute flow but involve forcibly two-way
predict cyclone performance; in this respect, it coupling. To this end, a deemed-modest number of
should be noted that the enormous amount of ten million parcels is tracked, assumed representa-
time-dependent data generated by LES precludes tive of a much larger number of real particles. The
post-processing and requires at once repeated simulation uses a standard SGS, with no proviso for
tracking of a large amount of identical particles. the particles effect on unresolved scales, which is
(The method is neither free of the necessity of a considered, also provisionally, adequate.
stochastic model of particle turbulent diffusion due One outcome of the investigation is that such a
to the unresolved part of the velocity.) Accordingly, computational load is indeed enormous and con-
several alternatives are proposed, based on average, verged results for gas and particle fields (in the
frozen and periodic LES-velocity fields. Although quasi-periodic sense) cannot be reported yet, so that
no definitive conclusions are drawn, the periodic conclusions are only tentative. However, while we
approximation seems to be the best, but also the wait for the end of the calculation, the main thesis
more expensive, logically. In any case, calculated of the paper is very interesting.
efficiencies compare much better with experiment This is that the dependence of separation
than former, classical Lagrangian calculations [32]. efficiency on inlet solid loading is not really caused
For interacting flows, the two-way transfer of by a singular separation process as postulated
momentum between particulate and fluid phases is classically by Muschelknautz, but the result of a
modeled by particle-source-in cell (PSI-Cell) meth- competition between two effects of a high particle
ods, originally developed in [124]. The flow field is concentration: attenuation of swirl, which lowers
calculated first without particle-phase source terms efficiency due to a lowered centrifugal force, and
until a converged solution is achieved. Then, a large attenuation of turbulence, which augments effi-
number of ‘‘parcels’’ (i.e. discrete particles repre- ciency through a decreased turbulent diffusion of
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 447

particles. The second effect having received very medium cyclone separators [131–135], with promis-
little attention in the traditional cyclone literature, ing, but not completely satisfactory, results. Aside
this is of course quite interesting. If definitive results for that, as for gas cyclones, these authors only
support the hypothesis (as the provisional ones seem know of two investigations. In [136], a mixed k– e/
to do), Muschelknautz’s views will become, at least algebraic RSTM, stationary gas flow model is used
for moderate solids loadings, another very useful, along with a simple Eulerian model of a diluted,
but purely imaginary, theory on cyclone operation. monodispersed particulate phase. The results,
clearly exploratory, predict several effects reason-
6.2.2. Eulerian– Eulerian models of cyclone ably, most notably the reduction of swirl leading to
operation a reduction of pressure drop.
The Eulerian–Eulerian approach treats both the The very recent work reported in [137] deals with
solid particles and the fluid as interpenetrating a simulation of two-phase cyclone flow in the dense
continua, each governed by conservation laws regime, with solids loadings up to 2 kg/m3 N. The
whose general structure is known albeit specific Fluent’s implementation of the multiphase algebraic
terms that must be modeled. These are those slip scheme is used (details on the gas flow model are
representing the interaction between phases, i.e., not given), with no account of interparticle effects.
the result of the forces the gas imposes on the Results on collection efficiency, as compared with
particles (average draft and turbulent diffusion, dedicated experiments, are not conclusive or even
acting on a continuum) on the one side, and the deviant. However, interestingly, the attenuation of
effect in the gas of the presence of solid particles on both swirl and turbulence with mass loading is
the other [126]. A generic inconvenient of Eulerian clearly noticed.
models is that they require multiple levels of
modeling (e.g., boundary conditions for the dis- 7. Concluding remarks and perspectives
persed phase, mutual interaction between particles,
interactions with the wall). On the other hand, Cyclones are robust devices, widely used in the
complex phenomena such as agglomeration and chemical and process industries, as well as for
break-up can be taken into account by using a heavy-duty hot gas-cleaning service in several
population balance model [128]. applications of combustion of solid particles. Much
Eulerian–Eulerian schemes have been used in ingenuity has been devoted to the prediction of their
many practical multiphase flow simulations, for performance parameters, collection efficiency and
instance, and typically, in fluidized-bed applications pressure drop, which has required a previous
[129], where they are specially suited. However, knowledge of their basic flow patterns.
