J . Fluid Mech. (1978), 2.01. 85, purl 3, pp.

497-518
Printed in Great Britain
497
Laser anemometer study of flow
development in curved circular pipes
By Y. AGRAWAL, L. TALBOT AND K. GONG
Department of Mechanical Engineering, University of California, Berkeley
(Received 12 October 1976 and in revised form 2 August 1977)
An experimental investigation was carried out of the development of steady, laminar,
incompressible flow of a Newtonian fluid in the entry region of a curved pipe for the
entry condition of uniform motion. Two semicircular pipes of radius ratios & and 3
were investigated, covering a Dean number range from 138 to 679. The axial velocity
and the component of secondary velocity parallel to the plane of curvature of the pipe
were measured using laser anemometry. It was obeerved that, in the upstream region
where the boundary layers are thin compared with the pipe radius, the axial velocity
within the irrotational core first develops to form a vortex-like flow. I n the downstream
region, characterized by viscous layers of thickness comparable with the pipe radius,
there appears to be tbee-dimensional separation at the inner wall. There is also an
indication of an additional vortex structure embedded within the Dean-type secondary
motion. The experimental axial velocity profiles are compared with those constructed
from the theoretical analyses of Singh and Yao & Berger. The quantitative agreement
between theory and experiment is found to be poor; however, some of the features
observed in the experiment are in qualitative agreement with the theoretical solution of
Yao & Berger.
1. Introduction
The problem of flow development in curved pipes is one of fundamental interest in
fluid mechanics. Although considerable effort hasgone into the study of fully developed
flows in curved pipes, the problem of flow development has only recently received com-
parable attention, in part owing to its relevance to such arterial phenomena as the
mixing of injected substances and the distribution of wall shear, which has been con-
jectured to be linked to atherogenesis by Car0 (1973), Fry (1973) and others. The pre-
sent study reports the results of an experimental investigation of this flow development.
The secondary motion in the fully developed as well as in the entry region of curved
pipes is usually explained on the basis of centrifugally induced pressure-gradient con-
siderations as, for example, in the second paper of Dean (1928). In the following, an
alternative description of the flow development process is given in terms of vorticity
arguments which will be of help in interpreting the results of our measurements. The
co-ordinate systems are shown in figure l ( a) . We shall employ the terms circular
motion and circumferential motion to denote, respectively, motion in the stream-
wise and @ directions.
We may imagine a vortex ring, concentric with and of only slightly smaller diameter
than the pipe, as being representative of the vorticity within the very thin boundary
I 7 F L U 85
498 Y . Agrawal, L. Talbot and K. Gong
( b)
FIGURE 1. (a) Co-ordinate systems. ( b) Transport of a vortex line in tho lower half of pipe.
layer which has formed just downstream of t,he entrance to the pipe. Initially, the
vorticity represented by this vortex ring has only one component, us, as indicated in
figure 1 ( b) . Consider now the transport of this vortex line, which being a material line
moves at the local fluid velocity. I n a time At, the segments of this vortex line travel
nearly equal linear distances of order WAt , being the average streamwise velocity.
Thus segments of this vortex line near the inner wall travel through a greater angular
distance 0 than those near the outer wall. The vorticity vector, being tangential to this
vortex line, now in general has two components, wII. and wg, the latter being the stream-
wise vorticity, as shown in figure 1 ( b) , which induces outward motion exterior to the
boundary layer. The presence of this wg, however, results in the production, at the pipe
wall, of vorticity of the opposite sign (see Lighthill 1963), which leads to circumferen-
tial motion in the boundary layer, from the outside of the bend to the inside of the bend.
The production of w, due to convective bending, being proportional to sin$, is
evidently strongest at the pipe top and bottom walls ($ = in, @r). It is therefore to be
expected that the circumferential motion would also be strongest here. The induced
circumferential mot.ion has the further effect of stretching vortex lines near the outer
Flow development i n curved circular pipes 499
wall, thug intensifying o$ and keeping the boundary layer thin, while at the same time
relaxing them in the inner wall region and causing a thickening of the boundary layer
there. The circumferential motion also causes the outer wall to behave in the manner
of a sink, drawing fluid from the core, which in turn counters the diffusive growth of the
boundary layer there. At the inner wall, in opposite manner, the boundary-layer diffu-
sive growth is augmented by source-like behaviour induced by the secondary motion.
We turn our attention now to the initial stages of flow development within the
inviscid core. Since uniform streamwise motion does not constitute a solution to the
inviscid flow problem in the toroidal geometry under consideration, an initially uniform
core flow must progress towards its asymptotic (0 = co) limit, the potential line vortex,
which has a core velocity distribution of W = constant/(R + r cos $). The core will
thus have a tendency to develop a streamwise velocity maximum near the inner wall,
contrary to the fully developed flow case. This requires streaming motion of the secon-
dary flow towards the inner wall in the core, exactly as in the boundary layer. Until
thickening boundary layers overpower this tendency, the wall shear will be greater at
