IMPORTANCE OF TEMPERATURE, MOISTURE CONTENT, AND SPECIES FOR THE CONVERSION PROCESS OF WOOD RESIDUES INTO FUEL PELLETS Niels Peter K. Nielsen* Research Scientist Chemical Engineering DONG Energy Power A/S Copenhagen, Denmark Douglas J. Gardner{ Professor Advanced Engineered Wood Composites Center University of Maine Orono, ME 04469 Torben Poulsen Pelletizing Specialist TP Pellets Consulting ApS Præstevænget 6 DK-7480 Vildbjerg, Denmark Claus Felby Professor Forest and Landscape University of Copenhagen Faculty of Life Sciences Copenhagen, Denmark (Received February 2009) Abstract. In wood pellet production, knowledge is needed about the raw material properties that affect the energy requirements for pelletizing and pellet quality. This study presents novel methods for this purpose, including analyses of influence of the raw material properties on the energy requirements in the sequence of subprocesses (compression, flow, and friction components) that constitute the pelletizing process, and the strength of the pellets. The methods were used to analyze the importance of pelletizing temperature and MC and the differences between sawdust of European beech (Fagussylva´tica L) and Scots pine (Pinus sylvestris L). Results showed that increasing temperature and MC decreased the energy requirements for all components of the pelletizing process and that beech required more energy than pine in all components. Beech produced the stronger pellets; increasing temperature resulted in stronger pellets, whereas increasing MC caused weaker pellets. Also, a method to quantify the energy require- ments for the combined pelletizing process is presented. The methods can be used to analyze the allocation of the energy requirements of pelletizing in the die and can be useful tools for analyzing the pelletizing properties of wood and other biomass residues. Keywords: Wood pellets, pelletizing, biofuel, compression, friction, viscosity, adhesion, extractives, densification, sawdust. INTRODUCTION The raw materials for wood pellet production in- clude wood residues (primarily sawdust) from pri- mary and secondary wood processing industries. * Corresponding author: [email protected]{ SWST member Wood and Fiber Science, 41(4), 2009, pp. 414–425 # 2009 by the Society of Wood Science and Technology
12
Embed
IMPORTANCE OF TEMPERATURE, MOISTURE …...IMPORTANCE OF TEMPERATURE, MOISTURE CONTENT, AND SPECIES FOR THE CONVERSION PROCESS OF WOOD RESIDUES INTO FUEL PELLETS Niels Peter K. Nielsen*
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IMPORTANCE OF TEMPERATURE, MOISTURE CONTENT,
AND SPECIES FOR THE CONVERSION PROCESS
OF WOOD RESIDUES INTO FUEL PELLETS
Niels Peter K. Nielsen*Research Scientist
Chemical Engineering
DONG Energy Power A/S
Copenhagen, Denmark
Douglas J. Gardner{Professor
Advanced Engineered Wood Composites Center
University of Maine
Orono, ME 04469
Torben PoulsenPelletizing Specialist
TP Pellets Consulting ApS
Præstevænget 6
DK-7480 Vildbjerg, Denmark
Claus FelbyProfessor
Forest and Landscape
University of Copenhagen
Faculty of Life Sciences
Copenhagen, Denmark
(Received February 2009)
Abstract. In wood pellet production, knowledge is needed about the raw material properties that affect
the energy requirements for pelletizing and pellet quality. This study presents novel methods for this
purpose, including analyses of influence of the raw material properties on the energy requirements in the
sequence of subprocesses (compression, flow, and friction components) that constitute the pelletizing
process, and the strength of the pellets. The methods were used to analyze the importance of pelletizing
temperature and MC and the differences between sawdust of European beech (Fagus sylvatica L) and
Scots pine (Pinus sylvestris L). Results showed that increasing temperature and MC decreased the energy
requirements for all components of the pelletizing process and that beech required more energy than pine
in all components. Beech produced the stronger pellets; increasing temperature resulted in stronger
pellets, whereas increasing MC caused weaker pellets. Also, a method to quantify the energy require-
ments for the combined pelletizing process is presented. The methods can be used to analyze the
allocation of the energy requirements of pelletizing in the die and can be useful tools for analyzing the
pelletizing properties of wood and other biomass residues.
