Flow Pattern Assessment in Tubes of Reciprocating Scraped

8/11/2019 Flow Pattern Assessment in Tubes of Reciprocating Scraped

1/13

Flow pattern assessment in tubes of reciprocating scrapedsurface heat exchangers

J.P. Solano a , * , A. Garca a, P.G. Vicente b, A. Viedma aa Universidad Politcnica de Cartagena, Departamento de Ingeniera Trmica y de Fluidos, Campus Muralla del Mar, 30202 Cartagena, Spainb Universidad Miguel Hernndez, Departamento de Ingeniera de Sistemas Industriales, Avenida de la Universidad, 03202 Elche, Spain

a r t i c l e i n f o

Article history:Received 28 May 2010Received in revised form20 October 2010Accepted 24 November 2010Available online 26 January 2011

Keywords:SSHEPIV Flow patternFriction factorCFD

a b s t r a c t

Flow pattern in the tubes of an innovative scraped surface heat exchanger with reciprocating scrapershas been experimentally investigated. The scraper consists of a concentric rod inserted in each tube of the heat exchanger, mounting an array of semicircular plugs that t the inner tube wall. A hydraulicpiston provides the scraper with constant-velocity reciprocating motion. Phase-averaged velocity eldshave been obtained with PIV technique for both scraping semi-cycles, with special emphasis on the effectof the scraping velocity (velocity ratio) and Reynolds number. Visualization results have been contrastedwith experimental data on Fanning friction factor, obtaining a clear relation between ow patterns,pressure drop augmentation and turbulence promotion. CFD simulations for quasi-steady laminar owprovide a further insight into the relation of the ow structures with wall shear stress, and the contri-bution of pressure forces to global head losses, for each scraping semi-cycle.

2010 Elsevier Masson SAS. All rights reserved.

1. Introduction

Scrapedsurface heat exchangers (SSHE) are commonly employedfor heat transfer and crystallization processes in the food, chemicaland pharmaceutical industries. These mechanically assisted devicesarespecially suitedfor productsthat areviscous,stickyor thatcontainparticulate matter. During operation, the product is brought incontact with a heat transfer surface that is continuously scraped,thereby exposing the surface to the passage of untreated product [1].

Highheat transfer coef cientsareachievedbecausethe boundarylayer is continuously replaced by fresh material, while heat transfersurfaces remain clean. In addition to maintainhigh and uniformheatexchange, the scraper blades also provide simultaneous mixing andagitation, of utmost importance for laminar ow heat transferenhancement [2]. Most commercial designs have a rotating shaft inthe center with the product being pumped through the annular gapbetween the shaft and the outer cylindrical heat transfer tube. Flowand shearing pro les have been widely studied since the rst worksof Trommelen and Beek [3]. In uence of rotational speed and pres-ence of blades has been analyzed in the remarkable works of Stran-zinger et al. [4] and Dumont et al. [5], clarifying three differentregimes characterized by pure shear ow, steady toroidal vortices

and unstable, wavy vortices. Extensive reviews covering designaspects, heat transfer characteristics and power consumption of SSHE s can be found in Harrod [6] and Rao and Hartel [1].

This work presents an innovative concept of scraped surfaceheat exchanger. Within each tube is a concentric reciprocating rod,mounting an array of semi-circular elements with a pitch P 5D(see Fig. 1). These elements t the internal diameter of the tubes.During the reciprocating motion, they scrape the inner tube wall.Additionally, the movement of the insert device generates macro-scopic displacements of the ow, that continuously mix coreregions with peripheral ow. As a result of the mentioned features,the reciprocating scraped surface heat exchanger (RSSHE) provideshigh overall heat transfer coef cients, and prevents down time forcleaning operations. Depending on the severity of the foulingphenomenon, the scraper can be either activated intermittently, ormove continuously with adjustable scraping frequencies.

Commerciallyavailable versions of this RSSHE are manufacturedby HRS-Spiratube S.L. under the brand UNICUS Dynamic HeatExchanger, and by Alfa Laval Inc, with the brand Viscoline DynamicUnit. Along with the industrial evidence of fouling reduction, themanufacturers claim for its higher heat transfer area, smallernumber of shutdowns, lower induced shear stress in the product,and particle integrity in food applications, compared to rotatingscraped surface heat exchangers. RSSHE s are progressively beingintroduced in the food industry, wastewater treatment processes,production of second-generation biofuels, etc.

* Corresponding author. Tel.: 34 968325938; fax: 34 968325999.E-mail address: [email protected] (J.P. Solano).

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

j ou rna l homepage : www.e l sev i e r. com/ loca t e / i j t s

1290-0729/$ e see front matter 2010 Elsevier Masson SAS. All rights reserved.

doi: 10.1016/j.ijthermalsci.2010.11.019

International Journal of Thermal Sciences 50 (2011) 803 e 815
mailto:[email protected]://www.sciencedirect.com/science/journal/12900729http://www.elsevier.com/locate/ijtshttp://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://dx.doi.org/10.1016/j.ijthermalsci.2010.11.019http://www.elsevier.com/locate/ijtshttp://www.sciencedirect.com/science/journal/12900729mailto:[email protected]


2/13

The present work is devoted to the analysis of the ow patternand the friction characteristics in tubes of reciprocating scrapedsurface heat exchangers. Particle Image Velocimetry (PIV) tech-nique is employed for obtaining the two-dimensional velocity eldin the symmetry plane of the tube in laminar regime. The separateperformance of the device in the two scraping semi-cycles isconsidered in the analysis, thus de ning counter-current and co-current stages. Flow patterns existing for different velocity ratios uare identi ed, establishing as reference condition the performanceof the still scraper. The in uence of Reynolds number on the owfeatures is assessed.

