Fuel Processing Technology 123 (2014) 122–126

Contents lists available at ScienceDirect

Fuel Processing Technology

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

Clarifying sub-processes in continuous ring die pelletizing through die

temperature control
Markus Segerström, Sylvia H. Larsson ⁎
Swedish University of Agricultural Sciences, Department of Forest Biomaterials and Technology, Division of Biomass Technology and Chemistry, SE-90183 Umeå, Sweden

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

Article history: A pilot scale pelletizer with a custom-made die temperature control system was used for pelletizing of a typical
Received 27 May 2013 Nordic softwood blend in an experimental design where die temperature, moisture content, and steam condi-
Received in revised form 7 February 2014 tioning were varied independently. Steam conditioning, expressed as material temperature, showed a strong
Accepted 7 February 2014
negative correlation with the pelletizer motor current, but had no significant effect on other responses. Die tem-
Available online 6 March 2014
perature was negatively correlated to bulk density and durability. This negative correlation is contradictory to re-
Keywords:
sults from a pilot scale study where die temperature co-varied with other factors, and to results from single
Moisture content pelletizing studies that do not mimic the friction originated pressure build-up that is required for pellet formation
Bulk density in a continuous process.
Durability © 2014 Elsevier B.V. All rights reserved.
Compaction
Biomass
Fiber

1. Introduction scale continuous pelletizing [i.e. 7–12]. Lab scale single pelletizing re-
quires small amounts of raw material and provides a high degree of con-
Biomass has a key role in transforming the world's energy systems trol over process parameters that can be varied independently. Raw
into orbits of renewable sources and energy carriers. The potential of material factors (material type, particle size, moisture content, etc.),
biological materials is huge, and by suitable pre-treatment – i.e. drying, die temperature, and compression force are varied and responses such
milling, and compaction – transportation, handling, storage and end- as pellet density, pellet strength, and friction force when pushing the
use are simplified [1]. Pellets are a standardized refined biomass pellet out of the die, are measured. The abovementioned responses are
product that is traded internationally. Standardized pellet quality pa- interpreted as indirect measures of pellet bulk density, pellet durability,
rameters with a practical effect on handling and storage are described and pelletizing energy consumption. Single pellet studies are cheap,
mainly in the following properties; bulk density, mechanical durability, fast, and controllable, but have their challenge in providing an experi-
and amount of fines. Bulk density should be high to ensure transport mental setup that can produce data with relevance for the continuous
and handling efficiency. The mechanical durability is important as ring die pelletizing process with industrial implication.
pellets should sustain being blown as bulk cargo and transported in In pilot scale continuous ring die pelletizing, process parameters
screw conveyors, lifts and through burners without breaking into cannot be varied independently and thereby the range of settings is lim-
small pieces and dust that can block the supply chain. The amount of ited. Pilot scale operations require an infrastructure that can provide
fines is restricted to be below a certain limit to ensure safe handling, large amounts of raw material, and thus, research is costly and time con-
thereby avoiding dust induced fire, explosion, and health hazards. suming. However, pilot scale pelletizing research has a clear advantage
Further, even though costs in pellet production are dominated by raw in being performed within the actual process relevant for industrial
material, drying, and personnel, pelletizing itself constitutes around applications.
10–15% of the pellet production costs [2] and guidance on how to re- In regular ring die pelletizing, die temperature and compression
duce the energy required in the pelletizing process while sustaining force cannot be controlled but are co-varying with material properties
high pellet quality is welcome to pellet producers. (i.e. moisture content) and process settings (i.e. die channel length).
Pelletizing research is typically performed in two distinctively differ- For each specific setting, the process will stabilize at a specific die
ent environments: i) lab scale single pelletizing [i.e. 3–6] and ii) pilot temperature and motor current (the latter giving an indication of the
applied compression force).
For this study, custom-made equipment for die temperature control
⁎ Corresponding author. Tel.: +46 90 786 87 90. in a ring die pelletizer was used to overcome the co-variation of die tem-
E-mail address: sylvia.larsson@slu.se (S.H. Larsson). perature with other process parameters. The setup had been used in

http://dx.doi.org/10.1016/j.fuproc.2014.02.008
0378-3820/© 2014 Elsevier B.V. All rights reserved.
 M. Segerström, S.H. Larsson / Fuel Processing Technology 123 (2014) 122–126 123

