DSpace at VNU: Fast pyrolysis of palm kernel cake using a fluidized bed reactor: Design of experiment and characteristics of bio-oil

Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 Contents lists available at SciVerse ScienceDirect Journal of Industrial and Engineering Chemistry journal homepage: www.elsevier.com/locate/jiec Fast pyrolysis of palm kernel cake using a fluidized bed reactor: Design of experiment and characteristics of bio-oil Thanh-An Ngo a, Jinsoo Kim b,*, Seung-Soo Kim c,** a Department of Chemical Engineering, Ho Chi Minh City University of Technology, 268 Ly Thuong Kiet, District 10, Ho Chi Minh City, Vietnam Department of Chemical Engineering, Kyung Hee University, Seocheon-dong Giheung-gu, Yongin, Gyeonggi-do 446-701, Republic of Korea c Department of Chemical Engineering, Kangwon National University, Joongang-ro, Samcheok, Gangwon-do 245-711, Republic of Korea b A R T I C L E I N F O Article history: Received 31 May 2012 Accepted 16 July 2012 Available online 24 July 2012 Keywords: Fast pyrolysis Palm kernel cake (PKC) Central composite rotatable design (CCRD) Optimization A B S T R A C T In this study, the central composite rotatable design (CCRD) was employed to investigate the effects of the feedstock feed rate, biomass particle size, pyrolysis temperature, and residence time on the fast pyrolysis of palm kernel cake A mathematical model for the liquid product yield was developed and applied to obtain a maximum yield of 49.5 wt% The GC–MS analyses of the bio-oils at the two different temperatures of 400 and 500 8C showed that they were a complex mixture composed of mostly oxygenated compounds including b-D-allose, derivatives of furan and phenol, and a considerable amount of fatty acids ß 2012 The Korean Society of Industrial and Engineering Chemistry Published by Elsevier B.V All rights reserved Introduction Among oils and fats, palm oil is currently produced at the highest volume worldwide [1] Accompanying the palm oil, palm kernel cake, an undesirable by-product primarily composed of mannan (cellulose-like biopolymer [2,3]), is also produced Due to its abundance, numerous attempts to convert this waste biomass into energy or fuel by direct burning, pyrolyzing, or gasifying have been carried out Of all the products obtained from processing palm kernel cake, bio-oil is preferable Since bio-oil currently receives much interest as an alternative to the shrinking fossilbased fuel reserves, much effort has been spent to find an effective and economically feasible technology to obtain as high a bio-oil yield as possible Fast pyrolysis using a fluidized bed reactor can achieve relatively high yields and is considered one of the most efficient approaches to acquire the highest yield of bio-oil In order to yield higher amounts of oil, the fluidized bed reactor must be operated at a residence time of less than one second [4] However, this does not mean that if the residence time is extremely short, the bio-oil yield will increase significantly In the case of a very short residence time, entrainment of biomass may occur during the pyrolysis process, thus causing loss of the feedstock An appropriate flow rate of the fluidizing gas can be * Corresponding author Tel.: +82 31 201 2492; fax: +82 31 202 1946 ** Corresponding author Tel.: +82 33 570 6544; fax: +82 33 570 6535 E-mail addresses: jkim21@khu.ac.kr (J Kim), sskim2008@kangwon.ac.kr (S.-S Kim) calculated from the residence time based on the inverse proportionality between residence time and flow rate With the residence time requirement of less than one second previously mentioned, the operating flow rate of the fluidizing gas should be changed to a specific range at which the fluidized condition can occur without entrainment In addition to the residence time, the bio-oil yield is also strongly affected by the pyrolysis temperature The feedstock will decompose to produce more gas and liquid products at higher pyrolysis temperatures At low temperatures below 250 8C, most of the components in biomass not degrade When the temperature is in the range of 250–520 8C, the pyrolysis process will occur efficiently due to the decomposition of hemicellulose, cellulose, and lignin, as reported