No. 12. pp 955 -963. 1983 oo?l-9290 x3 $3 00 -I~ .

oo
C 1983 Pcrgdmnn Pre\r Ltd

FLOW DISTURBANCE MEASUREMENTS THROUGH A

CONSTRICTED TUBE AT MODERATE REYNOLDS
NUMBERS

SAAD A. AHMED and DON P. G~DDENS

School of Aerospace Engineering, Georgia Institute of Technology. Atlanta, Georgia 30332. U.S.A

Abstract--Instantaneous velocities in the field distal to contoured axisymmetric stenoses were measured
with a laser Doppler anemometer. Upstream flow conditions were steady and spanned a range of Reynolds
numbers from 500 to 2000. Autocorrelation functions and spectra of the velocity were employed to descrtbe
the nature of fluid dynamic disturbances. Depending upon the degree of stenosis and the Reynolds number,
the flow field contained disturbances of a discrete oscillation frequency, of a turbulent nature. or both. If
turbulence was detected in a given experiment, it was always preceded upstream by velocity oscillations at
discrete frequency arising from vortex shedding. For mild degrees of stenosis (50”” area reduction or less) the
intensity offlow disturbances was relatively low until the Reynolds number exceeded 1000, thus highlighting
difficulties to be expected in employing flow disturbance detection as a diagnostic tool in the recognition of
early atherosclerosis in major arteries. In view of the relatively high noise levels inherent in noninvasive
Doppler ultrasound systems employed clinically, it seems unlikely that detection of stenoses of less than SO I’*,
area reduction is feasible unless the Reynolds numbers exceed 1000 or unless pulsatility introduces new
unsteady flow features beyond those studied here.

NOMENCLATURE work concentrates upon flow disturbances engendered

by these constrictions, with particular emphasis on the
d diameter of the constriction nature of these disturbances and their quantitative
D diameter of the tube IL2 description. Subsequent investigations will consider
E averaged-power spectral density -
2Af pulsatility in the proximal flow.
f frequency of the fluctuation One of the major characteristics of stenotic flow is
MI stenosis model with 25 5; area reduction
the presence of turbulence or other forms of disturb-
M* stenosis model with 509; area reduction
stenosis model with 75 7; area reduction ances downstream of the constriction. The extent to
M,
I radial coordinate measured from the centerline of which flow is disturbed depends upon: flow conditions
the tube proximal to the stenosis; Reynolds number and. in the
rl?I local radius of the section case of pulsatility, frequency parameter; the degree of
U’ axial disturbance velocity
blockage and contour of the constriction; and wall
u’/V disturbance intensity in the axial direction
u mean axial velocity elasticity. There have been several studies which con-
P mean value of the time averaged velocity across the firm the existence of disturbed flows in the cardiovas-
tube cular system of the dog, even in the absence of a
W’ circumferential disturbance velocity
vascular constriction, as demonstrated by Seed and
w’/V disturbance intensity in the circumferential
direction Wood (1971), by Nerem and Seed (1972), and by
z or Z/D nondimensional axial coordinate measured from Yamagouchi rt al. (1983). Giddens et al. (1976) and
the throat of the stenosis Khalifa and Giddens (1978) have shown that mild
stenoses in the dog aorta create both coherent and
random flow disturbances distal to the constriction,
INTRODUCTION Furthermore, the phenomenon of poststenotic turbu-
lence in the human carotid artery has been long
This investigation was motivated by an interest in
recognized by physicians, who employed the stetho-
detailed fluid dynamics associated in
with constrictions
scope to detect bruits as a method of diagnosing the
tubes, particularly under flow conditions similar to presence of carotid plaques. Generally, however, ,ifl
those found in large arteries. The biological counter- uito data are difficult to obtain under well controlled
part to this model is the phenomenon of localized conditions and with accurate instrumentation. particu-
arterial stenoses, such as may be encountered in larly for human subjects. Consequently, in l>itro a-
atherosclerosis or aortic coarctations. A previous perimentation which models important characteristics
paper (Ahmed and Giddens, 1982) described post- of the in vice situation have become extremely useful.
stenotic velocity profiles and wall shear stresses under Forrester and Young (1970) and Young and Tsai
steady conditions of the upstream flow. The present (1973) have experimentally investigated steady flow
through an axisymmetric, smoothly contoured steno-
sis in a rigid tube. Turbulence spectra downstream of
Recriced December 1982; in revised form May 1983. sharp-edged obstructions were determined by Kirn

