1
Pelletisation of canola meal by extrusion-spheronisation for ethanol dehydration
C. H. Niua,b, T. Baylak b, D. I. Wilson a,* and M. Zhang a,c
2
a
3
Department of Chemical Engineering and Biotechnology, New Museums Site, Pembroke St,
4
Cambridge CB2 3RA, UK
b
5
Department of Chemical and Biological Engineering, University of Saskatchewan, 57 Campus
6
Drive, Saskatoon, SK, S7N 5N9, Canada
c
7
School of Pharamacy, Health Science Center, Xi’an Jiaotong University, 76 Yanta Westroad,
Xi’an Shannxi 710061, PR China
8
9
10
ABSTRACT
11
Canola meal has been previously demonstrated to be an attractive biomaterial for dehydrating
12
wet ethanol vapours in bioethanol manufacture. Extrusion-spheronisation was employed to
13
prepare reasonably spherical pellets of canola meal for use in dehydration units. Canola meal
14
pastes were prepared at water volume fractions of 57-70% and extruded through single and
15
multi-holed dies with diameters 2, 3.5 and 4.5 mm. The pressure required to extrude the pastes,
16
size and shape distribution of pellets and strength of dried pellets were measured. Formulations
17
with a water volume fraction of 70% gave low extrusion pressures and highest pellet strength.
18
Die land diameters of 2 mm gave the best combination of specific surface area, size and shape
19
distribution, packing density and ethanol adsorption. Dehydration testing confirmed that the
20
canola meal pellets could dehydrate water/ethanol vapour from an ethanol mass fraction of
21
92.5% (below the azeotrope at 1 bara) to 99%. The equilibrium water loading of 47.3 mg water
22
per g adsorbent is larger than other biomass-based adsorbents reported for this application.
23
24
Keywords: Bioethanol, canola meal, dehdyration, extrusion, pelletization, spheronization,
25
*
Corresponding author: Dr D. Ian Wilson, Tel +44 1223 334 791; FAX +44 1223 334 796, E-mail diw11@cam.ac.uk
1
26
1. INTRODUCTION
27
Current concerns about energy supply have prompted the development of various renewable fuel
28
sources. Biofuels have received substantial attention because of the widespread availability of
29
biomass. Biofuels such as ethanol, methanol, isopropanol, and butanol generated by fermentation
30
routes require separation and purification in order to achieve fuel grade specifications, and the
31
current processing techniques have limitations.
32
33
For example, conversion of carbohydrates to ethanol via fermentation usually yield aqueous
34
solutions with ethanol mass fractions of 5-12% ethanol in water with other organics. Recovery of
35
ethanol from the fermentation broth to yield fuel grade ethanol is mainly performed at the
36
industrial scale by distillation, giving ethanol-water mixtures with ethanol mass fractions of 75-
37
92%, below the azeotrope (95.6 % at 1 atmospheric pressure), followed by adsorption using
38
adsorbents [1].
39
40
Biomaterials represent a potential source of biadsorbents. Corn grits have been reported to be
41
used in industry [2] but use of corn places pressure on food supply. Alternative bioadsorbents for
42
bioethanol dehydration included cellulosic materials such as canola meal [3], kenaf core [4], and
43
bleached wood pulp [4], and starchy materials such as cassava pearls [5] and corn meal [6].
44
Studies [3-6] have shown that surface area, density, porosity, particle size, and mechanical
45
strength of adsorbent particles have significant effects on industrial ethanol dehydration. Most of
46
these bio-sorbents are prepared by grinding the raw biomaterial and the resulting particle size and
47
shape distributions are not controlled. No results have been reported to date for biosorbents for
48
ethanol dehydration with controlled shape and particle size. Developing methods for pelletising
49
biosorbents with controlled size, shape and mechanical strength (friability) are important as these
50
properties determine whether the materials can be readily used in packed bed devices.
51
52
Extrusion-spheronisation is widely used in the pharmaceutical industry to manufacture pellets
53
with high sphericity from powder feedstocks [7, 8]. In this process, powders such as
54
microcrystalline cellulose (MCC) are combined with a liquid binder to produce a viscoplastic
55
paste (a highly-filled suspension) which is extruded to give cylindrical extrudates which are
2
56
subsequently broken up and rounded on a rotating friction plate. The pellets are then dried to
57
remove the free (not strongly absorbed) liquid. This technology has potential for making bio-
58
adsorbent pellets from cellulosic or starchy materials. Previous work in paste processing has
59
demonstrated that physical properties including particle size and shape, density, and porosity,
60
liquid content and die geometry all influence the performance of the extrusion and spheronisation
61
steps [9-11].
62
63
This paper investigates the use of extrusion-spheronisation to manufacture bioadsorbent pellets
64
from canola meal. Canola meal is obtained by grinding the canola seed cake after oil extraction.
65
It is an abundant by-product from canola oil extraction and biodiesel production. Its composition,
66
by mass, is36-40% crude protein, 12% moisture, 20% neutral detergent fibre consisting of
67
celluloses, hemicelluloses and lignins, 5% starch, 10% free sugar and non-starch polysaccharides,
68
4% crude fat, and 6% ash [12]. Water adsorption by biomaterials is reported to involve the polar
69
attraction between water and the cellulosic hydroxyl components and the protein carboxyl and
70
amine groups in the adsorbent [13, 14].
71
72
In a previous study, Baylak et al. [3] demonstrated that adsorbents prepared from raw canola
73
meal particles were able to dehydrate ethanol from solutions with mass fraction of 65-95%
74
ethanol to yield fuel ethanol at mass fraction higher than 99%. The meal adsorbs water from the
75
vapour phase, so that the ethanol-rich vapour can be condensed directly to give the product.
76
Regeneration of the meal yields water, which renders the regeneration step relatively safe. This
77
paper presents an investigation of pelletisation of canola meal by extrusion-spheronisation,
78
characterization of the pellets thus generated, and a short trial of their ethanol dehydration
79
performance.