Eulerian models cannot treat different particle sizes The classical view of a steady flow, that can be
directly, but considering every size class as a distinct calculated as a clean gas, single-phase flow corrected
phase, which is evidently a serious drawback for by mass loading effects has been relatively success-
practical problems. As a consequence, simplified ful. Many algebraic models and semi-empirical
Eulerian multiphase approaches have been devel- formulas have been developed throughout the years;
oped, mostly based on an algebraic velocity slip among them, the more complex models of Muschel-
between phases. This is adequate under conditions knautz [6] for pressure drop, and Trefz and
of a strong coupling, with particle relaxation times Muschelknautz [7] for collection efficiency give
much shorter that the characteristic time-scale of most accurate results as compared to experimental
the flow, which is frequently reasonable. At once, it data.
represents a significant computational saving, al- However, the complexity of the flow in cyclones is
lowing the sole integration of the mixture equations, due to instability, as modern experimental and
with the only additions of the volume fractions and numerical techniques have demonstrated. Specifi-
slip relationships [130]. cally, the double-vortex structure that makes up the
This kind of approaches is referred to as Mixture basic flow is essentially unstable, and develops a
Models in some commercial CFD codes. Although phenomenon of quasi-periodic oscillations known
specifically signaled as apt for modeling cyclone as a ‘‘precessing vortex core’’. This instability can be
separators, reports in the literature are very scarce. a threat to the cyclone objective of solids’ separa-
Simplified Eulerian–Eulerian models have been used tion, but also affects a basic flow feature, formerly
recently in the related field of hydro- and dense- thought of as stationary, called the cyclone natural
 ARTICLE IN PRESS
448 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

length. This is in turn a major influence in cyclone cyclones, the experience to date is so meager that we
velocity patterns, pressure drop and collection can only hope for simplified, algebraic-slip models
efficiency. that can compete in simulating dilute flows and then
As a result, it can be said that the flow in cyclones can be applied to the dense regime.
is not yet completely understood. Given the kind of In any case, the incipient experience with two-
phenomena involved, classic, algebraic models are phase flow simulations seems to be in the verge of
positively not well suited to a proper account; another revolution, in the sense that classical, well-
modern progress should evidently resort to CFD established theories on cyclone particle flow might
calculations. However, the very nature of the be compromised. This would be of course stimulat-
problem makes the task difficult. Unusual high ing, and surely a great step forward in our under-
precision in the numerical discretization and un- standing of cyclone operation.
steady simulation methods (LES or URANS) are
needed to possibly capture the whole complexity of
Acknowledgements
the unsteady flow of gas inside a cyclone. The
simplest of these methods is computationally very
Authors are grateful to Dr. José Velilla, now on
costly, much more than the usual steady-state
General Electric (USA), who spent several years at
simulation of industrial equipment. Perhaps this
CIRCE facilities (experimental and numerical)
has prevented up to now a closed account of several
untangling cyclone flow: Many thanks. A part of
related design aspects, such as a precise methodol-
our research on cyclones was financed by the
ogy to reasonably estimate the cyclone natural
Escatrón pressurized fluidized bed combustion
length, and thus the advisable geometric length, the
power station (Contract nos. 40.702/94, 40.815/95,
circumstances for and the effects of a loss of
40.590/96, 41.090/97, 41.114/97 and 40.700/98),
coherence of the vortex and the ensuing chaotic
then owned by Endesa Generación SA (Spain).
flow patterns, the effect of swirl-stabilization
The authors also wish to thank Prof. Alex C.
devices, y . More systematic research is obviously
Hofmann for his useful comments and suggestions.
needed in this direction.
On the other hand, more work should be devoted
to ascertain if, and under what circumstances, a
References
URANS solver with advanced turbulence models
(differential RSTMs) can supplant a more rigorous, [1] Álvarez M, Anthony EJ, editors. Pressurized fluidized bed
but more costly, LES or its variants. combustion. Blackie Academic & Professional; 1995.
