Sunteți pe pagina 1din 14

Experimental Thermal and Fluid Science 31 (2007) 825838 www.elsevier.

com/locate/etfs

The inuence of nozzle aspect ratio on plane jets


R.C. Deo *, J. Mi, G.J. Nathan
Turbulence Energy and Combustion (TEC) Research Group, School of Mechanical Engineering, The University of Adelaide, SA 5005, Australia Received 19 March 2006; received in revised form 14 August 2006; accepted 30 August 2006

Abstract This paper reports a systematic investigation of the eect of nozzle aspect ratio (AR) on plane jets. The aspect ratio AR ( w/h, where h and w are the nozzle height and width) was varied from 15 to 72. The present velocity measurements were performed using single hotwire anemometry, over a downstream distance of up to 85h and at a nozzle-height-based Reynolds number of Reh = 1.80 104. Results obtained reveal that both the extent and character of statistical two-dimensionality of a plane jet depend signicantly on AR. Most aspects of the near eld ow exhibit an asymptotic-like dependence on AR, but do not become independent of AR within the range of AR investigated. A region of statistically two-dimensional (2-D) mean velocity eld is achieved only for AR P 20, and its axial extent increases with AR. However, the centerline turbulence intensity in the far eld displays an asymptotic-like convergence only for AR P 30. In the self-similar region, both the mean decay and spreading rates of the jet increase as AR increases and do not reach an asymptotic value, even at AR = 72. The aspect ratio of the local jet (w/local velocity half-width) at the end of the 2-D region becomes asymptotically independent of nozzle aspect ratio, for approximately AR P 30. That is, the plane jet ceases to be statistically 2-D at a xed value of local jet aspect ratio for nozzle aspect ratios greater than 30. The skewness and atness factors also depend on AR. These results imply that independence of AR, even in the near eld, will require very much larger aspect ratios than have been investigated previously. Crown Copyright 2006 Published by Elsevier Inc. All rights reserved.
Keywords: Plane jet; Turbulent mixing; Eect of nozzle aspect ratio; Hot-wire measurement

1. Introduction Since the work of Schlichting [1], the plane jet has received signicant attention (e.g., Heskestad [2], Bradbury [3] and Gutmark and Wygnanski [4]). In experiments, a plane jet is produced by a slender rectangular slot and two parallel plates, known as sidewalls, attached to the slots short sides. The presence of sidewalls restrict the jet from developing in the spanwise direction (normal to the short sides) so that it is expected to achieve a statistically two-dimensional (2-D) ow over a suciently great downstream distance into the far eld. It follows that the nozzle aspect ratio, AR ( w/h, where w and h are the long and
* Corresponding author. Previous address: School of Engineering and Physics, Faculty of Science and Technology, The University of the South Pacic, Fiji. Tel.: +61 8 83035460; fax: +61 8 8303 4367. E-mail address: ravinesh.deo@mecheng.adelaide.edu.au (R.C. Deo).

short sides of the slot), if suciently high, is believed to have negligible inuence on the downstream development of a plane jet. Because of this, previous studies, e.g., Bashir and Uberoi [5], Van der Hegge Zijnen [6] have all claimed that they investigated the 2-D plane jet, although using quite dierent values of AR, varying from 20 to 144. The dominant mean motion of a plane jet is in the streamwise (x) direction and its secondary spread is in the lateral (y) direction. No entrainment can occur in the spanwise (z) direction due to the presence of sidewalls, placed in the xy plane, that are necessary to achieve the statistical two-dimensionality of a plane jet [13,14], over a reasonably large axial distance. These sidewalls dierentiate them from rectangular jets, a term that is commonly used for the case when the jet has no sidewalls, although such jets are only truly rectangular at the exit plane. The analytical investigations of George [7] and George and Davidson [8] and the experimental work on a round

0894-1777/$ - see front matter Crown Copyright 2006 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.expthermusci.2006.08.009

826

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

Nomenclature AR nozzle aspect ratio (AR = w/h) ARlocaljet local jet aspect ratio ( w/y0.5 local) at x2-D using local half width, y0.5local Fu centerline atness factor, F u hu4 i=hu2 i2 max Fu maximum value of centerline atness factor F1 asymptotic value of centerline atness factor u h height of a plane nozzle Ku decay rates of mean centerline velocity Ky jet spreading (widening) rate Re Reynolds number Reh  Uo,ch/t Su centerline skewness factor, Su = hu3i/(hu2i)3/2 min Su minimum value of centerline skewness factor S1 asymptotic value of centerline skewness factor u u uctuation component of mean velocity in streamwise (x) direction u0 root-mean-square (rms) of the velocity uctuation, u 0 = hu2i1/2 0 un;c normalized centerline turbulence intensity, u0n;c u0c =U c 0 uc;1 asymptotic value of centerline turbulence intensity Uc Uo,b Uo,c Un,c local mean velocity on the centerline exit bulk mean velocity mean exit centerline velocity normalized centerline mean velocity (Un,c = Uc/Uo,c) w width of a plane nozzle x01 virtual origin of the normalized mean centerline velocity x02 virtual origin of the normalized velocity halfwidth xp length of the jets potential core x2-D downstream distance that represents the end of the 2-D region y0.5 velocity half-width, calculated at the y-location at which U x 1 U c x 2 m kinematic viscosity of the air, m % 1.5 105 m2 s1 at 20 C ambient conditions x, y, z streamwise or axial (x), lateral (y) and spanwise or transverse (z) coordinate

and plane jets [911] have shown that the asymptotic state of turbulent ows, e.g., a round or a plane jet, depend upon their initial and boundary conditions. By normal convention, the choice of the value of AR becomes a boundary condition of a plane jet. Gouldin et al. [12] have assumed that the magnitude of AR may govern aspects of the downstream behavior. However, to our best knowledge, no systematic examination of the eect of aspect ratio on either the behavior or extent of the statistically 2-D ow region is currently available. Much more information on the eects of nozzle aspect ratio on downstream development is available for rectangular jets than for plane jets. This can provide useful insights into plane jets, since both these jets have an initial quasi-planar (2-D) ow [15], although the exit conditions of both ows are fundamentally dierent. Nevertheless, within this quasi-planar region, rectangular jets appear to behave broadly similarly to plane jets, thus they may well be used to study a plane jet behavior of varying AR. Trentacoste and Sforza [16] pioneered the investigation of aspect ratio eect on rectangular jets over the range 2.5 6 AR 6 100. Their study found an increase in length of the jets potential cores, a reduction in the decay of mean centerline velocity and a reduction in jet spreading rate with an increase in nozzle aspect ratio. A decade later, Sfeir [17] measured jets from rectangular nozzles of aspect ratios over the range 10 6 AR 6 60. He obtained similar results to Trentacoste and Sforza [16]. Marsters and Fotheringham [18] also obtained similar trends for smaller nozzle aspect ratios between 1.88 and 3.39. Then followed many other studies, such as those of Krothapalli et al. [19],