two previous studies, showing that die temperature control is efficient 2.3. Experimental design
for overcoming discontinuous production in straw material pelletizing
[9] and that pellet quality can be improved by manipulating the die tem- Controllable factors for the experimental design were; die tempera-
perature [7]. In the present study, emphasis was put on using the benefit ture (°C), raw material moisture content (%, w.b.), and material temper-
of die temperature control for evaluation of the true effects of moisture ature (°C). By using raw material temperature as a measure for steam
content, steam conditioning, and die temperature when die tempera- conditioning, continuous logging of the parameter was made possible,
ture was varied independently from other factors. and achieved factor settings could be followed up.
Further, the setup gave unique opportunities for gaining a deeper The experimental design chosen was a 2 × 2 × 2 factorial design.
understanding of the effects of steam conditioning in fuel pelletizing. Fixed levels were chosen according to the following; die temperature
Through steam conditioning, both heat and moisture are added to the at 40 °C—representing a low level only achievable with die cooling,
raw material and pre-conditioning of raw material by the use of steam and 100 °C—representing a medium high biofuel pelletizing die tem-
is widely used in fuel and feed pelletizing [13]. Steam conditioning is perature, raw material moisture content at 9.5 and 12.5% w.b. which is
in the fuel pelletizing process expressed in at least three vectors; mate- a typical moisture range for softwood pelletizing, and 23 °C and 55 °C
rial temperature, die temperature, and moisture content. The custom- for material temperature (equaling no steam conditioning and approx-
made experimental setup made it possible to isolate the material tem- imately 6 kg/h, which is a medium high steam conditioning level for
perature increment from the other factors, and to further investigate softwood pelletizing). A center point for die temperature and moisture
the impact of that vector. content was added at each of the low and high steam conditioning
For this study, a typical softwood blend was pelletized in an experi- levels. Thus, the total number of runs was 2 × 2 × 2 + 2 = 10. Because
mental design where die temperature, moisture content, and steam of the nature of the raw material, exact level settings were not achieved,
conditioning were varied independently. The aim was to 1) model the and thus, factors could not be coded at −1 and 1. Instead, range scaling
non-confounded effects of moisture content, steam conditioning, and [14] was used. Studied responses were typical pellet quality responses;
die temperature on pellet quality and energy consumption in continu- pellet bulk density (kg/m3), pellet durability (%), and amount of fines
ous ring die pelletizing and 2) make comparisons with results from (%). Pelletizer motor current (A) was used as a measure for energy con-
lab scale single pelletizing studies and with previous continuous ring sumption. Pellet temperature (°C) was also measured and modeled.
die pelletizing studies where die temperature co-varied with other
factors. 2.4. Data collection; measurements of factors and responses

