Magnetic resonance studies of jets in a gas-solid fluidised bed

particuology 10(2012)161-169Contents lists available at SciVerse Science DirectParticuologyELSEVIERjournalhomepagewww.elsevier.com/locate/particMagnetic resonance studies of jets in a gas-solid fluidised bedMeenal Pore, Thusara C Chandrasekera, DanielJ. Holland, Aining Wang, Fei WangQussai Marashdehb, Michael D Mantle, Andrew ]. Sederman, Liang-Shih Fanb, Lynn F Gladden.John S. Dennisb The william G. Lowrie Department of Chemical and Biomolecular Engimeering. Ohio State University. Columbus, OH 43210, USA omroke Street, Cambridge CB2 3RARTICLE INFOABSTRACTArticle hMagnetic resonance imaging( MRi) has been used to study the behaviour of jets at the distributor of aReceived 15 August 201150mm diameter fluidised bed of 0. 5 mm diameter poppy seeds. Two perforated-plate distributors wereReceived in revised form 13 October 20Accepred 21 October 2011examined, containing either 10 or 14 holes, each 1 mm diameter. Ultra-fast MR imaging was able toshow the transient nature of the upper parts of the jets, where discrete bubbles are formed. imaging in 3Dshowed that the central jets were the longest for flow rates below minimum fluidisation. Above minimumliaised bedfluidisation, the outer jets, nearest the wall of the fluidised bed, arched inward towards the central axis. Inthis latter case, interpretation of the time-averaged 3D image required the use of ultra-fast MR imaginECVTto identify the approximate height above the distributor at which discrete bubbles were formed. Thapparently continuous void extending along the central axis above this height in the time-averaged 3Dimage was thus identified, using ultra-fast MR imaging, as representing the averaged paths of releasedbubbles. Time-averaged MR velocity mapping was also used to identify dead zones of stationary particlesresting on the distributor between the jets. The dead zones could be observed when thesuperficial velocity Comparable images of a single jet through 1.2 mm diamdl. obseof the gas approached minimum fluidisation, but they were smaller than thoselower gasm MRiand electrical capacitance volume tomography (ECvT)are also demonstrated@2012 Published by Elsevier B V on behalf of Chinese Society of Particuology and Institute ofProcess Engineering, Chinese Academy of Sciences1. Introductioncontribute to poor gas-solid mixing. This poor mixing can lead tosintering. Depending on the design, the gas may issue from eachThe motion of particles in gas-solid fluidised beds, as well as hole in the distributor either as a permanent jet, which eventuallymany other fluidisation phenomena, are still poorly understood. spills off bubbles at its tip, or as a stream of bubbles directly fromThe design and scale-up of fluidised beds has thus been diffi- the orifice(Geldart Baeyens, 1985). where jets are formed, prob-cult, despite their widespread use for, e.g. the catalytic cracking lems might arise owing to the fact that jets can convey particlesof hydrocarbons, gasification and combustion of coal, and liquid at sufficiently high velocities to erode internal parts of the bed infuel synthesis by the Fischer-Tropsch process(Fan Zhu, 1998: the vicinity of the distributor(e.g. heat exchanger tubes and walls).Kunii& Levenspiel, 1991 ) Fluidised beds provide high rates of heat Understanding the behaviour of jets near the distributor is theregas, but, in some cases, the gas distribution can be poor(Thorpe, wamt portant for successful design and forms the subject of thisand mass transfer with a good mixing between the particles and fore imDavidson, Pollitt, Smith, 2002). Small bubbles are helpful forVarious techniques have been used to study the flow of solidsgood mixing of gas and solids. However, insufficient contacting can and fluids in fluidised beds. Low-cost, intrusive measurementsresult from bubble coalescence, forming large gas bubbles which are convenient for studying large-scale systems. Typical examplesquickly ascend the bed, bypassing many of the particles. Dead which have been applied to study jets include measurements byzones of defluidised particles can occur on perforated plate distrib- pressure sensors( Vaccaro, Musmarra, Petrecca, 1997). opticalutors between the gas jets issuing from the holes therein, and also probes(Blake, Webb, Sunderland, 1990; Wen, Deole, Chen1982)therm中国煤化工We&Bmes2009and2009: McMillan, ZhouCorresponding author. Tel: +44 1223 334787: fax: +44 1223 334796.AryaCNMH GAng, Fan,2000).Theseauap, vn jets. Internal probes als1674-2001/5-see front matter e 2012 Published by Elsevier B V on behalf of Chinese Sociery of Particuology and Institute of Process Engineering, Chinese Academy of Sciences.w. Pore et aL/ Particuology 10(2012)161-169Wang, Marashdeh, Fan, Warsito. 