Effect of Temperature, Sample Size, and Gas Flow Rate on Drying on

Jan 1, 1994 - Beulah-Zap lignite and Wyodak-Anderson coal (-100 and -20 mesh from the Argonne Premium. Coal Sample Program) were dried in nitrogen ...
0 downloads 0 Views 458KB Size
Energy & Fuels 1994,8, 320-323

320

Effect of Temperature, Sample Size, and Gas Flow Rate on Drying on Beulah-Zap Lignite and Wyodak Subbituminous Coal Karl S. Vorres Chemistry Division, Building 21 1, Argonne National Laboratory, Argonne, Illinois 60439 Received September 14, 1993. Revised Manuscript Received November 22, 199P

Beulah-Zap lignite and Wyodak-Anderson coal (-100 and -20 mesh from the Argonne Premium Coal Sample Program) were dried in nitrogen under various conditions of temperature (20-80"C), gas flow rates (20-160 cm3/min), and sample sizes (20-160 mg). An equation relating the initial drying rate in the unimolecular mechanism was developed to relate the drying rate and these three variables over the initial 80-85 % of the moisture loss for the lignite. The behavior of the WyodakAnderson subbituminous coal is very similar to that of the lignite. The nitrogen BET surface area of the subbituminous sample is much larger than the lignite. The larger area does not lead to a more rapid rate of drying.

Introduction Economicalproduction of synthetic coal liquids suitable for transportation fuels is a continuing goal of a number of coal conversion programs. The vast reserves of lowrank coals in the western part of the U.S.provide a plentiful input for potential processes. The low-rank coalsgenerally possess a high reactivity in many chemical reactions which also make them attractive feedstocks. One of the major detractors, however,is the highmoisture content associated with many of these fuels, ranging up to 40%. An economical means of removing the moisture without sacrificing the reactivity or oil yields is desirable. One of the goals of this work is to describe the kinetics so that an efficient means of moisture removal can be established. A number of studies have been carried out on the drying behavior of coals from a fundamental viewpoint.l4 The drying behavior has been found1-' to follow a unimolecular mechanism as shown by a plot of the logarithm of the water left in the sample against time in a flow of dry gas (nitrogen or carbon dioxide) over a range of temperatures, gas flow rates, and sample sizes. The mechanism proceeds through two stages as shown by two straight-line segments in the plots. The initial stage accounts for about 80-85 % of the moisture loss, while the second accounts for all but about 1% of the rest of the loss. The drying of coals has not been a uniform process, due in part to the nature of the water in the coal. Mraw and co-workers6J have examined the freezing of water in the Abstract published in Advance ACS Abstracts, January 1, 1994. (1)Vorres, K. S.;Kolman,R.; Griswold, T. Prepr. Pap.-Am. Chem. SOC.,Diu. Fuel Chem. 1988,33,(2),333. (2) Vorres, K. 5.: Kolman, R. ReDr. Chem. SOC.,Diu. Fuel - Pap.-Am. Chem. 1988,33,7.(3)Vorres, K.S.;Wertz, D. L.; Malhotra, V. M.; Dang, Y.;Joseph, J. T.; Fisher, R. Fuel 1992,71 (9),1047-1053. (4)Vorres, K.S. Prepr. Pap.-Am. Chem. Soc.,Diu. Fuel Chem. 1992, 37 (2).928. See also: Vorres. K.S.: Wertz. D.: JoseDh. J. T.: Fisher. R. Prepr:Pap.-Am. Chem.Soc.',Diu. h e 1 Chem.'1991,-33(3),853. VoGes, K. S.;Molenda, D.; Dang, Y.; Malhotra, V. M. Prepr. Pap.-Am. Chem. SOC., Diu. Fuel Chem. 1991,36 (l),108. (6)Abhari, R.; Isaacs, L. L. Energy Fuels 1990,4,448. (6)Mraw, S.C.; Silbemagel, B. G. Am. Inst. Phys. Proc. 1981,70,332. (7)Mraw,S.C.;Naae-O'Rourke,D.F. Science 1979,205,901;J. Colloid Interface Sci. 1982,89,268.

