See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/317248783

Experimental study of cyclone performance for blow-by gas cleaning

applications

Article  in  Journal of Aerosol Science · May 2017

DOI: 10.1016/j.jaerosci.2017.05.009

CITATION READS

1 229

4 authors, including:

Benoit Sagot
Ecole Supérieure des Techniques Aéronautiques et de Construction Automobile
11 PUBLICATIONS   127 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

blow-by gas cleaning View project

All content following this page was uploaded by Benoit Sagot on 19 October 2017.

The user has requested enhancement of the downloaded file.

 Experimental study of cyclone performance
for blow-by gas cleaning applications

B. Sagota,*, A. Forthommea, L. Ait Ali Yahiaa, G. De La Bourdonnayeb

(a) École Supérieure des Techniques Aéronautiques et de Construction Automobile

(ESTACA), 12 Avenue Paul Delouvrier, 78180 Montigny-le-Bretonneux, France

(b) Toyota Motor Europe, Hoge Wei 33 - Technical Centre, B-1930 Zaventem, Belgium

Corresponding author: Tel.: +33 (0)1 76 52 11 57

E-mail address: benoit.sagot@estaca.fr (B. Sagot).

Abstract

The aim of this experimental study is to investigate the separation performances of a new set of small cyclones.
The diameter of these cyclones spans the range of 20 to 45 mm, which is the typical size of these devices when they
are used for separating oil mist from blow-by gases in internal combustion engines. To reproduce the flow rates and
oil load of combustion engines, we developed a flow bench, with a polydisperse aerosol generator producing engine
oil droplets with diameter in a 0.3-10 µm range. Measurements of the aerosol particle concentrations upstream and
downstream of the cyclone permitted the determination of the particle collection efficiency, for the six devices that
were tested, with flow rates ranging from 20 to 200 Nl/min which are typical operating conditions. For the six
cyclones, the geometry is normalized and all internal dimensions within the cyclone are proportional to its diameter.
With the operating conditions of this study, the annular Reynolds number varied in the range 1.27 x 103 to 12.2 x
103. From the measured fractional collection efficiency curves, dimensionless cut-off aerodynamic diameters were
determined. These measurements have been compared with available correlations for the determination of this
dimensionless diameter as a function of the annular Reynolds number. Different correlations are proposed in this
study, for the evaluation of the cut-off diameter and for the shape of the collection efficiency. We were able to
establish that the collection efficiency curve dimensionless slope is correlated to a Reynolds number.

Keywords: Cyclone; cut-off diameter; blow-by gas; oil mist;

1. Introduction

Cyclone separators are commonly used in industry to remove solid or liquid particles from a gas stream. These
devices are frequently applied to large-scale processes, for both separation and drying applications. Small cyclones
have also found various applications, such as personal cyclone samplers applied to environment control. They are
also widely used in the automotive industry, particularly in internal combustion engines for the separation of oil
mist from blow-by gases. These so-called “blow-by gases” result from leakages between the combustion chamber
and the crankcase. These gases flow through the crankcase which contains lubricating oil and the resulting oil mist
has to be cleaned up, for pollution control, oil consumption reduction and also to avoid turbo charger oil coking.
This is an important feature of crankcase venting systems, and the design of the separation cyclone is of great
importance for its collection efficiency: a reduction of the cyclone diameter produces an increase of the collected
oil flow, but also results in increased pressure drops.
Various numerical or experimental studies have been carried out to determine the influence of the flow rate and
cyclone geometrical parameters on the separation performances. Cut-size diameters which correspond to 50%
collection efficiency have been reported in the literature, for a wide range of cyclones, together with theoretical
models. Cortés & Gil (2007) proposed a review of these models developed for the evaluation of the cyclone
separators collection efficiency. However, it should be noted that currently available studies on small cyclones with
liquid aerosol are limited.
In this experimental study, we propose to investigate the collection efficiency performances of a new set of six
small cyclones with diameters in the range of 20 to 45 mm (C-20 to C-45), which are typical of combustion engine
crankcase applications where oil droplets polydisperse aerosols are collected. This set of cyclones which are fully
homothetic will be referred to as the SFYB family.
1
2. Review of cyclone modeling

2.1 Cut-off diameter

Together with the pressure drop, the major characteristic of a cyclone flow is the cut-off diameter dae,50 that
corresponds to a 50% collection efficiency. Many semi-empirical models have been proposed in literature for the
prediction of the collection efficiency curve of a cyclone. To predict the cut-off size, Barth (1956), Dietz (1981) and
Mothes & Löffler (1988) developed theories based on a force balance between the centrifugal force and drag force
acting on particles. Another so-called “time-of-flight” modeling approach (Lapple, 1950; Leith and Licht, 1972)
considers the residence time of a particle traveling from an initial position to the cyclone wall. Both approaches
provide simple relationships between the Stokes number (St50, corresponding to the cut-off size) and one or several
ratios of cyclone dimensions.
Ccmd  p d ae,50
2
U
St 50  (1)
18 g L

with Ccmd the Cunningham-Millikan-Davies correction factor,  p the particle density, g the fluid viscosity, U
and L characteristic gas velocity and cyclone dimension which may vary between the theories. The Ccmd correction
factor is defined as a function of the particle Knudsen number Kn and is evaluated with the following equation:

Ccmd  1  Kn ( A  B  e C / Kn ) (2)

where A=1.257, B=0.40 and C=1.10 (Friedlander, 2000).

For low mass loading, the cut-off diameter can be estimated using Muschelknautz’s model, with the following
expression (Hoffmann and Stein, 2002):
18 g (0.9 Qv ) (3)
d 50,MM  xfact,MM
2 (  p   g ) v2
CS ( Lb  Lc  S )

Equation (3) is a variation of Barth’s modeling, taking into account a ‘roof leakage’, and a correction factor xfact,MM
that may be applied to force the computed cut-point to match that observed in practice, and that normally falls within
the range of 0.9 to 1.4 (Hoffmann and Stein, 2002).

Most of the results from the literature express the variation of the Stokes number as a function of the annular
Reynolds Number, which is defined as:
 g uin ( D  Do ) / 2 (4)
Reann 
g

With uin the average gas velocity at the cyclone inlet,  g the gas density, D the cyclone diameter and Do the
cyclone vortex finder diameter (figure 1).
However, it should be noted that the choice of the definition for the Reynolds number might be different from one
author to another. As an illustration, Beeckmans & Kim (1977) and John & Reischl (1980) used the cyclone
Reynolds number which is based on the cyclone diameter D and the inlet velocity uin, while Saltzman & Hochstrasser
(1983) presented their measurements using the outlet tube Reynolds number, based on the average axial air velocity
in the vortex finder tube and its internal diameter.
Overcamp & Scarlett (1993) brought together all available cut-point Stokes number results, and presented a graph
of these data as a function of the corresponding values of the cyclone Reynolds number, to show that experimental
points were scattered in this dimensionless representation. Later, Moore & McFarland (1996) suggested a new
presentation of the measurements, based on a dimensionless cut-off aerodynamic diameter  50 (see eq. (5)) which
is based on the cyclone body diameter D multiplied by the square root of the Cunningham-Millikan-Davies
correction factor:

Ccmd d ae,50 (5)

 50 
D

2
 With this new definition of a dimensionless cut-off aerodynamic diameter, Moore & McFarland (1996) established
that most of the experimental data were gathering on a straight line in a log-log plot, and they proposed a best-fit
correlation with two regression coefficients a and b:
ln ( 50 )  ln (a)  b ln ( Reann ) (6)

