Fuel Processing Technology 127 (2014) 80–87

Contents lists available at ScienceDirect

Fuel Processing Technology

journal homepage: www.elsevier.com/locate/fuproc

Hammer mill operating and biomass physical conditions effects on

particle size distribution of solid pulverized biofuels
Miguel Gil ⁎, Inmaculada Arauzo
Centre of Research for Energy Resources and Consumptions, University of Zaragoza, Mariano Esquillor 15, E-50018 Zaragoza, Spain

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

Article history: Milling is a required pre-treatment for the use of biomass as a pulverized solid biofuel in some thermochemical
Received 26 March 2014 technologies such as combustion, gasification and bioethanol production, as well as in densification processes.
Received in revised form 4 June 2014 The particle size plays a key role on these energy conversion technologies. Experimental tests for poplar and
Accepted 7 June 2014
corn stover were performed to obtain pulverized material at different physical conditions of the biomass
Available online 28 June 2014
(input particle size and moisture content) and operational parameters (opening sizes of the screen and angular
Keywords:
speed of hammers). Fourteen parameters related to size central trends, dispersion and shape of particle size dis-
Milling tribution (PSD) were calculated and analyzed by a novel data post-processing methodology, combining Artificial
Poplar Neural Networks and statistical analysis. Results show that the characteristic size of the product (geometric mean
Corn stover size) is mainly influenced by the classification of the screen with values from five to eight times lower than their
Particle size distribution openings size. The angular speed of the hammer governs the variability and dispersion of sizes. The higher the
Neural network angular speed, the lower the dispersion on particle size. Physical conditions of the biomass present a negligible
effect on PSD.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction With regard to goal 1, the studies on the characterization of the

The European Commission established the “20-20-20” targets in
order to achieve in 2020 a greenhouse gas emission reduction by 20%,
a renewable energy contribution of 20% on the final energy consump-
tion and an increase of the energy efficiency by the 20% [1]. In this
context, biomass should play a key role due to its huge potential for sub-
stitution of fossil fuels with significant reduction of carbon dioxide
emissions assessed between 55% to 98% [1]. However some economic,
social and policy barriers must be overcome.
Regarding biomass, energy conversion technologies based on
pulverized/powdered materials present important advantages such as a
higher area/volume ratio (specific surface area), improving the energy
conversion efficiency and the material behavior on handling labors. How-
ever, fine grinding is energy intensive and a balance needs to be struck be-
tween this cost and efficiency improvement. Research efforts are required
around the requirements of the particle size of milled biomass:
• A methodology to characterize the particle size and shape in depth
• Improvements on the characterization and control of milling process
to reduce the energy consumption and to obtain particles with the
desirable sizes.
• An accuracy description of the size requirements imposed by each
final user
⁎ Corresponding author. Tel.: +34 976 761863; fax: +34 976 732078.
E-mail addresses: miguelgc@unizar.es (M. Gil), iarauzo@unizar.es (I. Arauzo).
biomass particle morphology have been traditionally scarce. However,
the high diversity on shape and on size (from few microns to the highest
imposed size limit) due to the breakage process itself deserves a higher
effort in its description. In this sense, a few researchers [2–6] focused on
obtaining not only the particle size by standard procedure but also
the effects of non-sphericity on the description of biomass particle
morphology.
Most milling reports are focused on goal 2. In this regard, authors
focused on the influence of several operational variables and the kind
of biomass on milling energy consumption and the size distribution of
the milled product. Himmel et al. [7] tested poplar, wheat straw and
corn stover with hammer- and knife mill, varying the rejection screen
size, the feed rate and the underpressure on the mill discharge. Mani
et al. [8] and Bitra et al. [9] tested switchgrass, corn stover and wheat
straw with hammer mill at several operational conditions. In addition,
[9] also studied the knife mill varying the rejection screen size from
12.7 to 50.8 mm, the feed rate from 120 to 660 kg h−1 and the operating
speed from 250 to 500 rpm [10–12].Gil et al. [13] showed also the influ-
ences of input biomass conditions of poplar and corn stover and operat-
ing factors on milling energy consumption, on the drying effect of
biomass during milling and on other handling behavior indicators
such as bulk density and angle of repose of the ground particles.
Finally and related to goal 3, multiple researchers focused on deter-
mining the size requirements and effects of pulverized biomass not only
on the handling properties but also on the efficiency of mass and heat
transfer in combustion, gasification and bioethanol production. In

http://dx.doi.org/10.1016/j.fuproc.2014.06.016
0378-3820/© 2014 Elsevier B.V. All rights reserved.
 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87 81