the inner wall than at the outer wall. Interestingly, the development of the core inviscid
vortex has not been anticipated by previous workers. It is perhaps worth remarking
that the mass flux associated with the secondary motion is not conserved in the develop-
ing flow, since the secondary flow streamlines do not become closed until the flow is
fully developed.
Although there have been many theoretical and experimental investigations of
fully developed flow in curved pipes (see Collins & Dennis (1975) for a recent biblio-
graphy) relatively little information is available on the entry-flow problem. The two
theoretical investigations with which comparisons will be made are the analyses of
Singh (1974) and Yao & Berger (1975). One other related study is that by Smith (1976)
of the transition of Poiseuille flow from a straight pipe into a curved one. However,
the solution obtained by Smith is valid only for very small curvature a r O(Re-2))
where Re is the Reynolds number based on the pipe radius a, the mean velocity w
and the kinematic viscosity v (Re = wa / v ) , and thus this solution cannot be compared
usefully with the present results.
2. The experiment
given. Further details can be found in the report by Agrawal (1975).
A brief description of the major features of the experimental apparatus will be
The flow system
Figure 2 shows a schematic diagram of the flow system employed for the study. Two
constant-level reservoirs at a differential elevation of approximately 2.5 m, connected
t o a recirculating pump, provided a steady flow. Flow straightening was carried out
in a 30cm diameter stagnation chamber using a multiple orifice plate and a wire screen.
The circulating liquid was continuously filtered, but a partial bypass of the filter was
provided to reduce capture of particles added for laser anemometry. The stagnation
chamber was connected to the curved pipes by means of a trumpet-shaped inlet con-
traction section of empirical design.
Two curved pipes were studied. These both had an internal diameter of 1.5in. but
differed in radius ratio a = a/R, one having a = &, the other ct = +. The pipes were
17- 2
500
Y. Agrawal, L. Talbot and K. Gong
Overtlow
Filter
Bypass
Pump
Flow control value
Stagnation tank
3
Venturimeter
Heater
cth
FIGURE 2. Schematic diagram of the flow facility.
machined out of Plexiglas in two halves, split at the plane of symmetry.t Both the
circular internal and the planar external surface were polished carefully using a
sequence of polishing pastes and wool cloths. The circulating liquid employed was
a mixture of glycerin and water, in proportions chosen to match closely the index of
refraction of Plexiglas. Exact matching was not possible, thus requiring refractive
corrections in data processing, but approximate matching was required to ensure
intersection of the laser beams within the flow. Careful smoothing and polishing of the
inlet surfaces was carried out to ensure smooth laminar flow at the entrance to the
pipes. The dow exited through a modest contraction of area ratio 4: 1 into a venturi-
meter discharging into the lower reservoir.
Instrumentation and control
Although the actual flow rates used in computation of the Reynolds numbers were
obtained by numerical integration of the measured axial velocity profiles, a venturi-
meter type flow sensor was also employed to monitor and maintain the flow rate
constant to within 2 yo. A Hewlett-Packard differential pressure transducer of type
t This method of fabrication of the curved pipe served two purposes. It ensured that the
internal machined cross-section of the pipe was accurately circular, in contrast t o the coiled
pipes used by other investigators, which most likely had some ellipticity in cross-section due t o
the coiling process. Since the fabrication method produced a pipe of circular cross-section
contained within planar outer walls, the laser measurements were simplified because beam
refractive corrections were required for only one curved interface.
Flow development in curved circular pipes 501
( b)
FIQURE 3. Schematic diagram of the her-Doppler anemometer.
(a) Fringe-mode LDA optics. ( b) Details of probe volume.
267 BC powered by a 5V, 24kHz Schaevitz model SCM-026 source was used to sense
the venturi pressure drop. A specially designed electronically controlled time-propor-
tioning regulated heater was employed to keep the temperature of the circulating fluid
constant to within i 0.2 "C.
Velocity measurements at the entry section of the circular pipes were carried out by
means of hot-film anemometry, because this region was inaccessible optically. A DISA
model 55D01 anemometer was used with a TSI (Thermo-Systems, Inc., St Paul, Minn.)
model 1210 hot-film probe.
The viscosity of the circulating liquid was measured using a Fann V-G Model 35
viscometer. The refractive index of the liquid was determined using a Pulfrich refracto-
meter. Probable errors are estimated to be 5 yo for the viscosity and 0.1 % for the
refractive index.
All detailed fluid velocity measurements were carried out using a laser-Doppler
anemometer operated in the fringe mode (see Durst, Melling & Whitelaw 1972).
A schematic diagram of the LDA optics and signal processing electronics is shown in
figure 3 ( a) . For the axial velocity measurements the 15 mW He-Ne laser beam was
split by an optical flat into two beams 1.93cm apart and brought to intersection
through the side of the pipe by a lens of focal length 55 mm. For the secondary velocity
measurements the optical flat waa replaced by a 15-24 cm diameter rotating bleached
diffraction grating (Stevenson 1970) having 21 600 lines, which served the function
both of a beam splitter and of a frequency shifter. Rotational speeds corresponding to
frequency shifts of 20 kHz (for the lower Dean numbers) and 100 kHz (for the higher
Dean numbers) were used, and a mirror-lens arrangement was used t o bring the two