80
3
81
2. MATERIALS AND METHODS
82
2.1 Canola meal
83
The canola meal used in this work was purchased from Federated Co-Operatives Limited
84
(Saskatoon, Canada). The mass fractions of the major meal components (supplier assay) were:
85
crude protein 36.0%, crude fat 2.0-5.0%, crude fibre ((cellulose, hemicellulose, and lignin)
86
12.0%, moisture 12.0%, and non-starch polysaccharides, starch, ash etc. constituting the
87
remainder. The meal was sieved and particles passing through a 500 μm mesh were used for
88
making pellets. In this paper the term ‘particle’ refers to individual elements of the canola meal
89
feedstock and ‘pellet’ is used for the granulated assemblies of particles.
90
91
The moisture content of particles and pellets was determined gravimetrically. The weight of the
92
sample was measured before and after oven drying at 105oC for 24 h, or at reaching constant
93
weight if this was reached earlier. Surface area of particles and pellets was determined by a
94
Micromeritics ASAP 2020 surface area analyzer. The true density of canola meal particles, CM,
95
was measured by a pyconometer (Micromeritics AccuPyc 1330). The as-poured aerated and
96
tapped bulk densities of the meal particles were measured on an automated tap density analyzer
97
(Quantachrome Instruments, Reading, UK) using a measuring cylinder of internal diameter 26.4
98
mm. The cylinder was filled to the 100 ml mark and the mass of charge measured to give the as-
99
poured density. The accuracy of volume readings was ±0.5 ml. The tapped bulk density was
100
obtained at 800 taps when the change of bulk volume was invisible by eye observation.
101
102
2.2 Paste preparation
103
Canola meal pastes were prepared by mixing canola meal particles with known amounts of
104
reverse osmosis water in a Kenwood Chef domestic planetary mixer (Kenwood Ltd, UK). The
105
powder was first mixed at the lowest speed for 5 min and water added slowly by pouring on to
106
the powder bed. Then paste was then mixed at the highest speed for 10 min, pausing every 2 min
107
to remove the paste adhered on the wall and bottom of the mixing bowl with a plastic spatula.
108
The freshly mixed paste was stored in a sealed plastic bag in a refrigerator for 2 h to allow the
109
water to equilibrate throughout the mass. Control of laboratory conditions was important to limit
110
evaporation. The room temperature was 22.7 ± 1.3 C and relative humidity 0.55±0.03.
4
111
2.3 Extrusion and spheronisation
112
The water-canola meal pastes were extruded through a computer-controlled ram extruder based
113
on a Zwick/Roell 50 kN strain frame (Zwick Testing Machines Ltd., Leominster, UK). The
114
extruder comprised a cylindrical ram, cylindrical 316 stainless steel barrel of internal diameter
115
(D0) 25 mm and various concentric square entry dies. A detailed description of the unit is given
116
in Mascia et al. [15].
117
determined by the extrudate diameter, D, so different pellet sizes were obtained by using
118
different die sizes. The dies used in this work all featured cylindrical die lands, with length (L) to
119
diameter ratios (L/D) of 4~4.5. Single, centrally holed dies were used for initial tests, with die
120
land dimensions (L×D) 8 mm × 2 mm; 16 mm × 3.5 mm; 18 mm × 4.5 mm. A multi-holed die
121
was also used to generate 2 mm diameter extrudates: this had six 8 mm × 2 mm holes, with 5
122
spaced evenly in a ring around a central hole (see [16]). The multi-holed die was used as it is
123
more representative of industrial screen extrusion and pelletisation devices.
The average diameter of the pellets obtained by spheronisation is
124
125
About 70 g of paste was loaded into the barrel and tamped down by hand to an initial height of
126
about 160 mm. The charge was then pre-compacted with a blank die in place to a force of 650 N,
127
corresponding to a mean compressive stress (= force/barrel cross-sectional area) of 1.3 MPa. The
128
ram was released immediately once the force was reached. The blank die was then replaced and
129
the paste extruded at a constant ram speed of 1 mm s-1 for up to 100 mm ram displacement. The
130
force on the ram was monitored using a load transducer on the cross-head and the extrusion
131
pressure calculated as above.
132
133
Batches of about 25 g of extrudates were spheronised using a Caleva Spheroniser 120 (Caleva
134
Process Solutions Ltd, UK) fitted with a 120 mm diameter cross-hatched friction plate. The
135
rotation speed was increased gradually from 157 rpm to 750 rpm over a period of 12 min. The
136
pellets were oven dried at 105 C for 24 h. The average water mass fraction of the dried spheres
137
was less than 0.01 %.
138
139
5
140
2.4 Size and shape
141
The particle size and shape distributions of the canola meal particles were determined using a
142
Morphologi G3 automated microscopy system (Malvern Instruments Ltd., Worcestershire, UK).
143
The size and shape distributions of the canola meal pellets were determined by an automated
144
digital video system (JM Canty International Ltd., Ireland). Size data are reported in terms of the
145
equivalent circle diameter, dCE, and shape is quantified using the elongation ratio, ER, defined as
146
Lmax/Lmin, where Lmax is the longest dimension on a 2-dimensional projection of a particle or
147
pellet and Lmin is the length of the chord normal to it.
148
149
2.5 Pellet strength
150
The crushing strength of the canola meal pellets was measured using a Stable Microsystems TA-
151
XT2 Texture Analyser. Individual, near spherical, pellets were located between two horizontal
152
platens. The platens were moved together at an approach speed of 0.1 mm s-1 and the maximum
153
force at failure, Fc, recorded. The tensile crushing strength, Y, was calculated using [17]
Y  0.576
154
Fc
d2
[1]
155
where d is the diameter of the spherical pellet. At least 20 samples were measured from a given
156
batch.
