In-Situ Characterization of Coal Particle Combustion via LongWorking Distance Digital In-Line HolographyXuecheng Wu,† Longchao Yao,†,‡ Yingchun Wu,*,† Xiaodan Lin,† Linghong Chen,† Jun Chen,*,‡

Xiang Gao,† and Kefa Cen†

†State Key Laboratory of Clean Energy Utilization, Zhejiang University, Hangzhou 310027, China‡School of Mechanical Engineering, Purdue University, West Lafayette, Indiana 47907, United States

ABSTRACT: A long working distance digital holographic system is developed to study burning coal particles, which are lessthan 200 μm in diameter, in a methane-air flame. A general model for the lensed digital holography system is introduced first.An optimal lateral resolution of 3.9 μm is achieved while the object is located at a safe distance (>15 cm) from the opticalapparatus. The evolutions of particle morphology and devolatilization products are observed. Sizes and three-dimensionalvelocities of particles are measured and compared at the burner outlet, 10 and 24 cm above the burner. At the ignition, acondensed phase cloud is observed surrounding the coal particle due to the ejection of devolatilization products. Then thevolatile cloud accumulates into stringy tails after detaching from the parent particle. The surface of the coal particle has roughmorphology due to formation of porous structure after the reaction. Statistical size distributions at different heights above theburner show a 35% increase of peak size after combustion, and the fraction of the larger particles also significantly increased.Particles spread in the radial direction; meanwhile, particle velocities along the axial direction decrease with the height becauseof the drag force and gravity.

■ INTRODUCTION

Coal is one of the major fossil fuels to produce energy due toits low cost and vast deposits. There have been many strategiesto utilize coal including pulverized coal combustion, gas-ification, liquefaction, coal water slurry, and integratedgasification combined cycle,1,2 among which pulverized coalcombustion with the coal particles less than 200 μm in size hasremained the dominant method for decades. Despite theadvantages of coal based energy production, the large amountof greenhouse gas release and air pollutant emissions arebecoming an increasingly severe threat to the environment.3,4

Thus, there is an urgent need to improve the energy efficiencyand to reduce the pollutant emissions in coal combustion. Ascoal is a complex mixture of organic and inorganic substances,the combustion process is more complicated than those ofliquid and gas fuels. Pulverized coal undergoes pyrolysis,ignition, homogeneous combustion of the volatile matter, andthe heterogeneous combustion of the char.5,6 Meanwhile, theair pollutants such as SO2 and NOx are generated in thechemical process, and soot and fly ash are ejected andaccumulate into particulate matters.7,8 Understanding thecomplex physical and chemical process around and in thecoal particle is important to reveal coal combustionmechanism, which will further help the development of newcombustion models and optimization of combustion con-ditions. Knowledge of the coal particle morphologies, sizes,velocities, and their evolutions is essential for understandingcoal combustion.A variety of architectures such as tube drop furnace,6,9,10 flat

flame burner,11−14 and coal-laden jet burner15−17 have beendeveloped to simulate the high temperature and high heatingrate of the combustor in coal-fired boilers. It is convenient toimplement optical/laser diagnostic techniques to characterize

the flow and particle features with these architectures. LaserDoppler velocimetry (LDV) and particle image velocimetry(PIV) are two standard tools for flow field velocity measure-ment, and their applications to flame velocity are wellproven.18,19 However, the scattered laser signal is frequentlycontaminated by larger coal particle’s incandescence, whichintroduces noises in the Mie scattering images and deterioratesthe measurement accuracy. To measure coal particle velocity,particle streak velocimetry (PSV)15,20 records a long exposureimage of bright burning chars as streaks. Particle displacementsare determined by examining the start and end of the streaks. Itis more suitable for the stage of heterogeneous char burningwhen the particles are incandescent. In many research studies,particle morphologies were observed with scanning electronicmicroscopy (SEM)15,21 before and after the combustion. Thishelped the study of the particle size evolution and porousstructure formation on the particle surface. However, SEM isunable to capture images for in situ coal particles in the flameand thus is not applicable to the dynamic measurement. Directobservation of the coal particle behavior is one of the mostimportant ways to investigate events and parameters of coalcombustion. For this purpose, shadowgraphy,14 high-speedphotography,6,11,22 and holography17,23,24 were used in the coalcombustion study history. Shadowgraphy uses bright whitebackground light so that the coal particles and condensedphase of the volatile matters are recognized as a dark region inthe bright background. High-speed photography takes photo-graphs or shadowgraphs at a high frame rate at 1000 Hz orhigher and has become more prevalent than ever before with

Received: May 13, 2018Revised: July 18, 2018Published: July 19, 2018

Article

pubs.acs.org/EFCite This: Energy Fuels 2018, 32, 8277−8286

© 2018 American Chemical Society 8277 DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

Dow

nloa

ded

via

PUR

DU

E U

NIV

on

Oct

ober

3, 2

018

at 1

9:07

:20

(UT

C).

Se

e ht

tps:

//pub

s.ac

s.or

g/sh

arin

ggui

delin

es f

or o

ptio

ns o

n ho

w to

legi

timat

ely

shar

e pu

blis

hed

artic

les.

the fast development of ultrafast cameras and data storagetechniques. A dynamic process like volatile flame evolution,coal particle fragmentation, and movement can be recorded byhigh-speed photography. The drawback of the traditionalimaging techniques is that only particles in focus are useful foranalysis, while the defocused particles are blurred, making theobservation and measurement confined in a limited depth offield. In addition, most optical/laser techniques resolve onlyone parameter at one time, and the measurement area isusually limited to point wise, or two-dimensional (2D).Combining different techniques in one experiment is a feasiblebut sophisticated and expensive solution.15

