- Email: [email protected]
Moisture content and storage time influence the binding mechanisms in biofuel wood pellets
Applied Energy 99 (2012) 109–115
Contents lists available at SciVerse ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
Moisture content and storage time influence the binding mechanisms in biofuel wood pellets Robert Samuelsson ⇑, Sylvia H. Larsson, Mikael Thyrel, Torbjörn A. Lestander Swedish University of Agricultural Sciences, Unit of Biomass Technology and Chemistry, SE-901 83 Umeå, Sweden
h i g h l i g h t s " Pellet bulk density and energy consumption is mainly determined by moisture content. " Durability and fines are determined by moisture and storage time in a complex model. " A qualitative model for binding mechanisms in wood pellets is suggested.
a r t i c l e
i n f o
Article history: Received 20 December 2011 Received in revised form 4 April 2012 Accepted 4 May 2012 Available online 31 May 2012 Keywords: Pellet quality Storage time Extractive content Binding mechanism
a b s t r a c t In a pelletizing experiment the three factors sawdust moisture content, steam conditioning of the sawdust, and storage time of the raw material, were varied in a full factorial design with a total of 34 experiments to evaluate the influences on the pelletizing process and the pellet quality when producing biofuel wood pellets from pine sawdust. Moisture content of the sawdust was found to be the dominant factor for the bulk density and for the pelletizer motor current, both showing low values at high moisture contents due to the lubricating property of water that lowers the friction in the pelletizing process. More complex models were obtained for mechanical durability and the amount of fines, where all factors and most interactions and squared terms were significant. In order to explain the effect of the sawdust moisture content and the storage time on these response factors, a qualitative model for the binding mechanism is suggested. In this model water is supposed to be actively involved in the binding mechanism as hydrogen bonded bridges. The increase in binding strength with storage time is explained by the reduction of extractive content which contain molecules that can block binding sites on the material surface. Optimum pellet quality was obtained when the storage time exceeded 120 days and within a range of sawdust moisture content of 11–13%. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Biofuel pellet quality is a result of used feedstock and process settings in the pelletizing process. The industrial pelletizing process is partly quite rigid when it comes to process control. Process settings such as particle size distribution, die channel length, gap between rollers and die, cannot be altered without interrupting the production for adjustment. In most cases, the only real-time process controlling parameters, besides blending of raw materials, are feeding rate, raw material moisture content, and steam conditioning. Bulk density, mechanical durability and fines are among the most important quality parameters for fuel pellets. Bulk density is a measure of the extent of the compaction of the particles in the pellet, while the durability and the fines is a result of the ⇑ Corresponding author. Tel.: +46 90 786 87 94. E-mail address: [email protected] (R. Samuelsson). 0306-2619/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2012.05.004
bonding strength between the particles in the pellet. The latter are strongly correlated where high durability results in low amount of fines. Bonding mechanisms have been widely discussed in the literature [1–6]. Rumpf [1] was the first to describe possible mechanisms of bonding in granules and agglomerates and he proposed the following five mechanisms: (i) attraction forces between solid particles; (ii) interfacial forces and capillary pressure in movable liquid surfaces; (iii) adhesion and cohesion forces; (iv) solid bridges; (v) mechanical interlocking between particles. Attraction forces between solid particles, i.e. hydrogen bonds and van der Waals forces, are short range forces that are active only when particles are close together and the attraction decreases rapidly when the distance is increasing. Interfacial forces and capillary pressure in movable liquid surfaces are results of surface tension and capillary forces between the liquid and the particles. These forces create strong bonds between particles but disappear when the liquid evaporates. Cohesion is the attraction between
110
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
molecules of the same substance, while adhesion is the attraction between molecules of different kind. Solid bridges are formed by different mechanisms basically at high pressure and temperature by crystallization of dissolved substances, hardening of binders, and melting and sintering of various pellet components. Mechanical interlocking is a bonding mechanism where the particles fold about each other to form interlocking bonds. Water has a crucial role in the pelletizing process [3–5]. It is found to be the most important factor influencing the pellet quality and it can act as both a binding agent that affect mechanical durability and fines, and as a lubricant that lower the friction in the die resulting in low bulk density and energy consumption [5]. Inherent physical and chemical properties of raw materials can be manipulated to tailor the fuel pellet feedstock. An example is how process operators in plants producing wood pellets have found that seasoning of pine sawdust in piles for a period of time before pelletizing improves the pelletizing properties. It is well known that seasoning of softwood sawdust reduces the concentration of extractives and that of the two major Scandinavian softwood species Scots pine has higher extractive content than Norway spruce [7]. Since fresh pine sawdust show very different properties in the pelletizing process compared to stored pine sawdust and spruce sawdust it has been suggested that the extractive content is influential on pelletizing properties and pellet quality [8–10]. However, detailed studies have not been performed to quantify the influence of different process settings on pine sawdust with varying seasoning time and the effect of sawdust moisture and extractive content. In this study, the three factors sawdust moisture content, steam conditioning of the sawdust, and storage time of the raw material, were varied in a full factorial design with a total of 34 experiments to evaluate the influences on the pelletizing process and the pellet quality when producing biofuel wood pellets from pine sawdust. The objectives were (1) to determine the influences and interactions of these factors on pellet quality responses, (2) to quantify a required seasoning time for optimum pellet quality and (3) to develop a qualitative model for the binding mechanisms in pellets with different moisture and extractive contents. 2. Materials and methods 2.1. Biomaterials Scots pine (Pinus sylvestris L.) sawdust originating from latitude 64 °N in Sweden was delivered from a sawmill (SCA Timber AB in Holmsund, Sweden). About 18 tons of sawdust was transported and stacked on asphalt at the end of June at the Swedish national pilot plant Biofuel Technical Center (BTC), Swedish University of Agricultural Sciences (SLU), Umeå, Sweden. The pile (3 m high, 5 m wide at the bottom, and 10 m long) was stored for 160 days, during which extractive content and content of Klason lignin were monitored. At specific storage times (0, 46, 81, 117 and 160 days) batches were taken from the short end of the pile after removal of a protecting sawdust layer of about 2 m along the pile. The protecting sawdust layer was put back again after each sampling occasion. This was done to ensure that the batches were taken from the inside of the pile with a similar environment each storage time. 2.2. Experimental All pelletizing experiments were performed at the SLU BTC national pilot plant and the experimental set-up consisted of a horizontal batch drier, a combined mixer wagon and balance, conveyor systems, mill, steam treatment equipment, pelletizer
and pellet cooler. The full system is described in detail in [10]. The pelletizer (SPC PP300, Swedish Power Chippers, Borås, Sweden) with a capacity of 300 kg/h, was equipped with a fixed die (outer diameter 540 mm, press channel lengths of 55 mm) and two rotating press rollers (200 mm in diameter). The mill had a 4 mm sieve, and saturated steam was generated by a steam boiler to a temperature of 111.0 ± 0.76 °C. The steam conditioning of the sawdust material was carried out during 2 min just before reaching the pellet press. The pelletizer was controlled by a programmable logic controller (PLC) (Swedish Power Chippers, Borås, Sweden). Drying of the raw sawdust with approximately 50% moisture content was executed on a ventilated batch drier with a capacity of about 400 kg/batch and the temperature of the in going ventilating air was about 40 °C. The drying process was continued until the moisture content of the material was close to the design value, which was controlled by repeated moisture content measurements using a moisture balance (Sartorius MA30, Tillquist, Sweden). A drying cabinet (1600 800 700 mm) (Elvärmedetaljer, Sweden) calibrated to 105 ± 2 °C was used for oven drying of the samples collected from the pelletizing trials. Temperature calibrations were carried out using a thermometer (Technoterm 9300, Nordtec Instrument AB, Sweden) calibrated at an accredited laboratory (AREPA Mätteknik AB, Sweden). The size of the drying trays was 750 600 mm and the samples were weighed to 0.1 g accuracy with a balance (Mettler PE 16, instrument TEKNIK, Sweden). 2.3. Experimental design Pelletizing of the sawdust was executed in a full factorial design with three parameters: storage time at the five levels: 0, 46, 81, 117 and 160 days, moisture content at the three levels: 8.4 ± 0.48%, 10.7 ± 0.38% and 13.1 ± 0.38%, and steam treatment at the two levels: 2 and 6 kg/h. There were some difficulties to reach the intended design values of 8%, 11% and 14% for the moisture content, especially for the upper level, probably due to loss of moisture after the drying process was completed. For determination of experimental precision, three replicates of the experiments with moisture contents of 11% and storage time of 160 days were carried out for both steam levels, respectively. In all, the design for pelletizing consisted of 34 experiments in a randomized order except for storage time. About 75 kg of sawdust was milled and compressed into pellets in each experiment. The response factor parameters measured during each experiment were: bulk density and mechanical durability of produced pellets, the share of fines generated during the pelletizing process, and the energy consumption of the pelletizer recorded as the mean current during a 2 min measurement period. Each experiment lasted for about 30 min at a production rate of 154 ± 11 kg/h, where the first 20 min were used to stabilize the process and obtain a fairly constant die temperature. During the last 10 min of each experiment, pelletizer motor current was monitored and samples of 1 L milled sawdust, and 5 kg pellets were collected during two different measurement periods. All samples of sawdust and pellets were analysed for moisture content according to the CEN standard [11]. In Fig. 1 the correlation between sawdust moisture content and pellet moisture content for the 2 kg/h steam treatment is shown. The content of fines produced in the pelletizing process was quantified by manual sieving of an 8 L sample of pellets through a 3.15 mm sieve, and calculated as the percentage of the loss of the fine material to the total sample weight. Pellet bulk densities were determined using the CEN standard [12]. Mechanical durability of the pellets was measured by use of a pellet tester (Q-tester, Simon Heesen BV, Netherlands) according to the CEN standard [13]. The pellet length distribution was determined by subsequent dividing of the pellet sample until about 10 pellets remained followed by a length determination using a calliper.