first-order diffracted beams into intersection through the undersurface of the pipe. The
probe volume dimensions, defined in figure 3 ( b) , were estimated to be a, = b = 44pm,
h = 45pm and d = 256pm for both beam-splitter arrangements. The LDA laser and
optics were mounted on the x, y, z traverse of a milling machine with an accuracy of
502 Y . Agrawal, L. Talbot and K. Gong
traverse of 0.001 in. All of the optics mounted on this table were moved together to
make velocity measurements at different points. Initially, 1 pm polystyrene spheres
were used t o seed the flow. However, Dow Chemical plastic paint Pigment 722 was
found to be a much more economical and equally satisfactory seeding material, though
less monodisperse in size. A typical relaxation time for these micron-sized particles in
our flows was calculated to be of the order of 1ps (Vom Stein & Pfeiffer 1972). Thus
the particles could be expected to follow the flow velocity very closely.
The detection optics consisted of a 55mm f/l.8 Pentax camera lens, a 0.3mm
diameter pinhole, a 3 0 8 bandpass interference filter centred at 63288 and an ex-
tended 5-20 RCA type 4526 photomultiplier tube with a nominal 12% quantum
efficiency at 6328 8. The amplified signal was frequency-demodulated on a TSI model
1090 LDA frequency tracker, which was independently checked for frequency calibra-
tion and linearity against a signal generator and a precision digital clock-counter. The
accuracy was found t o be within the manufacturers specified limit of 0-4 % full scale.
Data acquisition and processing
Axial velocity measurements were made at five axial stations for each of the two
curved pipes. Secondary velocity (i.e. the velocity U in the x direction of figure 1 a)
measurements were carried out at six axial stations within the a = + pipe. Owing to
symmetry about the plane of the pipe axis, measurements are reported only in the
lower halves of the pipes. (The symmetry condition was verified.) Velocity data were
obtained along data lines. These are defined as follows. For axial velocity measure-
ments, the LDA optics were traversed at five different elevations, parallel to the
symmetry plane. Owing to refraction of the laser beams, the actual measurement
points lay on five straight lines making small angles with the symmetry plane. The
y intersections (in inches) of these data lines with the outer wall of the pipe and their
slopes dy/dx are recorded in the figure captions. For the secondary velocity measure-
ments, the data lines were similarly slight,ly rotated away from the normal to the plane
of symmetry, although in the plotted results these data lines are shown as vertical.
Individual axial velocity measurements are believed to be accurate to within
2 yo and secondary velocities to within 5 5 yo. Flow rates at different axial stations
obtained by integration of the axial velocity data were consistent to within 5 yo.
3. Results and discussion
Measurements were first carried out at the pipe inlet by means of a hot-film probe
to determine the nature of the inlet flow velocity profile. These data, obtained at two
different values of Re for each of the two pipes, are displayed in figure 4. Since only
relative values of the velocity were required, the probe output is reported in arbitrary
unit,s. It can be seen that the inlet velocity condition within the inviscid core is indeed
the uniform flow case. We were not able to position the hot-film probe close enough to
the pipe wall to obtain measurements within the very thin inlet boundary layer.
Axial velocity measurements were made by LDA for four different flow rates for
pipe 1 ( a = &) and for three flow rates for pipe 2 ( a = +).The corresponding Dean
numbers K = 2atRe were 139, 200, 251 and 565 for pipe 1 and 183, 278 and 372 for
pipe 2. Secondary velocity measurements were made within pipe 2 at three different
flow rates corresponding to values of K equal to 138, 205 and 679. For the purpose of
Flow development i n curved circular pi pes
. : : - - : : : : ; : . - .
- -
_ - _ - - - _ - -
I I I I I
r5 0.50 0.25 0 -0.25 -0.50 - 1
Horizontal distance from centre of pipe, x (in.)
503
75
FIGURE 4. Inlet axial velocity profiles. A, a = -&, Re 2: 1200; V, a Re N 300;
0, CY. =$, Re 2: 500; x , a = $, Re 1: 150.
this paper we have selected as representative of the axial velocity data the case
K = 565 from pipe 1 (figures 5a- e) and K = 183 from pipe 2 (figures 6a-e). Also, we
have selected as typical of the results for the secondary velocity those for the Dean
numbers K = 138 and 679 (figures 9a-f, IOa-f).
More complete axial velocity data can be found in Agrawals report. This report also
contains secondary velocity data which were obtained before the frequency-shifting
apparatus was available, and are therefore less accurate than the present secondary
velocity data.
Discussion of the experimental results
From the axial velocity profiles in figures 5 and 6, it is immediately clear that qualita-
tive differences in the flow field distinguish the upstream and downstream regions.
The upstream region, immediately adjacent to the entry section, is characterized by
thin boundary layers near the pipe wall and aW/ ay = 0 in the core. However, in the
direction parallel to the plane of symmetry, the streamwise velocity in the core
increases very nearly linearly towards the centre of pipe curvature and the maximum
axial velocity occurs near the inner wall. For a potential vortex, the quantity
G = ( -a/W,) ( a W/ ax) , is equal to the radius ratio a, where the subscript zero denotes
values at pipe centre-line. The values of G were estimated from our data to be 0.041,
0.045, 0.042 and 0.049 for pipe 1 at K = 139, 200, 251 and 565, respectively, while for
pipe 2 the value of G was 0.12 for all three cases, K = 183,278 and 372. Thus it appears
that the vortex waa fully developed, or very nearly so, by the time the flow reached
our first observation station.
7
0

