第 1４回数値流体力学シンポジウム

C04-1
Numerical Computation on Recirculation Flow Structures
in Co-Axial Confined Laminar Jets

P. L. WOODFIELD, K. NAKABE and K. SUZUKI,

Heat Transfer Laboratory, Department of Mechanical Engineering, Kyoto University,
Kyoto 606-8501, Japan
E-mail: [email protected]

‘Craya-Curtet’ number (Ct) has been described as one of the most suitable universal similarity parameters for co-
axial confined jets. This work shows that the formation of the recirculation zone is greatly influenced by Reynolds
number, and not by Ct alone under laminar flow conditions. If the length scale of the thermo-fluid machinery is
decreased, for example, in micro jet flame combustors, the corresponding Reynolds number, Re, becomes smaller.
In the case of Re > 50, the length of the recirculation zone is found proportional to Re number for a fixed value of Ct.
In the limit of very small Re number, the recirculation zone disappears, unlike turbulent flow cases.

1. Introduction
∂u ∂u ∂ 2 u µ ∂ æ ∂u ö ∂p
ρu + ρv =µ 2 + çr ÷ −  …(1)
Fundamental investigation into confined jet interactions at low ∂x ∂r ∂x r ∂r è ∂r ø ∂x
Reynolds number has particular value for the development of
micro-jet pumps and micro combustors. The co-axial confined
∂v ∂v ∂ 2v ∂ æ 1 ∂rv ö ∂p
ρu + ρv = µ 2 + µ ç
jet, which is perhaps one of the simplest possible configurations,
÷− …(2)
was selected for this study.
∂x ∂r ∂x ∂r è r ∂r ø ∂r
Figure 1 gives a schematic illustration for the computational
domain. Under certain conditions where the momentum of the Equation 3 gives the constant property continuity equation in
primary (center) jet is sufficiently in excess of the momentum of cylindrical coordinates.
the secondary (annular) jet, an axisymmetric recirculation zone
∂u 1 ∂
(rv ) = 0
appears on the wall of the tube as shown in the figure.
+ …(3)
∂x r ∂r
3. Boundary Conditions

2 The inlet velocity profiles are for fully developed tube flow
D (primary stream) and fully developed annular flow (secondary
1 stream) as shown in Fig. 1. The radial velocity component is
zero over the entire inlet cross-section (i.e. at x = 0). The
d computational domain extends to 33 tube diameters downstream
2
of the inlet, and a zero normal gradient boundary condition is
assumed at the exit.
x Separation Reattachment
4. Numerical Technique

For the solution of Eqs. (1) to (3), the method of finite volumes is
Fig. 1 Co-axial confined jet configuration (1: Primary jet; 2:
used in conjunction with Patankar’s SIMPLE technique1. For the
Secondary jet. Inlet velocity profiles are the fully developed
discretization of the convection and diffusion terms, a first-order
laminar profiles for circular and annular tubes).
hybrid differencing scheme is employed1. Code validation was
performed by comparison with fully developed pipe flow data and
experimental velocity distributions for the confined laminar jet by
For the present work, Reynolds number is based on the mean
Shavit and Lavan2.
velocity, U, averaged over the entire tube cross-section and on the
diameter of the tube, D.
A computational grid of 200 by 54 cells (x×r) is used to discretize
the domain, the grid spacing being progressively finer closer to the
2. Mathematical Model
inlet. For the very low Reynolds number calculations (Re<10),
In order to consider the simplest case, the flow is taken to be the length of the computational domain is reduced whilst
steady and laminar. All fluid properties are assumed constant. maintaining the same number of grid points to preserve the fine
Governing equations for axial and radial momentum are solved resolution.
together with the continuity equation.
5. Effect of Craya-Curtet Number
Equations 1 and 2 give the Navier-Stokes equations for
conservation of momentum. The Craya-Curtet number is a well-known similarity parameter for