111
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
60 °C) and acetone (90–10 v/v). The extraction time was 1 h and extractions were repeated twice for each sample [15]. The amount of Klason lignin was determined according to [16]. 2.4. Modelling and diagnostics
Fig. 1. Correlation between sawdust moisture content (%) and pellet moisture content (%) with the 2 kg/h steam conditioning.
The pelletizer motor current was recorded during the sampling of sawdust and pellets at a frequency of 1 Hz by the PLC using a data logger (Mitec, SatelLite-U (type 1), Mitec Instrument AB, Sweden). Motor current determination was restricted to pelletizing of stored sawdust only. Pelletizing of fresh sawdust gave rise to uneven pellet production due to feed layer losses (c.f. [14]), and thus, reliable pelletizer current data could not be gathered. The amount of extractives in the sawdust samples were determined by use of an extraction system (Universal Extraction System B-811 from Büchi Labortechnik AG, Flawil, Switzerland). The extraction solvent was a mixture of petroleum ether (bp 40–
All measured response factors in the factorial design were analysed by multiple linear regression (MLR) using the software MODDE 8.0.0.0 (Umetrics AB, Sweden). This software was also used in the design of the experiments. All values were centred before evaluation. Leave-one-out crossvalidation was used to calculate the residual for each validation round. These residuals (ft) between observed (y) and predicted responses (yp), i.e. ft = y yp, were ordered in a column (I 1) and used as diagnostics together with leverage of each observation. The different multivariate models were then evaluated using the coefficient of multiple determinations (Q2) and calculated as:
Q 2 ¼ 1 PRESS=ðyT yÞ where yTy is the total sums of squares in the data and PRESS is the leave-one-out cross-validation predictor sum of squares and calculated according to [17] as: 2
PRESS ¼ 1T ðf t ð1 diagðHÞÞ2 Þ where 1 is a column vector of I 1 ones; T denotes a transposed 2 vector or matrix; f t a column vector of I 1 squared residuals; diag(H) represents the leverage of each observation and is the I 1
Table 1 Summary of the design and experimental results. Analytical results are average values of the two samples collected during each run. Motor current values are the average of 1 Hz sampling during 2 min of pelletizing. Exp. name
N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 N14 N15 N16 N17 N18 N19 N20 N21 N22 N23 N24 N25 N26 N27 N28 N29 N30 N31 N32 N33 N34 a
Parameters
Responses
Storage (days)
Moisture (%)
Steam (kg/h)
Bulk density (kg/m3)
Durability (%)
Fines (%)
Current (A)
Extractives (%)
Klason lignin (%)
Pellet length (mm)
0 0 0 0 0 0 46 46 46 46 46 46 81 81 81 81 81 81 117 117 117 117 117 117 160 160 160 160 160 160 160 160 160 160
10.7 10.7 12.9 12.7 7.8 7.7 8.8 9.3 11.5 11.5 13.8 13.8 12.9 12.9 10.6 10.1 8.1 8.5 10.5 10.7 8.6 8.8 13.1 12.8 10.6 10.6 8.2 8.5 10.6 10.6 12.9 12.9 10.8 10.6
2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6
612 602 525 523 709 712 717 693 660 630 553 523 550 550 711 673 734 723 712 662 735 703 647 597 695 657 755 730 677 675 625 598 685 663
83.5 88.2 61.8 71.4 84.6 89.4 93.5 94.1 90.5 92.3 76.2 80.6 84.1 89.7 93.6 95.8 91.8 93.9 96.6 96.9 94.3 96.0 96.6 95.7 96.9 97.2 93.8 96.0 96.8 97.1 97.4 96.8 96.7 97.2
4.2 3.0 8.4 7.2 7.6 2.9 1.5 1.3 1.6 1.2 2.9 3.0 2.1 2.6 0.85 0.89 1.7 0.76 0.96 0.45 0.97 0.62 0.49 0.93 0.33 0.28 0.96 0.48 0.31 0.46 0.49 0.33 0.46 0.30
n.a.a n.a.a n.a.a n.a.a n.a.a n.a.a 29.4 28.4 26.5 26.5 25.2 25.1 25.1 26.0 26.0 25.1 29.0 28.6 27.1 26.3 27.4 26.6 24.7 24.4 27.0 26.9 29.6 27.8 27.4 26.8 25.2 24.7 27.1 26.0
3.44 3.17 4.04 4.03 3.81 3.79 3.06 2.91 2.89 2.87 2.56 3.05 2.50 2.19 2.15 2.29 2.22 2.32 2.21 2.16 2.29 2.16 2.20 1.88 2.06 1.91 2.05 1.81 1.66 1.74 1.74 1.69 1.61 1.57
24.0 25.0 24.6 24.2 24.1 23.4 24.7 23.6 24.6 25.4 25.3 25.9 25.4 25.5 25.3 25.5 25.7 25.5 24.8 25.3 24.7 24.9 25.1 25.4 25.8 25.8 24.9 25.3 27.2 26.9 26.8 26.6 26.3 26.5
26.0 28.5 8.9 22.0 22.4 24.7 23.7 27.0 24.0 28.3 21.3 22.1 24.8 13.2 26.8 24.7 25.4 21.3 23.6 23.8 21.7 28.9 28.9 31.6 26.6 32.7 18.0 23.8 34.0 30.9 33.9 33.6 31.2 35.9
n.a.: Not analyzed.
112
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
diagonal elements of the symmetric (I I) hat matrix H, that here is calculated as X(XTX)1XT with dimension I I containing the model parameters ordered in a X matrix. The Q2 value expresses how much of the variance in the response variable (y) that can be predicted and can at best be 1 and a value of 0 indicate no predictive capability at all. The number of factors used in the models was determined by optimization of Q2. 3. Results and discussion The experimental design and corresponding results are summarized in Table 1. In Table 2 the coefficients for the different factors in the regression analyses of the various responses are summarized and the following equations can be written for each response:
BD ¼ 660 99:4 Moi þ 32:9 Sto 11:5 Ste
Fig. 2. Bulk density (kg/m3) of pellets produced from sawdust with various moisture contents (%) at five different storage times (days) and a steam conditioning of 6 kg/h.
