Critical Condition for Flow Transition in a Full-

Developed Annulus Flow

Hua-Shu Dou,1,2 Boo Cheong Khoo,2 and Her Mann Tsai1

1. Temasek Laboratory, National University of Singapore, Singapore 119260
2. Department of Mechanical Engineering, National University of Singapore, Singapore 119260
tsldh@nus.edu.sg; huashudou@yahoo.com
The flow transition to turbulence still largely remains an unsolved fluid dynamics problem. The
stability of parallel flows is still a research subject for the turbulent transition for studies despite
much works in the literature. In previous works, we have shown that the flow instability in wall
shear flows depends on the relative magnitude of energy gradient and the energy loss for a given
disturbance. For pipe and plane Poiseuille flows, we have demonstrated that the subcritical
transition to turbulence for wall bounded parallel flows occurs at a consistent value of the energy
gradient parameter. In this study, the critical condition for subcritical transition in annulus flow is
calculated for various radius ratios. The critical flow rate and critical Reynolds number are given
for various radius ratios. These results have wide implications for many fluid delivery devices.

Keywords: Subcritical transition; Full-developed flow; Annulus; Critical flow rate; Radius ratio.

INTRODUCTION

Annulus flow passage is widely used in aero-engines, turbomachinery and various chemical industrial devices.
The flow dynamics in the annulus has significant influence on the performance and flow efficiency of the devices. The
flow in the annulus may be laminar or turbulent depending on the behaviour of the main flow and the disturbance
level. Generally, the disturbance level in turbomachinery is very high so that the flow is likely to be turbulence in most
circumstances. However, the flow may be kept laminar if the flow passage is carefully designed. The determination of
the critical condition for the flow transition in annulus is therefore of great interest.
Flow instability and turbulence transition have been challenging topics for fluid dynamists for more than a
century. This problem has not been solved fully owing to the many difficulties encounted [1-2]. There exist a few
theories for the analysis of flow instabilities. These are linear stability analysis, energy method, weekly nonlinear
method, secondary instability theory, and others [1-2]. Linear stability analysis was successful for some flows such as
Taylor-Coutte flow and Rayleigh-Benard flow. But, it is failed for some flows such as plane Poiseuille flow, pipe
Poiseuille flow, and plane Couette flow [1-2]. The pipe Poiseuille flow (Hagen-Poiseuille) is stable by linear stability
analysis for all the Reynolds number Re. However, experiments showed that the flow would become turbulence if Re
(=ρUD/µ) exceeds a value of about 2000. Linear stability analysis of plane parallel flow gives a critical Reynolds
number Re (=ρ u 0 h/µ) of 5772, while experiments show that transition to turbulence occurs at the Reynolds number of
order 1000 [3-4]. For plane Couette flow, the flow is stable for all Re from linear theory consideration, but the
experiments show a critical value of Re of 370.
The energy method uses the Reynolds-Orr equation and involves the integration of energy for the entire flow
domain. Using the energy method, the critical Re is 88, 26 and 40 respectively for plane Poiseuille flow, pipe
Poiseuille flow and plane Couette flow [1]. Therefore, the non-uniformity in using this method to predict the flow
transition is still imminent. On the other hand, the occurrence of stability is strictly a local behaviour and the flow
during the transition is intermittent. The first occurrence of the flow instability generally takes place in the most
“dangerous” positions as seen in the formation of turbulence spot, the cylinder wake, and the dynamic stall on the
airfoil with large attack angle. Hence, a method considering the local flow behaviour may be the correct approach.
The weekly nonlinear method [5] and the secondary instability theory [6] seem to give better results than the above-
mentioned other methods and can explain some phenomena; however there is still discrepancy with experiments.
Recently, Dou [7] proposed a new approach to study flow instability and turbulent transition. It is demonstrated
that the plane and the pipe Poiseuille flows have a consistent critical condition at the subcritical transition condition
determined by experiments. In this paper, following the proposed principle of energy gradient theory, the critical
condition for subcritical transition in annulus flow is calculated for various radius ratios. The critical flow rate and
critical Reynolds number are also obtained for various radius ratios.