on the final particle size in order to establish the conditions of milling

Nomenclature
process to obtain the desired final particle size.
Cg coefficient of gradation
2. Materials and methods
Cu uniformity coefficient
dgm geometric mean diameter (mm)
2.1. Biomass products
dp particle diameter (mm)
dtarget maximum particle size of obtained product from milling
Two biomass resources were tested: SRF poplar and corn stover.
process (mm)
They differ in their cultivation goals (forestry energy crop vs residue
d10 effective size (mm)
of agricultural crop), the kind of resource (woody vs herbaceous) as
d50 median diameter (mm)
well as in their chemical compositions and physical properties. Howev-
HSgm high graphic geometric standard deviation
er, both resources present one of the highest potential for biofuels due
hi input moisture content
to the high yields of production of poplar as short rotation energy crop
IS input particle size (mm)
and the abundance of corn stover, one of the most widespread residues
lRR Rosin–Rammler size parameter (mm)
worldwide.
Iu uniformity index
Tested SRF poplar (Populus spp.) was cultivated in Fuente Vaqueros
Kg geometric kurtosis
(UTM coordinates: 30 N 431306 4118679), province of Granada, south
lRR Rosin–Rammler size parameter
of Spain and corn plant (Zea Mays L.) was cultivated in Sariñena (UTM
LSgm low graphic geometric standard deviation
coordinates: 30 N, 738030 4629314) province of Zaragoza, northeast
MRs mass relative span
of Spain. Each resource was divided into four fractions. Each fraction
mRR Rosin–Rammler distribution parameter
was chipped and dried under different operational conditions, obtaining
rev experimental variable under analysis: angular speed
material at coarse-wet, coarse-dry, fine-wet and fine-dry conditions.
Sgm geometric standard deviation
Raw materials at different conditions of input particle size and moisture
Skg geometric skewness
content allow evaluating their effects on milling process. A complete
description about the physical conditions of each biomass fraction has
been reported on Gil et al. [13].
Acronyms
ANN Artificial Neural Network
2.2. Sample analysis
CCD central composite design
PSD particle size distribution
Standard specifications for sampling method (CEN/TS 14778–1:2005
RSM response surface methodology
EX [30]), moisture content (CEN/TS 14774–1:2004 [31]) and particle
RR Rosin–Rammler distribution
size analysis (CEN/TS 15149–2:2006 [32]) were used. Sieves (under
SRF short rotation forestry
standard ISO 3310–1:2000 [33]) with apertures of (0.1, 0.25, 0.5, 1, 2
and 5) mm were assembled on a vibrating screen machine (Orto Arlesa,
vibro model). Milled materials were screened for 20 min. Sieving time
was established in the pre-test to guarantee that any mass changes of
combustion processes, the optimal size is established according to several the size fractions must be below 0.3% min−1 (CEN/TS 15149–2:2006
criterions such as the residence time [14–16]; the thermodynamic con- [32]) related to the total sample mass.
version of the particle inside the combustion camera (e.g. Lu et al. [17] For the particle size analysis, probability density function and cumu-
showed a volatile yields decreasing with an increasing particle size and lative distribution function were calculated. Fourteen parameters from
Sudhakar and Kolar [18] found the strongest influence of size on both distributions are obtained to get a complete characterization of
devolatilization time for wood particle); the flame stability (very fine par- particle size distribution (PSD). Geometric method for mean diameter
ticles, dp b 0.1 mm, plays a key role on the stability of the flame in pulver- (dgm), standard deviation (Sgm), skewness (Skg) and kurtosis (Kg) are
ized wood burners [19,20]); and the chemical composition itself calculated according to Folk and Ward formulation [34]. Rosin and
(Bridgeman et al. [21] reported that the fine fractions have a significantly Rammler function [35] fit the cumulative distribution function with R2
higher concentration of inorganic matter, moisture and nitrogen content values around 99% (Section 3.5). F(dp) is the cumulative undersize per-
but lower carbon content and calorific value than larger particles for centage, mRR is the distribution parameter and lRR, the size parameter:
switchgrass and reed canary). Similar effects can be observed in gasifica-
    m
tion processes [22,23]. Regarding bioethanol production, a decreasing of F dp ¼ 1−exp − dp =lRR
RR
Þ: ð1Þ
particle size involves higher hydrolysis yields of the lignocellulose [24,
25], but a depth size characterization is required because some studies
[26] show that a size reduction further than 0.4 mm has little effect on hy- Rosin–Rammler distribution is used to obtain the required size per-
drolysis rates. centiles (dx) to calculate the rest of parameters to characterize particle
Other authors focused their works on the effects of particle size on size distributions: mean diameter (d50), effective size (d10), high and
handling behavior of milled biomass. Ileleji and Zhou [27] observed a low graphic geometric deviation (HSgm = d84 / d50; LSgm = d50 / d16),
better flow behavior of corn stover at lower particle size by means of mass relative span (MRS = (d90 − d10) / d50), uniformity index
angle of repose tests. Paulrud et al. [2] found that tendency to bridge (Iu = d95 / d5), uniformity coefficient (Cu = d60 / d10) and coefficient
was a combined effect between particle size and shape for wood pow- of gradation (Cg = d302 / (d60 − d10)).
der (dp b 2.25 mm). Gil et al. [28] also observed this tendency for
three kinds of milled cardoon at different particle size (dp b 0.5– 2.3. Data post-processing strategy: Artificial Neural Network and statistical
