Powder Technology 297 (2016) 106–114

Contents lists available at ScienceDirect

Powder Technology

journal homepage: www.elsevier.com/locate/powtec

Modelling of hydrocyclone classifiers: A critique of ‘bypass’ and

corrected efficiency
K. Nageswararao
Taijasa Consultants, 10-116, Visalakshi Nagar, Visakhapatnam 530 043, India

a r t i c l e i n f o a b s t r a c t

Article history: In early 1950s, Kelsall visualised the classification process in which a fraction of feed solids reaches underflow
Received 3 February 2016 without undergoing classification in a hydrocyclone. We bring to light that this notion of ‘bypass’ is beyond ex-
Received in revised form 29 March 2016 perimental reach and is purely imaginary. This implies that the corrected efficiency, which is a key performance
Accepted 9 April 2016
characteristic in all the successfully used hydrocyclone models and for which bypass is the foundation is also
Available online 16 April 2016
hypothetical.
Keywords:
We discuss the reasons for the continued acceptance of a physical meaning to ‘bypass’ among the hydrocyclone
Classification practitioners. Following a discussion on why the classification process as visualised by Kelsall is purely notional,
Actual efficiency we bring to light the lack of an experimental basis for ‘bypass’ and why no physical meaning can be attributed to
Centrifugal efficiency it. A mathematical interpretation of the method suggested by Kelsall for calculating corrected efficiency is also
Reduced efficiency presented.
Normalisation A general method for normalising any function y = f(x) where x and y vary between xmin to xmax and ymin to ymax
Empirical models respectively into Yn(x) where the range of Yn is from 0 to 1, is then proposed. We show that this normalisation
can be done in an infinite number of ways by choosing user defined normalising functions and demonstrate
our method with a numerical example. Taking into consideration that classifications function is a special case
of y = f(x), we show that it too can be normalised in an infinite number of ways. While illustrating our method
with examples, we show that the procedure suggested by Kelsall is the simplest method to normalise actual ef-
ficiency curves. The similarity between bypass and normalising functions is illustrated with examples; it is shown
that both of them are purely notional and have no physical significance.
We bring to light the possibility of development of improved hydrocyclone models using a normalised cut size
other than Kelsall cut size. Finally, we propose that the corrected efficiency be denoted as Kelsall efficiency and
the corrected classification size as Kelsall cut size, as a tribute to the landmark contribution of Kelsall towards
modelling of hydrocyclones.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction Simultaneously, there is an increased consensus among the mineral

A conspicuous feature of the efficiency curve, of a hydrocyclone clas-
sifier, that is, the size-recovery relationship of the feed particles to the
coarser product, is its non-zero minimum value. To explain this regular
observation, Kelsall [1] proposed an innovative and ingenious explana-
tion that a fraction of feed solids ‘bypass’ to underflow without undergo-
ing cyclone action. Based on his conceptualisation of the classification
process, a hypothetical ‘centrifugal efficiency’, more commonly referred
to as the corrected efficiency, Ec(d), could be calculated for which the
minimum starts at zero for near zero sized particles. Over the years,
the corrected efficiency curve and its derivative, the corrected cut size,
d50c have become standard parameters for comparing cyclone perfor-
mance at different design and operating conditions. They continue to
be key parameters in all the hydrocyclone models currently in use [2].
E-mail address: nageswararao.karri@gmail.com.
processors that bypass is an integral part of the true classification
process.
In this paper, we discuss how a perception regarding a physical
meaning to bypass evolved and strengthened. We bring to light that
the parameters such as tonnage of feed bypassing to underflow, feed
subjected to cyclone action and the feed reaching underflow due to cy-
clone action are only notional and cannot be determined experimen-
tally. We show that as such, bypass can only be notional as it is
derived from notional quantities and therefore cannot be part of the
classification process actually taking place within the cyclone. This is
followed by a mathematical analysis of the method suggested by Kelsall
for transforming the actual efficiency to corrected efficiency.
We propose a general method to normalise any function y = f(x)
where y varies between ymin to ymax and x between xmin to xmax to
Yn(x) such that the range of Yn is 0 to 1. We demonstrate that this
normalisation can be done in an infinite number of ways through user
defined normalising functions. It is shown that the technique for