6
0

5
0

v

E

.
-

$
4
0

h

8

?

'
$

3
0

-

d

r
d

2
0

1
0

0

I

C
e
n
t
r
e
-
p
l
a
n
e

1

-
-
-
-
-
*
-
-
-
-
-

O
u
t
e
r

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

7
0

6
0

5
0

5

v

h

.
-

6

4
0

8

-

0

r
d

-

.
-

<

3
0

2
0

1
0

0

1
2
3
u

C
e
n
t
r
e
-
p
l
a
n
e

_
_
_

7
5

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e
f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

F
I
G
U
R
E
S

5
(
a
,
h
)
.

F
o
r

l
e
g
e
n
d

E
W

p
.

5
0
6
.

1
2
3
L
-l
C
e
n
t
r
e
-
p
l
a
n
e

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

;3
u

C
e
n
t
r
e
-
p
l
a
n
e

F
I
G
U
R
E
S

5
(
c
,

d
)
.

F
o
r

l
e
g
e
n
d

s
e
e

f
o
l
l
o
w
i
n
g

p
a
g
e
.

o
r

0

o
r

606 Y. Agrawal, L. Talbot and K. Gong
5
Horizontal distance from centre of pipe, x (in.)
FIQURE 5. Axial velocity profiles for a = &-, K = 565, = 36.1 cm/s. (a) Z/a = 2.4, ( b) Z/a = 12.2,
(c) Z/a = 29.4, ( d) Z/a = 37.6, ( e) Z/a = 57.6. Profiles are shown along data lines 1-5. The initial
values of y at the outer wall of the pipe for these lines, their sIopes and the corresponding symbols
are: (1) A, -0.050, -0.0034; (2) 0, -0.200, -0.0141; (3) V, -0.350, -0.0270; (4),+, -0600,
- 0.0465; (5) x , - 0.650, - 0,0969.
We now turn t o the relatively more complex flow development in the downstream
region, where the viscous-layer thickness, at least near the inner wall, is O( a) , as
seen in figures 5 (b-e) and 6 (b-e). The remains of the inviscid core can be easily recog-
nized by its vortex-like velocity distribution. It may be observed from figure 5 ( b )
that a marked decrease in the axial velocity has occurred near the inner wall in the
vicinity of data line 1, and that the smallest value of the velocity gradient d W/ dx at
the pipe wall occurs nearest the plane of symmetry. Also, data line 2 exhibits a clearly
defined second maximum or peak in the axial velocity near the inner wall.
As the flow develops (figures 5c-e), one observes the continued erosion and velocity
increase of the core region, while the second maximum appears progressively further
away from the centre-plane, appearing at data lines 3 and 4 in figures 5( c- e) , and at
the same time becomes increasingly less pronounced. Far downstream (figure 5e) , the
axial velocity maximum appears on data line 1 near the outer wall. Within the stream-
wise distance available in our studies, the second maxima on data lines 3 and 4 did
not disappear but did become diminished. It may be noted that doubly peaked
velocity profiles observed here were also observed by Austin (1971) in his studies of the
C
e
n
t
r
e
-
p
l
a
n
e

I

I

r
-
-
-
-
-
*
-
-
-
-
-
-
J

L

7
5

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

C
e
n
t
r
e
-
p
l
a
n
e

1

t-
-
-
-
-
-
*
-
-
-
-
-
-
7

1
0

0

-
0
.
2
5

-
0
.
5
0

-
I

0
.
7
5

0
.
5
0

0
.
2
5

0

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

5

F
I
G
U
R
E
S

6
1
0
,

b
)
.
F
o
r

l
e
g
e
n
d

s
e
e

p
.