1’. :I-,
955
956 SAAD A. AHMED and
and Corcoran (1974). Clark (1976) studied steady flow
through a number of symmetrical and nonsymmetrical
stenoses represented by nozzles with several different
area ratios. Steady and pulsatile flow measurements
downstream of sharp-edged and contoured occlusions
were presented by Cassanova and Giddens (1978). All
of the studies referenced above employed hot-film
anemometry to measure fluid velocity. More recently,
Khalifa and Giddens (1981) have employed a laser
Doppler anemometer to measure centerline velocity in
a pulsatile poststenotic field and have reported the
existence of three types of disturbances: a coherent
starting structure reproducible with each cycle; lami-
nar oscillations arising in the shear layer; and turbulent
velocity fluctuations.
Although these and other investigators have con-
tributed to the understanding of stenotic flow, knowl-
edge in this area is far from complete. Few experiments
have been reported which give a detailed description of
the flow field, particularly in the range of Reynolds
numbers associated with large human arteries; and
most, though not all, data have been obtained with
hot-film probes which invade the flow field and which
are unreliable in regions of separated flow. Results are
presented here for flow disturbances measured with
laser Doppler anemometry throughout the post-
stenotic field of smoothly contoured stenoses with area
reductions of 25, 50 and 75% and at Reynolds
numbers of 500, 1000 and 2000, based upon mean
velocity and unconstricted tube diameter.
METHODS
Flow system and instrumentation
A description of the experimental apparatus has
been provided in a previous paper (Ahmed and
Giddens, 1982), and details will not be repeated here.
Briefly, a schematic diagram of the flow system and a
drawing of the 75 % constriction are shown in Fig. 1.
The flow system could be employed in either a steady
Couplin
Fig. 1. Schematic diagram of the flow system and the geometry
DON P. GIDDENS
or pulsatile mode, and the constriction geometry
followed a cosine contour. The area reductions for the
stenoses were 25, 50 and 75 “/, and the stenosis length
was constant at 4 in. (2 tube diameters). All data
presented are for steady flow conditions upstream of
the stenosis. Reynolds numbers employed were 500,
1000 and 2000, where Re = FDjv with 7 the mean
velocity, D the unoccluded tube diameter (2 in.), and v
the kinematic viscosity of the water/glycerine mixture.
Entrance length was 96 tube diameters.
Velocity measurements were taken with a Thermal
Systems, Incorporated two color, two component laser
Doppler anemometer, operated in the forward scatter
mode. The sample volume dimensions in the fluid were
0.39 mm by 0.052 mm, accounting for the differences
in the indices of refraction. The Doppler signals were
processed with frequency counters, and data rates
always exceeded 5000 Hz. Measurement of negative
velocities were possible by passing both beam pairs
through Bragg cells to provide moving interference
fringe patterns. Axial (u) and circumferential (w)
velocity components were measured simultaneously.
Radial components were not measured due to the
difficulties associated with locating the sample volume
precisely and determining accurately the beam inter-
section angle in the presence of tube curvature.
Data analysis
For these experiments it was assumed that the
velocity at any location was steady in the time mean
sense. Consequently, the measured velocity at a point x
in the flow was expressed as, using the axial component
u as an example,
u(x,t) = U(x)+u’(x,t)+u)N(x,t) (1)
where u(x, t) is the instantaneous velocity, U(x) is the
time averaged velocity, u’(x, t) is the true velocity
disturbance, and u’Jx, t) is a noise term which is
assumed to be uncorrelated with the fluid velocity. The
time averages of u’ and ui are, of course, zero. Using
equation (1) as the basis for analyzing the data, both
constant Head
wturn Tank
of the 75 “/Astenosis model
 Flow disturbance measurements 057