Tsuchiya et al. [20] and Quinn [21], all of which focused on aspect ratio issues in rectangular jets. Based on his velocity measurements, Quinn [21] found that the near-eld spreading rate increases with increasing nozzle aspect ratio from AR = 5 to AR = 20. He attributed a deduced increase in jet mixing to an increase in three-dimensionality of the entire jet. The ow visualization of rectangular jets (of AR ranging from 2 to 10) by Tsuchiya et al. [22] found a shorter potential core and an increase in spreading rate for AR = 10 relative to AR = 2. By contrast, the measurements of the mass ow rate of a rectangular jet at exit Mach number 0.95 by Zaman [23] for AR = 238 suggested that the entrainment rate increases signicantly with nozzle aspect ratio only for AR P 8. More recently, Mi et al. [15] found that rectangular jets of AR = 15120 can be characterized by three distinct zones: an initial quasi-plane jet zone, a transition zone and a nal quasiaxisymmetric-jet zone. Importantly, the extent of the quasi-plane zone was found to depend on nozzle aspect ratio. In synopsis, previous investigations of rectangular nozzles reveal a distinct dependence of the ow evolution on nozzle aspect ratio. There are only a few previous investigations that directly investigated the eect of nozzle aspect ratio on the evolution of plane jets. Bashir and Uberoi [5] provided by far the best indication of the eect of nozzle aspect ratio in a plane jet. Their measurements in the ow from a plane nozzle with a smoothly contoured exit at Re = 2770 and AR = 20, 44 and 144 show that the AR = 144 jet decays at the highest rate (Fig. 1) and its asymptotic value of turbulence intensity is the lowest of these jets (Fig. 2). However, they provide no

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

827

Sym.

Authors

AR

Re

25

20

(Uo,b/Uc)2

Heskestad [2] 120 36.900 Hitchman et al [26] 60 7.000 Bashir & Uberoi [5] 144 2.700 Bashir & Uberoi [5] 44 2.700 Bashir & Uberoi [5] 20 2.700 Bradbury [3] 48 30.000 Thomas & Goldschmidt [27] 46 6.000 Browne et al. [13] 20 7.700

15

10

0 0 20 40 60 80 100

x/h
Fig. 1. Streamwise evolutions of centerline velocity of previous investigations of plane jets.

0.3

uc'/ Uc

0.2
Sym. Authors AR Re 18.000 2.700 2.700 2.700 30.000 6.000 7.700

0.1

Heskestad [2] 120 Bashir & Uberoi [5] 20 Bashir & Uberoi [5] 40 Bashir & Uberoi [5] 144 Bradbury [3] 48 Thomas & Goldschmidt [27] 46 Browne et al [13] 20

0 0 20 40 60 80 100 120

x/h
Fig. 2. Streamwise evolutions of centerline turbulence intensity of previous investigations of plane jets.

data in the near eld, and their Reynolds number is well below the value of 10 000 argued by Dimotakis [24] to be necessary to achieve a fully turbulent state. A similar trend was found by Van der Hegge Zijnen [6], but over a much smaller range of aspect ratio for AR = 20 and 25, and again their data were limited to the far eld. These papers also do not provide other details, such as the eects of aspect ratio on the higher order statistics. There are also some apparent discrepancies. The plane nozzle of Heskestad [2] measured at AR = 120 produced the highest mean velocity decay (Fig. 1) and the largest asymptotic turbulence intensity (Fig. 2) of all previous investigations, although it is not quite the largest nozzle aspect ratio. This apparent inconsistency with Bashir and Uberois [5] ndings is probably explained by the use of a sharp-edged orice slot by Heskestad [3], but this dierence and others prevents a denitive comparison. In addition, although there is reasonably good agreement between the mean velocity decay of Bashir and

Uberoi [5] and Browne et al. [13] (both measured at AR = 20, Fig. 1), their turbulence intensities are discernibly dierent (Fig. 2). This is primarily due to dierences in Reynolds number (Namar and Otugen [25], Deo [9]), although dierences in other initial conditions could also play a signicant role. For instance, note that Bradbury [3] used the highest Reynolds number but obtained second lowest turbulence intensity. The apparent discrepancy may be attributable to the use of a co-ow (co-ow to jet velocity % 16%), use of a higher nozzle aspect ratio (AR = 48) or any associated dierences between the nozzle surface nish (for instance, rough or smooth). Given the expected dependence of a jet ow on all boundary conditions [710], and the range of dierent boundary conditions used by previous investigators, it is therefore not possible to obtain a detailed picture of AR eects on plane jets from previous investigations. Hence a primary aim of the present investigation is to determine in detail, the dependence of the evolution of a plane jet on nozzle aspect ratio by a systematic variation of only this parameter. Further issues that remain unexplored are deduced from careful examination of results of previous investigations. Fig. 1 shows the decay of mean centerline velocity of plane jets, as reported by previous investigators. As is well known, a relationship of the form Uc $ x1/2 (where the exponent % 1/2 is a necessary indicator of a 2-D jet) is evident in the self-similar region. However, a close examination reveals that axial extent over which data is reported increases with nozzle aspect ratio. For instance, Heskestad [2], Hitchman et al. [26], Bradbury [3] and Thomas and Goldschmidt [27] used AR > 45, and present velocity data for the range x/h 6 100. In contrast, the low-aspect ratio plane jets of Bashir and Uberoi [5] and Browne et al. [13], for AR = 20, only report data to the range x/h = 55 and 40, respectively. This suggests that they have only reported data for the axial range where the jet was statistically two-dimensional, perhaps prior to the secondary inuence of sidewalls becoming signicantly large. Based on the review of previous data, Gouldin [12] noted that the axial extent of two-dimensionality is expected to depend on AR. This notion is particularly relevant for the data of Bashir [28] and Bashir and Uberoi [5], who performed measurements using three nozzles of aspect ratios 20, 40 and 144, so their measurement range obviously appears not to be limited by experimental apparatus. For instance, for the case of AR = 20, their data extends up to x/h = 30 whereas for AR = 144, their data extends up x/h = 80. Were a dependence of the maximum planar ow (x2-D) region on AR be conrmed to be genuine, one could establish a possible relationship between the width of the local plane jet relative to the spacing between the sidewalls at which the transition from 2-D to 3-D ow occurs. Any such previously unidentied relationship would be of wide general signicance to other planar ows, e.g., 2-D wakes, and would be a useful test for two-dimensional models. Hence another aim of the present research is to provide some insight into this issue.