157
158
2.6 Ethanol dehydration
159
The lab-scale packed bed ethanol-water vapour dehydration apparatus described by Baylak [3]
160
was used to simulate the industrial ethanol dehydration process. An ethanol-water solution with
161
ethanol mass fraction of 92.5% is pumped at 0.044 L h-1 from a sealed stainless steel feed tank
162
through a section of electrically heated pipe which serves as an evaporator and passes to one of
163
two 316 stainless steel adsorption columns with dimensions (length 500 mm, i.d. 47.5 mm, wall
164
thickness 1.65 mm) containing randomly packed canola pellets. One column undergoes
165
regeneration while the other dehydrates. The temperature of the adsorption columns is
166
maintained at 85o±0.5 C by external heating tapes. System pressure and temperature are
167
monitored continuously. For dehydration, the average pressure at the inlet of the column was
168
123.6±0.8 kPa, and the average pressure drop across the column was 5.6±0.5 kPa.
6
169
170
The exit vapour is condensed and the product collected in a fraction collector at one minute
171
intervals. The water content is determined by an automated Karl-Fischer Titrator (Titroline KF,
172
Schott Instruments) and ethanol content by HPLC (Agilent, 1100 Series, Refractive Index
173
Detection). HPLC analysis showed no impurities beyond ethanol and/or water in all effluent
174
samples tested.
175
176
For regeneration, the water-saturated pellets were readily dried on-line by purging with inert gas
177
N2 at 98oC under vacuum. The pellets could also be oven dried at 105oC for 24 h and reused.
178
179
3. RESULTS AND DISCUSSION
180
3.1 Raw Material Characterisation
181
Figure 1(a) shows a photograph of the raw canola meal particles. The meal was obtained by
182
grinding canola seed cakes following oil extraction. The particles are irregular in shape, Figure
183
1(b). The particle size distribution in Figure 2 shows that sieving removes the majority of
184
particles larger than 500 μm, but some pass through the sieve diagonal. There are a substantial
185
number of fines and the elongation ratio (ER) values in Figure 2 show that these are relatively
186
angular and sharp, as expected for a milled product. The ER values of the larger particles show
187
that they are roughly spherical.
188
189
Table 1 summarises the canola meal properties, including specific surface area, moisture, density
190
and porosity. The porosity, , of the as-poured aerated bulk canola meal was determined from [18]
191
 1
A
1  w
 CM
[2]
192
where w is the water mass fraction of the meal and A is the aerated bulk density. The porosity
193
value determined by Equation [1] includes the inter-particle voids and pore space within the
194
particles. The value of 0.64 for the aerated bulk canola meal indicates a high packing voidage as
195
well as a significant amount of internal porosity.
196
7
197
The Hausner ratio, HR, is calculated from HR = T/A and gives a measure of the inter-particle
198
friction within a powder [19]. HR values less than 1.25 indicate a Geldart type A material which
199
is free-flowing and easy to fluidise, HR > 1.4 indicates a cohesive powder (Geldat type C) [20].
200
The canola meal HR value was 1.23, indicating that the particles were free flowing, which is
201
desirable for powder handling in devices used to prepare pastes, and also renders them suitable
202
for extrusion.
203
204
3.2 Extrusion
205
3.2.1 Effect of paste water content
206
The liquid volume fraction is a key factor affecting the extrudabilty of solid-liquid pastes.
207
Identifying an optimal liquid content is an important first step in formulating canola meal pastes.
208
The effect of water content was investigated by preparing pastes with volume fractions ranging
209
from 0.57–0.70, i.e. ranging from slightly unsaturated with respect to the measured porosity, , to
210
saturated with excess liquid over that needed to fill the pores and voids. Each paste was extruded
211
through the L/D = 16/3.5 single-holed die at a ram speed of 1 mm s-1. Figure 3 shows examples
212
of the extrusion pressure profiles. The initial rise in Pex is associated with paste filling the die
213
land, after which Pex is steady. The coefficient of variation in the steady Pex region shown in
214
Table 2 was 0.05 or less, indicating that Pex did not change significantly over the test.
215
216
The steady Pex values in Figure 3 are lower than those reported for MCC (Avicel PH101) by
217
Rough et al. [21]: they reported Pex > 12 MPa for an MCC/water paste with a water volume
218
fraction of 61.5%, whereas the corresponding canola meal value is 1.63 MPa. The high value for
219
MCC may be related to the biopolymer absorbing a large fraction of water so that there is little
220
free water in the formulation: canola meal absorbs a small amount of water. The canola meal
221
extrudates were smooth and free from surface defects. These results confirm that the paste route
222
is a suitable method for forming canola meal into shaped products.
223
224
8
225
3.2.2 Liquid phase migration
226
The steady values in extrusion pressure indicated that liquid phase migration, which can affect
227
other paste materials such as MCC at these velocities, was not occurring in these tests. The
228
absence of significant liquid phase migration was confirmed by measuring the water content of
229
the paste material remaining in the barrel (data not reported). The water mass fraction of this
230
material was very similar ( 2 %) to that of the freshly loaded paste. The cellulosic fraction of
231
the canola meal in the paste, at 12 wt%, is much lower the MCC used by Rough et al. [21] (mass
232
fractions of 35 - 55%). The lower cellulose content and larger quantity of proteins, fats and non-
233
cellulosic polysaccharides, giving different affinity for water, is likely to explain the absence of
234
liquid phase migration. A systematic investigation of these factors is needed to establish the
235
effects of the individual components of the canola meal on liquid phase migration. The absence
236
of liquid phase migration is a factor favouring pelletisation by this route.
237
238
3.2.3 Extrusion pressure and formulation
239
The steady Pex values from Figure 3 are plotted against water content in Figure 4. The data
240
exhibit a linear dependency on water content, which least squares regression analysis gave as
241
R2 = 0.9934
log Pex = -0.0702 w + 4.5146
[3]
242
Rough et al. (2002) reported a similar dependency of extrusion pressure on water content for the
243
ram extrusion of MCC/water pastes using an apparatus similar to that employed here that log Pex
244
[18]. Equation [3] is similar to the relationship relating total density of a solids mixture to
245
confining pressure reported by Lukasiewicz and Reed [22]
246
ρ=αlnPex+β
0.7 MPa <Pex < 4 MPa
[4]
247
where ρ is the density of the paste, Pex is the total applied stress (in MPa), and α and β are
248
constants. Rewriting  in terms of the volume fractions of the solid and liquid, i.e. assuming
249
saturation and negligible absorption, yielded a similar form to Equation [3]. The parameters α
250
and β were found to be 21 kg m-3 and 1118 kg m-3, respectively. This result is reported as
251
evidence that the material is extruded in the saturated state, as the extrusion pressure also
252
depends on die geometry and extrusion velocity, as described below.
