Falchi Provenzano Pietrogiacomi Romano 2006 EiF

7/29/2019 Falchi Provenzano Pietrogiacomi Romano 2006 EiF

1/13

R E S E A R C H A R T IC L E

M. Falchi G. Provenzano D. PietrogiacomiG. P. Romano

Experimental and numerical investigation of flow control on bluff bodiesby passive ventilation

Received: 25 July 2005 / Revised: 16 March 2006 / Accepted: 19 March 2006 / Published online: 14 April 2006 Springer-Verlag 2006

Abstract In this work, the so-called natural or passiveventilation drag reduction method is investigatedexperimentally and numerically. Passive ventilation isperformed by directly connecting the high pressure re-

gion at the front of a body to the lower pressure in thenear wake using a venting duct; in this manner, a netmass flux is established within the wake. In particular, inaerodynamic applications it appears suitable to attain aglobal reduction in the drag of a body moving in a fluidand a reduction in turbulence levels by means of a globalmodification of the body wake. Velocity field investiga-tions using particle image velocimetry measurementsand a Reynolds averaged numerical code are employedat moderately high Reynolds numbers to clarify theeffectiveness of drag reduction on a vented bluff body.The numerical and experimental results agree qualita-tively, but the amount of reduction for the vented body

(about 10%) is underestimated numerically. The effec-tiveness of drag reduction has been proved both forsmooth and rough (single strip) models. Direct balancemeasurements are used for comparisons.

1 Introduction

Drag reduction, particularly in automotive and aero-nautic applications, is one of the most investigatedtopics for its possible benefits in reduction in fuel con-

sumption and improvement in performance. In thesubsonic range, the total drag force exerted on a body isdue to the sum of the skin friction drag and the pressureor form drag. As the body shape changes, the relevanceof each of these two drags also changes; for streamlined

bodies, the skin friction drag is the main part (8590%),while for bluff bodies the form drag is the most impor-tant (8085%). In particular, a bluff body, immersed in afluid flow, is characterized by a large separated region in

the rear part of the body i.e. the wake. The wake isgenerated by the boundary layer detachment whichavoids the pressure recovery; the net result, consideringthe high values of the pressure in front of the body, is anincrease in the form drag.

Natural ventilation is a flow control technique whichmodifies the bluff body wake and reduces the total dragby acting on the form drag. In this technique, a ventingduct connects the high pressure region in front of thebody to the low pressure region in the near wake. In thisway, the pressure gradient sets up a flow inside the ductwith energy levels as high as the free stream, which en-ergises the wake. The destructive interaction of the

counter-rotating vortices from the venting duct and theexternal shear layer, generating a sort of aerodynamicstreamlining, partially destroys the wake and raises thepressure values (behind the body) allowing pressurerecovery. The main effect is a large reduction in the formdrag together with the reduction in the turbulenceintensity in the wake region (Roshko 1961; Aschenbach1972, 1974; Suryanarayana and Meier 1985; Gad el Hak2000). It is important to note that no external power isrequired to generate the improvement in comparisonwith most other drag reduction (active) techniques, e.g.base bleed, suction or near wake-heating (Monkewitz1992). Up to now, its effectiveness has been proved only

in the case of the sphere (Suryanarayana and Meier1985; Suryanarayana et al. 1993; Suryanarayana andPrabhu 2000). The results obtained indicate a 5060%total drag reduction in the supercritical range (i.e. forReynolds numbers larger than 3.5 105 with a fullyturbulent boundary layer), while practically no reduc-tion occurs for lower Reynolds numbers. The authors ofthe previously mentioned papers claim that this is due tothe boundary layer transition which helps in transferringmomentum from the side of the body and from theoutlet of the venting duct to the rest of the wake.

M. Falchi G. Provenzano D. PietrogiacomiG. P. Romano (&)Department of Mechanics and Aeronautics, University of Rome,La Sapienza, via Eudossiana n. 18, 00184 Rome, ItalyE-mail: [email protected]

Experiments in Fluids (2006) 41: 2133DOI 10.1007/s00348-006-0141-x


2/13

The main open question concerns the possibility ofsuccessfully applying the technique on bluff bodies otherthan the sphere, which are particularly interesting forindustrial applications (it should be considered thatmany investigations on this aspect are not fully availableto the scientific community due to their commercial use).Under this aspect, the typical Reynolds numbers forsome applications on vehicles could be lower than thecritical Reynolds numbers (for example, for motorcyclesor bicycles) so that the performances of natural venti-lation should be exploited in subcritical regimes also. Tothis aim, to attain results similar to those in supercriticalregimes, the use of tripping devices in subcritical regimesshould be considered.

