Evaluation of the Performance of Air Dense ... - ACS Publications

Aug 13, 2013 - This condition led to a clean coal ash content, recovery, and separation efficiency of 10.6, ... Particulate Science and Technology 201...
8 downloads 0 Views 2MB Size
Article pubs.acs.org/EF

Evaluation of the Performance of Air Dense Medium Fluidized Bed (ADMFB) for Low-Ash Coal Beneficiation. Part 1: Effect of Operating Conditions Ebrahim Azimi,† Shayan Karimipour,‡ Moshfiqur Rahman,‡ Jozef Szymanski,† and Rajender Gupta‡,* †

Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta T6G 2W2, Canada Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta T6G 2V4, Canada



ABSTRACT: Low-rank coals are widely used as fuel in coal-fired power plants. However, feeding such a coal in addition to lower efficiency generates a variety of problems mostly associated with the ash-forming minerals. An efficient dry coal cleaning method can offer a solution while avoiding problems associated with wet methods and generates partially dried coal. Air dense medium fluidized bed (ADMFB) can be an efficient and economical dry ash removal technology, but the comprehensive understanding on its performance is not yet openly available. In the present work, a detail study of the factors affecting the beneficiation of a low-ash lignite coal using ADMFB is provided. Response surface methodology supported by a central composite experiment design method is employed to study the effect of superficial air velocity (V), residence time (T), and bed height (H) on the performance of a batch ADMFB separator at three levels. Also, the effects of sample weights and particle size of lignite coal are separately studied. The system was found to effectively decrease the ash content of the clean coal product. The organic material recovery to clean coal product was affected negatively by H, V, and T, while T, H, and V were found to affect the separation efficiency positively. Various levels of interactions between parameters were also revealed and discussed. The optimum operating condition for maximizing the recovery was found to be 15 cm/s, 90 s, and 15 cm for V, T, and H, respectively. This condition led to a clean coal ash content, recovery, and separation efficiency of 10.6, 95.63, and 15.29%, respectively. The beneficiation test results also revealed that higher ash removal (23%) and recoveries (86%) are obtainable for coarser coal particles. Higher recovery and separation efficiency are achievable for larger sample weights (84.6 and 30.7%, respectively).

1. INTRODUCTION Coal is the second largest primary energy source in the world after oil.1 It is widely distributed all around the world and is the largest energy source used for electricity generation (up to 41% of the total in 2010).1,2 Increasing the energy demands necessitates the use of low-rank coals to continue the production of low-cost energy. About 66% of the total fossil fuel resources of Canada is coal, which is mostly located in Alberta (70%) and is directly burnt for power generation asmined.3 Direct run of mine (ROM) coals that are feeding the furnaces contain a substantial amount of ash-forming mineral matter and moisture and, in addition to reducing the efficiency of the power plants, generate additional particulate materials, SOx, and emission of trace elements. The clean coal technology campaign has started in most countries that are using coal as the main source of energy and is intended to increase coal use efficiency and decrease the carbon footprint and environmental/health-related pollutions. The clean coal technology is a collective term, covering all steps related to the energy produced from coal, i.e., coal mining, conversion processes, and flue gas treatment. The ROM coal preparation is an essential component of the clean coal technology.4−6 Even though it is not yet widely practiced by the power industry, coal beneficiation is receiving more and more recognition for its role in decreasing the costs and environmental impacts of coal combustion to generate electricity. Firing clean coal directly into boiler/gasifiers can contribute in decreasing higher costs of flue gas treatment, lower particulate © 2013 American Chemical Society

matters in the exhaust gas and furnace waste issues (fly ash handling/pounds), and increase furnace efficiency. Acceptable results are obtained for wet coal beneficiation methods, such as froth flotation, jigging, spiral separators, and heavy medium separation for years.7−10 However, high process costs, fresh water resources scarcity, and subsequent water recycling expenses, sliming of wastes, high clay content of the low-rank coals, high operating costs for coal and waste slurry treatment, and lower thermal efficiency because of the higher moisture content are some issues that motivate the implementation of dry coal cleaning methods.11−13,15 Considering the mentioned issues of the accompanying ash-forming minerals and wet beneficiation processes besides recent progresses of the dry coal beneficiation methods, such as air dense medium fluidized beds (ADMFBs), air jigs, magnetic separators, electrostatic separators, and pneumatic oscillating tables, application of dry beneficiation methods for cleaning ROM coal seems to be inevitable for the industry. At the same time, cold weather in Canada and other cold countries have made the wet methods less desirable for the industry. Dry methods usually take advantage of the differences between component hardness, density, surface characteristics Special Issue: Impacts of Fuel Quality on Power Production and the Environment Received: March 14, 2013 Revised: June 8, 2013 Published: August 13, 2013 5595

