Propagation of the Ignition Front in Beds of Wood Particles

VTT Energy, P.O. Box 1603, FIN-40101 Jyva ¨skyla ¨, Finland

J. J. SAASTAMOINEN* and R. TAIPALE

Lappeenranta University of Technology, Department of Energy Technology, P.O. Box 20, FIN-53851, Lappeenranta, Finland The speed of the propagation of the ignition front downwards against the air flow in a fixed bed of burning wooden particles has been studied. The ignition speed reaches a steady state soon after ignition at the surface, if the inlet air temperature is the same as the initial temperature of the fixed bed. Modeling of the quasi-steady-state speed of propagation of the ignition front is considered. The effects of the flow rate of air, moisture, particle size, density, and wood species on the velocity of the ignition front and on the maximum temperature in the bed are studied experimentally and interpreted by modeling. A maximum speed of propagation was observed with a specific air rate. Correlations for estimating the velocity of the ignition front and maximum temperature in the bed are presented. © 2000 by The Combustion Institute

M. HORTTANAINEN and P. SARKOMAA

NOMENCLATURE a c ˙Љ C Cٞ d ⌬H h K l M m ˙Љ Qٞ qЉ qٞ R r Sٞ T t thermal diffusivity, a ϭ ␭ /( ␳ c ), m s specific heat, J kgϪ1 KϪ1 ˙ Љ ϭ cm flux of heat capacity, C ˙ Љ, W Ϫ2 Ϫ1 m K heat capacity per unit volume, (1 Ϫ ⑀ ) ␳ p c p , J mϪ3 KϪ1 particle diameter, m effective heat of combustion, J kgϪ1 heat transfer coefficient, W mϪ2 KϪ1 effective radiation coefficient heat of vaporization or pyrolysis, J kgϪ1 molecular weight, kg/kmol mass flux, kg mϪ2 sϪ1 energy/volume, J mϪ3 4 heat flux, at ignition front q Љ 0 ϭ KT f , Ϫ2 Wm rate of heat generation per unit volume, W mϪ3 thickness of a plate, radius of a cylinder or a sphere, m coordinate inside a particle, m surface area of particles per unit volume of bed, S ٞ ϭ (1 ϩ ⌫ )(1 Ϫ ⑀ )/ R , mϪ1 temperature, K time, s
author. E-mail:
2 Ϫ1

U U* u v w x x* YO2

dimensionless ignition velocity, U ϭ ˙Љ C ٞ w ig / C g dimensionless ignition velocity, U * ϭ R 2␣ w / a p moisture content (dry basis), mass of water per unit mass of dry solid mass fraction of volatile matter in dry ash-free fuel velocity, m sϪ1 coordinate, m moving coordinate, m mass fraction of oxygen in gas (0.232 for air)

Greek Symbols

␣ ⌫ ␦ ⑀ ␩ ␽ ␪

␭ ␬

*Corresponding @vtt.fi

Jaakko.Saastamoinen

decay coefficient, mϪ1 particle shape factor, 0 for a plate, 1 for a cylinder, 2 for a sphere thickness, m porosity of bed coefficient for heat recovery to gas flow dimensionless gas temperature, ␽ ϭ ( T g Ϫ T ϱ )/( T ig Ϫ T ϱ ) dimensionless particle temperature, ␪ ϭ ( T p Ϫ T ϱ )/( T ig Ϫ T ϱ ) dimensionless parameter, ⌳ ϭ hS ٞ / ˙Љ (␣C g) thermal conductivity of dry layer, W mϪ1 KϪ1 coefficient for char temperature at reaction zone
COMBUSTION AND FLAME 123:214 –226 (2000) © 2000 by The Combustion Institute Published by Elsevier Science Inc.

0010-2180/00/$–see front matter PII S0010-2180(00)01144-9

EXPERIMENTAL Two pot furnaces were used: A at VTT Energy (VTT) and B at Lappeenranta University of Technology (LUT). since wood must not be transported too far due to high costs. 1. The ignition front propagates to the initially cool bed against the flow of air. as depicted in Fig. Average fuel moisture contents were specially prepared to have three different values. c . Experimental results . e ) were used with furnace A. In this paper. knowledge of ignition is important for optimizing gas-phase combustion above the bed to minimize emissions. the average measured bulk density of the beds ␰ ␳ ␴ ⌽ ␾ dimensionless coordinate. Wood and biomass combustion is feasible only in small and medium-scale furnaces. The results can also be applied to traveling grates.67 ⅐ 10Ϫ8 W mϪ2 KϪ4 dimensionless heat flux. kg mϪ3 Stefan-Boltzmann’s constant. where the fuel is ignited on top of the bed and the air flows upwards through the grate and the bed. d . volatile nitrogen species (NH3 and HCN). mass/volume. and NO formed from volatiles is directly related to the speed of the ignition wave inside the bed. with the beneficial side effect of producing energy. Pot furnace A (diam. 244 mm and height 300 mm) was on a balance. ⌽ ϭ ␣ q Љ 0R / 2 [ S ٞ ␭ p ( T ig Ϫ T ϱ )] ϭ ␣ q Љ 0 R /[(1 ϩ ⌫ )(1 Ϫ ⑀ ) ␭ p ( T ig Ϫ T ϱ )] dimensionless ignition parameter. Likewise. ␰ ϭ hS ٞ x * / ˙Љ C g density. flame propagation in underfed combustors is considered. 21] on a grate. 5. 17.IGNITION IN FUEL BEDS 215 and modeling of the flame speed in a bed of wooden particles are presented in this paper. there has been new interest in combustion or gasification [20. ␾ ϭ ( q ٞ / hS ٞ )/( T ig Ϫ T ϱ ) Subscripts 0 ϱ a b c f g ig max p vap vol w at ignition front initial. There is an analogy between a fixed bed and a moving bed [32] and the results can be applied to estimate ignition above traveling grates. i.7 wt % (wet basis) for fuel a . far from ignition front adiabatic dry bed char effective flame seen from the ignition plane gas ignition maximum dry particle vaporization.e. Knowledge of the factors affecting the speed of the ignition front is important. Recently. the combustion air flows against the propagation of the ignition front. the particles were quite long (ϳ35 mm) and thin (thickness ϳ1 mm). Pot furnace B had a square cross-section (150 mm ϫ 150 mm ϫ 900 mm). because of the need to decrease carbon dioxide emissions. Combustion of wood or biomass particles in fixed beds has been studied experimentally [24 –30] and by modeling [25.3. water vapor devolatilization. Ignition in the combustion of coal [9 –13] or wastes [14 –18] on a grate has been studied and also summarized [19]. 31]. Thus. volatiles liquid water INTRODUCTION There is a vast literature on both experiments and models of the propagation of forest and grass fires [1– 8]. and 29. inlet. Some models for the propagation of combustion in fixed beds [22. but their applicability to combustion in fuel beds used for energy production is limited.. One fuel (fuel a ) was used with furnace B and several fuels ( b . The speed of the ignition front with combustion on a grate is of practical interest. namely 8. incineration is becoming more important for solid waste. 23] have been presented.9. The use of wood and biomass as fuels is expected to increase in the future. since this velocity often determines the heat output from the grate’s area. Furthermore. as used for generating small and medium-scale heat and power. Waste disposal by landfill is expensive and often impossible. In addition. In furnaces the fuel beds are much denser. Fuel a consisted of thin wood chips (mixture of spruce and pine) produced when notches of construction baulks are cut. The release of volatiles.

