Flow mapping in bubble columns using CARPT

Chemical Engineering Science, Vol . 45, No . 8, pp . 228 5 -2291, 1990 . Printed in Great Britain .

0009-2509/90 $3 .00 + 0 .00 Q 1990 Pergamon Press plc

FLOW MAPPING IN BUBBLE COLUMNS USING CARPT

N . DEVANATHAN,*+, D . MOSLEMIANt and M .P . DUDUKOVIC* *Chemical Reaction Engineering Laboratory, Washington University, St . Louis, MO 63130 tDepartment of Mechanical Engineering, Florida Atlantic University, Boca Raton, FL 33431 +Currently with Amoco Oil R&D, Naperville, Illinois 60566

ABSTRACT A noninvasive Computer Automated Radioactive Particle Tracking (CARPT) Facility is used for the investigation of liquid recirculation and turbulence in a bubble column . The motion of a single neutrally buoyant radioactive particle is monitored by an array of scintillation detectors and analyzed by an on-line computer to map the flow field . Results for the mean flow patterns and turbulence in a 12" diameter column are reported for the air-water system . This communication marks the first application of CARPT for tracing the liquid flow in bubble columns .

KEYWORDS Two-Phase Flow ; Bubble Column ; Liquid Circulation ; Turbulence ; Particle Tracking .

INTRODUCTION Liquid circulation or gulf streaming is a phenomenon commonly encountered in bubble columns caused by nonuniform holdup profiles . It is primarily responsible for liquid phase mixing (through convection and turbulence) and affects interfacial mass transport (through coalescence dependent interfacial area) and heat transfer to the column walls and immersed tubes . The assumption of complete backmixing can lead to overdesign, if high conversions are desired, or to reduced selectivity in the case of complex kinetics . The accurate design and scale-up of bubble columns is, in part, contingent upon the ability to describe the non-ideal flow of the liquid phase . How this gulf streaming is affected by column size (diameter and height), distributor design, physical properties of the gas-liquid mixture and the operating conditions (superficial velocities) is presently not well understood . Neither a theory based on first principles nor abundant data are available . The inviscid models (Whalley and Davidson, 1974 and Joshi and Sharma, 1979) make use of the single phase vorticity transport equations which are not applicable for two phase flow . The "principle" of minimization of maximum vorticity used to close the problem has no physical basis . Joshi and Sharma proposed multiple circulation cells in tall columns (with cell height equal to column diameter) without confirming their existence . Further, based on the observation by numerous investigators that the continuous phase flows upward at the center and downward at the wall, they assigned the same clockwise rotation to cells above each other, which is physically impossible . While the information on the existence of such cells is vital for development of a model for liquid mixing, it can only be obtained by determining the overall flow patterns of the liquid phase . Liquid circulation has also been described by 1-D momentum balances for the gas and liquid phase by Ueyama and Miyauchi (1979) among others . These models require the void profile and eddy viscosity as model inputs . Different constitutive relations have been proposed for the eddy viscosity ranging from a purely empirical correlation (Ueyama and Miyauchi, 1979) to turbulence models based on single phase mixing length (Clark et al ., 1987) and von Karman's hypothesis (Anderson and Rice, 1989) . The applicability of these models can be assessed only if Reynolds stress measurements are available . While the currently available one-dimensional models are unable to predict the gas void fraction, turbulence information can be used for this prediction (Drew and Lahey, 1982) . Experimental measurements of liquid velocities are limited . The most useful data for model verification has been that of Hills (1974) who used a pitot tube . Hills velocity and voidage profiles satisfy the liquid continuity within 30 percent, (Devanathan, 1990) . Most of the current techniques are invasive and affect the flow pattern locally . Noninvasive techniques such as Laser Velocimetry which have worked very well in single phase flows, cannot be used in gas-liquid two phase flows at high gas holdup as the bubbles scatter the light making signal discrimi .nar_ion very CES 451* -W

2285



B15

N . DEVANATHAN et al.

2286

difficult . Except for some limited measurements reported by Lubbert (1983), turbulence information in bubble columns is virtually non-existent . Clearly, there is a need for systematic experimental investigation of mean recirculation and turbulence . The information could be used for evaluation of the existing models and development of new ones for reliable scale-up .