Experimental results of Fanning friction factor for a wide rangeof Reynolds numbers and velocity ratios are contrasted with thevisualization data, for both scraping semi-cycles. The contributionof the ow structures to the local shear stress is brie y analyzedwith the numerical simulation tool Fluent, complementing theexperimental data and providing further information towards theholistic understanding of the physical ow nature.

2. Experimental program

2.1. Visualization facility

The facility depicted in Fig. 2 was built in order to study theow pattern induced by a wide variety of insert devices in round

tubes [7] . The main section consists of a 32 mm diameter acrylictube installed between two reservoir tanks, that stabilize the ow.The ow temperature is regulated by an electric heater anda thermostat placed in the upper reservoir tank. The ow is

impelled from the lower calm deposit to the upper one by a gearpump, which is adjusted by a frequency converter. By usingmixtures of water and propylene-glycol at temperatures from20 C to 60 C, Reynolds numbers between 100 and 20,000 can beobtained. The tests presented in this work were carried outemploying a mixture of 90% propylene-glycol and 10% water, attemperatures from 25 C to 50 C, yielding Reynolds numbers inthe range from 36 to 378.

The test section has been placed ve scraper pitches down-stream (25D), this ensuring periodic ow conditions. To improveoptical access in this section, a at-sided acrylic box was placedaround. The box was lled with the same test uid that owsthrough the test section. Particle Image Velocimetry (PIV) has beenemployed for ow visualization.

PIV is a well-known technique to obtain global velocity infor-mation, instantaneously and with high accuracy [8]. In theseexperiments, planar slices of the ow eld containing thesymmetry plane of the inserted device were illuminated. The owwas seeded with polyamide particles of 57 mm mean diameter. Thecamera viewed the illuminated plane from an orthogonal directionand recorded particle images at two successive instants in time inorder to extract the velocity over the planar two-dimensionaldomain. 2D velocity elds along a scraper pitch were assembled toprovide an overall insight of the ow structure.

The spatial resolution of the measurement is 110 mm/pixel. A1 mm thick light sheet is created by a pulsating diode laser of 808 nm wavelength. A computer synchronizes the camerashutter opening and the laser shot at sampling frequenciesbetween 250 and 500 Hz, most appropriate for the test

Nomenclature

A annular cross-section (m 2)D tube inner diameter (m)Dh hydraulic diameter (m)d rod diameter (m)F force over a scraper pitch (N)

_m mass ow rate (kg s 1)P pitch of the insert devices (m)D p pressure drop across the test section (Pa)S scraping amplitude (m)T temperature (K)v velocity (m s 1)

Dimensionless groups f h fanning friction factorReh Reynolds numberu velocity ratio, vp/vf c F force coef cient, F =12r v f 2 AC f skin friction coef cient, s w =12r v f 2

Greek symbolsb blockage parameter, ( vf vp)/vf (e )G numerical surfacek radii ratio, d/D (e )m dynamic viscosity (Pa s)U numerical volumer uid density (kg m 3)s wall shear stress (N m 2)

Subscriptscc counter-currenteq co-currentf uidin tube inletmed averageout tube outletp scrapers smooth tubew tube wall z axial direction

Fig. 1. Detail of the scraper geometry inserted in the tube. Typical designs cover the total length of the tube for full cleaning.

J.P. Solano et al. / International Journal of Thermal Sciences 50 (2011) 803 e 815804


3/13

conditions. The sketch of the measurement arrangement is

similar to that employed by [7] .The high speed camera of the PIV system was employed sepa-rately to measure the piston velocity during a characteristicscraping cycle, as shown in Fig. 3. During both semi-cycles thepiston moves with constant velocity, but transient disturbancesshortly occur when the piston motion reverses. Velocity levels andduration of both semi-cycles are nearly identical. When the pistonmoves in the same direction of the ow ( vp > 0), the semi-cycle iscalled co-current, whereas the movement of the piston in oppositedirection to the ow ( vp < 0) is called counter-current. As thehydraulic piston moves with constant velocity, quasi-steady

velocity pro les inside the tube can be considered from a movingreference frame. Therefore, for each working condition, twovelocity elds co-currentand counter-current will be considered tofully characterize the ow pattern during the scraping cycle. Phaseaveraged PIV measurements are performed separately for eachsemi-cycle, following the methodology sketched in Fig. 4. A pho-tosensor (1) is located in the bottom of the acrylic tube. The end of the moving rod (2) activates this sensor for a certain duration overthe cycle, covering the reversal from counter-current to co-currentmotion. The appropriate con guration of the photo-sensor output(normally closed in counter-current measurements, and normallyopen in co-current tests), and the subsequent activation of a triggerallows capturing a pair of images at constant phase for eachcomplete scraping cycle. Ensembled-average of 50 pairs of imageswere done for obtaining the velocity eld over the symmetry planeof the tube in both semi-cycles.

A non-dimensional parameter, the velocity ratio u , is introducedto account for the movement of the device inside the tube:

u v p

v f (1)

The velocity ratio will be positive for co-current conditions, andnegative for counter-current mode. The derivation of vp is per-formed by means of the cycle duration measurement, Dt , obtainedwith a pulse timer, provided that the position of the end-of-strokeswitches ensure a known scraping amplitude S 2P . Fluid velocityvf was derived from ow rate measurements obtained with anelectromagnetic owmeter.

