Fuel Processing Technology 76 (2002) 23  39
www.elsevier.com/locate/fuproc
Particle and handling characteristics of wood fuel
powder: effects of different mills
a, b a
*
Susanne Paulrud , Jan Erik Mattsson , Calle Nilsson
a
Swedish University of Agricultural Sciences, Unit of Biomass Technology and Chemistry/BTK, P.O. Box 4097,
SE-904 03 Umeå, Sweden
b
Department of Agricultural Engineering, Swedish University of Agricultural Sciences, P.O. Box 66,
SE-230 53 Alnarp, Sweden
Received 1 January 2001; received in revised form 1 September 2001; accepted 1 December 2001
Abstract
Wood fuel powder cut in various types of impact mills and knife mills was characterised using
sieve analysis, laser diffraction, image analysis and a funnel method for measuring the tendency to
bridge. In addition, this study aims to find the relationship between particle characteristics, tendency
to bridge and the type of mill. The different types of mills produced wood powder with different
properties. Both sieve analysis and laser diffraction showed that impact mills produced wood powder
with more fine particles than wood powder produced by knife mills. Image analysis showed that
particle sizes and shapes differed between the two types of mills. The bridging tendency, measured as
the size of a funnel opening needed for the powder to pass freely, was higher for powder produced by
impact mills than by knife mills. The bridging tendency was found to be a combined effect of particle
size and shape. © 2002 Elsevier Science B.V. All rights reserved.
Keywords: Wood fuel powder; Impact mill; Knife mill; Sieving; Laser diffraction; Image analysis; Bridging
1. Introduction
Wood powder is a biofuel produced by raw materials such as sawdust, shavings, and
bark. In order to produce the most efficient fuel properties, the raw material is crushed,
dried, and fine milled. Because of the increasing amounts and varieties of wood powders
used in combustion plants in Sweden, handling and combustion characteristics need to be
*
Corresponding author. Tel./fax: [phone].
E-mail address: [e-mail] (S. Paulrud).
0378-3820/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S 0378-3820(02)00008-5
24 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
examined. Wood powder can be produced with a variety of different physical properties
such as particle size distribution, particle shape, and moisture content. Raw materials, the
type of mill, and the type of sieve affect these properties [1,2].
Biofuel producers use a variety of techniques, the most common wood powder
production techniques use impact mills. Impact mills use discs, hammers, swing beaters,
beater plates, or impeller plates to crush the material. Other producers use various types of
knife mills to reduce material between several knives.
Particle properties and moisture content affect a powder s tendency to bridge and flow
in transport and storage systems [3,4]. Furthermore, these parameters influence efficiency
and emission levels when burning fuel powders [5,6]. Various methods can determine
particle size distribution of powders [7]; sieving is the most common method used in
commercial fuel powder plants. Other techniques, such as laser diffraction, may also be
used [7,8]. In most methods, size distribution is the only parameter studied and particles
are often described as or assumed to be spheres. Particle shape, however, may influence
physical behaviour as well and previous studies have shown that particle shape
influences bulk behaviour such as bridging [9,10]. Image analysis identifies many
variables that can be used to describe the powder. Moreover, the method allows direct
observation and verification of single particles [7,11]. Dealing with the large number of
variables is a major problem. By using principal component analysis (PCA), the variation
in data is described in a few principal components or latent variables, orthogonal to each
other [12].
One of the most widely used methods to determine handling characteristics is to
measure the flow properties. This information can be obtained from tests carried out in
some form of shear cell [9]. However, these methods are based on the assumptions that the
material is a homogeneous and isotropic continuum and is fully plastic [13]. In other
words, the particles should be small compared to openings, should not be irregular in
shape, and should not be elastic. Consequently, these methods are not suitable for elastic
powders with fibrous and irregular particle shape, which usually is the case for biomass
powder. Different bridging apparatus methods can be used for biomass material where a
tendency to bridge over outlet openings is studied [10,14,15].
In order to control the production, handling, and combustion process of biomass
powder, informative and simple characterisation methods should be developed. This study
aims to characterise wood powder cut in different types of mills using four techniques:
sieve analysis, laser diffraction, image analysis, and a new funnel method for measuring
the tendency to bridge. In addition, this study aims to find the relationship between particle
characteristics, handling characteristics (tendency to bridge), and the type of mill.