Finally, two-phase flow simulation of cyclones is [2] Robertson A, Fan Z, Goldstein H, Horazak D, Newby R,
certainly at its very beginning and demands several Bose AC. Second generation PFB plant with super critical
pressure. In: Proceedings of International Pittsburgh Coal
advances. On the one hand, factors like interparticle
Conference, 2005.
phenomena and conditions at the wall have not [3] Venalainen I, Psik R. 460 MWe supercritical CFB boiler
received yet the attention they surely deserve. But design for Lagisza power plant. In: Proceedings of Power-
most importantly, modern two-phase flow simula- Gen Europe, 2004.
tions of cyclones have turned out to be too costly, [4] Velilla J. Study of the flow at a PFBC cyclone dipleg. PhD
due to the necessity of reproducing unsteadiness of thesis, University of Zaragoza, 2005.
[5] Gil A. Experimental cold flow model of a PFBC primary
the gas flow and combine it with the simulation of a cyclone. PhD thesis, University of Zaragoza, 2000.
poly-disperse particulate system. Even for dilute [6] Muschelknautz E. Die Berechnung von Zyklonabscheidern
flow, even for one-way coupling, present computer für Gase. Chem-Ing-Tech 1972;44:63–71.
capabilities are insufficient, at best making the CFD [7] Trefz M, Muschelknautz E. Extended cyclone theory for
calculation a costly expedient. gas flows with high solids concentrations. Chem Eng
Technol 1993;16:153–60.
Simplified schemes are obviously needed. To the [8] Dirgo J, Leith D. Design of cyclone separators. In:
possibilities explored in the literature of using Cheremisinoff NP, editor. Encyclopedia of fluid mechanics,
‘‘abridged’’ LES data to this end, we may add the vol. 4—gas–solid flows. Houston: Gulf Publishing Com-
necessity of directly testing URANS methods, not pany; 1986.
[9] Reznik VA, Matsnev VV. Comparing the characteristics of
yet realized. A development of the PSI-Cell techni-
the elements in batteries of cyclones. Therm Eng 1971;18:
ques is the sense of coping more efficiently with 34–9.
time-variable problems seems also, obviously, very [10] Hoffmann AC, Stein LE. Gas cyclones and swirl tubes.
convenient. As for the calculation of highly-loaded Principles, design and operation. Berlin: Springer; 2002.
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 449

[11] Alexander RMCK. Fundamentals of cyclone design and swirl tube, with emphasis on the flow pattern and
operation. Proc Australas Inst Min Metall 1949;152: separation efficiency. Energy Fuel 2006;20:1691–7.
203–28. [34] Peng W, Boot PJAJ, Hoffmann AC, Dries HWA, Kater J,
[12] Ter Linden AJ. Investigations into cyclone dust collectors. Ekker A. Flow in the inlet region in tangential inlet
Proc Inst Mech Eng 1949;160:233–51. cyclones. Ind Eng Chem Res 2001;40:5649–55.
[13] Barth W. Berechnung und Auslegung von Zyklonabschei- [35] Liu Z, Zheng Y, Jia L, Jiao J, Zhang Q. Stereoscopic PIV
dern auf Grund neuerer Untersuchungen. Brennst- studies on the swirling flow structure in a gas cyclone.
Waerme-Kraft 1956;8:1–9. Chem Eng Sci 2006;61:4252–61.
[14] Mothes H, Löffler F. Motion and deposition of particles in [36] Boysan F, Ayers WH, Swithenbank J. A fundamental
cyclones. Ger Chem Eng 1985;8:223–33. mathematical modeling approach to cyclone design. Trans
[15] Meissner P, Löffler F. Zur Berechnung des Strömungs- Inst Chem Eng 1982;60:222–30.
feldes im Zyklonabscheider. Chem-Ing-Tech 1978;50:471. [37] Boysan F, Ewan BCR, Swithenbank J, Ayers WH.
[16] Gil A, Romeo LM, Cortés C. Cold flow model of a PFBC Experimental and theoretical studies of cyclone separator
cyclone. Powder Technol 2001;117:207–20. aerodynamics. Inst Chem Eng Symp Ser 1983;69:305–19.
