Pulverized coal stream ignition delay under conventional and oxy-fuel combustion conditions

Page 1

Pc

YC

a

ARRAA

KIPCGO

1

hsemtitd

L

U

U

1d

International Journal of Greenhouse Gas Control 5S (2011) S36–S46

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

journa l homepage: www.e lsev ier .com/ locate / i jggc

ulverized coal stream ignition delay under conventional and oxy-fuelombustion conditions

inhe Liu1, Manfred Geier, Alejandro Molina2, Christopher R. Shaddix ∗

ombustion Research Facility, Sandia National Laboratories, Livermore, CA 94550, USA

r t i c l e i n f o

rticle history:eceived 1 February 2011eceived in revised form 13 May 2011ccepted 14 May 2011vailable online 12 June 2011

eywords:gnitionulverized coaloal combustionroup combustionxy-fuel combustion

a b s t r a c t

The coal stream ignition process is critical to the performance of modern pulverized coal burners, par-ticularly when operating under novel conditions such as experienced in oxy-fuel combustion. However,experimental studies of coal stream ignition are lacking, and recent modeling efforts have had to relyon comparisons with a single set of experiments in vitiated air. To begin to address this shortfall, wehave conducted experiments on the ignition properties of two U.S. and two Chinese coals in a laminarentrained flow reactor. Most of the measurements focused on varying the coal feed rate for furnace tem-peratures of 1230–1320 K and for 12–20 vol.% O2 in nitrogen. The influence of coal feed rate on ignitionwith a carbon dioxide diluent was also measured for 20 vol.% O2 at 1280 K. A second set of measurementswas performed for ignition of a fixed coal feed rate in N2 and CO2 environments at identical furnacetemperatures of 1200 K, 1340 K, and 1670 K. A scientific CCD camera equipped with a 431 nm imagingfilter was used to interrogate the ignition process. Under most conditions, the ignition delay decreasedwith increasing coal feed rate until a minimum was reached at a feed rate corresponding to a particlenumber density of approximately 4 × 109 m−3 in the coal feed pipe. This ignition minimum corresponds

to a cold flow group number, G, of ∼0.3. At higher coal feed rates the ignition delay increased. The igni-tion delay time was shown to be very sensitive to (a) the temperature of the hot coflow into which thecoal stream is introduced, and (b) the coal particle size. The three high volatile bituminous coals showednearly identical ignition delay as a function of coal feed rate, whereas the subbituminous coal showedslightly greater apparent ignition delay. Bath gas CO2 content was found to have a minor impact onignition delay.

© 2011 Elsevier Ltd. All rights reserved.

. Introduction

The topic of coal particle ignition, including considerations ofomogeneous (gas-phase, volatile) versus heterogeneous (particle-urface) ignition, has been reviewed by Essenhigh et al. (1989), Wallt al. (1991), and Annamalai et al. (1994). These reviews show thatost studies of coal particle ignition have focused on the minimum

emperature for ignition of a cloud of particles, irrespective of res-dence time. These studies have generated relatively low ignitionemperatures (associated with a long thermal ‘soak’ time) that areirectly relevant to evaluations of fire safety and coal mine explo-

∗ Corresponding author at: Sandia National Labs, MS 9052, 7011 East Avenue,ivermore, CA 94550, USA. Fax: +1 925 294 2276.

E-mail address: [email protected] (C.R. Shaddix).1 Permanent Address: School of Energy and Power Engineering, Xi’an Jiaotongniversity, Xi’an, China.2 Permanent Address: Bioprocesses and Reactive Flows, Faculty of Mines, Nationalniversity of Colombia, Medellín, Car. 80 No. 65-223, Medellín, Colombia.

750-5836/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.oi:10.1016/j.ijggc.2011.05.028

sions, but have little relevance to the problem of flame holding inpulverized coal (pc) burners, wherein high velocity streams of coolcoal particles turbulently mix with hot flame products and mustignite in tens of milliseconds. For application to pc burners then,the key issue is the characteristic ignition delay of coal particlesintroduced into hot surroundings.

To-date, most of the information on coal ignition delay has beengleaned from experiments and modeling of individual reacting coalparticles, in most cases using particles substantially larger thanthose characteristic of pulverized coal. Although single-particlestudies are undoubtedly useful for the study of ignition andcombustion of dilute particle streams, particle group effects arelikely important to the flame-holding process for practical burn-ers. Therefore, an understanding of the ignition characteristics ofa continuous flow of pulverized coal particles at different parti-cle mass loadings is needed to address actual industrial practice.

An improved understanding of particle stream ignition is impor-tant because it influences many aspects of pc burner performance,including NOx production, char burnout, flame stability, flameshape, and flame length.

Page 2

reenh

viIta2fltcolPTtovdwfll

haoeSspwpitwtm2ldaaaOfeCjpOdd

atitdiditobffd

Y. Liu et al. / International Journal of G

Despite the recognized importance of particle loading to the pul-erized coal ignition process, few studies have been conducted ongnition with a systematic, controlled variation in particle loading.n fact, whereas there have been several modeling efforts devotedo evaluating the influence of particle loading on coal ignition (Ryannd Annamalai, 1991; Du et al., 1995; Wang et al., 2007; Zhao et al.,007), only a single experimental study with continuous particleow has been reported (Ruiz et al., 1990). In that study, the igni-ion delay of a column of high-volatile, size-classified bituminousoal particles injected into a laminar furnace flow was determined,ver the range of 1023–1150 K, based on photographic pictures ofuminous emission from the burning particles. The coal used wasee Wee, an eastern U.S. hvA coal with 2.5% moisture and 7.8% ash.he O2 content of the vitiated air in the furnace was held constanthroughout these experiments, at 9 vol.%, and a significant amountf cool air coflow was supplied with the coal particles (to matchelocity profiles at the entrance of the furnace). Under these con-itions, it was determined that the ignition time first decreasedith increasing particle loading, reached a minimum (at a fuel massow rate of ∼3–6 g/min), and then increased with further particle

oading.In oxy-fuel combustion of pulverized coal, poor ignition quality

as often been noted during pilot-scale burner trials when oper-ting with substantial flue gas recirculation or with a syntheticxidant with CO2 diluent (Wang et al., 1988; Kiga et al., 1997; Liut al., 2005; Khare et al., 2008). Molina and Shaddix (2007) andhaddix and Molina (2009) measured the ignition delay of isolated,ize-classified subbituminous and high-volatile bituminous coalarticles when introduced into a laminar flow furnace. Experimentsere conducted with different oxygen concentrations and gas tem-eratures and filtered imaging was used to determine the point of

gnition. The ignition delay was shown to be quite sensitive to theemperature and oxygen content of the bulk flow and to increaseith the use of a CO2 diluent. Single-particle imaging demonstrated

he formation of high-temperature soot clouds around the bitu-inous coal particles during devolatilization (Shaddix and Molina,