2. Experimental
2.1. Raw materials
Fuel chips from mixed pine and spruce stemwood (without bark, moisture content of
12% and particle length of 20 50 mm) were the raw materials. A commercial wood
powder from SBE Ulricehamn, Sweden was used as a reference material (I5).
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 25
2.2. Milling
The wood chips were first cut in a shredder (Lindner FRM-Micromat 2000, Lindner,
Spittal an dr Drau, Austria) with a 15-mm sieve. The material was then cut in five
different industry mills: four impact mills (I1, I2, I3, I4) and one knife mill (K1). One
sample was cut in a laboratory knife mill (K2). The following mills were used: a Chips
grinder HK 26 type hammer mill with fan wheel (I1) (Franssons, Sundsvall, Sweden)
and with a 1.5-mm sieve; an Akron HS 40 type hammer mill without a fan wheel (I2),
(AB Akron-maskiner, Järpås, Sweden) and with a 1.5-mm sieve; a Mafa President 4K
(I3) type hammer mill without a fan wheel (Mafa, Ängelholm, Sweden) and with a 1.5-
mm sieve; a Kamas Bac-50 (I4), hammer mill without a fan wheel (Kamas, Malmö,
Sweden) and with a 1.0-mm sieve; a Rapid 1528 (R1)-type knife mill with 12 rotor
knives and 2 stationary knives (Rapid Granulator, Bredaryd Sweden) and with a 2.0-mm
sieve; a Brabender 880800 (R2) laboratory mill, type knife mill with six rotor knives
and four stationary knives and with a 1.5-mm sieve (Brabender GHG, Duisburg,
Germany). The reference powder I5 was produced using a hammer mill (Jesma Sprout
Multimill 1001) without a fan wheel (Sprout Matador, Denmark) and with a 2.25-mm
sieve.
2.3. Sampling
Approximately 3 l of powder was taken from several places in the powder sack (200
kg) with a sample drill. The powder was well mixed and a half litre was then divided into
eight parts in a spinning riffler, a Retsch DR 1000 (Retsch, Haan, Germany). One
container was used for each method to determine the particle size (sieving, laser
diffraction, and image analysis).
2.4. Sieve analysis
Samples of 20 g were sieved for 10 min using a sieving machine Fritsch analysette
with DIN sieves: 1.6, 1, 0.8, 0.5, 0.315 and 0.25 mm. The analyses were done in
triplicate.
2.5. Laser diffraction
Laser diffraction measurements were performed using a Malvern Mastersizer 2000 with
a measuring area of 0.02 2000 Am (Malvern Instruments Nordic, Uppsala, Sweden) and it
was equipped with a dry powder feeder, Scirrocco with 2.0 bar pressure. All samples were
measured two times and the sample time was 30 s. A tablespoon (a few grams) of powder
was used for each sample.
2.6. Image analysis
The powder sample used for image analysis (a few milligrams) was blown out on 20
object glass plates (76 26 mm) by using a pipette bulb and a glass tube filled with
26 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
powder. The powder was well spread on the glass plates to eliminate aggregates from
forming. To avoid dust, the glass plates were kept in a covered aquarium.
The samples were observed using a M420   makroskop  (Wild, Heerbrugg, Switzer-
land) with an apozoom objective using two magnifications, objective 5.8 (magnification
1, particle length from 391 to 6000 Am) and objective 35 (magnification 2, particle
length from 19 to 391 Am). The particles were filmed with a monochrome TV camera
CCD, model Ikegami, ICD-42 E (Ikegami Tsuskinki, Japan). The used image analysis
system was a Visor version 2.23, F/64 (Omnivisor KB, Rimbo, Sweden). To ensure the
statistical validity of the result, a relatively large number of particles were examined. The
number of observed particles were 500 for each magnification; stable distribution was
verified using v2 test. For each observed particle, five size parameters and four shape
factors were recorded. Shape characterisation describes shape in as much detail as possible
without any assumptions that could bias the description.
The glass plates were taken in random order. All fields were measured on each plate
with magnification 1. The same glass plates were used with magnification 2 but all fields
were not measured. The fields were measured and randomly taken one row at a time with
magnification 2. The number of measured glass plates varied, depending on the amount of
particles on the plates.
Measured size parameters include the following:
Area (A)  The sum of all pixels that are part of the object.
