Powder Technology 208 (2011) 702–712

Contents lists available at ScienceDirect

Powder Technology
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p ow t e c

New method for the investigation of particle mixing dynamic in a spout-fluid bed
Yong Zhang, Wenqi Zhong ⁎, Baosheng Jin
School of Energy & Environment, Southeast University, Nanjing 210096, PR China

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

Article history: Detailed knowledge of particle mixing dynamic is essential in successfully designing and reasonably operating
Received 30 November 2010 spout-fluid reactor. Over the years, numerous methods for the investigation of particle mixing have been
Received in revised form 13 January 2011 proposed. This paper presents a novel method for the investigation of particle mixing. The method was carried
Accepted 24 January 2011
out in a well-designed microwave heater, in which the lab-scale spout-fluid vessel constructed with
Available online 1 February 2011
microwave-transparent Plexiglass was placed. Inside the bed, a number of polar tracer particles continuously
Keywords:
heated by microwave energy were real-time tracked using a top-grade infrared camera. From this, the
Mixing progress concentration field of the tracer particle could be obtained with high accuracy. This technique has been
Shannon entropy adopted to investigate the mixing dynamic of binary mixture differing in the density in a flat-bottom spout-
Microwave heating fluid bed. The mixing quality was evaluated by the Shannon entropy mixing index. On this basis, the influence
Infrared thermal imaging of important process parameters on the mixing progress in the initial mixing stage has been taken into
account. The investigation results show that the mixing progress is improved when increasing the means gas
velocity and the difference in density of bed particle. In addition, an empirical correlation as a function of gas
velocity and particle density was developed for calculating the mixing degree. The errors between
experimental data and the obtained ones from this correlation were within 28%.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction To date, there are many good studies on mixing of particles in gas–
solid systems [12–18]. In these work, the mixing process is often
A spout-fluid bed is a unique fluid-particle-agitated bed which assessed by observing the movement of each component during
overcomes the limitations of the spouted and fluidized bed by operation, whereas the mixing result is evaluated by mapping up the
simultaneously maintaining both spouting and fluidizing in the same spatial concentration distribution. For this reason, different experimen-
bed. This technique has been accepted traditionally as a solid–fluid tal methods have been developed over the years, which can be broadly
contacting method for physical operations such as drying, coating and divided into two categories: static method and dynamic method.
granulation of granular solids [1–6]. In recent years, the application of The static method is only to assess the mixing results after a given
spout-fluid bed has extended to catalytic reactors and combustion and mixing time. Typical example includes the bed-frozen method, which
gasification of coal and biomass [7–10]. was early applied by many researchers [12–16]. It is performed by a
Mixing of particles is deemed as a complex process to obtain a sudden stop of gas flow to allow particles to fall down under gravity,
uniform mixture of ingredients distributed among each other [11]. followed by the use of a proper method to divide the whole bed into
Irrespective of physical operation or chemical conversions, this process several sections and measure the concentrations in each section. The
is closely related to the rates of mass, heat and moment transfers among advantage of this method is that an entire concentration field can be
reactants and products. In general, good mixing among reactors is obtained directly and non-intrusively. However, the experiments must
essential to avoid hot spots due to the heat released by the highly be repeated a number of times to obtain reliable average values of the
exothermic reactions and segregation of the larger particles at the concentration and it is time consuming and very laborious. Another
bottom of the reactor. Whereas poor homogeneity of the particulate disadvantage is that it may suffer a certain deal of uncertainty because of
mass can lower the overall process efficiency and complicate its thermal the rearrangements associated with the transition from a fluidized to a
control. Therefore, a better knowledge and a quantitative description of fixed bed. To avoid the rearrangement in the radial direction, the new
the particle mixing behavior are of essential importance to understand bed-frozen method combined with the image processing technology
and to control the specific process and thus to eventually produce the was developed and successfully used in our early work [17,18], in which
desired product with maximum quality and yield. the flashboard-box was applied to separate the sampling regions along
the radial direction before freezing the bed.
The dynamic method can not only obtain the extent of mixing but
⁎ Corresponding author. Tel.: + 86 25 83794744; fax: + 86 25 83795508. also review the whole mixing process. According to whether they are
E-mail addresses: zyong@seu.edu.cn (Y. Zhang), wqzhong@seu.edu.cn (W. Zhong). disruptive or non-disruptive of the flow conditions within the system

0032-5910/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2011.01.011
 Y. Zhang et al. / Powder Technology 208 (2011) 702–712 703

