Comprehensive study of ignition and combustion of single wooden particles

Comprehensive Study of Ignition and Combustion of Single WoodenParticlesMaryam Momeni,*,† Chungen Yin,*,† Søren Knudsen Kær,† and Søren Lovmand Hvid‡

†Department of Energy Technology, Aalborg University, DK-9220 Aalborg, Denmark‡DONG Energy, Kraftværksvej 53, 7000 Fredericia, Denmark

ABSTRACT: How quickly large biomass particles can ignite and burn out when transported into a pulverized-fuel (pf) furnaceand suddenly exposed to a hot gas flow containing oxygen is very important in biomass cofiring design and optimization. In thispaper, the ignition and burnout of the largest possible biomass (pine wood) particles in a pf furnace (a few millimeters indiameter) are studied experimentally in a single particle combustion reactor rig, in which the ambient gas temperature andoxygen concentration can vary in the ranges 1473−1873 K and 5−20%, respectively. A one dimensional (1D) transient model isalso developed to predict their conversion, in which the key processes inside the particle and in the boundary layer outside theparticle are properly considered. For the pine wood particles in which large temperature gradients exist, the primaryheterogeneous ignition is always detected for all the test conditions. As the particle is further heated and the volume-weightedaverage temperature reaches the onset of rapid decomposition of hemicellulose and cellulose, a secondary homogeneous ignitionoccurs. The model-predicted ignition delays and burnout times show a good agreement with the experimental results.Homogeneous ignition delays are found to scale with specific surface areas while heterogeneous ignition delays show lessdependency on the areas. The ignition and burnout are also affected by the process conditions, in which the oxygenconcentration is found to have a more pronounced impact on the ignition delays and burnout times at lower oxidizertemperatures.

1. INTRODUCTION

Ignition of pulverized solid fuel particles is a complex process inthe early stage of combustion, involving heat and mass transfer,fluid mechanics and chemical kinetics. It plays a vital role inpulverized fuel (pf) flames and has been studied since the1880s, as reviewed by Essenhigh et al. (1989).1 In the primaryignition, the particles can ignite homogeneously or heteroge-neously. For “large” coal particles (typically over 100 μm indiameter) under “slow” heating condition (<100 K/s), theprimary ignition mode may be homogeneous ignition,characterized by prior pyrolysis and subsequent ignition ofthe volatiles in a circumambient flame. For small particlesheated quickly, heterogeneous ignition is more likely theprimary ignition mode, by direct oxygen attack on the wholecoal particle. The majority of ignition experiments in literaturefocus on determination of the minimum gas temperature atwhich ignition occurs, irrespective of residence time.2 Suchstudies might be of less relevance with fuel particle ignition andcombustion in practical pf furnaces, in which relatively cold fuelparticles are suddenly exposed to a hot ambient gas and thenundergo a rapid heating and a sequence of conversionprocesses. Therefore, in practical pf furnaces, the more relevantquestion is how quickly the cold fuel particles can ignite andburn out after they are introduced into hot furnaces and howvarious factors affect the conversion times.Only a few studies on coal particle ignition in hot ambient air

can be found in the literature. For instance, Du and Annamalai(1994)3 presents a one dimensional (1D) model for ignition ofbituminous coal particles in a hot gas. The coal particles areassumed under isothermal condition and the domain outsidethe particle is divided into 10 spherical shells. For the coals

(particle sizes dp = 50−400 μm) under the ambient conditions(gas temperature T∞ = 1500 K, oxygen mass fraction YO2

=0.23), heterogeneous ignition always occurs first, followed bythe secondary homogeneous ignition. The primary ignitionmode shifts to homogeneous for larger particles (>400 μm).When lowering T∞ from 1500 to 1000 K, the primary ignitionis still heterogeneous for small particles and homogeneous forlarge particles, but the transition particle size at which theprimary ignition mode shifts to homogeneous increases from400 to 600 μm. The primary ignition also varies with O2concentration: heterogeneous under high O2 concentrationwhile homogeneous under low O2 concentration. Wendt et al.(2002)4 extend the work of Du and Annamalai (1994)3 bydeveloping a 2D model and accounting for all the intraparticleprocesses. Simulations are done for spherical bituminous coalparticles (dp = 50−1000 μm) suddenly inserted into hot air at1500 K. Good agreement with the model of Du and Annamalai(1994)3 is obtained only for small particles (dp ≤ 200 μm). Notransition in the primary ignition mode is found. The 2D modelis also used to study the impacts of particle shapes on ignitionof differently shaped brown coal particles of a few millimeters insize. Primary heterogeneous ignition is detected by the modelfor all the shapes and heterogeneous ignition delays are almostthe same for the slabs and cylinders. The secondaryhomogeneous ignition delays predicted by the model agreewell with their experiment, both showing the homogeneousignition delays scale with the particle specific surface areas.

Received: May 25, 2012Revised: January 30, 2013Published: January 30, 2013

Article

pubs.acs.org/EF

© 2013 American Chemical Society 1061 dx.doi.org/10.1021/ef302153f | Energy Fuels 2013, 27, 1061−1072

pubs.acs.org/EF

Zhang and Wall (1994)5 perform experiments and modelingfor ignition of some fuel samples. Ignition temperaturedecreases with an increase in the sample mass. As the heatingrate is increased from 4.2 × 103 to 5.5 × 103 K/s, a shift in theprimary ignition mechanism from homogeneous to heteroge-neous is predicted. Faundez et al. (2005)6 conduct a coalignition experimental study in an entrained flow reactor at highcoal feed rates (0.5 g/min). Under their conditions, the fourhigh-volatile bituminous coals show a similar behavior: ignitionvia a homogeneous mechanism (sequential ignition of volatilesand char). The subbituminous coal has a heterogeneousignition mechanism, involving the simultaneous ignition ofvolatiles and char. Shaddix and Molina (2009)7 study ignitionand devolatilization of a bituminous and a subbituminous coalvia single-particle imaging. For their conditions (T∞ = 1700 K,high heating rates, 12−36 vol % O2, high-volatile coals, dp = 100μm), the coal particles appear to undergo homogeneousignition. Increasing O2 concentration accelerates particleignition. Zhu et al. (2011)8 develop a 1D model to studyignition of two bituminous coals and a lignite coal (1.5−2 mmin diameter) in a hot furnace of 1123−1500 K undermicrogravity conditions. When intraparticle thermal conditionis considered, the model predictions agree better with availableexperimental data. The transition diameter is in the range of600−800 μm, depending on T∞ and whether or not theintraparticle thermal conduction is considered. Khatami et al.(2012)9 investigate experimentally ignition of pulverized coalsand sugar cane-bagasse (dp = 75−90 μm) in a drop-tubefurnace electrically heated to 1400 K. As O2 concentrationincreases, strong tendencies are observed for all the fuels toburn in one-mode (i.e., simultaneous volatile and charcombustion at particle surface). Increasing coal rank enhancesthe tendency of the coal particles to burn in two-mode (i.e.,homogeneous volatiles envelope flames, followed by heteroge-neous char oxidation).Comparatively, there are very few studies on ignition of