Copyright © 2000 by JSCFD

the co-axial confined turbulent jet. In the present work the effect tube wall. In case (C) of Fig. 2 with Ct = 0.85, the recirculation
of this parameter upon the confined laminar jet is examined. The does not appear. This is consistent with observations by Curtet
definition of Ct used3, is given in Eqs. (4) to (6). and others of the turbulent case where for Ct less than about 0.75
recirculation occurs downstream of the inlet on the duct wall.3,4 (It
Uo should be noted however, that the critical value of Ct for which the
Ct = …(4) recirculation appears, reported in the literature, ranges from 0.75
U d2 − 0.5U o2 to 0.976).3

Figure 3 shows the effect of the Craya-Curtet number on the

1 friction coefficient at the wall. The three curves represent the
U d2 =
Ao ò Ao
(u 2 − 0.5U 2f ) dA …(5) three cases given in the previous figure (Fig. 2) for a Reynolds
number of 50. ‘Cf’’ is defined as given in Eq. (8).

1 τ
Cf ≡
Ao òAo
Uo = udA …(6) …(8)
ρU 2 / 2

For the turbulent case, Uf is defined as the inlet velocity at the For fully developed flow, Cf takes on a value of 16/Re which
constant velocity part of the co-current stream. Here it is taken agrees to within 0.05% of the calculated downstream values
to be the mean velocity of the co-current stream as given in Eq. shown in Fig. 3.
(7). To maintain consistency with this assumption, when
evaluating the integral in Eq. (5), it is also assumed that the At the mixing zone entrance (i.e. at x = 0), the wall friction
velocity profile is flat over the entire secondary stream at the coefficient increases slightly as the diameter of the primary jet is
entrance. For the present calculations, the effect of the secondary increased. This is because the secondary and primary mass flow
stream profile is found to be small giving some justification to this rates are the same for the three cases, and hence the velocities in
assumption. the secondary annular jet increase as the cross-sectional area is
shrunken following the increase of the primary jet area.
In Eqs. (5) - (7), ‘Ao’ is the entire cross-sectional area of the tube
and ‘A2’ is the cross-sectional area of the secondary stream at the The points where the curves intercept the x-axis correspond to the
entrance (i.e. ‘Ao’ minus the primary jet cross-sectional area). separation and reattachment points for the recirculation zone.
Quite clearly in Fig. 3, the separation point moves only slightly
upstream with a decrease in Craya-Curtet number from 0.4 to 0.27.
1
A2 òA2
U f = U2 = udA …(7) The reattachment point, on the other hand, is affected drastically
by Ct moving from about 4.5 tube diameters downstream of the
entrance to about 8 tube diameters as Ct is varied from 0.4 to 0.27.
This is quite different to the turbulent case where the reattachment
Figure 2 shows the effect of Craya-Curtet number on the flow point moves only slightly as Ct is decreased and the separation
field for a Reynolds number of 50. For all three cases in Fig. 2, point moves remarkably upstream as Ct is decreased.
the mass flow rates of the two streams are kept constant, the ratio
of the primary flow to the secondary flow (m1/m2) being 0.32. Also, all three curves in Fig. 3, display a minimum (or maximum
The Craya-Curtet number is altered by adjusting the diameter of negative) friction coefficient located at a point which moves
the primary jet. For a smaller diameter and the same flow rate, downstream as Ct is decreased.
the momentum of the primary jet increases and hence Ct
decreases.

0.6
0.4 Ct = 0.85

0.2
0
Cf

-0.2
-0.4 Ct = 0.4
Ct = 0.27
-0.6
-0.8
0 2 4 6 8 10 12 14
x/D
Fig. 3 Wall friction coefficient – Effect of Ct (Re = 50 and m1/m2
Fig. 2 Effect of Ct on flow field (Re = 50, m1/m2 = 0.32, Craya-
= 0.32).
Curtet number is altered by varying the nozzle diameter of the
primary jet. (A) d/D = 0.074; (B) d/D = 0.111; (C) d/D = 0.222.
Mass flow-rates of the two streams are the same for all three cases.
6. Effect of Reynolds Number
Reynolds number is calculated on mean axial velocity and the
diameter of the tube).
For the case of the developing flow in a tube, the ‘hydrodynamic
entry length’ extends approximately in proportion to Reynolds
number. This proportionality breaks down if the flow becomes
In Fig. 2 (A) and (B), a weak recirculation zone appears on the
turbulent or under creeping flow conditions.5 Also, for the case of