A different formulation of the velocity ratio can be expressed as:

b v f v p

v f (2)

being b the blockage parameter, b 1 u , that can be computed inorder to assess the blockage effect ( b > 0) or dragging effect ( b < 0)of the scraper on the ow.

Fig. 2. Experimental setup for PIV: (1) reservoir tank and lter, (2) honeycomb, (3) testsection, (4) hydraulic piston, (5) reservoir tank, (6) electric heater, (7) electromagnetic

owmeter, (8) frequency converter, (9) centrifugal pump.

0 0.5 1 1.5 2 2.5

0.15

0

0.15

time (s)

v ( m / s )

cocurrent

countercurrent

Fig. 3. Scraper velocity for a characteristic scraping cycle.

Fig. 4. Optical sensor arrangement for phase-averaged PIV measurement in counter-

current and co-current motion.

J.P. Solano et al. / International Journal of Thermal Sciences 50 (2011) 803 e 815 805


4/13

Image processing was carried out with the software vidPIV , thatapplies a cross-correlation algorithm between consecutive images.The interrogation window size used for PIV processing was 32 32pixels, with a 50% overlap. Subsequent adaptive cross-correlationalgorithm [9] was applied with an interrogation window size of 16 16 pixels and 50% overlap. To obtain a clear velocity eld, aftertheimages were correlated,a globaland a local lterwere applied toremove outliers. The resulting vectors were averaged over ftyrealizations. Using the standard formula for 20-to-1 odds,

1.96 s N 0.5 , the uncertainty in PIV measurements is about 3%.Taking into account the uncertainty in the computation of pixelsdisplacements,the naluncertainty inPIVmeasurements rises to5%.

2.2. Friction factor measurement tests

A schematic diagram of the experimental setup is shown inFig. 5. The test section consists of a 20 1 mm smooth tube wherethe scraper is inserted.A secondarycircuit is used for regulating theworking uid temperature. Propylene-glycol is employed asworking uid.

Pressure measurement ports were separated a distancelp 20 P , and consisted of four pressure holes peripherally spacedby 90 . Test section was preceded by a development region of le 6 P length, in order to establish periodic ow conditions.

The scraping amplitude of the insert device was S 2P , so thatthis development region reduced to le 4P for the end-of-strokeco-current motion.

Fluid inlet and outlet temperatures, T in and T out were measuredby submerged type RTDs (Resistance Temperature Detector). Sincetest and developing ow regions were insulated to ensureisothermal conditions, the uid temperature was calculated byT f (T in T out )/2.

Single gauge piezorresistive transducers were coupled to eachend of the pressure test section, to acquire the time-dependentpressure drop occurring during the dynamic performance of the

scraper. A highly accurate, differential pressure sensor was addi-tionally employed for the motionless tests.

Unsteady pressure drop across the tube was computed as thedifference between both signals obtained with the piezorresistivetransducers, D p (t ) p1(t ) p2(t ).

A detail of this signal, for a characteristic scraping cycle, isshown in Fig. 6. The periodicity of the signal corresponds to thescraping frequency. During the counter-current semi-cycle,a higher pressure level is observed in the tube. When the scrapermoves in co-current direction, the pressure level decreases. As

a result, the pressure drop signal approaches to a square wave,in uenced by the constant velocity motion of the piston.

Fig. 5. Experimental set-up for pressure drop measurements: (1) reservoir tank, (2) positive displacement pump, (3) frequency converter, (4) control valve, (5) electrical heater,(6) Coriolis owmeter, (7) oval-wheel owmeter, (8, 9) inlet and outlet RTD probes, (10) differential pressure transducer, (11) hydraulic piston, (13) plate heat exchanger,

(15) diverter valve, (16) PID controller, (17) cooling water, (19) cooling machine, (12, 14, 18) centrifugal pumps.

5 10 15 20

0,3

0,35

0,4

time (s)

p

( t ) = p

1 p

2 ( b a r )

pmed

peq

pcc

tm

countercurrent

cocurrent

Fig. 6. Pressure drop signal across the test section for a characteristic scraping cycle:Reh 60, u 0.1.



5/13

Signal is split into counter-current and co-current frames inorder to derive the separate pressure drop occurring in the tubeduring each semi-cycle. Pressure drop is time-averaged overa suf cient number of cycles, typically n > 50, along Dt m intervalscentered in each semi-cycle (see Fig. 6).

Fanning friction factor for each semi-cycle, based on thehydraulic diameter of the annular section, was determined frommass ow rate measurements and time-averaged pressure dropresults as:

f h D p

lprp 2D d2D3h

32 _m2 (3)

Experimental uncertainty was calculated by following the Guideto the expression of uncertainty in measurement published by ISO[10] . Uncertainty calculations based on a 95 percent con dencelevel showed maximum values of 4% for Reynolds number, 3% forfriction factor in motionless conditions, 8.5% for friction factor indynamic conditions, and 7% for velocity ratio.

3. Numerical model and mesh

The geometry of the computational model wascreated using thesoftware Gambit. It consists of an annular duct to which semi-circular plugs are substracted every P /2 distance, which constitutesthe uid volume of the inserted tube. The scraper extends over fourpitches, being the ow solution in the third one subjected toanalysis. This ensured periodic ow conditions in the entrance and

ow free of outlet effects. The geometrical symmetry of theproblem proved to ensure also ow symmetry for the range of Reynolds number under study. Thus, only half domain wasmodeled. Structured, hexahedral mesh was employed (see Fig. 7). Adouble compression ratio was introduced in both sides of the plugs,where greater variations of the ow pattern exists due to thegeometry constriction, and in radial direction, to ensure bettersolution where higher velocity gradients where expected.

