INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS, VOL.

11, 1113-1 126 (1990)

NUMERICAL STUDIES O F FLUID FLOW THROUGH TUBES

WITH DOUBLE CONSTRICTIONS

T. S. LEE
Mechanical and Production Engineering Departments, Faculty of Engineering, National University of Singapore,
Singapore 051I

SUMMARY
The flow fields in the neighbourhood of double constrictions in a circular cylindrical tube were studied
numerically. The effects on the streamline, velocity and vorticity distributions as the flow passes through the
constrictions in the tube were studied in the Reynolds number range 5-200. Double constrictions with
dimensionless spacing ratios of 1,2, 3 and co were studied for a 50% constriction.
It is noted that when the Reynolds number is below 10, no recirculation region is formed in the above
constricted flow. For Reynolds numbers greater than 10, a recirculation region forms downstream of each of
the constrictions. For constriction spacing ratios of 1,2, and 3, when the Reynolds number is high, a
recirculation region spreads between the valley of the constrictions. The recirculation region formed between
the two constrictions has a diminishing effect on the generation of wall vorticity near the second constriction
area. In general, the peak value of wall vorticity is found slightly upstream of each of the constrictions. When
the Reynolds number is increased, the peak wall vorticity value increases and its location is moved upstream.
Maximum wall vorticity generated by the first constriction is found to be always greater than the maximum
wall vorticity generated by the second constriction. The extent of this spreading of the recirculation region
from the first constriction and its effects on the second constriction depend on the constriction spacing ratio
and the flow Reynolds number.

KEY WORDS Constrictions Recirculation Vorticity Streamline

INTRODUCTION
The flow field in the neighbourhood of constrictions in tubes has been of great interest to fluid
dynamicists because of its many engineering applications. This type of configuration is used in
heat exchangers in order to enhance heat transfer performance. A wavy configuration of tubes has
also been of great interest to biofluid dynamicists because of its relationship to localized stenoses,
blood and urinary flow, and for the optimal design of artificial organs. Similar viscous fluid flow
past wavy boundaries has also been of great interest to researchers because of its importance in
phenomena such as the generation of wind waves on water, the stability of a liquid film in contact
with a gas stream, the transpiration cooling of re-entry vehicles and rocket boosters, film
vaporization in combustion, fluid flow in pipes with fittings, etc.
An earlier numerical work on this type of problem was done by Lee and Fung' to study the
flow in locally constricted tubes in the Reynolds number range 0-25. A bell-shaped constriction
specified by a Gaussian normal distribution curve was used. Similar numerical studies were
carried out by Oberkampf and Goh.' An outflow-type boundary condition was used by Lee and
Fung, whilst Oberkampf and Goh used an infinity condition. Constrictions with other type of
profiles such as a sinusoidal function were used by Desphande et aL3 to model the steady laminar