828

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

2. Experiment details The plane nozzle facility, shown schematically in Fig. 3, consists of an open circuit wind tunnel driven by a 14.5 kW aerofoil-type centrifugal fan, a wide angle diuser, with ow straightening elements (including a honeycomb and screens) and a smooth contraction exit of area ratio 6:1. The honeycomb has its cells aligned in the ow direction and attempts to reduce mean or uctuating variations in spanwise velocity while producing minimal eect on streamwise velocity because of the small pressure drop. Likewise, the screens help reduce the velocity defect in the turbulent boundary layer that passes through it, further streamlining the incoming airow. The present screens have an open area ratio of % 60% and the smooth contraction is based on a polynomial curve. Two at plates were mounted to the end of the windtunnel contraction, with radially contracting long-sides, with two parallel plates, known as sidewalls, attached to the slots short sides, to create a plane nozzle. The height of the nozzle was xed at h = 10 mm for these measurements and the width (separation between the sidewalls) was varied from w = 150 mm to w = 720 mm to achieve seven aspect ratios of AR = 15, 20, 30, 40, 50, 60 and 72.

The inner radius of the long side of nozzle is r = 36 mm so that r/h = 3.60. The facility was mounted horizontally, with the plane nozzle located near to the mid point between the oor and ceiling in an experimental laboratory of dimensions 18 m (long) 7 m (wide) 2.5 m (high). The distance from the jet exit to the front wall of the laboratory was % 1400h and between the jet and the ceiling/oor was % 125h, allowing the unheated jet to discharge into still air freely. Based on the approach of Hussein et al. [29], the eects of room connement is estimated to produce less than 0.5% momentum loss for AR = 1572 at a downstream distance of 85h. Hence the present jets closely approximate plane jets in an innite environment. For all cases of the present investigation, the jet exit centerline mean velocity was xed at Uo,c % 27 m s1 which results in a Reynolds number Reh % 1.80 104. Velocity measurements were performed over the region 0 6 x/h 6 85 using single hot-wire anemometry in an isothermal laboratory of ambient temperature 20 C 0.1 C. To avoid aerodynamic interference of the prongs on the hot wire senor, the present probe was carefully mounted perpendicular with prongs parallel to the plane jet. In addition, the hot wire was aligned properly so that

Fig. 3. A schematic view of the present experimental setup. Dimensions are not to scale.

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

829

it corresponds closely to the streamwise component of the ow velocity to minimize directional ambiguity, although it is not possible to eliminate this eect completely. The single hot-wire (tungsten) sensor was 5 lm in diameter and % 0.8 mm in length, aligned in parallel to the long nozzle sides. The overheat ratio of the hot-wire sensor was 1.5, and a square wave test revealed a maximum frequency response of 15 kHz. Hot-wire calibration was conducted using a standard Pitot tube, placed side by side with the hot-wire probe, at the jets exit (x/h = 0), where the turbulence intensity was 6 0.5%, before and after measurements of each AR. Both calibration functions were tested for discrepancies, and if velocity drift exceeded 0.5%, the experiment was repeated. No further corrections were applied, thus it is expected that measurements away from the centerline (in the outer region of the jet) are signicantly in error, because of high-velocity uctuations relative to the mean value. Nevertheless, the central aim here is to compare the measurements of one experiment with another, with most data taken on jet centerline for each case of AR. While converting data points from voltages to velocities using a fourth order polynomial curve similar to the one proposed by George et al. [31], the average accuracy of each calibration function was 0.2%. Signals obtained were low-pass ltered with an identical cuto frequency of fc = 9.2 kHz to eliminate highfrequency noise at all the measured locations. The voltage signals were oset to a range of 03 V (as a precautionary measure that no clipping of the signal occurs [32]). They were then amplied appropriately through the circuits, and then digitized on a personal computer at fs = 18.4 kHz via a 16 channel, 12-bit PC-30 F A/D converter of signal input range 05 V. The sampling duration was about 22 s, in which at approximately 400 000 (instantaneous) data points were collected. Using the inaccuracies in calibration data and observed scatter in the measurements, the present random uncertainties correspond to a mean error of 4% at the outer edge of the jet and 0.8% on the centerline. The errors in the centerline rms velocities were found to be 1.8%, and in the skewness (Su) and atness (Fu) factors up to 3% and 8%, respectively. 3. Jet exit conditions The exit velocity proles were obtained at x/h = 0.2 in each isothermal air jet using nozzles of AR = 1572. Fig. 4(a) and (b) show that measurements near the exit plane conrm that these nozzles, like conventional smoothly-contracting ones, produce a quasi-top-hat mean velocity proles, which are uniform (U/Uo,c % 1) for all AR within 5% and a central-region turbulence intensity between 0.8% and 1.8%. Using these gures, the exit conditions are dened by calculations of the boundaryR h=2 layer displacement thickness, dd 0 1 U =U o;c dy and R h=2 momentum thickness, hm 0 1 U =U o;c U c =U o;c dy. As AR is varied from 15 to 72, the present values of dd and hm vary from 0.2h to 0.1h and 0.08h to 0.14h, respectively.