009). Jovanovic et al. (2011) measured the visible ignition standoffength of a Russian high-volatile bituminous coal when conductingrop tube experiments at furnace temperatures of 1073–1623 Knd with oxygen concentrations from 10 to 100% in both nitrogennd CO2 diluents. The standoff length initially dropped consider-bly with increasing oxygen content, but for concentrations of 50%2 or higher the decrease was fairly minor. The CO2 diluent was

ound to delay ignition, particularly for environments with mod-st levels of oxygen. Zhang et al. (2011) studied the influence ofO2 versus N2 diluent gas in the primary “air” stream of a coal

et burner operating in an oxy-fuel furnace with an overall sup-lied oxygen level of 40 vol.%. Over a wide range of primary stream2 concentrations, use of CO2 in the primary stream was found toecrease flame stability and increase the average flame stand-offistance.

To improve upon the existing datasets on coal stream ignitionnd also provide data of specific relevance to oxy-fuel combus-ion, this paper reports on experiments using several commerciallymportant U.S. and Chinese coals, over a range of oxygen con-ents, in both N2 and CO2 diluents. Furthermore, more refinediagnostic methods are implemented to determine the character-

stic ignition point than were used in the previous data set forifferent coal loadings (Ruiz et al., 1990). These new data give

nsight into the optimal particle loading for ignition and also onhe influence of coal type, oxygen concentration, and CO2 diluentn the ignition characteristics. These data should provide useful

enchmark tests for computational models used to predict oxy-uel combustion flame characteristics in pc boilers and provideundamental understanding of coal ignition phenomena for burneresigners.

ouse Gas Control 5S (2011) S36–S46 S37

2. Experimental

2.1. Gas composition and temperature

The experiments were carried out in the optical, laminarentrained flow reactor at Sandia National Laboratories. A detaileddescription of the experimental setup for the flow reactor has beenpreviously provided (Tichenor et al., 1984; Molina and Shaddix,2007). The reactor operates at one atmosphere and uses a diffusion-flamelet-based Hencken burner to provide a high temperature,high-velocity gas flow. Coal particles are injected at the bottomof the furnace centerline through a 0.75 mm stainless-steel tubeand flow upwards with the furnace gases. A 5 cm × 5 cm squarequartz chimney isolates the reacting particles and burner productsfrom the surrounding air and allows optical measurements to beperformed on the particles injected into the flow.

The effect of particle loading on coal ignition was evaluatedby varying the feeding rate of coal particles into burner productmixtures with N2 as balance gas at three different oxygen concen-trations (12 vol.%, 16 vol.% and 20 vol.%), while holding the watervapor concentration and CO2 concentration constant at 11.6 vol.%and 0.3 vol.%, respectively. The total flow of burner products was80 standard (298 K and 101,325 Pa) liters per minute (slpm) foreach condition, yielding a flow velocity along the central portionof the furnace of approximately 2.5 m/s. It should be noted thatthe use of moderate concentrations of O2 in this study is inten-tional, because ignition of coal particle streams occurs where thecoal stream mixes with hot, oxidative gases characteristic of a mix-ture of flame product gases and air (or a mixture of flame productsand oxygen diluted with recycled flue gas, in the case of oxy-fuelcombustion). Assuming a representative flame product tempera-ture of 2000 K (associated with near-zero excess oxygen) and airpreheat to 400 K, a gas mixture at 1100 K (about the minimum tem-perature required for rapid coal stream ignition) is associated withan oxygen concentration of 12%. To generate a similar temperaturegas mixture with 20% O2, flame products would need to mix withan oxidizer source consisting of approximately 40% O2.

For the experiments reported here, the coal feed rate was var-ied from a very low rate (0.005 g/min) to a high rate (1.0 g/min),which was found to be sufficient to cover the transition fromsingle-particle to group ignition modes. Steady coal feed at thesespecified rates was provided by a coal feeding system, similar tothat described by Quann et al. (1982), employing a test tube, electricdrive motor, and vibrator. The coal mass feed rate was calibratedfor each coal type and size fraction used. The maximum coal feedrate corresponded to a minimum overall stoichiometric ratio of 5in these experiments (for the lowest concentration of O2 in theoxidizer), such that the furnace flow itself was always fuel-lean.

The coal particles were entrained by a very low flow of0.033 slpm diluent gas (either N2 or CO2), to minimize the ther-mal shielding effect of the cold entraining gas once the particlesare injected in the furnace. We believe this approach gives aclose approximation to the rapid mixing and heating that injectedcoal streams experience in pulverized coal burners. In addition,by minimizing the quantity of cold coflow gas, we minimize theimpact of the difference in thermal diffusivities between N2 andCO2(˛CO2∼0.6 ˛N2 ) on the heating rate of the injected coal particles.

For the comparison of particle ignition delay for different condi-tions it is important to match the gas temperature profiles near thebottom of the flow reactor (where the particles undergo initial heat-ing and devolatilization). Fig. 1 shows the gas temperature profilesfor the experiments involving coal feed rate variations, measured

along the reactor centerline with a 25-�m, type-R fine-wire ther-mocouple, corrected for radiative loss (Shaddix, 1999). The profilesfor the three different oxygen concentrations agree very well andare nearly height-independent between 40 mm and 125 mm above

Page 3

S38 Y. Liu et al. / International Journal of Greenh

Fig. 1. Measured gas temperature profiles along the furnace centerline for studiesof ignition as a function of coal feed rate. Open symbols denote 1320 K N2 diluentconditions: circles = 12 vol.% O2, diamonds = 16 vol.% O2, squares = 20 vol.% O2. Filleds1w

ttiattw(t

ttlafa

2

vf

FoNtm

quares denote 1280 K CO2 diluent condition (20 vol.% O2) and filled triangles denote230 K N2 diluent conditions (12 vol.% O2). The range of heights over which ignitionas observed in this study is indicated with the hatching.