in literature [4,5] Many studies have also reported that this temperature range is preferable to obtain the highest bio-oil yield [6] Once the temperature exceeds 520 8C, the secondary cracking reaction of bio-oil will occur, leading to a reduced product yield [7] Since pyrolysis is a thermal process, heat transfer plays a critical role in the fluidized bed technology The particle size and feed rate of the feedstock are the two factors directly affecting the heat transfer To enhance the supply of energy from the reactor to deep inside the biomass particles, a small particle size is more desirable However, the biomass particle size should be carefully selected for use in fluidization because a small particle size will facilitate entrainment The feed rate of the feedstock not only affects the heat transfer, but also the pressure drop in fluidized bed reactor In practice, the feed rate of the feedstock may affect the initial height of material in the reactor, resulting in the pressure drop as well as fluidization behavior (smooth, particulate, bubbling or aggregative) [8] 1226-086X/$ – see front matter ß 2012 The Korean Society of Industrial and Engineering Chemistry Published by Elsevier B.V All rights reserved http://dx.doi.org/10.1016/j.jiec.2012.07.015 138 T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 Finally, there are at least four operating parameters which influence the fast pyrolysis using a fluidized bed reactor Although many researchers have studied the influence of the parameters on the bio-oil yield experimentally and numerically, it is not easy to obtain the optimum condition for higher yield of bio-oil without scanning all the parameters Therefore, in this research, an experimental design technique referred to as central composite rotatable design (CCRD) was utilized to organize the experimental runs This experiment design procedure is commonly used for planning experiments both in the laboratory and in industry [9] Following this design closely, a second-order regression equation describing the effects of all factors was obtained Based on this equation, an appropriate algorithm was then used to obtain the optimized conditions Experimental sections 2.1 Palm kernel cake The raw palm kernel cake received from Green Ocean Co (Malaysia) was ground with a knife mill and then exposed to air for 24 h The prepared sample was then sieved to obtain the desirable particle size for each experiment Before being used as a feedstock for the fast pyrolysis experiments, all the samples were dried again at 80 8C for 24 h in a dryer They were subsequently preserved in a closed bag prior to the experiment The weight of the remaining palm oil was quantified based on the difference of the weights of the raw sample before and after washing in pure acetone where the weight of the remaining sample was assumed to be palm oil 2.2 Fluidized bed reactor The fast pyrolysis of the palm kernel cake was conducted in a fluidized bed reactor, as schematically described in Fig The pyrolysis reactor, cyclone, and two condensers used in this research were made of Pyrex glass The reactor had a diameter of 3.6 cm with a height of 20 cm In order to avoid the product condensing on the wall of the cyclone and pipeline, the whole apparatus was carefully covered by a heat-insulating material For the purpose of applying various biomass particle sizes for the fast pyrolysis and maintaining the fluidized condition more stably another exit from the reactor, in addition to the usual product exit from the reactor to the cyclone (way 1), to a solid collecting bottle (way 2) was also designed The presence of way was helpful for drawing the char out of the reactor when a large particle size and small flow rate of fluidizing gas (at which the entrainment of solid product does not occur) were used and successfully resulted in balancing the input and output materials during the process For each run, 100 g of a sample was placed in the feed hopper In addition, the reactor was also charged with 15 g of silica sand in range of 250–300 mm as a pyrolysis medium When the furnace temperature reached the preset value, the MFC (mass flow controller) was switched on at a prescribed value (from 