dx.doi.org/10.1021/ef400456n | Energy Fuels 2013, 27, 5595−5606

Energy & Fuels

Article

Figure 1. Schematic diagram of the experimental setup.

the ADMFB separator. The separation tests were performed under both a constant air velocity and separation time of 3.5 L/ s and 30 s, respectively. The obtained Ep values showed a decreasing trend with the increase of the coal size until leveling off for particles larger than 25 mm (Ep = 0.04). The ADMFB separator was found to provide good results for the size range of 15−50 mm, with ash contents of 45%. The non-stability of the process (back mixing) was reported by Mohanta and others as a reason for lower separation efficiency of 4.75−13 mm coal. The highest combustible material recovery, ash rejection, and separation efficiency of 78.97, 63.22, and 42.20%, respectively, were obtained for the Hingula coal sample with the original ash content of 44.7%. Chikerema et al.5 verified the effect of the particle size and shape on the performance of a magnetite−silica bed. The bed height of 32 cm and separation times of 5−60 s were considered for the four particle size fractions within the range of 9.5−53 mm. A lower Ep value of 0.05 was obtained for the coarse particles (22−31.5 and 37−53 mm), while for the smaller particles, the Ep values of 0.07 and 0.11 were considered to be high and as an indication of the lower effectiveness of the separation process for finer particles. Also, blockish particles with a smaller surface area/volume ratio showed better separation compared to the other two classes of flat and sharp-pointed particles. Generally, increasing the separation time from 15 to 60 s was found to decrease the separation Ep values. The presence of the interaction between the tested parameters was not examined in this study either. In our current investigation, a systematic study is conducted on the effects of some main operating parameters, such as superficial air velocity (V), separation time (T), bed height (H), feed particle size, and sample weight along their mutual interactions, on the performance of an ADMFB separator when dealing with low-ash/rank coals and smaller particle sizes. More realistic measures for beneficiation test results, such as clean coal ash content, organic material recovery to clean coal, and separation efficiency of the system, are chosen as responses to evaluate and compare separation experimental results. On the basis of the obtained response functions, the optimum operating conditions providing the highest performance of the separator are proposed and examined.

(conductivity), particle size and shape, and their coefficient of friction when free falling in fluids.5,12−16 The ADMFB coal beneficiation method as an efficient coal cleaning method with the comparable separation performance to the wet methods has been used in industrial scale for more than a decade. An ADMFB coal preparation plant was established in 1994 in China to beneficiate coarse bituminous coal (6−50 mm) and operated successfully.6,17,18 Lower upward air flow rates and, consequently, less and smaller dust-collecting equipment with minimum possible moving parts and the possibility of using waste heat (low-quality heat) for simultaneous coal drying are some of the advantages of the ADMFB separators over the other dry methods. Although ADMFB is applied in industrial scale here and there, there is very little information available on the behavior of this method and how various operating parameters interact or affect the performance of these separators. The results of such studies would cause a definite improvement in the control and optimization of the separation process. The effect of the air flow rate, moisture content, stability of the fluidized bed, feed mixture composition, and coal particle size and shape are discussed in several studies, usually for high ash-coarse size samples.5,19−27 In most of the studies, the effect of mutual interactions between operating parameters (because of complex hydrodynamics associated with the fluidized beds) as well as the effective range of the parameters on the apparatus response is not considered accurately. Van Houwelingen and De Jong22 investigated the effect of the air velocity, composition, and moisture content of the coal on the performance of a 160 × 20 cm vibrating fluidized bed. Along the separation tests, mixtures of 20−30 mm coal particles and shale were separated in a bed of 220 μm (d50) sand. With one factor at a time method, they found that increasing air velocity increases the amount of rejected coal particles (versus shale) in the sink product, where a higher moisture content also intensified the coal losses at higher air velocities. They also reported that the rejected portion of coal decreases when mixed with high amounts of shale particles (85% shale and 15% coal particles). The existence of coal particles versus shale particles in the sink or float zone was used to evaluate the separation performance. Mohanta and colleagues19,20 used a 15 cm diameter column with 20 cm active bed height and 4.7 tons/m3 magnetite particles (d50 of 212 μm) as a fluidization medium to investigate the effect of the feed size (4.75−50 mm) on the performance of