The numerical problem was considered quasi-steady froma moving reference frame located on the scraper. This modelingassumption is supported by two main features of the ow in tubesof RSSHE s. The rst feature comes from the experimental evidenceof constant velocity motion of the scraper during both semi-cycles,as demonstrated in the instantaneous scraping velocity pro ledepicted in Fig. 3. Besides, the ow features at the commencementof each semi-cycle present a short duration transient, being thequasi-steady ow pattern arisen representative of the most part of each scraping semi-cycle. This short-duration phenomenon wasobserved during the visualization campaign, and numericallyquanti ed as 1/4 of the semi-cycle duration in the most unfavorablesituation ( u 2).

To account for this solution strategy, Multiple Reference Frame

(MRF) model [11] was employed in the nite volume code FLUENT

(commercially available software, version 6.3). Absolute velocitiesare used in momentum equations as dependent variables, yieldingto the absolute formulation of the governing equations of theproblem:

V $ v 0!

0 (4)

V $r v 0! v! V p V s (5)

where

v 0!

v! v p! (6)

The next boundary conditions are imposed through the MRFmodel:

v! Gin v! in v! p (7)

v! Gw v! p (8)

Control-volume storage scheme was employed where all variableswere stored at the cell center. Second order upwind scheme wasused in order to interpolate the face values of computed variables.Implicit segregated solver solved the governing equationssequentially. In this study, pressure-velocity coupling algorithmSIMPLE was used.

The Grid Convergence Index (GCI) method has been employedto compute the numerical uncertainty of the simulations (Celiket al. [12] ). Systematic grid re nement was assessed with threesigni cantly different sets of grids, the nest of which is employedto obtain the nal results ( Fig. 7). Maximum values of GCI of 3.49%have been found for the computation of the velocity, 0.85% for thepressure and 2.47% for the derivation of the force coef cient. Asummary of the results for Re h 80 and u 0.5, 1 and 2 isreported in Table 1 .

4. Flow description

During the counter-current semi-cycle, the scraper imposesa positive blockage to the ow, b > 0, with independence of thevelocity ratio. This positive blockage is also found for the motion-less scraper ( u 0), and comprises a common velocity pattern forall these working conditions.

Fig. 8 depicts the non-dimensional velocity eld, v! =v med , in thesymmetry plane of the scraper, for Re h 40 and u 1. Thiscondition is suf ciently representative of the positive blockagephenomenon. A horizontal plane containing the tube axis alsodivides the visualization eld into bottom and top sections, withodd-symmetry characteristics.

Following the sequence of the ow direction (left to right),

a region of low velocity is found in the top section, downstream the

Fig. 7. Detail of the structured mesh. Left: cross-sectional area. Right: Symmetry plane.



6/13

rstplug.The ow vectors enclosed to the picture move backwards,which is associated to ow separation and recirculation.

Afterwards, a high velocity region grows along the streamwisedirection, and nally impacts against the third plug, generatinga transverse vortex.

The mechanismyielding to the appearance of these structures isrelated to the meandering path of the ow: the presence of theplugs every P /2 distance forces the ow to sharply turn and movetowards the section not affected by the blockage. The resultingdeviation induces a local velocity increase, as the ow area isapproximately reduced by one half. The main stream reattaches tothe wall and proceeds towards the tube. As this stream expands tothe annular section, two counter-rotating recirculation bubbles arecreated behind the plugs. These bubbles converge in the symmetryplane, where the local structure with low velocity, reversed owvectors are measured. The reattached ow nally impacts againstthe next plug, creating a transverse vortex rolling over a curved axisparallel to the front side of the plug.

4.1. Counter-current mode

Fig. 9 represents the velocity eld in the symmetry plane of theinserted tube at Reynolds number Re h 80, for the limitingcondition of motionless scraper ( u 0), and for velocity ratiosu 0.5, 1 and 2. This picture allows analyzing the effect of thescraper movement on the ow structures.

Two main features of the recirculation bubbles are enhanced

during the counter-current motion: the characteristic size andthe local velocity level. The length of these bubbles reaches up toL/P z 0.3 in dynamic conditions. This value appears to beasymptotic, as the mean ow deviates along a characteristic tubestream to overcome the next plug. Additionally, the local velocitylevels of the reversed ow are higher than those found for themotionless condition ( v/vmed z 0.1), owing to the higher staticpressure de cit induced by the scraper movement. Values of theorder of v/vmed z 1 are reported for u 0.5, and v/vmed z 3 foru 2. Higher wall shear rates are thus expected downstreamthe plugs. The dynamics of the recirculation bubbles conveycolder uid from the center of the tube towards the wall, and willeventually contribute to increase the local heat transfer coef -cient in this region.

The transverse vortex is also depicted in Fig. 10, and contrastedwith the motionless solution. It is clear that the movement of theplugs promotes the appearance of the vortex, which maintainsa similar shape and size, as shown for Re h 80 and the testedvelocity ratios. Local velocities are in all dynamic conditions muchhigher than those found for u 0, of the order of v/vmed z 2.5.