A pot furnace experiment for the study of the combustion rate of a wood chip layer (fuel d ). 2. An example of temperature measurements inside the fuel bed is presented in Fig. and 75. which greatly affects the distance traveled before ignition occurs. where no flame is initially available in the neighborhood. The flame speed inside the fuel bed was calculated by measuring the times needed for temperature waves of 673. SAASTAMOINEN ET AL. Locations of temperature measurements 1.1. together with radia- Fig. 27. 148.8. J. In addition. 2. The main mechanism causing the ignition front to propagate is assumed to be radiation. Measured mass of the layer and temperatures inside the fuel bed as function of time. (b) Propagation of an ignition plane in a fixed bed.216 J. 5–20 mm) were 10.4. 773. The box drawn with the dashed line is a control volume which corresponds to a fixed bed. 145. The flame speed was calculated as the average of these values. After some time the ignition wave stabilizes to a constant velocity and the system can be treated as a wave in an infinite medium.8. ( ␳ b calculated for dry fuel) had one of three values: 58. when the solid’s surface reaches a given ignition temperature.5% and the corresponding dry bed densities were 157. The moisture contents of the medium size particles b (sieved spruce wood chips. The mass flow rate of air was 0. 2. Fig. respectively. The largest particles c (30 mm) were a mixture of spruce and pine. The shape of the other fuels was typical of wood chips with one dimension smaller than the other two. two particle sizes of both birch ( d ) and alder ( e ) were studied. . 142. a flame starts to propagate downwards into the bed against the air flow. A simplified formula for estimating the ignition distance of the top surface of the fuel. including the effect of moisture. and 140 kg mϪ3. The shape of c particles was approximately cubic. 57. 18. This system under study is presented in Fig. Again. wood contains a significant amount of moisture. there is a flame present and piloted ignition is assumed to take place. and 50 mm from the bottom of the bed. and 873 K to travel the distance between two thermocouples.5. In addition. and 3 are 250. since air is in counterflow. Turbulent mixing and combustion near the flame front. 1. respectively. After ignition at the surface. respectively. almost dry (2. due to the burning of volatiles from the surface. and 33.1% wet basis) and the dry bed density was 230 kg mϪ3. has been presented [33].8% (wet basis). 35] than for spontaneous ignition [36]. moisture 12. 2. In this case there is an ignition source (a flame) in the moving front and the temperature for ignition is much lower (ϳ630 K) in piloted ignition [34. but conduction and gas mixing near the flame front are also important.7 kg mϪ3. T ig . (a) Analogy between combustion in a fixed bed and on a grate. MODELING The distance from the fuel’s inlet at which the surface of a moving bed is ignited depends on the radiation to the bed and on the density and specific heat of the fuel.15 kg mϪ2 sϪ1 with sieved (SCAN-CM 40:88 standard) birch chips 13 mm.

The heating of the solid fuel in the bed is described by Cٞ ѨTp ϭ hS ٞ͑ T g Ϫ T p͒ ϩ q ٞ Ѩt (2) ͩ ͪ (10) The boundary condition far from the ignition front is T p ( ϱ ) ϭ T g ( ϱ ) ϭ T ϱ . q Љ ϭ q Љ 0 exp( Ϫ ␣ x * ). From a practical point of view. The wooden chips used in commercial boilers are irregular with a size distribution and random orientation. where the effect of radiation is included in the dependence ␭ ϳ T 3 . ϪU ϭ␽Ϫ␪ϩ␾ d␰ d␰ By eliminating ␪. it is assumed that the local absorption of energy in the bed is directly proportional to the local heat flux. which gives the local heat generation q ٞ ϭ Ϫ dq Љ /dx ϭ ␣ q Љ 0 exp( Ϫ ␣ x * ). The heating of the oxidizing gas flowing against the moving ignition wave is described by the energy equation ˙Љ C g ѨTg ϭ hS ٞ͑ T g Ϫ T p͒ Ѩx (1) dT p ϭ hS ٞ͑ T g Ϫ T p͒ ϩ q ٞ dx * 217 ϪC ٞ w ig (4) Equations 3 and 4 can be presented in dimensionless form as d␪ d␽ ϭ ␽ Ϫ ␪ . Thin Dry Solids Including the Effect of Convection The particles are so thin that their temperature can be assumed uniform. In the models presented here. ␭. etc. provided they have no negative curvature [1]. T f depends on the stoichiometry. Then the problem is reduced to a simple heat conduction model [16]. when the effective radiation temperature exceeds the critical value Tf Ͼ ͩ ˙Љ C g͑ T ig Ϫ T ϱ͒ K ␴ ͑ 1 ϩ 1/ ⌳͒ ͪ 1/4 (11) In this treatment. An alternative approach is to lump the complicated processes in the fuel bed in an overall thermal heat conductivity. Equation 10 predicts that the flame will propagate. where K ( Ͻ 1) is the effective radiation coefficient.IGNITION IN FUEL BEDS tion. even for a bed consisting of regular particles. We apply a new moving coordinate x * ϭ x Ϫ w ig t and obtain ˙Љ C g dT g ϭ hS ٞ͑ T g Ϫ T p͒ dx * (3) The radiative flux depends on the effective 4 temperature of the flame zone T f . it has been assumed that the temperature of the gas far from the ignition . q Љ 0 ϭ K␴Tf . are difficult to model. this gives the solution for the velocity of the ignition front in dimensionless form as U ϭ ␾ 0⌳ Ϫ ⌳ / ͑ 1 ϩ ⌳͒ which in dimensional form is w ig ϭ ˙Љ 1 C qЉ 0 g Ϫ ˙Љ C ٞ T ig Ϫ T ϱ 1 ϩ ␣ C g/ ͑ hS ٞ͒ (9) where the term on the right-hand side describes the convective heat exchange between the gas and particles. U ͑ 1 ϩ 1/ ⌳ ϩ 1/ U ͒ ␾ 0͑ 1 ϩ ⌳͒ exp͑Ϫ␰ / ⌳͒ ] U ͑ 1 ϩ 1/ ⌳ ϩ 1/ U ͒ (8) At the ignition front ␰ ϭ 0. The relationship ␣ ϭ S ٞ /4 has been presented for randomly arranged fuel elements. The location of the moving ignition front is propagating with a steady velocity w ig . the calorific value of the fuel. Then the heat flux from the burning particles through the bed decays exponentially with distance. we obtain U␽Љ Ϫ ͑1 ϩ U͒␽Ј ϭ ␾ Equation 6 can be integrated to yield: (6) (5) ␽ Ј Ϫ ͑ 1 ϩ 1/ U ͒ ␽ ϭ ␽ Ј͑ 0 ͒ Ϫ ͑ 1 ϩ 1/ U ͒ ␽ ͑ 0 ͒ ϩ ␾ 0⌳͓ 1 Ϫexp͑Ϫ␰ / ⌳͔͒ / U (7) The temperatures of the gas and solid can be derived in the form: ␽ϭ ␪ϭ ␾ 0⌳ exp͑Ϫ␰ / ⌳͒ . the solids are at the ignition temperature T p (0) ϭ T ig and ␪ ϭ 1. simple models with suitable experimental parameters to account for randomness seem more useful than highly sophisticated models.