1.2 1.0 0.8

U / Uo,c

0.6

Sym
0.4 0.2 0 0 0.1

AR 15 20 30 50 72
0.2 0.3 0.4 0.5 0.6 0.7

y/h

0.15

0.10

u'/Uo,c
0.05 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

y/h
Fig. 4. Lateral proles of the (a) normalized mean velocity and (b) turbulence intensity, measured x/h = 0.2.

Likewise, the corresponding shape factors, H = dd/hm, which often represents the atness (uniformity) of the mean velocity proles [33] lie between 2.4 and 2.5 for AR = 15 72, compared with a value of % 2.6 for a Blasius exit velocity prole. Thus the present plane nozzles may be characterized as having an initially laminar boundary layer since the shape factors are close to that of a Blasius prole (Schlichting [33]).

4. Results and discussion Fig. 5 presents, in loglog form, the mean streamwise velocity, Un,c = Uc/Uo,c along the jet centerline for all nozzle aspect ratios. Note that successive proles are vertically oset for clarity. The length of the jets potential core, xp, estimated by the maximum axial (x) distance at which Uc(x) = 0.98Uo,c is a function of nozzle aspect ratio. Apparently, as AR is increased from 15 to 72, xp increases asymptotically (as shown later in Fig. 18) from xp % 2h to xp % 5h. Since the magnitude of xp is one measure of the near-eld entrainment rate, this implies that the rate of

830

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838
170 150 130

Un,c= Uc / Uo,c [arbitrary scale]

110

x2-D / h

90 70 50 30
Heskestad [2] Bashir & Uberoi [5] Hitchman et al. [26] Thomash & Goldschmidth [27] Bradbury [3] Miller and Comings [35] Browne et al. [13] present data

Uc~ x

-1/2

AR = 15
20

10 0

30 40 50 60 72
1 10 100

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

AR
Fig. 7. The normalized maximum downstream distance, x2-D/h, up to which the mean velocity eld retains statistical two-dimensionality. Values from previous investigations are estimated from axial extent of data presented. (See also Refs. [2,5,26,27,3,35,13].)

x/h
Fig. 5. The evolution of centerline mean velocity, Un,c = Uc/Uo,c from nozzles of dierent AR. Note the y-ordinate is arbitrary.

1.0

AR = 15 30 50 72

0.8

0.6

0.4

0.2

0 0 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00

y/h
Fig. 6. Lateral proles of the mean velocity, U/Uc for AR = 1572 measured at x/h = 3.

entrainment of the present plane jets increases as AR decreases. This is further conrmed by lateral proles of the mean velocity in the potential core region (at x/ h = 3), shown in Fig. 6, where a higher jet spreading rate is obtained for AR = 15 than for AR = 72. Further interesting observations are made from Fig. 5, which shows that the jets of AR P 20 all exhibit four distinct regions of decay in Uc. Closest to the nozzle is the well-known non-decaying potential core region. The jet then undergoes a transition into the power-law-decaying region, where Uc $ x1/2. This is followed by the fourth region in which Uc does not appear to obey any systematic dependence on x. The departure from Uc $ x1/2 in the

fourth region is evidence of emerging three-dimensional eects, presumably induced directly and indirectly by sidewalls. For AR = 15, there appears to be no obvious region in which Uc $ x1/2. This reveals that an aspect ratio of 15 is perhaps insucient to achieve a statistically 2-D jet. That the present exponent (=1/2) is independent of AR stands in contrast to previous measurements of rectangular jets, since both Trentacoste and Sforza [16] and Sfeir [17] found the magnitude of their exponent to be dependent on both the nozzle geometry and aspect ratio. The reason for this apparent discrepancy is presently unclear, but it seems to underline the need for sidewalls to achieve a statistically 2-D jet. One must note from Fig. 5 that the axial extent of the 1/2-power-law region increases with AR, as indicated by the length of the dashed lines. In addition, it is also shown that AR has an inuence on the extent of other ow regions. To establish a relationship, if any, between the maximum planar ow region (denoted as x2-D) on AR, Fig. 7 plots x2-D/h against various AR. Note that for previous investigations, we have estimated x2-D from the axial range of their data, assuming that their Uc $ x1/2 was provided up to this downstream distance. It is clear from Fig. 7 that the present data set, in which only AR is varied, exhibits an excellent t to a linear scaling of x2-D/h $ AR. That is, the present data is well represented by a linear equation of the form x2-D mAR C, where m and C h are deduced to be 1.40 and 2.10, respectively. Some scatter between the present and previous measurement is evident, although this is quite expected, given that no previous investigators directly measured x2-D. Nevertheless, the agreement between present and previous data is sucient to establish condence that the scaling of x2-D $ AR is generic. Further, the actual values of m and C are unlikely to be universal, but rather can be expected to depend upon on

U/Uc

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838
10 18

831

present data
8

16 14

ARlocal jet= w / y0.5 local

Bashir [5]

12

AR = 15 20 30 40 50 60 72

(Uo,c / Uc)2
100

Van der Hegge Zijnen [6]


6

10 8 6

Heskestad [2]
4 2 4 10 20 50 0

AR (= w/h)
Fig. 8. The variation of local jet aspect ratio ARlocaljet against nozzle aspect ratio AR, calculated at x2-D for the present plane jets.

10

20

30

40

50

60

70

80

90

x/h
Fig. 9. The normalized proles of centerline mean velocity for nozzle aspect ratios 1572.