Differently, holography25 uses coherent light as theillumination source. Morphologies, sizes, and locations ofthree-dimensional (3D) distributed particles are encoded inthe interferometric fringes known as hologram and thendecoded in a reconstruction process. Holography is a real 3Dimaging technique that has the potential to obtain quantitative,multiparameter results of burning coal particles. It was adoptedfor pulverized coal combustion nearly 40 years ago23,24 withinteresting phenomena observed. But the optical holography atthat time was limited by the tedious recording andreconstruction processes, with only fundamental observationsbeing made. With the fast development of digital cameras andcomputer science, digital holography has begun to play a moreimportant role in engineering measurements.25,26 Hologramsare digitally recorded by a digital camera and numericallyreconstructed in the computer. In addition to particlemeasurement, it also shows some potential to retrieve thetemperature distribution of a flame.27,28 The digital in-lineholography (DIH) configuration, which is easy to implementwith higher sampling efficiency (a tilted reference beam in off-axis creates denser fringes that require smaller sampling stepfor the same imaging resolution),29 has demonstrated itsadvantages in studying particle dynamics30 and has beenapplied to measure solid particles,31 droplets,32−34 bubbles,35,36

crystals,37 micro-organisms,38 and cells.39 DIH was alsodemonstrated to image burning aluminum drops with highcontrast and clear borders. The image distortion caused by theindex of refraction gradient due to thermal variation seemednegligible for the laminar flame.40 Together with two-colorimaging pyrometry, dynamic parameters and the temperatureof burning aluminum could be measured simultaneously.41

Our group recently applied lensless DIH for burning coalcharacterization,16,17 in which the morphology of an individualcoal particle and its surrounding flame and wake plume wereextracted, and the number concentration of coal particles in a3D volume was retrieved. In these studies, a charge coupledevice (CCD) was used to record the hologram digitally.Holograms were reconstructed at different depth locations torefocus coal particles by simulating the wavefront’s back-propagation in the computer. Automatic particle detection andlocation algorithm allowed for fast measurement.42 A keychallenge of applying DIH to combustion diagnostics is thatthe camera should be put far away from the harsh and hightemperature flame, and thus the imaging resolution is limitedby a significantly reduced numerical aperture (NA). Usinglenses can not only improve the NA, but increase the effectiveworking distance as well in DIH.40 So far, many researchstudies have applied the single-lens model43 even when acomplicated lens combination was used. The resolutionlimitations from the lens aperture, recording distance,magnification, sensor size, and pixel size together have not

been discussed. A quantitative design method of the DIHsystem with the multilens system is essential to optimize theresolution and working distance.In this paper, we introduce a design of a long working

distance DIH system for investigating combustion behaviors ofpulverized coal particles. Spatial resolution of a few micro-meters is achieved when the optical apparatus is put more than15 cm away from the flame. The pyrolysis stage of pulverizedcoal is studied by observing particle morphology and volatilestructure at different residence times. Coal size distributionand 3D velocities are also obtained and compared at differentheights above the burner. The work begins with a generalmodel for lensed DIH with generic lens combinations. Theexperimental setup and method are then described. Theresolution and errors of the system are discussed, followed bydirect observation of particle morphology and quantitativeanalysis of particle size and velocity statistics.

■ A GENERIC MODEL FOR LENSED DIH SYSTEMFigure 1a shows a conventional design of a lensless DIHsystem. A plane wave travels through the particle field and is

scattered by particles. Scattered light (object wave, OW)interferes with the undisturbed wave (reference wave, RW) toform a hologram on the digital recording media (e.g., CCDsensor). When a particle is located at a distance za from therecording plane and paraxial approximation is employed, thecomplex amplitude of the wave field at the recording plane is25

U u vikz

i zU x y

ikz

x u y v x y

( , )exp( )

( , )

exp2

( ) ( ) d d

a

a

a

o

2 2

∫ ∫λ=

× [ − + − ]

−∞

∞

lmoonoo

|}oo~oo (1)

where Uo(x,y) and U(u,v) are the complex amplitudes at theparticle plane (PP, z = 0) and sensor plane (SP, z = za),respectively. Uo(x,y) is usually represented by the transmission

Figure 1. Schematics of digital in-line holography system: (a)conventional lensless configuration and (b) lensed configuration.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8278

function [1− O(x,y)] when the particle is approximated as anopaque screen (λ is the wavelength, and k = 2π/λ is thewavenumber).Moreover, the measurement accuracy of lensless DIH

system is limited by relatively large pixel pitch and overallsmall chip size. For example, the pixel pitch of a typical CCDsensor is close to 10 μm × 10 μm, meaning the smallestmeasurable particle is around 30 μm assuming a 3 × 3 pixelblock can describe the particle. The overall chip size is on theorder of 1 cm × 1 cm, suggesting the recording distance of ahologram should be less than 10 cm if a numerical aperture of0.05 is required. To investigate small objects, magnifying lenseswere consequently adopted in imaging, e.g., secondary dropletsizing,44 burning aluminum particles,40 and particle-ladenflow,45 in addition to the development of digital holographicmicroscope.43

Previous research revealed that the recorded hologram canbe seen as a magnified image of a virtual hologram located atthe conjugate plane (CP) of the sensor plane by applying asingle-lens model. A more general model containing multiplelenses and the aperture, shown in Figure 1b, is developed hereusing ray-transfer matrix and the generalized Huygens−Fresnelintegral.46,47 Generalized Huygens−Fresnel integral wasapplied in DIH in some earlier studies, showing itseffectiveness in astigmatic optical systems.48,49 FractionalFourier transform should be used to reconstruct particles.Hereby we will express the system in an equivalent lenslessway, and the traditional angular spectral method is able toreconstruct particle images. The complex amplitude of opticalfield at SP is

U u vikL

i BU x y

ikB

A x y D u v

xu yv x y

( , )exp( )

( , )

exp2

( ) ( )

2( ) d d

o

2 2 2 2{}

∫ ∫λ=

× [ + + +

− + ]

−∞

∞

(2)

where L is the axial optical path from PP to SP. A, B, and D are

the elements of the system ray-transfer matrix A BC DM =

ÄÇÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑ

that relates the ray coordinates of PP and SP:r r

MSPSP

PPPPθ θ=

ÄÇÅÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑÑ

ÄÇÅÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑÑ.