http://dx.doi.org/10.1016/j.powtec.2016.04.016
0032-5910/© 2016 Elsevier B.V. All rights reserved.
 K. Nageswararao / Powder Technology 297 (2016) 106–114
normalisation of actual efficiency to ‘corrected efficiency’ developed
earlier [3] is only a special case of the general method.
We illustrate the normalisation procedure of a typical actual effi-
ciency curve by generating different normalised curves from it using dif-
ferent normalising functions. We show that the method suggested by
Kelsall is the simplest. The analogy between Kelsall bypass and normal-
ising functions is also illustrated with a numerical example.
Evidently, lack of an experimental basis for bypass is not a constraint
for the normalisation of actual efficiency curves and its continued usage
in the modelling of hydrocyclones. In fact, it is now possible to choose a
normalising function so that the normalised curve is closer the actual ef-
ficiency curve than the Kelsall curve. That is, there is a scope for explor-
ing the possibility of using a normalised cut size other than Kelsall cut
size and the corresponding reduced efficiency curve for development
of improved hydrocyclone models.
Finally we propose that it would be highly appropriate that the
corrected efficiency be denoted as Kelsall efficiency and the correspond-
ing cut size as Kelsall cut size, as a tribute to the landmark contribution
of Kelsall towards modelling of hydrocyclones,
2. Centrifugal/corrected efficiency
In the development of modelling of hydrocyclone classifiers,
‘corrected efficiency’, the notion conceived by Kelsall [1], is unquestion-
ably a landmark contribution. It paved the way for development of com-
prehensive mathematical models of hydrocyclone classifiers. In this
section, we trace its origin; practical utility and how the concept of ‘by-
pass’ started gaining acceptance as a true representation of the classifi-
cation process. We re-examine this notion based on fundamental
principles and show that it is only a mathematical transformation. The
method for calculation of corrected efficiency proposed by Yoshioka
and Hotta [4] is also explained.
2.1. Origin of bypass and corrected efficiency
The classification process as visualised by Kelsall [1] is shown in
Fig. 1. Here is an extract: ‘… A knowledge of cyclone operation leads to
the conclusion that if x% is the fraction of the total water which is
discharged through the underflow then, independent of any centrifugal
forces acting on particles, x% of all particles must leave through the
underflow. The additional percentage of particles of a given size which
are discharged through the underflow is a measure of the efficiency of elim-
ination of solids due to the cyclone action.…’
The above conceptualisation on the mechanism of classification is
the basis for transforming the actual efficiency curve which varies
from Rf to 1 to a hypothetical ‘centrifugal/corrected efficiency’ for
which the range is from zero to 1. The more common nomenclature
for this performance characteristic is the ‘corrected’ efficiency.
Fig. 1. The concept of ‘bypass’ and centrifugal efficiency in the words of Kelsall [1].
107
The corrected efficiency curve and the corrected cut size could be
used conveniently and effectively to compare the performance of cy-
clones at different design and operating conditions and accordingly re-
ceived acceptance without any reservations.
Notably, Yoshioka and Hotta [4] too recognised the necessity for a
performance characteristic for comparison of actual efficiency curves
and independently proposed a procedure to calculate the corrected effi-
ciency. They are also the first to report the reduced efficiency curve, a
plot of the corrected efficiency against dimensionless particle size (d/
d50c).
2.2. Physical significance to bypass
Bradley and Pulling [5] are among the earliest who concurred with
Kelsall and Yoshioka and Hotta that actual efficiency needs to be
corrected for comparative purposes; the notation used by them is cen-
trifugal efficiency. They proposed an equation for the corrected classifi-
cation size as well. Although, Yoshioka and Hotta [4] are the first to
develop a comprehensive hydrocyclone model for which the corrected
cut size (d50c) and the reduced efficiency curve are key performance
characteristics, an extensive and effective usage of these parameters is
attributable to Lynch and colleagues [6–9]. While the Lynch and Rao
[6] hydrocyclone model is based on extensive test work on large
hydrocyclones (15 cm–50.8 cm units) with concentrated feed slurries,
the base data for the model developed by Yoshioka and Hotta is ob-
tained with dilute suspensions and laboratory size units (7.6 cm–
15.2 cm units).
Successful application of the Lynch–Rao model [6–11] in the opera-
tion and control of grinding circuits in industry marked the beginning of
a wide spread use of corrected cut size d50c and the reduced efficiency
curve A schematic representation of bypass or short circuit of feed to
underflow as presented by them [9] is shown in Fig. 2.
The Lynch-Rao model paved the way for continued usage of the
bypass concept and corrected classification size in all hydrocyclone
models developed since then [12–23] etc. The continual success of
these models in industrial applications for example, [24–27]
strengthened the perception that bypass is a true representation
of the classification process within the hydrocyclone. A widespread
usage of commercial software packages such as JKSimMet and
Limn® could be another important reason why a physical signifi-
cance to bypass and ‘corrected’ efficiency became an integral part
of the conventional wisdom in hydrocyclone practice. It is apt to
say that the classification process continues to be perceived by the
mineral processing community as visualised by Kelsall [1] and
Lynch and Rao [9]. In fact, these days a ‘physical’ meaning to ‘bypass’
is so widespread that it is part of standard text books in mineral pro-
cessing for example [27].
108 K. Nageswararao / Powder Technology 297 (2016) 106–114
Fig. 2. A graphical representation of the classification mechanism in hydrocyclones
proposed by Lynch and Rao [9]. This ‘mechanistic model’ reinforced a physical meaning
to the notion of ‘bypass’ proposed by Kelsall.
2.3. Size dependent bypass
Kelsall assumed the bypass fraction to be equal to water recovery for
particles of all sizes. Significantly, he did not offer any justification why
it should be same for all sizes. Subsequently, when reports of non-
monotonic efficiency curves started appearing in literature, Finch [28]
suggested a size dependent bypass, His proposal could give a tangible
explanation to the occurrence of ‘fish hook’ in efficiency curves. Others
who refined the notion of size dependent bypass include Kelly [29],
Del Vilar and Finch [30], Roldan-Villasana et al. [31] and Kraipech et al.
[32]. Flintoff et al. [13] too advocated fundamental studies on this
issue. Napier-Munn and Lynch [33] also opined that size independent
bypass is intuitively unlikely.
Significantly, the basic notion of bypass was not questioned by any.
This further strengthened the perception of a physical meaning to
bypass.
2.4. No rationale for bypass
The near zero sized particles (or ultra fines, say of size ε) just flow
with water. As such, a fraction of those particles equal to water recovery
report to underflow. This is a common observation in hydrocyclone
practice. Based on this observation only, Kelsall conceived the notion
of bypass by postulating that a fraction of particles of all sizes equal to
water recovery must leave through the underflow.
While it is reasonable to expect that ultra fines flow along with
water to underflow, it is inconceivable that particles of all sizes also
flow along with it in the same proportion. There is no rational basis
for such a hypothesis. It is significant that, Kelsall did not offer any ex-
planation on why coarser particles in the feed too should flow along
with water in the same proportion as ultra fines. In spite of this crucial
limitation, the basic notion of bypass retained its ground, though assum-
ing it to be constant for particles of all sizes was questioned [28–33].
Attributing zero efficiency to near zero (ε) sized particles is a conten-
tious outcome of the notion of bypass. We emphasise here that it is due
to ‘cyclone action’ that water splits into two streams and so do the fines
along with it. It is therefore realistic to visualise the efficiency of these
ultra fine particles to be always equal to water recovery under all condi-
tions of cyclone operation. That is, based on phenomenological consid-
erations perfect separation of near zero sized particles suspended in
water is unattainable due to ‘cyclone action’. This means that it is im-
possible to obtain experimental evidence of zero efficiency for near
zero sized particles under any real conditions of cyclone operation.
This leads us to an unambiguous conclusion that the corrected efficiency
is not based on any physical process actually taking place in the cyclone
and therefore is only a hypothetical performance characteristic. Accord-
ingly, no physical meaning could be attributed to it. Evidently, the above
physical considerations are in total disagreement with the mechanism
proposed by Kelsall (Fig. 1) according to which all the near zero (ε)
sized particles reach underflow due to bypass and therefore the effi-
ciency of these particles is zero due to ‘cyclone action’.
It is important to bring to light that Kelsall did not explain explicitly
why the corrected efficiency of near zero sized particles should be zero.
As such, we can only surmise possible reasons for his visualisation. One
plausible explanation could be a consequence of the near zero settling
velocities of these particles. Taking into consideration the general con-
sensus among hydrocyclone practitioners that settling velocity and effi-
ciency are related, Kelsall presumably perceived that near zero settling
velocities result in zero efficiency of fines.
Whether Kelsall suggested bypass purely as a justification to convert
the efficiency curve so that it is zero at near zero sizes or he presumed
the efficiency of zero particles to be zero and therefore proposed bypass
mechanism is only a matter of speculation. What is of interest is that by-
pass cannot be verified experimentally; zero efficiency of near zero
sized particles is unrealistic; Kelsall did not provide any justification
whatsoever for either of them. Obviously, physical significance to by-