Lidén & Gudmundsson (1997) extended Moore & McFarland results, by suggesting that the correlation
coefficients might depend on geometrical parameters. The regression models tested by Lidén & Gudmundsson
(1997) are similar to eq. (6), but with additional dimensionless terms that depend on dimensionless aspect ratios:
n (7)
ln ( 50 )  ln( A0 )  A1 ln ( Reann )   Ai ln ( Bi )
i 2

where the values of Bi denote various cyclone dimensional ratios such as D/Do or S/D and (Lb+Lc-S)/Do. A0, A1,
and Ai are the regression coefficients obtained, and n is the number of regression coefficients. Lidén & Gudmundsson
(1997) used least-square techniques to define their model and suggested to retain only regression coefficients
significant at a level of 5%. With this condition, authors retained only one geometrical parameter dependence which
is the dimensionless length of the vortex finder S/D. With a regression coefficient r2=0,981 and a GSD (geometric
standard deviation of the fit) of 1.12, the corresponding final model is:
ln ( 50 )  3.184  0.713 ln ( Reann )  0.172 ln (S/D) , with r2=0.981 and GSD=1.12 (8)

Kuo & Tsai (2001) proposed an analysis of a wide set of data including their measurements (Tsai et al., 1999), to
evaluate the accuracy of existing theories on the cyclone cut-off diameter. They concluded that no unique theory on
the cut-off aerodynamic diameter was fitting all the available data for all cyclone sizes and shapes, or flow Reynolds
number ranges. However, by using eq. (6) suggested by Moore & McFarland (1996), they proposed a new set of
coefficients for the evaluation of the dimensionless cut-off diameter, with a regression coefficient r2=0.98:
ln ( 50 )  3.17  0.74 ln ( Reann ) , with r2=0.98 (9)

This correlation (9) is based on a large set of measurements reported by Dirgo & Leith (1985-a), Iozia & Leith
(1989), Kim & Lee (1990), Moore & McFarland (1993), Lin (1996), Tsai et al (1996) and Hsiao (1997), and the
corresponding correlation coefficient R2 is 0.98.
Kuo & Tsai (2001) established that for Stairmand-type cyclones, an excellent correlation coefficient r2=0,994
could be obtained with measurements carried out by Dirgo & Leith (1985-a), Iozia & Leith (1989), Moore &
McFarland (1993) and Lin (1996), by using the following correlation:
ln ( 50 )  3.46  0.72 ln ( Reann ) , with r2=0.994 (10)

Hsiao et al. (2009) investigated the separation performance of two prototype mini-cyclones with cyclone body
diameter D=2.61 mm and D=4.48 mm. The performances of these two cyclones were compared with existing
models. They concluded that the cut-off aerodynamic diameter was underestimated when using the correlation
proposed (eq. (8)) suggested by Lidén & Gudmundsson (1997). As none of the available models was fitting correctly
with their measurements, Hsiao et al. (2009) developed a linear regression model (eq. (11), with a regression
r2=0.966) that is proposed to be used as a tool for the future design of mini-cyclones with the similar size and
configuration:
ln ( 50 )  3.219  0.656 ln ( Reann ) , with r2=0.966 (11)

In our experimental study, the six cyclones of the SFYB family have been tested over a wide range of Reynolds
number, and the experimental determinations of the cut-off aerodynamic diameter will be compared with existing
correlations.

2.2 Collection efficiency models

A cyclone collection efficiency expresses the dependence of separation performance upon particle diameter. The
corresponding curve is usually a S-shape that goes to zero for very small particles, and reaches a unit value for
particles much larger than the cut-off diameter. Based on the idea that for a particle of any size, the efficiency can
be determined from its ratio to the cut-off diameter, a variety of models have been proposed to represent this
collection efficiency function. Dirgo and Leith (1985-b) obtained a best fit of measurements with the following
simple model:

3
 1 (12)
 (d ae ) 
1  (d ae,50 / d ae ) 6.4

Later, Hoffmann and Stein (2002) extended this model, suggesting that the value 6.4 could be replaced by an
arbitrary value m, that could be fitted to match experimental results.
Lidén & Gudmundsson (1997) introduced a dimensionless particle size  relative to measured aerodynamic cut-
off diameter dae,50, which is normalized in such a way that 50% collection efficiency occurs for  = 0 :

Ccmd (d ae ) d ae (13)
Ξ 1
Ccmd (d ae,50) d ae,50

Lidén & Gudmundsson pooled together a wide range of available collection efficiency data, for four different
types of cyclones, including the Stairmand cyclone, Lapple cyclone, German Z cyclone, and Southern Research
Institute (SRI) cyclones II and III. They found that with a representation of the collection efficiency as a function of
this normalized aerodynamic diameter, these data could be fitted correctly with a quadratic (Blachman and
Lippmann, 1974), or a linear (Maynard and Kenny, 1995) interpolation, defined as:

E W  W Ξ  W2 Ξ 2 (14)
ln  0 1
1  E  W0  W1 Ξ

A collection efficiency steepness parameter 1, was defined by Lidén & Gudmundsson as the derivative at :
 1  W1 (15)

With the available data, Lidén & Gudmundsson were not able to establish which cyclone design parameters
determine the steepness of the efficiency curve, concluding that further experimental work was required.

2.3 Pressure losses in cyclones

The performances of a cyclone separator are evaluated through two major parameters which are the separation
efficiency and the pressure drop of the device. Based on Barth’s model, Muschelknautz and co-workers
(Muschelknautz and Kambrock, 1970; Muschelknautz and Trefz, 1990; Trefz and Muschelknautz, 1993;
Muschelknautz, 2010) developed an empirical model that is recognized as the most practical method for modelling
cyclone separators at the present time (Cortes and Gil, 2007; Hoffmann and Stein, 2002)
The pressure drop in cyclones, for “clean” gas with no particle is usually expressed as:
1 (16)
P  Euin g,in uin2
2
 g / (  g,in WH) ,
with Euin the inlet Euler number, uin the value of the average velocity in the inlet tube uin  m
 g,in the gas density evaluated at the inlet pressure, H and W respectively the height and width of the inlet section.
The so-called ‘‘pressure drop’’ or ‘‘pressure losses’’ is defined as the viscous dissipation of total pressure. However,
from an experimental point of view, it should be noted that pressure drops are usually measured by the means of
static pressure taps on the inlet and outlet duct walls (Zhu and Lee, 1999). In the inlet pipe where streamlines are
rectilinear, it is correct to consider the total pressure as the sum of the measured static pressure and the evaluated
dynamic pressure 0.5g,in uin2 . In the outlet pipe, the strongly swirling flow may induce an incorrect total pressure
evaluation. This point is clarified by Hoffman and Stein (2002), who report that the best way to evaluate pressure
drops is probably to let the cyclones exhaust at the atmosphere. Considering a complete loss of velocity head or
dynamic pressure is a good way to get around the problem of the cyclone outlet pressure interpretation. This
hypothesis of complete loss of dynamic pressure has been used to build the pressure drop evaluation model proposed
by Muschelknautz, which is currently reported as the most relevant for the evaluation of pressure drops. Our
measurements are compared with the results obtained by application of this model. We summarize its equations in
Table 1. In this modelling, the cyclone body Reynolds number ReR is used to evaluated the wall friction factor f
from the original graph reported by Muschelknautz and Trefz (1990 - fig. 10, page 252), also reported by Hoffmann
and Stein (2002 - fig. 6.3, page 101). In our experiments, the cyclone body Reynolds number ReR varied in the range
60 to 600, which is slightly lower than the typical values in large scale industrial applications. In this range, the
4
friction factor depends on both the Reynolds number ReR and on the relative roughness, ratio of the absolute
roughness to the cyclone internal radius. The absolute roughness was estimated to 50 µm for all cyclones.
All dimensions are defined in figure 1, and the position of the pressure tap used for the inlet static pressure
measurement is provided in figure 5. As observed, we measure the static pressure upstream the convergent and the
static pressure in the inlet section (with height H and width W) is then evaluated considering Bernoulli relation
p+0.5u2=constant between section 2 and the inlet section.