Each experimental setting (run) was divided into three, two minute
2. Material and methods long, measurement periods (e.g. M1:1, M1:2, M1:3) when data and ma-
terials were collected. During two minutes, a sample amount of approx-
2.1. Material imately 6 kg was produced, which was just enough for further pellet
quality analyses, and the data sampling period was long enough to
A sawdust mixture consisting of 50% Norway spruce (Picea abies L. H. achieve representative data. For modeling, mean values of the data
Karst) and 50% Scots pine (Pinus sylvestris L.) was delivered at 10% wet gathered at the three measurement periods were calculated to repre-
base (w.b.) moisture content from the Neova pellet mill, Främlingshem, sent each of the 10 runs. Achieved settings for the experimental design
Valbo, Sweden to the Biofuel Technology Centre, Swedish University of and corresponding measured values are shown in Table 1.
Agricultural Sciences, Umeå, Sweden. The sawdust was hammer milled Just before each measurement period, the milled raw material was
(Vertica Hammer Mill DFZK-1, Bühler AG, Uzwil, Switzerland) with a sampled before and after the steam conditioner and hot pellets just as
screen size of 4 mm. Moisture content after milling was around 9% they came out of the die. Samples were immediately sealed in plastic
w.b. For each point in the experimental design, approximately 150 kg bags and kept for moisture analysis. Sampling of the milled raw material
of material was prepared. Moisture contents were adjusted according after steam conditioning was difficult due to steam leakage and conden-
to the experimental design by adding water during mixing in a screw sation of steam on the sampling equipment. Hence, values for raw ma-
blender, and prepared materials were left overnight to reach moisture terial moisture content after steam conditioning should be treated with
equilibrium. caution. Pellets produced during each two-minute measurement period
were collected in 20 L open plastic trays and weighed to determine the
production rate, then left overnight to cool down where after samples
2.2. Experimental setup were sealed in plastic bags until further analysis (cool pellet moisture
content and pellet quality analyses).
Pelletizing experiments were performed using an SPC PP300 Die temperature and pelletizer motor current were measured and
Compact pelletizer (Sweden Power Chippers, Borås, Sweden), with a logged continuously (1 Hz) as described in [9]. Due to the die being sta-
maximum capacity of 300 kg/h. The pelletizer is equipped with a tionary, and the pelletizer not running at maximum production capaci-
steam generator (1 bar, 120 °C), and a cascade mixer for mixing steam ty, pellets were not produced in all sections of the die. Pellet production
homogenously into the material flow at ambient pressure. The die is sta- was concentrated to the left side of the die, and thus, die temperature
tionary with drive on the roll suspension instead of the die. The experi- measured at the left side was used for modeling. Pellet temperature
mental setup with die temperature control and measurement systems was measured throughout each measurement period with a hand
is described in detail by Larsson et al. [7]. Two circumferential slits held IR sensor (Optris CT laser 75:1, Optris GmbH, Berlin, Germany) di-
(width: 12 mm, depth: 12 mm) are cut out at both sides of die channel rected towards the pellet surface at the outlet of a die channel. The IR
rows in the die, into which copper coils (outer diameter: 12 mm, inner sensor was directed towards one specific pellet channel outlet until a
diameter 10 mm) for cooling and heating media are inserted. During stable pellet temperature value was obtained. The IR sensor was cali-
cooling, water was used as the medium and circulated through a large brated according to the manual and special caution was taken to avoid
tub where ice and snow were added continuously to keep a tempera- measurement disturbing vapor formation on the sensor. Die tempera-
ture of 0 °C. During heating, hydraulic oil was used as the medium ture, pelletizer motor current, and pellet temperature (when applied)
and the oil was heated with an electrical heater to a maximum temper- were logged continuously (1 Hz) with a data logger (PC-logger 3100i,
ature of 155 °C. Throughout the experimental series, a die channel Intab Interface-teknik AB, Stenkullen, Sweden). Moisture content anal-
length/width of 52.5/8 mm was used, and the pellet production rate yses for determining raw material moisture content before and after
was adjusted towards a set point of 180 kg/h. steam treatment, hot pellet moisture content, and cool pellet moisture
124 M. Segerström, S.H. Larsson / Fuel Processing Technology 123 (2014) 122–126

content, were performed according to the CEN standard [15] by drying

production
rate (kg/h)

180 (10.6)

173 (13.4)
180 (7.3)
184 (4.1)
186 (5.9)
182 (1.7)

179 (7.5)
182 (4.8)

181 (6.3)

180 (4.3)
the samples overnight at 105 °C in a drying cabinet.

Pellet
The amount of fines produced in the pelletizing process was quanti-
fied by manual sieving of pellet samples through a 3.15 mm sieve and
calculated as the percentage of the loss of the fine material to the total

content (%)
Cool pellet

9.4 (0.3)
8.6 (0.2)
9.4 (0.1)
6.5 (0.1)
7.5 (0.2)
10.2 (0.1)
12.1 (0.1)

7.5 (0.2)
8.1 (0.0)
9.2 (0.1)
sample weight.

moisture
Pellet bulk density was determined according to the CEN standard
[16] using a 5.4 L cylindrical bucket. Mechanical durability of pellets
was measured by use of a pellet tester (Q-tester, Simon Heesen BV,
content (%) after steam

Netherlands) according to the CEN standard [17] where 500 g of pellets

Raw material moisture

is tumbled for 10 min. After treatment, the sample is sieved using a

3.15 mm sieve and the weight percentage of the original pellet sample
Factors and responses not used for modeling

conditioning

that does not pass the sieve is used as a measure of mechanical

12.7 (0.2)
12.8 (0.1)
14.8 (0.1)
9.5 (0.0)

12.5 (0.2)
15.9 (0.2)

9.4 (0.1)
11.4 (0.1)

11.2 (0.1)
13.5 (0.1)
durability.