027 1-209 1/90/161113-14$07.O0 Received October 1989

0 1990 by John Wiley & Sons, Ltd. Revised January 1990
1114 T. S. LEE

flow through vascular stenoses. A survey of the numerical methods used for constricted physiolo-
gical flows was presented by M ~ e l l e r .The
~ separating flow through a severely constricted
symmetric tube was studied analytically by Smith.' The main separation was shown to take place
on the upstream constriction surface. In more recent work, Wille6 studied the pressure and flow
fields in arterial stenoses simulated by mathematical models. Sobey' studied numerically the flow
through furrowed channels and investigated the Reynolds number effects on the separated flow.
Stephanoff et a/.' compared the numerical results of Sobey with experimental observations
through flow visualization techniques. Velocity measurements in the neighbourhood of axisym-
metric constrictions in rigid tubes were investigated by Ahmed and Giddens' using laser Doppler
anemometry and hydrogen flow visualization techniques. Experimental flow visualization of
streamlines in steady flow through constrictions was also obtained by Siouffi et al. l o in a study on
the effects of unsteadiness on the flow through stenoses. Prata and Sparrow' obtained numerical
solutions for a periodic, fully developed regime in an annulus of varying cross-section of double
pipe in a heat exchanger. On the basis of the computed heat transfer coefficients and pressure
drops, the periodic annulus appears to be an attractive enhancement of the configuration relative
to the annulus of axially unchanging cross-section. Other related recent studies of constricted
flow include a study of laminar steady flow in sinusoidal channels by Tsangaris and Leiter" using
a perturbation technique. O'Brien and Ehrlich' studied numerically the pulsatile flow through a
constricted artery using conformal mapping. Experimental measurements and prediction of flow
through a replica segment of a mildly atherosclerotic coronary artery were done by Back et al.I4
In the present work the flow behaviour in a double constricted symmetrical tube was studied
numerically using a finite difference model. The double bell-shaped constricted tube was trans-
formed into a rectangular solution domain using a generalized mapping function. The dynamics
of the flow describing separation, reattachment and the formation of recirculation eddy were
revealed by the streamline, velocity and vorticity fields. Numerical results were obtained for
Reynolds numbers in the range 5-200 with a 50% constriction. Constrictions with dimensionless
constriction spacing ratios of 1,2, 3 and co were studied. The findings of the present study were
also compared with available experimental and numerical work of other investigators.

PROBLEM FORMULATION
The geometrical configuration of the tube with double constrictions and its co-ordinate system is
shown in Figure l(a). The unsteady governing equations are used to solve for the steady state
flow fields considered in this study. Constant fluid properties are assumed and the flow is
considered axisymmetric and laminar. The governing equations are:

momentum equations

continuity equation
av, +-0, + av,
- -= 0.
ar r az
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1115

r
I

Figure 1. Model of the double constriction tube and its co-ordinate systems: (a) The original physical space; (b) The
Transformed solution domain

The vorticity 5 is defined as

By eliminating the pressure terms in equations (1) and (2), and with the use of equation (3)
together with the definition of vorticity given by equation (4), the vorticity transport equation is
obtained as

The incompressible streamfunction $ is defined by

1 a* - a*
=--1
u =--
r az’ ’ r ar’
where v, and v, are the velocity components in the r- and z-directions respectively.
Hence equation (4) can be expressed as
[=- a2* a211/ la+
ar2 az2 r dr (7)

From the geometry of the constriction, the tube radius a, is identified as a characteristic length.
The inlet velocity v , on the axis of the tube far away from the constriction is taken as the
1116 T.S. LEE

characteristic velocity. The following dimensionless variables are defined:

r* = riao, z* = z lao, 0
: = v,Jv,, v: = vz/v,,
** = */(~,4% [* = [/(vmJao)9 t* = t/(ao/v,).

The dimensionless vorticity transport equation can then be expressed as

where the Reynolds number Re = v,ao/v.

The dimensionless vorticity-streamfunction equation becomes

and the dimensionless velocities are

(In subsequent expressions the asterisk is dropped for simplicity.)

In dimensionless form the geometry of the constrictions may be described by the following bell-
shaped Gaussian distribution profile:
f ( z ) = 1 - c 1 exp[ - c2(z - s ) ~ ] , (12)
where c1 is the constriction ratio ( = (D - d , ) / D ) , c2 is a shape constant and s is the dimensionless
distance of the constriction from the inlet plane.
Referring to Figure l(a), for the double constrictions considered here, c1 = 0.5, c2 = 04, s = s1
for z1 < z < z2, s = s2 for z3 < z < 24 and c1 = 0 elsewhere; z1 and z2 are the upper and lower
limits of the first constriction; z3 and z4 are the upper and lower limits of the second constriction;
s1 and s2 are the distances of the first and second constrictions from the inlet plane respectively.
The spacing between the two constrictions is given by S = s1 - s2, and sl = 2.0 in this study.
When the dimensionless spacing ratio S J D = 00, this is equivalent to a single constriction tube
with c1 = 0.5, c2 = 0.4, s = s1 for z1 < z < z2 and c 1 = 0 elsewhere.
For the present studies the incoming flow is assumed to be Poiseuillean and the outflow is
assumed to be unrestricted. A non-slip boundary condition is assumed for the tube wall. The flow
is assumed to be symmetrical about its axis, with v, = 0 axis, i.e.
at the inlet v, = 0, v, = 1 - r 2 ,
at the outlet a v , p z = avz/az= 0,
along the tube axis 0, = 0, av,iar = 0,

along the tube wall v, = v, = 0. (13)