It must also be noticed that in the sphere the constantcurvature allows the boundary layer detachment pointto move freely over the sphere surface following changesin the fluid dynamic field (in particular the energy of theshear layers in relation to the wake). This is not allowedin bluff bodies which are more or less truncated in therear part; that is why it is important to verify the effec-tiveness of the technique in this case. In particular, based

on results of preliminary measurements, the use of around shape rather than sharp edged bodies in the rearpart was considered; this allows comparison of thephenomena observed on the sphere.

In this paper, experiments by particle image veloci-metry (PIV) and numerical results by a commercialReynolds averaged code (RANS) are coupled to inves-tigate the velocity fields around vented and non-ventedbluff body models to clarify the above mentioned phe-nomena, with particular regard to subcritical regimesusing round edged tripped bluff bodies. An overallmeasurement of the drag of different models has beenperformed by acquiring direct dynamometric balance

data.

2 Experimental set-up and numerical simulations

The experimental investigations have been performed inthe low speed wind tunnel of the University of RomeLa Sapienza, Italy. The tunnel has an open round testchamber (0.9 m diameter, 1,2 m length), a maximumvelocity of 55 m/s, Re = 2.34 106, and a level of tur-bulence intensity


3/13

It is important to notice that all measurements on the

models have been taken both in the smooth or clean(no-strip) configuration and in a tripped rough config-uration (with strip); for the strip configuration, anadhesive strip is stuck to the model at the end of thespherical part, while the clean configuration is with-out the strip. This choice allows the boundary layer onthe model surface to exploit the transition from laminarto turbulent; thus, the comparison between clean andstrip models points out the importance of the boundarylayer transition. Moreover, rough models allow us totest the technique in more realistic conditions for prac-tical applications. The strip consists of 1.5 cm wideadhesive paper; in Fig. 1 a picture of the vented model in

strip configuration is given (the model is painted in blackto reduce reflections in PIV measurements).

The numerical simulations have been performed totest the robustness and suitability for advanced appli-cations of the commercial RANS code used (StarCD) bycomparing the data with the experimental ones. Thesimulations have been performed at a Reynolds numberequal to 2.38 105, on a fully three-dimensional do-main; a high Reynolds number finite volume K-e modelwas used with about 3 106 cells. The wall functionapproach with monotone advection and reconstruction

scheme (MARS) for average fields and upwind differ-

ences for turbulence evaluation have been used. Inletand outlet profiles have been given at boundaries. Pre-liminary grid sensitive tests have been performed and afinal mesh extending 10 D along all directions has beenselected; the residual was 105 for all variables.

Preliminary flow visualisations using a high speedcamera (250 frame/s with resolution equal to 480 420)and a 12W infrared laser, seeding the wind tunnel withsmoke have been performed. The Reynolds number forthese visualisations is equal to about 5 105; this valueis within the interval used for direct balance drag mea-surements and quite close to those used for the othermeasurement techniques. A couple of acquired images

are given in Fig. 2b and c. Although results frominstantaneous images must be treated with care (theinterval between frames is equal to only 0.4 D/U), thevented case clearly shows a different wake; in particular,the transverse amplitude of the wake seems to be re-duced with respect to non-vented conditions. From thesevisualisations, it is also observed how the venting jetextends up to about one body diameter (this is not al-ways the case for the other tested venting duct, i.e. withsmaller diameter or with variable duct geometry) and aclear interaction between the jet itself and the two shear

y

xa) b)

strip

c)

Fig. 1 Bluff body models tested: a non-vented or closed model, b vented model. The black painted vented bluff body in strip configuration (c)

23


4/13

layers developing from the body rear side. This inter-action reduces the size of the vortical structures in theshear layers in comparison with the non-vented case.

3 Results

3.1 Direct drag measurements

In Fig. 3, the drag coefficient measured with the dyna-mometric balance as a function of the Reynolds number

is depicted for the tested bluff body models; the dragcoefficient has been calculated, once the drag, Dr, hasbeen measured, by CD = 2Dr/qU

2 S, where q is the airdensity, U

is the free stream velocity and S is the model

cross-sectional area.As shown in the figure, when comparing non-vented

to vented conditions for the clean (no-strip) configura-tion models, an average drag reduction of about 78% isattained; this is slightly larger than the measurementuncertainty (about 5%) and so it must be confirmed byvelocity field measurements.