2. ADMFB FOR COAL CLEANING ADMFB separator uses a gas−solid fluidized bed as the separating medium. With the increase of the velocity of a fluid 5596

dx.doi.org/10.1021/ef400456n | Energy Fuels 2013, 27, 5595−5606

Energy & Fuels

Article

Represent the direct and interacting effects of the input parameters on the responses via two-dimensional (2D) or three-dimensional (3D) plots. (iv) Determine a set of input parameters based on the response surface to fulfill the optimum response condition. The optimum condition may not fall in the range of variables examined. (v) Verify the model by repeating the obtained optimal input set, to check for the prediction validity of the model as well as repeatability of the output response.32,35,36 Full factorial, partial factorial, and central composite rotatable design are the most commonly DOE methods used for process analyses. Central composite design could be an effective alternative for this study because it gives almost the same information as the threelevel factorial design while requiring many fewer tests than the full factorial design. Also, central composite design is one of the most common designs fitting second-order polynomials, which can reveal nonlinear interactions between parameters as well as linear interactions that are important in this study.32,37 The total number of tests in central composite design is a function of the standard factorial runs, 2k, axial runs, 2k, and replicate tests at center point, nc, where k is the number of input variables. Replication tests at center point of the design are very important because they provide an independent estimation of the experimental error. Two replications are often sufficient, but for a three-factor design, the recommended number of replication tests by Design Expert software 8.0.7.1, is six to obtain an acceptable accuracy when verifying the results.32,35,37,38 In the RSM, if the independent input parameters (x1, x2, ..., xk) are assumed to be continuous with a negligible measurement error, then the random response variable (y) could present functionally as in eq 133−36,39

passing a vessel containing a bed of solid particles, the bed expands and particles become suspended in a way that drag force imported by the upward fluid equals the weight of the particles. The fluid velocity in which onset of fluidization occurs is known as the minimum fluidization velocity.29 To achieve the highest separation in ADMFB, it is desirable to operate the bed above the minimum fluidization velocity but close enough to minimum fluidization to achieve full bed fluidization and stability and also avoid excessive bubbling or turbulence that results in back mixing of the materials by gas bubbles in the bed.29−31 2.1. Experimental Setup. The batch ADMFB unit used in the current experiments is made of a 40 cm tall and 20.2 cm inner diameter vertical Plexiglas cylinder. A metallic porous plate with an average pore size of 40 μm and thickness of 0.3 cm (Matt Corporation, Farmington, IL) is used as the air distributor. Air was adjusted by an Aalborg GFC67A thermal mass flow controller with an accuracy of ±1%. A Nederman filter box (Helsingborg, Sweden) is used to collect the entrained fine dust in the exhaust air. The filtered air was then discharged into the atmosphere. A schematic diagram of the experimental setup is presented in Figure 1. Silica sand with a density of 2.62 g/cm3 from SIL Industrial Minerals is used as the fluidization medium. Because the size of medium particles is one of the dominant parameters affecting the minimum fluidization velocity and then the fluidization regime in the bed, the sand was screened to a narrow particles size distribution. For this purpose, the two size fractions (350− 500 μm) of sand were selected and mixed to generate the final bed medium particles with an average particle size of 390 μm, which can be classified as Geldart group B particles.28 2.2. Coal Sample Preparation. The Boundary Dam (BD) lignite coal, provided by Sherritt, Inc., first crushed down to smaller than 13.2 mm and then divided into several 1−2 kg index samples by a riffle after homogenizing. Particles of each index sample were divided into four size fractions, namely, very fine (0−1 mm), fine (1−2.8 mm), middle (2.8−5.6 mm), and coarse (5.6−13.2 mm). Three size fractions (fine, middle, and coarse) were used in the current tests. Similar size fractions of index samples were mixed together, for homogenization purposes, to obtain the necessary amount of each size fraction to perform the separation tests. The average ash content of the fine, middle, and coarse feed samples prepared for the beneficiation tests was measured (ASTM Method D3174) to be 14.4, 12.5, and 11.7%, respectively. The average volatile matter (ASTM Method D3175) of all three size fractions was measured to be 40.5% (±0.5%), regardless of the size fraction. The moisture content (ASTM Method D3173) of the fine, middle, and coarse particles was 21.5, 20, and 17%, respectively.

