630
Ind. Eng. Chem. Process Des. Dev. 1981, 20, 630-636
Thermogravimetric Analysis of Biomass. Devolatilization Studies on Feedlot Manure Pattabhl Raman, Walter P. Walawender,' L. T. Fan, and Jerald A. Howell Department of Chemical Engineering, Kansas State Univemlty, Manhattan, Kansas 66506
The devolatilization reactions of feedlot manure were studied by means of a thermogravimetrlc analyzer. The effects of size fraction, heating rate, and purge gas rate on the thermogram characterfstk, such as the total weight percent devolatilized and the temperature at which the maximum rate of weight loss occurred, were analyzed statistically. The results indicated that the total weight percent devolatilized was mainly influenced by the size fraction, while the temperature at which the maximum rate of devolatilization occurred was mainly affected by the heatlng rate. The devolatilization data were fitted to a multireaction model, and the kinetic parameters were determined. The parameters Em,the mean value of the activation energy, and 6,the standard deviation of the activation energy, were analyzed statistically to assess their dependence on the heating rate, size fraction, and purge gas rate. The results indicated that Emwas not influenced by any of these variables, while 6 was influenced by the heating rate and the purge gas rate.
Introduction Shortages and the upward spiral in oil prices have given impetus to research on finding new sources of energy. Biomass is one such renewable, low sulfur energy resource that has merited attention. Thermochemical processes such as pyrolysis and gasification at atmospheric pressure have been proposed for converting biomass materials into useful energy products. When cellulosic materials, e.g., biomass, are heated to high temperatures in the absence of oxygen, a series of chemical reactions takes place that results in evolution of volatile products and a carbonaceous solid residue. Experimental investigations with pure cellulose have shown that these reactions take place in two principal steps (Antal et al., 1978). The first step is devolatilization, which produces volatile matter and char. The second step consists of secondary reactions involving the evolved volatiles in the gas phase and the solid char. The final product distribution for gasification is dictated by both the devolatilization and the secondary reactions. Thus, the design and development of a process for the gasification of manure will require a basic knowledge of kinetic data on the devolatilization as well as the secondary reactions involved. Furthermore, it is essential to know the temperature at which the devolatilizationreactions are initiated and terminated as well as the extent of devolatilization at a given temperature. Thermal analysis is one of the most commonly used methods for studying devolatilization reactions. In this method, slow heating is necessary to maintain stability within the instrument, since the weight loss of the sample is continuously monitored with respect to the change in temperature. In a thermal analysis experiment, a captive sample is heated or cooled according to a predetermined temperature-time program, while a physical property of the sample such as its weight is continuously monitored and recorded as a function of temperature. Though the sensitivities of the available thermal analysis equipment are excellent, the configuration of the instrument and experimental parameters are likely to influence the data generated. Heating rate, sample size, purge gas flow rate, and pressure are some of the factors that may be important. A survey of the literature on the kinetics of devolatilization of biomass has indicated that the kinetic data available are very limited. Antal et al. (1978) performed thermogravimetric studies on pure cellulose, newsprint, 0196-4305/81/1120-0630$01.25/0
hardwood, softwood, and cow manure in an inert environment as well as in steam. They used a Dupont 951 thermogravimetric analyzer with a temperature range of ambient to 873 K. Their data indicated that changing the heating rate resulted in a lateral shift in the thermograms. Havens et al. (1971) performed thermogravimetric analysis and differential scanning calorimeter studies on the pyrolysis of wood. They varied the heating rates from 20 K/min to 160 K/min. They also observed a lateral shift in their thermograms with respect to the heating rate. They developed a heat transfer-thermal decomposition model to explain their data. Maa and Bailie (1973) performed kinetic studies with wood samples in a fluid bed reactor as well as with a thermogravimetric system. They conducted experiments with 0.005 m diameter wood samples in an isothermal TGA and developed a mathematical model based on shrinking core kinetics to describe the results. They used 0.032 m diameter wood samples in the fluid bed reador to check the validity of their model based on the TGA data. They were able to predict the pyrolysis time of the samples with their model. Barooah and Long (1976) investigated the rates of pyrolysis of materials such as sawdust, granular celluloee, and sucrose-impregnated pumice. They used a 0.076 m (3 in.) i.d. fluidized bed operating with nitrogen as the fluidizing gas. Their work focused mainly on the char yields and their results indicated that the pyrolysis of these materials takes place in two stages. The first stage is a first-order reaction and the second stage a second-order reaction. Anthony et al. (1974) studied the devolatilization of coal in a rapid-heating furnace. The heating rates employed were around loo00 K/s and the mean particle size used ranged between 50 and lo00 pm. The heating rates used in this study were substantially higher than the ones used in a TGA (about 5 K/s) and fluidized beds (about lo00 K/s). They concluded that the volatiles yield increased with decreasing particle size and increased slightly with increasing heating rate. The devolatilization reactions were conducted over a temperature range of 673 to 1373 K. The devolatilization studies reported thus far for biomass materials have not taken into consideration the influence of the experimental parameters on the thermograms generated and the kinetic parameters. The total fraction of the biomass devolatilized and the temperature location of the maximum rate of weight loss are the two major characteristics of a thermogram. Also, the kinetic param0 1981 American Chemical Society
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 4, 1981 631
Table I. A Typical Analysis of Manure dry ash free wt %
FURNACE ASSEMBl
P W G E GAS
Figure 1. Perkin-Elmer TGS-2.