5 mm), and later reported the underlying mechanism that govern the analysis
handling behavior for poplar and corn stover (dp b 5 mm), partially,
due to effects of particle size and shape [29]. Artificial Neural Networks (ANN) are very useful mathematical
The aim of this study is to find the influence of the biomass conditions structures for characterizing and predicting complex problems in
(moisture content and size at mill inlet) and the mill operational vari- which multiple variables and their interactions play relevant role on
ables (opening size of output screen and angular speed of the hammers) final results. Taking into account that it is not suitable to perform
82 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87
experimental test with all the possible variable combinations, ANN [36,
37] provide a suitable model, able to reproduce complex multivariate
non-linear relationships even from noisy and incomplete data and, at
the same time, they are conceptually simple, computationally efficient,
and well supported by professional technical computing software
environments.
Relevant independent variables such as hammermill operational
parameters and physical conditions of biomass (Section 2.4) are the in-
puts of ANN models and the fourteen PSD parameters are the outputs
(Section 2.2). The individual architecture of neural networks is both
simple and regular, aiming to be computationally efficient and to
preserve the ability to generalize beyond training data: they are feed-
forward networks, with sigmoid neurons in the hidden layer and one lin-
ear neuron in the output layer. They are trained using back-propagation
(Levenberg–Marquardt) algorithms, which include an automated Bayes-
ian regularization method [38], allowing for fast training and good gen-
eralization. Feed-forward back-propagation algorithms consist of two
phases: training and testing [39]. In this study, two ANN were designed
and trained from experimental data for each biomass resource, one hav-
ing with five sigmoid neurons in the hidden layer for dgm, d50 and lRR as
outputs, and another one having with ten sigmoid neurons in the hidden
layer for d10, Sgm, HSgm, LSgm, MRS, Iu, Cu, Cg, Skg, Kg and mRR as outputs.
ANN models also allow illustrating the influences of inputs on output
variables by means of 3D predictive surface from ANN and super-
imposed points from experimental tests (see Fig. 2 shows dgm as a func-
tion of the hi and dtarget). However, a systematic analysis is required to
determine the effects of each input variable and their interactions on
each of the fourteen output parameters of PSD. For such propose, the
standard response surface methodology (RSM) by central composite
design (CCD) was selected to perform the statistical analysis with a con-
fidence level of 95%. CCD consists of 2n full factorial runs with 2n axial
runs and n center runs, in which the variable levels must be precisely
established and identified for statistical analysis by (− 1, 0, + 1) low,
medium and high levels, respectively. For this reason, experimental
data could not be analyzed directly by CCD due to the input moisture
content variability of the biomass. However, predictive model from
the ANN can generate the output results, at the desired levels of input
moisture content, in order to be further processed by CCD analysis.
2.4. Test planning
Experimental test campaign was carried out with a horizontal axis
hammer mill of 11 kW and 3000 rpm nominal angular speed. Input bio-
mass rate was regulated by means of the linear velocity of the feeding
conveyor belt as a function of the current electric consumption of the
mill, avoiding mill clogging (more details of the hammer mill and exper-
imental facility can be found in Gil et al. [40,13]). Tests were carried out
on stationary continuous regime in which around 15–20 kg of biomass
were milled.
In relation to the hammer mill, two operational variables may have
important effects on the milled product quality: 1) the opening sizes
of the screen surrounding the mill chamber, as a key factor on particle
classification and 2) the hammer angular speed and therefore the im-
pact energy between particle and hammer, as a key factor on particle
fracture. The other main factor group with repercussion on milling pro-
cess is the physical condition of biomass, mainly, its moisture content
and particles size.
Following, the four factors are enumerated:
• Biomass moisture content (h): wH2 O ¼ ð5 to 10Þ% and (25 to 30)%
• Biomass input particle size (is): previously chipped with (50 or 20)
mm mesh size.
• Process target diameter (dtarget): (5, 2) mm
• Angular speed on hammer mill rotor (rev): (2000, 3000) rpm
A total of 32 tests were performed for SRF poplar as result of 24 var-
iable combinations and a repeat for each test. In the Results section, it
can be read that the input particle size (within the analyzed range)
has no relevant effect on the particle size of the milled poplar. In addi-
tion, Gil et al. [29] showed that this variable had also no effects on ener-
gy consumption, drying of biomass during milling or some handling
properties of the milled poplar. Therefore this variable was not consid-
ered on the test campaign for corn stover in order to decrease the num-
ber of test from 32 to 16 and reduce the experimental campaign costs.
On the other hand, three new central point tests for ANOVA analysis
were required. In central point tests, the variables have the medium
value within their mentioned ranges: wH2 O ≈18%, mixture input particle
size at 50% on mass basis, dtarget = 3.5mm and rev = 2500 rpm. As a