discrete frequency oscillations and turbulence con-

tribute to u’. The level of the disturbance velocity, i.e.
Fib ,
!2
the rms value, and the frequency content are of interest “N
in describing the flow. The noise term comprises both
instrumentation noise and Doppler ambiguity
r?
(George and Lumley, 1973) and presents little difficulty U
m determining mean values, provided an adequate
signal to noise ratio is available. The situation is quite
different, however, if flow disturbance information is r
h
desired. Although the contribution of uh to the
disturbance level and to the power spectrum may Fig. 2. A sketch of an autocorrelation function.
frequently be neglected, this must be ascertained on an
ad hoc basis when performing laser Doppler measure-
essentially unaffected by the noise term. An alternative
ments. In these studies it was frequently necessary to
and equivalent approach is to subtract the noise level in
correct measurements for the effects of uh. This
the spectral domain.
correction was achieved in the following manner. First,
The data obtained from the LDA measurements
the time average value U is computed and is subtracted
were processed on a Hewlett-Packard 545 1B Fourier
from the instantaneous value u to form a stochastic
Analyzer system with hard disc storage. After con-
variable u* with zero mean
siderable experimentation with various sampling par-
u*(t) = u(t)-_ (2) ameters and examining measurements from laminar to
intensely turbulent flows, a data sampling rate of
or, from equation (1)
1024 Hz was selected with a data record duration of
u*(t) = u’(t) + t&(f). (3) one second. This gives a frequency resolution of 1 Hz
and a maximum frequency of 512 Hz. One hundred
Computing the autocorrelation of u*, assuming a
such data records were obtained for each computation
stationary, ergodic process, gives
of U and u’, and the signal was low pass filtered with a
T
48 db/octave filter (Khron-Hite) at 512 Hz to prevent
R*(T) = ,l\rnT& (u’(t)+I&l(t)J (u’(t +T)
s -7
aliasing. The processing of each set of data includes
calculation of the time average velocity U. the energy
+u:,(t +s)jidt. (4) spectrum E, the autocorrelation function, and the rms
If the noise term is uncorrelated with the flow value of the disturbance velocity. as illustrated in
fluctuation-and this assumption is usually invoked Fig. 3.
(e.g., George and Lumley. 1973wquation (4) results Numerous experiments in fully developed laminar
in and in fully developed turbulent pipe flows were
R*(t) = R’(r)+R;(s) (5) performed to verify the accuracy of the measurement
where techniques. These are documented by Ahmed (198 I).
T
R’(T) = lim L u’(t)u’(t + r)dt RESULTS
r+x2T I _T
and Velocity measurements
T
Flow visualization studies (Ahmed and Giddens,
R;(s) = lim L u;(t)u;(t + r)dt.
T-xc s _T 1982) demonstrated that the poststenotic velocity field

If the noise term is a white noise, the power spectrum is

flat and the autocorrelation is an infinite impulse at
‘I = 0 and is zero elsewhere. The autocorrelation func-
tion for the combined flow and noise is illustrated
schematically in Fig. 2. Consequently, the flow disturb-
ance level may be determined by evaluating R(r) as
K-0. In practice, of course, the noise is band-limited
and the expression for Rk(+assuming a sharp cutoff
at frequency o,,,-becomes
2a2
R)v(z) = T sm w,,,T (6)

where a2 = (noise power)/2w,,,. From the compu-

tational standpoint, for the data obtained in this study,
it was usually a straightforward matter to extrapolate
R(T) to’ T = 0, using values from 7 > 0 which were Fig. 3. Flow chart of LDV data reduction.
958 SAAD A. AHMED and DON P. GIDDENS