Perimeter ( P)  The perimeter of the structure represents the length of the
boundary line. Limiting factors are the scale, the camera resolution, and how much
the image is processed.
Feretperimeter ( F )  The perimeter of a polygon with 64 corners that circumscribe
the object.
Length (L)  The longest distance that can be measured between the middle points
of two contour pixels of the object.
Width (W )  The length of the object s projection on a plane, perpendicular to L-
axis.
Measured shape factors were:
Form-PE (PE) (also called circularity). Form-PE = 4k area/(perimeter)2. (PE) can
have values from 0 to 1. For a perfect circle, the value is 1 and for an elongated
particle the value is close to 0.
Form-AR (AR) is an area/rectangular shape factor. Form-AR=(4 area pixels)/(p
length width). For a rectangular shape, the value is close to 1.
Form-Feret (FF). Form-Feret = Min Feret/Max Feret [16]. For a more circular or
quadratic shape, the value is closer to 1 and for elongated narrow particles the value
is closer to 0.
Roughness (R). Roughness = Perimeter/Feretperimeter. (R) may have values higher
or lower than 1 depending on the fine structure of the boundary line of a particle.
This factor depends on the magnification used or the resolution obtained. A more
irregular shape gives a higher value.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 27
2.6.1. Principal component analysis (PCA)
The image analysis data was first classified into 4 classes. Class 1 represents
particles with low values and class 4 represents particles with high values. Every class
shows the number of particles as a percentage. Empirical mathematical modelling
using principal component analysis (PCA) was used to evaluate the image analysis
results.
Results are presented as component scores for the object and loadings for the variables.
The component scores are latent variables in the data matrix and the loading variable
indicates how important a variable is for the position in the score plot. The number of
statistical significant PCA components was determined using cross validation. The
variables were weighted by using unit variance (UV) scaling [12]. Simca 7.01 (Umetrics,
Umeå Sweden) program was used.
2.7. Bulk density
Poured or aerated bulk density (random loose packing) and tapped bulk density
(random dense packing) [17] was measured. The method used in this study involves
filling a graduated cylinder (250 ml) in the same way every time, and using the
weight and volume to determine the poured bulk density. Having determined the
poured bulk density, the cylinder was lifted 50 mm and dropped. This procedure was
performed 10 times to get the tapped bulk density and was repeated three times for
each sample.
Fig. 1. Apparatus used for bridging test of wood powder.
28 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
2.8. Tendency to bridge
The apparatus that was used for the bridging experiments is shown in Fig. 1. The analysis
measures the critical opening needed to avoid bridging. Two litres of powder was poured
into plastic funnels with different opening sizes and with a shutter in the bottom. When the
shutter was opened, the smallest bottom diameter where the powder could fall free was
registered. This procedure was repeated five times for each funnel and approximately four
funnels were used for each powder. The angle of the funnel was 60j and the bottom
diameter varied from 5 to 16 cm (0.5 cm between each funnel). The upper diameter of each
funnel was 25 cm and the plastic material used was propene. During these procedures, a
moisture content sample was taken and analysed according to Swedish Standard [18].
2.8.1. Partial least squares (PLS) projections to latent structures
To evaluate the relationship between the studied methods and the tendency to bridge,
partial least square projections to latent structures (PLS) was used. The method relates x
variables (characterisation result) to the y variables (bridging result) via a mathematical
model. The number of statistical significant PLS components was determined using cross
validation. The variables were weighted by using unit variance (UV) scaling. [12]. Simca
7.01 (Umetrics) program was used.
3. Result and discussion
3.1. Sieve analysis
Fig. 2 gives the cumulative size distribution for wood powder from 5 impact mills and 2
knife mills. The figure shows the cumulative weight fractions of particles smaller than
indicated size (relative standard deviation of three runs 0.5 50%). The powder samples
from the impact mills (I1, I2, I3, and I5) showed similar distributions and the material from
the knife mills (K1 and K2) showed slightly different distributions. The materials from the
knife mills resulted in less fines. Furthermore, the powder (I4) cut in a hammer mill with a
smaller sieve (1 mm) showed a different distribution with more fines.