For the strip configurations, a different situationarises; a 20% average drag reduction from non-ventedto vented conditions has been obtained. This differencebetween the no-strip and strip conditions depends on thefact that vented models (in both clean and strip config-urations) have almost the same drag, whereas non-ven-ted ones have different values. The non-vented stripmodel has about 10% higher drag (in comparison withthe non-vented clean model) due to forced transition tothe fully developed turbulent boundary layer which in-creases the skin friction drag. Thus, by considering that

in practical applications there is always some roughnesson the aerodynamic surfaces (which forces boundarylayer transition), the technique seems to work muchbetter just in such a situation. In any case, the effect ofthe venting jet overcomes the boundary layer perturba-tion due to the strip.

For a deeper understanding of the observed dragreduction, velocity measurements in front and in thewake of the models will be considered. In particular,velocity measurements in the wake allow us to detail thedifferences occurring in separation regions; on the other

b)

a)

c)

Laser

Model

WindTunnel

open testsection

Suspendingstrut

Lasersheet

Uo

Fig. 2 Experimental set-up: the dashed area corresponds to the measurement region (a). Smoke visualisation of the b non-vented and cvented models in strip configuration

24


5/13

hand, the velocity field in front of models is required tocompare results with a rather crude evaluation of thedrag ratio among vented to non-vented models (by

considering the gain in momentum which is obtained inthe venting jet). This evaluation gives as a result the ratioof cross-sectional front areas (about 98%); thus, theamount of theoretical drag reduction is only 2%. Thisvalue is lower than that observed in the present balancemeasurements both for clean and strip configurations.

3.2 Velocity measurements

In Fig. 4, instantaneous vector fields (with overlappingaxial velocity contours) measured by PIV in the wake ofthe models are shown for the non-vented and vented

conditions (with strip, similar diagrams are obtained forthe no-strip case); the upper one half of the model isconsidered (the black patterns on the left give indicationof the model and duct position and size). The referencesystem origin is located on the rear flat surface, on the

cylinder axis; the free stream velocity is directed fromleft to right. The overall (qualitative) impression fromthese plots is that the two wakes are very similar; closeinspection, however, reveals large vortical structuresdeveloping in the shear layer for the non-vented case.These structures, indicated by orange circles in Fig. 4a,contribute towards generating a wide wake even at x/D = 1.5. On the other hand, for the vented condition,the interaction between vortical structures from theventing duct and from the shear layer reduces the size ofthe vortices and the width of the whole wake.

In Fig. 5, averaged vector fields and axial velocitycontours in the wake of the models obtained by PIV aredepicted for the four tested conditions. In the non-ven-ted configuration, both for the clean and the strip con-figurations (Fig. 5a, c), the presence of a counter-flow inthe wake of the models from the base to approximatelyx/D = 1 is noticed; no large differences are seen be-tween the two configurations, except for a slight reduc-tion of the wake in the strip case. For the ventedconfigurations (Fig. 5b, d) the situation is different; infact, the venting jet decreases the extent of the recircu-

lation region by penetrating in the wake down to aboutx/D = 0.8 for the clean case, and to about x/D = 0.91.0 for the strip case (to appreciate this difference, con-sider the vertical velocity profiles around x/D = 1). Thedifference between strip and clean conditions is causedby the strip that, by forcing the laminar-turbulenttransition of the boundary layer, allows a slight down-stream displacement of the detachment point thusdecreasing the width and increasing the length of thewake (as can be seen from Fig. 1, the models haverounded rear edges to allow such a slight displacement).This is a key point with respect to previous tests per-formed with sharp edges.

It has to be noticed that in both strip and cleanconditions, the venting jet is not able to fully emergefrom the wake; thus a minor effectiveness of naturalventilation in comparison with the case of the spheremust be expected. This is confirmed by the fact that for

0

0,1

0,2

0,3

0,4

0,5

200000 300000 400000 500000 600000 700000

Re

Cd

unvented s trip

vented s trip

unvented clean

vented clean

Fig. 3 Direct drag coefficient balance measurements at differentReynolds numbers

x (x/D)

y

(y/D)

0.5 1 1.5 2

0

0.25

0.5

0.75

1

1.25

1.5

1.75 unvented

x (x/D)

y

(y/D)

0.5 1 1.5 2

0

0.25

0.5

0.75

1

1.25

1.5

1.75 venteda) b)

Fig. 4 Instantaneous vector fields and axial velocity contours from PIV: a non-vented, strip configuration; b vented, strip configuration.Re = 1.7 105

25


6/13

x/D > 1.5 higher velocities are obtained at the middleof the wake for non-vented models in comparison withthe vented. Thus, only some portions of the wake areaccelerated when ventilation is active; as a matter of fact,the whole configuration of the wake is changed incomparison with non-vented conditions.