could contain both random fluctuations and fluctu-
ations at discrete frequency. Although the output
voltage of the LDA system could often be categorized
similarly upon visual inspection, it was usually necess-
ary to examine the energy spectra of the disturbance
velocity in order to classify the flow according to
whether it comprised discrete frequency disturbances,
turbulence or both. Furthermore, under certain con-
ditions the flow possessed fluctuations of discrete
frequency, but the value of this frequency varied,
probably due to the intermittent occurrence of vortex
pairing. An example of this behavior in model M, is
shown in Fig. 4 for both u’ and w’ components at
z= 2.5 and Re = 2000. These data were taken at the
centerline of the tube and represent the disturbed
velocity components sampled for a one second dur-
ation but beginning at different times. Figures 4a and b
indicate fluctuations occurring at a discrete frequency
somewhat higher than those in Figs 4c and d.
Consequently, although this flow would be classified as
dominated by vortex shedding, the associated
frequency is not constant. This behavior results in the
presence of a fairly broad peak in the corresponding
energy spectrum. When velocity variations such as
these were recorded at a given location, fairly long
sampling times (150 s) were employed to obtain the
mean value.
Profiles of the mean axial velocity component were
measured at numerous locations in the poststenotic
field for the various models and flow conditions. These
were reported by Ahmed (1981) and Ahmed and
Giddens (1982). Although the present paper con-
centrates upon the disturbance velocity, Fig. 5 presents
axial velocity profiles for both the 257” and 750/,
stenoses at Re = 1000 to illustrate the behavior of the
mean velocity. The flow through the throat is nearly
uniform and the development of the shear layer may be
seen in the various velocity profiles. Although the effect
of the area reduction is weak for the 25 “/;,stenosis, the
development of a shear layer with inflection points in
the velocity profiles can be readily identified. The
velocity data indicated a stable laminar flow. The data
for the 75”!, constriction show a large recirculatory
flow region which begins immediately downstream of
the throat and extends at approximately z= 5. The
profiles in this region exhibit a jet-like character, being
rather blunt near the axis, while further downstream
the effect of turbulence can be seen in redistributing the
kinetic energy of the flow over most of the cross-
section.

Disturbance velocity

0.25 r
Velocity measurements and analysis for the 25’7”
I
stenosis at Re = 500 and 1000 indicated that the flows
under these conditions were laminar and stable. No
i/P 0.0
interesting flow disturbance phenomena were seen for
these cases. As will be discussed subsequently, only for
Re = 2000 did the 25 “,” stenosis create measurable
-0.25 flow disturbances. For the 50’5, stenosis, no disturb-

o.o>
0 25 f 1 ance phenomena were measured at Re = 500.
However, for Re = 1000 vortex shedding was ob-
served, and for Re = 2000 flow became distinctly
w’/p turbulent in the poststenotic region for z>, 5.0.
Figures 6a and b present profiles of the disturbance
I I 1 I I
-0.25 1

0.25 -

-0.25 -

-",250~0
T(s)
Fig. 4. Instantaneous disturbance velocity at time Tand time Fig. 5. Axial velocity profiles at different axial locations,
T+ 1.0 min at Z = 2.5, r/r, = 0.0, Model M2, R, = 2000. R, = 1000.
 Flow disturbance measurements 959

,E o.4
\
L 0.6

0.8

1.0
10 40 50
0
2oWiVyjO
Fig. 6. Profiles of the disturbance velocity of the SC0 area Fig. 7. Profiles of the disturbance velocity of the 75 “” area
reduction model, Rr = 2000 (a) u’/i? (b) w’/ V. reduction model. Re = 1000 (a) u’/F (b) WIT?