253
9
254
3.2.4 Die geometry
255
The effect of die diameter was investigated for the w = 0.70 paste by extruding through single-
256
holed 2 mm, 3.5 mm and 4.5 mm dies and the six-holed 2 mm diameter die. The extrusion
257
pressure profiles in Figure 5(a) show the same form as Figure 3. The average steady Pex values,
258
as well as the mean extrudate velocity, V, are summarised in Table 3. V was estimated by
259
assuming that the paste was incompressible, from V = ram velocity × (D0/D)2×1/N, where N is
260
the number of die holes.
261
262
The extrusion pressure decreases as D increases, which is expected from the model for paste
263
extrusion proposed by Benbow and Bridgwater [9], which for a cylindrical barrel and die in its
264
simplest form gives
265
A

L
Pex   Y ln  barrel   4  w
D
 Adie 
[5]
266
where A is the cross-sectional area of the barrel or die land(s), Y is a bulk yield stress and w is
267
the wall shear stress in the die land. The first term describes the deformation work at the die
268
entry and the second accounts for work done in the die land. The Pex data from Table 3 are
269
presented in the form suggested by Equation [5] in Figure 5(b) and the single holed die values
270
follow a roughly linear relationship. The Pex value for D = 2 mm is higher than expected for a
271
linear trend and this is attributed to Y and w being dependent on V (as reported for MCC by
272
Rough et al., 2000 [21]). These tests featured a constant ram speed so V was larger for smaller
273
die lands. A systematic investigation of the effect of velocity and die land parameters to quantify
274
the rheology of the canola meal pastes was not undertaken.
275
276
The extrusion pressure observed for the 6-holed die was similar to that obtained for the single-
277
holed die, which is not consistent with Equation [5]. Similar results were reported by Zhang et al.
278
[16] for extrusion of MCC/water pastes through multi-holed dies, although in that case they
279
varied the ram velocity so that the mean velocity through each die hole was the same. The flow
280
patterns in multi-holed dies are complex and are readily predicted by simple results such as
281
Equation [5].
282
10
283
3.3.Pellets
284
3.3.1 Effect of water on pellet tensile strength
285
The paste water content influences the mechanical strength of the pellet products as well as the
286
extrusion process. The tensile strength of pellets generated using the same extrusion conditions
287
was measured. The pellets were oven dried at 105C for 24 h, to a final water content of 0.01
288
wt%, before testing. The results are plotted on Figure 4 and show an increase in pellet strength
289
with increasing water content, in contrast to the effect of water content on Pex. Increasing the
290
water content reduces the amount of work done in extrusion but will require more energy to
291
remove the water in the drying stage. There is thus likely to be an optimal water content in
292
manufacturing which minimizes the total production cost.
293
294
3.3.2 Pellet size and shape
295
Extrudates generated through four different die sizes and configuration were spheronised, dried
296
to water mass fraction of lower than 0.5% and subjected to size and shape analysis. The yield of
297
pellets from spheronization was about 0.95. The photographs in Figure 6 show that the pellets
298
were rounded and the size increased with die diameter. Pellets prepared using the 6-holed D = 2
299
mm die were similar in appearance to those generated using the single-holed D = 2 mm die.
300
301
The size and shape distributions obtained for samples of approximately 1000 pellets are
302
presented in Figure 7. The equivalent circle diameter is used to quantify the pellet size and is
303
plotted as dCE/D as this provides a useful scaling of the data. The pellet shape is characterised
304
here using the elongation ratio: a number of other measures could be used. All four size
305
distributions exhibit a relatively narrow distribution with few fines and a small number of
306
oversized particles. There are a number of pellets with dCE/D > 1 although the number fraction of
307
such pellets is low (Figure 7(a)). The data were fitted to the normal distribution using the EasyFit
308
5.5 Professional software package and the results are summarised in Table 4. The data show that
309
all four distributions could be described by the normal distribution at the 95% level of confidence.
310
311
The pellet size is also affected by spheronisation conditions and particularly batch size. Figure 8
312
shows photographs of an extrudate from a 3.5 mm diameter die and typical pellets obtained after
11
313
spheronisation in batch sizes of 24 g and 58 g. All other spheronisation conditions were the same.
314
The larger batch size gave noticeably larger pellets, indicating that further investigation is
315
necessary to optimise and control pellet size.
316
317
The measured properties of the spherical pellets are summarised in Table 4. The mean dCE
318
values are slightly smaller than the die land diameter, D, which is attributed to shrinkage during
319
evaporation and drying. The elongation ratio decreases by a small amount with larger die
320
diameters. Chopra et al. [23] report a pellet elongation ratio of 1.25 as being satisfactory for use
321
in tabletting machines for pharamaceutical applications. The 2 mm pellets exhibit values larger
322
than this, so some further optimisation of spheronisation conditions would be required for this
323
pellet size.
324
325
The pellet specific surface area and tensile strength values in Table 4 both decrease with
326
increasing die land diameter (D). The tensile strength of the D = 2 mm pellets (around 3 MPa) is
327
almost twice that of MCC pellets reported by Sousa (2002) [24] but it is lower than the value of 8
328
MPa reported for commercial 3A molecular sieves used in ethanol dehydration [25].
329
Enhancement of pellet mechanical strength is an area for further development.
