- Email: [email protected]
Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking
Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking Anna K¨ohler, Alexander Rasch, David Pallar`es, Filip Johnsson PII: DOI: Reference:
S0032-5910(16)30973-1 doi:10.1016/j.powtec.2016.12.093 PTEC 12241
To appear in:
Powder Technology
Received date: Revised date: Accepted date:
14 May 2016 28 October 2016 30 December 2016
Please cite this article as: Anna K¨ohler, Alexander Rasch, David Pallar`es, Filip Johnsson, Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking, Powder Technology (2017), doi:10.1016/j.powtec.2016.12.093
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking Anna Köhler*, Alexander Rasch, David Pallarès and Filip Johnsson
IP
T
Dept. of Energy and Environment, Chalmers University of Technology, SE-412 96 Göteborg, Sweden
SC R
Tel: +46-(0)31-772-14 42
NU
Email: [email protected]
Abstract
MA
A Magnetic Particle Tracking (MPT) system is applied to a bubbling fluidized bed to study how axial mixing and segregation of fuel are influenced by the fuel density and operational conditions
D
(fluidization velocity, bed height and pressure drop across the gas distributor). The MPT system is
TE
used to determine the vertical distribution of the tracer particle in a fluid-dynamically down-scaled cold unit resembling a 0.74×0.74 m2 fluidized bed reactor operating at 800 °C. This work uses a tracer
CE P
particle of 10 mm in diameter, corresponding to a fuel particle of 44 mm. Different tracer particles are applied with solids density representing biomass, biomass char and that of the average bulk. The MPT
AC
system yields a spatial accuracy in the order of 10-3 m and a time resolution of 10-3 s. For the operational range investigated, three fuel segregation regimes can be identified from the MPT measurements: 1) A flotsam regime which occurs at low fluidization velocities and for low density tracer particles, 2) A transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of tracer particle at the bed surface and the splash zone remains constant with fluidization velocity. The transition velocities between the regimes depend on bed height and density of the tracer particle. Keywords: Fuel segregation; Fuel mixing; Solids mixing; Magnetic particle tracking; Fluiddynamical scaling; Fluidized bed
1
ACCEPTED MANUSCRIPT 1. Introduction Although previous research related to coal combustion in fluidized-bed (FB) boilers have given
T
valuable information on the mixing of gas and solids in fluidized beds, the application of the FB
IP
technology to biomass conversion requires more in-depth understanding of mixing, especially with
SC R
respect to mixing of the fuel. This, since biomass is a high volatile fuel where moisture and volatiles may be released before the fuel has been distributed evenly over the bed cross section and within the bed causing oxygen depleted zones and uneven combustion. In indirect biomass gasification, control
NU
of the fuel mixing process is of importance in order to optimize the gasification process while letting the remaining char be transferred to the combustion bed. Biomass particles are more prone to
MA
experience axial segregation from the bed material as they have a lower density and are larger in size than coal particles. An increase in axial segregation results in more fuel particles floating on the dense
D
bed surface, which implies enhanced bed-to-fuel mass transfer (due to the absence of particle-related
TE
tortuosity) and reduced bed-to-fuel heat transfer (due to absence of convection from bulk particles). The decreased bed-to-fuel heat transfer has been experimentally shown to yield less reactive char
CE P
particles [1]. Furthermore, fuel present in the splash zone will experience faster lateral mixing than if immersed in the bed, which is often beneficial in boilers but strongly undesired in other applications
AC
(indirect gasification, chemical looping combustion). Different types of particles are present in fluidized bed reactors used for combustion and gasification, but can be divided into finer and denser bulk bed material (e.g. fuel ashes, makeup sand, sorbent, catalytic material), and lighter and typically larger fuel particles. The latter are changing in both density and size during fuel conversion. In addition, there is a large variety of biomass fuels, with different shape, composition and density. Mixing and segregation is caused by the bubble activity in the bed and thereby highly influenced by operational conditions, i.e. by fluidization velocity, bed height, pressure drop across the gas distributor and by the physical properties of the solids [2] [3]. As biomass is a low-density fuel it tends to show a flotsam behaviour, concentrating at the top of the bed [4] [5]. Increasing fluidization velocity may decrease fuel segregation and allow sufficient axial mixing. There are several important experimental studies on solids mixing in literature. Rowe et al. [6]
2
ACCEPTED MANUSCRIPT studied binary particle mixtures and showed that particles of higher density and size tend to show a jetsam behaviour, i.e. to end up in the bottom of the bed. This work was followed by more recent studies [7] [8] discussing the presence of different fluidization regimes for beds of binary mixtures of
IP
T
solids. Nienow et al. [9] studied the segregation of a single large and light object immersed in a fluidized bed, and they found segregation to be mainly caused by buoyancy forces. Nienow et al.
SC R
created rising velocities of the object immersed in the bed by releasing it close to the distributor and noting the time it took for the object to reach the bed surface. Optical measurement methods have also
NU
been used to study solids mixing of large particles immersed in pseudo-2D beds [10-12]. At present, Radioactive (or positron emission) Particle Tracking (RPT) can be applied to follow tracers with high
MA
accuracy in all 3 dimensions, which has been used under cold conditions in laboratory units [13-15] and in what was referred to as a cold pilot-scale unit [16]. Moslemian et al. [17] used RPT in a labscale CFB. Mohs et al. [18] evaluated several particle tracking methods for use in laboratory scale (i.e.
TE
D
ambient conditions), and they choose Magnetic Particle Tracking (MPT) before RPT. The main benefits of MPT are that both position and orientation of the tracer particle can be measured in all 3
CE P
dimensions, while it is relatively cheap and does not require any extensive safety precautions [19-21]. Combined with fluid-dynamical scaling, MPT allows lab-scale measurements covering conditions, which are difficult to measure in commercial units. Thus, when investigating fuel mixing a wider
AC
range of density of the fuel tracer particle as well as a wider range of operational condition can be covered compared to an industrial boiler or gasifier but of course with the limitation that effects from the fuel conversion itself is not included. Yet, from experiments in a hot industrial scale biomass gasifier [22], focusing on fuel mixing on the bed surface, it was concluded that fuel buoyancy effects from volatile release seemed to be small compared to the mixing caused by the bubble flow. In a previous work by the authors of this work [23], it was shown that MPT can give continuous data with a spatial accuracy of ~1 mm and temporal resolution of up to 1 kHz. The determination of axial profiles and velocities in and above the bed can be directly derived from the measurement data after being smoothened with Kalman filtering [24].
3
ACCEPTED MANUSCRIPT While previous works in literature give valuable insights in the behaviour of FB solids mixing, they are all based on experiments in small lab-scale units operated at cold conditions, i.e. under conditions which cannot be directly translated to industrial conditions. The aim of the present work is to
IP
T
understand the basics of fuel mixing regimes for different operating conditions, focusing on the axial mixing of biomass fuel particles at different stages of conversion (i.e. different densities). In order to
SC R
resemble industrial operation, the system (including the tracer particle) is scaled to represent the density and size of such fuel but also other tracer densities are applied for comparison. The work
NU
applies the previously developed MPT technology described in [23, 25]. The work given in [23, 25] provides first measurements and a description of the MTP measurement technique, including the
MA
possibilities of implementing MPT in fluid-dynamically down-scaled conditions and a discussion of the limitations of the method. This contribution of the present work is to provide the first highlyresolved fluid-dynamically down-scaled measurements with the aim to increase the understanding of
fluidization velocity).
TE
D
axial solids segregation in fluidized beds (including the impact of tracer density, bed height and
CE P
2. Experimental setup
The experiments in the present work cover three tracer densities ranging within 350-1 230 kg/m3. In
AC
addition, a tracer with a density (2 130 kg/m3) similar to that of the bed solids was investigated but for the conditions studied the tracer sunk to the air distributor immediately after the onset of fluidization, where it then remained, i.e. exhibited a purely jetsam behaviour. The MPT experiments were carried out in a fluidized bed fluid-dynamically down-scaled according to a full set of scaling rules given by Glicksman et al. [26]. The bed, which is schematically shown in Figure 1, is fluidized with pressurized air entering a wind box and then the bed through a perforated plate. With a resulting length scaling factor of 4.4, the cold lab unit (0.17×0.17 m2) resembles a hot unit of 0.74×0.74 m2. Bronze particles were used to represent the inert bed material. Table 1 lists the scaling parameters. The bed was operated with two different bed heights, 0.18 and 0.305 m, and with superficial gas velocities ranging from 0.013 to 0.53 m/s (all values given on hot up-scaled basis). The investigated bed heights correspond to non-slugging beds. Thus, the height to width aspect ratio of the 4
ACCEPTED MANUSCRIPT dense bed, H0/D, is less than 1, which should not yield any significant wall effects influencing the bubble dynamics.
IP
T
Table 1: Overview of the fluid-dynamical scaling
SC R
All measurements use a single spherical tracer particle resembling an outer diameter of 44 mm in upscaled hot conditions. Figure 2 shows the design of the tracer particle consisting of a spherical NdFeBbased magnet embedded in a spherical plastic capsule with outer diameter of 10 mm, which can be
NU
filled with additional material such as bronze powder so as to adjust the density of the tracer to mimic a fuel particle density. Depending on the fuel density, magnet size Dmagnet was adjusted. The saturation
MA
magnetization, MS, was chosen to be as high as possible and was slightly different for the different magnets used (Table 2). Thus, in this way tracers with four different densities were designed,
D
corresponding to those of fresh and devolatilized biomass pellets used in the Chalmers 2-4 MWth
TE
gasifier [27], as well as reference particles with densities corresponding to the gas-solids emulsion at minimum fluidization as outlined in Table 2. Furthermore, the aim was also to apply a tracer of the
CE P
same density as the bulk solids, which is 815 kg/m3 at 0.27 m/s (up-scaled). But it was decided that this could be represented by the biomass tracer with a density of 800 kg/m3. A fourth tracer resembling
AC
the density of the bed solids was used in all measurements. As this tracer remained in the bottom of the bed (entirely jetsam) for almost all measurements, these results are not shown. Table 2: Tracer particle density
Figure 1: Measurement setup
Figure 2: Build up tracer
The gas distributor is a perforated plate made of a non-magnetic material in order not to disturb the MPT measurements. The influence of air-distributor pressure drop on fuel segregation was investigated by employing two air-distributor plates with different pressure drop characteristics given in Equations (1) and (2), where HP and LP stand for high and low pressure drop, respectively.