intensity for the axial and circumferential velocity

components. Very intense fluctuations are dem-
onstrated in these profiles with the disturbance in-
tensities in the shear layer reaching values as high as &!Zo.4
40”“. Energy spectra demonstrate that the disturbance \L 0.6
velocities are of a random turbulent nature for
Z 3 5.0, and the intensity values clearly show the
nonisotropic nature of these fluctuations. The peaks in
the disturbance profiles are gradually reduced and
spread due to turbulent diffusion until the cross-
section has, more or less, uniform turbulence energy.
Further downstream these high values of disturbance
intensity are not sustained as the production of
turbulence does not balance its dissipation.
The 75 I’0 stenosis produced flow disturbances at all
Reynolds numbers studied. At Re = 500 these were of
the nature of discrete frequency oscillations only, while
at Re = 1000 and 2000 both vortex shedding and
turbulence were demonstrated. Figures 7 and 8 present
disturbance intensity profiles for both axial and 0 20 80 100

circumferential components at Re = 1OOOand 2000. In

Fig. 7 (Rr = 1000) the disturbances at z= 2.5 were
caused by vortex shedding at discrete frequency, as
demonstrated by examination of the disturbance
energy spectra. At -i = 4.0 the flow has transitioned
into intense turbulence with &,,/Vexceeding 70 Onin
the shear layer. The disturbances are rapidly diffused
over the cross-section as evidenced in the profiles at
Z-= 6.0. For Re = 2000 turbulence existed at every
cross-section studied (Z- >, 1.5). Again, peak intensities
were extremely high in the shear layer and turbulent
diffusion was very rapid in acting to smear these peaks
as the axial distance increased.
Fig. 8. Profiles of the disturbance velocity of the 75’:” area
reduction model, Re = 2000 (a) u’/r (b) u(jr.
Centerline disturbance velocity
In a previous paper by Ahmed and Giddens (1982)
centerline mean velocity data were examined
thoroughly and important conclusions were reported.
The behavior of the centerline disturbance intensity is
equally interesting. Figure 9 presents data for u&/~
and wL,,/V measured distal to the 75”” stenosis at
960 SAAD A. AHMED and DON P. GIDDENS

100

60 - a u R,=SOO
0 I, =lOOO
$60 - v o p ~=ZOOo
v
v
.\ v
’ 0
40 - v
0
v v
20 - v

0 0 0
oo- 1 6 m I 0 I I P I
-1 0 1 2 3 4 5 6
z

100

0 h-500
60 - b 0 R, ~1000
$60 - 7 v 0 Rt12ow

3 40 - v v 0
v 0

20 - v 0 v v

0 0
00 I _
-1 0 1 2 3 4 5 6
z

Fig. 9. Axial variation of centerline disturbance intensity, 75 % area reduction (a) u’/Vi (b) w’/E

Re = 500, 1000 and 2000. There is a slight poststenotic
increase in disturbance intensity for Re = 500, but
examination of the corresponding energy spectra
reveal that this is due to the presence of velocity
fluctuations of discrete frequency and not to turbu-
lence. For Re = 1000 and 2000, intensity was low but
rose rapidly to a maximum, then decreased more
slowly and would eventually have dropped to a value
corresponding to the upstream value, in the far field, as
shown in Fig. 9. Data indicate that the maximum
usually takes place somewhere in the vicinity of the
reattachment point. This sharp rise in disturbance level
is due to transition into intense turbulence. It can be
seen that the locations of maxima in the centerline axial
and circumferential disturbances are in close cor-
respondence, although the nonisotropic nature of the
turbulence remains evident in the differences in magni-
tude of these components.
The effect of area reduction on the centerline
disturbance level is demonstrated in Fig. 10 which
graphs data measured at Re = 2000 for all three
stenosis models. Very little axial variation is seen for
the 25% data, and the disturbances for this case
contained evidence of vortex shedding (e.g., at z = 2.5)