he burner face (where ignition is seen to occur). Thus gas tempera-ures for ignition are essentially the same (1320 ± 5 K) for the threenvestigated oxygen levels. A limited set of data was also collectedt a lower temperature furnace condition (1230 K) to determinehe influence of gas temperature on the ignition behavior as a func-ion of particle feed rate. As indicated in Fig. 1, some experimentsere conducted with CO2 diluent at an intermediate temperature

1280 K) between the two temperatures used for investigating igni-ion in the N2 diluent.

A second set of experiments was conducted in which identicalemperature profiles for N2 and CO2 diluent gases were generated,o clearly delineate the effect of using a CO2 diluent over a range ofocal gas temperatures. This set of experiments was conducted forfixed (intermediate) coal feeding rate of 0.1 g/min into 20% O2 at

urnace temperatures of 1200 K, 1340 K, and 1670 K, respectively,s shown in Fig. 2.

.2. Coals

Two characteristic U.S. coals were investigated: Pittsburgh high-olatile bituminous coal and Black Thunder subbituminous coalrom the Powder River Basin. In addition, for investigations into the

ig. 2. Measured gas temperature profiles along the furnace centerline for studiesf the influence of CO2 diluent gas on coal stream ignition. Open symbols denote2 diluent and closed symbols denote CO2 diluent. 20 vol.% O2 was used for all

emperatures in this series of experiments: circles = 1200 K, squares = 1340 K, dia-onds = 1670 K.

ouse Gas Control 5S (2011) S36–S46

effect of coal particle loading, two typical high-volatile bituminousChinese coals were also investigated: Shenmu coal from north-ern Shaanxi Province in northwest China and Guizhou coal fromsouthwest China. Proximate and ultimate analyses of the inves-tigated coals are given in Table 1. The U.S. subbituminous coal isa low sulfur coal with 10% moisture, 5% ash, and 40% volatiles,whereas the Pittsburgh coal is a medium sulfur coal with 1% mois-ture, 7% ash, and 35% volatiles. Both Chinese coals contain 6%moisture, but differ dramatically in ash content, with the Shenmucoal having 9% ash and the Guizhou coal containing 32% ash. Pre-vious high-temperature single-particle ignition experiments withthe two U.S. coals demonstrated that the Pittsburgh coal particlesproduce large quantities of soot within the envelope flame aroundisolated, devolatilizing particles, whereas the Black Thunder coalhas a low sooting tendency (Shaddix and Molina, 2009). Note thatthe moisture content of all of the investigated coal samples is rel-atively low. Moisture is known to have a significant impact onparticle ignition delay on account of its endothermic vaporizationduring initial particle heat-up.

Coal particle size is known to have a strong influence on theignition process (Essenhigh et al., 1989). Therefore, in this studyspecific size cuts from commercially ground pulverized coal weregenerated by a commercial sieve shaker and set aside for subse-quent use. Most of the measurements were performed on particlesin the 75–105 �m size cut. Selected experiments were performedon the 54–74 �m and the 106–125 �m size fractions. The coal wastwice resieved, to assure that coal fines, which tend to adhere ontolarger particles, were removed from the samples.

2.3. Optical measurements

A commercial 8.1 megapixel digital camera was used to collectphotographic images of the broadband visible emission associ-ated with the particle stream ignition and combustion processes.This simple diagnostic corresponds to that used in previouslyreported measurements of particle stream ignition (Ruiz et al.,1990). Whereas this type of camera system is not optimal foranalyzing the ignition delay, it is useful for providing qualitativeinformation regarding particle stream ignition and burnout behav-ior.

A progressive-scan monochrome CCD camera (Roper ScientificCOOLSNAP fx, 1300 × 1030 pixels) was used to record images ofvisible light emission from ignited particles in the optical furnace.In an attempt to discriminate between gaseous volatile ignition (i.e.a definitive indication of homogeneous ignition) and ignition evi-denced by thermal emission from hot soot and/or char particles(which could result from either homogeneous or heterogeneousignition), a 431 nm bandpass filter was used in front of the cam-era to capture the chemiluminescent emission from electronicallyexcited CH radicals (i.e. CH*). CH* emission has been widely used asan indicator of the high-temperature flame front in hydrocarbon-fueled flames (Schefer, 1997). However, the commercially availablefilter at 431 nm has a 10-nm wide bandwidth, which allows someblackbody radiation to pass through. Furthermore, CO2* chemilu-minescent emission, generated from the oxidation of CO (whichitself is produced from oxidation of both volatiles and char par-ticles) extends over a broad spectral range and overlaps the CH*emission line, complicating the interpretation of the 431 nm sig-nals (Lauer and Sattelmayer, 2010). Comparison of the results fromusing this 431 nm optical filter and a spectrally adjacent bandpassfilter centered at 442 nm (which should not pass any CH* emis-sion) showed no noticeable difference in apparent ignition delay

with varying particle feed rate, suggesting that either CH* emis-sion is a minor component of the optical signals measured throughthe 431 nm filter, or that, for the investigated conditions, volatileignition occurs more or less synchronous with coal/char ignition. In

Page 4

Y. Liu et al. / International Journal of Greenhouse Gas Control 5S (2011) S36–S46 S39

Table 1Proximate and ultimate analysis of coals.

Proximate Coal type

Pittsburgh Black Thunder Shenmu Guizhou

wt.%, as rec’d wt.% dry wt.%, as rec’d wt.% dry wt.%, as rec’d wt.% dry wt.%, as rec’d wt.% dry

Moisture 1.4 10.8 5.7 5.7Ash 6.9 7.0 5.0 5.6 8.7 9.2 31.8 33.8Volatiles 35.4 35.9 40.4 45.3 35.1 37.2 22.8 24.1Fixed C 56.3 57.1 43.8 49.1 50.5 53.5 39.7 42.1

Ultimate Coal type

Pittsburgh Black Thunder Shenmu Guizhou

wt.% dry wt.% daf wt.% dry wt.% daf wt.% dry wt.% daf wt.% dry wt.% daf

C 77.2 82.9 60.9 64.1 78.8 86.8 55.6 84.0H 5.2 5.6 5.2 5.5 4.7 5.2 3.5 5.3Oa 7.2 7.7 27.6 29.1 4.8 5.7 5.0 7.6N 1.5 1.6 0.9 0.9 1.2 1.4 0.9 1.6

0.5

lflfiwna

viettoteitc

sptap2Ffbowddm

3

3

Brneet

S 2.0 2.2 0.4

a By difference.

ieu of a more definitive measurement of CH (e.g. by laser-induceduorescence), the CCD camera images with the 431 nm bandpasslter were used as the basis for determining coal stream ignition,hile recognizing that this filter likely measures a mixture of sig-als from CH* and CO2* as well as thermal radiation from hot sootnd char particles.