5
ACCEPTED MANUSCRIPT Δ𝑃𝐷𝑖𝑠𝐻𝑃 = 131.23 ∙ 𝑢02 [𝑘𝑃𝑎]
(1)
Δ𝑃𝐷𝑖𝑠𝐿𝑃 = 51.76 ∙ 𝑢02 [𝑘𝑃𝑎]
(2)
The LP plate (with 7×7 orifices of diameter 1.4 mm evenly distributed) providing a relatively low
IP
T
pressure drop resembling that in industrial units, while the HP plate (with 5×5 orifices of diameter
SC R
1.5 mm) represents gas distributors typically used in lab tests, i.e. having a higher pressure drop. In the present work, the MPT system applies four sensors assemblies mounted on each of the side wall of the bed at a height of 45 mm for tracking the tracer particle. Each sensor element contains a three-
NU
axis Anisotropic Magneto Resistive (AMR) sensor, which is powered by an external voltage source to be able to measure variation in the electrical resistance of the sensor. The magnetic field range of the
MA
sensors is ±0.6mT and each sensor module is fitted with a 4-element Wheatstone bridge and a first order analogue low pass filter. In order to compensate for built-in cross-axis effects, the sensors are
D
calibrated prior to mounting. Noise from possible magnetic fields surrounding the measurement device
TE
can be subtracted in the data analysis by taking a background measurement before each experiment. As each sampling point is measured by 4×3 sensors the system of equations with five variables is
CE P
overdetermined and position and orientation has to be solved by minimizing the squared difference between the modelled and measured magnetic field (see [23] for details).
AC
3. Results and discussion
It should be noted that all results are presented in values on an up-scaled basis. Figure 3 shows the probability density function (PDF) of the axial location of the tracer as obtained from the MPT measurements with the LP air distributor, with three tracer densities revealing the effect of fluidization velocity and bed height (given as the corresponding fixed bed height). For all tracer densities an increase in fluidization velocity results in a broader and flatter axial distribution of the tracer, i.e. a better axial mixing. This result is in line with previous findings, which point to an increase in bubble size resulting from an increase in gas flow is the reason behind the enhanced mixing within the bed [6] [2] [9] [14]. In addition, an increase in bed height has a similar effect, i.e. reducing the peaks and yielding a broader distribution of the tracer axial distribution. This is explained by the fact
6
ACCEPTED MANUSCRIPT that bubbles grow while rising through the dense bed, resulting in larger bubbles for taller beds, yielding an increase in bubble velocity which results in increased solids mixing [28]. With respect to the different tracer densities (Biochar, Biomass, Emulsion) it is seen that the lighter the tracer the
IP
T
higher the segregation with increased residence time at the bed surface. For a low bed height and low gas velocity (Figure 3a) the pellet char tracer shows a strong flotsam behaviour, while an increase in
SC R
gas velocity (Figure 3b) or in bed height (Figure 3c) improves mixing, and a combined increase of both (Figure 3d) has the strongest effect in enhancing axial mixing. In the latter case, the mixing is
NU
similar for all tracer densities.
As mentioned above the inert tracer with density similar to the bulk particles was also investigated
MA
under the conditions given in Figure 3. For the high bed height the inert tracer sunk directly to the bottom where after it remained at the gas distributor. This behaviour showed repeatability, indicating a
D
pure jetsam behaviour of the inert tracer. However, at low bed heights and high fluidization velocity
b.
d.
AC
c.
CE P
a.
TE
the tracer occasionally circulated within the bed, although this behaviour did not show reproducibility.
Figure 3: PDF of axial tracer location, LP plate, bin size: 3 mm, up-scaled values.
a) H0 =0.18 m, u0 = 0.24 m/s. b) H0 =0.18 m, u0 = 0.48 m/s. c) H0 =0.305 m, u0 = 0.24 m/s. d) H0 =0.305 m, u0 = 0.48 m/s.
Figure 4 gives the axial PDF of the Biochar and Emulsion tracer for the two different air distributors (HP and LP) for a fluidization velocity of 0.24 m/s and a bed corresponding to a (fixed) bed height of 0.305 m. Both air distributors show similar trends for the two tracer densities, although the HP yields a stronger flotsam behaviour for the lighter Biochar tracer. This should be due to that the HP gives smaller bubbles [29] resulting in less splashing [3], which in turn results in less strong axial mixing within the bed. Smaller bubbles yield a lower bubble velocity resulting in poorer mixing, of which mainly the light tracer is affected like it was shown before.
7
ACCEPTED MANUSCRIPT Figure 4: PDF of tracer axial distribution, comparing HP and LP plate, bin size: 15 mm, H0 =0.305 m, u0 = 0.24 m/s, upscaled values.
Figure 5 shows the dense bed height Hb as applied in Figure 6 and Figure 7. For velocities for which
T
the tracer was clearly floating on the bed surface the dense bed height was estimated from the axial
IP
location of the peak in the PDF (max PDF) of the axial tracer position (cf. Figure 3). Beyond a certain
SC R
fluidization velocity the tracer mixes into the dense bed, and then the max PDF estimation cannot be assumed to provide a robust estimation of Hb. Above this velocity, Hb is instead assumed constant and equal to the maximum value obtained from the max PDF location, as shown by the horizontal dashed
NU
lines in Figure 5. This, since it is assumed that the dense bed height remains constant with increasing velocity once it has reached its maximum according to the max PDF criteria. Considering the
MA
moderate velocities employed in this work, this is assumed to be a reasonable assumption. (Note: Applying the dense bed height defined by Johnsson et al. [30] – the height where the time averaged
D
pressure drop deviates from a straight line - the dense bed height should rather decrease than increase
CE P
TE
with increasing fluidization velocity for group B solids.)
Figure 5: Axial location of the maxima in the PDFs of vertical position of Emulsion tracer compared to dense bed height H b used for increasing fluidization velocity (up-scaled values).
Figure 6 shows the fraction of time spent by the tracer on or above the dense bed surface, Ffb, as a
AC
function of excess gas velocity for a low (H0 = 0.18 m in Figure 6a) and a high (H0 = 0.305 m in Figure 6b) fixed bed height, applying the LP distributor. The error bars indicate the effect from varying Hb ±1 cm, which results in ±5% variation in the values of the time spent on and above the bed surface. As seen in Figure 6, increasing the excess velocity reduces the difference in time spent above the dense bed between the three tracers. Thus, the two lighter bio-tracers, which are more or less fully flotsam at low gas velocities, are increasingly mixed into the bed with an increase in gas velocity. This is in contrary to the time spent above the dense bed by the tracer with the same density as the emulsion which, as expected, is not greatly influenced by an increase in velocity. For this tracer, the time spent in the freeboard becomes relatively constant above a certain velocity. The reason for the somewhat
8
ACCEPTED MANUSCRIPT lower values of residence time above the dense bed at low velocity is not obvious, but should be related to larger size of the tracer rather than the slight difference in density due to that scaling did not provide a perfect density match (cf. Table 2). Figure 6 underlines the findings given in Figure 3 that an
IP
T
increase in excess velocity enhances axial mixing, especially for tracers with low density. The higher the tracer density the sooner the time spent in freeboard settles on a constant value with an increase in
SC R
velocity. As can be seen from comparing Figure 6a and b, an increase in bed height enhances the effect of increased gas velocity on axial segregation. This should be due to that the bubbles can grow
NU
bigger with an increase in velocity, yielding higher bubble velocities and, thus, a more intense solids mixing. This is also in line with Figure 3d, where all tracers show similar distribution up through the
MA
bed, when applying a high fluidization velocity.
In summary, with an increase in fluidization velocity three regimes can be identified: 1) A purely
D
flotsam regime which occurs at low fluidization velocities and for the low density tracer particles, 2) A
TE
transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of
CE P
tracer particle at the bed surface and the splash zone remains constant with fluidization velocity.
b.
AC
a.
Figure 6: The average time spent by the tracer on and above the bed surface, LP plate, up-scaled values. a) H0 = 0.18 m b) H0 = 0.305 m
Rowe et al. [6] quantified axial solids segregation for jetsam particles in an FB by the mixing factor M, resulting in values between 0 and 1. The present work applies the mixing factor generically, defining M as:
𝑀=
maximum of tracer concentration in the bed tracer concentration in a well mixed bed
Thus yielding values larger than 1 for flotsam tracers.
9
(3)
ACCEPTED MANUSCRIPT Figure 7 shows the mixing factor of three different tracers for increasing u0-umf. Measurements were conducted with the LP distributor plate and conditions corresponding to a fixed bed height of 0.305 m. A perfectly mixed bed, i.e. M = 1 is indicated with a solid line. In line with what is given above, both
IP
T
Biomass and Biochar tracers behave clearly flotsam for low excess velocities. With increasing velocity the mixing factors approach 1 for all three tracers. For low excess velocity for which the average bed
SC R
density is higher, also the Emulsion tracer yields a somewhat flotsam behaviour, but with an increase in gas velocity, soon becomes well-mixed (M approximately 1).
NU
The surface fraction Ffb (Figure 6) defines the share of tracer observations on and above the dense bed surface, i.e. describes the share of fuel particles that will experience increased mass transfer and
MA
decreased heat transfer in comparison to immersed solids [1]. The mixing index M describes the quality of the axial mixing in the dense bed, indicating the maximum clustering of tracers. A low value
D
of M indicates a high local concentration of active particle, i.e. fuel particles, corresponding to a low
TE
effective mass transfer with the bed [31]. Note that Ffb and M give different information and thus can show opposite trends. This can be seen when comparing Figure 6b and Figure 7 for the Emulsion
CE P
tracer at low velocities, i.e. a stronger axial segregation of the tracer (high M) does not necessarily
AC
means that the tracer ends up as flotsam particle on the dense bed surface (low Ffb).
Figure 7: Mixing factor, LP plate, H0 =0.305 m, up-scaled values.