pores of coal and observed two types described as "freezable" and "nonfreezable". Their work involved calorimetric measurements over a range of temperatures. Miknis and co-workers studied drying by thermal, microwave, and chemical means.8 Microwave drying had earlier been showns to reduce reactivity of the coal for liquefaction when the microwave drying was continued beyond 75 % . Miknis' group indicated that the moisture loss was consistent for the initial about 75% but a transition was then observed leading to rapidly increasing temperatures, attributed to difficult alignment of the remaining water molecules in the radiation field. These remaining molecules were thought to be associated with the gel structure or pores of the coal. Chemical drying indicated that two types of water were removed sequentially. They indicated that free or surface sorbed water was removed initially, followed by loss of more tightly bound water as the drying agents diffused into the pores of the coal. Moisture is present in coal primarily in the pores and is also associated with various functional groups as well as inorganic species.1° Serio and co-workers studied the effects of moisture and cations on liquefaction of lowrank coals. They concluded that some of the moisture in a coal is associated with the cations. Samples containing carboxylate salts always exhibited water retention. Heating indicated a water evolution peak at 200 "C in TGFTIR experiments. This peak was obscure for demineralized samples. The peak was thought to be due to the evolution of moisture which is ionically bonded on the salt structure. The lowest rank coals contain water in a gel structure that is subject to irreversible change on drying.11 Drying and rehydration isotherms indicate large hysteresis due (8)Miknis,F. P.; Netzel, D. A.;Turner,T. F.Prepr. Pap.-Am. Chem.

SOC.Diu. Fuel Chem. 1993,38,609-617.

(9)Silver, H. F.; Frazee, W. S. Integrated Two Stage Coal Liquefaction Studies; EPRI report AP-4193;University of Wyoming, Laramie, WY, August 1986,460p. (10)Serio,M.A.;Kroo,E.;Teng,H.;Solomon,P.R.Repr.Pap.-Am. Chem. SOC.,Diu. Fuel Chem. 1993,38, 677-686. (11)Verheyen, T.V.; Perry, G. J. In The Science of Victorian Brown Coal:Structure,Properties and Coneequences for Utilization;Durie, R. A., Ed.; Butterworth-Heinemann Ltd.: Oxford, U.K., 1991;Chapter 6, p 279.

0887-0624/94/2508-0320$04.60/00 1994 American Chemical Society

Drying of Lignite and Coal in part at least to these changes.12 An understanding of the drying is linked t o an understanding of t h e structure which holds the water. Suuberg and co-workers indicated that a variety of sites are involved in the drying process.*3 As a result, the apparent activation energies are expected t o reflect a range of values and a simple model may be inappropriate. T h e y also pointed out that water is a good swelling agent for low-rank coals. Thus, drying represents a form of solvent swelling behavior. T h e usual assumptions about the process of moisture loss on drying of a coal sample involve diffusion from t h e pores as well as the external surface. The surface area obtained from BET measurements might then be expected t o correlate with the rate of drying. Larsen and Wernett, however, indicated that the pore structure of the Argonne Premium Coal Samples is not i n t e r ~ 0 n n e c t e d . l If ~ the pores are not connected, then the surface areas may have little to d o with t h e rate of drying. Wertz's X-ray diffraction studies indicated that the structure of the coal undergoesnotable changes on drying.3 He observed the peak corresponding t o t h e graphite-like 002 plane. The peak shape changed during the drying process, indicating a loss of a portion of the peak associated with larger dimensions for the peak. Toward completion of drying t h e diffractogram became sharper with reduced background fluctuation. T h e rates of drying are dependent on a number of parameters including temperature, gas flow rate, and sample thickness. T h e various types of equipment used by different workers led to a desire to indicate rates over a range of these parameters so that work could be more readily compared and possibly serve as a basis for extrapolation to larger scale work. The establishment of t h e relationships t o d o this is one of t h e goals of this work.

Experimental Section Coal drying was done with a modified Cahn Model 121 thermobalance or thermogravimetric analyzer (TGA) attached to an IBM PC/XT microcomputer. Vendor-supplied software was used to monitor the progress of individual runs and convert data files to a form that could be further studied with Lotus 123. The data were obtained as fiies of time, temperature, and weight at 10-8intervals. Run times varied from 7 to 23 h. Sample sizes typically started at about 80 mg, but varied from 20 to 165 mg. Runs were typically isothermal, with temperatures selected from 20 to 80 OC. The gas velocity moving upward past the sample was typically 80 cm3/min (but ranged from 20 to 160 cm3/min) in the 25 mm diameter tube. The thermobalance was modified so that all of the gas flow passed around the sample, rather than partly through the weighingmechanism. The sample was placed in a quartz flat bottom bucket of about 10mm internal diameter. The samples were the -100 mesh Argonne Premium Coal Samples: Beulah-Zap lignite and Wyodak-Anderson subbituminous.16 Samples were quickly transferred from ampules which had been kept in constant humidity chambers with water at room temperature (20 OC or 293 K). In the thermobalance system a (12) Allardice, D. J.In The Science of VictorianBrown CoakStructure, Properties and Consequences for Utilization; Durie, R. A,, Ed.; Butterworth-Heinemann Ltd.: Oxford, U.K., 1991; Chapter 3, p 103. (13) Suuberg, E. M.; Otake, Y.; Deevi, S. C.; Yun, Y. Proceedings of the 1991 International Conference on Coal Science; 16-20 S e p t , Uniuersity of Newcastle-upon-Tyne, United Kingdom; IEA Coal Research Newcastle-upon-Tyne, U.K., 1991. (14) Larsen, J. W.; Wernett, P. Prepr. Pap.-Am. Chem. SOC.,Diu. Fuel Chem. 1992,37, 848-855. (15) Vorres, K.S. Energy Fuels 1990,4 , 420-426.