Table 1 Muschelknautz model for the evaluation of the pressure drop coefficient Euin in cyclones.
Euin  (4 WH / ( Do2 )) 2 (b  o )
f Sf ( vθw vθCS )3/2
b  , o  2  3 (vθCS / vo ) 4/3  (vθCS / vo ) 2
0.9 Qv vo2
with
 2
Sf  D  Do2  4( DLb  Do S )  ( D  Dd ) 4L2c  ( D  Dd )2 

4  
Rin Rmvzw g
Cyclone body Reynolds number ReR 
( Lb  Lc ) g (1  (vzw / vm ) 2 )
 g / (  g,in WH) , vw  uin Rin / (R)
Rin  D / 2  W / 2 , Rm  ( Do / 2)( D / 2) , uin  m


1

 2
 2 2 
1  1  4 ( / 2)   / 2 1  (1   )( 2   ) / (1  c0 )  , with   W / (0.5D)

c0 corresponds to the ratio of particle mass flow to the air mass flow, equal to zero since pressure loss evaluation
are carried out with a pure gas flow.
0.9 Qv
vθm  vθw vθCS , vzw  , vo  Qv / ( Do2 / 4)
 (( D / 2) 2  Rm2 )
 f S f vw D / Do 
vCS  vw ( D / Do ) / 1  
 2Qv 

3. Design of the SFYB blow-by cyclone family

To design our cyclone family, some constraints are imposed:

- pressure drop should be less than 20 mbar, which is a current limit value in blow-by separation
applications

- the flowrate is comprised in a range of 10 and 200 Nl.min-1 (Normal Liter per Minute, so for a
corresponding density determined in the normal conditions T=0°C and P=101325 Pa)

The main dimension parameters of the cyclones design are presented in figure 1, and dimensionless ratios are
reported in Table 2, together with Stairmand, Lapple, German Z, and Southern Research Institute (SRI) cyclones.

5
 Do
vortex finder Dt,mod=30 mm
Section 2 20 cm
Section 1
Section 3
S
H Do
Lb
Lo
H W

D W S H
Lb
Lc
D
Dd
Lc

Dd
figure 1 : Cyclone geometrical parameters definition.

Dimensionless ratios H/D (figure 1), D/Do, S/D and W/D values are close to the values of the high efficiency
design of the Stairmand cyclone. However, the cyclones of this new family are designed to be integrated in engine
head covers, and the total height of the cyclones have been limited, with a value of the (Lb+Lc)/D parameter limited
to 2, while more classical values can reach twice this value.

We present in figure 2 the shape of the cyclone inlet pipe. The Section 1, 2 and 3 corresponds to three cross-
sections at different locations of the cyclone. For every cyclone, the angle of the convergent is constant, and the
length of this convergent is adjusted so that Section 1 is a square section of width W and height H. In Section 1, we
connect the square section of the cyclone inlet pipe to a circular conical part with the same cross section. As the
entering flow switches from a circular pipe to a square section, we maintained a stabilizing length of 4*H upstream
the convergent part of the inlet pipe. The conical part is connected to the upstream sampling probe with a constant
10 mm diameter.
Static pressure tap position conical connector
D Upstream sampling
12°
4*H

H
W
10 mm
Section 2 Section 1
figure 2 : Cyclone inlet pipe definition.

The connection of the cyclone outlet is presented on the right part of figure 1. The vortex finder length S
corresponds to the distance between the bottom of the vortex finder and the cyclone roof. We also represent the
length Lo which is adjusted for every cyclone of the family so that Lo/D=2. After Section 3 (see figure 1), a divergent
conical part connects the cyclone outlet tube to a large diameter pipe, with diameter Dt,mod=30 mm, where the
downstream sampling tube is located, 20 cm downstream of section 3, to limit any interaction between this sampling
tube and the vortex. Indeed, in an initial set-up (see Annex 1), we have been using a configuration with the same
sampling probes for both upstream and downstream sampling. So we were using a convergent conical part to connect
Section 3 with the constant diameter tube (Dt=10 mm) of the downstream sampling probe (see left side of figure 20,
in Annex 1). We observed that the contraction of the tube downstream the vortex finder produces a modification of
the vortex intensity inside the cyclone, thus leading to a bias in the measured collection efficiency performances of
a cyclone. This impact of the experimental setup on the cyclone cut-off diameter is reported in Annex 1.

6
Table 2 Cyclone geometrical parameters values.
Dimensionless Hsiao SFYB
Stairmand Lapple Z SRI
ratios Mini-cyclone (this study)
D/Do 2.0 2.0 4.0 3.5 4,1 2.23
H/D 0.5 0.5 0.45 (-) (-) 0.48
W/D 0.2 0.25 0.11 (-) (-) 0.22
D/Rinlet (-) (-) (-) 3.5 14,7-16,3 (-)
S/D 0.5 0.75 0.82 0.43 0.4 0.62
(Lb+Lc)/D 4.0 4.0 3.4 1.9 1.7 2
Lb/D 1.5 2.0 0.86 1.3 (-) 1

As reported in Table 3, the dimensions of the six tested cyclones C-20 to C-45 have varying dimensions, but are
fully proportional, all dimensionless ratios being constant.

Table 3 Cyclone geometry.

C-20 C-25 C-30 C-35 C-40 C-45

D (mm) 20 25 30 35 40 45
Common D/Do =2.23, H/D =0.48, W/D =0.22, S/D =0.62, (Lb+Lc)/D =2, Lb/D =1,
parameter Lo/D =2

4. Experimental set-up for the determination of the cyclone collection efficiencies

For this experimental study, we used two polydisperse aerosol generators PLG-2010 (PALAS), fed with clean
compressed air (only one is represented in figure 4). These devices have been designed for this kind of application
(blow-by gases simulation), with a laskin nozzle providing a wide range polydisperse aerosol with droplets diameters
in the range 0.3 to 10 µm with the engine oil SAE 0W-30 used in this experimental study. It should be noted that
the oil temperature is regulated in the generator, thus providing a time stable aerosol generation. To increase the
generation stability, the two PLG-2010 generators were fed with air via precision mass flow controllers (Brooks
5851S). The high concentration aerosol produced by these two generators is then diluted with a pure air flow,
produced by a third mass flow controller (Brooks 5853S), to achieve a global flowrate close to 200 Nl.min-1. One
of the generators is equipped with a downstream impactor to remove large diameter particles, and produces
distributions with a large fraction of fine particles. The cut-size of the impactor depends on the flow rate, thus
permitting to adjust the distribution. With the large range of measured cut-off diameter (1 to 4.5 µm) obtained, this
configuration permitted to adjust the particle distribution and get sufficient particle loading in the diameter range of
the measured cut-off diameter, as illustrated in figure 3. We present in this figure 3 a curve that corresponds to an
average of measurements carried out on a Euro 6 diesel engine, for 18 points of an engine map, where both engine
torque and rotation speed of the engine varied in a wide range. We observed during these measurements variations
of the mass concentration produced by the engine, while the mass-weighted relative size distribution was
approximately constant. On our experimental laboratory bench, we used two polydisperse aerosol generators to
reproduce as much as possible this engine distribution.

7
 100%
Cumulated mass
90% distribution (%)
80%
70%
60%
50% Average of engine bench
40% measurements

30% Average of laboratory C-20 to C-45

measurements
20%
10%
particle diameter (µm)
0%
0 1 2 3 4 5 6 7 8
figure 3 : Comparison of cumulated mass distribution. Black curve is an average of engine bench
measurements; red curve is the average of all cyclone measurements in this study; Grey area represents minimum
and maximum of distributions used in this study.

When the aerosol generation is stabilized and sufficiently diluted, the flow through the cyclone is varied by using
a bypass valve (figure 4) which permits to deviate part of the aerosol flow. Downstream of the bypass valve, a wall
flow trap is used to eliminate any oil deposited on walls, and to ensure that the cyclone is only fed with the aerosol.
It’s a simple volume that slows down the aerosol flow and captures the liquid fraction by gravity. This wall flow
trap is regularly emptied.

P l/min
Pin
Bypass
valve Jtec VF563

Mass flow controller

Compressed air Filter wall flow trap
Nl/min
Valve
5853S

Nl/min
Dryer 5851S upstream/downstream Welas particle
sampling probe
Mass flow controller sizer

Laskin nozzle aerosol Collected oil

generator
RS232

figure 4 : Experimental set-up.