Equation 14 predicts that ignition will take place provided that T f Ͼ ˙Љ ˙ vap ( T ig Ϫ T vap )]/ {[ ␩ g C g ( T ig Ϫ T ϱ ) ϩ ␩ vap C Љ ␴ K } 1/4 . the velocity of the ignition front will accelerate. SAASTAMOINEN ET AL. 0 and t ϭ 0 corresponds to the instant when the surface reaches the ignition temperature. which can be seen by comparing Eqs. If the air is preheated. Then the particle’s temperature is nonuniform. however. q Љ ϭ q Љ 0 exp( ␣ w ig t ). The effect of drying can also be estimated [38]. Effect of Moisture Wooden fuels usually contain moisture. can be described by Fourier’s equation: 1 Ѩ T p Ѩ 2T p ϭ ap Ѩt Ѩr2 (16) ˙ Љ T g describes the heat flux and Q ٞ where q Љ Ϫ C denotes the energy per unit volume required to heat the solids and evaporate the water. Effect of Particle Size Here it is assumed that the wooden chips are plates (so ⌫ ϭ 0). R ͒ ϭ T ig. 15). Here the fuel is assumed to consist of plates of a constant thickness. ig Ϫ T vap )/( T ig Ϫ T vap ) account for the fact that the gas temperature is lower than the ignition temperature of the solids at the ignition plane. moist wooden chips as ˙ g͑ T ig Ϫ T ϱ͒ w ig ϭ ͓ q Љ 0 Ϫ ␩ gC Љ ˙Љ Ϫ ␩ vapC vap͑ T ig Ϫ T vap͔͒ / Q ٞ ig where Qٞ ig ϭ ͑ 1 Ϫ ⑀ ͒ ␳ p͑ c p͑ T ig Ϫ T ϱ͒ ϩ u ϱ͓ l vap ϩ c w͑ T vap Ϫ T ϱ͔͒͒ (15) The coefficients ␩ g ϭ ( T g . 12 gives the ignition velocity in a fuel bed of thin. . The heat is assumed to be absorbed symmetrically on the surfaces of the plate (or perfect insulation is assumed on the plane of symmetry). ig Ϫ T ϱ )/( T ig Ϫ T ϱ ) Ͻ 1 and ␩ vap ϭ ( T g . In the limiting case. In the case u ϱ ϭ 0. A simplified problem is studied in order to obtain an analytical solution able to illustrate the effect of different parameters. r ͒ ϭ T ϱ. ␩ g ϭ ⌳ /(1 ϩ ⌳ ). temperature at the time of ignition. It is possible to estimate this effect by using formulae for transient temperatures in packed beds [37]. when convection is neglected. ␭ p ͩ ͪ ѨTp Ѩr rϭR ͩ ͪ ѨTp Ѩr ϭ ͑␣q Љ 0/ S ٞ͒ exp ͑ ␣ w igt ͒ . which are ignited more easily. For fuels other than a (see “Experimental” section) the particles are not thin. so Q ٞ ϭ ͑ 1 Ϫ ⑀ ͒ ␳ p͓ c pT p ϩ u ͑ l vap ϩ c wT p͔͒ (13) Integration of Eq. In practice. we obtain w ig ϭ q Љ 0/ [ C ٞ ( T ig Ϫ T ϱ )]. ␩ vap can be related to ␩ g by ␩ vap ϭ [ T ϱ Ϫ T vap ϩ ␩ g ( T ig Ϫ T ϱ )]/( T ig Ϫ T vap ). Also the radiative flux to the solids is not symmetrical in practice. T p͑ 0. The propagation of the ignition front is described by the quasi-steady energy balance ˙ Љ T g͒ d͑ q Љ Ϫ C dQٞ ϭ w ig dx dx (12) J. Heat transfer to the particles. this leads to numerical solutions of three-dimensional heat conduction problems for a specific configuration of the solids in the bed. J. can be solved by applying the finite Fourier cosine transform [39] . . In this case the energy required to heat the particle to the ignition state is less than Q ig (see Eq. final. The analysis could be extended to cylinders or spheres. the interior of the particle may still be wet. the fuel has a size distribution and wooden chips may have thin sharp edges. such as particle size. and only the outer surface of the particle is at the ignition (14) with the initial. 17. and boundary conditions T p͑Ϫϱ . 14 and 10. where the time t is defined in the range t ϭ Ϫϱ .218 plane is the same as that of the solid. ϭ0 rϭR (17) Equation 16 with the boundary conditions of Eq. if heat recovery by the gas flow is assumed negligible (⌳ 3 0).