2010). Whilst tomographictech-Nomenclatureniques, in general, provide a means of establishing void fractiondistributions, quantitative information on particle motion can beorifice diameter(m)difficult to obtainparticle diameter(mframe number in image sequenceMagnetic resonance imaging( MRi) provides spatially-resolvedinformation on chemical composition, diffusion and velocity. Sincestandard deviation of normalised image intensityit is possible to obtain time-averaged 3D images with resoluimage intensity (a.u.)average image intensity of un fluidised particletions 100 um. it is possible to map, for example, distributionsof voidage and particle velocity in particulate systems(Sederman,Gladden. Mantle, 2007). Examples of the application of magneticlage intensity over 50 frames(au)resonance(MR)techniques to study fluidisation phenomena canolumetric gas flow rate through an orifice(m /s)Qmf volumetric flow rate through an orifice at minimumbe found in Savelsberg, demco, Blumich and Stapf (2002). Mullerfluidisation(m /s)et al. (2008)and Holland et al. (2008). In terms of temporal res-olution, Holland, Muller, Dennis, Gladden, and Davidson(2010)Te echo time for magnetic resonance experiment(s)studied transient phenomena across various fluidisation regimesrepetition time for magnetic resonance experimentby acquiring 2D images at 40 frames per second; the acquisitionsuperficial gas velocity(m/s)of 1D profiles at -750 frames per second to study the rise of bubblesUminimum fluidisation superficial velocity( m/s)and slugs is demonstrated in muller et al. (2008 There is a tradeoff between the amount of information that an Mr experimentseeks to obtain and its acquisition time; i.e. higher spatial resolu-tion or velocity encoding necessitates lower temporal resolutioninterrupt the jet flows of interest, so non-intrusive methods are MRI experiments are generally limited to laboratory-scale fluidisedpreferable. Undistorted particle motions in fluidised flows can in beds: at Cambridge, the maximum diameter of bed which can begeneral be measured by positron emission particle tracking(PEPT) studied is currently 50 mm, bounded by the inner diameter of the(Stein, Martin, Seville, McNeil, Parker, 1997) and radioactive radiofrequency coil in the magnet. In addition to this geometricparticle tracking(RPT)(Larachi, Chaouki, Kennedy, Dudukovic, limitation, MR measurements are only applicable to non-magnetic997). Diffusing wave spectroscopy(DWS)can provide informa- and non-metallic systems. Another limitation is that MR signals cantion on particle collisions and velocity fluctuations in fluidised only be obtained from materials containing H nuclei (or other MRbeds( Menon Durian, 1997). Such particle monitoring meth- sensitive nuclei, such as F, 3Por C).ods can yield information on the movement of particles fromRees et al. (2006)were able to image jet structures and detectthe distributor, through the jets and around the rest of the bed. regions of defluidised particles at perforated-plate distributorsThese techniques, however do not provide information on void However, the time-averaged 3D images in that study were not ablefractionsto distinguish between permanent jets and streams of bubbles.ForImaging techniques can be particularly helpful when spatially- flow rates exceeding minimum fluidisation, it was thus not cerresolved measurements of size of jets, or a determination of voidage tain if the central core, which apparently formed above the mergedin their vicinity are required. Since granular systems are visually jets, was a permanent void. Muller et al. (2009)showed that jetopaque, optical measurements at any distance from the walls inside lengths depend on the start-up procedure, and provided evidencesuch 3D systems are difficult Whilst some indication of behaviour that, whilst the lower part of a single jet might be a permanent void,is available from optical studies of two-dimensional fluidised beds the top is transient for flow rates approaching minimum fuidisa(viz beds formed between two, flat, vertical walls spaced a short tion. Pore et al. (2010) demonstrated that jets can be stabilised bydistance apart horizontally)(Merry. 1975), the influence of wall the proximity of other jets or the bed wall. The stabilised jets wereeffects results in fluid and particle dynamics unrepresentative of longer and narrower than a jet from a single orifice, removed from3D beds(Rowe, MacGillivray, Cheesman, 1979). X-radiography the walls of the vessel, at the same orifice velocity. Since Pore et alCleaver. Ghadiri, Tuponogov, Yates, &Cheesman, 1995; Roweet al (2010)considered distributors with up to four orifices with flow1979: Yates, Bejcek, Cheesman, 1986)can be used to produce(2D) rates at or below minimum fluidisation, their study was restrictedphotographs of 3D systems. When spatial resolution is required to stable jets. Jets issuing from distributors with many orifices, atas a two-dimensional slice across the bed or in all three dimen flow rates near and above minimum fluidisation, may not have thesions, tomographic techniques are necessary. For example, x-ray same steady behaviour as that seen by Pore et al. (2010)with justtomography( XRT) gives 3D images of a fluidised bed as stacked a few holes. Care is thus required in the interpretation of timeD slices(Mudde, 2010). Other relevant tomographic techniques averaged 3D MR images in such cases. It is of interest to examineinclude gamma-ray tomography(GRT)(Baumgarten Pigford. what interactions occur between the jets issuing from multi1960:Clough Weimer, 1985), positron emission tomography orifice distributors, particularly when they are not necessarily( PET(Dechsiri et aL, 2005 ) electrical impedance tomography stable(EIT)(Williams et al, 1993), electromagnetic tomography (EMT)Comparisons between MRI and ECVt are necessary to pro-(Binns, Lyons, Peyton, Pritchard, 2001)and electrical capacitance vide cross-validation and facilitate confidence in measurements attomography(ECT)(Halow Nicoletti, 