result, 54 experimental tests (32 + 3 for SRF poplar and 16 + 3 for
corn stover) were performed.
Obviously, the predictive tool based on Artificial Neural Networks
and the statistical results have been demonstrated within the variable
ranges (enumerated previously) and any extrapolation outside these
ranges must require future works.
3. Results
Size is the most widespread studied descriptor of pulverized
particles by means of its size-probability distribution (PSD). Size
and cumulative size distributions can be used for knowing the
mass percentage or the cumulative undersize mass percentage for
each size (Fig. 1), showing a log-normal behavior for comminuted
biomass.
To characterize the effects of input variables on the particle size,
fourteen descriptor parameters of the size distribution were analyzed.
Multivariate statistical analysis allows for clustering them (in most
cases R2 N 99%) in three groups in order to describe different character-
istics of the size distribution of milled biomass.
3.1. Clustering of PSD parameters
The first group consists of parameters related to average sizes and
percentiles (dgm, d50 and d10) of the distribution. These parameters are
named as ‘size parameters’. Multivariate analysis shows very strong
correlation R2 99% between dgm and d50, and slightly lower with (d10)
effective size (R2 N 90%).
The second group is composed by parameters related to dispersion
or variety of particle sizes, named ‘dispersion parameters’. The most
widespread used is the geometric standard deviation (Sgm) (ISO
9276–2:2001), representing the dispersion between each particle size
and the geometric mean size. Sgm can be graphically divided into high
graphic Sgm(HSgm = d84 / d50) and low graphic Sgm(LSgm = d50/d16), cor-
responding to high-medium and medium-low fraction of the distribu-
tion, respectively. In addition, other dispersion parameters are used
mostly regarding to ratios between the medium-high and low fractions
of the distribution: mass relative span (MRs = (d90 − d10) / d50), unifor-
mity index (Iu = d95 / d5), uniformity coefficient (Cu = d60 / d10), and co-
efficient of gradation (Cg = d230 / (d60 − d10)). Multivariate analysis
shows strong correlations between them (R2 N 98%).
The third group, denominated ‘shape parameters’ is related to
the distribution shape in comparison with the log-normal distribution
regarding to the lack of symmetry (geometric skewness), as well as
peakness or flatness of the distribution (geometric kurtosis).
Once parameter clustering was performed, neural networks are
used to obtain the results of the 14 descriptor parameters at any pos-
sible combination of the input variables. Obtained results from ANN
are statistically analyzed and graphically visualized on the 3D sur-
face response to find the influences of each input variable on the
particle size distribution. Experimental data are represented over
these 3D surfaces to show the good agreement of the neural
networks.
 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87
Fig. 1. Examples of cumulative particle size distribution for poplar and corn stover at three
levels of dtarget (5, 3.5 and 2) mm and different conditions of hi, IS and rev.
3.2. Size parameters
3.2.1. Characteristic size
Two characteristic size parameters were considered: geometric and
median mean diameters (dgm and d50). As mathematically was expect-
ed, d50 (values between 0.32 mm and 1.12 mm) always was greater
than dgm (values from 0.26 mm to 0.93 mm) due to the fine skewness
of the distribution (see Section 3.4). In Fig. 2, trends and values of dgm
for poplar and corn stover can be observed. The same shape of the
ANN response surface is obtained ford50 (not shown), therefore it
means similar effects of the operational variables (dtarget and rev) and
biomass conditions (hi and IS) on both characteristic sizes.
As it can be expected, there are a great influence of dtarget on average
size (lower dtarget means lower dgm and d50). However, this target size is
only a physical restriction imposed in the process (opening sizes of the
screen) and its value is very different to the average sizes of the milled
biomass. For example, dgm always presents much lower values than
dtarget (between 6 and 8 times lower at dtarget = 5 mm and between 5
and 7 times lower atdtarget = 2 mm). As a conclusion, dtarget should be
considered only like a reference size imposed in the process, but not a
characteristic size of milled product. Mani et al. [8] obtained similar
gaps to dgm = (0.41, 0.26 and 0.19)mm as results from dtarget = (3.2,
1.6 and 0.8)mm for corn stover at wH20 = 6.22 %.
83
Fig. 2. Predictive surfaces of neural network and experimental data points for geometric
mean diameter for poplar and corn stover at different hi, dtarget and rev.
Other tested variables (hi, IS and rev) presented a much lower signif-
icance than dtarget; hi has a slight relevance for poplar and rev for corn
stover. Generally, higher rev and lower moisture content decreases
slightly the mean sizes. Bitra et al. [9] also observed this effect of the
rotational speed on milled corn stover.
3.2.2. Effective size
Effective size (d10) is a distribution percentile representing the fine
fraction. The influence of input variables on effective size are similar
than mean sizes. ANOVA analysis showed also the strongest influences
of dtarget for both resources and a slight effects of rev, hi and IS, depending
on the type of biomass. For poplar, higher rev and hi implies higher
effective size, and therefore lower percentages of fine particles and the
opposite effect on corn stover (more fine).
3.3. Dispersion parameters
3.3.1. Geometric mean deviation