Wood and Fiber Science, 41(4), 2009, pp. 414–425# 2009 by the Society of Wood Science and Technology
Cylindrical wood pellets have a 6 – 8 mm dia andare 5 – 40 mm in length and are used as fuel inresidential heating and centralized heating andpower production. In 2007, the worldwide marketwas approximately 8 million tonnes, which isexpected to increase to 15 million tonnes by 2010(Ljungblom 2007; Vinterback 2008). The rawmaterials can be characterized by the anatomicaland chemical properties of the wood species,temperature, MC, preparation method, storage,and particle size distribution. These parametersillustrate the heterogeneous character of theraw materials coming from different suppliersand types of wood industries. These variationscause significant differences in the costs of pelle-tizing when different types of raw materials areused. The different costs are related to variationsin the energy requirements to pelletize theraw materials because the pellet mill’s capacity(t/h) is limited by the specific maximum powerconsumption of the pelletizing equipment. There-
fore, raw materials that require more energyto pelletize lead to a lower production capacityalong with higher maintenance costs per tonne ofpellets. To minimize the cost of production and tooptimize the use of new raw materials for thegrowing wood pellet market, better knowledge isneeded on how the raw material’s individual phys-ical and chemical properties affect the energy re-quirement of the pelletizing process. The strengththat the pellets obtain is an important quality pa-rameter, and this also varies among raw materials.However, rapid and consistent methods are neededto analyze the raw material’s effect on the energyrequirements for pelletizing and pellet strength toincrease the understanding of the connection be-tween the wood properties and these factors in theprocess.
Figure 1 illustrates the pelletizing process,which essentially works by extruding (e.g., saw-dust) through cylindrical press channels, which
Figure 1. Illustration of the pellet mill and a press channel.
Nielsen et al—ANALYSIS METHODS FOR PELLETIZING PROPERTIES OF WOOD 415
are 6- to 8-mm-dia holes drilled through a ring-shaped steel die. The figure (right) illustrates apress-channel design, in which the first part ofthe channel has a cone-shaped opening that canbe 2.5 mm deep and with 60� angles and thetotal length of the active part of the channelcan be 30 – 70 mm. The die (left) is typically a150 – 250-mm-long steel cylinder with an innerdiameter of 800 mm and outer diameter of 1 m.The press channels are arranged radially inclose proximity in the die (see Fig 2). The presschannel openings are on the die’s inner surfaceand the channel exits on the outer surface. Thenumber of channels in one die can be severalthousands. Some thickness of the die cylinder isrequired for strength; therefore, the press chan-nel also contains an inactive (downstream) partthat does not contribute to pelletizing. Becauseof the distance between the channel openings onthe die’s inner surface, an area is present be-tween the openings that may take up more thanone-half of the die area depending on the presschannel design.
Two to three bearing-mounted stationary rollsare located on the die’s inner surface, and by
rotation of the die, the rolls thus run on the presschannel band. Sawdust is continuously added infront of the rolls, and the rotation of the dieforces the sawdust into and through the chan-nels. Because of the force required to press thesawdust into the channels and the friction, highpressure is built up under the rolls, which com-presses and bonds the sawdust. Each time achannel opening is passed by a roll, a new layeris pressed into the channel and the downwardmovement in the channels therefore occurs indiscrete steps. The process causes high-density(1200 – 1300 kg/m3) pellets to be pressed outand break off from the press channel exits. Thepelletizing process generates heat that maintainsthe temperature of the operating die at 110 –130�C (unpublished data). The rotation of thedie is the energy-requiring part of the process,whereas the rolls run passively on the presschannel band. In the present work, the specificenergy required to rotate the die for a givenraw material is termed the specific work of pel-letizing, Wspec.