10 to 15 L/min, depending on the residence time) to allow the fluidizing gas (N2) to enter the reactor The equipment setup was first run without feeding biomass for 30 until the flow rate of the fluidizing gas and furnace temperature became stable Subsequently, the controller of screw feeder was turned on to push the feedstock to the reactor at an accurate feed rate After pyrolysis, the small sized solid product was blown out and separated at the cyclone, while the larger sized particles were drawn out to the bottle below the reactor After passing through the cyclone, the mixture of Fig Schematic diagram of experimental apparatus T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 the pyrolyzed products (liquid and gas) were cooled down by the two condensers, which used a chiller with coolant at À20 8C Most of the liquid product was condensed and collected in two-neck flasks, while the other non-condensable compounds were retained at the cotton filter located at the end of equipment The weight of liquid product, Wliquid, is defined as the weight difference of the two condensers and the cotton filter before and after pyrolysis The liquid yield of the fast pyrolysis was then calculated based on the equation shown below, Liquid yield wt%ị ẳ W liquid  100% W biomass where Wbiomass is the weight of the biomass feedstock used in experiments 2.3 Product analysis The bio-oil was first dissolved in pure acetone and then analyzed by GC–MS The GC–MS analysis of pyrolyzing oil was performed using an Agilent Technologies 7890A GC equipped with an Agilent Technologies 19091S-433 column (30 m  0.25 mm  0.25 mm) and a mass spectrometer (Agilent Technologies 5975C) The carrier gas was helium at a flow rate of mL/min The column temperature was initially held at 40 8C for min, then gradually increased to 280 8C at 10 8C/min, and finally maintained at 280 8C for 10 The injector and detector temperatures were set at 250 and 280 8C, respectively The ash content of the bio-oil was determined using ASTM E1755-01 (2001) [10] The elemental analyses for C, H, O, N, and S were carried out using an automatic elemental analyzer (EA, Flash EA1112, CE Instruments) The water content was analyzed using a method previously reported in the literature [11] When 1.0 g of anhydrous CaCl2 was added into a solution of 10 mL of acetone containing water, the exothermic energy led to an increased temperature of the solution which depends on the water content Based on this principle, a calibration curve showing the relationship between the water content and the temperature rise was constructed Using this calibrating curve, it was easy to calculate the water content existing in the pyrolyzing liquid product 2.4 Experimental design Experimental design refers to the process of planning, designing, and analyzing experiments so that valid and objective conclusions can be efficiently made [12,13] In practice, the second-order design referred to as central composite rotatable design (CCRD) is commonly employed, especially in chemical engineering [9] For the design, a matrix of coded variables (X) is initially set up to plan the experiments The number of rows in this matrix or the total experiments to run (n) depends on the factor (k) according to the following expression: n ẳ 2k ỵ 2k ỵ no ẳ n j ỵ na ỵ no where no is the number of replicated experiment at the center point Therefore, the number of experimental runs includes nj = 2k experiments at the core points (X = Ỉ1), na = 2k experiments at the axial points (X = Ỉ a = Ỉ 2k/4), and no experiments at the center point (X = 0) Corresponding to each set of variables in this matrix, the response value from the experiments can be obtained Subsequently, the experimental data are fit to a polynomial 139 mathematical model of second order, as shown below Y ẳ bo ỵ k X k X k X k X i¼1 i¼1 j > i i¼1 bi X i þ bi j X i X j þ bii Xi2 (1) where Xi and Xj are the coded variables from the actual xi, and xj variables, respectively Y represents the response value from the experiment while bo is the value of the fitted response at the center point of the design bi, bii, and bij are the linear, quadratic, and interaction terms, respectively The coded variable was achieved