The drying of the surface layer of the solid can be described by the simplified equation qЉ ϭ ␭p d␦ ͑ T s Ϫ T vap͒ ϭ ␳ pu ϱl vap ␦ dt (24) when the temperature distribution in the dry surface layer (thickness ␦) is approximated as linear. t ϭ 0 and ␪ ϭ 1. In this case w ig is directly proportional to R . The functions F 1 ( ⌽ ). [ ␭ p ␳ p c p ( T ig Ϫ T ϱ ) 2 ]. Thick Moist Solids As a limiting case. w ig Ϸ ␣ q Љ 0 / 2 2 [ ␳ p c p ␭ p S ٞ ( T ig Ϫ T ϱ ) ].. R . F2(⌽). and we obtain the relationship ͱU* tanh͑ ͱU*͒ ϭ ⌽ (23) between the dimensionless ignition velocity U * ϭ R 2 ␣ w / a p and the dimensionless heat flux ⌽ ϭ ␣q Љ 0 R /[ S ٞ ␭ p ( T ig Ϫ T ϱ )]. The first term. The effect of this size can be seen more clearly by using the dimensionless velocity U * / ⌽ ϭ F 2 ( ⌽ ). which does not depend on R . 16 becomes ˜ p ␣q Љ 1 dT 0 ϭ cos͑ n ␲ ͒ exp͑ ␣ w igt ͒ a p dt Sٞ␭ ˜p Ϫ ͑ n ␲ / R ͒ 2T (19) The solution for the transformed transient temperature of the particles becomes The temperature distribution is found by applying the inverse transform [39] f͑r͒ ϭ 2 1 ˜ f͑n ϭ 0͒ ϩ R R nϭ1 ͸ ˜f͑n͒ cos͑n␲r/R͒ ϱ (21) giving ␪ϭ a p␣ q Љ 1 0 ϩ2 RS ٞ ␭ p͑ T ig Ϫ T ϱ͒ ␣ w ig ͩ nϭ1 ͸ ϱ ͑Ϫ1 ͒ n cos͑ n ␲ r / R ͒ exp͑ ␣ w igt ͒ ␣ w ig ϩ a p͑ n ␲ / R ͒ 2 exp͑ ␣ w igt ͒ ͪ ϭ RS ٞ ␭ p͑ T ig Ϫ T ϱ͒ ͱ␣ w ig/ a p sinh͑ R ͱ␣ w ig/ a p͒ ␣q Љ 0 cosh͑ r ͱ␣ w ig/ a p͒ (22) At ignition. This is the same result as given by Eq. Let us finally consider the case of parallel plates.IGNITION IN FUEL BEDS 219 ˜ϭ ␪ ͑Ϫ1 ͒ na p␣ q Љ 0/ ͓ S ٞ ␭ ͑ T ig Ϫ T ϱ͔͒ exp͑ ␣ w igt ͒ ␣ w ig ϩ a p͑ n ␲ / R ͒ 2 (20) ˜ f͑n͒ ϭ ͵ R f ͑ r ͒ cos͑ n ␲ r / R ͒ dr (18) 0 Using the transform ˜ f ( n ) Eq. is the radiation ab- . and F3(⌽). Then the solids can be approximated as semi-infinite. since S ٞ ϭ (1 ϩ ⌫ )(1 Ϫ ⑀ )/ R . U * 3 ⌽ . Then we obtain the relation wig ϭ q 0 ЉF2(⌽)/[(1 ϩ ⌫)Cٞ(Tig Ϫ Tϱ)]. if ␣ ϰ S ٞ ϰ 1/ R . very large particles are considered.e. U * 3 ⌽ 2 . if the porosity is constant. ␣ and U * / ⌽ 2 ϭ F 3 ( ⌽ ) do not depend on R and we obtain w ig ϭ ␣ ( q Љ 0/ 2 S ٞ ) 2 F 3 ( ⌽ )/[ ␭ p ␳ p c p ( T ig Ϫ T ϱ ) 2 ] ϭ ␣ q Ј g F 3 ( ⌽ )/ Fig. For large values of ⌽ 2 (3ϱ). i. The functions F1(⌽). Equation 23 gives the dependence U * ϭ F 1 ( ⌽ ) or w ig ϭ a p F 1 ( ⌽ )/( R 2 ␣ ).. 3. of the plates changes and radiation through the gas q Ј g ϭ qЉ 0 ( R ϩ R g ) is assumed to be independent of R . For small values of ⌽ ( 3 0). 3. kept constant. but the thickness.e. Both dimensionless quantities depend on the particle’s size. with the distance between the plates. q Љ . which does not depend on R if the porosity of the bed remains constant. In this case. F 2 ( ⌽ ) and F 3 ( ⌽ ) are presented in Fig. 10 for the case ⌳ ϭ 0. i. w ig Ϸ q Љ 0 /[ ␳ p c p S ٞ R ( T ig Ϫ T ϱ )]. R g . since only a thin layer has been dried at the time of ignition.

