Particuology 32 (2017) 82–88

Contents lists available at ScienceDirect

Particuology
journal homepage: www.elsevier.com/locate/partic

Transient parameter analysis of pneumatic conveying of fine particles

for predicting the change of mode of flow
Shijo J.S., Niranjana Behera ∗
School of Mechanical & Building Sciences, VIT University, Vellore, Tamilnadu 632014, India

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

Article history: During the process of pneumatic conveying of fine particles, the flow mode may change from dense to
Received 15 January 2016 dilute. In studying this mode change, we evaluate three parameters (Hurst exponent, Shannon entropy,
Received in revised form 12 May 2016 and phase-space attractor size) used in signal analysis. Experimental data of pneumatic conveying of fly
Accepted 6 July 2016
ash at three locations along a 173-m-long pipeline were used for this analysis. Variations in magnitude
Available online 11 November 2016
of the Hurst exponent, Shannon entropy, and size of the phase space attractor exhibit different trends in
their variations for dense and dilute mode of flow. From these trends it is possible to predict the change
Keywords:
from dense to dilute mode and the location along the length of the pipeline of this change in mode of
Fluidized dense
Hurst exponent
flow.
Shannon entropy
© 2016 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of
Phase space Sciences. Published by Elsevier B.V. All rights reserved.

Introduction
Fine particles or powders can be conveyed by fluidized dense
phase conveying in which the material flows in two distinct layers,
an upper dilute flow and a lower dense flow. Compared with the
dilute mode of flow, the dense mode of flow has many advantages
such as low power consumption, low pipe wear, and low particle
degradation. During conveying, the material which is in the dense
mode at the inlet of a long pipeline may change to the dilute mode
at a certain location along the length of the pipeline. This arises
from a gradual expansion of gas from the inlet to the outlet of the
pipeline. This change in mode of flow from dense to dilute depends
upon many factors such as length and diameter of the pipeline,
inlet pressure, and solids loading ratio. Solids loading ratio (m*) is
defined as the ratio of solids mass flow rate to air mass flow rate.
When the flow becomes dilute, the power consumption increases,
particles degrade, and pipes are subject to erosive wear. For dilute
flow, erosive wear is more significant at bends in the pipeline. Oper-
ating in dilute mode for any extended period in time may require
replacing pipes in these bends or even straight sections. If the parti-
cle is abrasive (for example, alumina or silica sand) then the erosive
wear may be significant. Hence it is necessary to trace the loca-
∗ Corresponding author. Fax: +91 416 224 3092; tel.: +91 9092656890.
E-mail addresses: shijocet@gmail.com (S. J.S.), beheranira@rediffmail.com,
niranjanabehera@vit.ac.in (N. Behera).
tion of flow mode transition for a given flow condition or pipeline
configuration.
Few researchers have investigated the tracing of the location
of transition in mode of flow. Williams, Jones, and Cenna (2008)
applied the wavelet technique to analyze variations in pulse ampli-
tude of the pressure signal along a length of pipeline and concluded
that pulse amplitude increases because gas expands along the
pipeline. Behera, Agarwal, Jones, and Williams (2012) analyzed the
variation of transient parameters in pressure data such as pulse
amplitude, pulse slope ratio, and pulse time ratio. The variation
of the pulse slope ratio at three locations along the pipeline pro-
vided an approximate location of the flow mode transition. Mittal,
Mallick, and Wypych (2014) studied the variation of parameters
such as standard deviation and Shannon entropy for analyzing the
flow in bends. They observed that both values tend to decrease
after a bend. Mittal, Mallick, and Wypych (2015) used signal analy-
sis parameters and a technique (Shannon entropy, Hurst exponent,
and phase space diagram) to study flow behavior of conveyors
with bend sections of pipeline. Pahk and Klinzing (2008) used
pressure fluctuations in pneumatic conveyors transferring poly-
mer pellets to observe the flow regimes. They used four different
analyses—power spectral density analysis, phase space diagram
analysis, rescaled range analysis, and wavelet analysis. Few inves-
tigations have been performed for pneumatic conveying of fine
particles to determine the transition location from dense to dilute
flow. In this paper, three signal parameters (Hurst exponent, Shan-
non entropy, and phase space diagram) are used to analyze their
variation along the length of the pipeline.