To average out any irregularities resulting from instantaneousariation of coal feeding rate, the camera shutter was set to collectmage data over a prolonged period of time, and multiple cam-ra exposures were summed together. The maximum permissibleime duration for camera exposure while avoiding pixel satura-ion varied significantly according to the particle feed rate and thexygen content of the furnace flow. Typically, at the lowest par-icle feeding rates, the camera shutter was set for three-minutexposures, and 30 separate images were collected for each exper-mental condition. At higher particle feeding rates, the exposureime was reduced to as low as 10 s, and 90 separate images wereollected.

For the investigation of the influence of CO2 diluent on coaltream ignition, a slightly different approach to determining theoint of ignition was adopted. For convenience, a coded aper-ure imaging system that collects light from a small probe volumelong the centerline of the reactor and focuses the light ontohotomultiplier tubes (pmts) was used (Murphy and Shaddix,006). The results from use of a 700 nm bandpass filter (40 nmWHM) are shown here, although similar results were foundrom use of a 550 nm filter. Radiant emission at 700 nm is clearlyroadband thermal emission (Planckian in character) and eitherriginates from hot soot or hot char particles. The pmt signalsere averaged over 10-second intervals for a given operating con-ition at a given reactor height. Movement of the reactor up orown allowed vertical profiles of average light emission to beeasured.

. Results and analysis

.1. Digital camera images

Fig. 3 shows photographs for the combustion of 75–105 �mlack Thunder coal particles in 12 vol.% O2 for different coal feedates and two different gas temperatures. Several trends can be

oted from these photographs. First, there is a marked differ-nce in the apparent ignition delay as a function of feed rate,specially for the lower gas temperature. (In this laminar flow reac-or, flame standoff height can be directly correlated with ignition

0.8 1.0 1.1 1.5

delay.) In particular, in agreement with the earlier findings fromRuiz et al. (1990), the ignition delay first decreases with increas-ing coal stream loading, reaches a minimum, and then increases.Interestingly, in the current experiment, the minimum ignitiondelay is reached for a feed rate of 0.05–0.10 g/min, in contrast tothe much higher mass feed rate required to reach the minimumignition delay in Ruiz’s experiments (3–6 g/min). The differencepresumably results from the much smaller diameter coal feed tubeand perhaps from the lower feed gas flow used in the presentexperiments.

Fig. 3 also indicates that ignition occurs much more rapidly forthe Black Thunder coal in a 1320 K thermal environment than fora 1230 K thermal environment. With the faster ignition at 1320 K,the influence of the coal stream particle density on ignition delayis significantly weaker. Otherwise, the same trends are present at1320 K as exist at 1230 K.

Fig. 4 shows images of 75–105 �m Pittsburgh coal particle igni-tion and combustion for the same conditions as shown in Fig. 3 forBlack Thunder coal. Similar overall trends are present, except thatfor the Pittsburgh coal there is only a minor decrease in the appar-ent ignition delay until a minimum is reached, again for a coal feedrate of 0.05–0.10 g/min. In contrast to the images of Black Thun-der combustion, for the Pittsburgh coal the visible light emission isclearly brightest just after ignition. This difference may presumablybe ascribed to the strong tendency of Pittsburgh coal to producesoot (which radiates at high temperatures) during devolatilization(Shaddix and Molina, 2009).

A comparison of the photographs of ignition and combus-tion of 75–105 �m Shenmu and Guizhou coal streams at 1320 K(not shown here) with those of Black Thunder and Pittsburghcoals shows more similarity to the Pittsburgh coal photographsthan those of Black Thunder, as one would expect based onthe similarity in rank of the Shenmu and Guizhou coals tothe Pittsburgh coal, as shown in Table 1. However, neither theShenmu nor Guizhou coals exhibits nearly as much bright radi-ation from coal volatiles (i.e. soot production) as Pittsburghcoal.

Fig. 5 shows photographs of ignition and combustion of differ-ent size fractions of Black Thunder coal in 12 vol.% O2 at 1320 K. Forlow fuel feed rates, smaller particles appear to ignite faster thanlarger particles. At higher fuel feed rates, however, this trend is

not nearly so clear and even shows signs of reversing. Faster par-ticle burnout is clearly favored by small particles. Interestingly,for larger particles, a distinct physical separation of thermal emis-sion from devolatilizing particles (i.e. soot radiation) and from

Page 5

S40 Y. Liu et al. / International Journal of Greenhouse Gas Control 5S (2011) S36–S46

Fig. 3. Photographs of 75–105 �m Black Thunder coal streams igniting and burning in Sandia’s laminar entrained flow reactor when flowing 12 vol.% O2 at two characteristictemperatures, as shown. Photographs were obtained with a common exposure time of 1/20 s and fixed aperture. Coal particles are injected along the centerline of the reactorand flow upwards. Blue emission from the flat flame at the base of the furnace is evident in some photos. Coal stream mass flow rate increases from left to right across eachi 30, 0.0

crolp

3

teoPe

mage sequence. The coal feed rates for images (a)–(i) are equal to 0.005, 0.015, 0.0

har combustion is evident, at least for low to moderate fuel feedates. This phenomenon may reflect the transient depletion ofxygen from volatiles combustion around individual particles atower fuel feed rates – an effect that will be favored by larger fuelarticles.

