Temperatures of Wood Particles in a Hot Sand Bed Fluidized by

the conditions typical of fast pyrolysis (bed temperatures of 800-1100 K and particle sizes of .... 10-1 W/m K, whereas that perpendicular to the fibe...
0 downloads 0 Views 156KB Size
Energy & Fuels 2003, 17, 247-254

247

Temperatures of Wood Particles in a Hot Sand Bed Fluidized by Nitrogen Colomba Di Blasi* and Carmen Branca Dipartimento di Ingegneria Chimica, Universita´ di Napoli “Federico II” P.le V. Tecchio, 80125 Napoli, Italy Received July 3, 2002

The thermal history undergone by cylindrical beech wood particles, injected in a sand bed fluidized by nitrogen has been recorded. Experiments have been carried out by varying particle diameter (d ) 2-10 mm) and bed temperature (Tr ) 712-1107 K). The rate of volatile release becomes significant for temperatures above 625-650 K and is always completed for temperatures below 675-825 K. Devolatilization causes a strong reduction in the heating rate which, at the particle center and for the most intense reaction activity, is comprised between 0 and 25 K/s. For the conditions typical of fast pyrolysis (bed temperatures of 800-1100 K and particle sizes of 2-6 mm), the yields of char are 10-18% and the devolatilization times (corresponding to a conversion of 95%) 18-45s. Furthermore, in qualitative agreement with previous analyses carried out for coal particles, these are well predicted by an empirical power-law relation: tv ) 0.8e1525/Trd1.2 s, over the entire range of experimental conditions examined.

Introduction The pyrolytic behavior of thick wood has been studied extensively for applications in the fields of fire safety science and fixed-bed thermochemical conversion of biomass (for instance, see refs 1-8). Wood heating has been accomplished through thermal radiation and several aspects of the process, such as details of the temperature and pressure profiles, conversion times, shrinkage and product yields, and different geometrical configurations, have been examined. The results of these studies are not applicable for the fast pyrolysis of wood carried out in fluid-bed reactors, given the much smaller particle sizes and the different mechanisms responsible for particle heating. On the other hand, a recent study,9 though using a fluid bed, is focused on high tempera* Corresponding author. Tel: 39-081-7682232. Fax: 39-081-2391800. E-mail: [email protected]. (1) Lee, C. K.; Chaiken, R. F.; Singer, J. M. Charring Pyrolysis of wood in Fires by Laser Simulation. In Sixteenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, 1976; pp 1459-1470. (2) Pyle, D. L.; Zaror, C. A. Heat Transfer and Kinetics in the LowTemperature Pyrolysis of Solids. Chem. Eng. Sci. 1984, 19, 147-158. (3) Chan, W. R.; Kelbon, M.; Krieger-Brockett, B. Single-Particle Biomass Pyrolysis: Correlation of Reaction Products with Process Conditions. Ind. Eng. Chem. Res. 1988, 27, 2261-2275. (4) Koufopanos, C. A.; Papayannakos, N.; Maschio G.; Lucchesi, A. Modelling of the Pyrolysis of Biomass Particles. Studies on Kinetics, Thermal and Heat Transfer Effects. Can. J. Chem. Eng. 1991, 69, 907915. (5) Bilbao, R.; Millera, A.; Murillo, M. B. Temperature Profiles and Weight Loss in the Thermal Decomposition of Large Spherical Wood Particles. Ind. Eng. Chem. Res. 1993, 32, 1811-1817. (6) Gronli, M. G.; Melaaen, M. Mathematical Model for Wood PyrolysissComparison of Experimental Measurements with Model Predictions. Energy Fuels 2000, 14, 791-790. (7) Di Blasi, C.; Gonzalez Hernandez, E.; Santoro, A. Radiative Pyrolysis of Single Moist Wood Particles. Ind. Eng. Chem. Res. 2000, 39, 873-882. (8) Di Blasi, C.; Branca, C.; Santoro, A.; Gonzalez Hernandez, E. Pyrolytic Behaviour and Products of Some Wood Varieties. Combust. Flame 2001, 124, 165-177.

