Advanced material distribution measurement in multiphase flows: A case study

7/30/2019 Advanced material distribution measurement in multiphase flows: A case study

1/14

ADVANCED MATERIAL DISTRIBUTION MEASUREMENT IN MULTIPHASE FLOWS: ACASE STUDYD. L. George and S. L. Ceccio

Departmentof Mechanical Engineering and Applied MechanicsUniversity of Michigan

Ann Arbor, Michigan 48109-2121 USAT. J. OHern, K. A. Shollenberger, and J. R. Torczynski

Engineering Sciences CenterSandia National Laboratories

Albuquerque, New Mexico 87185-5800 USAABSTRACT

A variety of tomographic techniques that have b m applied tomultiphase flow are described. The methods discussed includeelectrical impedance tomography (EIT): magnetic resonance imaging(MRI), positron emission tomography (PET), gamma-densitomcptomography (GDT), radiative particle tracking (RDT). X-ray imaging,and acoustic tomograph!.. Also prcscnted is a c s c study in whichmeasurements w r c made isith EIT and GDT in hvo-phase flows. Bothsolid-liquid and gas-liquid flows were examined. EIT and GD T wereapplied independently to predict mean and spatially resolved phasevolume fractions. T h e results from the two ?stems compared well.

INTRODUCTIONA dilemma encountcrcd in multiphase flow measurements is thatprobes or instruments should be placed outside the flow domain so as

not to disturb the flow itself, yet phase distributions cannot easily bemeasured Gom the flow b o u n d q . Tomography, the technique ofimaging plane or volume sections withiin an object, offers a possiblesolution, and many differcnt tomographic methods have been applied tothe measurement of multiphase flows. Most of these tomographicmethods have been derived from devices for medical applications, suchas electrical impedance omography (EIT), magnetic resonance imaging(MRI), and positron emission tomography (PET). Presented here is abrief review of a variety of tomographic techniques and related methodsthat can be used to determine the spatial distribution of the phases in amultiphase flow. In addition. a case study is presented that comparesand evaluates two tornographic methods: EIT and gamma-dcnsitometqtomography (GDT).

Sandid is a mult ipropm laboraton o p t d by Snndin Corporation, aLmkhheed Marun Company, for the United States Department of Energy undmContractDE-ACO4-94ALSjGW.

REVIEW OF TOMOGRAPHY METHODS FORMULTIPHASE FLOWS

Radiation-based TomoqraphyMany tomographc methods applied to multiphasc flows involv

radiation. such as neutrons. gamma rays. or x-res. that are onlpartially attenuated by the flow. Information about local density ophase lstributions can be obtained by measuring attenuation ofradiation beam through he domain or by trianbwlating the locations oradiation sourccs within the flow. These methods also have thadvantages of being nonintrusive and easily applied to industriaconditions but t?pically rcquire radiation shjelding and other methods opersonnel protection. Th c methods are slow since long counting timeare needed to obtain useful statistics, but images of high spatiaresolution can be readih rcconstructcd. The nest four subsectiondescribe a variety of radiation-based methods applied to the study omuhiphase flows.

Radioactive Particle Trackinq (RPTI. Radioactivparticles can be placed in a multiphasc flow and used as Lagrangiaflow markers. The particle is designed to be a marker of a particulaphase in the flow. For example, neutraily buoyant particles can be useto track a liquid element of the continuous phase, or the particle cabe one of many in a dispersed solid phase. Lin et ai. (1985): Lxachi eaf. (1994, and Roy er al. (1998) provide detailed descriptions of thtechnique. The radioactive emissions from the particle are measure


2/14

DISCLAIMERThis report wa s prepared as an account of work sponsored by an agency of th eUnited States Government. Neither the United States Government nor any agencythereof, nor any of their employes, makes any warranty, express or implied, orassumes any legal liability or responsibility for the accuracy, completeness, or use-fulness of any information, apparatus, product, or process disclosed, or representsthat its use would not infringe privately owned rights. Reference herein to any spe-cific commercial product, process, or %Mix by trade name, trademark, manufac-turer, or otherwise does not necessarily constitute or imply its endorsement, recom-mendation. or favoring by the United States Government or any agency thereof.'The views and opinions of authors expressed herein do not necessarily state orreflect those of the United S tates Governm ent or any agency thereof.


3/14

DISCLAIMERPortions of this document may be illegiblein electronic image products. Images areproduced from the best available originaldocument.