Here r and θ denote a ray’s elevation coordinate and directioncoordinate, respectively. By defining x′ = Ax and y′ = Ay andadopting AD − BC = 1 in free space, eq 2 becomes

U u vikL

i ABi

kCA

u v

AU

xA

yA

ikAB

x u y v x y

( , )exp( )

exp2

( )

1,

exp2

( ) ( ) d d

o

2 2

2 2{ }∫ ∫

λ= +

′ ′

× [ ′ − + ′ − ] ′ ′

−∞

∞

Ä

ÇÅÅÅÅÅÅÅÅ

É

ÖÑÑÑÑÑÑÑÑ

ikjjj

y{zzz

(3)

It is interesting to compare eqs 3 and 1, which model thehologram recording process in the lensed and lenslessconfigurations, respectively. First the constant phase terms,exp(ikz)a and exp(ikL), can be neglected since they onlyinfluence the absolute amplitude. If we introduce an equivalent

particle transmission function ( )U U ,A o

xA

yAeq

1= ′ ′and an

equivalent recording distance zeq = AB, the hologram recordedby a lensed system can be modeled as one by an equivalent

lensless system because the additional phase i u vexp ( )kCA2

2 2+ÄÇÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑ

does not affect the intensity distribution.The system ray-transfer matrix can be further expressed as

A BC D

A B

C Dz1

0 1b1 1

1 1=

Ä

ÇÅÅÅÅÅÅÅÅ

É

ÖÑÑÑÑÑÑÑÑ

Ä

Ç

ÅÅÅÅÅÅÅÅÅÅ

É

Ö

ÑÑÑÑÑÑÑÑÑÑ

Ä

Ç

ÅÅÅÅÅÅÅÅÅ

É

Ö

ÑÑÑÑÑÑÑÑÑ (4)

in which z10 1

bÄÇÅÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑÑ is the translation matrix (from PP to CP in

Figure 1) andA BC D

1 11 1

ÄÇÅÅÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑÑÑ is the matrix describing ray traveling

from CP to SP. From eq 4, it is easy to get A = A1 and C = C1.Since CP is the conjugate plane of SP, one can also get B1 = 0and B = A1zb. So the analytical complex amplitude at SP isexpressed in the frequency domain by applying theconvolution theorem as

U u v ikC

Au v U x y

ikz f f P f f

( , ) exp2

( ) ( , )

exp 12

( ) ( , )X Y X Y

2 2 1eq

eq

22 2λ

= + × [ ]

− +

−Ä

ÇÅÅÅÅÅÅÅÅ

É

ÖÑÑÑÑÑÑÑÑ

lmooonooo

lmoonoo

Ä

Ç

ÅÅÅÅÅÅÅÅÅ

É

Ö

ÑÑÑÑÑÑÑÑÑ

|}oo~oo

|}ooo~ooo (5)

where

U x yA

UxA

yA

( , )1

,oeq1 1 1

=ikjjjjj

y{zzzzz (6)

and

z A zeq 12

b= (7)

and 1− denote Fourier transform (FT) and inverseFourier transform (IFT), respectively. f X = u/(λA1zb) and f Y =v/(λA1zb) are the corresponding frequency components, and Pis the aperture function at the exit pupil. Thus, the hologramrecorded by lensed system can be described by an equivalentlensless system. A1 has nothing to do with the particle positionwhich theoretically means a uniform spatial magnification forall particles. By transferring to equivalent lensless DIH, we areable to design a DIH system with magnifying lenses includingthe limit of aperture. In a lensless DIH system, when an N × Npixel imaging sensor with the pixel size of dp is used, theminimum fringe spacing of a particle located at the center ofthe field of view is dmin = λ/sin(α) ≈ 2λza/(Ndp). α is the anglethat represents the direction of the highest frequency.According to the Nyquist theorem, 2dp ≤ d min is required tosample the fringes; otherwise, the aliasing problem will occur.50

This results in the relationship za ≥ Ndp2/λ to satisfy the

sampling condition. Decreasing za will increase the numericalaperture with NA = Ndp/(2za), so as to improve the resolution.However, when za is smaller than the critical distance zcr =Ndp

2/λ, the aliasing problem occurs and dense fringes near theedge of sensor can not be sampled properly. Thus, the effectiveNA and resolution will no longer be improved. At the criticaldistance, the best resolution proves to be Δx = dp by applyingΔx ≡ λ/(2NA) and NA = Ndp/(2zcr). Combined with eq 7,the critical zb for a general lensed DIH system is zbcr = Ndp

2/(A1

2λ) to achieve the best spatial resolution of dp/A1. On theother hand, the limit by the aperture could be calculated in thediffraction limited system as in ref 51. The lateral resolution

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8279

will thus be finalized by the stricter limit from both theaperture and the pixel size

x

d

Az

Nd

A

A zNd

zNd

A

,

,

pb

p

b

pb

p

1

2

12

12

12

λ

λλ

Δ =

≤

>

l

m

oooooooooo

n

ooooooooo (8)

Our long working distance DIH system here simply adopts a 4ftelescope unit consisting of two lenses and an iris as shown inFigure 2a. The focal lengths of L1 and L2 are f1 and f 2,respectively. The aperture diameter is changeable by adjustingthe iris which is put immediately in front of L1. The rear focalplane of L1 overlaps with the front focal plane of L2 and the SPof a CCD is located at the rear plane of L2, and thus the CP isat the front plane of L1. One advantage of this design is that theworking distance is extended from zb to f1 + zb compared to alensless hologram that has the same resolution by using CCDwith pixel size of dp/A1. It should be noted that the spatialresolution could be improved by using an L2 with larger focallength, but the working distance will always be longer than f1.

Additionally, the extra phase term i u vexp ( )A

kC2

2 2+ÄÇÅÅÅÅÅÅ

ÉÖÑÑÑÑÑÑ turns out

to be 1 in this system because C = 0. Though a telescope is notthe first time applied in a holographic system,52 we evaluate theresolution of the design that almost reaches the theoretical oneand improve the effectiveness of DIH in studying burning coalparticles. We also expect the developed model to providesolutions to more complicated DIH systems with lenscombinations such as holographic endoscope for measure-ments in a boiler in the future.