.2. CCD camera images

In comparison to the unfiltered, full-height digital camera pho-ographs, the 431 nm CCD camera images allow a more rigorous

valuation of the coal stream ignition process. Fig. 6 shows a seriesf average CCD camera images as a function of coal feed rate forittsburgh coal particles igniting in 12 vol.% O2 in N2 at 1320 K. Ifxamined closely, the images reveal a distinct progression in igni-

Fig. 4. Photographs of 75–105 �m Pittsburgh coal streams ign

50, 0.10, 0.20, 0.40, 0.67, and 1.00 g/min, respectively.

tion behavior as the coal particle stream increases in density. At thelowest particle densities, the particles are too far apart to ignite inconcert with one another and a statistical distribution of individ-ual particle ignition heights results over time. As the coal particlestream increases in particle density, particles tend to devolatilizeand ignite together, producing a sharp rise in the detected sig-nals. As the coal particle stream density continues to increase, theinnermost regions of the coal stream are thermally and chemicallyshielded from the hot, O2-containing ambient gas, so the first igni-tion occurs in an annular ring surrounding the center of the coal

stream and then moves inward as the outer coal volatiles are con-sumed. The overall ignition process is thus delayed relative to thatat more moderate particle loadings and also occurs over a muchlonger time period.

iting and burning, for the same conditions as for Fig. 3.

Page 6

Y. Liu et al. / International Journal of Greenhouse Gas Control 5S (2011) S36–S46 S41

ng and

3

3

mbcpflWttattoacd

tFs

Ftadr

Fig. 5. Photographs of the indicated size cuts of Black Thunder coal igniti

.3. Ignition time delay analysis

.3.1. MethodologyThe characteristic ignition delay was determined from the accu-

ulated CCD images for a given condition by extracting the radiallyinned vertical profile along the reactor, as shown in Fig. 7. Verti-al heights were translated into flow residence times by assuminglug flow in the reactor and dividing the volumetric flow rate ofat flame product gases by the cross-sectional area of the furnace.hile it is known that the flow in the furnace develops a Poiseuille-

ype pipe-flow profile with increasing height in the furnace (leadingo increases in the centerline gas velocity even as the gas temper-ture slowly decreases), the approximation used here to computehe flow residence time should be reasonably accurate near the cen-er of the furnace over the low furnace heights at which ignitionccurs. Gas temperature and flow velocity modifications associ-ted with heat absorption and volatiles ejection in the high-densityoal flows are also ignored in this analysis, in part because of theifficulty of accurately determining the magnitude of these effects.

There are many possible ways of deducing a characteristic igni-ion delay from the vertical CCD image profiles such as shown inig. 7. In this study we chose to use the criterion of the signal inten-ity reaching 50% of the peak intensity for a given experimental

ig. 6. False color mean CCD camera array readouts for 54–74 �m Pittsburgh coal streao the furnace axis, with coal particles entering from the left side of the array image. Wre displayed in the plots above and to the right of the CCD images (i.e. the top plot shoistribution at the height with the greatest mean signal). Coal stream mass flow rate varight image.

burning in 12 vol.% O2 at 1320 K for the same coal feed rates as in Fig. 3.

condition, as we reasoned that this approximately represented thelocation at which half of the injected particles had ignited. Anotherprospective criterion would be the location of the maximum ups-lope of the signal intensity profile. This latter criterion was foundto give the same trends as the “50% max signal” criterion, but withthe ignition heights moved slightly lower.

Fig. 8 presents the ignition delays determined for the Pittsburghand Black Thunder coals at the two different temperatures, usingthe “50% max signal” criterion. For Black Thunder coal the mini-mum ignition delay is seen to be approximately 20 ms at 1320 Kand 30 ms at 1230 K, whereas for Pittsburgh coal the minimumignition delays are 15 ms and 25 ms, respectively. The Pittsburghcoal dataset has also been analyzed for ignition delay on the basisof the locations where 25% of the maximum signal is reached andwhere 75% of the maximum signal is obtained. The results of thisinterquartile analysis of uncertainty in the derived ignition delayare shown as error bars in Fig. 8. Clearly, under the investigatedconditions the determination of ignition delay is fairly insensi-tive to the methodology by which one defines the characteristic

point of ignition. Another aspect of uncertainty is associated withthe repeatability of the measurements and with the long temporalaveraging that has been used to compute mean profiles. To analyzethe magnitude of these effects, the individual CCD camera frames

ms igniting in 12 vol.% O2 at 1320 K. The horizontal axis of the array correspondshite crosshairs indicate the chosen array row and column whose intensity valuesws the axial distribution along the centerline and the right plot shows the radialies from 0.005 g/min in left image to 0.050 g/min in center image to 1.00 g/min in

Page 7

S42 Y. Liu et al. / International Journal of Greenhouse Gas Control 5S (2011) S36–S46

Fig. 7. Mean, radially binned CCD array profiles along the furnace axis for (a) Pittsburghand (c) Shenmu coal in 20 vol.% O2 in CO2 at 1280 K. All profiles are for 75–105 �m coal pam

Fa1ft

g/min.

ig. 8. Characteristic ignition delays measured for Black Thunder (filled symbols)nd Pittsburgh (open symbols) coal as a function of coal mass feed rate injected into2 vol.% O2 in N2 at 1230 K (squares) and 1320 K (triangles). The error bars shownor the Pittsburgh coal correspond to the results for interquartile determination ofhe ignition delay, as described in the main text.

coal in 12 vol.% O2 in N2 at 1320 K, (b) Shenmu coal in 20 vol.% O2 in N2 at 1320 K,rticles. The legends give the coal mass feed rate for the different curves, in units of

were processed for two representative datasets for ignition delayusing the identical procedure as described above and these resultswere used to calculate standard deviation values for the derivedignition delays. The results of this analysis are shown in Fig. 9, whichcompares the results for the two different measures of uncertainty.In general, the interquartile uncertainty exceeds the uncertaintyassociated with image repeatability, except for the low tempera-ture condition at high particle feed rates. Even in this case, however,the worst-case (2�) uncertainty in repeatability is only 2.5 ms.