being studied, the dynamic methods are classified as intrusive and However, in spite of the rapidly growing number of articles
non-intrusive methods. Initial intrusive method is based on the written on the subject, particle flow and particle mixing are topics
thermal tracing technique [19–21]. Valenzuela and Glicksman [19] that are far from being well understood, especially in the area of
used heated particles as tracers to study solids mixing in a fluidized mixing dynamic. One of the major reasons is for the lack of the
system. Subsequently, they [20] again employed particles cooled by successful approach toward the understanding of such complex flows.
liquid nitrogen as tracers to study the mixing characteristics in a 1/4 Whereas it requires reliable experimental data, which, in turn,
scale model of a pressurized bubbling fluidized bed combustor. While depends on the implementation of sophisticated experimental
Bellgardt and Werther [21] utilized carbon dioxide as the tracer techniques, which are capable of fast response to intense mixing as
material in a bed of quartz sand. Then, the temperature distribution well as the ability to provide the required information over the entire
detected by an array of thermistors indicated the particle concentra- flow field in the dense gas–solid system.
tion. Another intrusive method was developed by Wei et al. [22] by For this purpose, a new method has been developed for the
means of a phosphor tracer technique to measure solids mixing in a investigation of particle mixing dynamic. This method combined
variety of CFB units. The technique used the phosphor particles that microwave heating with infrared thermal imaging technique. The
served as tracer particles, which was excited by a strong flash and can former was used to continuously heat a cluster of polar tracers and the
be captured by the detector. The tracer concentration can be obtained latter was adopted to track the tracer particles and further obtained
from the recorded light intensity signal. Obviously, the main the tracer distribution. To obtain an empirical correlation predicting
uncertainty of intrusive method is that the monitoring probe that is the mixing degree at an arbitrary time and gas velocity, the effect of
intrusive to the system disrupts the internal flow patterns and hence gas velocity and particle density on the mixing progress in the initial
will have some effect on the mixing behavior. Therefore, several mixing stage has been studied.
researchers have avoided the probe interference problem by
developing externally detecting technique. 2. New measurement technique
A common non-intrusive method is based on the visual observation
[23,24]. During experiments, a high-speed camera was used to record 2.1. Design description
the mixing of tracers, which were marked with dye or fluorescent
powder. The resultant images allowed the direct observation of particle Traditionally, microwave heating was used to heat food. In recent
mixing behavior and evaluated the final mixing degree by means of the years, this kind of heating method combined with the fluidization or
image processing technique. The low costs and easy handling contribute spouting technology were employed in the dry process of agriculture
to this method. But it is really only suitable for measurements close to product [29–31]. Microwave heating technology is based on the
the walls, limiting the application to a three-dimensional units. PEPT transformation of alternating electromagnetic field energy into
(positron emission particle tracking) is a very effective and popular thermal energy by affecting the polar molecules of a material. The
method to investigate the particle mixing. These techniques are based major advantages of using microwave heating are rapid heat transfer,
on detecting the pairs of back-to-back γ-rays arising from annihilation volumetric and selective heating, compactness of equipment, speed of
of the emitted positrons. Applications of PEPT to the motion and mixing switching on and off and pollution-free environment as there are no
of particles in the fluidized bed were reported by some authors [25,26]. products of combustion.
In a recent study [27], an electromagnetic method was applied to study Typical microwave heating equipment is shown in Fig. 1. Basically
the mixing and segregation of spherical solids by monitoring single the systems of a microwave oven can be divided into three dominant
particle trajectories in a bubbling fluidized bed. The method utilized the sections, the power section, the microwave generator section and the
interaction between a magnetic field imposed on a fluidized bed oven cavity. The power section serves to increase the typical
chamber and a single metal covered tracer particle moving inside the household voltage, of about 120 V, to approximately 3000 V. The
bed. In addition, a radioactive measuring technique was also adopted to second section consists of a magnetron, a microwave tube in which
study the mixing. For example, Mostoufi and Chaouki [28] studied the electrons generated from a heated cathode are affected by magnetic
local axial and radial solids mixing. The tracers were made of a mixture and electric fields in such a way as to produce microwave radiation
of gold powder and epoxy resin. The detectors were arranged around and a waveguide, which causes the microwave to propagate into the
the fluidized bed and two personal computers simultaneously regis- oven cavity. At the end of the waveguide lying opposite of the
tered the number of γ-rays detected by each detector in every sampling magnetron there is the so-called oven cavity, in which the microwave
period of 20 ms. works on the product to be heated. The oven cavity is always a

Fig. 1. Typical microwave heating set-up.

704 Y. Zhang et al. / Powder Technology 208 (2011) 702–712

Fig. 2. Configuration of the oven cavity. (a) Rectangle-shaped internal cavity; (b) arc-shaped internal cavity.

metallic cavity ensuring that the radiation is reflected and stays inside perforations in the metal screw are much less than the microwaves'
the oven. Some waves go directly through the supporter (insulating wavelength, most of the microwave radiation cannot pass through the
material) and toward the produce (absorbing material), other bounce viewing port, while visible and infrared light (with a much shorter
off the metal walls and flooring. Thus, the microwave energy reaches wavelength) can. At the same time, the infrared glass, acting as the
all surfaces of the product heated from every direction. front wall of the spout-fluid bed, enables the infrared light through
Since the primary source of infrared radiation is heat or thermal the glass, which has no effect on the flow of particle.
radiation, any object which has a temperature radiates in the infrared.
Furthermore, the hotter the object, the more infrared radiation it emits. 2.2.3. Tracer particle
Infrared thermography is the technique that uses an infrared imaging As for the tracer particle, the essential requirement was that it
camera to detect the electromagnetic energy radiated in one infrared should be made of polar material, so that it could be heated by a
spectral band by an object and convert it into an electronic video signal microwave. The microwave heating ability of a material depends on
[32]. Hence, thermography makes it possible to “see” an object with or the dielectric constant and dielectric loss factors. In comparison with
without visible illumination. So it would be rather easy to detect a tracer the common polar material, Silicon carbide (SiC) has excellent
heated consecutively by microwave from a bed material in a spout-fluid properties such as high strength-to-density ratios, a high dielectric
bed and track it by means of infrared thermal imaging. constant (26.66) and a high dielectric loss factor (27.99) [33].
Thereby, we selected a SiC material as the main material of tracer.
2.2. Design details Considering the size and shape properties of the bed particles used,
the homemade tracer sphere was designed by thoroughly mixing
2.2.1. Internal cavity resin adhesive and SiC powders. Thus, the as-prepared mixture was
An internal cavity is a closed metal structure that confines weighted, blended, shaped and sized. In this way, three kinds of
microwave fields and prevents the microwaves from escaping. The tracers with different densities were obtained, as shown in Fig. 4.
conventional internal cavity is designed as a four-sided plane figure
with four right angles (shown in Fig. 2 (a)), which offers an even 3. Overall experimental setup, materials and procedure
distribution of microwave energy inside the oven and a high volume
to place big matter within it. For the current study, the heated objects 3.1. Overall experimental setup
are some millimeter-sized particles and move with a high velocity.
However, the preliminary test has revealed that it is extremely A schematic of the overall experimental setup is shown in Fig. 5.
difficult to effectively heat such small particles when it is placed in the The used spout-fluid bed had a cross-section of 100 mm × 30 mm and
conventional internal cavity. It may be due to the fact that this type of an overall height of 300 mm. The bottom of the bed was a perforated
internal cavity fails to concentrate microwave beam toward moving plate as air distributor, which was welded with the windboxes. A
targets and converge the microwave energy on one point. In reality,
microwave heating can be deliberately uneven by design. On the basic
of it, one designed new internal cavity shaped like a curve or an arch,
as shown in Fig. 2 (b). It successfully resolves this problem and allows
the moving small particle to be heated quickly.