Another important feature ampli ed by the counter-currentworking mode is the global velocity level reached in the highvelocity region. Velocity elds in Fig. 9 allow identifying thisevolution, which reaches values as high as vmax /vmed z 4 foru 0.5 and vmax /vmed z 7 for u 2. This is contrasted with thelower level, vmax /vmed z 3, reported for u 0. A more detailedanalysis of the velocity pro le allows stating the in uence of Rey-nolds number on the ow nature. Fig. 11 presents the velocitypro le in a section located at a distance P /5 of the front side of theplug, for u 0, 0.5, 1 and 2. Results are contrasted with theanalytical solution for the concentric annuli with counter-currentmoving core and corresponding velocity ratio:

v

v med

1 u 1

2ln 1=k k2

1 k2

121 k4

1 k2 1 k2

ln1=k

1r R

2 1 k2

ln1=kln

Rr

u lnR=r ln1=k

9

In the motionless condition, the velocity pro les present a para-bolic shape for Re h 40 and 105. However, radial transport of momentum yields to a atter distribution for Re h 213. Theexperimental results for the counter-current velocity pro lespresent a similar behaviour, with an earlier onset to this velocitypattern, that plays a role on the appearance of low Reynoldsnumber turbulent ow characteristics. For u 2, a at velocitydistribution is found for Reynolds number Re h 40 and 80.

4.2. Co-current mode

During the co-current semi-cycle, three different ow patternsmay appear, depending on the sign of the blockage parameter b.Experimental results for Reynolds number Re h 80 and velocityratios u 0.5, 1 and 2 are presented in Fig. 12 and discussed below.

u 0.5 (b > 0). In this working mode, the scraper moves in thesame direction of the mean ow, but with lower velocityvp vf /2. Thus, the insert device imposes a blockage over the

ow, that is forced to deviate when it encounters the movingplugs. The nature of the ow is therefore similar to thatdescribed for counter-current conditions, although somefeatures do not exist for u < 1. Actually, no reversed ow is

Fig. 8. Assessment of mean

ow structures measured in the symmetry plane (PIV).

Table 1Grid convergence Index (%) for Re h 80 and different velocity ratios.

u 2 1 0.5 0.5 1 2

v 3.49 3.62 2.75 2.12 2.93 3.15 p 0.23 0.85 0.78 0.69 0.53 0.55c F 2.47 1.97 1.90 1.83 1.75 2.42



7/13

found in the rear side of the plugs, indicating the absence of recirculation bubbles when the ow expands, due to the co-current movement of the scraper. Likewise, no transversevortex appears in the front side of the plugs, as can be seen indetail in Fig. 13, for the tested velocity ratios. Nonetheless,transverse vortices have been reported in co-current mode,with the aid of numerical simulation, for very low velocityratios ( u 0.1) and high Reynolds numbers.The blockage of the plugs increases the global velocity level.Velocity pro les in the control axial position are shown inFig. 14 (left), for Reynolds numbers u 0.5 and Re h 80, 150and 250. A parabolic shape is reported in all the conditions,

with similar maximum velocity values, of the order of vmax /vmed

z 1.8. The absence of recirculation bubbles justi es the lowervelocity values in this acceleration region. This is the reason of the similar velocity levels encountered for all the Reynoldsnumbers. It is also noticeable the fact that no evolution of theparabolic shape towards a atter velocity pro le is found forthe higher Reynolds number tests. This could mean that thelower blockage of the plug in co-current conditions allowsstabilizing laminar ow conditions for a wider working regime,rather than acting as a turbulence promoter. u 1 (b 0). The scraper and the mean ow move with thesame velocity, thus minimizing the blockage imposed by theplugs. The ow is smoothly deviated in the vicinity of the plugs,

but a similar pattern is observed upstream and downstream

Fig. 9. PIV velocity eld along the tube symmetry plane in counter-current motion, for Re h 80 and u 0.5, 1 and 2.



8/13

these elements. The axial velocity pro le, depicted in Fig. 14(center), presents a parabolic shape with a fair agreementwith the analytical solution for the equivalent concentricannuli with co-current moving core. This tendency is preservedfor all the Reynolds numbers under investigation(40 Reh 250), indicating again that the ow keeps laminarfor this working condition in a wider range of Reynoldsnumbers, with respect to the motionless characteristics.

u 2 (b < 0). The scraper moves faster than the mean ow,dragging it locally in the surroundings of the plugs. Asa consequence, the ow presents a completely differentpattern to those studied so far. The ow in the front and rearside of the plugs has the same direction of the mean ow.However, as the scraper velocity is higher than the ow

velocity, vp 2 vf , the movement of the plugs not only drags

the ow over their front and rear sides, but also generatesa suction effect on the opposite tube section, yielding toa single recirculation bubble that extends backwards, with lowlocal velocity levels of the order of v/vmed z 1. Velocity pro lesare shown in Fig. 14 (right) for Reynolds number Re h 40 and80, contrasted with the analytical solution for the equivalentconcentric annuli.

5. Pressure drop results

5.1. Counter-current

Fanning friction factor results in counter-current conditionspresent a twofold interest: in the frame of this work, the evolution

of friction factor with Reynolds number and velocity ratio provides

Fig. 10. Experimental results of the transverse vortex ( v! =v med ) for Reh 80 and 2 u 0.

0

1

2

3

4

V z

/ V

m e

d

=0

0.5

0

1

2

3

4 =0.5

0.25 0.5 0.75 11

0

2

4

6

r/R

=1

V z

/ V

m e

d

0.25 0.5 0.75 12

0

2

4

6

8

r/R

=2

Reh=40

Reh=80

concentric annuli

Reh=80

Reh=150

concentric annuli

Reh=40

Reh=105

Reh=213

concentric annuli

Reh=40

Reh=80

Reh=150

concentric annuli

Fig. 11. PIV velocity pro les for u

0,

0.5,

1 and

2 and several velocity ratios.