The boundary streamfunction values follow closely from those of the velocity boundary
conditions described above. II/ = 0 is arbitrarily chosen along the axis of the tube. The inlet and
tube wall streamfunctions are then obtained from the assumed inlet velocity distribution. The
outlet streamfunction is obtained from the above assumption of unrestricted flow. From the
definition of vorticity given in equation (9), the vorticity boundary conditions follow those of the
streamfunction and the velocity boundary conditions.
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1117

NUMERCIAL SOLUTION
The cylindrical (I,z) co-ordinate system shown in Figure l(a) is not suitable for accurate
evaluation of the numerical solution near the irregular boundary. One of the most efficient
methods of dealing with this problem is to use a co-ordinate transformation as shown in
Figure l(b). The tube with the bell-shaped constrictions is mapped into a rectangular solution
domain and the flow fields are solved with a finite difference method. The new co-ordinate system
is defined as follows:
& = F,(z) = 2, ll = F’b.7 z) = r/f (4. (14)

Expressing the partial derivatives in the new co-ordinates, the governing equations become

+ -a[( -
a&
d2E)_ -
dZ2
;’I,

The velocities are

where z = E, r = v f (E) and def d r = 0, deldr = 1.

Similarly, the boundary conditions for the streamfunctions, velocities and vorticity are also
expressed in the e-q co-ordinate system.
The rectangular domain in the &-q co-ordinate system in Figure l(b) is overlaid with a regular
finite difference mesh. At the node points the finite difference solutions to equation (15) with their
boundary conditions are obtained through an alternating direction implicit (ADI) procedure
proposed by Samarskii and Andreev.” The successive overrelaxation (SOR) method with a
relaxation parameter o = 1.1 is used to solve the vorticity-streamfunction equation (16). All
spatial derivatives are approximated by second-order-accurate central differences.The convective
terms in equation ( 15) are approximated by the second-order upwind differencing method. Three-
point backward and forward difference formulae are used for the derivatives at the boundaries.
The vorticity boundary values are obtained by considering the Taylor series expansion of $ into
+
the solution region and taking into consideration the and the velocity at the boundary. The
mesh size used for the solutions illustrated here is 21 x 151. The steady state solution of equations
+
(15) and (16) is said to have converged when a difference of 0.01% of a referenced and q is
detected. This has prove satisfactory. The streamfunction contours, the velocity fields and the
vorticity contours are noted to be steady after the above criteria are satisfied.
1118 T.S . LEE

RESULTS AND DISCUSSION

Characteristics of the recirculating flow region and flow through the constrictions were investig-
ated for constriction spacing ratios of 1,2, 3 and co.The streamlines and vorticity distributions
for this investigation are shown in Figure 2. At a low Reynolds number of 5 (Figure 2(a)) the
streamlines and vorticity contours for constriction spacing ratios of 1,2, 3 and co are similar. The
flow around each constriction of the double constriction tube behaves in a manner similar to that
in an independent single constriction tube, i.e. without any interference of the flows with each
other. Hence the vorticity contours are also similar near each of the constrictions. No recircul-
ation region appears in the constricted tube at a Reynolds number of 5. As the Reynolds number
approaches 10 (Figure 2(b)),a recirculation eddy appears downstream of each constriction. At a
Reynolds number greater than 50 (Figure 2(c)) the vorticity fields are substantially altered and
the closed contours of the vorticity distribution are seen to be advected downstream from each
constriction. The recirculatory eddy from the upstream constriction is also spread downstream
and this affects the flow passing through the downstream constriction. A recirculation zone now
exists which fills part of the valley region between the two constrictions. Once the recirculation
flow field between the constrictions is established, there is a separation streamline that divides the