þ 16:0 ðMoi StoÞ 17:2 ðMoi MoiÞ 23:0 ðSto StoÞ Pellets produced from fresh sawdust had lower bulk density than pellets from sawdust that had been stored for 46 days or more. There was no obvious difference between bulk densities for pellets from sawdust with storage times longer than 46 days. No correlation between pellet length and bulk density was found (R2 = 0.04), which indicate that the eventual influence of the pellet length on bulk density is small in comparison to the effect from in particular the sawdust moisture content.
MD ¼ 94:7 5:5 Moi þ 8:6 Sto þ 1:2 Ste þ 7:5 ðMoi StoÞ 1:3 ðSto SteÞ 6:1 ðMoi MoiÞ 5:1 ðSto StoÞ Fi ¼ 0:78 þ 0:71 Moi 2:25 Sto 0:25 Ste þ 1:13 ðMoi MoiÞ þ 1:8 ðSto StoÞ PC ¼ 25:9 2:24 Moi þ 0:41 Sto 0:46 Ste
3.2. Mechanical durability
BD, MD, Fi and PC correspond to bulk density, mechanical durability, fines and press current, respectively. Moi, Sto and Ste are defined in Table 2. The standardized factor settings used in the calculations are 0.85 ± 0.16, 0.09 ± 0.12 and +0.69 ± 0.13 for the three levels of moisture content; 1, 0.425, +0.0125, +0.463 and +1 for the five levels of storage time; and 1 and +1 for the two levels of steam conditioning, respectively. 3.1. Bulk density The overall trend was a lowered bulk density with increasing moisture content within the interval of 8–14% moisture. From Table 2 it is shown that the most significant effect was obtained at low moisture contents and long storage times, which affected the bulk density in a positive manner. A Q2-value of 0.91 indicated an excellent prediction accuracy of the model. Moisture content was the parameter with the greatest impact on pellet bulk density (Fig. 2 and Table 2). There is an effect also from the steam treatment similar to the effect from the sawdust moisture content. However, it is much smaller and is probably due to the alteration of the moisture content of the material.
The analysis for mechanical durability (Table 2) showed a more complex pattern compared to that for bulk density. Storage time was found the most influential parameter and a long storage time resulted in high durability. However, no factor was dominant over the others and most interactions and squared terms were involved in the model building. The prediction ability of the model, Q2, was 0.85. The influence of moisture on the mechanical durability for various storage times is shown in Fig. 3. Four important conclusions was drawn from this result; (i) there was a large increase in mechanical durability with storage time, (ii) local maxima of mechanical durability within the studied range of moisture content were observed for all storage times, (iii) the durability maxima were shifted towards higher moisture contents the longer time the sawdust was stored and were found at 9.2%, 9.8%, 9.9%, 11.3% and 11.9%, respectively and (iv) the sensitivity for changes in moisture content on the mechanical durability decreased with storage time. The local maxima in mechanical durability were calculated from the 1st derivative of the polynomial regression line for each storage time.
Table 2 Summary of the coefficients for the different factors in the regression analysis of the various responses. Moi: moisture content, Sto: storage time, Ste: steam treatment, Q2: predicting power of the models.
Constant Moi Sto Ste Moi Sto Sto Ste Sto x Sto Moi Moi Q2 ns
Not significant.
Bulk density (kg/m3)
Durability (%)
Fines (%)
Current (A)
Extractives (%)
Klason lignin (%)
660 99.4 32.9 11.5 16.0
94.7 5.5 8.6 1.2 7.5 1.3 5.1 6.1 0.85
0.78 0.71 2.25 0.25ns
25.9 2.24 0.41ns 0.46
2.40 0.03ns 0.95 0.03ns
25.1 0.49 0.49ns 0.04ns
23.0 17.2ns 0.91
1.8 1.13ns 0.63
0.60
0.35
0.61ns
0.89
0.27
113
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
Fig. 5. Decrease in extractive content (%) with storage time (days).
Fig. 3. Mechanical durability (%) of pellets produced from sawdust with various moisture contents (%) at five different storage times (days) and a steam conditioning of 6 kg/h.
Fig. 4. Correlation between mechanical durability (%) and extractive content (%) at three levels of sawdust moisture contents (%).
A large increase in mechanical durability with decreasing extractive content was found (Fig. 4) and this effect was more pronounced at higher sawdust moisture levels. 3.3. Fines, energy consumption, extractive content and Klason lignin The regression analysis for the amount of fines collected after the pelletizing was very similar to that of the durability, except that the effects had the opposite direction. However, the Q2-value (0.61) was considerably lower. The analysis for the energy consumption of the pellet press was performed for the stored sawdust only and showed that the current was almost solely determined by the moisture content of the sawdust among the studied factors in this investigation, where high energy consumption was obtained at low moisture levels. The Q2-value of 0.60 indicated that the model was acceptable but not as good as for bulk density and mechanical durability. In general, extractives also act like lubricants in the die with low friction and low energy consumption as a consequence. This was also found for the fresh sawdust, which was indicated by significantly lower bulk density for pellets from fresh sawdust compared to pellets from sawdust stored for 46 days. However, due to the uneven pellet production for the fresh sawdust the unreliable pelletizer currents are not reported. For longer storage times a small increase in current with decreasing extractive content was observed, but the variation in current was too large for this effect to be significant.
In Fig. 5 the decrease of extractive content with storage time is shown and the model indicated that the storage time was the only factor that affected this response. As suggested earlier [4,9,10], the large increase in mechanical durability and decrease in the amount of fines with increasing storage time may be caused by a decrease in extractive content. To verify this, storage time was replaced by extractive content in the regression analysis for the two responses. This resulted in an increase of the Q2-value from 0.85 to 0.88 for mechanical durability and from 0.63 to 0.81 for the amount of fines (Table 3), indicating that the decrease in extractives during storage is the main reason to the effect. Klason lignin was measured to see whether the storage of the sawdust had increased the content of the lignin and if this might explain the changes in pellet quality. However, no predictive model was obtained for this response and only a small increase was found partly caused by the decrease in extractives. In addition, no correlation between the lignin content and the pellet quality was obtained. 3.4. Required seasoning time for optimum pellet quality The suggested quality criteria of pellets for non-industrial use is P600 kg/m3, P97.5% and 61.0% for bulk density, mechanical durability and fines, respectively [18]. From Figs. 2 and 3, and Table 1 it was found that quality requirements for bulk density and fines were fulfilled for all storage times longer than about 120 days. For mechanical durability, requirements could not be reached, but an optimum value of about 97% was obtained for the two longest storage times within the range of moisture content between 11 and 13%. This limitation was probably caused by the experimental setting of the pellet press, since further storage of the sawdust did not give any significant increase in the mechanical durability.
Table 3 Summary of the coefficients for the different factors in the regression analysis of mechanical durability and fines when the factor storage time is replaced by extractive content. Moi: moisture content, Ext: extractive content, Ste: steam treatment, Q2: predicting power of the models.
Constant Moi Ext Ste Moi Ext Ext Ste Moi Moi Ext Ext Q2 ns
Not significant.