To control the flowrate in the cyclone separator, a two-step procedure is used: first, the bypass valve is closed and
a pressure drop in the cyclone is measured for the different targeted flowrates, with a high precision differential
pressure sensor (Keller Druck PD-41X). The position of the pressure tap used for the inlet static pressure
measurement is provided in figure 2, and the downstream pressure tap is located in the vortex finder, at a distance
equal to 1.8*Lo from the vortex finder inlet, so close to Section 3 (see figure 20, in Annex 1), and upstream of
perturbations due to the sampling probe. As already pointed out, the flow in the downstream measurement section
is highly swirling, and we don’t search with these measurements for a determination of the cyclone pressure losses
to compare with the model’s predictions. These pressure loss evaluations will be carried with the set-up described
in section 5 of this manuscript. Then, when aerosol measurements are carried out, the bypass valve position is
adjusted to get a specific pressure drop and its corresponding flowrate through the cyclone. To cross check the
validity of this procedure, a flowmeter (J-Tec VF563) is located downstream of the separator. Since the sampling
flow rate is constant (5 l/min) in the downstream sampling probe, the actual mass flow rate in the cyclone is the sum
of the mass flowrate evaluated from the volumetric flowrate reported by the J-Tec VF563 and of the sampling mass
flow rate.

8
 The particle size distribution measurements are carried out with an optical particle counter WELAS 2100
(PALAS), which has been identified as a relevant measurement system for the polydisperse aerosols analysis (Heim
et al., 2008). In this system, the determination of a particle diameter is based on the measure of the scattered light
intensity sent out by the particle, and detected at an angle of 90° of the 77 µm*82 µm measurement volume. This
intensity depends on the refractive index of the particle material (refractive index m= 1.45 for DEHS, and m = 1.464
for engine oil SAE 0W-30 used in this study). A calibration curve is thus necessary to convert the scattered light
intensity into an optical diameter, which was provided by the WELAS manufacturer. However, during operation,
the sensor must be regularly cleaned and it is important to check and correct the scattered light intensity with known
diameter and optical particle properties. This is done using monodisperse polystyrene latex (PSL, refractive index
m = 1.59). Since the WELAS is a single particle sensor with a detection range of about 0.3–10 µm optical equivalent
diameter, the concentration of the sample flow is limited, to prevent coincidence errors. To avoid the use of a dilution
system, the particle number concentration in these experiments has been limited to a range of 0.4 to 0.7x105 part.cm-
3.
Upstream and downstream aerosol particle concentration measurements are carried out, to determine the fractional
collection efficiency of the cyclone separators. Two different sampling systems are used to collect the aerosol
upstream and downstream of the cyclone: the upstream sampling probe is connected to Section 1 (see figure 2). This
sampling probe is composed of a divergent-convergent geometry (angle of the cone is 16°), with a maximum
diameter in the middle equal to 29 mm. In this section, a sharp edge tube is used for the sampling. For the
downstream sampling, as already pointed out, it is important to avoid a reduction of the cross-section downstream
the vortex finder and upstream the sampling location, since this increase of the velocity in the pipe has an impact on
the vortex intensity, so on the measured collection efficiency (see Annex 1). For this reason, the downstream
sampling tube is simply introduced into the pipe downstream of the cyclone after section 3 and a divergent
connection to this outlet tube with diameter Dt,mod (see figure 1).
To obtain a representative sample of the aerosol, it is important to ensure for an isokinetic sampling, especially
for particles with diameter greater than 1 µm. In this study, the sampling flow rate of the optical particle counter is
constant, equal to 5 l/min, which fixes the sampling velocity. For a given cyclone configuration, when the flowrate
in the cyclone is varied, we should adjust the sampling tube diameter for every flowrate, to maintain an isokinetic
sampling. This was not possible, and we simply used four sets of upstream/downstream sampling device (probe
upstream, and direct sampling downstream), with a sampling internal diameter in the range 6 to 14 mm for both the
upstream and downstream sampling tubes (internal diameter of upstream probe is 29 mm, close to the diameter
value Dt,mod=30 mm of the outlet tube). With this sampling procedure, it was possible to ensure for a maximum of
30% difference between the sampling and mainstream velocities. We made some tests to check that this limited non-
isokinetic sampling has no influence on the results, by measuring in parallel the particle concentrations with
isokinetic and non-isokinetic conditions for pairs of probes. To convey the aerosol from the upstream and
downstream sampling probes to the optical particle counter cell, we used flexible hoses composed of Hytrel (Bev-
A-line) with inner diameter of 10 mm, specially designed for minimal particle loss, with large bending radius and
with the same length to ensure for an equivalent penetration rate of both upstream and downstream sampling trains.
We evaluated the diffusion losses in the sampling hoses, and with particles with diameter over 0.3 µm and a hose
length limited to 0.5 m, we concluded that these losses were negligible. The oil density is 845 kg.m-3, and we convert
geometrical particle diameters d into aerodynamic diameters, using eq. (17):

Ccmd (d p ) p (17)
d ae  dp
Ccmd (d ae )  water

5. Experimental configuration for the determination of pressure losses

The experimental configuration used for the pressure losses measurements is slightly different from the set-up
used for the determination of the cyclone collection efficiencies. We don’t measure any static pressure in the outlet
pipe, where the strongly swirling flow may induce an incorrect total pressure evaluation. As recommended by
Hoffman and Stein (2002), we let the cyclones exhaust at the atmosphere, considering a complete loss of velocity
head or dynamic pressure. We present in figure 5 the set-up we used: a mass flow controller (Brooks 5853S) imposes
the flow rate, and we simply measure the inlet static pressure with a high precision differential pressure sensor
(Keller Druck PD-41X).

9
 Patm
Section 3
Static pressure tap Patm Pin-Patm
Compressed air Filter
Nl/min Lo
5853S
Mass flow Section 1 Section 2
Dryer controller

Section 1 Section 2

W
H
side view
12°
4*H
Static pressure tap position top view
figure 5 : Set-up for pressure loss evaluation.

The measured pressure drop is thus simply the sum of the difference in static pressure and the dynamic pressure
0.5g,inuin2 :

pexp  ( pin  patm)  0.5g,inuin2 (18)

6. Result and analysis

We present in figure 6 the evolution of the grade-efficiency curve or so-called separation efficiency when varying
the cyclone flow rate, for the cyclone with a diameter D=35 mm. One can note that for each value of flow rate, the
collection efficiency increases with the particle diameter, to reach a value close to 100%. In most of the practical
applications, this grade-efficiency curve takes a smooth “S” shape, with low values of the collection efficiency for
small particle diameters.
100%
Collection efficiency (%)
90%
Qv =24 Nl/min
80% Qv =30 Nl/min
70% Qv =37 Nl/min
Qv =44 Nl/min
60%
Qv =52 Nl/min
50%
Qv =61 Nl/min
40% Qv =71 Nl/min

30% Qv =82 Nl/min

Qv =96 Nl/min
20%
Qv =116 Nl/min
10%

0%
0,1 1 10
dae(µm)
figure 6 : Collection efficiency measurements for the C-35 cyclone, with a 35 mm diameter.

10
 In the figure 6, solid points correspond to measurements of the collection efficiency, and a best fit with a simple
hyperbolic tangent model (solid curves) was used to determine the corresponding cut-off diameter. For this
determination of the cut-off diameter, measurement points at the upper and lower end of the collection efficiency
curve should not be taken into account. Indeed, as observed on the figure 7, we can even measure “negative”
collection efficiency values, which is due to a limited number of counted particles together with an imperfect
stability of the aerosol generation. To avoid taking into account such erroneous data, we simply decided to select
only the measurement points corresponding to a range of collection efficiency of 0.1 to 0.9, which is illustrated on
figure 7, for two values of the flow rate in cyclone C-35.
100%
Collection efficiency (%)
75%

50%
C-35, Qv=24 Nl.min-1

25% C-35, Qv=96 Nl.min-1

0%
0.1 1
dae (µm)
-25%
figure 7 : Collection efficiency fitting.