I I3
SD

( a ) REYNOLDS NUMBER = 5

Figure 2. (a)
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1119

SIC!. 00

[ b ) REVNOLOS NUMBER 2 10

Figure 2. (b)

flow into two parts: the recirculating flow field between the two constrictions and the main flow
field near the centre of the tube with relatively straight and parallel streamlines.
The separation and reattachment points of the recirculation eddies formed downstream of each
of the constrictions for different constriction spacing ratios are shown in Figure 3. It can be seen
that when the Reynolds number is increased, the separation point on the surface of the
constriction where the recirculation eddy begins to form moves slightly upstream of the throat.
The reattachment point where the recirculation eddy terminates on the surface of he constricted
tube spreads downstream of the throat. When a steady recirculation region is established between
the two constrictions, there is then little change to the separation and reattachment points for the
flow in the valley region. However, the reattachment point of the downstream constriction
spreads further as the Reynolds number is increased and eventually approaches that of a single
constriction corresponding t o S/D = 00.
The variations of the centreline velocities due to the influence of the constriction spacing ratios
and Reynolds numbers are shown in Figure 4. For a Reynolds number greater than 5 it is noted
that the maximum centreline velocity does not occur at the throat. At this point the fluid is still
experiencing an acceleration, so that the maximum centreline velocity in the field occurs slightly
downstream of each of the constrictions. At higher Reynolds number ( R e 2 25), where a zone of
1120 T.S. LEE

SIO.1

SIO m

[ c ) REVHOLDS NUMBER = 50
Figure 2. Streamline profiles for different constriction spacings: (a) Re = 5; (b) Re = 10; (c) Re = 50

200 -

i:
a
w
150 -
=
0 SECOND CONSTRlCllON [SIOzl 1
0
v)
A FIRST CONSlRlCTlON (S/O =2 I
p 100 - SECOND CDNSlRlCllON IUD-2 1
>
E
W
x FIRST CONSTRlCIlDN lUO.3 )
0 SECOND CONSTRICTION [S/O =3 1
50 -
0 I
0 1 c 6 11 10 12 1I
AXIAL OISTANCE
Figure 3. Separation and reattachment of flow for different constriction spacings
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1121

SID = 1. t,t2.O. I, .LO

0
I I

Y
0'
2
Y 0'
>

S/O=m,+=2.0,~=m
I I I

recirculation is established between the constrictions with S / D = 1,2 and 3, the centreline
velocity does not have the opportunity to recover fully before passing through the second
constriction. Note that the maximum centreline velocity shifts downstream as the Reynolds
number increases; the maximum value near the second constriction is also slightly higher than the
maximum value at the first constriction.
Characteristics of the flow through constrictions can also be described by the velocity vector in
the tube. The velocity vectors in Figure 5 show that as the fluid flows into the converging portion
of the constriction, the velocity increases in magnitude and the vectors point towards the axis of
the tube. The formation of the recirculation region in the flow is indicated by the negative velocity
vectors in a reversed flow. The wall vorticity values, which are related to the velocity distribution,
are also of considerable interest to researchers since they are related to the wall shear stress. As
shown in Figure 5, the magnitude of the wall vorticity value increases rapidly when the flow
approaches the constriction, and reaches a peak value near the maximum constricted area for a
Reynolds number less than or equal to 10. At higher Reynolds numbers the peak wall vorticity
value is found slightly upstream of the maximum constricted area. At a location downstream of
this peak value the wall vorticity decreases rapidly and will reverse to negative values when
1122 T. S . LEE

separation begins at the wall of the tube. It is observed that the peak wall vorticity value increases
with increasing Reynolds number. For the present investigation with S / D = co (Figure 5(c)) the
peak wall vorticity value is 24.02 at a Reynolds number of 5 ; this increases to 57.31 when the
Reynolds number is increased to 200. The location of the peak wall vorticity value tends to shift
upstream as the Reynolds number is increased. The negative wall vorticity values also give an
indication of the extent of the recirculation region in the constricted flow. For Reynolds numbers
greater than 10,negative wall vorticity values are found at the tube surface owing to the existence
of the recirculation eddy downstream of the constriction. The negative magnitude of the wall
vorticity value in the recirculation region increases when the Reynolds number is increased. At a
Reynolds number of 200 there is a large region of recirculatory flow in the constricted tube, as
shown by the extent of negative wall vorticity distribution along the tube wall. With S / D = 1,2
and 3 the recirculation eddy formed downstream of the first constriction has a diminishing effect