2.5. Data evaluation and modeling

content (%) before steam
Raw material moisture

For each response, a multiple linear regression (MLR) model was

created from range scaled x-variables, making it possible to directly
compare regression coefficients from different factors. Modeling and
conditioning

statistical evaluation were done with the software MODDE 9.1 [18].
9.4 (0.1)
12.8 (0.1)
12.8 (0.1)
9.5 (0.0)

12.5 (0.2)
12.5 (0.2)

9.4 (0.1)
9.4 (0.1)

11.2 (0.1)
11.2 (0.1)

By using hot pellet moisture content as a factor in the modeling,

moisture added in steam conditioning was included in the moisture
content factor. Hence, steam conditioning was represented solely as a
change in the raw material temperature. Other values for moisture con-
0.79 (0.2)
1.34 (0.0)
2.12 (0.6)
0.74 (0.5)

0.80 (0.2)
1.24 (0.2)

0.78 (0.2)
1.00 (0.2)

0.92 (0.3)
0.80 (0.2)
Fines (%)
Achieved factors and responses used and not used for modeling in the experimental design. Standard deviations within brackets, n = 3 for each value.

tent (milled raw material, milled raw material after steam conditioner,
or cool pellets) were also tested, but the best (highest Q2 values) and
simplest (lowest number of factors) models could be created when
hot pellet moisture content was used as the moisture content factor.
temperature

All measurement values were scaled and centered before evaluation.

86 (1.7)
100 (2.6)
103 (3.0)
97 (0.9)

79 (2.3)
76 (1.2)
104 (0.5)

88 (1.8)
101 (1.8)
99 (0.7)

Leave-one-out cross-validation was used to calculate the residual for

Pellet

(°C)

each validation round. The different multivariate models were evaluat-

ed using the coefficient of multiple determinations (Q2) according to
current (A)

Myers [19]. The Q2 value expresses how much of the variance in the re-
28.1 (0.6)
25.3 (0.1)

26.5 (0.3)

26.5 (0.2)
29.8 (0.3)
26.8 (0.6)
28.8 (0.1)
25.9 (0.5)

29.2 (0.5)
29.3 (0.6)
Pelletizer

sponse variable can be predicted and can at best be 1 and a value of 0 in-
motor

dicates no predictive capability at all. The Q2 value is thus similar to the

coefficient of determination (R2). The difference is that while R2 contin-
Responses used for modeling

uously increases with the number of terms, Q2 is improved by inclusion

93.1 (0.4)
89.9 (0.4)
94.3 (0.3)
96.0 (0.2)

96.9 (0.0)

95.5 (0.2)
96.9 (0.2)
94.7 (0.4)

96.3 (0.2)
95.5 (0.2)
durability

of terms until the model becomes over-fitted to the data in the leave-
Pellet

one-out cross-validation models, and then the predictions of the left

(%)

out observations are worsened. The number of factors used in the

models was determined by optimization of Q2 according to the follow-
Pellet bulk

640 (5.7)
611 (2.6)

647 (2.8)
706 (7.9)
694 (5.7)
677 (4.7)
611 (0.0)
673 (7.0)
717 (3.7)
682 (7.0)
(kg/m3)

ing procedure: For each response, a model was created from all factors
density

and interactions of die temperature (°C), raw material moisture content

(% w.b.), and raw material temperature (°C). Non-significant terms
were then deleted one by one from the model until a maximum value
content (%)
Hot pellet

10.8 (0.3)
12.6 (0.1)
7.9 (0.1)
9.1 (0.2)
12.1 (0.0)
14.6 (0.1)
11.1 (0.2)
8.8 (0.2)
9.7 (0.0)
11.4 (0.1)

for Q2 was obtained. Modeled responses were: pellet bulk density

moisture

(kg/m3), pellet durability (%), fines (%), pelletizer motor current (A),
and pellet temperature (°C).
Raw material temperature

2.6. Results and discussion

Achieved factor settings and measured responses for the 10 experi-

(°C) after steam

mental settings are shown in Table 1. Die temperatures could be well

conditioning

23.6 (0.5)
55.6 (0.8)
19.7 (0.1)
53.9 (2.5)
22.5 (0.3)
59.8 (2.6)
63.8 (0.8)
30.4 (1.9)
21.3 (0.3)
54.6 (2.1)

controlled at three levels. Steam regulation and hot pellet moisture con-
Factors used for modeling

tents were less exact—raw material temperature when steam was

added varied from 54 to 64 °C and the moisture content for the observa-
tion M6 was approximately 2%-units higher than planned. The pellet
temperature

production rate was held constant at 180 ± 7 kg/h.