Geometric mean deviation (Sgm) is the most widespread dispersion
parameter for a log-normal distribution. Mathematically, Sgm computes
the difference between each particle size in relation to the dgm. Values of
Sgm were similar for both biomass: from 2.25 to 3.05 for poplar and from
2.38 to 3.27 for corn stover. The main differences between resources
were:
• For both biomass, the higher the rotational speed the narrower the
PSD and therefore more particles close to characteristics size. Howev-
er, the effect was more noticeable for poplar due to its size heteroge-
neity decreased faster than that of the corn stover when the rev was
increased. For this reason, poplar presented higher size heterogeneity
at low angular speed (2000 rpm) whereas higher size heterogeneity
84 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87
was found at 3000 rpm for corn stover.
• At high moisture contents of the biomass, corn stover presented
higher size dispersion than poplar.
As it can be seen in Fig. 3, a dispersion parameter as Sgm showed ab-
solutely different behavior than size parameters (Section 3.2). ANOVA
analyses concluded that Sgm is mainly influenced by the angular speed
and input moisture content and very slight by the dtarget, conversely to
analyze for size parameters. In this case, lower rev and higher hi caused
less distribution dispersion and therefore more homogeneous sizes
around dgm. In Fig. 3, it can be observed the significant separation be-
tween surfaces at different rev (2000 rpm on top) and the scarce reper-
cussion of dtarget.
Within distribution dispersion, it is possible to analyze the disper-
sion related to coarse or fine particles. High graphic geometric deviation
(HSgm = d84/d50) defines the dispersion related to the high-medium
distribution region (coarse), as well as low graphic geometric deviation
(LSgm = d50/d16) for the medium-low region (fine). Predictive neural
network surfaces (not shown) show similar behavior than Sgm and
ANOVA analysis established the same influences and repercussions of
the input variables as well as multivariate analysis achieved also strong
correlations (R2 N 99%) between them. However, HSgm presents a higher
average dispersion than LSgm, 31% for poplar and 36% for corn stover,
therefore fine concentration is higher than coarse ones, showing a fine
skew distribution (Section 3.4) and therefore dgm b d50 in all cases.
Regarding to the first variable on dispersion size influences (rev), an
increase on angular speed involves more homogeneity size effect on
poplar than on corn stover. However, it is unknown in which regions
are exactly more affected: coarse, medium or fine. By means of
HSgm = (d84/d50), a stronger decrease of coarse fraction can be conclud-
ed since d50 decreases likewise (Section 3.2.1). Experimental data and
Fig. 3. Predictive surfaces of neural network and experimental data points for standard
geometric mean deviation for poplar and corn stover at different hi, dtarget and rev.
predictive 3D-NN surfaces (not presented) show higher coarse reduc-
tion for poplar than corn stover. In fine fraction, LSgm(d50/d16) also de-
crease in the same sense, thus it can be associated with two reasons: a
stronger decrease of medium particles than fine or a possible fine in-
crease. To find the cause, it can be assumed the same behavior for d16
than d10 (Section 3.2.2): higher rev implies higher d10 and thus lower
fine for poplar and opposite trend for corn stover. As a result, the LSgm
reduction for poplar is due to median size decrease (d50) as well as in-
crease of d16 (less fine) causing a particle concentration increase on
medium-low region. However for corn stover, the decrease of d50 and
d16 allow establishing a global decrease of particle sizes that is stronger
in mediums than in fine regions.
3.3.2. Mass relative span, uniformity index, uniformity coefficient and
coefficient of gradation
Mass relative span ((MRS = (d90/d84)/d50), uniformity index
(Iu = d95 / d5), uniformity coefficient (Cu = d60/d10), coefficient of gra-
dation (Cg = d230/(d60 − d10)) are dispersion parameters that evaluate
percentiles from high-medium against low regions of the distribution.
Predictive neural network surfaces of these parameters shows al-
ways the same trends and evolutions between them (e.g. MRS, Fig. 6),
as well as the same influences of input variables from ANOVA analysis
and strongest lineal correlations with R2 N 99% from multivariate analy-
sis. The influences of input variables present high similarities with Sgm.
For example, in the case of the variable with the highest influence on
dispersion size (rpm, rev) it is also concluded that higher angular
speed implies lower dispersion and a higher size homogenization on
poplar than on corn stover. In other words, generally the ratios between
high-medium and low regions decreases when rev increases.
Numerically, mass relative span (MRS, Fig. 4) was greater than 1.0,
which indicated a wide distribution, with values around 2 for both bio-
mass. The higher the MRS the higher the heterogeneity in particle sizes.
Coefficient of gradation was between 1 and 1.3, showing a well graded
particle size [10,9]. In Fig. 1, it can be observed that distributions present
a continuum evolution without irregularities or jumps. Cu showed
values from 3.1 to 6 for poplar and from 3.75 to 6.3 for corn stover.
Cu b 4 means a uniform mix of particles, and in the case of Cu = 1, all
particles have the same size.
3.4. Shape distribution parameters
Skewness, Skg, measures the asymmetry of a distribution and occurs
when significant quantities of particles, either in coarser or in finer re-
gions, exit or are removed from the region of predominant size. In all