4/14

with anmay of precisely calibrated detectors. Th e domain of the flowis divided into a number of volume elements, and the signals from thedetectors are used to locate the tracer particle nithin an element. Bytracking the particle as it moves across volume elements, the phasevelocie may be determined. By tracking the particle over many hours.the flow within a closed vessel can be accurately mapped. For themethod to be effective, the particle must move faithfully with the phaseof interest- .g . ,a solid tracer particle must have the same d en si e andsize as the remainder of the dispersed solid phase.

been used for some time to investigate multiphase flows, ancomprehensive reviews are provided by Munshi (1990), Beck et a(1993) and Simons (1995). The spatial resolution of GD T systems cabe very h e < 1 mm). At present, the long data acquisition times oGDT %stems (tens of minutes to hours) limits their usefdness to thdetermination of time-averaged properties. X-ray tomography may alsbe used in situations where lower-energy photons are acceptabl(Dyakowski,1996, and Torcqm ski er al.; 1997).

Positron Emission Tornocaraphv (PET). In positronemission tomography (PET), radioisotopes. such as fluorine-18.sodium-22, or gallium-68, are placed in the flow to act as tracers. Asan atom of the radioisotope d ec cs , it emits a positron, a particle withthe same mass as an electron but oppo site charge. Th e positron travelson& a few millimeters in solids or liquids before encountering anelectron: the resulting annihilation of the hvo particles produces hvogamma photons of the same e n e r g that travel in opposite directionsfrom the annihilation point along a single asis. Because the entireprocess of emission. annihilation. and photon travel is nearlyinstantaneous, the photons can be detected simultaneously by an arrayof radiation detectors around the flow. Tomograp hic reconstructionmethods can thn be used to locate the radioactive tracers and image thephasc distribution of the floiv.

Parker ei ai. 1993, 1995) describe a PET s).stcm with hvo large,position-scnsitive detectors and have prcsented tests of the s).stem ontagged lubricating oil in a bearing test rig. Using a weightedbacirprojection method, good images were obtained of the oildistribution along 64 parallel planes in the rig: the time required toobtain the data for a full thrccdimensional tomogram was about anhour. whereas t\vo-dimensional projections required about a minute.Similar images were obtained of dough hydraulically extruded into amold over a 40-minute period. McKee et ai. (1995) used PET tovisualize slurry mixing in a cylindrical vessel. using sand labelcd withfluorine-18 as the solid phase. Comparisons of PET solids profiles\vith data from a local conductivity probe showed qualitative agreement;data acquisition requircd a period of 30 minutes under steady-stateconditions. Even with projected advances in data acquisition, PET islikely to be too slow to be useful in situations with time scales below afov seconds. A variation of PET is positron emission p m c l e tracking,or PEPT (Simons, 1995; Parker ei a/., 1993 and 1995). PEPT issimilar to the RPT method described above. A single tracer is placed inthe flow, an d the photon emission from the particle is used totriangulate the particle position. Because hvo orthogonal photons arecmitted by the tracer particle, fewer detectors can be used compared toconventional RPT.

Gamma Rav Densitometrv Tomoqraphv (GDTI.Gamma ray tomography, or gamma-densitometry tomography (GDT),uses collimated radiation sources to project photon beams through theregion of interest and detection systems to measure the fraction ofphotons that pass through the region unattcnuatcd. Measurement of theattenuation along many different paths through the domain of interestcan be combined with linear tomographic reconstruction methods toproduce an image of the phase distributio n in the domain. GD T has

X - W Imacaina. A number of medical x-ray systems can bused to image multiphase flows, and real-time images of flows apossible. These images can be digitized and analyzed to obtain bubblvolumes. velocities, and population statistics. Yates and Cheesma(1992) describe a Tstem used to investigate gas bubble growth ancoordinated with the imaging system which can acquire full images 25 frames per second. The ofiline image processing %stem ha s beeused to produce silhouettes of gas voids within the solid rcgimc thmay be used to obtarn local gas volume fractions using a deconvolutiotechnique.

breahvp in fluidized beds. The x - r q SOUTCX produces p u l s

Maqnetic Resonance lmaainq (MRI)Magnetic resonance imaging (MRI. othenvise known as nuclemagnetic resonance or NM R) is another tomographic method that hits origins in mcdicine. In MRI a strong magnetic field is applicd to thflow. causing nuclei to emit radio signals by prcccssing in phase withe ficld. By applying a magnetic field with h o t n gradients, nuclei differcnt arcas of the field nil1 precess at different frequencies. allowinthe positions of the nuclci to be dctcrmincd. In medical applications tfield is tuned to cause resonance of hydrogen nuclei (protons): the samapproach can be used in multiphasc flows involving water hydrocarbons.MRI has becn used successfully to measurc rheological propcrti