pass is as realistic as perfect separation of near zero sized particles.
Nevertheless, acceptance of zero efficiency for zero sized particles is
so widespread that it is not a topic of discussion or research in literature.
In fact, simplified physical models to represent classification proposed
by Plitt [34] and Reid [35] are in complete agreement with the visualisa-
tion of Kelsall with regards to zero efficiency of zero sized particles.
While Plitt explicitly assumed bypass, Reid assumed dependence of ef-
ficiency on particle size to arrive at zero efficiency for zero sized parti-
cles. Suffice it to say that a physically unattainable perfect separation
of zero sized particles is a key feature of both the models.
In this context, it is appropriate to mention the observations of Schu-
bert and colleagues [36–39] that the classification process with bypass
and the reduced efficiency curve as key features does not reflect the
physical phenomena actually taking place within the cyclone. It is note-
worthy that when the separation function is modelled taking the phys-
ics of the process into consideration, the efficiency of near zero sized
particles turns out to be equal to water recovery [36–39] in tune with
experimentally observed facts. It is relevant to mention here that the
theoretical study of Criner [40] also predicts an efficiency of Rf for near
zero sized particles. Calculations using the equation for ‘drift velocity’
of solid particles derived by Bloor and Ingham [41] also yield the same
result. Also, we may note that the results of CFD simulations [42–45]
are expected to show the actual efficiency and not the hypothetical
corrected efficiency, as the CFD models for simulation are formulated
on the basis of flow of fluid and solids within the hydrocyclone.
The observations of Bradley and Pulling [5] are also relevant in this
regard. They rightly remark that the actual efficiency curve does not
pass through the origin and observe that: ‘… a positive efficiency is
shown for particles approaching zero particle size which experience
no centrifugal concentration. Allowance for this must be made for corre-
lation purposes or for comparison of the effects of those design vari-
ables…’. However, they too offered no justification why zero sized
particles should have (or should be assigned) zero centrifugal efficiency.
Nevertheless, they are in agreement with the correction to actual effi-
ciency suggested by Kelsall.
It is noteworthy that Yoshioka and Hotta [4] recommended a pro-
portional correction to the actual efficiency curve to calculate the
corrected efficiency curve with a maximum correction of Rf for near
zero sized particles. Conspicuously, they visualised equating the effi-
ciency of near zero sized particles to zero as a mathematical procedure
only.
The above issues apart, when the notion of bypass is applied to the
feed stream we get a curve for which the corrected efficiency varies
from 0 to1. Hydrocyclone practitioners readily accepted this
 K. Nageswararao / Powder Technology 297 (2016) 106–114
hypothetical efficiency curve for comparing the cyclone performance at
different design and operating conditions; Kelsall's argument is also in-
genious and innovative. It was convincing enough to explore the possi-
bility of its usage for analysing cyclone performance. Successful
industrial application [6,10,11,17,27] etc ensured a continuous usage
of the notion of bypass and the corrected efficiency curve.
2.5. No experimental basis for bypass
A crucial issue which is totally ignored while accepting the notion of
bypass is the lack of experimental evidence to verify it. Experimental
data generally obtained from a hydrocyclone are the mass flow rates
of solids and water in the feed, underflow and overflow streams in ad-
dition to the particle size analyses of each stream. Obviously, from
these data we cannot determine the mass flow rates (or fraction) of
feed solids which reach underflow due to bypass and which are sub-
jected to cyclone action. Similarly, we can determine only the total
mass flow rates to underflow and not those due to bypass and cyclone
action separately Moreover, even if bypass actually took place, we can-
not distinguish which of the particles or group of particles in the feed
have actually bypassed. More significantly, we cannot even conceive
an experiment by which we can determine whether a particle (or a
group of particles) which reached underflow reached there through by-
pass or due to centrifugal action.
Further, it is simply inconceivable to visualise that the centrifugal
forces acting on solid particles in the cyclone selectively ‘bypass’ a frac-
tion of particles and classify the remaining particles only.
Furthermore, intuitively, we can say that if the water reporting to
undertow could carry a fraction of feed solids without allowing them
to be subjected to classification, there is no reason why we should not
expect that water reporting to overflow also does the same. Clearly, in
such a situation the cyclone operates as a splitter only.
In this context, the observation of Napier-Munn and Lynch [33] on
the lack of an experimental consensus for a suitable function to describe
the variation of bypass with size is highly relevant. We can conjecture
that our inability to experimentally determine the bypass fraction is
the only reason why a consensus could not be reached on this issue.
All the above considerations lead us to conclude unequivocally that
bypass is beyond experimental reach; it is purely imaginary. Conse-
quently, there is no justification to attribute any physical significance
to it.
That is, the procedure suggested by Kelsall to construct the corrected
efficiency curve is to be regarded as a simple mathematical transforma-
tion only. This means that the purely notional bypass with a numerical
value equal to the minimum value of Ea (which is also equal to Rf), is
only a mathematical object used in transforming the actual efficiency
curve.
It is worthwhile to mention in this context, the demarcation crite-
rion proposed by Popper [46,47] to distinguish empirical sciences
from mathematics, logic and metaphysics. According to him, any state-
ment/theory to be classified as scientific (that is belonging to empirical
sciences) should be capable of getting tested. That is, we should be able
to design an experiment, the result of which could influence our under-
standing of the statement/theory one way or the other. It is important to
note here that the crucial concern of this criterion of testability is not the
result of the experiment, that is, whether the theory is true or false; but
can we say if it is so (true or false) through an experimental technique.
Evidently, the notion of bypass fails to meet this criterion. We cannot
conceive any experimental programme using the results of which we
can conclude whether bypass is a constituent of the classification pro-
cess or not.
The above scrutiny based on the demarcation criterion of Popper
gives credence to our conclusion that what Kelsall proposed is simply
a mathematical transformation. A discussion on the steps involved in
this transformation follows.
109
2.6. Mathematical analysis of calculation of corrected efficiency
The procedure suggested by Kelsall for constructing the corrected ef-
ficiency curve can be seen as a mathematical transformation carried out
in two simple steps as elaborated below.
2.6.1. Step I: reducing Ea of all sizes by its minimum value
The range of efficiency as a result varies from 0 to 1- Rf.
Kelsall is innovative; instead of simply suggesting that the efficiency
be reduced by its minimum value (=Rf), he suggested that
(i) a fraction of particles of all sizes bypass to underflow;
(ii) the bypass fraction is same for all sizes;
(iii) the bypass fraction is equal to the recovery of near zero sized
particles.
We reiterate that all of (i) to (iii) above are beyond experimental
reach. Nevertheless, they are taken for granted to represent the classifi-
cation process. The fraction which reaches underflow due to cyclone ac-
tion is then calculated as (Ea − Rf).
2.6.2. Step 2: readjust the range of efficiency as 0 to 1
After step 1, the efficiency values range between 0 and (1 − Rf). This
needs to be adjusted so that the range is 0 to 1. It is done by dividing the
resultant values after step 1 by (1 − Rf).
Kelsall suggested calculation of the corrected efficiency from the
fraction which reaches underflow as a result of cyclone action (that is,
Ea − Rf) and the feed which is subjected to classification (1 − Rf).
We can note that the calculation of corrected efficiency is a simple
two step mathematical procedure only. However, Kelsall explained
this procedure in terms of purely notional quantities such as
(i) fraction of particles reaching underflow due to bypass (ii) fraction
of feed particles subjected to cyclone action, and (iii) fraction of feed
particles reaching underflow due to cyclone action. We reiterate that
it is impossible to determine experimentally any of the quantities, (i)–
(iii). Therefore, no physical meaning can be attributed to any of them
or to the steps involved in the calculation of corrected efficiency. Ac-
cordingly, the procedure suggested by Kelsall should be treated as is a
mathematical transformation only.
Nevertheless, as detailed in Section 2.2, there is a widespread belief,
though mistaken, that bypass is a part of the classification process. This
is a consequence of the highly successful usage of corrected cut size, d50c
and the reduced efficiency curve in hydrocyclone practice.
2.7. Method of Yoshioka and Hotta for calculation of corrected efficiency
Yoshioka and Hotta [4] independently proposed a method to calcu-
late corrected efficiency (reduced fractional recovery is the notation
used by them) for correlating actual efficiency (fractional recovery)
curves which differ for different flow ratios (Rf) and operating
conditions.
Their method of calculation is a simple mathematical procedure; the
corrected efficiency of near zero sized particles is equated to zero (a cor-
rection equal to Rf) and no correction is applied for those particles for
which Ea = 1. A proportional correction is applied for all intermediate
efficiencies as shown in Fig. 3. The corrections suggested follow Eq.(1):
AB=AC ¼ DE=DF ¼ 1−R f ⋯: ð1Þ
From Fig. 3 we can note that, DE and DF correspond to 1 − Ea(d) and
1 − Ec(d) respectively. Hence we can write:
1−Ea ðdÞ
¼ 1−R f ⋯: ð2Þ
1−Ec ðdÞ
110 K. Nageswararao / Powder Technology 297 (2016) 106–114