Sieve analysis is a relatively simple and cheap method. However, according to several
authors there are problems and inaccuracies [7,19]. Too much powder will prolong the
sieving process and effect the result due to small particles binding up on the screening
surface and reducing the aperture size (blinding). The efficiency of grading a sample will
depend on the vibration of the sieves. The passage of particles through a sieve aperture
will depend on the shape of the particles and the shape of the aperture. Spherical particles
tend to blind the sieves and fibrous, narrow particles tend to go right through the sieves.
Thus, it is difficult to compare sieve analysis results from one user to another and compare
biomass powder from different raw materials. However, sieve analysis is a cheap method
for comparing the same type of material in the production process and it produces relative
data regarding changes in size distribution. To produce accurate results from sieve
analysis, sieve sizes, measurement times, and operating methods need to be standardised
for biomass powder.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 29
Fig. 2. Sieve analysis for wood powder cut in different mills: I1  I5 impact mills and K1  K2 knife mills.
3.2. Laser diffraction
Fig. 3 gives the cumulative size distribution for wood powder from five impact mills
and two knife mills. The figure shows the cumulative volume (percentage of the volume of
the equivalent sphere) of particles smaller than indicated size. The figure shows similar
distribution curves as for sieving; sample I4 had more fines as compared to the other
powders.
Laser diffraction, a rapid analytical method that produces results in a couple of minutes,
is a highly reproducible technique; however, as when using many other instruments the
Fig. 3. Laser diffraction analysis for wood powder cut in different mills: I1  I5 impact mills and K1  K2 knife
mills.
30 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
particles are described as or assumed to be spheres. The method underestimates large
particles if they are few and also a dependency of particle shape is obtained [20,21].
However, laser diffraction has the advantage of high speed and it should be possible to
adapt the technique in order to obtain more accurate results for a powder like wood
powder. A recent study has shown that PCA can improve the detection of large particles
[21].
It should be noticed that the particle size and shape effects all size measurement
techniques, which should be remembered when interpreting results.
3.3. Image analysis
3.3.1. Particle shape
Fig. 4 shows examples of different particle shapes and Table 1 shows the shape class
the particles represent. The image analysis data was interpreted using principal component
analysis (PCA) and the result can be seen in the score and loading plots in Figs. 5 and 6 for
magnification 1 (particle length 391 4000 Am) and Figs. 7 and 8 for magnification 2
(particle length 19 391 Am). The score vectors are called t1 and t2 and the loading vectors
p1 and p2. Three principal components (pc) explain 90% (pc1 54%, pc2 28%, pc3 8%) of
the variation in data for magnification 1 and 82% (pc1 47%, pc2 23%, pc3 12%) for
magnification 2.
The score plot (Fig. 5) is a map of how the seven powders relate to each other. Powders
close to each other, e.g. I2, I3, I4, and close to plot origin have similar and average
properties, whereas powders far from each other (K1 and I5) are dissimilar with respect to
shape factors.
The loading plot (Fig. 6) displays the relationships between all variables at the same
time. The further away from plot origin a variables lies, the more influence the variable has
on the model. This means that FF1, AR4, R4 (narrow, elongated, rectangular, regular
Fig. 4. Example of different particle shapes for wood powder.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 31
Table 1
Shape classes that the particles represent
Particle PE AR FF R
1 PE1 AR3 FF2 R2
2 PE4 AR3 FF4 R1
3 PE3 AR2 FF4 R2
4 PE4 AR3 FF4 R1
5 PE3 AR2 FF4 R4
6 PE2 AR1 FF4 R4
7 PE2 AR3 FF3 R2
8 PE1 AR4 FF1 R1
shapes) as well as FF4 and PE3 (quadratic, shorter, broader shapes) are important for the
separation of powder I5, I1, and K1, K2.
For magnification 2 (smaller particles, Figs. 7 and 8), powder I1 showed similar
properties to powder I2, I3, I4 when K1 and K2 were more dissimilar to each other. As for
magnification 1, elongated particles (FF1, FF2) are important for powder I5 and more
quadratic particles (FF4) are important for K1 and K2. For these smaller particles, powder I5
also contained more circular particles (PE4) and powder K2 also contained elongated
particles (PE1). Particles from I1 to I4 also showed less roughness (R1, R2). Combination of
magnification 1 and 2 (an even sample distribution assumed and correcting for the different
area measured at the two magnifications) gave a similar result as for magnification 2.
Fig. 5. Principal component vectors of scores t1 vs. t2, which summarise all variables entering the analysis. Figure
shows two first principal components for identification of shape factors (magnification 1) for wood powder cut in
different mills.