performed tests, particle size distributions for both resources showed
a ‘fine skewed’ according to higher concentrations on the left side of dis-
tribution, demonstrated in Section 3.3.1 by HSgm N LSgm. Corn stover pre-
sented slightly wider value range (−0.5 to −0.27) than poplar (−0.49
to −0.31). Partly, the asymmetry grows faster on corn stover due to the
finer regions increase at higher rev in comparison to the opposite trend
on poplar (effect analyzed in Section 3.3). In general, the higher the rev
the higher the negative skew due to, not only by the finer region, but
also by the coarser region is removed to lower size region. However,
in contrast with the dispersion parameters, ANOVA analysis allows sim-
ilar significance of the four input variables on final Skg without overall
effect of rev. Moreover, higher fine asymmetries are due to increases
on the operational variables (rev and dtarget) and reductions on hi and
IS (physical conditions of the biomass).
The other parameter related to the shape distribution is the geomet-
ric kurtosis (Kg) that measures the peakness or flatness of the distribu-
tion, relative to the log-normal. Kg presented values from 0.64 to 1.5 for
poplar and from 0.75 to 1.63 for corn stover, therefore they can be clas-
sified like very platykurtic [41]. By definition, the flatness of distribution
represents a higher diversity or heterogeneity on particle size, and
therefore an important relation with dispersion parameters. For this
reasons, ANN predictive response surfaces (Fig. 5) show similar
 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87
Fig. 4. Predictive surfaces of neural network and experimental data points for mass
relative span for poplar and corn stover at different hi, dtarget and rev.
behavior and trends like dispersion parameters but noticeable differ-
ences on the effects of input variables. In fact, dtarget presents again the
highest influences, following by rev or hi. In addition, multivariate anal-
ysis showed also moderate correlation between Kg and distribution pa-
rameters (R2 ≈ 0.73), lower one than agreements obtained between
dispersion parameters themselves (R2 ≈ 0.99).
3.5. Rosin–Rammler function
Rosin and Rammler [35] (definition in Section 2.2) obtained a distri-
bution equation to characterize the coal fineness. Subsequently, it was
observed that it was well suited to skewed distribution of particle sizes
and was used to fit hammer-milled switchgrass, wheat straw and corn
stover [9], as well as knife-milled corn stover [10] and switchgrass [11].
Distribution parameter (mRR) and size parameter (lRR) characterize
Rosin–Rammler distribution. The first one is the slope of the distribu-
tion. Lower mRR means lower slope and therefore a wider distribution
and therefore higher diversity of the particles sizes. The second one
corresponds mathematically to lRR = d63, therefore it is a distribution
percentile. According to the mathematical definition, both parameters
can be enclosed to the two main described groups of the analyzed
distribution parameters: dispersion and size parameters, respectively.
Multivariate analysis confirm the strong correlation (R2 N 99%)
between lRR and size parameters (Section 3.2) and between mRR and
dispersion parameters (Section 3.3). ANOVA analysis shows the same
influences and predictive 3D-NN surfaces (Figs. 6 and 7), following the
same evolutions and trends as function of the four input variables. In
the case of mRR, predictive surfaces present an inverse behavior since
higher slopes (mRR) of cumulative distribution involves lower disper-
sion on particle size, inversely to lower values of dispersion parameters.
85
Fig. 5. Predictive surfaces of neural network and experimental data points for geometric
kurtosis for poplar and corn stover at different hi, dtarget and rev.
Therefore, both RR parameters appear as useful parameters to describe
the central trend and dispersion of the PSD and may be used as input
variables for closed loop control of milling processes.
In order to easily check and visualize the evolution of PSD when
(mainly) opening size and angular speed are varied, Fig. 8 shows the
RR fitting of experimental PSD with extreme values of lRR and mRR
obtained during the experimental campaign.
4. Discussion
The developed characterization of PSD by means of a semi-empirical
tool allows the prediction of the guidelines to control the milling
process in order to adjust the particle size to the requirement of each
final application.
Energy conversion technologies such as combustion, gasification or
lignocellulose hydrolysis in which the residence time and the kinetic
of thermo-chemical conversion govern, require an optimum size to
achieve the total particle conversion. The ideal PSD would be a step
function with the 100% mass percentage for this size. However, particle
fracture by impact produce multiple particles with different morphol-
ogies and the step function is unachievable in practice. In order to obtain
a product that approaches requirements, the opening size of the mill
screen (dtarget) and the angular speed of the rotor (rev) can be con-
trolled. The former (dtarget) determines the mean sizes (Section 3.2.1);
in Fig. 8, PSD shows x-axis displacement of PSD as a result of variation
on dtarget. The latter (rev) mainly affects the dispersion above the
mean size (Section 3.3); the higher the rpm the lower the size disper-
sion, and therefore, the cumulative particle size distribution can be
slightly closer to the ideal step function around the characteristics size
at 3000 rpm. As a conclusion, the opening size of the screen should be
86 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87