we ensure that:

NðYÞ ¼ Y min at Y ¼ Y min …: ð6Þ

The above condition (Eq.(6)) would ensure that when Y(x) is mini-
mum that is, when it is equal to Ymin, the normalised function Yn(x) = 0.
From Eq.(5), we can note that maximum value of Yn (x) shall always be
1, as the maximum value of Y(x) is 1 (Eq.(4)).
We illustrated earlier [3] a procedure for generation of normalising
functions, N (Ea) for Ea by assuming different functional forms of dN
(Ea)/dEa. Since Ea and Y are similar functions, an exactly similar proce-
dure for normalisation of Y can be adopted by simple substitution of Y
for Ea.
Evidently, the simplest method to normalise is by choosing N(Y) as
constant. In that case, N(Y) can only take the unique value equal to
Ymin for all values of Y. Then

Y ðxÞ−Y min
Yn ðXÞ ¼ ⋯: ð7Þ
1−Y min

We illustrate the normalisation procedure of a typical function

y = f(x) which has two local maxima and two local minima (Fig. 4a).
For this function, the values of y varying from −15 to 90 for x ranging
from 0 to 500 are shown in Table 1. Also shown in the table are the
Fig. 3. Yoshioka and Hotta [4] suggested calculation of corrected efficiency using the values of Y, the dimensionless y as calculated using Eq.(4) and the
identity AB/AC = DE/DF = 1 − Rf. normalised efficiency, Yn using Eq.(7) for each of the values of x.
These are shown graphically in Fig. 4b.
The above example clearly shows that any function y = f(x) can be
On rearrangement of Eq.(2) we get: normalised to Yn(x) so that its range is from 0 to 1. Similarity to the
method of Kelsall is noteworthy.
Ea ðdÞ−R f
Ec ðdÞ ¼ ⋯: ð3Þ
1−R f 3.3. Classifier efficiency curves

Classifier efficiency function, Ea(d) can be viewed as a special case in

Noticeably, Yoshioka and Hotta did not attribute any physical mean-
which the independent variable, x, corresponds to d, the particle size.
ing to their method of calculation of corrected efficiency; they visualised
Obviously, there is no need to go through Step I described in section 3
it as a simple mathematical transformation only.
above. We use the simpler notation of Ea for Ea(d) and Ec for its normal-
ised form Ec(d). The first step in the normalisation procedure of Ea(d) is
3. A general method of normalisation
to define a normalising function. Let this function be denoted as N {Ea(d)}
or in its simpler form as N(Ea).
Let y = f(x) be a function defined in xmin to xmax. Let ymin and ymax
We can now rewrite Eq.(5) as:
are the limiting values of y. A method to transform this function to
Yn(x) such that its range is from 0 to 1 is discussed below.
Ea ðdÞ−N ðEa Þ
Ec ðdÞ ¼ ⋯: ð8Þ
3.1. Step I: reducing y to a dimensionless form 1−NðEa Þ

Let Y(x) be defined as Clearly, the range of Ec(d) shall be between 0 and 1 if we set the
lower limit of the normalising function N(Ea) as equal to the minimum
YðxÞ ¼ f ðxÞ=y max ⋯: ð4Þ value of Ea(d).

If y = f(x) represents a dimensional equation of a physical quantity,
then Y(x) is the dimensionless parameter of that quantity. Let us denote
ymin/ymax as Ymin. Evidently, the range of Y(x) is from Ymin to 1.
3.2. Step II: normalisation of Y(x)
We can now normalise Y(x) so that the range of the normalised
function Yn(x) varies from 0 to 1 by defining Yn (x) as:
Y ðxÞ−NðY Þ
Yn ðXÞ ¼ ⋯:
1−NðY Þ
where N(Y) is the normalising function. In principle, it is possible to
choose the normalising function in an infinite number of ways, provided
3.3.1. Constant normalising function, N(Ea)
This is the simplest option available for normalising Ea(d). Obvi-
ously, N(Ea) should be equal to the minimum value of Ea(d) for all
values of Ea(d). For monotonic efficiency curves of hydrocyclone classi-
fiers, this is equal to Rf. In this case, the normalisation method is the
same as calculating the Kelsall ‘corrected efficiency’ as below.
Ea ðdÞ−R f
Ec ðdÞ ¼ ⋯: ð9Þ
1−R f
ð5Þ
The important point to be noted is the above is nothing but a simple
mathematical transformation. Here, all the values of Ea (d) are reduced
by a numerical value equivalent to Rf and the axis of the efficiency is
scaled up by a factor of 1/(1 − Rf).
 K. Nageswararao / Powder Technology 297 (2016) 106–114 111

Fig. 4. a. A typical function y = f(x) which exhibits two local minima and two local maxima. The range of x is from 0 to 500 while y varies from −15 to 90 as shown in Table 1. b. The
function y is converted into a dimensionless form Y such that Y = y/ymax. The range of Y is now from −0.17 to 1.00. Y is then normalised as Yn = (Y − Ymin)/(1 − Ymin) so that its
range is from 0 to 1. Detailed calculations are shown in Table 1.