2.2.2. Viewing port

In order to easily observe the flow and mixing behavior of the
tracer, the front side of the microwave oven should be equipped with
a viewing port. The design of the viewing port must meet the
following requirements: 1) the infrared can penetrate freely through
the viewing port which enables infrared observation and recording of
the flow. 2) The viewing port must strictly hinder external leaking of
microwaves and maintain the shielding to prevent harm to human
health. 3) The bed hydrodynamics would not be altered due to the
presence of the viewing port. For this reason, a special design was
conducted, in which the infrared glass and metal screw were put
together and they were tightened by means of screws (as shown in
Fig. 3). Also, this design contained two gaskets for the purposes of
insuring that no leaking of fluidizing gas would take place at the joint
surface of the bed and the infrared glass. Because the size of the Fig. 3. Configuration of the viewing port.
 Y. Zhang et al. / Powder Technology 208 (2011) 702–712 705

minimum spouting and fluidizing velocity was 11.11 and 0.82 m/s,
respectively. The tracer particles had the same diameter of 2.8 mm
and different densities of ρt1 = 1248 kg/m3, ρt2 = 1647 kg/m3 and
ρt3 = 2615 kg/m3, respectively. During all experimental runs, a bed
initial height of 0.1 m was maintained. Each run was conducted for a
desired period of time before the gas was turned off.

3.3. Procedure

The mixing experiments were conducted at room temperature and

atmospheric pressure. Before runs, the bed particles and tracer
particle were poured together into the bed from the particle inlet.
Meanwhile, the microwave generator was opened and its power was
adjusted until the heated tracers could be observed in the infrared
camera. Then, the gas velocity was increased to a desired operating
condition. During the whole mixing process, the real-time thermal
images including the mixing information were obtained and trans-
mitted on-line to the computer for treatment.
How to get the particle distribution of a certain region in the bed is
the key to the investigation of particle mixing. In our previous work
[17,18], the particle concentration was measured by a realizable
method combining flashboard box method and image processing
Fig. 4. Typical tracer particle. technology. Although the concentration profile of the bed in the
steady state could be obtained reasonably, error existed inevitably, in
metal screen was placed above the spout nozzle, which also served to particular in the evolution of the concentration with the time. This is
support the bed material. The spouting and fluidizing air were probably due to the delay, which was chiefly induced by the
supplied by a compressor and flow rates are measured by a set of flashboard box method. The reason why this method was adopted
rotameters. All the gas was dried in the drier consisting of a fixed bed to detect the tracer in the annulus region was that the digital camera
of silica gel high enough to remove humidity in the air before being only captured the mixing information close to the walls. In the current
used. A ball valve mounted on the air feed line helped to minimize work, the method combining the microwave heating and infrared
pressure oscillations and achieve a steady flow. thermal imaging technique can solve this intractable problem. This
The microwave generator operated at 2450 MHz. The generator can be demonstrated by the comparison of the images obtained by
output power was regulated between 0 and 1 kW. Images of solid two methods, as shown in Fig. 6. Hence, the particle concentration can
flows in the spout-fluid bed were recorded using a top-grade be obtained by the box-counting method. The results presented below
performance infrared camera (TH9260). The camera provides high were obtained using a sampling box of 5 mm in width and 5 mm in
resolution 640 × 480-pixel detector, and it could continuously record height (shown in Fig. 7).
or video record thermal image with a speed of 30 frames/s. The digital
image reading and processing was carried out by a computer. 4. Results and discussion

3.2. Materials 4.1. Illustration of mixing process

The bed materials utilized for testing were Polymer beans with a Below, we will present the spatial distribution of the tracer group
particle diameter of 2.8 mm and a particle density of 900 kg/m3. The at different times. To assist the discussion, the spouting gas velocity,