tures (above 923 K) and thick particles (above 7 mm). For fast pyrolysis carried out by means of fluidized bed reactors, particle sizes are always below 10 mm (typical values are 0.1-6 mm10) and, for liquid fuel (oil) production, bed temperatures around 800 K are required.10 Fuel heating takes place through particle/particle contact, convection, and radiation.11 Given the importance of the process for industrial applications, numerous investigations have been carried out at a laboratory/pilot scale in order to determine the optimal conditions for liquid or gas production and to identify the key parameters for tar vapor degradation. Yields and composition of the products (char, liquids, and gas) of wood pyrolysis12 have been determined as functions of the bed temperatures.13-18 Models have also (9) De Diego, L. F.; Garcia-Labiano, F.; Abad, A.; Gayan, P.; Adanez, J. Modelling of the Devolatilization of Nonspherical Wet Pine Wood Particles in Fluidized Bed. Ind. Eng. Chem. Res. 2002, 41, 3642-3650. (10) Bridgwater, A. V. Principles and Practice of Biomass Fast Pyrolysis Processes for Liquids. J. Anal. Appl. Pyrolysis 1999, 51, 3-22. (11) Agarwal, P. K. Transport Phenomena in Multiparticle Systemss IV. Heat Transfer to a Large Freely Moving Particle in Gas Fluidized Bed of Smaller Particles. Chem. Eng. Sci. 1991, 46, 1115-1127. (12) Scott, D. S.; Majerski, P.; Piskorz, J.; Radlein, D. A Second Look at Fast Pyrolysis of BiomasssThe RTI Process. J. Anal. Appl. Pyrolysis 1999, 51, 23-37. (13) Scott, D. S.; Piskorz, J. The Continuous Flash Pyrolysis of Biomass. Can. J. Chem. Eng. 1984, 62, 404-412. (14) Scott, D. S.; Piskorz, J.; Bergougnou, M. A.; Graham, R.; Overend, R. P. The Role of Temperature in the Fast Pyrolysis of Cellulose and Wood. Ind. Eng. Chem. Res. 1988, 27, 8-15. (15) Beaumont, O.; Schwob, Y. Influence of Physical and Chemical Parameters on Wood Pyrolysis. Ind. Eng. Chem. Res. 1984, 23, 637641. (16) Agblevor, F. A.; Besler, S.; Wiselogel, A. E. Fast Pyrolysis of Stored Biomas Feedstocks. Energy Fuels 1995, 9, 635-638. (17) Horne, P. A.; Williams, P. T. Influence of Temperature on the Products from the Flash Pyrolysis of Biomass Fuel 1996, 75, 10511059. (18) Wehlte, S.; Meier, D.; Moltran, J.; Faix, O. The Impact of Wood Preservatives on the Flash Pyrolysis of Biomass. In Developments in Thermochemical Biomass Conversion; Bridgwater A. V., Boocock, D. G. B., Eds.; Blackie A& P: 1997; pp 206-229.

10.1021/ef020146e CCC: $25.00 © 2003 American Chemical Society Published on Web 12/21/2002

248 Energy & Fuels, Vol. 17, No. 1, 2003

Figure 1. Schematic of the fluid-bed reactor (A) and the sample holder (B): (1) fluidizing gas, (2) reactor, (3) reactor closing cap, (4) heating gas, (5) sample and thermocouple support, (6) wood sample, (7) furnace, (8) controller, (9) acquisition data set, (10) demister.

been developed,19,20 which give quantitative predictions of the three classes of products. However, measurements of particle temperatures and conversion times are not available for a better quantitative understanding of the process and to provide the basis for further development and validation of mathematical models. This study provides extensive measurements of the thermal response of small wood particles injected in a fluidized sand bed, as particle size and bed temperature are varied. The thermal history undergone by the particles is used to evaluate heating rates, reaction temperatures and devolatilization times. The charred particles, collected at the conclusion of the decomposition process, are weighed and examined to obtain information about the yields of char and the extent of shrinkage and/or structural changes. Furthermore, the parameters are determined of an empirical correlation similar to those applied for the devolatilization times of coal particles as functions of operating parameters (see, for instance, refs 21-24). Experimental Section Beech wood (chemical composition:7,8 78% holocellulose (hemicellulose and cellulose components), 20% lignin, 2% extractives, and 0.5% ash) was used for the experiments. These were performed in a laboratory scale system (Figure 1A) consisting of a fluidized bed (stainless steel reactor with 0.063 m internal diameter and 0.45 m length), electrically heated by a furnace. Nitrogen, fed through a jacket (internal diameter (19) Di Blasi, C. Modelling the Fast Pyrolysis of Cellulosic Particles in Fluid-Bed Reactors. Chem. Eng. Sci. 2000, 55, 5999-6013. (20) Di Blasi, C. Modelling Intra- and Extra-Particle Processes of wood fast pyrolysis. AIChE J. 2002, 48, 2386-2397. (21) Agarwal, P. K.; La Nauze, R. D. Transfer Processes Local to the Coal Particle: A Review of Drying, Devolatilization and Mass Transfer in Fluidized Bed Combustion. Chem. Eng. Res. Des. 1989, 67, 457-480. (22) Stubington, J. F.; Huang, G.; Scaroni, A. W. Devolatilization Times of mm-Size Coal Particles. Fuel 1991, 70, 1105-1108. (23) Lufei, J.; Becker, H. A.; Code, R. K. Devolatilization and Char Burning of Coal Particles in a Fluidized Bed Combustor. Can. J. Chem. Eng. 1993, 71, 10-19. (24) Ross. D. P.; Heindereich, C. A.; Zhang, D. K. Devolatilization Times of Coal Particles in a Fluidised Bed. Fuel 2000, 79, 873-883.