biomass particles. For example, Grotkjær et al. (2003)10 did aseries of pulse ignition experiments by placing 2 g of wheatstraw sample in an isothermal reactor and varying temperaturein the range 483−629 K while maintaining gas velocity as 14cm/s. Such conditions are similar to those of the lower part ofthe biomass bed in commercial grate-firing boilers. Homoge-neous ignition mechanism, initiated and facilitated by reactionson the particle surface, is concluded. Kuo and Hsi (2005)11

measured the ignition delays of single spherical particles ofvarious woods (dp = 20 and 50 mm) heated in hot air flow of673, 773, and 873 K. Among all factors tested, the airtemperature is found to be most influential. A 1D model is alsodeveloped and is found to perform well in predicting the massloss of a wooden sphere prior to ignition.The conditions in the existing biomass ignition studies are

still distinctly different from those in a traditional pf flame. Theobjective of this work is to experimentally and numericallystudy ignition and burnout of large pine wood particles underconditions simulating large pf furnaces. The focus is placed onthe ignition delays and mechanisms and burnout times as wellas the impacts of particle properties and process conditions onignition and burnout.

2. EXPERIMENTAL METHODExperiments on pine wood particle combustion are conducted in asingle particle combustion reactor rig, which mainly consists of areactor, a burner, a safety system, and a gas supply system. A large

number of injection nozzles (94) and a flat profile of the burner resultin good mixing of gases, a uniform gas flow distribution, and a flattemperature profile in the reactor center. The gas temperature, oxygenconcentration, and air velocity in the reactor are well controllable. Thetested gas temperature and oxygen concentration are in the ranges1473−1873 K and 5−20%, respectively, which are selected to be closeto the conditions in large pulverized wood fired boilers. The entirecombustion process is recorded using a high performance camera, andthe ignition, devolatilization, and char oxidation are characterized fromthe image analysis. In order to ensure the repeatability of the results,some tests are repeated 3 to 5 times. All the details about the test rigand experiments can be found in the experimental study ofcombustion characteristics of single biomass particles.12 Here, theexperimental results mainly serve as inputs for the modeling work tobe presented in this paper.

3. MATHEMATICAL MODELINGFor a cold biomass particle suddenly exposed to a hotenvironment, it will be heated and then undergo a series ofconversion processes (e.g., drying, devolatilization, char gas-ification, and oxidation). If all parameters at time t are known,mathematical modeling of a large, thermally thick biomassparticle conversion for the parameters at a new time t + Δt canbe subdivided into three coupled issues: (1) external heat andmass transfer to or away from the particle, (2) biomass particleconversion analysis, in which the key intraparticle processesneed to be properly considered for a thermally thick particle,and (3) particle dynamics. The first issue involves empiricalcorrelations, if the detailed profiles of temperature or species inthe boundary layer are not to be resolved. The second is tosolve some coupled partial differential equations to obtain thedetailed profiles within the particle. The third needs to addressa few ordinary differential equations to update particle velocityand position. There are some similar modeling works in theliterature.13−15

3.1. Governing Equations and Boundary/Initial Con-ditions. In this study, the issue of particle dynamics is left outsince the particle is suspended in the reactor without anymovement. The intraparticle conversion analysis is the mainchallenge, in which the following governing equations aresolved by using finite volume method to update the velocity,temperature, and gas species profile in a porous biomassparticle as a function of time.

∑ ∑

∑

ερερ

ημ

ρ ρ ερ

ερ

ερ ερ

∂

∂+ =

= −∂∂

∂∂

+ +

+

= ∇ + − ⎯→ +

∂∂

+ = −⎯→ +

⎧

⎨

⎪⎪⎪⎪⎪⎪⎪⎪

⎩

⎪⎪⎪⎪⎪⎪⎪⎪

tu S

upr

th h h

uh

k T h J S

tY uY J S

( )div( )

( )

div( )

div( ) div( )

( ) div( ) div( )

ii i

kk k

jj j

j j i Y

gg g

solid, liquid,g

g

eff h

g g j (1)

where ε, ρg, t, u , Sg, η, μ, p, r, h, keff, T, Jj, Sh, Yj, and SYj representporosity, gas density, time, gas velocity, source term to gas massconservation, permeability, gas dynamic viscosity, pressure,radius, sensible enthalpy, effective heat conductivity, temper-ature, diffusion flux of gas species j in the gas mixture, source

Energy & Fuels Article

dx.doi.org/10.1021/ef302153f | Energy Fuels 2013, 27, 1061−10721062

term to the energy equation, mass fraction of gas species j, andsource term to j-th gas species transport equation, respectively.The velocity u is only for gas phase and both solid and liquidphases are assumed to be stationary.Here, it has to be mentioned that a simplified momentum

equation (i.e., a Darcy law type equation) is used instead toevaluate gas velocity inside the porous particle. In the Darcyequation, the pressure is evaluated from the ideal gas law, p =(ρgRuT)/MW, in which Ru and MW denote the universal gasconstant and molecular weight of local gas mixture,respectively. A local thermal equilibrium is assumed amongthe different phases inside the particle so that a lumped energyequation for all the phases is employed. In the energy equation,the sensible enthalpy of the gas mixture is calculated by h =ΣjYjhj, where the sensible enthalpy of individual gas species isdefined as hj = ∫ Tref

T Cp,j(T) dT. The diffusion flux of gas species jin the gas mixture is evaluated by J = −ρgDj,m▽Yj, where Dj,m isthe effective mass diffusivity of species j in the gas mixture.Here, the general governing equations in terms of divergenceoperator, rather than dimension- or coordinate-dependentequations, are used. The former can be better and more easilyintegrated into finite volume method, provided the basicunderstanding of the governing equations and finite volumemethod as well as the basic knowledge in calculating distance,surface area and cell volume for various geometries.All the parameters (e.g., velocity, temperature, and species)

are assumed to vary only in the radial direction. So, it turns outto be a 1D transient problem, and the governing equations aresolved under the following free-stream (or process), initial, andboundary conditions (BCs):

ρ ρ

ϕ

ε σ

‐ = == = =

== = =

∂∂

=

| =

∂∂

= − +

−

∂∂

= −

∞ ∞

∞ ∞ ∞

=

∞

∞

⎧

⎨

⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪

⎩

⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪

T T u UY Y

up p T Y

r

p p

k ATr

h A T T

A T T

D AY

rh A Y Y

Free stream condition: , ,0.23, 0.77,

Initial condition in particle: 0,, 300 K, 0

BCs at particle center: 0 (symmetry)