1
ENERGY GRADIENT THEORY

Dou [7] proposed an energy gradient theory with the aim to clarify the mechanism of transition from laminar to
turbulence for wall bounded shear flows. Here, we shall give a short discussion /description for a better understanding
of the present work. In the theory, the whole flow field is treated as an energy field. It is suggested that the gradient of
total energy in the transverse direction of the main flow and the viscous friction in the streamwise direction dominate
the instability phenomena and hence the flow transition for a given disturbance. The energy gradient in the transverse
direction has the potential to amplify a velocity disturbance, while the viscous friction loss in the streamwise direction
can resist and absorb this disturbance. The flow instability or the transition to turbulence depends on the relative
magnitude of these two roles of energy gradient amplification and viscous friction damping to the initial disturbance. It
is noted that the energy loss ( ∂H / ∂s ) in unit length in streamwise direction due to viscous friction is equal to the
gradient of total energy ( ∂E / ∂s ) of unit volume fluid in streamwise direction for pressure driven flows. As such, a
new dimensionless parameter, K (the ratio of the energy gradient in the transverse direction to that in the streamwise
direction), is defined to characterize the stability of the base flow for the case where there is no work input,
∂E / ∂n
K= . (1a)
∂E / ∂s
1
Here, E = p + ρV 2 + ρgξ is the total energy for incompressible flows with ξ as the coordinate in the direction of
2
gravitational field, n denotes the direction normal to the streamwise direction and s denotes the streamwise direction.
Furthermore, ρ is the fluid density, g is the gravity acceleration, V is the velocity, and p is the hydrodynamic pressure.
For shear driven flows (there is work input), the calculation of K can be obtained by the ratio of the energy gradient in
transverse direction and the energy loss in unit length along the streamline [8],
∂E / ∂n
K= . (1b)
∂H / ∂s
As such, the parameter K in Eq.(1) is a field variable. Thus, the distribution of K in the flow field and the
characteristics of disturbance may serve as the means to describe if there is a disturbance amplification or decay in the
flow. It is suggested that the flow instability can first occur at the position of K max which is construed to be the most
“dangerous” position. Thus, for a given disturbance, the occurrence of instability depends on the magnitude of this
dimensionless parameter K and the critical condition is determined by the maximum value of K in the flow. For a
given flow geometry and fluid properties, when the maximum of K in the flow field is larger than a critical value K c ,
it is expected that instability can occur for certain initial disturbance [7]. For a given flow, K is proportional to the
global Reynolds number. A large value of K has the ability to amplify the disturbance, and vice versa. The analysis has
suggested that the transition to turbulence is due to the energy gradient and the disturbance amplification[7], rather
than just the linear eigenvalue instability type as stated in [3,4]. Both Trefethen et al. [3] and Grossmann [4]
commented that the nature of the onset-of-turbulence mechanism in parallel shear flows must be different from an
eigenvalue instability of linear equations of small disturbance. In fact, finite disturbance is needed for the turbulence
initiation in the range of finite Re as found in experiments [9]. Dou [7] demonstrated that the criterion obtained has a
consistent value at the subcritical condition of transition determined by the experimental data for plane Poiseuille flow,
pipe Poiseuille flow as well as plane Couette flow (see Table 1). The data in Table 1 may suggest that the subcritical
condition of transition for Poiseuille flows occurs at K c ≈ 385 and the subcritical transition in parallel flows takes
place at a value of K c ≈ 370-385. This finding further suggests that the flow instability is likely resulted from the
action of energy gradients, and not due strictly to the eigenvalue instability of linear equations. Using energy gradient
theory, it is also demonstrated that the viscous flow with an inflectional velocity profile is unstable for both two-
dimensional and axisymmetrical flows [10].
For plane Poiseuille flow, this said position where Kmax> Kc should then be the most dangerous location for flow
breakdown, which has been confirmed by Nishioka et al’s experiment [11]. Nishioka et al's [11] experiments for plane
Poiseuille flow showed details of the flow breakdown. The measured instantaneous velocity distributions indicate that
the first oscillation of the velocity occurs at y/h=0.50~0.62. For pipe flow, in a recent study, Wedin and Kerswell [12]
showed that there is the presence of the "shoulder" in the velocity profile at about r/R=0.6 from their solution of the
traveling waves. They suggested that this corresponds to where the fast streaks of traveling waves reach from the wall.
It can be construed that this kind of velocity profile as obtained by simulation is similar to that of Nishioka et al's
experiments for channel flows [11]. The location of the "shoulder" is about the same as that for K max . According to

2
the present theory, this "shoulder" may then be intricately related to the distribution of energy gradient. The solution of
traveling waves has been confirmed by experiments recently [13].