3.3.2. Ignition delay as a function of particle feed rateFor Pittsburgh coal, the trends in ignition delay as a function of

fuel loading agree with the previously discussed trends suggestedby the digital camera photos (Fig. 4). For the Black Thunder coalat the low-temperature condition (1230 K), however, the digitalcamera photos had suggested a sharp decrease in ignition delayas the coal feed rate increased from a low level (Fig. 3), whereasthe filtered CCD image data do not show such a decrease. In fact,the filtered CCD image data yield the same result for this condi-tion when using either a 442 nm or 431 nm filter, so the difference

in the results from the two cameras cannot be attributed to CH*emission. It is possible that CO2* emission, from combustion of thecoal volatiles, generated substantial signal on the CCD images butwas inconsequential compared to char combustion thermal emis-

Page 8

Y. Liu et al. / International Journal of Greenhouse Gas Control 5S (2011) S36–S46 S43

Fig. 9. Comparison of ignition delay uncertainties determined for the Pittsburghcoal introduced into 12 vol.% O2 in N2 on the basis of interquartile analysis (opensTr

steip

saaoiC(ecaNwqpt

Fsw

Fig. 11. Ignition delays for different size cuts of Pittsburgh coal as a function of coal

ymbols) and repeatability analysis (filled symbols), as described in the main text.riangles denote data collected in the 1320 K furnace environment and squares giveesults for the 1230 K furnace environment.

ion in the unfiltered digital camera images. It is also possible thathe poor statistical sampling that is represented in the digital cam-ra images happened to suggest a consistent trend of decreasinggnition delay at low particle loadings that would not have beenresent for sufficiently long temporal averaging.

Ignition delay trends with a CO2 diluent gas are shown for theseame two coals, but at a higher oxygen level, in Fig. 10. Results arelso presented for N2 diluent, but at a somewhat greater temper-ture (1320 K) than for CO2 (1280 K). With the strong dependencef ignition delay on temperature that was demonstrated in Fig. 8,t is unclear from the results of Fig. 10 whether the presence ofO2 played any role in the delayed ignition in the CO2 experimentsbut further insight on this is provided in the same-temperaturexperimental results to be discussed later). For the Black Thunderoal, the ignition trend for low particle feed rates in CO2 showssharp break in comparison to the smooth trend for ignition in2 environments. It is possible that the emission from CO2* is tooeak under these conditions (because of the thermal and/or radical

uenching effects of CO2; Williams et al., 2008) to effectively com-ete against the thermal emission, such that the deduced ignitionime corresponds to that of the coal char and not its volatiles.

ig. 10. Ignition delays for Black Thunder (filled symbols) and Pittsburgh (openymbols) coal as a function of coal mass feed rate injected into 20 vol.% O2 at 1280 Kith CO2 diluent (inverted triangles) and at 1320 K with N2 diluent (circles).

mass feed rate (top) and coal particle number density (bottom) injected into 12 vol.%O2 in N2 at 1320 K.

The influence of coal particle size on ignition delay is shown inFig. 11, expressed both in terms of fuel mass loading and in termsof particle number density in the cold coal feed flow. The num-ber density calculation is based on the mean particle size withinthe sieve cut and on the assumption of a coal particle density of1200 kg/m3 for all particle size cuts. A fuel stream composed ofsmaller particles ignites more quickly, as one might expect on thebasis of characteristic particle heat-up times. This trend was alsosuggested by the data of Ruiz et al. (1990). The plot against fuel massloading shows that the fuel mass feed rate for minimum ignitiondelay varies with particle size (i.e. it varies from 0.03 g/min for thesmallest size cut investigated to 0.10 g/min for the largest size cutinvestigated). For low coal feed rates, there is a pronounced sensi-tivity of ignition delay to particle size, showing nearly a factor oftwo difference for a size fraction of 54–74 �m compared to a sizefraction of 106–125 �m. At high coal feed rates, the ignition delayincreases rapidly for all coal size fractions, leading to smaller differ-ences between the profiles for different size fractions. When plottedagainst particle number density, it appears that the number den-sity for minimum ignition delay may vary somewhat with particlesize, albeit more weakly than on the basis of fuel mass loading (theposition of minimum delay appears to vary from 3 × 109 m−3 forthe largest size fraction investigated to 5 × 109 m−3 for the small-

est size fraction investigated). Interestingly, the coal stream particlenumber density for minimum ignition delay lies towards the upperend of the range of typical particle number densities in pulverized

Page 9

S reenhouse Gas Control 5S (2011) S36–S46

c0dd

3

ophiua

G

(dprr

f

w

G

wWip

fmsttorrtrtnoacrtiZreldf

Pccicvic

3.3.4. Effect of oxygen concentration and coal typePerhaps surprisingly, little influence of oxygen concentration

was found for coal stream ignition at 1320 K. Fig. 13 shows the

44 Y. Liu et al. / International Journal of G

oal burner feed pipes (Annamalai et al., 1994, quotes a range of.5–5.0 × 109 m−3). Also, the number density for minimum ignitionelay determined here agrees well with the value of ∼4.5 × 109 m−3

etermined from the study of Ruiz et al. (1990; Du et al., 1995).

.3.3. Group number analysisThough the concept of group number was originally devel-

ped for application to quasi-steady liquid spray combustion androbably has limited applicability to solid particle ignition, itas frequently been invoked in discussions and analysis concern-

ng ignition of coal particle streams. For a cylindrical column ofniformly sized particles (as approximated in this study), the char-cteristic group number, G, can be expressed by

= 3�gR2c

a2�p(mg/mp)= 3

(Rc

a

)2fv (1)

Chiu and Liu, 1977; Annamalai et al., 1994), where �g is the gasensity, �p is the particle density, Rc is the column radius, a is thearticle radius, mg is the mass flow rate of gas, mp is the mass flowate of particles, and fv is the volume fraction of particles. Using theelationship of

v = 43

�a3n (2)

here n is the number density of particles, one can also show that

= 2�nR2c dp (3)

here dp is the particle diameter, as expressed in the studies ofang et al. (2007) and Zhao et al. (2007). Thus, for a given character-

stic particle column radius, the group number is proportional to theroduct of the particle number density and the particle diameter.

The group number can be readily calculated within the cold coaleed flow, but estimating it at the point of ignition is substantially

ore problematic, because of uncertainties regarding the expan-ion of the fuel jet within the reactor as a consequence of heating ofhe cold gas, fluid mechanical mixing, and pre-ignition devolatiliza-ion of the particles (which leads to particle dispersion on accountf volatile jetting action). Thermal expansion of the cold feed gaseduces the coal particle number density and therefore tends toeduce the group number. However, if the thermal expansion ofhe cold gas occurs radially for the injected particle stream, theeduction in number density is offset by the increasing radius ofhe particle column, yielding no net change in the calculated groupumber. On the other hand, if the thermal expansion of the cold gasccurs vertically along the particle column (i.e. becomes manifests a higher gas flow velocity rather than a wider column of parti-les), then the group number is reduced, in direct proportion to theatio of the absolute temperature at the point of ignition relative tohe feed flow temperature (i.e. by approx. a factor of 4). The groupgnition modeling studies of Du et al. (1995), Wang et al. (2007), andhao et al. (2007) all compared computed ignition delays with theesults of Ruiz et al. (1990) on the basis of group numbers. The mod-ling results were presented for group numbers calculated underocal conditions (at the point of ignition), and the group numberserived from the Ruiz et al. (1990) study were calculated under colduel feed pipe conditions.