Figure 8 display PDFs of the axial upward and downward velocities of the tracer for measurements corresponding to a fixed bed height of 0.305 m and superficial gas velocity of 0.3 m/s, with Figure 8a showing the Biochar tracer and Figure 8b showing the Emulsion tracer. Median values, m, of each distribution are shown in the legend. The tracer velocity is calculated through the application of Kalman filtering, as explained elsewhere [23], i.e. the frequency employed for the calculation is the sampling rate of the measurement (20 Hz). Figure 8 gives that at low absolute velocities (< 0.25 m/s) both tracer types give a more narrow and centred PDF distribution in the downward velocities than in the upwards velocities. This is in line with previous findings, which show that the upwards motion of the tracer is governed by fast rising bubble paths [3, 32], whereas the downwards motion is following 10
ACCEPTED MANUSCRIPT the slower sinking bed material on the sides of the bubble paths. As indicated by its narrower distribution in Figure 8b, the downwards velocity has weaker fluctuations, which can be explained with the smoother motion of the sinking bed material in comparison to the strongly fluctuating
IP
T
upwards motion of the bubble flow, as previously reported in [12].
SC R
When comparing Figure 8a and b with each other the effect of tracer density can be observed. The Emulsion tracer with its higher density yields higher median velocities, i.e. exhibits higher velocities for both upwards and downwards motion, i.e. due to that its density is similar to the emulsion phase of
NU
the bulk material it mixes better into the dense bed rather than floating on the dense bed surface. This is in line with the results given earlier in this paper, where the Emulsion tracer segregates less on the
MA
bed surface (Figure 6) and shows a better axial mixing inside the bed (Figure 7), including in the cases
D
with lower fluidization velocities.
TE
a.
b.
CE P
Figure 8: PDF of upwards and downwards uZ in the dense bed for H0 =0.305 m, u0 = 0.3 m/s, bin size: 0.02 m/s. a) Biochar tracer b) Emulsion tracer
Figure 9 give the vertical profiles of average upward and downward tracer velocity, urise and usink, for a
AC
low (0.19 m/s) and a high (0.43 m/s) excess velocity (up-scaled values) for the Biochar (Figure 9a) and the Emulsion (Figure 9b) tracer. The bed is divided into slices of around 0.04 m in the axial direction. The bed height (0.305 m) is indicated as a solid line. It should be noted that some points are associated with a considerable scatter (up to 90%). This is due to the sampling frequency may be too low for velocity measurements as discussed in [25]. However, a clear trend is observed for all the curves: It can be seen that for both tracer densities the rising velocities are higher than the sinking velocities and that the influence from an increase in fluidization velocity is greater for the Biochar tracer (Figure 9a) than the Emulsion tracer (Figure 9b). The lowest average velocities are found in the bottom near the air distributor and around the height of the dense bed (the static bed height is indicated). The low axial velocities near the air distributor reflects that the tracer spent a certain time near the gas distributor, yielding low values of the average axial solids velocity. Similar, there is a
11
ACCEPTED MANUSCRIPT certain flotsam behaviour of both tracers, which results in an increase in residence time at the bed surface and, thereby, a reduction in the average axial velocities. This is especially the case for the Biochar tracer which spends significant time on the bed surface (Figure 9a). Once entrained into the
IP
T
freeboard, the velocities are again higher, representing the rising velocities caused by bubble eruption (upward) and the back-mixing into the dense bed (downward). As for the freeboard mixing it can be
SC R
assumed that the tracer solid is to a more or less extent immersed in the emulsion phase, i.e. follows
NU
the ballistic movement of clustered solids mentioned above.
MA
a.
b.
D
Figure 9: Vertical profile of mean urise and usink at two different u0-umf. H0 =0.305 m, up-scaled values.
TE
a) Biochar. b) Emulsion.
Figure 10 compares average axial tracer velocities obtained in this work (on an up-scaled basis) with
CE P
corresponding literature data for different excess gas velocities. The literature data are with conditions similar to those applied in the present work on an up-scaled basis, i.e. H0 = 0.15-0.5 m, ρs = 1 1002 940 kg/m3, dp = 212-800 µm. Figure 10a gives the upward velocities and Figure 10b the downward
AC
velocities, in both cases averaged over the entire dense bed (i.e. up to H0). As for downward velocities of immersed objects, not all of the above mentioned works provide such data. All literature applied tracers with a density similar to that of the average bulk solids, i.e. similar the Emulsion tracer in this work. From Figure 10 it can be seen that in the present work (×), the increase in upward and downward velocities with an increase in fluidization velocity levels out at the higher range of velocities investigated. The upward tracer velocity is around 50% of the theoretical bubble velocity calculated by the Davidson and Harrison’s model [28]. Corresponding values between 10% and 50% are mentioned in the literature [10] [12] [13]. It should be mentioned that the bubble velocity is a function of bed material properties, bed height and gas inlet velocity and therefore differs between the data given in Figure 10. Nienow et al. [9] and Rees et al. [33] used the rather simple measurement method of releasing the tracer close to the distributor in 3D beds and observing the time it takes for the 12
ACCEPTED MANUSCRIPT tracer to reach the bed surface. Soria-Verdugo et al. [12] and Lim et al. [10] measured with optical methods in 2D units. Both Fotovat et al. [14] and Stein et al. [13] used radioactive particle tracking methods. All works which applied particle tracking methods presented tracer particle velocities
IP
T
averaged over the dense bed, i.e. as done in this work. From Figure 10 it can be seen that literature values are generally lower than the values obtained in this study. It is believed that the higher tracer
SC R
rising velocity obtained for the present experiments is connected to the “higher temperature level”, resembled by the fluid-dynamic scaling. All data points from literature are measured in units operating
NU
at cold conditions, not applying fluid-dynamically downscaling. This hypothesis is supported by the available values of lateral solids dispersion coefficients under conditions relevant for industrial units:
MA
Olsson et al. [3] provide an experimentally determined lateral fuel dispersion coefficient in an industrial-scale unit operated under cold conditions, while Sette et al. [34] used a model operated according to scaling laws to resemble the unit used by Olsson et al. under hot conditions. From his set-
TE
D
up Sette et al. obtained lateral dispersion coefficients more than two times higher than under cold conditions. Thus, the smaller (and thereby slower) bubbles generated under cold conditions by Olsson
CE P
et al. induce less intense solids mixing than from larger and faster bubbles under the hot conditions by Sette et al. resembled through the application of fluid-dynamical scaling.
AC
With respect to the downward velocities, Figure 10b shows that data obtained in 2D beds [10] [12] at similar operational conditions as the up-scaled values in the present work but without applying fluid dynamic scaling, give considerable lower values than the ones obtained in the present work. As for the 3D data by Fotovat et al. [14] these are closer to those of the present work, although somewhat lower. As discussed above for the rising velocity, the effect of the temperature level can be argued to cause the difference between the non-scaled experiments and application of scaling laws, revealing the importance of using fluid-dynamical scaling in order to obtain experimental data of quantitative relevance for commercial units. Further, the effect of pressure drop over the gas distributor as well as the effect of reduced fixed bed height on rising and sinking velocities were evaluated (not shown in Figure 10). With increasing pressure drop (see Equations (1) and (2)) both rising and sinking velocities are reduced by around 13
ACCEPTED MANUSCRIPT 20%. This is in line with the findings made in connection to Figure 4, i.e. the HP distributor yielding a stronger flotsam behaviour, which are explained by the reduced bubble size resulting in a decreased bubble velocity. When reducing the fixed bed height H0 from 0.305 m to 0.18 cm the rising velocities
IP
T
are reduced by around 30%, while sinking velocities are reduced by around 40%. Since the bubbles grow up through the bed, shallower beds give smaller bubble size which result in lower bubble
SC R
velocities and, thus, less vigorous solids mixing, reducing the rising and sinking velocities.
NU
a.
MA
b.
Figure 10: Tracer average rising, urise, (a), and sinking, usink, (b), velocity in the dense bed (up to h=H0) as function of the excess gas velocity, u0-umf. Emulsion tracer. H0 =0.305 m, up-scaled values.
D
Compared data from literature: H0 = 15-50 cm, ρs = 1100-2940 kg/m3.
TE
4. Conclusions
CE P
Magnetic particle tracking (MPT) using single tracer particles is used in a fluid-dynamically downscaled bubbling fluidized bed to investigate how axial mixing depends on fluidization velocity, bed height and fuel particle density.
AC
The MPT results show that increase in bed height, fluidization velocity and in fuel tracer density enhance the axial fuel mixing. The axial mixing is, however, similar for the two different air distributors used, although the air distributor with the higher pressure drop yields a stronger flotsam behaviour. This should be due to that the HP gives smaller bubbles, resulting in less vigorous splashing which in turn gives less downward flow of solids dragging down the tracer in the bed. With an increase in fluidization velocity three axial mixing regimes of the fuel tracer can be identified: 1) A flotsam regime occurring at low fluidization velocities and for light tracer particles which, combined show a purely flotsam regime, 2) A transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of tracer particle at the bed surface and the splash zone
14
ACCEPTED MANUSCRIPT remains constant with fluidization velocity. The onset velocities between the regimes depend on bed height and tracer properties.
T
As expected, the upward and downward velocities of the fuel tracers depend on bubble properties – the
IP
larger the bubbles, the higher the bubble velocity and the more vigorous the axial solids mixing.
SC R
Comparing these velocities to literature data on an up-scaled basis, the velocities obtained in this work is considerably higher than corresponding values from literature. This is believed to be due to that the present data represent full-scale conditions, as compared to the literature data which were obtained
NU
under cold conditions not applying scaling laws. Yet, there is a need to also measure in real combustion or gasification conditions in order to derive to what extent the present data quantitatively
MA
correspond to axial mixing at industrial conditions.
5 Acknowledgement
D
This work is co-financed by the Swedish Energy Agency – project P38347-1 – and Valmet
TE
Corporation Oy, within the framework of the project “Experimental investigation of the gas-solids
References
CE P
flow in FB units”.