heating of combustion air to the reaction temperature. with combustion of the volatiles. and q 9 Љϭcc␬ (1 Ϫ v ) m ˙Љ p ( T vol Ϫ T a ) are the energies associated. The thickness of the zone. The heat balance is (25) Energy is consumed also in heating the combustion air.26O0. The values of B are readily calculated from the balances for C. The adiabatic temperature of the flame from burning the volatiles can be estimated from a simple heat balance on the moving flame zone. Equation 26 gives volv ϩ c c␬ ͑ 1 Ϫ v ͔͒ T vol ϩ c vapu ϱT vap͖ (27) Ta ϭ iϭ1 ˙Љ C ˙Љ gϩ m p͓͑ c volv ϩ c c␬ ͑ 1 Ϫ v ͒ ϩ c vapu ϱ͔ At low flow rates of air. J. The coefficient A ϭ ( Y O 2 m ˙Љ g 0 / M O 2 )/ (m ˙Љ vol / M vol ). The mass flux of dry fuel to the moving flame zone is related to the velocity of ˙ ЉT ϩ m ˙ Љ ͕͓ c ͸ qЉϩ C 5 i g 0 p iϭ1 ͸ qЉ ϭ 0 g (26) where the terms q 1 Љϭvm ˙p Љ⌬Hvol. A depends on the primary air mass flux m ˙Љ g and the mass flux of the volatiles generated. Eq. The flux of water vapor generated due to moisture is m ˙Љ vap ϭ (1 Ϫ ⑀ ) u ␳ p w ig . which is accounted for by assuming a value of 0. when the particles are small. SAASTAMOINEN ET AL. volatiles are released in a thicker zone and the difference between the effective radiating flame temperature T f and the adiabatic flame temperature T a will be greater. the pyrolysis reactions.99 and also to react by CH2. where t d is the drying and pyrolysis time of the particles as estimated by a model presented earlier [40]. as accounted for by the second term in Eq. and orientation of a particle affect the local turbulence. q 5 Љϭ ˙g cpm ˙p Љ (TϱϪTvol). q 6 Љ ϭC Љ(TϱϪTa). where simultaneous drying and pyrolysis take place. heating of particles to the devolatilization temperature. q 7 Љ ϭcvolvm ˙p Љ (TvolϪTa). heating of volatiles to the reaction temperature.99 ϩ A O2 3 B 1 CO ϩ B 2 CO2 ϩ B 3 H2O ϩ B 4 CH4. the mixing of fuelvolatile with oxygen. and the heating of char close to reaction temperature. respectively. Then the effects of air flow rate (reaction stoichiometry) and moisture on the temperature are adequately described by Eq. When the particles are large. heating of water vapor to the reaction temperature. here it can be approximately taken into account by using an effective heat flux q Љ 0 Ϫ ˙Љ ˙ ␩ gC ( T Ϫ T ) Ϫ ␩ C Љ ( T Ϫ T g ig ϱ vap vap ig vap ) instead of q Љ . as accounted for by the last term. q 4 Љ ϭcwuϱm ˙p Љ (TϱϪTvap). the ignition front m ˙Љ p ϭ (1 Ϫ ⑀ ) ␳ p w ig and the mass flux of the volatiles is m ˙Љ vol ϭ v (1 Ϫ ⑀ ) ␳ p w ig .220 sorbed at the surface. It is also that conducted through the dry surface layer. The energy of vaporization is assumed to be much greater than that for heating the material and the liquid water. 27. The volatiles are assumed to be CH2. when A is known. This is for a typical elemental composition of wood. the char does not reach the gas temperature. 24. can be calculated from the relationship L d ϭ t d w ig . and also combustion in the . shape.26O0. Due to a time lag. 0 Adiabatic Temperature of the Flame from the Volatiles Steam from drying and volatiles are released to the gas flow. q 2 ЉϭϪvm ˙p Љlvol. DISCUSSION The size. if the char is assumed to consist of carbon and its yield is 18 wt %.9 for the coefficient ␬. q3 Љ ϭϪuϱm ˙p Љ lvap. 24 can be solved to give 2 w ig ϭ ␣ ͑ q Љ 0/ S ٞ͒ / ͓ ␭ p␳ pu ϱl vap͑ T ig Ϫ T vap͔͒ J. but the accuracy can be improved if this sensible heat is accounted for by using an effective heat of vaporization. The coefficient A and the heat of decomposition ⌬ H vol are defined by the stoichiometry. H. the evaporation of water. heating of water to the temperature for evaporation. the temperature of the reaction zone can be assumed to be close to the adiabatic flame temperature (so T f Ϸ T a ). if it is assumed that the gases leave the zone at that temperature. q 8 Љϭcvapuϱm ˙p Љ(TvapϪTa). Finally it is the energy consumed in vaporization. and O.

T ig ϭ 650 K.5 const . At m g ϭ 0. R . moisture.714–2. fuel c dotted trendline. and orientation of the particles in the bed.008 m ˙Љ ˙Љ g .30 were obtained for small and large size fractions. all of which are important factors affecting ignition. and specific heat. one experiment with wooden pellets gave K ϭ 0. by ␦ ϳ m ˙Љ g R . For fuel e the values K ϭ 0. Calculated effective radiation coefficient K based on the measured ignition velocity and maximum temperature in the bed. The available data are too scarce to obtain a single expression for K accounting for the effects of a particle’s size and moisture content. since S ٞ ϳ 1/ R . If we roughly assume that the stoichiometric condition corresponds to the maximum flame temperature T f and that radiation also decays exponentially between the locations of maximum temperature and ignition. The best lines for the radiation coefficients are for fuel a : K ϭ 0.79. 14. which are included in Eq.25 were obtained for particle sizes 7–13. times for ignition and ignition temperatures in a bed of fuel vary from case to case due to the random size. More difficult is the determination of the parameters related to radiation and convection and also modeling of the particle size effects. 0. so that m ˙Љ vol ϳ R 1 Ϫ n . These effects are lumped into an experimental coefficient K and the maximum temperature T max .8 E. The following correlations for T f and K can be used in Eq. solid trendline. when comparing fuels a .1 ϫ. Thus. 27. 4.24. 13. Values of c p ϭ 2000 J kgϪ1. making ␦ smaller. 1). which is seen in Fig. 2 For fuel a T f ϭ 997. we obtain a simple relationship K ϭ K f exp( Ϫ ␣␦ ) Ϸ K f Ϫ 0. and 0. 4. and 30 mm. 14. since they are fuels used in practice. The time needed to heat a particle to the ignition temperature is directly proportional to the particle’s density and specific heat by theory. The particles used in the experiments do not have a specific shape and size. ϫ m ˙Љ . such as the kinetics of pyrolysis. 43]. This approximately linear relationship with respect to m ˙Љ g is seen in Fig. shape. The flux of volatiles at the surface of a particle depends on its size.626 m ˙Љ g .819–2. where the gas mixture reaches the stoichiometric fuel/oxygen ratio.IGNITION IN FUEL BEDS bed.4 Œ.28. for which a constant value of 2000 J kgϪ1 was assumed. 0. Fuel b (moistures are 10.5 gR [42. 14 by using measured values of w ig and T max .2 ϩ 5547 m ˙Љ g Ϫ 2 11. for fuel b : K ϭ 0.15 kg mϪ2 sϪ1 the values of K obtained from the trendline were 0.50 for fuels a . K f is a constant.100 m ˙Љ ˙Љ g . and c . In addition. It also predicts that K is higher for small particles. b . the products of pyrolysis (including ignition temperature). which could be due to differences in a fuel’s properties other than bed density or moisture. Thus the velocity of the ignition front is inversely proportional to the bulk density of the dry mixture ␳ b ϭ (1 Ϫ ⑀ ) ␳ p and specific heat of wood.002 m g . since ␣ ϳ 1/ R and n ϭ 1. The fuel’s properties (bed density. if the devolatilization time is ϳ R n . A simple analysis shows that the distance ␦ from the ignition front (see Fig. For fuel d the values K ϭ 0.8 ϩ 1415 m ˙Љ ˙Љ g Ϫ 1646 m g and for other fuels T f ϭ 903. Experiments with particles of different sizes but equal shape are needed to develop this. and specific heat) can be measured.41. there are rather large differences between the fuels. 14 in units of K and kg mϪ2 sϪ1. The coefficient K was evaluated from Eq. and ␩ g ϭ ␩ vol ϭ 0 were used in these calculations. and n the size of particles. and 33.4 F % in wet basis). when fuel moisture (wet basis) is below 30%. Here we suggest the use of Eq. 0. T ϱ ϭ 293 K. differences in shape and size distributions and orientation of the particles affect K . However. b . Consequently.30.53 and 0. m ˙Љ g .8 ‚. it is important to get some quantitative approximation for the velocity of the ignition front. which is assumed to correspond to the effective radiation temperature T f . respectively. and for fuel c : K ϭ 0. 11. .943–1. The largest c particles are practically dry and the release of volatiles be- 221 Fig. this is probably the reason for the relatively high values of K for this fuel. fuel a dashed trendline. 18. 4. and c . depends on the flow rate of air. comes faster than with the other fuels. In addition. respectively.