from the actual variable based on the following expression [14]: Xi ¼ xi À xoi Dx i (2) where xoi is the midpoint value of the actual variable xi ðxoi ¼ max ẵxi ỵ xmin =2ị, Dxi is the interval value of the actual variable xi i Dxoi ẳ ẵxmax ỵ xmin Š=2Þ, and xmax ; xmin are the high and low levels i i i i of the actual variable, respectively All the regression coefficients (bo, bi, bii, bij) were calculated as follows: bo ¼ a1 n k X n X X Y u À a2 Xiu Yu u¼1 bi ¼ a3 (3) i¼1 u¼1 n X X iu Y u (4) u¼1 bi j ¼ a4 nj X X iu X ju Y u (5) u¼1 bii ¼ a5 n k X n n X X X 2 Xiu Y u ỵ a6 Xiu Y u a7 Yu u¼1 i¼1 u¼1 (6) u¼1 where a1, a2, a3, a4, a5, a6, and a7 are constants determined from literature [14] In case of experiment with four factors, a1, a2, a3, a4, a5, a6, and a7 are 0.1428, 0.0357, 0.0417, 0.0625, 0.0312, 0.0037, and 0.0357 respectively Once all the regression coefficients were determined, their statistical significance was then estimated A regression coefficient is statistically significant if its absolute value is higher than the confidence interval Finally, the obtained regression model was then checked for a lack of fit by calculating the FR value If FR < FT (FT: tabular value of the Fisher-criterion), the regression equation is considered to be adequate Results and discussion Table shows the actual and coded levels of the process parameters The coded values were designated by À1 (minimum), (center), +1 (maximum), Àa, and +a The selection of levels for each factor was based on previous pyrolysis reports of other biomass using a fluidized bed reactor The investigated temperature range was from 400 to 500 8C since the biomass containing hemicellulose, cellulose, and lignin is effectively decomposed in this temperature range, as reported by Demirbas and Arin [5] The residence time, which was in the range of 0.6–0.9 s, was limited by the fact that the pyrolysis only effectively yields a high bio-oil yield if the residence time is less than s [4] The biomass particle size range selected was 300–600 mm in order to make the screw feeder operate well Because palm kernel cake contains residual palm oil at an amount of up to 10.7 wt% (determined by the acetone washing method previously mentioned), it will be sticky if the particle size is too small Depending on the size of the reactor, a feed rate in the range of 160–300 g/h was appropriate Table 140 T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 Table Factor variation intervals Factors Unit Feed rate, x1 Temperature, x2 Residence time, x3 Particle size, x4 g/h 8C s mm Variation levels Variation interval À2 À1 Dx 90 350 0.45 150 160 400 0.6 300 232 450 0.75 450 300 500 0.9 600 375 550 1.05 750 70 50 0.15 150 Table Experimental design matrix and response values Run X1 X2 X3 X4 Liquid yield (%) Experiment at core point 10 11 12 13 14 15 16 À1 À1 À1 À1 À1 À1 À1 À1 Run X1 X2 X3 X4 Liquid yield (%) 0 0 À2 0 0 0 0 À2 40.0 Ỉ 1.5 33.6 Ỉ1.6 13.0 Ỉ 0.6 36.5 Æ 0.5 40.2 Æ 0.5 52.3 Æ 2.3 44.0 Æ1.0 45.2 Ỉ 0.8 0 0 0 0 0 0 0 46.2 Ỉ 0.7 46.1 Ỉ 0.8 45.7 Ỉ 0.7 45.3 Ỉ 0.5 45.2 Æ 0.1 46.4 Æ 0.8 45.2 Æ 0.9 Experiment at axial point À1 À1 1 À1 À1 1 À1 À1 1 À1 À1 1 À1 À1 À1 À1 1 1 À1 À1 À1 À1 1 1 À1 À1 À1 À1 À1 À1 À1 À1 1 1 1 1 30.0 Ỉ 1.5 19.0 Ỉ 1.3 43.6 Ỉ 1.5 42.5 Æ 0.5 44.0 Æ 2.0 37.8 Æ 1.0 40.0 Æ 0.3 46.5 Ỉ 0.5 32.3 Ỉ 1.3 18.1 Ỉ 1.3 49.0 Ỉ 1.8 47.0 Ỉ 1.0 41.0 Ỉ 1.0 35.2 Æ 0.6 41.6 Æ 0.6 47.5 Æ 1.2 shows the parameters of the conducted experiments based on the design matrix The obtained liquid yield for each condition in this table was presented in form of the average value of twoexperiment run values and plus/minus the standard error 3.1 Checking the fitted models Table shows the regression coefficients calculated based on Eqs (3)–(6) and the F-test of the obtained models The significance 17 À2 18 19 À2 20 21 0 22 0 23 0 24 0 Experiment at center point 25 0 26 0 27 0 28 0 29 0 30 0 31 0 from the standard F-distribution with a confidence of 99% This means that the regression equation was adequate This adequacy can be revealed by the relationship between the actual and predicted values of the liquid yield It clearly shows that the model successfully correlates the process parameters to the liquid yield with a correlation coefficient