Figure 2 schematically illustrates the pelletizingprocess with a section of a channel row andshows that the process can be separated intocomponent sequences of compression, flow,and friction of e.g., sawdust. Therefore, Wspec
can be seen as the sum of the individual energyrequirements for these components and there-fore does not quantify or apportion the energyrequired for each component. The componentsmay be differently affected by sawdust proper-ties and therefore, to characterize where andhow much the raw material properties causevariations in Wspec, the properties must be ana-lyzed for each component.
The pelletizing process can further be describedas the ratio between the force applied by the roll(Froll) and the force required for downwardmovement of the compressed sawdust in thechannels (Fdie). When the roll approaches thedie surface and Fdie is highest, the sawdust tem-porarily forms a compressed layer on the diesurface. When Froll exceeds Fdie, the material inthe press channel is forced to move downward,and the sawdust in the compressed layer is
Figure 2. Illustration of the pelletizing process. A section
of a press channel row is used to illustrate the die/roll/
sawdust system. The components (see text) of the pelletiz-
ing process are allocated to the positions marked by the
white dashed lines. The lower part of the press channels
with larger diameter is not part of the pelletizing.
416 WOOD AND FIBER SCIENCE, OCTOBER 2009, V. 41(4)
distributed into the press channel openings. Thisprocess occurs under pressures ranging 210 –450 MPa (Leaver 2000), which was confirmedin the present study, and some of the material inthe compressed layer is located over the hori-zontal area in between the press-channel open-ings. This part of the layer must therefore bemoved horizontally to enter a channel opening.Thus, Fdie at this point is dependent on thestrength and viscosity of the highly compressedlayer because it must be separated to flow hori-zontally and to flow into the channel openings,and Fdie is also dependent on the friction be-tween the compressed sawdust and the presschannel walls.
In the present work as illustrated in Fig 2, thecompression component represents the energyto compress the sawdust (Wcomp), the flow com-ponent represents the energy required to forcethe compressed layer into the press channels(Wflow), and the friction component representsthe energy required to push the compressedsawdust in the channels (Wfric). These compo-nents are used to characterize pelletizing differ-ences between samples. The pressure andtemperature in the die cause the raw materialparticles to bond together probably by autoad-hesion processes (Back 1987) as no adhesivesare used in conventional pelletizing. The sur-face-to-surface bonding of the particles maycontribute significantly to the strength of thepellets, and the extent of the bonding may beaffected by extractives (Nielsen et al 2009a) andthe MC.
This study presents laboratory procedures formeasuring the raw material effects on the com-ponents combined with an analysis of the im-portance of die temperature and MC of thewood for pine and beech sawdust. Also, a meth-od to measure Wspec is presented.
MATERIALS AND METHODS
Materials
Sawdust from European beech (Fagus sylvatica L)and Scots pine (Pinus sylvestris L) was prepared by
chainsaw cutting across the grain producing saw-dust with longitudinal fiber orientation (Nielsenet al 2009b) of dried boards bought at a local lum-beryard. The chainsaw was operated without chainoil. The 1 – 4 mm particle size fraction was isolatedusing a Retsch sieve shaker to provide a homoge-nous sample. The samples were moisture-conditioned in desiccators with different saturatedsalt solutions for 14 da at 23�C. The MCs weremeasured with a Sartorius MA 30 moisture ana-lyzer (see Table 1). The samples were kept at–18�C before further analyses to minimize the stor-age effect (Hse and Kuo 1988; Back 1991; Nielsenet al 2009c).
Methods
The pelletizing components were measuredwith dies and pistons produced for the purpose(Fig 3) in an Instron 4485 testing machine witha 200-kN load cell and a data logging system.The Instron system was used to push the presspistons and simultaneously measure the re-quired position and force. Electrical heating(HSS braid coil, HT 30 regulator, Horst GmbH)was used to adjust the temperature of the dies.
Data for Wcomp and Wfric were measured as illu-strated by Fig 3a – b. The press channel diameterwas 8 mm, and a pellet was made by compres-sion of 0.750 g sawdust to 15-kN maximumforce (�300 MPa). Compression speed was 127mm/min and the maximum force held for 10 s(Fig 3a). After the pressure was released and thestop piston removed, the pellet was pushed
Table 1. Moisture content (oven-dry basis) after con-ditioning in desiccators.