http://dx.doi.org/10.1016/j.partic.2016.07.004
1674-2001/© 2016 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.
 S. J.S., N. Behera / Particuology 32 (2017) 82–88
For the analysis of fluidized bed of sand particles, Ahuja,
Agarwal, Sethi, and Raj (2005) used the Hurst exponent, which they
found to rise, peak and then fall with increasing superficial gas
velocity. Azizpour, Mostoufi, Zarghami, and Sotudeh-Gharebagh
(2011) used the Hurst exponent as a parameter for analyzing flu-
idized bed hydrodynamics near the transition point from bubbling
to turbulent fluidization regime. They observed a similar trend in
the Hurst exponent of macro structures. This trend in the Hurst
exponent has been correlated with a regime transition in the flu-
idized bed. Sheikhi, Sotudeh-Gharebagh, Mostoufi, and Zargham
(2013) divided signals into three levels of micro-, meso-, and
macro-scales. They observed micro-scale signals with all Hurst
exponents smaller than 0.5, meso-scale signals with some Hurst
exponents larger than 0.5 and some smaller than 0.5, and macro-
scale signal with Hurst exponent greater than 0.5. Sedighikamala
and Zarghami (2013) studied the behavior of a gas solid fluidized
bed and found that the behavior of macrostructures (bubbles) did
not change notably with increasing gas velocity because the change
in Hurst exponent is small.
Kang, Woo, Ko, Cho, and Kim (1999) analyzed pressure fluc-
tuation signals from three-phase fluidized bed. They observed in
pressure fluctuation data that the Shannon entropy of pressure fluc-
tuation data attained a local minimum by varying the liquid flow
rate. A flow transition in the fluidized solid particles was detected
quite conveniently using the variation in Shannon entropy. Zhong
et al. (2009) studied the transitions of flow patterns (under-
fluidization, steady fluidization, and turbulent fluidization) in a
biomass fluidized bed from variations in Shannon entropy. Zhong
and Zhang (2005) observed that for a spout fluid bed at high spout-
ing velocity the Shannon entropy at all bed locations increased
sharply accompanied by an asymmetric unstable flow. The Shan-
non entropies of different flow regimes are distinct, so they were
used to identify the flow regimes. Duan and Cong (2013) analyzed
pressure fluctuations resulting from the non-uniform flow behav-
ior of solid particles (particles classed as Geldart groups B and D) in
a fluidized bed using an analysis of Shannon entropy. Based on its
variation, they were able to observe transitions between bubbling
fluidization, turbulent fluidization, and fast fluidization.
Kim and Han (1999) studied the effects of coarse particles
and relative humidity of air on the flow behavior of polymer
powders–air suspension in the riser. They observed that the size of
the phase-space attractor increased with increasing solid circula-
tion rate. Similarly, Briongos and Soler (2005) used the attractor size
to characterize fluidization dynamics in slugging flow regime. Here
the increased number of particles may have increased the interac-
tions between particle–particle and particle–wall. Zhao, Jin, Gao,
Du, and Wang (2014) also used the phase space attractor to ana-
lyze different flow patterns in a vertical upward three-phase flow
of oil–gas–water.
Experimental
Pneumatic conveying experiments were conducted over a 173-
m-long test pipeline conveying fly ash. Fly ash has a particle density
of 2096 kg/m3 and mean particle diameter 14.91 ␮m. The calcu-