of R2 = 0.99 Finally, the predictive response equation containing all of the significant coefficients was determined, as shown below In coded units: Y liq ẳ 45:71 1:697X ỵ 6:142X ỵ 3:182X ỵ 0:446X ỵ 2:906X X ỵ 1:794X X 4:069X X ỵ 1:044X X À 0:894X X À 2:277X12 À 5:285X22 À 0:33X42 (7) In uncoded (actual) units: Y liq ẳ 529:179 0:31x1 ỵ 2:177x2 ỵ 243:574x3 0:017x4 ỵ 8:3 104 x1 x2 ỵ 0:171x1 x3 0:543x2 x3 ỵ 1:392 104 x2 x4 0:040x3 x4 À 4:647  10À4 x21 À 2:114  10À3 x22 À1:467  10À5 x24 (8) of each regression coefficient was also verified and is presented in Table The t-test values were obtained by dividing each coefficient by its standard error A coefficient was considered significant if the magnitude of its t-test value was larger than the standard t-distribution at a certain confidence In this study, a 95% confidence was selected A large t value implies that the coefficient is much greater than its standard error As can be seen in the t-test values for the liquid yield in Table 3, most of the coefficients were significant, except for the b14 and b33 coefficient Therefore, these two coefficients can be eliminated from the regression equation of the liquid yield Subsequently, the F-test value of the regression model was also calculated The obtained F-test value was smaller than the F-value 3.2 Effect of operating parameters on the fast pyrolysis performance From the regression equation, it can be seen that there are three types of effects of the variables on the response value: main, squared, and interaction effects corresponding to the bi, bii, and bij coefficients, respectively As for the main effects, the magnitudes of the t-test were obtained in the order as follows: b2 (59.06) > b3 (30.6) > b1 (À16.3) > b4 (4.3) It is also known that the higher the t-test value of a coefficient, the more significant the effect of the coefficient Therefore, it can be concluded that the X2 (pyrolysis temperature) and X3 (residence time) variables were the two most important factors having the strongest effects on the response value The order of the significant effect on the response value can T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 141 Table Statistical significance of regression coefficients bo b1 b2 b3 b4 b12 b13 b14 b23 b24 b34 b11 b22 b33 b44 FR FTa Coefficient t-test Significance 45.710 À1.697 6.142 3.182 0.446 2.906 1.794 À0.269 À4.069 1.044 À0.894 À2.277 À5.285 0.082 À0.330 3.32 7.87 238 À16.3 59.06 30.6 4.3 22.9 14.1 À2.1 À32 8.2 À7 À24.2 À56.2 0.87 À3.5 + + + + + + + À + + + + + À + + = significant; À = insignificant; degree of freedom f = 6, t-student(0.05) (6) = 2.45 a FT: referenced in [9] be arranged as follows: X2 > X3 > X1 > X4 Since there are four factors which affected the pyrolysis process, it is impossible to present all the effects on the same 3D graph As a result, two factors were held constant, while the others were varied leading to the 3D graphs shown in Figs 2–4 Indeed, pyrolysis is a cracking reaction to break a high molecular weight hydrocarbon chain into smaller compounds Pyrolysis is always an endothermic process Consequently, when the pyrolysis temperature increases, the product yield increases However, it should be noted that if the temperature exceeds the value at which the products decompose, the obtained product yield decreases Therefore, it was found that the liquid yield increases with pyrolysis temperature until the decomposition temperature is achieved, after which the yield decreases with increasing temperature In the experiments conducted at a low temperature of 400 8C, the receiving energy may not be enough to completely decompose the biomass, as shown in Fig If the residence time was decreased by increasing the flow rate of the fluidizing gas, more and more incompletely decomposed biomass was withdrawn from the reactor When feedstock is pushed out of the reactor, the biomass Fig Liquid yield at a feed rate of 160 g/h and a particle size of 300 mm Fig Liquid yield at a pyrolysis temperature of 400 8C and a residence time of 0.9 s conversion is obviously not complete, thus leading to a reduced liquid yield At the high temperature of 500 8C, the biomass can be completely decomposed Hence, if the residence time