■ EXPERIMENTAL SETUPA schematic of experimental setup is shown in Figure 2. A double-pulsed Nd:YAG laser beam (532 nm) is spatially filtered andcollimated to a plane wave of 50 mm diameter before illuminating theparticle field. The width of each laser pulse is 5 ns, which is shortenough to ”freeze” the fast moving coal particles. The energy of eachpulse is chosen as 30 mJ. Then the scattered OW and the undisturbedRW both travel through a 4f imaging unit whose usage is explained atthe end of last section. Here the focal lengths are f1 = 100 mm and f 2= 200 mm, respectively, while the iris aperture can vary between 3mm and 25 mm. The system magnification is 2. A band-pass filter(532 ± 10 nm) is mounted on a CCD camera (2048 × 2048 pixel

array with pixel size of 7.4 × 7.4 μm2) to block the unwanted flameradiation. Hologram pairs from the double pulses are captured at aframe rate of 7 Hz by the camera.

Coal particles (diameter less than 200 μm, supplied by a screwfeeder) are delivered by air stream and emitted to the center of anannular burner.17 They are then ignited by the surrounding methaneflame. The inset of Figure 2a exhibits the detail structure of theburner. The particle-laden air is fed through the inner tube with aninner diameter d0 = 5.5 mm, at a flow rate of 9.0 standard (298 K;101,325 Pa) liters per minute (slpm). The methane is introducedthough an annular slit with a thickness of 0.5 mm, at a flow rate of 2.0slpm. The burner is installed on a 3D motorized translation stage,which is able to relocate the burner to control the flame distance andsample volume with an accuracy of 10 μm. The origin of thecoordinate system is set at the center of the burner surface as shownin Figure 2a.

Particles of China Shanxi bituminous coal are studied, of which themain properties are listed in Table 1. The coal particles are ground

and sieved to a size fraction from 10 to 200 μm, with a peak diameterof about 40 μm, which is similar to the pulverized coal used in thermalpower plants. The temperature of the methane flame is first measuredwith a K-type thermocouple at some different heights as shown inFigure 2b that initiates with 600 K at the burner outlet and peaks1300 K at the height of 10−15 cm. Different phenomena are seen atvarious heights in the coal combustion process. A representativephotograph (Figure 2b) indicates that coal particles are ignited at y =4 cm and above, where both the homogeneous volatile mattercombustion and heterogeneous char burning are observed.

■ RESULTS AND DISCUSSIONCalibration of the Lensed DIH System. In this design,

the critical value of zb is chosen as zbcr = 52.7 mm according toeq 7 and ref 50. The sample volume is located between zbcr and1.6zbcr, indicating the working distance (from particle to L1) is152.7 mm to 192.3 mm. The centerline of the burner outlet islocated 170 mm away from L1. The imaging resolution of thesystem is calibrated with a USAF target set at different zb. Asdescribed by eq 8, the resolution remains the same when zb issmaller than zbcr, but it deteriorates in the zb > zbcr region.

Figure 2. (a) Schematic of the long working distance DIH system for measuring burning coal particles in a methane flame. (b) Representativephotograph of the burning particles (exposure time of 50 μs) with measured temperature distribution.

Table 1. Properties of the Coal Sample

proximate analysis moisture ash volatile matter fixed carbon

air-dried, % 6.36 21.29 28.13 44.22ultimate analysis carbon hydrogen nitrogen sulfur oxygen

air-dried, % 56.77 3.79 1.08 0.61 10.10

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8280

Figure 3a shows a resolution of 3.91 μm line width (group 7,element 1) at zb = 0.75zbcr, very close to the analytical one of

3.7 μm line width. The reconstructed target becomes veryblurred and hard to be recognized when zb gradually increasesto 2.28zbcr. Lower resolution will increase the measurementuncertainty. To evaluate the size error and depth positionerror, we use our model to simulate holograms of sphereparticles (regarded as opaque disks) with the same recordingparameters of the experimental setup. Particles are recon-structed and extracted from these holograms followed with acomparison with the preset size and position. The size error(normalized by true diameter) and depth position error(normalized by pixel size) are shown in Figure 3b,c. Both thesize error and depth position error increase after zb > zbcr. Thesize error is larger for smaller particles because a smallerparticle covers fewer pixels. However, the depth position errorshows little relevance with particle size. The measured regionin our experiment, zbcr to 1.6zbcr, has a relatively lowmeasurement uncertainty.Observation of Burning Particles and Volatile Matter

Release. By reconstructing 2000 depth planes from ahologram in a span of 40 mm, 3D distribution of particlesand volatile clouds are refocused. Figure 4a,b is a sample

hologram (recorded at y = 10 cm) and a reconstructed imageat z = 0 in which no particle is in focus, respectively. Theparticle in region 1, of about 60 μm, is refocused at z = −4 mmwith a clear border (Figure 4c). The larger particle in region 2is refocused at z = 6.4 mm, surrounded with an obvious volatilecloud (Figure 4d), where the condensed matter introducesslight blurring around the burning particle. Then the cloud,which encloses fine soot particles less than the resolution,accumulates into a stringy shape tail. Figure 4e shows the topof the tail refocused at z = 8.8 mm, indicating its 3D feature.Figure 5 compiles the reconstructed images of 16 selected