330
331
The specific surface areas for the pellets generated using the 2 and 3.5 mm diameter dies are
332
similar, and lie in the range of 0.5-1.0 m2 g-1. The value obtained for the 4.5 mm pellets is
333
noticeably smaller. The specific surface areas are considerably larger than the external surface
334
area of the pellets, which was calculated at 0.8-1.8 10-3 m2g-1. For comparison, the reproted
335
specific surface areas of 3A moleclular sieves for water adsorption are in the range of 45-800 m2
336
g-1 [1, 26].
337
12
338
3.4 Ethanol dehydration
339
Dehydration tests were performed with 43 g charges of each of the pellet types in Table 4 and
340
Table 5 summarises the packing characteristics of the packed beds. The increase in bed voidage
341
with dCE is expected as the bed internal diameter is 44.5 mm, which is less than 20 particle
342
diameters for the D = 3.5 mm and D = 4.5 mm pellets and wall effects will arise (reference
343
needed).
344
345
Figure 9(a) shows ethanol profiles in the effluent for each of the pellet types. All the pellets
346
tested were able to increase the ethanol vapour concentration from the feed ethanol mass fraction
347
of 92.5 % to one above the azeotrope, 95.6 %. The smaller pellets gave superior performance,
348
which is consistent with their higher specific surface area. The larger (D = 4.5 mm) pellets gave
349
undesirable scatter.
350
351
The performance of the smaller pellets was investigated further using a larger packed bed volume.
352
Figure 9(b) shows the performance of a bed 1.5× those used in Figure 9(a). Both pellet types
353
were able to concentrate the vapour up to an ethanol mass fraction of 99% ethanol, which is
354
suitable for fuel applications. There was no statistically significant difference between pellets
355
prepared using the single- and multiple-holed dies.
356
357
Adsorption is considered to reach equilibrium (saturation) when the ethanol content of the
358
effluent equals that of the feed. The equilibrium water uptake is defined as the mass of water
359
adsorbed per unit mass of packed dry adsorbent, and was calculated as the difference between the
360
total mass of water fed to the column minus the accumulated mass of water in the effluent,
361
divided by the dry net weight of the adsorbent in the column. This gave 47.3 ± 0.1 mg g-1 dry
362
adsorbent for the pellets prepared with D = 2 mm dies, which is almost twice the value of 25.2
363
mg g-1 reported by Baylak et al. for the raw canola meal particles with sieved size 0.43 mm - 1.18
364
mm [3]. Pelletisation via extrusion and spheronisation improved the water adsorption capacity of
365
the canola meal. The packing densities of the 2 mm spherical pellets, at 586 – 592 kg m-3, are
366
also larger than that of the raw canola meal particles (528 kg m-3), and the pressure drop across
367
the bed was similar.
368
13
369
Table 6 shows that the water uptake of the canola meal pellets is superior to other biomaterials.
370
Laboratory trials confirmed that the water saturated pellets were readily dried in situ by purging
371
with inert gas at 98oC under vacuum before reuse. However, it was also found that canola meal
372
adsorbes ethanol while adsorbing water. For the feed studied here (ethanol mass fraction 92.5% ,
373
water 7.5%) the respective uptake values were 47.3 mg water per gram dry adsorbent and 411 mg
374
ethanol per gram dry adsorbent, which represents a separation factor of water over ethanol of 1.4.
375
Recovery of ethanol in the effluent is 90 ± 0.01 g per 100 g input. Further work could be done on
376
enhancing ethanol recovery efficient by reducing the ethanol uptake of the adsorbent.
377
378
Topics for future work for determining the feasibility of conola meal-based processes include the
379
optimisation of pellet properties identified above, optimisation of deydration conditions,
380
regeneration by air or reduced oxygen mixtures, a life-cycle analysis of the manufacturing
381
process including water reuse and energy demand for drying, and cost-benefit analysis of this
382
route compared to molecular sieves and unmodified canola meal.
383
384
4. CONCLUSIONS
385
Extrusion-spheronisation has been successfully demonstrated as a route for manufacturing
386
nearly-spherical pellets from canola meal for ethanol dehydration. Dry, sieved canola meal was
387
found to be a free-flowing material which could be combined with water (water mass fraction of
388
57-70%, wet basis) to give cohesive pastes suitable for ram extrusion and spheronisation.
389
Extrudates generated using single- and multi-holed dies both yielded acceptable pellets. The
390
extrusion pressure decreased with increasing water content and depended on the die size. The
391
extrusion pressure for the single-holed dies fitted the trend suggested by the Benbow-Bridwater
392
model, whereas the results for the multi-holed die did not: this requires further attention. Liquid
393
phase migration was not found to be significant for the paste formulations and operating
394
conditions employed in these tests.
395
396
The pellet size distributions fitted the Normal distribution reasonably well and exhibited a
397
common trend when sizes were plotted as the ratio of dCE to die diameter, D. The average
398
elongation ratio was about 1.22 - 1.38, indicating that the pellets were acceptable for filling
399
operations such as loading into packed beds. It should be noted that the spheronisation step was
14
400
not optimised. Pastes with higher water content, when dried, were found to give pellets with
401
higher tensile strength, indicating lower friability. The 70 wt% water formulation was therefore
402
studied in greater detail and in dehydration tests. The latter tests confirmed that the canola meal
403
pellets were able to yield fuel grade ethanol from ethanol-water vapour mixtures. The 2 mm
404
spherical pellets gave a higher water uptake than that reported for raw canola meal and a number
405
of other starchy and cellulosic materials.
406
407
408
5. ACKNOWLEDGEMENTS
409
Authors Niu and Baylak were supported by the University of Saskatchewan Research Sababtical
410
travel fund, Agriculture Bioproduct Innovative Program of Canada, Saskatchewan Agricuture
411
Development Fund and Saskatchewan Canola Development Commission.
412
(Cambridge) performed pycnometry and particle analyses; Dr. Ramin Azargohar and Mr. Chuck
413
Oraedu (Saskatoon) determined the surface area of raw canola meal and pellets.