Flow type Re expression Eigenvalue analysis, Energy Experiments, Kmax at Re c (from

Re c method Re c experiments), ≡ Kc
Re c
Pipe Poiseuille Re = ρUD / µ Stable for all Re 81.5 2000 385
Plane Poiseuille Re = ρUL / µ 7696 68.7 1350 389
Re = ρu 0 h / µ 5772 49.6 1012 389
Plane Couette Re = ρUh / µ Stable for all Re 20.7 370 370
Table 1 Comparison of the critical Reynolds number and the energy gradient parameter Kmax for plane Poiseuille flow
and pipe Poiseuille flow as well as for plane Couette flow [7,8]. U is the averaged velocity, u 0 the velocity at the mid-
plane of the channel, D the diameter of the pipe, h the half-width of the channel for plane Poiseuille flow (L=2h) and
plane Couette flow. For Plane Poiseuille flow and pipe Poiseuille flow, the Kmax occurs at y/h=0.5774, and r/R=0.5774,
respectively. For Plane Couette flow, the Kmax occurs at y/h=1.0.

VELOCITY DISTRIBUTION AND FLOW RATE IN THE ANNULUS

The differential equation for the full-developed flow in an annulus (Fig.1) of an incompressible Newtonian fluid
(neglecting gravity force) in the cylindrical coordinates (r,z) is

∂p ∂  ∂u 
0 = −ρ + µ r z  (2)
∂z ∂r  ∂r 

where ρ is the fluid density, u z is the axial velocity, p is the hydrodynamic pressure, and µ is the dynamic viscosity of
the fluid.
Integrating the above equation and using the boundary condition at the walls of inner and outer cylinders, the
velocity distribution along the radius can be obtained as [14]

1 ∂p 2   r  r   r 2 r
2
1− k 2
uz = − R 1 −   + ln  = u 0 1 −   + b ln  , (3)
4µ ∂z   R  ln(1 / k ) R    R  R 

1 ∂p 2
where k=R1/R2 is the radius ratio (R2=R, R1=kR), and u 0 = − R . Here, u 0 is the velocity at the axis of the
4µ ∂z
( )
cylinder for the pipe Poiseuille flow and b = 1 − k 2 / ln (1 / k ) .

The flow rate can be integrated as [14]

) ( )
 1− k 2 
2

(
R2
u
Q = ∫ u z 2πrdr = πR 02
 1− k −
4
. (4)
R1 2  ln(1 / k ) 

The average velocity in the annulus is

U=
Q
πR 2 − π (kR) 2
=
u0 
2 
 (
1 + k 2
− ) (
1− k 2  )
.
ln(1 / k ) 
(5)

The Reynolds number based on average velocity is therefore

3
 (1 − k ) = 0 (1 − k )(1 + k 2 ) − (1 − k )  .
Uρ 2( R − R1 ) ρU 2 R ρu 2 R  2

Re = = (6)
µ µ 2µ  ln(1 / k ) 

CALCULATION OF K IN THE ANNULUS

The parameter K for the flow in an annulus is using Eq.(1)

du z  1 du z d 2 u z 
K = ρu z / µ  + .
 (7)
dr  r dr dr 2 

Introducing Eq.(3) into Eq.(7) and simplifying, the following equations for parameter K can be derived,

1 ρu 0 R   r 2 r  r b 
K= 1 −   + b ln   − . (8)
2 µ   R R   R 2( r / R) 
Let
ρu 0 R ρ (u 0 / 2)2 R
Re1 = = (9)
µ µ

be the Reynolds number of the virtual pipe flow, Eq.(8) can be further expressed as
1   r 2 r  r b 
K = Re1 1 −   + b ln   − . (10)
2   R  R   R 2(r / R) 