Salt solution Pine Beech
MgCl2 6.0 5.8
CH3COOK 8.1 7.9
BrNa (dried) 9.7 10.7
BrNa 11.3 12.1
NaCl 12.1 12.9
NaCl (moistened) 12.6 13.6
NaCl (moistened) 15.9 NA
“Dried” and “moistened” refer to drying (50�C) and water addition (aero-
sol) after conditioning, respectively.
NA, Not available.
Nielsen et al—ANALYSIS METHODS FOR PELLETIZING PROPERTIES OF WOOD 417
3.5 mm in the press channel (Fig 3b). Pelletspeed was 127 mm/min, and the pellet wassubsequently pushed out of the channel. After24-h cooling at 20 – 25�C and ambient RH, thepellet strength was measured by placing thepellet on the side and measuring the maxi-mum force required to fracture it (Nielsen et al2009b).
The Wflow measurement is illustrated in Fig 3c – f.The flow die was made to imitate the press chan-nel of a commercially available press-channeldesign with a channel diameter of 8.0 mm and a2.5-mm deep cone-shaped press channel open-ing with 60� angles. The cross-sectional area ofthe press piston chamber equaled the sum of thearea of a press-channel opening and the asso-ciated horizontal surface area based on a com-mercial design. This resulted in a rim ofhorizontal area surrounding the Press channelopening (see Fig 2). The diameters were 10.89mm (press channel opening) and 14.76 mm(press piston chamber). Press channel lengthwas 40.00 mm, including the 2.5-mm channelopening. In the first step (Fig 3c – d), a com-
pressed layer (CL) was formed in the press chan-nel opening by compressing 0.75 g of sawdust.The press piston was stopped 1.00 mm abovethe press channel opening giving a compressedlayer density of 1500 kg/m3. In the second step(Fig 2e – f), the stop piston was removed and anadditional 0.25 g (second layer) was similarlycompressed causing the compressed layer toflow into the press channel. The press pistonwas stopped 1.00 mm above the press channelopening. Compression speed for both steps was25.4 mm/min.
Continuous pelletizing (Fig 3g) was dome bysuccessively pressing new layers (0.25 g) intothe press channel, which eventually filled thechannel, thus producing pellets and simulatingthe continuous pelletizing process.
The influence of temperature was measuredwith die temperatures of 60, 75, 85, 95, 105,115, 125, 135, 145, and 160�C with pine andbeech at 11.3 and 12.1% MC, respectively.Likewise, the influence of MC was measured inthe range of 6 – 16% (Table 1) at a die temper-ature of 125�C. Continuous pelletizing was
Figure 3. Illustrations of the two dies and test procedures. (a – b) The procedure for compression and friction analysis;
(c – f) the flow analysis; (g) the continuous pelletizing.
418 WOOD AND FIBER SCIENCE, OCTOBER 2009, V. 41(4)
made with pine and beech at 11.3 and 12.1%MC, respectively, and with a 70/30 pine/beechmix, all at 125�C die temperature.
Data Analysis
The areas under the plots of the force vs presspiston position data were used to calculate theenergy required in the Wcomp, Wfric, and Wflow
measurements. Figure 4 illustrates the data(lines) and areas (gray) used for Wcomp andWfric, which is the energy used in each proce-dure. The larger the area, the more energy wasrequired to compress the pellet and to push it inthe channel. Figure 5 illustrates the pressing ofthe second layer in the flow procedure. Wflow
was calculated as the area under the plot in Fig5a. The point where compression of the secondlayer shifted into flow of the compressed layerwas located from the slope inflection calculatedfrom the Wflow plot in Fig 5a. Figure 5a shows aplot of the slopes in Fig 5a. The force at thispoint was used to calculate the correspondingpressure [force/(press piston area)] giving theflow initiation pressure (Pflow). Each layer inthe continuous pelletizing process was analyzedin the flow measurements. A plot was made toillustrate the effect on Wflow from adding more
layers to the press channel and the increase infriction from a longer pellet in the channel. Theerror bars in the figures illustrate standard errorof the means.