BCs at particle surface:

( )

( )

( )

j

r

jj

j j

f f

O , N , f

atm

0

s atm

eff ss

T s s emis

s rad4

s4

,m ss

M s , ,s

2 2

(2)

where the subscripts ∞ and s denote the free-stream and theparticle surface, respectively. As, εemis, σ, hT, Trad, and hMrepresent particle surface area, particle emissivity, Stefan−Boltzmann constant, heat transfer coefficient, radiation temper-ature, and mass transfer coefficient, respectively.To reliably handle the BCs at particle surface, empirical

correlations applicable to a certain particle shape andappropriately account for the Stefan flow effect need to beused to calculate the convective heat and mass transfercoefficients, hT and hM. The resistance to the convective heatand mass exchange between the particle surface and free-streamgas flow is assumed to be within a gas film (or boundary layer)of thickness, δT and δM. For example, for spherical particles

without Stefan flow effect, the heat and mass transfercoefficients can be calculated as follows:

≡ = +

≡ = +

⎧

⎨⎪⎪

⎩⎪⎪

h d

k

h d

D

Nu 2.0 0.64Re Pr

Sh 2.0 0.64Re Sc

0T p

g

0.5 0.33

0M p

g

0.5 0.33

(3)

where Nu0 and Sh0 represent Nusselt and Sherwood number,respectively. The average physical properties in the boundarylayer need to be used in eq 3. For example, the average gasconductivity kg, gas diffusion coefficient Dg, viscosity μg , Prandtl

number Pr, and Schmidt number Sc are all evaluated at somereference temperature and gas species mass fraction, forexample,

= + − = + −∞ ∞T T T T Y Y Y Y13

( );13

( )j j j jref s s ,ref ,s , ,s

(4)

The Reynolds number in eq 3 is calculated based on the free-stream density ρ∞ and the average viscosity μg in the boundary

layer, Re = (ρ∞|u∞ − up|dp)/(μg), since it is interpreted as a

ratio of inertia to viscous forces. The thickness of the boundarylayer for heat and mass transfer without Stefan flow effect, δT0and δM0, can be calculated as follows:

δ δ=−

=−

d d

Nu 2;

Sh 2T0p

0M0

p

0 (5)

In this study, the effect of the Stefan flow is accounted for inthe calculation of the heat and mass transfer coefficients. Asurface blowing results in the thickening of the heat and masstransfer boundary layer by a factor of FT and FM, respectively,

δδ

δδ

≡ = ++

≡ = ++

⎧

⎨⎪⎪

⎩⎪⎪

F BB

B

F BB

B

(1 )ln(1 )

(1 )ln(1 )

TT

T0T

0.7 T

T

MM

M0M

0.7 M

M (6)

where BT and BM represent the Spalding heat and mass transfernumbers, respectively. When the Stefan flow effect isconsidered, the Nusselt and Sherwood numbers are calculatedbased on the model of Abramzon and Sirignano (1989),16

which substantially refines the classical model,17

= + −+

= + −+

⎧

⎨⎪⎪

⎩⎪⎪

FB

B

FB

B

Nu (2 (Nu 2)/ )ln(1 )

Sh (2 (Sh 2)/ )ln(1 )

0 TT

T

0 MM

M (7)

from which the external heat and mass transfer coefficients, hTand hM, can be evaluated.To investigate ignition mechanisms, the boundary layer

domain with a thickness of δT and δM is also included anddiscretized in the modeling. Thus, the profiles of keyparameters of interest (e.g., temperature and gas species) inthe boundary layer are also numerically solved, using thefollowing boundary conditions instead:

| = | = | =δ δ≥ = + ∞ = + ∞p p T T Y Y; ;r r r r j r r jatm ,p p T p M (8)



During time marching in the model, the conservation of mass,heat, and species flux at the gas/solid interface (i.e., the particlesurface) is secured by representing the fluxes through theinterface in a consistent manner in the adjacent controlvolumes located at both sides of the particle surface.For cylindrical particles, the above process is followed, in

which the 1D model is still used assuming the main variationsonly in the radial direction. The same, simplified way is veryoften employed in modeling of cylindrical biomass particlecombustion in literature.13,15 One of the main differences inmodeling of cylindrical particle combustion is to replace eq 3with correlations that are more applicable to cylindricalparticles, as used in the similar modeling works.13,15

3.2. Particle Conversion Subprocesses and Mecha-nisms. In the governing equations, eq 1, the number of gasspecies transport equations and the various source termsdepend on how the physical conversion processes areconsidered. The species and the reactions including their rateexpressions and kinetic data are summarized in Table 1, fromwhich all the source terms in the transport equations can bereadily evaluated.For the drying processing, the release of both the free water

and bound water can be taken into account. Moisture contentabove the fiber saturation point (FSP) is assumed as free water

and exists in pores and cells in liquid form. Moisture below theFSP is considered as bound water, which exists as hydratespecies in the particle. The average FSP, the ratio of the weightof water in the wood to the weight of oven-dry wood, is about30%.21 The moisture content of the wood particle samples usedin this study is about 10%. Therefore, all the moisture has beenassumed as bound water in modeling and is released bychemical reaction, as shown in Table 1.When the particle is further heated, biomass starts to be

decomposed to noncondensable gases, condensable species(e.g., water and organic compounds), and char via differentreaction routines, in which the organic vapor degrade further toform chars, noncondensable gases, and water if held in contactwith the solid biomass undergoing devolatilization.22 Accord-ingly, a two-step devolatilization model is used in this work.The biomass is decomposed to light gases, tar, and char. Then,the tar can be further converted into light gases and char, asshown in Table 1, in which the composition of the light gasesand the three homogeneous reactions are also given.The char left in the biomass particle can be oxidized via

various ways, depending on the gas species available in the solidmatrix and at the particle surface. As shown in Table 1, threeheterogeneous reactions are considered here: char oxidationwith oxygen, carbon dioxide, and water vapor. Also from Table