Copyright © 2000 by JSCFD

the sudden expansion duct, the reattachment point of the eddy correlation between the velocity profiles at points located with the
forming in an axisymmetric backward facing step moves same value of x/(D.Re). Figure 6 compares the dimensionless
downstream in proportion to Reynolds number.6 Macagno and axial velocity distributions at various points for Re = 50 and 200,
Hung6 found the distance to the reattachment point to be almost downstream in the axial direction having the same value for
directly proportional to Reynolds number between 40 and 200 and x/(D.Re). In this figure, the solid lines are for a Reynolds
curving upward slightly for Re < 40 in which Re is based on the number of 50 and the dashed lines are for a Reynolds number of
diameter and mean velocity of the smaller duct). 200. At almost all of the positions considered, the agreement
between the two cases is so close that it is difficult to distinguish
between the two lines. Even at the axial position ‘C’, which is
0.4 located approximately half way between the separation and
reattachment points, the agreement for the reverse flow is also
0.3 very close.
0.2

0.1 40
0
Cf

-0.1 Re = 200 35
Re = 100
-0.2 A: (x/D)/Re = 0.0
30 B: (x/D)/Re = 0.0186
-0.3 Re = 50 C: (x/D)/Re = 0.0533
D: (x/D)/Re = 0.0950
-0.4 25 E: (x/D)/Re = 0.22
0 5 10 15 20 25 30 A
x/D
20
Fig. 4 Wall friction coefficient – Effect of Reynolds number (Ct =
u/U
0.4 and m1/m2 = 0.32).
15
B
The same kind of proportionality with respect to Reynolds number 10 D
is observed here for the case of the axisymmetric confined laminar C
jet. Figure 4 shows the skin friction coefficient at the wall for the
cases of Re = 50, 100 and 200 where the relative proportion of the 5
primary and secondary fluid flow is kept constant. The ratio of
d/D is also constant so that Ct is the same for all three cases.
0
Quite clearly in Fig. 4, for the range of Reynolds number E C
investigated, the distances to both the separation and reattachment -5
points increase in the downstream direction in a virtually linear 0 0.1 0.2
0.3 0.4 0.5
proportion with Reynolds number. This point is further r/D
highlighted in Fig. 5. Here, the x-coordinates are divided by Fig. 6 Axial velocity distributions (Ct = 0.4, m1/m2 = 0.32, Re = 50
Reynolds number, and the skin friction coefficient is normalized (solid lines) and Re = 200 (dashed lines)).
with respect to the value for fully developed tube flow (16/Re).
The three curves are remarkably similar with a slightly closer
correlation for the cases of Re = 100 and 200.
Having obtained such a close correlation for dimensionless axial
velocities, u/U, with respect to the dimensionless coordinate,
x/(D.Re), mass conservation would suggest that there will be some
1.5
similarity for the radial component of velocity for different
Reynolds number and the same Craya-Curtet number. Figure 7
1 gives the dimensionless radial velocity (v/U).Re in terms of axial
coordinate, x/(D.Re) for four different radial positions. Again
0.5 there can be seen a clear correlation between the two Reynolds
number cases except the largest differences in the near entrance
Cf.Re/16

0 region.

-0.5
It is also apparent from Fig. 7 that the close matching of the radial
Re = 50 velocity curves applies at different radial positions across the
Re = 100 diameter of the tube. In Fig. 7, the position ‘A’ (r/D = 0.042) is
-1 Re = 200
inside the primary jet (cf. (d/2)/D = 0.055) and hence the radial
component is positive for all downstream values of x, tending to
-1.5 zero as x becomes large. The positions ‘B’ and ‘C’ begin outside
0 0.05 0.1 0.15 0.2 0.25 the primary jet and hence the radial velocity takes on relatively
(x/D)/Re
large negative values just downstream of the entrance, increasing
Fig. 5 Normalized skin friction coefficient – Relation to Reynolds in closer proximity to the jet. A little further downstream, the jet
number (Ct = 0.4 and m1/m2 = 0.32). width grows to coincide with the radial positions ‘B’ and ‘C’ and
hence the velocities become positive. As the radial distance from
the jet center is increased, the influence of the jet upon the radial
The close matching of the curves in Fig. 5 suggests a strong velocity diminishes as shown by the line for position ‘D’ in Fig. 7.