RESULTS AND DISCUSSION
Compression Component
The effect of temperature (T) and MC on com-pression is illustrated in Fig 6a – b and it is seenthat both factors had a significant effect onWcomp. In the temperature plot, a minimum wasreached at approximately 105�C for both spe-cies. Pellet mass measurements (not presented)showed that T greater than 105�C caused signif-icant sample drying during the procedure, whichmay explain this minimum. Increased moisturedecreases the stiffness of the wood (Hillis andRozsa 1978; Gerhards 1982), and therefore, theincrease in Wcomp at T less than 105�C mostlikely related to sample drying, which seemsreasonable because the system was not sealed.This is supported by the importance of MCshown in Fig 7b, in which increasing MC causeda decrease in Wcomp probably also related tothe softening of lignin under these conditions(Goring 1963; Irvine 1984; Kelley et al 1987).
Figure 4. The data and areas used for Wcomp and Wfric. (a) The pellet density is plotted to illustrate the compression. The
area under the force vs position (mm) plot is used for Wcomp. (b) Approximately 200 N was required to start the pellet
movement, and the force vs position (mm) plot area used for (Wfric) is shown.
Nielsen et al—ANALYSIS METHODS FOR PELLETIZING PROPERTIES OF WOOD 419
Also, it can be seen that the decrease in Wcomp isnot caused by simple replacement of wood withwater because the increase in MC of 6 – 16% forpine caused a decrease in Wcomp of 38 – 28 J(26% decrease) and likewise 6 – 14% MC and45 – 37 J (18% decrease) for beech. A linear
regression of Wcomp (60 – 105�C) in Fig 6a illus-
trates a possible outcome in a sealed system withconstant MC.
A stress vs density analysis (not shown) showedfurther differences in compressive propertiesbetween the species. Beech required a higherstress for compression than pine for all densitiesin the range of 200 – 1550 kg/m3. Compressionof sawdust compared with solid wood elimi-nates the difference in initial density betweenthe samples, and therefore the analog of com-pression of pine and beech sawdust can be seenas the compression of solid pine and beech withthe same density. Therefore, the difference be-tween the two species cannot be explained bythe conventional relation between density andhardness (Kollmann 1968). The effect could berelated to the higher content of extractives inpine than beech (Fengel and Wegener 1989;Zule and Moze 2003; Josefsson et al 2006; Niel-sen et al 2009b), which are reported to act asplasticizers (Matsunaga et al 2000) and proba-bly lubricants in compression (Nielsen et al2009a).
Friction Component
Figure 7 shows that friction (Wfric) was alsominimized by increasing T and MC. The corre-lation between T and friction is supported byfindings with pine on steel (McMillin et al1970), but the correlation to MC contradictsother literature (Atack and Tabor 1958;Lemoine et al 1970; Guan et al 1983). Thesereports state that increasing MC increases thecontact area and thereby the friction of solidwood on a platen because of softening of thewood. However, the softening effect on a com-pressed pellet could be insignificant for the con-tact area compared with uncompressed wood.The pellet surface is smooth, similar to the presschannel wall in contrast to the rough and opencharacter of the cell structure that characterizesthe surface of uncompressed wood (Nielsen2009). The contact area between the pellet andthe press channel wall is therefore already sub-stantial from compression, which could mini-mize the effect of softening. More importantly
Figure 6. Importance of temperature (T) and MC for the
energy required for compression. The dashed regressions in
T are based on 60 – 105�C, r2pine = 0.95, r2beech = 0.95. In
the MC plot, r2pine = 0.94, r2beech = 0.88.
Figure 7. Importance of temperature (T) and MC for
Wfric. The dashed regressions in the T plot are based on
60 – 105�C, r2pine = 0.88, r2beech = 0.96.
Figure 5. Illustration of the flow component data analysis.
H is the height of the sawdust layer above the press channel
opening. (b) The slopes of the plots in (a) and the positions
of the slope optima, which are also marked in (a).