[17] Gil A, Cortés C, Romeo LM, Velilla J. Gas-particle flow [38] Boysan F, Swithenbank J, Ayers WH. Mathematical
inside cyclone diplegs with pneumatic extraction. Powder modelling of gas-particle flows in cyclone separators. In:
Technol 2002;128:78–91. Encyclopedia of fluid mech, vol. 4. Gulf Publishing Co.;
[18] Wang J, Bouma JH, Dries H. An experimental study of 1986 [chapter 42].
cyclone dipleg flow in fluidized catalytic cracking. Powder [39] Yuu S, Jotake T, Tomita Y, Yoshida K. The reduction of
Technol 2000;112:221–8. pressure drop due to dust loading in a conventional
[19] Reydon RF, Gauvin WH. Theoretical and experimental cyclone. Chem Eng Sci 1978;33:1573–80.
studies in combined vortex flow. Can J Chem Eng [40] Parida A, Chand P. Turbulent swirl with gas–solid flow in a
1981;59:14–23. cyclone. Chem Eng Sci 1980;35:949–54.
[20] Ogawa A. Separation of particles from air and gases, I and [41] Morweiser M, Bohnet M. Influence of temperature and
II. Florida: CRC Press; 1984. pressure on separation efficiency and pressure drop of
[21] Ogawa A. Mechanical separation process and low aerocyclones. In: Schmidt E, editor. High temperature gas
patterns of cyclone dust collectors. Appl Mech Rev 1997; cleaning. Institut für Mechanische Verfahrenstechnik und
50:97–129. Mechanik; 1996. p. 26–39.
[22] Comolet R, Bonnin J. Stabilité d’un vortex. In: Mécanique [42] Shepherd CB, Lapple CE. Flow pattern and pressure drop
Experimentale des Fluides. Paris: Masson; 1986. p. 417–23. in cyclone dust collectors. Ind Eng Chem 1939;31:972–84.
[23] Kirch R. Der Einfluss der Turbulenz auf die Partikelbewe- [43] Muschelknautz E, Kambrock W. Aerodynamische Bei-
gung im Gaszyclon, in VDI-Fortschritts-Berichte Reihe 7: werte des Zyclonabscheiders aufgrund neuer und verbes-
Strömungstechnik, vol. 145. Düsseldorf: VDI-Verlag; serter Messungen. Chem-Ing-Tech 1970;42:247–55.
1988. [44] Casal J, Martı́nez-Benet JM. Valencia, Cálculo y diseño de
[24] Hoffmann AC, Van Santen A, Allen RWK, Clift R. Effect ciclones. Ing Quim 1989;2:115–24.
of geometry and solid loading on the performance of gas [45] Gil A, Romeo LM, Cortés C. Effect of the solid loading on
cyclones. Powder Technol 1992;70:83–91. a PFBC cyclone with pneumatic extraction of solids. Chem
[25] Peng W, Hoffmann AC, Boot PJAJ, Udding A, Dries Eng Technol 2002;25:407–15.
HWA, Ekker A, et al. Flow pattern in reverse-flow [46] Briggs LW. Effect of dust concentration on cyclone
centrifugal separators. Powder Technol 2002;127:212–22. performance. Trans Am Inst Chem Eng 1946;42:511–26.
[26] Wang B, Xu DL, Chu KW, Yu AB. Numerical study of [47] Smolik J. Air pollution abatement, part I, Scriptum No.
gas–solid flow in a cyclone separator. Appl Math Model 401-2099, Technical University of Prague, 1975.
2006;30:1326–42. [48] Baskakov AP, Dolgov VN, Goldovin YuM. Aerodynamics
[27] Hoffmann AC, de Groot M, Hospers A. The effect of the and heat transfer in cyclones with particle-laden gas flow.
dust collection system on the flowpattern and separation Exp Therm Fluid Sci 1990;3:597–602.
efficiency of a gas cyclone. Can J Chem Eng 1996;74: [49] Chen J, Luo X, Shi M. Experimental research on the
464–70. pressure drops of PV-type cyclone separators at various
[28] Hoekstra AJ, Derksen JJ, Van Den Akker HEA. An cyclone concentrations. In: Proceedings of the Int. Sympo-
experimental and numerical study of turbulent swirling sium on Multiphase Flow, 1997. p. 332–4.
flow in gas cyclones. Chem Eng Sci 1999;54:2055–65. [50] Lewnard JJ, et al. Effect of design and operating
[29] Derksen JJ. Simulations of confined turbulent vortex flow. parameters on cyclone performance for circulating fluidized
Comp Fluids 2005;34:301–18. bed risers. In: Avidan AA, editor. Circulating fluidized bed
[30] Syred N. A review of oscillation mechanisms and the role of technology IV. New York: AIChE; 1994. p. 525–31.
the precessing vortex core (PVC) in swirl combustion [51] Ushiki K, et al. Experimental study on the performance of
systems. Prog Energy Combust Sci 2006;32:93–161. very small cyclones for the recycle of ultra-fine powders. In:
[31] Lucca-Negro O, O’Doherty T. Vortex breakdown: a Avidan AA, editor. Circulating fluidized bed technology
review. Prog Energy Combust Sci 2001;27:431–81. IV. New York: AIChE; 1994. p. 504–10.