32 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
Fig. 6. Principal component loading vectors p1 vs. p2, showing how the variables are combined to form the
scores. Figure shows two first principal components of shape factor data (magnification 1).
The differences in particle shape between the powders are clearer for the large particles
(magnification 1). For these particles it can also be seen visually that powder particles
from the knife mills are more quadratic in shape compared to powder I5 that had more
Fig. 7. Principal component vectors of scores t1 vs. t2, which summarise all variables entering the analysis. Figure
shows two first principal components for identification of shape factors (magnification 2) for wood powder cut in
different mills.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 33
Fig. 8. Principal component loading vectors p1 vs. p2, showing how the variables are combined to form the
scores. Figure shows two first principal components of shape factor data (magnification 2).
elongated particles. The smaller particles (magnification 2) showed a combination of
quadratic, elongated, and circular particles but the powders were still separated. However,
the large particles are few in number compared to small particles; this may explain why the
result is the same for magnification 2 as for magnification 1 and 2 combined.
3.3.2. Particle size
The particle length varied between 19 and 3900 Am and the width between 3 and 2300
Am. Figs. 9 and 10 show the cumulative size distribution of particle length and width
(magnification 1 and 2 combined). The largest difference is between powder K2 and I4, I5.
The image analysis results in this study were principally used to describe the shape and
size of the individual particles. It is an interesting technique because it is possible to look
directly at the particles in question. By using image analysis and measuring different shape
factors, it was possible to see some differences that were not described by the size
distribution methods, sieve analysis, and laser diffraction. However, it is a method limited
to two dimensions and because it is a very tedious method, it is not suitable as a production
control technique.
3.4. Bulk density
The bulk density of a powder changes depending on the way particles are packed.
Therefore, there is no unique value for a given powder. Fig. 11 shows poured and tapped
bulk densities (relative standard deviation 0.14 2.7%). All powders showed similar
results; powder K2 showed the highest value and powder I4 showed the lowest value.
34 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
Fig. 9. Cumulative size distribution (particle length Am) for wood powder measured by image analysis: I1  I5
impact mills and K1  K2 knife mills.
Analysis of variance (ANOVA) with Tamhane s T2 pairwise multiple comparison tests
(poured density) showed that K2 differed significantly from the other powders except for
I1. There was also a significant difference between powders I2 and I4. As expected the
tapped bulk density is higher than the poured bulk density. The order between powder
qualities is the same for both measures of the bulk density except for I1 and I2.
3.5. Tendency to bridge
Fig. 12 shows the result from the bridging experiments. The powders produced by the
impact mills (I1 I5) showed a higher tendency to bridge compared to the powders
produced by the knife mills (K1, K2). Table 2 shows the fuel moisture content during the
Fig. 10. Cumulative size distribution (particle width Am) for wood powder measured by image analysis: I1  I5
impact mills and K1  K2 knife mills.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 35
Fig. 11. Poured and tapped density (kg/m3) for wood powder produced in different mills. Mean values of three
experiments.
experiments. The moisture content is slightly lower compared to the raw material (12%)
except for powder K2.
The study indicates that funnels could be used to give a relative estimation of the
tendency to bridge. However, the method needs to be developed further. The powder must
be handled in a closed system in order to avoid loss of fine particles from the sample. It
was noticed that the bridging tendency decreased for some of the powders during the
experiments. This was especially noticeably for the powders that contained the most fine
particles. More attention should also be paid to the funnel material because it may effect
the friction and electrostatic charging, which also affects the behaviour of powders.
Finally, a bridging apparatus with an outlet opening that can be opened to different sizes
Fig. 12. Critical opening needed to avoid bridging (5) (passes through five out of five tries) and opening size
where bridging occurs (0) (no pass in five tries).
36 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
Table 2
Moisture content of wood powder during bridging experiments
Powder Moisture content (%)
K1 8.3
K2 11.9
I1 7.1
I2 7.7
I3 8.5
I4 4.7
I5 7.5
may be a better solution if the material is handled less and if less fine particles are lost as
dust.
3.6. Relationship between the studied methods to characterise powder and effect of mills
Sieving and laser diffraction gave different size distribution curves because the
techniques measure different size properties. The sieve analysis measures the size in sieve
diameter, i.e., the width of the minimum square aperture through which the particle passes.