Table 1. Chemical Reactions Considered: The Rate Expressions and Kinetic Data Used in k = ATb e−E/(RuT)

kinetic data and heat of reaction

subprocess reactions rate expressions A [1/s] b E [kJ/mol] ΔH [kJ/kg]

evaporation H2O(l, free) ↔ H2O(g) ρ ρρ

ρ ρ=∂∂

= −rt

S h Y( )1fw

afw

fw0 m,pore v

satv g

−2440

H2O(l, bound) → H2O(g) ρρ=

∂∂

=rt

k2bw

2 bw

5.13 × 1010 0 88

devolatilization biomass → light gas* ρρ=

∂∂

=rt

k3B

3 B

4.38 × 109 0 152.7 −418

biomass → tar ρρ=

∂∂

=rt

k4B

4 B

1.08 × 1010 0 148

biomass → char ρρ=

∂∂

=rt

k5B

5 B

3.27 × 106 0 111.7

tar → light gas ρε ρ=

∂∂

=rt

k Y6G

6 g T

4.28 × 106 0 107.5 42

tar → char ρε ρ=

∂∂

=rt

k Y7C

7 g T

1.0 × 105 0 107.5

light gas combustion CO + O2 → CO2 = ∂∂

=rt

k[CO][CO][O ] [H O]8 8 2

0.252

0.5 1 × 1012.35 0 167 10110

H2 + O2 → H2O =∂

∂=r

tk

[H ][H ][O ]9

29 2 2

1.42 1 × 1012.71 0 171.3 120900

CxHyOz + O2 → CO + H2 = ∂∂

=rt

k[HC][HC] [O ]10 10

0.52

T·P0.3·1 × 104.32 0 80.2 41600

char oxidation C + O2 → CO ρρ ρ ρ

ε=∂

∂=

+ +r

t

Sk

[O ][O ]11

2 a,char C

C B A11 2

0.658 m/(s·K) 1 74.8 9212

C + CO2 → CO ρρ ρ ρ

ε=∂

∂=

+ +r

t

Sk

[CO ][CO ]12

2 a,char C

C B A12 2

3.42 m/(s·K) 1 130 14370

C + H2O → CO + H2 ρρ ρ ρ

ε=∂

∂=

+ +r

t

Sk

[H O][H O]13

2 a,char C

C B A13 2

3.42 m/(s·K) 1 130 10940

*Composition of the light gas produced during devolatilization process (Thunman et al., 2001):18 CO = 0.396, CO2 = 0.209, H2 = 0.019, H2O =0.249, light hydrocarbon (C6H6.2O0.2) = 0.127 in mass fraction. Unit of [Gas]: kmol gas/m3. The sources for the kinetic data in the 1st step of thepine wood devolatilization (i.e., biomass → light gas, tar, and char): kinetic data in k3, k4 and k5 expressions,

19 and heat of reaction.20 Data source forother reactions which may be rather standard or general.13



1, the density of various solid and liquid species can be readilyupdated from the given rate expressions, which are ordinarydifferential equations.3.3. Physical Properties. After the detailed particle

conversion mechanism is addressed, as listed in Table 1, theremaining issue to close the system of equations and make themodel solvable is to evaluate physical properties of the rawbiomass particles and all the solid, liquid, and gas speciesinvolved in the conversion mechanism. The approximate andultimate analysis, heating value, and density of the pine woodare given in our experimental study.12 Other properties aboutthe pine wood particles, including porosity ε, emissivity εemis,permeability η, pore size dpore, heat conductivity k, specific heatCp, biomass particle specific surface area Sa (i.e., the ratio ofsurface area to particle volume, m−1), and char particle specificsurface area Sa,char, are considered in the same way as themodeling of wood particle combustion.13 During theconversion process, a constant porosity of 0.4 is used for theparticle domain13,14 and 1 for the boundary layer domain. Theparticle shrinkage or swelling during drying, pyrolysis, and charoxidation are also accounted for by using an empiricalcorrelation.13 All the physical properties of water and variousgas species considered in the model are readily available inhandbooks.23

3.4. Numerical Method. The particle is divided into anumber of spherical shells (or cells) in the radial direction. Thegoverning equations in eq 1 are discretized using the finitevolume method. For all the transport equations, the fullyimplicit scheme is used for the transient terms and the powerlaw scheme is employed for the convective-diffusion terms,both of which are readily implemented. Since the reactionsconsidered in the conversion mechanisms are highly nonlinearin regard to temperature and species concentration, the sourceterms in the governing equations are cumbersome. Appropriatetreatment of the source terms is crucial to ensure the stability ofthe model. The source terms are linearized by following therule of negative-slope linearization,24 as discussed anddemonstrated in discretizing the energy equation for theoutermost cell in a simplified single particle model.25

The grid-independence solution is always an important issuein numerical modeling and simulations. In this study, meshes ofvarious resolutions are tested to secure a grid-independentsolution. For example, in the 1D model of the 3 mm-in-diameter spherical particle, the particle from the center to thesurface is finally divided into 40 spherical shells and theboundary layer outside the particle surface is divided into 20spherical shells. A time step size of 1 ms is used in the transientmodeling.

4. RESULTS AND DISCUSSIONTwo groups of pine wood particles are studied.12 The firstgroup consists of one 3 mm-in-diameter spherical particle andfour cylindrical particles that have the same mass (or volume)as the spherical particle but different aspect ratios of 2, 4, 6, and8, respectively. The second group consists of one 3 mm-in-diameter spherical particle and three other cylindrical particlesthat have the same diameter of 3 mm but different lengths of 6,12, and 18 mm, respectively. The ignition delays and burnouttimes of the two groups of particles under process conditionsclose to those in a pf flame are studied both experimentally andnumerically.4.1. Model Verification. An extra comparison between the

model and ignition experiments under thermogravimetric (TG)

conditions is provided here, before the model is applied tosimulate ignition and burnout of the large, differently shapedpine wood particles under various conditions in the singleparticle combustion reactor. One has to be aware of that TGconditions are distinctly different from those in a pf furnace.Therefore, the TG experiment has nothing to do withproviding insight into the pine wood particle ignition in a pfflame and is only to collect data for a quantitative comparisonwith the model. In the TG experiment, 5 mg pine wood chips,pared by a knife from the regularly shaped pine woodparticles12 and used to simulate the pine wood chips actuallyfired in a power plant, are heated in a thermal analyzer from303 to above 1273 K with a heating rate of 500 K/min. Themass loss history is recorded.In order to reasonably compare the model prediction with

the experimentally recorded mass loss, the conditions in themodel are appropriately adapted to some extents. In the model,the radiation effect is removed. The ambient gas velocity is setto a negligible value, for example, 0.02 m/s. The ambient gastemperature is set as a linear function of the heating rate. Asingle spherical particle is assumed and used, whose diameter iscalculated based on the mass of the sample used in TGA (i.e., 5mg) and the density of pine wood. However, the kinetic dataassociated with pine wood conversion are kept unchanged inthe model. Figure 1 shows the comparison between the scaled

mass loss recorded in TG analysis and the model prediction,which shows acceptable agreement. One has to be aware of thata sample of 5 mg pine wood chips is used in the TG analysis,while a single spherical pine wood particle of equivalent mass isused in the modeling. It will have some impact on the ignitionand combustion process due to the cooperative effect in theformer associated with an increase in the concentration of thecombustible gases.5 It will also induce difference in the heattransfer process inside the wood particles and in the diffusionand accessibility of oxygen to the wood particles. Keeping inmind this difference and other uncertainties in preciselydefining the ambient conditions in the model, the discrepancyin Figure 1 may be understood. The remarkable difference in

Figure 1. Comparison of the scaled mass loss recorded in TG analysis(5 mg pine sample, YO2