[32] Derksen JJ. Separation performance predictions of a [52] Cankurt NT, Yerushalmi J. Some observations on the
stairmand high-efficiency cyclone. AIChE J 2003;49: performance of cyclones in a fast-fluidized bed system. J
1359–71. Powder Bulk Solids Technol 1978;3:29–35.
[33] Hoffmann AC, Peng W, Dries H, Regelink M, Foo K-K. [53] Lapple CE. Gravity and centrifugal separation. Ind Hyg
Effect of pressure recovery vanes on the performance of a Quart 1950;11:40–7.
 ARTICLE IN PRESS
450 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

[54] Leith D, Licht W. The collection efficiency of cyclone type [75] Hoffmann AC, Groot M, Peng W, Dries HWA, Kater J.
particle collectors—a new theoretical approach. AIChE Advantages and risks in increasing cyclone separator
Symp Ser 1972;68:196–206. length. AIChE J 2001;47:2452–60.
[55] Clift R, Ghadiri M, Hoffmann AC. A critique of two [76] Zhongli J, Xiaolin W, Minxian S. Experimental research on
models for cyclone performance. AIChE J 1991;37:285–9. the natural turning length in cyclones. In: Proceedings
[56] Dietz PW. Collection efficiency of cyclone separators. of Filtech Europa ’91, Karlsruhe, The Netherlands, 1991.
AIChE J 1981;27:888–92. p. 583–9.
[57] Lorentz T. HeiXgasentstaubung mit Zyklonen. Dusseldorf: [77] Zenz FA. Cyclone separators. In: Manual on disposal
VDI-Verlag; 1994. of refinery wastes, volume on atmospheric emissions.
[58] Reinhardt B, Cordonnier A, Florent P. Use of a isokinetic American Petroleum Institute Publication 961, 1975
sampling probe. Results in a cyclone. Powder Technol [chapter 11].
1999;101:81–90. [78] Bryant HS, Sylverman RW, Zenz FA. How dust in gas is
[59] Abrahamson J, Martin CG, Wong KK. The physical affects cyclone pressure drop. Hydro Proc 1983;62:87–90.
mechanisms of dust collection in a cyclone. Trans Inst [79] Syred N, Beér JM. Combustion in swirling flows: a review.
Chem Eng 1978;56:168–77. Combust Flame 1974;23:143–201.
[60] Abrahamson J, Allen RWK. The efficiency of conventional [80] Gupta AK, Lilley DG, Syred N. Swirl flows. Tunbridge
return-flow cyclones at high temperatures. AIChE Symp Wells, UK: Abacus Press; 1984.
Ser 1986;99:31–43. [81] Styles AC, Syred N, Najim SA. A study of modulatable
[61] Wheeldon JM, Burnard GK. Performance of cyclones in cyclone combustors using gaseous fuel. J Inst Energy
the off-gas path of a pressurized fluidized bed combustor. 1979;159:160–8.
Proc Filtration Soc Filtr Separat 1987;3:178–87. [82] Griffiths AJ, Yazdabadi PA, Syred N. Alternate eddy
[62] Stairmand CJ. The design and performance of cyclone shedding set up by the non-axisymmetric recirculation zone
separators. Trans Inst Chem Eng 1951;29:357–83. at the exhaust of a cyclone dust separator. J Fluids Eng
[63] Lidén G, Gudmundsson A. Semi-empirical modeling to 1998;120:193–9.
generalize the dependence of cyclone collection efficiency [83] Williams JM. Investigations into the precessing vortex core
on operating conditions and cyclone design. J Aerosol Sci phenomenon within dust sperating ciclones. Internal report
1997;28:853–74. no. 1772. Cardiff School of Engineering, Cardiff University
[64] O’Doherty T, Griffiths AJ, Syred N, Bowen P, Fick W. of Wales, 1993.