Laser diffraction measures the diameter of a rotating particle and assumes that this refers to
a sphere. Thus, the single particles are recorded as larger by laser diffraction than by sieve
analysis but the graphs show similar results (Fig. 13). Despite different measuring
techniques, they produced the same result: powder from the impact mills contained more
fines compared to the powders from the knife mills.
The size distribution curves from image analysis differs from sieving and laser. One
explanation is that image analysis measures the actual particle length and width in two
dimensions whereas the result from sieving and laser is very shape-dependent (length,
Fig. 13. Particle size distribution measured by sieve analysis and laser diffraction. Wood powder I4 and K1.
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 37
width, and roughness of particles). There is also a danger in combining two magnifications
if the distribution of sample over the measuring area is not even. Image analysis was the
only method used that could measure the different shapes of the particles. This analysis
also showed a difference between impact mills and knife mills. The impact mill produced
more elongated particles while the knife mill produced a shorter and broader particle. In
the literature, it has been reported that powder samples may change shape throughout their
size range [1], a trend seen when comparing results from magnification 1 and 2.
Particle size distribution by sieving, particle shape (magnification 1 and 2 combined),
moisture content, and bulk density were correlated to the bridging result through multi-
variate statistical analysis. The x variables include the following: sieve size ( < 250, > 250,
>315, >500 and >800 Am); particle shape (PE1, PE2, PE3, PE4, AR1, AR2, AR3, AR4,
FF1, FF2, FF3, FF4 and R1, R2, R3, R4), moisture content and poured density. The y-
variable includes the funnel diameter that allowed the powder to fall. The PLS modelling
yielded a three-component model explaining 83% of the variation in x and 99% of the
variation in y.
Fig. 14 shows PLS regression coefficients for the y variable. These express the relation
between the y variable (funnel diameter) and the x variables. X variables with large values
(positive or negative) are highly correlated with y. Positive values increase the tendency to
bridge and negative values decrease the tendency to bridge. The dominating factors that
effect the tendency to bridge are particle size (800 and 315 Am) and shape factors PE3
(circularity) and R2 (particle roughness). A higher weight percentage of smaller particle
sizes typical for I1 I5 increase the tendency to bridge (positive value) and larger particles
typical for K1 and K2 decrease the tendency to bridge. The two dominating shape factors
(R2 and PE3) are typical for powder I1 I4, especially I4. The moisture content and bulk
density showed minor influence on the model.
The difference in tendency to bridge seems to be a combined effect of particle size and
shapes-characteristics that are highly mill-type specific. Impact mills are designed to
Fig. 14. Regression coefficients of scaled and centered variables for the response, funnel diameter. Three model
components (Cum).
38 S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39
reduce particle size by impact energy transfer rather than by shearing, which usually
results in larger quantities of fines [2]. As particle size decreases the cohesion is expected
to increase [17] and the particles stick together more easily. Particle shape is also
important. The inter-particle force between a sphere and a plane surface is higher than
between two equally sized spheres [22]. The influence of particle size and shape on the
tendency to bridge is confirmed by other studies [9,10]. Another important parameter that
influences bridging is moisture content [10]. In this study, the moisture content was fairly
similar and of minor importance. However, it may explain the increased tendency to bridge
for powder K2 (12%) compared to K1 (8%).
4. Conclusions
Sieve analysis is a simple and cheap method. However, results depend not only on the
size of the particles but also on their shape and the operating methods. Laser diffraction
analysis is rapid and highly reproducible; however, the analysis costs are also higher. The
two methods gave similar results and gave the same relative picture of particle size
distribution in the tested powders.
Image analysis was the only method that could measure different shape classes of the
particles. However, it is limited to two dimensions and is too time consuming to be
suitable as a production control technique.
The tested method using a series of funnels with increasing openings to determine
bridging tendency is very similar to emptying powder from a silo. The principle is
promising but needs further development in order to work as a standard method.
The different types of mills produced wood powder with different properties. Both
sieve analysis and laser diffraction showed that impact mills produced wood powder with
more fine particles than wood powder produced by knife mills. Image analysis showed
that particle sizes and shapes differed between the two types of mills.
The bridging tendency, measured as the size of a funnel opening needed for the powder
to pass freely, was higher for powder produced by impact mills than by knife mills. The
bridging tendency was found to be a combined effect of particle size and shape.