Copyright © 2000 by JSCFD

For all radial positions, the boundary condition at x = 0 is that v = This result is consistent with the experimental findings of other
0. Because of the scaling and the large jump in radial velocity researchers for the case of developing laminar flow in a tube
immediately downstream of the entrance this is a little difficult to where the proportional lengthening of the flow field with respect
decipher from Fig. 7. In fact, this sudden jump in the radial to Reynolds number diminishes as Reynolds number becomes
velocity profile calls into question the practical possibility of the very small5. Macagno and Hung6 made the same observation
zero radial velocity inlet boundary condition. The radial velocity results in the case of a sudden expansion duct.
distribution within the jets upstream of the entrance to the mixing
region may in fact be influenced by the downstream conditions. It is interesting to note here, that for the case of Re < 10 in Fig. 8,
Hence, as suggested by Shavit and Lavan2, ideally the the length of the recirculation zone is less than one tube diameter.
computational domain should be extended a little upstream of the
jet nozzle exit. 8. Discussion

The expansion of the flow field in approximate proportion to

30 Reynolds number shown in Figs. 4 to 7 can be explained in part
by non-dimensionalizing the constant property Navier-Stokes
20 A equations in terms of the following dimensionless parameters.
B

10
x r u
x* = r* = u* =
(v/U).Re

0
D. Re D U …(9)
A: r/D = 0.042
v p t .U
-10 D B: r/D = 0.125
v * = × Re p* = t* =
ρU 2
C: r/D = 0.208
-20
C D: r/D = 0.375 U D. Re
-30 In terms of these dimensionless groups, the boundary conditions
for the three examples given in Figs. 4 to 7 are identical.
-40
0 0.05 0.1 0.15 0.2
This particular scheme used here for non-dimensionalization is
(x/D)/Re
almost similar to the one proposed by Pai and Hsieh7 (1972),
Fig. 7 Dimensionless radial velocity distributions (Ct = 0.4, m1/m2 which produces two-dimensional ‘boundary layer’ type equations
= 0.32, Re = 50 (solid lines) and Re = 200 (dashed lines)). (independent of Reynolds number) suitable for analysis of laminar
jets with and without free stream. Moreover, Langhaar8 (1942), for
the analysis of developing flow in a tube, used x/(a.Re) for the
7. Low Reynolds Number Limit dimensionless axial coordinate (where ‘a’ is the radius of the
tube).
Figure 8 gives the normalized skin friction coefficient against the
dimensionless axial distance, x/(D.Re), for low Reynolds number Using the above transformations the Navier-Stokes equations, Eq.
(1 < Re < 200) and a fixed Craya-Curtet number (Ct = 0.4). As (1) and (2) become as follows:
Reynolds number is reduced to below 30, the proportionality of
the eddy size with respect to Reynolds number diminishes,
∂u * * ∂u
*
* ∂u
*
although the remarkable similarity was shown in Fig. 5 for the
+ u + v =
cases, 50 < Re < 200. Between the cases of Re = 10 and 5, the ∂t * ∂x * ∂r *
recirculation zone vanishes altogether. If the flow rate is reduced  …(10)
1 ∂ 2 u * 1 ∂ æ * ∂u * ö ∂p *
even further to a Reynolds number of 2, the location of the
+ çr ÷−
minimum value for Cf moves downstream with respect to the Re 2 ∂x *2 r * ∂r * çè ∂r * ÷ø ∂x *
dimensionless distance, x/(D.Re).