Di Blasi and Branca 0.089 m) at the reactor top, was also preheated and distributed by a sintered metal plate supporting the bed. The bed material was calcined sand (at 1100K for 3 h) sieved to 180-250 µm size range, with a static bed of height 0.068 m and density 1650 kg/m3, and a minimum fluidization velocity at atmospheric pressure of 0.036 m/s. The volumetric gas flow rate was varied with the bed temperature so as to achieve in all cases a superficial velocity about 8 times higher than that at minimum fluidization conditions. Temperature profiles along the reactor axis were measured by seven thermocouples (chromel-alumel type, 500 µm diameter), with their tips exiting from a protective steel tube, at chosen distances from the flow distributor. The expanded bed (about 0.075 m) was isothermal at a temperature determined by a proper set point of the furnace (PID controller), but gradients in the upper part were high. However, this was not a drawback given that only the single particle characteristics were of interest. Once the desired thermal conditions were achieved, the reactor closing cap was substituted by another one equipped with a thin steel rod acting as a support for the thermocouple and the sample. In this way, the particle was suspended in the fluidized bed, with the center positioned, along the reactor axis, at a distance of 0.035 m from the flow distributor. Beech wood particles were prepared in the shape of cylinders (diameters comprised between 2 and 10 mm) starting from cubic rods cut from logs in the direction parallel to the fibers. The length of the cylindrical particle was determined so as the characteristic thermal size could be made coincident with the diameter. This is important because wood is an anisotropic medium25,26 with higher values of thermal conductivity and permeability to gas flow along the fibers. In particular, for beech wood,27 the thermal conductivity along fibers λp is 3.49 × 10-1 W/m K, whereas that perpendicular to the fibers λc is 2.09 × 10-1 W/m K. The ratio between the characteristic heating (conduction) times28 along (tp) and across (tc) the wood fibers, with reference to thermal conductivities of virgin wood, can be expressed as

R)

tp L2/λp L2 ) 2 ) 0.6 2 tc d /λ d

(1)

c

where d and L are the particle diameter and length, respectively. For d ) 10 mm, the highest diameter examined in this study, a value of L ) 20 mm already results in tp about 2.4 times longer than tc. Hence, all the tests were made with 20 mm long cylinders. Once prepared, a 0.5 mm hole was drilled from top to the center of the particles which were then predried (8-10 h at 373 K). The particle temperature T was measured using a 0.5 mm sheated type K thermocouple. As already outlined, the thermocouple was supported by a thin steel rood connected to the reactor closing cap. To minimize heat conduction across its length, the thermocouple (and the support) was insulated by means of a ceramic tube and refractory cement. The particle was caged by a 300 µm Nichrome wire (tied to the steel rod), to permit a direct contact with the bubbling bed (Figure 1B). In some cases, the temperature was also measured of the gas adjacent the external surface of the particle. More specifically, a thermocouple was positioned either at the median position (25) Di Blasi, C. Transition between Regimes in the Degradation of Thermoplastic Polymers. Polym. Degrad. Stab. 1999, 64, 359-367. (26) Di Blasi, C. Heat, Momentum and Mass Transfer through a Shrinking Biomass Particle Exposed to Thermal Radiation. Chem. Eng. Sci. 1996, 51, 1121-1132. (27) Gronli, M. G. A Theoretical and Experimental Study of the Thermal Degradation of Biomass. Ph.D. Thesis, NTNU, Trondheim, Norway, 1996. (28) Di Blasi, C. Physicochemical Processes Occurring inside a Degrading Two-Dimensional Anisotropic Porous Medium. Int. J. Heat Mass Transfer 1998, 41, 4139-4150.