Durability (%)
Fines (%)
91.5 6.7 10.6 1.4 7.5 1.7 5.8 3.7 0.88
1.98 0.51 3.23 0.36 0.58 1.87 0.38ns 0.81
114
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
3.5. Qualitative model Raw material moisture content was the single most important parameter for bulk density and energy consumption. As reported earlier [4,10,19], this may be explained by moisture altering the friction when biomaterial is compacted and pressed through the die channels, and consequently also the back pressure. To some extent, storage time also affected these response parameters, caused by the decrease in the content of extractives that act like lubricants in the pelletizing process. A much more complex mechanism was found for mechanical durability and amount of fines where all factors and most interactions and squared terms contributed to the models. Back discussed in a review [2] the bonding mechanism in the manufacture of hardboard from lingo-cellulosic fibres and he concluded that most of the inter-fibre bonding was due to secondary bonds, in particular hydrogen bonds but also to some extent van der Waals forces. The pressure used in the hardboard process is of the order of 5–15 MPa [20–22] and considering that the pressure used in the pelletizing process is much higher, 200–450 MPa, it is reasonable to assume a similar bonding mechanism in fuel pellet manufacture even if other mechanisms may contribute. The dominant hydrogen bonding group in the ligno-cellulosic material is the hydroxyl group, although the carbonyl, the carboxyl and the ether groups are also involved in the bonding. Even the very weak CAH O bond, bond strength 2.5 kJ mol1, may contribute to the bonding due to its multiplicity [23]. The bonding strengths for the hydrogen bond are of the order of 2.5– 120 kJ mol1 [23] and the bond lengths vary between 0.12 and 0.32 nm depending on the bond strength, where the shortest length correspond to the strongest bond, i.e. the FAH F bond. This means that by compressing the sawdust to reach a distance of less than 0.32 nm between the binding sites on the particle surfaces, hydrogen bonding may occur. If the distance between the active sites is too long no hydrogen bonding can take place. However, dipole molecules, like water, can act like bridges between these sites, with an increase in bonding strength as a result. This might be possible if the distance between the particles surfaces are in the range 0.5–0.74 nm, which correspond to twice the hydrogen bond length plus the OAH length in water (0.1 nm) assuming weak hydrogen bond strength (<20 kJ mol1) with bond lengths of 0.2–0.32 nm. Oleophilic low molecular weight material in the extractives, i.e. fatty acids and waxes, may interfere with the bonding mechanism by blocking and consequently reducing the number of binding sites on the surfaces [2,24] with low bonding strength between the particles as a result. This is due to the high adhesive force between these compounds and the wood surface, and the low cohesive strength of the oleophilic interface. The mechanism discussed above may explain the increase in mechanical durability with storage time as is shown in Fig. 3, since the extractive content is decreased during storage. The model may also explain the shift of the optimum in mechanical durability as a function of storage time; extractives block a number of binding sites in fresh sawdust, and thus, few hydrogen bindings between particles occur and only a small number of water molecules can act like bridges, resulting in low durability with a maximum at low moisture content. During storage the number of sites increases with increased mechanical durability as a result. In addition, more water molecules may be involved in the binding mechanism giving a maximum durability at higher moisture content. If the moisture content is too low, gaps where water molecules can bridge are not filled up and binding strength is lowered. On the other hand, if the moisture content is too high, water will attack the hydrogen bonds between the particles, resulting in decreased durability [2,4]. These effects will have a greater impact in fresh sawdust where the relative number of active binding sites are smaller, resulting in a larger
sensitivity to changes in moisture content on mechanical durability. 4. Conclusions – Moisture content was the dominant factor for bulk density and pelletizer motor current, which is explained by the alteration of the friction through the die due to the lubricating property of water. – For the mechanical durability and the amount of fines more complex models were obtained, where all factors and most interactions and squared terms were significant. A qualitative model for the binding mechanism suggests that extractives block binding sites for hydrogen bonding to occur and that water is actively involved in the bonding. – An optimum pellet quality was obtained at storage times longer than 120 days and within a range of moisture content of 11– 13%.
Acknowledgements We thank SCA Timber AB, Holmsund Sweden for the delivered sawdust. The Swedish Energy Agency, the Swedish Pellet Association (PiR), Swedish University of Agricultural Sciences and the Project Bio4Energy are thanked for financial support. Carina Jonsson is greatly acknowledged for carrying out the chemical analyses of the various biomaterials. References [1] Rumpf H. The strength of granules and agglomerates. In: Knepper WA, editor. Agglomeration. Interscience; 1962. p. 379–418. [2] Back EL. The bonding mechanism in hardboard manufacture – review report. Holzforschung 1987;41:247–58. [3] Mani S, Tabil LG, Sokhansanj S. An overview of compaction of biomassgrinds. Powder Handling Process 2003;15:160–7. [4] Nielsen NPK, Gardner DJ, Poulsen T, Felby C. Importance of temperature, moisture content, and species for the conversion process of wood residues into fuel pellets. Wood Fiber Sci 2009;41:414–25. [5] Kaliyan N. Morey RV. Factors affecting strength and durability of densified biomass products. Biomass Bioenergy 2009;33:337–59. [6] Stelte W, Holm JK, Sanadi AR, Barsberg S, Ahrenfeldt J, Henriksen UB. A study of bonding and failure mechanisms in fuel pellets from different biomass resources. Biomass Bioenergy 2011;35:910–8. [7] Assarsson A, Åkerlund G. Studies on wood resin, especially the change in chemical composition during seasoning of the wood. Part 4. The composition of the petroleum ether soluble nonvolatile extractives from fresh spruce, pine, birch and aspen wood. Svensk Papperstidn Nord Cellulosa 1966;69:517–25. [8] Bradfield J, Levi MP. Effect of species and wood bark ratio on pelleting of southern woods. Forest Prod J 1984;34:61–3. [9] Nielsen NPK, Gardner DJ, Felby C. Effect of extractives and storage on the pelletizing process of sawdust. Fuel 2010;89:94–8. [10] Samuelsson R, Thyrel M, Sjöström M, Lestander TA. Effect of biomaterial characteristics on pelletizing properties and biofuel pellet quality. Fuel Process Technol 2009;90:1129–34. [11] EN 14774–2:2009. Solid biofuels – methods for determination of moisture content – oven dry method. Total moisture: simplified method. [12] EN 15103:2009. Solid biofuels – determination of bulk density. [13] EN 15210–1:2009. Solid biofuels – determination of mechanical durability of pellets and briquettes. Pellets. [14] Larsson SH, Rudolfsson M, Thyrel M, Örberg H, Kalén G, Wallin M, et al. Temperature controlled feed layer formation in biofuel pellet production. Fuel 2012;94:81–5. [15] Arshadi M, Gref R. Emission of volatile organic compounds from softwood pellets during storage. Forest Prod J 2005;55:132–5. [16] Theander O, Westerlund EA. Studies on dietary fiber. 3. Improved procedures for analysis of dietary fiber. J Agr Food Chem 1986;34:330–6. [17] Myers RH. Classical and modern regression with applications. Boston, USA: Duxbury Press; 1986. [18] EN 14961–1:2010. Solid biofuels – fuel specifications and classes – Part 1: General requirements. [19] Lestander TA, Johnsson B, Grothage M. NIR techniques create added values for the pellet and biofuel industry. Bioresource Technol 2009;100:1589–94. [20] Angles NN, Montane D, Ferrando F, Salvadó J. Properties of binderless panels from thermomechanical aqueous vapor exploded residual softwood. In:
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115 Kopetz H, Weber T, Palz W, Chartier P, Ferrero GL, editors. Biomass for energy and industry; 1998, p. 435–8. [21] Salvadó J. Velásquez JA., Ferrando F. Binderless fiberboard from steam exploded Miscantus Sinensis: optimization of pressing and pretreatment conditions. Wood Sci Technol 2003;37:279–86. [22] Okuda N, Sato M. Manufacture and mechanical properties of binderless boards from kenaf core. J Wood Sci 2004;50:53–61.
115
[23] Walker J. Basic wood chemistry and cell wall ultrastructure. In: Walker J, editor. Primary wood processing, principle and practice. Springer; 2006. p. 23–30. [24] Stehr M, Johansson I. Weak boundary layers on wood surfaces. J Adhes Sci Technol 2000;14:1211–24.