y = f (x1, x 2 , ..., xk) + ε k

= β0 +

k

k

k

∑ βi xi + ∑ βiixii 2 + ∑ ∑ βijxixj + ε i=1

i=1

(1)

i T. The perturbation plot for recovery is presented in Figure 3b. The effect of all factors could be assessed simultaneously by the perturbation plot, which shows how the response behaves as each factor moves in the defined range, while all other factors are held constant at a reference point. A steep slope or curvature in a factor shows that the response is sensitive to that factor, while a relatively horizontal flat line shows insensitivity to that particular factor.35,38,43 It should be mentioned that coded values are used for this plot and the midpoint of all variables is chosen as the reference point while plotting. As seen from Figure 3b, all three factors inversely affect the organic material recovery to L1. This means that better recovery of coal to a clean product can be obtained at lower levels of the operating conditions. The bed height affects recovery more intensively as the range of variations imposed by changes of the bed height throughout the tested parameter range is the greatest compared to other parameters. The recovery increases from 64.6 to 84.6% when 15 cm H is replaced with 25 cm H, while T and V are kept at their central levels. As eq 4 indicates, there is a significant interaction between air superficial velocity and separation time at a 95% confidence level. This interaction is further illustrated in Figure 4a. An interaction occurs when the effect of one factor depends upon the level of the other factors.37,38 The non-parallel or crossed lines in such plots confirm the presence of an interaction between two variables. The general shape of the interaction graphs are not changing with the H level; therefore, the middle level of H is assumed when generating both interaction plots. As Figure 4a shows, there is an interaction between V and T in a sense that the effect of V on recovery is dependent upon T and vice versa. Because of this dependency, the recovery is higher for low separation times across the V range than the high separation time. The difference between high and low separation time recoveries decreases with an increasing V. According to Figure 4a, when both V and T are set at a lower limit, the recovery of 92% is achievable while becoming lower to 68% at higher levels of V and T, considering the central level of H. Considering the positive effect of interactions on the recovery and negative effect of parameters themselves, there seems to be an opposing effect involved with V and T. This opposing effect also might cause the optimum condition to not occur at exact lower levels of V and T. Thus, employing the optimization methods is necessary to make sure that the optimum condition is achieved. Figure 4b presents the effect of V and T and their mutual interaction on the recovery in a contour plot for a bed height of 20 cm. The recovery values for different sets of air velocities and separation times can be extracted from this figure. Figure 4b indicates that the recovery is more sensitive to a step size change of V and T at their lower levels, where small changes in

interpreting the results than the one factor at a time methods. Usually, a significance level of 95% or a p value less than 0.05 is considered in engineering analyses. Thus, terms with p values less than 0.05 are considered to be significant on the proposed response parameter. Terms with 0.1 p values are marginally significant (which are eliminated here). R2 evaluates the proximity of the model predictions to the experimental results. R2 is a relative value ranging between 0 and 1, where high values (very close to 1) imply accurate prediction of the experimental data by the model.31,41,40 For instance, a 0.9 R2 value indicates that 10% of the total variations are not explained by the model.40,41 Adjusted R2 is calculated for the model here to compensate for the increase of R2 by adding new terms.35,40 The lack of fit test is used to determine whether discrepancies between measured and expected values of the fitted model are originating from random or systematic error. The diagnosing procedure involves the comparison of residual error to the pure error based on the replicated run results, in a threshold 95% confidence level or p values of 0.05. A significant lack of fit means that the suggested model is not fitting all of the design points well.35,36,40,42 4.1. Effect of the Operating Parameters on the Organic Material Recovery. The combined results of all experiments (Table 2) were used to find the quadratic polynomials that best represent the effect of tested variables on the organic material recovery. Equations 4 and 5 present the normalized and actual correlations obtained for recovery as a function of three operating parameters and their interactions, respectively. In both models, the insignificant terms are eliminated. The X1, X2, and X3 symbols are normalized representatives of V, T, and H variables, respectively. R C% to L1 = 74.83 − 7.78X1 − 3.83X 2 − 9.9X3 + 2.71(X1X 2) + 2.88X12