Wood Particles in a Hot Sand Bed

Figure 2. Time profiles of the temperature at the particle center T and the corresponding time derivative h and temperatures along the lateral (Ts) and top (Tp) surfaces for d ) 4 mm and Tr ) 807 K (definitions of the characteristic parameters are included). along the lateral surface (Ts) or just above the top surface (Tp). This additional thermocouple used the same support and insulation of that for the particle center. Preliminary tests carried out for d ) 5 mm (L ) 5, 10, 20, and 30 mm) and d ) 10 mm (L ) 10, 20, and 30 mm) (bed temperature equal to 807 K) also confirmed that the thermal history undergone by the particle center is not affected by the cylinder length given L g 20 mm. After the temperature of the particle center attained the bed value, the furnace was kept at the selected set point for at least 30 min, then the power was turned off. The sample was left under a continuous nitrogen flow until the temperature lowered to about 450 K in order to determine yield and characteristics of char. Two sets of experiments were conducted to investigate the effects of the particle size (cylinder diameter) and bed temperature. The first set was carried out for bed temperatures of 807, 950, 1020, and 1107 K by varying the particle diameter between 2 and 10 mm. The second set was made for particle diameters of 4 and 8 mm by varying the bed temperatures in the range 712-1107 K. Each test was made at least in triplicate, showing acceptable reproducibility. Temperature at the particle center and yields of char were measured. The char particles were also examined to estimate the extent of shrinkage and/or structural failure. Furthermore, to facilitate this task, some tests (d comprised between 2 and 10 mm and a bed temperature of 807 K) were repeated with parallelepipeds. In this case, the particle temperature was not measured, thus avoiding the possible weakening of the wood structure caused by the thermocouple insertion.

Results Definition of Devolatilization Parameters and Process Dynamics. An example of the dynamics of the temperatures at the particle center T and the corresponding time derivative (heating rate) h is shown in Figure 2 for a bed temperature of 807 K and a diameter of 4 mm. Figure 2 also reports the temperatures registered along the lateral (Ts) and the top (Tp) surface of the particle. It can be observed that the time lag between the particle center and the surface temperatures, especially Ts, is very high, indicating heat transfer control. For very short times (below 4 s), both Ts and Tp present the same rate of increase. Then, while the former rapidly approaches an asymptotic value, the latter shows more complex dynamics caused by the flow of volatile products leaving the degrading wood particle. In fact, the top surface thermocouple is wrapped up by

Energy & Fuels, Vol. 17, No. 1, 2003 249

these products, whose prevailing velocity is along the cylinder axis (parallel to the wood fibers).26 The sudden decrease and the wide region of nearly constant temperature (about 600-625 K) for times of 5-25 s can be associated with the beginning and the occurrence of wood pyrolysis. Moreover, the temperature dynamics are likely to be the result of (a) degradation temperatures significantly lower than the bed temperature and (b) lack of thermal equilibrium between the volatile products and the solid char (presumably at a temperature higher than those of the reaction zone). It is wellknown that wood degradation, at least for thermogravimetric conditions, takes place within a relatively narrow range of temperatures29 (500-750 K), whereas the temperature of the charred region is determined by the external heat source.19,20 On the other hand, the high devolatilization rates, the relatively small size of the charred layer and the formation of cracks are factors opposing the attainment of a thermal equilibrium between the volatile products and the char. Several points of the variables h and T (heating rate and temperature of the particle center, respectively) can be used to characterize the thermal response of the particle. Those selected here are the local maxima h1 and h4, the local minimum h3, and a point of high variation h2 of the heating rate (and the corresponding temperature and times), the maximum temperature T5 (with the corresponding time), and the time t6 when the heating rate goes to zero and the particle temperature attains the bed value. The maximum heating rate at the particle center h1 is attained when degradation is localized along a superficial layer, as confirmed by the region of low Tp values (high devolatilization rates). Then, convective cooling, caused by the release of the hot volatile products, the global endothermicity of the process and the formation of a low-thermal conductivity char 19,20,25,29-31 hinder inward heat transfer so that the rate of temperature rise experiences a continuous decrease. Due to spatial gradients, when the devolatilization rate attains a high value, the center of the particle is still relatively cold (T1 ) 400 K). For the inner core of the particle, given the reduced heat transfer rates in the final stage of wood conversion, the endothermic degradation of holocellulose is completed before the slower and exothermic degradation of lignin, in accordance with the degradation dynamics of wood for the conditions of thermal analysis (for instance, see refs 32 and 33). Indeed, the beginning of wood conversion (endothermic degradation of holocellulose) at the particle center can be associated with a sudden change of slope in h (h2) for a temperature T2 of about 650 K, which is followed by the attainment of a local (29) Di Blasi, C.; Branca, C. Kinetics of Primary Product Formation from Wood Pyrolysis. Ind. Eng. Chem. Res. 2001, 40, 5547-5556. (30) Narayan, R.; Antal, M. J., Thermal Lag, Fusion, and the Compensation Effect during Biomass Pyrolysis. Ind. Eng. Chem. Res. 1996, 35, 1711-1721. (31) Di Blasi, C. Kinetic and Heat Transfer Control in the Slow and Flash Pyrolysis of Solids Ind. Eng. Chem. Res. 1996, 35, 37-47. (32) Shafizadeh, F. Pyrolytic Reactions and Products of Biomass. In Fundamentals of Biomass Thermochemical Conversion; Overend, R. P., Milne, T. A., Mudge, L. K., Eds.; Elsevier: London, 1985; pp 183-217. (33) Antal, M. J.; Varhegyi, G. Cellulose Pyrolysis Kinetics: The Current State of Knowledge. Ind. Eng. Chem. Res. 1995, 34, 703717.