Contents lists available at SciVerse ScienceDirect
Applied Energy journal homepage: www.elsevier.com/locate/apenergy
Moisture content and storage time influence the binding mechanisms in biofuel wood pellets Robert Samuelsson ⇑, Sylvia H. Larsson, Mikael Thyrel, Torbjörn A. Lestander Swedish University of Agricultural Sciences, Unit of Biomass Technology and Chemistry, SE-901 83 Umeå, Sweden
h i g h l i g h t s " Pellet bulk density and energy consumption is mainly determined by moisture content. " Durability and fines are determined by moisture and storage time in a complex model. " A qualitative model for binding mechanisms in wood pellets is suggested.
a r t i c l e
i n f o
Article history: Received 20 December 2011 Received in revised form 4 April 2012 Accepted 4 May 2012 Available online 31 May 2012 Keywords: Pellet quality Storage time Extractive content Binding mechanism
a b s t r a c t In a pelletizing experiment the three factors sawdust moisture content, steam conditioning of the sawdust, and storage time of the raw material, were varied in a full factorial design with a total of 34 experiments to evaluate the influences on the pelletizing process and the pellet quality when producing biofuel wood pellets from pine sawdust. Moisture content of the sawdust was found to be the dominant factor for the bulk density and for the pelletizer motor current, both showing low values at high moisture contents due to the lubricating property of water that lowers the friction in the pelletizing process. More complex models were obtained for mechanical durability and the amount of fines, where all factors and most interactions and squared terms were significant. In order to explain the effect of the sawdust moisture content and the storage time on these response factors, a qualitative model for the binding mechanism is suggested. In this model water is supposed to be actively involved in the binding mechanism as hydrogen bonded bridges. The increase in binding strength with storage time is explained by the reduction of extractive content which contain molecules that can block binding sites on the material surface. Optimum pellet quality was obtained when the storage time exceeded 120 days and within a range of sawdust moisture content of 11–13%. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Biofuel pellet quality is a result of used feedstock and process settings in the pelletizing process. The industrial pelletizing process is partly quite rigid when it comes to process control. Process settings such as particle size distribution, die channel length, gap between rollers and die, cannot be altered without interrupting the production for adjustment. In most cases, the only real-time process controlling parameters, besides blending of raw materials, are feeding rate, raw material moisture content, and steam conditioning. Bulk density, mechanical durability and fines are among the most important quality parameters for fuel pellets. Bulk density is a measure of the extent of the compaction of the particles in the pellet, while the durability and the fines is a result of the ⇑ Corresponding author. Tel.: +46 90 786 87 94. E-mail address: [email protected] (R. Samuelsson). 0306-2619/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apenergy.2012.05.004
bonding strength between the particles in the pellet. The latter are strongly correlated where high durability results in low amount of fines. Bonding mechanisms have been widely discussed in the literature [1–6]. Rumpf [1] was the first to describe possible mechanisms of bonding in granules and agglomerates and he proposed the following five mechanisms: (i) attraction forces between solid particles; (ii) interfacial forces and capillary pressure in movable liquid surfaces; (iii) adhesion and cohesion forces; (iv) solid bridges; (v) mechanical interlocking between particles. Attraction forces between solid particles, i.e. hydrogen bonds and van der Waals forces, are short range forces that are active only when particles are close together and the attraction decreases rapidly when the distance is increasing. Interfacial forces and capillary pressure in movable liquid surfaces are results of surface tension and capillary forces between the liquid and the particles. These forces create strong bonds between particles but disappear when the liquid evaporates. Cohesion is the attraction between
110
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
molecules of the same substance, while adhesion is the attraction between molecules of different kind. Solid bridges are formed by different mechanisms basically at high pressure and temperature by crystallization of dissolved substances, hardening of binders, and melting and sintering of various pellet components. Mechanical interlocking is a bonding mechanism where the particles fold about each other to form interlocking bonds. Water has a crucial role in the pelletizing process [3–5]. It is found to be the most important factor influencing the pellet quality and it can act as both a binding agent that affect mechanical durability and fines, and as a lubricant that lower the friction in the die resulting in low bulk density and energy consumption [5]. Inherent physical and chemical properties of raw materials can be manipulated to tailor the fuel pellet feedstock. An example is how process operators in plants producing wood pellets have found that seasoning of pine sawdust in piles for a period of time before pelletizing improves the pelletizing properties. It is well known that seasoning of softwood sawdust reduces the concentration of extractives and that of the two major Scandinavian softwood species Scots pine has higher extractive content than Norway spruce [7]. Since fresh pine sawdust show very different properties in the pelletizing process compared to stored pine sawdust and spruce sawdust it has been suggested that the extractive content is influential on pelletizing properties and pellet quality [8–10]. However, detailed studies have not been performed to quantify the influence of different process settings on pine sawdust with varying seasoning time and the effect of sawdust moisture and extractive content. In this study, the three factors sawdust moisture content, steam conditioning of the sawdust, and storage time of the raw material, were varied in a full factorial design with a total of 34 experiments to evaluate the influences on the pelletizing process and the pellet quality when producing biofuel wood pellets from pine sawdust. The objectives were (1) to determine the influences and interactions of these factors on pellet quality responses, (2) to quantify a required seasoning time for optimum pellet quality and (3) to develop a qualitative model for the binding mechanisms in pellets with different moisture and extractive contents. 2. Materials and methods 2.1. Biomaterials Scots pine (Pinus sylvestris L.) sawdust originating from latitude 64 °N in Sweden was delivered from a sawmill (SCA Timber AB in Holmsund, Sweden). About 18 tons of sawdust was transported and stacked on asphalt at the end of June at the Swedish national pilot plant Biofuel Technical Center (BTC), Swedish University of Agricultural Sciences (SLU), Umeå, Sweden. The pile (3 m high, 5 m wide at the bottom, and 10 m long) was stored for 160 days, during which extractive content and content of Klason lignin were monitored. At specific storage times (0, 46, 81, 117 and 160 days) batches were taken from the short end of the pile after removal of a protecting sawdust layer of about 2 m along the pile. The protecting sawdust layer was put back again after each sampling occasion. This was done to ensure that the batches were taken from the inside of the pile with a similar environment each storage time. 2.2. Experimental All pelletizing experiments were performed at the SLU BTC national pilot plant and the experimental set-up consisted of a horizontal batch drier, a combined mixer wagon and balance, conveyor systems, mill, steam treatment equipment, pelletizer
and pellet cooler. The full system is described in detail in [10]. The pelletizer (SPC PP300, Swedish Power Chippers, Borås, Sweden) with a capacity of 300 kg/h, was equipped with a fixed die (outer diameter 540 mm, press channel lengths of 55 mm) and two rotating press rollers (200 mm in diameter). The mill had a 4 mm sieve, and saturated steam was generated by a steam boiler to a temperature of 111.0 ± 0.76 °C. The steam conditioning of the sawdust material was carried out during 2 min just before reaching the pellet press. The pelletizer was controlled by a programmable logic controller (PLC) (Swedish Power Chippers, Borås, Sweden). Drying of the raw sawdust with approximately 50% moisture content was executed on a ventilated batch drier with a capacity of about 400 kg/batch and the temperature of the in going ventilating air was about 40 °C. The drying process was continued until the moisture content of the material was close to the design value, which was controlled by repeated moisture content measurements using a moisture balance (Sartorius MA30, Tillquist, Sweden). A drying cabinet (1600 800 700 mm) (Elvärmedetaljer, Sweden) calibrated to 105 ± 2 °C was used for oven drying of the samples collected from the pelletizing trials. Temperature calibrations were carried out using a thermometer (Technoterm 9300, Nordtec Instrument AB, Sweden) calibrated at an accredited laboratory (AREPA Mätteknik AB, Sweden). The size of the drying trays was 750 600 mm and the samples were weighed to 0.1 g accuracy with a balance (Mettler PE 16, instrument TEKNIK, Sweden). 2.3. Experimental design Pelletizing of the sawdust was executed in a full factorial design with three parameters: storage time at the five levels: 0, 46, 81, 117 and 160 days, moisture content at the three levels: 8.4 ± 0.48%, 10.7 ± 0.38% and 13.1 ± 0.38%, and steam treatment at the two levels: 2 and 6 kg/h. There were some difficulties to reach the intended design values of 8%, 11% and 14% for the moisture content, especially for the upper level, probably due to loss of moisture after the drying process was completed. For determination of experimental precision, three replicates of the experiments with moisture contents of 11% and storage time of 160 days were carried out for both steam levels, respectively. In all, the design for pelletizing consisted of 34 experiments in a randomized order except for storage time. About 75 kg of sawdust was milled and compressed into pellets in each experiment. The response factor parameters measured during each experiment were: bulk density and mechanical durability of produced pellets, the share of fines generated during the pelletizing process, and the energy consumption of the pelletizer recorded as the mean current during a 2 min measurement period. Each experiment lasted for about 30 min at a production rate of 154 ± 11 kg/h, where the first 20 min were used to stabilize the process and obtain a fairly constant die temperature. During the last 10 min of each experiment, pelletizer motor current was monitored and samples of 1 L milled sawdust, and 5 kg pellets were collected during two different measurement periods. All samples of sawdust and pellets were analysed for moisture content according to the CEN standard [11]. In Fig. 1 the correlation between sawdust moisture content and pellet moisture content for the 2 kg/h steam treatment is shown. The content of fines produced in the pelletizing process was quantified by manual sieving of an 8 L sample of pellets through a 3.15 mm sieve, and calculated as the percentage of the loss of the fine material to the total sample weight. Pellet bulk densities were determined using the CEN standard [12]. Mechanical durability of the pellets was measured by use of a pellet tester (Q-tester, Simon Heesen BV, Netherlands) according to the CEN standard [13]. The pellet length distribution was determined by subsequent dividing of the pellet sample until about 10 pellets remained followed by a length determination using a calliper.