3.3.2. Variable normalising functions and nNII (Ea), values of n = 2 is used. For the sake of comparison, the
It is possible to generate normalising functions which vary with Ea. Kelsall ‘corrected’ efficiency curve is also shown.
Since Ea is a function of particle size d, these could be thought of as size Of all the normalising functions, the normalised curve with nNII(Ea)
dependent normalising functions. The simplest way to generate such is closest to the actual efficiency curve. To illustrate the variation of
normalising functions is to assume different functions for dN(Ea)/dEa. the normalised curve with changes in the degree of the normalising
The boundary conditions for N(Ea) can be set as: function, n, calculations for the same data with different values of n
are shown in Fig. 5b. We can observe that for this type of normalising
function, the higher the value of n, the closer are the normalised and ac-
NðEa Þ ¼ R f at Ea ¼ R f
: ð10Þ tual efficiency curves. From Fig. 5a and b, we can clearly observe a wide
and NðEa Þ ¼ 0 at Ea ¼ 1 ⋯
variation in the ‘cut sizes’ ranging from about 103 μm to 125 μm, when
different functions are used for normalising the actual efficiency curve.
A few examples of normalising functions and corrections required to What is to be noted is that we cannot verify through any experimen-
Ea(d) are shown in Table 2. In the commonly used terminology in tal campaign if the fraction of bypass is equal to Rf as Kelsall suggested or
hydrocyclone practice, these assumed boundary conditions imply that it is equal to any of the values equal to the normalised functions given
the so called ‘bypass’ is efficiency dependent which in turn means it is Table 2, column 3. Suffice it to say that we can generate an infinite num-
dependent on particle size. ber of normalising functions which could be used to transform the ac-
tual efficiency curve. It is necessary to mention here that the
3.3.3. An illustrative example of normalisation normalising functions are not limited to only those shown in Table 2;
We show in Fig. 5a complete ‘normalised’ curves for a typical actual they are some typical functions and many more can be generated.
efficiency curve for which Rf is 25% and the actual cut size, d50a is We recall that Fig. 4b shows normalisation of y = f(x) following the
101 μm. These normalised curves are generated using the different nor- method of Kelsall. We may note that once y is reduced to its dimension-
malising functions shown in Table 2. For the calculations using nNI (Ea) less from, Y, it is possible to normalise it in an infinite number of ways by
using normalising functions similar to those in Table 2.
Table 1
An example of normalisation. 3.3.4. Normalisation of anomalous efficiency curves
x y Y Yn
Ever since Finch [28] reported the so called ‘fish hook’ in classifier ef-
ficiency curves, a large number of reports as well as theories to explain it
0 60 0.67 0.71
10 40 0.44 0.52
have appeared in literature for example, [30–32,48–58]. The existence
15 30 0.33 0.43 of fish hook phenomenon is disputed as well [59–64]. However, in fu-
20 10 0.11 0.24 ture, should the necessity arise, it is possible, to normalise such a non-
30 −10 −0.11 0.05 monotonic efficiency curve also by substituting the minimum value of
40 −15 −0.17 0.00
the actual efficiency for Rf in any of the normalising functions (Table 2
55 −13 −0.14 0.02
65 −4 −0.04 0.10 column 4). Obviously, in that case the normalised efficiency curve also
80 15 0.17 0.29 exhibits a fish hook, since the shape of the efficiency curve remains
105 38 0.42 0.50 the same after normalisation.
120 45 0.50 0.57 It is worthwhile to mention here that for classification of cement,
150 44 0.50 0.57
Altun and Benzer [57] have shown plots of corrected and reduced effi-
200 40 0.47 0.54
225 43 0.48 0.55 ciency curves exhibiting a fish hook. Although, they have not mentioned
275 56 0.62 0.68 the numerical value used for ‘bypass’ in calculating the corrected effi-
300 64 0.71 0.75 ciency, it is likely that they used the minimum value of Ea as the con-
325 72 0.80 0.83
stant normalising function.
350 81 0.90 0.91
400 90 1.00 1.00
450 88 0.98 0.98 3.3.5. Similarity between bypass and normalising functions
500 80 0.89 0.90 From Eqs. 8 and 9, we can observe that bypass and normalising func-
Dimensionless Y = y/ymax tion are analogous. Both have the same role in the transformation of the
Normalised Yn = (Y − Ymin)/(1 − Ymin) actual efficiency curve.
112 K. Nageswararao / Powder Technology 297 (2016) 106–114

Table 2
Examples of functions to normalise the actual efficiency, Ea(d).

S No dN(Ea)/dEa N(Ea) Normalised efficiency Correction to actual efficiency Remarks

1 0 Rf EKelsall
c
1−Ea
Rf 1−R Kelsall normalisation
f

2 K 1−Ea
Rf 1−R Eclin ð1−Ea Þ
Rf ð1−2
2
K is a constant
f R f þEa R f Þ
3 K/Ea logEa
Rf logR Elog
c Rf logR logE a
(1- Ea) –
f f −R f logE a

4 K Ena Rf 1−Enþ1
a
I
nE c Rf ð1−Enþ1
a Þ
(1-Ea) n ≠ −1
1−Rnþ1
f
ð1−Rnþ1
f
Þ−R f ð1−Enþ1
a Þ

5 K (1 − Ea)n nþ1
Rf ð1−Ea Þ nþ1
II
nE c Rf ð1−Ea Þnþ1
(1-Ea) n N −1
ð1−R f Þ ð1−R f Þnþ1 −R f ð1−Ea Þnþ1

As discussed earlier, the success of the corrected and reduced effi-
ciency curve based hydrocyclone models entrenched the perception
that bypass is a real physical phenomenon and an integral part of the
true classification process within the hydrocyclone. In fact, Wills and
Finch [27] illustrate the notion of bypass as a true physical phenomenon
with a numerical example. In their example, the recovery of water to
underflow is 25%; the mass flow rates of particles of a particular size
in feed and underflow are 16 t/h and 12 t/h respectively. That is the ac-
tual efficiency of particles of that size is 75%. Their calculations show by-
pass fraction as 4 t/h (25% of 16 t/h) and the particles reaching
underflow due to cyclone action as 8 t/h. They therefore calculated the
corrected efficiency to be (12–4)/(16–4), that is, 67%. These calculations
give an impression that 4 t/h of particles of this size physically short cir-
cuit to underflow without undergoing any classification.
We wish to emphasise here that in the above data, only the total
mass flow rates in feed (16 t/h), underflow (12 t/h) and recovery of
water to underflow (25%) can be determined experimentally. The
mass flow rate of feed which reaches underflow without undergoing
classification shown as 4 t/h is a calculated value. The only basis for
the calculation is the widespread belief among the hydrocyclone practi-
tioners that bypass is an integral part of the classification process as con-
ceived by Kelsall and that it is constant and equal to water recovery for
particles of all sizes.
It is important to note that there is no way we can verify experimen-
tally whether or not the bypass fraction is 4 t/h. In fact, we cannot even
function. The results of these calculations clearly show that we can
choose the value for ‘bypass’ (that is, normalising function) for
correcting the actual efficiency.
It is important to bear in mind that there is no method by which we
can evaluate whether any one of the normalising functions represents
the physical process more accurately than the others. The provision to
choose any value for bypass is a clear indication that it is only a notional
quantity, a mathematical object with no physical significance. Precisely
for this reason, the procedure involved in transforming the actual effi-
ciency to corrected efficiency curve is to be treated as a mathematical
transformation only.
3.3.6. General remarks
From an inspection of the normalising functions, shown in Table 2,
and Fig. 5a and b we can note:
1. The correction to Ea is dependent on Rf (or the minimum of Ea for
non-monotonic curves) and the actual efficiency only.
2. Any normalised curve generated by the functions in column 3, al-
ways lies between the actual efficiency curve, Ea(d) and the Kelsall
curve, Ec(d). This shall hold true even with other normalised func-
tions, which could be generated, provided it is ensured that the nu-
merical value of the normalising function is always less than the
minimum value of Ea. Subject to this condition, the normalised cut
size dN50c , follows the relationship given below.