0 2 1 6 10
(a) S/O = 1

Figure 5. (a)
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1123

0 2 1 6 8 10 I? 11

I b ) S/O : 3

Figure 5. (b)

on the generation of vorticities by the main stream near the second constriction area. The main
stream approaching the second constriction wall is straightened by the recirculation region
formed in the valley. Hence the maximum wall vorticity generated by the first constriction is
always greater than the maximum wall vorticity generated by the second constriction. The
maximum wall vorticity values obtained for spacing ratios of 1 , 2 and 3 when the flow passes
through the second constriction are always lower than those obtained for a constriction spacing
ratio of co.
The present study with S / D = co is also equivalent to a study of the single constriction tube.
This is similar to a study by Lee and Fung' where numerical results were obtained for Reynolds
numbers in the range 0-25. The corresponding streamlines, vorticity and velocity distributions
for the present investigation are shown in Figures 2 and 5(c) for the Reynolds number range
1124 T. S. LEE

-s
%
T
y :

> 2. Re = 5
= I-

- - - *s

It) S I D :m

Figure 5. Velocity vectors and wall vorticity distributions for different constriction spacings: (a) S / D = 1; (b) S / D = 3;
(c) S / D = a,

5-200. The streamlines, vorticity and velocity profiles show striking similarities between the two
investigations in the Reynolds number range 5-25. At Re = 9.9 an eddy was observed just
downstream of the constriction by Lee and Fung. A similar eddy was also observed in the present
study at Re = 10. The study of Lee and Fung showed that the location of the maximum wall
vorticity shifts slightly upstream of the throat as the Reynolds number is increased. This flow
behaviour was also observed in the present study. Lee and Fung obtained maximum wall
vorticities of 28.5 and 36.5 at Reynolds numbers of 10 and 25 respectively. The corresponding
values obtained in the present investigation were 24.62 and 29.84. The difference is believed to be
due to the manner in which the outlet boundary condition was formulated. Lee and Fung
assumed the outflow was Poiseuillean, whilst the present study assumed an unrestricted flow at
the outlet and allowed the flow profile to develop on its own. An experimental study of a similar
 FLOW THROUGH TUBES WITH DOUBLE CONSTRICTIONS 1125

constriction was presented by Bentz and Evans. l 6 Velocity profiles were obtained using a laser
Doppler velocimeter for steady state results with Reynolds numbers in the range 2-170. A
comparison of these experimental results with the results of the present study for Reynolds
numbers in the range 5-200 shows similar velocity profiles.

CONCLUSIONS
Numerical solutions to the flow fields in the neighbourhood of double symmetrical bell-shaped
constrictions in a circular cylindrical tube were obtained for a Reynolds number range 5-200 with
constriction spacing ratios of 1 , 2 , 3 and 00. For the present study with a constriction of 50%, a
recirculation eddy is found to occur downstream of the constriction when the Reynolds number is
increased to 10. When the Reynolds number is further increased, the point of separation of the
eddy moves upstream of the constriction and the reattachment point moves downstream. At
higher Reynolds numbers a recirculation zone fills the valley region between the two constric-
tions. A separation streamline then divides the flow into two parts: the recirculation region in the
valley, and a relatively straight and parallel flow near the centre of the tube. The maximum
centreline velocity shifts downstream as the Reynolds number increases, and the maximum value
at the second constriction is also higher than the maximum value at the first constriction. In
contrast, the local maximum wall vorticity value always occurs slightly upstream of each of the
constrictions, and the maximum wall vorticity value at the second constriction is always less than
the maximum wall vorticity value at the first constriction.