In Fig. 6, the vented configuration with strip andwithout it are compared (contours of vertical velocityare given to better point out the differences in the wake).The elongation of the wake for the strip in comparisonwith the no-strip case is here more clearly demonstratedby considering the regions of negative vertical velocitiesfrom the shear layer (in blue) and the positive vertical

velocity from the venting duct (in red).The experimental results can be coupled with the full

field results obtained with the numerical computations;in Fig. 7, such a field is shown for the non-vented andvented conditions. As can be observed from the figures,the general features of the velocity field are captured bythe code; however, different from experimental data, thewake is attenuated in the non-vented case (Fig. 7a),while in the case of venting duct, the jet exit velocityappears small (in comparison with outer velocity)throughout the wake (Fig. 7b). This result is the best

that can be obtained using different turbulence and wallmodels; also due to the results observed in the shearlayer, the sensitivity of the numerical code to the usedmodels seems to be high.

Velocity profiles can be derived from previous two-dimensional experimental and numerical data; in Fig. 8,the streamwise component of velocity along transverseand longitudinal directions in the wake of the model areshown. From the transverse profiles immediatelydownstream of the vented models at x/D = 0.25(Fig. 8a), the overlapping between the different tech-niques is poor. For HWA, this is due to the unresolvedvelocity direction in the wake (no negative velocities),

while for the Pitot tube this is due to the poor spatialresolution of the method (which smoothes the differ-ences in the profile). It is well known that these tech-niques are not useful to measure a velocity deficit in thewake of a bluff body where low or reversing velocitiescan occur; it is also important to consider that the datafrom the different methods are acquired in rather dif-ferent conditions (Reynolds number equal to about3.4 105 for the Pitot and HWA measurements andequal to 1.7 105 for PIV). This will be particularlyrelevant when considering rms profiles in the following.

unvented venteda) b)

unvented ventedc) d)

x (x/D)

y

(y/D)

0.5 1 1.5 2

x (x/D)

0.5 1 1.5 2

x (x/D)

0.5 1 1.5 2

x (x/D)

0.5 1 1.5 2

0

0.25

0.5

0.75

1

1.25

1.5

1.75

y(y

/D)

0

0.25

0.5

0.75

1

1.25

1.5

1.75

y

(y/D

)

0

0.25

0.5

0.75

1

1.25

1.5

1.75

y

(y/D)

0

0.25

0.5

0.75

1

1.25

1.5

1.75

Fig. 5 Averaged vector fields and axial velocity contours from PIV: a non-vented, no-strip configuration; b vented, no-strip configuration;c non-vented, strip configuration; d vented, strip configuration. Re = 1.7 105

26


7/13

It is interesting to compare PIV and numerical results(in this case the difference in Reynolds numbers is muchlower than before); from the last, there is a clear

underestimation of the absolute values of the negativevelocities in the separation region (about one half).Moreover, the profiles for y/D > 0.5 (external flow) are

no-strip stripa) b)

x (x/D)

y

(y/D)

0.5 1 1.5 2

x (x/D)

0.5 1 1.5 2

0

0.25

0.5

0.75

1

1.25

1.5

1.75

y

(y/D)

0

0.25

0.5

0.75

1

1.25

1.5

1.75

Fig. 6 Averaged vector fields and vertical velocity contours from PIV: a vented, no-strip configuration; b vented, strip configuration.Re = 1.7 105

Fig. 7 Vector fields and velocity magnitude (vector colours) in the wake of clean (no-strip) models from numerical computations: a non-vented and b vented conditions. Re = 2.38 105

-0,5

0

0,5

1

0,0 0,5 1,0 1

y/D

,5

x/D=0.25 Piv data

x/D=0.25 Pitot data

x/D=0.25 Hwa data

x/D=0.25 numerical data

-0,4

0

0,4

0,8

1,2

0 0,5 1 1,5 2 2

x/D

,5

u/Uu/U

Unvented clean

Vented clean

Unvented strip

Vented strip

a) b)

Fig. 8 Streamwise velocity profiles: a transverse profile at x/D = 0.25 obtained with different techniques for the clean vented model;b longitudinal profiles at y/D = 0 (centreline) obtained from PIV measurements for the different models

27


8/13

also different; the PIV data show a slight increase invelocity in the shear layer which is not displayed bynumerical results. The former result seems to be rea-sonable due to the increased velocity (in comparisonwith the free stream) around the bluff body model; thediscrepancy with the numerical result is probably due topoor resolution of the numerical grid to describe theexternal boundary layer in detail.