Fig. 5. Schematic diagram of the experimental system. (1) Compressor; (2) silica gel drier; (3) bypass valve; (4) control valve; (5) flow meter; (6) spout nozzle; (7) fluidizing gas
inlet; (8) windboxes; (9) ball valve; (10) spout-fluid bed volume; (11) infrared viewing port; (12) particle inlet; (13) infrared camera; (14) front door; (15) microwave generator;
(16) gas outlet; (17) computer; and (18) supporter.
706 Y. Zhang et al. / Powder Technology 208 (2011) 702–712

Fig. 6. Comparison of the digital image (a), which is from a previous study and the thermal image (b), which is from the current study.

fluidizing gas velocity and particle density are expressed as dimen- completely concentrated (t = 0 s) gradually gives way to one in which
sionless variables, given by: the tracer particles diffuse toward the central region of the bed and
begins to mix with the bed particles. As mixing proceeds, the spout


us = us = ums ð1Þ gas penetrates through the bed and two symmetrical circulation rolls
form at both sides of the bed. Following this, more and more tracer


uf = uf = umf ð2Þ particles travel into the spout region and along the spout, some of
them move to the fountain region, which are mainly caused by the


ρ = ρt = ρb ð3Þ entrancement of spouting gas. During this process, particle mixing is
only limited to the left of the bed and no tracers have been found on
in which us and uf are, respectively, the actual spouting and fluidizing the other side of the bed. Then, the tracer particles fall down at the
gas velocities; ums and umf are the minimum spouting and fluidizing surface of the annulus and run downwards with the bed particles in
gas velocities, respectively; ρt and ρb are the densities of the tracer the annulus. Once the tracer particles finish the journey of one
and bed particles, respectively. circulation, they mix vigorously with the bed particle.
Typical thermal images obtained during the experiment are shown Therefore, the whole process can be divided into two subsequent
in Fig. 8, which displays the single heated tracer at various regions. stages: the initial mixing stage and the stable mixing stage. In the first
The region that the tracer is located is determined by the digital photo stage, the circulation for the bed particle has been formed but it still
obtained from our previous study [17]. It is evidence that the particle has not been formed for the tracer particle. Here, most of the tracer
temperature in the annulus is higher than the one in the spout and particles travel only along the route of circulation roll, which has been
fountain, which is mainly induced by the cooling effect of air with a explained in our previous work [34]. After then, when the circulation
high velocity. There is no obvious difference in particle temperature for the tracer particle has been formed, nearly one half of the tracer
for these tracers in the same region. Provided that the tracer heated by particles have moved into the right side of the bed. After above initial
the microwave has a temperature over room temperature, it can be mixing stages, a dynamical equilibrium mixing state is established
detected by the infrared camera. Hence, whether the particle and the degree of mixing does not change with time.
temperature is high or low, it has no influence on the final results. As already mentioned, the final result of the particle tracking was the
Fig. 9 shows an illustrative example of the mixing for particles of distribution of those heated tracer particles. Considering that most
1647 kg/m3 in density and 2.8 mm in diameter, which generates some factors relating to the mixing behavior can be explained in terms of
visual impression of the mixing process. Obviously, initial state of probabilities, it could be appropriate to take up mixing phenomena from

Fig. 7. Dimension of samples in the bed.

 Y. Zhang et al. / Powder Technology 208 (2011) 702–712 707

The probability that a tracer particle goes into bin i can be

calculated by dividing ni by the overall system population:

pi = ni = nt : ð5Þ

Using the probability of Eq. (5), we then calculate the Shannon

entropy:

m
S = − ∑ pi lnðpi Þ: ð6Þ
i=1

When the tracer is existing only in one cell and there is no mixing
in the bed, the condition of minimum Shannon entropy corresponds
to this state, which can be given by:

Smin = 0: ð7Þ

When the tracer is distributed uniformly and its concentration is

uniform everywhere, the Shannon entropy achieves its maximum
values, which is defined as:

Smax = lnð jÞ: ð8Þ

Therefore, the rational measure of mixing can be defined as the

ratio of the Shannon entropy increase at a certain location to the
maximum possible entropy increase, as represented by the following
equation

Sðt Þ−Smin
S⁎ðt Þ = ð9Þ
Smax −Smin :

Thus, the quality of mixedness can vary from 0 to 1. The Shannon

entropy increase is a good indicator of mixing performance.