When b=0, the above equation simplifies to the pipe flow expression [14]. Equation (10) can be also re-written as

1 r 
K= Re 1 f  , k  , (11)
2 R 
where
r   r r  r
2
b 
f  , k  = 1 −   + b ln   − . (12)
 R    R  R   R 2(r / R) 

The distributions of f (r / R, k ) versus r/R for various k are shown in Fig.2. It is found that there is a maximum or
minimum of f (r / R, k ) along r/R for given k. Therefore, we know from Eq.(11) that there is a maximum or minimum
of K for given Re1 , which corresponds to those of f , respectively. Now, we refer these values to as K max or f max
which is larger in the magnitude between the maximum and minimum of them. We have from Eq.(11)

1 r  1 ρu 0 R  r 
K max = Re1 f  , k  = f  ,k . (13)
2  R  max 2 µ  R  max

CRITICAL REYNOLDS NUMBER OF THE ANNULUS

As we know from energy gradient theory, the dominating parameter determining the critical condition for
instability is K max . Dou [7] demonstrated that the critical value ( K c ) of K max at subcritical condition for wall
bounded parallel flows (pressure driven flows) is a consistent quantity of K c =385. Given this critical value K c for the
subcritical transition condition, the critical values of u 0 , U, and Re can be calculated for any radius ratio of the
annulus. As long as the values of these parameters are below the critical values, the flow in the annulus can be
maintained as a laminar flow regardless of the level of disturbance. Obviously, this observation has significant
implications in engineering.

4
 At the critical condition, from Eq(13), we have the critical value of u 0 given as

2µ K c
u0 c = . (14)
ρR f max
Thus, at the critical condition, the critical flow rate is obtained from Eq.(4) as

) ( )
1 2µ K c  1− k 2 
2

Qc = πR 2

2 ρR f max
(
 1− k −
4
.
ln(1 / k ) 
(15)

The average velocity at critical flow rate is from Eq.(5),

Uc =
Q
πR − πkR
2 2
=
1 2µ K c
2 ρR f max

 1(+ k 2
−) (
1− k 2 )
.
ln(1 / k ) 
(16)

The critical Reynolds number is from Eq.(6),

Re c =
Uρ 2( R − R1 )
=
ρU 2 R  R1 
1 −
K  R 
 = 2 c 1 − 1  (1 + k 2 ) −
(1 − k 2 ) . (17)

µ µ  R2  f max  R2   ln(1 / k ) 

At the critical condition, given as Kc=385, and f max can be found from Eq.(12) (see Fig.2). Thus the critical
values of Q, U and Re versus the radius ratio can be obtained. The critical flow rate and critical Reynolds number are
shown in Fig.3 and Fig.4 respectively. It is found that the critical value of the Re increases with the radius ratio (k) of
the annulus when the ratio k is larger than a value of 0.08. Below k=0.08, the critical value of the Re is less than that of
the Poiseuille flow of circular pipe. Therefore, we can say that the inner cylinder initiates instability for k<0.08 and
enhance stability for k>0.08. Similarly, the critical value of the Q increases with the radius ratio (k) of the annulus
when the ratio k is larger than a value of 0.10. Below k=0.10, the critical value of the Q is less than that of the
Poiseuille flow of circular pipe. The implication is that we can change the critical value of Re by inserting an inner
cylinder at the center of a pipe and thus control the occurrence of turbulence. In this way, turbulence flow in a pipe can
be caused to relaminarise in an annulus. This change may increase the averaged velocity in the passage, but the drag
force is not increased significantly if k is small. This is because the friction coefficient for laminar flow is lower than
that for a turbulent flow. As above discussed, the radius ratio of the inner to the outer cylinders has important
influence on the subcritical condition of transition to turbulence. This finding can be applied to many industrial
processes and fluid transportations.