8 wt % (wet basis). but not the limits of sustaining combustion. The calculated adiabatic temperature also predicts this.4%). otherwise combustion is extinguished. c .8% ϫ. the calculated adiabatic temperature becomes higher than the measured maximum temperature. The model predicts the effects of both the flow rate of air and moisture.7%. e (‚ small. At higher flow rates of air.4% ϩ. Similar trends are seen in Fig. 8. the maximum temperature in the bed was higher than for smaller ones. Fig. since the rate of devolatilization decreases.9% ***). 32.1%. It is also seen that the ignition front moves only for a specific range of flow rates of the primary air. d (ᮀ 7–13 mm. and wood pellets (I). 18. which becomes important at high temperatures. SAASTAMOINEN ET AL. Increase in the moisture content decreases the velocity of the reaction front. Fig.8% ‚. c (E 2. Air rate is 0. Effect of flow rate of air on the maximum temperature measured by the middle thermocouple. here the flame speed depends on the flow rate of air to some extent at moisture contents of 10. The measured velocities of the ignition front presented in the literature for counterflow systems are summarized in Fig.3% F. in spite of the increasing amount of energy consumed in vaporization.7% Œ). but was not included in the calculation of T a . 4 and 5 were calculated iteratively from Eqs. The range of flow rates for a moving front becomes narrower with increasing moisture content. when the air rate is small. Another reason might be a lower value of ␬ for larger nonisothermal particles. b (ϫ). Figure 6 shows that moisture seems to have no influence on the maximum temperature. The lower maximum temperatures of fuel consisting of smaller particles can partly be explained by the higher value of K (greater losses from the flame). when the moisture content of the fuel is below ϳ30% (wet basis). Ϫ27. The velocity of the ignition front was estimated by Eq. Fuels (and moisture contents in wet basis) are a (8. 11.9%. I. 14 and 27. 6. E 13 mm. b (10. changing the conditions more rich with oxygen. The upper solid line is the best-fit second-order polynomial for points for fuels b .3– 33. The adiabatic temperatures for fuel b with moisture 10. The temperatures in a bed of the thin particles a are clearly lower than those of other fuels with larger particles. The line is the calculated adiabatic temperature. 17. The velocity reaches a maximum at a specific flow rate of air for fuel a (thin wood chips) as shown in Fig. if average values of the points for each size are used.222 J.8 and 18. c (ϩ). ᮀ 33. The dashed line is the calculated adiabatic temperature.15 kg mϪ2 sϪ1. but is rather constant at a moisture . Fuels are a (Ϫ). 29. For larger alder particles. Figure 5 shows the effect of the flow rate of air on the measured maximum temperature inside the bed. Figure 4 shows that. which is believed to be due to the endothermic gasification reaction (C™CO2 and C™H2O). and d with moisture Ͻ30%. 7. 9 for fuel b with medium size particles. decreasing the temperature. trend line. Effect of fuel moisture on the maximum temperature measured by the middle thermocouple. 5. A similar trend was obtained for birch particles. The net result is that the temperature remains quite constant.1%) and d ({ 12.8% shown in Figs. Œ large). so the ignition front no longer propagates above a specific moisture level. 14 using the correlation presented above for K and assuming T f ϭ T a . J. the adiabatic temperature is close to the best-fit second-order polynomial. F 30 mm).

Summary of literature values of the velocity of the front in beds of wooden particles: Birch 11% [30] (F). since in the limiting situation of zero flow rate. solid trendlines) and filled ones for fuel c (mixture of spruce and pine. where n ϭ 1. 9) with moisture contents less than ϳ20%. the speed must be zero.3 (F).5 . softwood 13% [28] (}). 5–20 mm. content of 34. 31 mm. the ratio of fuel mass to oxidizer at the start of the flame zone is proportional to 1/ d 1. dashed trendline. almost dry. 57 (17. Then. 27.8 E. increasing the flow rate will lead to a .1% F on a wet basis. and 33. the ignition speed slightly increases with decreasing flow rates of air. 10% [25] (I. Open symbols are for fuel b (spruce.9). For the largest. It is not possible to increase significantly the combustion intensity from the devolatilization zone of the grate beyond the air rate giving the maximum ignition velocity. willow 40% [28] (Œ). Fig. 43]. The rate of pyrolysis depends on the particle size and the pyrolysis time is usually represented by a power law t ϳ d n . willow 28% [28] (‚). thin solid trendline) on a wet basis. 14. ␳ b ϭ 58 (8. 8) with a higher flow rate of air than for larger ones (Fig. solid trendline). Measured velocity of the ignition front as a function of the flux of primary air in a bed of wooden particles.IGNITION IN FUEL BEDS 223 Fig. ϫ based on visual observations). However. Also fewer particles are entrained from the grate. Then the maximum velocity of the ignition front is reached for small particles (Fig. when considering NO formation [41]. so the temperature falls.7%). since fuel-rich conditions are usually more favorable. and 29.9 (e).4% ‚ on a wet basis. birch 38% [30] (E). However. wood 30 mm. Solid lines are calculated by Eq. dashed trendline).3). In fact.7% (‚) on a wet basis.1 {. increasing the flow rate of air dilutes the gas in the flame. When the flow rate of the air and the temperature are constant. 2. the rates of char combustion and gasification above the flame zone increase with the flow rate of air. From a pollution point of view it is more favorable to choose a flow rate for air below that giving the maximum ignition speed. wood 10 mm. The ignition velocities for the results of Gort [25] have been calculated by assuming a dry bed density of 200 kg mϪ3. Fig. wood 10 mm. 9 shows no such maximum. The flame temperature depends on the local ratio of [fuel]/[O2] and reaches a maximum at a specific flow rate of air. the rate of generation of volatiles in the flame zone is not so intense. the rate of devolatilization is faster and more volatiles are available in the flame zone.4 wt %. particles c . the ignition speed must decrease. softwood 40% [28] (ϫ). it is clear that if the flow rate is further decreased. since the mass flux of volatiles released from the bed is directly proportional to the velocity of the ignition front.5 [42. If the particles are large. and 76 kg mϪ3 (29. ␩ ϭ 0. For small particles. 7. 17. 30% [25] (ϩ. 10% [25] (ᮀ. 8. moistures are 10. Fig. Measured and calculated velocity of the ignition front as a function of the flux of primary air in a bed of wood particles (fuel a ) with different moistures 8. 18.8 ᮀ. 9. After that.