250

Energy & Fuels, Vol. 17, No. 1, 2003

minimum h3 of about 6 K/s (and a temperature of about 697 K). Given that holocellulose contributes for about 78% in the chemical composition of beech wood, its initial degradation temperature, as reported by the centerline thermocouple, has been considered in several cases7,8 as being representative of the pyrolysis temperature. The subsequent exothermic degradation of lignin and the termination of the devolatilization process give rise to a rapid temperature rise (h4 ) 21 K/s and T4 ) 760 K) and a barely visible maximum in the temperature (T5 ) 810 K). For the corresponding time (t5 ) 32 s), the conversion process is practically terminated, but the time needed for the particle temperature to equal to the bed value is significantly longer (t6 ) 50 s). In reality, as a precise evaluation of this time is difficult because of small temperature oscillations often associated with structural changes, it will not be examined in the following. Though the temperature recorded by the external thermocouples cannot be directly applied to evaluate the heating rate at the particle surface, the information concerning the beginning of the degradation process (t ) 5 s) can be of use for such a purpose. Assuming an initial degradation temperature of about 650 K and taking into account that a possible maximum should coincide with the bed temperature (807 K), the (average) heating rate established for a thin superficial layer can also be estimated as roughly 70-100 K/s. The temperature profile of the particle center presents the same features, in particular the existence of a plateau followed by a rapid rise in the heating rate, observed for thick cylinders or spheres uniformly heated along the external surface (for instance, see refs 4, 5, 7, 8, 34, and 35). In this case, the temperature history measured at several spatial locations showed the existence of two different dynamics, corresponding to the degradation of the external layer, dominated by heat transfer effects, and the inner core of the particle, characterized by the sequential degradation of the main components. Reaction Temperatures and Heating Rates. The effects of the particle size (d ) 2-10 mm) on the thermal response of wood, for a bed temperature of 807 K, are shown through the temperature profiles of Figure 3. Their shape remains qualitatively the same, however, as expected, the characteristic times become successively longer as d is increased. As already anticipated, for sufficiently thick particles, the attainment of a maximum temperature higher than the bed temperature is observed, as a consequence of the higher mass of lignin degrading in the final stage of the process. Figures 4-6 report the characteristic heating times t1-t5, reaction temperatures T1-T4, and heating rates h1-h4, previously introduced (Figure 2), as functions of the particle diameter for Tr ) 807 K. The times t1-t5 present a power law dependence on the particle diameter in accordance with analyses conducted for coal particles.24 Owing to a successively stronger internal heat transfer resistance, the center temperature indica(34) Roberts, A. F.; Clough, G. Thermal Decomposition of Wood in an Inert Atmosphere. Ninth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, 1963; pp 158-166. (35) Tinney, E. R. The Combustion of Wooden Dowels in Heated Air. Tenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, 1965; pp 925-930.

Di Blasi and Branca

Figure 3. Time profiles of temperatures at the particle center for several particle sizes and Tr ) 807 K.

Figure 4. Characteristic heating times t1-t5 at the particle center as functions of the particle diameter for Tr ) 807 K.

Figure 5. Characteristic temperatures T1-T4 at the particle center as functions of the particle diameter for Tr ) 807 K.

tive of the beginning of significant reaction activity at the surface T1 decreases (500-375 K) and the corresponding time t1 increases (7-20 s) as the particle size is varied from 2 to 10 mm. It is also worth noting that the external heat transfer coefficient is strongly reduced as the particle size is increased.19 On the other hand, the effects of internal and external heat transfer resistance are barely visible on the reaction temperatures T2 and T3, with values varying in the range 650-638 and 717-680 K, respectively. As already well-known from kinetic analysis29,33 and confirmed by detailed process simulation,19,20 this is due to the narrow range of temperatures where lignocellulosic fuels degrade. Hence, the external (bed) temperatures, when higher than those of wood degradation, essentially affect char

Wood Particles in a Hot Sand Bed

Energy & Fuels, Vol. 17, No. 1, 2003 251

Figure 6. Characteristic heating rates h1-h4 at the particle center as functions of the particle diameter for Tr ) 807 K.

Figure 8. Characteristic heating times t1-t4 at the particle center as functions of the bed temperature for d ) 4 mm.

Figure 7. Time profiles of temperatures at the particle center for several bed temperatures and d ) 4 mm.