5
0
9
.

C
e
n
t
r
e
-
p
l
a
n
e

1

r

-
-
-
-
-
*
-
-
-
-
-
-
7

2
3
w

4
0

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

o
f

p
i
p
e
,

x

(
i
n
.
)

4
0

C
e
n
t
r
e
-
p
l
a
n
e

I

r
-
-
-
-
-
*
-
-
-
-
-
-
1

2
-

5

0
.
7
5

0
.
5
0

0
.
2
5

0

-
0
.
2
5

-
0
.
5
0

-
C

H
o
r
i
z
o
n
t
a
l

d
i
s
t
a
n
c
e

f
r
o
m

c
e
n
t
r
e

p
i
p
e
,

x

(
i
n
.
)

5

F
'
I
G
U
R
E
S

6
(
c
,

d
)
.

F
o
r

l
e
g
e
n
d

s
e
e

f
o
l
l
o
w
i
n
g

p
a
g
e
.

Flow development in curved circular pipes 509
40
-
rn .
5 30
h
2
8
v
c
.-
-
0
-
. ; 20
d
10
0
C
23%s
5 0.50 0.25 0 -0.25 -0.50 -( 75
Horizontal distance from centre of pipe, x (in.)
FIGURE 6. Axial velocity profiles for a = i, K = 183, v = 18-1 cm/s. ( a) Z/a = 1.84, ( b ) Z/a =
3.68, (c) Z/a = 7.34, (d) Z/a = 13.44, ( e) Z/a = 19.54. Profiles are shown along data lines 1-5. The
initial values of y at the outer wall of the pipe for these lines, their slopes and the corresponding
data symbols are: (1) A, -0.050, -0.0023; (2) 0, -0.200, -0.0096; (3) V, 0.350, -0.0183;
(4) +, -0.500, -0.0313; (5) X , -00.650, -00636.
flow development starting from a parabolic entry profile. I n addition, the same type
of profiles has been observed by Pedley, Schroter & Sudlow (1971) and Brech &
Bellhouse (1973) in studies of t,he entry region downstream of a branch in a vessel,
where flows with curvature of the core-flow streamlines are similarly involved.
Flow development in the downstream region for the low-K case is substantially the
same as that for high K, but a few qualitative differences are t o be noted. Most im-
portant, the doubly peaked axial velocity profiles just discussed are not observed
(see figures 6c-e).
Figure 7 shows a plot of the isovelocity contours constructed from the data in
figure 5 ( c) (K = 565, .?/a = 29-4). One observes from these isovels that the second peaks
in the axial velocity profiles are due t o a ridge-like zone of high axial velocity which
persists in the lower part of the inner region of the pipe, although the high axial velocity
fluid in the region of the midplane has shifted towards the outer wall. At low Dean
numbers, the isovels in the core region are more or less vertical and do not double back
on themselves, thus the second maxima do not appear.
I n figure 8 the axial velocity profile measured close to the plane of symmetry (data
line 1) for the case 01 = 3, l l a = 19.54 and K = 372 is compared with the fully developed
510 Y . Agrawal, L. Talbot and K. Gong
45
FIGURE 7. Axial isovelocity contours constructed from the data in figure 5( c) .
Numbers assigned to the isovels are the velocities in cm/s.
(x + a) / a
FIGURE 8. Axial midplane velocity profile comparisons. - a - , fully developed flow computed
by Collins &Dennis at K = 369.5; o - . * 0 , measured by Adler at K = 372; x - x , present measure-
ments at K = 372, l / a = 19.54, data line 1.
flow measurements of Adler (1934) for K = 369.5 and the calculated results of Collins &
Dennis (1975) for K = 372. The profiles are very similar in shape, but it is evident that
the flow has not yet become fully developed in the present case. The entry length 1, for
a straight pipe at a Reynolds number wa / v = 492 (corresponding to the data of
figure 8) can be calculated from the results of Fargie & Martin (1971) to be
lJa = 0.196Re = 96, which is considerably greater than the value of t/a = 19.54
corresponding to the present data. According to the calculations of Yao & Berger,
the entry flow length 1, for flow development in a curved pipe for the conditions of
figure 8 is about 10 yo greater than for a straight pipe. Evidently, the final stages of
flow development occur rather slowly, since although 1/a is much smaller than Isla or
l Ja, the developing velocity profile is not much different from the fully developed one.
I
n
s
i
d
e

I
n
s
i
d
e

~
E
A

I
n
s
i
d
e

O
u
t
s
i
d
e

(
4

f

'

F
I
G
U
R
E
S

9
(
a
-
c
)
.