AC
[1] L. Lundberg, P.A. Tchoffor, D. Pallarès, R. Johansson, H. Thunman, K. Davidsson, Influence of surrounding conditions and fuel size on the gasification rate of biomass char in a fluidized bed, Fuel Processing Technology, 144 (2016) 323-333. [2] P.N. Rowe, A.W. Nienow, Particle mixing and segregation in gas fluidised beds. A review, Powder Technology, 15 (1976) 141-147. [3] J. Olsson, D. Pallarès, F. Johnsson, Lateral fuel dispersion in a large-scale bubbling fluidized bed, Chemical Engineering Science, 74 (2012) 148-159. [4] R. Bilbao, J. Lezaun, M. Menendez, J. Abanades, Model of mixing—segregation for straw/sand mixtures in fluidized beds, Powder Technology, 56 (1988) 149-155. [5] R. Bilbao, J. Lezaun, M. Menendez, M. Izquierdo, Segregation of straw/sand mixtures in fluidized beds in non-steady state, Powder Technology, 68 (1991) 31-35. [6] P.N. Rowe, A.W. Nienow, A.J. Agbim, The mechanisms by which particles segregate in gas fluidised beds - binary systems of near-spherical particles Transactions of the Institution of Chemical Engineers 50 (1972) 310-323. [7] A. Marzocchella, P. Salatino, V. Di Pastena, L. Lirer, Transient fluidization and segregation of binary mixtures of particles, AIChE Journal, 46 (2000) 2175-2182. [8] G. Olivieri, A. Marzocchella, P. Salatino, Segregation of fluidized binary mixtures of granular solids, AIChE Journal, 50 (2004) 3095-3106. [9] A.W. Nienow, P.N. Rowe, T. Chiwa, Mixing and segregation of a small proportion of large particles in gas fluidized beds of considerably, AIChE Symposium Series, 74 (1978) 45-53. [10] K.S. Lim, P.K. Agarwal, Circulatory motion of a large and lighter sphere in a bubbling fluidized bed of smaller and heavier particles, Chemical Engineering Science, 49 (1994) 421-424. 15
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
NU
SC R
IP
T
[11] A. Soria-Verdugo, L.M. Garcia-Gutierrez, N. García-Hernando, U. Ruiz-Rivas, Buoyancy effects on objects moving in a bubbling fluidized bed, Chemical Engineering Science, 66 (2011) 2833-2841. [12] A. Soria-Verdugo, L.M. Garcia-Gutierrez, S. Sanchez-Delgado, U. Ruiz-Rivas, Circulation of an object immersed in a bubbling fluidized bed, Chemical Engineering Science, 66 (2011) 78-87. [13] M. Stein, Y.L. Ding, J.P.K. Seville, D.J. Parker, Solids motion in bubbling gas fluidised beds, Chemical Engineering Science, 55 (2000) 5291-5300. [14] F. Fotovat, R. Ansart, M. Hemati, O. Simonin, J. Chaouki, Sand-assisted fluidization of large cylindrical and spherical biomass particles: Experiments and simulation, Chemical Engineering Science, 126 (2015) 543-559. [15] G.M. Rios, K. Dang Tran, H. Masson, FREE OBJECT MOTION IN A GAS FLUIDIZED BED, Chemical Engineering Communications, 47 (1986) 247-272. [16] C.E. Weinell, K. Dam-Johansen, J.E. Johnsson, Single-particle behaviour in circulating fluidized beds, Powder Technology, 92 (1997) 241-252. [17] D. Moslemian, N. Devanathan, M.P. Dudukovic, Radioactive particle tracking technique for investigation of phase recirculation and turbulence in multiphase systems, Review of Scientific Instruments, 63 (1992) 4361-4372. [18] G. Mohs, O. Gryczka, S. Heinrich, L. Mörl, Magnetic monitoring of a single particle in a prismatic spouted bed, Chemical Engineering Science, 64 (2009) 4811-4825. [19] J. Halow, K. Holsopple, B. Crawshaw, S. Daw, C. Finney, Observed Mixing Behavior of Single Particles in a Bubbling Fluidized Bed of Higher-Density Particles, Industrial & Engineering Chemistry Research, 51 (2012) 14566-14576. [20] K.A. Buist, A.C. van der Gaag, N.G. Deen, J.A.M. Kuipers, Improved magnetic particle tracking technique in dense gas fluidized beds, AIChE Journal, 60 (2014) 3133-3142. [21] K.A. Buist, T.W. van Erdewijk, N.G. Deen, J.A.M. Kuipers, Determination and comparison of rotational velocity in a pseudo 2-D fluidized bed using magnetic particle tracking and discrete particle modeling, AIChE Journal, 61 (2015) 3198-3207. [22] E. Sette, T. Berdugo Vilches, D. Pallarès, F. Johnsson, Measuring fuel mixing under industrial fluidized-bed conditions – A camera-probe based fuel tracking system, Applied Energy, 163 (2016) 304-312. [23] E. Sette, D. Pallarès, F. Johnsson, F. Ahrentorp, A. Ericsson, C. Johansson, Magnetic tracerparticle tracking in a fluid dynamically down-scaled bubbling fluidized bed, Fuel Processing Technology, 138 (2015) 368-377. [24] D. Simon, The discrete-time Kalman filter, Optimal State Estimation, John Wiley & Sons, Inc.2006, pp. 121-148. [25] A. Köhler, D. Pallarès, F. Johnsson, Magnetic tracking of a fuel particle in a fluid-dynamically down-scaled fluidized bed Submitted for publication, (2016 ). [26] L.R. Glicksman, M.R. Hyre, P.A. Farrell, Dynamic similarity in fluidization, International Journal of Multiphase Flow, 20 (1994) 331-386. [27] H. Thunman, M.C. Seemann, First experiences with the new chalmers gasifier, 2009, pp. 659663. [28] J.F. Davidson, H. D., Fluidised particles Cambridge University Press New York 1963. [29] A. Svensson, F. Johnsson, B. Leckner, Bottom bed regimes in a circulating fluidized bed boiler, International Journal of Multiphase Flow, 22 (1996) 1187-1204. [30] F. Johnsson, S. Andersson, B. Leckner, Expansion of a freely bubbling fluidized bed, Powder Technology, 68 (1991) 117-123. [31] F. Scala, Particle-fluid mass transfer in multiparticle systems at low Reynolds numbers, Chemical Engineering Science, 91 (2013) 90-101. [32] D. Pallarès, P.A. Díez, F. Johnsson, Experimental analysis of fuel mixing patterns in a fluidized bed, The 12th International Conference on Fluidization, ECI, Vancouver, Canada, 2007, pp. 929-936. [33] A.C. Rees, J.F. Davidson, J.S. Dennis, A.N. Hayhurst, The rise of a buoyant sphere in a gasfluidized bed, Chemical Engineering Science, 60 (2005) 1143-1153. [34] E. Sette, D. Pallarès, F. Johnsson, Influence of bulk solids cross-flow on lateral mixing of fuel in dual fluidized beds, Fuel Processing Technology, 140 (2015) 245-251.
16
ACCEPTED MANUSCRIPT Vitae Anna Köhler
Anna Köhler is a doctoral student at Energy and Environment in the division of
T
Energy Technology at Chalmers University of Technology in Gothenburg. Her
IP
research focuses on the fluid dynamics in fluidized beds, which she addresses
SC R
with both experimental work and modelling. Her work aims at an improved use of non-fossil fuels such as biomass in fluidized bed units.
Alexander Rasch is a Bachelor student in Mechanical Engineering, working
Rasch
with energy engineering related issues. Completing his thesis at Chalmers
NU
Alexander
MA
University of Technology in Gothenburg, he conducted experiments and analysis
David Pallarès is an Associate Professor at Chalmers University of Technology.
TE
David Pallarès
D
related to axial fuel mixing.
His research focuses on experiments and modeling of fluidized bed phenomena
CE P
of relevance for different fuel conversion processes (e.g. air-combustion, oxyfuel-combustion, chemical looping combustion and indirect gasification)
AC
with the aim to assess the design, scale-up and performance of such processes.
Filip Johnsson
Filip Johnsson is Professor in Sustainable Energy Systems in Department of Energy and Environment, Chalmers University of Technology. Filip Johnson’s research area comprises Fluidized bed combustion and gasification, oxyfuelcombustion for CO2 capture energy systems analysis. The fluidized-bed related research focus on fluid dynamics and mixing.
17
Fig. 1
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
18
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 2
19
Fig. 3a
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
20
Fig. 3b
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
21
Fig. 3c
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
22
Fig. 3d
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
23
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
Fig. 4
24
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
Fig. 5
25
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 6a
26
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 6b
27
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 7
28
Fig. 8a
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
29
Fig. 8b
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
30
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 9a
31
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 9b
32
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 10a
33
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 10b
34
Anna Khler
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
35
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
Alexander Rasch
36
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
NU
David Pallares
37
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
Filip Johnsson
38
ACCEPTED MANUSCRIPT Table 1: Parameter
Unit
Hot model
Cold model ideal
Temperature
°C
800
Gas density
kg/m3
0.32
Bed dimensions
m×m
L = 0.74 × 0.74
Bed height
m
𝐻0 = 0.18; 0.305
Superficial velocity
m/s
𝑢0 = 0.01 – 0.5
Bed material density
kg/m3
2 600
Bed material size
µm
250
Cold model actual
T
20
IP
1.2
𝐿
AC
CE P
TE
D
MA
NU
SC R
4.4
39
𝑢0 ∙
= 0.17 × 0.17
𝐻0
4.4
= 0.04; 0.07
1 √4.4
= 0.006 – 0.236
9 530
8 900
57
60
ACCEPTED MANUSCRIPT Table 2:
ρhot
ρdownscaled
ρactual
(kg/m3)
(kg/m3)
Biochar
350
Biomass Emulsion
Error
Dmagnet
MS
(kg/m3)
(%)
(m)
(Gauss)
1 280
1 470
-14.6
800
2 930
2 980
-1.7
1 230
4 500
4 320
downscaled
0.05
13050
0.08
12550
0.08
12550
IP
SC R 4.0
NU MA D TE CE P AC
40
T
Tracer density
ACCEPTED MANUSCRIPT
NU
SC R
IP
T
Graphical abstract
MA
Magnetic Particle Tracking is used in a fluid-dynamically down-scaled bubbling fluidized bed to study axial mixing and segregation of fuel particles. The system provides 3D trajectories of the fuel tracer
AC
CE P
TE
D
particle, highly resolved in time and space.