1992). Electrical capaci- industrial scales, made possible with ECVT. The bubbling bed studye volume tomography(ECVT)(Marashdeh, Fan, Du, Warsito. of Holland et al. (2009)considered Group a particles and -30mm2008: Wang, Marashdeh, Fan, & Williams, 2009: Wang, Marashdeh, diameter bubbles. That work demonstrated quantitatively com-Warsito, Fan 2008: Wang, Yu, Marashdeh, Fan, 2010: Warsito, paralMarashdeh, &Fan, 2007) improves on the axial spatial resolution using中国煤化 Tency and voidage obtainedvailable from the stacked 2D images obtained from conventional et alCNMHGs with approximate radialECT, to image in 3D. Sequences of these D images can be produced symmetry arer tne ouooie size nor its shape could bewith a time resolution of 12.5 ms ECVT has, for example, been inferred directly from the ECvt observations. It remains uncertainused to observe a jet of gas entering horizontally into a bubbling what geometric information can be obtained by ECVT, especiallygas-solid fluidised bed of 0.3 m diameter(Wang Yu, et al, 2010: for features which may not have symmetry in all three spatialM. Pore et al. Particuology 10(2012)161-169dimensions. Detection of relatively small objects is a challenge for64 mmCVT, given the smoothness assumptions needed for the difficultprocess of underdetermined reconstructionThis paper addresses three unresolved issues from the aforementioned MRI and ECVT studies. The first question concernsParticleswhether a central core extending from several apparently mergedjets in time-averaged 3D imaging(Rees et al., 2006)correspondsr.f. coilto a permanent void or a bubbling region. The second questionconsiders to what extent the observations of Pore et al. (2010)generalise to distributors containing many holes and the jet-jetDistributorinteractions observable in such systems. The third question is concerned with what aspects of jet geometry can be determined byCVT observations. The present study examines distributors withten and fourteen holes. To interpret time-averaged 3D structureimaging correctly, it is necessary to consider ultra-fast imaging to60detect bubble detachment. MR was used to produce maps of par-ticle velocities around these jets to investigate particle movement43 mmbetween the jets and to detect any dead zones. Additionally, a firstcomparison of magnetic resonance imaging(MRi)with electricalcapacitance volume tomography(ECVT)is also presented to exam-ine jet structure. for the case of a distributor containing a singleGas inlet2. Experimental21. Fluidised bedI mmI mmIn this study, a bed of poppy seeds was fluidised by air. Thebed was contained in a 50 mm i d PerspexPlexiglas, polymethyl-methacrylate) column, shown schematically in Fig. 1(a). Two typesof distributor were used, as shown in Fig. 1(b)(1)Drilled plate distributors with 14 and 10 orifices(orifice diam-eter, do, of 1 mm with a pitch of 11 mm)(2)Asingle orifice distributor with an orifice diameter, d o, of 4 mm.4 mmThe plenum chamber below each multi-orifice distributor hada volume of -200 cm'to dampen fluctuations in gas pressurefrom the compressed air supply(regulated at 1 bar g). The fracion of the distributor area occupied by the holes was 0.40% for the10-hole distributor and 0.56% for the 14-hole distributor. The airsupplied to the single orifice distributor was humidified by passingnmit through a bubbler containing water at 20 C, to mitigate effects onExperimental jetting bed apparatus, and (b) fourteen, ten- and sin-the ECvt measurements arising from the accumulation of electrodistributors. The pitch of the 1 mm diameter holes in the multi-orificestatic charge on the particles. the bubbler was also used during mRdeterminations, for consistency. In addition, the bubbler dampenedluctuations in the air pressure. Poppy seeds were used in this study.because oftheir high content of oil with mabile ' nuclei detectableby MRI (Muller et al, 2008). Seeds of diameter 0.5 mm(Geldart used with Group D particles, except that the air was flowing atGroup B)and 1.2 mm(Group D)were used, with measured mini- -1.30mf for consistency between the mRI and ECVT experimentsmum fluidisation velocities in air at 298 K, Umf, of0 13 and 0.30 m/s, The high flow rate ensured that the Group d particles were fluand particle densities of -1060 and kg/m, respectively. The idised on entry into the bed with the single orifice distributorGroup B seeds were used with the multi-orifice distributors and despite the poorer air distribution in comparison with the multithe Group D seeds were used with the single orifice distributor. orifice distributors. Measurements were acquired for sequentiallyThe height of the bed when tapped was constant at 200 mm for the decreasing flow rates of air, allowing -100s for the bed to stabiliseGroup B seeds and 150 mm for the Group D seeds. In the case of the after each decrement in flow rate. The ratio of the pressure dropGroup B seeds, the height of the bed was greater than the maxi- across the distributor to the pressure drop across the bed of Groupmum spoutable bed depth predicted using the model of Lefroy and B seedculated to be 145% for theDavidson(1969). The empty bed was loaded with Group B particles 10-hole中国煤化士 ole distributor. The ratiowhilst air was being passed through it at the minimum fluidisation of presflow rate, Qms, sufficient for minimisation fluidisation of the parti- pressuCN MH Distributor relative to thecles being added. This standard protocol was used in view of the 443%. These calculations assumed a voidage at minimum fluidifindings of Muller et al. (2009)that the length of jets depended sation of 0.45 and a discharge coefficient of 0.62 for each orificeon how the bed was initially established. The same protocol was (Rees et al., 2006)2. 