burning particles. The volatile matters also display x − y

movement (No. 2), because of the nonuniform emission ofdecomposition product and attachments of volatile clouds aremost possibly observed right above the ignition point. Thecloud rises up with the gas and gradually detaches from theparent particle (Nos. 3−5) because of the solid−gas slip. In thearea 10 cm < y < 24 cm, the detached stringy objects arecommonly observed. The lengths of stringy objects vary fromthe longest (several millimeters, e.g., Nos. 6, 9, and 10) to theshortest (tens of micrometers, e.g., No. 16). The coal particlesbecome bare and the char starts to burn. The stringy objectsare formed by the detached volatile cloud and move with thegas flow before being oxidized. Some other tails, e.g., Nos. 8and 12, break up into smaller ones. Previous investigationswith shadowgraphy or conventional holography also observedsimilar phenomena.14,53 This can be explained by the fact thatbituminous coals are rich in hydrocarbons and tars whichaccounts for soot formation.Figure 6 shows the morphology evolution of coal particles

along y direction. Particles larger than 50 μm are collectedfrom five holograms for each of the three cases: Figure 6a,burner outlet (y ≈ 0); Figure 6b, y = 10 cm (y/do ≈ 18.2);Figure 6c, y = 24 cm (y/do ≈ 43.6). These particles are selectedby the autodetection algorithm with good representativeness ofthe volatile clouds. Particles at the burner outlet have smoothedges, although their morphologies vary dramatically, whichare similar to the raw coal particles. The edges of particles at y= 10 cm become ragged, and it can be explained by that the

Figure 3. Calibration results of the lensed DIH system. (a) USAFtarget recorded and reconstructed at different distances. (b) Diametererrors of simulated particles. (c) Depth position errors of simulatedparticles. Colors of the scatters represent particle diameter.

Figure 4. Sample images of reconstructed burning particles andvolatile matters. (a) A cropped raw hologram, 580 × 650 pixels. (b)Reconstructed image of the same burning particle in (a) at z = 0. (c)Reconstructed burning particle in region 1 of (b) at z = −4 mm. (d)Reconstructed volatile cloud around the particle (region 2 in (b)) at z= 6.4 mm. (e) Reconstructed soot particle accumulation (region 3 in(b)) at z = 8.8 mm.

Figure 5. Gallery of reconstructed images of 16 selected particlesshowing the ejection and evolution of condensed-phase volatile.Different particles are numerically labeled.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8281

volatile matters are ejected from the surface of particles in thedevolatilization process and the ragged edges roughly reflectthe side view of blow holes. The particle morphologies alsobecome more irregular. A larger proportion of particles at y =24 cm display morphologies significantly different from theones at y = 10 cm. The particles in Figure 6, panels b and c aremore blurred compared with Figure 6a, it is probably due tothe lower hologram quality caused by extensively releasedvolatiles that increase the shadow density. As described inMalek et, al.,54 larger shadow density would decrease thesignal-to-noise ratio. The particle velocities at y = 10 cm and y= 24 cm are about 8 m/s (detailed below), so the residencetimes in these two regions are about 13 and 30 ms,respectively. Since the whole process of the devolatilizationmay last 100 ms,53 only the early stage of this process isobserved in the present study. The fast devolatilization at thisstage causes fast changes of particle morphologies. Furtherstudies attaining longer residence time should be carried out infuture studies to characterize the entire evolution during thedevolatilization process.The observation of relatively small particles from the

cracking of volatile tar tail (e.g., No. 12 of Figure 5) shouldbe attributed to the improved resolution of this system. In ourprevious research with the same camera but without the long

working distance unit,17 only the volatile cloud and wake in alarger scale were imaged. Similarly, the improved resolution ofthe long distance system enables us to study the morphologychange in Figure 6. The improvement can be easily seen bycomparing Figure 6 and another previous study (also Figure 6in that paper).16

Particle Size Distribution. In this study, particle size isdefined as the diameter of the circle that has the same particlecross-section area. The distributions of particle sizes at theburner outlet, y = 10 cm, and y = 24 cm are obtained by 50independent holograms for each case with a total number of5593, 3216, and 1396 identified particles, respectively. Figure7a−c shows the histograms and logarithmic normal fits of thethree cases, with peaks at 35 μm, 48 μm, and 45 μm,respectively. A log-normal fit is often used to represent sizedistributions of pulverized coal particles.55 The fits agree wellwith the histograms in the measurements. A comparison of theprobability density functions (PDFs) (Figure 7d) demon-strates an averaged size increase in the devolatilization process.The obvious size increase at in the residence time of 13 ms(from y = 10 cm) suggests that particles experience swelling inthe very early stage of the devolatilization process. A very slightdecrease of peak size from y = 10 cm to y = 24 cm might resultfrom char burning out. However, the char burning out time isnearly 10 times as the devolatilization time,5 and this influenceshould be too small to be determined in the present study. Inaddition, the size distribution tendency along the longitudinaldirection is of good accordance with that observed from themorphology evolution (Figure 6).

Particle Velocity. To extract the particle velocities, thecamera is operated in a double-exposure mode andsynchronized with the double-pulsed laser by a programmabletiming unit. A pair of holograms are sequentially recorded witha time interval of 20 μs. Particles in the hologram pair are firstidentified and then matched in a 2D way by using theHungarian algorithm,56 which minimizes the total distance toyield the globally most possible pairing of particles insequential frames. The x − y displacement is calculated by asimple subtraction, while the z displacement was calculateddirectly with a focus correlation method because of the lower zlocation accuracy.57 The particle velocity is then derived fromthe displacement and the time interval (vx = δx/δt, vy = δy/δt,vz = δz/δt). Figure 8 demonstrates the 2D and 3D velocityfield from a pair of hologram recorded at the outlet of theburner. In Figure 8a, the base is a depth-of-field extendedimage with all particles focused using a wavelet-based imagefusion algorithm.42 Particles in the first frame are marked byred circles, while those in the second frame are marked by bluecrosses, as shown in the inset of Figure 8a. The vectors presentthe magnitudes and directions of the particle velocities. Figure8b visualizes the 3D velocity field. Particles are represented bythe colored spheres. Both the diameter and the color of thespheres indicate the particle sizes. The vectors show the threecomponents and magnitudes of velocities. 3D particle velocitymeasurement relies highly on the z location accuracy, which isactually determined by the imaging resolution. We did notreport velocity measurement in our two previous researchstudies16,17 because the uncertainly was relatively high.Empirically, locating and tracking burning coal particles aremore difficult than cold particles, and thus needs higherphysical resolution. It may be caused by the gas densityfluctuation in the flame, as well as the deteriorative signal-to-noise ratio after the condensed-phase volatiles appear. That is

Figure 6. Representative morphologies of particles at differentheights. (a) Close to the outlet of the burner. (b) y = 10 cm (∼13ms residence time). (c) y = 24 cm (∼30 ms residence time).