F
o
r

l
e
g
e
n
d

s
e
e

f
o
l
l
o
w
i
n
g

p
a
g
e
.

I
n
s
i
d
e

n

w

O
u
t
s
i
d
e

I
n
s
i
d
e

n

F
I
G
U
R
E

9
.

S
e
c
o
n
d
a
r
y

v
e
l
o
c
i
t
y

p
r
o
f
i
l
e
s

f
o
r

a

=

?
,

K

=

1
3
8
.

(
a
)
l
/
a

=

1
-
8
3
,

(
b
)
Z
/
a

=

3
.
6
7
,

(
c
)

l
/
a

=

6
.
1
1
,

(
d
)
l
/
a

=

9
.
1
6
,

(
e
)

Z
/
a

=

1
2
-
8
3
,

(
f
)

1
/
a

=

1
6
.
4
9
.

T
h
e

v
e
l
o
c
i
t
i
e
s

a
r
e

p
l
o
t
t
e
d

o
n

E
L

s
c
a
l
e

f
o
r

w
h
i
c
h

t
h
e

d
i
s
t
a
n
c
e

b
e
t
w
e
e
n

a
d
j
a
c
e
n
t

d
a
t
a

l
i
n
e
s

c
o
r
r
e
s
p
o
n
d
s

t
o

5
c
m
/
s
.

I
n
s
i
d
e

n

F
I
G
U
R
E
S

1
0
(
a
-
c
)
.

F
o
r

l
e
g
e
n
d

s
e
e

f
o
l
l
o
w
i
n
g

p
a
g
e
.

-
l
-
-
-
-
l

I
n
s
i
d
e

O
u
t
s
i
d
e

(
d
)

I
n
s
i
d
e

n

I
n
s
i
d
e

n

F
I
G
U
R
E

1
0
.

S
e
c
o
n
d
a
r
y
v
e
l
o
c
i
t
y

p
r
o
f
i
l
e
s

f
o
r

a

=

+
,
K

=

6
7
8
.

(
a
)
l
/
a

=

1
.
8
3
,

(
b
)
l
/
a

=

3
.
6
7
,

(
c
)

l
/
a

=

6
-
1
1
,

(
d
)
1
/
a

=

9
.
1
6
,

(
e
)
1
/
a

=

1
2
.
8
3
,

(
f
)

l
/
a

=

1
6
.
4
9
.

T
h
e

v
e
l
o
c
i
t
i
e
s

a
r
e

p
l
o
t
t
e
d

o
n

a

s
c
a
l
e

f
o
r

w
h
i
c
h

t
h
e

d
i
s
t
a
n
c
e

b
e
t
w
e
e
n

a
d
j
a
c
e
n
t

d
a
t
a

l
i
n
e
s

c
o
r
r
e
s
p
o
n
d
s
t
o

1
0
c
m
/
s
.