eten for devolatilization,namely, the order of reaction, the preexponential factor, and the activation energy, have not been analyzed to assess their dependence on the experimental parameters. The objectives of the present work were: (1) to study the devolatilization characteristics of a biomass material, namely, feedlot manure, with the thermogravimetric analyzer (TGA); (2) to assess the influence of the operating parameters of the instrument as well as the size fraction of the material on the thermograms generated; (3) to determine kinetic parameters for the devolatilization reaction; and (4) to assess the influence of the operating conditions and size fraction on the kinetic parameters. Experimental Section Equipment. A Perkin-Elmer Model TGS-2 thermogravimetric analyzer was used in this study to continuously monitor weight changes in manure due to devolatilization as the sample followed a linear heating program. This instrument was chosen on the basis of its sensitivity (0.1 pg max), range of heating rates (0.3125 to 320 K/min), convenience of operation, and ability to use any type of gaseous environment. The furnace area, shown in Figure 1, was separated from the weighing chamber in minimize any effect of temperature on the weighing mechanism. The weighing mechanism operated on the null-balance principle. Balance response time could be selected by using a low, medium, or high filter and corresponding response times on the TGS-2 were 0.7, 1.2, and 4 s, respectively. This flexibility permitted less noise at the higher sensitivities. The instrument was capable of handling samples up to 1300 mg with six full-scale range settings from 0.01 mg to lo00 mg. The flow path of the purge gas through the system is shown in Figure 1. Nitrogen waspused as the purge gas and a positive pressure was maintained through the weighing chamber in order to protect the balance mechanism from the condensables formed. The furnace assembly consisted of a platinum wire element wound around a cylindrical aluminum support. This element acted alternately as a heater and temperature sensor. The sample was placed in a platinum pan, 5.8 mm in diameter and 1.8 mm deep. The sample material was heated by both radiation and convection by means of the purge gas entering the furnace chamber. A chromel-alumel thermocouples was positioned 1 to 2 mm below the pan. The low internal mass of this furnace configuration allowed rapid temperature equilibrium and, therefore, permitted the instrument to attain what is probably the fastest heating rate available today for equipment of this kind. Data collection was accomplished using a Bascom-Turner Model 8110 plotting microprocessor, which allowed permanent data storage as well as data manipulation.
carbon hydrogen nitrogen oxygen moisture ash
50.1 6.9 4.0 39.0 8.0 (as received basis) 10.7 (as received basis)
T e m p e r a t u r e Calibration. The thermocouple mounted in the furnace was calibrated so that it monitored the temperature not only of the furnace enclosure but also of the actual location at which the sample would be placed. This calibration was performed by placing a sample of a ferromagnetic metal with a precisely known Curie point temperature in the sample pan. The weight of this metal was electrically suppressed. The furnace tube was then surrounded by a magnetic field which pulled down on the metal sample. Upon heating, the point at which the sample lost its observed weight indicated that the contents of the sample pan were at the Curie point temperature. The thermocouple gain potentiometer was adjusted to indicate this temperature at the point where the weight loss occurred. Two metals were used for calibration; alumel with a Curie point temperature of 436 K and iron with a Curie point temperature of 1073 K. Sample Preparation. The manure used in this study was collected from paved feedlots a t Kansas State University’s Beef Research Center. It was flash dried to reduce its moisture content to about 8%, ball milled, and then size classified. Three size fractions, namely, -325 +400 mes, -120 +170 mesh, and -20 +40 mesh, were chosen for this study. The samples used for experimentation were dried in an oven for 24 h at 378 K and stored in a desiccator. A typical ultimate analysis of the manure used is presented in Table I. Procedure. The devolatilization experiments were conducted in a nitrogen atmosphere by placing a dried manure sample of about 1.0 mg on the sample pan. To reduce possible heat and mass transfer effects, the material was spread in a uniform monolayer in the sample pan. The sample was heated from 403 to 1223 K at a preset heating rate and purge gas