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8282

why the particle images at y = 10 cm and y = 24 cm seem moreblurred compared to the nonignited particle at the burneroutlet having lower gas temperature.At each of three cases, particle velocities are examined. The

radial direction velocity is characterized by vx in Figure 9a−c.At the burner outlet, the radial velocity is below 1 m/s for mostparticles, while the average velocity is around 0, indicating theparticles are moving dominantly upward with the gas jet. At y =18.2d0, the vx in the outer region (larger x) tends to be largerthan that in the core region. The outer particles fled from thecenter at an average velocity of about 0.5 m/s, which suggestsan expansion of the flame in radial direction. At y = 43.6d0,however, the vx difference between the inner and outer regionsis not evident. Instead, there is strong velocity fluctuation,induced by unsteady flame motion. One is reminded that dueto the limited field of view in x direction the fast particles mayexit the sample volume, while slow particles occupy the marginof the field of view. The distributions of axial velocities (vy) are

shown in Figure 9d−f. Because the z span is much larger thanthe x − y span of the measurement volume in hologramreconstruction, we adopt z as the radial direction tocharacterize vy. At the burner outlet, all particles are confinedwithin the gas jet. Average vy presented a profile of pipe flowwith the center velocity of 8 m/s. The large standard deviationis attributed to the difference of drag force on particles ofdifferent size. At y = 18.2d0 particles are more widely spread inthe radial direction, due to mixing of burning gas with theambient air. The average velocity peaks at 7.5 m/s at thecenter. Particles at y = 43.6d0 spread even wider and theaverage vy keeps on decreasing.

■ CONCLUSIONS

In this study, a model for describing a general lensed DIHsystem is first developed by expressing it to an equivalentlensless DIH. A 4f unit is adopted to achieve a workingdistance longer than 15 cm, with spatial resolution of 3.9 μm

Figure 7. Size distributions of coal particle at locations: (a) close to the burner outlet (y ≈ 0 cm), (b) y = 10 cm (∼13 ms residence time), (c) y =24 cm (∼30 ms residence time), and (d) a comparison of log-normal fitted distributions.

Figure 8. Demonstrated snapshot of velocity measurement: (a) field-of-view extended images showing the particles and their velocities (inset: 2Dparticle matching) and (b) a 3D visualization of the measured particle field and velocity field.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8283

for DIH. Meanwhile, the lens aberration is suppressed. A directobservation of coal particle combustion in a pulverized coalflame shows that volatile cloud is ejected from the coal particleat the ignition stage and then detached from the parent particledue to the velocity difference between the coal particle and thebulk gas. The devolatilization products accumulate into stringyobjects that are explained as soot emission from the reaction ofhydrocarbons in the literature.14,53 Examining the recon-structed particles at different heights above the burner showsthat the morphologies change after the combustion process.Ragged borders of particles indicate the formation of blowholes in the devolatilization. Particle size statistics indicate thatparticles experience swelling at first because of the devolatiliza-tion and then reduce slightly in size after very short charburning out. Both the radial and longitudinal velocity aremeasured at different heights. They visualize the influences ofthe flow expansion, the drag force, and the gravity on coalparticles. The particle morphology, size, spatial distribution,and velocity can be quantitatively measured in the 3D volume.This work demonstrates the effectiveness of a long working

distance DIH system for particle measurement in a harshenvironment like the methane flame. With this method, weimprove the measurement accuracy while retaining a safeworking distance. The progress hereby presented enables us tobetter study the burning coal particles in three aspectscompared to previous works:16,17 First, smaller soot agglom-eration (∼10 μm) from volatile combustion can be observed.Second, morphology change before and after devolatilization isshown, and it will be helpful to characterize the volatilebubbling. Third, 3D velocity of coal particles is obtained withthe improved resolution. The data are valuable for optimizingflow condition and validating models. We also expect thetechniques described in this work can be used to develop anoninvasive holographic probe for measuring in situparameters in the coal-fired burner or fuel engine in the nearfuture.

■ AUTHOR INFORMATIONCorresponding Authors*(Y.W.) E-mail: wycgsp@zju.edu.cn.*(J.C.) E-mail: junchen@purdue.edu.ORCIDXuecheng Wu: 0000-0001-9897-8776Yingchun Wu: 0000-0001-8914-1931NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis work is partially supported by the National NaturalScience Foundation of China (51576177), the Major Programof National Natural Science Foundation of China (51390491),the Innovative Research Groups of the National NaturalScience Foundation of China (51621005), and ChinaScholarship Council (201606320173).

■ REFERENCES(1) Chen, W.; Xu, R. Clean coal technology development in China.Energy Policy 2010, 38, 2123−2130.(2) Ding, L.; Dai, Z.; Guo, Q.; Yu, G. Effects of in-situ interactionsbetween steam and coal on pyrolysis and gasification characteristics ofpulverized coals and coal water slurry. Appl. Energy 2017, 187, 627−639.(3) Jia, L.; Geddis, P.; Madrali, S.; Preto, F. Determination ofEmission Factors for Co-firing Biomass and Coal in a SuspensionFired Research Furnace. Energy Fuels 2016, 30, 7342−7356.(4) Zheng, C.; Wang, L.; Zhang, Y.; Zhang, J.; Zhao, H.; Zhou, J.;Gao, X.; Cen, K. Partitioning of hazardous trace elements among airpollution control devices in ultra-low-emission coal-fired powerplants. Energy Fuels 2017, 31, 6334−6344.(5) Smith, I. Symp. (Int.) Combust., [Proc.] 1982, 19, 1045−1065.(6) Khatami, R.; Stivers, C.; Joshi, K.; Levendis, Y. A.; Sarofim, A. F.Combustion behavior of single particles from three different coalranks and from sugar cane bagasse in O 2/N 2 and O 2/CO 2 atm.Combust. Flame 2012, 159, 1253−1271.