111
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
60 °C) and acetone (90–10 v/v). The extraction time was 1 h and extractions were repeated twice for each sample [15]. The amount of Klason lignin was determined according to [16]. 2.4. Modelling and diagnostics
Fig. 1. Correlation between sawdust moisture content (%) and pellet moisture content (%) with the 2 kg/h steam conditioning.
The pelletizer motor current was recorded during the sampling of sawdust and pellets at a frequency of 1 Hz by the PLC using a data logger (Mitec, SatelLite-U (type 1), Mitec Instrument AB, Sweden). Motor current determination was restricted to pelletizing of stored sawdust only. Pelletizing of fresh sawdust gave rise to uneven pellet production due to feed layer losses (c.f. [14]), and thus, reliable pelletizer current data could not be gathered. The amount of extractives in the sawdust samples were determined by use of an extraction system (Universal Extraction System B-811 from Büchi Labortechnik AG, Flawil, Switzerland). The extraction solvent was a mixture of petroleum ether (bp 40–
All measured response factors in the factorial design were analysed by multiple linear regression (MLR) using the software MODDE 8.0.0.0 (Umetrics AB, Sweden). This software was also used in the design of the experiments. All values were centred before evaluation. Leave-one-out crossvalidation was used to calculate the residual for each validation round. These residuals (ft) between observed (y) and predicted responses (yp), i.e. ft = y yp, were ordered in a column (I 1) and used as diagnostics together with leverage of each observation. The different multivariate models were then evaluated using the coefficient of multiple determinations (Q2) and calculated as:
Q 2 ¼ 1 PRESS=ðyT yÞ where yTy is the total sums of squares in the data and PRESS is the leave-one-out cross-validation predictor sum of squares and calculated according to [17] as: 2
PRESS ¼ 1T ðf t ð1 diagðHÞÞ2 Þ where 1 is a column vector of I 1 ones; T denotes a transposed 2 vector or matrix; f t a column vector of I 1 squared residuals; diag(H) represents the leverage of each observation and is the I 1
Table 1 Summary of the design and experimental results. Analytical results are average values of the two samples collected during each run. Motor current values are the average of 1 Hz sampling during 2 min of pelletizing. Exp. name
N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 N14 N15 N16 N17 N18 N19 N20 N21 N22 N23 N24 N25 N26 N27 N28 N29 N30 N31 N32 N33 N34 a
Parameters
Responses
Storage (days)
Moisture (%)
Steam (kg/h)
Bulk density (kg/m3)
Durability (%)
Fines (%)
Current (A)
Extractives (%)
Klason lignin (%)
Pellet length (mm)
0 0 0 0 0 0 46 46 46 46 46 46 81 81 81 81 81 81 117 117 117 117 117 117 160 160 160 160 160 160 160 160 160 160
10.7 10.7 12.9 12.7 7.8 7.7 8.8 9.3 11.5 11.5 13.8 13.8 12.9 12.9 10.6 10.1 8.1 8.5 10.5 10.7 8.6 8.8 13.1 12.8 10.6 10.6 8.2 8.5 10.6 10.6 12.9 12.9 10.8 10.6
2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6 2 6
612 602 525 523 709 712 717 693 660 630 553 523 550 550 711 673 734 723 712 662 735 703 647 597 695 657 755 730 677 675 625 598 685 663
83.5 88.2 61.8 71.4 84.6 89.4 93.5 94.1 90.5 92.3 76.2 80.6 84.1 89.7 93.6 95.8 91.8 93.9 96.6 96.9 94.3 96.0 96.6 95.7 96.9 97.2 93.8 96.0 96.8 97.1 97.4 96.8 96.7 97.2
4.2 3.0 8.4 7.2 7.6 2.9 1.5 1.3 1.6 1.2 2.9 3.0 2.1 2.6 0.85 0.89 1.7 0.76 0.96 0.45 0.97 0.62 0.49 0.93 0.33 0.28 0.96 0.48 0.31 0.46 0.49 0.33 0.46 0.30
n.a.a n.a.a n.a.a n.a.a n.a.a n.a.a 29.4 28.4 26.5 26.5 25.2 25.1 25.1 26.0 26.0 25.1 29.0 28.6 27.1 26.3 27.4 26.6 24.7 24.4 27.0 26.9 29.6 27.8 27.4 26.8 25.2 24.7 27.1 26.0
3.44 3.17 4.04 4.03 3.81 3.79 3.06 2.91 2.89 2.87 2.56 3.05 2.50 2.19 2.15 2.29 2.22 2.32 2.21 2.16 2.29 2.16 2.20 1.88 2.06 1.91 2.05 1.81 1.66 1.74 1.74 1.69 1.61 1.57
24.0 25.0 24.6 24.2 24.1 23.4 24.7 23.6 24.6 25.4 25.3 25.9 25.4 25.5 25.3 25.5 25.7 25.5 24.8 25.3 24.7 24.9 25.1 25.4 25.8 25.8 24.9 25.3 27.2 26.9 26.8 26.6 26.3 26.5
26.0 28.5 8.9 22.0 22.4 24.7 23.7 27.0 24.0 28.3 21.3 22.1 24.8 13.2 26.8 24.7 25.4 21.3 23.6 23.8 21.7 28.9 28.9 31.6 26.6 32.7 18.0 23.8 34.0 30.9 33.9 33.6 31.2 35.9
n.a.: Not analyzed.
112
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
diagonal elements of the symmetric (I I) hat matrix H, that here is calculated as X(XTX)1XT with dimension I I containing the model parameters ordered in a X matrix. The Q2 value expresses how much of the variance in the response variable (y) that can be predicted and can at best be 1 and a value of 0 indicate no predictive capability at all. The number of factors used in the models was determined by optimization of Q2. 3. Results and discussion The experimental design and corresponding results are summarized in Table 1. In Table 2 the coefficients for the different factors in the regression analyses of the various responses are summarized and the following equations can be written for each response:
BD ¼ 660 99:4 Moi þ 32:9 Sto 11:5 Ste
Fig. 2. Bulk density (kg/m3) of pellets produced from sawdust with various moisture contents (%) at five different storage times (days) and a steam conditioning of 6 kg/h.