Principles of CARPT Originally developed by Lin et al . (1985) the CARPT Facility has undergone a number of refinements as summarized by Devanathan (1990) . A single radioactive particle emitting gamma radiation of constant energy, which is dynamically similar to the recirculating phase, is introduced into the column . As it moves along with the recirculating phase, the particle is tracked using an array of scintillation detectors located around the column . The frequency of gamma rays arriving at each detector decreases with increasing distance between the source and the detector . The photon count rate (the intensity) is related to the distance between the source and the detector using pre-established calibrations . The instantaneous position of the tracer is then accurately calculated from the distances using an optimized linear regression scheme . Prudent use of the purposely introduced redundancy in distance measurements helps overcome the intrinsic noise associated with the quantized nature of the gamma emission . Time differentiation of the displacements yields Local velocities . Correspondingly, ensemble averaged velocity distributions and other turbulence quantities can be computed after acquiring the data for a sufficient length of time . The radioactive particle was fabricated by embedding a Scandium cylinder (1 mm diameter, long) in a 2 .4 mm polypropylene sphere such that the entire particle had a density of 1 making it neutrally buoyant in water . The particle was then sent to the Reactor Research University of Missouri, Columbia, and activated to Sc-46 with a source strength of 200pCi half life of 84 days .

0 .7 mm .01 gm/cc, Facility . and a

The intensity (I) vs . distance (r) calibration curves are obtained by placing the source at various points in the column and recording the time averaged intensities of each detector . The following polynomial is used : 2 r i = f (i ) = a0i + all (!T-) + a?1 (-I!--) +

(1)

The error due to void fluctuations is minimized by performing in-situ calibration for each superficial gas velocity . As the tracer moves around the column, its position is monitored continuously using the array of 16 scintillation detectors . In a typical experiment, the sampling rate is 33 Hz and the data are collected for a period of 5 hours . Using the calibration curves, the intensities recorded by the detectors are converted to distances between the tracer and each detector . The distance between the tracer and i-th detector is given by : (~E-xi)2 + ( n-y i ) 2 + (r,-z i ) 2 = ri

(2)

Here x i , y i and z i are the coordinates of the center of the Nat crystal of the detector . The redundancy in data from multiple detectors is used to obtain the optimized coordinates of the tracer . E, rl and S . A weighted linear regression scheme used for this purpose is described by Lin et al . (1985) and is not repeated here for brevity . Once the successive tracer locations are obtained the instantaneous velocities are calculated by time differentiation . The column is divided into sampling compartments . By recording the repeated appearance of the tracer in a compartment, the ensemble average of each component of the velocity can be computed . A measure of stagnancy of the liquid within the column can also be obtained by accumulating the total number of groups of tracer occurrences in a compartment for which the occurrences are successive in time . TheBubble ColumnFacility The CARPT Facility, shown in Fig . 1, consists of the column, the detector support structure and the signal processing and data acquisition system . The column makes use of a Plexiglas plenum which can accommodate test sections of various diameters . A 0 .635 porous plate distributor (stainless-steel or tetra-glass, with an average pore size of 40 pm) is sandwiched between the two sections . A positioning device was fabricated and attached to the top of the column for calibration purposes . The versatile detector support structure can accommodate different configurations . Air to the column is supplied by a compressor and the flow rate is monitored by three parallel rotameters . More information on the column, detector support structure and the positioning device can be found elsewhere (Devanathan, 1990) .

B15

Flow mapping in bubble columns using CARPT

Fig . 1 .

2287

Overview of the CARPT Facility showing the column and the detector support structure .

SignalProcessing and Data Acquisition System Gamma photons striking the detector are converted to current pulses (one pulse for each incident photon) which are amplified by fast filter amplifiers . The amplified pulses are then filtered by a discriminator which removes the noise due to secondary gamma radiation generated by the interaction of primary gamma radiation with the column and its contents . A binary pulse counter then counts the pulses which are then transferred to the hard-disk of the computer under machine language software . The data acquisition system makes use of the modular CAMAC (Computer Automated Measurement and Control, IEEE-583) high speed data acquisition in conjunction with the GPIB (General Purpose Interface Bus) for rapid acquisition (Devanathan, 1990) . RESULTS AND DISCUSSION Operating Conditions Results are presented for a 0 .292 m internal diameter bubble column operated in the batch liquid mode with tap water as the liquid phase . The distributor used is the porous stainless steel plate mentioned earlier . The static liquid height was 0 .584 m giving an aspect ratio of 2 . The superficial air velocity was 0 .105 m/s . Under these conditions, the liquid height in the column was 0 .711 m yielding an average bed void fraction of 0 .18 . Visually, the column appeared to be operating in the churn-turbulent flow regime with a wide bubble size distribution . Calibration For calibration, the tracer was positioned at 185 different spatial locations within the column and the time averaged intensities were recorded . Figure 2 shows a typical intensity distance relationship and the corresponding polynomial curve fit . Once the calibration was completed, the particle was dropped into the column and it was allowed to move along with the liquid . Mean Liquid Circulation Profiles Figure 3 presents the velocities vector and stream function time averaged over the column crosssection . The existence of a single recirculation cell with the liquid ascending along the column center and descending along the wall is apparent . This observation is contrary to the hypothesis on the existence of multiple circulation cells with cell height equal to the column diameter in bubble columns (Joshi and Sharma, 1979) . The maximum velocity occurs at the column center at a height of 0 .58 m and is 0 .52 m/s . By examining the flow pattern along two sectional planes in the column, it was possible to conclude that the flow in a bubble column is in general asymmetric, although cylindrical averaging makes the data tractable and clearly indicates mean flow patterns . Similar results were obtained in smaller diameter bubble columns (0 .114 and 0 .190 m internal diameters) at high superficial gas velocities (Devanathan, 1990) . However, at gas velocities less than 0 .05 m/s, two recirculation cells were observed . This is not shown pictorially here



2288

N. DEVANATHAN

Fig .