= 5%, 500 K/min, from 303 to above 1273 K)and the model prediction.



char oxidation process is most likely from the char oxidationrates. The objective of the modeling study is to investigate theignition and burnout of pine wood particles underrepresentative conditions in a pf flame. Fast pyrolysis chars,as used in the model kinetics, are significantly more reactivethan slow pyrolysis chars from the TG experiment.26 Therefore,the model predicts a faster char oxidation than the experimentalplot from the TG analysis.4.2. Particle Ignition. A study on fuel particle ignition

often involves ignition delay, ignition temperature, and ignitionmechanism, and their dependency on fuel properties andprocess conditions.As an example, Figure 2 shows the ignition process of a 3

mm-in-diameter spherical pine wood particle in the reactorunder the ambient air temperature of 1473 K, molar fraction ofoxygen of 21%, and superficial air velocity of 1.5 m/s. Thecommonly used ignition indicators in experimental studiesinclude detection of a light flash, increase in the luminous flux,change in the mass loss, increase in particle temperature, rapiddecrease in CO and O2 combined with increase in CO2 andNO, increase in CO2/CO ratio, exothermic peaks ondifferential thermal analysis curves combined with mass loss,and use of devolatilization images.1,5−7,10,11,27 Ignitionindicators can have a significant impact on the estimatedignition delay, ignition temperature, or even interpretedignition mechanism. In this experimental study, detection of alight flash is used as the ignition indicator, since neither theparticle temperature/mass nor the evolved gases are measured.From the light flash itself, it is hard to distinguish the ignitionmechanism under the experimental conditions, in which boththe upward gas velocity (∼1.5 m/s) and gravity-inducedbuoyancy play important roles in the shape and position of theflame. It is also difficult to distinguish whether the light flash isdue to volatile matter ignition or char ignition. From theimages, one may only conclude that the ignition delay is about0.1318 s; that is, the first image in Figure 2 corresponds to theonset of ignition.To identify the ignition mechanism, simulations are done

using the 1D transient particle model, from which much moredetailed information can be extracted, for example, theinstantaneous temperature and species profiles inside theparticle and in the boundary layer, reaction rates, and massloss of the particle. Once quantitative ignition criteria areestablished, the ignition mechanism and the possible shiftbetween them could be detected.For a fuel particle dropped into hot ambient air with constant

gas temperature Tg and radiation temperature θR, the particletemperature Tp will increase but the heating rate dTp/dt willdecrease before any remarkable conversion process occurs (i.e.,

dmp/dt = 0), as indicated by eq 9. At this stage, only the firstterm on the right-hand side contributes to the particle heatingprocess. For a dry fuel particle, once heterogeneous ignitionoccurs on the particle surface, the third term on the right-handside in eq 9 will be an extra heat source term. This extra source,together with a decreasing particle mass, renews an increase inthe particle heating rate. It is often called the inflectioncondition, as given in eq 10, and commonly used as thecriterion for the onset of heterogeneous ignition.1,3−5,8

ε σ θ=

− + −

+ +( ) ( )

T

t

hA T T A T

m C

H

m C

H

m C

d

d

( ) ( )

m

tm

t

p p g p p p R4

p4

p p

d

d evapevap

p p

d

d reacreac

p p

p p

(9)

= ≥d T

dt

T

t0 and

d

d0

2p

2p

(10)

In this study, the pine wood particles contain about 10 wt %moisture. Such a heterogeneous ignition criterion is stillexpected to work. Once evaporation starts, the evaporationheat sink gives rise to a further decrease in the particle heatingrate, as shown in eq 9. However, the reduced particle mass hasan opposite impact on the heating rate. Both the effects mightbe offset to some extent so that the particle heating ratedecreases slightly or stays nearly unchanged before a sudden,detectable sharp increase in the heating rate which correspondsto the onset of heterogeneous ignition.Homogeneous ignition occurs when the gas temperature,

oxygen, and volatile concentration reach the flammability limit.In this case, ignition will occur in a narrow region surroundingthe particle, since the oxygen and volatiles are diffusing intoeach other. Therefore, a fuel particle is said to have ignitedhomogeneously if the gas temperature of a shell is greater thanthe temperature of both its adjacent shells, that is, a gastemperature peak founded at a radial location in the boundarylayer,

> >+Δ −ΔT T T Tandr r r r rg,r g, g, g, (11)

This homogeneous ignition criterion is also commonlyused.1,3−5,8

In this study, the temperature profile along the radialdirection from the center of the particle to the outer boundaryof the gas film surrounding the particle is produced by themodel. Then, the criteria, eq 10 and 11, can be evaluated. Thecriterion that is satisfied first during the transient ignition,

Figure 2. Ignition of a 3 mm-in-diameter spherical pine wood particle in the reactor (T∞ = 1473 K and O2 = 21%).



assisted by the variations in gas species and heterogeneous massburning rate, determines the primary ignition mechanism.Figure 3 shows the model predictions of the ignition process

of a 3 mm-in-diameter spherical pine wood particle. It has beenwell established that the ignition mechanism will shift fromheterogeneous ignition for small particles to homogeneousignition for large particles.1,3,8,9,27 The transition particle size atwhich such a shift occurs strongly depends on the ambientconditions, fuel properties, and even the ignition indicators.Howard and Essenhigh (1967)28 propose an empiricalcorrelation to estimate the transition particle size, dt, for agiven set of combustion conditions and fuel properties, asshown in eq 12. If a particle is smaller than this size, it issubjected to heterogeneous ignition on the particle surface.Otherwise, the particle is more likely subjected to primary

homogeneous ignition of the evolved volatiles in the boundarylayer, which is followed by heterogeneous burning of chars.

ρ=

· · −∞Δd

D P x x

y f RT

12 ( )V V

t it

O , O ,sd( / )

d F v v

2 2

v 0

(12)

where D, P, xO2,∞, xO2,s, ΔVv, V0, ρF, yv, f v, R, and Ti denote thebinary diffusion coefficient of O2 in N2 at the background gastemperature and pressure [cm2/s], total pressure of thebackground gas [atm], molar fraction of O2 in the ambientgas, molar fraction of O2 on the outer surface of the flame sheetsurrounding a particle, volatile matters evolved from solid fuelas pyrolysis products [g], initial volatile matters in the fuelparticle [g], fuel density [g/cm3], mass fraction of volatilematters in the original fuel, moles of O2 required to burn a unitmass of volatiles [mol/g], ideal gas constant [(cm3·atm)/

Figure 3. Conversion of a 3 mm-in-diameter spherical pine particle in the ambient air flow (T∞ = 1473 K, O2 = 21%, upward speed 1.5 m/s): (a)particle center, surface, and volume-weighted average temperatures and mass burning rate of char; (b) moisture, volatiles, and char left in the particlescaled by initial particle mass and CO, CO2 mass fractions at particle surface; (c) instantaneous temperature profiles from the particle center to theouter boundary of the gas film.