4.2. Effect of gas velocity

Little attention has been devoted in the mixing literature to the

effect of the mixing time on the mixing process and, as mentioned
above, previous studies have focused on the effects of gas velocity on
the mixing behavior [35–37]. In these simulations, four operational
conditions are used, which are case 1 of u⁎s = 0.82 and u⁎ f = 0.78, case
2 of u⁎s = 0.87 and u⁎ f = 1.14, case 3 of u⁎ s = 0.69 and u⁎f = 0.26 and
case 4 of u⁎s = 1.33 and u⁎ f = 0.66.
Fig. 10 (a) shows the evolution of the Shannon entropy increase
with time at different gas velocities for tracer 2 with a particle density
of 1647 kg/m3. It can be found that for all four cases the evolution of
the Shannon entropy increase follows the similar trend. That is, the
Shannon entropy increase first increases from zero to an equilibrium
Fig. 8. Typical thermal images obtained during the experiment. (a) In the annulus; value and then fluctuates around a high value. Based on the definition
(b) in the spout; and (c) in the fountain. in our previous study, the former stage is called as the initial mixing
stage and the later the stable mixing stage. In the meantime, it can also
be noted that the gas velocities exert a great influence on the
the standpoint of probability theory. To evaluate the degree of scatter development of the Shannon entropy increase. On the one hand, the
from the distribution of heated tracer particles, we introduce the maximum Shannon entropy increase implying the mixing degree at
Shannon entropy or information entropy, which is a measure of the the dynamical equilibrium state varies with the operating parameter.
uncertainty associated with a random variable. Furthermore, the In early studies, the change of spouting gas velocity or fluidizing gas
Shannon or information entropy has been used extensively in many velocity always leads to the monotonic change of mixing degree. But
different scientific areas as a measure of disorder or lack of information. for such condition with different spouting and fluidizing gas velocity,
The bed is divided into many bins i, i = 1, 2…m, with a size of the entropy increase changes non-monotonically. For example, the
5 mm × 5 mm and for each effective bin i, i = 1, 2…j (1 ≤ j ≤ m), we maximum Shannon entropy increase reduces considerably from 0.82
evaluate the number ni, of the tracer particle. The total number of to 0.52 when the u⁎ f is increased from 0.66 to 0.78. It then increases
tracer particles is given by: from about 0.68 to 0.82 when the u⁎ f is increased from 0.78 to 1.14.
On the other hand, the mixing time required for the entropy
increase to enhance from 0 to a mixing equilibrium in the initial
j
nt = ∑ni : ð4Þ mixing stage varies with the gas velocity. When us⁎ = 0.69 and
i=1 u⁎f = 0.26, it seems to take 11 s for the Shannon entropy increase to
708 Y. Zhang et al. / Powder Technology 208 (2011) 702–712

Fig. 9. Instantaneous snapshots of particle mixing process with time (u*s = 1.11, and u*f = 0.23). (a) t = 0 s; (b) t = 0.767 s; (c) t = 1.633 s; and (d) t = 3.967 s.

enhance from 0 to a mixing equilibrium. But at us⁎ = 1.33 and 4.3. Effect of particle density
u⁎f = 0.66, it only takes about 5 s to reach the mixing equilibrium state.
Due to the difference in the maximum entropy increase, it is very In these experiments, the particle diameter is kept at 2.8 mm while
difficult to significantly compare the mixing rate at variant operating the particle density of the tracer particle is increased from 1248 to 1647,
conditions. To make the Shannon entropy increase of the different gas then to 2615 kg/m3. The change of the Shannon entropy increase with
velocities regular and consistent, we normalize the Shannon entropy time at us⁎ = 0.87 and u⁎f = 1.14 is displayed in Fig. 11 (a). It can be
increase as the dimensionless form P = S* / S*e, in which S*e is the observed obviously that though a similar trend is found among these
maximum Shannon entropy increase at the mixing equilibrium state. curves, some quantitative differences exist, which depend on the tracer.
Thus, the Shannon entropy increases for all the cases examined are That is to say, an increase in the density of the tracer particle results in a
brought into the range 0 to 1 in the first stage and kept at drastic change in the variance of the Shannon entropy increase with
approximately 1 in the second stage. The results are illustrated in time.
Fig. 10 (b). Clearly, the mixing time for case 4 is the shortest, implying For one thing, the maximum entropy increase reduces from round
that the mixing rate for case 4 is higher than that for other cases. This about 0.72 to 0.59 when the density of the tracer particle is increased
may be attributed to the influence of the spouting gas velocity, which from 1248 to 2615 kg/m3. This indicates that the achievable
is the main driving force for particle circulation. While the mixing homogeneity is strongly dependent on the density difference between
time for case 3 is the longest and the mixing time for cases 1 and 2 lie the bed particle and the tracer particle. There is a significant decrease
between case 3 and 4. in the mixing degree when the density of the tracer increases. This is
 Y. Zhang et al. / Powder Technology 208 (2011) 702–712 709

Fig. 10. Evolution of (a) Shannon entropy increase and (b) dimensionless Shannon Fig. 11. Evolution of (a) Shannon entropy increase and (b) dimensionless Shannon
entropy increase with time at different gas velocities. entropy increase with time for different tracers.

corroborated by observations in the experiment that the heavier 4.4. Quantifying the dynamic of mixing
tracer particles have been found to frequently congregate at the upper
inner part of the annulus. These phenomena become more pro- Since the mixing degree changes significantly in the initial mixing
nounced with increasing the difference in particle density of the bed stage, we limit our discussion in the initial mixing stage of interest and
particle and tracer particle. Such phenomena coincide with the result take into account the mixing progress in this stage. A clue to elucidate
obtained by means of the combination of a box-counting method and mixing progress, considering the results shown in Figs. 10 (b) and 11 (b),
digital image processing technique [38]. The cause of such a mixing it is appropriate to assume a relationship between the dimensionless
pattern can be better explained by the fact that the heavier tracer Shannon entropy increase, P, and time elapsed, t, as follows:
particles are prone to land at smaller radial positions and thus to
subsequently achieve a shorter path through the annulus. P = 1−expð−kt Þ: ð10Þ
For another, there seems to be a difference in the mixing time in
the initial mixing stage for the three tracers. However, it is very hard Here, k is an important parameter, which has a relation to the
to clearly discern the difference between these curves, especially for operating condition. Note that, in this equation, the relationship can
the quantitative difference. Fig. 11 (b) presents the change of be made linear by transforming. In an attempt to calculate the
dimensionless form P with time for the different tracer particles. It coefficient k, we can take the logarithm of each side of Eq. (10),
can be noted that the mixing time for tracer 3 is smaller than those for making the relation a bit clearer. Thus, this equation can also be
tracers 1 and 2. Thus, the mixing speed for tracer 3 is much higher expressed as:
compared with tracers 1 and 2. That is to say, the heavier particles
have a higher mixing rate in the initial mixing stage. These finds are lnð1−P Þ = −kt: ð11Þ
qualitatively supported by our previous visual observation [38]. This
can be attributed to the following reasons. As indicated above, the And, after calculating the average P value of several measurements
heavier tracer particles are prone to land at smaller radial positions. As at each operating condition, it is confirmed that Eq. (11) shows clearly
a result, these particles pass through a small cycle roll, even a short the relationship between P and t by plotting ln(1 − P) vs. t.
cycle [39], which covers a short route. What's more important is that Fig. 12 exhibits the evolution of ln(1 − P) with time. It can be
most of the regions traveled by the tracer particle lie in the spout observed that although data are scattered, the general trend is that the
region and the internal annulus, where the velocity of particle is value of ln(1 − P) decreases linearly with the increase of time.
higher compared with other regions [34]. Therefore, by linear fitting a series of scattered data, we can determine
710 Y. Zhang et al. / Powder Technology 208 (2011) 702–712