During these measurements, we checked the stability of the aerosol generation, by carrying upstream, then
downstream and a final upstream sampling. The typical measurement time of a sampling was 120 s, corresponding
to an order of magnitude of 105 counted particles, but the measurement time was adjusted to ensure a good
determination of the collection efficiency: for low flow rate, a longer measurement time was necessary, due to low
loading with high diameter particles.

6.1. Cut-off diameter

For all cyclones of the SFYB family C-20 to C-45, collection efficiency curves were determined with 10 values
of the flow rate. The flow rate values were adjusted to cover a range from 0.5 to 20 mbar, which justifies the higher
flow rate values when the cyclone diameter increases from 20 to 45 mm. We present in figure 8 the corresponding
measurement results by a representation of the evaluated cut-off diameter as a function of the flow rate. The solid
points correspond to measurements and solid curves to power law trend lines.
5,5
5
dae,50 (µm) C-20
4,5 C-25
4 C-30
3,5 C-35
3 C-40
2,5 C-45
2
1,5
1
0,5 Qv (Nl/min)
0
0 50 100 150 200
figure 8 : Cut-off diameter as a function of the flow rate.
11
 Since inertial force is the primary mechanism for cyclone separation, we observe a logical decrease of the cyclone
cut-off particle size with the increase of the flow rate, which is attributed to the stronger centrifugal force created
for the higher flowrates. The same for a given flowrate: we observe a reduction of the cut-off diameter when reducing
the cyclone diameter, which is attributed to an increase of the internal velocity thus producing higher centrifugal
forces and improving the collection efficiency.

These cut-off diameter values were converted into dimensionless cut-off aerodynamic diameter  50 (eq. (5)), to
be plotted as a function of the annular Reynolds Number Reann. We present our measurements in figure 9, together
with measurements reported by Lidén & Gudmundsson (1997), and with the correlation proposed by Kuo & Tsai
(2001).
1E-03

C-20 C-30
50
C-35 C-40
1E-04 C-45 C-25

Stairmand – wall-attached periodic

1E-05 Stairmand – bottom-attached turbulent
Lapple
Z
SRI
ln(50)=-3.46-0.72 ln(Reann), Kuo & Tsai (2001)
1E-06
5E+02 5E+03 5E+04 Reann
figure 9 : Moore-McFarland's particle cut-off size parameter 50 versus Reann based on cyclone annulus.

We can conclude from figure 9 that our experimental determinations of the dimensionless cut-off aerodynamic
diameter are in quite good agreement with previous measurements, and an interpolation can be proposed, with a
regression coefficient r2=0,984 and a residual standard deviation of 0.060, for the Reynolds number range of this
experimental study:
ln ( 50 )  2.496  0.842 ln ( Reann ) , with r2=0,984 and residual standard deviation=0.060 (19)

We present our measurements in figure 10, together with this regression. To establish that there is no extra cyclone
size influence on the non-dimensional cut-size 50, we carried out a regression analysis incorporating the
dimensionless number (Do/Dt,mod), expressed as ln ( 50 )  A  k ln ( Reann )  m ln ( Do /Dt,mod ) . It should be noted
that in this statistical analysis Dt,mod could be exchanged for any length in the experimental setup that is kept constant
during the experiment, and the overall statistical result would be the same. The regression results in a regression
coefficient r2=0,986, and a residual standard deviation of 0.057, with A=-2.663, k=-0,83, and m=-0,08. We can
conclude to a non-significant coefficient for the extra ratio of dimensions.

12
 1,E-03

50

1,E-04
C-20 C-25
C-30 C-35
C-40 C-45
ln(50)=-2.496-0.842 ln(Reann)
Reann
1,E-05
1000 10000
figure 10 : Cut-off size parameter 50 versus Reann with the improved measurement procedure.

Kenny & Gussman (1997) investigated the variation of the cut-size as a function of the flow rate, for two groups
of cyclones (SRI and GK) of different sizes, but fully scaled for the GK family (except the GK 1.52X due to an error
during manufacturing). We compared our measurements to their results in figure 11, and as observed, the slope and
intercept of the linear regression between ln(d50/D) and ln(Reann) determined by Kenny & Gussman for the GK
family are in good agreement with our determinations.

1,E-03

d50/D

1,E-04
SRI-IIIa
SRI-V
GK(lp)
GK(sp)
GK1.52X Reann
1,E-05
1000 10000
figure 11 : Comparison of dimensionless cut-off diameter with existing data.

We also compared our measurements to Muschelknautz’s model (eq. (3)), and for each individual combination of
flow rate and cyclone body size, we calculated the correction factor xfact,MM that should be applied to force the
computed cut-point to match our experimental determinations. These correction factor values are plotted in figure
12, as a function of the annular Reynolds number.

13
 xfact,MM

C-20 C-25

C-30 C-35

C-40 C-45
ln(xfact,MM)=1,896-0.261 ln(Reann)
Reann
0,1
1000 10000
figure 12 : Muschelknautz’s model correction factor xfact,MM, as a function of the Reynolds number.

As observed in figure 12, the value of the xfact,MM correction factor is correlated to the Reynolds number, and an
interpolation can be proposed, with a regression coefficient r2=0,814 and a residual standard deviation of 0.069:
ln ( xfact,MM )  1.896  0.261 ln ( Reann ) , with r2=0,814 and residual standard deviation=0.069 (20)

This strong correlation between the value of the correction factor xfact,MM and the annular Reynolds number seems
to be an indication that the influence of the annular Reynolds number is not correctly taken into account within
Muschelknautz’s model. Based on our experiment, the range of xfact,MM extends down to almost half the value of the
lower limit stated by Hoffmann and Stein (2002), 0.56 versus 0.9.
We carried out a regression analysis incorporating the dimensionless number (Do/Dt,mod), expressed as
ln ( xfact,MM )  A  k ln ( Reann )  m ln ( Do /Dt,mod ) . We observe an increase of the regression coefficient r2=0,893, and
a decrease of the residual standard deviation down to 0.052 (with k=-0,303, and m=0,18). This result indicates that
there might be an effect of the cyclone body size over the correction factor xfact,MM. But it is difficult to conclude
since the friction factor f based on an evaluation of the absolute roughness has a large influence on the cut-off
diameter estimated using Muschelknautz’s model. For cyclone C-45 and a Reynolds value of 7540, the estimated
value of xfact,MM is 0,732, and an increase of the estimated absolute roughness of 75 µm instead of 50 µm would lead
to a reduction of 10% of the correction factor, with xfact,MM=0,660.
Our experimental determinations of the cut-off diameter were also compared to the results obtained by application
of the model developed by Mothes and Löffler (1988), based on the work of Dietz (1981). This model considers a
diffusive flux of particles between the inner and outer part of the swirl. In the original paper, Mothes and Löffler
suggest the use of a constant diffusivity Dp=0.0125 m2.s-1. We evaluated the correction factor xfact,ML that should be
applied to force the computed cut-point to match our experimental determinations. These correction factor values
are plotted in figure 13, as a function of the annular Reynolds number. As observed, this model also overpredicts
the cut-off diameter when compared to our experiments. But the range (0,66< xfact,ML <0,94) of the correction factor
factor xfact,Ml to Mothes and Löffler’s model is more limited than for Muschelknautz’s model (0,56< xfact,MM <1,05).

14
 xfact,ML

C-20 C-25

C-30 C-35

C-40 C-45

ln(xfact,ML)=0.663-0.112 ln(Reann) Reann

0,1
1 000 10 000
figure 13 : Comparison of experimental determination of the cut-off diameter to Mothes and Löffler’s model.