41
ACCEPTED MANUSCRIPT Highlights
CE P
TE
D
MA
NU
SC R
IP
T
Magnetic particle tracking provides resolved 3D trajectories of tracer particle The impact of fuel density, bed height and fluidization velocity is shown Fluid-dynamical downscaling yields higher mixing than non-scaled cold units
AC
42
S0032-5910(16)30973-1 doi:10.1016/j.powtec.2016.12.093 PTEC 12241
To appear in:
Powder Technology
Received date: Revised date: Accepted date:
14 May 2016 28 October 2016 30 December 2016
Please cite this article as: Anna K¨ohler, Alexander Rasch, David Pallar`es, Filip Johnsson, Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking, Powder Technology (2017), doi:10.1016/j.powtec.2016.12.093
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Experimental characterization of axial fuel mixing in fluidized beds by magnetic particle tracking Anna Köhler*, Alexander Rasch, David Pallarès and Filip Johnsson
IP
T
Dept. of Energy and Environment, Chalmers University of Technology, SE-412 96 Göteborg, Sweden
SC R
Tel: +46-(0)31-772-14 42
NU
Email: [email protected]
Abstract
MA
A Magnetic Particle Tracking (MPT) system is applied to a bubbling fluidized bed to study how axial mixing and segregation of fuel are influenced by the fuel density and operational conditions
D
(fluidization velocity, bed height and pressure drop across the gas distributor). The MPT system is
TE
used to determine the vertical distribution of the tracer particle in a fluid-dynamically down-scaled cold unit resembling a 0.74×0.74 m2 fluidized bed reactor operating at 800 °C. This work uses a tracer
CE P
particle of 10 mm in diameter, corresponding to a fuel particle of 44 mm. Different tracer particles are applied with solids density representing biomass, biomass char and that of the average bulk. The MPT
AC
system yields a spatial accuracy in the order of 10-3 m and a time resolution of 10-3 s. For the operational range investigated, three fuel segregation regimes can be identified from the MPT measurements: 1) A flotsam regime which occurs at low fluidization velocities and for low density tracer particles, 2) A transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of tracer particle at the bed surface and the splash zone remains constant with fluidization velocity. The transition velocities between the regimes depend on bed height and density of the tracer particle. Keywords: Fuel segregation; Fuel mixing; Solids mixing; Magnetic particle tracking; Fluiddynamical scaling; Fluidized bed
1
ACCEPTED MANUSCRIPT 1. Introduction Although previous research related to coal combustion in fluidized-bed (FB) boilers have given
T
valuable information on the mixing of gas and solids in fluidized beds, the application of the FB
IP
technology to biomass conversion requires more in-depth understanding of mixing, especially with
SC R
respect to mixing of the fuel. This, since biomass is a high volatile fuel where moisture and volatiles may be released before the fuel has been distributed evenly over the bed cross section and within the bed causing oxygen depleted zones and uneven combustion. In indirect biomass gasification, control
NU
of the fuel mixing process is of importance in order to optimize the gasification process while letting the remaining char be transferred to the combustion bed. Biomass particles are more prone to
MA
experience axial segregation from the bed material as they have a lower density and are larger in size than coal particles. An increase in axial segregation results in more fuel particles floating on the dense
D
bed surface, which implies enhanced bed-to-fuel mass transfer (due to the absence of particle-related
TE
tortuosity) and reduced bed-to-fuel heat transfer (due to absence of convection from bulk particles). The decreased bed-to-fuel heat transfer has been experimentally shown to yield less reactive char
CE P
particles [1]. Furthermore, fuel present in the splash zone will experience faster lateral mixing than if immersed in the bed, which is often beneficial in boilers but strongly undesired in other applications
AC
(indirect gasification, chemical looping combustion). Different types of particles are present in fluidized bed reactors used for combustion and gasification, but can be divided into finer and denser bulk bed material (e.g. fuel ashes, makeup sand, sorbent, catalytic material), and lighter and typically larger fuel particles. The latter are changing in both density and size during fuel conversion. In addition, there is a large variety of biomass fuels, with different shape, composition and density. Mixing and segregation is caused by the bubble activity in the bed and thereby highly influenced by operational conditions, i.e. by fluidization velocity, bed height, pressure drop across the gas distributor and by the physical properties of the solids [2] [3]. As biomass is a low-density fuel it tends to show a flotsam behaviour, concentrating at the top of the bed [4] [5]. Increasing fluidization velocity may decrease fuel segregation and allow sufficient axial mixing. There are several important experimental studies on solids mixing in literature. Rowe et al. [6]
2
ACCEPTED MANUSCRIPT studied binary particle mixtures and showed that particles of higher density and size tend to show a jetsam behaviour, i.e. to end up in the bottom of the bed. This work was followed by more recent studies [7] [8] discussing the presence of different fluidization regimes for beds of binary mixtures of
IP
T
solids. Nienow et al. [9] studied the segregation of a single large and light object immersed in a fluidized bed, and they found segregation to be mainly caused by buoyancy forces. Nienow et al.
SC R
created rising velocities of the object immersed in the bed by releasing it close to the distributor and noting the time it took for the object to reach the bed surface. Optical measurement methods have also
NU
been used to study solids mixing of large particles immersed in pseudo-2D beds [10-12]. At present, Radioactive (or positron emission) Particle Tracking (RPT) can be applied to follow tracers with high
MA
accuracy in all 3 dimensions, which has been used under cold conditions in laboratory units [13-15] and in what was referred to as a cold pilot-scale unit [16]. Moslemian et al. [17] used RPT in a labscale CFB. Mohs et al. [18] evaluated several particle tracking methods for use in laboratory scale (i.e.
TE
D
ambient conditions), and they choose Magnetic Particle Tracking (MPT) before RPT. The main benefits of MPT are that both position and orientation of the tracer particle can be measured in all 3
CE P
dimensions, while it is relatively cheap and does not require any extensive safety precautions [19-21]. Combined with fluid-dynamical scaling, MPT allows lab-scale measurements covering conditions, which are difficult to measure in commercial units. Thus, when investigating fuel mixing a wider
AC
range of density of the fuel tracer particle as well as a wider range of operational condition can be covered compared to an industrial boiler or gasifier but of course with the limitation that effects from the fuel conversion itself is not included. Yet, from experiments in a hot industrial scale biomass gasifier [22], focusing on fuel mixing on the bed surface, it was concluded that fuel buoyancy effects from volatile release seemed to be small compared to the mixing caused by the bubble flow. In a previous work by the authors of this work [23], it was shown that MPT can give continuous data with a spatial accuracy of ~1 mm and temporal resolution of up to 1 kHz. The determination of axial profiles and velocities in and above the bed can be directly derived from the measurement data after being smoothened with Kalman filtering [24].
3
ACCEPTED MANUSCRIPT While previous works in literature give valuable insights in the behaviour of FB solids mixing, they are all based on experiments in small lab-scale units operated at cold conditions, i.e. under conditions which cannot be directly translated to industrial conditions. The aim of the present work is to
IP
T
understand the basics of fuel mixing regimes for different operating conditions, focusing on the axial mixing of biomass fuel particles at different stages of conversion (i.e. different densities). In order to
SC R
resemble industrial operation, the system (including the tracer particle) is scaled to represent the density and size of such fuel but also other tracer densities are applied for comparison. The work
NU
applies the previously developed MPT technology described in [23, 25]. The work given in [23, 25] provides first measurements and a description of the MTP measurement technique, including the
MA
possibilities of implementing MPT in fluid-dynamically down-scaled conditions and a discussion of the limitations of the method. This contribution of the present work is to provide the first highlyresolved fluid-dynamically down-scaled measurements with the aim to increase the understanding of
fluidization velocity).
TE
D
axial solids segregation in fluidized beds (including the impact of tracer density, bed height and
CE P
2. Experimental setup
The experiments in the present work cover three tracer densities ranging within 350-1 230 kg/m3. In
AC
addition, a tracer with a density (2 130 kg/m3) similar to that of the bed solids was investigated but for the conditions studied the tracer sunk to the air distributor immediately after the onset of fluidization, where it then remained, i.e. exhibited a purely jetsam behaviour. The MPT experiments were carried out in a fluidized bed fluid-dynamically down-scaled according to a full set of scaling rules given by Glicksman et al. [26]. The bed, which is schematically shown in Figure 1, is fluidized with pressurized air entering a wind box and then the bed through a perforated plate. With a resulting length scaling factor of 4.4, the cold lab unit (0.17×0.17 m2) resembles a hot unit of 0.74×0.74 m2. Bronze particles were used to represent the inert bed material. Table 1 lists the scaling parameters. The bed was operated with two different bed heights, 0.18 and 0.305 m, and with superficial gas velocities ranging from 0.013 to 0.53 m/s (all values given on hot up-scaled basis). The investigated bed heights correspond to non-slugging beds. Thus, the height to width aspect ratio of the 4
ACCEPTED MANUSCRIPT dense bed, H0/D, is less than 1, which should not yield any significant wall effects influencing the bubble dynamics.
IP
T
Table 1: Overview of the fluid-dynamical scaling
SC R
All measurements use a single spherical tracer particle resembling an outer diameter of 44 mm in upscaled hot conditions. Figure 2 shows the design of the tracer particle consisting of a spherical NdFeBbased magnet embedded in a spherical plastic capsule with outer diameter of 10 mm, which can be
NU
filled with additional material such as bronze powder so as to adjust the density of the tracer to mimic a fuel particle density. Depending on the fuel density, magnet size Dmagnet was adjusted. The saturation
MA
magnetization, MS, was chosen to be as high as possible and was slightly different for the different magnets used (Table 2). Thus, in this way tracers with four different densities were designed,
D
corresponding to those of fresh and devolatilized biomass pellets used in the Chalmers 2-4 MWth
TE
gasifier [27], as well as reference particles with densities corresponding to the gas-solids emulsion at minimum fluidization as outlined in Table 2. Furthermore, the aim was also to apply a tracer of the
CE P
same density as the bulk solids, which is 815 kg/m3 at 0.27 m/s (up-scaled). But it was decided that this could be represented by the biomass tracer with a density of 800 kg/m3. A fourth tracer resembling
AC
the density of the bed solids was used in all measurements. As this tracer remained in the bottom of the bed (entirely jetsam) for almost all measurements, these results are not shown. Table 2: Tracer particle density
Figure 1: Measurement setup
Figure 2: Build up tracer
The gas distributor is a perforated plate made of a non-magnetic material in order not to disturb the MPT measurements. The influence of air-distributor pressure drop on fuel segregation was investigated by employing two air-distributor plates with different pressure drop characteristics given in Equations (1) and (2), where HP and LP stand for high and low pressure drop, respectively.