flow rate. The weight of the sample was recorded for every 1.64 K of temperature rise over the temperature range. For each experiment, 500 data points were collected. At the end of every run, oxygen was introduced into the system to burn the residual char in order to determine the ash content of the sample. Operating Conditions. The principal experimental variables which could affect the devolatilization results in a TGA are the pressure, the purge gas flow rate, the heating rate, and the size fraction of the sample used. As the focus of this study was on the devolatilization of manure at atmospheric pressure, the pressure used in this study was atmospheric and the purge gas used was nitrogen. Two purge gas rates, namely 50 and 150 cm3/min, were chosen since they represented two extremes that were possible in the experimental setup. The heating rates of 40 and 160 K/min were chosen since they represented the extreme values under which stable thermograms could be obtained. The size fractions were chosen in such a way that the size variation was large enough to detect the influence on the thermograms. Also, the particle size was chosen small enough so that the influence of heterogeneity of sample makeup was minimized. In the case of -20 +40 mesh fraction only three of four particles made up a sample. The uniformity of the sample was maintained by using particles of the same type. The operating parameters used are presented in Table 11. Two thermograms were
632 Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 4, 1981 Table 11. Experimental Conditions 40 K/min and 160 K/min heating rates -20 +40 mesh (420 t o 840 pm) particle size ranges -120 +170 mesh (88 t o 125 pm) -325 +400 mesh (37 to 44 pm) dry nitrogen purge 50 and 150 cm3/min rates pressure atmospheric 0.5 to 1.5 mg sample sizes
Table 111. Average Values of Parameters heating rate, particle size, mesh K/min means maximumrate of weight loss temp, K total devolatilization
160
-325 +400
-120
40
+170
-20 +40
595.0
608.8
599.0
607.6
599.1
85.0
84.6
87.1
85.4
81.9
15.0
15.4
12.9
14.6
18.1
%
char daf, %
Table IV. Statistical Analysis of Thermograms. F Ratio (Critical F Ratio: 4.75 for a = 0.051
ac
factor
variable
maximum rate 9.14ga of weight loss temperature total 0,157 devolatilization percentage (daf)
20t
C4 i-
:
4rC
5i^
T C
9OC
6OC
-ENPERATUSE
,
i
; _ _ t _ _
ECC
X0
OC
heating rate
l2GC
particle size
nitrogen purge rate
1.558
0.191
6.557*
0.098
2. Letting x = -Y
- + - + - - + -E+ - + x x2 x3 x4 x5
x6
HI\
(A7)
x7
or
The coefficients B through H of the polynomial in eq A7 are the same as those of eq A6. The value of A' is -0.0000035. Replacement of the temperature integral in eq A6 by the approximation in eq A8 results in the final form of the correlation
wo-w,=v * - v = , A Hooke and Jeeves (1961)pattern search technique was used to estimate the parameters E,,, and u. The objective function to be minimized was the sum of the squares of the differences between the experimental weights remaining and those predicted from eq A9. Nomenclature E = activation energy, kcal/mol E, = mean activation energy, kcal/mol f ( E )= activation energy distribution k = rate constant, s-l ko = frequency factor, s-l R = gas constant, kcal/mol K t = time, s t ' = exponential term, -E/RT T = temperature, K To = initial temperature, K V = volatile matter, % = [( Wo- W)/ Wo]x 100 V* = potential volative matter, % = [(W,- W f ) Wo] / X 100 W = weight remaining, % Wo = initial wt, % Wf = final wt, % x = exponential term, -E/RT y = exponential term, -E/RT Greek Letters fi = linear heating rates, K/s u = standard deviation from mean activation energy, kcal/mol Subscripts i = ith parallel reaction Literature Cited Antal, M. J.; Edwards, W. E.; Friedman, H. L.; Rogers, F. E. "A Study of the Steam Gaslflcatlon of Organic Wastes": Project Report to EPA, W. W. Llberlck Project Officer, 1978. Anthony, D. 6.; Howard, J. 6.; Hottel, H. C.; Melssner, H. P. "Rapid Devolatc llzation of Pulverized Coal"; Fifteenth Symposium (International) on Com
838
Ind. Eng. Chem. Process Des. Dev.
bwtlon, The Combustlon Institute, Pittsburgh, Pa., 1974,p 1303. Barooah, J. N.; Long, V. D. Fuel 1976, 55, 116. Benson, S. "Thermochemical Kinetics"; Wiiey: New York, 1968. Berkowitz, N. Fuet 1960, 39, 47. Cheong, P. ti. Ph.D. Thesis, California Institute of Technology, Pasadena,
1976. Havens, J. A.; Weiker, J. R.; Sliepcevich, C. M. J. fire Flammsbi/#y1971,2 ,
321. W e , R.; Jwves, T. A. Assoc. Comput. Mach. J . 1961, 8 , 212. Howell, J. A. M.S. Thesis, Kansas State University. Manhattan, 1979. Jahnke. E.; Emde, E.; Losch, F. "Tables of Hlgher Functions", 6th ed.; McGraw-Hill: New York, 1960.