lated terminal velocity of fly-ash particles is 0.013 m/s. Based on
Geldart’s fluidization diagram (a plot between particle–gas density
ratio and mean particle diameter), the fly ash used in this exper-
iment was classified as Group A. Group A materials have higher
air retention capability and hence are suitable for fluidized dense
phase flow. The particle size distribution for this fly ash material
has been presented in Fig. 1. The experiments were performed by
pre-pressurizing the conveying material in a 1-m3 top discharge
blow tank or feeder with the help of supplied air from a controlled
bank of sonic nozzles before feeding into the pipeline (53-mm pipe
83
Fig. 1. Particle size distribution of fly ash.
diameter, 173-m length, see Fig. 2 for a schematic diagram of the
pipeline). One pressure transmitter was located near to the supply
point of air at the bottom of the blow tank (primary air) to measure
the primary air pressure. Part of the air is supplied near to the feed-
ing point of the blow tank (secondary air) at which the pressure
is measured using a pressure transmitter; this pressure is referred
to as the secondary air pressure. Twenty-two pressure transmit-
ters were fixed on the top and bottom of the pipeline at different
locations along its length to measure the gauge pressure values. A
SCXI-1102 data acquisition system (Yokogawa, Japan) was used to
record electrical signals from the load cells and pressure transmit-
ters. This system has 32 differential channels. Pressure data were
recorded by each of the pressure transmitters at a sampling fre-
quency between 80 and 100 Hz. The solids mass flow rates from
the blow tank and into the receiver were measured using a series
of load cells. Before conducting the experiment, the pressure trans-
mitters were calibrated using a Barnett deadweight tester and the
load cells were calibrated by placing different known weights upon
it. The experiments were performed over a 12–50 range in solids
loading ratios for fly ash. The inlet gas velocity for fly ash varied
between 3–12 m/s.
Transient parameter analysis
After a bend, particles are subjected to deceleration that damp-
ens particle turbulence. Nevertheless, after a certain distance,
particles are reaccelerated in the flow. Pressure signals near a bend
are affected (Mittal et al., 2014). Hence transmitter locations are
chosen sufficiently distant (approximately 8 m) from a bend section
of pipe. Pressure transmitters T1, T15, and T17 were maintained at
distances 27, 100, and 129 m, respectively from the blow tank. The
location of transmitter T1 was chosen as the flow there reaches a
steady state. The transmitters T15 and T17 were located well apart
from each other and also T1 so that different signal data can be
captured for each transmitter.
For the purpose of analysis, three distinct flow conditions were
chosen as:
Case P (kPa) m*
1 233 26.3
2 235 29.6
3 186 14.2
The Hurst exponent, Shannon entropy, and phase-space attrac-
tor size were calculated for the time-varying pressure data
registered by the transmitters at the three locations (T1, T15, T17)
for each of the flow conditions. Pressure signals for the three flow
conditions at T1 are presented in Fig. 3. Variations of the calculated
parameters were also analyzed and given below. For the present
analysis, the flow is considered as dense if the superficial gas veloc-
ity is below 15 m/s and dilute if this velocity is above 15 m/s.
84 S. J.S., N. Behera / Particuology 32 (2017) 82–88