∂v * * ∂v
*
* ∂v
*

1.5 + u + v =
Re = 2
∂t * ∂x * ∂r * …(11)
Re = 3 1 ∂ 2v* ∂ æ1 ∂ * * ö 2 ∂p
*
1 + ç ( r v ) ÷ − Re
Re = 2
Re 2 ∂x *2 ∂r * è r * ∂r * ø ∂r *
Re = 5
0.5
Cf.Re/16

Re = 10
The time dependent terms are retained here to show that the
analysis will be applied also to the unsteady case.
0
Re = 20 The continuity equation (Eq. (3)) transforms very simply into the
Re = 30 following equation.
-0.5
Re = 200
∂u * 1 ∂ * *
-1
0 0.05 0.1 0.15 0.2 0.25
+
∂x * r * ∂r *
(
r v =0 ) …(12)

x/(D.Re)
Fig. 8 Normalized skin friction coefficient – Effect of further In Eqs. (10) to (12), there are three terms which still contain
reducing Reynolds number (Ct = 0.4 for all cases). Reynolds number:

Copyright © 2000 by JSCFD

 1 ∂ 2u * 1 ∂ 2v* 2 ∂p
*
, , − Re Figure 11 shows the calculated flow and mixing patterns for the
Re 2 ∂x *2 Re 2 ∂x *2 ∂r * inlet geometry illustrated in Fig. 10. Three Reynolds number
cases are considered – Re = 50, 100 and 200. For all cases, the
If these particular terms can be neglected, or canceled each other ratio of the primary and secondary mass flows is identical (m1/m2
out, perfect agreement for the flow field should be found in terms = 0.0465). This value corresponds to the mass flow rates for a
of x*, u* and v* for the same boundary conditions, irrespective of methane-air flame with an equivalence ratio of 0.8. Calculations
the Reynolds number. In Langhaar’s analysis of the developing were performed for constant density three-dimensional laminar
flow in a tube8, the second derivative of velocity in the axial flow. The selected plane in the θ direction is between two
direction and the radial pressure gradient were neglected. Also, adjacent secondary stream jets.
Pai and Hsieh7 neglected all the pressure terms and the second
derivative of the velocity gradient in the axial direction for the
analysis of the free laminar jet and the laminar jet with free stream.
Following their treatment, the terms listed above disappear for the
cases of the ‘free laminar jet’ and ‘developing laminar flow in a
tube’. If we can assume that the sum of these terms containing
Reynolds number is also small (or also increased in proportion to D
Re) for the conditions considered here, we can gain some insight
as to the reason behind the lengthening of the flow field in
approximate proportion to Reynolds number.

One would expect these three terms to be largest at the very

entrance to the tube – which may explain the differences in the Fig. 10 Possible inlet configuration for multiple jet miniature can-
dimensionless radial velocities near the entrance shown in Fig. 7. type combustor.

Re = 50 0.13
R 0.12
1.5 0.11
0.10
Re = 10 0.09
0.08
1 (all terms included) 0.07
0.06
0.05
Re = 200 0.04
Cf.Re/16

0.5 both with and without

0.03
0.02
2 2 2 2 0.01
d u/dx , d vdx terms 0.00
0
0 0 X1 X2 1.5 D
Re = 10
2 2 2 2 Re = 100 0.13
-0.5 d u/dx , d v/dx terms R
X3 0.12
0.11
neglected 0.10
0.09
-1 0.08
0.07
0 0.05 0.1 0.15 0.2 0.25 0.06
0.05
x/(D.Re) 0.04
0.03
0.02
Fig. 9 Effect of neglecting second derivatives of velocity in the 0.01
0.00
0
axial direction. (Solid lines – all terms included; dotted lines – 0 X1 X2 1.5 D
second derivatives of velocity neglected. For Re = 200, the 0.13
Re = 200
dotted line falls on top of the solid line). X3 0.12
0.11
R
0.10
0.09
0.08
0.07
Quite clearly, for creeping flow conditions the second derivative 0.06
0.05
velocity terms containing Re in Eqs. (10) and (11) cannot be 0.04
0.03
neglected, as the size of these terms may increase in proportion to 0.02
0.01
the inverse square of Reynolds number. 0
0.00
0 X1 X2 1.5 D

Figure 9 shows the effect of neglecting the second derivatives of

Fig. 11 Flow and mixing patterns for geometry given in Fig. 10
velocity in the axial direction. For Re = 10, the effect is
(Contours represent mass fraction of gas originating in the primary
remarkable but for Re = 200 the terms have negligible effect on
(center) jet: X1 = upstream stagnation point, X2 = downstream
the wall friction coefficient. Hence, the increasing importance of
stagnation point, and X3 = Reattachment point of recirculation
these terms explains why the recirculation disappears as Reynolds
near wall).
number becomes very small.