design an experimental programme to prove whether or not any parti-

cle(s) reached underflow through bypass mechanism. Understandably,
this dilemma is a consequence of attributing a physical significance to
the experimentally indeterminate bypass. If however, the calculation N
d50a bd50c bd50c
of corrected efficiency is treated as a simple mathematical procedure
without attributing any physical significance to bypass as proposed by
Yoshioka and Hotta [4], this quandary is avoided.
To illustrate this issue further, we calculated normalised efficiencies 3. Of all the normalising functions, the normalised curve with nNII(Ea)
for the data of Wills and Finch using the normalising functions shown in is closest to the actual efficiency curve; the higher the value of n,
column 3 Table 2. The results are shown in Table 3. We can observe that the closer are the normalised and actual curves. Obviously, the satu-
the hypothetical quantities of bypass vary from 4 t/h (Kelsall) to as low ration limit shall be dependent on material characteristics.
as 0.44 t/h when 1NII (row 8, column 3, Table 3) is the normalising 4. Development of improved hydrocyclone models

Fig. 5. a. An example of generating normalised efficiency curves by using different normalising functions (Table 2, column 3). The degree the normalising functions is 2 for nNI and nNII. The
actual efficiency curve and Kelsall curve are also shown for comparison. b. Normalised efficiency curves obtained by using normalising function of Type II (nNII) of different degrees. The
degree of the function (n = 1, 2, 3, 5) is shown as a subscript.
 K. Nageswararao / Powder Technology 297 (2016) 106–114 113

Table 3 2. There is no rationale for bypass; it is purely notional. Further, an

Calculations showing hypothetical mass flow rates of bypass for different normalising outcome of the notion of bypass is equating the corrected efficiency
functions for the example given by Will and Finch [27]. The mass flow rates of solids in
feed and underflow are 16 t/h and 12 t/h respectively.
of near zero sized particles to zero. However, phenomenological
considerations show that these ultra fines split with water and
S N(E Hypothetical Correction Normalised Remarks† their recovery shall always be equal to that of water. This means
No. a) Bypass t/h to Ea Efficiency
that perfect separation of these particles is unattainable. As such,
1 0.250 4.00 0.083 0.667 Kelsall method, it is impossible to obtain experimental evidence of zero efficiency
EKelsall
c
for near zero sized particles under any conditions of cyclone
2 0.083 1.33 0.023 0.727 Elin
c
3 0.052 0.83 0.014 0.736 Elog operation.
c
4 0.067 1.07 0.018 0.732 n = − 0.5 3. We showed that it is not possible to determine experimentally the
5 0.117 1.87 0.033 0.717 n = 1.0 with nEIc fraction of feed which reaches underflow due to bypass or which
6 0.100 1.60 0.028 0.722 n = 0.5 reaches there due to cyclone action. This means that the so called
7 0.144 2.31 0.042 0.708 n = −0.5
8 0.028 0.44 0.007 0.743 n = 1.0 with nEIIc
bypass is purely notional. Significantly, we cannot even conceive
9 0.048 0.77 0.013 0.737 n = 0.5 an experimental programme to test bypass.
4. The method suggested by Kelsall for calculating corrected efficiency
Water recovery to underflow is 25%.
†For S. Nos. 1–3, the normalising functions correspond to those shown in column 3, is shown to be a simple mathematical procedure.
Table 2; for S. Nos 4–6, and S Nos. 7–9, the functions used are those in rows 4 and 5 5. Yoshioka and Hotta independently proposed a method to calculate
respectively. the corrected efficiency. They visualised the procedure as a simple
mathematical transformation only and did not attribute any physi-
Fortunately, lack of an experimental basis for bypass is not a con- cal significance to the steps involved.
straint for the normalisation of actual efficiency curves. The present 6. A general method for normalising any function y = f(x), for which x
well established model structure of hydrocyclones consists of equations and y vary between xmin to xmax and ymin to ymax respectively, is
for: (i) pressure-throughput (ii) recovery of water to underflow (iii) proposed. It is shown that we can transform y to its normalised
corrected cut size, d50c and (iv) the reduced efficiency curve. form Yn(x) such that the range of Yn is from 0 to 1, in an infinite
As shown earlier, there is no physical significance to any of the nor- number of ways. Our method, which utilises a user defined normal-
malising functions and hence to the normalised efficiency curves. As ising function, is illustrated with an example.
such, it is inconceivable that theoretical models for any of the normal- 7. The classification function is a special case where the actual effi-
ised cut sizes including the Kelsall corrected classification size, d50c ciency, Ea(d) varies between Rf to 1 for positive values of d. Clearly
could be developed from first principles. Extending the logic we could this too can be normalised in an infinite number of ways such that
say that properly formulated phenomenological models are more likely the ‘corrected’ (normalised) efficiency varies between 0 and 1.
to give better predictions of normalised cut sizes which are closer to the 8. The method suggested by Kelsall to derive the ‘centrifugal/
actual cut sizes. corrected efficiency’ is the simplest method of normalisation. Here
There is no doubt that the Kelsall method, which has been in contin- the normalising function is constant with a numerical value equal
ual use for more than six decades, is highly satisfactory. However, we to Rf for all values of Ea(d).
may note that normalised curves closer to the actual efficiency curve 9. Normalisation of a typical actual efficiency curve is illustrated using
can be generated by a judicious choice of the normalising function and different normalising functions with examples. The analogy be-
the degree of the normalising function, n. This offers a possibility of tween Kelsall bypass and normalising functions is also illustrated
using a normalised cut size other than Kelsall cut size and the corre- with a numerical example.
sponding reduced efficiency curve in future hydrocyclone models. It is 10. The possibility of development of improved hydrocyclone
particularly relevant here to recall that assuming the reduced efficiency models using a normalised cut size other than Kelsall cut size is
curve to be material specific is an approximation only, though highly re- explained.
liable for all practical purposes [65]. As such, it is worthwhile to explore 11. It is proposed that the corrected efficiency be denoted as Kelsall ef-
if normalisation other than by Kelsall method gives even a better ap- ficiency and the corresponding cut size as Kelsall cut size, as a trib-
proximation. There is a scope for development of an improved ute to the landmark contribution of Kelsall towards modelling of
hydrocyclone model which could give better predictions. hydrocyclones.