of Coucttc and Poiseuillc flows (Abbott er ai., 1991: Altobelli et a1991). In the latter study sleadl;-state pipe flow were analyzed fvariations in axial fluid velocity and particle conccntration: the workinfluid \vas a suspension of pol>m er particles in SAE gear lubricant. Ovelocity profiles were in agreement with anal!-tical predict ions , anparticle distributions in suspcnsion flows could be esplained in part bknown veloc ih effects and the velocity data. Flow velocities of lo-I d s ave been measurcd by MRI with a spatial resolution of IO-m turbulent conditions (Kosc, 1992). Altobelli and Mondy (1998) hasuccessfully used onedimensional imaging to measurc the settling glass microbubbles in flotation experiments. The microbubbles wethoroughly misedwith a suspending liquid of glycerol and water invertical qlin der. As the glas s spheres rose to the top of the suspensiover several hours, MRI data were taken continuously to measure tvolume fraction axial distributions and phase velocities with timMeasured solids settling functions were in agreement with both existicorrelations and hematic theory. Powell (1998) described an Msystem used to visualize flows of pulp suspensions, including velocprofiles in pipe flows. Generally, the length of time required to obtadata for a single image in MRI is greater than 30seconds.


5/14

Acoustic TomographyUltrasound can also be used to probe and image a multiphase flow.Dense liquid-solid flows have been visualized with acoustic holography.Shekaniz er ai. ( 1 998) describe such a system that has been developedfor examining dense particle suspensions. Unlike medical ultrasound

imaging systems, the plane that is visualized is perpendicular to thedirection of insonification. An acoustic beam is transmitted through themultiphase flow with a liquid bath as a coupling medium. The f o n w dscattered sound is focused onto a free surface and mked with the soundfrom a reference transducer that is in phase with the object beam, andthe acoustic intensity of the mixed signal is visualized as the surfacewave pattern The free surface becomes a diffraction grating withv q i n g amplitude proportional to the acoustic intensie. The surface isthen interrogated with the coherent light of a laser, and the resultingholographc image is captured for processing. The spatial resolution ofthis system is dependent on the acoustic wavelength of the interrogationbeam. For a 5-MHz interrogation beam, an in-plane spatial resolutionof < 1 mm is reported vith a depth of field of 6 mm (Shekarriz andBrenden. 1995: Shekarriz et ai., 998). Care must be taken to selectthe appropriate interrogation frequency for the multiphase flow understudy.

Electrical Impedance Tomoqraphv (EITlIn reccnt years the medical communi& has pioneered the USC ofelectrical impedance tomography (EIT) as a less expensive alternativeto conventional m d c d omographic systems. A sun.? of the basics ofthe mcthod and its application to medical lagnos tics is found in thebook Electrical Impedance Tomography. edited by J . C. Webster

(1990). R w c h e r s are now attempting to implement such qstexns forthe study of multiphase flows: Plaskowski et ai. (1995) describesseveral potcntial areas for application of the method. EIT employsmeasurements of the voltage at the boundary of a test domain toreconstruct the impedance dmribution within the domain. For ACelectrical conduction with field frequencies on the order of tens ofmegahertz or lower. the electrical potential within the domain. i': isrelated to the complex electrical conductivity of the domain, D. by

Current injections and voltage measurements are perfomed at a finitenumber of boundary electrodes. If N clectrodes arc used, thc domaincan be modeled as an N-port impedance network. The total number oflinearly independent voltage measurements, R.\r, is then given byN(N - )R.y = 2

Several simplifications are generally employed in EIT, w hich makeit possible to reconstruct the domain using discrcte finite elementmethods. These include the assumption that current travels only in iltwo-dimensional domain, the simple rcpresentation of conductivity by apiecewise constant function, and mathematical models to represent thevoltage in the domain. It is also assumed in EIT that the im pedancels t r ibut ion nithin the domain does not change significantly ove r timewhile voltage projections are acquired. As a result, EIT systems mustacquire projections rapidly to image most multiphase flows. T o

reconstruct the image of the domain, a "candidate" rcpresentation othe impedane distribution is first constructed. A se t of voltagprojections is then compu ted from the candidate distribution. Thcandidate projections are compared to voltages measured on thboundary during current injection. and their difference is characterizeby some error criterion. Finally. the candidate distribution is modifiebased on the error, and a new set of candidate projections is computeThis process continues until some minimum error criterion is satisfieEIT reconstruction is an ill-posed problem; consequently, threconstructed impedance "solution" may not satis& requirements oexistence or continuous dependence on the problem data. A prioknowledge of the impedance distribution may be used to aid irestoration, but in most multiphase f lows such information may not breadily available.D i c k et al. (1993) and Jmes er al. (1992, 1993, 1994) report thdevelopment of EIT systems implemented for the study of multiphaflows. A recent rcvicw is also provided by Ceccio and George (1996EIT qstems have been developed for both conducting and noconducting mixtures. and the number of electrodes ranges from 8 to 3Complete projection sets for 1Zilectrode systems can be acquired ovtime int e nd s of milliseconds.