The value of the xfact,ML correction factor is also correlated to the Reynolds number, and an interpolation can be
proposed, with a regression coefficient r2=0,568 and a residual standard deviation of 0.054:
ln ( xfact,ML )  0.663  0.112 ln ( Reann ) , with r2=0,568 and residual standard deviation=0.054 (21)

6.2. Collection efficiency

For every cyclone C-20 to C-45, collection efficiency curves were determined for 10 values of the flow rate.
Collection efficiency curves were plotted against  (eq. (13)), the slip-corrected particle size normalised to the
experimentally determined slip-corrected cut-off sizes. As illustrated in figure 14 for an intermediate cyclone size
C-35, a linear Blachman-Lippmann (eq. (14)) model can be used to interpolate the data, with a reasonable scattering.
In this linear version expressed as ln( E /( 1  E ))  W0  W1 , we observed that W0 could be simply fixed to 0, and
we only tuned the slope W1, which is a measure of the cyclone separator “sharpness”, to obtain a best fit of the
measurements.
100%

90%
Collection efficiency (%)

80% Blachman-Lippmann
Qv=24 Nl/min
70%
Qv=30 Nl/min
60% Qv=37 Nl/min
Qv=44 Nl/min
50% Qv=52 Nl/min
Qv=61 Nl/min
40%
Qv=71 Nl/min
30% Qv=82 Nl/min
Qv=96 Nl/min
20% Qv=116 Nl/min

10%


0%
-1 -0,75 -0,5 -0,25 0 0,25 0,5 0,75 1

figure 14 : Cyclone collection efficiency versus normalized particle size  for the 35 mm diameter cyclone C-35.

For each of the six cyclones C-20 to C-45, the value of this slope W1 was determined. Finally, we present in figure
15 each of the six best fit curves, together with a cloud of points of all the experimental determinations of the
collection efficiency.

15
 100%

90% Collection efficiency (%)

80%

70%

60%

50%

40%

30%

20%

10%

0%
-1 -0,75 -0,5 -0,25 0 0,25 0,5 0,75  1
figure 15 : Cyclone collection efficiency versus normalized particle size , for the six cyclones C-20 to C-45.

We observe in this figure 15 a change in the dimensionless slope W1 of the collection efficiency curve when
varying the cyclone body diameter. But as already observed in figure 10, when changing the cyclone size and the
flow rate range, we also modified the Reynolds number range. So we determined the value of this dimensionless
slope W1 for each individual combination of flow rate and cyclone body size, to plot it as a function of the annular
Reynolds number. As observed in figure 16, the dimensionless slope W1 can obviously be correlated to the Reynolds
number, with a decrease of the cyclone “sharpness” when increasing the Reynolds number.
10

W1

C-20 C-25
C-30 C-35
C-40 C-45
ln(W1)=3.263-0.181 ln(Reann)
Reann
1
1000 10000
figure 16 : Dimensionless slope W1 as a function of the annular Reynolds number.

An interpolation can be proposed, to evaluate the value of this dimensionless slope W1:
ln (W1 )  3.263  0.181 ln ( Reann ) , with r2=0.662 and residual standard deviation=0.079 (22)

The values of the two coefficients in Eq. (22) were obtained by the least squares method of all experimental results,
with a regression coefficient r2=0,622 and a residual standard deviation of 0.079. We carried out a regression
analysis incorporating a dimensionless ratio (Do/Dt,mod), expressed as ln (W1 )  A  k ln ( Reann )  m ln ( Do /Dt,mod ) .
But the increase of r2=0,681 and decrease of the residual standard deviation down to 0.073 are limited, and do not
permit to conclude to an influence from the experimental setup on the results.
With this correlation (22), we can estimate to 1.13 the reduction factor of the slope of the non-dimensional
separation curve if the flow through the cyclone separator is doubled. Our experimental determinations of the slope
W1 are consistent with previous data reported by Lidén & Gudmundsson (1997), although these data were too
scattered to make a conclusion on the dependence of the slope W1 on the Reynolds number.
With the proposed correlations (19) and (22) for the evaluation of the dimensionless cut-off aerodynamic diameter
and dimensionless slope W1 respectively, it is possible to evaluate the collection efficiency E, using the transformed
efficiency ln( E /(1  E ))  W1 Ξ for given values of the annular Reynolds number and particle diameter. However,
16
we observe in figure 6 that for high flow rates and for small particle sizes, collection efficiency curves have non-
zero collection efficiency. We investigated this point to determine if, within the experimental uncertainty, the
transformed separation curve ln ( E / (1  E)) has a significant deviation from the linearity, described by W1*.
The experimental collection efficiency E for a given particle diameter is defined as E  (ni  no ) / ni , with ni and
no the counted particle numbers with the upstream and downstream sampling respectively. For each cyclone, we
measured the upstream number concentration 10 times. So it is possible to estimate the mean and standard deviation
as a function of the particle diameter. Since the generators produce a small amount of large diameter particles, we
observe an increase of the standard deviation when the particle diameters are higher than 4~5 µm. However, based
on our measurements, the average relative standard deviation of ni remains below  5% for particles with diameter
lower than five microns. Considering the same relative standard deviation for ni and no , and using the Gauss formula
for error propagation for these variables considered as independent, we were able to determine the experimental
uncertainty of the transformed efficiency. These uncertainties are reported in figure 17, for both the maximum and
minimum values of the flowrate, since the slope of the transformed efficiency is a function of the Reynolds number
(eq. (22)).
7,0
ln(E/(1-E))
Qv=24 Nl/min Qv=30 Nl/min
Qv=37 Nl/min Qv=44 Nl/min
Qv=52 Nl/min Qv=61 Nl/min
Qv=71 Nl/min Qv=82 Nl/min
Qv=96 Nl/min Qv=116 Nl/min
0,0
-1 -0,5 0 0,5 1


linear interpolation, for Qv=24 Nl/min

linear interpolation, for Qv=116 Nl/min

-7,0
figure 17 : Transformed efficiency and experimental uncertainties, for the C-35 cyclone.

From figure 17, we could suspect an increase of the collection efficiency for small particle sizes, which is a
phenomenon described by Hoffmann & Stein (2002) as a “tail”. Cortés & Gil (2007) reported that this tail was a
common observation, which could be attributed to an “adhesion of the small particles to larger ones”, or to the
“sweep of the small particles by the larger particles inside the vortex”. However, it is not possible to state that there
is actually such an increase of the collection efficiency for small particle sizes in this study, since there are almost
no measurement points outside the measurement uncertainty of the linear interpolations.

6.3. Pressure losses

In figure 18, we present a comparison between the experimental determinations of pressure losses obtained with
the set-up described in figure 5, together with the results obtained by using Muschelknautz’s model, for the six
cyclones of this SFYB family. For this confrontation, we used the experimental procedure described earlier, with an
imposed pure gas flow rate, a measure of the upstream static pressure corrected as an absolute pressure, and letting
the cyclone exhaust to the atmosphere. For the determination of the pressure loss as evaluated with Muschelknautz’s
model, we applied equations summarized in Table 1.

17
 20
P (mbar)

15

10

Experiments
5 C-20 C-25 C-30
C-35 C-40 C-45

Solid curve : Muschelknautz model

0
0 25 50 75 100 125 150 175 Q200 225
v (Nl/min)

figure 18 : Pressure drop - experimental results compared with Muschelknautz’s modeling.

We can see that a good agreement is obtained between our measurements and the pressure losses evaluated with
Muschelknautz’s model, on the entire flow range. From eq. (16), we can determine the experimental values of the
non-dimensional number for the pressure drop, the Euler number Euin, and plot them as a function of the annular
Reynolds number in figure 19.
30,0

ln(Euin) = -0.648+0.319 ln(Reann)

Euin C-20
C-25
C-30
C-35
C-40
C-45

Reann
3,0
1000 10000
figure 19 : Euler number.

We observe that the Euler number is highly correlated to the Reynolds number. An interpolation can be proposed,
to evaluate the Euler number:
ln ( Euin )  0.648  0,319 ln ( Reann ) , with r2=0.950 and residual standard deviation=0.040 (23)

The values of the two coefficients in eq. (23) were obtained by the least squares method, with a regression
coefficient r2=0,950 and a residual standard deviation of 0.040. During the pressure drop measurements, the cyclone
is directly vented to the ambient atmosphere (see figure 5), and there seems to be only one possible external
dimension that might influence the measurements, namely, the circular tube with 10 mm inner diameter upstream
of Section 1. So we carried out a regression analysis incorporating the dimensionless number (Do/Dt), expressed as
ln ( Euin )  A  k ln ( Reann )  m ln ( Do /Dt ) . The regression results in a regression coefficient r2=0,979, and a
residual standard deviation of 0.026. These variations indicate that there might be a remaining influence of the set-
up on the measured value of the Euler number, which should be further investigated.