2. Magnetic resonance imagingby the dashed line along the 10-hole distributor schematic ing The MR experiments were conducted in a 4.7T(199.7 MHz ' H ig. 1(b). The flow rate of air was 18L/min, as measured at 293Kand 1 bar, giving U/Umf"1.2 for the 0.5 mm Group B seeds. Therequency)magnet equipped with a Bruker DMX spectrometer. The standard deviation map of normalised image intensity over 50actively shielded gradient system was capable of producing a max- frames(2.5s total duration)is also shown. The intensity(n)forimum gradient strength of 0. 139T/m. A birdcage radiofrequency each frame(k) used in the standard deviation calculation was(r f )coil(63 mm i.d. was used to excite and detect the ' H nuclei normalised by the average intensity of unfluidised particles(lur).within the poppy seeds three pulse programs were usedThe values of resulting normalised intensity. I//uf, ranged from0 for complete void with 1 corresponding to unfuidised parti(1)A FLASH(fast, low-angle shot)sequence was used to image cles. The standard deviation(s)at each position(x,z)was thusrapidly a 5mm thick, 2D slice(x, plane)at 32 pixel x 32 pixel calculated aswith a field of view of 55 mm x 70mm, giving a spatial reso-lution of 1.72 mm x 2. 19 mm. the images were acquired atI(x, z, k I(x, z)temporal resolution of 50 ms per image. Here, the z direction is s(x, z)=rE0-1Eluf(1)vertically upwards( 2)A velocity-encoded, spin-echo sequence was used to acquire where the mean intensity at each position is given bytime-averaged 2D 64 x 64 velocity maps(xz plane)of the jetting region the image was acquired with an acquisition time of-20min and a slice thickness of 2 mm with a 55 70mm (x,2-502(x,z, k)field of view, which gave a resolution of 0. 86 mm x 1. 1 mmFour averages were acquired per slice with a repetition time,TR=1.5s and echo time, TE=2msAs a guide, if a pixel spent 50% of the duration as unfluidised()A multi-slice, spin-echo sequence(Muller et al, 2009)was particles and 50% as void, it would have a standard deviation ofused to acquire time-averaged, 3D images of the distributor 0. 5. Since the orifices are 1 mm diameter and the horizontal res-region. In the axial (z)dimension, 41 slices were acquired, olution is 1. 72 mm per pixel, the entrance of the four jets intoach I mm thick, to give an axial spatial resolution of 1 mm. the bed is partially resolved(seen as an intensity reduction)justFour averages were acquired per slice with a repettiabove the distributor in the image sequence. As the jet widthsTR=2.5S, and echo time, TE-2.5 ms For the drilled plate dis- increase with height, the contrast improves such that the uppertributors each 2D (xy plane)64x32 slice was acquired with ends of the jets are most visible. Up to this height, the jet stalks55 mm x 55 mm feld of view, which gave an in-plane resolu- are similar across the frames- there does not seem to be anytion of0.86 1.72 mm. The acquisition time was 5min for significant variation in intensity in this region( s-0.05), accordall 41 slices. For the single orifice distributor each 2D(xy plane) ing to the map of standard deviation shown in Fig. 2. Also, in this128x128 slice was acquired with 55 mm x 55 mmfield of view, region the jets appear to be steady. The left three jets are mergedwhich gave an in-plane resolution of 0.43 mm x 0.43 mm. The at the upper end: in this region and higher, the intensity distriacquisition time was --20 min for all 41 slicesbution changes across the image frames and the correspondinghigh values of standard deviation(s>0.1)indicate significant vari-It is possible to acquire 2D FLASH images at higher temporal res- ation in the voidage distribution. There does not seem to be mucholution(25 ms) with lower spatial resolution(Muller et aL, 2008). Intensity variation around the jet on the right, apart from at itsThe 50 ms interval images here were acquired so that the heights upper end. The FLASH images show the typical release of a bubbleat which discrete bubbles detach could be established with -2 mm from where the jets merge. a bubble appears to be released everyvertical resolution. Such information can be particularly helpful in 100 ms. The tops of these jets are transient, whereas the stalksthe analysis of the time-averaged imagesbelow are steady. These images are sensitive to changes in the jet(solid-void )boundaries, but they do not distinguish between mov-ing and stationary particles directly. Any dead zones resting on the2.3. Electrical capacitance volume tomographydistributor, between the jets, are thus not apparent in these FLASHhe ECvT experiments with the single orifice distributor wereTime-averaged images of (i)intensity and (ii)the vertical com-performed with PTL (UK)DAM200-TP-G 12-channel data acquisi- ponent of velocity are shown for the Group B particles with (a)tion hardware. A 3D sensor with 12 rectangular electrode plates, 10L/min(U/Umt -065)and(b)15 L/min(U/Umf =0.98)through thedistributed over three planes( four electrodes in each plane), was 10-hole distributor in Fig 3. The jets appear segregated in theselocated around the bed to image up to-100 mm above the distribu- time-averaged images, although there is significant blurring intor. The sensor geometry is illustrated in Holland et al.