initial conditions [79], which perhaps contributes to the scatter between the present and past values of x2-D. Next we assess the AR dependence of aspect ratio of the local jet at the end of the 2-D region (x2-D). To facilitate this, we characterize the jet width by its half-width, y0.5, the location at which U x; y 0:5 1 U c x, and dene the 2 local half-width as the value of y0.5local at the location x2-D. Hence the local aspect ratio of the jet at which it ceases to be 2-D is ARlocaljet = w/y0.5local at x2-D. Fig. 8 presents the dependence of ARlocaljet on AR. Two distinct regimes, separated by a clearly dened elbow point become immediately evident. As the nozzle aspect ratio is increased from 20 to 30, ARlocaljet decreases rapidly from 9.08 to 7.16, with ARlocaljet achieving an approximately constant value for AR P 30. Some comparisons of the present measurements with previous investigations are possible, if we retain the assumption used previously to obtain Fig. 7 (i.e., that ARlocaljet is obtained from their furthest downstream data-point). These measurements include Van der Hegge Zijnen [6] for a nozzle of AR = 20, Heskestad [2] for a nozzle of AR = 120 and Bashir and Uberoi [5,28] for a nozzle of AR = 144. The present values of ARlocaljet are in good agreement with two of the three estimates derived from previous work [5,6] and moderate agreement with the other. The observed scatter is consistent with the rather crude method for extracting x2-D from previous measurements. Hence it is clear from Fig. 8 that, for AR P 30, the plane jet undergoes a transition from statistically 2-D to a 3-D behavior at a xed value of local jet aspect ratio. This physically explains that the transition from a 2-D to a 3-D jet occurs when the width of the jet, which grows with axial distance, reaches a critical value relative to the spacing between the sidewalls. In plane jets, the 2-D roller-like counter-rotating large-scale structures, which are symmetric about the jet centerline, dominate the mixing layers which bound the potential core, the interaction region and the self-preserving region (Browne et al.

[34]). The transition from a 2-D to a 3-D jet perhaps implies that at downstream distances greater than x2-D, these large-scale structures, which naturally scale with the width of the jet, undergo a substantial change, say from being roller-like 2-D to more three-dimensional (e.g., similar to those found in rectangular jets). Such a change will probably occur when the size of the large-scale structures matches the spanwise extent of the plane jet. If we consider that the ratio of the half-width to the outer edge of a jet is about three times at x2-D, then the transition from 2-D to 3-D occurs when the spanwise extent of the jet and lateral extent become almost equal. In other words, the spacing between the sidewalls controls this transition, which occurs when the actual aspect ratio of the local jet is approximates to unity. Although Uc $ x1/2 within the 2-D region for all tested cases, the decay of centerline mean velocity for various AR are discernibly dierent (Fig. 9). Clearly, an inversely square relationship of the form (Uo,c/Uc)2 = Ku(x/h + x01/h) is demonstrated, where Ku is a measure of the centerline mean velocity decay rate and x01 is the jets virtual origin. The mean centerline velocity becomes self-similar when x/h P 20 for AR P 20. Consistent with Fig. 5, no obvious 2-D region is evident for AR = 15. The present dependence of Ku on AR is plotted in Fig. 10, along with Ku of some previous investigations. The present values of Ku reveal a clear trend. As AR is increased from 20 to 72, Ku increases approximately linearly from 0.115 to 0.180, with subtle, but denite dierences in Ku for AR = 60 and 72. Such a trend was also found by Bashir and Uberoi [5] for AR = 20144, although their dependence of the Ku on AR appears weaker. The obvious higher values of Ku from Bashir and Uberoi [5] at similar aspect ratios are, at least in part, explained by dierences in Reynolds number. Both Namar and Otugen [25] and Deo [9] found that a low-Reynolds number plane jet produces a higher value of Ku than a high-Re jet when measured

832
0.30

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

0.25

Heskestad [2] Bashir & Uberoi [5]

0.20

0.15

Hitchman etal. [26]


0.10

present
0.05 0 15 30 45 60 75 90 105 120 135 150

AR
Fig. 10. The decay rate of the centerline mean velocity for aspect ratios 1572.

using identical nozzles. Hence, both the present and past measurements reveal a consistent trend that the mean velocity decay depends upon nozzle aspect ratio, even at AR = 72. Lateral proles of the mean velocity, normalised by the centreline value are shown in Fig. 11(a)(e). The maximum axial location of the measured proles span across till x/ h 6 20 for AR = 15 and x/h 6 40 for AR = 20 and 30. For AR = 50 and 72, the proles are measured up to the maximum achievable distance of x/h = 80. In the self-similar region, the proles conform closely to the Gaussian form Un = exp[ln 2(yn)2]. Fig. 11 clearly demonstrates that the distance at which self-similarity is reached decreases with increasing nozzle aspect ratio. For example, the mean velocity proles become self-similar at x/h = 20 for AR = 15, much further downstream than the x/h = 5 for AR = 72. As mentioned previously, and revealed from Figs. 5 and 7, the increasing inuence of the sidewalls, with decreasing nozzle aspect ratio, causes three-dimensional eects to dominate at axial locations closer to the nozzle exit for nozzles of smaller aspect ratios. Correspondingly, a higher-AR jet is associated with a larger axial distance over which the self-similar state is achieved. Fig. 12 presents the streamwise variations of the normalised velocity half-widths, y0.5/h. The half-width y0.5 varies linearly with x, a trend consistent with well-established self-similar principles [2]. The half-widths agree closely with a far eld relation of the form y0.5/h = Ky(x/h + x02/h), where Ky is the spreading (widening) rate and x02 is the virtual origin of jet spread. As AR is decreased from 72 to 15, the jets spread at decreased rates such that AR = 15 has the smallest spreading rate (see insert of Fig. 12). This is consistent with the trend in decay rate of Uc discussed above (Figs. 9 and 10). This nding is consistent with Zaman [23] who found that a jet issuing from a nozzle of higher aspect ratio rectangular nozzle entrains ambient uid at a higher rate than a nozzle of lower aspect ratio. Since Zaman [23]