Flow development i n curved circular pipes 516
We consider now the secondary velocity data, shown in figures 9 (a) - ( f ) for K = 138
and figures l O( a) - ( f ) for K = 678, both sets of data having been obtained in the
a = 3 pipe. The K = 138 data exhibit about what one would anticipate. At the first
measurement station, l / a = 1.83 (figure 9a) , the potential vortex has already been
established, and the flow is now outward within the core and inward within the cir-
cumferential wall boundary layer. As the flow proceeds (figure 9b, l / a = 3-67) the core
velocity increases and the wall boundary layer thickens, and this trend is evident also
at 1/a = 6-11, figure (9c). At 1/a = 9.16 (figure 9d) there is an indication that separa-
tion of the secondary flow boundary layer? may have occurred (on the innermost data
line), and there is an enhancement of the outward secondary flow veIocity in a region
between the midplane and the pipe wall perhaps caused by the fluid returning from the
wall boundary layer at the curve of separation. This behaviour is evident also at
l / a = 12.83 (figure 9e) and at l / a = 16.49 (figure 9f). One notes that the magnitude
of the outward secondary motion, as, for example, in the vicinity of the midplane, is
a decreasing function of l / a in the far downstream region. This is what one expects in
the flow development process. I n the early stages of flow development, there must be
a net outward mass flux associated with the movement of the axial velocity maximum
towards the outer wall. As the flow approaches the fully developed limit, this net mass
flux diminishes until finally it vanishes when the secondary flow streamlines become
closed in the fully developed limit. One notes also that the maximum circumferential
velocity within the wall boundary layer occurs at about $ = Qrr, as anticipated on the
basis of the vorticity arguments given earlier.
The secondary flow data obtained at K = 678 (figures lOa-f) are more difficult to
interpret. Although the secondary velocities at 1/a = 1.83 (figure 10a) and l / a = 3.67
(figure l ob) exhibit the same qualitative features as are observed at the lower Dean
number, namely more-or-less uniform outward motion within the core and strong
inward motion within the narrow wall boundary layer, the situation changes quite
dramatically further downstream. One observes additional flow reversals along the
second data line (counting from the inner wall), first at 1/a = 6.11 in figure 1O(c) and
then along both the second and the third data line at .!/a = 9.16 in figure lO(d). The
pattern persists to 1/a = 12.83 (figure 10e) and vestiges of it still remain at l / a = 1649
(figure 10 f ). It also appears from figure 10 ( f ) that separation of the secondary flow
boundary layer may have occurred at the first data line.
A possible interpretation of the flow revexsals along the second and third data lines
as being due to two embedded helical motions is sketched in figures lO(c) and (d).
However, we have thus far been unable to devise a satisfactory explanation for their
existence. Although both clockwise and counterclockwise streamwise vorticity is
present in the circumferential boundary layer, as discussed in the introduction, it is
not apparent how this vorticity could be the source of that contained in the embedded
vortical region. Intuitively, one might expect that such motions could appear at the
inner wall, as a result of sepaIation, and indeed there is the hint of the possible exist-
t I n laminar, fully devcloped curved-pipe flow, according to our present understanding (see
Smith (1975) for a discussion of the existence of an attached flow solution for a circular cross-
section) the boundary curve given by $ = n, r = a is the envelope of limiting streamlines on the
inner wall. We use the term separation to describe the existence of two envelopes of limiting
streamlines on the inner wall symmetrically disposed about the midplane at (not necessarily
constant) values of $ 77.
516 Y. Agrawal, L. Talbot and K. Gong
FIGURE 1 1 . Comparisons between measured axial velocities at K = 565, a = $6 and theoretical
predictions of Singh ( - - - - -) and Yao & Berger (----). Left edge of each profile corresponds to
outer wall of pipe. Profiles in a horizontal row are along data lines 1-5, from left to right, at a
given Zla. The values of l / a from top to bottom are 2.4, 12.2, 29.4, 37.6 and 57.6.
ence ofvortex motion, albeit very weak, in the region ofthe first data line in figure 10 (d).
The embedded vortex region does coincide more or less with the ridge of high axial
velocity (figure 7), and the recirculation within this region may thus be the cause of
the persistence of the high axial velocity there. Evidently, if our conception of the
fully developed flow is correct, both the embedded vortex region and the separated
flow region must eventually be engulfed by the Dean-type secondary motion. It is
of interest to note that Scarton, Shaw & Tsapoges (1977) report the observation, by
means of a dye flow-visualization technique, of two counter-rotating trapped inner
wall vortices. From the description given, it would appear that these trapped vortices
are the result of flow separation.
Another interesting feature of the secondary flow field seen in figures 10 ( c) - ( e) is the
relatively narrow region of strong outward flow above the hypothesized embedded
vortex region. Conceivably this region is analogous t o the central viscous layer
postulated by Adler and Yao & Berger, although they placed its location at the mid-
plane of the pipe. A similar region, though much less pronounced, was noted for the
K = 183 data.
Comparison with theory
Axial velocity profiles were constructed from the theoretical solutions of Singh and