ACKNOWLEDGEMENT

This work was sponsored by the University of Singapore under research grants RP 25/85 and RP
890633.

APPENDIX: NOMENCLATURE
co-ordinate variables in the original cylindrical co-ordinate system
radius of the tube having a constant cross-section
diameter of the tube having a constant cross-section
opening of the constriction
length of the tube
distance of first and second constrictions from inlet plane
spacing between constrictions, S = s2 - s1
radial velocity component
axial velocity component
axial velocity at infinity
limits of first constriction
limits of second constriction
co-ordinate variables in the transformed co-ordinate system
streamfunction
vorticit y
time
Reynolds number
pressure
density
1126 T. S. LEE

V kinematic viscosity
Av incremental step in radial direction of transformed mesh
A& incremental step in axial direction of transformed mesh
0 relaxation factor in the SOR method

REFERENCES
1. J. S. Lee and Y.F. Fung, ‘Flow in locally constricted tubes at low Reynolds numbers’, A S M E J . Appl. Mech. 37,9-16
(1970).
2. W. L. Oberkampf and S. C. Goh, ‘Numerical solution of incompressible viscous flow in irregular tubes’, Proc. Int. Cot$
on Computer Methods in Non-linear Mechanics, University of Texas, Austin, 1974, pp. 569-579.
3. M. D. Desphande, D. P. Giddens and R. F. Mabon, ‘Steady laminar flow through modelled vascular stenoses’,
J. Biomech., 9, 165-174 (1976).
4. T. J. Mueller, ‘Application of numerical methods to physiological flows’, in H. J. Wirz and J. J. Smolderen (eds),
Numerical Methods in Fluid Dynamics, McGraw-Hill, New York, 1978, pp. 89-154.
5. F. T. Smith, ‘The separating flow through a severely constricted symmetric tubes’, J. Fluid Mech., 90,725-754 (1979).
6. S. 0.Wille, ‘Pressure and flow in arterial stenoses simulated in mathematical models’, Appl. Math. Model. 4,483-488
(1980).
7. I. J. Sobey, ‘On flow through furrowed channels. Part 1: Calculated flow patterns’, J. Fluid Mech. 96, 1-26 (1980).
8. K. D. Stephanoff, I. J. Sobey and B. J. Bellhouse, ‘On flow through furrowed channels, Part 2: Observed flow patterns’,
J . Fluid Mech. 96, 27-32 (1980).
9. S. A. Ahmed and D. P. Giddens, ‘Velocity measurements in steady flow through axisymmetric stenoses at moderate
Reynolds number’, J. Biomech., 16, No. 7, 505-516 (1983).
10. M. Siouffi, R. Pelissier, D. Farahifar and R. Rieu, ‘The effects of unsteadiness on the flow through stenoses and
bifurcations’, J. Biomech. 17, 299-315 (1984).
11. A. T. Prata and E. M. Sparrow, ‘Heat transfer and fluid flow characteristics for an annulus of periodically varying
cross section’, Numer. Heat Transfer, 7, 285-304 (1984).
12. S. Tsangaris and E. Leiter, ‘On laminar steady flow in sinusoidal channels’, J. Eng. Math. 18, 89-103 (1984).
13. V. O’Brien and L. W. Ehrlich, ‘Simple pulsatile flow in an artery with a constriction’, J . Biomech. 18, 117-127 (1985).
14. L. H. Back, J. R. Radbill and Y.I. Cho, ‘Measurement and prediction of flow through a replica segment of a mildly
atherosclerotic coronary artery of man’, J. Biomech. 19, 1-17 (1986).
15. A. A Samarskii and V. B. Andreev, ‘On high accuracy difference scheme for an elliptic equation with several space
variables’, USSR Comput. Math. Phys. 3, 1373-1382 (1963).
16. J. C. Bentz and N. A. Evans, ‘Hemodynamic flow in the region of a simulated stenosis’,A S M E Winter Annual Meeting,
Houston, 30 November-4 December 1975, ASME Paper No. 75-WA/BAlO-10.