(2009). Each the intensity images in the regions around each jet. The blurringelectrode was 30 mmin height and theelectrodes were spaced with indicates particle motion during the -20 min averaging tinmm gaps between them. Capacitance frames(each consisting of vertical velocities corresponding to these blurred intensit80 frames per second. The NN-MOIRT algorithm(Warsito Fan, utor plate, between the jets, are regions of particles that seem tobe at rest one of these apparent dead zones has been outlined inimage with 50 mm x 50 mm x 100 mm(x y,z)field of vieweach of Fig. 3(a)ii) and Fig. 3(b(ii). The zones are shorter whenthe a中国煤化工0 L/min, whereas a largertranszones at the higher flowrateC MH Golids predominantly move3.1. Motion near the distributor and jetsdownwards in tne regions Detween tne jets and above the deadzones. Falling solid can still be observed above the 1 mm diameterA sequence of FLASH images is shown at 50ms intervals in orifices, at the bases of the jets. This solid corresponds to partiFig. 2 for the 10-hole distributor The slice orientation is given cles either side of the orifice, which have been included within theM. Pore er al. Particuology 10(2012)161-16955 mmSolidR50 ms100ms150ms200ms0.0250ms350msStandard deviationFig. 2. Sequential central (x z)slices acquired using ultra-fast FLASH MR imaging Data are shown for the 10-hole distributor. The flow rate is 18L/min through the GrouB seeds(U/Umr -1. 2).The time resolution is 50ms. The slice position corresponds to the faint dashed line through the 10-hole distributor layout in Fig. 1(b). The standarddeviation of normalised intensity over 50 frames is also shown.2 mm slice thickness. The voidage within the jets is too high for the 3. 2. Structure of, and merging of, jetsfast rising particles within the jets to be detected, Rising solid can,however, be detected above the top of each of the time-averagedjetVertical 2D(xz)slices situated at the vertical axis of the bedvoidsobtained by 3D multi-slice MRL, are shown in Fig 4 for air flowGas0.05m/s中国煤化工005msCNMHGFig-3. Time-averaged(xz)()intensity and (i)vertical component velocity images of Group B seeds with(a)104/min(UJUmr-065)and(b)154/min(U/Umr-098)throughthe 10-hole distributor. Positive velocities correspond to rising particles, negative velocities correspond to falling particles Regions where insufficient solid is detectable toneasure velocities in(ii)are masked in grey. The middle dead zones are outlined in(ii) The slice position corresponds to the faint dashed line through the 10-hole distributorlayout in Fig. 1(b).M. Pore et al Particuology 10(2012)161-16955 mm(b)mm50mFig. 4. Central vertical 2D(x.z)slices extracted from 3D multi-slice MRI of GroupB seeds for(a)10Umin(U/Umf-065)(b)15y/min(U/Umf=0.98)and(c)20L/min(C(UJUmf-13)through the 14-hole distributor. The slice position corresponds to thefaint dashed line through the 14-hole distributor layout in Fig. 1(b).rates (at 293 K and 1 bar)of(a)104/min(U/Umf =0.65).(b)15 L/minyUmf-0.98)and(c)20L/min(U/Umf"1.3)through the 14-holedistributor. The slice-plane orientation is indicated by the dashedline across the diagram of the 14-hole distributor in Fig. 1(b). Theseimages indicate the average structure over-5 min At U/Umf=0.65,the jets appear segregated, whereas merging behaviour is seen forone of the centraljets at U/Umf=0.98, probably arising from interac-50 mmion with at least one other jet out-of-plane; these will be discussedshortly. The apparent channel extending above this longestjet givesan indication of the averaged paths of released bubbles over the-5min acquisition time. FLASH imaging with 50ms frame inter-vals(not shown)demonstrated similar bubbling behaviour to Fig. 2,confirming that the channel was not a permanent void. The imagefrom the multi-slice experiment for U/ Umf=1. shows each jet toin that the merged jet seen in Fig. 4(b)is no longer seen. The void at 104min(U/Umr-065)(b)154/min(U/Umr-098)and(c)20/min(U/Umt-1.3)through the 14-hole distributor, orientation shown in(d). The particles are Groupthe top of Fig. 4(c), separated from the in-plane jets below, possibly B seeds. The blue clipped areas at the top of subplots(b) and (c) correspond tosuggests the entrance of voids originating out-of plane. The spatial the upper image boundaryresolution and contrast is sufficient to subject the 3D images to a (verified with time-averaged 2D MR images, also not shownbinary gate to infer the shapes of the voids. The gating thresholdwas set at three times the background noise level (measured fromimaged regions outside the bed ) Accordingly, examples of 3 D rep- time-averaged void, in fact, extends beyond that height(verifiedresentations of the average void boundaries are given in Fig. 5 for with time-averaged 2D MR images, albeit not shown). One of thair flow rates(at 293 K and 1 bar)of(a)10L/min(U/ Umf 0.65),(b) outer jets in the foreground also appears to be interacting with the15L/min(U/Umf"0.98)and(c)20L/min(U/Umf=1.3)through the four central jets. the observed interactions are consistent with thedering is indicated in Fig. 5(d). The dashed line through Fig. 5(d) 20L/r中国煤化工 to the wall appear toba14-hole distributor The orientation of