is decreased (or the flow rate of the fluidizing gas is increased), the product can be pushed out of the reactor as quick as possible in order to avoid decomposition of the bio-oil, which therefore results in increasing the liquid yield Fig shows that the feed rate has more effect on liquid yield than the particle size At a certain pyrolysis temperature, residence time and particle size, the increase of feed rate results in more biomass accumulated inside the reactor Consequently, the efficiency of the heat transfer inside the reactor is lowered, leading to incomplete decomposition of the biomass Therefore, it is observed that higher liquid yield is obtained at the lower feed rate On the contrary, the effect of particle size on liquid yield is not noticeable With increasing the particle size from 300 to 600 mm, the liquid yield increases less than 5% Fig Liquid yield at a residence time of 0.9 s and a particle size of 600 mm 142 T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 Fig shows the effect of temperature and feed rate on the liquid yield at the residence time of 0.9 s and particle size of 600 mm Both temperature and feed rate influence significantly on the liquid yield in form of a second-order polynomial Fig also shows that the liquid yield is inversely proportional to pyrolysis temperature In contrast, the liquid yield rises to a maximum and then decreases with increasing the feed rate from 160 to 300 g/h 3.3 Optimization of operating parameters The analysis of the response surface from the mathematic model of liquid yield yielded the stationary point (X1 = 0.149, X2 = 0.597, X3 = 0.356, and X4 = 1.139) at which the derivative of the equation is zero and is located outside the surveyed region Therefore, the optimal response values may be determined at one of the boundaries of the investigated variable ranges [9] In order to specify the optimal response values, a trial and error algorithm was applied for the equation of the liquid yield in coded units A procedure using Matlab was set up as a loop to search for the condition in the surveyed range at which the response function achieved the highest value Once obtained, the optimal condition (X1 = À0.1, X2 = 1, X3 = À1, and X4 = 1) was then converted into the respective uncoded (actual) units using the formulas shown in Eq (2) Finally, the optimal values of the feed rate, pyrolysis temperature, residence time, and particle size were found to be 225 g/h, 500 8C, 0.6 s, and 600 mm, respectively, corresponding to the highest liquid yield of 49.5 wt% In order to check the accordance with the regression model, the experiment at optimal condition was also performed The liquid yield obtained was 50.3 Æ 1.4% This result shows that the experiment was well fitted with the model 3.4 Characteristics of pyrolyzing liquid product The physical properties of bio-oil are useful in the evaluation of treatment technology and the selection of process equipment directly affects the application and efficiency of bio-oil The liquid product obtained from experiment run at temperature of 500 8C, feed rate of 300 g/h, particle size of 300 mm, and residence time of 0.6 s was used for characteristic analysis, including ash and water contents, density, and amounts of carbon, hydrogen, oxygen, and nitrogen The results are shown in Table As can be seen in this table, the ash content is zero, which suggests that this bio-oil has obvious advantages as a clean fuel oil The data also show that the water content, a product of the dehydration reaction occurring during the pyrolysis [15], was rather consistent and in the range of 15–30%, as reported by Czernik and Bridgwater [15] From the elemental analysis results, the heating value was obtained using the formula shown below [16], HHVMJ=kgị ẳ f33:5ẵCỵ142:3ẵH15:4ẵO À 14:5½NŠg  10À2 (9) where [C], [H], [O], and [N] are the contents (wt%) of carbon, hydrogen, oxygen, and nitrogen, respectively The calculated heating value is just 13.9 MJ/kg, which is rather small in comparison with other conventional fuels such as petroleum (43 MJ/kg), LPG (45.75 MJ/kg), or kerosene (41 MJ/kg) [16] The reason for this result is mainly due to the high content of oxygen and nitrogen remaining in the