Fig. 2. Pneumatic conveying set up.

tions. To calculate the exponent for a time-series of observations

(let X = X1 , X2 , X3 ,. . ., Xn ), the following parameters need to be cal-
culated:
1 n
Mean: X̄ = Xi , (1)
n i=1

Departure of time series: Y (t) = X (t) − X̄, (2)

t
Cumulative deviation series: Z (t) = Yi for t − 1, 2, .....n, (3)
i=1

Range : Rn = max (Z1 , Z2 , . . ., Zn ) − min (Z1 , Z2 , . . ., Zn ) , (4)


1 n  2
Standard deviation: Sn = Xi − X̄ . (5)
n i=1

The ratio of Rn to Sn gives a parameter called the rescaled range.

Fig. 3. Pressure signals for the three flow conditions in flow at transmitter location
T1.
Hurst exponent analysis
Hurst (1951) used a technique that was specifically meant
for modeling reservoirs. The technique required an estimate of a
parameter called the Hurst exponent for a time span of observa-
For calculating the Hurst exponent (H), the total number of data
points is taken in the form 2N . The rescaled range is calculated for
each sub-set. These values are plotted against the number of data
points. The Hurst exponent is calculated by plotting a linear regres-
sion line through these points on a logarithmic scale. The slope of
the line gives the magnitude of Hurst exponent.
The magnitude of the Hurst exponent indicates the amount
of correlation in the time-series data. Three instances may occur
(Ahuja et al., 2005): if 0 < H < 0.5, negatively correlated time-
series data exist; if H = 0.5, uncorrelated time-series data exist; if
 S. J.S., N. Behera / Particuology 32 (2017) 82–88
Fig. 4. Variation of Hurst exponent with distance along the length of pipeline.
0.5 < H < 1, positively correlated time-series data exist. If H > 1, more
persistent data exist that exhibit some trends.
A decrease in value of H indicates that higher disorder exists in
the time-series data and higher pulsations occur in the flow. Con-
trastingly, an increase in value of H indicates lower disorder in the
time-series data and turbulence or pressure fluctuations decrease
(Mittal et al., 2015).
Case 1 (P = 233 kPa, m* = 26.3): Fig. 4 shows the variation of
H along the length of the pipeline. The magnitude of H increases
from 0.88 at the first location to 0.89 at the second while the flow
remains in the dense mode. Physically however the flow gets aer-
ated between these two locations but could not be identified from
the variation in H. From the second to the third location, a decrease
in the value of H from 0.89 to 0.87 occurs. This decrease indicates an
increase in disorder in the time series data. The value of H is lower at
low solids loading ratios. Hence somewhere near the third location
(near 120 m) the flow changes to dilute mode. The experimental
data also confirmed that the superficial gas velocities at each of the
locations are in sequential order 7.3, 10.7, and 12.9 m/s. At the third
location the mode is tending toward dilute. Once the mode of flow
changes to dilute, the flow cannot revert to its previous dense mode
because at this section the gas velocity is quite high because gas
expands considerably through this uniform cross-section pipeline.
However, flow can revert to dense mode by stepping down the pipe
diameter at the location of transition.
Case 2(P = 235 kPa, m* = 29.6): in Fig. 4, the value of H
increases slightly from 0.89 at the first location to 0.9 at the second
and it remains unchanged up to the third. This shows that the flow
remains in dense mode up to the third location.
Case 3(P = 186 kPa, m* = 14.2): here the flow starts in the
dilute mode (with superficial gas velocity 11.83 m/s). The value of
the Hurst exponent decreases from 0.84 to 0.8 between first and
second locations. A decrease in its H value indicates an increase in
disorder in the time-series data and therefore the flow is compar-
atively more dilute at the second location compared with the first.
Also at the second location, the value of the Hurst exponent is low-
est of all three flow conditions. The value of H increases to 0.81 at
the third location, for which there is no physical explanation.
To study the influence of solids loading ratio on the Hurst expo-
nent, a plot of these two parameters is given (Fig. 5) showing that
the Hurst exponent increases with increasing solids loading ratio.
The trend line indicates that the data points are scattered on both
sides with a R2 value of 0.7127.
Shannon entropy analysis
Statistical entropy was introduced by Shannon (1948) as a basic
concept in information theory, measuring the average missing
information on a random source. Shannon entropy quantifies the
85
Fig. 5. Dependence of Hurst exponent on solids loading ratio.
unevenness of the probability distribution p. It is expressed in units
called bits. The higher the value of Shannon entropy, the higher
will be the disorder in the system because of the complex and
chaotic nature of flow resulting from turbulence and the intensive
particle/wall/air interactions.
According to the second law of thermodynamics, an increase
in thermodynamics entropy indicates that an event transits from
order to disorder. Nevertheless, for Shannon entropy, the value of
entropy expresses the degree of uncertainty involved in a certain
system. Hence, as for an increment in thermodynamics entropy, an
increase in Shannon entropy can also be considered to represent
an event that transits from order to disorder.
For a random variable X with values in a finite set X1 , X2 ,...,Xn ,
entropy is defined as:

n
E(X) = − pi logb pi , (6)
i=1
where n is the length of the signal, b = 2, and pi is the probability for
the ith component in the signal, which satisfies the constraint,