100 (2.2)
106 (0.9)
94 (2.4)
105 (2.0)
34 (0.4)
37 (0.7)
39 (0.7)
39 (0.2)
73 (4.1)
82 (1.5)

Response measurements for pellet bulk density varied from 611 to

717 kg/m3. Pelletizer motor current varied from 25.3 to 29.8 A. Out of
(°C)
Die

this, 17 A was required for idle running. Thus, the specific pelletizer cur-
rent varied from 8.3 to 12.8 A. Pellet durability varied from 89.9 to 96.9%
Name
Table 1

and fines varied within a range of 0.74–2.1%. The pellet temperature

M10
M1
M2
M3
M4
M5
M6
M7
M8
M9

was in between 76 and 104 °C.

 M. Segerström, S.H. Larsson / Fuel Processing Technology 123 (2014) 122–126 125

Table 2
Summary of regression model statistics.

Model components Pellet bulk density (kg/m3) Pellet durability (%) Fines (%) Pelletizer motor current (A) Pellet temperature (°C)

Number of observations 10 10 10 10 10
R2 0.999 0.922 0.606 0.996 0.912
Q2 0.979 0.672 0.067 0.988 0.774
p-value of model 0.00 0.02 0.04 0.00 0.00
F-value [degrees of freedom] 551 [3] 9.50 [4] 5.38 [7] 299 [5] 20.7 [6]

Coefficients: Range scaled [unscaled] (p-value of coefficient)

Constant 665 [759] (0.00) 95.6 [74.2] (0.00) 1.08 [−1.038] (0.00) 27.5 [32.3] (0.00) 92.9 [110] (0.00)
Moisture content (Moi) −63.5 [−4.80] (0.00) −2.26 [1.90] (0.02) 0.42 [0.130] (0.03) −1.11 [−0.143] (0.00) −2.46 [−3.06] (0.26)
Material temperature (Ste) 7.35 [−1.69] (0.01) 0.498 [0.501] (0.34) n.s.a −1.28 [−0.0580] (0.00) n.s.
Die temperature (Die) −9.10 [0.743] (0.00) −1.80 [0.133] (0.01) 0.245 [0.00684] (0.08) −0.328 [0.0270] (0.00) 10.8 [−0.112] (0.00)
Moi × Ste −11.8 [−0.165] (0.01) −2.71 [−0.0380] (0.03) n.s. n.s. n.s.
Moi × Die −13.1 [−0.113] (0.00) −1.69 [−0.0145] (0.05) n.s. −0.333 [−0.00287] (0.01) 3.82 [0.0329] (0.16)
Ste × Die 8.16 [0.0103] (0.01) n.s. n.s. n.s. n.s.
a
Non-significant.