attributed increased entrainment at higher aspect ratios to the streamwise vortical structures, it is likely that changes to the nozzle aspect ratio modify the large-scale structures emanating from the present plane jet. Fig. 13 presents the evolution of centerline turbulence intensity, u0n;c u0c =U c in the present jets. The proles are oset vertically for clarity, so that an arbitrary ordinate scale is used. A near eld plateau in u0n;c is clearly visible for plane jets of AR P 30, although with a greater scatter than for Un,c, as one might expect. There is a at region in the far eld for AR P 30, leading to the widest plateau for the highest AR. That is, the width of the plateau increases with aspect ratio. Over the entire x/h, the values of u0n;c are strongly dependent on AR. Between x/h = 8 9, possible humps in centerline turbulence intensity are evident for both, AR = 15 and 20, although it is most pronounced for the smaller aspect ratio. Further downstream from the hump, u0n;c approaches an asymptotic value for AR P 30, whose magnitude (Fig. 14), u0c;1 is strictly dependent on AR. Consistent with Fig. 5, the case of AR = 15 shows that u0n;c does not achieve a constant value at all, further indicating strong three-dimensional eects from the sidewalls. Since u0n;c is found not to be constant for AR = 20, this indicates that an aspect ratio of 20 is also insucient to achieve a fully developed 2-D turbulent ow. This is broadly consistent with Figs. 5, 7 and 10, which reveal a very small region of x/h for which the mean velocity eld remains statistically two-dimensional. For AR P 30, u0n;c maintains asymptotic values of u0c;1 to the present ranges of x/h. Taken together, these observations suggest that a fully developed 2-D ow is truly achieved only for AR P 30 for the present ranges of x/h. Fig. 15(a)(e) presents the lateral proles of turbulence intensity, u 0 /Uc for jets of dierent AR. As clearly evident, the turbulence intensity proles for the cases AR = 15 and 20 never become truly congruent, consistent with the other data presented so far. However, for AR = 30, turbulence intensity proles become congruent for x/h P 20. As with the case of mean velocity proles, the distance required for the turbulence intensity proles to become congruent decreases with AR, so that for AR = 50 and 72, the proles become congruent for x/h P 10. This, again, is further evidence that the two-dimensionality of the fully developed ow increases with AR throughout the investigated regime. The dependence of ow statistics on AR may be further investigated using the moments of higher order uctuations. Figs. 16 and 17 present the centerline evolutions of the skewness, Su = hu3i/(hu2i)3/2, and atness, Fu = hu4i/ (hu2i)2. Note that the values of Su and Fu were determined from % 400 000 data points of U (t) at each axial location, so the convergence of the calculations is reasonably satisfactory, and an appropriate oset applied to the A/D sampling board ensured that Su and Fu were not truncated due to clipping eects arising from the nite range of our sampling board (Tan-Atichat et al. [32]). Clear trends emerge, both Su and Fu appear to be AR-dependent. The factors evolve from nearly Gaussian value (Su, Fu) = (0, 3) at the

Ku

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

833

1.00

x/h = 3 5 10 20

1.00

0.75

0.75

x/h = 3 5 10 20 40

Un= U / Uc

0.50

U n= U / U c

0.50

Gaussian
0.25

0.25

0 0 0.5 1.0 1.5 2.0 2.5

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5

yn= y / y0.5

1.00

0.75

x/h = 3 5 10 20 40
Un= U / Uc

1.00

0.75

Un= U / Uc

x/h = 3 5 10 20 40 80

0.50

0.50

0.25

0.25

0 0 0.5 1.0 1.5 2.0 2.5

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5

yn= y / y0.5

1.00

0.75

Un= U / Uc

x/h = 3 5 10 20 40 80

0.50

0.25

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5
Fig. 11. Lateral proles of the normalized mean velocity U/Uc for (a) AR = 15, (b) AR = 30, (c) AR = 50, (d) AR = 60, and (e) AR = 72. A Gaussian curve is also shown in each plot.

nozzle exit to highly non-Gaussian values around x/h = 3 5. Such were the observations of Browne et al. [34], whose passive temperature uctuations at the exit of a plane jet were Gaussian, while at the potential core region (between 3 and 5h) were highly non-Gaussian. The present large values of Su and Fu around the potential core region (say

between x/h = 3 and 5) perhaps reects the occasional intrusions of low-velocity ambient uid into the jet (Browne et al. [34]), due to the growth of the large-scale roller-like structures in the mixing layers. An estimate of the relationship between atness and intermittency is given by Batchelor and Townsend [30] as Fu = hu4i/(hu2i)2 % 3/c,

834

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838
0.25

0.15

10
0.10 Ky

0.05 0 20 40 AR 60 80

0.23

y0.5 / h

0.21

u'c,
2

AR = 15 20 30 50 72
0 10 20 30 40 50 60 70 80 90

0.19

0.17

x/h
Fig. 12. Streamwise evolutions of mean velocity half-width for AR = 15 72. Note that inserted gure shows the dependence of jet spreading rates, Ky on AR.

0.15 0 15 30 45 60 75

AR
Fig. 14. Variations of far eld asymptotic turbulence intensity, u0c;1 with AR.

AR = 15 20 30 40 50 60 72

20

40

60

80

100

x/h
Fig. 13. Streamwise evolutions of locally normalized turbulence intensity, u0c =U c for AR = 1572. Note that the y-axis values are arbitrary.

where c is the intermittency factor, which denes the proportion of time for which the velocity signal is turbulent. This derivation assumes that all the turbulent uid has a atness of 3, which may not necessarily be true. However, it suggests that the high-near-eld values of Fu are associated with low-turbulence intermittency. Specic observations show that Su exhibit a local maximum, jS max j, and a local minimum, jS min j, (Fig. 16 insert) at u u x/h = 3 and 5, respectively, while Fu reaches a maximum jF max j at x/h = 5 (Fig. 17 insert). Figs. 16 and 17 also u show that the magnitudes of jS min j and jF max j at x/h % 5 u u depend on AR. The reason for this stems from the nding in Figs. 5 and 18. Note from those gures that the length of the potential core decreases with AR. A shorter potential core implies a more rapid development of large-scale, coherent structures through its shear-layer, and increased

uid entrainment and, so too, perhaps more coherent large-scale structures. As with earlier results, while the near eld values of jS min j and jF max j appear to exhibit an u u asymptotic-like dependence on AR, they retain subtle but signicant dierences in their magnitudes. Moving downstream from x/h = 5, both factors Su and Fu approach near Gaussian (0, 3) in the far eld, albeit, not reaching truly Gaussian values. This again is consistent with Browne et al. [34], whose far eld passive temperature distributions did not approximate a Gaussian value. Greater deviations of both factors from their Gaussian values are notable for the cases AR 6 20, as conrmed by the probability density functions (pdf) of the centerline velocity uctuations [9]. To examine whether or not, there is an asymptotic-like convergence of near eld ow, Fig. 18 plots the virtual origins, x01 and x02 and the potential core lengths xp for various AR. Although with some scattering, there is sucient evidence to conclude that the near eld mean statistics show a consistent dependence on AR, however, the ow retains subtle dierences from being truly asymptotic, forbidding a complete elimination of the eect of exit conditions. The eect of the notable departures from its true asymptotic state in the near eld is conrmed by the significant dierences in ow properties in the far eld. In other words, the near eld eect does propagate into the far eld, causing the ow to be signicantly dependent upon the nozzle aspect ratio. It is revealed the maximum axial distance (x) required for the attainment of self-preservation depends on AR, and more specically, on the spanwise dimension (w) of the nozzle if h is constant. It is thus deduced that a lowAR-rectangular jet (without sidewalls) will be initially quasi-2D but will eventually become axisymmetric at a small value of x relative to a high-AR-rectangular jet which

u'c/Uc [arbitrary scale]