9/13

indirect information on the nature of the ow, allowing to identifypure laminar, transitional and turbulent ow. It also gives infor-mation on the in uence of the working regime in maximumpressure drop. This is of primary importance, for example, inapplications where a minimum mass ow must be continuously

pumped despite the adverse conditions present during counter-current semi-cycles.

Fig.15 presents Fanning friction factor results in counter-currentconditions for velocity ratios u 0.5, 1 and 2 and maximumvalues of Reynolds number of Re h z 1000. Results for the

Fig. 12. PIV velocity eld along the tube symmetry plane in counter-current motion, for Re h 80 and u 0.5, 1 and 2.

Fig. 13. Experimental results of the velocity

eld in the front side of the plug, for Re h

80 and 0u

2.



10/13

motionless scraper ( u 0) are also includedas a reference, togetherwith the analytic laminar solution for the equivalent smooth tubeand the concentric annuli. The analytical solution of the latter withinner moving rod reads:

f h 161 k2

Reh 266641 u

12ln 1=k

k2

1 k2

1 k4

1 k2 1 k2

ln1=k

37775(10)

For u 0.5, a laminar region is identi ed, where friction factordependence on Reynolds number is similar to motionless results, f h,cc a Reh 0.75 . However, the change of this tendency occurs at Re h z100,which implies earlier transitional ow (the onset in motionlessconditions is reported for Re h z 200). Pressure drop augmentationsof 70% are found in laminar regime, referenced to u 0, and 150% inturbulent ow (Re h z 700), as a result of the turbulence advance.

A similar trend is found for u 1 and u 2. However, theincrease of velocity ratio yields to a lower in uence of Reynoldsnumber in friction factor in the laminar regime, f h,cc a Reh 0.66 . Thetransition from laminar to turbulent ow can not be inferred fromthe slope change detected in the data. The intersection of laminar-side and turbulent-side trend lines allows estimating transitionalregime at Re h z 70 for u 1 and Re h z 40 for u 2. Theseresults con rm the effect of the scraping velocity on turbulencepromotion in counter-current conditions.

Similarly, pressure drop increases for increasing velocity ratios.Maximum augmentations of 260% are found for u 2 in laminarregime, whereas turbulent ow augmentations of 900% are can beinferred from the experimental data, with respect to the motionlessconditions.

5.2. Co-current

Experimental results of Fanning friction factor are representedin Fig. 16, for u 0, 0.5 and 1. The co-current movement of thescraper decreases the pressure drop across the tube for allthe conditions with respect to the motionless scraper, as a result of the lower blockage imposed by the moving plugs, and thedecreased shear rate resulting in the ow.

The data trend for u 0.5, covering the range 15 < Reh < 300,presents the characteristic behaviour of laminar ow, with f h,cca Reh 0.85 . Unlike the results for u 0, no onset of transition isobserved for the Reynolds number range tested, following that the

ow pattern induced by the co-current movement of the scraperdelays the appearance of turbulence phenomena. This conclusionfully agrees with the velocity pro les of Fig. 14, that keep a para-

bolic shape even for Re h 250.

The results for u 1 are also included in Fig. 16, together withthe solution of Fanning friction factorfor the concentric annuliwithco-current moving core and u 1. The fair agreement between theexperimental solution for moving scraper and the analytical solu-tion for concentric annuli reveals the negligible blockage of thescraper when moving with same velocity and direction than themean ow. This fact is also proved by the similarity of the experi-mental and analytical velocity pro les for this velocity ratio.

The performance of the scraper for u 2 presents a differentpattern, owing to the negative blockage imposed by the scraper,that moves faster than the uid.

Actually, in these conditions the scraper addsmechanicalenergyto the ow, working as a pump inside the tube. As a result, a pres-sure increase is measured across the test section for u 2, yieldingsubsequent negative friction factors. The accuracy of the piezorre-sistive pressure sensors prevents from obtaining reliable results forthe magnitudes measured in this working condition, that nallyretrieve spurious friction factor values. The numerical simulation

0.25 0.5 0.75 10

1

2

r/R

V z

/ V

m e

d

=0.5

0.25 0.5 0.75 10

0.5

1

1.5

r/R

=1

0.25 0.5 0.75 10

1

2

3

r/R

=2

Reh=40

Reh=80

concentric annuli

Reh

=40Re h=80Re

h=150

Reh=250

concentric annuli

Re h=80Re

h=150

Reh=250

concentric annuli

Fig. 14. PIV velocity pro les for u 0.5, 1 and 2 and several velocity ratios.

10 1 10 2 10 3

101

100

101

Re h

f h

countercurrent

=0 =0.5 =1 =2

concentric annuli ( =0)

smooth tube

Fig.15. Experimental results of Fanning friction in counter-current mode, for u 0.5,1 and 2. Comparison with motionless results ( u 0) and analytical solutions for

smooth tube and concentric annuli.



11/13

will be employed next to overcome this inconvenience and providea physical insight into this phenomenon.

6. Numerical wall shear pro les

Numerical simulations of the ow were performed in the rangeof Reynolds number between 15 and 170, under the assumption of laminar ow. Solutions for counter-current and co-current motionwere obtained with velocity ratios u 0.5, 1 and 2. Velocitypro les in the radial section located at P /5 of the front side of theplugs are extracted for Re h 80 for counter-current and co-currentmotion. The comparison with experimental results is representedin Fig. 17, showing a satisfactory agreement. Higher discrepanciesare found in the vicinity of the tube wall and rod, as a result of theexperimental restrictions to properly illuminate those regions of

the insert tube. On the contrary, the numerical simulationsuccessfully retrieves the slip condition in the scraper rod, r /R k,for all the operating conditions.