Fig. 12 shows the ignition delays of the different size fractions ofittsburgh coal plotted against the group number for the injectedolumn of particles. The group number has been calculated for theold coal feed flow. For the three different particle sizes, the min-mum ignition delay occurs for a common value of G of ∼0.3. In

omparing Fig. 12 to Fig. 11, it is apparent that the use of G pro-ides a more consistent correlation parameter for the minimumgnition delay than the coal mass feed rate, though the use of theoal particle number density seems to also correlate quite well.

Fig. 12. Ignition delays for different size fractions of Pittsburgh coal in 12 vol.% O2

in N2 at 1320 K as a function of the characteristic group number, G.

In contrast to the relatively low critical value of G determinedin this study, simulations of particle cloud ignition have typicallyfound minimum ignition delay for G values of 10–200 (Ryan andAnnamalai, 1991; Du et al., 1995; Zhao et al., 2007; Wang et al.,2007). The experimental data from Ruiz et al. (1990) yielded a min-imum ignition delay for a G value of approximately 10. As notedpreviously, the particle number densities for minimum ignitiondelays for Ruiz et al. agreed with the current results, so the differ-ence in the critical G values is directly attributable to the differenceof feed pipe diameters used in the two studies and the R2

c factorin the calculation of G. It would be interesting to see if simulationsof ignition of the smaller dimension feed column used here wouldgive predictions of the critical value of G that agree with the currentdata or with the results for the larger diameter feed pipe (i.e. testingwhether the R2

c scaling for G is actually appropriate to describe coalstream ignition). The agreement seen here in the value of particlenumber density for minimum ignition delay for a range of particlesizes and also in agreement with the Ruiz et al. (1990) data sug-gests that the particle number density may be a more meaningfulproperty than the group number in correlating coal particle ignitionbehavior.

Fig. 13. Ignition delays for 75–105 �m Pittsburgh coal streams as a function of coalparticle number density for different oxygen levels in N2 at 1320 K.

Page 10

Y. Liu et al. / International Journal of Greenh

Fr

rOaw

ttrerTdttelcpnlspdd

Fsfa

ig. 14. Ignition delays for 75–105 �m coal streams as a function of coal mass feedate for the four different coals investigated for 12 vol.% O2 in N2 at 1320 K.

esults for 75–105 �m Pittsburgh coal particles as a function of2 (12–20 vol.%) concentration in N2, and clearly the trending asfunction of fuel loading is the same for different oxygen levels,ith only a minor influence on the absolute value of ignition delay.

The three high-volatile bituminous coals produced nearly iden-ical ignition delays at 1320 K, as shown in Fig. 14. It is surprisinghat Guizhou coal, with 32% ash and only 23% volatiles, ignites asapidly as the other two high-volatile bituminous coals. At the high-st coal feed rates investigated, however, Guizhou does show aapidly increasing ignition delay. The subbituminous coal (Blackhunder) has a somewhat longer apparent ignition delay. Thiselayed ignition may be a result of the higher moisture level inhis coal (11 wt.%, versus 6 wt.% or less for the other coals). In fact,he overall trending is consistent with the analyzed moisture lev-ls of the coals, with Pittsburgh coal having the lowest moistureevel (<2 wt.%) and the fastest ignition. However, one needs to alsoonsider that our attempts to directly measure indications of gas-hase (volatile) ignition (through CH* or CO2* emission) have notecessarily been very successful, such that the apparent ignition

ocation may be mostly dependent on thermal emission from hotoot (produced by bituminous coal volatiles) or hot (ignited) char

articles. For the subbituminous coal, which does not tend to pro-uce soot during devolatilization, this dearth of measurable signalsuring devolatilization could unduly delay the indication of igni-

ig. 15. . Ignition delays for 75–105 �m Pittsburgh coal and Black Thunder coaltreams fed at 0.10 g/min for 20 vol.% O2 with N2 or CO2 diluent at three differenturnace temperatures. Filled symbols are for Black Thunder coal and open symbolsre for Pittsburgh coal.

ouse Gas Control 5S (2011) S36–S46 S45

tion. Another factor that may affect the ignition delay is radiationtransfer upstream from the ignited particles, particularly when sub-stantial soot formation occurs, which would tend to decrease theignition delay. This factor may be responsible for the somewhatsmaller ignition delays for Pittsburgh coal, which the photographssuggest produces the most radiant soot.

3.3.5. Effect of CO2 diluentThe results of the experiments with Pittsburgh and Black

Thunder coals that explicitly compared ignition in N2 and CO2 envi-ronments at the same temperature are shown in Fig. 15. For ignitionat a low temperature (1200 K), ignition is substantially delayed,whereas it continues to occur sooner at higher temperatures. Forthe Pittsburgh coal, the use of CO2 diluent produces a small delay inignition relative to a N2 environment, more or less independent ofgas temperature. As discussed previously, the high volumetric heatcapacity of CO2 is likely the dominant reason for this additional igni-tion delay with a CO2 diluent, by virtue of inhibiting local thermalrunaway (Molina and Shaddix, 2007; Shaddix and Molina, 2009).However, under group ignition conditions, in which diffusionaltransport of oxygen into interior regions may be an important fac-tor for ignition, the reduced diffusivity of oxygen in CO2 may alsolead to delayed ignition. In general, the Black Thunder coal showsa slightly greater ignition delay than the Pittsburgh coal, as previ-ously seen during the study of particle loading effects. However,the low temperature N2 condition yielded the same ignition delayfor Black Thunder coal as for Pittsburgh coal.