414
15
Zlako Saraçevic
415
6. NOMENCLATURE
416
Roman
417
A
cross-sectional area of the duct, m2
418
d
diameter of spherical pellets, m
419
dCE
equivalent circle diameter, m
420
D0
internal diameter of the barrel, m
421
D
internal die land diameter, m
422
ER
elongation ratio, -
423
Fc
force at crushing
424
HR
Hausner ratio, ρT/ρA, -
425
L
length of die land, m
426
Lmax
longest dimension on a 2-dimensional projection, m
427
Lmin
length of the chord normal to Lmax, m
428
N
number of die holes, -
429
Pex
extrusion pressure, MPa
430
R2
coefficient of determination
431
V
extrudate mean velocity, m s-1
432
w
water mass fraction of the canola meal
433
Y
tensile strength, MPa
434
435
Greek symbols
436
α
constant, kg m-3
437
β
constant, kg m-3
438
ρ
density of canola meal paste, kg m-3
439
ρA
as-poured aerated bulk canola meal density, kg m-3
440
ρT
tapped bulk canola meal density, kg m-3
441
ρ CM
density of canola meal determined by pycnometer, kg m-3
442
Y
bulk yield stress, MPa
443
w
wall shear stress in the die land, MPa
444

bed voidage, or porosity, -
445
16
446
7. REFERENCES
447
448
449
[1] Simo M, Sivashanmugam S, Brown CJ, Hlavacek V. Adsorption/desorption of water and
ethanol on 3A zeolite in near-adiabatic fixed bed. Ind. Eng. Chem. Res. 2009; 48(20):
9247-60.
450
451
452
[2] Beery KE, Gulati M, Kvam EP, Ladisch MR. Effect of enzyme modification of corn grits on
their properties as an adsorbent in a skarstrom pressure swing cycle dryer. Adsorpt. J. Int.
Adsorpt. Soc. 1998; 4(3-4): 321-35.
453
454
[3] Baylak T, Kumar P, Niu CH, Dalai A. Ethanol dehydration in a fixed bed using canola meal.
Energy Fuels 2012; 26(8): 5226-31.
455
456
[4] Benson TJ, George CE. Cellulose based adsorbent materials for the dehydration of ethanol
using thermal swing adsorption. Adsorpt. J. Int. Adsorpt. Soc. 2005; 11: 697-701.
457
458
[5] Kim Y, Hendrickson R, Mosier N, Hilaly A, Ladisch MR. Cassava starch pearls as a
desiccant for drying ethanol. Ind. Chem. Eng. Res. 2011; 50(14): 8678-85.
459
460
461
[6] Chang H, Yuan X, Tian H, Zeng A. Experimental investigation and modeling of adsorption
of water and ethanol on cornmeal in an ethanol-water binary vapor system. Chem. Eng.
Technol. 2006; 29(4): 454-61.
462
463
[7] Vervaet C, Baert L, Remon JP. Extrusion-spheronisation. A literature review. Int. J. Pharm.
1995; 116(2): 131-46.
464
465
[8] Wilson DI, Rough SL. Extrusion-spheronisation. In: Salman AD, Hounslow MJ, Seville JPK,
editors. Handbook of Powder Technology: Granulation, vol. II.: Elsevier, London; 2007.
466
467
[9] Benbow, J.J. and Bridgwater, J. Paste Flow and Extrusion. Oxford, UK: Clarendon Press;
1993.
468
469
470
[10] Vervaet C, Remon JP. Influence of impeller design, method of screen perforation and
perforation geometry on the quality of pellets made by extrusion-spheronisation. Int. J.
Pharm. 1996; 133(1-2): 29-37.
471
472
473
[11] Newton JM. The preparation of pellets by extrusion/spheronisation. In: Augsburger, L.L.
and Hoag, S.W., editors. Pharmaceutical Dosage forms: Tablets. 3rd Ed. Volume 1: Unit
Operation Mechanical Properties. New York: Informa Healthcare USA, Inc; 2008.
474
475
[12] Canola Council of Canola. Canola Council of Canola, Canola Meal Feed Industry Guide
http://www.canolacouncil.org/. 2009; Accessed on May 24, 2012.
476
477
478
[13] Anozie AN, Okuhon EE, Osuolale FN, Adewole JK. Dehydration of ethanol-water mixture
using activated carbons from sawdust and palm kernel shells. Sep. Sci. Technol. 2010;
45(10): 1482-9.
17
479
480
481
[14] Chang H, Yuan XG, Tian H, Zeng AW. Experimental study on the adsorption of water and
ethanol by cornmeal for ethanol dehydration. Ind. Eng. Chem. Res. 2006; 45(11): 391621.
482
483
484
[15] Mascia, S., Patel, M.J., Rough, S.L., Martin, P.J., Wilson, D.I. Liquid phase migration in the
extrusion and squeezing of micro-crystalline cellulose pastes. Eur. J. Pharm. Sci. 2006; 29: 22–
34.
485
486
487
[16] Zhang M, Rough SL, Ward R, Seiler C, Wilson DI. A comparison of ram extrusion by
single-holed and multi-holed dies for extrusion-spheronisation of microcrystalline-based
pastes. Int. J. Pharm. 2011; 416(1): 210-22.
488
489
[17] Dexter AR, Kroesbergen B. Methodology for determination of tensile strength of soil
aggregates. J. Agric. Eng. Res. 1985; 31(2): 139-47.
490
491
[18] Rough SL, Wilson DI, Bridgwater J. A model describing liquid phase migration within an
extruding microcrystalline cellulose paste. Chem. Eng. Res. Design 2002; 80(7): 701-14.
492
493
494
[19] Rough SL, Wilson DI, Bayly A, York D. Tapping characterisation of high shear mixer
agglomerates made with ultra-high viscosity binders. Powder Technol. 2003; 132(2-3):
249-66.
495
496
[20] Geldart D, Harnby N, Wong AC. Fluidization of cohesive powders. Powder Technol. 1984;
37(1): 25-37.