CASE STUDY: VALIDATION OF AN ELECTRICAIMPEDANCE TOMOGRAPHY SYSTEMThis section prcscnts a series of studies in which an EIT systedevcloped a1 Sandia National Laboratories (O'Hem er al., 199Torcq-nski et ai., 1996a) and a GDT qstem were simultancousapplied to the same two-phase flows. Since the GDT system halready been successfully applied to multiphase flows (Adliins et a19 96 ; T o x i n s k i ai., I996b: Shollcnbcrger r f ai., 1997a), it wused to assess the accuracy of the EIT s)stem for two-phase flomcasurements. The Sandia EIT s)stem and the tomographreconstruction method are briefly described- and numerical aexperimental validation tests of the reconstruction algorithm apresented. The final sections of this paper describe the cxpcrimentconditions of the two-phase flow tests and co mpare results from and GDT. Both liquid-solid f lo w and air-water flows are examinewith close attention paid to the length cales of the dis pme d phase.

Electrical Impedance Tomoaraphv SvstemA description o f the Sands EIT system is presented in Torcqme! ai. (1997) and George et ai. (1998). The system consists of electrode array, or probe ring; a signal generator; a voltagecontrollcurrent source; multiplexers connecting the probe M g to the curre

source, ground. and measurement electronics: an instrumentatiamplifier and phase sensitive demodulators; and a data acquisition ca(Fig. I) . The data acquisition card contains an analog-to-digiconverter that measures the demodulated DC signals and a digicontroller that can be used to select electrodes for current injectioground. and measurement. The card also acts as an interface to a Pthat operates the entire system. During operation the EIT system injea controlled current through one electrode, "sinks" a second electrodegro und and measures voltages at all eleclrodes relative to ground. Tdomain is excited \kith a 50-M.z AC electric field: at 50 kHz, tresistive component of impedance dominates for air-water system


6/14

Figure 1. Block diagram of Sandia EIT system.

The voltage signals from each clectrode are passed through an amplifierto a phase-sensitive demodulator that separates the signal into hvocomponents: one in phase with the EIT carrier signal. the other aquadrature component. These components are lo\v-pass filtered to >ieldthe resistive and capacitive components. respectively. of the mea suredimpedance.Probe rings werc fabricated for us e in validation tests and inesperiments conducted in a transparent bubble column. Disk electrodes1.27 cm in diametcr and square electrodes 1.27 cm on a side werefabricated from stainless steel. These were mounted at equal azimuthalintervals in Lucite cylinders with an inner diameter of 19.05 cm. a m l lthickness of 0.64 cm, and a height of 38 . I aq wice the inner diameter(Fig. 2a). Another set of electrodes was fashioned from stainless steelstrips 0.64 cm wide. 7.62 cm high- and 0.0076 cm thick. mounted in aLucite cylinder 12.7 cm high (Fig. 2b). While the strip electrodes can intheory produce relatively two-dimensional electric fields, which arecasier to model: the point electrodes have higher mechanical integrityand good size-scaling properties. The m erits of both hpes of electrodeswere compared during tests and will be hscussed in following sections.

The reconstruction algorithm used with the Sandia EIT %stem hasbcen described in detail by Torcnnski er ai. 1996a. 1997) and is basedon the YWT method described by Yorkq er al. (1987 ). In thereconstruction algorithm, the electrical conductivity of the flowingmedium is treated as purely resistive (no capacitive contribution), whichis reasonable for the air-water ?stems considered here. The medium issurrounded by an insulating b o u n d q through which current is injectedor wi th dr aw a t discrete electrodes. A finite-element method (FEM)representation of the voltage equation is generated. and the electricalconductivity is represented as a function of position and one or moreconductivity parameters. To implement the N e u m m boundaryconditions, current flow is specified evenwhere on the boundary(specifically at the electrodes) such that the net current into and out ofthe domain is zero.Th e finite element equations are solved to find both the predictedvoltages at the electrodes and the derivatives of the electrode voltageswith respect to the conductivity parame ters. After the electrodevoltages are mapped as fimctions of the conductivitl; parameters, aNewton-Raphson algorithm is used to adjust the parameters tominimize the least-squares difference behveen the computed and

Figure 2. EIT electrode cylinders: (a) circular pointelectrodes, (b ) strip electrodes. The cylinders have the

same diameter; the bottom scale is in inches.

esperimental electrode voltages. With the converged parameters, conductivity distribution can be constructed, and phase informatiosuch as the domain-average gas volume fraction a, an be found froa constitutive model. Th e hw- r three-dimensional Maxwell relati(Maxwell, 1881) is most often used to relate the effective conductivof the mixture to the gas volume fraction of the mixture.Separate computer codes have bcen written to implement talgorithm in hvo and three dimensions. The hvo-dimensional coFEMEIT models arbitrary domains, including multiply connectdomains and domains with internal probes. Mesh nodes, which considered to be mathematical points. are used to represent