bio-oil In fact, it can be observed from formula (9) that the heating value decreases with increases of both the oxygen and nitrogen contents The oxygen can exist in the form of water or oxygenated compounds such as ketones, phenols, and esters Due to the high oxygen content, challenges remain in the further utilization of the bio-oil as a fuel oil Therefore, additional upgrading including deoxygenation is necessary Table Physical and chemical properties of pyrolytic oil obtained from an experiment at 500 8C Property Bio-oil Ash content (%) Water content (%) Elemental analysis (%) Carbon Hydrogen Oxygen Nitrogen Empirical formula Heating value (MJ/kg) free 12.1 33.8 8.08 54.72 3.23 C12.25H35.13O14.87N 13.9 Table shows the GC–MS analysis data of the bio-oil obtained from pyrolysis experiments at temperatures of 400 and 500 8C at the same feed rate (300 g/h), particle size (300 mm), and residence time (0.6 s) According to the data, the detectable components of the bio-oil were acetic acid, ketones, derivatives of furan, phenolics, esters, b-D-allose, and fatty acids As reported by Yaman [17], the bio-oil from pyrolysis of lignocellulosic biomass was largely composed of alcohols, aldehydes, ketones, organic aicds, ester, phenolics, and levoglucosan Obviously, it can be realized that there exists some differences in the product distribution obtained in this study and that of other researchers The most noticeable difference was the presence of bD-allose (a C3 epimer of glucose [18]) at an amount of up to around 20% while some compounds such as alcohols, aldehydes, and phenolics were present at smaller quantities These results can be explained by the differences in the biomass compositions Indeed, the major composition of palm kernel cake is mannan, a biopolymer formed by a great number of sugar mannose (C6H12O6) units (C2 epimer of glucose [18]) Basically, when a polymer is thermally decomposed, its chain is cracked to release lower molecular compounds including monomer, dimer, trimer, tetramer, oligomer, and some other compounds depending on the reaction conditions During this cracking process, various types of reactions may occur such as transposition of C, H, or cleavage of the C–C linkage [19] For a monomer, there may be a change in the molecular structure compared to its initial structure unit However, the molecular formula may not change From the Table GC–MS analysis of bio-oil obtained from pyrolysis of palm kernel cake at the optimal conditions No 10 11 12 13 14 15 16 17 18 19 20 Compound Acetic acid 1-hydroxy-2-propanone Furfural 4-hydroxy-4-methyl-2-pentanone 2-Furanmethanol (5H)-Furananone 1, 2-cyclopentanedione 2-furancarboxaldehyde, 5-methylphenol 2-furancarboxylic acid, hydrazide Benzenecarboxylic acid maltol 1, 2-benzenediol 2, 3-o-acetonemannosan b-D-allose Dodecanoic acid Tetradecanoic acid Hexadecanoic acid 6-octadecenoic acid (Z) 1, 2-benzenedicarboxylic acid, bis (2-ethylhexyl) ester 400 8C 500 8C Area (%) Area (%) 11.6 7.83 1.68 10.3 4.04 2.03 1.34 2.84 1.91 1.82 2.51 3.05 19.23 3.77 0.88 0 9.34 5.0 3.1 3.0 5.0 0 2.3 2.9 0.85 2.7 2.8 23.2 11.2 3.6 1.5 2.6 14.9 T.-A Ngo et al / Journal of Industrial and Engineering Chemistry 19 (2013) 137–143 molecules of b-D-allose and mannose, it can be seen that their structures are rather similar and have the same molecular formula Therefore, it can be concluded that b-D-allose was the product obtained from the decomposition of mannan component in the palm kernel cake Apart from b-D-allose, a considerable amount of fatty acids were also detected in the pyrolyzed bio-oil of PKC mainly including dodecanoic and tetradecanoic acid As reported by Demirbas [20], oils deriving from plants are built up with triglyceride molecules When this molecule is decomposed under thermal conditions, the three branches making the structure of the molecule are randomly cleaved to generate fatty acids and other smaller molecules Based on this result, the fatty acids can be attributed to the pyrolysis of the remaining palm oil in the raw biomass feedstock Moreover, as can be seen in Table 5, the acetic acid, 1-hydroxy2-propanone, and 2-furanmethanol contents decreased with increased pyrolysis temperature This means that the