Multilinear regression (MLR) models were created and model through the die at the high friction created by low moisture contents.
validities were evaluated (Table 2). Excellent models with high R2 and Single pelletizing studies where compression force was varied indepen-
Q2 values could be created for pellet bulk density, pelletizer motor cur- dently from moisture content showed i) negative correlations (within
rent, and pellet temperature. The model for pellet durability had excel- the range of the designs) between pellet density and moisture content
lent explanatory ability (R2) and fair predictive ability (Q2). The amount [3,6] and that ii) pellet density was positively correlated to compression
of fines could not be modeled satisfactorily, probably due to a low vari- force [4,6].
ation in the response, and the model for fines was not considered for For the softwood mix used in this study, maximum durability was
further discussion. Pellet temperature could be well modeled using found at the low end of the moisture range. In single pelletizing of
only two factors: die temperature (positive correlation) and moisture softwood, pellet hardness was found to be negatively correlated with
content (negative correlation). moisture content [3,5,6]. Durability is a complex response, where an
In this study, confounding effects regarding die temperature were optimum moisture content for maximum durability is believed to
eliminated through die temperature control but compression force exist for every single material [11].
still co-varied with other factors. Compression force confounding can The negative effect from die temperature on pellet bulk density and
be reduced through a careful adaption of different channel lengths but durability can be due to a decrease in wall friction and compression
since we had only one custom-made die that was not possible to do. force when raw material in contact with the die is softened. Hence,
Under the assumption that there exists a linear relationship between die temperature and compression force are confounded and low values
compression force and pelletizer current, relative effects of moisture for bulk density could be due to low compression forces. Contrary to the
content, die temperature, and material temperature on compression results from this study, single pelletizing studies show positive correla-
force can be predicted by the MLR-model for pelletizer current (equa- tions for die temperature with density [3,4,6] and compressive strength
tion using unscaled coefficients in Table 2). Such an analysis shows [3–6]. Also, in one of our previous pilot scale pelletizing studies [10]
that the same decrement in pelletizer motor current/compression where die temperature was not varied independently, die temperature
force is obtained through any of the following actions: increasing the was found to have a positive correlation with both bulk density and
moisture content with 1%-unit, increasing the die temperature with durability. When now looking further into the results, we found that
22 °C, or increasing the material temperature with 6 °C. the positive correlations were due to a negative co-variation of moisture
Major effects, both negatively correlated for both pellet bulk density content and die temperature—high die temperatures were found when
and durability, were moisture content and die temperature (Fig. 1). A low moisture, high frictional, materials were pelletized. Increased die
negative correlation between bulk density and moisture content is in temperatures probably decreased wall friction, but that effect was less
line with previous results from pilot or industrial scale softwood pellet- potent than that of the low moisture.
izing [11,12,20]. However, moisture content and compression force fac- Temperature correlations found in single pelletizing and in ring die
tors are confounded in ring die pelletizing. The improved pellet density studies with confounded die temperatures have led to conclusions
could be due both to better bonding properties at lower moisture con- that, by increasing the die temperature in continuous ring die pelletiz-
tents and to the higher compression forces required to push material ing, pellet quality is improved and energy consumption is reduced.
Pelletizer motor current (A)
Pellet bulk density (kg/m3)

0
Pellet durability (%)

0
0 -1

-50 -2
-5
-100 -3

-10
-4
Moi
Moi*Die
Moi*Ste
Die
Ste*Die
Ste

Moi*Ste

Moi

Die

Moi*Die

Ste

Ste

Moi

Moi*Die

Die

Fig. 1. Effect plots for pellet bulk density (kg/m3), pellet durability (%), and pelletizer motor current (A). Plots show the effect on the response when changing factors from their lowest to
their highest value within the range of the design while other factors are held constant at their average value. Error bars show 95% confidence intervals.
126 M. Segerström, S.H. Larsson / Fuel Processing Technology 123 (2014) 122–126