For the longitudinal profiles (Fig. 8b), it can be ob-served that the results for non-vented profiles are inde-pendent of the presence or not of the strip (as noticedalso in Fig. 5). On the other hand, the strip ventedprofiles clearly show an increase in velocity in the nearwake (x/D < 1) in comparison with the clean ventedone (i.e. an increased length of the wake as also derivedfrom Fig. 5). Moreover, for all vented conditions thewake appears to be more extended along the longitudi-nal than in non-vented (wake recover is obtained atlarger values of x/D).

To compare velocity data with direct force measure-ments, the drag, Dr, has been computed from measuredtransverse mean velocity profiles as those presented in

Fig. 8a (on cross-sections S (y, u), u being the azimuthangle) using the Maskells formula, modified by Cum-mings et al. (1996):

Dr q

Z

S

u U1 u dS 1

2q

Z

S

u2 U21 v

2 w2

dS

1

where u, v, w are the velocity components evaluated onthe given cross-section S in the wake (the w componentis unknown for 2 D measurements) and q is the fluiddensity. The drag coefficients have been computed usingvertical profiles (along y), assuming axial-symmetry onaverage (i.e. all mean velocity components are indepen-dent on the azimuth u, the spanwise velocity w 0 oneach y profile and the cross section is given by dS = dyy du). These coefficients have been derived at differentaxial distances (x/D = 1, 1.5, 2); this has been donebecause the distance at which it is correct to perform theevaluation given by (1) is not always specified in theliterature (Oertel 1990). In Table 1, the data are pre-sented for all the models.

Considering the approximate level of the previousrelation (1), the data, for the non-vented configuration,can be considered in good agreement with those from

the balance. On the other hand, for the vented config-urations the differences are much stronger. However,even in this case, a reduction of 5% between vented andnon-vented clean models can be derived from measuredprofiles. It is interesting to note that the data resultingfrom the velocity profiles decreases as the distance fromthe model increases; this is obvious due to the recoveryof the velocity downstream of the body. In practice, theprofiles in the range x/D = 1 to x/D = 1.5 seem to bebetter suited for drag measurements for non-ventedconditions, while for vented conditions the profiles be-tween x/D = 1.5 and x/D = 2 more closely resemblethe data from direct balance measurements. This isreasonable when considering that the wake for ventedconditions is more extended than for non-vented (asconfirmed by Fig. 8b). Thus, this confirms findings fromother authors, i.e. that drag measurements from velocityprofiles in the wake must be derived at locations whichdepend on the model geometry.

The observed differences depend on the pressure field,not evaluated in the present experiments; pressure playsan important role in the evaluation of the drag from

velocity profiles (pressure fields derived from numericalsimulations indicate that pressure recovery is far to beobtained at x/D = 2). For this reason, pressure mea-surements are required in future experiments.

A question arises if the observed overall drag reduc-tion in vented conditions, derives from reduction in formdrag or in skin friction (or both). Indeed, for the presentbluff body models, as pointed out in Sect. 1, the majorcontribution should be expected from the former.However, it is still interesting to consider the effects ofthe venting jet on the boundary layer behaviour. To dothis, the momentum thickness (h) has been evaluatedfrom the streamwise velocity profiles in the shear layer

(y/D = 0.5), from the rear part of the body ( x/D = 0)to the near wake (x/D = 0.5). In Fig. 9, this thickness isreported for the different conditions; from this figure, itis clear that at the separation point (x/D = 0) themomentum thickness (non dimensional by the modeldiameter) is lower for strip in comparison with no-stripconditions

h/D % 0.0146 strip ventedh/D % 0.0149 strip non-ventedh/D % 0.0165 vented cleanh/D % 0.0152 non-vented clean

This means that the shear stress at the wall (and thefriction drag) should be higher for strip conditions dueto laminar-turbulent boundary layer transition. In par-ticular, the lowest value is attained for the strip ventedcondition, i.e. the one which exhibits the largest totaldrag reduction.

Concerning the rate of increase of momentum thick-ness, it is lower for the vented conditions in comparisonwith the non-vented. However, while for the strip ventedcase the rate is constantly lower than the strip non-ventedcase, for the clean vented case the rate is initially as high

Table 1 Comparison between drag coefficients obtained from PIVvelocity profiles for the different models. Direct balance measure-ments are reported for comparisons

Distances Non-ventedclean

Ventedclean

Non-ventedstrip

Ventedstrip

1D 0.384 0.332 0.364 0.3561.5D 0.281 0.275 0.254 0.2652D 0.190 0.211 0.178 0.200Balance measurements 0.33 0.27 0.37 0.23

28


9/13

as non-vented condition and only from x/D % 0.25 itbecomes lower, probably where effects of the venting jetare sensed. This means that when ventilation is activethere is not a separation as large as without ventilation(in agreement with the findings of smaller vorticalstructures for the vented case). Thus, there is a net de-crease in form drag and a slight increase in friction dragfor the vented conditions. Of course, for a bluff body theincrease in friction drag is much less relevant than thedecrease in form drag, so that the net effect is a reductionin total drag.