Yao & Berger, along the data lines of our measurements, and are shown in figure 11.
Since the region of validity of Singhs solution extends only a streamwise distance
O( a) from the entry, we have compared his results only at the first axial measurement
station. The results of Yao & Berger are compared at all stations.
Evidently, Singhs solution does not predict the initial formation of the vortex-type
inviscid core. This stems from the expansion procedure employed by him, where the
effect of curvature appears only through a term of order al Re3, implying uniform
flow as the inviscid solution. Further, since he did not compute flow acceleration in the
Flow development in curved circular pipes 517
core t o account for boundary-layer displacement, the normalized mass flux from his
profile appears to be less than unity.
It was not possible from our experiments t o determine the location of the cross-
over point, which is the axial distance from the entrance at which the wall shear at
the outer wall becomes equal t o that at the inner wall, and was found by Singh to be
at z = 1 . 9 ~ and by Smith t o be 1*51a, in both cases independent of Re. The velocity
gradients at the inner and outer walls on data line 1 at the first measurement station
for both pipes would seem t o indicate that the cross-over has occurred upstream of
these measurement stations, in order-of-magnitude agreement with the predictions,
though of course determination of wall shear by extrapolation of a velocity profile is
not an accurate procedure.
The analysis of Yao & Berger predicts a slower rate of flow development than is
observed, with respect t o the displacement of the velocity profile maximum towards
the outer wall. The flow model used by Yao & Berger involved the assumptions of linear
variation with x of the axial velocity in the core and secondary flow streamlines parallel
t o the plane of symmetry (following the analysis of Barua (1963) for the fully developed
case), and the boundary layer was calculated by a momentum-integral method. These
assumptions obviously preclude the possibility of obtaining from the analysis detailed
information on the velocity profiles either in the core or in the boundary layer. How-
ever, the analysis does predict separation of the secondary flow in the inner wall region
as is suggested by the experiments. The integrated normalized mass flux obtained
from the Yao-Berger analysis is less than unity, probably because in their mass
conservation condition they use an average value for the circumferentially varying
thickness of the boundary layer. Some of these shortcomings in the Yao-Berger
analysis undoubtedly could be rectified so as to give improved agreement with
experiment.
4. Concluding remarks
The measurement of curved-pipe entry flow reported here exhibit two features of
particular interest. The first is the transition immediately downstream of the entry
of the inviscid axial velocity profile from a uniform velocity distribution to a vortex-
type distribution. Although a simple explanation based on inviscid flow behaviour can
be given for this transition, its existence seems to have been overlooked in the analytical
investigations which have been made of the entry flow pIoblem. The second feature of
interest is the observation, for the highest Dean numbers investigated, of the develop-
ment in the downstream region of pronounced axial velocity maxima near the inner
wall. These maxima move away from the plane of symmetry and decrease in strength
as the flow proceeds downstream. They appear to be linked t o the presence of an
embedded vortex pair. Separation of the circumferential boundary layer on the inner
wall is also indicated.
The theoretical analyses which are at present available for curved-pipe entry flow
do not predict adequately the flow development process, but it is hoped that the
experimental results reported here will be helpful in suggesting directions in which
these analyses can be improved.
518 Y . Agrawal, L. Talbot and K. Gong
This work was supported by the National Science Foundation under grant no.
ENG-73-03970. The authors are indebted to Professor S. A. Berger for many helpful
discussions.
REFERENCES
ADLER, M. 1934 2. angew. Math. Mech. 14, 257.
AGRAWAL, Y . C. 1975 Laser velocimeter study of entrance flows in curved pipes. Univ. CaZ;,f.
Coll. Engng Berkeley, Rep. FM-75-1.
AUSTIN, L. 1971 The development of viscous flow within helical coils. Ph.D. thesis, University
of Utah.
BARUA, S. N. 1963 Quart. J . Mech. Appl . Math. 16, 61.
BRECH, R. & BELLHOUSE, B. J. 1973 Cardiovasc. Res. 7, 593.
CARO, C. J. 1973 In Atherogenesis: Initiating Factors, Ciba Syrnp. (New Ser.), 12, 127. Elsevier:
COLLINS, W. M. & DENNIS, S. C. R. 1975 Quart. J . Mech. Appl . Math. 28, 133.
DEAN, W. R. 1928 PhiZ. Mag. 5, 673.
DURST, F., MELLING, A. & WHITELAW, J. H. 1972 J . Fluid Mech. 56, 143.
FARGIE, D. & MARTIN, B. W. 1971 Proc. Roy. Soc. A 321, 461.
FRY, D. L. 1973 In Atherogenesis: Initiating Fuctors. Ciba Syrnp. (New Ser.), 12, 93. Elsevier:
LIGHTHILL, M. J. 1963 In Laminar Boundary Layers (ed. L. Rosenhead), p. 84. Oxford Uni-
PEDLEY, T. J., SCHROTER, R. C. & SUDLOW, M. F. 1971 J . Fluid Mech. 46, 365.
SCARTON, H. A. , SHAW, P. M. & TSAPOGAS, M. J. 1977 Relationship of the spatial evolution
of secondzry flow in curved tubes to the aortic arch. In Mechanics in Engineering, pp. 111-
131. Univarsity of Waterloo Press.
Excerpta Medica.
Excerpta Medica.
versity Press.
SINCH, M. P. 1974 J . Fluid Mech. 65, 517.
SMITH, F. T. 1975 J . Fluid Mech. 71, 15.
SMITH, F. T. 1976 Proc. Roy. Soc. A 351, 71-87.
STEVENSON, W. H. 1970 Appl , Optics 9, 649.
VOM STEIN, H. D. & PFEIFFER, J. J. 1972 AppZ. Optics 11, 305.
YAO, L. S. & BERGXR, S. A. 1975 J . Fluid Mech. 67, 177.