Figure 9. Measured velocities of particles at different heights: (a−c) radial components vx and (d−f) axial component vy. Scattered dots representthe results from individual particles. Curves are the averaged value at a chosen x or z location while the error bars give the rms value.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8284

(7) Hu, Y.; Naito, S.; Kobayashi, N.; Hasatani, M. CO 2, NO x andSO 2 emissions from the combustion of coal with high oxygenconcentration gases. Fuel 2000, 79, 1925−1932.(8) Linak, W. P.; Yoo, J.-I.; Wasson, S. J.; Zhu, W.; Wendt, J. O.;Huggins, F. E.; Chen, Y.; Shah, N.; Huffman, G. P.; Gilmour, M. I.Ultrafine ash aerosols from coal combustion: Characterization andhealth effects. Proc. Combust. Inst. 2007, 31, 1929−1937.(9) Maffei, T.; Khatami, R.; Pierucci, S.; Faravelli, T.; Ranzi, E.;Levendis, Y. A. Experimental and modeling study of single coalparticle combustion in O 2/N 2 and oxy-fuel (O 2/CO 2)atmospheres. Combust. Flame 2013, 160, 2559−2572.(10) Marek, E.; Swiatkowski, B. Reprint of ”Experimental studies ofsingle particle combustion in air and different oxy-fuel atmospheres.Appl. Therm. Eng. 2015, 74, 61−68.(11) Lee, H.; Choi, S. An observation of combustion behavior of asingle coal particle entrained into hot gas flow. Combust. Flame 2015,162, 2610−2620.(12) Lee, H.; Choi, S. Motion of single pulverized coal particles in ahot gas flow field. Combust. Flame 2016, 169, 63−71.(13) Molina, A.; Shaddix, C. R. Ignition and devolatilization ofpulverized bituminous coal particles during oxygen/carbon dioxidecoal combustion. Proc. Combust. Inst. 2007, 31, 1905−1912.(14) McLean, W.; Hardesty, D.; Pohl, J. Symp. (Int.) Combust.,[Proc.] 1981, 18, 1239−1248.(15) Rajasegar, R.; Kyritsis, D. C. Experimental investigation of coalcombustion in coal-laden methane jets. J. Energy Eng. 2015, 141,C4014012.(16) Wu, Y.; Wu, X.; Yao, L.; Xue, Z.; Wu, C.; Zhou, H.; Cen, K.Simultaneous particle size and 3D position measurements ofpulverized coal flame with digital inline holography. Fuel 2017, 195,12−22.(17) Wu, Y.; Yao, L.; Wu, X.; Chen, J.; Grehan, G.; Cen, K. 3Dimaging of individual burning char and volatile plume in a pulverizedcoal flame with digital inline holography. Fuel 2017, 206, 429−436.(18) Tokarev, M.; Sharaborin, D.; Lobasov, A.; Chikishev, L.; Dulin,V.; Markovich, D. 3D velocity measurements in a premixed flame bytomographic PIV. Meas. Sci. Technol. 2015, 26, 064001.(19) Balusamy, S.; Schmidt, A.; Hochgreb, S. Flow field measure-ments of pulverized coal combustion using optical diagnostictechniques. Exp. Fluids 2013, 54, 1534.(20) Toth, P.; Draper, T.; Palotas, A. B.; Ring, T. A.; Eddings, E. G.Three-dimensional combined pyrometric sizing and velocimetry ofcombusting coal particles. I. Velocimetry. Appl. Opt. 2015, 54, 4049−4060.(21) Chen, Y.; Wang, G.; Sheng, C. Comparison of particle sizeevolution during pulverized coal combustion in O2/CO2 and O2/N2atm. Energy Fuels 2014, 28, 136−145.(22) Bu, C.; Liu, D.; Chen, X.; Pallares, D.; Gomez-Barea, A.Ignition behavior of single coal particle in a fluidized bed under O2/CO2 and O2/N2 atmospheres: a combination of visual image andparticle temperature. Appl. Energy 2014, 115, 301−308.(23) Trolinger, J. D.; Heap, M. P. Coal particle combustion studiedby holography. Appl. Opt. 1979, 18, 1757−1762.(24) Samuelsen, G. S.; Trolinger, J. D.; Heap, M. P.; Seeker, W. R.Observation of the behavior of coal particles during thermaldecomposition. Combust. Flame 1981, 40, 7−12.(25) Jueptner, W.; Schnars, U. Digital Holography: Digital HologramRecording, Numerical Reconstruction, and Related Techniques; Springer,2005.(26) Meng, H.; Pan, G.; Pu, Y.; Woodward, S. H. Holographicparticle image velocimetry: from film to digital recording. Meas. Sci.Technol. 2004, 15, 673.(27) Agarwal, S.; Kumar, V.; Shakher, C. Temperature measurementof wick stabilized micro diffusion flame under the influence ofmagnetic field using digital holographic interferometry. Optics andLasers in Engineering 2018, 102, 161−169.(28) Sharma, S.; Sheoran, G.; Shakher, C. Digital holographicinterferometry for measurement of temperature in axisymmetricflames. Appl. Opt. 2012, 51, 3228−3235.