þ 16:0 ðMoi StoÞ 17:2 ðMoi MoiÞ 23:0 ðSto StoÞ Pellets produced from fresh sawdust had lower bulk density than pellets from sawdust that had been stored for 46 days or more. There was no obvious difference between bulk densities for pellets from sawdust with storage times longer than 46 days. No correlation between pellet length and bulk density was found (R2 = 0.04), which indicate that the eventual influence of the pellet length on bulk density is small in comparison to the effect from in particular the sawdust moisture content.
MD ¼ 94:7 5:5 Moi þ 8:6 Sto þ 1:2 Ste þ 7:5 ðMoi StoÞ 1:3 ðSto SteÞ 6:1 ðMoi MoiÞ 5:1 ðSto StoÞ Fi ¼ 0:78 þ 0:71 Moi 2:25 Sto 0:25 Ste þ 1:13 ðMoi MoiÞ þ 1:8 ðSto StoÞ PC ¼ 25:9 2:24 Moi þ 0:41 Sto 0:46 Ste
3.2. Mechanical durability
BD, MD, Fi and PC correspond to bulk density, mechanical durability, fines and press current, respectively. Moi, Sto and Ste are defined in Table 2. The standardized factor settings used in the calculations are 0.85 ± 0.16, 0.09 ± 0.12 and +0.69 ± 0.13 for the three levels of moisture content; 1, 0.425, +0.0125, +0.463 and +1 for the five levels of storage time; and 1 and +1 for the two levels of steam conditioning, respectively. 3.1. Bulk density The overall trend was a lowered bulk density with increasing moisture content within the interval of 8–14% moisture. From Table 2 it is shown that the most significant effect was obtained at low moisture contents and long storage times, which affected the bulk density in a positive manner. A Q2-value of 0.91 indicated an excellent prediction accuracy of the model. Moisture content was the parameter with the greatest impact on pellet bulk density (Fig. 2 and Table 2). There is an effect also from the steam treatment similar to the effect from the sawdust moisture content. However, it is much smaller and is probably due to the alteration of the moisture content of the material.
The analysis for mechanical durability (Table 2) showed a more complex pattern compared to that for bulk density. Storage time was found the most influential parameter and a long storage time resulted in high durability. However, no factor was dominant over the others and most interactions and squared terms were involved in the model building. The prediction ability of the model, Q2, was 0.85. The influence of moisture on the mechanical durability for various storage times is shown in Fig. 3. Four important conclusions was drawn from this result; (i) there was a large increase in mechanical durability with storage time, (ii) local maxima of mechanical durability within the studied range of moisture content were observed for all storage times, (iii) the durability maxima were shifted towards higher moisture contents the longer time the sawdust was stored and were found at 9.2%, 9.8%, 9.9%, 11.3% and 11.9%, respectively and (iv) the sensitivity for changes in moisture content on the mechanical durability decreased with storage time. The local maxima in mechanical durability were calculated from the 1st derivative of the polynomial regression line for each storage time.
Table 2 Summary of the coefficients for the different factors in the regression analysis of the various responses. Moi: moisture content, Sto: storage time, Ste: steam treatment, Q2: predicting power of the models.
Constant Moi Sto Ste Moi Sto Sto Ste Sto x Sto Moi Moi Q2 ns
Not significant.
Bulk density (kg/m3)
Durability (%)
Fines (%)
Current (A)
Extractives (%)
Klason lignin (%)
660 99.4 32.9 11.5 16.0
94.7 5.5 8.6 1.2 7.5 1.3 5.1 6.1 0.85
0.78 0.71 2.25 0.25ns
25.9 2.24 0.41ns 0.46
2.40 0.03ns 0.95 0.03ns
25.1 0.49 0.49ns 0.04ns
23.0 17.2ns 0.91
1.8 1.13ns 0.63
0.60
0.35
0.61ns
0.89
0.27
113
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
Fig. 5. Decrease in extractive content (%) with storage time (days).
Fig. 3. Mechanical durability (%) of pellets produced from sawdust with various moisture contents (%) at five different storage times (days) and a steam conditioning of 6 kg/h.
Fig. 4. Correlation between mechanical durability (%) and extractive content (%) at three levels of sawdust moisture contents (%).
A large increase in mechanical durability with decreasing extractive content was found (Fig. 4) and this effect was more pronounced at higher sawdust moisture levels. 3.3. Fines, energy consumption, extractive content and Klason lignin The regression analysis for the amount of fines collected after the pelletizing was very similar to that of the durability, except that the effects had the opposite direction. However, the Q2-value (0.61) was considerably lower. The analysis for the energy consumption of the pellet press was performed for the stored sawdust only and showed that the current was almost solely determined by the moisture content of the sawdust among the studied factors in this investigation, where high energy consumption was obtained at low moisture levels. The Q2-value of 0.60 indicated that the model was acceptable but not as good as for bulk density and mechanical durability. In general, extractives also act like lubricants in the die with low friction and low energy consumption as a consequence. This was also found for the fresh sawdust, which was indicated by significantly lower bulk density for pellets from fresh sawdust compared to pellets from sawdust stored for 46 days. However, due to the uneven pellet production for the fresh sawdust the unreliable pelletizer currents are not reported. For longer storage times a small increase in current with decreasing extractive content was observed, but the variation in current was too large for this effect to be significant.
In Fig. 5 the decrease of extractive content with storage time is shown and the model indicated that the storage time was the only factor that affected this response. As suggested earlier [4,9,10], the large increase in mechanical durability and decrease in the amount of fines with increasing storage time may be caused by a decrease in extractive content. To verify this, storage time was replaced by extractive content in the regression analysis for the two responses. This resulted in an increase of the Q2-value from 0.85 to 0.88 for mechanical durability and from 0.63 to 0.81 for the amount of fines (Table 3), indicating that the decrease in extractives during storage is the main reason to the effect. Klason lignin was measured to see whether the storage of the sawdust had increased the content of the lignin and if this might explain the changes in pellet quality. However, no predictive model was obtained for this response and only a small increase was found partly caused by the decrease in extractives. In addition, no correlation between the lignin content and the pellet quality was obtained. 3.4. Required seasoning time for optimum pellet quality The suggested quality criteria of pellets for non-industrial use is P600 kg/m3, P97.5% and 61.0% for bulk density, mechanical durability and fines, respectively [18]. From Figs. 2 and 3, and Table 1 it was found that quality requirements for bulk density and fines were fulfilled for all storage times longer than about 120 days. For mechanical durability, requirements could not be reached, but an optimum value of about 97% was obtained for the two longest storage times within the range of moisture content between 11 and 13%. This limitation was probably caused by the experimental setting of the pellet press, since further storage of the sawdust did not give any significant increase in the mechanical durability.
Table 3 Summary of the coefficients for the different factors in the regression analysis of mechanical durability and fines when the factor storage time is replaced by extractive content. Moi: moisture content, Ext: extractive content, Ste: steam treatment, Q2: predicting power of the models.
Constant Moi Ext Ste Moi Ext Ext Ste Moi Moi Ext Ext Q2 ns
Not significant.