Fig. 6. Predictive surfaces of neural network and experimental data points for lRR for poplar
Fig. 7. Predictive surfaces of neural network and experimental data points for mRR for pop-
and corn stover at different hi, dtarget and rev.
lar and corn stover at different hi, dtarget and rev.

from five to eight times higher than the required size (Section 3.2.1) and between 5 to 8 times lower. Therefore, dtarget must be only considered as
the hammermill should be operated at higher rev to decrease the size a reference size imposed in the process, but not as a characteristic size of
dispersion and obtain a PSD focused around the mean size. milled product.
Other technologies such as pelletization of ground materials may Dispersion parameters confirm that comminuted biomass presents
require higher dispersion on particle size since smaller particles refill wide and well graded distribution with a considerable variety and
the holes between the bigger ones and make the cohesion for pelletizing dispersion on particle size, uniform mix of particles, predominance of
easier. If the handling labors are also taken into account, particles with medium-fine particles (fine skew) and very platykurtic distribution. In
similar sizes contribute to improve the relative movement between par-
ticles due to the lack of smaller ones between them that would increase
the frictional effects.
As a conclusion, the operational parameters, dtarget for central trends
and rev for dispersion size, can be adjusted in order to approach the
requirements established by the final energy conversion technology.

5. Conclusions

A set of 54 experimental tests of hammer-milling for SRF poplar and

corn stover were performed in order to determine the influences of the
physical conditions of the biomass (moisture content and chip size) and
the operational variables (opening size of the screen and the angular
velocity of the hammers) on the size of the milled particles. Data post-
processing methodology, combining the neural networks and statistical
analysis, allows to determine the influences of each input variable and
also to obtain a semi-empirical tool to predict the particle size of the
milled product as a function of the process variables.
A total of 14 parameters were analyzed and statistically clustered in
three groups in order to describe different aspects of the particle size
distribution: size central trends, dispersion and shape of the particle
size distribution.
The main average mean size, geometric diameter, was mainly influ- Fig. 8. Example of the effects of dtarget and rev on the cumulative particle size distribution of
enced by the opening size of the internal screen (dtarget), showing values the milled biomass.
 M. Gil, I. Arauzo / Fuel Processing Technology 127 (2014) 80–87
milling process, the angular speed (rev) of rotor governs the variability
and dispersion of the particle size. An increase of rev produces higher
homogenization on particle size for poplar than that for corn stover
due to higher particle concentration around central sizes.
Rosin–Rammler function perfectly fits the cummulative particle size
distribution of comminuted biomass (R2 N 99%) and, in addition, both
Rosin–Rammler parameters (lRR and mRR) represent the main character-
istics of their distributions (central sizes and dispersion).
As a conclusion, a computational tool based on neural networks
was carried out to predict the screen opening size and the rotor angular
speed required to approach the particle size to the requirements
established by the final users. Opening size and angular speed
govern the central trends and the dispersion of PSD, respectively.
Rosin–Rammler parameters (lRR and mRR) appear as useful key parame-
ters indicators of these PSD characteristics and may be used as inputs for
control systems.
Acknowledgments
This work was partially supported by Spanish Ministry of Education,
Culture and Sports (project ENE2008-03358/ALT), funded by the Minis-
try of Education, Culture and Sports, Spain. The authors are grateful to
Mr. Óscar Puyó for his extensive support in the laboratory work and to
the agricultural cooperative of Sariñena for the corn stover supply.
References
[1] E. Commission, Europe 2020, 2010.
[2] S. Paulrud, J.E. Mattsson, C. Nilsson, Particle and handling characteristics of wood
fuel powder: effects of different mills, Fuel Processing Technology 76 (2002) 23–39.
[3] C. Igathinathane, L. Pordesimo, E. Columbus, W. Batchelor, S. Methuku, Shape iden-
tification and particles size distribution from basic shape parameters using ImageJ,
Computers and Electronics in Agriculture 63 (2008) 168–182.
[4] C. Igathinathane, L. Pordesimo, E. Columbus, W. Batchelor, S. Sokhansanj, Sieveless
particle size distribution analysis of particulate materials through computer vision,
Computers and Electronics in Agriculture 66 (2009) 147–158.
[5] Q. Guo, X. Chen, H. Liu, Experimental research on shape and size distribution of
biomass particle, Fuel 94 (2012) 551–555.
[6] M. Gil, E. Teruel, I. Arauzo, Analysis of standard method for milled biomass through
image processing, Fuel 116 (2014) 328–340.
[7] M. Himmel, M. Tucker, J. Baker, C. Rivard, K. Oh, K. Grohmann, Comminution of
biomass: hammer and knife mills, Biotechnology and bioengineering symposium,
vol. 15, Wiley, J, New York, 1986, pp. 39–58.
[8] S. Mani, L.G. Tabil, S. Sokhansanj, Grinding performance and physical properties of
wheat and barley straws, corn stover and switchgrass, Biomass and Bioenergy 27
(2004) 339–352.
[9] V.S.P. Bitra, A.R. Womac, N. Chevanan, P.I. Miu, C. Igathinathane, S. Sokhansanj, D.R.