the distributor in each ren- suggested out-of plane interactions associated with Fig. 4(b)corresponds to the orientation of the corresponding slices in Fig. 4. growtAt 10L/min(U/Umf =0.65), it is seen that the jets are clearly sep- aplCN MH Gated configuration seen inarate with the four central jets being slightly longer than the Fig 4(). FLASH imaging (not shown)suggests that bubbles detachuter jets At 15L/min(U/Umf=0.98), the voids in the centre have from the outer jets at a height of- 22 mm above the distributor. Thegreater height and appear to merge into a central core which forms extension of the voids above this approximate height indicates thealong the bed axis. the image is clipped at the top, although the averaged paths followed by the spilled bubbles towards the centralresolve with ECVT-instead the side appears vertical. The location(a)and height of the jet is broadly similar to that seen by mrL.The detection of bubbling by ultra-fast imaging is important forthe interpretation of time-averaged images, which provide highern(50 ms)of theto identify bubbling behaviour, albeit at the expense of spatial res-olution. The standard deviation map in Fig. 2 suggests that littlevariation in the jet structure is seen in the region up to -6mmabove the distributor(for U/Umf =1. 2), in comparison to what is30seen above that region. the bubbles in Fig. 2 are seen to rise fromthe merging region at the top of the left three jets. The FLASH technique can thus determine where there is movement of the voidand bubble release Ultra-fast imaging distinguishes between a per-manent void and a stream of bubbles( muller et al, 2009), both ofwhich can appear as continuous voids in time-averaged images.Whilst time-averaged images, such as those presented in Figs. 3-5,or in Rees et al. (2006), cannot show individual bubbles, they canindicate the major paths they take up the bed.In the case of U/ Um>1, FLASH imaging(50 ms intervals]demontrated that the apparently continuous void extending above the20mm level along the central axis, corresponded. in fact, to theaveraged paths taken by released bubbles as they ascended thebed. This distinction was not revealed by time-averaged 3D imaginged by rees et al. (2006). Given the FLASH observations,it is likely that the central core observed for U/Umf=1.0-1,4 in thetime-averaged 3D images of Rees et al. (2006)corresponds to theaveraged paths of released bubbles, rather than a permanent void.It should be noted that this situation did not feature in the observation of interactions between two or three neighbouring jets inPore et al. (2010).The 3D rendered images for the 14-hole distributor here showFig.6. Images of the estimated boundary of a single jet froma mm diameter orifice similar behaviour to the observations of Rees et al. (2006)for dis-tributors with at least 15 holes. Separate jets are seen in this workas well as in that of Rees et al. (2006), for U/Umf <0.7. The four cen-axis, where they appear to merge into a single average channel. This tral jets are each surrounded by six jets (other jets are surroundedcentral core corresponds to the void seen at the top of Fig 4(c). The by fewer). These four central jets appear to experience the mostbling channel extends above this level, although it is not showFigs. 4(a)and 5(a). The voids appear longer in Figs. 4(b) and 5(b),for(also verified with time-averaged 2D MR images, also not shown) UjUmf-1 The apparent merging of the voids in the time-averagedThis apparent central core and the bending of the wall jets towards images indicates interactions between the jets For U/Umt>1. thethe axis of the bed are consistent with the observations of Rees et al jets near the walls arch inwards. towards the central axis, as seen in(2006)for values of U/Umt between 1.0 and 1.4Fig 5(c). The slice in Fig 4(c)shows that the jets, which correspondto the orifices along the dashed line in Fig. 5(d), appear segregatedfrom each other. The 3D reconstruction in Fig. 5(c)shows that the3.3. Comparison of MRi and ECVTinteractions between jets are mostly out of the plane of this cen-tral slice. There are three main hypotheses which might explainFig. 6 shows 3D rendered images of a single jet, emanating from these observations. First, the flow pattern set by the start-up pro-a 4mm diameter orifice at 14.2 L/min(U/Umf=0. 4). through the cedure affects which jets are seen to interact and merge DifferentGroup D seeds. The ECVT image was interpolated using a cubic- configurations of merged jets are thus possible. the bend of thespline in all three dimensions to 50 x 50 x 100(from 20 x 20 x 20) outer jets towards the central axis has been observed in the Reesto give a resolution of 1 mm in each direction, prior to binary gating. et al. (2006 )study, for U/Umf=1.0-1.4, whilst some of the centralInterpolationot necessary for the mr multi-slice data. The jets were higher than the other jets for U/Umf 51.0. Second, thehreshold for the mr data was chosen according to the noise level, axial and radial dissipation of gas momentum affects the structureas described for the 14-hole distributor image in Fig. 5. The choice of a中国煤化工 ation by their surroundingof gating threshold for the ecvt image was based on comparison deawith the MR image. A systematic method for choosing this ECVT theCNMH Ges around the jets fluidisegating threshold will be presented in a future paper( Chandrasekera (U/Umf+n. ine resistance to tne air rlow further reduces as theet al. submitted for publication). The MR image in Fig. 6 shows the voidage near the orifice increases. Hence, longer central jets thangrowth in jet width from the base towards the top of the jet and outer jets cannot be sustained for U/Umf >1. third, the extent tothe shape of the head of the jet. This detail in shape is difficult to which the wall affects the outer jets is uncertain. The observationsM. Pore et aL Particuology 10(2012)161-169of Pore et al. (2010 ) showed that when the outer edge of any part ofThe 50 ms frame rate used in this study was sufficient to detecta jet impinges on the wall, the wall provides a significant stabilisinghe approximate height of bubble release with 2 mm vertieffect and the jet length increases. In that study, the jets, from 2 mm cal spatial resolution. It was thus shown that the central corediameter orifices within a critical distance of 6 mm from the wallobserved above apparently merged jets was a bubbling region,extended (in width)to touch the wall. As these jets moved close torather than a permanent void, resolving the ambiguity noted inthe wall, they became asymmetric with the top of each jet curvingRees et al. (2006)in towards the centre of the bed. These jets broke through the bed (2) Spatially-resolved velocity measurements demonstrated theforming spouts along the wall when U exceeded Umf, although thepresence of dead zones of defluidised particles resting on thebed depth used was less than the maximum spoutable bed depthdistributor between the jets, when U/Umf=0.98. The height ofpredicted by the model of Lefroy and Davidson( 1969). The outerthese dead zones increased when the flow rate was reducedI mm diameter holes in the multi- orifice distributors used in this(U/Umf=0.65) Imaging in 3D indicated that, for U/Um >1, therestudy were farther than the critical distance from the wall observedwere interactions between the jets nearest the wall of the flu-by Pore et al. (2010). The pitch of the distributors in Rees et al.dined bed These jets arched inwards towards the central axis2006)also exceeded that critical distance. The jets emanating fromThe central jets appeared longer for U/Umt<1. Three explanathe outer orifices thus did not extend to the wall, as can be seen fortions exist to account for these observations, but further studiesU/Umf-13 in Fig. 5(b). The inward arching of the outer jets seems are needed to determine the correct one. Thus, the behaviourto have occurred at a greater distance from the wall than would bemight be(i)a result of the flow patterns established as the bedexpected for jet-wall stabilisation effects, as studied by Pore et alwas started-up by loading the particles at minimum fluidisa-(2010). A combination of ultra-fast imaging, quantitative voidagetion and subsequently decreasing the gas flow to the requiredand velocity mapping with MR, similar to Holland et al. (2010)rates, (ii) a consequence of the relative rates of axial and radialcould be used to investigate the extents to which these three issuesdissipation of the momentum of the gas entering through thecontribute to the observed jet behaviour.holes into the bed or(iii) the influence of possible interactionsThe time-averaged velocity maps( Fig 3(ii)) identify the locabetween the jets and the wall of the tube.tions of falling or stationary particles. It is thus possible to detect (3)Imaging of a single jet from a 4 mm orifice distributor bywhere dead zones form between multiple jets and observe theirECVt gave a jet height consistent with images obtained by MRreduction in size with increasing flow rate. The rising particlesalthough the ECVT resolution was insufficient to recover detailsin the low voidage regions are typically not observed, but alof the jet shape.instead inferred to be within the jets. ldentification of the tran-sitions between stationary and falling particles is easiest with thevelocity images in Fig 3(). Such transitions are difficult to detect Acknowledgementprecisely by inspection of blurred and non-blurred regions in theThis work was funded by the Engineering and Physical Sciencesintensity images in Fig 3(0), although the blurring in those images Research Council( Grant number EP/F041772/1).can be used as a basic indicator of movementAn important question to be answered is the extent to whichjet behaviour varies with the size of the fluidised bed. As noted Referencesearlier, our MR measurements are currently limited to a maximumdiameter system of 50 mm, but techniques such as ECVT could pro- Baumgarten, P.K.& Pigford RL( 1960). Density fluctuations in fluidized beds. AIChEvide a means to explore scale. The thresholding method used for Berruti. F. Da we M.& Briens. C (2009). Study of gas-liquid jet boundaries in athe jet in Fig. 6(b)and compare it with the mr version in fig. 6(a). Binns, R, Lyons, A.R. A, Peyton, A.J.