Experimental analysis of rotating instabilities in swirling [84] Fick W, Griffiths AJ, O’Doherty T. Visualization of the
and cyclonic flows. Dev Chem Eng Mineral Process precession vortex core in an unconfined swirling flow. Opt
1999;7:245–67. Diagn Eng 1997;2:19–31.
[65] Qian F, Zhang J, Zhang M. Effects of the prolonged [85] Yadzabadi PA, Griffiths AJ, Syred N. Axial and tangential
vertical tube on the separation performance of a cyclone. J velocity components at the exhaust end of highly complex
Hazard Mater 2006;B136:822–9. cyclonic flows. In: Developments in laser techniques and
[66] Peng W, Hoffmann AC, Dries HWA, Regelink MA, Stein applications to fluid mechanics. Berlin: Springer; 1996. p.
LE. Experimental study of the vortex-end in centrifugal 67–82.
separators: the nature of the vortex end. Chem Eng Sci [86] Claypole TC, Evans P, Hodge J, Syred N. The influence of
2005;60:6919–28. the PVC on velocity measurements in swirling flows. In:
[67] Peng W, Hoffmann AC, Dries HWA, Regelink M, Foo K-K. Proceedings of the 3rd international symposium on
Neutrally buoyant tracer in gas cleaning equipment: a case application of laser techniques to fluid mechanics, Lisbon,
study. Meas Sci Technol 2005;16:2405–14. Portugal, 1986.
[68] Caplan KJ. Source control by centrifugal force and gravity. [87] Gorton-Hülgerth A. Messung und Berechnung der Gesch-
Air Pollut 1968;3:359–407. windigkeitsfelder und Partikelbahn im Gaszyclon, VDI-
[69] Geiger FJA, Dries HWA, Bakker E. Erosion, fouling and Fortschritt-Berichte, Reihe 7, Nr. 357, 1999.
cracking of FCC cyclone systems resolved. VDI-Berichte [88] Derksen JJ, Akker VD. Simulation of vortex core preces-
1511. Düsseldorf: VDI-Verlag; 1999. sion in a reverse flow cyclone. AIChE J 2000;46:1317–30.
[70] MacAuley, R. Dries H. FCC cyclone systems: a vital [89] Solero G, Cogher A. Experimental fluid dynamics char-
element for FCC profitability. In: European refining acterization of a cyclone chamber. Exp Therm Fluid Sci
technology conference, ERTC, Rome, 2000. 2002;27:87–96.
[71] Hoffmann AC, De Jonge R, Arends H, Hanrats C. [90] Griffiths AJ, Yazdabadi P, Syred N. The use of
Evidence of the ‘‘natural vortex length’’ and its effect on centre bodies and de-swirl vanes in the exhaust of
the separation efficiency of gas cyclones. Filtr Separat cyclone dust separators. Proc Inst Mech Eng 1996;210:
1995;32:799–804. 193–203.
[72] Montavon CA, Grotjans H, Hamill IS, Philips HW, Jones [91] Hankalic K. Closure models for incompressible turbulent
IP. Mathematical modelling and experimental validation of flows (lecture series). In: Beeck JPAJ, Benocci C, editors.
flow in a cyclone. BHR conference on cyclone technologies, Introduction to turbulence modeling. Von Karman In-
Warwick, 2000. stitute for Fluid Dynamics; 2004. p. 1–75 /http://
[73] Smith JL. An analysis of the vortex flow in the cyclone www.vki.ac.be/educat/lect-ser/2004/turbulence2004/hanja-
separator. J Basic Eng 1962; 609–18. lic.pdfS.
[74] Büttner H. Dimensionless representation of particle separa- [92] Jakirlic S, Hanjalic K. Modeling rotating and swirling
tion characteristics of cyclones. J Aerosol Sci 1999;30: turbulent flows: a perpetual challenge. AIAA J 2002;40:
1291–302. 1984–96.