pyrolysis process preferably produced these compounds at lower temperatures In contrast, the contents of two major compounds, b-Dallose and 1, 2-benzenedicarboxylic acid, bis (2-ethylhexyl) ester, increased proportionally to the pyrolysis temperature As explained previously, b-D-allose is a product of mannan decomposition Therefore, its content only increases if the degradation reaction of mannan occurs more completely In addition, when the pyrolysis temperature increased, the cracking reaction of triglyceride was also favored to produce fatty acids, as reported by Demirbas [20] As a result, it was found that the content of each fatty acid increased with increasing pyrolysis temperature Conclusions The fast pyrolysis of palm kernel cake was conducted by applying the central composite rotatable design in a fluidized bed reactor over a range of variables as follows: a feed rate of 160– 300 g/h, a pyrolysis temperature of 400–500 8C, a residence time of 0.6–0.9 s, and a particle size of 300–600 mm The results demonstrated that the experimental data were well fitted with a second-order regression equation Therefore, the effects of the 143 four operating parameters on the fast pyrolysis were elucidated The order of significance of each operating parameter factor was obtained as follows: pyrolysis temperature > residence time > feed rate > particle size The highest liquid yield achieved was 49.5 wt% at a feed rate of 225 g/h, a pyrolysis temperature of 500 8C, a residence time of 0.6 s, and a particle size of 600 mm The pyrolysis was conducted at temperatures of 400 and 500 8C and the obtained products were subsequently analyzed in order to determine the reaction characteristics The pyrolytic oil contained a considerable amount of oxygen, which led to a reduced heating value This bio-oil largely contained 1, 2-benzenedicarboxylic acid-bis (2-ethylhexyl) ester, b-D-allose, and fatty acids References [1] Foreign Agricultural Service, Indonesia and Malaysia Palm oil production (Commodity intelligence report), United States Department of Agriculture, 2007, see also http://www.pecad.fas.usda.gov/highlights/2007/12/Indonesia_palmoil/ [2] M.J Daud, M.C Jarvis, Phytochemistry 31 (1992) 463 [3] E.M Dusterhoft, M.A Posthumusb, A.J.V Voragena, Journal of the Science of Food and Agriculture 59 (1992) 151 [4] D Mohan, C.U.J Pittman, P.H Steele, Energy and Fuels 20 (2006) 848 [5] A Demirbas, G Arin, Energy Sources 24 (2002) 471 [6] A.V Bridgwater, D Meier, D Radlein, Organic Geochemistry 30 (1999) 1479 [7] A.V Bridgwater, Journal of Analytical and Applied Pyrolysis 51 (1999) [8] D Kunii, O Levenspiel, Fluidization Engineering, Second ed., ButterworthHeinemann, Newton, 1991 [9] Zˇ.R Lazı´c, Design of Experiments in Chemical Engineering, Wiley-VCH, Weinheim, 2004 [10] Annual Book of ASTM Standards, 1997 [11] M.A Orzherovskii, Chemistry and Technology of Fuels and Oils (1969) 609 [12] G.E.P Box, K.B Wilson, Journal of the Royal Statistical Society B 13 (1951) [13] D.C Montgomery, Design and Analysis of Experiments, Fifth ed., John Wiley & Sons Inc, Hoboken, 2001 [14] N.R Draper, H Smith, Applied Regression Analysis, Third ed., John Wiley & Sons Inc, Hoboken, 1998 [15] S Czernik, A.V Bridgwater, Energy and Fuels 18 (2004) 590 [16] A Demibras, Fuel 76 (1997) 431 [17] S Yaman, Energy Conversion and Management 45 (2004) 651 [18] V.S.R Rao, P.K Qasba, P.V Balaji, R Chandrasekaran, Conformation of Carbohydrates, Harwood Academic Publishers, Amsterdam, 1998 [19] J McMurry, Organic Chemistry, Seventh ed., Thompson, New York, 2008 [20] A Demirbas, Energy Conversion and Management 44 (2003) 2093  ... weight of the remaining sample was assumed to be palm oil 2.2 Fluidized bed reactor The fast pyrolysis of the palm kernel cake was conducted in a fluidized bed reactor, as schematically described... the wall of the cyclone and pipeline, the whole apparatus was carefully covered by a heat-insulating material For the purpose of applying various biomass particle sizes for the fast pyrolysis and. .. 5975C) The carrier gas was helium at a flow rate of mL/min The column temperature was initially held at 40 8C for min, then gradually increased to 280 8C at 10 8C/min, and finally maintained at 280