4. Proposal to denote corrected efficiency as Kelsall efficiency

Nomenclature
Undoubtedly, the concept of ‘corrected’ efficiency curve introduced
by Kelsall caused a paradigm shift in quantifying the performance of d particle size, μm
hydrocyclone classifiers. The cut size, d50c derived from this curve con- d50a size of particle with recovery of 50% on the gross/actual
tinues to be a key performance characteristic of all successful efficiency curve
hydrocyclone models currently in use. Thus, it would be most appropri- d50c size of particle with recovery of 50% on the Kelsall ‘corrected’
ate to denote the corrected efficiency as Kelsall efficiency and the efficiency curve
corrected cut size, d50c as Kelsall cut size as a tribute to his landmark Ea(d), Ea actual recovery of particles of size, d, to underflow
contribution [66]. Ec(d), Ec normalised recovery of particles of size, d, to underflow
n degree of the normalising functions for nNI(Ea) and nNII(Ea)
5. Summary and Conclusions N(Y) function to normalise Y
N {Ea (d)}, N (Ea) function to normalise actual efficiency
1. Kelsall visualised the classification process as one in which a frac- Rf recovery of water to underflow.
tion of the feed solids bypasses to underflow and the remaining xmin, xmax minimum and maximum values of x
fraction only is subjected to classification action. This notion of by- y function of x
pass led to the concept of corrected efficiency, which continues to ymin, ymax minimum and maximum values of y
be the foundation of all hydrocyclone models currently in use. At Y dimensionless y; equal to y/ymax
present, hydrocyclone practitioners perceive bypass as an integral Ymin minimum value of Y; equal to ymin/ymax
part of the classification process in hydrocyclones ε very small positive number close to zero
114 K. Nageswararao / Powder Technology 297 (2016) 106–114
Acknowledgements
I am extremely grateful to Satya Simha for introducing the Works of
Sri Aurobindo on Nature, which have a profound influence on my un-
derstanding of particulate systems.
References
[1] D.F. Kelsall, A further study of hydraulic cyclone, Chem. Eng. Sci. 2 (6) (1953)
254–272.
[2] K. Nageswararao, D.M. Wiseman, T.J. Napier-Munn, Two empirical hydrocyclone
models revisited, Miner. Eng. 17 (5) (2004) 671–687.
[3] K. Nageswararao, Normalisation of the efficiency curves of hydrocyclone classifiers,
Miner. Eng. 12 (1) (1999) 107–118.
[4] N. Yoshioka, Y. Hotta, Liquid cyclone as a hydraulic classifier, Chem. Eng. Japan 19
(12) (1955) 632–641 http://doi.org/10.1252/kakoronbunshu1953.19.632.
[5] D. Bradley, D.J. Pulling, Flow patterns in the hydraulic cyclone and their interpreta-
tion in terms of performance, Trans. Inst. Chem. Eng. 37 (1959) 34–44.
[6] A.J. Lynch, T.C. Rao, Digital computer simulation of comminution systems, Proc. 8th
Comm. Min. Metall. Congr., 6 1965, pp. 597–606 Aust. NZ.
[7] A.J. Lynch, W.J. Whiten, N. Draper, Developing the optimum performance of a multi-
stage grinding circuit, Trans. Inst. Mining Met. (Sect. C: Miner. Process. Extract. Met.
76 (1967) C169–C182.
[8] A.J. Lynch, T.C. Rao, Studies on the operating characteristics of hydrocyclone classi-
fiers, Ind. J. Technol. 6 (4) (1968) 106–114.
[9] A.J. Lynch, T.C. Rao, Modelling and Scale-up of hydrocyclone classifiers, Proc. of XIth
Int. Miner. Process, Cong, Cagliari 1975, pp. 245–269.
[10] A.L. Mular, Mathematical models for optimum design of grinding circuits, CIM Bull.
64 (708) (1971) 34–41.
[11] J.D. Pitts, S.R. Holsinger, N.W. Johnson, D.E. Crowell, Direct Digital Computer Control
of a Primary Grate-Discharge Ball Mill-Hydrocyclone Circuit at ASARCO's Silver Bell
Unit, AMIRA Symp. on Optimisation and Control, Brisbane, July, 1974.
[12] L.R. Plitt, A mathematical model of the hydrocyclone classifier, CIM Bull. 69 (776)
(1976) 114–123.
[13] B.C. Flintoff, L.R. Plitt, A.A. Turak, Cyclone modelling - A review of present technol-
ogy, CIM Bull. 80 (905) (1987) 39–50.
[14] K. Nageswararao, A generalised model for hydrocyclone classifiers, Proc. of Austral-
asian Inst. of Mining and Metal. 300 (2) (1995) 21.
[15] T. Braun, M. Bohnet, Influence of feed solids concentration on the performance of
hydrocyclones, Chem. Eng. Technol. 13 (1) (1990) 15–20.
[16] J.J. Cilliers, A.L. Hinde, An improved hydrocyclone model for backfill preparation,
Miner. Eng. 4 (7–11) (1991) 683–691.
[17] M.A.N. Doheim, M.A.M. Ahmed, S.A. Sayed, Hydrocyclone Investigation and Model-
ling at Medium Feed Solids Concentration of Mixed Minerals, 1041995 C37–C44.
[18] A.K. Asomah, T.J. Napier-Munn, An empirical model of hydrocyclones, incorporating
angle of cyclone inclination, Miner. Eng. 10 (3) (1997) 339–347.
[19] W. Chen, N. Zydek, F. Parama, Evaluation of hydrocyclone models for practical appli-
cations, Chem. Eng. J. 80 (1–3) (2000) 295–303.
[20] M.A.Z. Coelho, R.A. Medronho, A model for performance prediction of
hydrocyclones, Chem. Eng. J. 84 (1) (2001) 7–14.
[21] L.M. Tavares, L.L.G. Souza, J.R.B. Lima, M.V. Possa, Modeling classification in small-
diameter hydrocyclones under variable rheological conditions, Miner. Eng. 15 (8)
(2002) 613–622.
[22] B. Firth, Hydrocyclones in dewatering circuits, Miner. Eng. 16 (2) (2003) 115–120.
[23] M. Narasimha, A.N. Mainza, P.N. Holtham, M.S. Powell, M.S. Brennan, A semi-
mechanistic model of hydrocyclones – Developed from industrial data and inputs
from CFD, Int. J. Miner. Process. 133 (2014) 1–12.
[24] T.J. Napier-Munn, S. Morrell, R.D. Morrison, T. Kojovic, Mineral Comminution Cir-
cuits - their Operation and Optimisation, JKMRC Monograph Series, Julius
Kruttschnitt Mineral Research Centre, University of Queensland, 1996.
[25] Honeywell International Inc, Optimizing Grinding Circuits: Technical Information
NoteTN 633July 2011.
[26] S. Botha, I.K. Crag, J.D. Le Roux, Switching cyclones to increase product particle size
range for ore milling circuits, IFAC (International Federation of Automatic Control)
Hosting by Elsevier Ltd, IFAC Papers Online, 48-17 2015, pp. 092–097 http://doi.
org/10.1016/j.ifacol.2015.10.084.