logarithm of each side of Eq. (12). As a result, Eq. (12) could be

changed into the following formation:



lnðkÞ = lnðaÞ + b lnðum Þ + c ln ρ : ð13Þ

For the given tracer, it is constant for the third item of the right side
of Eq. (13). If ln(um) is considered as the independent variable, ln(k) is
the dependent variable. Hence, Eq. (13) is deemed as the linear
equation, whose graph in the Cartesian coordinate plane is a straight
line. The constant b is obtained by calculating the slope of the straight
line. The scatter point for the growth of ln(k) with ln(um) is presented
in Fig. 13. By means of least-squares linear fitting, a straight line can be
plotted, and whose slope is just the value of b. It can be seen from
Fig. 13 that there are different fitting straight lines for each tracer
particle. But the slope for each fitting straight line is almost the same.
The real value of b is by calculating the average of these slopes so as to
minimize the error. Using the same method, c can also be determined
for fixed operating condition, which is shown in Fig. 14. Thus, Eq. (12)
can be given

0:48
0:43
k = aum ρ : ð14Þ

Fig. 15 reveals clearly the relationship between k and u0. 48

m ρ*
0.43
. It
can be found that the value of k increases gradually with the increase of
u0. 48
m ρ*
0.43
. This trend agrees with the results in Fig. 10 (b) and 11 (b).
That is to say, at fixed gas velocity, the index of mixing rate, k, grows
progressively when increasing the density of the tracer particle. At the
same time, for the fixed tracer particle, k increases with the increase of
gas velocity. Doing least-squares linear fitting these scatter date, we can
determinate the value of a, which is just the slope of the fitted straight
line and also shown in Fig. 15.
Thus, Eq. (14) could be expressed as:

0:48
0:43
k = 0:548um ρ : ð15Þ

Finally, a model predicting the mixing progress could be obtained

Fig. 12. Evolution of ln(1 − P) with time. (a) At various gas velocities; and (b) for
different tracers.  
0:48
0:43
P = 1−exp −0:548um ρ t : ð16Þ

The comparison of experimental results at various tracer condi-

tions with the calculations by Eq. (16) is given in Fig. 16. The results
show that most of the predictions calculated by Eq. (16) are in
the value of k, which is the slope of the fitted line and also shown in agreement with the present experimental data with a deviation
Fig. 12. From Fig. 12 (a), one can find that the value of k depends within 15%, but some of the predictions have relatively large deviation
greatly on the operating condition. For case 4, it is 1.84 and greater that is in the range of 15%–28%. Under the condition of high P value,
than that of the other cases. For cases 2 and 1, it is 1.57 and 1.28, the predictions only limit within the range of 5%. On the whole, the
respectively, greater than that of case 3, which is the lowest. The
results agree with the findings in Fig. 10.
It can be observed from Fig. 12 (b) that the density of the tracer
particle exerts an effect on the value of k. When using the heaviest
tracer particle 3, k achieves the maximum value of 2.04. It is on the
contrary for tracer 1. The results agree with the findings in Fig. 11. The
above statement shows that k is able to quantitatively indicate the
mixing rate.
According to the analysis above, spouting gas velocity, fluidizing
gas velocity and particle density take effect on the development of
mixing degree with time. Thus, the mixing coefficient, k, should be the
function of these effects, which can be given as the following
formation:

b
c
k = aum ρ ð12Þ

where, a, b and c are the constants and um is the mean gas velocity,
which is a function of spouting and fluidizing gas velocity. Eq. (12)
allows us to determine the value of these constants by taking the Fig. 13. Effect of mean gas velocity on the mixing coefficient k.
 Y. Zhang et al. / Powder Technology 208 (2011) 702–712 711

1.0
+5%
Tracer 1
+15% -5%
0.8 Tracer 2
Tracer 3
-15%

[S*/S*e]cal (-)
0.6 +28%

0.4
-28%

0.2

0.0
0.0 0.2 0.4 0.6 0.8 1.0
[S*/S*e]exp (-)

Fig. 14. Effect of tracer density on the mixing coefficient k.