CONCLUSIONS

The critical condition for the instability of full-developed flow in an annulus is given. The criterion is based on
the energy gradient theory for parallel flow instability. The critical flow rate and critical Reynolds number increase
with the radius ratio for large radius ratio annulus, but there is a tuning point in the range of very small radius ratio of
about k<0.1. It is clear that inner cylinder creates instability for k<0.1; and enhance stability for k>0.1. Following this
principle, the flow transition can be controlled by changing the radius ratio for a given outer radius. The idea can be
employed in the design of fluid flow devices to control the flow status like drag reduction (by keeping to laminar flow)
or increase the mixing of the fluid media (by changing to turbulent flow).
(Blank line : 10 pt)
REFERENCES
(Blank line : 10 pt.)
[1] Schmid, P.J., Henningson, D. S.: Stability and transition in shear flows, Springer-Verlag, (2000)
[2] Drazin, P. G., Reid, W. H.: Hydrodynamic Stability, Cambridge University Press, 2nd Ed., (2004)
[3] Trefethen, L. N., Trefethen, A. E., Reddy, S. C., Driscoll, T. A.: Hydrodynamic stability without eigenvalues,
Science, Vol.261, pp.578-584, (1993)
[4] Grossmann, S.: The onset of shear flow turbulence, Reviews of modern physics, Vol.72, pp.603-618, (2000)
[5] Stuart, J.T.: Nonlinear Stability Theory, Annu. Rev. of Fluid Mech., Vol.3, pp.347-370, (1971)
[6] Bayly, B. J., Orszag, S. A.: Herbert Th. Instability mechanism in shear-flow transition, Annu. Rev. Fluid Mech.,
1988, 20: 359-391

5
[7] Dou, H.-S.: Energy gradient theory of hydrodynamic instability, Technical Report of National University of
Singapore, 2002. Also, presented at The Third International Conference on Nonlinear Science, Singapore, 30 June
-- 2 July, 2004. http://arxiv.org/abs/nlin.CD/0501049
[8] Dou, H.-S., Khoo, B.C., Phan-Thien, N., and Yeo, K.S.: Instability of plane Couette flow,
http://arxiv.org/abs/nlin.CD/0501048
[9] Darbyshire, A.G., Mullin, T.: Transition to turbulence in constant-mass-flux pipe flow, J. Fluid Mech., Vol.289,
pp.83-114, (1995)
[10] Dou, H-S.: Viscous instability of inflectional velocity profile, Recent Advances in Fluid Mechanics, Proc. of the
4th Inter. Conf. on Fluid Mech., July 20~23, 2004, Dalian, China; Tsinghua University Press & Springer-Verlag,
pp. 76-79, (2004)
[11] Nishioka, M., Iida, S., Ichikawa, Y.: An experimental investigation of the stability of plane Poiseuille flow, J.
Fluid Mech, Vol.72, pp.731-751, (1975)
[12] Wedin, H., Kerswell, R.R.: Exact coherent structures in pipe flow: travelling wave solutions, J. Fluid Mech.,
Vol.508, pp.333-371, (2004)
[13] Hof, B., van Doorne, C.W.H., Westerweel, J., Nieuwstadt, F.T.M., Faisst, H., Eckhardt, B., Wedin, H., Kerswell,
R.R., Waleffe, F.: Experimental observation of nonlinear traveling waves in turbulent pipe flow, Science, Vol.305,
pp.1594-1598, (2004)
[14] Papanastasiou, T.C., Georgiou, G.C., Alexandrou, A.N.: Viscous fluid flow, CRC Press, (1999)

0.8

0.6

0.4
2f(r/R)

0.2

k=0.
-0.2 k=0.1
k=0.2
Fig.1 Sketch of the flow in an annulus. k=0.4
k=0.6
R2=R; R1=kR -0.4
k=0.8

0 0.25 0.5 0.75 1

r/R

Fig.2 Function f(r/R) along the radius ratio r/R

for various radius ratios.
3500
2.5
unstable
3000

2
unstable
2500

stable
2000 1.5
stable
Rec

Qc

1500
1

1000

0.5
500

0 0.2 0.4 0.6 0.8 1 0

0 0.2 0.4 0.6 0.8 1
k k

Fig.3 Reynolds number at critical condition Fig.4 Flow rate at critical condition for annulus
for annulus flow at various radius ratio k. The flow at various radius ratio k. The outer radius is
out radius is kept constant. kept constant.