7/14

electrodes; current-bearing electrodes are modeled as line sources orsmks, extending to infinity out of the plane of the mesh. The globalconducti\iity functions are selected &om a library of choices in asubroutine that includes a constant conductivity, a circular insulatingregion at an arbitrw position and analytical conductivity distributions.FEMEIT generates and solves the FEM equations using theconductivity libraq and applies the Newton-Raphson algorithm todetermine the fmal conductivity parameters. For three-dimensionalcomputations the FEM equations are generated and solved using thef i nke l emen t code FIDAP (Fluid Dqnamics International, 1995).Fundamental FEM solutions have been determined for a range of valuesof the conductivity shape parameters, and library files have been createdthat map the electrode voltages in parameter space. To keep he size ofthe library manageable, only radial variations in the conductivity fieldare considered. The library file is used with the codes EITCON andEITAXI, which use cubic-spline interpolation and a Newton-Raphsonalgorithm to determine the final values of the shape parameters.EITCON solves for the gas volume Gaction EG by assuming a uniformcond uctiv ic across the domain, whereas EITAXI solves for coeflicientsof other conductivity distnbutions that can be described by theparameters, such as a second- or fourth-order conductivity profile, or aql ind nc al insulating inclusion ccntered in the domain. Thereconstruction algorithms have been validated using both anal!ticallyand physically derived projection sets. The complete system was alsovalidated with the detection of a series of inclusions in the test domain.Thcse results have been reported in Torcqnski et ai. 1996a, 1997).

Gamma-Densitometrv Tomonraphv SvstemA gamm ade nsi to mw tomography (GDT) ?stem was developedat Sandia for studies of industrial-scale multiphase flows (Torcgnsk i eraf., 996b: S hollenberger et al.. 1997a). The GDT systcm. s h o w inFig. 3: employs a j-curic 13Cs gamma sourcc, a sodium-iodidescintillation detector vstem, a computercontrolled traverse to positionthe source and detector. and data acquisition hardware and software.Measurements of gamma-ray attenuation along many different beampaths are translated into a gamma attenuation coefficient. p- averagedalong each path. Attenuation by the testbed walls is subtracted fromthis raw data. then the time-averaged gas volume fraction distribution inthe domain is reconstructed using the Abel transform - . g . , Vest(1985) - nd the assumption of an as isynm etric phase distribution,wh ch is reasonable for these experiments.

measurements. Controlled amounts of sodium chloride lvere added t\viita to obtain the desired conductivie. F~~ resistance tdomina& the mdummpedance, the folloI,,ng conslraint b

(3

Experimental Measurements of Two-Phase FlowsThe EIT system was applied to in situ measurement of two-phaseflows. Both liquid-solid and gas-liquid flows were studied to determine

the ability of EIT to measure volume-averaged phase fractions. In theliquid-solid tests the misture was assumed to have properties thatvaried smoothly over scales much larger than the size of the dispersedsolid phase but much smaller than the size of the test vessel. In the air-water bubbly flows. the scale of the dispersed phase (air bubbles) waslarger than in the solid-liquid misture but stili small compared to thescale of the apparatus. This represents a progression toward theapplication of EIT to industrial-scale multiphase flows.Liquid conductivity in these tests was selected to ensure Lhatresistive effects would dominate over capacitive effects in EIT

where f s the AC frequency, 2 is the normalized permittivity (odielectric constant) of water, and a s the permittivity of a vacuum . Fof = 50 I ~ H L n water. this constraint is satisfied if >> 2 pS/cmEsperimental conductivities were chosen to be at least one hundretimes lugher than this value and were adjusted to compensate for thdifferent impedances of strip and point electrodes.

Solid-liquid experiments. The first comparisons of GDand EIT (Shollenberger et al., 1997b) involved a closed container oconducting liquid with a flow of insulating particles. Schematdiagrams of the esperimental setup are shorn in Fig. 4. The testbewas created from the 16-point electrode ql ind er , which was capped


8/14

-9 cm -.7 cmFigure 4. Schematic of test geometry for the liquid-solid

experiments showing the EIT electrodes and impellergeometry.

the bottom and the top. A Sargent-Welch miser was inserted into theqlinder to stir the contents and generate a relatively uniform solidsdistribution throughout most of the flow inside the cylinder. The mixersystem consisted of a compact impeller assembly positioned 1 cmabove the lower cylinder cap, a motor mounted ab ove the testbed, and ashaft of 0.8-cm diameter connecting the impeller to the motor. Theshaft passed through a small center hole in the top cap of the cylinderaround which an "overflow" volume was placed to ensure the absence ofkce-surface effects- .g.. a vortical "funnel"- n the cylinder interiorduring mixing.