. .
.oo L 8 . 8 : : . , 7 T
-1 0 1 2 3 4 5 6
z

Fig. 10. Axial variation of centerline disturbance intensity, Re = 2000 (a) u’/r, (b) w’/r
 Flow disturbance measurements 961

and of very low levels of turbulence (e.g., at z = 4.0) as IC

I I
---- 7;1.5
evidenced by the shapes of corresponding energy
- - 7=25
spectra. On the other hand, the 50 ‘4 and 75 ‘>istenoses
-.-- f=40
produce intense poststenotic turbulence in addition to
~ f=bO
discrete frequency vortex shedding. The tendency is for
the peak value of intensity to move upstream as the IC

degree of stenosis increases at a fixed Re and as Re

increases at a fixed percent stenosis.
E*
Energy spectra and correlation functions
The energy spectrum and its inverse Fourier trans- 10.
form, the autocorrelation function, give vital informa-
tion in classifying the nature of fluid dynamic disturb-
ances. The energy spectrum (termed power spectrum
in electrical engineering applications) E(f) as applied
in the present study is a measure of the frequency 10.
distribution of the kinetic energy associated with the
disturbance velocity. It is convenient to employ the
normalized spectrum

and Strouhal number

N=Znfd Fig. 11. Normalized centerline disturbance energy spectra,

5
v, 75 “j; area reduction, R, = 2000, u’~.

where E(f) is the averaged-power spectral density

u’* (f )/2AA f is the frequency of the fluctuation, u’* is discussed in the Methods section. Figure 12 gives an
the mean square of the velocity fluctuation, and d and example of this application for the 50 y,,, stenosis and
V, are reference values for length and velocity. The Re = 1000. Note that the disturbance intensity is less
‘correct’ choice for d and 7, is not unambiguous for than 5”” and the frequency of oscillations was es-
poststenotic flows, as discussed by Cassanova and timated to be approximately 1.3 Hz from the graph.
Giddens (1978). However, in the present study we have
taken d to be the constriction diameter and v0 to be the
DlSCUSSlON
mean velocity at the throat of the constriction, in-
asmuch as the spectral data presented are obtained The data presented here, coupled with that reported
from a region dominated by the jet characteristics by Ahmed (1981) and Ahmed and Giddens (1982),
arising from the throat rather than in regions further provide a detailed description of the poststenotic
downstream where the tube diameter and bulk velocity velocity field at Reynolds numbers comparable to
would be more appropriate as characteristic those found in large arteries of man. A two-component
parameters. laser Doppler anemometer was employed to obtain
The utility of the energy spectrum in characterizing velocity measurements noninvasively so that the
the nature of flow disturbances is illustrated in Fig. 11 region of study included flow separation immediately
where the normalized centerline energy spectrum of a’*
is presented for several axial stations for model M3 and
Re = 2000. At 7 = 1.5 the spectrum displays a narrow T---P I
peak at N, = 4.2, corresponding to a vortex shedding i; 4
frequency of 15 Hz. Further downstream, at z = 2.5, R(o) ’
this peak is still seen, together with a much broader
peak at N, = 2.3. As discussed in an earlier section, this
is probably due to intermittent vortex pairing as flow is
becoming unstable. The spectra at z = 4.0 and 6.0 no
longer contain evidence of discrete frequency phenom- u’?
-
ena, and their shapes are characteristic of turbulent R (01
flow. It is on the basis ofsuch spectral information that
disturbances were classified as of a discrete frequency, L-z I
turbulent or both.
00 T(S) 1.25