To better understand the fluid dynamic fields sur-rounding the models, PIV measurements in front of thebodies have also been performed (Fig. 10). Differencesare observed only just in front of the models, at the ductinlet (x/D = 0); far upstream the differences in the fluiddynamic field are negligible; as already reported, themomentum differences between the two cases leads to a

difference not higher than 2% (difference in cross-sec-tions). These results agree almost perfectly with numeri-cal simulations in front of the bodies as shown in Fig. 11.

Therefore, the measured differences between ventedand non-vented models indicate that the drag reduction

mechanism in the vented models is only partially due tomodifications in front of the body, whereas the maineffect is in the bluff body wakes.

3.3 Turbulence and Reynolds stress measurements

It is important to have information about the modifi-cation of the turbulent kinetic energy in the wake; inFig. 12, the streamwise rms velocity profiles (normalisedwith the free-stream velocity) are given for measure-ments with HWA and PIV. In the near wake (x/D = 0.25), the strong additional peak for the ventedcase (at y/D = 0.1, Fig. 12c) reveals an increase in tur-bulence of the near wake in comparison with the non-vented case (Fig. 12a). The shear layer peak (at y/D = 0.5) is slightly reduced in the vented case, thusconfirming an overall redistribution of the kinetic en-ergy. There are quite strong differences between the re-sults obtained with HWA and PIV; the latterunderestimate the level of fluctuations of about 20%.

This is due to the already considered limitations inHWA techniques in measuring wake flows and to dif-ferences in Reynolds numbers between PIV and HWAtests (about a factor 2).

For the profile at x/D = 2 (Fig. 12b, d), the turbu-lence levels are reduced when the venting duct is active;in the non-vented case, the maximum (at y/D = 0.35) isabout 0.16 (for HWA) and 0.14 (PIV), while for thevented case it is about 0.15 (HWA) and 0.13 (PIV) (andalso moved to y/D = 0.3, i.e. slightly closer to thecentreline). This reduction is observed also at thecentreline and in the outer layer; the overall picturedisplays a reduction in the wake turbulence levels and

amplitude thus confirming the observed reduction ofturbulence in the wake for the case of the sphere.However, for a bluff body, the amount of reduction islower than for the sphere even if still interesting forpractical applications.

0,15 0,30 0,45

0,010

0,015

0,020

0,025

0,030

0,035

x/D

/D

vented

non-vented

strip vented

strip non-vented

Fig. 9 Momentum thickness behaviour along the streamwisedirection obtained from PIV velocity measurements for thedifferent models

0

a) b)

x (x/D) x (x/D)

y

(y/D)

0.51 00.51

0

0.25

0.5

0.75

1

1.25

y

(y/D)

0

0.25

0.5

0.75

1

1.25

Fig. 10 Averaged vector fields and axial velocity contours in front of clean (no-strip) models from PIV: a non-vented and b ventedconditions. Re = 1.7 105

29


10/13

In Fig. 13, the Reynolds stress obtained from PIV forthe different configurations are shown. Two main pointscan be noted. First, the vented configurations presentsmoothed local minima with respect to the non-ventedconfigurations (compare Fig. 13b, d with Fig. 13a, c).Secondly, a complete redistribution of the stresses overthe whole wake (and a general reduction in the wakeamplitude) takes place.

In particular, a slight downstream displacement canbe noted for the local maxima and minima of the stressin vented strip configuration with respect to the ventedno-strip one (compare Fig. 13d with b). These findingsconfirm the reduction in turbulence fluctuations and inthe amplitude of the wake already observed from othermeasurements.