 ARTICLE IN PRESS
C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452 451

[93] Wilkes NS, Clarke DS. Turbulent flow predictions using [113] Sommerfeld M. Theoretical and experimental modelling of
algebraic stress models. Technical report, AERE-R 12694, particulate flows. Lecture series 2000–2006, von Karman
Harwell Laboratory, 1987. Institute for Fluid Dynamics, 2000.
[94] Clarke DS, Wilkes NS. The calculation of turbulent flows [114] Chiesa M, Mathisen V, Melheim JA, Halvorsen BM.
in complex geometries using a differential stress model. Numerical simulation of particulate flow by the Euler-
Technical report, AERE-R 13428, Harwell Laboratory, ian–Lagrangian and the Eulerian–Eulerian approach with
1989. application to fluidised bed. Comput Chem Eng 2005;29(2):
[95] Hogg S, Leschziner MA. Computational of highly swirling 291–304.
confined flow with a Reynolds stress turbulence model. [115] Loth E. Numerical approaches for motion of dispersed
AIAA J 1989;27:57–63. particles, droplets and bubbles. Prog Energy Combust Sci
[96] Ferziger JH. The physics and simulation of the turbu- 2000;26:161–223.
lence. In: VKI lecture series, introduction to turbulence [116] Ho CA, Sommerfeld M. Modelling of micro-particle
modeling. von Karman Institute for Fluid Dynamics, agglomeration in turbulent flow. Chem Eng Sci
Springer; 2002. 2002;57:3073–84.
[97] Prasad R, Bakker A. Cyclone modeling using LES. Fluent [117] Gosman AD, Ioannides E. Aspects of computer simu-
users group meeting. Massachusetts: Danvers; 1999. lation of liquid-fuelled combustors. AIAA Paper 81-0323,
[98] Laurence D. Applications of Reynolds averaged Navier 1981.
Stokes equations to industrial flows. In: Introduction to [118] Berlemont A, Desjonqueres P, Gouesbet G. Particle
turbulence modeling. Bruxelles: VKI von Karman Institute Lagrangian simulation in turbulent flows. Int J Multiphase
for Fluid Dynamics; 2002. Flow 1990;16:19–34.
[99] Speziale SG. Turbulence modeling for time-dependent [119] Sommerfeld M, Kohnen G, Rüger M. Some open questions
RANS and VLES: a review. AIAA J 1998;36:173–84. and inconsistencies of Lagrangian particle dispersion
[100] Spalart PR. Strategies for turbulence modeling and models. In: Proceedings, 9th symposium on turbulent shear
simulations. Int J Heat Fluid Flow 2000;21:252–63. flows, Kyoto, Japan, Paper 15-1, 1993.
[101] Guo B, Langrish TAG, Fletcher DF. Simulation of [120] Ma L, Ingham DB, Wen X. Numerical modelling of the
turbulent swirl flow in an axisymmetric sudden expansion. fluid and particle penetration through small sampling
AIAA J 2001;39:96–102. cyclones. J Aerosol Sci 2000;31:1097–119.
[102] Dyakowsky W. Modelling turbulent flow within a [121] Frank Th, Wassen E, Yu Q. Lagrangian prediction of
small-diameter hydrocyclone. Chem Eng Sci 1993;48: disperse gas particle flow in cyclone separators. In: Third
1143–52. international conference on multiphase flow, ICMF ‘98,
[103] Griffiths WD, Boysan F. Computational fluid dynamics Lyon, France, 1998.
(CFD) and empirical modelling of the performance of a [122] Frank Th. Application of Eulerian-Lagrangian prediction
number of cyclone samplers. J Aerosol Sci 1996;27: of gas particle flows to cyclone separators. In: Theoretical
281–304. and experimental modelling of particulate flow, Lecture
[104] Averous J, Fuentes R. Advances in the numerical simula- Series 2000–2006. von Karman Institute for Fluid Dy-
tion of hydrocyclone classification. Can Metall Quart namics.
1997;36:309–14. [123] Narasimha M, Brennan MS, Holtham PN, Napier-Munn
[105] Jo Y, Tien C, Ray MB. Development of a post cyclone to TJ. A comprehensive CFD model of dense medium cyclone
improve the efficiency of reverse flow cyclones. Powder performance. Miner Eng 2006, DOI:10.1016/j.mineng.