Smith, Direct mechanical energy measures of hammer mill comminution of switch-
grass, wheat straw, and corn stover and analysis of their particle size distributions,
Powder Technology 193 (2009) 32–45.
[10] V.S.P. Bitra, A.R. Womac, Y.T. Yang, P.I. Miu, C. Igathinathane, S. Sokhansanj, Mathe-
matical model parameters for describing the particle size spectra of knife-milled
corn stover, Biosystems Engineering 104 (2009) 369–383.
[11] V.S.P. Bitra, A.R. Womac, Y.T. Yang, C. Igathinathane, P.I. Miu, N. Chevanan, S.
Sokhansanj, Knife mill operating factors effect on switchgrass particle size distribu-
tions, Bioresource Technology 100 (2009) 5176–5188.
[12] V.S.P. Bitra, A.R. Womac, Y.T. Yang, P.I. Miu, C. Igathinathane, N. Chevanan, S.
Sokhansanj, Characterization of wheat straw particle size distributions as affected
by knife mill operating factors, Biomass and Bioenergy 35 (2011) 3674–3686.
87
[13] M. Gil, I. Arauzo, E. Teruel, Influence of input biomass conditions and operational
parameters on comminution of short-rotation forestry poplar and corn stover
using neural networks, Energy & Fuels 27 (2013) 2649–2659.
[14] S. Van Loo, J. Koppejan, The Handbook of Biomass Combustion and Co-firing,
Earthscan Publications Ltd, London, 2008.
[15] J. Pallarés, A. Gil, C. Cortés, C. Herce, Numerical study of co-firing coal and Cynara
Cardunculus in a 350 MWe utility boiler, Fuel Processing Technology 90 (2009)
1207–1213.
[16] J. Steer, R. Marsh, A. Griffiths, A. Malmgren, G. Riley, Biomass co-firing trials on a
down-fired utility boiler, Energy Conversion and Management 66 (2013) 285–294.
[17] H. Lu, E. Ip, J. Scott, P. Foster, M. Vickers, L.L. Baxter, Effects of particle shape and size
on devolatilization of biomass particle, Fuel 89 (2010) 1156–1168.
[18] D.R. Sudhakar, A.K. Kolar, Experimental investigation of the effect of initial fuel par-
ticle shape, size and bed temperature on devolatilization of single wood particle in a
hot fluidized bed, Journal of Analytical and Applied Pyrolysis 92 (2011) 239–249.
[19] J. Reuther, G. Karsner, S. Jack, Plane flame furnace combustion studies of pulverized
wood, in: 1Fundamentals of Thermochemical Biomass Conversion Conference.
[20] R.J. Gravelsins, Studies of Grinding of Wood and Bark-Wood Mixtures with the
Szego Mill, Ph.D. thesi University of Toronto, 1998.
[21] T.G. Bridgeman, L.I. Darvell, J.M. Jones, P.T. Williams, R. Fahmi, Influence of particle size
on the analytical and chemical properties of two energy crops, Fuel 86 (2007) 60–72.
[22] F.V. Tinaut, A. Melgar, J.F. Pérez, A. Horrillo, Effect of biomass particle size and air
superficial velocity on the gasification process in a downdraft fixed bed gasifier. An ex-
perimental and modelling study, Fuel Processing Technology 89 (2008) 1076–1089.
[23] J.J. Hernández, G. Aranda-almansa, A. Bula, Gasification of biomass wastes in an
entrained flow gasifier: effect of the particle size and the residence time, Fuel
Processing Technology 91 (2010) 681–692.
[24] A. Hendriks, G. Zeeman, Pretreatments to enhance the digestibility of lignocellulosic
biomass, Bioresource Technology 100 (2009) 10–18.
[25] V.B. Agbor, N. Cicek, R. Sparling, A. Berlin, D.B. Levin, Biomass pretreatment:
fundamentals toward application, Biotechnology Advances 29 (2011) 675–685.
[26] V. Chang, B. Burr, M. Holtzapple, Lime pretreatment of switchgrass, Applied
Biochemistry and Biotechnology 63–65 (1997) 3–19.
[27] K.I. Ileleji, B. Zhou, The angle of repose of bulk corn stover particles, Powder Technol-
ogy 187 (2008) 110–118.
[28] M. Gil, I. Arauzo, E. Teruel, C. Bartolome, Milling and handling Cynara Cardunculus L.
for use as solid biofuel: experimental tests, Biomass and Bioenergy 41 (2012)
145–156.
[29] M. Gil, D. Schott, I. Arauzo, E. Teruel, Handling behavior of two milled biomass: SRF
poplar and corn stover, Fuel Processing Technology 112 (2013) 76–85.
[30] CEN/TS 14778–1, Solid biofuels. Sampling. Methods for sampling, European
Committe for Standardization, Brussels, 2005.
[31] CEN/TS 14774–1, Solid biofuels. Methods for the determination of moisture content.
Oven dry method, Part 1: Total moisture. Reference method, European Committe for
Standarization, Brussels, 2004.
[32] CEN/TS 15149–2, Solid biofuels. Methods for the determination of particle size
distribution, Part 2: Vibrating screen method using sieve apertures of 3.15 mm
and below, European Committe for Standarization, Brussels, 2006.
[33] ISO3310-1, Technical requirements and testing — part 1: test sieves of metal wire
cloth, International Organization for Standardization, Geneva, 2000.
[34] R.L. Folk, W.C. Ward, Brazos River bar: a study in the significance of grain size pa-
rameters, Journal of Sedimentary Petrology 7 (1957) 3–26.
[35] P. Rosin, E. Rammler, Laws governing the fineness of powdered coal, Journal of the
Institute of Fuel 7 (1933) 29–36.
[36] C.M. Bishop, Neural Networks for Pattern Recognition, Oxford University Press,
New York, 1995.
[37] M. Hagan, H. Demuth, M. Beale, Neural Network Design, PWS Publishing, 1996.
[38] H. Demuth, M. Beale, M. Hagan, Neural Network Toolbox, For Use with Matlab, The
MathWorks, 2001.
[39] D. Rumelhart, G. Hinton, R. Williams, Learning Internal Representations by Error
Propagation, MIT Press Cambridge, USA, 1986.
[40] [40] M. Gil, I. Ramos, I. Arauzo, J. Román, Characterization of a biomass milling
pilot plant, in: G. F. De Santi (Ed.), 17th European Biomass Conference, Hamburg,
pp. 447–454.
[41] S.J. Blott, K. Pye, GRADISTAT: a grain size distribution and statistics package for the
analysis of unconsolidated sediments, Earth Surface Processes and Landforms 26
(2001) 1237–1248.