Fig. 11 Vector fields and velocity magnitude (vector colours) in front of clean (no-strip) models from numerical computations: a non-vented and b vented conditions. Re = 2.38 105

Absolute turbulent intensity I - x/D = 0.25

0

0,05

0,1

0,15

0,2

0,25

0,3

0 0,4 0,8 1,2 1,6

(y/D)

I

I Piv data

I Hwa data

Absolute turbulent intensity I- x/D = 2

0

0,05

0,1

0,15

0,2

0,25

0,3

0 0,4 0,8 1,2 1,6

(y/D)

I

I Piv data

I Hwa data

a) b)

Absolute turbulent intensity I - x/D = 0.25 Absolute turbulent intensity I - x/D =2

0

0,05

0,1

0,15

0,2

0,25

0,3

0 0,4 0,8 1,2 1,6

(y/D)

I

I Piv data

I Hwa data

0

0,05

0,1

0,15

0,2

0,25

0,3

0 0,4 0,8 1,2 1,6

(y/D)

I

I Hwa data

I Piv data

c) d)

Fig. 12 Turbulence intensity profiles measured by PIV and HWA; non-vented models at a x/D = 0.25 and b x/D = 2; vented models at cx/D = 0.25 and d x/D = 2

30


11/13

3.4 Spectra

From HWA data it is also possible to evaluate powerspectral densities of the velocity at different points of thewake (the data rate of the HWA data is sufficiently highto detect vortical structures from the shear layer andfrom the venting duct). In Fig. 14, three spectra ob-tained from measurements in the shear layer (x/D = 0.5, y/D = 0.45) (x/D = 1, y/D = 0.45) and atthe centreline (x/D = 0.5, y/D = 0) are given for thestrip configuration in vented and non-vented conditions(in the horizontal axis the Strouhal number, St = fD/Urather than the frequency f alone is given). In the shearlayer close to the model (Fig. 14a), for the non-ventedcase, a distinct peak at St % 0.22 is obtained; this is justthe value expected for wake vortices released behindbluff bodies, i.e. for vortex shedding conditions. Theflow visualisations in Fig. 2b also confirm the presenceof vortex shedding patterns at the considered HWAmeasurement position (x/D = 0.5, y/D = 0.45). More-over, as already observed, the instantaneous vector plotsfrom PIV measurements (Fig. 4) allow to detect (al-

though only qualitatively) large scale vortices in the nearwake (indicated by the orange circles in Fig. 4a). On theother hand, in the vented case, the peak disappearsindicating that large scale shedding vortices are sup-

pressed; this is also confirmed by visualisations (Fig. 2c)and PIV instantaneous measurements (Fig. 4b). Fromthe analysis of Fig. 14a, in the vented case, the energyrelated to large scale vortices seems to be redistributedtowards high frequency vortical structures. To quantifythis redistribution, spectral slopes are computed fromthe data; differences among vented and non-ventedconditions are revealed. For the non-vented case (whichhas a single shear-layer), a 5/3 slope definitely appears,thus indicating some approach to local equilibrium. Forthe vented case (with interaction of a outer and innershear layer due to the presence of the vented jet), no suchequilibrium is evident, as shown by the smaller slope(about 0.7), indicating energy production at differentwave numbers.

For the non-vented data obtained at the secondlocation downstream in the shear layer (Fig. 14b), theprimary peak observed in the non-vented case is atten-uated, while the second peak at St % 0.16 (probably dueto secondary vortical structures developing in the shearlayer) is emphasised. Also the extension of the region inwhich the 5/3 slope is observed is reduced. Moreover,

the whole large scale contribution for the vented data isattenuated. The picture is similar to that derived fromthe previous location although at this position thespectral redistribution of the kinetic energy takes place

Fig. 13 Contours of Reynolds stress measured by PIV: a non-vented, no-strip; b vented, no-strip; c non-vented, strip; and d vented, strip

31


12/13

for both vented and non-vented (on a lower extent)conditions.

In the spectra obtained at the centreline (Fig. 14c),large scale contributions almost disappear for the non-vented case while are still present for the vented case; thisis due to vortical structures appearing in the second shearlayer generated at the borders of the vented jet. As pre-viously, this aspect is qualitatively confirmed by theanalysis of flow visualisations (Fig. 2) and PIV instan-taneous measurements (Fig. 4b). Thus, the results fromspectral analysis point out a strict link between thebehaviour in space (which differs from the external to theinternal shear layer) and the redistribution of kineticenergy in frequency domain for the vented in comparisonto the non-vented case; this confirms the findings derived

from direct balance and PIV-HWA measurements.

4 Conclusions

In this work, the passive (or natural) ventilation tech-nique, applied to a bluff body has been studied by meansof different experimental and numerical techniques. Forthe models investigated (without and with turbulenceinducing strip on the surface) a drag reduction fromdirect balance measurements has been valuated both for

the clean (78%) and the strip configuration (20%).Therefore, the technique seems to work better on rough

(i.e. strip) rather than smooth (clean) surfaces; this isquite important for practical applications where aero-dynamic surfaces are not really smooth and there is al-ways some degree of roughness. The amount of dragreduction is also dependent on the Reynolds numberinvolved; the reduction is particularly effective for Rey-nolds numbers around 5 105. This is an importantpoint to be considered for applications to relatively lowspeed vehicles.