Fig. 16. Calculated mixing degree using Eq. (16) vs. the results of the experiments of this
work.
model prediction gives a reasonable agreement with the experimental
data. obtained at an arbitrary time and gas velocity. The errors between
The above discussion indicates that particle density and gas experimental data and the obtained ones from this correlation were
velocity take effect on the development of mixing degree. Other within 28%. Further investigations are scheduled for this new method
parameters such as particle size, shape, the ratio between bed amount to determine the effect of varying a set of operating conditions such as
and tracer amount, and configuration of the fluidizing gas distributor particle size and shape on mixing characteristics of the bed particles.
affect the mixing behavior of the particle [13,15,40], and affect the
mixing dynamic. However, the present work is not able to perform List of symbols
this valuable step since experiments cover larger range of bed k mixing coefficient (−)
geometry and operating conditions are needed. nt total number of tracer particles (−)
pt probability that a tracer particle goes into bin i (−)
5. Conclusions P dimensionless Shannon entropy increase (−)
S Shannon entropy (−)
A new method for particle tracking in the gas–solid system by Smin minimum Shannon entropy (−)
means of microwave heating and infrared thermal imaging technique Smax maximum Shannon entropy (−)
is presented and shown to be an efficient tool for characterization of S* Shannon entropy increase (−)
the particle mixing process. In comparison with most existing um mean gas velocity (m/s)
methods, the new approach has the ability to penetrate into dense umf minimum fluidizing gas velocity (m/s)
particulate systems and further provides the real-time information uf fluidizing gas velocity (m/s)
about particle concentration over the entire flow field, which u*f dimensionless fluidizing gas velocity (−)
significantly reduces the time consumed and labor intensity. This ums minimum spouting gas velocity (m/s)
technique has been adopted to investigate the mixing dynamic of us spouting gas velocity (m/s)
binary mixture differing in the density in a flat-bottom spout-fluid us* dimensionless spouting gas velocity (−)
bed. The influence of two parameters on the mixing progress in the ρt tracer density (kg/m3)
initial mixing stage has been studied. The investigation results show ρb bed particle density (kg/m3)
that the mixing progress is improved when increasing the means gas ρ* dimensionless particle density (−)
velocity and the difference in density of the bed particle. Based on it,
the empirical correlation for calculating the mixing degree was

Acknowledgement

Financial support from the National Natural Science Foundation of

China (50706007, and 51076029), the National Key Program of Basic
Research in China (2010CB732206), the Foundation of Excellent
Young Scholar of Southeast University (4003001039) and the
Collaboration Project of China and British (2010DFA61960).

References
[1] J. Link, C. Zeilstra, N. Deen, H. Kuipers, Validation of a discrete particle model in a
2D spout-fluid bed using non-instrusive optical measuring techniques, Can. J.
Chem. Eng. 82 (2004) 30–36.
[2] M. Zielinska, M. Markowski, Drying behavior of carrots dried in a spout-fluidized
bed dryer, Dry Technol. 25 (2007) 261–270.
[3] I. Białobrzewskia, M. Zielińskaa, A.S. Mujumdarb, M. Markowskia, Heat and mass
transfer during drying of a bed of shrinking particles-simulation for carrot cubes
dried in a spout-fluidized-bed drier, Int. J. Heat Mass Tran. 51 (2008) 4704–4716.
[4] K.B. Park, J.L. Plawsky, H. Littman, J.D. Paccione, Mortar properties obtained by dry
premixing of cementitious materials and sand in a spout-fluid bed mixer, Cement
Fig. 15. Evolution of the mixing coefficient k with u0.
m ρ
48 * 0.43
. Concr. Res. 36 (2006) 728–734.
712 Y. Zhang et al. / Powder Technology 208 (2011) 702–712