497
498
[21] Rough SL, Bridgwater J, Wilson DI. Effects of liquid phase migration on extrusion of
microcrystalline cellulose pastes. Int. J. Pharm. 2000; 204(1-2): 117-26.
499
500
[22] Lukasiewicz SJ, Reed JS. Character and compaction response of spray-dried agglomerates.
American Ceramic Society Bulletin 1978; 57(9): 798-805.
501
502
503
[23] Chopra R, Podczeck F, Newton JM, Alderborn G. The influence of pellet shape and film
coating on the filling of pellets into hard shell capsules. Eur. J. Pharm. and Biopharm.
2002; 53(3): 327-33.
504
505
[24] Sousa JJ, Sousa A, Podczeck F, Newton JM. Factors influencing the physical characteristics
of pellets obtained by extrusion-spheronization. Int. J. Pharm. 2002; 232(1-2): 91-106.
506
507
[25] Delta Adsorbents. 3A molecular sieves. http://www.deltaadsorbents.com/3a-4x8-molecularsieve-desiccant-1lb/t101c18p43.aspx, accessed on July 18, 2013.
508
509
[26] Al-Asheh S, Banat F, Al-Lagtah N. Separation of ethanol–water mixtures using molecular
sieves and biobased adsorbents. Chem. Eng. Res. Design 2004; 82(7): 855-64.
510
511
18
512
Tables
513
Table captions
514
Table 1 Canola meal particle properties
515
Table 2. Effect of water content on steady extrusion pressure
516
Table 3. Average extrusion pressure and extrudate velocity versus die geometry.
517
Table 4. Summary of pellet properties. Sample size 1000 pellets.
518
Table 5. Bed voidage and packing density using the spherical pellets made by the dies of
519
520
different geometry
Table 6. Comparison of water uptake by biomaterials
521
19
522
Table 1 Canola meal particle properties
523
Specific surface area
2.5±1.3 g m-2
Moisture content, w
11.6±0.4 vol%
As-poured aerated bulk density, CM
528±4 kg m-3
Tapped bulk density, T
652±14 kg m-3
Particle solid density, CM
1334±3 kg m-3
Estimated bed voidage, 
0.64±0.01
524
525
526
527
Table 2. Effect of water content on steady extrusion pressure
Water volume fraction Average extrusion Standard deviation Coefficient of
pressure, Pex
in Pex
variation
(vol%)
(MPa)
(MPa)
(%)
57.1
3.03
0.09
2.9
61.5
1.63
0.03
2.0
65.8
0.85
0.04
5.0
70.0
0.37
0.01
3.2
528
529
20
530
Table 3. Average extrusion pressure and extrudate velocity versus die geometry.
531
Die
L/D = 8/2
L/D = 16/3.5 L/D = 18/4.5
single 6-hole
single
single
Average Pex (MPa)
0.6
0.58
0.38
0.33
Standard deviation (MPa)
0.03
0.03
0.01
0.01
Coefficient of variation
3.2
4.1
2.0
2.4
Average extrudate velocity V (mm s-1)
156
26
51
31
532
533
534
Table 4. Summary of pellet properties. Sample size 1000 pellets.
535
Die, L×D
Property
Number of holes
8×2
8×2
16×3.5
18×4.5
1
6
1
1
Mean size
dCE (mm)
1.59±0.06 1.49±0.06 2.83±0.11 3.74±0.13
Scaled size dCE/D
Mean (-)
0.80
0.75
0.81
0.83
Standard deviation
0.24
0.20
0.12
0.11
Shape
Elongation ratio (-)
1.38±0.29 1.38±0.29 1.25±0.23 1.22±0.14
Specific surface area
(m2 g-1)
0.95±0.30 0.57±0.09 0.52±0.10 0.07±0.01
Bulk density
(kg m-3)
Tensile strength
(MPa)
592 ± 8
587 ± 8
486 ± 3
440 ± 3
2.96±0.20 3.13±0.19 1.42±0.10 1.14±0.05
536
21
537
538
Table 5. Bed voidage and packing density using the spherical pellets made by the dies of
different geometry
539
Die
8×2
8×2
16×3.5
18×4.5
1
6
1
1
dCE (mm)
1.59±0.06
1.49±0.06
2.83±0.11
3.74±0.13
Bed voidage (-)
0.57±0.01
0.55±0.01
0.63±0.00
0.65±0.00
592±8
587±8
487±3
440±3
Property
L×D
N
Packing density (kg m-3)
540
541
542
543
544
Table 6. Comparison of water uptake by biomaterials
545
Equilibrium water loading
Biomaterial
(mg per g adsorbent)
Canola meal spherical pellets (2 mm) (this work)
47.3
Raw canola meal particles [3]
25.2
Corn meal [6]
30.5
Cassava pearls [5]
26.0
Kenaf core [4]
6.3
Bleached wood pulp [4]
11.6
546
547
22
548
Figures
549
Figure captions
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
Figure 1. Photographs of canola meal particles (a) bulk powder (scale divisions are spaced 1 mm
apart); (b) individual particles, imaged by Morphologi G3 automated microscope system
using a 2.5 × magnification lens.
Figure 2. Raw canola meal particle size analysis showing size distribution (number fraction) and
predicted shape distribution (elongation ratio). Data truncated below 5 μm.
Figure 3. Effect of water content on extrusion pressure profiles. Conditions: single-holed die,
L/D =16 mm/3.5 mm, ram velocity of 1 mm s-1, temperature 22.7 ± 1.3 C, humidity
55.1±3.9 (%).
Figure 4 Effect of paste water content on (i) extrusion pressure, L/D = 16/3.5 single holed die,
ram velocity 1 mm s-1 (log scale); and (ii) tensile strength of dried pellets generated by
spheronisation of the extrudates (linear scale). Trend lines added as guide to the eye.