Deionized water with a known amount of dissolved sodiumchloride \vas chosen for the liquid phase. For the solid phase, glasspheres with a mean diameter of 80 pn were used. Glass is ainsulator with respect to salt water, and because of its higher densityglass attenuates gamma photons more strongly than water; thesproperties make it possible for both EIT and GDT to discriminatbem een the solid and liquid phases. Glass spheres are also rugged ancan easily be separated from water by settling. Variations in the locsolids density during mixing were on a scale much larger than the 80pm diameter of the particles themselves; this provided a smoothlv q i n g phase distribution.

To obtain a nominal qlinder-averaged solid volume fractioE S ' ~ ~ ' ,he required mass of spheres [vas computed from the volume othe cylindrical testbed, the volume of the spheres, and the knodensity of the glass. This mass was introduced into the testbed, anwater was added to f i l l the remaining volume. A mixer speed of 60rpm was applied for 30 minutes to all solid loadings, and a roughluniform distribution of particles was observed visually within the liquduring measurements. For solid volume fractions much in exccss o0.03, large fluctuating motions and variations in solids distributiocould be visually detected. so this study was restricted to valuescsroa'no larger than abo ut 0.03 (some solid volume fraction variationwere discernible. cven at this loadingj

The prescncc of the miser shaft posed problems for borcconstruction techniques. For GDT it produced extra attenua tion whethe gamma beam passed through the testbed centcrline. Thesanomalous points a.ere not used in performing the reconstruction. FEIT. placing a good conductor in the center of the testbed wousignificantly distort the electric field lines. so the steel shaft animpeller were coated with a la)er of insulating paint to mitigate thcffect. The presencc of a sma lldiam cter. a..is)mmetric. insulateinclusion was expected to have only a small effect on the electricbehavior of the ?stem: EIT measurements nit h and without the mixshaR in place. uith no particles prescnt. verified this assumption.For each test condition EIT vokage measurements were repeatduring mixing over a span of ten minutes, for a t o d of 640measurements at each electrode. GDT measurcments were also talrduring mixing for another four minutes. Mixing was then terminateand the spheres were allowed to settle to the bottom of the water-fillq l indcr , a proccss that was completed after five minutes. Followinthis settling period EIT was applied again. The second Emeasurement was necessary for calibration purposcs because tconductivity of the water was altered by soluble contaminan

introduced as the spheres were add ed Although these tracontaminants had a negligible effect on GDT, their effect on the watconductivity was comparable to that of the suspended solid particlduring mixing.Table 1 summarizes the measurements by GDT and EIT for thrdifferent solid volume fractions. Analysis of the GDT dademonstrated that the radial solids profiles were relatively uniform,desired. Th e ratio of the cylinder-average d miVture gamm a attenuaticoefficient to the liquid attenuation coefficient p~ was found increase monotonically with the nominal solid volume fractionsimilarly the ratio of the average miyture conductivity to the liquconductivity, @/aL) decreased monotonically with ncreasing &?OMBy assuming a spatially uniform gamma attenuation coefficient, GD


9/14

Table I.Comparisonof nominal solid volume fractionsand values measured by GDT and EIT in the liquid-solid

experiment..5sXVoM pirlpr sGDT 1 ala{. SmI -0.010 1.015 0.011 0.982 0.0120.020 1.030 0.021 0.968 0.02 10.030 1.057 0.040 0.940 0.04 1

0.05E0c.

0.042E.ea2 0.030>

-0 - EIT................. ................

.....

8_ _ . . . . . . . . . ./;. . . . . . . . . . . . .~ . . . . . -U301 0.01mE

0.000.00 0.01 0.02 0.03 0.04 0.05

nominal solid volume fractionFigure 5. Comparison of solid volume fractions

measured by EIT and GDT.

Lt D

Figure 6. Schematic diagram of air-water bubble columnused fo r EIT and GDT comparisons.

fraction was lcss than the nominal value near the top of the ql in de r anlarga than nominal toivard the bottom.data were converted to a qlinder-av erage. solid yolume fraction c$DTthrough the formula

where the attenuation cocffcients of the water and glass spheres werepreviously measured. The t h r e c h e n s io n a l Maxwell relation wasused to transformEIT data to an average solid volume fraction am :

The conductivity distribution was assumed to be uniform across thedomain.The solid volume fractions determined by GD T and EIT are seento be in close agreement with each other for all cases (see Fig. 5) andwith the nominal values for the first two cases. The case of CS ~ =0.03 is interesting in that the GD T and EIT values are in agrementwith each o ther but are somewhat higher than the nominal value. Themixingwas not strong enough to produce a uniform axial distribution ofglass spheres throughout the cylinder in this case; the solid volume