5
ACCEPTED MANUSCRIPT Δ𝑃𝐷𝑖𝑠𝐻𝑃 = 131.23 ∙ 𝑢02 [𝑘𝑃𝑎]
(1)
Δ𝑃𝐷𝑖𝑠𝐿𝑃 = 51.76 ∙ 𝑢02 [𝑘𝑃𝑎]
(2)
The LP plate (with 7×7 orifices of diameter 1.4 mm evenly distributed) providing a relatively low
IP
T
pressure drop resembling that in industrial units, while the HP plate (with 5×5 orifices of diameter
SC R
1.5 mm) represents gas distributors typically used in lab tests, i.e. having a higher pressure drop. In the present work, the MPT system applies four sensors assemblies mounted on each of the side wall of the bed at a height of 45 mm for tracking the tracer particle. Each sensor element contains a three-
NU
axis Anisotropic Magneto Resistive (AMR) sensor, which is powered by an external voltage source to be able to measure variation in the electrical resistance of the sensor. The magnetic field range of the
MA
sensors is ±0.6mT and each sensor module is fitted with a 4-element Wheatstone bridge and a first order analogue low pass filter. In order to compensate for built-in cross-axis effects, the sensors are
D
calibrated prior to mounting. Noise from possible magnetic fields surrounding the measurement device
TE
can be subtracted in the data analysis by taking a background measurement before each experiment. As each sampling point is measured by 4×3 sensors the system of equations with five variables is
CE P
overdetermined and position and orientation has to be solved by minimizing the squared difference between the modelled and measured magnetic field (see [23] for details).
AC
3. Results and discussion
It should be noted that all results are presented in values on an up-scaled basis. Figure 3 shows the probability density function (PDF) of the axial location of the tracer as obtained from the MPT measurements with the LP air distributor, with three tracer densities revealing the effect of fluidization velocity and bed height (given as the corresponding fixed bed height). For all tracer densities an increase in fluidization velocity results in a broader and flatter axial distribution of the tracer, i.e. a better axial mixing. This result is in line with previous findings, which point to an increase in bubble size resulting from an increase in gas flow is the reason behind the enhanced mixing within the bed [6] [2] [9] [14]. In addition, an increase in bed height has a similar effect, i.e. reducing the peaks and yielding a broader distribution of the tracer axial distribution. This is explained by the fact
6
ACCEPTED MANUSCRIPT that bubbles grow while rising through the dense bed, resulting in larger bubbles for taller beds, yielding an increase in bubble velocity which results in increased solids mixing [28]. With respect to the different tracer densities (Biochar, Biomass, Emulsion) it is seen that the lighter the tracer the
IP
T
higher the segregation with increased residence time at the bed surface. For a low bed height and low gas velocity (Figure 3a) the pellet char tracer shows a strong flotsam behaviour, while an increase in
SC R
gas velocity (Figure 3b) or in bed height (Figure 3c) improves mixing, and a combined increase of both (Figure 3d) has the strongest effect in enhancing axial mixing. In the latter case, the mixing is
NU
similar for all tracer densities.
As mentioned above the inert tracer with density similar to the bulk particles was also investigated
MA
under the conditions given in Figure 3. For the high bed height the inert tracer sunk directly to the bottom where after it remained at the gas distributor. This behaviour showed repeatability, indicating a
D
pure jetsam behaviour of the inert tracer. However, at low bed heights and high fluidization velocity
b.
d.
AC
c.
CE P
a.
TE
the tracer occasionally circulated within the bed, although this behaviour did not show reproducibility.
Figure 3: PDF of axial tracer location, LP plate, bin size: 3 mm, up-scaled values.
a) H0 =0.18 m, u0 = 0.24 m/s. b) H0 =0.18 m, u0 = 0.48 m/s. c) H0 =0.305 m, u0 = 0.24 m/s. d) H0 =0.305 m, u0 = 0.48 m/s.
Figure 4 gives the axial PDF of the Biochar and Emulsion tracer for the two different air distributors (HP and LP) for a fluidization velocity of 0.24 m/s and a bed corresponding to a (fixed) bed height of 0.305 m. Both air distributors show similar trends for the two tracer densities, although the HP yields a stronger flotsam behaviour for the lighter Biochar tracer. This should be due to that the HP gives smaller bubbles [29] resulting in less splashing [3], which in turn results in less strong axial mixing within the bed. Smaller bubbles yield a lower bubble velocity resulting in poorer mixing, of which mainly the light tracer is affected like it was shown before.
7
ACCEPTED MANUSCRIPT Figure 4: PDF of tracer axial distribution, comparing HP and LP plate, bin size: 15 mm, H0 =0.305 m, u0 = 0.24 m/s, upscaled values.
Figure 5 shows the dense bed height Hb as applied in Figure 6 and Figure 7. For velocities for which
T
the tracer was clearly floating on the bed surface the dense bed height was estimated from the axial
IP
location of the peak in the PDF (max PDF) of the axial tracer position (cf. Figure 3). Beyond a certain
SC R
fluidization velocity the tracer mixes into the dense bed, and then the max PDF estimation cannot be assumed to provide a robust estimation of Hb. Above this velocity, Hb is instead assumed constant and equal to the maximum value obtained from the max PDF location, as shown by the horizontal dashed
NU
lines in Figure 5. This, since it is assumed that the dense bed height remains constant with increasing velocity once it has reached its maximum according to the max PDF criteria. Considering the
MA
moderate velocities employed in this work, this is assumed to be a reasonable assumption. (Note: Applying the dense bed height defined by Johnsson et al. [30] – the height where the time averaged
D
pressure drop deviates from a straight line - the dense bed height should rather decrease than increase
CE P
TE
with increasing fluidization velocity for group B solids.)
Figure 5: Axial location of the maxima in the PDFs of vertical position of Emulsion tracer compared to dense bed height H b used for increasing fluidization velocity (up-scaled values).
Figure 6 shows the fraction of time spent by the tracer on or above the dense bed surface, Ffb, as a
AC
function of excess gas velocity for a low (H0 = 0.18 m in Figure 6a) and a high (H0 = 0.305 m in Figure 6b) fixed bed height, applying the LP distributor. The error bars indicate the effect from varying Hb ±1 cm, which results in ±5% variation in the values of the time spent on and above the bed surface. As seen in Figure 6, increasing the excess velocity reduces the difference in time spent above the dense bed between the three tracers. Thus, the two lighter bio-tracers, which are more or less fully flotsam at low gas velocities, are increasingly mixed into the bed with an increase in gas velocity. This is in contrary to the time spent above the dense bed by the tracer with the same density as the emulsion which, as expected, is not greatly influenced by an increase in velocity. For this tracer, the time spent in the freeboard becomes relatively constant above a certain velocity. The reason for the somewhat
8
ACCEPTED MANUSCRIPT lower values of residence time above the dense bed at low velocity is not obvious, but should be related to larger size of the tracer rather than the slight difference in density due to that scaling did not provide a perfect density match (cf. Table 2). Figure 6 underlines the findings given in Figure 3 that an
IP
T
increase in excess velocity enhances axial mixing, especially for tracers with low density. The higher the tracer density the sooner the time spent in freeboard settles on a constant value with an increase in
SC R
velocity. As can be seen from comparing Figure 6a and b, an increase in bed height enhances the effect of increased gas velocity on axial segregation. This should be due to that the bubbles can grow
NU
bigger with an increase in velocity, yielding higher bubble velocities and, thus, a more intense solids mixing. This is also in line with Figure 3d, where all tracers show similar distribution up through the
MA
bed, when applying a high fluidization velocity.
In summary, with an increase in fluidization velocity three regimes can be identified: 1) A purely
D
flotsam regime which occurs at low fluidization velocities and for the low density tracer particles, 2) A
TE
transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of
CE P
tracer particle at the bed surface and the splash zone remains constant with fluidization velocity.
b.
AC
a.
Figure 6: The average time spent by the tracer on and above the bed surface, LP plate, up-scaled values. a) H0 = 0.18 m b) H0 = 0.305 m
Rowe et al. [6] quantified axial solids segregation for jetsam particles in an FB by the mixing factor M, resulting in values between 0 and 1. The present work applies the mixing factor generically, defining M as:
𝑀=
maximum of tracer concentration in the bed tracer concentration in a well mixed bed
Thus yielding values larger than 1 for flotsam tracers.
9
(3)
ACCEPTED MANUSCRIPT Figure 7 shows the mixing factor of three different tracers for increasing u0-umf. Measurements were conducted with the LP distributor plate and conditions corresponding to a fixed bed height of 0.305 m. A perfectly mixed bed, i.e. M = 1 is indicated with a solid line. In line with what is given above, both
IP
T
Biomass and Biochar tracers behave clearly flotsam for low excess velocities. With increasing velocity the mixing factors approach 1 for all three tracers. For low excess velocity for which the average bed
SC R
density is higher, also the Emulsion tracer yields a somewhat flotsam behaviour, but with an increase in gas velocity, soon becomes well-mixed (M approximately 1).
NU
The surface fraction Ffb (Figure 6) defines the share of tracer observations on and above the dense bed surface, i.e. describes the share of fuel particles that will experience increased mass transfer and
MA
decreased heat transfer in comparison to immersed solids [1]. The mixing index M describes the quality of the axial mixing in the dense bed, indicating the maximum clustering of tracers. A low value
D
of M indicates a high local concentration of active particle, i.e. fuel particles, corresponding to a low
TE
effective mass transfer with the bed [31]. Note that Ffb and M give different information and thus can show opposite trends. This can be seen when comparing Figure 6b and Figure 7 for the Emulsion
CE P
tracer at low velocities, i.e. a stronger axial segregation of the tracer (high M) does not necessarily
AC
means that the tracer ends up as flotsam particle on the dense bed surface (low Ffb).
Figure 7: Mixing factor, LP plate, H0 =0.305 m, up-scaled values.