n
pi = 1. (7)
i=1
The pressure data considered for analysis are rounded off to the
nearest integers. With the rounded-off pressure data, the number of
times a particular value of data is calculated. Hence the probability
for an individual value (pi ) can be calculated by dividing the number
of times a particular value of data occurs relative to the total length
of the data series.
Shannon entropy can be calculated using the expression in Eq.
(6). In a calculation, each data set (consisting of approximately 4000
data for analysis) is divided into sub-sets each consisting of 250
data. The reason for considering these subsets is to investigate the
magnitudes of their Shannon entropy. For a given flow condition,
the Shannon entropy has been calculated for 16 data sub-sets at
the three locations along the length of the pipeline. For ease in
identification, the symbols star, square, and diamond are used to
distinguish the magnitudes at the three locations T1, T15, and T17
at 27, 100, and 129 m, respectively, along the length of the pipeline.
Case 1 (P = 233 kPa, m* = 26.3): Fig. 6(a) shows the Shannon
entropy calculation for Case 1 flow condition at the three locations
for 16 data sub-sets. For 14 data sub-sets the Shannon entropy
increases from the first to second locations and then decreases from
the second to third location. Higher Shannon entropy indicates high
uncertainty because randomness in the signal increases. Hence a
dense flow has a higher value of Shannon entropy because the flow
is dominated by significant collisions and has higher uncertainty.
From the first to the second location, it increases because the bed
material is aerated inside the pipeline and the flow remains in the
86 S. J.S., N. Behera / Particuology 32 (2017) 82–88
Fig. 6. Variations in Shannon entropy across various data subsets at different loca-
tions in the pipeline under different flow conditions. (a) Case 1 (P = 233 kPa,
m* = 26.3), (b) Case 2 (P = 235 kPa, m* = 29.6), (c) Case 3 (P = 186 kPa, m* = 14.2).
dense mode. However, beyond the second location, i.e., at the third
location, the value of the Shannon entropy decreases because some-
where near this location the mode of flow changes to dilute. The
experimental data also confirms that the superficial gas velocity
at the third location is 12.9 m/s, corresponding to the dilute mode
regime. The calculated solids volume fraction at the third location
is 0.026, again indicating that the flow is dilute.
Case 2 (P = 235 kPa, m* = 29.6): Fig. 6(b) shows the Shannon
entropy under Case 2 flow condition at all three locations for 16 data
sub-sets. For 11 data sub-sets, the Shannon entropy increases from
the first to the third location. The uncertainty hence increases and
the flow remains in the dense mode and shows no sign of change
in mode unlike the results for Case 1.
Case 3 (P = 186 kPa, m* = 14.2): Fig. 6(c) shows the results for
the Shannon entropy calculation under Case 3 flow condition at
all three locations using 16 data sub-sets. For 14 data sub-sets, the
Shannon entropy decreases from the first to the second location
and then increases from the second to the third location. The flow
is dilute at the first location but has become more dilute on reach-
ing the second location as indicated by the decrease in Shannon
entropy or uncertainty. For a dilute flow, the collision frequency
is lower compared with a dense flow. However, from the first to
the third location, the decrease in Shannon entropy has no physical
significance.
Phase space analysis
In a nonlinear analysis of the time series data, the time devel-
opment of the each of the signals is plotted. Based on the method
of delays, the time-varying pressure data are plotted with its own
time-lagged signal. The pressure signal data are also plotted in state
space to reveal the attractor. In practice, the selection of the time
lag is more important in these types of plots. The plots are between
the original time series P (t) versus its delayed versions, i.e., P(t + ),
where ␶ is the delay time to collect one data point. For the present
analysis, two pressure data sets, each one set having 1000 data
points, were considered for the analysis. Taking 1000 data points
is sufficient as it allows more data points to be plotted at once and
helps in identifying the difference between the two plots for two
flow conditions or two locations.
For the Case 1, the size or area of the phase space attractor
increases from transmitter location T1–T15 but then decreases
from location T15–T17 (Fig. 7(a)). A significant increase in the area
of the phase space attractor at T15 indicates that there is more
uncertainty in the pressure time data and flow becomes more tur-
bulent. The phase space attractor data shows greater eccentricity
at T15 compared with the centric or binding state found at T1.
At T15, the mode of flow changes from dense to dilute, which is
also confirmed from the magnitude of experimental superficial gas
velocity of 12.9 m/s. Beyond T15, i.e., at T17, the phase space attrac-
tor remains eccentric and the flow remains in the dilute mode,
showing that at T1 where the flow is in the dense mode, the larger
particle concentration suppresses turbulence in comparison with
the state at T15 or T17.
For Case 2, the area of phase space attractor increases from T1 to
T15 and then further increases from T15 to T17 (Fig. 7(b)). However,
the increase in size is not as large compared with that for Case 1.
Although there is an increase in the size of phase space attractor
up to location T17, the flow remains in dense mode for all three
locations T1, T15, and T17.
For the Case 3, the size of the phase space attractor does not
vary greatly from the location T1–T17 (Fig. 7(c)). This indicates that
the dilute mode of the flow at location T1 has persisted through
T15–T17. The phase space plots indicate that there is no specific
range in values for size or area of the attractor for a dense or dilute
mode of flow.
Comparison of the three methods
The phase space method is able to predict the change in mode of
flow from the change in size of the phase space attractor. A better
 S. J.S., N. Behera / Particuology 32 (2017) 82–88 87