In counter-current motion, maximum values of the skin frictioncoef cient are found in the front side of the plugs, towards the partof the tube where the bulk ow is deviated (see Fig.18). The regionof high shear stress extends towards the reduced cross-sectionregion and further downstream. Progressively, the boundary layergrows along the tube wall side opposite to the plugs, yielding toa reduction of the local shear stress. Conversely, a huge region of low wall shear stress is found in the rear side of the plugs. This areaisin uenced by the lower local velocity of the recirculation bubbles

101

102

103

101

100

Re h

f h

cocurrent

=0 =0.5 =1



smooth tube

Fig. 16. Experimental results of Fanning friction in co-current mode, for u 0.5 and 1.Comparison with motionless results ( u 0) and analytical solutions for smooth tube,motionless concentric annuli and concentric annuli with internal co-current movingcore ( u 1).

0.25 0.5 0.75 10

0.5

1

2

3

r/R0.25 0.5 0.75 1

2

10.5

0

2

4

6

8

V z

/ V

m e

d

r/R

PIVCFD

= 2

= 1

= 0.5

countercurrent cocurrent

= 2

= 0.5

= 1

Fig. 17. Numerical velocity pro les for Re h

80 and several velocity ratios, in co-current and counter-current modes. Comparison with experimental data.

Fig. 18. Skin friction coef cient in the tube wall in counter-current motion, forReh 80 and u 0.5, 1 and 2.



12/13

generated when the ow expands downstream the plugs. Thisdistribution is basically repeated for the three velocity ratios.

In co-current motion, the existence of three different owpatterns implies different shearing pro les, which are presented inFig. 19. For u 0.5, the skin friction coef cient distribution followsa similar pattern to the one obtained above for positive blockagesituations presented above: lower shear stress are found down-stream the plugs, although there are no recirculation bubblesowing to the parallel motion of the scraper and uid. However, theexpansion of the ow that overcomes the plugs yields to lowervelocities in this area. The acceleration of the ow and furtherdeviation generates a high local shear stress eld as well. For u 1,only higher distributions are found in the deviation of the ow

opposite to the plugs. The region in-between presents a fairlyconstant shear eld, owing to the slight ow disruption.

For u 2, the higher shear stresses are found upstream anddownstream the plugs, owing to the acceleration of the ow forcedby the scraper movement. Conversely, lower shear stress are foundon the opposite part of the tube where the plugs are located, due tothe recirculation bubble created by the dragged ow.

The in uenceof pressureand viscous forceson theglobalpressuredrop can also be accounted for with the aid of numerical simulation.Both effectshave beenanalyzed integrating the viscous forces, whichare dominant in the tube and rod walls, and pressure forces, thatmainly appear on the plug surfaces normal to the ow direction.

To account for this analysis, a periodic control volume extendedover the scraper pitch is chosen. Two plugs are contained in thisvolume, thus contributing with net pressure forces over their rearand front faces. Momentum conservation over the pitch allowsestablishing the next forces relation:

Z s 2

s 1

p n! ds

|fflfflfflfflfflfflffl{zfflfflfflfflffltotal force

Z wall

s n! ds Z rod

s n! ds

|fflfflfflfflfflfflfflfflfflfflviscous force Z

plugs

p n! ds

|fflfflfflfflfflpressure force(11)

where s 1 and s 2 are the annular cross-sections of the tube limitingthe domain.

Fig. 20 presents the pressure, viscous and total force coef cientsover a scraper pitch, obtained for the numerical simulations of counter-current motion. Pressure forces exceed the viscouscontribution in all working conditions. For u 0.5, these forcesare of the same order for the lowest Reynolds numbers (Re h z 15),whereas for Re h z 150, the contribution of pressure forces to totalforce is of the order of 80%. Experimental results of Fanning frictionfactor are contrasted with total force coef cient by means of

c F f h4P

Dh(12)

Excellent agreement of experimental data with total force numer-ical computation is found for u 0.5. For u 1, the sametendency is observed, with a contribution of pressure forces of 85%at Re h z 150. For u 2, pressure forces play a fundamental roleeven for the lowest Reynolds numbers, highlighting the negativeeffect of the scrapers motion in terms of loss of mechanical energy.In this working condition, pressure forces contribution is as high as70% for Reh z 15 and 90% for Re h z 150.

Fig. 19. Skin friction coef cient in the tube wall in counter-current motion, forReh 80 and u 0.5, 1 and 2.

0 50 100 1500

40

80

120

160

Re h h

c F

=0.5 countercurrent

pressure force (CFD )viscous force (CFD )

total force (CFD )total force (Experimental)

0 50 100 150

Re

=1 counter current

0 50 100 150

Re h

=2 counter current

Fig. 20. Numericalcomputation ofpressure, viscousand totalforcecoef cientsovera scraper pitch, incounter-currentmotion.Comparisonwith experimentaldatafrom pressure droptests.



13/13

Agreement of numerical results with experimental data fails forReh > 80, which may indicate the existence of a different ownature, with turbulent and unsteady characteristics. The numericalstudy of these ow features is beyond the scope of this work.

Numerical results in co-current motion are presented in Fig. 21.Viscous force for u 0.5 is dominant for the range of Reynoldsnumber under investigation. For Re h z 150, viscous and pressureforces are of the same order, and the data trend allows inferringthat, for higher Reynolds numbers, pressure forces would exceedviscous effects. Agreement with experimental results derived fromFanning friction factor is excellent.