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838
0.4 0.4

835

0.3

x/h = 3 5 10 20

0.3

x/h = 3 5 10 20 40

u'/Uc

0.2

0.1

0 0 0.5 1.0 1.5 2.0 2.5

u'/Uc

0.2

0.1

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5
0.4 0.4

yn= y / y0.5

0.3

x/h = 3 5 10 20 40

0.3

x/h = 3 5 10 20 40

u'/Uc

u'/Uc

0.2

0.2

0.1

0.1

0 0 0.5 1.0 1.5 2.0 2.5

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5
0.4

yn= y / y0.5

0.3

x/h = 3 5 10 20 40 80

u'/Uc

0.2

0.1

0 0 0.5 1.0 1.5 2.0 2.5

yn= y / y0.5
Fig. 15. Lateral proles of the turbulence intensity, u 0 /Uc for (a) AR = 15, (b) AR = 20, (c) AR = 30, (d) AR = 50, and (e) AR = 72.

will be quasi-2D up till a larger value of x and it becomes axisymmetric too. However, with sidewalls; the latter jet is expected to experience the inuence of the wall-induced secondary ow into the jet centerline, causing increased

three-dimensionality. This is a useful subject for further investigation and can be undertaken by comparison of two jets from identical rectangular nozzles but congured with and without sidewalls.

836

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

1.0 0.9
min

0.30 0.25 0.20 0.15 0.10 0 15 30 45 60 75

|Su

0.8 0.7 0.6

Su= <u3> / (<u2>)3/2

2 1 0
0 0 0 0 0 0 -1 -2 0 10 20 30 40

AR

Gaussian

50

60

70

80

|S u |
90

x/h
Fig. 16. Streamwise evolutions of the skewness, Su, for AR = 1572. The insert shows the AR-eect on the near-eld minimum of Su and on far eld asymptotic Su. Note that each prole is shifted vertically by unity relative to its neighbour for clarity, and symbols are identical to Fig. 13.

5. Conclusions In summary, the present systematic study reveals that all the normalized mean and turbulent properties of a plane jet are dependent on nozzle aspect ratio even at AR = 72. It is also suggested that asymptotic independence from AR will require much larger aspect ratios than have been investigated previously. The present results are consistent with previously reported data, but provide a much greater detail than has previously been available. An asymptotic-like convergence of many near-eld aspects of the velocity eld, as AR approaches 72, is noted, but substantially highvalues of AR are required to achieve true convergence. In the near eld, an increase in AR leads to an asymptotic-like increase in the length of the jets potential core, the jet virtual origins, x01 and x02, and the magnitude of the near eld troughs in skewness and spikes in atness. In contrast, a true asymptotic state is not attained in the far eld for present ranges of AR, with subtle, but denite dierences found even for the cases of AR = 60 and 72. This is indicated by rates of jet-spreading and centerline velocity decay, which continue to increase approximately linearly with AR and by the far eld skewness and atness factors. Since the presence of large-scale structures, if any, typically

control the overall-spreading and decay rate, this dependence implies an AR-dependence of the underlying largescale structures on nozzle aspect ratio. It is proposed that in the self-similar eld, a statistically 2-D jet is attained only for AR P 30 over an axial distance of up to 85h. This is evidenced, for example, by the mean centerline velocity scaling as Uc $ x1/2, by the normalized turbulence intensity approaching a constant value and by skewness and atness factors of velocity uctuations approaching near-Gaussian values for these cases. Further, the axial extent of the region of statistical 2-D jet increases approximately linearly with AR. It must be noted that there is often a misconception when dierentiating between a large AR-rectangular jet and a true plane jet. It must be recognized that the latter is only achievable using a rectangular nozzle with sidewalls, however, if sidewalls are not used, the ow eld of the former case will be signicantly dierent from the latter case, even in the near eld quasi2D region, where aspect ratio eects are believed thought to be small. For plane jets of AR P 30, the axial location at which the jet undergoes a transition from 2-D to 3-D is found to occur at a xed local jet aspect ratio. This corresponds approximately to the condition where the outer edge of the jet equals the sidewall spacing, i.e., when

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838

837

12
6
max

3.0 2.8 Fu

F
5

4
4

2.6 0 15 30 45 AR 60 75

Fu= <u4> / (<u2>)2

3 3

3 3

Gaussian
2 0 10 20 30 40 50 60 70 80 90

x/h
Fig. 17. Streamwise evolutions of the atness, Fu for AR = 1572. The insert shows the AR-eect on the near-eld maximum of Fu. Note that each prole is shifted vertically by unity relative to its neighbour for clarity, and symbols are identical to Fig. 13.

10 0

x02 x01
xp

r2 = 0.95

-4

r2 = 0.95

-8

-12

potential core length, xp

virtual origins, x01, x02

present study, along with the recent measurements by Deo [9] therefore provides further support for the analytical hypotheses proposed by George [7] and George and Davidson [8], that the downstream development of a plane jet is entirely governed by initial conditions. Hence, consistent with established knowledge for round jets, the classical hypothesis, which argues that all jets should become asymptotically independent of the source conditions, at suciently large distances from the source, does not hold true for a plane jet. Acknowledgements

0 0 15 30 45 AR 60 75 90

Fig. 18. The dependence of the virtual origins, x01, x02 and lengths of the potential core, xp on AR.

the actual aspect ratio of the local jet is approximately unity. Taken together, the statistical dependence of the velocity eld on AR is attributable to dierences in the underlying large-scale structures that propagate downstream, emanating from nozzles of diering boundary conditions. The

We acknowledge that this original research article reports work from the rst authors Ph.D. thesis completed through the support of Endeavor International Postgraduate Scholarship, Adelaide Achievers Scholarship and an ARC Discovery Grant in collaboration with FCT. We also thank the two anonymous reviewers whose comments have improved the overall clarity of our paper. References
[1] H. Schlichting, Laminare strahlausbreitung, Z. Angew. Math. Mech. 13 (1933) 260263.