Figure 9. Characteristic temperatures T1-T4 at the particle center as functions of the bed temperature for d ) 4 mm.

heating, after the degradation process has already been completed. The temperature T4 is practically constant (760 K), probably because the increased effects of exothermic lignin degradation (higher sample mass) are counteracted by an increased size of the char thickness. For all the particle sizes examined, the heating rates at the particle center are functions of the conversion level and, apart from the final stage of the process (h4), they become successively slower as d is increased. The maximum is attained during the inert heating stage either of wood (h1 for d < 6 mm) or char (h4 for d > 6 mm). For the conditions considered here, they are on the order of 80-25 K/s. The heating rates h3, corresponding to the fastest decomposition rate at the particle center, are roughly comprised between 18 and 0K/s. Figure 7 reports the profiles of the temperature at the particle center for several bed temperatures and d ) 4 mm. The shape of the temperature profile presents the features already discussed for the case of Figure 2. In general, the characteristic process times become successively shorter together with the duration of the plateau, as the heating conditions are made successively more severe. Figures 8-10 show the characteristic heating times t1-t4, the reaction temperatures T1-T4 and the heating rates h1-h4, previously introduced (Figure 2), as functions of the bed temperature for d ) 4 mm (the time t5 is not plotted because it cannot be determined for temperatures below 725 or above 900 K). The characteristic heating times are highly affected by the bed

Figure 10. Characteristic heating rates h1-h4 at the particle center as functions of the bed temperature for d ) 4 mm.

temperature only for values below 800 K, then they tend to almost constant values. Again, this is due to the narrow range where wood degrades, as also confirmed by the weak dependence of temperatures T2 (630-655 K) and T3 (680-707 K) on the bed temperature. As expected, the heating of the inert char particles is significantly affected by the bed temperature, as indicated by T4 varying from about 700 to 845 K. The maximum heating rate at the particle center is attained during the heating stage of inert wood (h1 for Tr < 850 K) or char (h4 for Tr > 850 K). Only for the latter case are quite high values (40-110 K/s) shown, whereas those attained during conversion are always relatively low (h3 below 25 K/s). Finally, spatial temperature gradients appear to be nonnegligible, as confirmed by

252

Energy & Fuels, Vol. 17, No. 1, 2003

Di Blasi and Branca Table 1. Shrinkage, Expressed as Percent of the Initial Length, along the Direction Parallel (Sp) and Perpendicular (Sc) to the Wood Fibers for Several Particle (parallelepiped) Thicknesses (Tr ) 807 K) (Values Averaged over Four Tests)

Figure 11. Char yields, expressed as percent of the initial dry wood mass, as functions of the particle diameter (Tr ) 807 K and Tr ) 1107 K) and the bed temperature (d ) 4 mm and d ) 8 mm).

T1 values (390-445 K) significantly lower than the bed temperatures. Char Yields and Particle Shrinkage/Failure. Figure 11 reports the char yields Yc, expressed as percent of the initial dry wood mass, as functions of the bed temperature (d ) 4 and d ) 8 mm) and the particle diameter (bed temperature of 807 and 1107 K). In qualitative agreement with previous measurements, carried out for thick wood radiatively heated3,7,8 and continuous fluid-bed pyrolysis of small wood particles,12-18 and numerical simulations,19,20 the yields of char decrease as the bed temperature becomes successively higher and/or the particle size smaller. These trends are wellknown to be the consequence of both primary and secondary reactions of char formation becoming successively less favored. Indeed, both the actual reaction temperatures (see Figures 5 and 9) slightly increase and the intraparticle residence time of volatiles become shorter (in consequence of higher devolatilization rates and gas velocities, and shorter sizes of the charred region). From the quantitative point of view, the char yields are significantly dependent on the bed temperature only for values of these below 850 K and then approach constant values (values comprised between 30 and 10%). Also, for the higher temperatures examined (above 800 K), the influences of the particle size are relatively small (variations comprised between 18 and 8%). Good quantitative agreement is obtained with the results of continuous reactors, in particular those reported for some hardwood species.13-15,18 Heat and mass transfer processes are affected by volume variation and structural changes undergone by wood while degrading, though these effects on the conversion characteristics are difficult to quantify. For cylindrical particles (hosting a thermocouple), the formation of wide cracks, along the fibers at the median section, has always been observed for d > 2 mm. For the smallest size they are not visible and the char particle remains integral. For sizes above 4 mm, independently of the bed temperature, structural failure always occurs along the direction parallel to the wood fibers resulting, in the majority of the tests, into two semicylindrical fragments. For sizes above 8 mm, three fragments have been

d [mm]

Sp [%]

Sc [%]