2.

B15

et al.

Calibration data and polynomial fit for detector 7 .

0

In r

0 0 Co

f

.

t

f

.

t

t

.

t

f t f

. . .

f

t t E Q U

X Q

i

r

t t

t t t t t t

.

R

4

4

1



.

I

.

.

,

.





-,

.

T 1 0

1

+ . r . .

4 t

0U y 0 d

. .

. .

r r . e

. . ,

M

. .

4

O 0 5

10

Radial Position, cm

Fig . 3 .

15

0

5

10

16

Radial Position, cm Vmax = 52 cm/s

Streamlines and velocity vectors in a 12" column at 0 .105 m/s gas superficial velocity .

due to space limitations . In the lower cell (which is confined to the entry region), the liquid ascends at the wall and descends at the column center . The flow reverses itself in the upper cell (which occupies the rest of the column), a behavior which was also observed in gas fluidized beds (Lin et al ., 1985) . The lower cell completely disappears at higher velocities . Figure 4(a) shows the axial velocity profiles at various axial locations . These profiles are qualitatively similar to the data reported in a smaller column by Hills (1974) . Profiles are relatively flat close to the distributor and the free surface . The entry region for the establishment of fully developed profiles is approximately one column diameter . Radial velocities are typically small except in the reversal regions close to the distributor and the free surface (Fig . 4b) . This fact supports the assumption that the flow in a bubble column can be simplified as one-dimensional if end effects are neglected .



2289

Flow mapping in bubble columns using CARPT

B15

(b)

(a)

0 0 m

Z Levels, cm 0 - 5 .43 a = 19.36 + = 33.33 x = 47 .26 0 = 6123

.r, y~ o ~+ 0

Axial Position . cm

0 0 0 .0

5 .0 10 .0 Radial Position, cm

15.0

Fig . 4 . (a) Radial variation of axial velocity at various z locations, and (b) Axial variation of radial velocity at various r locations .

Figure 5 shows the stagnancy pixel plot . Each compartment was assigned a particular color depending on the total number of tracer occurrences in a given comparment for which occurrences are successive in time . In the black and white photo shown below the darkest shade represents the most stagnant region followed by lighter shades in the order of decreasing stagnancy . Unfortunately, the black and white representation is a poor substitute for the color version . The map shows that stagnant regions are confined to the middle of the circulation cell at the top and at the bottom . The central portion of the column which is well aerated by the bubbles is active . This diagnostic capability of CARPT is useful for detecting pathological behavior in systems with different internals and distributors .

Fig . 5 . Stagnant region map in a bubble column .



2290

B15

N. DEVANATHAN et al.

Liq id Pha e T

b lence

eloci ie , he fl c a ing eloci Once he mean eloci ie a e comp ed f om he in an aneo componen a e ob ained . The e can be ed o comp e he Re nold ' e e , he kine ic ene g he in i of b lence . A one dimen ional anal i of he da a a pe of b lence and en med b a e aging he flo a iable in he a ial di ec ion abo e he en egion and belo fo .62 m) . Fig e 6(a) ho he a iall a e aged he f eeface egion (be een = 0 .38 m and 0 e i h he da a of Hill (1974) . The an i ion poin liq id feloci p ofile hich compa ell adial loca ion he e liq id a ial eloci become e o) occ a = 0 .1025 m o (i .e ., ell i h he da a of Hill (1974) . Fig e 6(b) ho he a e aged /R = 0 .72 and ag ee e e p ofile fo ingle pha e flo Re nold ' p ofile hich peak clo e o he Re nold ' e 6(c) depic he edd i hich h o gh a pipe (Bi d e al ., 1960) . Fig i co a comp ed e he eloci g adien ( he g adien i elf a comp ed b b di iding he Re nold ' b fi ing he eloci p ofile o a pol nomial and diffe en ia ing i ) . Nik ad e' da a fo 10 6 (f om Schlich ing, ingle pha e flo in a moo h pipe fo a Re nold ' n mbe of 1 .1 in beha io i ema kable gi en he 1979) i al o ho n in he ame fig e . The imila i he e ogeneo na e of o pha e flo . (a)