9. Practical Implications
The flow field shown in Fig. 11 is very complex with recirculation
zones near the center of the tube and near the tube wall. In front
Presently at the authors’ laboratory of Kyoto University a multiple
of the primary jet a stagnation point, X1, appears. Further
jet can-type combustor is being investigated for micro applications.
downstream, another stagnation point, X2, appears, which is also
Figure 10 shows a possible inlet geometrical configuration for the
associated with the recirculation at the center. The position ‘X3’
miniature combustion chamber. The fuel enters through the jet at
gives the reattachment point for the recirculation at the wall of the
the center and air through the surrounding six jets. Should such a
tube. For the Re = 50 case, the flow pattern near the wall is
device be suitably designed to operate under low Reynolds
different to the other cases so ‘X3’ is not marked.
number laminar flow conditions, the present results would have a
number of implications.
In Fig. 11, the effect of changing Reynolds number from 100 to

Copyright © 2000 by JSCFD

200 is to move the positions ‘X2’ and ‘X3’ further downstream
roughly in proportion to the Reynolds number. The upstream
stagnation point, ‘X1’, however, moves only slightly with an
increase in the flow rate for all three cases. This is perhaps
because the second derivatives of velocity are very important in
this region.

The change in the patterns of both velocity and mass fraction for
the Re = 50 case in the vicinity of wall are further evidences of the
low Reynolds number effects discussed in this paper.

10. Conclusions

1. While the ‘Craya-Curtet’ number appears to be an important

parameter for the co-axial confined laminar jet, Reynolds
number also plays a key role.
2. The primary effect of Reynolds number on the laminar cases
investigated is to expand the flow field in the downstream
direction with an increase of the Reynolds number. Close
proportionality with Reynolds number has been demonstrated
for the co-axial confined jet with recirculation (Ct = 0.4) for
Reynolds number ranging from 50 to 200 (based on the outer
tube diameter and bulk mean velocity).
3. In the low Reynolds number limit, the recirculation zone
disappears for a Reynolds number between 5 and 10 under the
Craya-Curtet number of 0.4. This is shown to be connected
with an increasing importance of the second derivatives of
velocity in the axial direction at such low Reynolds numbers.

Acknowledgement

This numerical work is one of the results of “Micro Gas Turbine /

Fuel Cell Hybrid-Type Distributed Energy System” which is
supported by Department of CREST in the program of JST.

References

1. Patankar, S. V., “Numerical Heat Transfer and Fluid Flow,”

Hemisphere Publishing Corporation, (1980).
2. Shavit, G. and Lavan, Z., “Analytical and experimental
investigations of laminar mixing of confined heterogeneous
jets,” AIAA Journal, 11 (1973), pp. 352-358.
3. Kang, Y. M., “A Study on Heat Transfer in Recirculating
Flows,” PhD thesis, Kyoto University, (1980).
4. Rajaratnam, N., “Turbulent Jets,” Elsevier Scientific Publishing
Company, (1976).
5. White, F. M., “Viscous Fluid Flow,” McGraw-Hill Inc., (1974).
6. Macagno, E. O. and Hung, T., “Computational and
experimental study of a captive annular eddy,” Journal of Fluid
Mechanics, 28 (1967), pp. 43-64.
7. Pai, S. I. and Hsieh, T., “Numerical solution of laminar jet
mixing with and without free stream,” Applied Scientific
Research (The Hague), 27 (1972), pp. 39-62.
8. Langhaar, H. L. “Steady flow in the transition length of a
straight tube,” Trans. ASME: Journal of Applied Mechanics, 64
(1942), pp. A55-A58.

Copyright © 2000 by JSCFD