2 4 5 8 10

11.2 10.6 10.2 9.8 10.0

27.7 22.2 24.0 24.3 19.0

observed in several cases. These also present additional cracks or are broken into two or more pieces along the direction perpendicular to the fibers. The qualitative features remain the same also for the parallelepipeds, apart from a reduction in the cracks/fragments in the upper part of the particle (probably due to the absence of thermocouple hole). Previous literature35,36 attributes primary fragmentation of wood particles to the pressure build-up, when the rate of volatiles production internally is faster than their escape rate through the pores in the charred wood. The heating characteristics, the rate and heat of reactions and the thermophysical properties are indicated as the factors responsible for the details of structural behavior. From these studies, it appears that, at both low and high temperatures, the pressure peaks before the center temperature exceeds the external temperature. Then, particle break-up suddenly occurs and the pressure drops quickly to the ambient value. It is worth noting that fragmentation of coal particles, fed to fluidized-bed combustors, is also concentrated around the end of the devolatilization regime.23 It is plausible that, despite the slow devolatilization rates established for the inner core of the particles, the outward flux rates of volatile products are even slower, owing to the presence of a thick char layer. In this way the conditions are created for the pressure to increase. On the other hand, as overpressures attain a maximum at near complete conversion, it can be postulated that a large part of the thermal history registered by the thermocouple at the particle center (and the devolatilization characteristics) is not highly affected by structural failure. On the contrary, the effects of primary fragmentation are important for the subsequent char oxidation in fluidized bed combustors.23 Shrinkage enhances internal heat transfer and reduces the conversion time.20 It has been evaluated both along (Sp) and across (Sc) the wood fibers (using the unbroken side of the fragments for sizes above 4 mm). No significant quantitative differences have been observed between cylinders and parallelepipeds (Table 1 summarizes the results obtained in the latter case). The shrinkage Sp (direction parallel to the wood fibers) is only weakly affected by the particle thickness/diameter. Indeed, it is roughly comprised between 11 and 10% (L is equal to 20 mm for all the tests). Shrinkage Sc (direction perpendicular to the wood fibers) is more important and roughly comprised between 28 and 19%, as d is varied from 2 to 10 mm. In accordance with the increase of the char yield with the particle size, it can (36) Hastaoglu, M. H.; Kahraman, R.; Syed, M. Q. Pellet Breakup Due to Pressure Generated during Wood Pyrolysis. Ind. Eng. Chem. Res. 2000, 39, 3255-3263.

Wood Particles in a Hot Sand Bed

Energy & Fuels, Vol. 17, No. 1, 2003 253

Figure 12. Times corresponding to the second local maximum of the heating rate tv as functions of the particle diameter (Tr ) 807 K and Tr ) 1107 K) and the bed temperature (d ) 4 mm and d ) 8 mm): symbols for the experiments and solid lines for the correlation (eq 2).

be postulated that it is directly related to the amount of volatiles released during the conversion process. Empirical Correlations. Several definitions have been used for the devolatilization time of wood particles, based either on gas evolution (time when 95% of the total gas mass has been released37), rate of weight loss (time when it reduces to 1/10 of the maximum value7), or time history of the temperature at the particle center (time when the temperature attains a maximum8 associated with lignin decomposition) (a review of different definitions used for coal is reported in ref 24). For thick wood cylinders radiatively heated, the three definitions produce very small variations on the estimated values.7,8 However, the temperature measurements presented here show that a maximum, higher than the bed temperature and caused by lignin degradation, can be observed only for d > 3 mm and 725 < Tr < 900 K. Hence, to make a comparison of the characteristic times for the entire range of conditions examined, the time t4, indicated in the following as tv, is considered. As discussed above, this corresponds to the second local maximum in the heating rate of the particle center and, for thick particles,7,8 it is the time when conversion is about 95% (temperature range 750-950 K). Also, it has been observed that, for bed temperatures up to 875 K, tv is coincident with the time when the temperature at the particle center becomes equal to 0.95Tr. However, owing to the high heating rates attained by the particle at near complete devolatilization, the corresponding temperatures are significantly different from T4, especially for very low and very high Tr. For Tr above 875 K, tv becomes slightly longer than the time when T ) 0.95Tr (differences of about 4s for Tr ) 1107 K). Figure 12 reports tv as a function of the bed temperature (d ) 4 and d ) 8 mm) and the particle diameter (bed temperature of 807 and 1107 K), as measured (symbols) and obtained by the following empirical correlation (solid lines):

Figure 13. Logarithm of the parameter A (eq 3) as a function of 1000/Tr.