For u 1, pressure forces are negligible for the whole range of Reynolds numbers. This result is expected from the fact that thescraper and the uid move with similar velocity, thus minimizingthe blockage imposed by the plugs. Total pressure drop in thisworking condition is thereforecaused exclusivelyby viscous effects.For u 2, a completely different behaviour is found: pressure forcecoef cient is negative,as a result of the faster velocity of the scraperwithin the ow. Higher pressure elds are reported in the rear facesof the plugs, thus resulting in net pressure forces opposite to themean ow direction. For this velocity ratio, this pressure contribu-tion yields to negative total force coef cients. Contrast with exper-imental data are also reported. A reasonable agreement is found, asexpected from the negative Fanning friction coef cients obtained inthe testbench. As discussed above, sensorsaccuracyprevented fromobtaining more reliable experimental data for u 2.

7. Conclusions

1. Flow pattern and shearing pro les investigation is carried outin tubes of heat exchangers with reciprocating scrapers. PIV measurements are complemented with pressure droptests andCFD computations, at low Reynolds numbers and velocityratios u 0, 0.5, 1 and 2, in order to obtain a globalunderstanding of the ow behaviour.

2. Flow structures in counter-current motion are similar to thosefound for the motionless condition, owing to the positiveblockage imposed by the scraper. Velocity ratio plays animportant role on the radial transport of momentum, yieldingto the promotion of turbulent ow characteristics.

3. Flow behaviour in co-current mode presents three differentpatterns depending on the blockage parameter: the ow maypresent a meanderingpath forpositive blockage,no disturbances

for zero-blockage, and a dragging pattern for negative blockage.

4. Fanning friction factor during both scraping semi-cycles havebeen experimentally obtained, for different velocity ratios.Pressure drop augmentations in counter-current mode havebeen assessed, as well as earlier transition to turbulence. Co-current motion of the scraper have proved to preserve laminar

ow for a wider range of Reynolds number.5. Numerical simulation of the ow provides an insight into

the wall shear stress, and allows identifying the contribution of the local ow structures to the pressure losses mechanism. Thehigher contribution of drag losses to global pressure drop hasbeen assessed for counter-current conditions. In co-currentmode, the in uence of the blockage on head losses is deter-mined, and the drag effect for u 2 is clearly de ned.

Acknowledgements

This research has been partially nanced by the DPI2007-66551-C02 grant of the Spanish Ministery of Science and thecompany HRS Spiratube . The authors are grateful to SEDIC-SAIT(UPCT) for providing the technical resources for CFD computations.

References

[1] C.S. Rao, R.W. Hartel, Scraped surface heat exchangers, Critical Reviews inFood Science and Nutrition 46 (2006) 207 e 219.

[2] A.E. Bergles, S.D. Joshi, Augmentation Techniques for Low Reynolds Numberin-tube Flow, Low Reynolds Number Flow Heat Exchangers. Hemisphere,Washington, DC, 1985.

[3] A.M. Trommelen, W.J. Beek, Flow phenomena in a scraped surface heatexchanger(Votator-type),Chemical Engineering Science 26 (1971) 1933 e 1942.

[4] M. Stranzinger, A. Bieder, K. Feigl, E. Windhab, Effects of ow incidence andsecondary mass ow rate on ow structuring contributions in scraped surfaceheat exchangers, Journal of Food Process Engineering 25 (2002) 159 e 187.

[5] E. Dumont, F. Fayolle, V. Sobolik, J. Legrand, Wall shear rate in the Tay-lor e Couette e Poiseuille ow at low axial Reynolds number, International Journal of Heat and Mass Transfer 45 (2002) 679 e 689.

[6] M. Harrod, Scraped surface heat exchangers: a literature survey of owpatterns, mixing effects, residence time distribution, heat transfer, and powerrequirements, Journal of Food Process Engineering 9 (1986) 1 e 62.

[7] A. Garca, J.P. Solano, P.G. Vicente, A. Viedma, Flow pattern assessment intubes with wire coil inserts in laminar and transition regimes, International Journal of Heat and Fluid Flow 28 (2007) 516 e 525.

[8] M. Raffel, C. Willer, J. Kompenhans, Particle Image Velocimetry: A PracticalGuide, rst ed. Springer, 2000.

[9] F. Scarano, M.L. Reithmuller, Advances in iterative multigrid PIV image pro-cessing, Experiments in Fluids 29 (2000) 51 e 60.

[10] ISO, Guide to the Expression of Uncertainty in Measurement, rst ed. Interna-tionalOrganization for Standarization, Switzerland,1995, ISBN 92-67-10-188-9.

[11] J.Y.Luo,R.I.Issa,A.D. Gosman,Predictionof impeller-induced owsin mixingvesselsusing multipleframes of reference, IChemESymposiumSeries 136(1994)549 e 556.

[12] I.B. Celik, U. Ghia, P.J. Roache, C.J. Freitas, H. Coleman, P.E. Raad, Procedure forestimation and reporting of uncertainty due to discretization in CFD appli-

cations, Journal of Fluids Engineering 130 (2008) 078001-1-4.

0 50 100 150

20

0

20

40

60

Re

c F

=0.5 cocurrent

pressure force (CFD)viscous force (CFD)total force (CFD)total force (EXP)

0 50 100 150

Rehh

=1 cocurrent

0 50 100 150 200

Reh

=2 cocurrent

Fig. 21. Numericalcomputationof pressure,viscousand total forcecoef cients overa scraper pitch, in co-currentmotion. Comparisonwithexperimentaldatafrom pressuredrop tests.


Documents

Flow Pattern Assessment in Tubes of Reciprocating Scraped