Liquidqas experiments. Liquid-gas tats of the EIT?stem \%ere conducted in a transparent bubble column assembled aSandia as a tcstbcd for optical, electrical. and radiation-basemultiphase flow diagnostics (Torcqnski et al . , 1997). The Lesacolumn. show schematically in Fig. 6, has an inner diameter D o19.05cm, wall tfuchess of 0.64cm, nd is built from interchangeablsections so that different diagnostic tools c an be placed in the columnIn this stu& the EI T electrode sections were placed near the center othe column. The point electrodes formed a plane at a distance L o109.86 cm above the column base, whereas the 16 strip electrodes wercentered on a plane L = 97.16 cm above the base. The column wafilledwith water to a depth H, f 1.45 tq for a height-todiameter ratiof 7.6: this depth placed the region of EIT sen sitiv ic completely undewater. Air can be introduced at flow rates up to GOO L/min through onof several interchangeable spargers at the base of the column. Thspargers are electrically isolated to prevent interference with the EIsystem. The column operates at ambient conditions, but the water isubject to evaporative cooling as air is bubbled through; the liquitemperature was held constant by active heating to within kO.2Cduring experiments. which limited variations in conductivity wittemperature to 20.3%.The goal for the air-water tests was to validate EIT in a flow witvariations on a larger scale than the liquid-solid flows. First, a nearl


10/14

Figure 7. Ring sparger used in chum-turbulent flowexperiments.

homogeneous bubbly flow was produced within the column. Twodiffercnt sparger designs for homogeneous flows wxe tested. andcontrolled amounts of surfactant were added to the working liquid toreduce surface tension and prevent coalescence. A truly homogeneousbubbh- flow could not be attained with either sparger. GDTreconstructions of the radal gas volume fraction profiles showed aquartic pattcm. indcating a low gas volume fraction at the columnCenter from both sparg er designs. Vortical motions were also observedat the higher flow rates. indcating a transition from homogeneous tochurn-turbulent flow..Xmt. churn-turbulent flows were examined withEIT and GDT. Such flows in vertical columns normally have parabolicvolume fraction profiles that can be reconstructed by both the GDT andEIT algorithms. The sparger used to produce chum-turbulent flows isshorn in Fig. 7. This sparger is a hollow stainless steel toroid with a10.16-cm ccntcrline diameter. an inner tube diameter of 0.95 c m andten holes of diameter 0.16 cm facing downward.For each flow condtion 25 full EIT projection sets were averagedto acquire the voltage data for reconstructions; each projection setrequired less than 0.75 seconds. This averaging was required in orderto cornpan: the EIT results with the time-averaged GDT results. Theconductivity of the water used in the final liquid-gas experiments was5 6 8 ~ 6 S/cm for point electrode tcsts and 285?5 pS/cm for stripelectrode tests. The source-detector plane of the GD T system wasplaced to avoid interference from the elcctr&. For tests using pointelectrodes, all gamma paths were in a plane located at L = 96.00 cmabove the column base. 13.86 cm below the EIT electrode plane; themeasurement chords were parallel and spaced 1 cm apart. During testswith the strip electrodes, GDT measurements were taken in a plane81 OO cm above the co~umnloor.Air-water experiments were performed using the ring sparger atfive flow rates: 25 > 50 . 75 > 100 and 150 Wmin. The correspondingsuperficial gas velocities ranged from 1.5 to 8.8 c d s . Figure 8 showsthe flow conditions in the column for the minimum and maximum flowrates. A t the lowest flow condition, a range of spherical and coalescednonspherical bubble sizeswas evident. A central vortical bubble streamwa s dso seen. similar to the vortical-spiral flow regime described byChen tv al. (1994) between dispersed bubbly flows and full turbulence.

Figure 8. Chum-turbulent flow conditions in thetransparent bubble column at minimum and maximum flow

rates: (a) Q = 25 Umin, (b) Q = 150 Urnin.


11/14

Table 2. FundamentalEITvoltage solutionsfor 16 strip electrodes in the transparent bubble column.Experimental value

0.2946540.1454480.08029700.0443 1860.009430460.002274740.0225977

Fundamental voltage Computational value Experimental value X 1.00688Vg (injection) 1.32746 1.23243 1.24091Vlr=,V5I.;v3

t'l

0.2929550.1443450.07962770.04392 160.0225 1900.009480880.00229490

0.2949700.1453380.080 17540.04422370.02267390.009546090.0023 1069

At rates of 75 Umin and above, the flow was completely opaque andturbulent to the q e .As an aid in maluating the results. the fundamental EIT voltagesolutions for an axinmetr ic conduct iv ih prof i le wre computed.[The reader is referred to Torcqnski et al . (1997) for a complete