1981, 20, 636-640
Mea, P. S.; Bailie, R. C. Combust. Sci. techno/. 1973. 6 , 1.
Pitt, G. J. Fue/1962, 41, 267. Raman, K. P.; Walawender, W. P.; Fan, L. T. Prepr ACS Div. Fuel Chem., 1980, 25(4),233. Snedecor, G. W.; &&ran, W. G. "Statistical Methods";Iowa State University Press, Ames, Iowa, 1978. Whistler, R. L.; Smart, C. L. "Polysaccharide Chemistry"; Academlc Press: New York, 1963.
Received for review August 5, 1980 Accepted June 22, 1981
Formation of Drops and Bubbles in Flowing Liquids Yoshlnorl Kawase and Jaromlr J. Ulbrecht' Department of Chemical Engineering, State University of New York at Buffalo, Buffalo, New York 14260
A model of the process when bubbles and drops are formed at a nozzle submerged vertically in a continuous phase flow has been developed by simulating the influence of the continuous phase flow by a virtual inclination of the nozzle. Predictions were compared with available experimental data and the agreement was found to be very satisfactory. The model has two versions, one for low flow rates of the dispersed phase and the other for high flow rates. The model developed in this work provides a simple but a reasonably accurate means of estimating the diameter of drops and bubbles formed in a flow normal to the nozzle axis.
Introduction Most of the literature dealing with the formation of fluid spheres (drops and bubbles) is concerned with the fluid release from a nozzle or an orifice surrounded by a liquid at rest (see the review by Kumar and Kuloor, 1970). Fluid spheres, however, are in most cases formed in liquids moving past nozzles or orifices of various types of equipment, such as perforated tray column, valve-cap tray column, bubble aerator, mixer settlers, fermentation reactors, and others. In spite of this fact less attention has been given to the formation of fluid spheres in flowing liquids and very little is known about the effect of the motion of the liquid on the formation of fluid spheres. Chuang and Goldschmidt (1970) and Sada et al. (1978) considered bubble formation in co-flowing streams, i.e. for the case when the streamlines of the continuous phase are parallel with the axis of the nozzle. The effect of the velocity of a liquid flowing past a horizontal, submerged orifice on the formation of air bubbles was investigated, and dimensionless empirical equations were proposed by Sullivan et al. (1964). Kumar and Kuloor (1970) suggested that the reduction in bubble size due to the momentum transport from the moving continuous phase could be attributed to an extra upward drag force which adds to the bubble's buoyancy and they presented an equation to estimate the final bubble volume. Itoh et al. (1979a) investigated drop formation in a uniform flow and compared the experimental data with their empirical model. In the case of co-flowing streams, their model provided an acceptable agreement with the experimental results. However, in the case where uniform stream flows normal to the nozzle axis, their model was not in satisfactory agreement with the experimental data. The formation of drops in a stirred tank was analyzed by Itoh et al. (1979b) using their empirical model for drop formation in a uniform flow mentioned above. The agreement between the predicted and the measured drop sizes was reasonable.
The purpose of this work is to develop a new model for the formation of drops and bubbles in a uniform stream flowing normal to the nozzle and to compare the predictions of this model with available experimental data. Formation of Drops and Bubbles in Flowing Liquids. Consider a drop being formed at a nozzle in a flow normal to its axis. This is recognized as being a fundamental flow situation for the formation of drops in a plate column or in a stirred tank. In the following, we shall initially follow the line of thoughts proposed by Kumar and Kuloor (1970). This model, which assumes that the bubble is formed in two stages, the expansion stage and the detachment stage, has been used to investigate the formation of a drop in a quiescent liquid and it agrees well with experimental data (Kumar and Kuloor, 1970). The expression for the final drop volume may be written as the sum of the volumes obtained from the two stages. Thus v = VE + Qt, ( 1) where VE is the force-balance drop volume and Qt, is the volume entering the drop during detachment. There are four major forces which act on a drop during the process of formation at a nozzle submerged in a quiescent liquid: the buoyancy force due to the density difference between the two fluids and the inertial force associated with fluid flowing out of the nozzle which acta to separate the drop from the nozzle, while the interfacial tension force at the nozzle tip and the drag force due to the upward movement of the expanding drop act to keep the drop attached to the nozzle. The expansion stage is assumed to end when the upward forces are equal to the downward forces. (buoyancy force) + (inertial force) = (interfacial tension force) + (viscous drag force) (2) From this equation, the force-balance drop volume VE is determined. Substituting the quantitative expressions for
om-4305iaiii 120-0636~01.2sio 0 1981 American Chemical Society