& Pritchard, w D N(2001 ) Imaging moltenHolland et al. (2009) have previously demonstrated that compara- Blake. T.R. Webb, H& Sunderland, P B(1990). The nondimensionalization ofble voidage distributions in a bubbling bed of Group A particlebe obtained with MRi and ECVt. where the ECvt measurementsuations describing fluidization with application to the correlation of jet pen-were interpreted as providing locally averaged measurements.has been seen in these complementary measurements that ECVtet al. a com paris n of magnetic resonance imaging and electrica capacitanceprovided an image of a single jet that was comparable in size with Cleaver, J.A. S.Ghadiri,M,Tuponogov V.G. Yates.]G,&Cheesman. D, J(1995).that detected by Mri, despite lacking the resolution to recoverfluidised beds. Powder Technology, 85, 221-226.detailed comparisons between MRI and ECVT are the subject of Dechsiri, C, Gione, A van de Wiel, F Dehling, H G, Paans, A M ],&Hoffman, A.C.ongoing work by the authors( Chandrasekera et al, submitted forGeldart, D, Baeyens, J (1985). The design of distributors for gas-fluidised beds.5. Conclusionsowder Technology, 42, 67-78Halow, J S.& Nicoletti, P (1992). Observations of fluidized bed coalescence using(1)Acombination of ultra-fast imaging with time-averaged veloc- Holland. D.J. Marashdeh, Q- Muller, C.R. Wang F- Dennis J.S. Fan, LS et al.(2009).of ECVt andthe transient(upper)and permanent(lower)jetting regions HollanIndustrial and Engineering Chemistry Research, 48. 172-181n, L F, Davidson, ) F(2010near multi-orifice drilled plate distributors(with 10 and 14中国煤化工sntrial and engineeringholes). Discrete bubbles were seen to be released from theHollatransient region where the jets were prone to merging forCNMHG.MD, Gladden, LF. DavidsonU/Umr >1. ldentification of the averaged paths of released bub-Theoretical Foundations of chemical Engineering, 42(5). 469-478bles, which appeared as a continuous void in the time-averaged Kuni, D.& Levenspiel, o (199t ). Flutdtzarion engineering(2nd ed.)London:facilitated by determiningapproximate height of bubble release with ultra-fast imaging.Larachi, F, Chaouki, L, Kennedy, G, Dudukovic, M. P ( 1997). Radioactive particleacking in multiphase reactors: Principles and applications In J Chaouki, F.M Pore et al. Particuology 10(2012)161-169Larachi, M P. Dudukovic(Eds h Non-invasive monitoring of multiphase fows Stein, M. Martin, T. W Seville, J.P. K,McNeil, P.A.&Lefroy. G. A, Davidson J F (1969). The mechanics of spouted beds. Transactionsers and other applications, In J. Chaouki, F. Arachiof the instring of multiphase fiows(pp 161-184). Amsterdam:Marashdeh, Q, Fan, L S, Du, B.& Warsito, w.(2008) Electrical capacitancetomography-A perspective. industrial and Chemical Engineering Research.Thorpe, R B, Davidson, J F, Pollitt, M, &Smith, (2002) Maldistribution in fluidised3708-3719ds. Industrial and Engineering Chemistry Research, 41(23)582005). Characterization Vvaccaro, S, Musmarra, D Petrecca. M. (1997). Evaluation of the jet penetrationetween liquid spray dropletdes in a fluidized bedized beds by pressure signal analysis. Intemational Jourmal ofutg Cemistry Researd4931-4939Menon, N,& Durian, D J (1997), Particle motions in a gas-fluidized bed of sand, Wang, F. Marashdeh, Q. Fan, L S,& Warsito, w.(2010). Electrical capacitancePhysical Review Letters, 79, 3407-3410olume tomography: Design and applications, Sensors, 10, 1890-1917Merry, I.M.D. (1975). Penetration of vertical jets into fluidized beds, A/ChE Jourmal. Wang. F. Marashdeh, Q, Fan, L.S. Williams. R A (2009). Electrical capaelectrical resistance, and positron emission tomography techniques andMudde. R F(2010). Time-resolved X-ray tomography of a fluidized bed. Powderlications in multi- phase flow systems. In I H u(Ed. Advances in chemi-Technology, 199(1)55-59.gineering: Characterization of flow, particles, and interfaces(pp. 179-222).D, I., J. F. Dennis. ]S, Gladden. L F. Hayhurst. A.Amsterdam: Academic PressN, et al. (2009). Geometrical and hydrodynamical study of gas jets in packeangS(2008) Imaging gas/solid jetd fluidized beds using magnetic resonance. The Canadian Joumal of Chemicalpenetration in a gas-solid fluidized bed using electrical capacitance volumeerng87,517-525tomography In AIChE Annual Meeting (Section 03B01) Philadelphia.Maller. C.R. Holland, D J, Sederman, A. Mantle. M.ladden. L F& Davidson, Wang. F. Yu. Z, Mardshdeh, Q. Fan, L s.(2010) Horizontal gas and gas/soli. F (2008) Magnetic resonance imaging of Fluidized beds. Powder Technology.neration in a gas-solid fluidized bed. Chemical Engineering Science. 65,ore, M, Holland, D, J, Chandrasekera, T C, Mulier, C R, Sederman, A], Dennis. j Warsito, W fan, L..(2001). Neural nerwork baS, et aL. (2010) Magnetic resonance studies of a gas-solids fluidised bed: Jet-jetsing electrical capacitance tomography. Measurement Science and TechnologyRees, A. C- Davidson, J. F. Dennis, J. 5, Fennell, P S, Gladden. L F Hayhurst, A N,etal (2006). The nature of the flow just above the perforated plate distributor of Warsito, w, Marashdeh, Q&Fan, LS(2007). Electrical capacitance volume tomog-y IEEE Sersors JournaL, 7(4). 525-535Science,6l,6002-601wen, C.Y, Deole, N. R. Chen, L.H. (1982). A study of jets in a three-dimensionalRowe. P NR, Dickin, F. 3. Ilyas, O. M, Ying. P, Edwards, R. B. et al.velsberg. R, Demco, D E. Blumich, B,& Stapf. S (2002) Particle motion in gas-vessels. AIChE Symposiauidized granular systems by pulse- feld gradient nuclear magnetic resonance. Yates, J G. Bejcek, V.& Cheesman. D J ( 1986)on into fluidised bedsSederman, A.J. Gladden. L F. Mantle, M. D (2007). Application of magnetic86). New York: Engiging techniques to particulate systems. Advanced Powder TechneZhu C Wang,& Fan, Ls. (2000) Effect of solids concentration on evaporative181)23-38.liquid jets in gas-solid flows. Powder Technology. 111(1-2)79-82中国煤化工CNMHG