Figure 5 Extrusion through die lands of different diameter. Single holed dies, L/D = 4-4.5; w = 70
vol%, 22.7±1.3 oC, humidity 55.1±3.9%, ram speed 1 mm s-1. (a) Extrusion pressure
profiles; (b) steady Pex values plotted as Benbow-Bridgwater model, Equation [5]
Figure 6 Photographs of pellets generated from different dies, w = 70 vol%. (a) D = 2 mm, L = 8,
mm single hole; (b) D = 2 mm, L = 8 mm, 6-hole die; (c) D = 3.5 mm, L = 16 mm single
hole die; (d) D = 4.5 mm, L = 18 mm single hole die. Ruler scale shows 1 mm divisions.
Figure 7 Pellet (a) size and (b) shape distributions for w = 70 vol% pastes extruded through
different dies.
Figure 8 Photographs of extrudate and pellets obtained using D = 3.5 mm die. Ruler scale
divisions are 1 mm apart. A - extrudate; B - typical pellet, 24 g batch in spheroniser; C typical pellet, 58 g batch.
Figure 9 Ethanol dehydration performance of canola meal pellets: (a) 43 g bed charge; (b) 65 g
bed charge, D = 2 mm pellets only.
23
583
(a)
584
585
(b)
586
587
588
589
590
Figure 1. Photographs of canola meal particles (a) bulk powder (scale divisions are spaced 1 mm
apart); (b) individual particles, imaged by Morphologi G3 automated microscope system
using a 2.5 × magnification lens.
24
591
1
3
0.9
Number fraction (%)
0.7
2
0.6
0.5
1.5
0.4
1
0.3
0.2
Elongation ratio (-)
2.5
0.8
0.5
0.1
0
1
592
10
100
1000
0
10000
d CE (μm)
593
594
595
596
Figure 2. Raw canola meal particle size analysis showing size distribution (number fraction) and
predicted shape distribution (elongation ratio). Data truncated below 5 μm.
597
598
25
4
3.5
w = 57.1 v/v(%)
P ex / MPa
3
2.5
2
w = 61.5 v/v(%)
1.5
1
w = 65.8 v/v(%)
0.5
w = 70.0 v/v(%)
0
0
20
40
60
80
100
ram displacement (mm)
599
600
601
602
Figure 3. Effect of water content on extrusion pressure profiles. Conditions: single-holed die,
603
L/D =16 mm/ 3.5 mm, ram velocity of 1 mm s-1, temperature 22.7 ± 1.3 C, humidity
604
55.1±3.9 %.
605
606
607
26
608
1.6
10
Pex
Tensile strength
1.4
1.2
0.8
1
Y /MPa
P ex /MPa
1
0.6
0.4
0.2
0
0.1
55
60
65
70
75
w /v/v%
609
610
611
Figure 4 Effect of paste water content on (i) extrusion pressure, L/D = 16/3.5 single holed die,
612
ram velocity 1 mm s-1 (log scale); and (ii) tensile strength of dried pellets generated by
613
spheronisation of the extrudates (linear scale). Trend lines added as guide to the eye.
614
615
27
616
(a)
0.8
0.7
P ex (MPa)
0.6
0.5
0.4
0.3
D = 2 mm 6 hole
0.2
D = 2 mm 1 hole
D = 3.5 mm 1 hole
0.1
D = 4.5 mm 1 hole
0.0
0
20
40
60
80
100
120
ram displacement (mm)
617
618
(b)
0.8
single hole
0.7
6-holed die
P ex /MPa
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1
2
3
4
5
6
ln(A barrel/A die)
619
620
621
622
623
Figure 5 Extrusion through die lands of different diameter. Single holed dies, L/D = 4-4.5; w = 70
vol%, 22.7±1.3 oC, humidity 55.1±3.9%, ram speed 1 mm s-1. (a) Extrusion pressure
profiles; (b) steady Pex values plotted as Benbow-Bridgwater model, Equation [5]
28
624
D = 2 mm
L = 8 mm
Single hole
625
D = 2 mm
L = 8 mm
6 hole
(a)
(b)
D = 3.5 mm
L = 16 mm
Single hole
626
(c)
D = 4.5 mm
L = 18 mm
Single hole
(d)
627
628
Figure 6 Photographs of pellets generated from different dies, w = 70 vol%. (a) D = 2 mm, L =
629
8 mm single hole; (b) D = 2 mm, L = 8 mm, 6-hole die; (c) D = 3.5 mm, L = 16 mm
630
single hole die; (d) D = 4.5 mm, L = 18 mm single hole die. Ruler scale shows 1 mm
631
divisions.
632
633
29
634
(a)
16
4.5 mm
Number fraction (%) .
14
3.5 mm
12
2 mm 6-hole
2 mm single hole
10
8
6
4
2
0
0
0.5
1
1.5
2
2.5
3
d CE /D
635
636
(b)
3
4.5 mm
Elongation ratio (-) .
3.5 mm
2.5
2 mm 6-hole
2 mm single hole
2
1.5
1
0.5
0
0
0.5
1
1.5
2
2.5
3
d CE / D
637
638
639
640
Figure 7 Pellet (a) size and (b) shape distributions for w = 70 vol% pastes extruded through
different dies.
30
641
A
B
C
642
643
644
645
646
Figure 8 Photographs of extrudate and pellets obtained using D = 3.5 mm die. Ruler scale
647
divisions are 1 mm apart. A - extrudate; B - typical pellet, 24 g batch in spheroniser; C -
648
typical pellet, 58 g batch.
649
650
31
651
(a)
Ethanol concentration in effluent (wt%)
100
azeotrope
95.6 wt%
95
feed
92.5 wt%
90
2 mm 6 hole die
2 mm single hole die
85
3.5 mm single hole die
4.5 mm single hole die
80
0
20
60
40
80
100
time (min)
652
653
(b)
Ethanol concentratrion in effluent (wt%)
100
D = 2 mm single - hole die
99
D = 2 mm 6 - hole die
98
97
96
azeotrope
95.6 wt%
95
94
93
0
654
655
656
20
40
60
80
100
time (min)
Figure 9 Ethanol dehydration performance of canola meal pellets: (a) 43 g bed charge; (b) 65 g
bed charge, D = 2 mm pellets only.
32