However, this is a serious misinterpretation. In batch-wise single pellet- supporting this work. Neova is acknowledged for supplying the raw
izing, where pellets are compressed by a piston against a closed end, material. Pelletizing studies were performed at the Biofuel Technology
temperature induced friction reduction will not influence pellet quality Centre, Umeå, Sweden.
responses, but in continuous ring die pelletizing where friction builds
up the pressure that forms the pellets, increasing the die temperature References
will decrease the compression force and give an effect of lower bulk
[1] C.N. Hamelinck, R.A.A. Suurs, A.P.C. Faaij, International bioenergy transport costs and
density and durability, as shown in Fig. 1. energy balance, Biomass and Bioenergy 29 (2005) 114–134.
Steam conditioning, expressed and material temperature, was the [2] G. Thek, I. Obernberger, Wood pellet production costs under Austrian and in
most influential factor for pelletizer motor current but had little or insig- comparison to Swedish framework conditions, Biomass and Bioenergy 27 (2004)
671–693.
nificant effect on other responses (Fig. 1). In pilot scale fuel pelletizing [3] M. Finell, M. Arshadi, R. Gref, T. Scherzer, W. Knolle, T. Lestander, Laboratory-scale
studies, where steam addition co-varied with die temperature and raw production of biofuel pellets from electron beam treated Scots pine (Pinus silvestris
material moisture content, steam addition showed negative correlations L.) sawdust, Radiation Physics and Chemistry 78 (2009) 281–287.
[4] P. Gilbert, C. Ryu, V. Sharifi, J. Swithenbank, Effect of process parameters on
with pelletizer motor current [11] and bulk density [10–12] and ambigu- pelletisation of herbaceous crops, Fuel 88 (2009) 1491–1497.
ous or positive correlations with durability [10–12]. Varying die and ma- [5] N.P.K. Nielsen, D.J. Gardner, T. Poulsen, C. Felby, Importance of temperature,
terial temperatures independently show that there is a potential to moisture content, and species for the conversion process of wood residues into
fuel pellets, Wood and Fiber Science 41 (2009) 414–425.
reduce particle-to-particle friction and particle stiffness by increasing
[6] C. Rhen, R. Gref, M. Sjöström, I. Wästerlund, Effects of raw material moisture con-
the material temperature and to thereby reduce energy consumption. tent, densification pressure and temperature on some properties of Norway spruce
However, to maintain high pellet quality, sufficient compression force pellets, Fuel Processing Technology 87 (2005) 11–16.
has to be obtained and this can be done by keeping a low die temperature. [7] S.H. Larsson, M. Rudolfsson, Temperature control in energy grass pellet production—
effects on process stability and pellet quality, Applied Energy 97 (2012) 24–29.
Through die temperature control, we have been able to separate [8] S.H. Larsson, M. Rudolfsson, M. Nordwaeger, I. Olofsson, R. Samuelsson, Effects of
several sub-processes in ring die pelletizing. Still, to reach a fully moisture content, torrefaction temperature, and die temperature in pilot scale pel-
non-confounded experimental setup, studies need to be performed letizing of torrefied Norway spruce, Applied Energy 102 (2013) 827–832.
[9] S.H. Larsson, M. Rudolfsson, M. Thyrel, H. Örberg, G. Kalén, M. Wallin, T.A. Lestander,
using temperature controlled dies with different channel lengths pro- Temperature controlled feed layer formation in biofuel pellet production, Fuel 94
viding the ability to vary both die temperature and compression force (2012) 81–85.
independently. [10] S.H. Larsson, M. Thyrel, P. Geladi, T.A. Lestander, High quality biofuel pellet produc-
tion from pre-compacted low density raw materials, Bioresource Technology 99
(2008) 7176–7182.
[11] R. Samuelsson, S.H. Larsson, M. Thyrel, T.A. Lestander, Moisture content and storage
3. Conclusions
time influence the binding mechanisms in biofuel pellets, Applied Energy 99 (2012)
109–115.
By controlling the die temperature and using pellet moisture content [12] R. Samuelsson, M. Thyrel, M. Sjöström, T.A. Lestander, Effect of biomaterial characteris-
in the modeling, the effects of steam conditioning could be separated tics on pelletizing properties and biofuel pellet quality, Fuel Processing Technology 90
(2009) 1129–1134.
from the effects of moisture content and die temperature. Steam condi- [13] M. Thomas, D.J. van Zuilichem, A.F.B. van der Poel, Physical quality of pelleted
tioning, expressed as ingoing material temperature, showed strong nega- animal feed. 2. Contribution of processes and its conditions, Animal Feed Science
tive correlation with pelletizer motor current, but had no significant effect and Technology 64 (1997) 173–192.
[14] R.H. Myers, D.C. Montgomery, Response surface methodology: process and product
on other responses. Die temperature was negatively correlated to bulk optimization using designed experiments, Second ed. John Wiley & Sons, Inc., USA,
density and durability. This is contradictory to a previous pilot scale 2002.
study where die temperature co-varied with other factors, and to single [15] EN 14774–2, Solid biofuels—methods for determination of moisture content—oven
dry method, Total moisture: simplified method, 2009.
pelletizing studies that do not mimic a friction originated pressure [16] EN 15103, Solid biofuels—determination of bulk density, 2009.
build-up for pellet formation that is required in a continuous process. [17] EN 15210–1, Solid biofuels—determination of mechanical durability of pellets and
briquettes, Pellets, 2009.
[18] Umetrics, MODDE, 9.1.0.0 ed, 2011. [Umeå, Sweden].
Acknowledgements [19] R.H. Myers, Classical and Modern Regression with Applications, Duxbury Press,
Boston, USA, 1986.
[20] M. Arshadi, R. Gref, P. Geladi, S.-A. Dahlqvist, T. Lestander, The influence of raw
We thank Bio4Energy, a strategic research environment appointed material characteristics on the industrial pelletizing process and pellet quality,
by the Swedish government, and the Swedish Energy Agency for Fuel Processing Technology 89 (2008) 1442–1447.