420 WOOD AND FIBER SCIENCE, OCTOBER 2009, V. 41(4)
for the friction is that it may be related to bond-ing or interaction mechanisms between thewood and the surface of the steel (Atack andTabor 1958). Therefore, the effect of increasingMC as seen in Fig 7 could be related to wateracting as a mobile layer or lubricating agent(Bowden and Tabor 1966) between the pelletand the press channel wall.
The lower friction of the pine pellets may becaused by a higher extractives content leadingto different friction properties and bondingstrength between the sawdust particles com-pared with beech (Nielsen et al 2009a). Therole of extractives on frictional propertieshas been previously described (McKenzie andKarpovich 1968; Lemoine et al 1970; McMillinet al 1970; Nielsen et al 2009b) in which ex-traction with various solvents increased thefriction and that high extractives contentinvolved low frictional properties. Scots pinecontains more lipophilic extractives than Euro-pean beech, 5 vs less than 1%, respectively(Fengel and Wegener 1989; Zule and Moze2003; Josefsson et al 2006; Nielsen et al2009b), which supports the findings in Fig 7.The strength of the beech pellets was signifi-cantly higher than the pine pellets (see belowand Fig 10), which may affect the stress ap-plied by the pellet to the press channel wall(stresswall). As stress was applied in the com-pression component, stresswall also arose in thepellet because of strain in the wood perpendic-ular to the direction of the stress. It is assumedthat the two sawdust types had the same fiberorientations because of the same cutting proce-dure and that the Poisson ratios (Bodig andGoodman 1973) were in the same range. How-ever, as a result of the higher strength of thebeech pellets, they maintained more stresswallcompared with the pine pellets after compres-sion, which increased the friction of beech.The assumption is supported by Nielsen et al(2009b) showing that pellet strength may affectstresswall. However, the two explanations maybe problematic to estimate because extractivesaffect both pellet strength and the frictionalproperties (Nielsen et al 2009a).
Pellet Strength
The effect of temperature and MC for pelletstrength is presented in Fig 8. The bonding pro-cesses are most likely a combination of sawdustparticle entanglement and autoadhesive surfacebonding processes (Back 1987, 1991; Gardner2006). Contact between the available sites forhydrogen bonding may therefore be crucial forthe pellet strength. Water sorbs at hydrogen-bonding sites (Schneider 1980; Hartley et al1992; Berthold et al 1996; Haygreen andBowyer 1996) thus occupying sites for particle-to-particle bonding and may therefore decreasethe pellet strength as seen in Fig 8 MC. Thiscould also cause the increase in strength inFig 8 T that may be related to drying as previ-ously discussed. The difference between the spe-cies is assumed to be caused by differences inthe extent of additional blocking of hydrogen-bonding sites by extractives forming a weakboundary layer (Stehr and Johansson 2000;Nussbaum and Sterley 2002). This phenomenonis also known as “surface inactivation,” whichrefers to a time-dependent surface deposition ofextractives, preventing access to the high-energybonding sites on the structural wood surface(Back et al 2000; Gindl et al 2004). Surfacedeposition of extractives on sawdust has previ-ously been shown (Nielsen et al 2009c), and inthe present work, the deposition may have oc-curred during the moisture conditioning proce-dure. This effect of extractives can be higher forScots pine because of the higher extractivescontent, which can explain the lower pellet
Figure 8. The effect of temperature and MC on the pellet
strength.
Nielsen et al—ANALYSIS METHODS FOR PELLETIZING PROPERTIES OF WOOD 421
strength compared with beech. It is also seen inFig 8 MC that the effect of increasing MC ishigher for beech (steeper slope) than for pinebecause water adsorption on a less contaminatedsurface (beech) may affect the strength morethan on a more contaminated surface (pine)where a greater number of high-energy bondingsites are covered by extractives.
Flow Component
Figure 9 illustrates the second part of the flowprocedure, which is used for Wflow, which is acombination of compression of the second layerand flow and friction of the compressed layerin the press channel opening. Therefore Wflow
as illustrated in Fig 10 is the sum of the energyrequired for these processes. The shift betweencompression and flow occurs when the pressurereaches Pflow (see Fig 5), and Fig 11 shows thata negative correlation was found between Pflowand T and MC. A significant effect of drying isalso seen in the T plot. Pflow may be closelyrelated to the pellet strength, because the shiftfrom compression to flow requires breaking ofbonding between the particles in the com-pressed layer. This is supported by the increasein Pflow caused by drying in Fig 11, T, which
correlates to the importance of MC on pelletstrength (Fig 8, MC).