[5] J.L. Plawsky, H. Littman, Design and simulation of a spout fluid bed coating system, [24] L.H. Shen, J. Xiao, F. Niklasson, F. Johnsson, Biomass mixing in a fluidized bed
Excerpt from the Proceedings of the COMSOL Users Conference, 20068, Boston. biomass gasifier for hydrogen production, Chem. Eng. Sci. 62 (2007) 636–643.
[6] J.M. Link, W. Godlieb, N.G. Deen, J.A.M. Kuipers, Discrete element study of [25] M. Van de Velden, J. Baeyens, J.P.K. Seville, X. Fan, The solids flow in the riser of a
granulation in a spout-fluidized bed, Chem. Eng. Sci. 62 (2007) 195–207. Circulating Fluidised Bed (CFB) viewed by Positron Emission Particle Tracking
[7] M.S.J. Arnold, M.K. Laughlin, The British coal spout-fluid bed gasification process, (PEPT), Powder Technol. 183 (2008) 290–296.
Can. J. Chem. Eng. 70 (1992) 991–997. [26] M. Van de Velden, J. Baeyens, K. Smolders, Solids mixing in the riser of a
[8] C.J. Lim, A.P. Watkinson, G.K. Khoe, S. Low, N. Epstein, J.R. Grace, Spout, fluidized, circulating fluidized bed, Chem. Eng. Sci. 62 (2007) 2139–2153.
spout-fluid bed combustion of bituminous coal, Fuel 67 (1988) 1211–1217. [27] M. Wirsum, F. Fett, N. Iwanowa, G. Lukjanow, Particle mixing in bubbling fluidized
[9] R. Xiao, M.Y. Zhang, B.S. Jin, Y.J. Huang, H.C. Zhou, High-temperature air/ beds of binary particle systems, Powder Technol. 120 (2001) 63–69.
steamblown gasification of coal in a pressurized spout-fluid bed, Energ. Fuel. 20 [28] N. Mostoufi, J. Chaouki, Local solid mixing in gas–solid fluidized beds, Powder
(2006) 715–720. Technol. 114 (2001) 23–31.
[10] W.Q. Zhong, M.Y. Zhang, B.S. Jin, X.P. Chen, Flow pattern and transition of [29] H. Feng, J. Tang, R.P. Cavalieri, Combined microwave and spouted bed drying of
rectangular spout-fluid bed, Chem. Eng. Process. 9 (2006) 734–746. diced apples: effect of drying conditions on drying kinetics and product
[11] J. Gyenis, Assessment of mixing mechanism on the basis of concentration pattern, temperature, Dry Technol. 17 (1999) 1981–1998.
Chem. Eng. Process. 38 (1999) 665–674. [30] H. Feng, J. Tang, Microwave finish drying of diced apples in a spouted bed, J. Food
[12] W.C. Yang, D.L. Kealrns, Rate of particle segregation in a gas fluidized bed, Ind. Eng. Sci. 63 (1998) 679–683.
Chem. Fundam. 21 (1982) 228–235. [31] Z.H. Wang, G. Chen, Theoretical study of fluidized-bed drying with microwave
[13] A. Marzocchella, P. Salatino, V. Di Pastena, L. Lirer, Transient fluidization and heating, Ind. Eng. Chem. Res. 43 (2004) 3433–3438.
segregation by size difference of binary mixtures of particles, AIChE J. 46 (2000) [32] T. Astarita, G. Cardone, G.M. Carlomagno, C. Meola, A survey on infrared
2175–2182. thermography for convective heat transfer measurements, Opt. Laser Technol.
[14] G. Olivieri, A. Marzocchella, P. Salatino, Segregation of fluidized binary mixtures of 32 (2000) 593–610.
granular solids, AIChE J. 50 (2004) 3095. [33] A. Chatterjee, T. Basak, K.G. Ayappa, Analysis of microwave sintering of ceramics,
[15] M.C. Leaper, J.P.K. Seville, N. Hilal, S.W. Kingman, A.S. Burbidge, Investigating the AIChE J. 44 (1998) 2302–2311.
dynamics of segregation of high jetsam binary batch fluidised bed systems, Chem. [34] Y. Zhang, B.S. Jin, W.Q. Zhong, R. Bin, R. Xiao, DEM simulation of particle mixing in
Eng. Process. 43 (2004) 187–192. flat-bottom spout-fluid bed, Chem. Eng. Res. Des. 88 (2010) 757–771.
[16] S.H. Schaafsma, T. Marx, A.C. Hoffmann, Investigation of the particle flowpattern [35] M.J. Rhodes, X.S. Wang, M. Nguyen, P. Stewart, K. Liffman, Study of mixing in gas-
and segregation in tapered fluidized bed granulators, Chem. Eng. Sci. 61 (2006) fluidized beds using a DEM model, Chem. Eng. Sci. 56 (2001) 2859–2866.
4467–4475. [36] G. Kwant, W. Prins, W.P.M. van Swaaij, Particle mixing and separation in a binary
[17] Y. Zhang, B.S. Jin, W.Q. Zhong, Experiment on particle mixing in flat-bottom spout- solids fluidized bed, Powder Technol. 82 (1995) 279–291.
fluid bed, Chem. Eng. Process. 48 (2009) 126–134. [37] Y.Q. Feng, A.B. Yu, Microdynamic modeling and analysis of the mixing and
[18] Y. Zhang, B.S. Jin, W.Q. Zhong, Experimental investigations on the effect of the segregation of binary mixtures of particles in gas fluidization, Chem. Eng. Sci. 62
tracer location on mixing in a spout-fluid bed, Int. J. Chem. React. Eng. 6 (2009) (2007) 256–268.
A46. [38] B.S. Jin, Y. Zhang, W.Q. Zhong, R. Xiao, Experimental study of the effect of particle
[19] J.A. Valenzuela, L.R. Glicksman, An experimental study of solids mixing in a freely density on mixing behavior in a spout-fluid bed, Ind. Eng. Chem. Res. 48 (2009)
bubbling two dimensional fluidised bed, Powder Technol. 38 (1984) 63–72. 10055–10064.
[20] L.R. Glicksman, E. Carr, P. Noymer, Particle injection and mixing experiments in a [39] W.Q. Zhong, Y. Zhang, B.S. Jin, Novel method to study the particle circulation in a
one-quarter scale model bubbling fluidized bed, Powder Technol. 180 (2008) 284. flat-Bottom spout-fluid bed, Energ. Fuel. 24 (2010) 5131–5138.
[21] D. Bellgardt, J. Werther, A novel method for the investigation of particle mixing in [40] R. Escudie, N. Epstein, J.R. Grace, H.T. Bi, Effect of particle shape on liquid-fluidized
gas solid systems, Powder Technol. 48 (1986) 173. beds of binary (and ternary) solids mixtures: segregation vs. mixing, Chem. Eng.
[22] F. Wei, Y. Jin, Z.Q. Yu, W. Chen, S. Mori, Lateral and axial mixing of the dispersed Sci. 61 (2006) 1528–1539.
particles in Cfb, J. Chem. Eng. Jpn 28 (1995) 506–510.
[23] S. Karlsson, I.N. Bjorn, S. Folestad, A. Rasmuson, Measurement of the particle
movement in the fountain region of a Wurster type bed, Powder Technol. 165
(2006) 22–29.