(4)

R C% to L1 = 610.3 − 50.71Vcm/s − 0.32Ts − 1.98Hcm + 0.017(Vcm/sTs) + 1.28Vcm/s 2

(5)

Table 4 exhibits the detail of the ANOVA tests on the collected responses and the model terms. A very low value of the Fisher’s F test (i.e., p value < 0.0001) indicates that the model is statistically significant. Also, the lack of fit of the model was found to be non-significant because the p value was equal to 0.3573.41,42 Therefore, the model is suitable for predicting the target response within the range of variables tested here. Table 4. ANOVA Test Results for Coal Recovery source model X1, V X2, T X3, H X1X2 X12 residual lack of fit pure error corrected total

sum of squares

degree of freedom

mean square

1831.7 605.16 146.73 979.8 58.67 41.33 133.65 96.56 37.09 1965.35

5 1 1 1 1 1 14 9 5 19

366.34 605.16 146.73 979.8 58.67 41.33 9.55 10.73 7.42

F value

p value, prob > F

38.37 63.39 15.37 102.63 6.15 4.33

T, where only the V and T interaction is significant at a 95% confidence level. (3) Two interactions between V and T or H are found to be significant for the system separation efficiency. Main operating parameter effectiveness is in the order of T > H > V. (4) Because of the low ash content of feed samples and narrow range of clean coal ash contents, the designed experiments were not able (sensitive enough) to reveal possible interactions or suggest a mathematical equation for predicting the clean coal ash content. Ash removal of 13.8−27.6% relative to each test feed is obtained for conducted experiments. (5) Optimization of the mathematical equation for maximizing organic material recovery resulted in the optimum condition of V = 15 cm/s, T = 90 s, and H = 15 cm, predicting organic material recovery to be 97.36%. Also, the optimum condition for maximizing system separation efficiency was found to be V = 15 cm/s, T = 300 s, and H = 25 cm, predicting ADMFB separation efficiency 5605

dx.doi.org/10.1021/ef400456n | Energy Fuels 2013, 27, 5595−5606

Energy & Fuels

Article

(39) Aslan, N. Modeling and optimization of multi-gravity separator to produce Celestite concentrate. Powder Technol. 2007, 174, 127− 133. (40) Sen, R. Response surface optimization of the critical media components for the production of Surfactin. J. Chem. Technol. Biotechnol. 1997, 68 (3), 263−270. (41) Vazifeh, Y.; Jorjani, E.; Bagherian, A. Optimization of reagent dosages for copper flotation using statistical technique. Trans. Nonferrous Met. Soc. China 2010, 20, 2371−2378. (42) Borges, G. D.; Vieira, F. G.; Copetti, C.; Gonzaga, L. V.; Fett, R. Optimization of the extraction of flavanols and anthocyanins from the fruit pulp of Euterpe edulis using the response surface methodology. Food Res. Int. 2011, 44, 708−715. (43) Liong, M. T.; Shah, N. P. Optimization of cholesterol removal by probiotics in the presence of prebiotics by using a response surface method. Appl. Environ. Microbiol. 2005, 71 (4), 1745−1753.