discussion of fundamental voltages.] Because of a s i p m m e e acomplete set of R.Y voltage measuremaus \vi11 yield only N/2independent pieces of information in the fundamental solution. Theexperimental voltages were also reduced to fundamental values forcomparison nith the computed solutions. The reduced voltages fromthe point electrodes were in poor a g r m e n t itith predicted values: at allnoninjection. nongound electrodes. the cspenmental voltages weresimilar. indicating poor sensitivity to variations in the conductivivprofile. Conversely. data measured with the strip electrodes were inexcellent agcement with the predicted values and were much moresensitive to the conductiviv profile. As seen in Tablc 2, when amultiplicative factor of 1.00688 is uscd the computational andexperimental voltages a g e to almost th m significant figures. Adifference of 0.7% in computational and expenmental voltages at allelectrodes is reasonable. since experimental martainty in AC currentand liquid conductivity are about 0.5% . and computational accuracy issimilar. Basedon these results strip electrode data was used exclusivelyto reconstruct radial profiles of thechum-turbulent flows.In both the EIT and GDT reconsuuctions. the radial volumefraction distribution \\-as epresented by panmetric formulas. Theconductivity distribution in EIT reconstructions was assumed to beparabolic across he column. taking the form

crGL b

I + cf2(r/R)' - ]- =

where h is a scale parameter, c is the parabolic shape parameter, and Ris the radws of the bubble column. Similarly chord-averaged gasvolume fractions computed from GDT data were fit to a second orfourth-order polynomial before the Abel transform was taken. Withboth EIT and GDT, the column-average gas volume fraction EC wasdetermined by analytically averaging the resulting profile over thecolumn area. Figure 9 shows tqpical gas volume profiles and thecolumn-average gas volume fractions as a function of flow rate,determined from both methods. The EIT profile tends to lie slightlyabove the G DT profile at each flow rate. The GDT and EIT gas volumefractions are in excellent agreement, however. id1 lying within 0.01

(absolute volume fraction) of each other at all radial positions, despivariations in the local volume fraction across the column as large as 0.Because of the difference in collection t h e s for the hvo metho(about 23 minutes for GDT but less than 20 seconds for EIT), it wdecided to look for slow cyclic behavior in the flows which might baveraged out by GDT but not by EIT. Separate flow measurementwere laken with an impcdance-based bulk void fraction meter. s h o w Fig. 10. This instrument has two rectangular clcctrodcs 3.8 cm tawhich subtend 120" each on opposite sides of the column. A 50-kAC signal is applied to the electrodes. and the voltage across thelectrodes ma). be rccorded with a digitizing oscilloscope for analysIn earlier work h s meter was uscd to measure bulk gas volumfractions through comparisons of signals from flow and no-flocondrtions (Torcq-nsk et ai.. 1997). Hen: Fourier spectra and signato-noise ratios were computed from the rccorded voltages to determithe frequencies and magnitudes of void fraction fluctuations.Voltages were recorded over periodsof ten minutes under he samflow conditions as thc GDT-EIT tests. Signal spectra revealed ncoherent periodic behavior in the flow over the range of z r o to 12Hz: thii dlcviatcd the concern that EIT rncasurements were capturindata over only a portion of slow flow qc le s. Signal-to-noise ratios flow and no-flow conditions. along with the assumption of Gaussierror propag ation were used to compute voltage variations due to flofluctuations. By correlating mean voltages \sith average gas volumkactions. variations in a ver the course of the measurement were thestimated. Table 3 presents the estimated fluctuations at each flow raexcept for the lowest flow rate, fluctuations are higher than tdiscre panq behveen EIT and GDT, again suggestingthat both methoare incorporating such variations in theirmeasurements.In s u m r n q , the EIT %stem has been validated for measuremeof solid volume fractions in solid-liquid flows, and gas volume fractioand radial profiles in gas-liquid flows. Future work will include tmeasurement of gas-liquid-solid flows in the bubble column o examithe effect of the solid phase on hydrodynamic behavior. The study wemploy a solid phase with conductive properties similar to air anddensity similar to water, so that EIT wi l l detect both the solid and gphases but gamma attenuation will only b e d u e n c e d by air,as befoThe difference in the radial profiles Erom the GDT and Ereconstructions udl yield the radial solids distribution in the three-phaflow. The solid volume fraction in these three-phase tests will be wabove 0.01, based on the difference in gas volume fraction results fiotwo-phase tests.


12/14

0.30

0.25

0.20

0.15

0.10

0.05

0.00-1.0 -0.5 0.0 0.5 1.0

r/R(a1

0.30 I ' . ' I f0.20

0.15

0.10

0.05

0.00-1.0 -0.5 0.0 0.5 1.0

rl R(c)

0.30 I J0.25

0.20

' 0.160.10

0.05

0.00-1.0 -0.5 0.0 0.5 1.0

r/R(e1

-.--:--..

I " ' I; El ;. . . . . . -_.__... . ... .. . .... . . . . . , . . . . . .

Date post:	14-Apr-2018
Category:	Documents
Upload:	uka-oguh
View:	222 times
Download:	0 times

Advanced material distribution measurement in multiphase flows: A case study

Documents