Technol 2000;113:97–108. 2006.10.004
[106] Grotjans H. Application of higher order turbulence models [124] Crowe CT, Sharma MP, Stock DE. The particle-source-in-
to cyclone flows, VDI-Berichte 1511. Düsseldorf: VDI- cell (PSI-cell) model for gas-droplet flows. J Fluid Eng
Verlag; 1999. p. 175–82. 1977;99:325–32.
[107] Meier HF, Mori M. Anisotropic behavior of the Reynolds [125] Kohnen G, Rüger M, Sommerfeld M. Convergence
stress in gas and gas-solid flows in cyclones. Powder behaviour for numerical calculations by the Euler/
Technol 1999;101:108–19. Lagrange method for strongly coupled phases. In: Numer-
[108] Witt PJ, Mittoni LJ. Validation of a CFD model for ical Methods in Multiphase Flows, ASME Summer Meet-
predicting gas flow in a cyclone. CHEMECA 99, New- ing, FED-vol. 185, Lake Tahoe, USA, 1994. p. 191–202.
castle, Australia, 1999. [126] Crowe CT, Troutt TR, Chung JN. Numerical models for
[109] Slack MD, Prasad RO, Bakker A, Boysan F. Advances in two-phase turbulent flows. Annu Rev Fluid Mech
cyclone modelling using unstructured grids. Trans I Chem 1996;28:11–43.
2000;78:1098–104. [127] Derksen JJ, Sundaresan S, van den Akker HEA. Simula-
[110] Peng W, Boot P, Udding A, Hoffmann AC, Dries tion of mass-loading effects in gas–solid cyclone separators.
HWA, Ekker A, et al. Determining the best modelling Powder Technol 2006;163:59–68.
assumptions for cyclones and swirl tubes by CFD and [128] Ibsen CH, Helland E, Hjertager BH, Solberg T, Tadrist L,
LDA. International Congress for Particle Technology; Occelli R. Comparison of multifluid and discrete parti-
2001. p. 1–8. cle modelling in numerical predictions of gas particle flow
[111] Liu Z, Jiao J, Zheng Y, Zhang Q. Investigation of in circulating fluidised beds. Powder Technol 2004;149:
turbulence characteristics in a gas cyclone by stereoscopic 29–41.
PIV. AIChE J 2006;52:4150–60. [129] Enwald H, Peirano E, Almstedt AE. Eulerian two-phase
[112] Elghobashi SE. On predicting particle-laden turbulent flow theory applied to fluidization. Int J Multiphase Flow
flows. Appl Sci Res 1994;52:309–29. 1996;22:21–66.
 ARTICLE IN PRESS
452 C. Cortés, A. Gil / Progress in Energy and Combustion Science 33 (2007) 409–452

[130] Manninen M, Taivassalo V, Kallio S. On the mixture [134] Brennan MS, Narasimha M, Holtham. Multiphase model-
model for multiphase flow. Espoo, Finland: VTT Publica- ling of hydrocyclones—prediction of cut-size. Miner Eng
tions 288; 1996. 2006 DOI:10.1016/j.mineng.2006.10.010.
[131] Narasimha M, Sripriya R, Banerjee PK. CFD modelling of [135] Grady SA, Wesson GD, Abdullah M, Kalu EE. Prediction
hydrocyclone—prediction of cut size. Int J Miner Process of 10-mm hydrocyclone separation efficiency using compu-
2005;75:53–68. tational fluid dynamics. Filtr Separat 2003;40:41–6.
[132] Brennan MS. Multiphase CFD simulations of dense medium [136] Meier HF, Mori M. Gas–solid flow in cyclones: the
and classifying hydrocyclones. In: Proceedings of the third Eulerian–Eulerian approach. Comput Chem Eng
international conference on CFD in the minerals and process 1998;22(Suppl 1):S641–4.
industries, CSIRO Melbourne Australia, 2003. p. 59–63. [137] Qian F, Huanga Z, Chena G, Zhang M. Numerical study
[133] Narasimha M, Brennan MS, Holtham PN. Numerical of the separation characteristics in a cyclone of different
simulation of magnetite segregation in a dense medium inlet particle concentrations. Comput Chem Eng 2006,
cyclone. Miner Eng 2006;19(10):1034–47. DOI:10.1016/j.compchemeng.2006.09.012