(mol·K)], and ignition temperature of the fuel in a givenbackground gas [K], respectively. Although the transitionparticle sizes greatly depend on fuel properties and processconditions and the concrete values vary from study to study,they are generally on the order of several hundred micro-meters.1,3,8,9,27

The spherical pine wood particle, 3 mm in diameter, is muchlarger than the most commonly reported transition particlesizes in literature. Moreover, the pine wood particle contains77.4% volatiles, which favors a shift to homogeneous ignitionsince the flammability condition in the boundary layer can beeasily achieved. For instance, Du and Annamalai (1994)3 findthe primary ignition of coal particles becomes homogeneouswhen the volatile content is increased above 70%. The ambientgas temperatures in this study (1473, 1673, and 1873 K) arewell above the autoignition temperatures of various hydro-carbons and H2 (874 K) or CO (882 K),11 which also favors ahomogeneous ignition. Therefore, one may expect a primaryhomogeneous ignition mechanism for the large pine woodparticles that are suddenly exposed to the hot ambient airstream.However, a primary heterogeneous ignition is still detected

from the modeling results. It could be understood. When thepine wood particles are heated, the three major components inthe particles decompose to release volatile matters at differenttemperatures, in which hemicellulose, cellulose, and lignintypically decompose in the temperature ranges 473−533 K,513−623 K, and 553−773 K, respectively.22 The chemicalcomposition of the pine wood particles may be seen from thedata in literature.29,30 Pines from different sources are found toshow comparatively invariant chemical composition: containing65−69 wt % holocellulose (total hemicellulose and cellulose, inwhich cellulose content is 42 ± 2 wt %), 24−30 wt % lignin,and 5−7 wt % extractives. For the pine particles of severalmillimeters in diameter, as used in this study, there exist largetemperature gradients inside the particles, which can be seenfrom Figure 3(a). When the particle surface is heated to above573 K, which is enough for a remarkable heterogeneous surfaceburning, the particle center, as well as the majority of theparticle, is still quite cold. The decomposition of hemicelluloseand cellulose only occurs at the outer radius of the particle,slowly releasing a small amount of volatiles, which can neitherscreen the solid from direct oxygen attack nor form a flammablegas mixture in the boundary layer. Therefore, the onset ofignition detected from the first light flash in experiments isconcluded to be the primary heterogeneous ignition, which iscommonly defined as remarkable solid oxidation and/orvolatile combustion at the particle surface. The solid at suchlow surface temperatures (ca. 600 K) does not emit visibleradiation. As a result, it may be concluded that the light flashdetected in experiments is from ignition of volatiles at theparticle surface, facilitated by local char oxidation. The primaryheterogeneous ignition can also be confirmed by the onset ofthe heterogeneous mass burning of char, as shown in Figure3(a). At t ≈ 0.4 s, the primary heterogeneous reaction ratereaches a local maximum peak. However, it may beinappropriate to interpret t ≈ 0.4 s as the onset of the primaryheterogeneous ignition. Instead, the moment at which aremarkable char oxidation starts at the particle surface isinterpreted as the primary ignition delay, t ≈ 0.15 s,corresponding to a particle surface temperature of 550 K.The carbon oxidation rate of the reaction, C + 1/2O2 = CO, canbe estimated from

≈ × − ××

× ×

× × ≈·

ε

ρ

−

⎛⎝⎜

⎞⎠⎟r 0.658 550 exp

74.8 108315 550

0.4 10

0.1 0.004 0.005kmol O

m s

k

S11

6

, in the unit of m/s

6

,m

ratio [O ]

23

11

a1

2

So, the carbon oxidation rate is estimated to about 0.005 × 2 ×12 = 0.12 kg C/(m3·s) at Ts = 550 K. The detailed kinetics ofthis reaction can be seen in Table 1. The similar value of theprimary carbon oxidation rate can be used to characterize theonset of the primary heterogeneous ignition of other cases inthis study.When the particle is further heated, the evaporation front,

pyrolysis front, and char oxidation front eventually move insequence from the particle surface to the particle center. Thevolatiles released from the outer radius pass through the solidmatrix and burn at the particle surface. This, together with themore remarkable heterogeneous burning at elevated temper-atures at the outer radius, contributes to the primaryheterogeneous ignition.When the entire particle is heated to a moderately high

temperature, e.g., the volume-weighted average temperatureabove 523 K, significant decomposition of hemicellulose andcellulose starts rapidly in the entire particle, as seen from theplot of volatiles in Figure 3(b). A large amount of volatiles arerapidly released from the interior of the particle, which not onlyeffectively screens the solid from direct oxygen attack but alsoforms a flammable gas mixture in the boundary layer.Therefore, the primary heterogeneous ignition quenches anda secondary homogeneous ignition starts. This can also be seenfrom the sharp decrease in the burning rate of char as well asthe sharp decrease in CO and CO2 concentration at the particlesurface. The secondary homogeneous ignition time can also beestimated from the instantaneous temperature profiles in theradial direction, as shown in Figure 3(c). Here, thehomogeneous ignition time is estimated to be 0.4 s,corresponding to the moment when a temperature peak isfirst detected in the boundary layer.The experiments and modeling are repeated for differently

shaped pine wood particles. The ignition delays andmechanisms are detected in the similar way, as summarizedin Figure 4. For all the large pine particles exposed to suchambient conditions (air temperature of 1473 K and oxygenfraction of 21%), the primary heterogeneous ignition is alwaysdetected, which is quenched later on and replaced by asecondary homogeneous ignition. The secondary homogeneousignition will continue until all volatiles are released, which isfinally followed by heterogeneous char oxidation. Figure 4shows the effect of particle shapes on the ignition delays. Forthe secondary homogeneous ignition, both the experimentaland modeling results show the similar trend as theoreticalanalysis. The homogeneous ignition delays somehow scale withthe particle specific surface areas, i.e., the larger the specificsurface area is, the shorter the ignition delays will be.Comparatively, the heterogeneous ignition delays show lessdependency on the particle specific surface area, which may beconsistent with the literature.4 The impact of particle shape onparticle ignition may not be explained simply in terms ofspecific surface areas since ignition depends on complexinteractions involving external and internal heat and masstransfer and chemical reactions. Particularly for irregularlyshaped particles, the release and burning of volatiles, as well as



surface char oxidation, are more likely to start at a certain edgeor corner of the particles.4.3. Particle Burnout. The same as in our experimental

study,12 a number of times are defined and used here. The timeelapsed from the moment when the cold pine wood particle isexposed to the hot ambient gas flow to the moment when theflame around the particle disappears is termed as thedevolatilization time. The burnout time is defined as the entireperiod from the cold pine particle being exposed to the hotambient stream to the complete burnout of the particle (i.e.,only ash left and no remarkable particle shrinkage observed).