Figure 8 display PDFs of the axial upward and downward velocities of the tracer for measurements corresponding to a fixed bed height of 0.305 m and superficial gas velocity of 0.3 m/s, with Figure 8a showing the Biochar tracer and Figure 8b showing the Emulsion tracer. Median values, m, of each distribution are shown in the legend. The tracer velocity is calculated through the application of Kalman filtering, as explained elsewhere [23], i.e. the frequency employed for the calculation is the sampling rate of the measurement (20 Hz). Figure 8 gives that at low absolute velocities (< 0.25 m/s) both tracer types give a more narrow and centred PDF distribution in the downward velocities than in the upwards velocities. This is in line with previous findings, which show that the upwards motion of the tracer is governed by fast rising bubble paths [3, 32], whereas the downwards motion is following 10
ACCEPTED MANUSCRIPT the slower sinking bed material on the sides of the bubble paths. As indicated by its narrower distribution in Figure 8b, the downwards velocity has weaker fluctuations, which can be explained with the smoother motion of the sinking bed material in comparison to the strongly fluctuating
IP
T
upwards motion of the bubble flow, as previously reported in [12].
SC R
When comparing Figure 8a and b with each other the effect of tracer density can be observed. The Emulsion tracer with its higher density yields higher median velocities, i.e. exhibits higher velocities for both upwards and downwards motion, i.e. due to that its density is similar to the emulsion phase of
NU
the bulk material it mixes better into the dense bed rather than floating on the dense bed surface. This is in line with the results given earlier in this paper, where the Emulsion tracer segregates less on the
MA
bed surface (Figure 6) and shows a better axial mixing inside the bed (Figure 7), including in the cases
D
with lower fluidization velocities.
TE
a.
b.
CE P
Figure 8: PDF of upwards and downwards uZ in the dense bed for H0 =0.305 m, u0 = 0.3 m/s, bin size: 0.02 m/s. a) Biochar tracer b) Emulsion tracer
Figure 9 give the vertical profiles of average upward and downward tracer velocity, urise and usink, for a
AC
low (0.19 m/s) and a high (0.43 m/s) excess velocity (up-scaled values) for the Biochar (Figure 9a) and the Emulsion (Figure 9b) tracer. The bed is divided into slices of around 0.04 m in the axial direction. The bed height (0.305 m) is indicated as a solid line. It should be noted that some points are associated with a considerable scatter (up to 90%). This is due to the sampling frequency may be too low for velocity measurements as discussed in [25]. However, a clear trend is observed for all the curves: It can be seen that for both tracer densities the rising velocities are higher than the sinking velocities and that the influence from an increase in fluidization velocity is greater for the Biochar tracer (Figure 9a) than the Emulsion tracer (Figure 9b). The lowest average velocities are found in the bottom near the air distributor and around the height of the dense bed (the static bed height is indicated). The low axial velocities near the air distributor reflects that the tracer spent a certain time near the gas distributor, yielding low values of the average axial solids velocity. Similar, there is a
11
ACCEPTED MANUSCRIPT certain flotsam behaviour of both tracers, which results in an increase in residence time at the bed surface and, thereby, a reduction in the average axial velocities. This is especially the case for the Biochar tracer which spends significant time on the bed surface (Figure 9a). Once entrained into the
IP
T
freeboard, the velocities are again higher, representing the rising velocities caused by bubble eruption (upward) and the back-mixing into the dense bed (downward). As for the freeboard mixing it can be
SC R
assumed that the tracer solid is to a more or less extent immersed in the emulsion phase, i.e. follows
NU
the ballistic movement of clustered solids mentioned above.
MA
a.
b.
D
Figure 9: Vertical profile of mean urise and usink at two different u0-umf. H0 =0.305 m, up-scaled values.
TE
a) Biochar. b) Emulsion.
Figure 10 compares average axial tracer velocities obtained in this work (on an up-scaled basis) with
CE P
corresponding literature data for different excess gas velocities. The literature data are with conditions similar to those applied in the present work on an up-scaled basis, i.e. H0 = 0.15-0.5 m, ρs = 1 1002 940 kg/m3, dp = 212-800 µm. Figure 10a gives the upward velocities and Figure 10b the downward
AC
velocities, in both cases averaged over the entire dense bed (i.e. up to H0). As for downward velocities of immersed objects, not all of the above mentioned works provide such data. All literature applied tracers with a density similar to that of the average bulk solids, i.e. similar the Emulsion tracer in this work. From Figure 10 it can be seen that in the present work (×), the increase in upward and downward velocities with an increase in fluidization velocity levels out at the higher range of velocities investigated. The upward tracer velocity is around 50% of the theoretical bubble velocity calculated by the Davidson and Harrison’s model [28]. Corresponding values between 10% and 50% are mentioned in the literature [10] [12] [13]. It should be mentioned that the bubble velocity is a function of bed material properties, bed height and gas inlet velocity and therefore differs between the data given in Figure 10. Nienow et al. [9] and Rees et al. [33] used the rather simple measurement method of releasing the tracer close to the distributor in 3D beds and observing the time it takes for the 12
ACCEPTED MANUSCRIPT tracer to reach the bed surface. Soria-Verdugo et al. [12] and Lim et al. [10] measured with optical methods in 2D units. Both Fotovat et al. [14] and Stein et al. [13] used radioactive particle tracking methods. All works which applied particle tracking methods presented tracer particle velocities
IP
T
averaged over the dense bed, i.e. as done in this work. From Figure 10 it can be seen that literature values are generally lower than the values obtained in this study. It is believed that the higher tracer
SC R
rising velocity obtained for the present experiments is connected to the “higher temperature level”, resembled by the fluid-dynamic scaling. All data points from literature are measured in units operating
NU
at cold conditions, not applying fluid-dynamically downscaling. This hypothesis is supported by the available values of lateral solids dispersion coefficients under conditions relevant for industrial units:
MA
Olsson et al. [3] provide an experimentally determined lateral fuel dispersion coefficient in an industrial-scale unit operated under cold conditions, while Sette et al. [34] used a model operated according to scaling laws to resemble the unit used by Olsson et al. under hot conditions. From his set-
TE
D
up Sette et al. obtained lateral dispersion coefficients more than two times higher than under cold conditions. Thus, the smaller (and thereby slower) bubbles generated under cold conditions by Olsson
CE P
et al. induce less intense solids mixing than from larger and faster bubbles under the hot conditions by Sette et al. resembled through the application of fluid-dynamical scaling.
AC
With respect to the downward velocities, Figure 10b shows that data obtained in 2D beds [10] [12] at similar operational conditions as the up-scaled values in the present work but without applying fluid dynamic scaling, give considerable lower values than the ones obtained in the present work. As for the 3D data by Fotovat et al. [14] these are closer to those of the present work, although somewhat lower. As discussed above for the rising velocity, the effect of the temperature level can be argued to cause the difference between the non-scaled experiments and application of scaling laws, revealing the importance of using fluid-dynamical scaling in order to obtain experimental data of quantitative relevance for commercial units. Further, the effect of pressure drop over the gas distributor as well as the effect of reduced fixed bed height on rising and sinking velocities were evaluated (not shown in Figure 10). With increasing pressure drop (see Equations (1) and (2)) both rising and sinking velocities are reduced by around 13
ACCEPTED MANUSCRIPT 20%. This is in line with the findings made in connection to Figure 4, i.e. the HP distributor yielding a stronger flotsam behaviour, which are explained by the reduced bubble size resulting in a decreased bubble velocity. When reducing the fixed bed height H0 from 0.305 m to 0.18 cm the rising velocities
IP
T
are reduced by around 30%, while sinking velocities are reduced by around 40%. Since the bubbles grow up through the bed, shallower beds give smaller bubble size which result in lower bubble
SC R
velocities and, thus, less vigorous solids mixing, reducing the rising and sinking velocities.
NU
a.
MA
b.
Figure 10: Tracer average rising, urise, (a), and sinking, usink, (b), velocity in the dense bed (up to h=H0) as function of the excess gas velocity, u0-umf. Emulsion tracer. H0 =0.305 m, up-scaled values.
D
Compared data from literature: H0 = 15-50 cm, ρs = 1100-2940 kg/m3.
TE
4. Conclusions
CE P
Magnetic particle tracking (MPT) using single tracer particles is used in a fluid-dynamically downscaled bubbling fluidized bed to investigate how axial mixing depends on fluidization velocity, bed height and fuel particle density.
AC
The MPT results show that increase in bed height, fluidization velocity and in fuel tracer density enhance the axial fuel mixing. The axial mixing is, however, similar for the two different air distributors used, although the air distributor with the higher pressure drop yields a stronger flotsam behaviour. This should be due to that the HP gives smaller bubbles, resulting in less vigorous splashing which in turn gives less downward flow of solids dragging down the tracer in the bed. With an increase in fluidization velocity three axial mixing regimes of the fuel tracer can be identified: 1) A flotsam regime occurring at low fluidization velocities and for light tracer particles which, combined show a purely flotsam regime, 2) A transition regime over which an increase in fluidization velocity results in the presence of fuel particles at the bed surface decreases rapidly, and 3) A fully developed mixing regime in which the presence of tracer particle at the bed surface and the splash zone
14
ACCEPTED MANUSCRIPT remains constant with fluidization velocity. The onset velocities between the regimes depend on bed height and tracer properties.
T
As expected, the upward and downward velocities of the fuel tracers depend on bubble properties – the
IP
larger the bubbles, the higher the bubble velocity and the more vigorous the axial solids mixing.
SC R
Comparing these velocities to literature data on an up-scaled basis, the velocities obtained in this work is considerably higher than corresponding values from literature. This is believed to be due to that the present data represent full-scale conditions, as compared to the literature data which were obtained
NU
under cold conditions not applying scaling laws. Yet, there is a need to also measure in real combustion or gasification conditions in order to derive to what extent the present data quantitatively
MA
correspond to axial mixing at industrial conditions.
5 Acknowledgement
D
This work is co-financed by the Swedish Energy Agency – project P38347-1 – and Valmet
TE
Corporation Oy, within the framework of the project “Experimental investigation of the gas-solids
References
CE P
flow in FB units”.