It reproduces the same functionality of those proposed for the devolatilization times of coal particles (see, for instance, refs 22-24) and in one case for wood,9 that is,

tv ) Adn

(3)

(2)

The value of n ) 1.2 in eq 2 is comprised in the range 1-2 reported in the literature24 with variations caused by the different fuel, technique of measurement and definition of devolatilization time. On the other hand, the use of simple thermal theories38 applied to combustible spheres exposed in a hot gas environment leads to devolatilization times proportional to either d or d2, depending on external or internal heat transfer control. The dependence of tv on other parameters, such as bed temperature, coal rank, oxygen concentration, etc., is usually incorporated in the parameter A. In particular, a treatment of the devolatilization times by regression analysis, assuming a power-law function of the operating variables, resulted in an exponent for the bed temperature varying from -1.8 to -3 for different coal varieties.22-24 Also, an Arrhenius temperature dependence was postulated, as it fits equally well the experimental data with activation temperatures varying in the range 1122-3396 K. In this study, an Arrhenius plot has been constructed of ln(A) (Figure 13) as a function of 1/Tr giving A ) 0.8e1525/Tr (eq 1). As already noted,23,24 the very low value of the corresponding activation energy (about 12.7 kJ/mol) is indicative of rate control by transport phenomena. The relative importance of the different mechanisms can be approximately evaluated through the characteristic numbers. External heat transfer coefficients (900670 W/m K for d ) 2-10 mm and Tr ) 807 K) estimated according to Prins,39 average values for the medium properties7,8 (λ ) 1.4 × 10-1 W/m K, F ) 450 kg/m3, and cs ) 1.25 kJ/kg K), reaction temperatures equal to the T3 values previously introduced and kinetic parameters derived from Di Blasi and Branca29 have been used for the evaluation of the Biot (Bi) and the thermal Thiele (Th) numbers.7,8 Estimates lead to Bi values, based on the particle radius, comprised between about 7-25 (d ) 2-10 mm and Tr ) 807 K) or 10-14 (Tr ) 646-1107 K and d ) 4 mm). That is, internal heat transfer is

(37) Di Blasi, C.; Signorelli, G.; Di Russo, C.; Rea, G. Product Distribution from Pyrolysis of Wood and Agricultural Residues. Ind. Eng. Chem. Res. 1999, 38, 2216-224.

(38) Kanury, A. M. Combustion Characteristics of Biomass Fuels. Combust. Sci. Technol. 1994, 97, 469-491. (39) Prins, W. Fluidised Bed Combustion of a Single Carbon Particle. Ph.D. Thesis, Twente University, 1987.

tv ) 0.8e1525/Trd1.2

254

Energy & Fuels, Vol. 17, No. 1, 2003

controlling and it also becomes successively more important as Tr or d are increased. Computed values of Th are always very small. They, however, increase with d and/or Tr, as a consequence of internal heat transfer becoming relatively more important than chemical kinetics. Conclusions In this study experimental results are presented concerning reaction temperatures, heating rate, devolatilization times and char yields of single (cylindrical) wood particles injected in a hot sand bed fluidized by nitrogen. The experimental conditions examine ranges of particle sizes and bed temperatures of 2-10 mm and 712-1107 K, respectively. Measurements reveal that the heating rate at the particle center is a strong function of the conversion level; that is, pyrolysis of wood for the conditions typically established in fluidized-bed reactors is controlled by the rate of internal heat transfer. Decomposition is always associated with a strong reduction in the heating rate (values comprised between 0 and 25 K/s), so that temperature variations are small especially for the stage associated with the decomposition of hemicellulose and cellulose components (typical values of about 625-725 K). Consequently, heating rates attain a maximum either before decomposition begins or after decomposition is completed, depending on the external temperatures being comparable or significantly higher than the characteristic reaction values. Though already shown by mathematical modeling, 19,20,30,31 this study provides the first experimental evidence of this special feature of lignocellulosic fuel degradation under fast external heat transfer rates. Also, it should be noted that the actual reaction temperatures are those established in thermal analysis, though heating rates may be significantly higher.

Di Blasi and Branca

From the quantitative point of view, this study provides extensive measurements of conversion times, not available in the literature to verify the validity of simple empirical correlations, already extensively applied for coal devolatilization, and/or to validate detailed transport models currently successfully verified only for product yields.19,20 As for the first point, it has been found that a highly simplified expression, based on the same functional dependence extensively applied for coal, of the devolatilization times on the particle diameter and the bed temperature applies to wood as well. The second point will be examined in a successive analysis. Nomenclature A: parameter (eq 3) cs: heat capacity [kJ/kg K] d: particle diameter [mm] h: heating rate at the particle center [K/s] L: particle length [mm] n: parameter (eq 3) Sc,p: shrinkage across (c) or along (p) the direction of the wood fibers [%] of the initial length T: temperature at the particle center [K] Tp: temperature of the gas adjacent the top surface of the particle [K] Ts: temperature of the gas adjacent the lateral surface of the particle [K] Tr: bed temperature [K] t: time [s] tc,p: characteristic conduction time across (c) or along (p) the direction of the wood fibers [s] tv: conversion time [s] Yc: char yield [%] of the initial dry wood mass λc,p: thermal conductivity across (c) or along (p) the direction of the wood fibers [W/m K] Fs: solid density [kg/m3] EF020146E