18
7. Synthesis

A new set of six cyclones (SFYB family) have been developed, which are fully homothetic. They have been tested
on a wide flow rate range, with the only constraint of a limited pressure drop of 20 mbar, which is a current limit in
blow-by separation applications, for a corresponding flowrate range of 10 to 200 Nl.min-1.
The aim of the study was to provide measurements for collection efficiency curves, corresponding cut-off
diameters, and pressures losses. The conclusions of this experimental study can be summarized as follows.
1. We were able to produce measurements of the dimensionless cut-off aerodynamic diameters that compare
well to previous measurements. We also produced correlations that fit these measurements, and that would
be reliable tools for practical sizing engineering applications.
2. We determined that in the Muschelknautz model for the evaluation of the cut-off diameter, the correction
factor xfact,MM is directly correlated to the annular Reynolds number. A correlation of the factor as a function
of the Reynolds number has been proposed.
3. One of the most important findings emerging from the results of this study is that the dimensionless slope W1
of the collection efficiency curve can be correlated to the annular Reynolds number, which is a result that had
not been clearly established to our knowledge.
4. The present research has led to correlations between the annular Reynolds number and both the cut-off point
and the collection efficiency dimensionless slope. A model for the pressure loss evaluation has been validated.
These results will permit a designer to determine the size of a cyclone which will satisfy a target flow rate,
cut-point requirement and corresponding pressure loss.
5. We were able to detect an unwanted influence from the initial experimental setup on the measured results,
detailed in Annex 1. This led to a redesign of the setup, which we describe in the main core of the manuscript.
However, it seems that there still is something in the setup that influences the results with an extra factor
(D/L)a in the dependence of the non-dimensional slope of the collection efficiency and the Euler number on
the Reynolds number. L is a dimension outside of the cyclone, but which is so far unknown. Further
investigations will be carried out to investigate the point.

Acknowledgements

This project was supported by Toyota Motor Europe. The authors would like to thank Franck Faux and Michel
Mailly for making the cyclones. Thanks to Joël Giardi for assistance with experimental testing, to Franck Buron for
productive discussions on cyclone modelling. Thanks to the different students from ESTACA who contributed to
the project. The authors also wish to thank Prof. Göran Lidén for his useful comments and suggestions.

Nomenclature

a first regression coefficient (dimensionless cut-off aerodynamic diameter), dimensionless

Ai regression coefficient (Lidén & Gudmundsson (1997)), dimensionless
b second regression coefficient (dimensionless cut-off aerodynamic diameter), dimensionless
Bi cyclone dimensional ratios (Lidén & Gudmundsson (1997)), dimensionless
Ccmd Cunningham-Millikan-Davies coefficient, dimensionless
co ratio of particle mass flow to the air mass flow, dimensionless
d particle geometric diameter, (m)
dae particle aerodynamic diameter, (m)
d50,MM cut-off diameter evaluated using Muschelknautz’s model, see eq. (3), (m)
D cyclone body diameter, see figure 1, (m)
Do cyclone vortex finder diameter, see figure 1, (m)
Dt 10 mm tube diameter used in connecting parts, see figure 2, (m)
Dt,mod Dt,mod=30 mm, large tube diameter in improved set-up, see figure 20, (m)
Dp particle diffusivity (Mothes and Löffler’s model), (m2.s-1)
E collection efficiency, dimensionless
Euin Euler number, see eq. (16), dimensionless
f wall friction factor in Muschelknautz’s model, see Table 1, dimensionless
H height of cyclone inlet slit, see figure 1, (m)
Kn Knudsen number, dimensionless
L characteristic cyclone length, (m)
Lb height of cylindrical part of cyclone, see figure 1, (m)
Lc height of conical part of cyclone, see figure 1, (mm)

19
 Qv volumetric flow rate, Nl.min-1 (Normal Liter per Minute, Normal conditions being 0°C and 1.01325 Pa)
Rin radial position of the center of the inlet in Muschelknautz’s model, see Table 1, (m)
Rm geometric mean radius in Muschelknautz’s model, see Table 1, (m)
ReR cyclone body Reynolds number in Muschelknautz’s model, see Table 1, dimensionless
Reann annular cyclone Reynolds number, dimensionless
Re Reynolds number, dimensionless
S vortex finder length, see figure 1, (m)
Sf friction surface in Muschelknautz’s model, see Table 1, (m2)
St Stokes number, see eq. (1), dimensionless
U characteristic gas velocity in cyclone, (m.s−1)
uin average gas velocity at cyclone inlet, (m.s−1)
vo outlet axial velocity in the vortex finder, in Muschelknautz’s model, see Table 1, (m.s−1)
vzw “wall axial velocity” in Muschelknautz’s model, see Table 1, (m.s−1)
vCS “spin velocity near the wall” in Muschelknautz’s model, see Table 1, (m.s−1)
vm “geometrical mean rotational velocity” in Muschelknautz’s model, see Table 1, (m.s−1)
W width of cyclone inlet slit, see figure 1, (m)
W1 dimensionless slope, see eq. (15), dimensionless
xfact,MM correction factor in Muschelknautz’s model, see eq. (3), dimensionless

Greek letters

 entrance “constriction coefficient” in Muschelknautz’s model, see Table 1, dimensionless
 collection efficiency, dimensionless
1 steepness parameter for cyclone efficiency, see eq. (15), dimensionless
µg gas dynamic viscosity, (kg.m−1.s−1)
g gas density, (kg.m−3)
p particle density, (kg.m−3)
b cyclone body pressure loss coefficient, based on vo in Muschelknautz’s model, see Table 1, dimensionless
o outlet tube pressure loss coefficient, based on vo in Muschelknautz’s model, see Table 1, dimensionless
 W/(0.5D) in Muschelknautz’s model, see Table 1, dimensionless
50 dimensionless cut-off aerodynamic diameter or cut-off parameter, see eq. (5), dimensionless

Subscripts
50 at 50% efficiency
in at the inlet conditions
MM corresponding to Muschelknautz’s model, see Table 1
ML corresponding to Mothes & Löffler’s model

Annex 1: description of the initial set-up, and impact on the measurement results

In an initial step of this experimental study, we used a configuration with the same sampling probes for both
upstream and downstream sampling, so using a convergent conical part to connect Section 3 with the constant
diameter tube (Dt=10 mm) of the downstream sampling probe (see left side of figure 20).

20
 Downstream
sampling

Dt, mod=30 mm
20 cm
Dt=10 mm

Section 3
0,2 Lo
Do
Lo Lo
Static pressure Do
tap position

S S

D D

Initial set-up Improved set-up

figure 20 : Correction of the downstream sampling procedure.

We present our experimental determinations of the dimensionless cut-off diameter obtained with this initial set-
up in figure 21, together with the correlation proposed by Kuo & Tsai (2001). We observe that it is not possible to
fit all the measurements with a unique correlation, and it seems that the annular Reynolds number does not explain
all the variations of  50 .

50

1,E-04

C-20 C-25
C-30 C-35
C-40 C-45
ln(50)=-3.46-0.72 ln(Reann), Kuo & Tsai (2001)
Reann
1,E-05
1000 10000
figure 21 : Cut-off size parameter 50 versus Reann based on cyclone annulus.

The evident extra dependence of the non-dimensional cut-size on the cyclone size has been investigated. All the
internal dimensions of each cyclone have been scaled up to its internal body diameter D, but as it is clearly visible
on left side of figure 20, we have a conical part at the cyclone outlet, connecting the cyclone outlet with internal
diameter Do to the sampling probe, which has a constant inlet diameter Dt =10 mm. We report in Table 4 the
dimensionless ratio Do/Dt for the different cyclone sizes.