Use of the autocorrelation function to correct for Fig. 12. Autocorrelation function at z= 4.0, r/r, = 0.0,
noise in the measurement of low disturbance levels was Model MZ, Re = 1000.
962 SAAD A. AHMED
downstream of the stenosis. Cassanova (1975) and
Cassanova and Giddens (1978) conducted a series of
experiments similar to the present ones employing a
hot-film anemometer. These works also reported
vortex shedding phenomena and presented results for
the Strouhal number of shedding distal to contoured
stenoses. Figure 13 presents comparisons of those data
and the present study for 50 and 75 y0 constrictions at
similar Reynolds numbers. Both investigations dem-
onstrated a decreasing Strouhal number with increas-
ing axial distance from the throat.
The spectra of turbulent flow disturbances
measured in the present study are compared in Fig. 14
with the data of Clark (1976), .who reported on
measurements for moderate and severe stenoses ob-
tained with hot-film anemometry, and with that of
Resch (1970), also obtained with a hot-film probe in a
fully developed turbulent pipe flow at Re = 6100. In
Clark’s study the velocity data were high pass filtered
at 22.4 Hz to obtain turbulence velocities, and hence
the value of u” is expected to be smaller as a
consequence of excluding low frequency contri-
butions. This will result in slightly larger values of E*
in the spectral results. Clark’s data were obtained in
regions of maximum turbulence intensity. The spectra
of the present study which are graphed in Fig. 14 were
likewise obtained at locations corresponding approxi-
mately to the highest centerline disturbance intensity
and at Re = 2000. It can be seen that the spectra for the
three models agree well with the dimensionless results
from other investigators despite the fact that the
turbulence levels themselves vary over an order of
magnitude from 25 o/oto 75 ‘A stenosis.
The present investigation, because of the use of laser
Doppler anemometry, has provided additional infor-
mation on turbulence intensities and the separated
flow region that has not been available heretofore. In
contrast to hot-film anemometry studies reported
previously, the LDA can separate turbulence com-
ponents in the presence of intense velocity fluctuations.
00 I 2 3 atherosclerotic plaques. In view of the frequency
i response limitations of Doppler ultrasound and of the
Fig. 13. Comparison between the present results and
Cassanova’s (1976) results.
and DON P. GIDDENS
IO
..*l.‘....“.. Res&(,g,fJ
------- Clark(l976)
-.- Model M,
-..- Model M,
- Model M 3
IO
E'
10.
10.
Fig. 14. Comparison between the present normalized spectra
and results of Clark (1976).
Of particular interest in biofluid dynamics is the
relationship of these results to the problem of non-
invasive measurement of poststenotic flow fields
created by localized atherosclerotic plaques. These
results show that very little in the way of disturbed flow
phenomena occur for a smoothly contoured 25%
stenosis until the Reynolds number exceeds 2000, and
even then the low levels of turbulence intensity might
well be masked by the inherent noise level in Doppler
ultrasound systems (Giddens and Khalifa, 1982), the
technique most likely to be employed clinically. For
50% stenosis disturbances fluctuating at discrete
frequencies could be detected although, again, these
were of relatively low amplitude. It was not until the
Reynolds number reached 2OOtlthat high values for
flow disturbances were measured. At Re = 2000 and
50 % stenosis both discrete vortex shedding and turbu-
lence were easily detected. When the degree of constric-
tion reached 75 %, it was found that vortex shedding
and turbulence were both present at Re = 1000 and
2000. However, at Re = 500 the disturbances tended to
be much weaker in intensity and more difficult to
characterize.
Even though limited to steady flow, the present
investigation points out the difficulty to be encoun-
tered in employing flow disturbance detection as a
diagnostic tool in the recognition of mild to moderate
physiological variability to be expected in blood flow
patterns, particularly in areas of bifurcations which are
 Flow disturbance measurements 963