Following the shift from compression to flow, thecompressed layer must be forced through thecone-shaped press-channel opening, and there-fore its viscosity and friction affect the energyrequirement for the flow. Figure 10 shows thatWflow was negatively correlated with T and MCand that drying also increased Wflow at T greaterthan 105�C. The viscosity and pellet strength mayalso be closely related because the bonding be-tween the particles must be continuously brokenas the layer of compressed sawdust is forced tochange its shape, when it is pressed into the chan-nel. However, in addition to the particle bonding,the viscous properties of the lignocellulose mayalso contribute to the Wflow because significantcell wall bending and deformation will be part ofthe process.
Figure 10. Importance of temperature (T) and MC for
the Wflow. The dashed regressions in T are based on 60 –
105�C, r2pine = 0.95, r2beech = 0.96. The plots illustrate the
energy required to compress and press 0.25 g of sawdust
into the press channel.
Figure 9. Illustration of the process in Fig 3e – f. (a) The
second layer and the shape of the compressed layer.
(b) Force is applied and the second layer is compressed
and starts to force the compressed layer into the press
channel opening and the press channel. (c) The press piston
has been stopped 1 mm above the press channel opening.
The white line refers to the initial position of the sawdust in
the compressed layer.
Figure 11. Correlation between Pflow and temperature (T)/
MC. The dashed regressions in T are based on 60 – 105�C,r2pine = 0.69, r2beech = 0.85.
422 WOOD AND FIBER SCIENCE, OCTOBER 2009, V. 41(4)
If the values for Wcomp, and Wfric are subtractedfrom Wflow, the energy requirement for the vis-cous change in shape of the compressed layercan be estimated (Wvisc). For beech at T = 60�C,Wcomp = 46 J, Wfric = 2.8 J (both for 0.75 g),and Wflow = 72 J (for 0.25 g). Wvisc can then beestimated by:
Wvisc ¼ 72J � 46J � 0:250:75
� 2:8J
0:75¼ 53J ð1Þ
Note that in the Wflow procedure, 0.25 g (secondlayer) is compressed and 1.0 g (compressed lay-er + second layer) is pushed 3.5 mm (notshown), which is also the distance for the Wfric
measurement. For pine at T = 60�C, Wflow = 38 Jvs Wvisc = 25 J. This corresponds to 74 and66% of Wflow for beech and pine, respectively,and shows that the majority of Wflow is allocatedto the process of changing the shape of the com-pressed sawdust as it enters and passes the presschannel opening. The difference between thespecies in Wflow (Fig 10) is therefore mainlycaused by the difference in Wvisc. This is sup-ported by the species difference in Pflow (Fig 11)in which higher stress was required to initiatethe flow of beech; hence, the physical strengthof the pellet (bonding between the particles)was an important factor that may be closelyrelated to the viscous properties of the com-pressed sawdust.
Continuous Pelletizing
In Fig 12, the continuous pelletizing process isillustrated as the increase in Wflow by succes-sively pressing layers of sawdust into the chan-nel. For pine, and the 70/30 pine/beech mix,a steady-state condition was reached after 15 –17 layers, when the press channel was full andpellets were produced (Fig 3g). At steady statefor pine, the pressure was 260 MPa when thepress piston was 1 mm above the press channelopening and likewise 315 MPa for the 70/30pine/beech mix. This is within the range of210 – 450 MPa in a conventional pellet mill(Leaver 2000). The steady-state condition wasnot reached with beech because of mechanicallimits of the test equipment when the force
required on the press piston exceeded 120 kN(�700 MPa).