4. Conclusions

The ignition properties of two characteristic U.S. coals and twoChinese coals have been investigated for a range of pulverizedcoal particle feed rates and for three particle size fractions over arange of oxygen concentrations and gas temperatures, with bothN2 and CO2 diluents. Filtered CCD camera imaging with exten-sive time-averaging was used to interrogate the ignition process.In agreement with the only previous study similar to this in theliterature, the ignition delay decreased with increasing coal feedrate until a minimum was reached, corresponding to a particlenumber density of approximately 4 × 109 m−3 in the coal feedpipe. At higher coal feed rates the ignition delay increased. Theignition delay showed strong sensitivity to the furnace gas tem-perature and to coal particle size. For most conditions, the differentcoals showed similar ignition trends, with the subbituminous coalexhibiting somewhat greater apparent ignition delay. The ignitiondelay increased slightly with the use of a CO2 diluent.

Acknowledgments

This research was sponsored by the U.S. Department of Energy(DOE) through the National Energy Technology Laboratory’s PowerSystems Advanced Research Program, managed by Dr. RobertRomanosky. Support for Y. Liu’s visit at Sandia was mainly providedby China Scholarship Council and partly by the Natural Science Fundof China (50806058). Sandia is a multi-program laboratory oper-ated by Sandia Corporation, a Lockheed Martin Company, for DOE’sNational Nuclear Security Administration under Contract DE-AC04-94AL85000.

References

Annamalai, K., Ryan, W., Dhanapalan, S., 1994. Interactive processes in gasificationand combustion – Part III: Coal/char particle arrays, streams and clouds. Prog.Energy Combust. Sci. 20, 487–618.

Chiu, H.H., Liu, T.M., 1977. Group combustion of liquid droplets. Combust. Sci. Tech-nol. 17, 127–142.

Page 11

S reenh

D

E

J

K

K

L

L

M

M

Q

R

46 Y. Liu et al. / International Journal of G

u, X., Gopalakrishnan, C., Annamalai, K., 1995. Ignition and combustion of coalparticle streams. Fuel 74, 487–494.

ssenhigh, R.H., Misra, M.K., Shaw, D.W., 1989. Ignition of coal particles: a review.Combust. Flame 77, 3–30.

ovanovic, R., Milewska, A., Swiatkowski, B., Goanta, A., Spliethoff, H., 2011.Numerical investigation of influence of homogeneous/heterogeneous igni-tion/combustion mechanisms on ignition point position during pulverized coalcombustion in oxygen enriched and recycled flue gases atmosphere. Int. J. HeatMass Transfer 54, 921–931.

hare, S.P., Wall, T.F., Farida, A.Z., Liu, Y., Moghtaderi, B., Gupta, R.P., 2008. Factorsinfluencing the ignition of flames from air-fired swirl pf burners retrofitted tooxy-fuel. Fuel 87, 1042–1049.

iga, T., Takano, S., Kimura, N., Omata, K., Okawa, M., Mori, T., Kato, M., 1997. Char-acteristics of pulverized-coal combustion in the system of oxygen/recycled fluegas combustion. Energy Convers. Manage. 38, 129–134.

auer, M., Sattelmayer, T., 2010. On the adequacy of chemiluminescence as a mea-sure for heat release in turbulent flames with mixture gradients. J. Eng. Gas Turb.Power 132, 061502.

iu, H., Zailani, R., Gibbs, B.M., 2005. Comparisons of pulverized coal combustion inair and in mixtures of O2/CO2. Fuel 84, 833–840.

olina, A., Shaddix, C.R., 2007. Ignition and devolatilization of pulverized bitu-minous coal particles during oxygen/carbon dioxide coal combustion. Proc.Combust. Inst. 31, 1905–1912.

urphy, J.J., Shaddix, C.R., 2006. Combustion kinetics of coal chars in oxygen-enriched environments. Combust. Flame 144, 710–729.

uann, R.J., Neville, M., Janghorbani, M., Mims, C.A., Sarofim, A.F., 1982. Mineral mat-ter and trace-element vaporization in a laboratory-pulverized coal combustionsystem. Environ. Sci. Technol. 16, 776–781.

uiz, M., Annamalai, K., Dahdah, T., 1990. An experimental study on group ignitionof coal particle streams. In: Grosshandler, W.L., Semerjian, H.G. (Eds.), ASME

ouse Gas Control 5S (2011) S36–S46

HTD-Vol. 148, Heat and Mass Transfer in Flames and Combustion Systems. , pp.16–26.

Ryan, W., Annamalai, K., 1991. Group ignition of a cloud of coal particles. J. HeatTransfer 113, 677–687.

Schefer, R.W., 1997. Flame sheet imaging using CH chemiluminescence. Combust.Sci. Technol. 126, 255–270.

Shaddix, C.R., 1999. Correcting thermocouple measurements for radiation loss: acritical review. In: Proc. of 33rd National Heat Transfer Conference, AmericanSociety of Mechanical Engineers , Albuquerque, NM, USA, HTD99-282.

Shaddix, C.R., Molina, A., 2009. Particle imaging of ignition and devolatilizationof pulverized coal during oxy-fuel combustion. Proc. Combust. Inst. 32, 2091–2098.

Tichenor, D.A., Niksa, S., Hencken, K.R., Mitchell, R.E., 1984. Simultaneous in situmeasurement of the size, temperature and velocity of particles in a combustionenvironment. Proc. Combust. Inst. 20, 1213–1221.

Wall, T.F., Gupta, R.P., Gururajan, V.S., Zhang, D.-K., 1991. The ignition of coal parti-cles. Fuel 70, 1011–1016.

Wang, C.S., Berry, G.F., Chang, K.C., Wolsky, A.M., 1988. Combustion of pulverizedcoal using waste carbon dioxide and oxygen. Combust. Flame 72, 301–310.

Wang, S., Lu, H., Zhao, Y., Mostofi, R., Kim, H.Y., Yin, L., 2007. Numerical study ofcoal particle cluster combustion under quiescent conditions. Chem. Eng. Sci. 62,4336–4347.

Williams, T.C., Shaddix, C.R., Schefer, R.W., 2008. Effect of syngas composition andCO2-diluted oxygen on performance of a premixed swirl-stabilized combustor.Combust. Sci. Technol. 180, 64–88.

Zhang, J., Kelly, K.E., Eddings, E.G., Wendt, J.O.L., 2011. Ignition in 40 kW co-axial turbulent diffusion oxy-coal jet flames. Proc. Combust. Inst. 33, 3375–3382.

Zhao, Y., Kim, H.Y., Yoon, S.S., 2007. Transient group combustion of the pulverizedcoal particles in spherical cloud. Fuel 86, 1102–1111.