(b)

0 O O 0

0 O m

0 O

Line : Polynomial Fit

O 0 O a

LEGEND ⊂ = Single

Pha e Da a e = T o Pha e Da a

0 0

~E 2

N

e

N

E

(c)

al

N d O O

O

N d O

U 0 d O

0 0 _ O N

. I t

o U O

X

T a LU w

o

0 CC Q a] O

0

N

0

O O a

0 .0

I 3 .0

1 6 .0

1 9 .0

1 12 .0

Radial Po i ion. cm

O O

15 .0

0 .0

3 .0

6 0

8 .0

12 .0

Radial Po i ion, cm

15.0

I

00

3 .0

I 6 .0

9 .0

15.0

120

Radial Po i ion. cm

Fig . 6 . One dimen ional ep e en a ion of da a, a) a ial eloci p ofile, b) Re nold hea e p ofile and c) edd i co i p ofile . O he q an i ie ha ha e been comp ed (De ana han, 1990) incl de he Lag angian a ob len di pe ion coefficien co ela ion , he in eg al ime cale and he adial and a ial ( hich acco n fo liq id mi ing b he eddie ) . SUMMARY ing he CARPT Facili o map he flo field in a b bble col mn, i ha been demon a ed he p ima flo pa e n i a ingle eci c la ion cell . Ho e e , a mall econda cell ha ma e i clo e o he di ib o a ce ain eloci ie (po ibl d e o di ib o effec ) . In addi ion, econda comple a mme ic flo ma e i . The en egion i de e mined o be app o ima el one col mn diame e . The one-dimen ional eloci p ofile compa e ell i h he e p ofile i imila o he co e ponding ingle pha e da a of Hill (1974) . The Re nold ' flo p ofile al ho gh he magni de of he e i m ch la ge . We a e c en l in e iga ing pe ficial ga eloci on eci c la ion and he effec of col mn diame e and b lence . B

ACKNOWLEDGMENT Thi e ea ch i ppo ed b Na ional Science Fo nda ion h o gh he g an CBT-8707843 and CBT-8820555 . The a ho g a ef ll ackno ledge he e ice of he Machine Shop a he Depa men of Mechanical Enginee ing a Flo ida A lan ic Uni e i .

-

REFERENCES Ande on, K .G . and R .G . Rice, (1989), "Local T b lence Model fo P edic ing Ci c la ion Ra e in B bble Col mn ", AIChE J ., 35, 514-518 . Bi d, R .B ., W .E . S e a and E.N . Ligh foo , (1960), T an po Phenomena, John Wile , Ne Yo k . Cla k, N .N ., C .M . A kin on and R .L .C . Flemme , (1987), "T b len Ci c la ion in B bble Col mn ", AIChE J ., 33, 575-578 .

B15

Flo

mapping in b bble col mn

ing CARPT

2291

De ana han, N ., (1990), In e iga ion of Liq id H d od namic in B bble Col mn Via A Comp e A oma ed Radioac i e Pa icle T acking Facili , D .Sc . The i , Wa hing on Uni e i , S . Lo i , MO . D e , D .A . and R .T . Lahe , (1982), "Pha e Di ib ion Mechani m in T b len Lo -Q ali T oPha e Flo in a Ci c la Pipe," J . Fl id Mech .,117, 91-106 . Hill , J .H ., (1974), "Radial Non-Unifo mi of Veloci and Voidage in a B bble Col mn," T an . In . Chem . Eng ., 52, 1-9 . Jo hi, J .B . and M .M . Sha ma, (1979), "A Ci c la ion Cell Model fo B bble Col mn ," T an . In . Chem . Eng ., 57, 244-251 . Lin, J .S ., M .M . Chen and B .T . Chao, (1985), "A No el Radioac i e Pa icle T acking Facili fo Mea emen of Solid Mo ion in Ga Fl idi ed Bed ," AIChE J ., 31, 465-473 . L bbe , A ., (1983), "T b lence Mea emen in B bble Col mn ," in Ma T an fe i h Chemical Reac ion in M l ipha e S em , (ed : E . Alpe ), Ma in Nijhoff P bli he , Bo on, MA, 553-564 . Schlich ing, H ., (1979), Bo nda La e Theo , McG a -Hill, Ne Yo k, pp . 602-609 . ama, K . and T . Mi a Ue chi, (1979), "P ope ie of Reci c la ing T b len T o Pha e Flo in Ga B bble Col mn ," AIChE J ., 25, 258-265 . Whalle , P .B . and J .F . Da id on, (1974), "Liq id Ci c la ion in B bble Col mn ," P oc . S mp . T o Pha e Flo S em , In . Chem . Eng . S mp . Se ., 38, Pape J5, 1-29 .