Fig. 7. Variations in phase space attractor at three transmitter locations in the pipeline under different flow conditions. (a) Case 1 (P = 233 kPa, m* = 26.3), (b) Case 2
(P = 235 kPa, m* = 29.6), (c) Case 3 (P = 186 kPa, m* = 14.2).

prediction is possible only if an accurate value for the time delay is value of data occurs) will be different in each case. Accordingly, the
considered. value of the Shannon entropy will be different and may be different
The increase in the Hurst exponent for the dilute mode of flow from one study to another depending on the rounding off. Thus, one
from the second to the third location is unexplained. Because for should be careful to compare with the literature results without
each increase in value of Hurst exponent, the flow cannot be treated knowing such information.
as a dense, this method may not be suitable to use compared with
the other two methods.
For the Shannon entropy method, the Shannon entropy for each
Conclusions
data set was calculated at three locations. For all data sets, the
variation in entropy at these locations was analyzed. The types of
Three signal analysis parameters—Hurst exponent, Shannon
variation that occurs most of the time in all the data sets are typi-
entropy, and size of phase space attractor—have been calculated at
fied by the variation occurring at these three locations. This method
three locations along the length of a pipeline for three distinct flow
is more accurate in predicting the transition but still has certain
conditions in the pneumatic conveying of fine particles. The mag-
limitations, which are discussed below.
nitudes of these parameters significantly vary for the two modes of
From the pressure fluctuations analyzed by Kang et al. (1999)
flow, namely dense and dilute modes of flow.
for a three-phase fluidized bed of glass beads (particle sizes ranged
The variation of these parameters also suggests that for a cer-
from 1 to 6 mm), the Shannon entropy was found to lie between
tain flow condition, the dense mode of flow changes to the dilute
2.3 and 3. The Shannon entropy was between 0.5 and 3 for differ-
mode at a certain location within the pipeline. The values of the
ent fluidization regimes of biomass particles (size 2.6 mm diameter
Hurst exponent and Shannon entropy decrease when the mode of
and 6 mm in length) analyzed by Zhong et al. (2009). Mittal et al.
flow changes from dense to dilute, while the phase space attractor
(2014) found the Shannon entropy to be between 2 and 4.5 for two
becomes more eccentric. The change in mode of flow can also be
powders (fly ash (2300 kg/m3 particle density and 30-␮m particle
confirmed from the magnitude of gas velocity calculated from the
diameter) and white powder (1600 kg/m3 particle density and 50-
experimental data at a specific location.
␮m particle diameter)) used in conveying. For the present work,
These transient parameters were calculated from the pressure
the values lie between 2 and 6. The conclusion is that the range in
signal data from transmitters set at different locations along the
Shannon entropy is different for different particle sizes. Also, if the
length of the pipeline. Their variation from one location to another
range of Shannon entropy is not known for a certain particle, the
helped in finding the approximate location of the mode transition
value should not be considered solely in predicting order or disor-
for pneumatic conveying through long pipelines.
der and hence the fluidization regime or mode of flow. This is the
From these three methods, the Shannon entropy analysis pro-
main limitation of this method.
vides a better indication of a change in mode of flow. This is because,
In this work, the pressure value is rounded off to the nearest
for different data sets, the variation in Shannon entropy at the three
integer (i.e., 160 kPa if p = 160.125 kPa and 163 kPa if p = 162.56 kPa).