(29) Xu, L.; Miao, J.; Asundi, A. K. Properties of digital holographybased on in-line configuration. Opt. Eng. 2000, 39, 3214−3220.(30) Katz, J.; Sheng, J. Applications of holography in fluid mechanicsand particle dynamics. Annu. Rev. Fluid Mech. 2010, 42, 531−555.(31) Pu, S.; Allano, D.; Patte-Rouland, B.; Malek, M.; Lebrun, D.;Cen, K. Particle field characterization by digital in-line holography:3D location and sizing. Exp. Fluids 2005, 39, 1−9.(32) Yang, Y.; Kang, B. Measurements of the characteristics of spraydroplets using in-line digital particle holography. Journal of mechanicalscience and technology 2009, 23, 1670−1679.(33) Gao, J.; Guildenbecher, D. R.; Reu, P. L.; Kulkarni, V.; Sojka, P.E.; Chen, J. Quantitative, three-dimensional diagnostics of multiphasedrop fragmentation via digital in-line holography. Opt. Lett. 2013, 38,1893−1895.(34) Seifi, M.; Fournier, C.; Grosjean, N.; Mees, L.; Marie, J.-L.;Denis, L. Accurate 3D tracking and size measurement of evaporatingdroplets using in-line digital holography and ”inverse problems”reconstruction approach. Opt. Express 2013, 21, 27964−27980.(35) Tian, L.; Loomis, N.; Domínguez-Caballero, J. A.; Barbastathis,G. Quantitative measurement of size and three-dimensional positionof fast-moving bubbles in air-water mixture flows using digitalholography. Appl. Opt. 2010, 49, 1549−1554.(36) Liu, H.; Yu, J.; Wang, T.; Yang, Y.; Wang, J.; Zheng, R. Digitalholography experiment of 3D detection of underwater bubble fields.Chin. Opt. Lett. 2013, 11, S20901.(37) Fugal, J. P.; Shaw, R. A.; Saw, E. W.; Sergeyev, A. V. Airbornedigital holographic system for cloud particle measurements. Appl. Opt.2004, 43, 5987−5995.(38) Javidi, B.; Moon, I.; Yeom, S.; Carapezza, E. Three-dimensionalimaging and recognition of microorganism using single-exposure on-line (SEOL) digital holography. Opt. Express 2005, 13, 4492−4506.(39) Choi, Y.-S.; Lee, S.-J. Three-dimensional volumetric measure-ment of red blood cell motion using digital holographic microscopy.Appl. Opt. 2009, 48, 2983−2990.(40) Guildenbecher, D. R.; Cooper, M. A.; Gill, W.; Stauffacher, H.L.; Oliver, M. S.; Grasser, T. W. Quantitative, three-dimensionalimaging of aluminum drop combustion in solid propellant plumes viadigital in-line holography. Opt. Lett. 2014, 39, 5126−5129.(41) Chen, Y.; Guildenbecher, D. R.; Hoffmeister, K. N.; Cooper,M. A.; Stauffacher, H. L.; Oliver, M. S.; Washburn, E. B. Study ofaluminum particle combustion in solid propellant plumes using digitalin-line holography and imaging pyrometry. Combust. Flame 2017, 182,225−237.(42) Wu, Y.; Wu, X.; Yang, J.; Wang, Z.; Gao, X.; Zhou, B.; Chen, L.;Qiu, K.; Grehan, G.; Cen, K. Wavelet-based depth-of-field extension,accurate autofocusing, and particle pairing for digital inline particleholography. Appl. Opt. 2014, 53, 556−564.(43) Sheng, J.; Malkiel, E.; Katz, J. Digital holographic microscopefor measuring three-dimensional particle distributions and motions.Appl. Opt. 2006, 45, 3893−3901.(44) Guildenbecher, D. R.; Cooper, M. A.; Sojka, P. E. High-speed(20 kHz) digital in-line holography for transient particle tracking andsizing in multiphase flows. Appl. Opt. 2016, 55, 2892−2903.(45) Buchmann, N.; Atkinson, C.; Soria, J. Ultra-high-speedtomographic digital holographic velocimetry in supersonic particle-laden jet flows. Meas. Sci. Technol. 2013, 24, 024005.(46) Collins, S. A. Lens-system diffraction integral written in termsof matrix optics. J. Opt. Soc. Am. 1970, 60, 1168−1177.(47) Yura, H. T.; Hanson, S. G. Optical beam wave propagationthrough complex optical systems. J. Opt. Soc. Am. A 1987, 4, 1931−1948.(48) Verrier, N.; Coetmellec, S.; Brunel, M.; Lebrun, D. Digital in-line holography in thick optical systems: application to visualization inpipes. Appl. Opt. 2008, 47, 4147−57.(49) Sentis, M. P.; Bruel, L.; Charton, S.; Onofri, F. R.; Lamadie, F.Digital in-line holography for the characterization of flowing particlesin astigmatic optical systems. Optics and Lasers in Engineering 2017,88, 184−196.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8285

(50) Zhang, Y.; Shen, G.; Schroder, A.; Kompenhans, J. Influence ofsome recording parameters on digital holographic particle imagevelocimetry. Opt. Eng. 2006, 45, 075801−075801.(51) Goodman, J. W. Introduction to Fourier Optics; Roberts andCompany Publishers, 2005.(52) Sinha, A.; Barbastathis, G. Volume holographic telescope. Opt.Lett. 2002, 27, 1690−1692.(53) Seeker, W.; Samuelsen, G.; Heap, M.; Trolinger, J. Symp. (Int.)Combust., [Proc.] 1981, 18, 1213−1226.(54) Malek, M.; Allano, D.; Coetmellec, S.; Lebrun, D. Digital in-lineholography: influence of the shadow density on particle fieldextraction. Opt. Express 2004, 12, 2270−2279.(55) Ulusoy, U.; Igathinathane, C. Particle size distributionmodeling of milled coals by dynamic image analysis and mechanicalsieving. Fuel Process. Technol. 2016, 143, 100−109.(56) Munkres, J. Algorithms for the assignment and transportationproblems. J. Soc. Ind. Appl. Math. 1957, 5, 32−38.(57) Wu, Y.; Wu, X.; Yao, L.; Wang, Z.; Chen, L.; Grehan, G.; Cen,K. Direct particle depth displacement measurement in DHPTV usingspatial correlation of focus metric curves. Opt. Commun. 2015, 345,71−79.

Energy & Fuels Article

DOI: 10.1021/acs.energyfuels.8b01685Energy Fuels 2018, 32, 8277−8286

8286