Char burning time is assumed to be the time for consuming outthe char alone (without flame). However, for such largeparticles, the char burning process is inevitably overlapped withother processes. So, the char burning time is underestimatedand only serves as reference here.Figures 5 shows how the particle mass loss, char burning, and

burnout times change with particle shapes for the five pinewood particles that have the same mass (or volume). For theparticles of the same volume, the particle specific surface areaincreases with the increase in the aspect ratio. The largerspecific surface area favors a faster heating of the particle andthus a faster conversion. Moreover, a high aspect ratio alsoeffectively reduces the temperature gradients in the radialdirection, which also favors a faster conversion. In Figure 5, themodel-predicted burnout times are also plotted against theexperiments-derived burnout times, which shows a goodagreement and the trend is also coincident with theexpectations. Here, it has to be mentioned that the burnouttimes might be too long for a practical pf furnace, which can beattributed to two factors. All the five particles have an equi-volume diameter of 3 mm, which may represent the maximumparticle size in a pulverized wood-fired boiler. In addition,though the heating rate and temperature levels in the reactorare selected to be close to the realistic conditions in largepulverized wood-fired boilers, other conditions in the reactor,such as flow, turbulence and radiation, might still be distinctlydifferent from those in a real furnace.From a practical point of view, biomass particles prepared for

cofiring in power plants are more likely to have a similardiameter (rather than similar mass) and different lengths, dueto the milling process. Therefore, the second group of pinewood particles is also studied: a 3 mm-in-diameter sphericalparticle and three cylindrical particles that have the samediameter of 3 mm but different lengths (L = 6, 12, and 18 mm,respectively). For this group of particles, the particle specificsurface area decreases with the increase in the aspect ratio.Therefore, one would expect lower conversion rates when the

Figure 4. Conversion of the five pine particles with the same mass(volume) but different aspect ratios in the ambient air flow (T∞ =1473 K, O2 = 21%, upward speed 1.5 m/s) in which aspect ratio of 1corresponds to a 3 mm-in-diameter spherical particle (the higher theaspect ratio, the larger the specific surface area): effect of particleshapes on ignition delays.

Figure 5. Conversion of the five pine particles with the same mass (volume) but different aspect ratios in the ambient air flow (T∞ = 1473 K, O2 =21%, upward speed 1.5 m/s) in which aspect ratio of 1 corresponds to a 3 mm-in-diameter spherical particle: (a) model-predicted particle mass losshistory; (b) char burning and burnout times as a function of aspect ratio.



particle length increases. The three cylindrical particles areexpected to have more or less similar conversion rates becauseof the similar specific surface area. The model well capturesthese trends, as shown in Figure 6. The model-predictedconversion times are also quantitatively compared with theexperimental data, which also shows a good agreement.Comparatively, the model slightly overpredicts the burnouttime, which becomes more remarkable for longer cylindricalparticles. It is mainly because the model only considers theradial variations (i.e., 1D model) while for a long, slimcylindrical particle the 1D assumption may give rise to somederivations. One would expect an earlier and easier conversionat the edges or corners of the cylindrical particles, whichinitializes and speeds up the complete conversion of the entireparticle.Process conditions also play an important role in fuel particle

burnout. Figure 7 shows the effects of both the ambient gastemperature and oxygen concentration on burnout times of acylindrical pine wood particle (diameter of 1.65 mm, length of6.60 mm, and aspect ratio of 4). For burning of pure charparticles, increasing oxygen concentration and ambient gastemperature both always have a favorable impact: they willremarkably speed up the burning process, as revealed by theexperimental and modeling results in Figure 7. For large pinewood particles containing moisture volatiles, char, and ash, theignition, devolatilization, volatiles combustion, and charburning are inseparable processes. To better interpret theimpacts of oxygen concentrations and ambient temperature onburnout times of the pine wood particle, one has to look intotheir effects on all the individual processes. Increasing O2concentration accelerates particle ignition,3,7 which in turnaccelerates heat release, particle heating, and devolatilization.Increasing O2 concentration also speeds up homogeneouscombustion of the released volatiles and, in turn, furtherintensifies the local heat release. Therefore, higher oxygenconcentrations are expected to remarkably shorten thedevolatilization time. As discussed in detail in our experimentalstudy,12 increase in O2 concentration at low temperatures mayhave double, positive impacts in enhancing char oxidation:

greatly speeding up particle heating because of the acceleratedparticle ignition and intensified volatiles combustion whicheffectively reduces the kinetic resistance in char oxidation, andlargely enhancing char oxidation because of richer oxygenavailability. Therefore, increasing O2 concentration at lowtemperatures is expected to have a more pronounced impact onchar oxidation at low temperatures than at high temperatures.All these trends are well reproduced from the experiments andalso relatively well predicted by the model, as seen in Figure 7.

5. CONCLUSIONS

Ignition and burnout of large, differently shaped pine woodparticles (a few millimeters in size) in hot ambient gas stream(1473−1873 K) containing oxygen (5−20%) are investigatedexperimentally and numerically. The following conclusions canbe drawn.For all the test conditions, primary heterogeneous ignition is

always detected, though primary homogeneous ignition mightbe expected based on the literature. The large temperaturegradients inside the particles play a very important role in theignition process, in both ignition delays and ignitionmechanisms. When a large pine wood particle is suddenlyexposed to a hot gas stream, the surface temperature increasesquickly, initializing the primary heterogeneous ignition at theparticle surface. However, the majority of the particle is stillunder low temperatures. Only when the volume-weightedaverage temperature is raised above 523 K does significantdecomposition of hemicellulose and cellulose starts, whichrapidly releases a large amount of volatiles from the particle.The released volatiles effectively screen the solid from directoxygen attack (which quenches the primary heterogeneousignition) and also form a flammable gas mixture in theboundary layer (which starts the secondary homogeneousignition).The model-predicted ignition delays and burnout times show

a generally good agreement with the experimental results for allthe test cases. Homogeneous ignition delays are found to scalewith the particle-specific surface areas, that is, the larger the

Figure 6. Conversion of the four pine particles having the same diameter of 3 mm but different lengths (L = 6, 12, and 18 mm) in the ambient airflow (T∞ = 1473 K, O2 = 21%, upward speed 1.5 m/s): (a) model-predicted particle mass loss history; (b) devolatilization and burnout times as afunction of specific surface area.



specific surface areas, the shorter the homogeneous ignitiondelays. Comparatively, heterogeneous ignition delays show lessdependency on the specific surface areas.The ignition and burnout are also affected by the process

conditions. Higher oxygen concentration and higher oxidizertemperature can greatly accelerate ignition, devolatilization, andchar burning, among which oxygen concentration is found tohave a more pronounced impact on the ignition delays andburnout times at lower oxidizer temperatures.