Energy &Fuels, Vol. 8, No. 2, 1994 321

85 80

a 75 E

v

5 70 .65 60

55

50

1

'

0

I

I

200

400

I 600 800 Time (minutes)

loo0

1200

Figure 1. Wyodak (WY) and Beulah-Zap (ND) coals dried in nitrogen. Conditions were 40 "C, -100-mesh coal, 80 cma/min gas flow, and about 80 mg sample weight. period of about 5 min was used to stabilize the system and initiate data acquisition. A condenser was made to replace the usual quartz envelope and furnace that surrounds the sample. Water was circulated from a constant temperature bath through the condenser to maintain constant temperature during the experiments. This was more stable than the original furnace and provided uniform temperature control during the experiments. The atmosphere for the nitrogen gas rum was cylinder nitrogen (99.99%) or "house" nitrogen from the evaporation of liquid nitrogen storage containers used without further purification. Data were analyzed on a separate microcomputer as reported earlier.I4 Regression analysis was used to obtain the kinetic constants in terms of mg of water lost/g of sample per 10-8time interval. Lotus 123was used for analysis of individual run data. In some runs, approximations to a first and second derivative of the rate expressions were made in Lotus 123 by averaging over a number of the 10-8 time intervals before and after the point of interest. These derivatives were plotted with the rate data to aid in identifying the beginning of the transitions from an initial phase of drying to a seocnd, slower phase. The initial drying kinetic data from a run at a temperature different from room temperature were collected over a time in which the sample was reaching the temperature of the system. These runs indicated that the initial rate increased for about 10-30min until a constant value for the first phase was reached. Figure 1indicates typical weight changes over the duration of a run for the Beulah Zap lignite and the Wyodak-Anderson subbituminous samples. Note that the lignite has a higher initial moisture content and therefore can lose more water. Surface area measurements were carried out to complement the drying kinetic runs. The surface area was assumed to be important in that it determines the amount of exposure to the gaseousphase. The greater the surface area the greater the overall reaction rate will be, assuming a similar reactivity per unit of surface and comparable access to exterior and interior surface. A Quantasorb Jr. instrument was used for nitrogen BET (Brunauer-Emmett-Teller) measurements of the surface, A series of values were obtained on three seta of samples. One set was prepared by drying in the vacuum oven at a range of temperatures. Another set was nitrogen dried samples, prepared both in the vacuum oven with a flow of nitrogen at atmospheric pressure. The third set included the products of some of the thermobalance runs. The typical run gives the data from the desorption cycle. Samples of nitrogen in helium (10, 20, and 30% nitrogen) were allowed to interact with the surface of the sample. Adsorption was carried out at liquid nitrogen temperatures. The amount of adsorbed nitrogen was determined by calibration of a thermal conductivity bridge.

Vorres

322 Energy & Fuels, Vol. 8,No. 2, 1994 Table 1. Surface Area Measurements. runno.

200

0 100

400

300

600

800 lo00 1200 700 900 1100 1300 Time (minutes)

500

Figure 2. Rate plot for Wyodak (WY) and Beulah-Zap (ND) coals dried in nitrogen. Conditions were the same as those given for Figure 1.

Results and Discussion A typical run indicates a rapid initial moisture loss, followed by significantly reduced rates. A plot of the logarithm of the water remaining as a function of time gives two straight line segments followed by a downward slope for the loss of the last 1% . See Figure 2 for typical lignite and subbituminous coal runs. The first line segment has been correlated with the loss of “freezable”water, while the latter has been associated with “nonfreezable” water as identified by Mraw and co-~orkers.~~8 The rate constants for the initial rate loss can be correlated with an Arrhenius plot to give an activation energy for the initial stage of water loss, but the second stage could not be correlated. A multiple regression analysis of the initial rate with the lignite values of the absolute temperature, gas flow, and sample weight was carried out with the data obtained at 10-s intervals from 51 runs. For an equation of the form log initial rate = c1 X Wemperature (K) + c2 X gas flow (cm3/min) + c3 X sample weight (mg) + c4 After the multiple regression analysis it was noted that the R2 value was 0.87. The value seemed low. Plots of the calculated value of the three independent variables as well as log k versus the calculated values of the variables or log k from the regression equation revealed that the scatter was worst for the plot involving flows. The data included 15 runs for which the gas flow rate was zero. The points for the zero flow rates were furthest from the expected values. These runs were removed from the data set, implying that this model will not apply to static moisture loss. Another multiple regression analysis was carried out on the remaining 36 runs. The best fit was given for (value (standard error)) c1 = -1831 (144); ~2 = O.OO0 699 (O.OO0 528); ~3 = -0.005 97 (O.OO0 482); c4 = 3.97, with an R2value of 0.939. These values cover experiments in the range of 20-80 OC, 20-160 cm3/min gas flow, and 20-169 mg sample weight. The activation energy for the first constant corresponds to 8.4 kcal/mol which is close to the heat of vaporization of water. The gas flow term represents an approach to the measurement of the mass transfer coefficient. The constant