SAASTAMOINEN ET AL.3. ␳ p ϭ 500 kg mϪ3. this decreases the flame temperature and the propagation rate. measurements and the simple analysis indicate K decreasing with increasing size. ref ϭ 150 kg mϪ3 by multiplying the measured ignition velocity with ␳ b / ␳ b . Measured and calculated effect of fuel moisture content on the normalized ignition velocity. 140 kg mϪ3. However. m ˙Љ ig ϭ ␳ b w ig . Equation 23 (and also Eq. 8 mm. dashed line is Eq. In practical grates. 10. depending on the moisture content of the fuel. 30 mm. e (‚. are of equal magnitude for a wide range of dry bed densities (56 to 612 kg mϪ3) as seen in Fig. resulting in a decrease in the velocity of the ignition front after the optimal stoichiometric condition. J. 25) gives the same trend. Equation 23 (and also Eq. but at a higher level.7 mm. The ignition mass fluxes. 13 mm. decrease in the flame temperature. defined as . the flame temperature remains too low. 25 with effective heat of vaporization [ ϭ l vap ϩ c p ( T ig Ϫ T ϱ )/ u ].5 mϪ1. especially when the size of the particles is large. 54 kg mϪ3). 25) predicts that the velocity of the ignition front increases with a particle’s size provided that K is constant. the conditions close to the ignition front change from fuel-rich to Fig. thereby increasing the speed of the front. 25) predicts that the ignition velocity is lower for materials with a higher thermal conductivity. but this is mainly due to the difference in bulk density of the bed. c p ϭ 2000 J kg K. so extinction takes place. 10. 210 kg mϪ3. Modeling the effects of the flow rate of air. 9 by comparing fuels b (medium size particles) and c (large particles). and properties of wood (density. An increasing moisture content of the wood particles decreases the flame speed in two ways: the temperature of the ignited region decreases due to dilution. Figure 10 shows no clear experimental effect of the size of a particle on ignition velocity. 141 kg mϪ3) and wood pellets (Ϫ. 23 with R ϭ 0. Mass flux of air is 0. Œ. At very low flow rates of air. S ٞ ϭ 150 mϪ1. 230 kg mϪ3).150 kg mϪ2 sϪ1. 7–13 mm. the moisture content of the fuel. 14 with ␩ ϭ 0. This takes place at a lower flow rate of air than for smaller particles. As the flow rate of air is increased. A sustained ignition wave is obtained over a specific range of flow rates of air. The direction of the combustion air can also be such that it entrains hot combusting gases to heat the nonignited fuel convectively. ref . d (ᮀ. thermal conductivity) is discussed and used for interpreting the experimental measurements. Measurements: fuels and average dry bed densities ( ␳ b ) are a ({. where the ignition velocities have been normalized to the reference dry density ␳ b . The maximum ignition speed for the thin particles (Fig. ␣ ϭ 37.01 m. ig ϭ S ٞ ␭ p ( T ig Ϫ T vap )/( ␣ q Љ 0 ) ϭ 1. particle size. 612 kg mϪ3). due to dilution with excess air. 9). the thickness of the dried layer at the moment of ignition is ␦ p . small. which was small for thin particles. T f ϭ 1473 K. In the calculations K ϭ 0. specific heat. 196 kg mϪ3. 8) is greater than that for the larger ones (Fig. the movement of parts of a grate relative to the others or vibrations can increase the velocity of propagation of the ignition front. and ␭ p ϭ 0. E. large. F. this has the opposite effect on the velocity. 14 predict the effect of moisture content quite well. due to a lack of oxygen. the oxygen content is lower and more oxidizing gas is required for the same stoichiometric conditions.14 WmϪ1 KϪ1. c (ϩ. b (ϫ. 216 kg mϪ3). point I is Eq. 149 kgϪ3). If flue gas recirculation is applied. Calculated results: solid line is Eq. The normalized velocity is clearly lower for the dense wooden pellets with a high thermal conductivity (caused by the greater density). The treatment of the solids as thermally thick (Eq. as seen from Fig. Figure 10 shows that the calculations using Eq. CONCLUSIONS Ignition in beds of wood particles has been studied for the incoming air and propagating ignition wave in counterflow. and the energy needed to heat (and dry) the particles to the ignition temperature increases.224 J.