(16) Sampaio, C. H.; Aliagaa, W.; Pacheco, E. T.; Petter, E.; Wotrub, H. Coal beneficiation of Candiota mine by dry jigging. Fuel Process. Technol. 2008, 89, 198−202. (17) Luo, Z.; Zhao, Y.; Tao, X.; Fan, M.; Chen, Q.; Wei, L. Progress in dry coal cleaning using air-dense medium fluidized beds. Int. J. Coal Prep. Util. 2003, 23, 13−20. (18) Lubin, W. Mechanism and application of coal dry beneficiation with air-dense medium fluidized bed. J. Cent. South Univ. Technol. (Engl. Ed.) 1998, 5 (2), 100−103. (19) Mohanta, S.; Chakraborty, S.; Meikap, B. C. Influence of coal feed size on the performance of air dense medium fluidized bed separator used for coal beneficiation. Ind. Eng. Chem. Res. 2011, 50 (18), 10865−10871. (20) Mohanta, S.; Daram, A. B.; Chakraborty, S.; Meikap, B. C. Applicability of the air dense medium fluidized bed separator for cleaning of high-ash Indian thermal coals: An experimental study. S. Afr. J. Chem. Eng. 2011, 16 (1), 50−62. (21) Mak, C.; Choung, J.; Beauchamp, R.; Kelly, D. J. A.; Xu, Z. Potential of air dense medium fluidized bed separation of mineral matter for mercury rejection from Alberta sub-bituminous coal. Int. J. Coal Prep. Util. 2008, 28, 115−132. (22) Van Houwelingen, J. A.; De Jong, T. P. R. Dry cleaning of coal: Review, fundamentals and opportunities. Geol. Belg. 2004, 7 (3−4), 335−343. (23) Luo, Z.; Zhao, Y.; Chen, Q.; Tao, X.; Fan, M. Separation lower limit in a magnetically gas−solid two-phase fluidized bed. Fuel Process. Technol. 2003, 85, 173−178. (24) He, Y.; Zhao, Y.; Chen, Q. Fine particle behavior in air fluidized bed dense medium dry separator. Int. J. Coal Prep. Util. 2003, 23, 33− 45. (25) Yang, X.; Zhao, Y.; Luo, Z.; Song, S.; Duan, C.; Dong, L. Fine coal dry cleaning using a vibrated gas-fluidized bed. Fuel Process. Technol. 2013, 106, 338−343. (26) Sahu, A. K.; Tripathy, A.; Biswal, S. K.; Parida, A. Stability study of an air dense medium fluidized bed separator for beneficiation of high-ash Indian coal. Int. J. Coal Prep. Util. 2011, 31, 1−22. (27) Sahu, A. K.; Tripathy, A.; Biswal, S. K. Study on particle dynamics in different cross sectional shapes of air dense medium fluidized bed separator. Fuel 2013, 111, 472−477. (28) Geldart, D. Gas Fluidization Technology; John Wiley and Sons, Inc.: Hoboken, NJ, 1986; ISBN: 0471908061. (29) Kunii, D.; Levenspiel, O. Fluidization Engineering, 2nd ed.; Butterworth-Heinemann: Oxford, U.K., 1991; ISBN: 9780409902334. (30) Rhodes, M.; Introduction to Particle Technology, 2nd ed.; John Wiley and Sons, Inc.: Hoboken, NJ, 2008; ISBN: 978-0-470-01427-1. (31) Delebarre, A. B.; Morales, J.; Ramos, L. Influence of the bed mass on its fluidization characteristics. Chem. Eng. J. 2004, 98, 81−88. (32) Aslan, N. Application of response surface methodology and central composite rotatable design for modeling and optimization of a multi-gravity separator for chromite concentration. Powder Technol. 2008, 185, 80−86. (33) Oraon, B.; Majumdar, G.; Ghosh, B. Application of response surface method for predicting electroless nickel plating. Mater. Des. 2006, 27, 1035−1045. (34) Vejahati, F.; Katalambula, H.; Gupta, R. Entrained-flow gasification of oil sand coke with coal: Assessment of operating variables and blending ratio via response surface methodology. Energy Fuels 2012, 26 (1), 219−232. (35) Karimipour, S.; Gerspacher, R.; Gupta, R.; Spiteri, R. J. Study of factors affecting syngas quality and their interactions in fluidized bed gasification of lignite coal. Fuel 2013, 103, 308−320. (36) Pathirana, C. L.; Shahidi, F. Optimization of extraction of phenolic compounds from wheat using response surface methodology. Food Chem. 2005, 93, 47−56. (37) Montgomery, D. C. Design and Analysis of Experiments; John Wiley and Sons, Inc.: Hoboken, NJ, 2009; ISBN: 978-0-470-12866-4. (38) Stat-Ease, Inc. Design-Expert, Version 8.0.7.1; Highly Efficient Design of Experiments (DOE)User’s Guide; Stat-Ease, Inc.: Minneapolis, MN, 2010. 5606

dx.doi.org/10.1021/ef400456n | Energy Fuels 2013, 27, 5595−5606