■ AUTHOR INFORMATION

Corresponding Author*Phone: +45 99409279. Fax: +45 98151411. E-mail: [email protected]; [email protected].

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

This work was financially supported by DONG energy. Theauthors would like to thank Hong Lu for sharing his results.

■ REFERENCES(1) Essenhigh, R. H.; Misra, M. K.; Shaw, D. W. Ignition of coalparticles: A review. Combust. Flame 1989, 77, 3−30.(2) Wall, T. F.; Gupta, R. P.; Gururajan, V. S.; Zhang, D. K. Theignition of coal particles. Fuel 1991, 70, 1011−1016.(3) Du, X.; Annamalai, K. The transition of isolated coal particle.Combust. Flame 1994, 97, 339−354.(4) Wendt, C.; Eigenbrod, C.; Moriue, O.; Rath, H. J. A model fordevolatilization and ignition of an axisymmetric coal particle. Proc.Combust. Inst. 2002, 29, 449−457.(5) Zhang, D. K.; Wall, T. F. Ignition of coal particles: The influenceof experimental technique. Fuel 1994, 73, 1114−1119.

Figure 7. Devolatilization and burnout times of a cylindrical pine wood particle (diameter of 1.65 mm, length of 6.60 mm) as a function of oxygenconcentration: (a) T∞ = 1473 K, (b) T∞ = 1673 K, (c) T∞ = 1873 K, in all of which the ambient air stream has a upward speed of 1.5 m/s.



mailto:[email protected]



(6) Faundez, J.; Arenillas, A.; Rubiera, F.; García, X.; Gordon, A. L.;Pis, J. J. Ignition behavior of different rank coals in an entrained flowreactor. Fuel 2005, 84, 2172−2177.(7) Shaddix, C. R.; Molina, A. Particle imaging of ignition anddevolatilization of pulverized coal during oxy-fuel combustion. Proc.Combust. Inst. 2009, 32, 2091−2098.(8) Zhu, M.; Zhang, H.; Zhang, Z.; Zhang, D. A numerical modelingstudy of ignition of single coal particles under microgravity conditions.Combust. Sci. Technol. 2011, 183, 1221−1235.(9) Khatami, R.; Stivers, C.; Joshi, K.; Levendis, Y. A.; Sarofim, A. F.Combustion behavior of single particles from three different coal ranksand from sugar cane bagasse in O2/N2 and O2/CO2 atmospheres.Combust. Flame 2012, 159, 1253−1271.(10) Grotkjær, T.; Dam-Johansen, K.; Jensen, A. D.; Glarborg, P. Anexperimental study of biomass ignition. Fuel 2003, 82, 825−833.(11) Kuo, J. T.; Hsi, C. L. Pyrolysis and ignition of single woodenspheres heated in high-temperature streams of air. Combust. Flame2005, 142, 401−412.(12) Momeni, M.; Yin, C.; Kær, S. K.; Hansen, T. B.; Jensen, P. A.;Glarborg, P. Experimental study on effects of particle shape andoperating conditions on combustion characteristics of single biomassparticles. Energy Fuels 2013, 27, 507−514.(13) Lu, H.; Robert, W.; Peirce, G.; Ripa, B.; Baxter, L. L.Comprehensive study of biomass particle combustion. Energy Fuels2008, 22, 2826−2839.(14) Yang, Y. B.; Sharifi, V. N.; Swithenbank, J.; Ma, L.; Darvell, L. I.;Jones, J. M.; Pourkashanian, M.; Williams, A. Combustion of a singleparticle of biomass. Energy Fuels 2008, 22, 306−316.(15) Haseli, Y.; van Oijen, J. A.; de Goey, L. P. H. A detailed one-dimensional model of combustion of a woody biomass particle.Bioresour. Technol. 2011, 102, 9772−9782.(16) Abramzon, B.; Sirignano, W. A. Droplet vaporization model forspray combustion calculations. Int. J. Heat Mass Transfer 1989, 32,1605−1618.(17) Sazhin, S. S. Advanced models of fuel droplet heating andevaporation. Prog. Energy Combust. Sci. 2006, 32, 162−214.(18) Thunman, H.; Niklasson, F.; Johnsson, F.; Leckner, B.Composition of volatile gases and thermochemical properties ofwood for modeling of fixed or fluidized beds. Energy Fuels 2001, 15,1488−1497.(19) Di Blasi, C.; Branca, C. Kinetics of primary product formationfrom wood pyrolysis. Ind. Eng. Chem. Res. 2001, 40, 5547−5556.(20) Chan, W. C. R.; Kelbon, M.; Krieger, B. B. Modeling andexperimental verification of physical and chemical processes duringpyrolysis of a large biomass particle. Fuel 1985, 64, 1505−1513.(21) Forest Products Laboratory United States Department ofAgriculture Forest Service. Physical properties and moisture relationsof wood. In Wood Handbook: Wood as an Engineering Material; ForestProducts Society: Madison, WI, 1999; Chapter 3, pp 3−5.(22) Yin, C. Microwave-assisted pyrolysis of biomass for liquidbiofuels production. Bioresour. Technol. 2012, 120, 273−284.(23) Perry, R. H.; Green, D. W. Perry’s Chemical Engineers’ Handbook,7th ed.; McGraw-Hill: New York, 1997.(24) Patankar, S. V. Numerical Heat Transfer and Fluid Flow.Hemisphere Publishing Corporation: New York, 1980.(25) Yin, C.; Kær, S. K.; Rosendahl, L.; Hvid, S. L. Co-firing strawwith coal in a swirl-stabilized dual-feed burner: Modeling andexperimental validation. Bioresour. Technol. 2010, 101, 4169−4178.(26) Dall’Ora, M.; Jensen, P. A.; Jensen, A. D. Suspensioncombustion of wood: Influence of pyrolysis conditions on char yield,morphology, and reactivity. Energy Fuels 2008, 22, 2955−2962.(27) Chen, Y.; Mori, S.; Pan, W. Studying the mechanisms of ignitionof coal particles by TG-DTA. Thermochim. Acta 1996, 275, 149−158.(28) Howard, J. B.; Essenhigh, R. H. Mechanism of solid-particalcombustion with simultaneous gas-phase volatiles combustion. Symp.(Int.) Combust., [Proc.] 1967, 11, 399−408.(29) Di Blasi, C.; Branca, C.; Santoro, A.; Hernandez, E. G. Pyrolyticbehavior and products of some wood varieties. Combust. Flame 2001,124, 165−177.

(30) Grønli, M. G.; Varhegyi, G.; Di Blasi, C. Thermogravimetricanalysis and devolatilization kinetics of wood. Ind. Eng. Chem. Res.2002, 41, 4201−4208.



Documents

Comprehensive study of ignition and combustion of single wooden particles