[27] B.A. Wills, J.A. Finch, Wills' Mineral Processing Technology, eighth ed. Butterworth
and Heinemann, Oxford, 2015.
[28] J.A. Finch, Modelling a fish-hook in hydrocyclone selectivity curves, Powder Technol.
36 (1) (1983) 127–129.
[29] E.G. Kelly, The significance of bypass, Miner. Eng. 4 (1) (1991) 1–7.
[30] R. Del Villar, J.A. Finch, Modelling the cyclone performance with a size dependent
entrainment factor, Miner. Eng. 5 (6) (1992) 661–669.
[31] E.J. Roldan-Villasana, R.A. Williams, T. Dyakowski, The origin of fish-hook effect in
hydrocyclone separators, Powder Technol. 77 (3) (1993) 243–250.
[32] W. Kraipech, W. Chen, F.J. Parma, T. Dyakowski, Modelling the fish-hook effect of the
flow within hydrocyclones, Int. J. Miner. Process. 66 (1) (2002) 49–65.
[33] T.J. Napier-Munn, A.J. Lynch, The modelling and computer simulation of mineral
treatment processes- Current status and future trends, Miner. Eng. 5 (2) (1992)
143–167.
[34] L.R. Plitt, The analysis of solid - solid separations in classifiers, CIM Bull. 64 (708)
(1971) 42–47.
[35] K.J. Reid, Derivation of an equation for classifier - reduced efficiency performance
curves, Can. Metall. Q. 10 (3) (1971) 253–254.
[36] H. Schubert, T. Neesse, The role of turbulence in wet classification, Tenth Int. Mineral
Process. Congr. 1973 Instn, Min. & Metall, London 1974, pp. 213–239.
[37] H. Schubert, Th. Neeße, A Hydrocyclone Separation Model in Consideration of the
Multi-Phase Flow, Int. Conf. Hydrocyclones, BHRA Fluid Engineering, Cambridge,
1980 23–36.
[38] H. Schubert, Which demands should and can meet a separation model for
hydrocyclone classification? Int. J. Miner. Process. 96 (1) (2010) 14–26.
[39] J. Dueck, The sedimentation velocity of a particle in a wide range of Reynolds num-
bers in the application to the analysis of the separation curve, Adv. Powder Technol.
24 (1) (2013) 150–153.
[40] H.E. Criner, The Vortex Thickener, Int. Conf. on Coal Preparation, Paris, 1950, Rev. de
le Ind. Minerie, Special Issue No. 51951 628–643.
[41] M.I.G. Bloor, D.B. Ingham, Theoretical investigation of the flow in a conical
hydrocyclone, Trans. Inst. Chem. Eng. 51 (1) (1973) 36–41.
[42] J.A. Delgadillo, R.K. Rajamani, A comparative study of three turbulence-closure
models for the hydrocyclone problem, Int. J. Miner. Process. 77 (4) (2005) 217–230.
[43] K.U. Bhaskar, Y.R. Murthy, M.R. Raju, S. Tiwari, J.K. Srivastava, N. Ramakrishnan, CFD
simulation and experimental validation studies on hydrocyclone, Miner. Eng. 20 (1)
(2007) 60–71.
[44] G. Zhu, J.L. Liow, A. Neely, Computational study of the flow characteristics and sep-
aration efficiency in a mini-hydrocyclone, Chem. Eng. Res. Des. 90 (12) (2012)
2135–2147.
[45] A. Davailles, E. Climent, F. Bourgeois, A.K. Majumder, Analysis of swirling flow in
hydrocyclones operating under dense regime, Miner. Eng. 31 (2012) 32–41.
[46] K.R. Popper, The Logic of Scientific Discovery, Routledge, London, (Logik der
Forschung, First Published, 1935 by Verlag von Julius Springer, Vienna), 1992.
[47] K.R. Popper, Conjectures and Refutations: The Growth of Scientific Knowledge,
Harper and Row, New York, 1968.
[48] M. Frachon, J.J. Cilliers, A general model for hydrocyclone partition curves, Chem.
Eng. J. 73 (1) (1999) 53–59.
[49] A.K. Majumder, P. Yerriswamy, J.P. Barnwal, The “fish-hook” phenomenon in cen-
trifugal separation of fine particles, Miner. Eng. 16 (10) (2003) 1005–1007.
[50] B. Wang, A.B. Yu, Computational investigation of the mechanisms of particle separa-
tion and “fish-hook” phenomenon in hydrocyclones, AICHE J. 56 (7) (2010)
1703–1715.
[51] F. Bourgeois, A.K. Majumder, Is the fish-hook effect in hydrocyclones a real phenom-
enon? Powder Technol. 237 (2013) 367–375.
[52] J.G. Dueck, L.L. Minkov, E.V. Pikushchak, Modeling of the “fish-hook” effect in a clas-
sifier, J. Eng. Phys. Thermophys. 80 (1) (2007) 64–73.
[53] L.L. Minkov, J.H. Dueck, Numerical modelling of a non monotonic separation
hydrocyclone curve, J. Eng. Phys. Thermophys. 85 (6) (2012) 1317–1326.
[54] G. Zhu, J.L. Liow, Experimental study of particle separation and the fish hook effect in
a mini-hydrocyclone, Chem. Eng. Sci. 111 (2014) 94–105.
[55] M. Farghaly, Th. Neesse, The theoretical partition curve of the hydrocyclone, Miner.
Eng. 62 (2014) 25–30.
[56] T.R. Vakamalla, K.S. Kumbhar, R. Gujjula, N. Mangadoddy, Computational and exper-
imental study of the effect of inclination on hydrocyclone performance, Sep. Purif.
Technol. 138 (2014) 104–117.
[57] O. Altun, H. Benzer, Selection and mathematical modelling of high efficiency air clas-
sifiers, Powder Technol. 264 (2014) 1–8.
[58] L. Abdollahzadeh, M. Habibian, R. Etezazian, S. Naseri, Study of particle's shape fac-
tor, inlet velocity and feed concentration on mini-hydrocyclone classification and
fishhook effect, Powder Technol. 283 (2015) 294–301.
[59] K. Nageswararao, A critical analysis of fish-hook effect in hydrocyclone classifiers,
Chem. Eng. J. 80 (1–3) (2000) 251–256.
[60] K. Nageswararao, Comment on: ‘Is the fish-hook effect in hydrocyclones a real phe-
nomenon?’ by F. Bourgeois and A. K. Majumder [Powder Technology 237 (2013)
367-375], Powder Technol. 262 (2014) 194–197.
[61] K. Nageswararao, R.A. Medronho, Fish hook effect in centrifugal classifiers – a fur-
ther analysis, Int. J. Miner. Process. 132 (2014) 43–58.
[62] K. Nageswararao, Comment on: ‘Experimental study of particle separation and the
fish hook effect in a mini-hydrocyclone’ by G. Zhu and J. L. Liow [Chemical Engineer-
ing Science 111 (2014) 94–105], Chem. Eng. Sci. 122 (2015) 182–184.
[63] K. Nageswararao, Comments on ‘fish hook effect in classifier efficiency curves’ in re-
cent publications in Powder Technology, Powder Technol. 286 (2015) 468–470.
[64] K. Nageswararao, B. Karri, Fish hook in classifier efficiency curves: an update, Sep.
Purif. Technol. 158 (2016) 31–38.
[65] K. Nageswararao, Reduced efficiency curves of industrial hydrocyclones — an analy-
sis for plant practice, Miner. Eng. 12 (5) (1999) 517–544.
[66] A.B. Whitehead, In Memoria, Powder Technol. 33 (1) (1982) 1–2.