Table 4 Do/Dt values in the initial set-up

Cyclone C-20 C-25 C-30 C-35 C-40 C-45
Do /Dt 0.9 1.1 1.3 1.6 1.8 2

21
 Though being external to the cyclones and downstream the vortex finder, the connection is part of the experiment.
When we increase the cyclone size, the ratio Do/Dt increases, and the acceleration of the flow might be the reason
for an increase of the intensity of the vortex in the outlet tube, which would justify the observation of a reduction of
the dimensionless cut-off diameter when the cyclone size increases, for a fixed annular Reynolds value.
In order to check this idea, we carried out a regression analysis incorporating this dimensionless number (Do/Dt).
ln ( 50 )  A  k ln ( Reann )  m ln ( Do /Dt ) (24)

To produce a linear plot in a logarithmic-logarithmic diagram, we can rewrite this equation as

ln ( 50 )  A  k ln ( Reann  ( Do /Dt ) m / k ) .
The regression results in a regression coefficient r2=0,987 and a residual standard deviation of 0.061. If only the
Reynolds number is incorporated in the regression, the regression coefficient decreases down to 0.949, and the
residual standard deviation is multiplied by two, equal to 0,120
The corresponding result is shown in figure 22. It can be seen that there is no longer any systematic bias due to
the cyclone size between measured values and the correlation.
1,E-03

50

1,E-04

C-20 C-25
C-30 C-35
C-40 C-45
ln( 50)=-2.32-0.857 ln(Reann)-0.42 ln(Do/Dt) Reann*(Do/Dt)0.5
1,E-05
1000 10000

figure 22 : Cut-off size parameter 50 versus Reann  ( Do /Dt )0.5 .

As it was clearly established that this initial measurement set-up was generating an extra cyclone size dependence,
an improved version of the measurement procedure was used, with a modification of the downstream connection
and sampling system, as detailed in part 3 (figure 1). It should be noted that a similar extra dependence (Do/Dt)0.5
was found for the non-dimensional slope of the collection efficiency, which is not shown.

References

Barth, W. (1956) Berechnung und Auslegung von Zyklonabscheidern. Brennstoff wärme kraft 8, 1-9.
Beeckmans, J. M. and Kim, C. J. (1977) Analysis of the efficiency of reverse flow cyclones. Canadian Journal of
Chemical Engineering 55, 640-643.
Blachman, M. W. and Lippmann, M. (1974) Performance characteristics of the multicyclone aerosol sampler. Am.
Ind. Hyg. Ass. J. 35, 311-326.
Cortés, C. and Gil, A. (2007) Modeling the gas and particle flow inside cyclone separators. Progress in Energy and
Combustion Science 33, 409-452.
Dietz, P. W. (1981) Collection Efficiency of Cyclone Separators. American Institute of Chemical Engineers Journal
6, 888-892.
Dirgo, J. and Leith, D. (1985) Cyclone Collection Efficiency: Comparison of Experimental Results with Theoretical
Predictions. Aerosol Science and Technology 4, 401-415.
Dirgo, J. and Leith, D. (1985) Performance of theoretically optimised cyclones. Filtration and Separation
March/April issue :119-125.
Friedlander, S. K. (2000) Smoke, Dust and Haze. Fundamentals of Aerosol Dynamics 2nd Edition. Oxford, New
York.
Heim, M., Mullins, B., Umhauer, H., & Kasper, G. (2008). Performance evaluation of three optical particle
counters with an efficient ‘‘multimodal’’ calibration method. Journal of Aerosol Science, 39, 1019–1031.
22
 Hoffmann, A. C. and Stein, L. E. (2002) Gas Cyclones and Swirl Tubes. Principles, Design and Operation. Springer,
New York.
Hsiao, H. K. (1997) Accuracy of a Personal Respirable Dust Sampler, Master Thesis, Institute of Environmental
Engineering, National Chiao Tung University.
Hsiao, T.-C. Chen, D.-R. and Son, S.Y. (2009) Development of mini-cyclones as the size-selective inlet of miniature
particle detectors. Journal of Aerosol Science 40, 481-491.
Iozia, D. E. and Leith, D. (1989) Effect of Cyclone Dimensions on Gas Flow Pattern and Collection Efficiency.
Aerosol Science and Technology 10, 491-500.
John, W. and Reischl, V. (1980) A cyclone for Size-Selective Sampling. Journal of the Air Pollution Control
Association 30, 872-876.
Kenny, L.C. and Gussman, R.A. (1997) Characterization and modelling of a family of cyclone aerosol preseparators.
Journal of Aerosol Science 28 (4), 677-688.
Kim, J. C. and Lee, K. W. (1990) Experimental Study of Particle Collection by Small Cyclones. Aerosol Science
and Technology 12, 1003-1015.
Kuo, K.-Y. and Tsai, C.-J. (2001) On the theory of particle Cutoff and Collection Efficiency of Cyclones. Aerosol
and Air Quality Research 1, 47-56.
Lapple C. E. (1950) Gravity and centrifugal separation. Ind Hyg Quart, 11:40–7.
Leith, D. and Licht, W. (1972) The collection efficiency of cyclone type particle collectors-A new theoretical
approach. American Institute of Chemical Engineers Symposium Series 68, 196-206.
Lidén, G. and Gudmundsson, A. (1997) Semi-Empirical modelling to generalise the dependence of cyclone
collection efficiency on operating conditions and cyclone design. Journal of Aerosol Science 5, 853-874.
Lin, Q. R. (1996) Estimating the Collection Efficiency of Electro-Cyclones. Master Thesis, Institute of
Environmental Engineering, National Chung Hsing University.
Maynard, A. D. and Kenny, L. C. (1995) Performance assessment of three personal cyclone models, using an
aerodynamic particle sizer. J. Aerosol Sci. 26, 671-684.
Mothes, H. and Löffler, F. (1988) Prediction of particle removal in cyclone separators. Int. Chem. Eng., 28, 231–
240.
Moore, M. E. and McFarland, A. R. (1990) Design of Stairmand-type sampling cyclones. American Industrial
Hygiene Association Journal 51, 151-159.
Moore, M. E. and McFarland, A. R. (1993) Performance modeling of single- inlet aerosol sampling cyclones.
Environmental Science and Technology 27, 1842-1848.
Moore, M. E. and McFarland, A. R. (1996) Design Methodology for Multiple Inlet Cyclones. Environmental Science
and Technology 30, 271-276.
Muschelknautz, E. (2010) L3.4 Cyclones for the Precipitation of Solid Particles. 1226-1237, VDI Verlag GmbH
Heat Atlas, 2ème. ed, Ed Springer, Düsseldorf.
Muschelknautz, E. and Kambrock W. (1970) Aerodynamische Beiwerte des Zyclon abscheiders aufgrund neuer und
verbesserter Messungen. Chem. Ing. Technik, 42, 247–255
Muschelknautz, E. and Trefz, M. (1990) Design and calculation of higher and highest loaded gas cyclones.
Proceedings of Second World Congress on Particle Technology, Kyoto, Japan, pp. 52–71.
Overcamp, T. J. and Scarlett, S. E. (1993) Effect of Reynolds number on Stokes number of cyclones. Aerosol Science
and Technology 19, 362-370.
Saltzman, B. E. and Hochstrasser, J. M. (1983) Design and performance of miniature cyclones for respirable aerosol
sampling. Environmental Science and Technology 17, 418.
Stairmand, C. J. (1951) The design and performance of cyclone separators. Transactions of the Institution of
Chemical Engineers 29, 356-383.
Trefz, M. and Muschelknautz, E. (1993) Extended cyclone theory for gas flows with high solids concentrations.
Chemie Ingenieur Technik 16, 153–160.
Tsai, C. J. Shiau, H. G. Lin, K. C. and Shih, T. S. (1999) Effect of Deposited Particles and Particle Charge on the
Penetration of Small Sampling Cyclones. Journal of Aerosol Science 30, 313-323.
Zhu Y. and Lee K.W. (1999) Experimental study on small cyclones at high flowrates. Journal of Aerosol Science
30, 1303-1315.

23

View publication stats