1486. J... p. M. M. and Fenton.. Aho. Pittsburgh.. R. Heat Mass Transf. R.. Combust. T. Ruimin. and Reid. J. P. Nicolella. p. and Peterson. D. 29. 75:31 (1991). J. 1965. Massachusetts Institute of Technology. 1975..IGNITION IN FUEL BEDS oxygen-rich. M. A. 11:149 (1987). Incin.. 2... 312. Cooke. and Sage. Bureau of Mines. M. 20. Proc.. J. and Berge. 288.. 19. M. Drew. 1972. The effect of the principal factors.. W. Carrier.. (1972). However.. M. J.. Philadelphia. Rothermel. E. 110 –111:123 (1995). P. 1971.. Stumbar.... and Brem.. 18. and Steward. (1958). Combust. 1970. and Grumell. Sci. the moisture content of the fuel. H. H.. 927. J. M. Pittsburgh. 22. Harold.. 129. 1137. Berland. G. J. Cozzani.. Technol. Vol. Int. 30. p. Technol. A. Flame 14:123 (1970). Fuel.. Hesheng. and Steward. 14. M. Inst. C. Hong Kong.. p. p. decreasing the radiation from the hot combusting zone. C. M. and Shultz. Weintraub. 35:90 (1996). Carman. 225 Pagni. p. Orning. 7. S. R. Trans. 21. J.. M. 15. P. but increasing the particle size increased the maximum temperature in the bed to some extent. B. No effect of particle size on the ignition front’s speed was found. Sci. Sci. F. C. F. The velocity of the front was found to be inversely proportional to the density of the fuel bed and the specific heat of wood. and Frandsen. 33. Combust. Tenth Symposium (International) on Combustion. C. M... Proc. A. H. W. Research Paper INT-115. Gort.. H. 10th European Conference and Technology Exhibition. R. Kuo. 1099. Carman.. C. Williams. Inst. F. Conf. F. E. E.. Torero. L.. A. 11:129 (1937). Several experiments studied the effects of the flow rate of air. (1938). M.. A. Energy 65:137 (1992). 1995. G. G. E. (1987). L. Then the gases are diluted with excess air. Rogers. 11. C. X. Espoo.. contract 40760 (MH) is acknowledged. Graf. L. J. Rovatti. C. Conf. (1995). A.. 16. Correlations for the maximum bed temperature and the effective radiation coefficient in the bed are presented... the particle size. Sarofim. R. and Landau. H. Hong Kong 1995. The Combustion Institute.. 32.D... University of Twente. Cooke. The Combustion Institute. Gort.. Fourteenth Symposium (International) on Combustion. and Kja ¨ldman.. J. 9. Anal. 40:2607 (1997). Technol. p. Fuel Proc. C. Rothermel.D. Pfeiffer. New York. paper 58-12. V. L. 6. Support from the Commission of the European Community through contract JOR3-CT96-0059 (JS & RT) and from the Technology Developing Centre of Finland (TEKES). G. Pyrol.. 12:87. 3rd AsianPacific International Symposium on Combustion and Energy Utilization. 2 (K.. Zhou. 12. 26. The Combustion Institute.. Fuel. Goudeau. 47:139 (1986). and Schwartz. R. C. D. Ind. Department of the Interior. D. R. Ford. and the type of wood. H. N. J. J.. W. Dunningham. Technol. T. p.. and Essenhigh.. Inst.. Res. J. Ph. J. (1967). J.. Mayers.. E. Ph.. 24. and the density of fuel on the ignition velocity can all be explained by modeling. 10.. 1972. p. in Computational Fluid Dynamics ’98. G.. D. Sarofim. Williams. P. J. Saastamoinen. p. and Bregeon.. and Corey. A. Rogers. (1973).. J. J. so temperatures decrease. L. 17. and Howard. when the moisture content (wet basis) was less than 30 wt %. 1. 3. 51st Annual Meeting of APCA. A. moisture content. Fifteenth Symposium (International) on Combustion. thesis. A. F. 31. J. 13. J. Larfeldt. and Kaviany. Combust.. Technol.. ASME New York. M. B.. A. 23. Howard. Fang. E. Sci. W. Moisture was found to significantly lower the speed of the ignition front. Valk. Combust. Orning.. thesis. U. 27. J. Aho. Combust. J. S. Also the velocity of the ignition wave and the maximum temperature in the fuel bed were found to reach maximum values at specific flow rates of air. 25.. Eng. Technol. R. T. Fatehi. REFERENCES Hottel.. Crumell. F. ASME. 37:285 (1984). 32:563 (1940). 4th Nat. 4. N.. G. 135. M. B. Steward. 4:177 (1971). C. 3rd Asian-Pacific Symposium on Combustion and Energy Utilization. Increasing this flow rate also moves the region of the maximum temperature farther away from the ignition front. The Combustion Institute.. Fendell. L. Chem. Appl. A. Chem... such as the flow rate of air. Sci. 5th Nat. 997. resulting in extinction. the density of the bed. F. ASME 67:425 (1945). Biomass for Energy and Industry. Berland. Rogers. J. Combustion 67:59 (1954). Flame 13:392 (1969). J. Kuwata.. Eng. Combust. Stubington.. A. and Dunningham.. E. .. Z. and Dezhen. C.S.. moisture did not have any noticeable effect on the maximum temperature in the bed. F. 8. Pittsburgh. and Pettit. Research Reports 465. Ind. 1998. Pittsburgh. 28. 84. Ford. Technical Research Centre of Finland. RI 6908. when a specific flow rate of air is reached.. W. A Mathematical Model for Predicting Fire Spread in Wildland Fuels. 36:55 (1993).. F. E. and Wolff. N. Incin. and Tognotti. C. 13th Symposium (International) on Combustion.. USDA Forest Service. and Fine.. Huttunen. 5.

R. S. and Scaroni.. Energy Resources Technol. 39.. Kno ¨tzer. 38. Huang.. and Roberts. Tsahalis. Riaz.. 1972. Simms. McGraw-Hill Kogakusha. Flame 80:94 (1990). J. D. M. S. Received 21 June 1999. J. G. and Impola. R.. J.. and Law. Combust. L. Saastamoinen. ASME J. H.. C. 36. ASME J..). S. Energy 4:341 (1979). 42. 35. 40. 15:1919 (1997).. John Wiley. J. A. Drying Technol. Heat Transfer 99:489 (1977). Tokyo. Eds. Fire Mater.-R. F. J. 15:151 (1991). and D. I. p. J. W. J. Janssens. 37.... and Richard. Russel... Stubinton. Operational Mathematics. J. V. Combust. A. Churchill.. M. accepted 16 March 2000 . and Wichman. Pe ´riaux. D... 3rd ed. Combust. Flame 106:288 (1996). 34.. Chichester. L. A. 107:285 (1985). 1998.226 Papailiou. Fuel 70:1105 (1991).. Saastamoinen. M.. and Weiss. 814.. E. Ltd. 41. Tzeng. K. 43. Ragland. Flame 11:377 (1967). revised 6 March 2000. W. Atreya. SAASTAMOINEN ET AL.