838

R.C. Deo et al. / Experimental Thermal and Fluid Science 31 (2007) 825838 [21] W.R. Quinn, Turbulent free jet ows issuing from sharp-edged rectangular slots: The inuence of slot aspect ratio, Expl. Therm. Fluid Sci. 5 (1992) 203215. [22] Y. Tsuchiya, C. Horikoshi, T. Sato, M. Takahashi, A study of the spread of rectangular jets: (The mixing layer near the jet exit and visualization by the dye methods), JSME Int. J. 32 (Series II (1)) (1989) 1117. [23] K.B.M.Q. Zaman, Spreading characteristics of compressible jets from nozzles of various geometry, J. Fluid Mech. 383 (1999) 197 228. [24] P.E. Dimotakis, The mixing transition in turbulent ows, J. Fluid Mech. 409 (2000) 6998. [25] I. Namar, M.V. Otugen, Velocity measurements in a planar turbulent air jet at moderate Reynolds numbers, Expts. Fluids 6 (1988) 387 399. [26] G.J. Hitchman, A.B. Strong, P.R. Slawson, G. Ray, Turbulent planar jet with and without conning walls, AIAA J. 28 (10) (1990) 1699 1700. [27] F.O. Thomas, V.W. Goldschmidt, Structural characteristics of a developing turbulent plane jet, J. Fluid Mech. 163 (1986) 227 256. [28] J. Bashir, Experimental study of the turbulent structure and heat transfer of a two dimensional heated jet. Doctoral Dissertation, University of Colorado, 1973. [29] H.J. Hussein, S.P. Capps, W.K. George, Velocity measurements in a high Reynolds number, momentum-conserving axisymmetric turbulent jet, J. Fluid Mech. 258 (1994) 3160. [30] G.K. Batchelor, A.A. Townsend, Proc. R. Soc. Lond. A 199 (1949) 238255. [31] W.K. George, P.D. Beuther, A. Shabbir, Polynomial calibrations for hot wires in thermally varying ows, Expl. Therm. Fluid Sci. 2 (1989) 230235. [32] J. Tan-Atichat, W.K. George, S. Woodward, Use of Computer for Data Acquisition and Processing, in: A. Fuhs (Ed.), Handbook of Fluids and Fluids Engineering, vol. 3, Wiley, NY, 1996, pp. 1098 1116, Sec. 15.15. [33] H. Schlichting, Boundary Layer Theory, McGraw-Hill, 1968 (Chapter 7). [34] L.W.B. Browne, R.A. Antonia, S. Rajagopalan, A.J. Chambers, The interaction region of a turbulent plane jet, J. Fluid Mech. 149 (1984) 355373. [35] D.R. Miller, E.W. Comings, Static pressure distribution in a free turbulent jet, J. Fluid Mech. 3 (1957) 116.

[2] G. Heskestad, Hot-wire measurements in a plane turbulent jet, Trans. ASME, J. Appl. Mech. 32 (1965) 721734. [3] L.J.S. Bradbury, The structure of a self-preserving turbulent planar jet, J. Fluid Mech. 23 (1965) 3164. [4] E. Gutmark, I. Wygnanski, The planar turbulent jet, J. Fluid Mech. 73 (3) (1976) 465495. [5] J. Bashir, S.M. Uberoi, Experiments on turbulent structure and heat transfer in a 2-D jet, Phys. Fluids 18 (4) (1975) 405410. [6] B.G. Van der Hegge Zijnen, Measurements of the distribution of heat and matter in a plane turbulent jet of air, Appl. Sci. Res. A7 (1958) 277292. [7] W.K. George, The self preservation of turbulent ows and its relation to initial conditions, in: Recent Advances in Turbulence, Hemisphere, New York, 1989, pp. 3973. [8] W.K. George, L. Davidson, Role of initial conditions in establishing asymptotic behaviour, AIAA J. 42 (3) (2004) 438446. [9] R.C. Deo, Experimental investigations of the inuence of Reynolds number and boundary conditions on a plane air jet. Ph.D. thesis. School of Mechanical Engineering. The University of Adelaide, South Australia, 2005. [10] P. Bradshaw, The eect of initial conditions on the development of a free shear layer, J. Fluid Mech. 26 (1966) 225236. [11] P. Bradshaw, Eects of external disturbances on the spreading rate of a plane turbulent jet, J. Fluid Mech. 80 (1977) 795797. [12] F.C. Gouldin, R.W. Schefer, S.C. Johnson, W. Kollmann, Nonreacting turbulent Mixing ows, Prog. Energy Combust. Sci. 12 (1986) 257303. [13] L.W.B. Browne, R.A. Antonia, S. Rajagopalan, A.J. Chambers, Interaction region of a two-dimensional turbulent plane jet in still air, in: Proc. of Structure of complex turbulent shear ows, IUTAM Symp., Marseille, 1982, pp. 411419. [14] S.B. Pope, Turbulent Flows, Cambridge University Press, UK, 2002. [15] J. Mi, R.C. Deo, G.J. Nathan, Characterization of turbulent jets from high-aspect-ratio rectangular nozzles, Phys. Fluids 17 (6) (2005). [16] N. Trentacoste, P. Sforza, Further experimental result for threedimensional free jets, AIAA J. 5 (5) (1967) 885890. [17] A. Sfeir, Investigation of three-dimensional turbulent rectangular jets, AIAA J 17 (10) (1979) 10551060. [18] G.F. Marsters, J. Fotheringham, The inuence of aspect ratio on incompressible turbulent ows from rectangular slots, Aeronaut. Quart. J 31 (4) (1980) 285305. [19] A. Krothapalli, D. Bagano, K. Karamcheti, On the mixing of rectangular jet, J. Fluid Mech. 107 (1981) 201220. [20] Y. Tsuchiya, C. Horikoshi, T. Sato, On the spread of a rectangular jets, Expts. Fluids 4 (1985) 197204.

S-ar putea să vă placă și