drying tepp, surface method C area,mVg

ND173 RM-59-1 RM-59-1 RM-68-2

nitrogen vacuum vacuum vacuum

22 22 22 22

1.993 3.221 2.889 2.969

ND175 RM-67-1 ND177 RM-62-1 RM-63-1 RM-65-1 RM-65-1 RM-66-1

nitrogen vacuum nitrogen nitrogen vacuum nitrogen vacuum vacuum

40 40 60 110 110 150 150 150

1.811 2.141 1.887 1.819 1.768 1.865 1.900 1.764

RM-68-1a vacuum

25

remarks

reproducibility check reproduced from different sample

repeat above after heating overnight in N)at 150 O C

10.61

For comparison a run was made using Wyodak subbituminous coal. c2 for the gas flow term is closeto the value for the standard error for the constant and is small. Another regression analysis was made with the same 36 runs to obtain a fit without the gas flow term. The best fit was given for (value (standard error) c1 = -1862 (144); c3 = -0.005 86 (0.000 479); c4 = 4.11, with an R2value of 0.936. A smaller series of runs (6) made with the -100-mesh Wyodak subbituminous sample at gas flow rates of 80 cm3/minwere analyzed with multiple regression. The best fit over the range of 20-60 “C and weights of 40-80 mg were given by

log k, (initial rate in mg of water left/g of sample/lO-s interval) = -2361 X l/temperature (K) 0.0064 X sample weight (mg) + 5.777 The standard errors for the constants were 194,0.0016, and 0.059, respectively. The R2value = 0.983. The latter value is greater than that for the lignite but does not include the effect of the variation of the gas flow rate. The activation energy corresponding to the coefficient of the 1/ T term is 10.8kcal/mol, close to the heat of vaporization of water. The net effect of these constants is that the rates are very close, almost within the experimental error for a given run, for the two coals at a given set of conditions. The rate of drying may be affected by the surface areas of the samples. The surface area measurement data are given in Table 1. Drying in vacuum results in a surface area that is larger than that for samples dried in nitrogen as long as the temperature of the sample is below 40 OC. Increasing the temperature in nitrogen had only a small effect. Clearlythe subbituminous coal has a much larger surface area, reflecting a greater order in the coal particles and better developed pore structure. The great similarity in rates of drying for the lignite and subbituminous samples indicates that the surface areas of these coals are not proportional to the rates. The lack of an extended interconnected pore network as described by Larsen14may be responsible. Conclusions Generalizations on Lignite Drying. 1. A complete understanding of the rate of lignite drying must include the effects of a number of variables. Some of these are

Drying of Lignite and Coal inherent in the material itself, others depend on the processing given to the material, and still others have to do with mass-transfer effects. 2. The rate of drying is affected by the temperature, gas flow, sample thickness, and history. For the experiments conducted in this study, the rate of drying from the 10 mm diameter container can be expressed by log rate = -1831 X Utemperature (K)+ 0.000699 X gas flow (cm3/min)- 0.00597 sample weight (mg) + 3.97 3. There is a unimolecular mechanism for the initial 60-85 7% of the weight loss. A transition occurs to a slower mechanism until about 1%of the water remains. 4. There are at least two kinds of water present in the low-rank coals. The terms freezable and nonfreezable have been applied to these. The kinetic data support this concept and extend the perception to exchange between the two forms. The initial water loss corresponds to freezable water, while the later loss corresponds to nonfreezable water.

Energy & Fuels, Vol. 8, No. 2, 1994 323 5. The activation energy for the moisture loss of the freezable water is similar to the heat of vaporization of water. The second step, vaporization of the nonfreezable water, does not give data for the unimolecular expression that fit an Arrhenius plot. 6. The method described here is not suitable for the analysis of moisture loss rates in a static system. Generalizationson SubbituminousDrying. 1. The behavior of subbituminous coal on drying is very similar to that of lignite. 2. The mechanism of drying is a unimolecular process. There is a transition after about 80% of moisture loss to a slower unimolecular process. 3. The activation energy for the first step in drying is slightly more than that observed for lignite.

Acknowledgment. The author is grateful for the support of the Office of Fossil Energy, U.S.Department of Energy in this work. The summer assistance of Randy Pippen and Larry Jefferson, high school teachers, is appreciated.