Each layer in Fig 12 resulted in an increase inpellet length of approximately 3.5 mm; thus,based on the friction findings in Fig 7, it wouldbe expected that Wflow would increase with15 (layers) � 1.0 J (the energy required to pushthe compressed beech 3.5 mm) and correspond-ingly 15 � 0.4 J for pine, assuming that Wfric
could be linearly extrapolated. However, Fig 12shows that the increase in Wflow is not linearlycorrelated to the number of layers. The findingsupports previous work (Holm et al 2006, 2007)in which data and models incorporating thePoisson ratios of wood show that the stresson the press channel increases exponentiallywith increasing pellet length in the channel,thereby increasing the friction exponentially asseen in Fig 12.
Figure 12 also shows that the difference betweenspecies can be separated between the differencesin layer numbers 1 – 7 and after layer 7. Thismay indicate that the difference in layers 1 – 7was dominated by the difference in Wvisc, whichwas constant, and the small increase up to layer7 was more or less linearly correlated with theWfric data in Fig 7. After layer 7, however, the
Figure 12. The increase in Wflow when pressing layers
successively into the press channel.
Nielsen et al—ANALYSIS METHODS FOR PELLETIZING PROPERTIES OF WOOD 423
Wflow of beech increased more than pine. Themodel in Holm et al (2006) explains this differ-ence between species by their individual corre-lations between the stress to move the pellet(pressure) and the compression ratio (pelletlength/diameter). The model predicts a smallerratio (shorter pellet) of beech compared withpine before a steep increase occurs in the pres-sure for pellet movement, and layer 7 appears toinitiate this for beech in the present study.Therefore, a press-channel length resulting inpellets after layers 7 – 9 for beech could resultin a similar Wspec for pelletizing than for pinewith the present press channel length.
The continuous pelletizing method enables aquantitative analysis to compare the specificenergy requirements for pelletizing (Wspec). Fig-ure 12 shows that at steady-state pelletizing ofpine, Wflow of processing 0.25 g was approxi-mately 45 J, and this corresponds to �180 MJ/t,which is agreement with conventional pellet millenergy measurements by the authors (not pre-sented). Likewise, the 70/30 pine/beech mixwould require approximately 22 MJ/t based onFig 12, which is a 20% increase compared withpine. Also, it can be noted that for pine at 125�C,Wvisc was 20 J (not shown), which correspondsto 44% of the Wflow (45 J) for the continuouspelletizing process at steady state. This showsthat almost one-half of Wspec is used in the pro-cess allocated to the press channel openingwhere the compressed sawdust layer must flowinto the channels.
CONCLUSIONS
This article has presented tools and methodsthat separate the pelletizing process into com-pression, flow, and friction components to mea-sure the importance of raw materials propertiesfor the energy requirements of pelletizing andpellet strength. The study showed that in allcomponents, European beech sawdust requiredmore energy to process than Scots pine and thatincreasing temperature and MC decreased theenergy requirements for both species. It wasshown that a significant part of the energy for
the pelletizing process is allocated to the pro-cess of forcing the compressed sawdust fromthe surface of the die into the press channels. Itwas also shown that beech produced the stron-gest pellets, and it is suggested that the saw-dust’s bonding properties are an importantfactor for the energy required in the pelletizingprocess. The high bonding strength of beechwas reflected in the energy required to forcethe compressed sawdust into the channels andto high friction in the press channels. The mag-nitude of these may be correlated with a greaternumber of high-energy bonding sites (e.g., hy-droxyl groups) for bonding in beech than pinebecause of the lower extractives content inbeech. Also, a method to measure the energyrequirement in the continuous pelletizing pro-cess was shown to provide data comparable toan industrial pellet mill.
ACKNOWLEDGMENTS
The University of Copenhagen, DONG Energy,StatoilHydro, and The Danish Ministry of Sci-ence are gratefully acknowledged for facilitat-ing the presented work. Andritz Feed andBiofuel, Denmark is also gratefully acknowl-edged for manufacturing the test dies. The USNational Science Foundation Forest Biopro-ducts Initiative grant number EPS-0554545 isacknowledged for providing partial financialsupport for this project.
REFERENCES
Atack D, Tabor D (1958) The friction of wood. P Roy Soc