locations is significant. From the Hurst exponent analysis, the vari-
However, if the rounding off is given to 1 decimal figure (160.125
ation of the Hurst exponent is not significant although for every
becomes 160.1) or 2 decimal figures (160.125 becomes 160.13),
increase in value of the Hurst exponent, the flow is less likely to be
then the probability of occurrence (the number of times a particular
in the dense mode.
88 S. J.S., N. Behera / Particuology 32 (2017) 82–88
References
Ahuja, P., Agarwal, P. H., Sethi, A. K., & Raj, U. (2005). Chaotic analysis of pressure fluc-
tuations in a gas-solid fluidized bed. International Journal of Chemical Technology,
12, 212–219.
Azizpour, H., Mostoufi, H. N., Zarghami, R., & Sotudeh-Gharebagh, R. (2011). Dynamic
characteristics of bubbling and turbulent fluidization using a Hurst analysis tech-
nique. 10th International Conference on Circular Fluidizing Beds and Fluidized Beds
and Fluidization Technology CFB-10.
Behera, N., Agarwal, V. K., Jones, M. G., & Williams, K. C. (2012). Transient parameter
analysis of fluidized dense phase conveying. Powder Technology, 217, 261–268.
Briongos, J. V., & Soler, J. G. (2005). State space analysis, a tool for progress. Computer
Aided Chemical Engineering, 20, 43–48.
Duan, F., & Cong, S. (2013). Shannon entropy analysis of dynamic behavior of Geldart
Group B and Geldart Group D particles in a fluidized bed. Chemical Engineering
Communications, 200(4), 575–586.
Hurst, H. E. (1951). Long-term storage capacity of reservoirs. Transactions of the
American Society of Civil Engineers, 116, 770–808.
Kang, Y., Woo, K. J., Ko, M. H., Cho, Y. J., & Kim, S. D. (1999). Particle flow behav-
ior in three-phase fluidized beds. Korean Journal of Chemical Engineering, 16(6),
784–788.
Kim, S. H., & Han, G. Y. (1999). An analysis of pressure drop fluctuation in a circulating
fluidized bed. Korean Journal of Chemical Engineering, 16(5), 677–683.
Mittal, A., Mallick, S. S., & Wypych, P. W. (2014). An investigation into flow mode tran-
sition and pressure fluctuations for fluidized dense-phase pneumatic conveying
of fine powders. Particuology, 16, 187–195.
Mittal, A., Mallick, S. S., & Wypych, P. W. (2015). An investigation into pressure fluc-
tuations for fluidized dense-phase pneumatic transport of fine powders. Powder
Technology, 277, 163–170.
Pahk, J. B., & Klinzing, G. E. (2008). Assessing flow regimes from pressure fluctuations
in pneumatic conveying of polymer pellets. Particulate Science and Technology,
26, 247–256.
Sedighikamala, H., & Zarghami, R. (2013). Hydrodynamic characterization of bub-
bling fluidization by principal component analysis of pressure fluctuations.
Particulate Science and Technology, 31(1), 51–57.
Shannon, C. E. (1948). A mathematical theory of communication. The Bell System
Technical Journal, 27, 379–423.
Sheikhi, A., Sotudeh-Gharebagh, R., Mostoufi, N., & Zargham, R. (2013). Frequency-
based characterization of liquid–solid fluidized bed hydrodynamics using the
analysis of vibration signature and pressure fluctuations. Powder Technology,
235, 787–796.
Williams, K. C., Jones, M. G., & Cenna, A. A. (2008). Characterization of the gas pulse
frequency, amplitude and velocity in non-steady dense phase pneumatic con-
veying of powders. Particuology, 6, 301–306.
Zhao, J. Y., Jin, N. D., Gao, Z. K., Du, M., & Wang, Z. Y. (2014). Attractor compari-
son analysis for characterizing vertical upward oil gas water three-phase flow.
Chinese Physics B, 23(3), 034702.
Zhong, W., & Zhang, M. (2005). Characterization of dynamic behavior of a spout-fluid
bed with Shannon entropy analysis. Powder Technology, 159(3), 121–126.
Zhong, W. Q., Jin, B. S., Zhang, Y., Wang, X. F., Zhang, M. Y., & Xiao, R. (2009). Shannon
entropy increment analysis on dynamic behavior of a biomass fluidized bed.
Journal of Physics: Conference Series, 147, 012071.