prone to atherosclerosis (Bharadvaj et al., 1982a, b), it Steady flow in a model of the human carotid bifurcation:
seems unlikely that reliable detection of stenoses of less Part I-flow visualization. J. Biomechanics 15, 349-362.
Bharadvaj, B. K., Mabon, R. F. and Giddens, D. P. (1982b)
than 50 “/; area reduction is feasible unless (1) the
Steady flow in a model of the human carotid bifurcation:
Reynolds number approaches 2000 or (2) pulsatility Part II-laser Doppler anemometer measurements.
introduces new unsteady features beyond those J. Biomechanics 15, 363-378.
studied here. An example of (2) is the coherent starting Cassanova, R. A. (1975) An experimental investigation of
structure reported by Khalifa and Giddens (1981). It steady and pulsatile flow through partial occlusions in a
rigid tube. Ph.D. Thesis, Georgia Institute of Technology.
appears indeed that new breakthroughs in hemo- Cassanova, R. A. and Giddens, D. P. (1978) Disorder distal to
dynamic understanding of poststenotic flow and in modelled stenoses in steady and pulsatile flow. J. Biv-
analysis of velocity measurements derived from mechanics 11, 441453.
Doppler ultrasound methods will be required before Clark, C. (1976) Turbulent velocity measurements in a model
early atherosclerosis can be detected reliably by flow of aortic stenosis. J. Biomechanics 9, 677487.
Forrester, J. H. and Young, D. F. (1970) Flow through a
disturbance analysis. Additional experiments under converging-diverging tube and its implications in occlusive
pulsatile flow conditions have been performed and will vascular disease. J. Biomechanics 3, 297-305.
be reported separately. George, W. K. and Lumley, J. L. (1973) The laser Doppler-
Insofar as fluid dynamic factors are concerned the velocimeter and its application to the measurement 01
turbulence. J. Fluid Mech. 60. 321L362.
present investigation has demonstrated the following:
Giddens, D. P. and Khalifa, A. M. (1982) Turbulence
1. Flow separation and associated intense turbu- measurements with pulsed Doppler ultrasound employing
lence can be expected to occur in the immediate a frequency tracking method. Ultrasound Med. Biol. 8.
poststenotic field of locally constricted arteries prior to 427437.
stenoses becoming flow-restricting or ‘hemodynami- Giddens, D. P., Mabon, R. F. and Cassanova, R. A. ( 1976)
Measurements of disordered flow distal to subtotal vas-
tally significant’.
cular stenoses in the thoracic aortas of dogs. Circularion
2. A region of relatively constant centerline velocity Res. 39, 112-I 19.
occurs in the poststenotic field. This velocity rapidly Khalifa, A. M. A. and Giddens, D. P. (1978) Analysis of
decreases when transition to turbulence occurs. disorder in pulsatile flows with application to poststenotic
3. In the region immediately downstream of the blood velocity measurement in dogs. J. Biomechanic:: 11.
129-141.
constriction, mean velocity profiles exhibit a jet-like Khalifa, A. M. A. and Giddens, D. P. (1981) Characterization
character with large velocity gradients. Flow disturb- and evolution of poststenotic flow disturbances. J. BIO-
antes originate in this shear laver , for steadv flow mechanics 14. 279-296.
conditions upstream. Kim, B. M. and Corcoran, W. H. (1974) Experimental
measurements of turbulence spectra distal to stenosis.
4. Turbulence was always preceded by flow disturb-
J. Biomechanics 7, 335-342.
ances at discrete oscillation frequencies as would be Nerem, R. M. and Seed, W. A. (1972)An in vivo study ofaortic
expected from stability theory. flow disturbances. Cardiovasc. Rex 6, l-14.
Resch, F. J. (1970) Hot-film turbulence measurements in
water flow. J. H,draul. Dir. Am. So<. cir. Enars. HY?.
787-800.
REFERENCES Seed, W. A. and Wood, N. B. (1971) Velocity patterns in the
aorta. Cardiovasc. Res. 5, 319-330.
Ahmed, S. A. (1981) An experimental investigation of steady Yamagouchi, T., Kikkawa,.S., Yoshikawa, T.. Tanishlta, K
and pulsatile flow through a constricted tube. Ph.D. Thesis, and Sagawara. M. (1983) Measurement of turbulence
Georgia Institute of Technology. intensity in the center of the canine ascendmg aorta with ri
Ahmed, S. A. and Giddens, D. P. (1983) Velocity measure- hot-film anemometer, J. biomwh. Enyny 105, IV 1XT.
ments in steady flow through axisymmetric stenoses at Young, D. F. and Tsai, F. Y. (1973) Flow characteristics in
moderate Reynolds numbers. J. Biomechanics 16,50>516. models of arterial stenoses-1. Steady flow. J BIW
Bharadvaj, B. K., Mabon, R. F. and Giddens, D. P. (1982a) mechanics 6. 395410.