Durability (%)
Fines (%)
91.5 6.7 10.6 1.4 7.5 1.7 5.8 3.7 0.88
1.98 0.51 3.23 0.36 0.58 1.87 0.38ns 0.81
114
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115
3.5. Qualitative model Raw material moisture content was the single most important parameter for bulk density and energy consumption. As reported earlier [4,10,19], this may be explained by moisture altering the friction when biomaterial is compacted and pressed through the die channels, and consequently also the back pressure. To some extent, storage time also affected these response parameters, caused by the decrease in the content of extractives that act like lubricants in the pelletizing process. A much more complex mechanism was found for mechanical durability and amount of fines where all factors and most interactions and squared terms contributed to the models. Back discussed in a review [2] the bonding mechanism in the manufacture of hardboard from lingo-cellulosic fibres and he concluded that most of the inter-fibre bonding was due to secondary bonds, in particular hydrogen bonds but also to some extent van der Waals forces. The pressure used in the hardboard process is of the order of 5–15 MPa [20–22] and considering that the pressure used in the pelletizing process is much higher, 200–450 MPa, it is reasonable to assume a similar bonding mechanism in fuel pellet manufacture even if other mechanisms may contribute. The dominant hydrogen bonding group in the ligno-cellulosic material is the hydroxyl group, although the carbonyl, the carboxyl and the ether groups are also involved in the bonding. Even the very weak CAH O bond, bond strength 2.5 kJ mol1, may contribute to the bonding due to its multiplicity [23]. The bonding strengths for the hydrogen bond are of the order of 2.5– 120 kJ mol1 [23] and the bond lengths vary between 0.12 and 0.32 nm depending on the bond strength, where the shortest length correspond to the strongest bond, i.e. the FAH F bond. This means that by compressing the sawdust to reach a distance of less than 0.32 nm between the binding sites on the particle surfaces, hydrogen bonding may occur. If the distance between the active sites is too long no hydrogen bonding can take place. However, dipole molecules, like water, can act like bridges between these sites, with an increase in bonding strength as a result. This might be possible if the distance between the particles surfaces are in the range 0.5–0.74 nm, which correspond to twice the hydrogen bond length plus the OAH length in water (0.1 nm) assuming weak hydrogen bond strength (<20 kJ mol1) with bond lengths of 0.2–0.32 nm. Oleophilic low molecular weight material in the extractives, i.e. fatty acids and waxes, may interfere with the bonding mechanism by blocking and consequently reducing the number of binding sites on the surfaces [2,24] with low bonding strength between the particles as a result. This is due to the high adhesive force between these compounds and the wood surface, and the low cohesive strength of the oleophilic interface. The mechanism discussed above may explain the increase in mechanical durability with storage time as is shown in Fig. 3, since the extractive content is decreased during storage. The model may also explain the shift of the optimum in mechanical durability as a function of storage time; extractives block a number of binding sites in fresh sawdust, and thus, few hydrogen bindings between particles occur and only a small number of water molecules can act like bridges, resulting in low durability with a maximum at low moisture content. During storage the number of sites increases with increased mechanical durability as a result. In addition, more water molecules may be involved in the binding mechanism giving a maximum durability at higher moisture content. If the moisture content is too low, gaps where water molecules can bridge are not filled up and binding strength is lowered. On the other hand, if the moisture content is too high, water will attack the hydrogen bonds between the particles, resulting in decreased durability [2,4]. These effects will have a greater impact in fresh sawdust where the relative number of active binding sites are smaller, resulting in a larger
sensitivity to changes in moisture content on mechanical durability. 4. Conclusions – Moisture content was the dominant factor for bulk density and pelletizer motor current, which is explained by the alteration of the friction through the die due to the lubricating property of water. – For the mechanical durability and the amount of fines more complex models were obtained, where all factors and most interactions and squared terms were significant. A qualitative model for the binding mechanism suggests that extractives block binding sites for hydrogen bonding to occur and that water is actively involved in the bonding. – An optimum pellet quality was obtained at storage times longer than 120 days and within a range of moisture content of 11– 13%.
Acknowledgements We thank SCA Timber AB, Holmsund Sweden for the delivered sawdust. The Swedish Energy Agency, the Swedish Pellet Association (PiR), Swedish University of Agricultural Sciences and the Project Bio4Energy are thanked for financial support. Carina Jonsson is greatly acknowledged for carrying out the chemical analyses of the various biomaterials. References [1] Rumpf H. The strength of granules and agglomerates. In: Knepper WA, editor. Agglomeration. Interscience; 1962. p. 379–418. [2] Back EL. The bonding mechanism in hardboard manufacture – review report. Holzforschung 1987;41:247–58. [3] Mani S, Tabil LG, Sokhansanj S. An overview of compaction of biomassgrinds. Powder Handling Process 2003;15:160–7. [4] Nielsen NPK, Gardner DJ, Poulsen T, Felby C. Importance of temperature, moisture content, and species for the conversion process of wood residues into fuel pellets. Wood Fiber Sci 2009;41:414–25. [5] Kaliyan N. Morey RV. Factors affecting strength and durability of densified biomass products. Biomass Bioenergy 2009;33:337–59. [6] Stelte W, Holm JK, Sanadi AR, Barsberg S, Ahrenfeldt J, Henriksen UB. A study of bonding and failure mechanisms in fuel pellets from different biomass resources. Biomass Bioenergy 2011;35:910–8. [7] Assarsson A, Åkerlund G. Studies on wood resin, especially the change in chemical composition during seasoning of the wood. Part 4. The composition of the petroleum ether soluble nonvolatile extractives from fresh spruce, pine, birch and aspen wood. Svensk Papperstidn Nord Cellulosa 1966;69:517–25. [8] Bradfield J, Levi MP. Effect of species and wood bark ratio on pelleting of southern woods. Forest Prod J 1984;34:61–3. [9] Nielsen NPK, Gardner DJ, Felby C. Effect of extractives and storage on the pelletizing process of sawdust. Fuel 2010;89:94–8. [10] Samuelsson R, Thyrel M, Sjöström M, Lestander TA. Effect of biomaterial characteristics on pelletizing properties and biofuel pellet quality. Fuel Process Technol 2009;90:1129–34. [11] EN 14774–2:2009. Solid biofuels – methods for determination of moisture content – oven dry method. Total moisture: simplified method. [12] EN 15103:2009. Solid biofuels – determination of bulk density. [13] EN 15210–1:2009. Solid biofuels – determination of mechanical durability of pellets and briquettes. Pellets. [14] Larsson SH, Rudolfsson M, Thyrel M, Örberg H, Kalén G, Wallin M, et al. Temperature controlled feed layer formation in biofuel pellet production. Fuel 2012;94:81–5. [15] Arshadi M, Gref R. Emission of volatile organic compounds from softwood pellets during storage. Forest Prod J 2005;55:132–5. [16] Theander O, Westerlund EA. Studies on dietary fiber. 3. Improved procedures for analysis of dietary fiber. J Agr Food Chem 1986;34:330–6. [17] Myers RH. Classical and modern regression with applications. Boston, USA: Duxbury Press; 1986. [18] EN 14961–1:2010. Solid biofuels – fuel specifications and classes – Part 1: General requirements. [19] Lestander TA, Johnsson B, Grothage M. NIR techniques create added values for the pellet and biofuel industry. Bioresource Technol 2009;100:1589–94. [20] Angles NN, Montane D, Ferrando F, Salvadó J. Properties of binderless panels from thermomechanical aqueous vapor exploded residual softwood. In:
R. Samuelsson et al. / Applied Energy 99 (2012) 109–115 Kopetz H, Weber T, Palz W, Chartier P, Ferrero GL, editors. Biomass for energy and industry; 1998, p. 435–8. [21] Salvadó J. Velásquez JA., Ferrando F. Binderless fiberboard from steam exploded Miscantus Sinensis: optimization of pressing and pretreatment conditions. Wood Sci Technol 2003;37:279–86. [22] Okuda N, Sato M. Manufacture and mechanical properties of binderless boards from kenaf core. J Wood Sci 2004;50:53–61.
115
[23] Walker J. Basic wood chemistry and cell wall ultrastructure. In: Walker J, editor. Primary wood processing, principle and practice. Springer; 2006. p. 23–30. [24] Stehr M, Johansson I. Weak boundary layers on wood surfaces. J Adhes Sci Technol 2000;14:1211–24.