Acknowledgements
The staff at Malvern Instruments (Uppsala, Sweden) and Rapid Granulator (Bredaryd,
Sweden) is gratefully acknowledged. This study was financially supported by the Swedish
National Energy Administration.
References
[1] M.G. Scanlon, J. Lamb, Fracture mechanisms and particle shape formation during size reduction of a model
food material, J. Mater. Sci. 30 (1995) 2577  2583.
[2] M. Himmel, M. Tucker, J. Baker, K. Rivard, K. Grohmann, Comminution of biomass: hammer and knife
mills, Biotechnol. Bioeng. Symp. 15 (1985).
S. Paulrud et al. / Fuel Processing Technology 76 (2002) 23 39 39
[3] K. Ridgway, R. Rupp, The effect of particle shape on powder properties, J. Pharm. Pharmacol. 21 (1969)
30  39.
[4] E. Teunou, J.J. Fitzpatrick, E.C. Synnott, Characterisation of food powder flowability, J. Food Eng. 39
(1999) 31  37.
[5] J. Duxbury, G.B. Welford, The effect of particle size on NOx emissions during the firing of pulverised coal
in a shell boiler, J. Inst. Energy 452 (1989) 147  151.
[6] J. Kramlich, W. Seeker, G. Samuelsen, Observations of chemical effects accompanying pulverized coal
thermal decomposition, Fuel 67 (1988) 1182  1189.
[7] A. Rawle, The Basic Principles of Particle Size Analysis, Malvern Instruments, Malvern, UK, http://
www.malvern.se/laboratory/psintro.htm (2001).
[8] Z. Ma, H. Merkus, J. de Smet, C. Heffels, B. Scarlett, New development in particle characterization by laser
diffraction: size and shape, Powder Technol. 111 (2000) 66  78.
[9] C.R. Woodcock, J.S. Mason, Bulk Solids Handling. An Introduction to the Practice and Technology,
Leonard Hill, London, 1987, ISBN: 0-249-44169-1.
[10] J.E. Mattsson, Basic handling characteristics of wood fuels: angle of repose, friction against surfaces and
tendency to bridge for different assortments, Scand. J. For. Res. 5 (1990) 583  597.
[11] M. Pons, H. Vivier, K. Belaroui, B. Bernard-Michel, F. Cordier, D. Oulhana, J.A. Dodds, Particle morpho-
logy: from visualisation to measurement, Powder Technol. 103 (1) (1999) 44  45.
[12] L. Eriksson, E. Johansson, N. Kettaneh-Wold, S. Wold, Introduction to Multi- and Megavariate Data
Analysis using Projection Methods (PCA and PLS), Umetrics, Sweden, 1999.
[13] G.G. Enstad, A novel theory on the arching and doming in mass flow hoppers. Thesis. NTH, Bergen,
Norway. 1981.
[14] R.K. Eckoff, P.G. Leversen, A further contribution to the evaluation of the Jenike method for design of mass
flow hoppers, Powder Technol. 10 (1974) 51  58.
[15] W. Bernhardt, Handling properties of sugar cane bagasse Part III: The phenomenon of bridging, Zuck-
erindustrie 124 (1) (1999) 456  461.
[16] Manual to Visor version 2.21. Omnivisor KB, Rimbo, Sweden (1997).
[17] E.C. Abdullah, D. Geldart, The use of bulk density measurements as flowability indicators, Powder Technol.
102 (1999) 151  165.
[18] Swedish Standards Institution, SS18 71 70:2. Solids and peat Determination of total moisture content,
1990.
[19] K. Leschonski, Sieve analysis, the cinderella of particle size analysis methods, Powder Technol. 24 (1979)
115  124.
[20] M. Natito, O. Hayakawa, K. Nakahira, H. Mori, J. Tsubaki, Effect of particle shape on the size distribution
measured with commercial equipment, Powder Technol. 100 (1998) 52  60.
[21] Z. Ma, H. Merkus, J. de Smet, C. Heffels, B. Scarlett, New developments in particle characterization by laser
diffraction: size and shape, Powder Technol. 111 (2000) 66  78.
[22] J.C. Williams, Mixing and segregation in powders, in: M. Rhodes (Ed.), Powder Technology, Wiley,
England, 1995, pp. 71  90.