The use of round rear edge models allows a dis-placement of the separation point to take place. This isalso another key point for applications of the techniqueand it can be readily achieved due to the small curvature

radius involved.Experimental and numerical results on the velocity

fields in front and in the wake of the models clarified thephenomena involved; a reduction of the wake width, anincrease of the wake length and a redistribution of thekinetic energy and turbulent (Reynolds) stress have beenmeasured in the vented conditions in comparison to thenon-vented case. This phenomena are mainly due to theinteraction between the external shear layer and the jetflowing out of the venting duct; the net effect is areduction in size and energy of the external vortices by

non-vented

vented

-5/3

-0,74

non-vented

vented

-5/3

-0,72

a) b)

c)

1,0E-04

1,0E-03

1,0E-02

1,0E-01

1,0E-04

1,0E-03

1,0E-02

1,0E-01

1,0E-04

1,0E-03

1,0E-02

1,0E-01

1,0E-03 1,0E-02 1,0E-01 1,0E+00 1,0E+01

1,0E-03 1,0E-02 1,0E-01 1,0E+00 1,0E+01 1,0E-03 1,0E-02 1,0E-01 1,0E+00 1,0E+01

non-vented

vented-5/3

-0,73

Fig. 14 Power spectral densities (and derived slopes) as a function of Strouhal number obtained from HWA in strip conditions ata x/D = 0.5 and y/D = 0.45, b x/D = 1 and y/D = 0.45, c x/D = 0.5 and y/D = 0, non-vented (blue) and vented (red) conditions.Re = 3.3 105

32


13/13

coupling with the counter rotating vortical structuresfrom the jet. The analysis of spectra in the wake of themodels confirms that in the vented case kinetic energy issubtracted from the low frequency vortical structurestowards the high frequency ones except for the regionclose to the venting jet (where another shear layerdevelops). Negligible differences have been observed infront of the vented or non-vented models, so that the gainin momentum due to the venting jet can account only fora small portion of the observed drag differences (a crudeevaluation gives about 2% gain for the vented case). Onthe other hand, the external boundary layer and the shearlayers in the wake seem to be dependent on the presenceof the venting jet. A slight increase in friction drag shouldbe expected in venting conditions; however, this increaseis very small in comparison to the strong decrease in formdrag (as expected for a bluff body).

There is a substantial agreement between the experi-mental and numerical results (thus confirming theapplicability of commercial advanced codes to this sortof problems), even if some underestimation of the wakedefect and of the shear layer thickness is obtained from

the numerical results.The evaluation of the force made from the velocity

profiles (considering that the applicability of this evalu-ation is still debated especially from the point of view onthe downstream distance to be considered) confirms theobserved reduction, although by a smaller magnitude

(about 5%); in any case, the amount appears interestingfor practical applications. Pressure measurements overthe whole model surface are planned for the future forbetter quantifying the drag reduction mechanism.

References

Achenbach E (1972) Experiments on the flow past spheres at very

high Reynolds numbers. J Fluid Mech 54:565575Achenbach E (1974) Vortex shedding from spheres. J Fluid Mech

62:209221Cummings RM, Giles MB, Shrinivas GN (1996) Analysis of ele-

ments of drag in three-dimensional viscous and inviscid flow,AIAA, paper no 962482

Gad El Hak M (2000) Flow control: passive, active, and reactiveflow management. Cambridge University Press, Cambridge

Monkewitz PA (1992) Wake control, in bluff body wakes; dynamicand instabilities. In: Iutam symposium, Gottingen

Oertel H Jr (1990) Wakes behind blunt bodies. Annu Rev FluidMech 22:539562

Rae WH, Pope A (1984) Low-speed wind tunnel testing. WileyInterscience, New York

Roshko A (1961) Experiments on flow past a circular cylinder atvery high Reynolds number. J Fluid Mech 10:345356

Suryanarayana GK, Meier GEA (1985) Effects of ventilation onthe flow field around a sphere. Exp Fluids 19:7888

Suryanarayana GK, Pauer H, Meier GEA (1993) Bluff body dragreduction by passive ventilation. Exp Fluids 16:7381

Suryanarayana GK, Prabhu A (2000), Effect of natural ventilationon the boundary layer separation and near-wake vortex shed-ding characteristics of a sphere, Exp Fluids 29:582591

33

Falchi Provenzano Pietrogiacomi Romano 2006 EiF

Documents

Transcript of Falchi Provenzano Pietrogiacomi Romano 2006 EiF