AC
[1] L. Lundberg, P.A. Tchoffor, D. Pallarès, R. Johansson, H. Thunman, K. Davidsson, Influence of surrounding conditions and fuel size on the gasification rate of biomass char in a fluidized bed, Fuel Processing Technology, 144 (2016) 323-333. [2] P.N. Rowe, A.W. Nienow, Particle mixing and segregation in gas fluidised beds. A review, Powder Technology, 15 (1976) 141-147. [3] J. Olsson, D. Pallarès, F. Johnsson, Lateral fuel dispersion in a large-scale bubbling fluidized bed, Chemical Engineering Science, 74 (2012) 148-159. [4] R. Bilbao, J. Lezaun, M. Menendez, J. Abanades, Model of mixing—segregation for straw/sand mixtures in fluidized beds, Powder Technology, 56 (1988) 149-155. [5] R. Bilbao, J. Lezaun, M. Menendez, M. Izquierdo, Segregation of straw/sand mixtures in fluidized beds in non-steady state, Powder Technology, 68 (1991) 31-35. [6] P.N. Rowe, A.W. Nienow, A.J. Agbim, The mechanisms by which particles segregate in gas fluidised beds - binary systems of near-spherical particles Transactions of the Institution of Chemical Engineers 50 (1972) 310-323. [7] A. Marzocchella, P. Salatino, V. Di Pastena, L. Lirer, Transient fluidization and segregation of binary mixtures of particles, AIChE Journal, 46 (2000) 2175-2182. [8] G. Olivieri, A. Marzocchella, P. Salatino, Segregation of fluidized binary mixtures of granular solids, AIChE Journal, 50 (2004) 3095-3106. [9] A.W. Nienow, P.N. Rowe, T. Chiwa, Mixing and segregation of a small proportion of large particles in gas fluidized beds of considerably, AIChE Symposium Series, 74 (1978) 45-53. [10] K.S. Lim, P.K. Agarwal, Circulatory motion of a large and lighter sphere in a bubbling fluidized bed of smaller and heavier particles, Chemical Engineering Science, 49 (1994) 421-424. 15
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
NU
SC R
IP
T
[11] A. Soria-Verdugo, L.M. Garcia-Gutierrez, N. García-Hernando, U. Ruiz-Rivas, Buoyancy effects on objects moving in a bubbling fluidized bed, Chemical Engineering Science, 66 (2011) 2833-2841. [12] A. Soria-Verdugo, L.M. Garcia-Gutierrez, S. Sanchez-Delgado, U. Ruiz-Rivas, Circulation of an object immersed in a bubbling fluidized bed, Chemical Engineering Science, 66 (2011) 78-87. [13] M. Stein, Y.L. Ding, J.P.K. Seville, D.J. Parker, Solids motion in bubbling gas fluidised beds, Chemical Engineering Science, 55 (2000) 5291-5300. [14] F. Fotovat, R. Ansart, M. Hemati, O. Simonin, J. Chaouki, Sand-assisted fluidization of large cylindrical and spherical biomass particles: Experiments and simulation, Chemical Engineering Science, 126 (2015) 543-559. [15] G.M. Rios, K. Dang Tran, H. Masson, FREE OBJECT MOTION IN A GAS FLUIDIZED BED, Chemical Engineering Communications, 47 (1986) 247-272. [16] C.E. Weinell, K. Dam-Johansen, J.E. Johnsson, Single-particle behaviour in circulating fluidized beds, Powder Technology, 92 (1997) 241-252. [17] D. Moslemian, N. Devanathan, M.P. Dudukovic, Radioactive particle tracking technique for investigation of phase recirculation and turbulence in multiphase systems, Review of Scientific Instruments, 63 (1992) 4361-4372. [18] G. Mohs, O. Gryczka, S. Heinrich, L. Mörl, Magnetic monitoring of a single particle in a prismatic spouted bed, Chemical Engineering Science, 64 (2009) 4811-4825. [19] J. Halow, K. Holsopple, B. Crawshaw, S. Daw, C. Finney, Observed Mixing Behavior of Single Particles in a Bubbling Fluidized Bed of Higher-Density Particles, Industrial & Engineering Chemistry Research, 51 (2012) 14566-14576. [20] K.A. Buist, A.C. van der Gaag, N.G. Deen, J.A.M. Kuipers, Improved magnetic particle tracking technique in dense gas fluidized beds, AIChE Journal, 60 (2014) 3133-3142. [21] K.A. Buist, T.W. van Erdewijk, N.G. Deen, J.A.M. Kuipers, Determination and comparison of rotational velocity in a pseudo 2-D fluidized bed using magnetic particle tracking and discrete particle modeling, AIChE Journal, 61 (2015) 3198-3207. [22] E. Sette, T. Berdugo Vilches, D. Pallarès, F. Johnsson, Measuring fuel mixing under industrial fluidized-bed conditions – A camera-probe based fuel tracking system, Applied Energy, 163 (2016) 304-312. [23] E. Sette, D. Pallarès, F. Johnsson, F. Ahrentorp, A. Ericsson, C. Johansson, Magnetic tracerparticle tracking in a fluid dynamically down-scaled bubbling fluidized bed, Fuel Processing Technology, 138 (2015) 368-377. [24] D. Simon, The discrete-time Kalman filter, Optimal State Estimation, John Wiley & Sons, Inc.2006, pp. 121-148. [25] A. Köhler, D. Pallarès, F. Johnsson, Magnetic tracking of a fuel particle in a fluid-dynamically down-scaled fluidized bed Submitted for publication, (2016 ). [26] L.R. Glicksman, M.R. Hyre, P.A. Farrell, Dynamic similarity in fluidization, International Journal of Multiphase Flow, 20 (1994) 331-386. [27] H. Thunman, M.C. Seemann, First experiences with the new chalmers gasifier, 2009, pp. 659663. [28] J.F. Davidson, H. D., Fluidised particles Cambridge University Press New York 1963. [29] A. Svensson, F. Johnsson, B. Leckner, Bottom bed regimes in a circulating fluidized bed boiler, International Journal of Multiphase Flow, 22 (1996) 1187-1204. [30] F. Johnsson, S. Andersson, B. Leckner, Expansion of a freely bubbling fluidized bed, Powder Technology, 68 (1991) 117-123. [31] F. Scala, Particle-fluid mass transfer in multiparticle systems at low Reynolds numbers, Chemical Engineering Science, 91 (2013) 90-101. [32] D. Pallarès, P.A. Díez, F. Johnsson, Experimental analysis of fuel mixing patterns in a fluidized bed, The 12th International Conference on Fluidization, ECI, Vancouver, Canada, 2007, pp. 929-936. [33] A.C. Rees, J.F. Davidson, J.S. Dennis, A.N. Hayhurst, The rise of a buoyant sphere in a gasfluidized bed, Chemical Engineering Science, 60 (2005) 1143-1153. [34] E. Sette, D. Pallarès, F. Johnsson, Influence of bulk solids cross-flow on lateral mixing of fuel in dual fluidized beds, Fuel Processing Technology, 140 (2015) 245-251.
16
ACCEPTED MANUSCRIPT Vitae Anna Köhler
Anna Köhler is a doctoral student at Energy and Environment in the division of
T
Energy Technology at Chalmers University of Technology in Gothenburg. Her
IP
research focuses on the fluid dynamics in fluidized beds, which she addresses
SC R
with both experimental work and modelling. Her work aims at an improved use of non-fossil fuels such as biomass in fluidized bed units.
Alexander Rasch is a Bachelor student in Mechanical Engineering, working
Rasch
with energy engineering related issues. Completing his thesis at Chalmers
NU
Alexander
MA
University of Technology in Gothenburg, he conducted experiments and analysis
David Pallarès is an Associate Professor at Chalmers University of Technology.
TE
David Pallarès
D
related to axial fuel mixing.
His research focuses on experiments and modeling of fluidized bed phenomena
CE P
of relevance for different fuel conversion processes (e.g. air-combustion, oxyfuel-combustion, chemical looping combustion and indirect gasification)
AC
with the aim to assess the design, scale-up and performance of such processes.
Filip Johnsson
Filip Johnsson is Professor in Sustainable Energy Systems in Department of Energy and Environment, Chalmers University of Technology. Filip Johnson’s research area comprises Fluidized bed combustion and gasification, oxyfuelcombustion for CO2 capture energy systems analysis. The fluidized-bed related research focus on fluid dynamics and mixing.
17
Fig. 1
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
18
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 2
19
Fig. 3a
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
20
Fig. 3b
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
21
Fig. 3c
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
22
Fig. 3d
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
23
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
Fig. 4
24
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
Fig. 5
25
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 6a
26
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 6b
27
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 7
28
Fig. 8a
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
29
Fig. 8b
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
30
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 9a
31
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 9b
32
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 10a
33
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
Fig. 10b
34
Anna Khler
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
35
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
Alexander Rasch
36
SC R
IP
T
ACCEPTED MANUSCRIPT
AC
CE P
TE
D
MA
NU
David Pallares
37
AC
CE P
TE
D
MA
NU
SC R
IP
T
ACCEPTED MANUSCRIPT
Filip Johnsson
38
ACCEPTED MANUSCRIPT Table 1: Parameter
Unit
Hot model
Cold model ideal
Temperature
°C
800
Gas density
kg/m3
0.32
Bed dimensions
m×m
L = 0.74 × 0.74
Bed height
m
𝐻0 = 0.18; 0.305
Superficial velocity
m/s
𝑢0 = 0.01 – 0.5
Bed material density
kg/m3
2 600
Bed material size
µm
250
Cold model actual
T
20
IP
1.2
𝐿
AC
CE P
TE
D
MA
NU
SC R
4.4
39
𝑢0 ∙
= 0.17 × 0.17
𝐻0
4.4
= 0.04; 0.07
1 √4.4
= 0.006 – 0.236
9 530
8 900
57
60
ACCEPTED MANUSCRIPT Table 2:
ρhot
ρdownscaled
ρactual
(kg/m3)
(kg/m3)
Biochar
350
Biomass Emulsion
Error
Dmagnet
MS
(kg/m3)
(%)
(m)
(Gauss)
1 280
1 470
-14.6
800
2 930
2 980
-1.7
1 230
4 500
4 320
downscaled
0.05
13050
0.08
12550
0.08
12550
IP
SC R 4.0
NU MA D TE CE P AC
40
T
Tracer density
ACCEPTED MANUSCRIPT
NU
SC R
IP
T
Graphical abstract
MA
Magnetic Particle Tracking is used in a fluid-dynamically down-scaled bubbling fluidized bed to study axial mixing and segregation of fuel particles. The system provides 3D trajectories of the fuel tracer
AC
CE P
TE
D
particle, highly resolved in time and space.
41
ACCEPTED MANUSCRIPT Highlights
CE P
TE
D
MA
NU
SC R
IP
T
Magnetic particle tracking provides resolved 3D trajectories of tracer particle The impact of fuel density, bed height and fluidization velocity is shown Fluid-dynamical downscaling yields higher mixing than non-scaled cold units
AC
42